Chemically Enabled CO2- Enhanced Oil Recovery in Multi- Porosity, Hydrothermally Altered Carbonates in the Southern Michigan Basin Final Technical Report DE-FE-0031792 DOE FOA-1988 Prepared by: Battelle 505 King Avenue Columbus, Ohio 43201 Principal Investigator: Dr. Neeraj Gupta Deputy Principal Investigator: Aubrey Collie Authors: Aubrey Collie, Derrick James, Timothy Vance, Stuart Skopec, Jared Hawkins, Samin Raziperchikolaee, Evan Byrnes, Matt Young, Neeraj Gupta Submitted to: The U.S. Department of Energy, National Energy Technology Laboratory May 20, 2026 Battelle does not engage in research for advertising, sales promotion, or endorsement of our clients’ interests including raising investment capital or recommending investments decisions, or other publicity purposes, or for any use in litigation. Battelle endeavors at all times to produce work of the highest quality, consistent with our contract commitments. However, because of the research and/or experimental nature of this work the client undertakes the sole responsibility for the consequence of any use or misuse of, or inability to use, any information, apparatus, process or result obtained from Battelle, and Battelle, its employees, officers, or Trustees have no legal liability for the accuracy, adequacy, or efficacy thereof. Battelle | May 2026 TBR. Task 8 – Development Plan: Based on learnings from Tasks 5 through 7, performance assessment and modeling were used to generate high-level economic analysis of CO2-EOR for the TBR trend, including CO2 sourcing, transportation, and storage. This task also created a development strategy plan to guide project development in the TBR. The plan was intended to be utilized by individual operators because production in the region is conducted by small to medium, independent producers. Additionally, the plan included geologic characterization, required oil properties, necessary surface infrastructure, Underground Injection Control (UIC) permitting and regulation, meeting greenhouse gas reporting requirements, qualifying for 45Q tax credits, and developing monitoring, reporting, and verification plans to facilitate wide-scale deployment in the TBR. Summary and Path Forward: The TBR is a large oil and gas field that has produced oil since the 1970s. Drilling activity and production have fallen off significantly in the last decade, despite United States Geological Survey (USGS) estimation that half of known reserves remain in the reservoir, and an additional 800 million barrels of oil may be as-yet undiscovered. This is in large part due to the highly complex, heterogeneous nature of the hydrothermal dolomite reservoir. Geological characterization and lab studies completed during the project confirm that a dual-porosity system dominates production within the TBR trend, with the majority of oil stored in small but abundant matrix porosity, but the majority of the production coming from large, highly permeable vugs and fractures that make up the most transmissive, easily-swept portions of the reservoir. Based on lab testing of surfactant effectiveness and core floods with CO2 and surfactant, a stimulation plan was developed utilizing CETAC-30, a foaming surfactant, and CO2. In order to test the efficacy of chemically enabled CO2-EOR in the TBR trend, West Bay Exploration’s Lee 26 field, a small, linear reservoir developed with four wells, was repressurized using water. The project then implemented a huff & puff-style stimulation at the Fischhaber 2-35 well in the Lee 26 field, injecting CETAC-30 laden water and 2,000 tons of CO2 across two distinct injection phases. After a soak period where CO2 and the surfactant were allowed to interact with the reservoir and residual fluids, the well was production tested for approximately two months. The production test did not yield any additional hydrocarbons from the Fischhaber 2-35. However, hydrocarbons were recovered from an offset well, indicating at least partial success in mobilizing residual oil via chemically enhanced CO2 stimulation. Laboratory and field work associated with this project was completed with no injuries, no unintended release of chemicals or hydrocarbons, and no damage to the environment. Based on the data recovered from the geological characterization program, laboratory studies, modeling and simulation Battelle | May 2026 iii efforts, and field testing, the project team completed an economic analysis of CO2-EOR as applied to the entire TBR trend and propose a development plan that can be used by small local operators to effectively implement CO2-EOR within their TBR assets. Battelle | May 2026 iv Contents Notice ........................................................................................................................................ 2 Acknowledgements .................................................................................................................... 1 Executive Summary ................................................................................................................... 2 Contents .................................................................................................................................... 5 List of Figures............................................................................................................................. 7 List of Tables .............................................................................................................................17 1.0 Task 1: Project Management & Planning ........................................................................ 1 1.1 Introduction .......................................................................................................... 1 1.2 Project Objectives ................................................................................................ 1 1.3 Scope of Work ...................................................................................................... 3 1.4 Organization & Project Partners............................................................................ 3 1.5 Project Timeline and Changes to Project Structure ............................................... 4 1.6 Schedule & Milestones ......................................................................................... 7 2.0 Trenton-Black River Geologic Overview.......................................................................... 9 3.0 Task 2: Advanced Field Characterization & Machine Learning Based Data Integration..27 3.1 Subtask 2.1: Data Compilation, Review, and Analysis .........................................30 3.2 Subtask 2.2: Risk Assessment.............................................................................70 3.3 Subtask 2.3: Advanced Field Characterization .....................................................72 3.4 Subtask 2.4: Integrated Physics-Based Machine Learning and Advanced Data Analytics ..............................................................................................................77 4.0 Task 3: Laboratory Experiments ....................................................................................78 4.1 Subtask 3.1: Core Characterization .....................................................................78 4.2 Subtask 3.2: MMP Determination.........................................................................83 4.3 Subtask 3.3: CO2 Core Flood .............................................................................88 4.4 Subtask 3.4: CO2 Foam Flood ............................................................................91 4.5 Subtask 3.5: Lab-Scale Compositional Modeling .................................................96 5.0 Task 4: Design and Simulations of Injection Scenarios.................................................101 5.1 Subtask 4.1: Development of Static Earth Models .............................................101 5.2 Subtask 4.2: Upscaling and Ranking of Static Earth Models. .............................108 5.3 Subtask 4.3: Material Balance Study .................................................................108 5.4 Subtask 4.4: Dynamic Simulations and Analysis ................................................110 6.0 Task 5: Field Injection Tests ........................................................................................142 6.1 Subtask 5.1: Site Preparation at the Marshell 1-35, Lee 26 Field, Calhoun County 142 6.2 Subtask 5.2: Baseline Site Characterization and Monitoring ..............................145 6.3 Subtask 5.3: Injection Operations ......................................................................147 7.0 Task 6: Production Monitoring......................................................................................173 7.1 Subtask 6.1: Fluid Sampling ..............................................................................173 7.2 Subtask 6.2: Fluid Flow .....................................................................................175 7.3 Subtask 6.3: Pressure Analysis: Identifying the Placement of CO2 in the Reservoir ...........................................................................................................180 7.4 Subtask 6.4: Additional Monitoring Analyses .....................................................184 7.5 Subtask 6.5: Integration of Monitoring Technologies ..........................................186 Battelle | May 2026 v 8.0 Task 7: Performance Assessment................................................................................188 8.1 Subtask 7.1: Performance Assessment .............................................................188 8.2 Subtask 7.2: Modeling Update and Scale-Up.....................................................190 9.0 Task 8: Development Plan ...........................................................................................205 9.1 Subtask 8.1: Economic Analysis ........................................................................205 9.2 Development Plan Discussion ...........................................................................237 10.0 Summary and Outlook on CO2-EOR for the Trenton-Black River Trend in Southern Michigan ......................................................................................................................253 11.0 References ..................................................................................................................254 Appendix A: Supporting Data for Task 8 Economic Analysis ....................................................257 Attachment A: Advanced Field Characterization and Machine Learning Based Data Integration, Task 2 Final Report......................................................................................................259 Attachment B: UT Austin Laboratory Experiments to Optimize CO2-EOR Fluids for Complex Carbonate Reservoirs ..................................................................................................263 Attachment C: UT Austin Bureau of Economic Geology Seismic Interpretations of Fractured Trenton-Black River Carbonate in Southern Michigan .................................................264 Battelle | May 2026 vi List of Figures Figure 1. Project goals, objectives, and timeline.......................................................................... 2 Figure 2. Battelle is the project leader, while West Bay Exploration is the primary project partner, providing cost share as well as field site access. Additional cost share and field site access was provided by Innova Exploration. Analytical and technical expertise has been provided by Core Energy, Western Michigan University, UT Austin, and UT BEG. ....................................................................................................................... 3 Figure 3. Producer Price index for CO2 production, 2016-2026. Price increased significantly and unforeseeably since the project was approved. Source: U.S. Bureau of Labor Statistics via FRED® https://fred.stlouisfed.org/series/PCU3251203251204 ............. 5 Figure 4. Block flow diagram showing CO2 offtake and delivery process from local CO2 emissions source. .................................................................................................... 5 Figure 5. Schematic example of a CO2 huff and puff operation. CO2 is injected into the reservoir via a target well (1). CO2 is allowed to remain in the reservoir to ‘soak’ for some time, mobilizing previously immobile oil (2). The target well is then produced, ideally generating an uplift in oil production (3). .................................................................. 6 Figure 6. Regional depth-to-Precambrian basement map illustrating the shape of the Michigan Basin. From Catacosinos and Daniels, 1991 ........................................................... 9 Figure 7. Paleozoic stratigraphy of the Michigan Basin. Modified from Swezey et al., 2015.......10 Figure 8. Map showing distribution of major known hydrothermal dolomite trends in and around the Michigan Basin. Total production from the Albion-Scipio trend (approximately 135 MMBO), Stoney Point field (approximately 12 MMBO) and the Indiana-Lima trend (>500 MMBO) are illustrated. Modified from Hurley and Budros (1990). ........12 12 Figure 9. Location of the TBR fields, area of interest, Innova Exploration property, and pipeline rights of way. Images provided by Innova Exploration, Inc. ....................................12 Figure 10. Idealized facies track distribution in TBR rocks in the Michigan Basin. This cross- section from nearshore (left) to offshore (right) depositional energy. Texture code letters correspond to Dunham (1962) categories, M-Mudstone, W-Wackestone, P- Packstone, G-Grainstone. From Robinson (2012). .................................................13 Figure 11. Schematic diagram illustrating the general northwest-southeastern trend of faults and folds in the Michigan Basin. Thin, NW trending lines are a schematic representation of this trend. Bold lines named structural features: AS = Albion-Scipio trend, LM = Lucas-Monroe Anticline, HL= Howell Anticline, SC = Sanilac fault. Scale in miles (left) and kilometers (right)......................................................................................16 Figure 12. Diagram showing various associations of hydrothermal dolomite related to extensional tectonics. Modified from Davies (2000, unpublished data). Cited in Davies and Smith, 2006. ........................................................................................17 Figure 13. Example model for hydrothermal dolomitization with dolomitizing fluids concentrated along fault-controlled conduits and typically dolomitize rocks close to the fault plane. Note however, how the lateral extent of reservoir dolomite is illustrated schematically to vary in a vertical sense. ................................................................18 Figure 14. Map of major TBR Fields with greater than 100,000 barrels of oil cumulative production...............................................................................................................22 Figure 15. Capillary pressure versus percent Hg saturation of pore volume with injected mercury (Hg). For most samples, pressures of Mercury of 1,000 to 10,000 psi are required to move through more than 20% of the sample pore volume (from Sattler, 2015). .....23 Figure 16. Core photo of 10-foot interval of core from the Marathon-Mann #6 well showing a 2.5 centimeter thick shale bed. Approximately four vertical feet of dolomite lies directly Battelle | May 2026 vii below the shale as indicated by the purple highlights. All adjacent strata are undolomitized limestone (from Feutz, 2012). ..........................................................24 Figure 17. Two-dimensional theoretical model of Anderson Oil Co.-Whitaker #2 role of shale baffles in helping dolomitize strata. Dolomitizing fluids invade permeable facies away from vertical faults, illustrating the role heterogeneity of the original rock can influence fluid migration Red arrows = the migration of hydrothermal fluids, purple = dolomitized zones, yellow = Facies 1, green = Facies 3, red-orange = Facies 4. Drawing is not to scale (from Feutz, 2012). ............................................................25 Figure 18. Shale baffle on wireline logs. Logs are from the West Bay Exploration-Kilbourn Farms #5-2 well in Rice Creek Field Northern portion of Albion- Scipio field area, paired with the respective borehole image log illustrate what is interpreted to represent a thin shale seam (at 4,290 ft in the upper Trenton Formation) with dolomite diagenetic alteration directly below (from Feutz, 2012). ...........................26 Figure 19. Workflow illustrating the flow of results from data compilation and analysis to machine learning and static earth modeling. .........................................................................28 Figure 20. Map of all wells compiled in Albion-Scipio and analogous fields ................................31 Figure 21. Example formation tops for the Utica Shale (A) and the Trenton (B). The top of Utica in this well comes in at about 3285 ft measured depth as progressive downward transition to higher gamma ray values; the top of the Trenton comes in at about 3650 ft measured depth as a sharp transition from very high to very low gamma ray values.....................................................................................................................32 Figure 22. Locations of wells with routine core analysis (blue). ..................................................33 Figure 23. Photograph of workshop attendees observing a TBR whole core. .............................34 Figure 24. Map of 12 wells with core viewed during the workshop..............................................34 Figure 25. Pie chart of lithology types observed in cored intervals..............................................35 Figure 26. Pie chart of texture/matrix type (left), and bar chart of texture type by lithology/rock type. .......................................................................................................................36 Figure 27. Pie chart of fracture observations by lithology. ..........................................................37 Figure 28. Pie charts of vugs by type and lithology (upper), and vug type by lithology (lower). ...38 Figure 29. Pie chart of observed WMU developed facies. ..........................................................39 Figure 30. Density ridge plots of porosity (left) and permeability (right) for all wells with core data showing a variety of property distributions. .............................................................40 Figure 31. Porosity and permeability scatterplot showing poor exponential fit.............................40 Figure 32. Density ridge plots of porosity (upper) and permeability (lower) by lithology. .............41 Figure 33. Location of type wells (red) in Albion-Scipio (left), Stoney Point (middle), and Napoleon (right) fields.............................................................................................43 Figure 34. Violin plots of thickness (left), cap dolomite thickness (middle), and dolomite thickness (right) for each field. ...............................................................................................43 Figure 35. Bubble map of cap dolomite thickness for each field with Albion-Scipio on the left, Stoney Point in the middle, and Napoleon to the right. ...........................................44 Figure 36. Bubble map of the dolomite thickness for each field with Albion-Scipio on the left, Stoney Point in the middle, and Napoleon to the right. ...........................................44 Figure 37. Violin plots of average effective porosity for the total TBR and dolomitic intervals (upper) and mean permeability (lower). ..................................................................45 Figure 38. Violin plots of the net thickness. ................................................................................46 Figure 39. Map of study area wells selected with CT data, CBIL, and both. ...............................47 Figure 40. The CTViewer 2D interface with feature-segmented overlay on CT data. ..................49 Figure 41. Example snapshots of image logs before processing (right) and after processing (left) showing improved resolution of the images. ...........................................................51 Figure 42. Locations of wells with image logs in the Napoleon field, colored by priority. Red circles indicate down-selected wells. ......................................................................52 Battelle | May 2026 viii Figure 43. Vug classification based on percentage of vugular occurrences at depth. Unknown (5) indicates a data gap or image distortion. ................................................................53 Figure 44. Fracture classification based on open, partially filled, filled, or induced fractures. Unknown (5) indicates a data gap or image distortion interval. ...............................53 Figure 45. Fracture occurrences in the Curtis, Lantis, Swank, and Whalen Byron wells showing fractures at the base of the Utica and top of the Trenton formations.......................54 Figure 46. Fracture occurrences across all image log analysis wells, indicating open, partially filled, filled, and induced fractures. .........................................................................55 Figure 47. Open fractures at the base of the Utica in the Hilden and West Bay wells. ................56 Figure 48. Vuggy occurrences across all image log analysis wells, indicating percentage of vuggy occurrences in the well. ...............................................................................57 Figure 49. Example results of the seismic well tie for the Travis 1-18 well showing log responses on the left and synthetic seismogram on the right. ..................................................59 Figure 50. Example horizons traced using the results from the seismic well tie. .........................59 Figure 51. Example horizons picked along a W-E line through the 3D seismic volume. ..............60 Figure 52. Horizon maps of the Utica (left) and Trenton (right) showing overall trend and heterogeneity. Cool colors are deeper, hot colors are shallower. ...........................60 Figure 53. Horizon maps of the Back River (left) and Glenwood (right) showing overall trend and heterogeneity. Cool colors are deeper, hot colors are shallower. ...........................61 Figure 54. Results of the RMS amplitude showing areas of high amplitude along known structural features. High amplitudes are yellow; low amplitudes are red. ...............62 Figure 55. Results of the Instantaneous frequency attribute showing low frequency to the north. High amplitudes are cool colors, low amplitudes are warm colors. .........................63 Figure 56. Variance attribute showing areas of greater variance. Grayscale colors are low variance, warm colors are high variance. ...............................................................63 Figure 57. Acoustic impedance of the Black River showing areas of low impedance in red which could be associated with high porosity. ..................................................................64 Figure 58. Interpreted seismic W-E line showing complex fault system. .....................................65 Figure 59. Interpreted seismic N-S line showing complex fault system. ......................................65 Figure 60. Interpreted faults on ant-track slice showing a series of faults throughout the seismic (left), and 3D view of faults (right). ..........................................................................66 Figure 61. Interpreted seismic section showing interpreted surfaces. Note the Trenton Formation top has a strong peak on seismic reflection (in black), whereas the Utica Formation top has a generally negative peak (trough) characteristic (in white)........................67 Figure 62. An example of the Utica surface interpolation and clean up. The first image on the left shows the map before processing, and the one on the right shows the map after processing..............................................................................................................67 Figure 63. An example of a single well-based inversion. The blue curve in the 1st track shows original impedance, and the red curve shows the seismic-inverted impedance. The error, the difference between the seismic trace and synthetic, is minimal (~0.07). ..68 Figure 64. P-impedance from seismic inversion in the Lee field. There are two wells inserted in the cross-section. Inversion-based impedance matches with well-log-based impedance. Impedance curves in both wells and seismic have the same scale and color pattern for ......................................................................................................69 Figure 65. Comparison between well-log-based impedance and seismic inversion-based impedance, showing overall good match. ...............................................................69 Figure 66. Estimated porosity from seismic inversion in the Lee field. ........................................70 Figure 67. Description of DarkVision's Hades-R and Hades-T casing evaluation tool capabilities. From Dark Vision. ..................................................................................................73 Figure 68. Casing wall loss features identified by Hades tools. None of these features is a Battelle | May 2026 ix concern for wellbore integrity..................................................................................73 Figure 69. Wireline logging at the Fischhaber 2-35 well pad during December, 2024. ................74 Figure 70. Carbon/Oxygen log comparison between 2024 and 2025 logging run. Depth range is 4,500-4,616 feet measured depth, covering the primary reservoir zone. Note hardware effects from tubing in the well in 2024 (tracks 3-6) as well as transitional effects from water-diesel interface in 2025 log (tracks 3&7). ...................................76 Figure 71. CT-scan of core sample #6-44 showing the largest pore present on the sample which measured 11.102mm. ............................................................................................79 Figure 72. Mercury porosimetry curves for core sample #1-40 ...................................................80 Figure 73. Mercury porosimetry curves for core sample #6-41 ...................................................80 Figure 74. Capillary pressure versus mercury saturation for core samples #1-40 and #6-41 ......80 Figure 75. Oil-water relative permeability curves. .......................................................................83 Figure 76. Schematic of the set up for MMP determination apparatus using slim tube. ..............85 Figure 77. The relationship between oil recovery and pressure (Michigan oil) ............................86 Figure 78. The relationship between oil recovery and pressure (Napoleon oil) ...........................86 Figure 79. The relationship between oil recovery and pressure (McCann live oil) .......................87 Figure 80. The oil recovery at 1,750 psi at different situations ....................................................88 Figure 81. The oil recovery at 1,750 psi with or without additive ME-2 .......................................88 Figure 82. Schematic of the set up for CO2 flooding experiments ..............................................89 Figure 83. Oil recovery, water cut, and relative pressure drop for the continuous CO2 flooding experiment..............................................................................................................90 Figure 84. Oil recovery, water cut, and relative pressure drop for WAG flooding experiment ......91 Figure 85. Soloterra843 (0.5wt%) surfactant, brine solution (30,000 ppm), and oil foam stability test. Foam breakdown >50% occurs after 10 minutes. ...........................................92 Figure 86. CETAC-30 (0.5wt%) surfactant, brine solution, and oil foam stability tests. ...............92 Figure 87. LMDO (0.5wt%) surfactant, brine solution, and oil foam stability tests. ......................93 Figure 88. Graph showing surfactant stability half-life results in brine only and brine with oil solutions. Soloterra843 had the least foam stability in higher density brines. LMDO showed greater stability in higher density brines. CETAC-30 is comparable to LMDO, however, there is less stability in higher density brines. ..............................93 Figure 89. A model of a pure CO2 flood to determine approximate oil recovery in the Lee 26 field. ...............................................................................................................................94 Figure 90. A model of a CETAC-30 and CO2 flood to determine approximate oil recovery during a foam flood scenario. ............................................................................................94 Figure 91. Foam evolution versus time for both foam cases in the presence and absence of methanol ................................................................................................................95 Figure 92. Foam height versus time for both foam cases in the presence and absence of methanol ................................................................................................................95 Figure 93. The slim tube simulation results for live oil and pure CO2 ..........................................97 Figure 94. Composition of Hoffman stock tank oil sample. .........................................................98 Figure 95. Compositional fluid model match to experimental data for the 350 GOR Hoffman oil. This model was used as the basis of the black oil fluid model used in history matching efforts in Q4 FY24. ..................................................................................99 Figure 96. Modeled stock tank viscosity compared with lab-measured values. ...........................99 Figure 97. Black oil properties for the fluid model used in simulation. Top: solution gas-oil ratio (Rs) and oil formation volume factor (Bo); Middle: gas formation volume factor (Bg); Bottom: viscosity of oil and gas (Viso and Visg, respectively). ..............................100 Figure 98. Pressure buildup analysis for the Marshell well on 1/30/2009 (left) and 4/28/2009 (right). Shown are Horner plots (top) and typecurves (bottom) for each test, as well as the overlain model fit........................................................................................103 Figure 99. Isopach map of the Trenton Formation showing thin ‘sag’ features (black dashed Battelle | May 2026 x outlines) in ............................................................................................................104 Figure 100. A- Arbitrary line through 3D seismic volume showing interpreted wrench faults bounding the sag features shown in the figure above. ..........................................105 Figure 101. Dip of maximum similarity seismic attribute showing Lee field extent. ....................106 Figure 102. Cross-sectional view of model showing zones, layering, and faults bounding the dolomitized sag feature. .......................................................................................107 Figure 103. Cross-sectional view of initial porosity distribution in the Lee Field, with higher porosity zones present to the SE near the Fischhaber well. .................................108 Figure 104. Material balance estimates of OOIP and OGIP for the Lee 26 reservoir. ...............109 Figure 105. Depiction of the initial gas saturation in the Lee field tank model. Model dimensions are 4,030 ft x4,030 ft x35 ft. ..................................................................................111 Figure 106. Pressure data and history match...........................................................................112 Figure 107. Cumulative Oil (green) and Gas (maroon) production data and history matches. ..112 Figure 108. Placement of the horizontal well within the gas cap of the reservoir tank model. Coloring corresponds to gas saturation in November 2019. .................................113 Figure 109. Pressure (orange) and injected water volume (blue) in the history matched tank model with repressurization. In this scenario, water is injected at 50,000 bbls/mo. until pressure reaches 1,000 psi. ..........................................................................114 Figure 110. Image of the updated volumetric model for the Lee field with approximate locations for the four wells. Orientation is NW to SE............................................................114 Figure 111. Lee field history-matched model results (lines) with field data (points) for Cumulative Oil (black), Cumulative Gas (green), and Reservoir Pressures (orange). .............115 Figure 112. Image illustrating the sector to be repressurized (pink) in the volumetric model. The orientation of the cross-section is NW to SE. ........................................................115 Figure 113. Average pressures in the sector shown in Figure 112 for a variety of injection scenarios. Target pressure is shown by the pink line, and three scenarios manage to attain this pressure. ..........................................................................................116 Figure 114. Three-dimensional initialization of Lee Field static model in CMG-GEM. Image shows porosity throughout the model. ..................................................................117 Figure 115. Portions of the Lee field used in the CO2 volume estimation: (top) lower region of the field within which the Lee field reservoir lies; (bottom) SE sector of the field, highlighted in pink. ...............................................................................................117 Figure 116. Three-dimensional rendering of the dynamic model, viewed from the south. The upper render indicates the distribution of matrix porosity; the lower render indicates the distribution of fracture and vuggy porosity. White/blank space indicates limestone matrix that is not pressure connected to the reservoir. ..........................119 Figure 117. Dynamic reservoir model filled with reservoir fluids. The model indicates a reservoir filled with gas-saturated oil, with an overlying gas cap. The lower portion of the reservoir is a water leg, although there are no indications it provided any water drive. ....................................................................................................................120 Figure 118. Field-scale production history match showing comparison between gas/oil production and pressure of the simulation and field data. .....................................120 Figure 119. Well-by-well history match at the Lee 26 field, showing gas (blue) and oil (black) production at each well in the simulation (solid line) vs. field data (open circles). .121 Figure 120. Pressure response at the Fischhaber (left) and Marshell (right) wells due to water injection. The model response does not match observations................................121 Figure 121. Pressure distribution (top) and fluid saturations (bottom) in the Lee field model after 15 months of water injection. ................................................................................122 Figure 122. Field-wide history match of current best-fit Lee 26 model. Match quality is slightly decreased from prior models in favor of injection data match. ..............................123 Figure 123. Well-by-well history match of current best-fit Lee 26 model. ..................................123 Battelle | May 2026 xi Figure 124. Injection pressure data match of current best-fit Lee 26 model. Downhole gauge data (points) are coarsened from the high resolution output of the gauges. ..........124 Figure 125. Saturation distributions in the vug/fracture network (top) and matrix (bottom) at field depletion in the Lee 26 model. The orange box highlights a column of oil in the “fracture” network at field depletion. ......................................................................125 Figure 126. Average reservoir pressure in each of the 5 huff and puff scenarios during pressurization, CO2 injection, and production. The colors range from low to high pressure as follows: orange, blue ,black, green, purple. .......................................126 Figure 127. Cumulative oil (top) and oil production rate (bottom) for each of the five huff and puff scenarios. Curves are synchronized for comparison purposes. Colors correspond to the low-high pressure colors described in Figure 118. ..........................................126 Figure 128. History matching results for the various shoebox model cases assessed in the sensitivity analysis. ...............................................................................................128 Figure 129. Cumulative recovery in the unfractured shoebox model for AOF production scenarios .............................................................................................................................129 Figure 130. Cumulative recovery in the unfractured shoebox model for rate-limited (600 bpd) production scenarios. ...........................................................................................130 Figure 131. Cumulative recovery in fractured shoebox model for AOF production strategy. Water re-injection was done through all three offset wells. .............................................130 Figure 132. Cumulative recovery in fractured shoebox model for AOF production strategy. Water re-injection was done through only the Marshell well. ...........................................131 Figure 133. Cumulative recovery in the fractured shoebox model for rate-limited (600 bpd) production strategy. Water re-injection was done through all three offset wells. ...131 Figure 134. Sensitivity results for low matrix poro/perm and low fracture poro/perm cases at two different pressure levels. The left panel shows an AOF production strategy and right shows rate-limited ................................................................................................132 Figure 135. Oil production rate (top) and cumulative production (bottom) for CO2 treatment vs. water-only treatment across five different repressurization scenarios. ..................134 Figure 136. Average reservoir pressures in each of the seven different repressurization scenarios run to test the impact of miscibility. Colors are consistent across scenarios in Figure 129 and Figure 130. ..............................................................134 Figure 137. Oil production rate in the first 250 days for the 7 repressurization scenarios in Figure 128. ......................................................................................................................135 Figure 138. Cumulative production versus production time (top) and versus date (bottom) across the seven repressurization scenarios from Figure 136. ........................................135 Figure 139. Oil rate (top) and cumulative recovery (bottom) for 4 repressurization cases, both with and without well stream recycling. Top panel is plotted in real time, while bottom panel is synchronized to start of production. .........................................................137 Figure 140. Water cut in four production scenarios with and without well stream recycling (reinjection and no reinjection, respectively). ........................................................138 Figure 141. Cumulative recovery in water-only scenarios with and without water reinjection (orange and blue respectively). ............................................................................138 Figure 142. Oil rate (top) and cumulative recovery (bottom) for huff and puff scenarios with four different CO2 volumes (0, 1,000, 2,000, 3,000 tonnes). ........................................139 Figure 143. Oil rate comparison of 2000t huff and puff with three different injection rates (1,000 tpd, 500 tpd, 250 tpd). ..........................................................................................139 Figure 144. Oil rate (top), cumulative oil (middle) and water cut (bottom) for four different production scenarios of a 2000t huff and puff. ......................................................140 Figure 145. Process Flow Diagram of proposed Production Facility .........................................143 Figure 146. Marshell Production Facility as of March 4, 2025 ..................................................144 Figure 147. Baker Hughes crew preparing to run wireline logs at the Fischhaber 2-35 well. .....146 Battelle | May 2026 xii Figure 148. Water injection rate data from the Marshell 1-35 water injection well. Note rate decrease during the second half of 2024. Injection rate subsequently increased after repair of purge water pump at water source..........................................................147 Figure 149. Wellhead pressure data collected during water injection phase. Spikes in Jacoby data correspond to well work operations. .............................................................148 Figure 150. Bottom-hole pressure data from the Fischhaber (orange), Marshell (blue), Jacoby (black), and Beattie (red) wells during water injection. ..........................................149 Figure 151. Datum-corrected bottom-hole pressure data from the Fischhaber (orange), Marshell (blue), Jacoby (black), and Beattie (red) wells during water injection. ...................150 Figure 152. Bottomhole temperature data across the Lee field during water injection. ..............151 Figure 153. Fluid level shots from non-repressurization wells in the Lee 26 field, July 2024 through December 2024. ......................................................................................152 Figure 154. Fluid tracer injection and sampling diagram, Lee 26 field. Tracers were injected into the Marshell well, and sampling will occur at the remaining 3 wells to understand flow pathways of injected water across the field. ..................................................153 Figure 155. Tracer concentration (parts per trillion) in bi-weekly samples from the Fischhaber 2- 35, Jacoby 1-26 HD2, and Beattie 1-27 HD1 wells. Note the logarithmic scale on the y-axis. ..................................................................................................................153 Figure 156. Salinity data (parts per million) captured from the Fischhaber 2-35, Jacoby 1-26 HD2, and Beattie 1-27 HD1 wells. ........................................................................154 Figure 157. January 7th production dry-run meeting, including representatives from West Bay Exploration, Battelle Memorial Institute, Baker-Hughes, Messer Americas, Hadra Safety Consulting, and M&S Oil Well Services. ....................................................155 Figure 158. Safety and operational walkdown of the Fischhaber 2-35 site. Well pad has been improved with gravel to support the weight of CO2 storage tanks. The wellhead is inside the fenced enclosure in the left of the image. .............................................155 Figure 159. Array of 80 ton CO2 storage tanks at the Fischhaber 2-35 pad..............................156 Figure 160. CO2 storage operations at the Fischhaber 2-35 pad. .............................................156 Figure 161. Delivery of CETAC-30 surfactant. .........................................................................157 Figure 162. Storage tank and pumping unit used for dilution and injection of surfactant-laden water. ...................................................................................................................157 Figure 163. Gauges on surfactant pump truck during surfactant injection. Injection completed at an average of 2.3 barrels per minute, with the well taking all fluid on vacuum. .....158 Figure 164. Fischhaber wellhead connected to CO2 injection spread. ......................................159 Figure 165. CO2 injection spread (left) and wellhead (right) during CO2 injection. Frost build-up on these components during injection is due to condensation and freezing of ambient moisture caused by cold cryogenic liquid and/or Joule-Thompson cooling. .............................................................................................................................159 Figure 166. Pumping pressure and total injected volume for Stage 1 of CO2 huff and puff at the Fischhaber 2-35 well. ...........................................................................................160 Figure 167. Messer Americas storage hand making connection between CO2 transport tank and on-site CO2 storage. .............................................................................................161 Figure 168. Example of normal, small-volume CO2 venting during storage and stand-by. ........161 Figure 169. Pumping pressure and total injected volume for Stage 2 of CO2 huff and puff at the Fischhaber 2-35 well. ...........................................................................................162 Figure 170. Datum-corrected bottomhole pressures at all four wells during the CO2 injection and soak periods. ........................................................................................................163 Figure 171. Tubing pressures at all four wells during CO2 injection and soak periods. Note that Fischhaber data is plotted on a different axis. ......................................................163 Figure 172. bottom hole gauge pressure data provided ...........................................................164 Battelle | May 2026 xiii Figure 173. Derivative plot for first falloff cycle .........................................................................165 Figure 174. Derivative plot for second falloff cycle ...................................................................167 Figure 175. Bottomhole temperature data during CO2 injection and soak periods. ...................168 Figure 176. Wellbore pressure and temperature gradients in the Fischhaber 2-35 on April 16th, 2025. ....................................................................................................................170 Figure 177. Fluid density as a function of depth within the Fischhaber 2-35 on April 16, 2025. .170 Figure 178. Pressure and temperature data from the Fischhaber 2-35 on April 16, 2025, overlaid on a CO2 PVT diagram. .......................................................................................171 Figure 179. Wellbore pressure and temperature gradients in the Fischhaber 2-35 on April 18, 2025. ....................................................................................................................172 Figure 180. Pressure and temperature data from the Fischhaber 2-35 on April 18, 2025, overlaid on a CO2 PVT diagram. .......................................................................................172 Figure 181. Example gas chromatograph output during production (9/16/25). ..........................173 Figure 182. Composition of Beattie oil sample. ........................................................................174 Figure 183. Relative difference in mole percent by component between Beattie oil and Hoffman oil. ........................................................................................................................174 Figure 184. Design schematic for original Electric Submersible Pump. ....................................176 Figure 185. ESP production data indicating inconsistent performance of the first ESP. Gas lock may have been an issue, however, assembly error was identified as the primary failure mode. ........................................................................................................177 Figure 186. (A) Fractured radial bushing in one ESP stage and (B) impeller hub with excessive wear and tip damage (lower). ...............................................................................178 Figure 187. Detailed schematic of optimized replacement ESP. Note shroud (gray cone) over intake ports...........................................................................................................179 Figure 188. Fluid rates (top) and cumulative volumes (bottom) observed during the production test period.............................................................................................................180 Figure 189. Datum-corrected bottom-hole pressure data from the Fischhaber (orange), Marshell (blue), Jacoby (black), and Beattie (red) wells from May 28, 2025 through the end of available data. ......................................................................................................181 Figure 190. Wellhead pressure data at all four wells during production testing. ........................182 Figure 191. Downhole temperature data from the production testing period. ............................183 Figure 192. Wellbore pressure and temperature gradients in the Fischhaber 2-35 on June 9, 2025. ....................................................................................................................184 Figure 193. Conceptual diagram showing the reactions occurring during a single-well partitioning tracer test; diagram is from tracer test provider Tracerco. ....................................185 Figure 194. Example of production curves used to determine residual oil saturation by single well tracer test, diagram is from tracer test provider Tracerco. ....................................186 Figure 195. Reservoir zones of the simplified reservoir model used in sensitivity analysis. .......191 Figure 196. Well-by-well primary production volumes for base scenario (top) and aggregate of sensitivity scenarios (bottom). In aggregate plot, dashed line shows best fit to data, while shaded region shows range of results. ........................................................196 Figure 197. Field-wide fluid volumes and pressure values during primary production for base scenario (top) and aggregate of sensitivity scenarios (bottom). In aggregate plot, dashed line shows best fit to data, while shaded region shows range of results....196 Figure 198. Well-by-well bottomhole pressure data during injection for base scenario (top) and aggregate of sensitivity scenarios (bottom). In aggregate plot, dashed line shows best fit to data, while shaded region shows range of results. ................................197 Figure 199. Fluid volumes from production test for base scenario (top) and aggregate of sensitivity scenarios (bottom). In aggregate plot, dashed line shows best fit to data, while shaded region shows range of results. ........................................................197 Figure 200. Tornado chart showing sensitivity of Fischhaber oil production post-CO2 treatment Battelle | May 2026 xiv to various variables. ..............................................................................................198 Figure 201. Jacoby downhole gauge data (points) and model data from Scenario 0.1 with a dynamic opening and closing fracture. Note the pressure increase in early January in both datasets. ...................................................................................................199 Figure 202. Fitted time decline (left) and q vs. Q (right) curves for Albion-Scipio field using EGLE data. .....................................................................................................................202 Figure 203. Subset of production rate scenarios (dashed lines) generated for the Albion-Scipio field. .....................................................................................................................203 Figure 204. Oil produced, CO2 injected and CO2 produced for a single scenario at the Albion- Scipio field. ...........................................................................................................204 Figure 205. Schematic of the cost-benefit analysis framework used in this work (adapted from Fukai et al., 2015). ................................................................................................206 Figure 206. Oil Rates at the Albion-Scipio field used as input to economic models. .................208 Figure 207. Simplified diagram showing components of CO2 injection and recycling operations. .............................................................................................................................212 Figure 208. CAPEX (left) and average annual OPEX (right) for the 12-year project in the Group 1 simulations of the Albion-Scipio Oifield. Note: Although the well plugging and abandonment is slated to occur at the end of the last year of the project, the cost is averaged over the entire project period. ...............................................................218 Figure 209. Total Taxes/Royalties (red) and net revenue without 45Q (light green) and with 45Q (dark green) at oil prices of $60 per STB (patterned fill) and $80 per STB (solid fill) for all Group 1 simulations of a 12-year project in the Albion-Scipio Oilfield. .........218 Figure 210. CAPEX (left) and average annual OPEX (right) for the 12-year project in the Group 2 simulations of the Albion-Scipio Oifield. Note: Although the well plugging and abandonment is slated to occur at the end of the last year of the project, the cost is averaged over the entire project period. ...............................................................219 Figure 211. Total Taxes/Royalties (red) and net revenue without 45Q (light green) and with 45Q (dark green) at oil prices of $60 per STB (patterned fill) and $80 per STB (solid fill) for all Group 2 simulations of a 12-year project in the Albion-Scipio Oilfield. .........219 Figure 212. CAPEX (left) and average annual OPEX (right) for the 20-year project in the Group 1 simulations of the Albion-Scipio Oifield. Note: Although the well plugging and abandonment is slated to occur at the end of the last year of the project, the cost is averaged over the entire project period. ...............................................................222 Figure 213. CAPEX (left) and average annual OPEX (right) for the 20-year project in the Group 2 simulations of the Albion-Scipio Oifield. Note: Although the well plugging and abandonment is slated to occur at the end of the last year of the project, the cost is averaged over the entire project period. ...............................................................222 Figure 214. Total Taxes/Royalties (red) and net revenue without 45Q (light green) and with 45Q (dark green) at oil prices of $60 per STB (patterned fill) and $80 per STB (solid fill) for all Group 1 simulations of a 20-year project in the Albion-Scipio Oilfield. .........223 Figure 215. Total Taxes/Royalties (red) and net revenue without 45Q (light green) and with 45Q (dark green) at oil prices of $60 per STB (patterned fill) and $80 per STB (solid fill) for all Group 2 simulations of a 20-year project in the Albion-Scipio Oilfield. .........223 Figure 216. CAPEX (left) and average annual OPEX (right) for the 12-year project in the Napoleon Oifield. Note: Although the well plugging and abandonment is slated to occur at the end of the last year of the project, the cost is averaged over the entire project period. ......................................................................................................225 Figure 217. Total Taxes/Royalties (red) and net revenue without 45Q (light green) and with 45Q (dark green) at oil prices of $60 per STB (patterned fill) and $80 per STB (solid fill) for all simulations of a 12-year project in the Napoleon Oilfield. ............................226 Figure 218. CAPEX (left) and average annual OPEX (right) for the 20-year project in the Battelle | May 2026 xv Napoleon Oifield. Note: Although the well plugging and abandonment is slated to occur at the end of the last year of the project, the cost is averaged over the entire project period. ......................................................................................................227 Figure 219. Total Taxes/Royalties (red) and net revenue without 45Q (light green) and with 45Q (dark green) at oil prices of $60 per STB (patterned fill) and $80 per STB (solid fill) for all simulations of a 20-year project in the Napoleon Oilfield. ............................228 Figure 220. Net cumulative discounted cashflow for Group 1 (top) and Group 2 (bottom) Simulations without 45Q tax credits (left) and with 45Q tax credits (right) for 12-Year projects with oil price of $60 per STB. Simulations include the Low RF (black), Mod RF/long to peak (red), Mod RF/short to peak (orange), and High RF (blue). .........230 Figure 221. Net discounted cumulative cashflow for Group 1 (top) and Group 2 (bottom) Simulations without 45Q tax credits (left) and with 45Q tax credits (right) for 12-Year projects with oil price of $80 per STB. Simulations include the Low RF (black), Mod RF/long to peak (red), Mod RF/short to peak (orange), and High RF (blue). .........231 Figure 222. Net discounted cumulative cashflow for Group 1 (top) and Group 2 (bottom) Simulations without 45Q tax credits (left) and with 45Q tax credits (right) for 12-Year projects with oil price of $60 per STB. Simulations include the Low RF (black), Mod RF/long to peak (red), Mod RF/short to peak (orange), and High RF (blue). .........233 Figure 223. Net discounted cumulative cashflow for Group 1 (top) and Group 2 (bottom) Simulations without 45Q tax credits (left) and with 45Q tax credits (right) for 12-Year projects with oil price of $80 per STB. Simulations include the Low RF (black), Mod RF/long to peak (red), Mod RF/short to peak (orange), and High RF (blue). .........234 Figure 224. Net discounted cumulative cashflow for the Napoleon Field Simulations without 45Q tax credits (left) and with 45Q tax credits (right) for 12-Year projects with oil price of $60 per STB. Simulations include the Low RF (black), Mod RF/long to peak (red), Mod RF/short to peak (orange), and High RF (blue). ............................................235 Figure 225. Net discounted cumulative cashflow for the Napoleon Field Simulations without 45Q tax credits (left) and with 45Q tax credits (right) for 12-Year projects with oil price of $80 per STB. Simulations include the Low RF (black), Mod RF/long to peak (red), Mod RF/short to peak (orange), and High RF (blue). ............................................236 Figure 226. Net discounted cashflow for the Napoleon Field Simulations without 45Q tax credits (left) and with 45Q tax credits (right) for 20-Year projects with oil price of $60 per STB. Simulations include the Low RF (black), Mod RF/long to peak (red), Mod RF/short to peak (orange), and High RF (blue).....................................................237 Figure 227. Net discounted cumulative cashflow for the Napoleon Field Simulations without 45Q tax credits (left) and with 45Q tax credits (right) for 20-Year projects with oil price of $80 per STB. Simulations include the Low RF (black), Mod RF/long to peak (red), Mod RF/short to peak (orange), and High RF (blue). ............................................238 Figure 228. Active TBR wells colored by distance to nearest neighbor. High-quality cross-well seismic can be acquired in wells <250 ft apart, moderate quality in wells <500 ft apart, and low quality in wells 500-1,000 ft apart. .................................................240 Figure 229. CT scan of the largest pore identified in laboratory-tested core samples (sample #6- 44). .......................................................................................................................241 Figure 230. Generalized block flow diagram showing typical CO2-EOR production facility setup. .............................................................................................................................244 Figure 231. Existing rights-of-way (ROWs) (left) and sensitive areas (right) for the project area (U.S. EPA, 2024; United States Census Bureau, 2023; Pipeline and Hazardous Materials Safety Administration (PHMSA), 2019, 2018, 2017; Energy Information Administration (EIA), 2022, 2020; Federal Railroad ..............................................245 Battelle | May 2026 xvi List of Tables Table 1. Summary of Project Tasks, Objectives and Outcomes .................................................. 2 Table 2. Cost estimates for CO2 offtake facility. .......................................................................... 6 Table 3. TBR facies classification for strata observed in selected Michigan cores from Robinson (2012) .........................................................................................................................14 Table 4. Summary of key activities performed under Task 2 ......................................................28 Table 5. Data received and integrated by type and source .........................................................30 Table 6. Summary of flags developed for cored intervals ...........................................................35 Table 7. List of core intervals selected for additional data collection. ..........................................48 Table 8. Wells with image log data, including depths and formations logged. .............................51 Table 9. Summary of risk assessment objectives. ......................................................................71 Table 10. Summary of preliminary risk assessment conclusions. ...............................................72 Table 11. Basic information of reservoir core samples from Napoleon field ................................78 Table 12. The properties of the core mount ...............................................................................81 Table 13. The properties of crude oil..........................................................................................81 Table 14. The properties of formation brine ...............................................................................81 Table 15. Calculated results for oil and water relative permeabilities ..........................................82 Table 16. Crude oil sample data used for MMP calculations (after Brannigan et. al., 1960)........83 Table 17. Statistical MMP values for the three samples. ............................................................84 Table 18. Properties of crude oils ..............................................................................................84 Table 19. Material and conditions ..............................................................................................94 Table 20. Slim tube model parameters ......................................................................................96 Table 21. Gas and oil composition .............................................................................................96 Table 22. Lumped components and composition of recombined Lee 26 fluid model ..................98 Table 23. Cumulative production and pressure data from the Lee field. ...................................111 Table 24. Scenario Matrix for Shoebox Model Sensitivity Analysis ...........................................127 Table 25. Downhole gauge installation and retrieval schedule at the Lee 26 field.....................149 Table 26. Input parameters into model.....................................................................................165 Table 27. Variables considered in model sensitivity analysis ....................................................192 Table 28. Scenario Matrix for Sensitivity Analysis ....................................................................193 Table 29. RMSE values for sensitivity scenarios. Red/Green colors show RMSE increase/decrease. Light/medium/dark color show change by 1-5%, 5-20%, and >20%, respectively. ..............................................................................................................195 Table 30. Relative change in RMSE for same scenarios as Table 28. Color coding is the same as prior table. ............................................................................................................195 Table 31. Parameters used for production scenario generation................................................201 Table 32. Relevant reservoir criteria for the Napoleon and Albion-Scipio Oilfields. ...................207 Table 33. Summary of outputs from reservoir performance models..........................................207 Table 34. Inputs and definitions of parameters included in the revenue model. ........................208 Table 35. Cost of CO2 Capture of industrial and electric power sources from DOE/NETL (2023a, b) and DOE/NETL (2022a, b), respectively. An adder of 15% was added to simulate a profit collected by the CO2 capture entity. ................................................................209 Table 36. Electric power and industrial facilities and the distance to the center of the Lee, Napoleon, and Albion-Scipio Oilfields. ......................................................................209 Table 37. Cost model parameters, equations, and associated references. Where needed, capital costs escalated from 2006$ to 2023$ using a factor of 1.317 based on the S&P Global Upstream Capital Cost Index (UCCI). Capital costs escalated from 2009$ to 2023$ using a factor of 1.095 based on the S&P Global UCCI. (S&P Global, 2024). ...........211 Table 38. Other Capital and Operating Costs used in the model. .............................................213 Battelle | May 2026 xvii Table 39. Financial metrics used for modeling efforts...............................................................214 Table 40. Economics of a 12-year project in Albion-Scipio Oilfield for Group 1 (newly drilled and existing wells, top) and Group 2 (existing wells only, bottom). All dollars are 2026$. 217 Table 41. Economics of a 20-year project in Albion-Scipio Oilfield for Group 1 (newly drilled and existing wells) and Group 2 (existing wells only). All dollars are 2026$. ....................221 Table 42. Economics of a 12-year project in Napoleon Oilfield. ................................................225 Table 43. Economics of a 20-year project in Napoleon Oilfield. ................................................227 Table 44. Cash Flow, IRR, NPV, and Break-Even Price of a 12-year project in Albion-Scipio Oilfield for Group 1 (newly drilled and existing wells) and Group 2 (existing wells only). .................................................................................................................................229 Table 45. Cash Flow, IRR, NPV, and Break-Even Price of a 20-year project in Albion-Scipio Oilfield for Group 1 (newly drilled and existing wells) and Group 2 (existing wells only). .................................................................................................................................232 Table 46. Cash Flow, IRR, NPV, and Break-Even Price of a 12-year project in Napoleon Oilfield using existing wells only. ..........................................................................................235 Table 47. Cash Flow, IRR, NPV, and Break-Even Price of a 20-year project in Napoleon Oilfield using existing wells only. ..........................................................................................237 Table 48. Results of 12-year project simulations where the purchase of purchase of 17,358 tonnes per year for all 12 years for the Low RF simulations and 135,423 tonnes per year for all 12 years for the High RF simulations are required...................................239 Table 49. Results of 20-year project simulations where the purchase of purchase of 17,358 tonnes per year for all 20 years for the Low RF simulations and 135,423 tonnes per year for all 20 years for the High RF simulations are required...................................239 Table 50. Electric power and industrial facilities and the distance to the center of the Lee, Napoleon, and Albion-Scipio Oilfields. ......................................................................244 Table 51. Sensitive features to consider when siting project infrastructure. The protective protocol and rationale are examples from Battelle (2024). ........................................246 Table 52. Potential timeline for a Class II application in Michigan, including individual permitting steps, and the timeline in instances where no valid comments or objections are made and instances where comments or objections are deemed valid. .............................248 Table 53. Table of MiSTAR reporting requirements for UIC Class II injection wells in the State of Michigan (EGLE, 2026) ............................................................................................251 Battelle | May 2026 xviii 1.0 Task 1: Project Management & Planning 1.1 Introduction The primary objective of the proposed research was to carry out a comprehensive laboratory experiment, computer modeling, and field testing-based evaluation of chemically enabled carbon dioxide enhanced oil recovery (CO2-EOR) in the Southern Michigan Basin (SMB) with a focus on the conventional Trenton/Black River (TBR) play. This work was conducted in the complex multi-porosity (i.e., matrix, vuggy, and fracture porosities) and/or hydrothermally altered dolomite (HTD) reservoirs developed along fault systems. Heterogeneities, compartmentalization, and the presence of thief zones make these reservoirs especially challenging for EOR, so they require customized CO2 injection programs where injection of pure CO2 alone may not be effective for satisfactory incremental oil recovery. These challenges are addressed through advanced field characterization, integrated physics-based machine learning (ML) and data analytics, laboratory process development, and optimized field tests. The research concept involved the integration of multiple data types to evaluate fields in the study area to determine which have the lowest technical and environmental risk and are optimal settings for EOR. Laboratory experiments were used to optimize a CO2 flood specific to HTD rock properties and subsequently design and simulate injection scenarios that offer wettability alteration, foaming, and reduced surface tension. The goal of this approach is to improve oil recovery from matrix porosity and mitigate the impact of fracture zones. The results provided strategies to improve oil recovery in complex carbonate formations and development plans for historical plays, thus increasing the likelihood of the application of EOR methods for similar reservoirs and numerous operators across a wide range of reservoirs and fields. The project consisted of eight tasks organized into three budget periods for planning and budget tracking purposes, with an overall period of performance (POP) from October 1, 2019 through March 31, 2026. The total budget for this project was $7,999,998 with$2,153,667 (21.2%) coming from cost share provided by the Project Team. 1.2 Project Objectives The project supported a comprehensive research program with the following novel and innovative features/goals, as identified in the project management plan and illustrated in Figure 1: • Develop and apply methods for improved geologic definition and characterization of HTD reservoirs to support CO2-EOR efficiency evaluation • Develop and apply physics-based ML and data analytics tools to fully integrate multiple datasets and accurately predict key reservoir properties in the TBR fields • Perform detailed laboratory experiments to optimize chemically enabled CO2-EOR fluids in a multi-porosity carbonate system • Develop and apply reservoir simulation workflows to design an efficient pilot test with integrated insights from geologic characterization, ML, laboratory tests, and simulations • Perform a field injection test in a conventional, TBR well to demonstrate engineered CO2-EOR design, production, and monitoring strategies in a field setting • Prepare a Development Strategy Plan to promote field-wide deployment of optimized EOR design in the test site Battelle | May 2026 1 These goals were accomplished through the execution of all eight technical tasks, described in Table 1. Figure 1. Project goals, objectives, and timeline. Table 1. Summary of Project Tasks, Objectives and Outcomes Specific Outcome from this Task Task Description Objectives from the FOA Project Navigated movement between Ensure satisfaction of technical, research sites, partners, and 1 Project Management schedule, and budget objectives and test approaches while meeting requirements project objectives. Advance approaches for First-of-a-kind comprehensive characterization database and characterization of the TBR in southern Improve methods (including but not Michigan Geological limited seismic) to detect, image, and Understanding of the 2 Characterization and characterize faults and fractures. distribution and extent of the Machine Learning Application of new/improved ML and vugs and fractures in the TBR deep algorithms to classify and reservoir using traditional predict fractures and reservoir machine learning and deep properties to improve EOR. learning techniques Improved composition of CO2-EOR Laboratory experiment-driven Laboratory fluid to reduce conformance issues, improved composition of CO2 3, 4 Experiments and lower miscibility pressure, and that targets multi-porosity, Modeling reduce effects of multi-porosity complex carbonate reservoirs reservoir and improves flood efficiency Battelle | May 2026 2 Table 1 (continued). Summary of Project Tasks, Objectives and Outcomes Specific Outcome from this Task Task Description Objectives from the FOA Project Modeling and injection simulations to reduce uncertainty in EOR process Field Injection Testing Demonstration of the proposed 5, 6, Evaluation of the viability of and Performance improvement over the state-of-the- 7 chemical CO2-EOR for Monitoring art in in-situ stress determination. stranded oil recovery in the 8 Development Development strategy plan for TBR and similar HTD plays, Strategy Plan scaling-up new/advanced EOR along with a development plan methodologies for wide-scale field for the field deployment 1.3 Scope of Work The project included tasks aimed at characterizing the TBR play through the use of existing data, collection of new geologic data, and data integration through the application of ML/DA techniques. Laboratory experiments were performed to determine optimum composition of CO2 and flood design to enhance recovery of oil. The geologic characterization and laboratory experiment results were used to design and implement a field injection test at the Lee 26 field in Calhoun County, Michigan. Based on the results of this work, a full-field development plan, including economic analysis, is proposed in Task 8. 1.4 Organization & Project Partners This project was led by Battelle Memorial Institute (Battelle) in Columbus, Ohio. The Project Team included West Bay Exploration, Inc., Messer LLC., Hadra Consulting, Baker Hughes, Innova Exploration, Inc., Core Energy, LLC., Western Michigan University/Michigan Geological Survey, and University of Texas at Austin, and the University of Texas Bureau of Economic Geology. Other partners included Sasol providing in-kind cost share, Valero providing advisory support, and DTE Energy. Figure 2. Battelle is the project leader, while West Bay Exploration is the primary project partner, providing cost share as well as field site access. Additional cost share and field site access was provided by Innova Exploration. Analytical and technical expertise has been provided by Core Energy, Western Michigan University, UT Austin, and UT BEG. Battelle | May 2026 3 1.5 Project Timeline and Changes to Project Structure Project work was initiated in late 2019, focusing on data compilation, review and advanced field characterization. Work proceeded through 2020, including three-dimensional (3D) geophysical analysis, image log analysis, core analysis, and laboratory testing of rock, fluid, and surfactant. Field injection testing was also planned with project partner Innova Energy, which intended to drill a development well, the McCann 1-20, in the Napoleon field. The project assisted with permitting and risk assessment for this well. On completion of drilling, the intention was to test a two-well, inter-well field flood scenario in virgin or near-virgin reservoir conditions, targeting injection of 10,000 tons of CO2. By comparing the production of the injector-producer pair to the production of offset wells, the team would be able to establish the value of surfactant-enhanced CO2 injection in multi-porous systems. The McCann 1-20 well was successfully drilled and completed in early 2021, but petrophysical analysis and production testing of the well indicated sub-economic porosity and oil saturation, and Innova Exploration chose to halt production. Given this poor reservoir character, it was determined that the well was not a sufficient analogue to the rest of the Trenton-Black River Trend and CO2 flood would not deliver the data quality necessary for project success. The decision not to proceed with injection in the McCann 1-20 well triggered a search for a pair of already-drilled wells that may be a suitable for replacement. Multiple well pairs were evaluated. West Bay Exploration, another TBR field operator, approached the project with the idea of injecting in one of its fully depleted fields, the Lee 26 field. The field was quickly evaluated, determined to be suitably analogous to the rest of the trend, and accepted. Work proceeded at the Lee 26 field until project completion, with West Bay Exploration taking on the role as the project’s primary field partner. This pivot to West Bay Exploration’s project site necessitated a no-cost time extension, which extended the project’s period of performance until September 30, 2024. In the time between project inception and the start of CO2 procurement efforts, changes to CO2 supply and demand in the United States domestic market resulted in a significant increase in the CO2 price (Figure 3), reducing the project’s purchasing power and necessitating a reduction in planned injection volume to 2,000 tons of CO2. The project reviewed possibilities for circumventing commercial price pressure by building a CO2 offtake facility at a local Core energy facility, but a facility of the correct size was too expensive (see block flow diagram and cost breakdown in Figure 4 and Table 2). For a well-to-well flood to be effective, injected CO2 volume needs to be greater than 50% of the pore volume between the wells. Given the broad well spacing at the Lee 26 field, 2,000 tons of CO2 was substantially less than 50% of the flooded pore volume, so the Lee 26 field was unable to be efficiently flooded. It was determined that switching to a huff-and-puff strategy (where a single well is used first for CO2 injection and for production, see example in Figure 5) would still provide sufficient data to understand the performance of surfactant-enhanced CO2. Combined with the external supply chain challenges, the pivot to a huff-and-puff operation required the second no-cost time extension for the program, which extended the program through its end date of March 31, 2026. The project proceeded to repressurize the targeted reservoir between August 2023 and December 2024. The goal was to increase reservoir pressure to above the CO2-oil miscibility point at approximately 1,200 pounds per square inch (psi), where interactions between oil and liquid CO2 are enhanced. This pressure was not achieved. However, dynamic reservoir modeling indicated that injecting gas-phase CO2 would be similarly effective. Battelle | May 2026 4 The project proceeded with injection in February 2025, which triggered the transition into the third and final budget period. Due to increased CO2 cost and the timing of Michigan Frost Laws, the project team was limited to two, approximately 1,000 tonne injection intervals and, in April 2025, injection operations were completed and followed by a 6-month soaking period. In September, West Bay Exploration installed an electronic submersible pump (ESP) into the Fischhaber well and planned to produce the well until December. After two months of producing a combination of reservoir and injected water, West Bay Exploration halted production early and removed the ESP to perform swab tests on the Beattie and Jacoby wells. The final four months of the project were spent on the performance assessment, including updates to the models and development planning tasks. Figure 3. Producer Price index for CO2 production, 2016-2026. Price increased significantly and unforeseeably since the project was approved. Source: U.S. Bureau of Labor Statistics via FRED® https://fred.stlouisfed.org/series/PCU3251203251204 Figure 4. Block flow diagram showing CO2 offtake and delivery process from local CO2 emissions source. Battelle | May 2026 5 Table 2. Cost estimates for CO2 offtake facility. Capex Component Low High Access $ 10,000 $ 30,000 Compression $ 5,000 $ 35,000 Liquefaction $ 2,000 $ 20,000 Storage (Both sites) $ 600,000 $ 1,200,000 Loading/unloading $ 70,000 $ 110,000 Well Pump $ 3,000 $ 15,000 Piping/gauges $ 138,000 $ 423,000 Engineering/Construction $ 55,200 $ 169,200 Management Total $ 828,000 $ 1,833,000 Figure 5. Schematic example of a CO2 huff and puff operation. CO2 is injected into the reservoir via a target well (1). CO2 is allowed to remain in the reservoir to ‘soak’ for some time, mobilizing previously immobile oil (2). The target well is then produced, ideally generating an uplift in oil production (3). Modified from Linde Gas. Battelle | May 2026 6 1.6 Schedule & Milestones Battelle | May 2026 7 2.0 Trenton-Black River Geologic Overview Michigan Basin- Formation, Structure, Stratigraphy The Michigan Basin is a roughly circular cratonic sag structure of unclear origin. The basin is approximately 400 km in diameter and underlies the Lower Peninsula of Michigan (Figure 6). Basin fill consists primarily of sediments deposited during three structural downwarping sequences: a Late Cambrian to Early Ordovician sequence deposited in a trough-shaped basin, a Middle Ordovician to Middle Silurian sequence deposited in a broad sag basin, and an Early Devonian to Mississippian sequence deposited in a symmetrical to eastern-tilting basin (Howell and van der Pluijm, 1999). The TBR Group and overlying Utica Shale that comprise the reservoir system for this study were deposited during the second of these sequences (Figure 7). Minor overlying sediments were deposited in the Pennsylvanian and a thick, unconsolidated glacial cover deposited during the Pleistocene (Kirst and Lusch, 2004). Figure 6. Regional depth-to-Precambrian basement map illustrating the shape of the Michigan Basin. From Catacosinos and Daniels, 1991 Battelle | May 2026 9 Figure 7. Paleozoic stratigraphy of the Michigan Basin. Modified from Swezey et al., 2015. The Trenton-Black River Group The Ordovician-age TBR Formations are a thick stack of shallow shelf carbonate rocks occurring in the subsurface of Michigan’s Lower Peninsula. The formations are primarily limestone, except in localized areas of dolomitization across the southern three tiers of counties and in a broad belt along the western margin of the Lower Peninsula. These dolomitized areas Battelle | May 2026 10 contain isolated commercial accumulations of oil and gas in discrete fields. Forty classified Michigan TBR fields have documented production, all in the SMB (Figure 8). Cumulative production from all fields exceeds 160 million barrels of oil and 303 billion cubic feet of natural gas. The most notable is the Albion-Pulaski-Scipio Field with more than 135 million barrels of oil (MMBO) cumulative production. All of these fields are classic examples of geometrically complex dolomite reservoirs described and modeled by the HTD reservoir facies concept (Davies and Smith, 2006). The TBR Formations are predominantly micritic limestone in the Michigan Basin (Catacosinos and Daniels, 1991) although dolomite is both stratigraphically and aerially significant. The Trenton is lithologically similar to the Black River Formation and generally indistinguishable in well logs. Prominent, regionally correlative Middle Ordovician K-bentonites are used to separate the Trenton from the Black River in Michigan and elsewhere in the eastern United States. These bentonites represent ash falls throughout the Michigan and Appalachian basins resulting from Taconic orogenic events and volcanism (Kolata et al., 1996). The Trenton is overlain by the Utica/Maquoketa shale in the Michigan Basin and adjacent cratonic arch areas (Fisher et al., 1988). This shale unit is an important top seal for the hydrocarbon reservoirs. The TBR succession increases in combined stratigraphic thickness from 400 ft to 1,000 ft from west to east sides in the Michigan Basin. In the basin, the Trenton consists of open marine limestone facies deposited in an intermittently high energy, subtidal or deeper water ramp environment (Keith, 1985), and lacking basin margin or tidal flat facies (Wilson and Sengupta, 1985). Local intervals of grain-rich fabrics are interpreted as either carbonate shoals or storm (tempestite) deposits (Fara and Keith, 1988). TBR hydrocarbon accumulations occur mainly as stratigraphic traps resulting from porosity and permeability variations between porous, hydrothermal dolostone that comprises the reservoir and tight, unaltered regional limestone that forms the lateral and (occasionally) top seals to the reservoirs. The Albion-Pulaski-Scipio-Stoney Point trend, which was discovered in 1957 (Figure 9) is by far the largest field in the Michigan Basin, consisting of large interconnected reservoir compartments (and numerous adjacent, isolated compartments) of heavily fractured and dolomitized strata occurring at all stratigraphic levels within the TBR formations. Hurley and Budros (1990) showed that production in the Albion-Scipio trend is from fracture-controlled dolostone reservoirs related to northwest-southeast fault and fold trends running parallel to a regional structural trend. Although these formations are present in the subsurface throughout the Lower Peninsula and in parts of the Upper Peninsula, commercial discoveries have only been found in the southern part of the Lower Peninsula of Michigan, adjoining parts of Indiana and Ohio and in portions of southwestern Ontario, Canada. Battelle | May 2026 11 Figure 8. Map showing distribution of major known hydrothermal dolomite trends in and around the Michigan Basin. Total production from the Albion-Scipio trend (approximately 135 MMBO), Stoney Point field (approximately 12 MMBO) and the Indiana-Lima trend (>500 MMBO) are illustrated. Modified from Hurley and Budros (1990). Figure 9. Location of the TBR fields, area of interest, Innova Exploration property, and pipeline rights of way. Images provided by Innova Exploration, Inc. Battelle | May 2026 12 Depositional Facies and Stratigraphic Architecture The TBR rocks of the Michigan Basin were formed in a widespread, Middle Ordovician, carbonate-rich deposition system of a shallow marine shelf, platform and gently-dipping ramp that occurs in the subsurface and as outcrops throughout the upper Midwest and Eastern United States. The primary carbonate depositional facies in these formations can be descriptively classified using the system of Dunham (1962) which is based on texture characteristics of the carbonate particles comprising the deposit. The Dunham classification uses the ratio of coarser grains to fine-grained carbonate mud as a proxy for depositional energy of the sediments. Mudstones contain mostly fine-grained material with less than 10% coarser grains. Wackestones also contain considerable mud, but more than 10% grains. Packstones contain mostly grains with lesser amounts of mud. Grainstones are mostly grains with less than 10% mud. Layers with these different textures or facies are stacked vertically and variably distributed laterally in response to changes in the energy conditions in the regional environment. An example of an idealized diagram of the regional distribution of these facies was used by Robinson (2012) to reconstruct stacking of these facies through time and to interpret their lateral distribution (Figure 10). Figure 10. Idealized facies track distribution in TBR rocks in the Michigan Basin. This cross-section from nearshore (left) to offshore (right) depositional energy. Texture code letters correspond to Dunham (1962) categories, M-Mudstone, W-Wackestone, P-Packstone, G-Grainstone. From Robinson (2012). Robinson (2012) made detailed descriptions of cores in the collection of the Michigan Geological Repository for Research and Education at Western Michigan University and used the classification of Dunham (1962) to group the rock layers into similar packages that reflected Battelle | May 2026 13 their original depositional environmental conditions. He interpreted the mud-rich rocks as forming in distal, deep water marine settings. Coarser grained facies were presumed to have formed in shallower, more proximal, higher energy settings. Major storms may have occurred frequently and would temporarily increase the seabed energy and sort and redistribute the coarser grains. Robinson (2012) examined cores from 14 wells in the southern Michigan basin and subdivided the strata into six facies based on textures, grain type and sedimentary structures (Table 3). Table 3. TBR facies classification for strata observed in selected Michigan cores from Robinson (2012) Diagnostic Attributes F6 F5 F4 F3 F2 F1 Dominant facies texture packstone- packstone- grainstone Packstone- wackestone mudstone- grainstone wackestone grainstone wackestone Gross thickness (ft) 16.8 100.6 54.1 575.2 584.0 28.4 Average thickness (ft) 8.4 4.6 1.3 6.9 6.9 1.7 Number of intervals 2 22 43 84 84 17 Grain bed mean 0.7 0.2 n-a 0.2 0.15 <0.1 thickness (ft) Grain bed mean 1.3 0.9 n-a 0.4 0.8 1.1 frequency (ft-1) Dominant grain bed packstone packstone n-a packstone- packstone wackestone texture grainstone Burrow fill sediment grain-mud grain grain grain mud-grain mud-grain (dominant) mix mix mix Detailed stratigraphic and petrographic analysis of TBR facies is available in Attachment A. Structural Elements, Thermal Evolution, and Hydrothermal Mineralization in the Michigan Basin Studies of the structural evolution of the Michigan Basin (summarized by Fisher et al., 1988) have identified a predominantly northwest-southeast structural grain composed of faults and folds transecting the basin (Figure 11). A subordinate, antithetic southwest/northeast component is thought to have developed due to periodic reactivation of basement faults and left lateral wrenching (Sanford et al., 1985; Prouty, 1988). Fault reactivation timing is unclear, although stratigraphic studies indicate that the basin periodically hosted isolated, positive bathymetric features throughout the early Paleozoic. These features were probably the result of syndepositional growth faults (Fisher et al., 1988). Regional stratigraphic studies (Lillienthal, 1978) and structural analysis (Prouty, 1988) suggest that the most widespread and intense period of basement fault reactivation occurred at the end of the Mississippian (~350 mybp), coincident with Alleghenian orogenesis in the Appalachian orogen. This coincides with a basin-wide diagenetic event (age dated at about 346 +/- 11 mybp), that includes the formation of “burial” dolomite in the Ordovician St. Peter Sandstone and siliciclastic-dominated Glenwood Formation (Barnes et al., 1992; Girard and Barnes, 1995). Models for hydrothermal mineralization associated with basement fault reactivation have long been used to explain linear fault and fracture-related oil fields (Hurley and Budros, 1990), as well as pervasive epigenetic dolomitization and related Mississippi Valley Type (MVT) Battelle | May 2026 14 mineralization (Budai and Wilson, 1991). Battelle | May 2026 15 Figure 11. Schematic diagram illustrating the general northwest-southeastern trend of faults and folds in the Michigan Basin. Thin, NW trending lines are a schematic representation of this trend. Bold lines named structural features: AS = Albion-Scipio trend, LM = Lucas-Monroe Anticline, HL= Howell Anticline, SC = Sanilac fault. Scale in miles (left) and kilometers (right). The global association of giant oil fields, hydrothermal dolomite, MVT type mineralization, and basement-involved wrench tectonics is documented by Davies (2000). The “hydrothermal dolomite reservoir facies” model and major hydrocarbon accumulations world-wide are thought to be controlled by the localization of hydrothermal fluid movement and mineralization, including hydrothermal dolomitization associated with faults (Figure 12), especially extensional or transtensional faults (Davies, 2000). Numerous fracture-based giant hydrocarbon reservoirs controlled by reactivated basement faults exist, including the Albion-Scipio pool and other Ordovician TBR fields (Tedesco, 1994). Battelle | May 2026 16 Figure 12. Diagram showing various associations of hydrothermal dolomite related to extensional tectonics. Modified from Davies (2000, unpublished data). Cited in Davies and Smith, 2006. Studies of the thermal evolution of the Michigan Basin (summary in Catacosinos and Daniels, 1991) propose at least three different models to explain the thermal maturity of hydrocarbon source rocks, and temperatures and timing of formation of authigenic minerals in the Michigan Basin relative to current depths of burial (no more than approximately 10,500 ft at the deepest for the Ordovician TBR Formations). Nunn et al. (1984) suggested that anomalously elevated geothermal gradients existed during the early Paleozoic throughout the Michigan Basin. Cercone (1984) hypothesized erosional stripping of thousands of feet of late Paleozoic strata. Hogarth and Sibley (1985) and Hurley and Budros (1990) emphasize the importance of spatially isolated hydrothermal fluid conduits that created hydrothermal chimneys, explaining both the spatial distribution of thermal maturity and the distribution of authigenic minerals, including hydrothermal dolomite (Figure 13). Girard and Barnes (1995) suggested that a Late Devonian- Mississippian (345 +/- 11 mybp) episode of authigenic mineralization in the Ordovician St. Peter Sandstone, including hydrothermal dolomitization, suggested anomalously high thermal regimes near faulted and fractured Precambrian basement within the Mid-Continent Rift in the central Michigan Basin. Battelle | May 2026 17 Figure 13. Example model for hydrothermal dolomitization with dolomitizing fluids concentrated along fault-controlled conduits and typically dolomitize rocks close to the fault plane. Note however, how the lateral extent of reservoir dolomite is illustrated schematically to vary in a vertical sense. Fluid inclusions in calcite cements in the Mississippian Bayport Limestone of Eaton County, Michigan, are present with mean homogenization temperature of 146 +/- 12° C (Blaske, 2003 and 2018). The calcite cements are found in conjunction with pyrite, fluorite, and other fracture- filling minerals suggestive of hydrothermal mineralization. Other examples of highly fractured upper Paleozoic bedrock formations in the surface and shallow subsurface are known to contain exotic suites of authigenic minerals not in chemical equilibrium with modern, fresh groundwater (Westjohn and Weaver, 1996). Occurrence of Dolomite in the Trenton-Black River The spatial distribution of limestone versus dolomite in the subsurface can be mapped using conventional wire-line log data. Cross plot of formation density an compensated neutron log responses, in conjunction with gamma ray logs, can be used to distinguish dolomite from limestone and provides a basis for preliminary geological models relating sedimentary facies and structural deformation to dolomite occurrences and hydrocarbon reservoirs on a regional scale. Geological models for the spatial distribution of dolomite based on conventional, subsurface, wire-line log mapping and rudimentary well sample analysis, however, typically provide only incomplete and equivocal evidence to distinguish the geological origin of dolomite. Battelle | May 2026 18 Confident prediction of dolomite reservoir geometry and distribution, on all relevant scales in any hydrocarbon system, requires rigorous, scientific testing and development of geological models for dolomite genesis. Petrographic and analytical methods are required to differentiate dolomite types and infer mechanisms of dolomite formation. Reliable geological models for the spatial distribution of dolomite reservoirs and dolomitized hydrocarbon reservoir rock properties is fundamentally dependent on data relating geological conditions and controls during dolomite genesis to the spatial distribution of dolomitized reservoirs. Dolomite, of various origins, is an important lithologic component of the TBR Formations in the Michigan Basin, and contains distinct spatial and stratigraphic distribution as well as variable petrologic character. Well-documented petrographic and analytical characterization of dolomite types in the TBR have resulted in at least three distinct dolomite types recognized in the Michigan Basin. These are: 1. a dense, finely crystalline, ferroan “cap dolomite” that occurs in the top 10 to 40 ft of the Trenton below the Utica Shale throughout much of the basin (Taylor and Sibley, 1986). This dolomite contains moderately well-preserved primary fabrics. Oxygen isotopic composition, trace element chemistry, and stratigraphic/cross cutting relationships in this dolomite are interpreted to indicate a relatively early, burial origin related to diagenesis and fluid expulsion from shales in the overlying Utica/Maquoketa formation (Taylor and Sibley, 1986; Coniglio et al., 1994). 2. a coarsely crystalline, “saddle” dolomite that typically fills fractures and vugs, that is common in highly faulted and fractured zones such as the Albion-Pulaski-Scipio trend. Petrologic and analytical data (Allan and Wiggins, 1993; Coniglio, et al., 1994), including those from oxygen isotopes and primary fluid inclusions, clearly indicate that this “fracture related” dolomite was produced during “burial” from fluids with elevated temperatures (> at least 100°C), that were most likely transported through fracture systems and involved formation fluids of complex origins. 3. a medium- to coarsely-crystalline, variably ferroan, “regional”, fabric destructive dolomite that may be volumetrically more significant compared to “cap” and fracture- related dolomites discussed above. Genesis of this dolomite type is somewhat enigmatic. Taylor and Sibley (1986) initially identified this “regional” dolomite to be of early diagenetic origin, and thought it to be distributed mainly in the margins to the south and west of the Michigan Basin. Later studies (Fara and Keith, 1988; Coniglio et al., 1994; Budai and Wilson, 1991) related Taylor and Sibley’s “regional dolomite” to “fracture related” burial dolomite with distribution controlled by proximity to fractures. This dolomite type is typically matrix replacive, may or may not preserve primary limestone textures, and probably constitutes substantial volumes of porous and permeable reservoir facies in fracture-trend TBR fields. Routine petrographic analysis is typically insufficient to discriminate multi-phase dolomite that results from differing mechanisms of formation. Complex, multistage, diagenetic alteration including partial to complete replacement or recrystallization of pre-cursor limestone and dolomite textures can superimpose petrographic characteristics that are genetically ambiguous. Analytical methods to distinguish the conditions and mechanisms responsible for dolomitization, including the temperature and composition of dolomitizing fluids, are required to relate dolomite occurrences to mechanisms of dolomite formation and hence the prediction of spatial distribution and properties of dolomitized hydrocarbon reservoirs within the framework of primary Battelle | May 2026 19 depositional fabrics. Geochemical analyses (Allan and Wiggins, 1993) of core material could provide more detailed information for characterization of diagenetic alteration through the TBR interval. Taylor and Sibley (1986) and Budai and Wilson (1991) examined the geochemical characteristics of the three main types of dolomite in the TBR rocks using carbon and oxygen isotopic analyses. All types of dolomite have delta O18 values considerably lighter than Ordovician or Silurian seawater, suggesting that they were all formed at elevated temperatures relative to surface or shallow burial conditions. Although there is some overlap in oxygen isotopic values, Taylor and Sibley (1986) found that these three types of dolomite could be reasonably distinguished from each other by the oxygen isotopic values. The “regional” dolomite had oxygen isotopic values of -5.5 to -7.5 per mil, whereas the “cap” dolomite had values ranging from -6.5 to -8.5 per mil. The most distinctive was the “fracture-fill” or “saddle” dolomite with values from -7.5 to -11.5 per mil. It is interpreted that the larger negative value for oxygen, the higher the temperature of formation. Variations in Dolomitization and Reservoir Geometries in the Michigan Basin Fundamentally different mechanisms of dolomitization will result in distinctly different reservoir geometry and rock properties. Primary depositional facies controls result in regional scale geometric distribution and rock properties in early diagenetic dolomite that are distinctly different from burial, epigenetic dolomite typically related to structural controls and fracturing (Zempolich and Hardie, 1997; Melim and Scholle, 2002). From a reservoir perspective, outcrop studies such as that by Zempolich and Hardie (1997) offer key insight into the preferential control on distribution of dolomite by primary sedimentary facies and resulting variability in the spatial distribution of dolomitized reservoirs. For example, Zempolich and Hardie (1997) showed that the size and distribution of fracture-related dolomite bodies may range from kilometers in areal distribution and hundreds of meters thick when fractures cut across certain primary sedimentary facies/fabrics, to only tens to hundreds of meters in area and only a few meters thick. All mechanisms for dolomitization require fluid flow and mass transport that is imposed or imprinted on pre-cursor facies patterns and primary facies-controlled fluid flow pathways. A well- documented geological model for the volumetric significance, spatial distribution, and the various mechanisms of dolomitization in a petroleum system is essential to decreasing risk and supporting new exploration, step-out and enhanced recovery activities. For the Michigan Basin, existing geological models for the origin of dolomite in Paleozoic TBR hydrocarbon reservoirs are diverse. In different regions and stratigraphic intervals of the Michigan Basin (and surrounding areas), the origin of dolomite is interpreted to result from a variety of mechanisms and processes: As a result of the apparent complexity and diversity of dolomite origins in these units, and a general lack of understanding of the spatial distribution of dolomite facies basin-wide, the implications of dolomite genesis on hydrocarbon reservoir distribution and quality is poorly, or at least incompletely, understood in the Michigan Basin. In the absence of well-documented geological models for the origins and spatial distribution of disparate dolomite types in these units, predictive models for the spatial distribution, field-scale geometry, and properties of hydrocarbon reservoir rocks are not well established. Although patterns of spatial distribution and types of dolomite have been studied in these reservoir units individually and in selected Battelle | May 2026 20 fields, only Coniglio et al. (1994) have suggested that cross-formational mechanisms common to the TBR zones are probably important to the creation of many/most dolomite reservoirs in these units. Trenton-Black River Production History TBR discovery and development drilling began in 1884 at Findlay, Ohio, along the Findlay- Kankakee Arch in Indiana and Ohio (Davies, 2000) resulting in the first TBR discoveries. This led to the drilling of over 100,000 wells and the production of 500 MBO across the region. The first commercial TBR discovery in the State of Michigan occurred in 1920 in Monroe County. This resulted in the development of the Deerfield Field located along the Lucas-Monroe monocline, an extension of Bowling Green Fault zone from Ohio. Reservoir quality dolomite lenses in the upper 125 ft of the Trenton Group has produced approximately 768,000 barrels of oil from 40 wells drilled on 360 acres. Two small oil fields were discovered in the 1940s in southern Wayne County about 20 miles northeast of the Deerfield field. Sumpter field discovered in 1941 has only one producing well that had very limited production (227 barrels) before being abandoned in 1942. New Boston field, 3.8 miles farther to the northeast, was discovered in 1943 and had two producing wells. It had a cumulative total of 2,349 barrels of oil produced before being abandoned in 1949. The Northville field, in portions of Oakland, Washtenaw and Wayne Counties, was discovered in 1947 for the Dundee pool. The 1954 discovery of the Salina-Niagaran and TBR pools in Northville expanded the commercial potential of this field which sits astride the Howell Anticline, a large northwest-trending basement-rooted structure to the northeast of the Lucas-Monroe monocline. At least 50 producing wells were drilled in the TBR reservoir and cumulative production exceeds 1.1 million barrels of oil. The small Freedom field (1 Trenton well) was also discovered in 1954 and produced around 8500 barrels of oil. The Albion-Scipio Field, a giant (>135 MMBO) TBR field located in Calhoun, Jackson and Hillsdale Counties, Michigan, was discovered in 1957 (Ells, 1962; Davies, 2000). The Scipio portion of the field discovery well was the Houseknecht No. 1 (Sec 10, T5S-R3W – Hillsdale Co.) which was originally drilled for Devonian gas but proved dry. Based on consultation with a local psychic, the well was then deepened and oil was encountered oil at 3,900 feet. The well was completed and flowed 140 barrels of oil per day (BOPD) with “considerable” gas. The Albion Field was discovered by the Rosenau No. 1 (Sec 23, T3S-R4W, Calhoun Co.) and completed for 200 BOPD. Subsequent drilling extended production to the northwest and southeast to become part of the Albion-Scipio-Pulaski Trend. Over 961 wells were drilled by 1986, although most wells were plugged and abandoned by the early 2000s. Currently, the State of Michigan divides the Albion-Scipio-Pulaski Trend into seven sections for production reporting purposes: Albion-Scipio 1 South through Albion Scipio 7 South. In 2006, renewed drilling in and around the Albion-Scipio-Pulaski producing wells commenced based on high resolution 3D seismic surveys. More than 70 new wells have been drilled opening several new compartments in the old field. An additional 4.4 million barrels of oil has been produced through the end of 2020 from these new wells. Stoney Point Field (5 miles east and sub-parallel to Albion-Scipio) was not discovered until 25 years later in 1982 when the JEM Casler No. 1-30 (Sec 30, T4S-R2W, Jackson Co.) encountered dolomite reservoir 115 ft into the Trenton at 3,910 feet. This well hit lost circulation at 4,248 ft. Casing was set and the well was tested at 2,000 BOPD from perforations at 4,161 to Battelle | May 2026 21 4,179 ft. The bottom hole pressure (BHP) drop never exceeded 3 psi and the well was put on production at 220 BOPD. Two hundred and ten wells were drilled around the Stoney Point Trend between 1983 and 1987. Seventy-five wells were oil and gas producers. Estimated oil-in-place figures for these older fields are difficult to accurately calculate due to difficulties in establishing a reliable porosity number as well as total reservoir volume, however, Buehner and Davis (1969) estimated the trend to have 320 to 340 million barrels of oil originally in place. Since the discovery of new compartments in and around the Albion-Scipio-Pulaski Trend, the application of 3D seismic surveying has provided for the discovery of at least 17 new fields, six of which have already produced at least 200,000 barrels of oil. Forty TBR fields are classified by the State of Michigan. Fourteen of these fields have produced at least 100,000 barrels of oil, five fields have cumulative production greater than 1 million barrels, two fields produced greater than 10 million barrels and one field greater than 100 million barrels. Figure 14. Map of major TBR Fields with greater than 100,000 barrels of oil cumulative production Confining System The ultimate top seal for TBR reservoirs is the stratigraphically, immediately overlying Utica Shale. This shale is up to 400 ft thick and is predominantly a silty claystone or silty marl lithology (Sattler and Barnes, 2017). The Utica Shale has very low intrinsic porosity and permeability due to the compact fabric and small size of the mineral grains (Sattler, 2015, Sattler and Barnes, 2017). Sattler (2015) Battelle | May 2026 22 reported porosity of 0.77% and permeability of 0.003 millidarcies (mD) for the Thompson #1-30 well core at a depth of 2,277 ft. The use of mercury injection-capillary pressure (MICP) analysis is a useful indicator of the capacity of fluids to pass through a rock or, conversely, to resist the passage of fluid through the sample. MICP measures the pressure necessary to inject a fluid (mercury [Hg]) into various sized pores in a sample. This provides a quantitative measurement of pore throat sizes which directly affects permeability. Sattler (2015) analyzed five samples from throughout Michigan using MICP (Figure 15). Sattler (2015) was analyzing the capacity of the Utica Shale to act as a seal to retain CO2 that might be injected into underlying units and therefore function as a seal for preventing vertical migration of CO2. The same principle applies to being a seal for retaining hydrocarbons in the reservoir. Figure 15. Capillary pressure versus percent Hg saturation of pore volume with injected mercury (Hg). For most samples, pressures of Mercury of 1,000 to 10,000 psi are required to move through more than 20% of the sample pore volume (from Sattler, 2015). MICP analysis of selected samples of Utica Shale in Michigan show that pressures greater than 10,000 psi for Hg and 2,000 psi for CO2 are required to pass fluid through more than 70% of the pore throats in these tested samples. That suggests that the Utica Shale is able to retain a thick column of hydrocarbons in the TBR reservoir. Multiple thin “shaley” beds are interbedded within the Trenton and Black River carbonate layers. The Black River Shale, near the contact of the base of the Trenton with the top of the Black River, is the best known (Hurley and Budros, 1990; Kolata et al., 1996), however some may serve as minor internal seals acting as vertical barriers to fluid movement and dolomitizing fluids. These beds may be true shales with compositions similar to the Utica Shale or they may be volcanic ash beds (K-bentonites) such as the informally named “E”-shale near the top of the Trenton Formation (Hurley and Budros, 1990). Feutz (2012) documented the mineralogy of Battelle | May 2026 23 these thin shaley beds and their capacity to serve as vertical barriers to vertical movement of dolomitizing fluids (Figure 16). Figure 16. Core photo of 10-foot interval of core from the Marathon-Mann #6 well showing a 2.5 centimeter thick shale bed. Approximately four vertical feet of dolomite lies directly below the shale as indicated by the purple highlights. All adjacent strata are undolomitized limestone (from Feutz, 2012). Feutz (2012) suggests a model for dolomitization of selected intervals of the TBR in portions of the formation that are laterally distanced from the main fault and fracture conduits that moved the dolomitizing fluids up from deeper in the geologic section (Figure 17). Additional seals are present in the very tight “cap” dolomite just beneath the Utica Shale and the limestone lithologies of non-reservoir layers within the reservoir and as a lateral seal at the margins of the reservoir. As an example, the “cap” dolomite in the ARCO-Conklin #1-31 core has an average porosity of 2.0% and an average permeability of 0.01 mD in all three dimensions over the top 25 ft of the Trenton Formation. These values are very similar to data from a limestone core at the margin of the field. Battelle | May 2026 24 Figure 17. Two-dimensional theoretical model of Anderson Oil Co.-Whitaker #2 role of shale baffles in helping dolomitize strata. Dolomitizing fluids invade permeable facies away from vertical faults, illustrating the role heterogeneity of the original rock can influence fluid migration Red arrows = the migration of hydrothermal fluids, purple = dolomitized zones, yellow = Facies 1, green = Facies 3, red-orange = Facies 4. Drawing is not to scale (from Feutz, 2012). It is also possible to recognize the occurrence of these thin shale layers that may serve as barriers to fluid migration by using modern wireline logs and borehole imaging logging (Figure 18). Battelle | May 2026 25 Figure 18. Shale baffle on wireline logs. Logs are from the West Bay Exploration-Kilbourn Farms #5-2 well in Rice Creek Field Northern portion of Albion- Scipio field area, paired with the respective borehole image log illustrate what is interpreted to represent a thin shale seam (at 4,290 ft in the upper Trenton Formation) with dolomite diagenetic alteration directly below (from Feutz, 2012). Battelle | May 2026 26 3.0 Task 2: Advanced Field Characterization & Machine Learning Based Data Integration Task 2 involved a systematic geologic characterization of the TBR play in the SMB which included the development of comprehensive datasets, advanced data analytics, risk assessment, and piggyback field characterization. These activities aimed to address the complex carbonate systems by evaluating the extent of fractures, dolomitization, facies, and reservoir properties with the primary objective of informing the static and dynamic modeling, field injection test, and providing input into the development strategy plan. The task was divided into four subtasks: • Subtask 2.1 – Data Compilation, Review, and Analysis • Subtask 2.2 – Risk Assessment • Subtask 2.3 – Advanced Field Characterization • Subtask 2.4 – Integrated Physics-Based Machine Learning and Advanced Data Analytics Several activities were planned and executed under Task 2 to meet the project objectives. These are summarized in Table 4, including subtask, approach, and outcome. Figure 19 illustrates the flow of results between subtasks, which will lead to the development of a static earth model (SEM) in Task 4. This began with the database compilation of geological and geophysical data from multiple wells and fields. Three different analysis tracks included geophysical (seismic), petrophysical (wireline logs and core), and vug and fractures (advanced data). Each of the analyses produced results which fed into the ML database for the prediction of electrofacies, permeability, and vugs and fractures. All information was fed into the static earth modeling conducted in budget period 2, under Task 4. Battelle | May 2026 27 Figure 19. Workflow illustrating the flow of results from data compilation and analysis to machine learning and static earth modeling. Table 4. Summary of key activities performed under Task 2 Activity Description Approach Outcome Mapped Subtask Database Identification Data management practices Integrated, 2.1 compilation and compilation including documentation, comprehensive of data of quality checking, and database to be multiple types correlation used for from multiple characterization and sources modeling efforts Petrophysical Development of Identification of data High-quality 2.1 Analyses petrophysical challenges, development of database of wireline workflows and workflows for correction, logs, computed analysis of normalization, and statistics and wireline logs computation, and executing reservoir properties, workflows on existing geologic maps and wireline logs cross sections Core Analyses Review of Review existing data, core Geologic 2.1 available core workshop with partners to descriptions and data and core review cores, integration of characterization of analyses core data with wireline log reservoir facies, data database of porosity and permeability, and correlated database Battelle | May 2026 28 Table 4 (continued). Summary of key activities performed under Task 2 Activity Description Approach Outcome Mapped Subtask Vug and Development of Advanced 3D computed Catalog of vug and 2.1, 2.4 Fracture a catalog of tomography (CT) scanning fractures and Characterization vugs and of selected core for integration into fractures and quantitative assessments wireline log characteristics and image log analyses for database flag development Preliminary Review and Seismic well tie, horizon Identification of 2.4 Seismic analysis of mapping, attribute analysis, potential attributes Analysis existing 3D and fault analysis which correlate to seismic volume reservoir properties Wellbore Evaluation of Review of construction and Screening-level 2.2 Integrity wellbore integrity plugging records, evaluation assessment of Evaluation for wells within of cement bond logs, and wellbore integrity to the study area scoring of wells identify potential leakage pathways Geohazard Identification Review of known Identification of 2.2 Assessment and evaluation geohazards such as potential risks of potential structure and seismicity geohazards Risk screening- Application of Application of FEP and Overall screening 2.2 based analysis Features, “what if” analysis assessment of and activity- Events, and methodologies to study area potential risks based “what if” Processes related to CO2-EOR analysis (FEPs) review in the study region and “what if” analysis methodologies to study area McCann 1-20 Collection and Collection of full suite of Site specific 2.3 Well analysis of new basic and advanced wireline assessment for the Characterization data logs for geologic planned injection characterization and test site baseline monitoring Advanced Data type Linking wireline logs, core Clean and complete 2.4 statistical integration into analyses, and vug and database, and analyses comprehensive fracture databases into a identified database and comprehensive database, electrofacies exploratory exploration of data to statistics product descriptive statistics, identify electrofacies, and data imputation Machine Application of Explore machine learning Predicted properties 2.4 learning based machine methodologies to produce at all well locations predictive learning best predictive models for models techniques to electrofacies, permeability, develop and vugs and fractures predictive models for key reservoir properties Battelle | May 2026 29 3.1 Subtask 2.1: Data Compilation, Review, and Analysis Data compilation, review, and analysis was one of the most time and effort intensive portions of this study. Review of basic processes and results is provided below. Full reporting of this effort is contained in the Task 2 Final Report that was delivered to DOE in June 2021. Database Development Numerous data types were compiled and integrated from public and private databases in collaboration with project partners Innova Exploration, Inc. and Western Michigan University. Datasets, detailed in Table 5, included historical well records, reservoir and production data, digital wireline logs (LAS), raster logs (images of wireline logs), well header and formation top data, image logs, core data, and seismic data. Additional data were collected by NETL which was comprised of CT scans, high-resolution x-ray fluorescence (XRF), and high-resolution velocity data for five wells in the study area. Overall, data were compiled for 607 wells in the Albion-Scipio (AS) and analogous fields (Napoleon, Stoney Point, Northville) and was integrated into a comprehensive database in Geographix. A map of all wells in the database are shown in Figure 20. Table 5. Data received and integrated by type and source Data Type Description Purpose Originator Historical Access database including well production, Mapping Innova and Records reservoir characteristics, lease production, Machine Learning partners drilling data, drilling notes, completion data, Correlations etc. compiled for 970+ wells in the AS field. Dynamic Modeling Inputs Historical Paper copies and scans of field reports, well Mapping WMU/MGRRE Records reports, pressure, production, etc. available Machine Learning from donated property to MGRRE. MGRRE Correlations scanned a report from 1960 which had Dynamic Modeling valuable historical reservoir performance Inputs data and maps. Oil and Database of known oil and gas wells with key Risk Assessment MDEQ, Gas well construction, status, and production previous registry information. 2,587 wells (AS Field) 53,801 project work- wells (State) AS Field and State of Michigan public sources LAS Digitized wireline log data for hundreds of wells Structural WMU/MGRRE covering the AS field and surrounding area Framework Petrophysics Integration into SEM LAS Hundreds of wells with digitized data in AS Structural Innova Field, Napoleon Field, and other small TBR Framework Exploration Inc. fields and wildcat wells. Data includes well Petrophysics header information and formation tops where Reservoir Facies available. ML Training Wells, Integration into SEM Raster Raster logs, including cement bond logs, in AS Digitization for Data MDEQ- public Logs Field and the surrounding area. Gaps source Risk Assessment Image Image logs from 26 wells in the Napoleon Field Fracture, fault, and Innova and Logs and 3 wells in Albion-Scipio Field. vug identification Partners Training wells Battelle | May 2026 30 Table 5 (continued). Data received and integrated by type and source Data Type Description Purpose Originator Core Core analyses for 60 wells in the TBR in Reservoir properties WMU/MGRRE, Analysis southern MI, including ~20 wells from the AS Log corrections Innova field. This includes porosity, permeability, Reservoir Facies density, and fluid saturations. Integration into SEM 3D 13 mi2 3D seismic dataset with post stack Advanced attribute Innova and Seismic attribute processing and post stack relative analysis partners acoustic impedance inversion conducted. High ML to map reservoir resolution, clean dataset with minimal near properties and surface issues. facies ML to map fracture network Formation Tops Formation tops were provided by project partners (Innova Exploration and Western Michigan University) for the AS field and Napoleon Field, and tops were quality checked to ensure consistency. Tops for the formations of interest were picked for wells in the nearby Stoney Point field, which focused on wells with good spatial coverage and quality of well logs. Guidelines were developed for consistent formation top picks for the Utica Shale, Trenton, and Black River Shale. The Utica was picked at the base of the first consistently high gamma ray (Figure 21a). The Trenton was picked at the base of the high gamma ray, where is drastically decreases indicating clean carbonate (Figure 21b). The Black River Shale was picked occasionally as many logs ended above this. It was picked where the gamma ray spikes. The formation top data compilation and top picking processes resulted in a total of 579 wells with formation tops in the database. Figure 20. Map of all wells compiled in Albion-Scipio and analogous fields Battelle | May 2026 31 Figure 21. Example formation tops for the Utica Shale (A) and the Trenton (B). The top of Utica in this well comes in at about 3285 ft measured depth as progressive downward transition to higher gamma ray values; the top of the Trenton comes in at about 3650 ft measured depth as a sharp transition from very high to very low gamma ray values. Core Analysis Core measured properties obtained from physical whole core and/or sidewall cores are critical information across multiple activities as depicted in Figure 22. The data were used for wireline log correlation, reservoir characterization, permeability prediction, and vug and fracture characterization, and integrated into the 3D modeling efforts including seismic analysis and property modeling. Analysis of core measured properties were compiled for 48 wells in the Albion-Scipio, Stoney Point, and Napoleon fields, and wells in nearby small fields and wildcat wells. Data included porosity, permeability, fluid saturations, grain and bulk density, and flags for facies, vugs, fractures, and lithology. A map of wells with routine core analysis is shown in Figure 22. Battelle | May 2026 32 Figure 22. Locations of wells with routine core analysis (blue). Anomalous data points can occur where analyses were conducted on a section of core with a fracture or large and/or isolated vug, resulting in porosity and permeability values that are outliers compared to other values in the same intervals. Damaged sample materials, poor analytical work, and/or outdated techniques are other possible sources of data quality problems to be considered. These data points were reviewed to determine whether the measurements are reliable. Porosity and permeability points were compared to fracture and vug flags. In cases where anomalously high values corresponded with isolated vugs or fractures, the points were removed from the quality dataset. Core Workshop A core workshop was held at Western Michigan University (WMU) including team members from Battelle, WMU, Innova Exploration, Inc., and University of Alaska (now BEG) (Figure 23). Using previous work completed by WMU, facies were linked from whole core to wireline log signatures to ultimately predict the location and extent of each facies throughout the field using data analytics. WMU studied seven unique lithofacies in the TBR, the associated core- measured reservoir properties, and the extent of dolomitization as associated to the facies. Resources available at WMU and the Michigan Geological Repository for Research and Education (MGRRE) include about 60 wells with TBR core samples and over 1,300 wells with digitized logs in the Albion-Scipio area. The workshop served as a knowledge and data sharing Battelle | May 2026 33 opportunity and allowed the Project Team to observe and discuss trends in core and log data from TBR reservoirs in Albion-Scipio, Napoleon, Stoney Point, and Northville fields. Cores evaluated during the workshop are shown in Figure 24. Figure 23. Photograph of workshop attendees observing a TBR whole core. Figure 24. Map of 12 wells with core viewed during the workshop. Battelle | May 2026 34 A series of flags was created for the cored intervals which included lithology, textures, features, and facies. Information was compiled from multiple sources including core descriptions, core photographs, and graduate theses. Of the 48 wells with available core data, only 20 wells had enough information to generate flags. The flags are summarized in Table 6. Table 6. Summary of flags developed for cored intervals Type Description Options # of Observations Lithology Presence of rock types including Limestone, Dolostone, Dolomitic- 1,079 dominant and trace occurrences Limestone, Shale Texture Structure of matrix rock Fine, Medium, and Coarse 1,308 indicating reservoir quality Crystalline Features Presence of and type of Small to pinpoint vugs, medium to 1,325 secondary porosity features large vugs, pervasive vugs, and presence of fractures Facies WMU assigned facies based on 1-6 as described in core workshop 807 different rock textures and types Lithology Dolostone was the dominant lithology observed, making up ~71% of the data (excluding unknowns). This was followed by limestone and the fewest observations in dolomitic-limestone (Figure 25). Lithology dolomitic-limestone dolostone limestone unknown Figure 25. Pie chart of lithology types observed in cored intervals. Texture Shale is commonly used as a flag or classification for lithology, but due to its occurrence as thin beds or traces, it has been grouped in the texture category to represent rock texture with low porosity. There were 1,079 total observations with texture evidence, of which fine-crystalline was the most common (56%). This was followed by shale (26%), medium crystalline (16%), and coarse crystalline (1%). Battelle | May 2026 35 Dolomite lithology was dominated by fine-crystalline textures with the highest number of medium-crystalline and shale observations. Dolomitic-limestone was fine to medium crystalline with one occurrence of shale. Limestone was dominated by fine crystalline with few observations of medium and shale. Coarse crystalline occurred in intervals with unknown lithology (Figure 26). Figure 26. Pie chart of texture/matrix type (left), and bar chart of texture type by lithology/rock type. Battelle | May 2026 36 Features Fractures were observed 135 times over the cored intervals. Fractures dominantly occurred in dolostone (84%) (Figure 27). Few occurrences were noted in dolomitic-limestone (4%), limestone (7%), and unknown lithologies (5%). Figure 27. Pie chart of fracture observations by lithology. Vugs were observed 1,325 times over the cored intervals and were dominantly present in dolostone lithology (89%). Only 1% of observations were in dolomitic-limestone, zero in limestone, and remaining 31% in unknown lithology (Figure 28). The majority of vugs observed were pinpoint to small, followed by medium to large. A small percentage were pervasive (4%) which occurred in dolostone lithology. Battelle | May 2026 37 Figure 28. Pie charts of vugs by type and lithology (upper), and vug type by lithology (lower). Facies There were only 806 observations available for WMU developed facies (Figure 29) for the cored intervals. Facies 2 had the greatest number of observations (46%) closely followed by facies 3 (39%). The remaining facies had few observations occurring from <1 to 6%. Battelle | May 2026 38 Facies 0 1 2 3 4 5 6 Figure 29. Pie chart of observed WMU developed facies. Porosity and Permeability Porosity and permeability values were measured for more than 2,800 samples from the cored intervals in 48 wells. Overall, the values ranged from 0.1 to 20.7% porosity and 0.01 to 939 mD permeability. Individually, many wells followed a similar distribution of values, with a subset ranging into higher values (Figure 30). The porosity to permeability relationship was poor due to broad scatter, likely influenced by lithology and secondary porosity changes (Figure 31). The dataset was further explored using data analytics and ML. Battelle | May 2026 39 Figure 30. Density ridge plots of porosity (left) and permeability (right) for all wells with core data showing a variety of property distributions. Trenton-Black River Porosity-Permeability 1000.000 100.000 Permeability (mD) 10.000 1.000 0.100 y = 0.0 886e0.3789x R² = 0.0139 0.010 0 5 10 15 20 25 Porosity Figure 31. Porosity and permeability scatterplot showing poor exponential fit. Battelle | May 2026 40 Dolostone lithology had the highest average porosity at 3.3% with a range from 0.3 to 12.6%. The resulting permeability ranged from 0.09 to 772 mD with an average of 8.5 mD (Figure 32). Dolomitic limestone had a slightly lower average at 2.9% with bi-modal peaks near 1% and 4%. The limestone had the lowest average at 1.5% and the narrowest distribution. Figure 32. Density ridge plots of porosity (upper) and permeability (lower) by lithology. Battelle | May 2026 41 The core analysis activities aimed to catalog and characterize the TBR reservoir through use of physically measured properties. The team began with a core workshop to familiarize themselves with the geology and heterogeneity of the system. During this time, the team selected core to be sent to NETL for advanced scanning including CT scans, XRD, and high-resolution velocity. A set of flags was generated for the cores based on their physical descriptions to capture changes in lithology, textures, secondary porosity features, and facies. Porosity and permeability were evaluated for all wells to determine a trend for use in petrophysical analyses. In summary, the TBR was observed to be highly heterogeneous which varied from tight limestone to porous dolomite, with varying occurrences of fractures and vugs. Petrophysical Analysis The goal of the petrophysical analysis of the wells across the fields is to characterize the reservoir (TBR) to better understand the distributions of porosity, permeability, thickness, and facies. These characteristics will later be integrated into a 3D conceptualization of the Albion- Scipio field and used to understand the feasibility of CO2-EOR in the TBR play. Approach The petrophysical analysis approach involved multiple steps including zone definition, type well selection, log preparation, core data preparation, final log computation, and zone calculations. Reservoir intervals, or facies, were defined for each well and later refined through a series of cross sections to ensure consistency. The results were mapped in a series of bubble maps and statistical plots. Zone Definition A zone is defined as an interval of rock with a top and a base. For the purposes of the petrophysical analyses, the zone was defined as the top of the Trenton and the base of the Black River. In many cases, the wellbore did not fully penetrate the Black River due to the oil- water contact. In these instances, the base of the zone is defined as the last wireline log measurement. When calculations are dependent on thickness or represent the thickness of the zone, it is important to note that the calculations only capture what is logged and not the true thickness of the TBR. Formation tops were provided by project partners (Innova Exploration and Western Michigan University). Tops were quality checked to ensure consistency. Type Well Definition Several wells were selected from each field (Albion-Scipio, Stoney Point, and Napoleon) to be designated as the type wells for the study. Figure 33 shows the geographic location of the wells used. The wells were selected for their data availability, with preference given to wells having core data. Additionally, the lithology and features were considered to ensure the wells captured the expected geologic variability. When available, these rock properties were compiled from driller’s logs and previous studies. Battelle | May 2026 42 Figure 33. Location of type wells (red) in Albion-Scipio (left), Stoney Point (middle), and Napoleon (right) fields. Overall, the total logged thickness (true vertical) was greatest in the Stoney Point field with similar values in the Albion-Scipio and Napoleon fields (Figure 34). Cap dolomite thickness was similar between the three fields with Albion-Scipio field having slightly lower than Napoleon and Stoney Point. Napoleon field had more regular occurrences of thicker cap dolomite than the other fields (Figure 35). The total thickness of dolomitic intervals was greatest in the Albion- Scipio field while Napoleon and Stoney point were similar. Dolomite was thickest in the northern part of the Albion-Scipio field, southern part of the Stoney Point field, and southern part of the Napoleon field (Figure 36). Figure 34. Violin plots of thickness (left), cap dolomite thickness (middle), and dolomite thickness (right) for each field. Battelle | May 2026 43 Figure 35. Bubble map of cap dolomite thickness for each field with Albion-Scipio on the left, Stoney Point in the middle, and Napoleon to the right. Figure 36. Bubble map of the dolomite thickness for each field with Albion-Scipio on the left, Stoney Point in the middle, and Napoleon to the right. Battelle | May 2026 44 The average effective porosity for the TBR interval was highest in the Napoleon field and slightly lower in the Albion-Scipio field, with the Stoney Point field have the lowest average porosity (Figure 37). Dolomitic intervals had higher average porosity in the Napoleon field with Albion- Scipio and Stoney Point showing similar values. Dolomitic porosity was highest in the northern portion of the Albion-Scipio field and southern portion of the Napoleon field. Similarly, the mean permeability of the dolomitic intervals was highest in the Napoleon field with and northern Albion-Scipio field. Figure 37. Violin plots of average effective porosity for the total TBR and dolomitic intervals (upper) and mean permeability (lower). The net thickness was higher in the Napoleon field than observed in Albion-Scipio or Stoney Point (Figure 38). The Napoleon field had frequent wells with high net thickness while Stoney Point had a localized area of higher values. Albion-Scipio was highest in the northern portion. Porosity feet followed the same trend as net thickness. The resulting net to gross ratio was highest in Napoleon field which was closely followed by the Albion-Scipio field. Stoney point had Battelle | May 2026 45 the lowest values and fewer occurrences greater than 20%. The northern portion of the Albion- Scipio had several wells with higher ratios and Napoleon had consistently higher values. Figure 38. Violin plots of the net thickness. The three fields have noticeable differences in reservoir quality. The number of dry holes in the Albion-Scipio field skews the results to be more reflective of non-reservoir intervals. The Napoleon field was developed using modern approaches and technology which resulted in fewer total wells, and fewer percentage of wells in non-reservoir intervals. Stoney Point has poorer reservoir quality across multiple attributes than the other two fields. Vug and Fracture Characterization Vugs and fractures play a critical role in the reservoir development of the HTD reservoirs in Michigan. Secondary porosity, such as vugs and fractures, contributes significantly to the total porosity, permeability, and storage potential throughout the reservoirs. However, these variable features have not been fully understood or predicted. The goal of the vug and fracture flag analysis was to detail the distribution of vug and fracture networks observed in image logs and CT scans and incorporate results into a ML/data analytics workflow for enhanced classification and prediction of reservoir property distributions. ML techniques, such as deep learning (e.g., convolutional neural network) and causality-based algorithms (e.g., Bayesian Networks) will be implemented to identify features of interest such as faults and fractures within the 3D space of the seismic volume and predict features where well data are missing. Additionally, key reservoir facies and characteristics will be predicted and mapped in the 3D space using regression and classification techniques. CT scans from five wells and image logs from 10 wells were selected across the study area (Figure 39). The objectives of this task were to: • Compile a dataset of CT scans and image logs that represent the variability in lithology, secondary porosity features across Albion-Scipio and analogous fields, • Observe secondary porosity features (i.e., vugs and fractures) and trends visible in the core and log data, • Quantify secondary porosity features in CT data using the CTViewer, • Develop vug and fracture flags from core and log data that will be used to train selected wells with ML and data analysis (DA) techniques. Battelle | May 2026 46 Figure 39. Map of study area wells selected with CT data, CBIL, and both. CT Data Overview Detailed two-dimensional (2D) and 3D image analysis of CT scans can provide the necessary data for quantifying and predicting secondary porosity and understanding trends in rock properties. CT scans are performed by emitting an x-ray source on a helical trajectory around longitudinal sections of cores to derive 3D images of density variations for each core volume. Sample densities are expressed in gray-scale color variations such that light-colored, higher- density areas can be distinguished from darker-colored, low-density areas to identify internal fractures, pore distributions/geometries, and silicate versus carbonate minerals. In the past, CT scans were used qualitatively to interpret characteristics in the whole core for each 2D cross-sectional slice via 2D images and descriptions of diagenetic features. The Battelle-developed CT scan analysis tool, CTViewer, processes raw CT scan data to produce not only 2D slices, but also a 3D visualization of the data. Specific rock density ranges can be isolated using CTViewer to quantify the percent of the features of interest in a given interval of core. The tool has provided a platform where CT scan data can be analyzed and quantified in 3D to better characterize the formation rocks of interest. The raw CT data can be supplemented by various logs and core analysis providing data on physical properties such as lithology, composition, and increases and decreases in density and porosity. Images of CT scans are also provided and can be useful for conducting qualitative descriptions and analysis of the core and CT logs. Vug concentration and size distribution, fracture sizes and orientations, occurrences of salt and anhydrite plugging, core density, core porosity, and major changes in lithologies are all major features that can be interpreted from CT scan images. Matrix features such as clasts, laminations, bedding, and features of biologic origin can also be observed with qualitative analysis. Battelle | May 2026 47 Battelle worked with MGRRE and Innova to select 124 ft of TBR core from Albion-Scipio, Napoleon, Stoney Point, and Northville fields. Table 7 details the intervals of CT scan data available for each of the five wells across the TBR fields. Well locations are shown in Figure 39. Table 7. List of core intervals selected for additional data collection. Selected Core Well Field Lithology Key Features Interval (ft) Consumers - Zebra fabric, vugs, Northville 4,421-4,451 Dolomite Nerreter #2 fractures Limestone, Stoney 4,081-4,087; 4,113- Limestone and mixed JEM - Casler #5-30 Dolomitic Point 4,118 lithologies sample Limestone Marathon - Albion- 3,799-3,827 Limestone Limestone sample Wooden #5L Scipio 3,719-3,,722; Ohio Oil Co. - W. Albion- Fractures, vugs, zebra 3,778-3,780; 3,859- Dolomite Rowe No. A-2 Scipio fabric, cyclical porosity 3879 West Bay Expl. - Dolomite, Some Matrix porosity, small Napoleon 4,069-4,099 West Bay #1–22 Limestone vugs, pinpoint porosity NETL conducted core analyses for the five wells, including CT scans and multi-sensor core logging (with X-ray diffraction, P-wave velocity, gamma density, and magnetic susceptibility). Cores were boxed and shipped from MGRRE to NETL for data collection. NETL acquired the CT scan data after core samples were inventoried and loaded into core barrels. Each barrel was scanned once by the low-resolution medical CT scanning machine, resulting in DICOM and tiff stacks with an approximate resolution of 0.1 mm. Concurrent to industrial CT scanning, NETL performed multi-sensor core logging analyses on competent samples to produce raw XRF, p- wave, and gamma density measurements. Approach One of the key goals of the CT scan analysis was to characterize the reservoir heterogeneity with regards to vugs and fractures in the TBR using recovered core. Detailed 2D and 3D image analysis of CT scans can provide the necessary data for quantifying secondary porosity and fracture porosity and predicting reservoir quality. CT scans were collected from core samples from five wells within TBR reservoirs with a variety of reservoir characteristics to ensure a representative dataset. The Battelle-developed CT scan analysis tool, CTViewer, was used to processes raw CT scan data and conduct the vug and fracture analysis. After the raw data were acquired, they were processed using CTViewer for three major features: 1) void space (i.e., vugs and fractures), 2) matrix, and 3) anhydrite plugs. Specific density ranges can be isolated using the tool to conduct image segmentation and quantify features of interest. Thresholds for each feature of interest were identified using the slice-by-slice 2D CT number images (lower CT number values/darker shades are lower density material and higher values/lighter shades are higher density material). CTViewer determines the percentage of each feature in every slice of data and generates a percent threshold versus depth curve that can be exported into Excel. Figure 40 shows an example of the CTViewer interface with a 3-ft section of core loaded and segmented. The features and their corresponding density thresholds, derived from the CT number in grayscale, are listed in the table in the upper right corner of the interface, and the CT image is overlaid with the feature-segmented layer. Below the table is the Battelle | May 2026 48 percent of each feature by depth (left) and a segmented 2D slice down the middle of the core (right). Figure 40. The CTViewer 2D interface with feature-segmented overlay on CT data. The percent threshold curves from CTViewer require manual quality assurance/quality control (QA/QC) to flag erroneous data associated with missing core, the endcaps for each core interval, and drilling-induced fractures to create a secondary porosity curve. During the QA/QC process, flag curves to be used in subsequent tasks were created for vugs and fractures. Two binary flag curves were developed for each well to indicate whether vugs were present and whether fractures were present. Additionally, zero to three fracture curves were developed for each well to indicate if a fracture was present and if the fracture was natural, induced, or if both were present at a depth point. These features were detected and validated using the exported percent threshold curves, the CTViewer interface, and core images. Once the curves were finalized, they were ready to incorporate into statistical correlation with well log and seismic data to aid in training wells to predict reservoir characteristics where data is not available. Full CT core analysis is included in the Task 2 Final Report, included as an attachment to this document. CT Scanning Conclusions Battelle’s CTViewer allowed for qualitative and quantitative CT data analysis of vugs and fractures, resulting in the creation of several vug and fracture flags. These flags assisted with ML and data analytics efforts in subsequent tasks to predict secondary porosity features across TBR reservoirs. The key observations of the CT data analysis are as follows: • The dolomitized core samples from Nerreter #2, Rowe #A-2, and West Bay #1-22 showed more vugs and natural fractures than the limestone wells, resulting in relatively higher secondary porosities in quality checked data. Battelle | May 2026 49 • The limestone wells, Casler #5-30 and Wooden #5, showed few to no vugs and natural fractures compared to the dolomitized wells, resulting in relatively lower secondary porosities in quality checked data. These wells, however, did show high levels of induced fracturing. Core analysis provided evidence of moderate matrix porosity. • The 3D rendering tool was able to detect smaller porosity features not included in flags for both dolomitized and limestone wells. Many of these small features could represent pinpoint porosity or could be associated with larger vugs and fractures where the CTViewer did not detect connectivity. • The CTViewer’s ability to differentiate between void space within the core (i.e., porosity) and on the edge of the core was hindered by effects of core sizes and/or core placement, resulting in difficulty differentiating between vugs and fractures during 3D rendering. Relative secondary porosity curves were created for these wells. Analysis of intensity and size distributions were conducted with uncertainty and supplemented with qualitative analysis. Image Log Analysis Acoustic and resistivity-derived image logs provide high-resolution, 360-degree wellbore images that can be used to derive structural and sedimentary features for detailed reservoir characterization. Planar features such as faults, fractures, and bedding can be identified, as well as stress fields, soft sediment deformation, and pore spaces. Image logs are instrumental in identifying and characterizing vugular porosity in carbonates, as these pore types are difficult to discern from other basic and advanced log signatures. Image logs can also be used to differentiate breakout and drilling-induced fractures from natural formation features. A set of image logs was used from Albion-Scipio and Napoleon fields to characterize the secondary porosity features (vugs and fractures), and to develop flags for machine learning. If successful, the flags would be integrated with the basic wireline log data to develop a predictive model. Data Set and Processing Image log data were limited in the Albion-Scipio field due to the age of the wells. Two were available from non-producing horizontal wells, which cross over from limestone to dolostone and capture changes in the secondary porosity. Napoleon field had 27 image logs collected, most of which did not undergo advanced processing. The Napoleon image logs were reviewed and ranked to identify the top priority wells to be processed by the original vendor. Processing is important to enhance the resolution of the images for more accurate analysis of the secondary porosity (Figure 41). Well records and mud logs were reviewed for each well to describe the type of well, lithology, and features observed. Wells which produced/producing oil, dolomitic, and had noticeable vugs and fractures were given a rating of 1, or high priority. Priority 2 was given to wells which had dolostone present, but could also have limestone, and produced oil. There may or may not have been visible vugs and fractures. Priority 3, the lowest priority, was given to non-producing limestone wells. In total, 11 wells were assigned priority 1, 8 wells were given priority 2, and 8 wells were given priority 3 (Figure 42). The results were reviewed with partners to down select, as highlighted in Figure 42 with red circles. Additionally, an image log was collected on the McCann 1-20 wells (producing/monitoring well for field injection test) and was used in this analysis. Altogether, nine wells were used in the analysis, and are summarized in Table 8. Battelle | May 2026 50 Table 8. Wells with image log data, including depths and formations logged. Well Name Logged Interval Utica Trenton C&L Farms 1-35 4,541-5,813   Curtis 1-32 4,093-4,501   Hilden-Rovsek Etal 1-15 4,250-4,687   Lantis 1-29 4,190-4,760   Potts 1-29 4,295-4,794   Swank 1-22 3,699-4,252   West Bay 1-22 3,750-4,365   Whalen Byron Etal 1-16 3,905-4,433   Figure 41. Example snapshots of image logs before processing (right) and after processing (left) showing improved resolution of the images. Battelle | May 2026 51 Figure 42. Locations of wells with image logs in the Napoleon field, colored by priority. Red circles indicate down-selected wells. Battelle | May 2026 52 Approach Static and dynamic imaging of the logs was done to determine vugular porosity, bedding, fractures, and dip of these features. Each log was analyzed per foot for the entire logged interval. Identified vugs were evaluated by the percentage of vugular porosity (Figure 43) and indicated as 0-5 according to the percentage of vuggy occurrences at the identified depth. Fractures were classified as no fractures, open fractures, partially filled, filled, induced, or unknown (Figure 44). Figure 43. Vug classification based on percentage of vugular occurrences at depth. Unknown (5) indicates a data gap or image distortion. Figure 44. Fracture classification based on open, partially filled, filled, or induced fractures. Unknown (5) indicates a data gap or image distortion interval. Quality image data were determined by evaluating the static and dynamic data as well as the borehole conditions (caliper and gamma ray) of the well. If the static and dynamic image data was clear and the caliper and gamma ray were consistent (borehole conditions) then the image log data were considered adequate for the vug and fracture log analysis. Full image log analysis is included in the Task 2 Final Report, included as an attachment to this document. Battelle | May 2026 53 Summary Fractures in the Trenton are primarily partially opened. Analysis of fracture occurrences by depth across the wells indicates that they typically start at approximately 100 ft below the top of the Trenton. However, the Lantis 1-29 and Whalen Byron et al 1-16 wells have consistent fractures that start from the base of the Utica through the top of the Trenton formation (Figure 45). Swank and Hilden wells have little to no fractures that occur in the Trenton. Wells with fractures are primarily partially filled with little to no filled occurrences (Figure 46). Indications of open fractures are low with the majority occurring in the Utica. The image log measurements did not cover the entire Utica in the wells, so the assessment of the Utica is limited. The Curtis, McCann, Hilden, West Bay, and Whalen image logs all indicate one or more open fractures that measure approximately 0.5 inches in width between 40 to 150 ft from the top of the Trenton formation (Figure 47). Figure 45. Fracture occurrences in the Curtis, Lantis, Swank, and Whalen Byron wells showing fractures at the base of the Utica and top of the Trenton formations. Battelle | May 2026 54 Figure 46. Fracture occurrences across all image log analysis wells, indicating open, partially filled, filled, and induced fractures. Battelle | May 2026 55 Figure 47. Open fractures at the base of the Utica in the Hilden and West Bay wells. Battelle | May 2026 56 Vugs in the Trenton are primarily between 10% and 20% and occur approximately 100 to 200 ft below the top of the Trenton (Figure 48). Most of the wells have few occurrences of greater than 40% vuggy occurrences in the wells. However, vugs occur throughout the Swank and Hilden wells which have little to no fractures in the well. Additionally, wells with a greater occurrence of vuggy percentage over 10% have fewer fractures in the Trenton. Figure 48. Vuggy occurrences across all image log analysis wells, indicating percentage of vuggy occurrences in the well. Two advanced data sets were analyzed to characterize the vugs and fractures in the TBR. This included the use of medical CT scans which provided a 3D representation of selected cores. Thresholds were defined to isolate void spaces and quantify their occurrences. The second dataset included image logs from the Albion-Scipio and Napoleon fields which provided a 2D representation of the logged interval. The image logs were analyzed to identify and characterize fractures and vugs, following a defined classification scheme. Overall, vugs and fracture frequency occurred, with a higher concentration in dolomitic intervals. Battelle | May 2026 57 Seismic Analysis, Albion-Scipio Field This section describes the preliminary seismic analysis of the Albion-Scipio field, which was completed ahead of drilling the McCann well. Seismic analysis at the Lee 26 field, completed by the University of Texas Bureau of Economic Geology, is reviewed below. Project field partner Innova Exploration, Inc. previously collected a high-resolution 3D seismic volume over 13 mi2 in the Albion-Scipio field, covering the planned injection site. A preliminary seismic interpretation was conducted in collaboration between University of Alaska, Innova Exploration Inc., and Battelle to lay the groundwork for more advanced seismic analysis and database integration. Approach The project team first developed a project database including wireline logs and 3D seismic data. The wells were linked to the seismic through a well-tie analysis, which correlates seismic horizons in time to formation tops in depth. Next, key seismic horizons were identified and picked throughout the 3D volume to map the extent and changes in formations in the time domain. Seismic attributes were then explored to determine which attributes might best represent reservoir changes. Finally, a seismic inversion was conducted to invert amplitude to acoustic impedance for a preliminary analysis of reservoir changes. Additionally, a fault and fracture analysis was conducted to better understand the structure in the study area. Seismic Well Tie A seismic well-tie is important to recognize the geologic horizons on seismic data. This is done by generating a synthetic seismogram from the wireline log data. A synthetic seismogram is the result of forward modeling the seismic response based on formation properties such as density and velocity. The Travis 1-18 well in Figure 49 shows an example of the results. Conventional logs were used to generate a synthetic seismogram to recognize the geologic horizons. An Ormsby filter (10/15-50/70-70/90 Hz) was designed and convoluted with reflectivity derived from the density and sonic logs to generate a synthetic seismogram. In this example, no phase rotation was applied. The results show that the Trenton has the strongest positive peak, which is observed in the seismic data. The Trenton reflector was used as the key marker to reference and pick the overlying Utica Shale and underlying Black River, Glenwood, and basement horizons. Figure 50 illustrates the geologic horizons traced based on the correlation to wireline log data. Battelle | May 2026 58 Figure 49. Example results of the seismic well tie for the Travis 1-18 well showing log responses on the left and synthetic seismogram on the right. Figure 50. Example horizons traced using the results from the seismic well tie. Seismic Horizon Interpretation Four seismic reflectors, or horizons, were mapped which included the Utica, Trenton, Black River, and Precambrian Basement. The Trenton was a recognizable reflector due to its strong peak and overall lateral continuity. The subsurface above the Utica seemed laterally homogeneous based on continuous seismic reflectors. However, the TBR interval was complex and showed both vertical and lateral complexity, likely due to the presence of vugs and fractures. Figure 51 illustrates picked horizons on a west-to-east line through the 3D volume. Battelle | May 2026 59 Figure 51. Example horizons picked along a W-E line through the 3D seismic volume. The Utica Shale is shallower in the southern portion of the volume and deepens the north and is continuous across the region, making it an effective confining unit (Figure 52). The Trenton horizon follows a similar trend, shallower in the south and deeper in the north. However, there is more heterogeneity due to the changes in reservoir properties and the influence of tectonic structures (Figure 52). The Black River and Glenwood show similar trends and lateral heterogeneities, likely due to faults (Figure 53). Figure 52. Horizon maps of the Utica (left) and Trenton (right) showing overall trend and heterogeneity. Cool colors are deeper, hot colors are shallower. Battelle | May 2026 60 Figure 53. Horizon maps of the Back River (left) and Glenwood (right) showing overall trend and heterogeneity. Cool colors are deeper, hot colors are shallower. Attributes Seismic attributes are mathematical quantities, geometries, or statistical values derived from seismic data using a set of equations or transformations. Extracting attributes from seismic data allows geoscientists to understand structural, lithological, mineralogical, and other characteristics of the subsurface that are too subtle to be evaluated via simple seismic reflection. There are several types of seismic attributes, of which more than 12 were applied to the seismic volume. This included root mean square (RMS) amplitude, instantaneous phase, frequency, curvature, variance, sweetness, envelope, and ant tracking. Complex trace attributes, such as RMS amplitude, instantaneous phase, and frequency are used to visualize changes in reflection characteristics which could be related to lithology and/or porosity changes. Geometric attributes, such as curvature and variance, reveal structural and stratigraphic changes such as faults. Ant tracking is an additional method that can be applied to the original volume or computed attribute to highlight structural features in more detail. During the preliminary analysis, three attributes were identified which highlighted heterogeneity in the TBR including RMS amplitude, instantaneous frequency, and variance (Ming and Zhao, 2014). The RMS displays amplitude values. This attribute can be used to highlight hydrocarbon indicators or higher porosity intervals. The resulting RMS attribute shows areas of high values towards the north, middle, and scattered throughout the south (Figure 54). In the middle of the field is a structural feature known as the Pulaski break, which is along the high RMS amplitude. Additionally, known faults coincide with the north and south areas of high amplitude. The areas of high amplitude extend away from the faults which could indicate zones of dolomite and/or porosity. Battelle | May 2026 61 Figure 54. Results of the RMS amplitude showing areas of high amplitude along known structural features. High amplitudes are yellow; low amplitudes are red. Instantaneous frequency is the time rate of change of the instantaneous phase. The attribute is a known indicator for low impedance beds, fracture zone, and hydrocarbons. The frequency is lowest in the northern section of the seismic, which correlates with the area highlighted by RMS. However, it does not capture the larger structural features observed at the Pulaski break and the division between the North and Central Scipio pools (Figure 55). Battelle | May 2026 62 Figure 55. Results of the Instantaneous frequency attribute showing low frequency to the north. High amplitudes are cool colors, low amplitudes are warm colors. The variance attribute measures the similarity between waveforms or adjacent traces over given lateral and vertical windows. It can image discontinuity, such as that caused by faulting. The variance is highest towards the middle of the field near the Pulaski break and in scattered intervals throughout the rest of the volume (Figure 56). Figure 56. Variance attribute showing areas of greater variance. Grayscale colors are low variance, warm colors are high variance. Battelle | May 2026 63 Preliminary Inversion Seismic inversion is the process of inverting the seismic amplitude traces to acoustic impedance. Acoustic impedance is a characteristic which represents changes in rock properties more distinctly than regular seismic amplitude. It is the multiplication of acoustic velocity and density of the rocks, resulting in highlighting changes in lithology and porosity. Figure 57 shows the acoustic impedance for the Black River in 3D, with areas of low acoustic impedance or suggested higher porosity (red). Figure 57. Acoustic impedance of the Black River showing areas of low impedance in red which could be associated with high porosity. Fault and Fracture Identification The southern Michigan area has known complex, wrench fault systems. There are primary fault stems that propagate upward from the basin into the TBR, with several branches coming out of the main faults. Manual fault interpretations were conducted along with the use of ant tracking to pick smaller features. Figure 58 illustrates interpreted faults on a west-to-east section and Figure 59 on a north-to-south section with potential sag fill. At least 50 faults were interpreted during this analysis. Few large faults originated in the basement, but most were smaller faults that branched from the main faults. The majority of faults are oriented in the northwest- southeast direction (Figure 60). Battelle | May 2026 64 Figure 58. Interpreted seismic W-E line showing complex fault system. Figure 59. Interpreted seismic N-S line showing complex fault system. Battelle | May 2026 65 Figure 60. Interpreted faults on ant-track slice showing a series of faults throughout the seismic (left), and 3D view of faults (right). Preliminary analyses revealed the complex structural nature of the Albion-Scipio field. Attribute analyses are a promising application to the high-quality seismic volume and will be used to correlate attributes with changes in rock properties such as porosity, vugs, and fractures. The results will be integrated into a 3D static earth model to better predict and understand the changes in the reservoir. Seismic Analysis, Lee Field Seismic analysis at the Lee 26 field was completed by Dr. Shuvajit Bhattacharya at the University of Texas Bureau of Economic Geology, using an 8 mi2 3D seismic dataset provided to the project by West Bay Exploration along with two wells used to generate synthetic seismic ties. Approach The project team first developed a project database including wireline logs and 3D seismic data. Key seismic horizons were identified and picked throughout the 3D volume to map the extent and changes in formations in the time domain. Next, the wells were linked to the seismic through a well-tie analysis, which correlates seismic horizons in time to formation tops in depth. Seismic attribute analysis was explored to determine which attributes might best represent reservoir changes. Finally, a seismic inversion was conducted to invert amplitude to acoustic impedance for a preliminary analysis of reservoir changes. Additionally, a fault and fracture analysis was conducted to better understand the structure in the study area. Battelle | May 2026 66 Seismic Horizon Interpretation Utica, Trenton, Black River, and Glenwood surfaces were interpreted across the seismic volume (Figure 61) and QA/QC was performed using various interpolation approaches (Figure 62), which were also used to complete the horizons in areas of the volume with high complexity or noise. Figure 61. Interpreted seismic section showing interpreted surfaces. Note the Trenton Formation top has a strong peak on seismic reflection (in black), whereas the Utica Formation top has a generally negative peak (trough) characteristic (in white). Figure 62. An example of the Utica surface interpolation and clean up. The first image on the left shows the map before processing, and the one on the right shows the map after processing. Battelle | May 2026 67 Seismic Inversion The surfaces and well data were then used to generate a low-frequency model of the subsurface, followed by a post-stack seismic inversion to derive P-impedance (Figures 63 and 64). Figure 64 shows the relationship between inversion-based impedance and log-based impedance. One well was used for training and another for blind testing in the inversion process. The relationship between well-log-based porosity and P-impedance was used to convert the seismic-derived impedance to porosity (Figure 65). Figure 63. An example of a single well-based inversion. The blue curve in the 1st track shows original impedance, and the red curve shows the seismic-inverted impedance. The error, the difference between the seismic trace and synthetic, is minimal (~0.07). Battelle | May 2026 68 Figure 64. P-impedance from seismic inversion in the Lee field. There are two wells inserted in the cross- section. Inversion-based impedance matches with well-log-based impedance. Impedance curves in both wells and seismic have the same scale and color pattern for Figure 65. Comparison between well-log-based impedance and seismic inversion-based impedance, showing overall good match. Battelle | May 2026 69 Seismic Porosity Estimation The seismic-based to log-based impedance relationship was used to produce the final porosity volume from seismic inversion in the Lee field. Because the porosity log curves provided did not correct for clay bound water effect in the Utica Shale, artificially high porosity of >10-16 porosity units (p.u.) was displayed in some places, whereas the Trenton Formation showed negligible porosity in places. Importantly, the TBR section was parameterized in the seismic inversion process, while the Utica overlying section was not. Therefore, a filter was applied in the Utica and above to address its apparently high porosities relative to the TBR section. This helped better visualize reservoir property variation in the TBR section. Figure 66. Estimated porosity from seismic inversion in the Lee field. The possibility of porosity prediction from multiple seismic attributes was also explored using a probabilistic neural network (PNN). Multiple seismic attributes were used as inputs to the PNN model, such as average frequency, coherence, and seismic data decomposed in different frequency bands, as well as inversion-based P-impedance to predict porosity (target). However, the results from PNN-based porosity prediction were poor after several iterations, with a correlation coefficient of 0.63. Lee Field Seismic Analysis Conclusions This subsurface characterization workflow revealed the structural and reservoir complexities in the TBR in both fields, which are critical for understanding potential sweet spots for CO2-EOR deployment. Seismic inversion-based impedance highlighted the potential reservoirs. It is anticipated that the workflow and lessons learned will lay the groundwork for future oil and gas and CO2-EOR-related work in these HTD reservoirs in Michigan. 3.2 Subtask 2.2: Risk Assessment McCann Well Pre-drill Risk Assessment For the McCann well drilling, a preliminary site-based risk assessment was completed to identify potential items that may affect project performance, safety, and/or the environment for a Battelle | May 2026 70 chemically enhanced CO2-EOR project in the Albion-Scipio field. The risk analysis addressed overall project performance risk factors, wellbore integrity concerns, and the identification of potential geohazards in the area such as seismicity or environmental concerns. The preliminary assessment focused on the immediate area being considered for the project test at a scale of approximately 15,000 metric tons CO2 injection. A summary of the risk assessment objectives is presented in Table 9. An expanded assessment was also completed to evaluate risk factors related to larger scale application of the EOR technology for the larger Albion-Scipio trend in southern Michigan. Table 10 summarizes conclusions of the preliminary risk assessment. Overall, no high-impact items were identified in the risk analysis. The caprock over the Trenton reservoir is thick and impermeable, which provides containment for upward migration of injected chemically altered CO2. Some fractures near the contact of the Trenton and Utica rock formations should be considered. No widespread or prevalent well construction or well plugging deficiencies were present in Albion-Scipio wells. However, the risk assessment did find that some wells had incomplete records or a limited number of smaller cement plugs. For example, three wells were identified in the project area with marginal plugging and abandonment records that may need consideration if they intersect the reservoir. A more detailed inspection of well conditions and/or testing of selected legacy wells may be required for expanded EOR operations. Subsurface stress conditions indicate a mainly strike-slip regime with stress magnitudes typical for the general region. If very high injection pressures are used for EOR operations, the fractured nature of the CO2-EOR reservoir presents risks related to EOR performance and geomechanical deformation. Results of the risk assessment may be used to guide site characterization, injection testing design, monitoring, and evaluation of further applications of chemically enhanced CO2-EOR along the Albion Scipio, and potentially across the TBR play in southern Michigan. Table 9. Summary of risk assessment objectives. Item Objectives Methods Site wellbore Determine condition of wellbores in the Review of well records, construction integrity review field site area to evaluate potential for CO2 specifications, geology in field area migration along legacy oil and gas wells Site cement Evaluate cement quality for wells in the Complete cement bond log analyses bond log field test site area for wells within the study area of the analysis Albion-Scipio field Site risk Identify any risk items that may be Conduct risk screening-based screening necessary to address in site analysis (FEP) and activity-based characterization, monitoring, system “what if” analysis of technical risks design, and safety systems Site geohazard Identify any surface features and geologic Review of potential geohazard related review features that may be affected by CO2-EOR risks such as seismicity and system or field activities environmental concerns within the study area Scale-up flow Evaluate risks related to potential flow Describe caprocks, wellbores, faults, pathway analysis pathways for CO2 migration in TBR play in fractures pathways for CO2 migration Southern Michigan Scale-up Investigate impact of CO2 injection on Estimate subsurface stress conditions fracture fractured reservoirs in TBR in Southern in TBR reservoir based on image log pressure Michigan analysis, compare results to CO2 analysis injection pressures The full risk assessment approach, data analyses, results, and documentation of results are Battelle | May 2026 71 included as an attachment to this final report. Table 10. Summary of preliminary risk assessment conclusions. Item Objectives Conclusions Wellbore integrity Determine condition of wellbores in Wells were built in 1960s with two review the field site area to evaluate casing strings, all wells plugged and potential for CO2 migration along abandoned, wells API 2107522652, legacy oil and gas wells 2107526713, and 2107521273 had two plugs or unknown plugging details. Cement bond log Evaluate cement quality for wells in Many wells had inadequate CBLs, the analysis the field test site area wells with modern CBLs had cement rating of 41-100% and all wells had intervals of 95-100%. Site risk screening Identify any risk items that may be No major risk items were identified in necessary to address in site the FEPs screening. The fractured characterization, monitoring, system matrix nature of the reservoir was design, and safety systems identified for CO2-EOR performance risk. The activity-based risk screening identified typical drilling activities. Geohazard review Identify any surface features and Fractured reservoir has geomechanical geologic features that may be risks related to injection. The test area affected by CO2-EOR system or is mostly cropland, pasture, and mixed field activities. forest with wetlands. Expanded Risk Review caprock, well integrity, fluid Thick Utica caprock is present in the Analysis for TBR migration pathways, and subsurface TBR play, there may be more frequent Play in southern stress conditions in southern fractures at the Trenton-Utica contact, Michigan Michigan TBR some wells have fewer plugs or incomplete well construction records, subsurface stress conditions are typical strike-slip for general region 3.3 Subtask 2.3: Advanced Field Characterization Fischhaber 2-35 Wellbore Integrity Evaluation Ahead of the CO2 injection campaign, to ensure that there was no possibility of underground CO2 leakage from the operation, Battelle and West Bay concluded that it was necessary to evaluate the condition of the Fischhaber 2-35 wellbore. To understand the current corrosion status of the Fischhaber 2-35 casing, DarkVision’s Hades-R and Hades-T microacoustic casing analysis tools were run, which provide sub-millimetric radial measurements of casing thickness and shape (Figure 67). These tools were run from the base of the well at 4,580 to 2,925 ft measured depth. Six casing joints with Class 1 or Class 2 features were noted (Figure 68), and based on DarkVision’s analysis, none of these features constitutes a significant risk for later corrosion breach or reduces the well’s maximum operating pressure enough to warrant concern. One casing joint was noted with full breach; these are the production perforations and do not constitute a risk. One item of note is that the Fischhaber 2-35 completion report indicates that perforations should exist from 4,551 to 4,561 ft measured depth, however, perforations appear to be missing from 4,555 to 4,560 ft measured depth. This reduced perforation coverage is not believed to be an issue; the existing perforations should be sufficient to deliver CO2 to the reservoir. A follow-up run for these tools is planned for approximately six months into the well production period to evaluate the impact of CO2 corrosion within the wellbore. Battelle | May 2026 72 Figure 67. Description of DarkVision's Hades-R and Hades-T casing evaluation tool capabilities. From Dark Vision. Figure 68. Casing wall loss features identified by Hades tools. None of these features is a concern for wellbore integrity. Spectral Pulsed Neutron (SPN) Logging In December 2024, ahead of CO2 injection into the Fischhaber 2-35, Baker-Hughes ran its SPN tool to evaluate saturations of oil, gas, and water in the pore space surrounding the well (Figure 69). The tool was run successfully from 4,580 to 4,300 ft. The data established a baseline against which data collected after CO2 injection can be compared. Battelle | May 2026 73 Figure 69. Wireline logging at the Fischhaber 2-35 well pad during December, 2024. Follow-up logging was completed between June 19 and 22, 2025. Initial, minimally processed logs were delivered on June 25, 2025, and indicate changes in oil, gas, and water saturation within the reservoir as compared to the baseline logs gathered in December, particularly in the portion of the reservoir immediately surrounding the perforations. Baker Hughes delivered final logs with detailed modeling, interpretation, and reporting on August 27. The primary conclusions of this time-lapse logging effort are as follows: 1. Differences in wellbore completion (caused by well control concerns) between the 2024 and 2025 logging runs significantly impacted the ability to compare data between the pre-injection and post-injection SPN runs (Figure 70). a. 2024 logging was completed with production tubing in the well and water occupying the casing annulus and tubing. b. 2025 logging was completed without tubing and with the wellbore loaded with diesel to maintain well control. The lower portion of the wellbore, below 4,555 ft measured depth, was filled with reservoir water. c. Baker-Hughes’ wellbore modeling struggled to resolve the differences in metal mass between the two runs, resulting in significant uncertainty in calculated saturation. d. Differences in wellbore fluid between the two runs also resulted in uncertainty, particularly in the carbon/oxygen measurements. e. There are indications that CO2 was bubbling up through the fluid column during the follow-up run, adding uncertainty to density measurements. f. Future efforts should plan for a wellbore configuration that can be replicated both before and after CO2 injection, regardless of differences in reservoir pressure and fluid saturations. Battelle | May 2026 74 2. No significant change in reservoir fluid saturation was detected in either the vuggy or matrix porosity systems. a. There does not appear to be significant differences in gas saturation values between the 2024 and 2025 runs. b. Elevated gas saturation detected below the lowest perforations are likely due to differences in wellbore configuration. c. The reservoir directly adjacent to the well perforations was strongly saturated with CO2 after injection. Lack of gas saturation observed by SPN logging is due to the extended soak period between injection and logging. d. It is not clear why there are no indications of increased gas saturation in the upper parts of the reservoir. 3. Low porosity values throughout the reservoir resulted in high uncertainty when modeling the logged response, leading to low confidence in the output. a. Baker-Hughes recommends the suite of tools be used in reservoirs with greater than 10% average porosity. GasView has acceptable uncertainty down to 5% porosity, but the tool’s vertical resolution is too low to effectively resolve the vuggy intervals within this reservoir. Battelle | May 2026 75 Figure 70. Carbon/Oxygen log comparison between 2024 and 2025 logging run. Depth range is 4,500- 4,616 feet measured depth, covering the primary reservoir zone. Note hardware effects from tubing in the well in 2024 (tracks 3-6) as well as transitional effects from water-diesel interface in 2025 log (tracks 3&7). Battelle | May 2026 76 3.4 Subtask 2.4: Integrated Physics-Based Machine Learning and Advanced Data Analytics The data integration, ML, and data analytics subtask aimed to integrate all datasets and apply ML techniques to predict key reservoir properties and compare different TBR fields using a workflow for enhanced classification and predictions. The first part of the task focused on wireline log data from the Albion-Scipio field in southern Michigan, then integrated other data types to build predictive models. The key objectives included: 1. Preprocessing of the wireline log LAS files a. Conversion to consumable formats for Python to process b. Linking formations to each depth in all wells 2. Exploratory data analysis of the wireline log data for the Albion-Scipio field a. Remove observations/rounding of log data which fall outside of geologic constraints b. Checking histograms and distributions of the cleaned dataset to ensure geologic soundness 3. Identification of electrofacies a. Use cluster analysis approaches including K-Means and hierarchical clustering 4. Imputation of missing data a. Determine which logs have highest correlations to others b. Determine evaluation metrics for models being used to impute missing data c. Compare distributions between imputed and observed data 5. Core data integration and exploration 6. Vug and fracture data integration and exploration 7. Build predictive models for sparsely sampled reservoir characteristics using machine learning a. Includes permeability, fractures, vugs, and seismic attributes There are many challenges when evaluating carbonate reservoirs, especially when they have undergone significant diagenetic alteration. This introduces complex heterogeneity which cannot always be captured by routine data and analysis. Additionally, the vintage of the data in the Albion-Scipio fields adds another challenge when trying to link wireline log signatures to predict reservoir features. Because of this, the predictive models for permeability and vugs and fractures were not successful. However, the task yielded important results which will be useful in upcoming tasks to model the reservoir in 3D. This includes the classification and prediction of electrofacies, distributions of porosity and permeability by electrofacies, estimated permeability for each well, and the frequency of secondary porosity features observed by electrofacies. Moving forward, the data will be integrated with the 3D seismic to link seismic attributes to areas of enhanced porosity and permeability to inform the modeling efforts. Full reporting on the ML and advanced data analytics efforts are provided in the Task 2 Final Report as an attachment to this report. Battelle | May 2026 77 4.0 Task 3: Laboratory Experiments The pore structure of the TBR is complex and the rock may be oil-wet/mixed-wet. Laboratory experiments were conducted at the University of Texas (UT) Austin to optimize CO2-EOR fluids to be effective in this complex carbonate reservoir. One reservoir core was characterized by CT scanning, mercury porosimetry was completed on two cores, wettability testing was completed on three cores, and relative permeability was completed on two cores. Slim-tube displacements were conducted to determine the minimum miscibility pressure (MMP) with CO2. CO2 and Water-Alternating-Gas (WAG) floods were conducted to determine the efficiency of the CO2 flood. Foam floods with wettability altering agents were conducted on three cores to see improvement in CO2 flooding with foam. These laboratory experiments were modeled using a compositional simulation approach, generating a calibrated model that was used to inform the design of the chosen EOR process to the field pilot scale. Results from these experiments are summarized below; the full study report from UT Austin is summarized in Attachment B and provided as an attachment to this report. 4.1 Subtask 3.1: Core Characterization The University of Texas completed a laboratory test to determine petrophysical characteristics of reservoir core samples. The project team obtained 13 1.5-inch diameter core samples (Table 11) from the Napoleon field and performed four tests to better understand the conditions in the reservoir: (1) CT scan, (2) mercury porosimetry, (3) oil/water relative permeability test, (4) imbibition tests. Table 11. Basic information of reservoir core samples from Napoleon field Sample Depth Length Diameter Mass (g) Air permeability No. (ft) (cm) (cm) (mD) 1 1-5 4,063.6 4.14 3.81 124.6 11.0 2 1-24 4,082.95 4.17 3.81 103.0 200.9 3 1-31 4,089.1 4.04 3.81 114.1 111.4 4 1-40 4,098.65 4.11 3.81 115.1 42.5 5 1-45 4,103.95 4.2 3.81 121.3 61.1 6 1-50 4,108.7 4.12 3.81 110.9 185.0 7 6-41 4,300.2 4.02 3.81 116.4 5.0 8 6-42 4,301.2 4.05 3.81 117.3 9.0 9 6-43 4,302.1 4.04 3.81 119.6 10.1 10 6-44 4,303.8 4.05 3.81 104.6 67.2 11 6-48 4,307.95 3.96 3.81 112.6 29.6 12 6-51 4,310.15 4.00 3.81 115.6 10.8 13 6-54 4,313.6 4.00 3.81 111.5 91.5 Battelle | May 2026 78 CT Scan Core Sample #6-44 was scanned in this test to identify the vug and fracture content of the core sample using a micro-CT with a resolution of 34.6 μm (Figure 71). The conclusion drawn from this test was that there are many vugs of varying sizes but no obvious signs of fractures in the cores that were scanned. Figure 71. CT-scan of core sample #6-44 showing the largest pore present on the sample which measured 11.102mm. Mercury Porosimetry Two core samples, #1-40 and #6-41, were used for mercury porosimetry. The air permeability of #1-40 was 34.6 mD and the air permeability of #6-41 was 3.7 mD. Intrusion volume of mercury versus applied pressure was obtained, and the pressures were converted into pore throat sizes. Figure 72 shows the mercury porosimetry curves for core sample #1-40, including the cumulative intrusion and incremental intrusion versus pore throat diameter. It indicates that core sample 1-40 has a large set of vugs that are accessed at a pore throat size of 15 μm. For the rest of the pores, it has another dominant pore throat size, 81 μm, which might be attributed to vugs present in the core sample. Figure 73 shows the mercury porosimetry curves for core sample #6-41. It has one dominant pore throat size, 26.9 μm; core sample #6-41 is more homogeneous than core sample #1-40. Figure 74 shows the capillary pressure versus mercury saturation for these two core samples. For mercury saturation of less than 0.75, the two curves are similar. The capillary pressure curve of #1-40 is higher than that of #6-41 at higher mercury saturation, which means that core sample #I-40 has many smaller pore throats and is more heterogeneous at the pore scale. Battelle | May 2026 79 Figure 72. Mercury porosimetry curves for core sample #1-40 Figure 73. Mercury porosimetry curves for core sample #6-41 Figure 74. Capillary pressure versus mercury saturation for core samples #1-40 and #6-41 Battelle | May 2026 80 Oil/Water Relative Permeability Test One composite core mount was used to determine oil and water relative permeabilities by the unsteady state method. The core mount was composed of three core samples, #1-54, #1-45, and #6-48. The properties of this core mount are shown in Table 12. Its absolute permeability is 53 mD, which is close to the real reservoir average permeability. The properties of crude oil and formation brine are shown in Table 13 and Table 14, respectively. Table 12. The properties of the core mount Core Mount1: 6-54, 1-45, 6-48 Length (cm) 12.16 Diameter (cm) 38.1 Mass (g) 345.4 Absolute permeability (mD) 53 Porosity (%) 11 Initial oil saturation (%) 70.3 Table 13. The properties of crude oil Crude oil Napoleon oil Gravity (API°) 42.6 Viscosity (25°C, cp) 5.5 Viscosity (40°C, cp) 4.25 Density (25°C, g/ml) 0.805 Molecular weight (g/mol) 202.21 Table 14. The properties of formation brine Brine Density (25°C, g/ml) 1.16 pH (25°C) 6.38 Viscosity (25°C, cp) 1.66 Viscosity (40°C, cp) 1.26 TDS (mg/L) 247,526.5 Cations (mg/L) Na+ 65,307.84 Ca+2 20,800 Mg+2 6,100 Anions (mg/L) Cl- 155,245.66 HCO3- 73 In the unsteady state method, the core mount is first saturated with crude oil at the irreducible water saturation and aged at 100°C for two weeks. Next, crude oil is injected into the core mount at several different flow rates to calculate the oil permeability at the irreducible water saturation. Then, formation brine is injected into the core mount at a constant flow rate of 0.15 ml/min (5 ft/day) to displace crude oil. The effluents coming from the core mount and the Battelle | May 2026 81 pressure difference between the inlet and outlet are monitored during the process. Finally, the water relative permeability at the residual oil saturation is calculated, when no more oil comes out. The calculated results are shown in Table 15. Figure 75 shows the oil-water relative permeability curves from the unsteady state experiment. It should be noted that the relative permeability curves are obtained over a limited saturation range of 0.458≤Sw≤0.64. The relative permeability curves between Swirr=0.298 and Sw = 0.458 cannot be obtained from the unsteady state method. The Corey-correlation (Equations 1 and 2) is used to match the experimental results. The parameters in Corey-correlation are shown in Table 15 and the Corey type relative permeability curves are shown in Figure 75. The oil relative permeability curve is matched very well; however, the water relative permeability curve match is worse. Table 15. Calculated results for oil and water relative permeabilities PVI Vo (ml) Vw (ml) △P (psi) Sw kro krw 0.2980 0.83396 0.00000 0.36 0.30 1.20 5.21 0.4591 0.05191 0.13702 0.46 0.10 1.40 4.91 0.4643 0.04505 0.14893 0.55 0.10 1.40 4.87 0.4697 0.04010 0.15858 0.75 0.20 3.10 4.58 0.4811 0.03334 0.17375 1.14 0.30 5.70 3.76 0.4983 0.02561 0.19509 1.76 0.30 9.40 3.64 0.5154 0.01936 0.21736 2.35 0.10 8.80 3.56 0.5203 0.01599 0.23240 2.94 0.10 9.10 3.43 0.5256 0.01375 0.24413 3.54 0.10 8.80 3.3 0.5311 0.01214 0.25375 4.13 0.10 8.90 3.24 0.5368 0.01091 0.26190 4.72 0.10 8.90 3.12 0.5425 0.00994 0.26897 5.31 0.30 8.60 3.06 0.5615 0.00915 0.27520 5.90 0.20 8.80 3.05 0.5740 0.00849 0.28079 7.67 0.20 26.80 2.92 0.5856 0.00704 0.29466 9.44 0.20 26.80 2.84 0.5975 0.00606 0.30561 12.06 0.20 39.40 2.75 0.6092 0.00506 0.31847 13.83 0.10 26.90 2.64 0.6150 0.00457 0.32560 15.60 0.10 26.90 2.65 0.6208 0.00418 0.33183 16.49 0.01 13.40 2.57 0.6212 0.00401 0.33467 The water relative permeabilities are > 0.1; the end point relative permeability is 0.58. The oil relative permeabilities are low, <0.1 and the crossover saturation is below 0.5. The relative permeability data shows that the core in oil-wet or mixed-wet. The oil exponent is 4.3 and the water exponent is 2.8, although curve fit is relatively weak. Battelle | May 2026 82 Figure 75. Oil-water relative permeability curves. 4.2 Subtask 3.2: MMP Determination MMP is defined as the lowest pressure at which the injected gas and the reservoir oil become miscible through a multi-contact process at the reservoir temperature. Most CO2 floods are conducted at pressures greater than MMP to maximize oil interaction with CO2, maximizing oil swelling, minimizing oil viscosity, and improving sweep efficiency. MMP was determined for TBR reservoir oils using both statistical correlation and laboratory analysis. These methods are described below and returned MMP values that are in strong agreement. Statistical Correlation MMP for TBR crude oil was calculated using three statistical correlations developed by Valluri et al. (2017), Cronquist (1978), and Yellig and Metcalfe (1980). This analysis required reservoir temperature and the API gravity of the crude oil, which were obtained from a report written by Brannigan et al. (1960) that contained information on three samples collected from the Albion, Pulaski, and Scipio fields. Table 16 lists the properties of the three crude oil samples used for the analysis and Table 17 shows the results of the statistical MMP analysis. Table 16. Crude oil sample data used for MMP calculations (after Brannigan et. al., 1960). Sample Parameter Albion Sample Pulaski Sample Scipio Sample Date 5/1/1959 10/22/1959 8/18/1958 Formation Trenton Black River Trenton Depth (feet) 3,390 4,100 3,800 Bottomhole pressure (psi) 1,737 1,732 1,542 Bottomhole temperature (deg. 101 106 107 F) Oil API gravity 41.2 41.4 41.9 Battelle | May 2026 83 Table 17. Statistical MMP values for the three samples. MMP (psi) Oil Sample Yellig and Metcalfe Valluri et. al. (2017) Cronquist (1978) (1980) Albion 1,255.0 1,103.6 1,216.0 Pulaski 1,323.8 1,159.2 1,309.8 Scipio 1,312.5 1,148.5 1,309.8 Laboratory MMP Determination Three different kinds of oils were used to determine MMP, with properties shown in Table 18. These fluids were freshly sampled from active wells in the trend for laboratory analysis. The sample termed “Michigan oil” was taken from the Travis 1-18 well in the Albion-Scipio field in November of 2019; the “Napoleon oil” was taken from the West Bay 1-22 well in the Napoleon field in mid-2021. Finally, the “McCann oil” was sampled from the McCann 1-20 well drilled during project execution in early 2021, also in the Albion-Scipio field. Slim tube experiments were conducted to determine MMP for these oils. The slim tube setup comprised a 20 ft coiled slim tube made of stainless steel with an inner diameter of 0.26 inches and thickness of 0.063 inches which was packed with 100-150 mesh sand. System porosity was 37.4%, and tests were run at pressures of 1,000 to 2,300 psi and temperature of 40°C, consistent with reservoir environments. Full experimental setup is shown in Figure 76. The effects of CO2 additives were also studied by slim tube experiments. Two additives from Sasol Company, ME-1 and ME-2, were tested for lowering MMP. The slim tube experiments were conducted with dead oil and the additives were mixed with CO2 separately prior to the experiment. Table 18. Properties of crude oils Battelle | May 2026 84 Figure 76. Schematic of the set up for MMP determination apparatus using slim tube. Results (1) Michigan Oil (Albion-Scipio Field) Slim tube experiments were conducted with Michigan oil at seven different pressures. The 1.2 pore volumes injected (PVI) recovery factors are plotted as a function of pressure in Figure 77. Oil recovery increased with pressure. The first five oil recoveries (1,000 to 2,000 psi) were lower than 95%. However, the last two recoveries (2,100 and 2,300 psi) surpassed 95% with a slight difference. Therefore, MMP was determined as 2,100 psi at 40°C. (2) Napoleon Oil (Napoleon Field) Slim tube experiments were conducted with Napoleon oil at three different pressures, and the corresponding oil recovery factors are presented in Figure 78. At 1,100 psi, the oil recovery was 78.3%, which is below 95%, indicating an immiscible flooding. At 1,300 psi, the oil recovery notably improved to 93.4%. The increase of pressure to 1,500 psi resulted in an oil recovery of 94.6%, which was close to the previous one. Therefore, the MMP for Napoleon oil and CO2 is approximately 1,300 psi at 40°C. (3) McCann Live Oil (Albion-Scipio Field) Slim tube experiments were performed using McCann live oil, composed of McCann dead oil combined with methane at a gas-oil ratio (GOR) of 220 scf/bbl. Results are shown in Figure 79. At 1,100 psi, the oil recovery was 68%, indicating an immiscible flooding. At higher pressures (1,300 psi, 1,500 psi, and 1,700 psi), the oil recovery was higher than 95%, indicating a shift to a miscible flooding. Consequently, the MMP for McCann live oil and CO2 was determined as 1,300 psi at 40°C. Battelle | May 2026 85 Figure 77. The relationship between oil recovery and pressure (Michigan oil) Figure 78. The relationship between oil recovery and pressure (Napoleon oil) Battelle | May 2026 86 Figure 79. The relationship between oil recovery and pressure (McCann live oil) Additive Results Four separate slim tube experiments were conducted using ME-1, with the additive combined with CO2 at between two weight percent and five weight percent, and then injected together into the slim tube at 1,750 psi. Oil recovery results are shown in Figure 80. These results suggest that the introduction of additive ME-1 did not yield an improvement in oil recovery, thus no improvement in miscibility performance between oil and CO2. A single slim tube experiment was performed using ME-2 as an additive. Five weight percent ME-2 was combined with CO2 and injected into the slim tube to displace oil. The resulting oil recovery, shown in Figure 81, was 84.7%. There is a 2.2% decrease compared to 86.9% recovery achieved by CO2 flooding without any additives. Hence, the introduction of ME-2 as an additive did not contribute to an increase in oil recovery at the tested pressure of 1,750 psi. Slim tube experiments were conducted for reservoir oil with CO2. The following conclusions can be drawn from the results. • The MMP for CO2 with Michigan oil was 2,100 psi at 40°C. • The MMP for CO2 with Napoleon oil was 1,300 psi at 40°C. • The MMP for CO2 with McCann live oil was 1,300 psi at 40°C. • Additives ME-1 and ME-2 cannot contribute to an improvement in miscibility performance between Michigan oil and CO2. Battelle | May 2026 87 Figure 80. The oil recovery at 1,750 psi at different situations Figure 81. The oil recovery at 1,750 psi with or without additive ME-2 4.3 Subtask 3.3: CO2 Core Flood The objective of this task was to assess the efficiency of CO2 flooding in EOR. To investigate the effectiveness of CO2 flooding for EOR, core flooding experiments were conducted at a pressure of 1,200 psi and a temperature of 40°C using the setup shown in Figure 82. Full Battelle | May 2026 88 methodology is included in Appendix to be determined. Both continuous CO2 flooding and WAG flooding were simulated. Figure 82. Schematic of the set up for CO2 flooding experiments Continuous CO2 Flooding Figure 83 shows the oil recovery, water cut, and relative pressure drop data versus pore volume injected (PVI) for continuous CO2 flooding. The experimental procedure involved injection of four pore volumes (PV) of simulated pore water (SPW), followed by 0.5 PV of CO2, and then another 1.5 PV of SPW injection. During the water flooding, breakthrough of water occurred at 0.16 PVI, and a total of 36.5% of the original oil in place (OOIP) was recovered at the end of water flooding. Subsequently, during the injection of CO2, breakthrough of CO2 was observed at 0.1 PVI of CO2 injection. At the end of the experiment, the incremental oil recovery was 14.4% OOIP, which was attributed to the multi-contact miscible behavior of CO2 and oil. Not all of the oil was recovered because of the viscous fingering and high-level heterogeneity of the core sample. The relative pressure drop was calculated based on the pressure drip generated by single-phase oil flow at the same flow rate of 1 ft/day. The relative pressure drop during CO2 flooding was found to be small, less than 1 psi, indicating that the CO2 tended to bypass the smaller pores in the core samples. Battelle | May 2026 89 Figure 83. Oil recovery, water cut, and relative pressure drop for the continuous CO2 flooding experiment. Water Alternating with Gas Flooding Figure 84 shows the data on oil recovery, water cut, and relative pressure drop versus PVI during WAG flooding. The water flood was followed by a WAG flood with the same amount of CO2. The water flood recovered 42.5% OOIP, higher than that in water alone. The tertiary WAG flood yielded an additional 20.1% oil, indicating that the WAG flood could improve oil recovery over the continuous CO2 flood in this core mount. Moreover, breakthrough of CO2 was observed at 0.25 PVI of WAG injection, and the relative pressure drop during WAG injection was higher than that of CO2 flooding, which indicated that WAG could help control CO2 mobility. Battelle | May 2026 90 Figure 84. Oil recovery, water cut, and relative pressure drop for WAG flooding experiment Continuous CO2 flooding with 0.5 PV CO2 injection could improve 14.4% OOIP over water flooding. WAG injection could help control CO2 mobility and increase the oil recovery to 20.1% OOIP. 4.4 Subtask 3.4: CO2 Foam Flood The University of Texas developed and conducted several foam stability and rheology formulation tests to determine the preferential alteration and wettability of rock to water-wet conditions over oil-wet conditions for EOR for the Lee 26 field. Foam stability tests were conducted to observe and determine foam performance, while rheology tests were used to determine the steady state of foam and the apparent differential flow rate viscosity. The testing parameters used included: • Reservoir Temperature = 40° • Pressure = 1,200 psi • Surfactant Concentration = 0.5 wt% • Foam Quality (gas volume fraction) = 80% • Foam Generation:  Pre-flush surfactant solution into the system and increase the pressure to 1,200 psi  Inject surfactant and CO2 simultaneously (qsurf = 0.4 cc/min, qCO2 =1.6 cc/min) Foam stability surfactant tests focused on three surfactants at weight concentrations of 0.5% (Soloterra843, CETAC-30, and LMDO). Each surfactant was tested under four conditions, 1) saltwater density of 30,000 parts per million (ppm) (SWD1), 2) saltwater density 165,500 ppm Battelle | May 2026 91 (SWD2), 3) SWD1 plus oil, and 4) SWD2 plus oil, to determine foam stability. Soloterra843 showed a breakdown of stability by more than 50% of foam volume at 10 minutes in a SWD1 and oil solution and complete breakdown at 40 minutes after solution injection (Figure 85). The CETAC-30 surfactant had significant foam generation in all four solution scenarios (Figure 86). The half-life results showed that foam breakdown of 50% was reached after 30 minutes and 80% after 90 minutes. In the higher salinity solution plus oil scenario, foam breakdown of >70% occurred after 30 minutes and was >85% after 60 minutes (Figure 87). LMDO tests showed that slightly less foam was generated than the CETAC-30 surfactant (Figure 88). Figure 85. Soloterra843 (0.5wt%) surfactant, brine solution (30,000 ppm), and oil foam stability test. Foam breakdown >50% occurs after 10 minutes. Figure 86. CETAC-30 (0.5wt%) surfactant, brine solution, and oil foam stability tests. Battelle | May 2026 92 Figure 87. LMDO (0.5wt%) surfactant, brine solution, and oil foam stability tests. Figure 88. Graph showing surfactant stability half-life results in brine only and brine with oil solutions. Soloterra843 had the least foam stability in higher density brines. LMDO showed greater stability in higher density brines. CETAC-30 is comparable to LMDO, however, there is less stability in higher density brines. CETAC-30 and LMDO showed the best foam generation in testing. However, CETAC-30 optimized pore surface wettability. Data from the foam stability tests were used to compare pure CO2 flooding versus surfactant and CO2 foam flooding to determine the most effective oil recovery scenario. Approximately 50% of the oil recovery resulted from a pure CO2 flood (Figure 89). However, the CETAC-30 surfactant with CO2 foam flood provided approximately 13% better oil recovery than pure CO2 (Figure 90). Battelle | May 2026 93 Figure 89. A model of a pure CO2 flood to determine approximate oil recovery in the Lee 26 field. Figure 90. A model of a CETAC-30 and CO2 flood to determine approximate oil recovery during a foam flood scenario. Methanol effect on foam stability was also tested at room conditions using CETAC-30 surfactant (0.5 wt%). Foam stability of the surfactant in the presence (5 wt%) and absence of methanol was investigated. Foam stability and foamability of the generated foam was investigated in each scenario as the main criteria for evaluation. Foam stability was indicated by the foam half-life, which is the time required for the foam to reach half of its original height. Foamability was quantified by the initial foam height. Table 19 shows the material and conditions during this project. Table 19. Material and conditions Parameter Value Temperature (°C) 20 Pressure (psi) 14.7 Gas Air Surfactant CETAC-30 Water DI water Foam quality (%) 80 Battelle | May 2026 94 Fifteen centimeters (cm) long and one centimeter diameter glass vials were filled with three milliliters (ml) of the surfactant solution prepared in deionized (DI) water. 0.5 wt% CETAC surfactant was utilized. Two cases were investigated. The first one without adding methanol and the second one with adding 5 wt% methanol to the aqueous solution. Finally, foam was generated by shaking the vials vigorously. Both foam stability and foamability were measured for both cases. Figure 91 shows the foam evolution versus time for both foam cases in the presence and absence of methanol. Figure 92 further illustrates both cases showing a higher foamability (12.6 cm) attained in the absence of methanol. In addition, foam stability was much longer with a half- life of 58 minutes. On the other hand, adding 5 wt% of methanol resulted in a lower foamability of 9.3 cm. Moreover, the foam stability was significantly reduced with a half-life of only seven minutes. Figure 91. Foam evolution versus time for both foam cases in the presence and absence of methanol Figure 92. Foam height versus time for both foam cases in the presence and absence of methanol Battelle | May 2026 95 The conclusions from the investigation of methanol impact on foam stability are as follows: • In the absence of methanol, the generated foam had a high foamability of 12.6 cm and better foam stability with half-life of 58 minutes. • In the presence of 5 wt% methanol, the generated foam had a lower foamability of 9.3 cm with much lower stability with half-life of only seven minutes. • This demonstrates the adverse effect of methanol on the generated foam by disrupting the foam structure and accelerating its collapse. 4.5 Subtask 3.5: Lab-Scale Compositional Modeling Simulation of Slim Tube Experiments Slim tube experiments were simulated in the commercial reservoir simulator, CMG-GEM, to investigate the effect of introducing rich gas in CO2 on MMP. The model parameters are shown in Table 20. The model size is the same as the slim tube size, and 40 grids were set in horizontal direction. Both live oil and dead oil were considered in the simulation. The rich gas, live oil and dead oil compositions which are based on Napoleon oil are shown in Table 21. Table 20. Slim tube model parameters Grid 40 x 1 x 1 Size (ft x ft x ft) 1 x 0.01345 x 0.01345 Porosity 0.439 Permeability (D) 6 Temperature (oC) 40 Table 21. Gas and oil composition Component Rich gas (mol %) Live oil (mol %) Dead oil (mol %) N2 0.0465 2.802 0 CO2 0.0121 0.092 0 CH4 2.4219 24.239 0 C2H6 10.8444 9.529 0 C3H8 28.4433 8.491 0 IC4 6.7034 1.460 0 NC4 21.6646 4.529 0 IC5 5.4296 1.335 1.625 NC5 10.1573 2.789 3.863 C6+ 14.2769 - - C6 - 3.524 6.643 C7+ - 41.209 87.87 Slim tube simulations were carried out using live oil and pure CO2 at several different pressures, and the resulting oil recovery factors are shown in Figure 93. From the simulation results, the MMP between live oil and CO2 was identified as 1,444 psi, the intersection where live oil and CO2 became miscible. To investigate the effects of introducing rich gas in CO2, separate simulations were conducted using 10 mol% and 20 mol% rich gas combined with CO2, and similar plots to Figure 93 were created to calculate MMP. Introducing 10 mol% rich gas led to an MMP decrease to 1,200 psi, while a higher concentration of 20 mol% rich gas resulted in a further reduction to 977 psi. This procedure was repeated using dead oil compositions rather than live oil, for which MMP values with pure CO2, 10 mol% rich gas and 20 mol% rich gas were found to be 1,373 psi, 1,190 psi, and 837 psi, respectively. These observations indicate that the Battelle | May 2026 96 introduction of rich gas in CO2 can help decrease MMP in both live and dead oil, with higher mole content of rich gas resulting in lower MMP. Figure 93. The slim tube simulation results for live oil and pure CO2 Lee 26 Oil Compositional Modeling Analogous in-situ reservoir fluid was sampled from a flowing well near the Lee 26 field that was producing similar fluid based on initial producing GOR and gas compositional analysis – the Hoffman 1-2 well. Two oil samples and three gas samples were taken from the Hoffman 1-2 well and analyzed via: • Compositional analysis of separator oil/gas • Oil viscosity measurements at surface conditions • Recombination of separator oil and gas to reservoir conditions based on producing GOR  Two recombinations were performed at two separate GOR values – 350 and 512 scf/bbl • Constant composition expansion (CCE) analysis of each recombined fluid Composition of the Hoffman stock tank oil is shown in Figure 94. The compositional analyses of separator fluids and recombined fluids were used to create a compositional fluid model for the Lee 26 reservoir. The CCE analyses provided bubble point pressures for each recombined fluid: 2,400 psia for the 512 scf/bbl fluid, and 1,750 psia for the 350 scf/bbl fluid. Discovery pressure at the Lee Field was around 1,620 psia, meaning that the 350 scf/bbl fluid is a more accurate representation of the saturated oil present at discovery within the reservoir. Using this fluid composition in combination with gas samples from the Lee field, a compositional fluid model was created for the reservoir at discovery. The full compositional analysis through C30 had to be simplified through component lumping to make the fluid model computationally tractable. The lumping scheme and final composition of the fluid model is shown in Table 22. Battelle | May 2026 97 Figure 94. Composition of Hoffman stock tank oil sample. Table 22. Lumped components and composition of recombined Lee 26 fluid model Component Composition (mol %) CO2 0.02 N2 3.17 CH4 24.84 C2-C3 9.56 C4-C6 7.87 C7-C10 19.32 C11-C15 15.86 C15-C29 14.73 C30+ 4.63 Using this lumped composition, the fluid model used in reservoir simulation was updated from the previous 520 GOR model to a fluid with 350 GOR that is more consistent with laboratory data. The model above was tuned using results from the CCE experiments run on the 350 GOR recombined fluid. The match to CCE results is shown in Figure 95, with laboratory data in red and simulation data in green/blue. All of the relevant properties match quite well, with the exception of oil compressibility. Stock tank viscosity also matches quite well, as shown in Figure 96. This compositional model was then combined with a model of free gas sampled from the Lee field to create a black oil fluid model using functionality in the CMG software, WinProp. Black oil properties of the model are shown in Figure 97. Battelle | May 2026 98 Figure 95. Compositional fluid model match to experimental data for the 350 GOR Hoffman oil. This model was used as the basis of the black oil fluid model used in history matching efforts in Q4 FY24. Figure 96. Modeled stock tank viscosity compared with lab-measured values. Battelle | May 2026 99 Figure 97. Black oil properties for the fluid model used in simulation. Top: solution gas-oil ratio (Rs) and oil formation volume factor (Bo); Middle: gas formation volume factor (Bg); Bottom: viscosity of oil and gas (Viso and Visg, respectively). Battelle | May 2026 100 5.0 Task 4: Design and Simulations of Injection Scenarios 5.1 Subtask 4.1: Development of Static Earth Models No SEM was developed for the McCann well area. This section focuses on SEM development for the Lee 26 field. Three formations were characterized for input into the SEM: the Utica, Trenton and Black River. The Utica Shale is the primary caprock in the system, while the Trenton Limestone and Black River Shale are both reservoir formations subject to dolomitization. The dolomitized zones, present near fluid-transporting faults, typically contain the highest porosity and permeability sectors. Therefore, facies and fault characterization are key factors influencing the development of the SEM. Analysis of oil production across the four wells within the field (Beattie, Jacoby, Fischhaber, and Marshell) is another indicator of the area with higher-quality reservoir properties. Well tops were picked in producing and non-producing wells in the area of the Lee 26 field for the Utica (model top), Trenton, Black River, and Glenwood (model base) formations based on known gamma ray log responses for each formation, selecting facies, and generating effective porosity and permeability logs. Four primary facies comprise the three zones of interest: shale, cap dolomite, limestone, and dolomite. The Utica Formation contains a uniform shale facies, although more heterogeneity exists in the Trenton and Black River formations. Typically, cap dolomite is present in the upper 5 to 50 ft of the Trenton formation and is believed to have occurred due to dewatering of the Utica. This interval typically has lower porosity and permeability than the conventionally dolomitized zones below it. Within the lower parts of the Trenton and the Black River Formations, facies are divided between limestone and dolomite. These facies were selected based on the neutron porosity and photoelectric (PeF) logs. Higher porosity, lower PeF intervals indicate dolomite presence, whereas low porosity and high PeF zones indicate limestone presence. The dolomite-rich zones are more commonly present in the southeast portion of the field, as indicated by the Fischhaber well logs and overall production in the Marshell well. These facies selections will guide the eventual property distributions that indicate reservoir quality. Based on the trend of the field from northwest-southeast, the facies major direction orientation (azimuth) is approximately -22o. Effective porosity was calculated using neutron porosity and shale volume log from gamma ray. The two permeability distributions were generated based on the Timur (1968) equation, which relates effective porosity to irreducible water saturation in sandstones and carbonates. Varying irreducible water saturation values of 0.15 and 0.5, derived in Task 2 “Advanced Field Characterization and Machine Learning Based Data Integration” and based on knowledge of the fields in the Albion-Scipio trend, were utilized due to there being no core data from the Lee 26 field. Pressure buildup data provided from the Marshell (easternmost) well were also examined. Buildup tests were conducted at the well on January 30, 2009 and April 28, 2009. Examining the way in which pressures evolve throughout these tests provides insight on reservoir characteristics and properties (including permeability). This information was invaluable in the history matching process. All analysis was done within the IHS Markit WellTest software. Screenshots of the buildup curves, along with fitted models, are shown in Figure 98. Battelle | May 2026 101 Results from the pressure buildup analysis are consistent with a dual-porosity system, affirming the conceptual model of the field. Single porosity models were a poor fit to the pressure data. Ranges for effective permeability, a composite of matrix and fracture (or in this case, moldic) permeability were between 250 to 1,400 mD. Estimates for interporosity flow coefficient (the effectiveness of flow between matrix and fracture) were between 10-5 and 10-6 which are average values in dual porosity systems. Results on storativity ratio, the ratio of oil stored in the network of vugs as compared to the matrix, were inconclusive, but ranged between 0.5% and 40%. Estimates based on the amount of dolomitization seen within the SEM are closer to this 40% mark. These ranges provide a starting point in history matching the SEM to fit field production data. Battelle | May 2026 102 Figure 98. Pressure buildup analysis for the Marshell well on 1/30/2009 (left) and 4/28/2009 (right). Shown are Horner plots (top) and typecurves (bottom) for each test, as well as the overlain model fit. Seismic Inputs to Static Model A 3D seismic volume provided by West Bay Explorations was used to characterize the structures controlling reservoir facies distribution in the field area. Horizons were picked above the target zones, within the sealing formation (Utica) and reservoir (Trenton-Black River), as well as below the reservoir (Glenwood and basement). Twelve horizons were mapped in total. Horizon-based and volume-based seismic attributes including RMS amplitude and average instantaneous frequency were generated, as well as a series of isopach maps. These attributes and isopachs maps were used to identify lineaments and ‘sag’ areas thought to be indicative of the faulting that introduced dolomitizing fluids to the reservoir interval (Figure 99). The sag features are clustered primarily in the vicinity of the reservoir and are oriented generally NNW- SSE. The sags and lineaments, visible both on maps and in seismic cross-sections, guided the placement of wrench-style faults throughout the seismic volume (Figure 100). The Lee Field outline was determined was defined during West Bay’s initial seismic analysis of the field using a dip of maximum similarity seismic attribute. This attribute identified the presence of faults that bound and seal the field based on where there is a continuous similar dip on a surface in the reservoir. The four producing wells in the Lee field fall within the bounding feature and several dry holes, such as the JF Booten well penetrate the TBR just outside of the bounding feature (Figure 101). The hotter colors bounding the Lee field wells in the figure correspond to where there is a continuous dip of maximum similarity, which indicates the presence of a fault providing lateral compartmentalized within this feature. Battelle | May 2026 103 Figure 99. Isopach map of the Trenton Formation showing thin ‘sag’ features (black dashed outlines) in relation to the reservoir (white solid outline). Battelle | May 2026 104 Figure 100. A- Arbitrary line through 3D seismic volume showing interpreted wrench faults bounding the sag features shown in the figure above. Battelle | May 2026 105 Figure 101. Dip of maximum similarity seismic attribute showing Lee field extent. The Fischhaber well contains both bulk density and sonic logs, which were used to generate a synthetic seismogram, allowing us to tie the well and formation tops to the seismic. Based on this well tie, the horizons for the four model formations were picked. The structural model was generated at a grid size to represent the vertical and lateral heterogeneity within the reservoir zones. The model comprises three zones representing the three formations of interest: the Utica, Trenton, and Black River. The depths and thicknesses of these zones were produced based on the well top picks and tied to their corresponding seismic horizons to guide structure (Figure 102). The Trenton formation is the key reservoir and therefore has the finest vertical resolution, followed by the Black River and Utica formations. The faults that were picked in seismic were integrated into the pillar grid. These faults bound the sag feature and are largely linear features within the model domain. Battelle | May 2026 106 Figure 102. Cross-sectional view of model showing zones, layering, and faults bounding the dolomitized sag feature. In addition to the faults bounding the field and identified using the dip of maximum similarity attribute, the faults transecting the Lee Field were incorporated into the model structure and used to guide the facies and property distribution. These faults are perpendicular to the bounding faults and in contrast are interpreted to serve as hydrothermal fluid conduits that dolomitize the reservoir and develop secondary porosity and permeability. A normalized distance to fault property was generated and co-kriged to the facies distribution, with dolomite presence being tied to proximity to these faults (Figure 103). The higher porosity zones are present within the dolomite, with the porosity distribution being guided by the facies. The property model is based on the effective porosity and permeability distributions, with these distributions to be guided by the facies and structural modeling. The well logs were upscaled to fit within the grid model, and a random gaussian property distribution was utilized to populate the grid cells. The facies and faults are intertwined, with the dolomite facies expected to be most commonly present near fault features that transport high-temperature brine. Based on the initial property model, the high porosity and permeability zones follow this trend (Figure 103). Battelle | May 2026 107 Figure 103. Cross-sectional view of initial porosity distribution in the Lee Field, with higher porosity zones present to the SE near the Fischhaber well. 5.2 Subtask 4.2: Upscaling and Ranking of Static Earth Models. SEMs generated in Subtask 4.1 were upscaled for use in CMG-GEM. The inability of the ML and data analysis work in Task 2.4 to resolve the relationships between hydrothermal dolomite facies and key geological properties caused the project to be unable to generate a broad and simultaneously data-driven set of static model realizations. Instead, the project focused its efforts on accurately representing geological property distributions based on the placement of structural features in the Lee 26 seismic volume as part of Subtask 4.1. 5.3 Subtask 4.3: Material Balance Study A series of material balance analyses (MBAs) were conducted using production and pressure data from the Lee 26 field in order to estimate initial hydrocarbon in place at the time of discovery –OOIP and original gas in place (OGIP). An accurate estimate of these values is crucial for estimating field recovery and residual oil volumes, as well as constraining dynamic reservoir simulations. MBAs are conducted using the material balance equation, which relates initial fluid in place (OOIP and OGIP) to produced oil and gas by accounting for volume gain/loss, fluid expansion/compression, solution gas evolution, and (where relevant) pore volume expansion/compression. The concept of MBA is explained in detail in Ahmed (2019). The Lee 26 field is a saturated oil reservoir with a gas cap showing a combination drive with little to no aquifer support and fairly incompressible rock. As a result, the MBA focuses on the oil and gas terms in the equation, ignoring pore volume compaction and water influx. An initial MBA was conducted by project partner West Bay Exploration, the operator of the Lee 26 field, using data available for the field in 2019. Estimates of the OOIP were calculated at various times throughout the production lifetime of the field using standard correlations for fluid properties (Vasquez and Beggs, 1977; Standing and Katz, 1942) alongside field data. An initial gas cap fraction of 20% was assumed, and solution GOR of 512 scf/bbl. The OOIP and OGIP estimates obtained from this analysis at various stages of field depletion are shown in Figure Battelle | May 2026 108 104. After the results of the PVT analysis conducted on Hoffman 1-2 oil in Task 3 were received showing an initial solution GOR of 350 scf/bbl, a second MBA was conducted using the same correlations with updated oil properties and final production volumes from the field. Using the same assumed 16% free gas volume fraction, the updated MBA suggested OOIP and OGIP of 3.06 million stock tank barrels (mmstb) and 1.50 billion cubic feet (Bcf), respectively. The OGIP is consistent with initial estimates. Given the level of depletion in the field, nearly all of the OGIP will have been produced (i.e. OGIP roughly equals produced gas volume). OOIP increases in this case by ~35% from the average of initial MBA calculations, suggesting a substantially lower recovery fraction than initially calculated. Figure 104. Material balance estimates of OOIP and OGIP for the Lee 26 reservoir. This analysis technique inherently has some uncertainties. Some of these uncertainties are relatively small, such as the assumption of no pore volume compaction. Any system with connate water saturation, which the Lee 26 field is expected to have, will exhibit some degree of pore volume compaction simply from the expansion of this connate water. Given the relative incompressibility of water, this uncertainty was deemed inconsequential. The volume of produced water in this system is also a source of small uncertainty. Given the lack of available water production data, the MBA ignored this piece of information. The best available estimate of between 175,000 and 200,000 stb of produced water serves to change the OOIP estimate by about 0.3%. Several significant sources of uncertainty do exist, however. One such uncertainty is the free gas volume fraction in the reservoir at discovery. While it is known that a gas cap exists based on observations during field development, the complex geometry of the Lee 26 field makes it difficult to obtain a better estimate than the 20% value used. Another significant uncertainty is the implicit assumption of MBA techniques that the system is a static tank. Pressure data obtained during water injection operations at the Lee 26 field suggests a dynamic system, with a high-conductivity fracture that closed at some point during depletion, decreasing connectivity to certain sections of the reservoir. This would have changed the pressure and production profiles Battelle | May 2026 109 of the system in a way that would be misrepresented by MBA calculations. Despite this uncertainty, the MBA gives a reasonable estimate for total reservoir volume that can be used as a starting point for dynamic modeling work. 5.4 Subtask 4.4: Dynamic Simulations and Analysis Dynamic modeling work for the Lee 26 field began early in the project lifetime and evolved in both complexity and scope over the course of the project. Initial modeling efforts consisted of tank models built to estimate the volume of water needed to pressurize the depleted field to MMP and estimate necessary CO2 volumes for a chemically-enhanced CO2-EOR flood. After this, a history-matched model was built from the full-complexity SEM by incorporating lab analysis results, multi-porosity systems and field production and pressure data. This history- matched model was used to forecast expected production from the CO2 treatment. A series of shoebox models was built and assessed based on this history-matched model to better assess the efficacy of low-pressure CO2-EOR suggested by the history-matched model. Finally, a specific set of forecasting scenarios were assessed on the history-matched model to inform production strategy during field testing operations. The modeling efforts are presented chronologically below to illustrate how the reservoir modeling evolved and built on existing work throughout the project lifetime. Tank Model – Early Repressurization Screening Early modeling efforts focused on estimating water injection needed to facilitate a miscible or pseudo-miscible flood at the Lee 26 field given the measured MMP of ~1,300 psi. A target pressure of 1,000 psi was investigated as a tradeoff between lower recovery efficiency and decreased repressurization time. A basic repressurization scenario in a homogeneous, layer cake reservoir was modeled in the commercial simulator CMG-IMEX using reservoir property data provided by project partner West Bay. Model dimensions were constrained by a material balance study performed by West Bay that estimated OOIP as 2,294,000 bbls. Using a formation volume factor of 1.26 (also from the material balance study) and the OOIP estimate, estimated pore volume occupied by oil at discovery as 2,890,000 bbls at reservoir conditions. The reservoir gas cap is estimated to take up roughly 16% of the volume occupied by oil, giving us a final hydrocarbon pore volume (HCPV) estimate of 3,352,000 bbls at reservoir conditions. Reservoir dimensions were then calculated using volumetric analysis techniques with the equation: 𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻 = 𝐴𝐴 ∗ ∆𝑧𝑧 ∗ 𝜙𝜙 ∗ 𝑆𝑆𝐻𝐻𝐻𝐻 where A is the surface area spanned by the reservoir, ∆𝑧𝑧 is the thickness of the reservoir, 𝜙𝜙 is porosity and 𝑆𝑆𝐻𝐻𝐻𝐻 is the initial saturation of hydrocarbons. An initially saturated reservoir with connate water saturation of 0.15 was assumed. The surface area was found by examining the footprint of wells within the reservoir and was estimated to be 0.6 mi2. Average reservoir porosity was estimated from core samples to be ~3.9%. Using these values and the HCPV estimate gives a reservoir thickness of 35 ft. These values were used to build the tank model illustrated in Figure 105. The reservoir model has a gas cap in the top layer which accounts for ~14% of the total volume. Reservoir permeability was set to 50 mD with vertical to horizontal anisotropy factor of 0.01. Battelle | May 2026 110 Figure 105. Depiction of the initial gas saturation in the Lee field tank model. Model dimensions are 4,030 ft x4,030 ft x35 ft. Using production data provided by industry partners, a history match was attempted using the simplified model with production data for the Lee field as shown in Table 23 below. Obtaining a suitable history match was a difficult process with the tank model. The complexity of the Lee field and its completions were difficult to replicate without a more sophisticated model. However, a suitable match to the data was obtained by adjusting well locations and orientations and modifying reservoir parameters. The match to field pressure data are shown in Figure 106, and the matches to both oil and gas production are shown in Figure 107. Table 23. Cumulative production and pressure data from the Lee field. Date BHP (psia) # of CUM Oil CUM Gas Wells (bbl) (MCF) 6/2/2008 1,612.7 1 91 251 4/28/2009 1,428.6 3 56,213 47,468 11/10/2009 1,394.5 3 79,355 121,011 11/13/2010 1,170.5 4 156,233 252,952 11/17/2010 1,363.5 4 157,050 254,235 6/13/2012 990.8 4 266,335 440,546 9/25/2014 433 4 338,877 970,034 12/31/2014 412 4 341,331 1,017,116 12/31/2018 65 3 351,290 1,464,267 Battelle | May 2026 111 Figure 106. Pressure data and history match. Figure 107. Cumulative Oil (green) and Gas (maroon) production data and history matches. Part of the difficulty in history matching stems from a change in production conditions in the Lee field. Starting around mid-2013, oil production tapered off and gas production increased. Several changes to the tank model were necessary to force additional gas production while keeping oil production at a steady decrease. These changes included decreasing vertical permeability, lowering oil mobility after June 2013 (by lowering the oil relative permeability), and inserting a horizontal production well into the gas cap in June 2013. Figure 108 shows the placement of this horizontal well within the reservoir model, as well as the gas saturation near the end of production (November 2019). Battelle | May 2026 112 Figure 108. Placement of the horizontal well within the gas cap of the reservoir tank model. Coloring corresponds to gas saturation in November 2019. These artificial changes provide insight into the actual production processes that were likely occurring in the field. Production data indicate that beginning sometime in 2013, production from the field was primarily gas which likely originated from the gas cap. This high amount of gas production would also explain the large drop in pressure from 2015 to 2019, as producing primarily gas lowers pressure much faster than producing a more oil-heavy mixture. In addition, the production rates necessary to match production data after 2013 caused a significant amount of coning in the simulations. The presence of this coning coupled with the low amount of oil production during this time is further evidence that much of the late-stage production was likely from within the gas cap. These lessons learned were applied in all following model development efforts. Reservoir repressurization volume estimates were then calculated using the history matched tank model by simulating water injection into the depleted reservoir until formation pressures reached 1,000 psi (a conservative estimate of the miscibility pressure for CO2). Injecting water at 50,000 bbls per month (~49 gpm), it took 28 months and an injection volume of 1.4 MMB of water to reach 1,000 psi. The results of this simulation are shown in Figure 109. A scenario was also modeled where water was injected with an injection well blowhole pressure maintained at 90% of the formation fracture pressure (assuming a fracture gradient of 0.8 psi/ft). In this scenario, the reservoir took less time (~14 mos.) to repressurize, but the injection volume remained consistent at 1.4 MMB. Battelle | May 2026 113 Figure 109. Pressure (orange) and injected water volume (blue) in the history matched tank model with repressurization. In this scenario, water is injected at 50,000 bbls/mo. until pressure reaches 1,000 psi. Given the high estimate of water volumes for repressurization, the team began exploring scenarios where only a section of the reservoir was repressurized (around a single well or across a specified well pair), theoretically allowing for chemical EOR on only the repressurized zone. An additional layer of sophistication was built into the tank model to facilitate this effort. To better represent the geometry of the Lee field, volumetric model dimensions of the tank model were adapted to be significantly longer in one areal dimension than the other (more consistent with field boundaries determined from seismic). Reservoir properties were maintained from the initial tank model. The four wells in the system were added at locations consistent with the four wells in the Lee field. The updated volumetric model is shown in Figure 110, with well traces of each of the four new wells shown. The figure is oriented NW (left) to SE (right). The model was history-matched to field production and pressure data, the results of which are shown in Figure 111. Figure 110. Image of the updated volumetric model for the Lee field with approximate locations for the four wells. Orientation is NW to SE. Battelle | May 2026 114 Figure 111. Lee field history-matched model results (lines) with field data (points) for Cumulative Oil (black), Cumulative Gas (green), and Reservoir Pressures (orange). With a suitable match obtained, modeling efforts shifted focus toward pressurizing a sector of the field that spanned across the identified injector-producer pair (Marshell and Fischhaber) to at least 1,300 psi, the minimum miscibility pressure of the formation fluid. The location of this sector is illustrated in Figure 112. Initial scenarios for repressurization were modeled in the history-matched domain with different wells (and number of wells) used for injection and variable injection rates. Jacoby was unable to be used for injection due to well conditions, and modeling results showed that injection through Beattie provided minimal benefit to repressurization of the eastern sector. Injecting water through Fischhaber, the designated producer during CO2 EOR flooding, would lead to a high water cut during the production phase of the project. Thus, the only viable scenario was to use Marshell to inject the water. Figure 112. Image illustrating the sector to be repressurized (pink) in the volumetric model. The orientation of the cross-section is NW to SE. Various injection rates and lengths of injection were simulated for a single-well repressurization scenario at the Marshell, the results of which are shown in Figure 113. Three different scenarios Battelle | May 2026 115 were able to attain an average sector pressure of 1,300 psi – injecting at 80,000 bbls/mo. for either 7 or 10 months or injecting at 50,000 bbls/mo. for 10 months. The models were also limited to a well BHP that is 90% of the fracture gradient, causing injection rates to drop off in later months. The injection volumes for these scenarios ranged from about 450,000 bbls to 580,000 bbls of water. These estimates were deemed more logistically feasible than the calculated 1.4 million barrels of water necessary to repressurize the entire field, and a strategy of sector repressurization was pursued. Figure 113. Average pressures in the sector shown in Figure 112 for a variety of injection scenarios. Target pressure is shown by the pink line, and three scenarios manage to attain this pressure. Full-Complexity Model – CO2 Volume Scoping Upon completion of the detailed SEM in Petrel, the model was imported into CMG. Shown in Figure 114 is the porosity distribution of the model in three dimensions. The Utica is the caprock at the top, followed by the Trenton, Black River-1 and Black River-2 in increasing depth. The oil- water contact was initialized at 3,735 ft true vertical depths, subsea (TVDSS), consistent with geologic interpretation, and a connate water saturation of 20% was applied across the initial model. Given the increases in the pricing of CO2 at this stage of the project, the first modeling goal was to obtain an estimate of the volume of CO2 necessary for a half-pore volume (PV) flood of the southeast sector of the field. Based on well log data, the majority of the Lee field reservoir is in the top section of the Black River-2, with some parts of the reservoir lying in Black River-1 and the bottom portion of the Trenton. To remove the PV above this region, null values were applied to blocks above the reservoir interval, as shown in the top panel of Figure 115. PV was then calculated in CMG for the field as a whole and for the sector of interest (bottom panel of Figure 115). Battelle | May 2026 116 Figure 114. Three-dimensional initialization of Lee Field static model in CMG-GEM. Image shows porosity throughout the model. Figure 115. Portions of the Lee field used in the CO2 volume estimation: (top) lower region of the field within which the Lee field reservoir lies; (bottom) SE sector of the field, highlighted in pink. Battelle | May 2026 117 The HCPV estimate of the full field was in alignment with values obtained from the production- based material balance analysis on the field – 3.35 million reservoir barrels. The HCPV of the sector of interest was 7.6 million cubic feet (at reservoir conditions). Based on this volume and the density of CO2 at 1,200 psi – a sub-optimal repressurization target – the CO2 mass requirement for a 0.5 PV flood of this section of the reservoir was calculated to be 29,850 metric tonnes. Full Complexity Model – Multi-Porosity Integration and Production History Matching Initial dynamic modeling was conducted on the single porosity model used in CO2 volume estimation. Given the large gap between early results and field data, along with the dual porosity pressure signatures observed in pressure transient analysis, the decision to implement a full dual porosity model was made. Although many depositional and diagenetic facies are identifiable within TBR reservoirs, the reservoir can be effectively represented for modeling purposes with three primary ‘producing facies’: tight limestone, matrix dolomite, and vugs/cavernous dolomite. In addition, fractures are expected to be present in the system; these are grouped with the high-conductivity vug/cavernous dolomite facies (referred to hereafter as vug/fracture). Facies were initially split within the SEM into limestone and dolomite. Laboratory and field data shared by industry partners suggest that the limestone facies are not pressure connected to the reservoir and do not contribute to fluid flow. As such, model blocks defined as limestone were eliminated fom the dynamic model by setting those cells as null. The dolomite facies was further divided by using a porosity cutoff. Within the dolomitic facies blocks, the dual porosity system was identified as: 1. Matrix porosity – small pores and generally lower porosity and permeability 2. Vug/fracture porosity – higher permeability and generally higher porosity In the dynamic model, porosity values of <10% were assigned as matrix dolomite, and values >10% assigned as vug/fracture. The matrix and fracture (really vug/fracture) porosity values were calculated from total porosity (ΦT) as defined in the SEM for each block using the following equations (for Χ=0.1): Implicit in this assignment of values is that the vug/fracture system is present in every block within the model. Blocks assigned as matrix dolomite are assumed to have a very small contribution from the vug/fracture system. Permeability for each system was calculated using the effective permeability from the SEM, volume fraction and dual porosity factors using the approach of Warren and Root, 1963. Grid blocks are modeled as each containing a single vertical fracture in each of the x and y directions. After some algebraic manipulation based on this geometric assumption and the use of shape factors defined by Warren and Root, the final equations used for permeability calculation are as follows: Battelle | May 2026 118 The resulting model porosities are shown in Figure 116. The dual porosity option was used in CMG-IMEX to determine matrix and fracture grid connectivity. Figure 116. Three-dimensional rendering of the dynamic model, viewed from the south. The upper render indicates the distribution of matrix porosity; the lower render indicates the distribution of fracture and vuggy porosity. White/blank space indicates limestone matrix that is not pressure connected to the reservoir. Once the porosity and permeability distribution were populated, fluid saturations were assigned to the model. The hydrocarbon fluids were represented with a black oil fluid model for the Napoleon oil assessed in Subtask 3.2 (Section 4.2). Fluid contact depths were assigned based on data provided by the field operators. A 15% minimum connate water saturation was used Battelle | May 2026 119 throughout the reservoir, and the top boundary of the reservoir was determined using OOIP, OGIP and free gas fraction constraints from the material balance analysis. The resulting fluid saturations are shown in Figure 117. Figure 117. Dynamic reservoir model filled with reservoir fluids. The model indicates a reservoir filled with gas-saturated oil, with an overlying gas cap. The lower portion of the reservoir is a water leg, although there are no indications it provided any water drive. Initial history matching to production data was done on this model using the original 520 scf/bbl solution GOR assumption. A fairly strong match was obtained by varying local pore volumes, local permeability, and grid connectivity in specific locations. Once PVT analysis was completed on the analogous Hoffman oil sample, suggesting a lower solution GOR, the existing model was history matched again to field production data using the new black oil model for Hoffman oil (see Subtask 3.1, Section 4.1). The prior model provided a good starting point for history matching efforts, and a match was obtained by modifying local-scale pore volumes, permeabilities and relative permeabilities throughout the reservoir. In addition, during history matching the gas cap over the Marshell well was removed from the model, a modification supported by revisiting log analyses. A good quality match was obtained for the complete field history as shown in Figure 118. The well-by-well history match, shown in Figure 119, was also fairly consistent. The Fischhaber and Jacoby wells showed the largest variability from observed production data. Cumulative volumes show a good match at each well (including Jacoby and Fischhaber), but production rates are not always well matched. Figure 118. Field-scale production history match showing comparison between gas/oil production and pressure of the simulation and field data. Battelle | May 2026 120 Figure 119. Well-by-well history match at the Lee 26 field, showing gas (blue) and oil (black) production at each well in the simulation (solid line) vs. field data (open circles). Full Complexity Model – Injection History Match and Forecasting Models Using the production history-matched model, further simulations were run to attempt to match pressure data obtained during water injection operations. Initial results showed that pressure buildup at the Marshell well was significantly higher than observed in the field while pressures at the Fischhaber were much lower than observed. This is illustrated in Figure 120. In an attempt to increase pressure transmission between Marshell and Fischhaber within the model, permeability was increased across the entire model. This had the result of reducing total pressure buildup in the field, with hardly any pressure increase observed at either well. This suggested a more complex explanation for the lack of a match in injection data despite a good production history match. Figure 120. Pressure response at the Fischhaber (left) and Marshell (right) wells due to water injection. The model response does not match observations. One possible explanation for the model pressure responses was gleaned from analysis of the fluid saturations and pressure distributions after water injection. Figure 121 shows these properties in the reservoir after 15 months of injection following the field injection schedule. Pressure primarily builds up in a compartment surrounding the Marshell well (eastern side of the field), and high pressures do not propagate very far outside of this region. Despite this, injected water does manage to travel throughout the entire reservoir, reaching the far western end of the field. This is consistent with sample analysis and observations of chemical tracers in all three offset wells. There is a clear discrepancy in inter-well conductivity between the Marshell and Fischhaber (and to a lesser extent, the western portion of the field). Battelle | May 2026 121 Figure 121. Pressure distribution (top) and fluid saturations (bottom) in the Lee field model after 15 months of water injection. A solution that was identified was the presence of a high conductivity channel between Marshell and Fischhaber that continued deeper into the field. A high conductivity channel that spans the entire field from E-W was added to the model at a depth of ~3,705 ft TVDSS (4,653 ft TVD at Fischhaber). This depth was chosen based on image logs at the Fischhaber well, which indicated the presence of vugs, and porosity logs at the Marshell well showing high porosity around this depth. The throughgoing channel was given a constant horizontal permeability of 1,000 mD and vertical permeability of 500 mD. The porosity was set to 0.5% in blocks that did not already show the existence of vugs/fractures and unchanged in those that did. In addition, a vertical connection was established at the Fischhaber well between the deeper and shallower sections of the reservoir. This connection was expected to exist based on the lack of agreement in pressure data from the previous history-matched (HM) model. A 3x3 section of the model was added surrounding the Fischhaber well trace with high vertical permeability (100 mD) and low horizontal permeability (1 mD). Porosity was calculated in the same way as the throughgoing channel. This vertical connection was modeled in the immediate wellbore vicinity to approximate the impact of an acid treatment done on the well shortly after completion. The addition of these two features, as well as further modifications to local pore volumes (both fracture and matrix) and permeabilities, resulted in a new field history match. The production data match was worse than the prior model, but injection pressure data match was significantly better. The team chose to weigh the injection data more heavily due to the high resolution nature of the data and its proximity in time to planned CO2 injection operations. The results of the final history-matched model are shown in Figure 122 (field-wide production history match), Figure 123 (well-by-well production match), and Figure 124 (injection pressure data match). The overall match is fair, but the lack of agreement particularly in production data suggests the Battelle | May 2026 122 subsurface interpretation shown in the model may not fully capture the complexity present in the Lee 26 reservoir. Figure 122. Field-wide history match of current best-fit Lee 26 model. Match quality is slightly decreased from prior models in favor of injection data match. Figure 123. Well-by-well history match of current best-fit Lee 26 model. Battelle | May 2026 123 Figure 124. Injection pressure data match of current best-fit Lee 26 model. Downhole gauge data (points) are coarsened from the high resolution output of the gauges. The saturation distribution within this history-matched model did not quite agree with the team’s mental model of the reservoir. The expectation was that the vug/fracture network within the reservoir would be entirely gas-saturated at field depletion, as supported by the very high GOR in the later years of production. Matrix oil saturations were expected to be high within the oil leg, accounting for practically all of the residual oil in the field. However, the numerical model suggests a substantial column of oil in the secondary porosity network below the perforations in the wellbores. Figure 125 (top) shows this column (highlighted with an orange box) in the model. Also shown in the figure are matrix saturations, which are essentially unchanged from discovery. This mostly agrees with the team’s understanding of the reservoir. Despite the disconnect between the team’s mental model and the numerical model, this match was deemed satisfactory in balancing field operation timelines, resource allocation, and diminishing returns from further modifications to the history-matched model. Battelle | May 2026 124 Figure 125. Saturation distributions in the vug/fracture network (top) and matrix (bottom) at field depletion in the Lee 26 model. The orange box highlights a column of oil in the “fracture” network at field depletion. In order to assess the utility of further field repressurization operations prior to CO2 injection (i.e., more water injection), a series of huff and puff simulations was assessed on the best-fit field history model. Five scenarios were run in which the field was pressurized to various target pressures, after which the anticipated field CO2 injection operations were modeled. The scenarios were designed as follows: 1. Inject water through Marshell at rates consistent with field data (1,800 bbl/d afterward) 2. Shut-in field for two weeks, assess average reservoir pressure  This is the representative pressure for scenario definition  Five target pressures: 550, 750, 950, 1,150, 1,350 (psia) 3. Inject CO2 through Fischhaber in three 1,000 tonne slugs  Inject at 1,000 tonnes/d  Follow CO2 with a wellbore volume of water  Soak for 12 days following an injection phase (including the final phase) 4. Produce Fischhaber well at 600 bpd liquid target rate CO2 injection was added to the model using the pseudomiscible option in CMG-IMEX (keyword MISNCG). The input solvent PVT and miscibility table was generated using results of the MMP lab testing results. The average field pressures for all five scenarios (color-coded) are shown in Figure 126. Oil recovery for each of the scenarios with the same color coding is shown in Figure 127. Battelle | May 2026 125 Figure 126. Average reservoir pressure in each of the 5 huff and puff scenarios during pressurization, CO2 injection, and production. The colors range from low to high pressure as follows: orange, blue ,black, green, purple. Figure 127. Cumulative oil (top) and oil production rate (bottom) for each of the five huff and puff scenarios. Curves are synchronized for comparison purposes. Colors correspond to the low-high pressure colors described in Figure 118. Battelle | May 2026 126 The results of these simulations show only small differences between the performance of low- and high-pressure huff and puffs, an unexpected observation based on CO2 huff and puff operations in more conventional settings. Moreover, the lower pressure scenarios show higher total oil recovery (at least within the first 6 months of production) than the higher-pressure scenarios. This result initially seemed to contradict the team’s understanding of CO2-EOR operations. Reaching MMP, or at least getting close to it, by repressurizing the field with water was a high priority in order to optimize the volume of swelled oil resulting from dissolved CO2. It appears in this case that the benefit from a higher volume of CO2 at reservoir conditions (i.e., the lower pressure scenarios) – which allows for more reservoir volume to be touched by the injected CO2 – outweighs the benefits of miscibility in total recovery. Motivated by the unexpected result, a literature review was conducted to identify previous work on the impacts of miscibility in CO2 huff and puff operations in fractured reservoirs. One such study was discovered with results that largely agree with the modeling results discussed above. The paper, Ding et al. (2019), describes bench-scale experiments of CO2 floods conducted on manufactured core with various fracture parameters and pressure regimes. The overarching result suggests that low/medium pressure (pseudomiscible) floods perform as well as or better than high pressure (fully miscible) floods in fractured settings. Shoebox Model – Huff and Puff Operational and Reservoir Parameter Sensitivity Additional simulation work was conducted to validate the findings that a low-pressure huff and puff operation would perform as well or better than a high-pressure huff and puff. The first step toward validating the low-pressure injection results was to build a simplified model, which was termed the “shoebox model,” to enable rapid sensitivity analyses. A simulation matrix was developed to determine which variables would be assessed in the sensitivity analyses, as shown in Table 24. Table 24. Scenario Matrix for Shoebox Model Sensitivity Analysis Variables Unit Value 1 Value 2 Value 3 Value 4 Notes Matrix Permeability mD 1 10 100 k direction = i- direction * 0.1 Matrix Porosity % 4 8 12 poro/perm linked (i.e. same Value #) Fracture Permeability mD 100 200 1,000 No k direction = i- fractures direction *2 Fracture No poro/perm linked (i.e. Porosity % 0.1 0.1 1 Fractures same Value #) Repressurized average reservoir psi 200 500 800 1,200 pressure pressure CO2 volume tonnes 100 1,000 Water Injection Well N/A Marshell All 3 well(s) used during Offsets repressurization Production BHP Rate N/A Constraint (AOF) (600 bpd) The shoebox models were set up as a rectangular prism (longer than it is wide, similar to the real field and prior rectangular tank model) with pore volume and original fluids in place (OFIP) Battelle | May 2026 127 similar to the full complexity reservoir model. Porosity and permeability in each porosity system were homogeneous across the model domain. Since porosity varied across the different scenarios, specific scenarios were fine-tuned with different volume multipliers for each unique porosity combination to match OFIP. The fluid contacts were kept at the same depths as in the complex model. The wells were in roughly the same locations as in the complex model, with some slight variation due to the lower resolution of the shoebox model. Relative permeability curves and black oil fluid model were identical to the full complexity model. In order to ensure each simplified model was representative of the depletion history of the Lee field, a full-field history matching exercise was conducted for each unique porosity combination. Proxy production wells located strategically throughout the different fluid columns (oil/gas) were used to generate the full-field matches since a well-by-well production match would be infeasible with this simplistic model. This resulted in a series of depleted reservoir models that were sufficiently representative of the true depleted reservoir state given the level of simplicity. The history matching results for each scenario are co-plotted in Figure 128. The dots show field data for production volumes and pressures, while the various curves show different scenarios (unlabeled due to the high number of scenarios). Green curves/dots show cumulative oil volume, red curves/dots show cumulative gas volume, and orange shows average reservoir pressure. All of the gas and pressure curves show a reasonably close match. The oil curves are not as well matched in some cases, indicating that specific fracture/matrix porosity combination is likely not perfectly representative of the reservoir. The scenario corresponding to the oil curve that drastically underestimates oil production corresponds to the unfractured scenario. It is understandable that this does not match well, since the reservoir is dual porosity, but it was important to have an unfractured scenario in the sensitivity analysis for purpose of comparison. Figure 128. History matching results for the various shoebox model cases assessed in the sensitivity analysis. Next, the repressurization period was simulated on each of the history-matched models. Depending on the scenario, water was either injected into just the Marshell well or into all three offset wells (Marshell, Jacoby, Beattie) at a field injection rate of 3,000 barrels per day (bpd). Water injection continued until the average reservoir pressure reached the value cited in Table 24 for a given scenario. Once the model reached target pressure, a CO2 slug was injected into Battelle | May 2026 128 the Fischhaber well – either 100 tonnes or 1,000 tonnes – at a rate of 200 tonnes per day (tpd). After a four-week soak period, the Fischhaber well was turned over to production with either (1) a bottom hole pressure constraint corresponding to absolute open flow (AOF – zero gauge pressure at bottom hole) or (2) a rate constraint of 600 bpd, with a secondary bottom-hole pressure (BHP) constraint corresponding to AOF. In the second case, the well would produce at 600 bpd until the BHP reached 0 psig, at which point the constraint would flip to BHP. A total of 46 unique scenarios were run in the sensitivity assessment. From the results of this sensitivity assessment, several conclusions are draw, and addressed in turn below: 1) In a single porosity reservoir, higher pressure leads to higher recovery. Figures 129 and 130 and show the results of models run on a single porosity model with porosity 8% and permeability 10 mD. Across all of these scenarios, attaining a higher reservoir pressure prior to CO2 injection leads to a higher cumulative recovery. This is true regardless of the volume of CO2 injected or production strategy (AOF versus rate-limited) used. These results align with the conventional understanding of CO2-EOR – higher pressures yield better results due to the impact of CO2-oil miscibility and the fact that a higher pressure reservoir has more energy to move fluid into the production wellbore. This indicates that the dual-porosity nature of the Lee field contributes to the efficacy of low-pressure huff and puff scenarios. Figure 129. Cumulative recovery in the unfractured shoebox model for AOF production scenarios Battelle | May 2026 129 Figure 130. Cumulative recovery in the unfractured shoebox model for rate-limited (600 bpd) production scenarios. 2) In a dual porosity reservoir producing at AOF, higher pressure leads to higher recovery. The trend observed for the unfractured scenarios remains true in the fractured scenarios where an AOF production strategy is used. Figure 131 below shows the results from a fractured model (matrix poro/perm same as above; fracture porosity = 1%, fracture permeability = 1,000 mD) where the field was repressurized using all three offset wells. Figure 132 shows the same, but for cases repressurized through only the Marshell (1,000 t CO2 injection). The observed trend is true regardless of the volume of CO2 injected. In contrast to the single porosity cases, however, oil volume plateaus after a point as water cut approaches 100%. The high permeability fractures/vugs allow for injected water to bypass oil remaining in the matrix system. Figure 131. Cumulative recovery in fractured shoebox model for AOF production strategy. Water re- injection was done through all three offset wells. Battelle | May 2026 130 Figure 132. Cumulative recovery in fractured shoebox model for AOF production strategy. Water re- injection was done through only the Marshell well. 3) In a dual porosity reservoir, a rate-limited production strategy increases hydrocarbon yields and performs similarly at both low and high pressures. Figure 133 shows the results for the 1000t injection, three-well repressurization, fractured scenarios from above, but with a rate-limited (600 bpd) production strategy. Unlike the AOF case, similar performance is seen across different pressure cases. Production in the first 200 days or so does follow the established trend, where higher pressures correlate with substantially higher recovery. However, over the course of the first 1.5 years, recovery is approximately the same across the four scenarios pictured, varying by ~1,000 bbls between the lowest and highest recovery scenarios at the 1.5 year mark. The same qualitative behavior is observed in scenarios with 100 tonnes of CO2 injected, and in scenarios where water is injected exclusively at the Marshell well (not pictured). In addition, the yields in the rate- limited case are substantially higher than in the AOF case. The high water cuts observed at later times in the AOF case do not appear to be an issue in the rate-limited case, so oil continues to be produced. Figure 133. Cumulative recovery in the fractured shoebox model for rate-limited (600 bpd) production strategy. Water re-injection was done through all three offset wells. 4) The heterogeneity of the full complexity model contributes to the efficacy of low-pressure huff and puff scenarios. Battelle | May 2026 131 It should be noted that while the distinction between low- and high-pressure scenarios is less apparent (i.e., similar performance) in the rate-limited, fractured scenarios, better performance is seen from high pressure than from low pressure cases. In the simulations discussed that were run on the full complexity model (see Figure 129), low pressure cases outperformed the high-pressure cases. This would suggest that the heterogeneity present in the full complexity model is at least partially responsible for the efficacy of a lower-pressure huff and puff. The contrast in reservoir parameters between the two porosity networks is also a factor, as was discovered within this sensitivity analysis. Figure 134 shows the results of a low porosity, low permeability case for both the fracture and matrix networks (matrix: 4%, 1 mD; fracture: 0.1%, 100 mD). The left panel shows results from an AOF production strategy, while the right panel shows results from a rate-limited strategy. In the AOF case, the high-pressure scenario (1,150 psi) outperformed the low pressure scenario (450 psi) as with the first porosity/permeability combination tested. However, oil production does not plateau as it did in the other case. It is unclear why this is the case, but it could be due to the lower degree of contrast between matrix and fracture permeabilities. In the rate-limited case, the low pressure treatment does indeed outperform the high pressure treatment. Qualitatively similar results are seen for a high matrix poro/perm (12%, 100 mD), low fracture poro/perm (0.1%, 200 mD) case (not pictured). While it is hard to draw exact quantitative relationships based on these results, this does show that for certain reservoir parameter distributions – even homogeneous ones – low pressure CO2 huff and puffs can outperform their high pressure counterparts. In summary, a combination of reservoir heterogeneity (both the multi-porosity nature of the reservoir and heterogeneity within each porosity network) and a rate-limited production strategy appear to be responsible for the efficacy of low-pressure huff and puff scenarios suggested from full-complexity models. These findings helped to further inform the results of additional simulations run on the full complexity reservoir model. Figure 134. Sensitivity results for low matrix poro/perm and low fracture poro/perm cases at two different pressure levels. The left panel shows an AOF production strategy and right shows rate-limited Full Complexity Model – Operational Sensitivities Additional full-complexity models were run to assess the impact on recovery of various operational considerations – specifically CO2 injection rate and volume, recycling of non-oil well stream components, and production rate – to further investigate the low-pressure huff and puff Battelle | May 2026 132 scenarios and to help inform field operations. The first set of simulations was run to assess the uptick from CO2 over a pure water injection. To compare with the five scenarios illustrated in Figure 126 (3,000 tonne huff and puff at 550, 750, 950, 1,150, 1,350 psi target pressures), five additional scenarios were run where the CO2 injection phase was skipped. The results of these 10 scenarios are co-plotted in Figure . The addition of CO2 provides a substantial recovery uptick at early times over water-only scenarios. This observation is true at all modeled pressures, indicating that even in the absence of full miscibility, CO2 treatment is effective at mobilizing oil. In the 550 psi scenario, the closest to actual field implementation, the difference in cumulative oil over the first year with CO2 vs. only water is about 6,000 barrels, a recovery increase of ~60%. The second set of simulations was run to investigate the contribution from miscible behavior of CO2 and oil in the reservoir. In addition to the five huff and puff scenarios already reported on, two additional runs were made with higher average reservoir pressures prior to CO2 injection (1,650 psi and 1,900 psi). The reservoir pressures during the repressurization (water injection) phase are shown in Figure , with the lab-obtained minimum miscibility pressure of 1,300 psi co- plotted. The times at which miscible phase behavior is contributing to oil mobility correspond to the time the average reservoir pressure spends above the horizontal minimum miscibility pressure (MMP) line, so the degree to which miscibility impacts overall recovery can be understood by comparing the higher pressure scenarios (green), with the lower pressure scenarios (blue). The oil rate corresponding to each of these scenarios is shown in Figure . The vertical axis is cutoff to better show the differences in the qualitative behaviors of each scenario. Initial production (first ~day) in the two highest pressure scenarios exceed 100 bpd, with the 1,900 psi scenario showing oil rates as high as 175 bpd. Cumulative recovery is shown in Figure 138. Cumulative production versus production time (top) and versus date (bottom) across the seven repressurization scenarios from Figure 136., with the top panel showing the cumulative curves synchronized and the bottom panel showing them in real time. Battelle | May 2026 133 Figure 135. Oil production rate (top) and cumulative production (bottom) for CO2 treatment vs. water-only treatment across five different repressurization scenarios. Figure 136. Average reservoir pressures in each of the seven different repressurization scenarios run to test the impact of miscibility. Colors are consistent across scenarios in Figure 129 and Figure 130. Battelle | May 2026 134 Figure 137. Oil production rate in the first 250 days for the 7 repressurization scenarios in Figure 128. Figure 138. Cumulative production versus production time (top) and versus date (bottom) across the seven repressurization scenarios from Figure 136. As can be seen in Figure 137, the rate curves appear to have two distinct features – a “miscible displacement” peak and an “immiscible displacement” peak. The miscible displacement peak occurs at early times in the scenarios where miscibility is reached and maintained for at least some portion of production (green curves – see Figure 136). This period of production shows very high oil rates until the reservoir depletes past MMP (1,300 psi), at which point production falls off. The miscible displacement period in the highest pressure scenario (dark green curve) lasts around 50 days, while this period in the 1,650 psi and 1,350 psi scenarios lasts a shorter amount of time. The blue curves, which correspond to scenarios that never exceed MMP at any point, do not have this miscible displacement period. Battelle | May 2026 135 In scenarios with a miscible displacement period of production, an immiscible displacement period follows. In Figure 137, this presents as a second peak in the oil rate. Immiscible production is the only peak in scenarios where MMP is not reached. The local maximum in this peak corresponds to the point at which the production well switches from a rate constraint (600 bpd) to a BHP constraint (AOF) – in other words, the point at which the near wellbore region is depleted to the point that the reservoir can no longer deliver 600 bpd to the producing well. Based on the figure, for higher pressure scenarios, the immiscible displacement peak appears to both have a lower local maximum and to occur later than observed in the lower pressure scenarios. The 550 psi scenario peaks around 60 bpd on day 55, while the 1,900 psi scenario peaks around 48 bpd on day 195. One might expect, then, that based on the high recovery seen during the miscible displacement period, the higher pressure scenarios would see greater cumulative recovery. However, as shown in the top panel of Figure 138, this is not the case. The scenarios with miscible displacement do have a higher cumulative recovery after the first 50 days or so, but the immiscible displacement in the lower pressure scenarios still results in a higher cumulative recovery over the first year of production. Starting with the lowest pressure scenario (dark blue), increasing pressure serves to decrease cumulative recovery, likely due to the higher volume of gas (and therefore greater reservoir volume touched) in the lowest pressure scenario. This trend reverses itself once miscibility comes into play (between the 1,150 psi and 1,350 psi scenarios – light blue to light green). From 1,300 psi and higher, increasing pressure serves to increase total recovery, causing the 1,900 psi scenario to have the second-highest total recovery of all of the models run after one year. Presumably, if the project could continue increasing the average reservoir pressure prior to CO2 injection, a high pressure scenario could overtake the 550 psi case as the optimal recovery scenario. However, given the expected formation fracture pressure within the reservoir, obtaining an average reservoir pressure higher than 1,900 psi would be difficult, leaving the 550 psi case as the optimal EOR strategy for the field based on modeling results. The time component of these different treatment scenarios is also an important piece of the puzzle. Until now, synchronized versions of the recovery curves have been the focus, looking only at time since initial production. However, when considering the additional time spent on repressurization (water injection) for the higher pressure scenarios, the 550 psi case becomes a clear-cut optimal scenario. The bottom panel of Figure 130 shows the cumulative recovery curves plotted in real time, where the timeline to repressurization is considered. Since lower pressure scenarios begin production at an earlier time, the operator is able to realize production revenue earlier. At the beginning of 2026, when the highest pressure scenario would just be turning over to production, the low pressure case (550 psi) has already netted 12,000 barrels of oil. Not only does this case result in the highest cumulative production, but that production occurs significantly earlier than the higher pressure cases, providing significantly higher value to the operator. It is clear, however, that in this dual-porosity model, miscible CO2 behavior causes a significant recovery uptick. The periods of miscible displacement in Figure 136 correspond to oil production rates that are as high as or higher than the peak rates seen in immiscible displacement. If pressures above MMP could be maintained, that scenario would almost certainly see the highest incremental recovery. One possible avenue for maintaining high reservoir pressure is recycling the non-oil components of the production stream – that is, pumping all produced fluid (except for oil) back into the reservoir to slow the rate of pressure depletion. Four additional Battelle | May 2026 136 simulations were conducted to analyze the impact of a recycled well-stream, with pre-CO2 injection reservoir pressures of 550, 750, 950 and 1,900 psi. The results of these simulations, co-plotted with results without recycling, are shown in Figure 139. Figure 139. Oil rate (top) and cumulative recovery (bottom) for 4 repressurization cases, both with and without well stream recycling. Top panel is plotted in real time, while bottom panel is synchronized to start of production. As can be seen in the figure, reinjection of the well stream (dashed lines) leads to a significant decrease in oil recovery compared to scenarios without recycling (solid lines). Oil rates are reduced by as much as 50% in matching pressure scenarios, and cumulative volume is reduced by as much as 8,000 barrels over the first year – twice that much over two years. This is true regardless of reservoir pressure – all four of the scenarios show a substantial decline in production. The primary reason for this is the high conductivity of the fracture/vug network. Re- injected fluid flows into this high-permeability system and, because the well creates a pressure sink during production, that fluid channels back to the production well, bypassing oil volumes in the matrix system. In a pure huff and puff scenario, as the fracture/vug system continues to deplete, fluids flow out of the matrix and into the high conductivity channels where they can easily flow into the production wellbore. Figure 140, which shows the water cut in the production stream for the same scenarios, illustrates this point. In the scenarios without recycling, water cut gradually decreases as matrix oil continues to be produced. In the scenarios with recycling, water cut holds steady around 96-98% throughout the entire production cycle. This suggests that any sort of flood scenario without significant mobility control is likely a sub-optimal recovery strategy. Further simulations investigating the efficacy of a waterflood without mobility control confirm this finding. Figure 141 shows cumulative recovery for four water-only scenarios with and without well stream recycling, and the same conclusion can be reached from these results. Battelle | May 2026 137 Figure 140. Water cut in four production scenarios with and without well stream recycling (reinjection and no reinjection, respectively). Figure 141. Cumulative recovery in water-only scenarios with and without water reinjection (orange and blue respectively). Full Complexity Model – Simulations to Inform Field Operations A final set of simulations was run on the history-matched model to help inform injection and production strategies during field testing. The first set of simulations was motivated by a loss in available CO2 volumes due to logistical challenges encountered during field testing. Three scenarios were run to assess the impact on recovery due to lower injected CO2 volumes. Figure 142 shows the oil rate (top) and cumulative oil (bottom) comparison of a huff and puff conducted with 3,000, 2,000, 1,000 and 0 tonnes of CO2 at an expected reservoir pressure of 550 psi (3,000 and 0 tonne scenarios correspond to the dark blue, black curves in Figure 143). As one might expect, a lower CO2 volume does lead to less incremental recovery than a higher CO2 volume. However, even the 1,000 tonne scenario shows an increase in recovery compared to the baseline case with no CO2, indicating the potential for project success even with a lower volume of CO2. Battelle | May 2026 138 Figure 142. Oil rate (top) and cumulative recovery (bottom) for huff and puff scenarios with four different CO2 volumes (0, 1,000, 2,000, 3,000 tonnes). Once the final procured volume of CO2 (2,000 tonnes) was determined, additional simulations were conducted to determine the optimum rate at which the CO2 should be injected into the reservoir. Figure 143 shows the oil rate resulting from a 2,000 tonne huff and puff where the CO2 was injected at 1,000 tonnes per day (tpd, black, solid line), 500 tpd (gray, dashed line) and 250 tpd (light gray, dotted line). A higher injection rate during the “huff” phase does appear to slightly increase peak oil rates during the “puff” due to slightly higher oil cuts (likely pushing more water away from the wellbore during gas injection). The difference between the scenarios was minimal, and it was determined that CO2 injection should proceed at as high a rate as reasonably practicable while staying well below estimated reservoir fracture pressure. Figure 143. Oil rate comparison of 2000t huff and puff with three different injection rates (1,000 tpd, 500 tpd, 250 tpd). A final suite of simulations was conducted to inform an optimal production strategy. Using the 2,000 tonne, 550 psi huff and puff scenario, four different production target rates were assessed – 600 bpd, 400 bpd, 300 bpd, and 200 bpd. These rates correspond to the anticipated pump Battelle | May 2026 139 operations, with the overall constraint on production rates still being the AOF condition (i.e., once the reservoir can no longer deliver the target rate, well constraint becomes BHP of 0 psig). Figure 144 shows the oil rate (top), cumulative oil (middle) and water cut (bottom) of these four production scenarios. The results indicate that a higher production rate produces more cumulative oil. The 400 bpd and 600 bpd perform somewhat similarly, but lower production rates perform much worse over the first 1.5 years. In some cases, a lower rate constraint is able to reduce water cut at a production well, but that does not appear to be the case within the Lee field. Water cut is actually lowest for the highest production rate scenario. Additional model outputs were assessed to help plan for impending field operations. The gas production rate over time for each case was assessed to inform surface compression unit procurement and operation, and the operating BHP was assessed to inform pump operation at the production well (since artificial lift was needed). Figure 144. Oil rate (top), cumulative oil (middle) and water cut (bottom) for four different production scenarios of a 2000t huff and puff. Modeling efforts accomplished in Task 4 provided useful results to help inform field operations. The sensitivities assessed ensured that operational planning proceeded with minimal delays and functional changes when necessary. Key uncertainties that remain include the exact impact of compositional simulation and surfactant inclusion on the results. Relative permeability, capillary pressure curves and vertical to horizontal permeability anisotropy could all be better Battelle | May 2026 140 constrained by field data. In addition, the impact of dynamic fracture systems was not investigated and may serve to play a critical role in accurately modeling these systems. Future modeling efforts would do well to assess these remaining uncertainties. Battelle | May 2026 141 6.0 Task 5: Field Injection Tests Reporting on this task, which focused on implementation of the CO2-EOR design to inject chemically enabled CO2 into the TBR, collect advanced characterization data from the injection well, and establish baseline measurements, will focus exclusively on activities at the Lee 26 field since no injection was attempted at the McCann 1-20 well. 6.1 Subtask 5.1: Site Preparation at the Marshell 1-35, Lee 26 Field, Calhoun County Details of field selection method, stimulation method, and CO2 procurement are described in Section 1.0 of this document. This section will focus on work completed to prepare the Lee 26 field for surfactant-enhanced CO2 stimulation. Production Facility Design Surface production equipment required to produce the Fischhaber 2-35 following CO2 injection was sited at the adjacent Marshell 1-35 well pad. This pad already housed oil and gas processing facilities from the primary producing phase of the field, and much of the equipment already in place was re-purposed for project needs. The purpose of the production facility was to control the rate and pressure of flowback and to separate the well stream into produced oil (for sale), produced water (for reinjection or disposal) and produced gas (for reinjection). Additionally, the facility was designed to control and capture gas emissions resulting from pressure let-down as oil reaches the stock tanks. Mainly, the facility consists of the following process units: choke valve, mainline compressor, heater treater, stock tanks, and vapor recovery compressor. The specification of each process unit was based on its operating pressure and flow capacity. The notional process flow diagram of the facility is given in Figure 145. The production facility controls the rate of production and reduces pressure using a choke valve near the wellhead. The choke valve has a variable orifice that can be adjusted manually or automatically, based on feedback from a flow or pressure measurement device. The well stream directly enters the heater treater, which was already located on the Marshell pad. The heater- treater operates similarly to the high pressure separator, but includes a heated compartment to drive volatile components out of the liquid stream, with such volatiles mixing into the gas phase exiting the top of the vessel. Produced gas effluent from the top of the vessel is sent to a multistage compressor for re-injection directly into the Marshell well. The heater-treater is fitted with float operated liquid dump valves for both water and oil, and the resulting liquid streams are sent to their respective stock tanks. From the stock tanks, the water may be transferred by pump for re-injection into the Beattie 1-27 and/or Marshell 1-35, and the oil may be trucked to market. Tank vapors, if combustible, are routed to a flare stack with a standing pilot for safety. This approach maximizes the amount of flash gas that can be recovered from the well and all but eliminates venting and flaring losses. Battelle | May 2026 142 Figure 145. Process Flow Diagram of proposed Production Facility For the most part, the facility is conventional in design and operation philosophy and presents low technical risk. Because the well requires artificial lift to bring the produced oil to surface, the initial rate of production was estimated based on reservoir dynamic model simulations and all downstream equipment was sized based on that initial production rate. Additionally, the production facility located at the Marshell pad rather than the Fischhaber pad, which will result in a flowline of 0.5 miles in length. The well was expected to produce paraffin waxes, which may occlude the flowline and require periodic maintenance (flushing) to clear. These issues are mostly operational and not out of the ordinary, but the paraffin mitigation plan required special attention to avoid plugging the underground flowline, which is not easily serviceable. The most prevalent unknown during facility design was the behavior of CO2 throughout the process, and how the complex phase behavior of CO2 will influence the performance of lift, separation, and compression equipment. Ultimately, since relatively low volumes of water and CO2 were produced, CO2 phase behavior was not an issue during the production period. The Marshell production facility was largely completed in the first quarter of 2025. The facility includes all the elements of the process flow diagram of Figure 145. Battelle | May 2026 143 Figure 146. Marshell Production Facility as of March 4, 2025 Facility Design Outcomes and Lessons Learned Because the Fischhaber 2-35 well did not produce measurable oil, there was limited data available to analyze the performance of the facility implemented at the Marshell 1-35 pad. However, the discussion below reflects lessons learned during this study that can be applied to future CO2-EOR development. For pipeline and facility designs at scales from individual pads to full-field systems, a process simulation should be initiated prior to construction, preferably using a Pressure-Volume- Temperature (PVT) study from offset production when available. In the absence of such data, an estimated fluid composition and anticipated production rates should be developed. The primary objective of the process is the effective separation of gas, oil, and water through the application of gravity, residence time, and heat. Gas is routed to the pipeline system, while oil and water are directed to storage tanks. For oils with higher API gravity, the final oil–water separation stage typically requires heat application to remove components that would otherwise vaporize at atmospheric pressure. Vapors generated from treaters, towers, and tanks must be captured and sent either to a flare system or to flash compression, depending on economic considerations. The design and installation of production facilities requires extensive experience, as proficiency is generally achieved only after multiple iterations and sustained field exposure, so facility designs that have worked well in the TBR trend in the past should be emulated, but updated and optimized for safe CO2 handling. Foundational concepts associated with effective facility design are widely documented and remain universally applicable across production operations. Within industry, many individuals with production backgrounds possess familiarity with the established methods; however, organizations lacking seasoned expertise may require additional guidance from engineers and facility operators with experience in the TBR as well as engineers and operators with experience handling CO2. Carbon dioxide introduces unique challenges due to limited industry-wide experience with its production and reinjection. In the near term, flow assurance represents the primary concern. Carbon dioxide, water, and methane have a tendency to form hydrates under elevated pressure Battelle | May 2026 144 and moderately low temperature conditions, particularly near control valves or in downhole environments where CO₂ may undergo a phase change from liquid to vapor, resulting in temperature drops below hydrate formation thresholds. Common mitigation strategies include gas dehydration (up to the critical point), methanol injection, heat application, and staged pressure reduction. Over longer periods, corrosion emerges as a key issue when CO₂ and water coexist. The choice between a materials-based solution and a monitoring-based solution should be driven by equipment age, evaluation of safety thresholds, and project economics. Carbon dioxide also presents additional hazards due to its non-combustible nature and its ability to displace oxygen. Vapors typically managed safely through combustion can transition between combustible and non-combustible states as CO₂ concentration increases, potentially extinguishing pilot flames and allowing dangerous accumulations of flammable gases. To address this risk, facility designs should incorporate methods for removing CO₂ from the solution prior to emissions control systems. Indirect heat application within an oil heater-treater and the installation of a low-pressure vapor recovery tower, upstream of storage tanks, can be effective approaches to ensure tank vapors remain combustible and suitable for flaring. Because CO₂ is an asphyxiant, sufficient air quality monitoring is required to protect personnel. When present as a cryogenic liquid, CO₂ will generate significant pressure increases if warmed while confined, reaching pressures of approximately 750 psi. Therefore, relief devices must be properly installed to safeguard equipment. Additionally, liquid CO₂ released to ambient conditions can rapidly form dry ice, which may temporarily block a line break before violently discharging as internal pressure increases. Carbon dioxide equipment must be depressured to gas phase and warmed prior to maintenance activities. Operational personnel should receive appropriate training to understand and manage these hazards. From a safety perspective, the most common risks for CO2-EOR operations involve loss of containment and the accumulation of flammable gas. API Recommended Practice 14C provides comprehensive guidance on these hazards for typical installations and represents a durable reference for ensuring the safe operation of production facilities. 6.2 Subtask 5.2: Baseline Site Characterization and Monitoring Spectral Pulsed Neutron Logging: Baseline data collection, January 2023 SPN logging was planned to evaluate two aspects of the Lee field to evaluate pre-injection reservoir saturations of oil, gas and water, to be assessed using Baker-Hughes' SPN tool, with the intention of making repeated runs with the tool during water and CO2 injection periods to understand change in reservoir saturation over time. Logging commenced during the week of January 23, 2023. Baker’s logging crew mobilized to Fischhaber 2-35 and ran in hole with the deep shear wave tool (Figure 147). During down- logging (running into the tool with sensors on to provide tool calibration), the tool encountered an unexpected plug at 4,600 ft measured depth, 18 ft above the expected plug-back total depth. A fluid column was expected in the Fischhaber 2-35, but no fluid was present upon commencement of logging operations. Both the SPN and deep shear wave tools require a fluid column to gather data, so logging of the Fischhaber well was put on hold, and the logging crew mobilized to the Marshell 1-35. Battelle | May 2026 145 Similar conditions were encountered at the Marshell 1-35, with insufficient fluid column to acquire logs. An attempt was made to trickle water down the hole to establish a temporary fluid column, but this method was unsuccessful. Figure 147. Baker Hughes crew preparing to run wireline logs at the Fischhaber 2-35 well. Baker-Hughes was able to run the SPN tool in a gas environment with a proprietary sleeve to run the tool inside. That sleeve was shipped to Michigan, and SPN logs were acquired in a gas- filled Marshell 1-35 wellbore on February 7, 2023. Calibrated logs were delivered to Battelle and West Bay on March 2, 2023, and revealed a lack of cement behind casing across most of the logged interval. That lack of cement created a fluid gap behind casing, thwarting the SPN tool’s ability to accurately assess reservoir saturations. With this well condition, Battelle and West Bay concluded that running repeat logs with the SPN tool over the course of repressurization and EOR would not yield useful data in the Marshell 1-35, so no further logs were run in this well. However, SPN logs were collected in the Fischhaber 2-35 both before and after CO2 injection, as discussed later in Section 6.2. Lee 26 Field Baseline Pressure Testing Baseline pressure monitoring was conducted at the Lee field in early 2023 to establish baseline conditions prior to the start of water injection. This was done through the use of fluid shots to test fluid levels in the wells and dip-in pressure readings with wireline gauges. Fluid shots taken at the Marshell well on January 6 and January 26, 2023, showed consistent fluid levels of 4,812 ft and 4,843 ft TVD. Assuming a freshwater gradient, this corresponds to a bottomhole pressure of between 30 and 52 psig, suggesting pressure did not decrease substantially since the readings taken in 2018. Fluid level was also measured at Jacoby on March 15, 2023, at 4,018 ft TVD. Assuming freshwater gradient, this corresponds to a bottomhole pressure of 350 psig. This measurement was fairly consistent with fluid levels taken at the well in 2020. At the time, these values were regarded as anomalous, likely due to wellbore completion elements skewing the fluid shot response. While this still may be the case, the low density fluid column observed in the Fischhaber well during gradient runs later in the project (see Section 5.3) suggests these fluid shots could have been seeing a column of this same low density condensate. Battelle | May 2026 146 Dip-in pressure measurements were taken at the Marshell, Beattie, and Jacoby wells on June 1, 2023. These measurements are much more accurate than pressures interpreted from fluid levels. Pressures of 3 psig, 107 psig, and 40 psig were measured at the Beattie, Jacoby and Marshell wells, respectively. The low pressure reading at Beattie was taken at a depth much shallower than the reservoir top and was likely not submerged due to low reservoir pressures. The reading at Marshell was consistent with the fluid level interpretations taken in January. Pressure at the Jacoby well was still anomalously high relative to the rest of the field, though not nearly as high as suggested by fluid level data. This elevated pressure may have been due to constrictions in the well from completion elements, or it may have been caused by a lack of connection to the rest of the field after closure of a high conductivity pathway that served as the primary connection to the rest of the reservoir. Existence of this dynamic fracture pathway is suggested by pressure data gathered during water injection operations, as discussed in the next section. 6.3 Subtask 5.3: Injection Operations No injection operations were attempted at the McCann 1-20 well. At the Lee 26 field, injection of CO2 was completed in March and April of 2025, preceded by approximately17 months of water injection to repressurize the field prior to CO2-EOR operations. The Lee field injection operations are detailed below. Lee 26 Water Injection Due to the low pressures present in the depleted Lee 26 reservoir, water injection was conducted to repressurize the field prior to CO2 injection. The goal of this water injection was to raise average pressure of a sector of the field (between Marshell and Fischhaber, the proposed injector-producer pair) above MMP to facilitate miscible CO2 treatment of the residual oil. Water injection began at the Marshell well at the end of September 2023, using water sourced from a nearby remediation well. Injection operations were conducted only through the Marshell to minimize watering out of residual oil in the field. Water injection concluded in February 2025, in preparation for CO2 injection. Daily water injection data from the Marshell 1-35 is shown in Figure 148. Injection was maintained at a rate between 1,500-2,500 bpd; a total of 957,870 barrels of water were injected over 17 months. Figure 148. Water injection rate data from the Marshell 1-35 water injection well. Note rate decrease during the second half of 2024. Injection rate subsequently increased after repair of purge water pump at water source. Battelle | May 2026 147 Real time wellhead pressure (WHP) monitoring data was collected via a supervisory control and data acquisition (SCADA) network at each of the wells. The SCADA system came online in September 2023, concurrent with the start of water injection. Tubing pressure data collected during water injection is shown in Figure 149. During the first six months of injection (9/2023 through 3/2024), while all wells besides Marshell were shut in, the WHP data provided an excellent view of pressure front migration through the reservoir. The first marked increase in pressure was observed at the Beattie well in early October 2023, indicating the presence of a high-conductivity channel between the Marshell and Beattie wells. WHP at the Fischhaber began rising slowly in mid-late October 2023, and pressures at the Jacoby never responded to water injection during this period. In March 2024, field operators began the sampling campaign for water samples from offset wells, requiring well blowdowns and other manipulations, at which point wellhead data became an unreliable proxy for subsurface pressures. Pressures at Marshell were negative for the entirety of water injection, as it was continually injecting on a vacuum. Casing pressures were also monitored during this period to ensure well integrity – nothing of note was observed in this data. Figure 149. Wellhead pressure data collected during water injection phase. Spikes in Jacoby data correspond to well work operations. Downhole pressure monitoring was also conducted at the Lee 26 field from September 2023 through the end of the production testing in November 2025. Bottomhole gauges were installed in each of the four wells at different times and were pulled periodically throughout injection operations to retrieve data. In addition, the ESPs installed in the Fischhaber 2-35 well during production testing (see Section 6.2) provided pressure data during the post-treatment production period. The final realized schedule of gauge installation and retrieval is shown in Table 25. Raw pressure data from the downhole gauges during water injection are shown in Figure 150, and a datum corrected version of the same data (flattened on the pressure prior to production startup on August 28, 2025) is shown in Figure 151. Pressure steadily increased at the injecting Marshell well from the beginning of data collection. A marked response was not observed at the Fischhaber well until March of 2024, at which point the Marshell and Fischhaber pressure curves increased in parallel (likely indicative of boundary-dominated flow). Large drops in pressures at various times during injection correlate with operations on surface infrastructure. Battelle | May 2026 148 Table 25. Downhole gauge installation and retrieval schedule at the Lee 26 field. Well Beattie 1-27 Jacoby 1-26 Fischhaber 2-35 Marshell 1-35 Gauge Installed: 1/2/2025 11/5/2024 9/13/2023 11/6/2023 Gauge 4/21/2025 3/5/2025 1/2/2024 1/2/2024 Retrieved: 9/23/2025 4/21/2025 4/11/2024 4/11/2024 11/6/2025 9/23/2025 7/10/2024 7/10/2024 11/6/2025 11/5/2024 11/5/2024 12/6/2024 9/23/2025 1/27/2025 11/6/2025 3/3/2025 4/16/2025 5/13/2025 6/16/2025 *6/26/25 - 11/4/25 monitored by ESP Figure 150. Bottom-hole pressure data from the Fischhaber (orange), Marshell (blue), Jacoby (black), and Beattie (red) wells during water injection. Battelle | May 2026 149 Figure 151. Datum-corrected bottom-hole pressure data from the Fischhaber (orange), Marshell (blue), Jacoby (black), and Beattie (red) wells during water injection. A significant change occurred in the pressure response across the field in early January 2025. Prior to this date, pressures at the Jacoby well are substantially lower than elsewhere in the field (including at the Beattie well). The direction of pressure gradients within the field suggests that fluid was flowing through a channel from Marshell and Fischhaber toward the Beattie well, then flowing back to the Jacoby. The combination of lower pressures and muted pressure responses (i.e. slower rates of change) at Jacoby before January 2025 suggests lower conductivity connection with the reservoir, as compared to the rest of the wells. From January 8-10, 2025, a 250 psi increase was observed at the Jacoby well, correlating with a slope change in pressure increases at the other three wells. This likely corresponds to the opening of a high conductivity fracture between the Jacoby section of the reservoir and the area around Fischhaber. After this date, pressures at Jacoby and Fischhaber were nearly identical, pressure changes were nearly instantaneous with changes at Marshell, and fluid flows from east to west through the field, according to pressure gradients. Pressure fell off substantially in the weeks preceding CO2 injection, while the field was shutin, indicating a >200 psi difference pressures while injecting and static reservoir pressures. This is likely due to fluid flow from the high conductivity fractures and vugs into the matrix dolomite porosity network. Bottomhole temperature was also monitored using the downhole gauges in each well. Temperature data from the water injection period is shown in Figure 152 below. Temperatures at both Beattie and Jacoby were roughly constant for the period when gauges were installed and show the impact of installation depth in their temperature values (i.e. the shallowest gauge, Beattie, shows the lowest temperatures of all non-injection wells). Temperatures at Marshell were much lower than elsewhere in the field (due to water injected from surface) and increased rapidly during periods of low/no injection. These temperatures also cycle seasonally from the injected water. Fischhaber temperatures fluctuated slightly throughout injection but mostly remained around 100-102 degrees Fahrenheit. Battelle | May 2026 150 Figure 152. Bottomhole temperature data across the Lee field during water injection. Fluid level shots were also collected from the three non-injection wells during the second half of 2024 and are shown in Figure 153. Fluid levels at the Fischhaber 2-35 well and Jacoby 1-26 well were elevated relative to the fluid level at the Beattie 1-27 at the beginning of the fluid shot campaign in July, but the Beattie level rapidly rose to meet the level of the other two wells in August. The three wells, more or less, followed the same trend from this point forward. The lagging fluid level in the Beattie 1-27 in July 2024 may be the result of a closed fracture restricting fluid flow to that region of the reservoir, or the result of fluid preferentially flowing toward the Fischhaber well at that point during repressurization. Interestingly, this timeline does not correlate with the fracture opening in early January 2025 observed in bottomhole pressure data. This could indicate that there was more than one fracture opening and closing in the system based on different pressure states in the reservoir. Battelle | May 2026 151 Figure 153. Fluid level shots from non-repressurization wells in the Lee 26 field, July 2024 through December 2024. Given the failure of baseline SPN logging to provide context on reservoir saturation changes, the team deployed water tracers during the repressurization period. These water-soluble tracers, provided by ProTechnics, are designed to provide data around reservoir connectivity and the transient time of water from the injection well to the non-injection wells in the field. Figure 154 illustrates the geographic relation of the wells to each other and shows the conceptualized flow pathway from the Marshell injection well to the other wells in the field. Twenty-four gallons of tracer was injected into the Marshell well on September 29, 2023, upon commencement of repressurization operations. After five months of injection, water levels rose sufficiently to allow fluid samples to be captured at the Fischhaber well in March 2024, and sampling was initiated in the Jacoby 1-26 well a month later in April 2024. The first successful water samples to be collected at the Beattie well occurred in September 2024. Fluid samples were collected bi-weekly. The resulting tracer concentrations (in parts per trillion) and salinity concentrations (in parts per million) are shown in Figures 155 and 156, respectively. Tracers were present in every sample from all three wells, with the highest concentrations of both tracer and salinity at the Jacoby well over the 10-month sampling period. These results indicate strong transmissibility across the reservoir and likely also indicate that the main tracer slug had already passed through the Fischhaber well before the wellbore built a sufficient water column to be sampled. Likewise, with fluid salinity concentrations significantly lower at the Fischhaber well than at Jacoby, it is believed that the fresh repressurization water injected at the Marshell well had reached and diluted the near wellbore brines in the Fischhaber reservoir before fluid sampling occurred. Concentrations of both tracer and salinity at the Beattie well were lower than the Jacoby and higher than the Fischhaber until the final sample collection, Battelle | May 2026 152 indicating that injection fluids moved past the Beattie well and were beginning to dilute with continued injection. It may also indicate that the Beattie and Fischhaber wells are in more direct hydraulic communication with one another, and that the Jacoby well received an isolated pulse of tracer-laden fluid early on during repressurization, but that the conduit that had previously been open has now been closed or blocked, isolating the Jacoby wellbore from continued freshwater flow. Tracer concentration at the Jacoby well dropped by two orders of magnitude from October 15 to November 5, while salinity increased substantially at both Jacoby and Beattie over the same time period. Salinities varied substantially at Jacoby from that time through the end of sampling. Figure 154. Fluid tracer injection and sampling diagram, Lee 26 field. Tracers were injected into the Marshell well, and sampling will occur at the remaining 3 wells to understand flow pathways of injected water across the field. Figure 155. Tracer concentration (parts per trillion) in bi-weekly samples from the Fischhaber 2-35, Jacoby 1-26 HD2, and Beattie 1-27 HD1 wells. Note the logarithmic scale on the y-axis. Battelle | May 2026 153 Figure 156. Salinity data (parts per million) captured from the Fischhaber 2-35, Jacoby 1-26 HD2, and Beattie 1-27 HD1 wells. Tracer concentrations and salinity variations in these wells over the course of the injection period lend further weight to the idea that the Lee 26 field (and likely TBR fields generally) have highly complex, fracture- and vug-dominated plumbing. The fact that water moved across the field before effective sample collection could be completed also indicates that highly permeable pathways extend across the field. It is also possible that the large shifts in both salinity and tracer concentration over the course of the repressurization period (particularly the major salinity shifts from September 2024 to January 2025, Figure 156) were caused by pressure-dependent opening and closing of fractures within the reservoir and/or other complex flow pathways (spillpoints, etc.) throughout the system. Lee 26 CO2 Injection Carbon dioxide injection operations commenced at the Lee 26 field with an injection dry run meeting during the first week of January 2025. The meeting constituted two parts: a morning ‘injection on paper’ session to review technical details including safety, operational procedure, and pressure monitoring (Figure 157), and an afternoon visit to the Fischhaber 2-35 pad to walk down the CO2 huff & puff site and the Marshell 2-35 pad to walk down the production and CO2 reinjection site (Figure 158). Battelle | May 2026 154 Figure 157. January 7th production dry-run meeting, including representatives from West Bay Exploration, Battelle Memorial Institute, Baker-Hughes, Messer Americas, Hadra Safety Consulting, and M&S Oil Well Services. Figure 158. Safety and operational walkdown of the Fischhaber 2-35 site. Well pad has been improved with gravel to support the weight of CO2 storage tanks. The wellhead is inside the fenced enclosure in the left of the image. Battelle | May 2026 155 CO2 Injection Mobilization Installation of packer and tubing in the Fischhaber 2-35 well was initiated on February 11, 2025 followed by a successful mechanical integrity test on February 17, 2025. Messer America started mobilization of CO2 storage tanks, injection skid, hoses, and personnel trailer to the site on February 18, 2025 and completed mobilization on February 21, 2025 (Figure 159). Figure 159. Array of 80 ton CO2 storage tanks at the Fischhaber 2-35 pad. CO2 Storage Operations Stage 1 of CO2 deliveries was initiated on February 22, 2025 and ceased on February 27, 2025 to protect local roads during spring frost thaw. During those five days, Messer delivered 915 tonnes of CO2 (Figure 160). Figure 160. CO2 storage operations at the Fischhaber 2-35 pad. Battelle | May 2026 156 Surfactant Delivery & Dilution Baker-Hughes Petrolite delivered 1,300 kg CETAC-30 surfactant on February 28, 2025 (Figure 161). The CETAC was diluted to 0.5% weight concentration using approximately480 barrels of water (Figure 162). Figure 161. Delivery of CETAC-30 surfactant. Figure 162. Storage tank and pumping unit used for dilution and injection of surfactant-laden water. Battelle | May 2026 157 Stage 1 Surfactant and CO2 Injection Operations Injection operations (the ‘huff’ portion of huff and puff) commenced on March 3, 2025, with injection of 422 barrels of surfactant laden water into the Fischhaber 2-35 well. Injection was completed at 2.3 barrels per minute (bpm), with the well taking the entire volume on vacuum (Figure 163). Messer then pressure tested the CO2 injection lines and initiated CO2 injection, ramping up from one bpm to two bpm. An issue was identified with the injection flow meter, and operations were shut down for the day to address the issue, which was found to be a dry ice blockage. Injection resumed on March 4, 2025 and proceeded with no further operational pauses (Figures 164 and 165). Injection was completed March 5, 2025 (Figure 166), with injection pressure below 200 psi for the majority of the injection period. Figure 163. Gauges on surfactant pump truck during surfactant injection. Injection completed at an average of 2.3 barrels per minute, with the well taking all fluid on vacuum. Battelle | May 2026 158 Figure 164. Fischhaber wellhead connected to CO2 injection spread. Figure 165. CO2 injection spread (left) and wellhead (right) during CO2 injection. Frost build-up on these components during injection is due to condensation and freezing of ambient moisture caused by cold cryogenic liquid and/or Joule-Thompson cooling. Battelle | May 2026 159 Figure 166. Pumping pressure and total injected volume for Stage 1 of CO2 huff and puff at the Fischhaber 2-35 well. Upon completion of Stage 1 injection, the Fischhaber 2-35 was shut in, the injection spread was sealed, and downhole gauges were re-installed in the well. The pad was then abandoned until initiation of operations for Stage 2 starting on April 7, 2025. Stage 2 Surfactant and CO2 Injection Operations Stage 2 CO2 delivery and storage operations were initiated on April 7, 2025 and proceeded through April 13, 2025, with a total of 1,011 tons of CO2 delivered and stored (Figure 167). A short standby period was needed before the team was able to pull bottom hole gauges from the Fischhaber (Figure 168). Stage 2 injection operations commenced on April 16, 2025, with removal of the downhole gauges. On April 17, 2025, 422 barrels of surfactant laden water into the Fischhaber 2-35 well, followed by CO2 injection. Initial injection was completed at about three bmp, with the well taking the entire volume at between 250 to 300 psi discharge pressure, well within the expected range based on performance during Stage 1 (Figure 169). Metering on the injection spread failed after about 550 tons of CO2 had been injected, and the Messer team manually monitored tank pressure and volume to complete the remainder of injection at three to four bmp . Injection was completed April 18, 2025. Battelle | May 2026 160 Figure 167. Messer Americas storage hand making connection between CO2 transport tank and on-site CO2 storage. Figure 168. Example of normal, small-volume CO2 venting during storage and stand-by. Battelle | May 2026 161 Figure 169. Pumping pressure and total injected volume for Stage 2 of CO2 huff and puff at the Fischhaber 2-35 well. Upon completion of Stage 2 injection on April 18, 2025, the Fischhaber 2-35 was shut in and the injection spread was demobilized from the pad, with full demobilization completed on April 24, 2025. The pad was then abandoned and the CO2 allowed to soak in the reservoir until initiation of flowback operations on May 28, 2025. Pressure Monitoring Pressure monitoring continued through CO2 injection using both the installed downhole gauges and SCADA WHP monitoring system. Datum-corrected BHP data (using the same datum correction as the water injection data) is shown in Figure 170. Fischhaber bottomhole data is missing during the actual injection as the gauge had to be removed during this period. BHP at the Fischhaber well increased to as high as 1,050 psi during the first phase of CO2 injection, with a long pressure falloff observed during the soak periods. Pressures at the other three wells displayed very similar qualitative behavior, indicating high connectivity across the formation. Pressures at the Marshell increased further than at Jacoby, likely due to the Marshell’s proximity to the boundary of the field. Pressures at Beattie decreased by the lowest amount due to being furthest from the injector. Pressures remained elevated at the Fischhaber well relative to the rest of the field during the falloff period until approximately one month after injection, at which point pressure decreases to a level more consistent with the other three wells. This may be due to the presence of foam near the Fischhaber well and subsequent foam breakage one month after injection. Beneath the pressure signatures due to CO2 injection, a background pressure decrease is observed in the data from continued falloff from water injection and fluid flow between the vug/fracture and matrix systems. For this reason, pressures during the second injection period were not as high as those observed during the first injection. Battelle | May 2026 162 Figure 170. Datum-corrected bottomhole pressures at all four wells during the CO2 injection and soak periods. Wellhead pressure data from the same period is shown in Figure 171. Note that the Fischhaber data is shown on a separate y-axis from the rest of the wells. WHP at the Fischhaber reached 650 psi during the first phase of CO2 injection and roughly 600 psi during the second phase. The high variability at the Fischhaber during the first injection period is due to poor signal connectivity with the sensors. Pressure at the Jacoby well is increasing from the start of CO2 injection, likely indicating gas influx into the wellbore. A sharp increase is seen coincident with the injection operations, which falls back off toward the background increasing trend. At the Beattie well, a similar (though smaller) response was observed coincident with the first injection phase, but no substantial response is seen during the second injection phase. No significant pressure response is observed at Marshell. Figure 171. Tubing pressures at all four wells during CO2 injection and soak periods. Note that Fischhaber data is plotted on a different axis. Battelle | May 2026 163 Pressure Transient Analysis Using gathered pressure data, a pressure transient analysis (PTA) of two falloff periods following CO₂ injection into the Fischhaber well was conducted. The analysis focused on identifying the dominant flow regime observed during each falloff period using rapid pressure- derivative diagnostics. Emphasis was placed on flow-geometry interpretation (e.g., linear versus radial behavior) and discussion of plausible flow scenarios consistent with the observed response, based solely on the pressure data, in order to characterize the pressure-transient behavior associated with CO₂ injection at this well. The PTA analysis was performed using IHS WellTestTM 2024.1. The primary input for this analysis was downhole pressure data acquired from the gauge installed in the Fischhaber well. Figure 172 shows the pressure data used in both falloff periods for analysis. Pressure measurements were recorded at one-minute intervals during the falloff periods. CO₂ injection rates were used as input in units of tonnes per minute. Pressure data during the active injection periods were not available; therefore, as an assumption, the first recorded pressure at the start of each falloff period was used to represent the pressure at injection shut-in. Additional input data included reservoir properties such as temperature, thickness, post-waterflood fluid saturations, and oil properties, which are summarized in Table 26. These properties were derived from a combination of field data and modeling results from Task 4. Fluid and reservoir properties were assumed to be constant over the analyzed falloff periods. Figure 172. bottom hole gauge pressure data provided Battelle | May 2026 164 Table 26. Input parameters into model Input Value/Range Reservoir Temperature 98 F Gas Saturation 7% Oil Saturation 30% Water Saturation 63% Oil API 40 Bubble Point Pressure 1620 psi Gas Oil Solution 350 scf/bbl Thickness 523 ft Porosity 10% Figure 173 shows the pressure-derivative and type-curve match for the first CO₂ injection falloff period. Pressure-derivative and type-curve plots are diagnostic tools used in pressure transient analysis to identify flow regimes based on the slope and shape of the pressure response as a function of time. The slope of the pressure derivative is particularly important because it reflects the underlying flow geometry controlling pressure diffusion in the reservoir. During the main portion of the falloff period, the pressure derivative exhibits a 1/2-slope behavior, consistent with a linear-flow regime, indicating flow toward a high-conductivity planar feature rather than matrix- dominated radial flow. At late time, the derivative trend steepens toward a slope approaching unity, which may suggest the influence of a boundary or transition to a different flow regime. However, due to data scatter and limited late-time stabilization, this interpretation should be considered indicative rather than conclusive. Figure 173. Derivative plot for first falloff cycle While a zero slope in the pressure-derivative plot is diagnostic of matrix-dominated radial flow, such behavior is not observed in this dataset. Instead, the derivative response is dominated by a sustained 1/2-slope, indicating linear flow geometry throughout the main portion of the falloff period. A 1/2-slope linear-flow response may be associated with flow toward a planar fracture or a linear channel system, and PTA alone cannot uniquely distinguish between these two scenarios. In linear channel flow, a transitional radial-flow period is often expected prior to the onset of linear behavior; however, no clear zero-slope radial-flow transition is observed here. This suggests either the presence of a very narrow channel that suppresses observable radial flow or flow toward a dominant planar high-conductivity feature, such as a pre-existing or Battelle | May 2026 165 injection-induced fracture. Battelle | May 2026 166 A similar pressure-derivative response is observed during the second CO₂ injection falloff cycle (Figure 174), confirming the observations from the first falloff period. Early-time data are affected by noise and do not exhibit a stable diagnostic trend. Similar to the first cycle, a zero- slope derivative indicative of matrix-dominated radial flow is not observed. Instead, the derivative response is dominated by a sustained 1/2-slope, consistent with a linear-flow regime throughout the main portion of the falloff period. Both linear fracture and linear channel type- curve models provide reasonable matches to the data, reflecting the inherent non-uniqueness of linear-flow interpretation using PTA alone. The absence of a clearly identifiable transitional radial-flow regime suggests either a narrow channel geometry that suppresses observable radial flow or flow toward a planar high-conductivity feature, such as a pre-existing or injection- induced fracture. The repeatability of this response across both falloff cycles supports the conclusion that linear flow dominates the pressure-transient behavior at the Fischhaber well. Figure 174. Derivative plot for second falloff cycle In conclusion, the pressure transient analysis of both CO₂ injection falloff periods at the Fischhaber well consistently indicates a dominant linear-flow regime, as evidenced by a 1/2- slope pressure-derivative response. The absence of a zero-slope radial-flow signature suggests that matrix-dominated flow does not control the pressure behavior during the analyzed periods. The repeatability of the linear-flow response across both falloff cycles increases confidence that linear flow is the governing flow geometry during CO₂ injection and falloff at this well. The sustained 1/2-slope linear-flow response observed in both falloff cycles indicates flow toward a high-conductivity planar or elongated feature. Such behavior may be explained by either a planar fracture or a narrow linear channel system, as PTA cannot uniquely distinguish between these geometries. However, linear channel flow typically exhibits a transitional radial-flow regime prior to the onset of linear behavior, which is not observed in the analyzed data. The absence of a discernible radial-flow transition suggests either a very narrow channel geometry that suppresses observable radial flow or flow toward a dominant planar fracture. Both linear fracture flow and narrow channel flow reflect the presence of a high-permeability, high- transmissibility pathway that dominates pressure diffusion relative to the surrounding matrix. The analysis of transient pressure data following CO2 injection supports the subsurface interpretation of the Lee reservoir and other TBR fields. In a reservoir with a system of interconnected, high-conductivity vugs and fractures flowing through low permeability matrix, the 1/2-slope behavior could be created by these high conductivity channels. The feature that is connected to the Fischhaber appears to be one of these high conductivity channels, within which matrix-dominated flow does not appear in the month of data analyzed. If the residual oil is Battelle | May 2026 167 all or mostly located within the matrix porosity system, as is suspected, the lack of matrix flow could help explain why no oil was produced from the Fischhaber well. While the observed 1/2- slope response clearly supports linear-flow behavior, PTA alone cannot uniquely distinguish between a planar fracture and a linear channel system as the underlying high-transmissibility feature. This uncertainty could be reduced by acquiring continuous downhole pressure data during the injection periods, which would enable direct analysis of injection period and improve early-time diagnostic interpretation. Additional improvement could be achieved by applying similar high-resolution pressure data and PTA workflows to other wells in the field to assess the consistency of flow regimes across the reservoir and better constrain the dominant flow mechanisms. Also, PTA analysis during falloff period of waterflooding (prior to CO2 injection) could provide additional information. Temperature Monitoring Bottomhole temperature data from the CO2 injection period are shown in Figure 175. The data illustrates the impact of CO2 injection operations on formation temperature. The Beattie and Jacoby wells show very little variation in temperature because the gauges are located within the vertical section of those wellbores. As a result, any temperature impacts seen at bottomhole must diffuse through the fluid column in the wellbore before they are seen by the gauges. There are small increases in temperature observed at these wells coincident with the observed pressure increases from injection. This is due to warmer fluid moving up the column in response to the increased bottom-hole pressure. Temperatures at Fischhaber decrease sharply coincident with cold liquid CO2 injection and warm rapidly back to initial formation temperatures (low temperatures in mid-February 2025 are from a wellbore gradient survey and are not representative of bottom-hole conditions). Temperatures at the Marshell well show a significant response to CO2 injection. Temperatures decrease coincident with observed pressure gain, indicating arrival of cold CO2, and remain low for approximately one week before beginning to warm back toward geothermal temperatures. These temperature responses likely indicate that most of the injected CO2 flowed to the Marshell well and either stayed there or continued to transit elsewhere in the formation. There is also an overall trend of warming in the temperature data that is a return to subsurface temperature conditions after water injection. Figure 175. Bottomhole temperature data during CO2 injection and soak periods. Battelle | May 2026 168 Wellbore Gradient Surveys Downhole pressure and temperature gauges were pulled from the Fischhaber 2-35 on April 16, 2025. On the way out of the hole, five-minute-long stops were made every 100 ft to measure pressure and temperature within the tubing, allowing the team to assess CO2 density and phase behavior in the wellbore (Figure 176). The pressure curve appears to be a straight line, which would suggest a constant density fluid column. However, upon further inspection, this was not the case. Figure 177 shows the density of fluid calculated using NIST RefProp with measured pressure and temperature as inputs. Also plotted is the slope of the best fit line to the pressure data. This shows that the density is, in fact, variable throughout the column. The temperature data shows variable temperature gradients from the top to the bottom of the well. These inflection points correlate with changes in lithology (and more importantly, thermal conductivity) within the stratigraphic column near the wellbore. The temperature anomaly from 4,000 to 4,500 ft is due to a measurement error in the gauge. The gauge had been pulled to the surface prior to this gradient run, then dropped back to the bottom of the well. As a result, the thermometer within the gauge had not yet equilibrated to downhole conditions at the start of measurement. Figure 178 shows this pressure and temperature data overlaid on a CO2 PVT diagram. The colormap shows the fluid density at a given pressure and temperature, and the red line and dot show the vapor pressure line and critical point, respectively. The white dots are the measured data from the wellbore gradient run. As shown on the plot, the data is entirely within the vapor phase section of the diagram (below the vapor pressure line), indicating that the entire column of CO2 is in the vapor phase from the top to the bottom of the well. Battelle | May 2026 169 Figure 176. Wellbore pressure and temperature gradients in the Fischhaber 2-35 on April 16th, 2025. Figure 177. Fluid density as a function of depth within the Fischhaber 2-35 on April 16, 2025. Battelle | May 2026 170 Figure 178. Pressure and temperature data from the Fischhaber 2-35 on April 16, 2025, overlaid on a CO2 PVT diagram. On April 18, 2025, after the completion of CO2 Injection Stage 2, the downhole pressure and temperature gauges were run back into the Fischhaber 2-35 well. On the way into the hole, five- minute-long stops were made every 100 ft to measure pressure and temperature within the tubing (opposite direction of the gradient run on April 16, 2025). The pressure and temperature data from this gradient run are shown in Figure 179. Compared to the previous gradient run, this version shows a much higher pressure at the bottom of the well. Temperatures at the wellhead are substantially lower due to recent injection of cold liquid CO2. There is also a temperature anomaly near the top of the well, which arises from slow thermal equilibration of the gauge when moving from ambient temperature to the cold tubing. The inflection point in both the temperature and pressure data around 3,500 ft appears to correspond to a phase change (given the rapid change in fluid column weight). Since the average pressure gradient below 3,500 ft corresponds to a fluid density of around 40 lb/ft3, which is lower than that of fresh water (62.4 lb/ft3), this is most likely a transition from gaseous CO2 to liquid CO2. Figure 180 shows the pressure and temperature data overlaid on the same PVT diagram as Figure 178. All of the gauge data (excluding the outliers from the shallow temperature anomaly) lies almost exactly on the vapor pressure line, pointing to a dynamic system that is actively undergoing phase change. The interpretation of this data is that the wellbore was filled with CO2, with a column of liquid CO2 at the bottom of the well that was actively boiling off into gaseous CO2. In other words, the liquid level encountered was likely dropping with time as more phase transition occurred, and the gauges crossed through this dynamic liquid level on the way to the bottom of the well. Battelle | May 2026 171 Figure 179. Wellbore pressure and temperature gradients in the Fischhaber 2-35 on April 18, 2025. Figure 180. Pressure and temperature data from the Fischhaber 2-35 on April 18, 2025, overlaid on a CO2 PVT diagram. Battelle | May 2026 172 7.0 Task 6: Production Monitoring At project inception, Task 6 focused on improving technologies for monitoring flood fronts and evaluating parameters affecting breakthrough times and CO2 sweep efficiency during and after CO2 injection. In the final huff and puff iteration of the project, monitoring was intended to capture placement of the injected CO2 plume within the reservoir, and efficiency of residual oil mobilization by CO2 and surfactant. Density, temperature, pressure, and gas composition measurements were used to inform models, understand fluid behavior within the reservoir, and revise reservoir architecture concepts. 7.1 Subtask 6.1: Fluid Sampling Fluid sampling during the production test was accomplished via multiple techniques including: 1. In-line gas chromatography at the Marshell 1-35 facility. 2. Water sampling at the separator to analyze density and salinity. 3. Planned wellhead oil sampling (no oil was produced, so no samples were gathered) The in-line gas chromatograph enabled the team to accurately monitor changes in CO2 and natural gas production in real time. An example of the chromatograph output taken on September 16, 2025 is shown in Figure 181. The CO2 fraction in the produced gas stayed around 97-98% for the entirety of the production period. Figure 181. Example gas chromatograph output during production (9/16/25). Water at the separator was consistently weighed to track density and salinity of the produced fluid. Water weight fluctuated between 8.33 (fresh) and 8.8 ppg (mix of injected fresh water and production water) for the entirety of the test, averaging around 8.7 ppg (roughly 65,000 ppm TDS). No clear correlation between water weight and production rate was observed. The fact that water weight was changing indicates non-uniform mixing of the injected fluid within the formation. Additional sampling was conducted during swabbing efforts at the Beattie well after the conclusion of the production test. Gas was sampled at the wellhead on November 22, 2025, prior to any blowdown of the well. This gas underwent compositional analysis, identifying it as an extremely wet gas (~25% methane, 70% C2-C5) with a specific gravity of 1.28. This is substantially heavier than gas sampled during primary production in the field and could correlate with a light condensate observed during wellbore gradient surveys conducted as part of this project (see Section 6.3). Battelle | May 2026 173 Swabbing operations also returned a small volume of oil (refer to Section 6.2 for further details). This oil was sampled and lab analysis consisting of compositional analysis, density measurement and API gravity. The composition of the sample (by mole percent) is shown in Figure 182. The API gravity of the sample was 36.9, making it the heaviest sample from the Southern Michigan Trenton-Black River trend analyzed as part of this project. Figure 182. Composition of Beattie oil sample. By comparing the composition of this sample with the Hoffman oil that is analogous to in situ fluid at the Lee 26 field at discovery, changes between live reservoir oil and the residual oil sampled post-CO2 treatment can be ascertained. One significant difference is in the makeup of the heavy hydrocarbons (C30+) between the samples. The average molecular weight of the C30+ components in the Hoffman oil is 570 g/mol, while in the Beattie oil is heavier at 600 g/mol. Figure 183 shows the relative difference in mole percent for each carbon number between the Hoffman and Beattie oils. A positive ratio indicates that a given component is more present in the Beattie oil than the Hoffman oil, while a negative ratio indicates the opposite. The light hydrocarbons are substantially depleted in the Beattie oil, which is not surprising – being a residual oil from an ultra-depleted reservoir, almost all of the light hydrocarbons would have evaporated and been produced as gas. The C4-C10 components are moderately depleted, with the exception of C5. This could be indicative of further mobilization of light-medium hydrocarbons during primary production, differences in the analog oil from Lee reservoir fluids, or some combination of both. Figure 183. Relative difference in mole percent by component between Beattie oil and Hoffman oil. Battelle | May 2026 174 From C12-C29, components are generally far more prevalent in the residual oil, with C24-C26 showing similar to slightly lower levels. Some level of increase is expected here given the fact that the Beattie oil is depleted of lighter hydrocarbons. However, the consistent level of increase observed across a wide range of molecular weight hydrocarbons appears to be indicative of preferential mobilization. The fact that C30+ components are much less prevalent in the Beattie oil further supports this theory (since CO2 floods tend to cause heavy ends to precipitate; Hartono et al, 2024), as does the fact that CO2 tends to mobilize these medium-heavy hydrocarbons left behind during primary production. This analysis supports the fact that CO2 operations in the field did, in fact, mobilize residual oil, though not within the anticipated regions of the reservoir. 7.2 Subtask 6.2: Fluid Flow Initial Production Tests On May 28, 2025, the Fischhaber well was opened to flow with no artificial lift in the well. The goal of this operation was to examine how long tubing pressures would stay elevated while flowing and to ensure surface gas recycling infrastructure was operating properly. The well free- flowed for one week, during which time the well produced CO2 that was reinjected in the Marshell. The well was then shut-in again in preparation for swab testing. No liquid reached the surface during these operations. On June 10, 2025, swab testing commenced at the Fischhaber well. These operations ran from June 10 – June 14, during which time a small volume of liquid was produced – mostly 8.4-8.5 ppg water with trace amounts of oil. The swab was run to 4,500 ft and was consistently bringing back around 4 bbls of liquid per run. On the first run-in, operators reported a liquid level around 2,700 ft but returning no liquid to surface. After this, liquid levels were consistently found around 3,500 ft to 3,700 ft from surface. At the conclusion of swab testing, the bottomhole gauge was retrieved from the Fischhaber well in preparation for pump installation. Electric Submersible Pump Installation, Troubleshooting, and Replacement At the start of the intended production test period, an ESP was installed in the Fischhaber 2-35 to lift reservoir fluids to surface. West Bay Exploration worked to design this pump with Summit ESPs (a Halliburton company). The pump was designed for flowrates up to 600 bbl/day of combined liquids, with a downhole gas separator to promote efficient pump operations (Figure 184). The pump was expected to consume 20 to 50 horsepower and powered by a diesel generating set on the Fischhaber pad and was controlled by a variable speed drive (VSD) that adjusted the pump operating speed to provide consistent flow and pressure to the Marshell production facility. West Bay installed the summit pump and drive equipment and found that the pump needed to be operated at maximum frequency (pump speed) in order to bring liquid to surface, well outside the intended operating window (Figure 185). This discrepancy may in part have been a result of the reservoir pressure being approximately 400 psi lower than expected due to the extended shut-in period. Summit attempted to troubleshoot the pump in position but were unable to identify a solution. It was hypothesized that insufficient liquid was entering the pump due to its location above the lowest perforations in the well. To improve liquid intake, the pump was lowered below the perforations and Summit attempted to operate the pump again. The pump ran successfully for about three days prior to repeatedly faulting out on either low or high frequency. Summit again attempted to troubleshoot the issue, but was unable to re-start the Battelle | May 2026 175 ESP. The Fischhaber was shut in on July 10, and the pump was removed from the well on July 17. On removal, the ESP was inspected and found to have failed radial bearings, which had caused the pump to mechanically lock and cease functioning. Summit initially concluded that sediment ingestion triggered the failure, but full disassembly of the pump identified several assembly defects that left the pump out-of-tolerance, destroying the bearings and subsequently the impellers (Figure 186). Figure 184. Design schematic for original Electric Submersible Pump. Battelle | May 2026 176 Figure 185. ESP production data indicating inconsistent performance of the first ESP. Gas lock may have been an issue, however, assembly error was identified as the primary failure mode. Battelle | May 2026 177 Figure 186. (A) Fractured radial bushing in one ESP stage and (B) impeller hub with excessive wear and tip damage (lower). The team leveraged the downtime caused by the pump failure to work with Summit to find a pump design that was optimized for the real-time reservoir conditions. The pump was re- designed to handle a wider range of fluid rates, as well as changes in fluid column density and percentage of gas content. Impeller shape and stage count were selected to control fluid density changes within the ESP due to condensation of CO2 as it moved upward through the pump. Battelle | May 2026 178 This replacement pump was intended to be placed below the perforations, so a shroud was added to shield the intake ports from sediment ingestion. A detailed schematic is provided in Figure 187. Figure 187. Detailed schematic of optimized replacement ESP. Note shroud (gray cone) over intake ports. The replacement ESP was delivered on August 26, 2025, installed on August 27, and started up on August 28, 2025. The pump ran within design tolerances and produced fluids consistently within the targeted production ranges for the entirety of the production test. Production Test Results Fluid flow rates were monitored by the field operators, project partner West Bay Exploration, throughout the production testing cycle. Swabbing tests to assess flowback operations began at the Fischhaber well on August 28, 2025 after installation of the second ESP and ceased on Battelle | May 2026 179 October 30, 2025. All produced gas was reinjected at the Marshell well, and produced water was diverted to an offsite disposal well. A total of ~42,000 bbls of water and ~3,300 mcf of gas (98% CO2) were produced during the 62-day production test period. This amounted to ~170 tonnes of CO2, or 9% of the total injected mass. No oil was produced from the Fischhaber during this period. The water and gas rates, along with cumulative rates, during the production test period are shown in Figure 188. Figure 188. Fluid rates (top) and cumulative volumes (bottom) observed during the production test period. After the Fischhaber well was shut in and the ESP removed, the Jacoby and Beattie wells were swabbed to investigate potential fluid production. The Jacoby did not produce any fluid after several days of swabbing due to formation infiltration downhole. Swabbing at the Beattie well produced 31 bbls of oil and five bbls of 8.8 ppg water (mix of production water and injected fresh water) over three days of swabbing. Another two days of swabbing was conducted after the release of packer elements in the well, during which time production fluid was all water and skim oil. The oil produced during initial swabbing runs was sampled and sent for laboratory analysis. 7.3 Subtask 6.3: Pressure Analysis: Identifying the Placement of CO2 in the Reservoir Pressure data from late May 2025 to the end of production testing in November 2025 shows the impact of production operations on field pressures. Datum-corrected pressure data from the production period is shown in Figure 189. The period from May 28 to June 15 shows decreased Battelle | May 2026 180 pressures at the Fischhaber well during swabbing and flowing wellbore conditions. Pressures at Marshell are elevated in response to the reinjection of gas produced from the Fischhaber well. During initial flowback, bottomhole pressure dropped initially around 30 psi, continuing a much slower drop for the remainder of that production period. Pressures fluctuated across a four to five psi range, likely indicating a dynamic system at the bottom of the well. The team’s interpretation of this data is that percolation of CO2 at the bottom of the well through a fluid column within the wellbore is being observed. During the shut-in period, these fluctuations gradually decrease in magnitude. The difference in producing BHP and shut-in BHP is around 10 psi. During swabbing operations much larger fluctuations are observed due to the production and subsequent backfilling of liquid volumes in the wellbore. The overall background trend of pressure drop during swabbing and production phases is roughly consistent with one another. During shut-in periods, the BHP drops much slower than during production phases, illustrating the impact of fluid production on reservoir pressure. Over the course of the approximately two weeks of production and swabbing, BHP at the Fischhaber well dropped by approximately 50 psi. Pressures from June 26-July 9, corresponding to the operation of the first ESP, are skewed due to the failed pump, but do show a pressure decrease from pump operations. This first production period ended with the failure of the installed ESP, at which point the field was shut-in until a new ESP could be installed. Bottomhole pressures at the Marshell dropped at this point, as injection ceased. Figure 189. Datum-corrected bottom-hole pressure data from the Fischhaber (orange), Marshell (blue), Jacoby (black), and Beattie (red) wells from May 28, 2025 through the end of available data. Production was restarted with the new pump on August 28, 2025. At this point, pressure at the Marshell increased (again due to gas reinjection), and pressures elsewhere in the field decreased. The spike in pressures across the field on September 20-21, 2025 corresponds to a brief downtime in the ESP. Pressures appeared to decrease much slower at the Jacoby well than at the other three wells, indicating that the fracture that opened in early January 2025 was likely closed, and the lower conductivity flow pathways between the Jacoby portion of the reservoir and the rest of the reservoir were again dominant. In early October 2025, the Jacoby pressures stabilized and remained relatively constant until the end of data collection. This is likely when the gauges were no longer submerged due to fluid levels dropping in the well as reservoir pressures decreased. Pressures at the other three wells continued to decrease until Battelle | May 2026 181 the reservoir was shut-in on October 30, at which point buildup behavior in the Fischhaber (consistent with end of production), falloff behavior in the Marshell (consistent with ceasing gas re-injection), and stable, constant pressures at the Beattie well were observed. WHP data during production testing is shown in Figure 190. Marshell wellhead pressures were elevated during periods of injection but fell off as pressures declined in the field. During shut in periods, WHP at the Marshell was constant, corresponding to the backpressure control at the injection facility. Fischhaber WHP was extremely variable over the course of production. Jacoby WHP gradually increased for the entire production test, increasing faster in early October 2025 before leveling off and staying constant through the remainder of available data. This increase likely corresponds to continuous CO2 influx throughout the production period. The average column density in the Jacoby well at the end of production correlates to vapor-phase CO2, which further supports this theory. Finally, pressures at the Beattie well remained stable at zero psi until the end of September 2025, at which point it increased by about five psi over the next month and a half. These low pressures likely indicate a lack of gas in the Beattie well. Figure 190. Wellhead pressure data at all four wells during production testing. Temperature Monitoring Bottom-hole fluid temperature data was also collected during this period from the bottomhole gauges and ESP. These data are shown in Figure 191. Temperatures stayed mostly constant throughout this period at the Beattie and Jacoby wells. Temperatures at Fischhaber were slightly elevated as the water flowing past the motor of the ESP and was absorbing heat from the pump. Temperatures at Marshell also increased slightly during periods of injection, likely due to gas heating during compression at the surface. Battelle | May 2026 182 Figure 191. Downhole temperature data from the production testing period. Wellbore Gradient Survey On June 9, 2025 – prior to commencement of swabbing activities – a third wellbore gradient run was conducted with downhole pressure/temperature gauges at the Fischhaber well. The resulting pressure and temperature data are shown in Figure 192. As with prior gradient runs, there appears to be a temperature anomaly in the shallow portion of the well. However, given the high ambient temperatures at the time of measurement and the observed increase in pressure over the same interval, these could be real data. The most striking feature of the data is the inflection point in pressure around 2,700 ft, seemingly indicating a liquid level at that point. Shallower than 2,700 ft, pressure is roughly constant around 20-30 psi. Deeper than this point, there is a relatively linear pressure gradient to the bottom of the well. The slope of this line (0.226 psi/ft) corresponds to a 4.3 ppg fluid. This is substantially lighter than any water that would be seen in the reservoir and substantially heavier than a column of CO2 at these pressure/temperature conditions. Hydrocarbon fluid at this density would correspond to a hot condensate (liquid at depth, vapor at surface). These data are consistent with what was observed in the first swabbing run. The fluid level detected by the operators corresponds to the same depth seen in this data, and the resulting production of zero fluid could correspond to a fluid that evaporated on the way up the wellbore. Future swabbing runs that consistently found liquid around 3,500-3,700 ft indicate a column density consistent with that of fresh to lightly saline water, which is what was produced during those runs. Battelle | May 2026 183 Figure 192. Wellbore pressure and temperature gradients in the Fischhaber 2-35 on June 9, 2025. 7.4 Subtask 6.4: Additional Monitoring Analyses In addition to the monitoring described above, the project team evaluated a number of additional technologies to monitor and evaluate the surfactant-enhanced CO2 stimulation. Technologies evaluated include: • CO2 and oil tracers: Intended to be injected along with CO2, these tracers have the potential to shed light on oil and CO2 mobility and placement within the reservoir. The project team worked with Baker-Hughes ProTechnics to understand applications within the scope of the study, but tracers were found to be too costly for this application. • One-spot tracer testing: Single-well chemical tracer testing (also known as one- spot tracer testing) is a method for in-situ evaluation of reservoir fluid saturation. During the test, a small amount of ester is injected into the reservoir, where it reacts with oil to generate alcohol. Because the ester reacts with the oil and is partitioned when the well is later produced back, the ester lags behind the alcohol (Figure 193), generating two distinct production profiles (Figure 194). The difference in time between the peaks of the two profiles is indicative of residual oil saturation. This test would have been conducted twice: once to determine native residual oil saturation, and a second time after a small injection of surfactant-enhanced CO2. The second test would reveal how much oil had been mobilized by the CO2, allowing the project team to very accurately determine the oil production uplift generated by this stimulation strategy. However, due to high cost, operational complexity related to reservoir pressure, and scheduling conflicts with test providers, the project team chose not to proceed with this method of surveillance. • Time-lapse seismic monitoring: The project team worked with Dr. Shuvajit Bhattacharya at UT BEG to scope a time-lapse seismic monitoring program for the Lee field. The team evaluated an integrated workflow developed by Bhattacharya et al. (2024) to conduct a pre-feasibility assessment of the potential for 4D seismic Battelle | May 2026 184 detection of CO2 injected into the Lee 26 field in Calhoun County, Michigan, using existing 3D seismic, well logs, and a CMG dynamic flow simulation model of the field to generate a predicted seismic amplitude anomaly generated by the injected CO2 plume. After injection, the team would shoot serial 2D seismic lines across the field, process them, and complete attribute analysis to identify the plume in the new seismic lines. However, based on dynamic modeling results, the project team determined that the induced anomaly generated by the relatively small 2,000 ton injection would likely be undetectable given the reservoir depth and thickness. Figure 193. Conceptual diagram showing the reactions occurring during a single-well partitioning tracer test; diagram is from tracer test provider Tracerco. Battelle | May 2026 185 Figure 194. Example of production curves used to determine residual oil saturation by single well tracer test, diagram is from tracer test provider Tracerco. 7.5 Subtask 6.5: Integration of Monitoring Technologies The integration of selected monitoring technologies (surface and bottom-hole pressure and temperature gauges; water tracers; SPN logging; fluid sampling & analysis) is demonstrated in Subtask 6.3 and in Task 7. This section of the report deals with the effectiveness of the selected monitoring technologies and recommendations for monitoring of future CO2-EOR efforts in the TBR trend. • Water Tracers (and water salinity): During the repressurization phase of the project, water tracers were found to be relatively effective at characterizing movement across the reservoir. By pairing tracer response and water salinity measurement, the team was able to identify slugs of relatively fresh, tracer-laden water arriving at wells in the field. Adding this context to bottom-hole pressure measurements allowed for the identification of distinct events in the reservoir which is believed to be related to fracture opening/closing or the initiation of a spill point within the vug systems in the reservoir. This gave the team a view into the extreme reservoir heterogeneity in the Lee 26 field and provides context for future TBR CO2-EOR. One area of improvement is that the tracers were injected into the field before water samples from the wells were accessible. This caused the team to fail at identifying the first arrival of tracers at each well, which would have been very useful data. In future projects, operators should wait until they are certain they will be able to acquire water samples before implementing their tracer study and should plan to sample for tracers at a high frequency (every day if possible) during the first few weeks after tracer injection. • Spectral Pulsed Neutron Logging (SPN): SPN was not effective in this application. A combination of poor cement condition (in the Marshell 1-35), lack of sufficient porosity and pay zone thickness (at the Fischhaber 2-35), and changes in wellbore configuration between the baseline and follow-up logging runs at the Fischhaber 2-35 resulted in the project acquiring no useful data from these logging operations. The project does not recommend Battelle | May 2026 186 SPN for reservoir saturation evaluation in depleted TBR fields. Instead, developers should utilize reservoir engineering techniques such as history matching and material balance analysis to determine which areas of their fields have sufficient residual oil saturations to be appropriate for CO2-EOR. Single-well partitioning tracer studies may also help answer this question, but it is possible that the high reservoir heterogeneity in the TBR could cause this technique to fail. • Downhole Pressure Gauges: This is the key technology for tracking reservoir behavior. Downhole pressure gauges are by far the most cost-effective method for directly monitoring reservoir fluid movement and tracking inter-well connectivity. The downhole pressure data gathered during the project provided the basis for the field history match during repressurization and production, meaningfully influencing the assessment of porosity distribution, fracture and vug connectivity, and heterogeneity. The downhole gauges also provided temperature sensing, which proved useful during CO2 injection as first arrival of the cold CO2 was easily observable. • Because the gauges were on hand (reducing cost) and were able to be rapidly deployed, the decision was made to use gauges with built-in memory rather than gauges that were connected to the surface and could be monitored in real time. In retrospect, this resulted in higher operating costs than expected since payment was needed for slickline removal of the gauges on a regular basis to collect stored data. Wired gauges will likely be selected in the future. • Surface Pressure Gauges: Surface tubing and annular pressure gauges are a requirement for every operating wellbore, therefore are strongly recommend for future projects. These gauges also proved useful for approximate fluid column density tracking, giving the project insight into wellbore contents. They also provided a real-time, continuous proxy for downhole pressure data, allowing for a method to observe some pressure differences during repressurization, CO2 injection, and production. • Gas Chromatography: In-line gas chromatography was built into the production facility to track changes in gas composition during production. This data would have been crucial for identifying changes in mobilized constituents between initial flowback and later sustained production, providing insight into the utility of both CO2 and CETAC-30 surfactant, as well as providing valuable CO2 concentration data for facility operation and optimization. Because no hydrocarbons were produced (outside of trace methane), the gas chromatograph was not useful. However, it is still recommended to include gas chromatography in facilities that CO2-EOR operations flow through. Battelle | May 2026 187 8.0 Task 7: Performance Assessment At project inception, Task 7 was designed with a focus on ongoing performance and assessment of the efficiency of the CO2-EOR injection. As initially planned, the period from the end of injection to project closeout was on the order of one year. Given the shortened timeline as the project played out, as well as the lack of production and subsequent field shutin after CO2 treatment, assessment of the sustained performance of the reservoir is no longer relevant. As a result, discussion surrounding the performance assessment (Subtask 7.1) focuses on the team’s overall evaluation of the efficacy of CO2 treatment in the Lee reservoir and explanations for production trends (or lack thereof) observed. Subtask 7.2 focuses on integration of field data to the existing reservoir model in the form of a sensitivity assessment, as well as simplified modeling to assess scale-up across the Southern Michigan trend and potential recovery in the TBR. 8.1 Subtask 7.1: Performance Assessment Despite challenges related to timeline, winter weather, wireline logging, and artificial lift malfunctions, field testing at the Lee 26 field was an operational success. Approximately 2,000 tons of CO2 were injected into the reservoir, and then the reservoir was production tested for two months with no injuries, no unintended release of hydrocarbons of CO2, and no harm to the environment. During these operations, the team was able to gather valuable data about reservoir performance that inform the economic potential of future chemically enabled CO2-EOR tests within the TBR trend. That being said, oil production (and therefore economic productivity) resulting from chemically enhanced CO2 treatment did not reach the intended target. Over the 63 days of production testing, several basic trends were revealed: 1. Bottom hole pressure declined predictably. Some portion of the pressure decline is likely related to continued bleed-off of pressure into the matrix, but the majority of pressure decline appears to be caused by production. 2. Water rate varied predictably based on pump frequency. This indicates that the replacement ESP functioned as intended, and that permeability is sufficiently high that pump rate is the limiting factor on deliverability, not reservoir conditions. 3. Produced CO2 to natural gas ratio was stable, at about 98% CO2 and 2% natural gas. 4. Water to total gas ratio remained stable. Minor increases in pump rate and water production are accompanied by similar increases in gas production. Since the perforations in the Fischhaber 2-35 are in the water leg of the reservoir, this provides at least some evidence that there is not a purely stratified fluid column in the reservoir-there is at least some gas that exists within the water column. 5. Water weight varied between 8.3 and 8.8 pounds per gallon, with no obvious trend. Water used for repressurization was approximately 8.4 pounds per gallon, while in- situ reservoir water is approximately 9.7 pounds per gallon, so the produced water indicates heterogeneous mixing between injected and in-situ waters. Forecasting simulations predicted a much larger return of residual fluid (>1,000 bbls of oil over the production period). There are a number of possible causes for the lack of oil production observed at the Fischhaber 2-35. These include: Battelle | May 2026 188 1. Carbon Dioxide and surfactant did not effectively penetrate into the matrix, where modeling indicates most residual oil should be. a. It is possible that the capillary entry pressure needed to move useful volumes of gas-phase CO2 into the matrix was not achieved. The apparent movement of injected water into the matrix does not conflict with this notion: if water exists as a continuous phase in the matrix, it does not need to overcome capillary entry pressure, whereas CO2 would have to disrupt the water phase in order to gain entry to the matrix. b. If this is the case, it is possible that gas-phase CO2 could have been effectively injected into the matrix if higher pressure had been achieved. However, it is likely that liquid or supercritical CO2 pressure would be necessary to generate sufficient entry force. 2. Carbon Dioxide and surfactant did penetrate into the matrix during high-pressure injection but because the vugs and fractures returned to a water-saturated state during the soak period, a relative permeability block exists, keeping CO2 and oil from flowing out of the matrix. a. It is possible that if the reservoir were to continue to be drawn down, matrix production might turn on, increasing CO2 production and initiating oil production. 3. Oil was successfully mobilized and the protracted shut-in period between injection and production has caused oil to buoyantly move upward. With perforations currently positioned in the water column, access to oil is precluded. Continued water withdrawal will cause water level to drop, ultimately providing oil production. a. A related scenario is that oil has moved buoyantly up-dip, away from the Fischhaber and towards the Jacoby and Beattie. Continued water withdrawal at the Fischhaber may draw oil back toward the well, providing production. 4. Carbon Dioxide injection swept CO2 and oil away from the Fischhaber and towards other parts of the reservoir, reducing the effectiveness of the ‘puff’ production phase. a. Related to this, the injection of water may have swept oil away from the Fischhaber well. Pressure gradients from bottom-hole gauge data suggest that fluids were flowing away from the Fischhaber well towards the Beattie and Jacoby wells for the majority of water injection. 5. Local pore volume around the Fischhaber well is smaller than initially predicted. Revisiting of log-interpreted dolomite presence, seismic interpretations, and recently acquired pressure/temperature data suggest that the majority of reservoir pore volume, at least in the eastern portion of the field, exists around the Marshell well. A smaller local pore volume would necessitate a smaller volume of residual oil around the Fischhaber well. a. If this hypothesis is true, then the large volume of injected CO2 would have pushed any residual oil away from the Fischhaber well (see scenario 4). 6. Primary production was so efficient that there is minimal residual oil. a. The technical team does not believe that this scenario is likely; it implies that all primary production was from fractures and vuggy porosity, which is inconsistent with the team’s understanding of production in the remainder of the trend. Based on interpretations of field data from the entire lifetime of the field, the most likely Battelle | May 2026 189 explanation for the disparity between model predictions and operational data is a combination of hypotheses four and five above. The intent of the surfactant in huff and puff operations, in addition to providing wettability changes within the reservoir to enhance oil mobilization, was to provide mobility control of the injected CO2 to allow it to act on residual fluids surrounding the Fischhaber well. If local pore volume was limited in this area of the reservoir (as the team believes it was), the large injected volume of CO2 would have flown away from the well, regardless of any mobility control, simply due to a lack of available void space. The temperature signals at the Marshell well, coincident with CO2 injection, support the idea that CO2, and therefore any mobilized oil, was swept away from the Fischhaber well. Fluid column stratification that would have happened in the high conductivity vug/fracture network during the soak phase of the huff and puff may then have caused fluid reorganization in such a way to prevent oil migration back toward the Fischhaber well during the “puff” phase. Based on laboratory analysis of the oil recovered from the Lee field during swabbing operations (see Section 6.1), it does appear that oil was a residual oil within the reservoir. Since sampling was conducted at the Beattie well throughout the entirety of water injection and no oil was detected in any of the samples, it appears that this residual oil was, in fact, mobilized by the chemical CO2 treatment. In essence – the chemical CO2 mixture succeeded in mobilizing residual oil, but the injection and production plan proved unsuccessful in predictably controlling the migration and recovery of this mobilized fluid. The solution to this problem is not clear – huff and puff operations appear to be ineffective, but the high conductivity pathways in TBR systems make effective flooding a challenge. It is possible that, if a concentrated zone of high pore volume could be identified, a huff and puff recovery strategy could be more effective. Even then, injected fluid will be prone to migrate away from the near-wellbore area through high conductivity paths. Substantial mobility control will be required to successfully implement any huff and puff or flooding tertiary EOR scheme. Another possibility involves a field implementation design that is agnostic to reservoir geometry. This may include something like a “full field huff and puff” where the entire field is repressurized with CO2, allowed to soak, and then turned to production. The practicality of such a scenario is unclear, especially given the delay in realized revenues from oil production. However, given that operations and data collected in this project support that mobilization of the residual fluid is possible, and that a substantial amount of residual resource exists in these fields, additional exploration and optimization of recovery strategies is worthwhile. 8.2 Subtask 7.2: Modeling Update and Scale-Up As part of this task, additional dynamic simulations were conducted that integrated pressure and fluid flow data obtained during field testing operations. Ideally, this exercise would have involved a full integration of the gathered data, including detailed pressure transient and inter-well flow analysis, updates to the SEM, and a new history match in the dynamic model. However, a major takeaway of the CO2 huff and puff operations was the large impact that small scale heterogeneities present in the reservoir have on fluid flow patterns. The level of complexity present in the reservoir severely inhibits effective correlations between wells, even with the relatively small well spacing present in the Lee field. As a result, the geostatistical property modeling techniques used in SEM construction and property manipulations used in history matching efforts struggle to effectively converge on a representative model. The disagreement between forecasting simulations and field observations – particularly in oil production rates – motivated the team to focus model update activities on gaining a better understanding of what parameters most contribute to the migration of fluids within these types of Battelle | May 2026 190 reservoirs. Rather than seeking a single history match to this extremely complex field, which may have little application to other fields in the trend, the goal was to delineate key subsurface factors which could lend some level of insight to operations in other HTD fields. A sensitivity assessment was conducted that examined large variability in key subsurface parameters and the resulting impact on performance metrics in the field – specifically pressure response and production volumes. With this approach, the utility of the existing reservoir model could be maximized despite a poor match to field test data. To facilitate rapid modeling of a large set of scenarios, the existing model was simplified into various sectors corresponding to sections of the reservoir associated with a specific well penetration. The divisions in the reservoir are depicted in Figure 195, corresponding to the oil/water legs near each well (Beattie, Jacoby, Fischhaber, Marshell) and the gas caps located above the Jacoby and Fischhaber wells. Within each block, homogeneous porosity and permeability distributions were assigned based on average values from the history-matched reservoir model. Fluid properties and contacts, fluid models, and relative permeability models were left unchanged. Matrix permeability was set to a single value (0.6 mD) across the entire model due to the low variability in this parameter in the history matched model. Figure 195. Reservoir zones of the simplified reservoir model used in sensitivity analysis. A list of key subsurface variables were identified for analysis based on their impacts in initial history matching, interpretations from field pressure and production data, and the level of uncertainty in existing data. The list of variables and their corresponding values are shown in Table 27. These variables primarily cover the distribution of pore volumes across the field and conductivity of the connected fractures and vugs across the field. The base variable values correspond to the average values for each zone in Figure 195, while the low and high values were chosen to assess the full range of variability expected within the parameters. Relative pore volumes were modified by changing pore volume modifiers across specific sections, rather than by changing porosity. The porosity of matrix dolomite is fairly well constrained from the geologic assessments, but the amount of dolomitization in each part of the reservoir (particularly between wells) is an unknown of interest. As such, matrix porosity was set to a constant value of 7.5% across the entire model. Similarly, fracture porosity was set to a constant 1.6%, corresponding to the 21% storativity ratio used in the history matched model. Total pore volume was held constant in the relative pore volume scenarios. In addition, the Beattie and Jacoby reservoir zones were combined for certain sensitivities due to the level of uncertainty in this section of the Battelle | May 2026 191 reservoir (specifically around the Jacoby completions). Table 27. Variables considered in model sensitivity analysis Variable Unit Base (-) Low (L) High (H) Notes Relative PV - Matrix Jacoby PV Beattie + Jacoby % 33.9 16 50 multiplier = (VMm B+J) 4.4*Beattie Relative PV - Matrix % 33.7 16 50 Fischhaber (VMm F) Relative PV - Matrix % 32.5 16 50 Marshell (VMm M) Permeability - Fracture Beattie + Jacoby (kf mD 385 128 1,155 B+J) Permeability - Fracture mD 285 95 855 Fischhaber (kf F) Permeability - Fracture mD 1,237 412 3,711 Marshell (kf M) Relative PV - Fracture Jacoby PV Beattie + Jacoby % 28.3 16 50 multiplier = (VMf B+J) 3.0*Beattie Relative PV - Fracture % 34.8 16 50 Fischhaber (VMf F) Relative PV - Fracture % 36.9 16 50 Marshell (VMf M) Throughgoing Vug Vug Train Vug Train No Train with - with k by Throughgoing Presence/Perm (k ft M- k=3,711 zone Vug Train mD B) Jacoby matrix PV multiplier = Free Gas Fraction 4.0*Fischhaber % 21.6 10.6 32.6 (VM gc J) Fracture PV multiplier = 0.8*matrix A total of 23 scenarios were run to assess the sensitivity of these variables – a base case run with the average property values, and 22 scenarios where a single variable was changed to either its high or low value. Contrary to the history-matched model, produced CO2 was reinjected into the Marshell to better reflect field setup. The scenario matrix for the sensitivity analysis is shown in Table 28. In addition, a single scenario (Scenario 0.1 in Table 28) was run with average property values where the fracture presumed to exist between the Jacoby and Fischhaber wells was dynamically opened and closed at a corresponding fracture closure pressure (FCP). This was done by applying transmissivity multipliers of zero across the gas and oil legs between these sections, while maintaining connectivity in the water leg. Battelle | May 2026 192 Table 28. Scenario Matrix for Sensitivity Analysis VM k k VM k VM m VM VM f f k f VM VM ft M- gc Scenario B+J mF m M B+J F fM B+J fF fM B J FCP 0 (Base) - - - - - - - - - - - - 0.1 - - - - - - - - - - - Y 1 L - - - - - - - - - - - 2 H - - - - - - - - - - - 3 - L - - - - - - - - - - 4 - H - - - - - - - - - - 5 - - L - - - - - - - - - 6 - - H - - - - - - - - - 7 - - - L - - - - - - - - 8 - - - H - - - - - - - - 9 - - - - L - - - - - - - 10 - - - - H - - - - - - - 11 - - - - - L - - - - - - 12 - - - - - H - - - - - - 13 - - - - - - L - - - - - 14 - - - - - - H - - - - - 15 - - - - - - - L - - - - 16 - - - - - - - H - - - - 17 - - - - - - - - L - - - 18 - - - - - - - - H - - - 19 - - - - - - - - - L - - 20 - - - - - - - - - H - - 21 - - - - - - - - - - L - 22 - - - - - - - - - - H - The results of each scenario were then compared to field datasets for 22 different metrics of field history and performance. The metrics were as follows: • Field cumulative production of oil and gas (2) • Field average pressure during primary production (1) • Oil and gas cumulative production at each well – including initial Jacoby completions – during primary production (12) • Bottomhole pressure response at each well for duration of injection operations (4) • Oil, water and gas production at Fischhaber post-treatment (3) As a proxy for goodness of fit, the root mean square error (RMSE) was calculated for each scenario across each of these metrics using the equation ∑𝑁𝑁 (𝑦𝑦𝑖𝑖 − 𝑦𝑦𝑦𝑖𝑖)2 √ 𝑖𝑖=1 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 = 𝑁𝑁 Where 𝑦𝑦𝑦𝑖𝑖 are the measured data points, 𝑦𝑦𝑖𝑖 are the synchronized model outputs corresponding to Battelle | May 2026 193 measured data, and N is the number of data points for a given metric. In the case of bottomhole pressure data, where minute resolution data was available, data were upscaled to daily resolution. RMSE values for each metric and each scenario are given in Table 29. The RMSE values for the history matched model (HM) are also included for comparison. Note that the “Fischhaber Gas Post-Inj” metric is significantly smaller in the HM case than in most of the other cases because of the lack of gas reinjection. In addition, RMSE values for the sensitivity scenarios were compared to the base case RMSE for each metric to assess whether a given variable changed improved or worsened the fit to field data and by what magnitude. The values in the table are color coded according to the following: • White – RMSE changed by <1% • Light Red/Green – RMSE increased/decreased by 1-5% • Medium Red/Green – RMSE increased/decreased by 5-20% • Dark Red/Green – RMSE increased/decreased by >20% Several of the metrics show little to no variation across scenarios (e.g., gas at vertical Jacoby completion during primary production). These metrics typically represent primary well constraints for a given well within the model, and as such the production values are identical across scenarios as long as the deliverability of the reservoir near that well is sufficiently high. While the raw RMSE values allow for effective comparison across scenarios within a given metric, they do not allow for comparison across metrics because of the large difference in raw values (e.g. comparing gas production, which is in billions of scf, to pressures in tens to hundreds of psi). To facilitate such a comparison, the relative change in RMSE for each metric (relative to base case RMSE, as a decimal) is presented in Table 30. The degree of change observed in each metric for a given variable can help inform what aggregate changes would drive better model performance. Visual inspection of the modeling results, relative to field data, can provide additional context around the RMSE values and improvements (or lack thereof) in goodness of fit across metrics. Figures 196 through 199 show visual comparison of model results co-plotted with field data for all 22 metrics for the base scenario (top panels) and the aggregated results of all sensitivity scenarios (bottom panel). The dashed lines in the bottom panels show the best fit scenario to a given metric (not necessarily the same scenario across metrics), while the shaded region shows the range of all sensitivity results. For all metrics (except those used as primary constraints), at least one case exists where RMSE has decreased. However, there is not a single scenario that improves fits across the board. This is illustrative of the complexity of the system, indicating that a combination of these factors is responsible for the pressure and fluid migration observed in the Lee field. Battelle | May 2026 194 Table 29. RMSE values for sensitivity scenarios. Red/Green colors show RMSE increase/decrease. Light/medium/dark color show change by 1-5%, 5-20%, and >20%, respectively. Field Beattie Jacoby Fischhaber Marshell Fischhaber Sensitivity Pressure Field Oil Field Gas Beattie Oil Beattie Gas Jacoby vert Jacoby vert Jacoby HD1 Jacoby HD1 Jacoby HD2 Jacoby HD2 Fischhaber Fischhaber Marshell Oil Marshell Gas pressure post- pressure post- pressure pressure post- Fischhaber Fischhaber Water post- Scenario # Parameter Value primary primary primary primary primary Oil primary Gas primary Oil primary Gas primary Oil primary Gas primary Oil primary Gas primary primary primary inj inj post-inj inj Gas post-inj Oil post-inj inj HM 66.5 22357.4 85378.4 15400.9 48116.1 76.4 160.1 253.7 89.8 22090.9 228.2 21053.8 3380.4 1215.2 59921.5 108.4 111.1 119.9 265.0 110.3 1961.4 2372.2 0 - - 74.7 20382.3 74476.5 14669.0 29547.7 153.0 2641.0 176.3 1273.8 22575.1 2509.6 24814.6 25090.5 2222.0 93359.6 179.2 211.3 175.2 202.8 1350.0 2284.9 667.3 0.1 FCP - 71.5 22834.0 68883.7 14704.8 29547.7 153.0 2641.0 176.3 1273.8 21013.0 2509.6 21726.0 25090.5 2222.0 86925.0 202.6 234.6 158.4 278.1 139.1 2959.7 910.7 1 VMm Beat/Jac L 95.0 16552.7 110072.5 8757.5 11543.5 76.3 2641.0 252.1 1273.8 21871.9 2509.6 20831.3 15624.1 2222.0 153905.0 107.1 134.7 143.3 191.9 2233.0 2374.7 668.0 2 VMm Beat/Jac H 86.4 24809.5 58864.4 15145.7 27569.1 487.3 2641.0 77.9 1273.8 26553.2 2509.6 28060.5 34177.6 2222.0 50533.1 304.2 328.2 234.7 225.6 272.7 2185.2 679.6 3 VMm Fisch L 74.2 20860.3 72265.7 16567.0 29356.5 360.6 2641.0 183.3 1273.8 23607.0 2509.6 28094.4 30970.1 2222.0 95011.7 187.3 222.6 166.7 192.6 1636.5 1551.4 1025.1 4 VMm Fisch H 76.6 18546.6 78484.0 12054.4 26687.7 55.8 2641.0 223.2 1273.8 21739.3 2509.6 23459.8 21828.0 2221.9 99231.4 169.4 198.3 177.9 225.7 1055.0 2939.9 896.7 5 VMm Marsh L 84.3 22177.0 59597.3 14271.4 29059.8 335.2 2641.0 182.8 1273.8 25101.3 2509.6 28843.9 35078.3 2221.5 54665.2 271.0 305.4 234.4 247.9 209.8 2943.8 904.4 6 VMm Marsh H 99.6 15401.3 119863.2 11443.4 25863.0 56.0 2641.0 223.2 1273.8 21673.1 2509.6 25188.4 20070.1 2221.6 155848.0 98.5 124.1 123.2 194.3 2784.1 1640.5 960.6 7 kf Beat/Jac L 125.2 13098.5 103970.4 15843.1 9301.9 129.9 2641.0 150.1 1273.8 9407.8 7436.2 29775.1 24374.5 2222.0 89500.7 393.4 367.1 264.4 160.8 543.9 2327.2 754.7 8 kf Beat/Jac H 88.6 42557.6 106318.3 17253.2 41538.1 65.6 2641.0 250.0 1273.8 42592.5 2509.6 13879.2 22381.5 2221.4 125358.5 184.7 216.2 157.2 230.2 2912.7 2044.4 736.9 9 kf Fisch L 82.4 20396.0 52335.1 11906.6 18389.2 411.0 2641.0 169.1 1273.8 15307.3 2509.6 20460.7 41637.4 2222.0 52901.7 261.8 297.7 128.6 755.2 966.9 1676.3 11059.2 10 kf Fisch H 79.7 24095.9 91208.3 17152.5 35587.5 63.8 2641.0 223.4 1273.8 26249.4 2509.6 24643.6 4741.2 2222.0 92175.3 188.1 216.4 223.1 171.2 567.7 1433.1 1119.5 11 kf Marsh L 115.5 17997.6 95588.7 11434.1 27259.5 168.3 2641.0 170.7 1273.8 18783.6 2509.6 27920.2 26620.7 2222.0 89738.7 349.4 376.1 290.7 227.9 322.2 2582.4 1171.7 12 kf Marsh H 117.3 24243.6 141259.8 16236.5 31132.3 130.1 2641.0 182.7 1273.8 26433.3 2509.6 23203.2 21782.7 2221.8 179532.4 88.8 104.4 106.9 251.4 2705.4 1651.6 951.7 13 VMf Beat/Jac L 98.1 18126.9 116568.8 13238.3 24513.0 103.3 2641.0 194.4 1273.8 20242.9 2509.6 19765.9 21603.3 2221.9 154433.7 103.3 125.5 117.1 249.7 2488.4 2103.0 699.7 14 VMf Beat/Jac H 104.2 24629.2 78592.9 14540.2 29097.2 247.8 2641.0 233.3 1273.8 24162.4 2509.6 27370.4 28751.0 2221.9 66861.7 360.0 371.9 280.8 191.7 1236.4 2216.7 743.5 15 VMf Fisch L 79.1 23040.1 86257.7 16261.8 29323.9 218.3 2641.0 152.4 1273.8 24893.2 2509.6 26571.0 29387.2 2221.2 114484.4 156.8 188.8 162.7 194.7 1191.5 2445.6 648.2 16 VMf Fisch H 75.0 18239.6 54882.6 12893.4 27829.5 115.3 2641.0 188.7 1273.8 19788.1 2509.6 23499.6 20828.8 2221.9 55922.1 240.3 276.4 223.3 201.3 919.6 2663.9 738.7 17 VMf Marsh L 108.5 20911.9 86411.8 11957.8 30523.1 175.5 2641.0 157.0 1273.8 22123.8 2509.6 27851.6 23079.4 2221.7 84684.7 355.2 375.5 282.4 210.6 1106.1 2513.3 968.9 18 VMf Marsh H 106.6 20782.0 127942.6 14207.0 26909.9 142.2 2641.0 184.9 1273.8 23617.7 2509.6 22601.1 25857.7 2222.1 170186.9 95.2 115.7 113.7 249.6 2530.0 1978.0 748.4 19 full-field path L 113.5 16390.9 78841.8 10923.8 13725.6 234.2 2641.0 175.2 1273.8 9958.2 2509.6 30144.9 32584.2 2221.9 50842.8 374.6 389.4 282.8 232.9 255.2 2592.3 1015.6 20 full-field path H 121.5 60897.0 145002.1 25626.7 47638.4 64.7 2641.0 191.1 1273.8 51125.7 2509.6 9146.3 6542.3 2221.9 151408.2 82.1 96.8 132.3 121.7 1047.1 1143.2 1364.0 21 VM gas cap L 91.6 50374.9 134766.1 19619.2 29370.8 5802.7 2641.0 181.3 1273.8 33780.5 2509.6 13042.7 4781.1 2221.9 153078.6 184.5 211.9 155.7 190.4 2722.8 2653.8 742.7 22 VM gas cap H 138.7 12993.8 144006.8 9642.9 18453.9 68.4 2641.0 233.6 1273.8 13751.7 4594.9 31780.1 48926.2 1911.4 130127.1 309.0 344.3 282.6 204.7 404.1 1913.1 1573.6 Table 30. Relative change in RMSE for same scenarios as Table 28. Color coding is the same as prior table. Field Beattie Jacoby Fischhaber Marshell Fischhaber Sensitivity Pressure Field Oil Field Gas Beattie Oil Beattie Gas Jacoby vert Jacoby vert Jacoby HD1 Jacoby HD1 Jacoby HD2 Jacoby HD2 Fischhaber Fischhaber Marshell Oil Marshell Gas pressure post- pressure post- pressure pressure post- Fischhaber Fischhaber Water post- Scenario # Parameter Value primary primary primary primary primary Oil primary Gas primary Oil primary Gas primary Oil primary Gas primary Oil primary Gas primary primary primary inj inj post-inj inj Gas post-inj Oil post-inj inj HM 66.5 22357.4 85378.4 15400.9 48116.1 76.4 160.1 253.7 89.8 22090.9 228.2 21053.8 3380.4 1215.2 59921.5 108.4 111.1 119.9 265.0 110.3 1961.4 2372.2 0 - - 74.7 20382.3 74476.5 14669.0 29547.7 153.0 2641.0 176.3 1273.8 22575.1 2509.6 24814.6 25090.5 2222.0 93359.6 179.2 211.3 175.2 202.8 1350.0 2284.9 667.3 0.1 dynamic fracture - -0.044 0.120 -0.075 0.002 0.000 0.000 0.000 0.000 0.000 -0.069 0.000 -0.124 0.000 0.000 -0.069 0.131 0.110 -0.096 0.371 -0.897 0.295 0.365 1 PVm Beat/Jac L 0.271 -0.188 0.478 -0.403 -0.609 -0.501 0.000 0.431 0.000 -0.031 0.000 -0.161 -0.377 0.000 0.649 -0.402 -0.363 -0.182 -0.054 0.654 0.039 0.001 2 PVm Beat/Jac H 0.156 0.217 -0.210 0.032 -0.067 2.185 0.000 -0.558 0.000 0.176 0.000 0.131 0.362 0.000 -0.459 0.698 0.553 0.340 0.113 -0.798 -0.044 0.018 3 PVm Fisch L -0.007 0.023 -0.030 0.129 -0.006 1.357 0.000 0.040 0.000 0.046 0.000 0.132 0.234 0.000 0.018 0.045 0.053 -0.048 -0.050 0.212 -0.321 0.536 4 PVm Fisch H 0.025 -0.090 0.054 -0.178 -0.097 -0.635 0.000 0.266 0.000 -0.037 0.000 -0.055 -0.130 0.000 0.063 -0.055 -0.062 0.016 0.113 -0.218 0.287 0.344 5 PVm Marsh L 0.128 0.088 -0.200 -0.027 -0.017 1.191 0.000 0.037 0.000 0.112 0.000 0.162 0.398 0.000 -0.414 0.512 0.445 0.338 0.222 -0.845 0.288 0.355 6 PVm Marsh H 0.333 -0.244 0.609 -0.220 -0.125 -0.634 0.000 0.266 0.000 -0.040 0.000 0.015 -0.200 0.000 0.669 -0.450 -0.413 -0.297 -0.042 1.062 -0.282 0.440 7 kf Beat/Jac L 0.676 -0.357 0.396 0.080 -0.685 -0.151 0.000 -0.148 0.000 -0.583 1.963 0.200 -0.029 0.000 -0.041 1.196 0.737 0.510 -0.207 -0.597 0.019 0.131 8 kf Beat/Jac H 0.186 1.088 0.428 0.176 0.406 -0.571 0.000 0.419 0.000 0.887 0.000 -0.441 -0.108 0.000 0.343 0.031 0.023 -0.102 0.135 1.158 -0.105 0.104 9 kf Fisch L 0.103 0.001 -0.297 -0.188 -0.378 1.686 0.000 -0.041 0.000 -0.322 0.000 -0.175 0.659 0.000 -0.433 0.461 0.409 -0.266 2.724 -0.284 -0.266 15.573 10 kf Fisch H 0.066 0.182 0.225 0.169 0.204 -0.583 0.000 0.268 0.000 0.163 0.000 -0.007 -0.811 0.000 -0.013 0.050 0.024 0.273 -0.156 -0.579 -0.373 0.678 11 kf Marsh L 0.545 -0.117 0.283 -0.221 -0.077 0.100 0.000 -0.031 0.000 -0.168 0.000 0.125 0.061 0.000 -0.039 0.950 0.780 0.660 0.124 -0.761 0.130 0.756 12 kf Marsh H 0.569 0.189 0.897 0.107 0.054 -0.150 0.000 0.036 0.000 0.171 0.000 -0.065 -0.132 0.000 0.923 -0.504 -0.506 -0.390 0.240 1.004 -0.277 0.426 13 VMf Beat/Jac L 0.312 -0.111 0.565 -0.098 -0.170 -0.325 0.000 0.103 0.000 -0.103 0.000 -0.203 -0.139 0.000 0.654 -0.424 -0.406 -0.332 0.231 0.843 -0.080 0.048 14 VMf Beat/Jac H 0.394 0.208 0.055 -0.009 -0.015 0.619 0.000 0.324 0.000 0.070 0.000 0.103 0.146 0.000 -0.284 1.009 0.760 0.603 -0.055 -0.084 -0.030 0.114 15 VMf Fisch L 0.058 0.130 0.158 0.109 -0.008 0.427 0.000 -0.135 0.000 0.103 0.000 0.071 0.171 0.000 0.226 -0.125 -0.107 -0.071 -0.040 -0.117 0.070 -0.029 16 VMf Fisch H 0.004 -0.105 -0.263 -0.121 -0.058 -0.246 0.000 0.071 0.000 -0.123 0.000 -0.053 -0.170 0.000 -0.401 0.341 0.308 0.275 -0.007 -0.319 0.166 0.107 17 VMf Marsh L 0.452 0.026 0.160 -0.185 0.033 0.147 0.000 -0.109 0.000 -0.020 0.000 0.122 -0.080 0.000 -0.093 0.983 0.777 0.612 0.038 -0.181 0.100 0.452 18 VMf Marsh H 0.426 0.020 0.718 -0.031 -0.089 -0.071 0.000 0.049 0.000 0.046 0.000 -0.089 0.031 0.000 0.823 -0.468 -0.453 -0.351 0.231 0.874 -0.134 0.122 19 full-field path L 0.519 -0.196 0.059 -0.255 -0.535 0.531 0.000 -0.006 0.000 -0.559 0.000 0.215 0.299 0.000 -0.455 1.090 0.843 0.615 0.148 -0.811 0.135 0.522 20 full-field path H 0.626 1.988 0.947 0.747 0.612 -0.577 0.000 0.084 0.000 1.265 0.000 -0.631 -0.739 0.000 0.622 -0.542 -0.542 -0.245 -0.400 -0.224 -0.500 1.044 21 VM gas cap L 0.226 1.472 0.810 0.337 -0.006 36.925 0.000 0.029 0.000 0.496 0.000 -0.474 -0.809 0.000 0.640 0.030 0.003 -0.111 -0.061 1.017 0.161 0.113 22 VM gas cap H 0.856 -0.362 0.934 -0.343 -0.375 -0.553 0.000 0.326 0.000 -0.391 0.831 0.281 0.950 -0.140 0.394 0.725 0.629 0.613 0.009 -0.701 -0.163 1.358 Battelle | May 2026 195 Figure 196. Well-by-well primary production volumes for base scenario (top) and aggregate of sensitivity Figure 197. Field-wide fluid volumes and pressure values during primary production for base scenario (top) and scenarios (bottom). In aggregate plot, dashed line shows best fit to data, while shaded region shows range of aggregate of sensitivity scenarios (bottom). In aggregate plot, dashed line shows best fit to data, while shaded results. region shows range of results. Battelle | May 2026 196 Figure 198. Well-by-well bottomhole pressure data during injection for base scenario (top) and aggregate of Figure 199. Fluid volumes from production test for base scenario (top) and aggregate of sensitivity scenarios sensitivity scenarios (bottom). In aggregate plot, dashed line shows best fit to data, while shaded region shows (bottom). In aggregate plot, dashed line shows best fit to data, while shaded region shows range of results. range of results. Battelle | May 2026 197 A large amount of information can be gleaned from the RMSE and relative RMSE tables. A discussion of a few key observations is provided here to illustrate the utility of the information contained. Figure 200 displays the relative change in RMSE for all variables across a selection of the key metrics – primary gas production, pressure at three of the wells during injection operations, and hydrocarbon production post-CO2 treatment – with the white bar showing change in the “high” case for that variable, and orange showing the same for the “low”. The plots are truncated at both a 100% increase and decrease in RMSE, though some of the RMSE changes are greater than this. The charts convey the largest contributing variables to each of these metrics, both in raw change and in terms of improving RMSE. A large improvement in RMSE across all of these metrics is observed when permeability of the throughgoing fracture/vug train is increased (the “high” case of full-field path variable). This pathway is clearly an important piece of the puzzle in determining subsurface behavior in this reservoir, but it does not tell the whole story. The presence of this high conductivity channel substantially decreases gas production (thereby increasing RMSE of primary gas production), which causes the depleted state of the reservoir to be far too high. The degree of observed improvements does seem to indicate a high conductivity pathway connects the entire feature to some extent. The presence – or lack thereof – of this throughgoing pathway is responsible for some of the largest raw changes to RMSE observed across metrics, both positive and negative. Clearly, identification and understanding of any such feature in other TBR fields will be crucial in controlling fluid migration for any future EOR operations. Figure 200. Tornado chart showing sensitivity of Fischhaber oil production post-CO2 treatment to various variables. Another key driver appears to be the size of the gas cap within the field. As with the throughgoing fracture, changes to the amount of free gas present at field discovery cause Battelle | May 2026 198 substantial changes across all metrics. For production of gas and CO2 at the Fischhaber well post-treatment, these changes are pronounced. The “high” case for this variable, which increases the relative pore volume of the gas cap, significantly improves RMSE for this metric by decreasing the volume of gas/CO2 produced. This makes sense, as the presence of a large amount of pore volume shallower than the location CO2 was injected provides more space for the CO2 to migrate into during both huff and puff operations and gas recycling. Any free gas in the system would then be further from the production well when it is turned on, thereby decreasing the volumes observed at the well. A decrease in free gas fraction has the opposite impact, and the change in RMSE resulting from this is even more pronounced (>100% increase in RMSE). Accurate representation of the gas cap in these saturated reservoirs is clearly important, indicating the gas leg of any TBR reservoirs should be adequately characterized to ensure success of CO2-EOR operations. One major observation was noted in this modeling process that is not captured simply by looking at the RMSE tables. The inclusion of dynamic fracture behavior in the system appears to explain some of the behavior observed post-treatment at the Lee field. Having the fracture open in early January, corresponding to a closure pressure of ~800 psi, does cause a marked increase in pressure at the Jacoby well. Figure 201 shows the model data co-plotted with field observations for downhole pressures while gauges were installed at this well. The qualitative similarity in the data in January 2025 – specifically the noted increase – provides strong evidence for the substantial increase in pressures being due to opening of a fracture. The fit itself is poor, indicating that the permeability of this fracture is likely too low. Some combination of changes in the other independent variables is necessary to truly improve the fit. This serves to indicate that these dynamic flow pathways are critical to understanding subsurface flow in these systems. Should any future CO2-EOR operations be conducted at other TBR fields, pressures should be closely monitored during initial injection operations to identify the presence of any dynamics in the system. Figure 201. Jacoby downhole gauge data (points) and model data from Scenario 0.1 with a dynamic opening and closing fracture. Note the pressure increase in early January in both datasets. This approach has its limitations. Sensitivities to some key variables in dual porosity systems, namely storativity ratio and interporosity flow coefficient, are not inherently assessed. The ratio of fracture to matrix porosity and permeability does implicitly depend on these variables, so there is some degree of analysis, though it is not explicit. Parameters like fracture density and Battelle | May 2026 199 fracture aperture size are not assessed. However, given the assumptions on geometry for the system (single fracture plane in each lateral dimension), any changes in fracture density or aperture would be reflected as a change to the ratio of matrix to fracture porosity in a given grid block. Sensitivity to relative permeability models is also not assessed due to lack of data. Core flooding experiments on field-specific cores would provide useful information on the relative permeability behaviors of these systems. Capillary entry pressure to the matrix system was not assessed in any of the modeling efforts and could potentially play an important role in determining to what extent CO2 penetrates the matrix system. This was also not assessed due to lack of data. It is also possible that non-Darcy flow impacts play a key role in fluid flow within these systems. Given that some of the cavernous vugs observed in image logs within certain TBR reservoirs are on the order of mm-cm in diameter, there is opportunity for laminar or turbulent flows to occur within the reservoir. Both in initial modeling and sensitivity assessments, these limitations were acknowledged and maintained in an effort to avoid overcomplicating a model that was built on sparse data. In addition, regarding the sensitivity results specifically, there is no guarantee when a combination of factors are changed within the model, that the change in RMSE will follow the same trends observed in each variable individually. However, these results provide some insight into what may be missing from the existing reservoir model, what controls the effectiveness of CO2 treatments in these HTD reservoirs, and what geologic characterization will be most important to screen candidate reservoirs. Scale Up to other TBR Reservoirs Given the lack of oil production observed at the Lee 26 field after CO2 treatment, alternative modeling approaches were pursued to estimate the feasibility of CO2-EOR in other TBR reservoirs. A series of simplified reservoir performance models were developed for the two largest fields in the TBR trend – Napoleon and Albion-Scipio – resolving volumetric parameters such as the potential incremental oil production (stock tank barrels [STB], one thousand STB [MSTB], or million STB [MMSTB]), associated water production (STB, MSTB, or MMSTB), and CO2 production and recycling (tonnes or millions tonnes [MMt]). A substantial amount of residual oil remains in these fields, and the field testing has shown that mobilization of the residual oil in this trend is possible. All that is needed is an appropriate field optimization scheme to tap into these resources. Due to the remaining uncertainty in how effectively such a field program might perform, a range of production scenarios were considered to explore the variety of possible outcomes for trend-scale implementation. The primary variables of interest across the various production scenarios are defined as follows: 1. Recovery Factor (RF) during primary production o Percentage of OOIP produced to date 2. Incremental RF due to CO2-EOR o Percentage of OOIP produced during hypothetical CO2-EOR operations 3. Time to peak oil production o Time from start of CO2 injection to peak oil production rate 4. CO2 utilization ratio o Barrels of oil produced per tonne of CO2 injected The high and low values for each parameter are shown in Table 31. Primary constraints used for scenario generation are denoted in the table. The range of values for the primary recovery factor Battelle | May 2026 200 was constrained using existing literature (Verma, 2017; Hurley and Budros, 1990) and material balance modeling conducted as part of this project for the Lee 26 and Napoleon fields. Incremental recovery factors were based on literature (Verma, 2017) and core flooding experiments conducted in Task 3. Time to peak production values were based on trends observed in literature (Verma, 2017) and collaboration with domain experts. Finally, CO2 utilization values were constrained by literature (Verma, 2017; Miller et al., 1994; Thomas, 1989; Haskin and Alston, 1989; Yang et al., 2010; Mohammed-Singh et al., 2006). For all parameters, the results of field testing at the Lee 26 field were weighed against compiled values to determine final ranges of values. Production scenarios were generated using a decline curve forecasting method loosely following the analysis done in Verma (2017), Chapter C for existing CO2-EOR floods. The scenario generation process is outlined step-by-step below. Table 31. Parameters used for production scenario generation Parameter Low Value High Value Primary constraint? Primary RF 15% 40% X Incremental RF 5% 15% X Time to Peak 1 yr 5 yrs X CO2 Utilization 2 stb/ton 12 stb/ton 1. Fit decline curves to monthly production data For each field, a decline curve was fit to monthly oil production data pulled from the Michigan EGLE Oil and Gas database (MI EGLE, 2026) using least-squares regression. Two parallel decline curve analysis methods are used to describe recent production trends – one fitting oil production rate versus time (assuming exponential decline), and another linear fit to production rate versus cumulative production. For clarity, these are referred to as the “time decline” curve and the “q versus Q” curve, respectively. The fitted curves for the Albion-Scipio production data are shown in Figure 202 (note that only data since 1982 is used for the fits). The time decline method is useful for estimating how long a particular field will be economically viable, but cannot provide an ultimate production number due to the asymptotic behavior of the exponential fit. The time decline curve for each field gives a baseline production case to which the forecasting scenarios can be compared. Conversely, the q versus Q curve provides an estimate of ultimate recovery but cannot describe the time-dependent behavior of the production rate. Ultimate cumulative recovery is calculated for each field using the q versus Q curves (the x-intercept of the fit) and used as an input to the next calculation step. Battelle | May 2026 201 Figure 202. Fitted time decline (left) and q vs. Q (right) curves for Albion-Scipio field using EGLE data. 2. Calculate OOIP and incremental oil production Using the ultimate cumulative recovery volumes from step 1, the OOIP for each field is calculated by dividing ultimate recovery by the assumed primary RF. Incremental recovery is then calculated by multiplying the OOIP value by assumed incremental RF. The incremental recovery volume is used to constrain the CO2-EOR decline curves. 3. Model forecasted decline curves for CO2-EOR The sum of the incremental and ultimate primary recovery gives a zero-rate point on the q versus Q graph, assumed to be the final cumulative volume after EOR operations. The slope of the CO2-EOR decline curve is estimated from field data reported in Verma (2017), Chapter C by examining the slope ratio of the CO2-EOR curve to the primary/secondary decline curve. Average values for this ratio range 0.53-0.65; to keep estimates conservative, the lower slope ratio is chosen. The CO2-EOR q versus Q decline curve is then generated by multiplying the slope of the primary production q versus Q curve by the slope ratio (0.53), and using the zero rate, final cumulative volume point to constrain the line. Using sampled points on this q versus Q decline curve, time series data (production rate versus time) can be calculated by dividing change in cumulative volume between two points by the higher of the two rates – essentially a left Riemann approximation. An exponential decline is then fitted to the calculated time series data, resulting in a bounding decline curve for a particular pair of primary RF and incremental RF. In order to vary the time to peak production across scenarios, a straight line is drawn (in rate versus time space) from the last date of production data to the bounding curve. This straight line represents a period of defined duration (1-5 years) where production rate increases to a peak rate, after which exponential decline begins. Figure 203 shows a subset of the scenarios generated via this method for the Albion- Scipio field. The red line shows the primary production decline curve. Black lines show the bounding exponential models for two separate primary recovery factors (15%, 40%), and the dashed lines show the six scenarios (three for each bounding curve) generated through variation in time to peak production. Battelle | May 2026 202 Figure 203. Subset of production rate scenarios (dashed lines) generated for the Albion-Scipio field. 4. Calculate CO2 demand Rather than using an assumed CO2 utilization ratio to calculate demand, the necessary mass of CO2 is derived from assuming a voidage replacement flood strategy (volume in = volume out) in the hydrocarbon column. The mass of injected CO2 is based on the peak production rate observed for each scenario and held constant throughout the entire project lifetime. Miscible flooding is assumed; hence the reservoir volume associated with peak production (assumed oil formation volume factor of 1.0 for simplicity) is converted to CO2 mass using a density value associated with the minimum miscibility pressure of Napoleon oil measured in Task 3. A subsurface temperature of 100oF is assumed, consistent with measured bottom hole temperatures at both the Lee field and the Napoleon field. Prior to peak production, while rates are ramping up, the volume of injected CO2 exceeds production volumes to approximate pressurization of the field. After peak production is reached and production rates begin to decrease, the makeup volume of the production stream is assumed to be made up of CO2. This results in an increasing volume of CO2 produced as the project continues. The time-series of oil production rate, CO2 injection rate and production rate for a specific scenario at the Albion- Scipio field is shown in Figure 204. Battelle | May 2026 203 Figure 204. Oil produced, CO2 injected and CO2 produced for a single scenario at the Albion-Scipio field. An implicit assumption of subsurface dynamics consistent with a CO2 flood constrains this approach. Initial injection of CO2 generates pressure in the field, pushing some residual oil toward the production well(s). As CO2 spreads further into the reservoir, multi-contact miscibility mobilizes, increasing volumes of residual oil which are pushed toward the production well. Immediately following peak production, the CO2 front arrives at the production well, after which the production well stream is a mix of oil and CO2. Despite the pressure maintenance of a voidage replacement flood strategy, CO2 is only able to mobilize a finite volume of oil. The production decline period is a result of decreasing CO2 utilization ratio over time as this finite volume is extracted. The overall CO2 utilization for these scenarios fell between 2.5 and 5.3 stb/tonne which is consistent with the range of utilization ratios observed in literature. Some additional assumptions concerning fluid volumes are used in this economic model. All of the produced CO2 is assumed to be recycled back into the injection stream. Water production is assumed to occur at a 2:1 ratio with produced oil. This is somewhat lower than observed water cut in other CO2-EOR operations; however, most of these projects are preceded by a water flood, which is not the case here. In addition, these Trenton-Black River fields have minimal water drive (Hurley and Budros, 1990), meaning the volume of water in the system is relatively finite. This 2:1 water-oil ratio is consistent with the Napoleon and Albion-Scipio fields based on EGLE production data (MI EGLE, 2026). Produced water is assumed to be sent offsite for disposal. A total of 24 modeling realizations were generated – 12 for each field – by varying the primary constraints of Table 30 (a time to peak production of 3 years was also considered). A subset of these results were used to constrain the economic analysis of trend-wide deployment. Battelle | May 2026 204 9.0 Task 8: Development Plan The work completed over the course of this study has yielded significant new information about the behavior of the Trenton-Black River hydrothermal dolomite play, including new insights into reservoir architecture, seismic evaluation of reservoir presence and quality, reservoir response to stimulation with surfactant-enhanced CO2, and operational lessons learned before, during, and after CO2 injection. These learnings are synthesized below into a development plan that can be used by operators to optimize CO2-EOR operations in the TBR and other hydrothermal dolomite plays. 9.1 Subtask 8.1: Economic Analysis This section presents the economic assessment of the Albion-Scipio and Napoleon Oilfields. The cost-benefit analysis uses results from the reservoir model, described in Section 8.1, in conjunction with a revenue and cost model to determine the net present value, internal rate of return, and CO2 break-even cost (Figure 205). This analysis framework was used in previous work by Battelle described by Fukai et al. (2015) and relies on work conducted by Advanced Resources International (ARI) and researchers at Carnegie Mellon University (CMU) in collaboration with the DOE’s NETL (ARI, 2006; McCoy and Rubin, 2006; McCoy, 2009; McCoy and Rubin, 2009; U.S. DOE-NETL, 2011, 2014a). Calculation schemes for the revenue and cost models were selected from these studies and verified using project-specific data and internal expertise. Reservoir data was derived from a dynamic reservoir model based on a SEM developed using site-specific data and analysis. CO2-EOR performance results from the reservoir model, described in Section 8.1, are integrated into separate revenue and cost models that calculate net revenue and total project costs, respectively. The resulting net cash flow is then discounted to derive the net present value (NPV) and internal rate of return (IRR) of the project. The break-even cost of CO2 is calculated to represent the maximum price a CO2-EOR operator could pay for CO2 without incurring loss on the project, and is calculated as a function of oil price to help determine a viable market price for captured CO2 based on the prevailing price of oil. CO2 storage is treated as a byproduct of EOR operations and is not directly associated with incurrence of project costs. As such, injection wells considered in this assessment can be categorized as Class II EOR wells, rather than Class VI under the United States Environmental Protection Agency (U.S. EPA) underground injection control (UIC) program (U.S. EPA, 2015). The value of 45Q tax credits was considered in all the scenarios to determine if claiming the credits could improve the economics in scenarios where the value of oil production is insufficient to generate a positive NPV. The costs of CO2 site monitoring and verification were ignored, as this could be reasonably assumed to be available from initial development and production of the oilfield. Liability was also not included in the analysis. The cost model assumes the CO2 initially delivered to the site via trunkline will be of sufficient purity and pressure for miscible injection. A per tonne cost of CO2 is assigned based on the maximum amount of CO2 required for field development. The costs and revenue associated with the recovery of natural gas and natural gas liquids were not included in the analysis. Battelle | May 2026 205 Figure 205. Schematic of the cost-benefit analysis framework used in this work (adapted from Fukai et al., 2015). Cost-Benefit Analysis Methodology Reservoir Performance Model Reservoirs of Interest In order to approximate scale-up across the entire TBR trend in southern Michigan, the two largest fields in the trend – Albion-Scipio and Napoleon – were considered for economic analysis. These two fields account for roughly 86% of oil production to date within the trend, hence providing a reasonable metric for full trend scale-up. Field development history, including historical and current well completions, are used to characterize the Napoleon and Albion-Scipio field geometries for flood design. Specifically, this includes: the number of active and total wells in the field; and the number of injection wells, production wells, and injection patterns needed for a full-field development scenario. These values were generated using data from the Michigan Department of the Environment, Great Lakes and Energy (MI EGLE, 2025) and are shown for each field in Table 32. Battelle | May 2026 206 Table 32. Relevant reservoir criteria for the Napoleon and Albion-Scipio Oilfields. Parameter Unit Napoleon Albion-Scipio No. active wells (all active production wells) Count 55 186 No. wells in field (plugged and abandoned Count 58 673 and active) No. Injection Wells Needed for full field development (No. Injection Patterns for full Count 11 135 field development) No. Production Wells Needed for full field Count 44 540 development No. Injection Patterns if only active wells are Count 11 37 used Model Realizations As described in Task 3.18.2, a series of simplified reservoir performance models were developed for the Napoleon and Albion-Scipio fields, resolving volumetric parameters such as the potential incremental oil production (stock tank barrels [STB], one thousand STB [MSTB], or million STB [MMSTB]), associated water production (STB, MSTB, or MMSTB), and CO2 production and recycling (tonnes or millions tonnes [MMt]). Modeling results from the scale-up analysis were used as input to the economic assessment. Four of the twelve model realizations at each field, covering the full range of performance outcomes, were selected for economic modeling. These scenarios are: • Medium-Low Scenario: 15% Primary RF, 5 year to peak production, 5% Incremental RF • Low Scenario: 40% Primary RF, 5 year to peak production, 5% Incremental RF • High Scenario: 15% Primary RF, 1 year to peak production, 15% Incremental RF • Medium-High Scenario: 40% Primary RF, 1 year to peak production, 15% Incremental RF The results of these four modeling scenarios are summarized in Table 33. The oil production rates for these scenarios at the Albion Scipio field are shown in Figure 206, along with field production data since 1980 and modeled decline curve to this production data (similar curves were produced for the Napoleon field). It should be noted that the Albion-Scipio field was mostly depleted by 1980, and peak production rates at the field were more in line with the modeled rates post-CO2 treatment. Table 33. Summary of outputs from reservoir performance models Model Outputs Units Albion-Scipio Napoleon Oil Production (per injection pattern) MSTB 129-1,020 249-1,250 Oil Production (per field) MMSTB 17.3-137 2.74-13.8 Water Production (per injection pattern) MSTB 258-2,040 498-2,500 Water Production (per field) MMSTB 34.6-273 5.48-27.5 Annual CO2 Injected (average, per injection 1,580- pattern) tonnes/yr 1,590-13,800 12,300 Annual CO2 Injected (average, per field) MMt/yr 0.213-1.85 0.017-0.135 CO2 Recycle Rate (average) Percent 59.3-72.5 51.7-64.2 Ultimate CO2 Demand (per injection pattern) MMt 0.019-0.114 0.023-0.132 Ultimate CO2 Demand (per field) MMt 2.60-15.3 0.252-1.45 CO2 Utilization Ratio (barrels per tonne of CO2) STB/tonne 2.46-2.71 3.39-5.26 Battelle | May 2026 207 Figure 206. Oil Rates at the Albion-Scipio field used as input to economic models. Revenue Model The revenue model calculates net revenue from oil recovered and sold at a specified oil price, with relevant tax deductions (Table 34). Royalties comprising 15% of the gross revenue were included in the calculation, derived from the average royalty percentage employed in the ARI and CMU models. The current Michigan severance tax on conventional crude oil production of 6.6% and an oil and gas fee of 1% was applied (Michigan Department of the Treasury, 2026). An Ad Valorem tax of 3.3% was determined, based on the average tax percentage levied on mineral properties in Lee Township of Calhoun County, Michigan (Transnation Title Agency, 2024). Severance and Ad Valorem taxes were deducted from the remaining revenue after royalty taxes. The revenue model calculation scheme produces a time series of incoming net cash flows, where the resulting net revenue of the project can be evaluated incrementally during specific periods of operation, and cumulatively at the end of the project life. Table 34. Inputs and definitions of parameters included in the revenue model. Revenue Model Input/Definition Parameter Volume of incremental or cumulative oil produced during reservoir performance Oil Produced (STB) simulations Market price of oil. Simulations are for the market price of oil to be $40/bbl, Oil Price (USD/STB) $60/bbl, and $90/bbl Gross Revenue Revenue from oil recovered and sold at the specified market price; oil (USD) recovered (STB) x oil price ($/STB) Royalty (%) 15% of gross revenue; 0.15 x gross revenue ($) Severance Tax (%) 6.6% of revenue after royalty tax; 0.066 x gross revenue ($) Oil and Gas Fee (%) 1% of gross cash market value; 0.01 x gross revenue ($) Ad Valorem Tax (%) 3.3% of revenue after royalty tax; 0.033 x gross revenue ($) Revenue after deduction of royalty, severance, and ad valorem taxes; Gross Net Revenue (USD) revenue ($) - Royalty ($) - Severance ($) - oil and gas fee ($) - Ad Valorem ($) Cost Models Cost models were used to determine the price of an integrated CO2-EOR project. The processes for determining the costs of capture, transport, CO2 injection, and oil production are detailed in this section. Battelle | May 2026 208 Capture Cost Model The model assumes that CO2 is provided at a price per tonne from an outside entity. In this way CO2 capture and transport costs are assumed to be a fee for service and the costs are incurred during the operational period only. This approach assumes that CO2 will be available from industrial and/or electric power facilities with CO2 capture facilities. As such, an additional 15% is added to the reported capture cost to simulate the profit collected by the capture entity. Table 35. Cost of CO2 Capture of industrial and electric power sources from DOE/NETL (2023a, b) and DOE/NETL (2022a, b), respectively. An adder of 15% was added to simulate a profit collected by the CO2 capture entity. Capital Purchased Fixed Variable Total Source Type Costs Power/Fuel O&M ($/t) O&M ($/t) ($/t) ($/t) ($/t) Ammonia 6.1 3.9 2.7 6.3 19.0 Ethylene Oxide 9.4 9.8 1.7 5.2 26.0 Ethanol 14.1 9.2 1.7 6.8 31.8 Natural Gas Processing 6.2 3.4 1.5 5.0 16.1 Coal-to-Liquids 2.0 0.7 0.3 2.6 5.6 Gas-to-Liquids 2.9 1.2 0.3 1.9 6.4 H2 Refinery, 99% Capture 21.3 14.4 5.1 16.5 57.3 H2 Refinery, 90% Capture 22.8 15.6 5.3 16.2 59.9 Cement, 99% Capture 21.8 10.6 5.9 22.6 60.8 Cement, 90% Capture 22.8 11.1 6.1 22.6 62.7 Iron/Steel (Retrofit), 99% Capture 27.8 9.3 5.6 22.6 65.4 Iron/Steel (Retrofit), 90% Capture 28.0 9.5 5.7 22.6 65.9 Subcritical Pulverized Coal power plant1 19.2 5.7 6.1 6.2 37.2 Natural Gas Combine Cycle power 31.0 10.3 6.5 10.2 57.9 plant1 Notes: 1. Average cost of CO2 capture. Table 36. Electric power and industrial facilities and the distance to the center of the Lee, Napoleon, and Albion-Scipio Oilfields. Straight-Line Distance to Field Assumed Facility Capturable Transport (miles) Facility Cost2 Type CO2 (MMt) Mech.1 Albion- ($/tonne) Lee Napoleon Scipio Anderson’s Marathon Ethanol 0.186 Pipeline 42.32 7.8 31.2 13.3 Carbon Green Ethanol 0.068 Truck 151.57 31.1 62.4 51.9 POET Ethanol 0.090 Truck 151.57 107.0 100.3 114.8 Air Products Hydrogen 0.363 Pipeline 71.65 87.1 54.5 78.5 Dan E. Karn Power Power 2.978 Pipeline 72.34 103.8 105.9 115.9 Plant Plant Midland Generation Power 3.253 Pipeline 72.34 91.1 100.5 105.9 Plant Dearborn Industrial Power 2.713 Pipeline 72.34 87.1 54.5 78.5 Plant Notes: 1. Transport mechanisms add $5 or $100 per tonne for pipeline and truck, respectively. Assumed cost is the cost of capture + assumed cost of transport with a 15% added cost to simulate profits collected by the capture and transport entities. Battelle | May 2026 209 Injection/Storage Cost Model The cost model calculated total costs for the CO2-EOR project and are broken out by capital costs and operating costs. Capital costs include well-related costs and CO2 distribution and recycling equipment. Capital costs are accrued in Year 1 of the project. Operating costs are accrued for the remainder of the project. Total Operations and Maintenance (O&M) costs account for consumables, services, and labor associated with ongoing site operation and maintenance (surface and subsurface), liquid lifting/handling, general and administrative (G&A) expenses, and plugging and abandonment for all wells and site restoration for all equipment and piping. Capital Cost calculations used to estimate injection and production well equipment costs and CO2 related costs (i.e. CO2 distribution and recycling plant capital costs) are outlined in Table 37. This process has been used by Battelle before in a study funded by the Ohio Coal Development Office (OCDO) as described by Fukai et al. (2015). Equations established by the previous studies were used to calculate costs. All capital and operating costs were adjusted to represent constant 2023 U.S. dollars (USD) using S&P Global Upstream Capital Cost Index (UCCI) (S&P Global, 2024). These values were then escalated to 2026$ equivalents (the year the project is assumed to start) using an escalation rate of 3.3%. Well-related costs account for drilling and completion (D&C), equipment, services, labor, and site-related capital costs associated with new injection well equipment and existing well conversion. Well D&C costs were calculated for all newly installed wells using Equation 1 in Table 38. New producing equipment for all newly installed production wells and a percentage of the existing wells (see Other Cost Items, below) using Equation 2. Injection well equipment is calculated for all injection wells (i.e., newly drilled injection wells and converted existing production wells) using Equation 3. 20% of existing wells are assumed to be converted to injection wells in all scenarios. These costs are calculated using Equation 4. CO2 costs include the capital costs associated with on-site CO2 distribution and CO2 recycling. Onsite distribution costs are based on the maximum amount of CO2 to be distributed per day multiplied by the length of the lines (i.e., the length of the field). These costs are calculated using Equation 6. The CO2 recycling rate in standard cubic feet (scf; 60°F, 1 atm) per day was used to estimate the capital costs of the CO2 recycling facility (Equation 4). CO2 recycling, including on-site separation, processing, and recompression as shown in Figure 207. Battelle | May 2026 210 Table 37. Cost model parameters, equations, and associated references. Where needed, capital costs escalated from 2006$ to 2023$ using a factor of 1.317 based on the S&P Global Upstream Capital Cost Index (UCCI). Capital costs escalated from 2009$ to 2023$ using a factor of 1.095 based on the S&P Global UCCI. (S&P Global, 2024). Original Reference cited Cost Model Parameter Equation by Fukai et al. (2015) Well Capital Costs Drilling and Completion (2026$) ($360 x depth [ft]) + $360,000 (1) Estimate New Producing Well Equipment ($110 x depth [ft]) + $130,000 (2) Estimate (2026$) Injection Well Equipment (EQi) (2026$) ($183 x depth [ft]) + $78,000 (3) Estimate Well Conversion (2006$) ($6.24 x depth [ft]) + $8,950 ARI (2006), adjusted with Well Conversion (2023$) ($8.21 x depth [ft]) + $11,800 (4) UCCI CO2 Capital Costs $150,000 (Fixed) + Variable (pipeline mi x cost/mi): -$80,000/mi for <15 MMcf/d CO2 Distribution (2006$) -$120,000/mi for 15 to 35 MMcf/d -$160,000/mi for 35 to 60 MMcf/d -$200,000/mi for >60 MMcf/d ARI (2006), adjusted with $188,000 (Fixed) + UCCI Variable (pipeline mi x cost/mi): -$100,000/mi for <15 MMcf/d CO2 Distribution (2023$) -$150,000/mi for 15 to 35 MMcf/d -$200,000/mi for 35 to 60 MMcf/d -$250,000/mi for >60 MMcf/d (5) $700,000 x max.CO2 recycling rate CO2 Recycling Plant Capital (2006$) (MMcf) ARI (2006), $922,000 x max.CO2 recycling rate adjusted with CO2 Recycling Plant Capital (2023$) UOCI (MMcf) (6) Total O&M Costs Periodic O&M (2009$) $33,684e0.0001xdepth x no. of wells McCoy, 2009; McCoy & Rubin, Periodic O&M (2023$) $36,884e0.0001xdepth x no. of wells (8) 2009 Liquid Lifting Costs (2006$) $0.25 x (produced water + oil) ARI (2006), adjusted with Liquid Lifting Costs (2023$) $0.27 x (produced water + oil) (9) UOCI CO2 Recycling O&M (2006$) ARI (2006), (0.01 x USD/STB oil) x CO2 recycled adjusted with CO2 Recycling O&M (2023$) (7) UOCI General and Administrative (G&A) 0.20 x ($O&M + $Liquid Lifting) (10) ARI (2006) (2006$ and 2023$) Battelle | May 2026 211 Figure 207. Simplified diagram showing components of CO2 injection and recycling operations. Other Capital Cost Items: Additional cost items that are not directly related to well drilling, well equipment, CO2 distribution, or CO2 recycling were also considered in the analysis (presented as 2023$ in Table 38). The model predicts costs for well testing and rehabilitation of existing wells. All existing wells will be tested with a four-armed caliper and mechanical integrity test (MIT). An estimated 25% of existing wells are assumed to need well rehabilitation. The same percentage of wells are assumed to need replacement production well equipment. Newly drilled wells have additional costs beyond drilling and completion and production or injection well equipment, calculated above. Because the Albion-Scipio and Napoleon Fields Characterization of new wells is assumed to be $100,000 per well and includes a triple combo log and cement bond log (CBL). Battelle | May 2026 212 Table 38. Other Capital and Operating Costs used in the model. Cost Category Price Year $ Reference Other Capital Costs Four-Armed Caliper (per existing well) $7,000 2026$ Internal Expertise MIT (per existing well) $7,500 2026$ Internal Expertise Well Rehabilitation (per existing well needing $78,400 2026$ Estimate rehabilitation) Well Rehabilitation Factor (% existing wells 25% NA Estimate needing rehabilitation) Well Production Equipment Factor (% existing wells needing new production equipment) 5% NA Estimate Characterization Costs for newly drilled wells $100,000 2026$ Estimate (per well) Other Operating Costs Contracted Water Disposal Costs (per MSTB) $1,000 2026$ Reference1 From DOE CO2 storage cost model Well Plugging and Abandonment (per well) $150,000 2023$ (injection well), with escalations using the UOCI. Equipment Removal and Site Restoration ($) $1,000,000 2026$ Estimate Approximate CO2 price (2023$) given CO2 Price ($/tonne)1 $40/$70 2023$ nearby sources (largely natural gas combined cycle with some ethanol) Notes: 1. A price of $40 per tonne was used for the Napoleon Field because the CO2 requirements could be satisfied from a nearby ethanol plant. A price of $70 per tonne was used for the Albion-Scipio because more CO2 is required and will likely need to come from a larger, more dilute source such as a nearby natural gas combined cycle power plant. Total O&M costs: Total O&M costs for the site include periodic O&M costs, liquid lifting costs, and G&A expenses. All operating costs data were escalated to 2023$ equivalents using the Upstream Operation Cost Index (UOCI) (S&P Global, 2024). These values were then escalated to 2026$ equivalents (the year the project is assumed to start) using an escalation rate of 3.3%. Periodic O&M costs reflect ongoing expenses associated with the operation and maintenance of the CO2-EOR site at the required capacity over a given time, including electricity, consumables, periodic well work over, surface and subsurface repairs (Equation 8, Table 37). Periodic O&M costs are also dependent on the number and depth of wells used in the project. Liquid lifting costs are estimated as $0.27 per volume of oil and water produced (Equation 9, Table 37). CO2 operating costs include CO2 purchased for injection (see Table 39) and operations of the CO2 recycling plant. Ongoing O&M costs for the CO2 recycling plant are determined by the volume of CO2 recycled (scf) per time series set at 1% of the price of oil per STB to reflect the cost of on-site generation of the energy required for operation (Equation 7, Table 38). Non- injection or production operational costs include off-site water disposal, well plugging, and site restoration. Water disposal costs were calculated on a per barrel basis and assumed to be $1 per STB (or $1,000 per MSTB) of water. Well plugging and abandonment costs are per well costs obtained from the DOE/NETL Saline Storage Cost Model (DOE/NETL, 2023). Equipment removal and site restoration are based on estimates and are single lump-sum cost. Plugging and abandonment and equipment removal and site restorage area incurred at the end of the project. G&A costs are calculated as 20% of the sum of periodic O&M and liquid lifting costs (Equation 10). 1 https://earthworks.org/wp-content/uploads/2021/09/MicrosoftwordProducedWaterFactsheet2.pdf Battelle | May 2026 213 Financial Metrics Financial metrics used in the model are outlined in Table 39. All costs are reported in 2026$. A nominal discount factor of 15% (and associated real discount factor of 11.3%) and an escalation factor of 3.3% was used. The current 45Q a rate of $85 per tonne for CO2 utilization was used and escalated from the base year of 2026 using the escalation rate of 3.3%. The credit is only claimed for the first 12 years of the project, and only 90% of the total CO2 stored for each of these 12 years is assumed to be eligible. This reduction simulates the lifecycle CO2-equivalent emissions from the project, which would reduce the value of the credit. Table 39. Financial metrics used for modeling efforts. Metric Value Reference Base Year for Costs 2026 NA Project Start Year 2026 NA Nom. Discount Factor (%) 15.0% Standard Discount Factor Calculated from the nominal discount factor Real Discount Factor (%) 11.3% using the Fisher conversion Escalation Factor (%) 3.30% Standard Escalation Current value of 45Q tax credits. Claimed for 45Q Value ($/tonne) $85 the 12 first twelve year of project only. % of CO2 that can be claimed for 45Q 90% Assumption for Lifecycle penalty Analysis Scenarios and Evaluation Metrics Price Scenarios The market scenarios examined for the cost-benefit analysis of potential CO2-EOR co-storage projects in the oilfields of interest reflected an oil price of $60 per barrel. A CO2 price was determined by the nearby emitters and annual CO2 demand. The analysis assumes that the $60/STB of oil (current oil price) was defined as the lower price scenario and $80/STB of oil as the higher price scenario in the NPV calculated in the analysis. The effect of discount rate on CO2-EOR project economics was also evaluated. The discount rate represents the percentage difference between the present and future value of an expected cash flow, accounting for the time-dependent value of money and risk associated with a future investment. A discount rate of 15% was used. The analysis considers two operating periods, 12 years and 20 years, for all scenarios. Net Present Value The value of future cash flows accrued incrementally and cumulatively over the duration of the CO2-EOR project is represented by the NPV. The NPV is calculated as the sum of cash inflows and outflows, discounted to account for the time value of money and the risk/uncertainty associated with future cash flows. The net revenues (incoming cash flows) from the revenue model were integrated with total costs (outgoing cash flows) calculated by the cost model to derive a time series (years) of NPV and IRR estimates. The cumulative NPV calculated at the end of the operation was then used to evaluate the profitability of the CO2-EOR operation, such that positive values represent scenarios of profit and negative values represent financial losses incurred by investors in the project. Battelle | May 2026 214 CO2 Break-Even Cost The break-even cost of CO2 is the CO2 purchase price that results in a NPV equal to zero, and can be interpreted as the maximum price an EOR operator could pay for CO2 without incurring a loss on the project (McCoy, 2009; McCoy and Rubin, 2009). The break-even cost of CO2 was calculated as a function of various oil prices to define the specific price conditions at which CO2 becomes a viable purchase for EOR, and a marketable asset for capture facilities. The overall trend observed in the CO2 break-even cost calculated over the range of oil prices was used to derive the unit value of CO2 relative to oil for the oil field of interest. It should be noted that CO2- EOR requires higher upfront capital and operational investments, such that the break-even cost represents a much different timeline compared to primary oil production. Effect of 45Q Tax Credits The effect of the 45Q tax credits on project economics was determined for all scenarios. The tax credits were assumed to be $85/tonne based on recent changes enacted in the One Big Beautiful Bill. The tax credits are assumed to be claimed for 12 years, are escalated 3.3% a year from a base year of 2026, and are claimed for only 90% of the CO2 stored, according to the model, to simulate losses during the CO2 lifecycle analysis. Effect of Existing Infrastructure For each scenario of the Albion-Scipio Field, two possible simulation groups were modeled. The first considers the full field development, which requires the drilling of additional injection and production wells to match the number of active and plugged and abandoned wells in the field and, thus, the reservoir modeling domain. The second considers only the wells in the field that are currently active. For this scenario, only 37 of the possible 135 injection well patterns were available (approximately 27.41%). For the scenarios that consider existing infrastructure only, the reservoir model inputs (e.g. CO2 injection, oil production, and CO2 recycling rate) will be multiplied by a factor of 27.41% to account for the smaller field area covered by these scenarios. This process was not completed for the Napoleon Field because nearly all of the wells in the field are still active. Application of Methodology to the Oilfields of Interest The economic assessment focused on two oilfields of interest, the Albion-Scipio and the Napoleon Oilfields. This section presents each oilfield separately and considers their development over a 12-year project and a 20-year project. For each scenario, the revenue was calculated at an oil price of $60 per STB (approximate current price) and $80 per STB (last reached in 2023) (macrotrends, 2026). All costs and revenue are presented as real costs/revenue in 2026$. Four scenarios were presented for each Simulation Group. They are bound by a scenario with low recovery factor (Low RF) and high recovery factor (High RF) to show the pessimistic and optimistic case, respectively, and are presented as Scenarios 6 and 7, respectively. Two scenarios with moderate recovery factors were also considered, one with a longer time to achieve peak oil production (Mod RF, long peak) and another with a shorter time to achieve peak oil production (Mod RF, short peak). These scenarios are presented as Scenarios 3 and 10, respectively. Battelle | May 2026 215 Albion-Scipio Economic model results for the Albion-Scipio Oilfield are presented for a 12-year project and a 20-year project in Table 40 and Table 41, respectively. This section discusses the results of the 12-year projects followed a discussion of the results of the 20-year projects. As of 2023, the Albion-Scipio Field has 186 active wells and 487 abandoned wells (673 wells total) that produced oil. Because of the number of legacy wells, two different simulation groups were completed for the field. The first considers the economics of redrilling several hundred wells to support 135 injection patterns of four producing wells to every injection well (675 wells total). This set of simulations, labeled Group 1, most closely matches the results of reservoir modeling because the entirety of the oilfield is produceable by the 135 injection patterns. The second set of simulations, referred to as Group 2, considers only the 186 active wells in the field, yielding 37 injection patterns of four producers to every injection well. Because there is no evidence that the existing wells are in a part of the reservoir that produces either a higher or lower than the average amount of oil, the analysis was simplified by assuming that the proportion of oil the wells produce is equal to the ratio of injection patterns in Group 2 compared to Group 1 (i.e., 27.41%). Twelve-Year Projects Capital Expenses: Total capital expenses (CAPEX) are similar across all four simulations for Group 1 (Figure 208). This is because the costs that are consistent across all these simulations (e.g., drilling and completion [D&C] costs and new wells characterization, production well equipment costs, and injection well equipment costs) account for 80% or more of the costs for each simulation, meaning that capital costs that vary across the simulations (CO2 recycling plant and CO2 distribution network) are less important for Group 1. Capital costs were more variable for the simulations for Group 2. Capital expenses in these simulations varied by more than two times, from $53.8 million to $135.5 million. CO2 Costs and OPEX: CO2 costs, which are considered operational expenses, vary by 6.5 times for simulations with Group 1 (Table 41). CO2-related expenses were dominated by the purchase of CO2, which outstrips CO2 recycling expenses by eight to more than 20 times. CO2 expenses for Group 2 (Table 41) vary by the same amounts; however, the overall CO2 expenses are more than three times lower for these scenarios compared to the simulations for Group 1. Total non-CO2 OPEX vary by more than 1.5 times for simulations within Groups 1 and simulations within Group 2. Well operations and maintenance (O&M) expenses account for around half to more than three-quarters of the non-CO2 OPEX in Group 1, depending on the simulation, and two-fifths to nearly two-thirds of the non-CO2 OPEX in Group 2, depending on the simulation. G&A costs are 16.7% of all non-CO2 OPEX costs in all simulations. The remainder of the non-CO2 OPEX is spent on Water Disposal (4 to 23% for simulations of Group 1 and 3 to 20% for simulations of Group 2), well plugging and abandonment and site restoration (11 to 16% for simulations of Group 1 and 10 to 14% for simulations of Group 2), and liquid lifting (less than 10% for all simulations). Battelle | May 2026 216 Table 40. Economics of a 12-year project in Albion-Scipio Oilfield for Group 1 (newly drilled and existing wells, top) and Group 2 (existing wells only, bottom). All dollars are 2026$. Total Capital Expenses (Real$) Total CO2 Costs (purchasing and recycling) (Real$) New Exist. Well Prod. Inj. Annual CO2 Max. CO2 Simulation New Well Well CO2 CO2 CAPEX Well Rework Well CO2 CO2 Well Charact- Surf. Recyc. Distri- Demand Recycling (total) Testing/ – Prod. Surf. Purchase Recycling D&C erizaton Equip. Plant bution (purchased) (MMcf/ ($mil) Rehab. to Inj. Equip. ($mil) ($mil) ($mil) ($mil) ($mil) ($mil) ($mil) (tonne) day) ($mil) ($mil) ($mil) Group 1 Avg: 486,417 Mod RF, 1,354 71.1 9.0 69.980 465.3 20.0 Max: long to 567,187 peak Avg: 182,493 1,308 26.7 7.3 26.255 174.6 7.5 Low RF Max: 212,795 880.2 48.9 6.7 1.8 227.1 109.4 Avg: 1,180,469 1,589 305.6 9.0 300.703 1,129 146.4 High RF Max: 1,853,648 Avg: 442,791 1,398 114.6 9.0 112.793 423.5 54.9 Mod RF, Max: short to 695,298 peak Group 2 Avg: 133,327 Mod RF, 67.7 19.5 5.5 19.182 127.5 4.1 Max: long to 567,187 peak Avg: 50,021 53.8 7.3 3.7 7.196 47.8 2.1 Max: 58,327 Low RF Avg: - - 6.7 1.8 4.2 30.0 323,566 135.5 83.8 9.0 82.423 309.5 40.1 Max: 508,085 High RF Avg: 121,369 81.4 31.4 7.3 30.917 116.1 11.3 Max: Mod RF, 190,581 short to peak Total non-CO2 Operating Expenses (Real$) Revenue at two Oil Prices (Real$) Non- Oil ($60/STB) Oil ($80/STB) CO2 Water Well Net Net Royalties, Net Net Simulation O&M Lifting G&A Royalties, OPEX Disp. P&A Revenue Revenue Fees, Revenue Revenue ($mil) ($mil) ($mil) Fees, Taxes (total) ($mil) ($mil) (no 45Q) (w. 45Q) Taxes (no 45Q) (w. 45Q) ($mil) ($mil) ($mil) ($mil) ($mil) ($mil) ($mil) Group 1 Mod RF, 795.5 30.1 61.6 132.6 478.7 1,370 1,791 638.3 1,826 2,247 long to peak Low RF 605.3 11.7 23.9 121.3 185.4 530.5 688.5 247.2 707.3 865.3 High RF 957.6 468.9 108.1 221.0 102.3 180.1 1,717 4,913 5,888 2,290 6,551 7,526 Mod RF, 711.1 40.6 83.1 139.0 645.3 1,846 2,212 860.5 2,462 2,827 short peak Group 2 Mod RF, 219.9 8.3 16.9 36.7 131.2 375.4 490.9 175.0 500.6 616.0 long peak Low RF 201.4 3.2 6.5 33.6 50.8 145.4 188.7 67.8 193.9 237.2 High RF 298.0 129.2 29.6 60.6 28.9 49.7 470.7 1,347 1,614 627.6 1,796 2,063 Mod RF, 229.5 11.1 22.8 38.3 176.9 506.1 606.3 235.9 674.8 775.0 short peak Net Revenue: For Group 1, net revenue ranges from $530.5 million to $4,913 million without 45Q and $688.5 million to $2,462 million with 45Q for across all simulations when the oil price is $60 per STB (Figure 209). If the oil price increases to $80 per STB, net revenue increases to $707.3 million to $6,551 million without 45Q and $865.3 million to $7,526 million with 45Q. For Group 2, net revenue ranges from $145.4 million to $1,347 without 45Q and $188.7 million to $1,614 million with 45Q for across all simulations when the oil price is $60 per STB (Figure 211). If the oil price increases to $80 per STB, net revenue increases to $193.9 million to $1,769 million without 45Q and $237.2 million to $2,063 million with 45Q. Battelle | May 2026 217 $2,000 $300 CAPEX (million real$) Avg. Annual OPEX (million real$) $250 $200 $1,000 $150 $100 $50 $0 Mod RF, Low RF High RF Mod RF, $0 long to short to Simulation Mod RF, Low RF High RF Mod RF, peak peak long to peak short to CO2 Distribution CO2 Plant Simulation peak Inj. Well Surface Equip. Prod. Well Surface Equip. CO2 Purchase CO2 Recycling Well O&M Well Rework - Prod. to Inj. Exist. Well Testing/ Rehab. Lifting Water Disp. Well P&A New Well Charact. New Well D&C G&A Figure 208. CAPEX (left) and average annual OPEX (right) for the 12-year project in the Group 1 simulations of the Albion-Scipio Oifield. Note: Although the well plugging and abandonment is slated to occur at the end of the last year of the project, the cost is averaged over the entire project period. $10,000 $9,000 $8,000 $7,000 (million real$) $6,000 $5,000 $4,000 $3,000 $2,000 $1,000 $0 Mod RF, long to peak Low RF High RF Mod RF, short to peak Figure 209. Total Taxes/Royalties (red) and net revenue without 45Q (light green) and with 45Q (dark green) at oil prices of $60 per STB (patterned fill) and $80 per STB (solid fill) for all Group 1 simulations of a 12-year project in the Albion-Scipio Oilfield. Battelle | May 2026 218 $150 $60 Avg. Annual OPEX (million real$) CAPEX (million real$) $50 $100 $40 $30 $50 $20 $10 $0 $0 Mod RF, Low RF High RF Mod RF, Mod RF, Low RF High RF Mod RF, long to peak short to long to peak short to Simulation Simulation peak peak CO2 Distribution CO2 Plant CO2 Purchase CO2 Recycling Well O&M Inj. Well Surface Equip. Prod. Well Surface Equip. Lifting Water Disp. Well P&A Well Rework - Prod. to Inj. Exist. Well Testing/ Rehab. G&A Figure 210. CAPEX (left) and average annual OPEX (right) for the 12-year project in the Group 2 simulations of the Albion-Scipio Oifield. Note: Although the well plugging and abandonment is slated to occur at the end of the last year of the project, the cost is averaged over the entire project period. $2,500 $2,000 (million real$) $1,500 $1,000 $500 $0 Mod RF, long to peak Low RF High RF Mod RF, short to peak Figure 211. Total Taxes/Royalties (red) and net revenue without 45Q (light green) and with 45Q (dark green) at oil prices of $60 per STB (patterned fill) and $80 per STB (solid fill) for all Group 2 simulations of a 12-year project in the Albion-Scipio Oilfield. Twenty-Year Projects Capital Expenses: CAPEX are similar across all four simulations for the Group 1 20-year project. The CAPEX for the 20-year project is also similar to the CAPEX the 12-year project Battelle | May 2026 219 (Figure 212). The difference is limited to the CAPEX for the CO2 plant, which are higher for the 20-year project compared to the 12-year project. This is because the maximum CO2 recycle rate is higher in years 13 through 20 compared to years one through 12, requiring a higher capacity recycling plant. This increases the CAPEX for the project by 4% or less. CO2 Costs and OPEX: CO2 costs, which are considered operational expenses, vary by 6.5 times for simulations with Group 1 (Figure 212). Like the 12-year project, CO2-related expenses for the 20-year project were dominated by the purchase of CO2, which outstrips CO2 recycling expenses by eight to more than 20 times. CO2 expenses for Group 2 vary by the same amounts (Figure 213); however, the overall CO2 expenses are more than three times lower for these scenarios compared to the simulations for Group 1. The total costs for CO2 for the 20-year project are higher than the 12-year project because of the additional CO2 either purchased or recycled in years 13 to 20 of the project. Total non-CO2 OPEX vary by more than 1.4 times for simulations within Group 1 and for simulations within Group 2. Well O&M expenses account for around half to 70% of the non-CO2 OPEX in Group 1, depending on the simulation, and half to more than two-thirds of the non-CO2 OPEX in Group 2, depending on the simulation. G&A costs are 16.7% of all non-CO2 OPEX costs in all simulations. The remainder of the non-CO2 OPEX is spent on Water Disposal (3 to 17% for simulations of Group 1 and 1% to 17% for simulations of Group 2), well plugging and abandonment and site restoration (11% to 16% for simulations of Group 1 and 10% to 14% for simulations of Group 2), and liquid lifting (less than 10% for all simulations). Net Revenue: For Group 1, net revenue ranges from $707.5 million to $5,771 without 45Q and $865.5 million to $6,746 million with 45Q for across all simulations when the oil price is $60 per STB (Figure 214). If the oil price increases to $80 per STB, net revenue increases to $943.3 million to $7,695 million without 45Q and $1,101 million to $8,670 million with 45Q. For Group 2, net revenue ranges from $193.9 million to $1,582 without 45Q and $237.2 million to $1,849 million with 45Q for across all simulations when the oil price is $60 per STB (Figure 215). If the oil price increases to $80 per STB, net revenue increases to $258.6 million to $2,109 million without 45Q and $301.9 million to $2,376 million with 45Q. Battelle | May 2026 220 Table 41. Economics of a 20-year project in Albion-Scipio Oilfield for Group 1 (newly drilled and existing wells) and Group 2 (existing wells only). All dollars are 2026$. Total CO2 Costs (purchasing and recycling) Total Capital Expenses (Real$) (Real$) New Exist. Well Prod. Inj. New Well Well CO2 Annual CO2 Max. CO2 CO2 CAPEX Well Rework Well CO2 CO2 Well Charact- Surf. Distri- Demand Recycling Recy- (total) Simulation Testing/ – Prod. Surf. Plant Purchase D&C erizaton Equip. bution (purchase) (MMcf/ cling ($mil) Rehab. to Inj. Equip. ($mil) ($mil) ($mil) ($mil) ($mil) ($mil) ($mil) ($mil) ($mil) (tonnes) day) ($mil) Group 1 Avg: 352,260 Mod RF, 1,389 105.4 9.0 103.707 561.6 55.3 Max: long to 567,187 peak Avg: 212,795 1,308 26.7 7.3 38.908 210.7 27.7 Low RF Max: 132,160 880.2 48.9 6.7 1.8 227.1 109.4 Avg: 818,149 1,651 368.0 9.0 362.042 1,304.3 259.0 Max: 1,853,648 High RF Avg: 306,886 1,421 138.0 9.0 135.801 489.2 97.2 Max: 695,298 Mod RF, Group 2 short to Avg: 96,554 peak 77.1 28.9 5.5 Max: 28.426 153.9 15.2 115,446 Avg: 36,225 74.7 9.3 Mod RF, 53.8 10.8 3.7 10.665 Max: 58,327 long to - - 6.7 1.8 4.2 30.0 Avg: peak 224,255 135.5 83.8 9.0 82.423 357.5 71.0 Low RF Max: 508,085 Avg: 84,117 87.8 37.8 7.3 Max: 37.223 134.1 26.6 High RF 190,581 Mod RF, short to peak Total Operating Expenses (Real$) Revenue at two Oil Prices (Real$) Non- Oil Price of $60 per STB Oil Price of $80 per STB CO2 Water Well Royalties, Net Net Royalties, Net Net Simulation O&M Lifting G&A OPEX Disp. P&A Fees, Revenue Revenue Fees, Revenue Revenue ($mil) ($mil) ($mil) (total) ($mil) ($mil) Taxes (no 45Q) (w. 45Q) Taxes (no 45Q) (w. 45Q) ($mil) ($mil) ($mil) ($mil) ($mil) ($mil) ($mil) Group 1 Mod RF, long to 1,209 40.5 82.8 201.4 643.6 1,841 2,263 858.2 2,455 2,876 peak Low RF 1,117 15.6 31.8 186.2 247.3 707.5 865.5 329.7 943.3 1,101 High RF 1,524 781.5 127.0 259.6 102.3 254.1 2,017 5,771 6,746 2,690 7,695 8,670 Mod RF, short to 1,235 47.7 97.5 205.8 757.8 2,168 2,534 1,010 2,891 3,256 peak Group 2 Mod RF, long to 333.7 11.1 22.7 55.6 176.4 504.7 620.2 235.2 673.0 788.4 peak Low RF 308.7 4.3 8.7 51.4 67.8 193.9 237.2 90.4 258.6 301.9 High RF 420.3 215.4 34.8 71.2 28.9 70.0 552.9 1,582 1,849 737.2 2,109 2,376 Mod RF, short to 340.0 13.1 26.7 56.7 207.7 594.3 694.5 277.0 792.4 892.6 peak Battelle | May 2026 221 $2,000 $300 CAPEX (million real$) Avg. Annual OPEX (million real$) $250 $200 $1,000 $150 $100 $50 $0 Mod RF, Low RF High RF Mod RF, $0 long to short to Simulation Mod RF, Low RF High RF Mod RF, peak peak long to peak short to CO2 Distribution CO2 Plant Simulation peak Inj. Well Surface Equip. Prod. Well Surface Equip. CO2 Purchase CO2 Recycling Well O&M Well Rework - Prod. to Inj. Exist. Well Testing/ Rehab. Lifting Water Disp. Well P&A New Well Charact. New Well D&C G&A Figure 212. CAPEX (left) and average annual OPEX (right) for the 20-year project in the Group 1 simulations of the Albion-Scipio Oifield. Note: Although the well plugging and abandonment is slated to occur at the end of the last year of the project, the cost is averaged over the entire project period. $150 $80 Avg. Annual OPEX (million real$) $70 CAPEX (million real$) $60 $100 $50 $40 $30 $50 $20 $10 $0 $0 Mod RF, Low RF High RF Mod RF, Mod RF, Low RF High RF Mod RF, long to peak short to long to peak short to Simulation Simulation peak peak CO2 Distribution CO2 Plant CO2 Purchase CO2 Recycling Well O&M Inj. Well Surface Equip. Prod. Well Surface Equip. Lifting Water Disp. Well P&A Well Rework - Prod. to Inj. Exist. Well Testing/ Rehab. G&A Figure 213. CAPEX (left) and average annual OPEX (right) for the 20-year project in the Group 2 simulations of the Albion-Scipio Oifield. Note: Although the well plugging and abandonment is slated to occur at the end of the last year of the project, the cost is averaged over the entire project period. Battelle | May 2026 222 $10,000 $9,000 $8,000 $7,000 (million real$) $6,000 $5,000 $4,000 $3,000 $2,000 $1,000 $0 Mod RF, long to peak Low RF High RF Mod RF, short to peak Figure 214. Total Taxes/Royalties (red) and net revenue without 45Q (light green) and with 45Q (dark green) at oil prices of $60 per STB (patterned fill) and $80 per STB (solid fill) for all Group 1 simulations of a 20-year project in the Albion-Scipio Oilfield. $2,500 $2,000 (million real$) $1,500 $1,000 $500 $0 Mod RF, long to peak Low RF High RF Mod RF, short to peak Figure 215. Total Taxes/Royalties (red) and net revenue without 45Q (light green) and with 45Q (dark green) at oil prices of $60 per STB (patterned fill) and $80 per STB (solid fill) for all Group 2 simulations of a 20-year project in the Albion-Scipio Oilfield. Napoleon Economic model results for the Napoleon Oilfield are presented for a 12-year project and a 20- year project in Table 42 and Table 43, respectively. This section discusses the results of the 12- Battelle | May 2026 223 year projects followed a discussion of the results of the 20-year projects. Because the field has many of the original producing wells, the only scenario considered was one in which existing infrastructure is used to produce the whole field. There are 55 existing wells, yielding 11 injection patterns of four producers to every injection well. Twelve-Year Projects Capital Expenses: Capital costs were more variable for the simulations for the Napoleon field compared to the CAPEX for Group 1 of the Albion-Scipio Oilfield. CAPEX in the Napoleon Oilfield varied by more than two times, from $18.1 million to $36.9 million (Figure 216). CAPEX in the Napoleon Oilfield varied by more than two times, from $18.1 million to $36.9 million. Purchasing injection well surface equipment is the largest capital cost for three of the four simulations ($8.9 million for all simulations). For the High RF simulation, however, the CO2 recycling plant was the largest capital expense ($18.9 million, or more than half of the $36.9 million CAPEX costs). This is because this simulation requires a plant with the capacity to handle more than 18 MMscf of CO2 per day compared to a capacity of less than 10 MMscf of CO2 per day for the other simulations. CO2 Costs and OPEX: Carbon dioxide costs, which are considered operational expenses, vary by 8.4 times for simulations with the Napoleon Oilfield (Figure 216). CO2-related expenses were dominated by the purchase of CO2, which outstrips CO2 recycling expenses by 5.6 to more than 20 times. Carbon dioxide recycling expenses vary by nearly 30 times between the simulations. Total non-CO2 OPEX vary by around 1.4 times for simulations of the Napoleon Oilfield. Well O&M expenses account for around 40% to 60% of non-CO2 OPEX, depending on the simulation. G&A costs are 16.7% of non-CO2 OPEX costs in all simulations. The remainder of the non-CO2 OPEX is spent on water disposal (6% to 21%), well plugging and abandonment and site restoration (10% to 14%), and liquid lifting (11% or less for all simulations). For most simulations, well O&M is the largest annual operational expense (Figure 216). The High RF simulation is the exception. The amount of CO2 that must be purchased makes this cost slightly larger in this compared to the well O&M costs. Net Revenue: Net revenue ranges from $89 million to $440 million without 45Q and $102million to $522 million with 45Q for across all simulations when the oil price is $60 per STB (Figure 217). If the oil price increases to $80 per STB, net revenue increases to $118 million to $587 million without 45Q and $132 million to $669 million with 45Q. Battelle | May 2026 224 Table 42. Economics of a 12-year project in Napoleon Oilfield. Total Capital Expenses Total CO2 Costs (purchasing and recycling) Prod. Exist. Well Inj. Max. Well CO2 Annual CO2 CAPEX Well Rework Well CO2 CO2 CO2 CO2 Sim. Surf. Distri- Demand (total) Testing/ – Inj. to Surf. Plant Recyclin Purchase Recyclin Equip bution (purchase) ($mil) Rehab. Prod. Equip. ($mil) g (MMcf/ ($mil) g ($mil) . ($mil) (tonne) ($mil) ($mil) ($mil) day) ($mil) Mod RF, Avg: 34,261 20.2 3.8 3.7 3.738 18.7 1.0 long to peak Max: 38,394 Avg: 15,490 Low RF 18.1 1.7 3.7 1.690 8.5 0.4 Max: 17,358 2.0 0.5 1.3 8.9 Avg: 96,665 High RF 36.9 18.7 5.5 18.387 63.4 11.3 Max: 135,423 Mod RF, Avg: 39,442 short to 24.0 7.6 3.7 7.502 21.6 3.5 Max: 55,257 peak Operating Expenses Revenue at two Oil Prices Non- Oil Price of $60 per STB Oil Price of $80 per STB CO2 Water Well Royalties, Net Net Royalties, Net Net Simulation O&M Lifting G&A OPEX Disp. P&A Fees, Revenue Revenue Fees, Revenue Revenue ($mil) ($mil) ($mil) (total) ($mil) ($mil) Taxes (no 45Q) (w. 45Q) Taxes (no 45Q) (w. 45Q) ($mil) ($mil) ($mil) ($mil) ($mil) ($mil) ($mil) Mod RF, long to 69.7 3.5 7.2 11.6 56 159 189 74 212 242 peak Low RF 64.1 1.9 4.0 10.7 31 89 102 41 118 132 38.2 9.3 High RF 92.3 9.7 19.8 15.4 154 440 522 205 587 669 Mod RF, short to 71.0 4.1 8.4 11.8 65 186 219 87 248 281 peak $40 $16 Avg. Annual OPEX (million real$) $14 CAPEX (million real$) $30 $12 $10 $20 $8 $6 $10 $4 $2 $0 $0 Mod RF, Low RF High RF Mod RF, Mod RF, Low RF High RF Mod RF, long to peak short to long to peak short to Simulation Simulation peak peak CO2 Distribution CO2 Plant CO2 Purchase CO2 Recycling Well O&M Inj. Well Surface Equip. Prod. Well Surface Equip. Lifting Water Disp. Well P&A Well Rework - Prod. to Inj. Exist. Well Testing/ Rehab. G&A Figure 216. CAPEX (left) and average annual OPEX (right) for the 12-year project in the Napoleon Oifield. Note: Although the well plugging and abandonment is slated to occur at the end of the last year of the project, the cost is averaged over the entire project period. Battelle | May 2026 225 $800 $700 $600 (million real$) $500 $400 $300 $200 $100 $0 Mod RF, long to peak Low RF High RF Mod RF, short to peak Figure 217. Total Taxes/Royalties (red) and net revenue without 45Q (light green) and with 45Q (dark green) at oil prices of $60 per STB (patterned fill) and $80 per STB (solid fill) for all simulations of a 12- year project in the Napoleon Oilfield. Twenty-Year Projects Capital Expenses: Capital costs were more variable for the simulations for the Napoleon field compared to the CAPEX for Group 1 of the Albion-Scipio Oilfield (Figure 218). CAPEX in the Napoleon Oilfield varied by more than two times, from $18.1 million to $36.9 million. Purchasing injection well surface equipment is the largest capital cost for three of the four simulations ($8.9 million for all simulations). For the High RF simulation, however, the CO2 recycling plant was the largest capital expense ($24.4 million, or more than half of the $42.6 million CAPEX costs). This is because this simulation requires a plant with the capacity to handle nearly 24 MMscf of CO2 per day compared to a capacity of less than 10 MMscf of CO2 per day for the other simulations. CO2 Costs and OPEX: CO2 costs, which are considered operational expenses, vary by 8.4 times for simulations with the Napoleon Oilfield. CO2-related expenses were dominated by the purchase of CO2, which outstrips CO2 recycling expenses by 5.6 to more than 20 times. CO2 recycling expenses vary by nearly 30 times between the simulations. Total non-CO2 OPEX vary by around 1.4 times for simulations of the Napoleon Oilfield. Well O&M expenses account for around 40% to around 60% of non-CO2 OPEX, depending on the simulation. G&A costs are 16.7% of all non-CO2 OPEX costs in all simulations. The remainder of the non-CO2 OPEX is spent on Water Disposal (6% to 21%), well plugging and abandonment and site restoration (10% to 14%), and liquid lifting (11% or less for all simulations). For most simulations, well O&M is the largest annual operational expense (Figure 218). The High RF simulation is the exception. The amount of CO2 that must be purchased makes this cost slightly larger in this compared to the well O&M costs. Net Revenue: Net revenue ranges from $89 million to $440 million without 45Q and $102 million to $522 million with 45Q for across all simulations when the oil price is $60 per STB (Figure 219). If the oil price increases to $80 per STB, net revenue increases to $118 million to $587 million without 45Q and $132 million to $669 million with 45Q. Battelle | May 2026 226 Table 43. Economics of a 20-year project in Napoleon Oilfield. Total Capital Expenses (Real$) Total CO2 Costs (purchasing and recycling) (Real$) Exist. Well Prod. Inj. Annual CO2 CAPEX Well Rework Well Well CO2 CO2 Max. CO2 CO2 CO2 Simulation Distri- (total) Testing/ – Inj. to Surf. Surf. Plant Demand Recycling Purchase Recycling bution ($mil) Rehab. Prod. Equip. Equip. ($mil) (purchase) (MMcf/day) ($mil) ($mil) ($mil) ($mil) ($mil) ($mil) ($mil) (tonne) Mod RF, Avg: 26,543 long to 22.7 6.2 3.7 Max: 6.140 24.2 3.1 peak 38,394 Avg: 12,000 Low RF 19.2 2.8 3.7 Max: 2.776 10.9 1.9 17,358 2.0 0.5 1.3 8.9 Avg: 71,975 High RF 42.6 24.4 5.5 Max: 23.993 65.6 21.3 135,423 Mod RF, Avg: 29,368 short to 26.4 9.9 3.7 Max: 9.790 26.8 6.5 peak 55,257 Total Operations Expenses (Real$) Revenue at two Oil Prices (Real$) Non Oil ($60/STB) Oil ($80/STB) -CO2 Net Net OPE Wate Well Revenu Net Revenu Net O&M Liftin G&A Royalties Royalties Simulation X r P&A e (no Revenu e (no Revenu ($mil g ($mil , Fees, , Fees, (tota Disp. ($mil 45Q) e (w. 45Q) e (w. ) ($mil) ) Taxes Taxes l) ($mil) ) ($mil) 45Q) ($mil) 45Q) ($mil) ($mil) ($mi ($mil) ($mil) l) Mod RF, 104. 4.6 9.3 17.4 73 208 238 97 277 307 long to peak 2 Low RF 96.4 2.4 5.0 16.1 39 111 124 52 147 161 132. High RF 63.7 12.2 24.9 9.3 22.0 194 554 636 258 739 821 1 Mod RF, 106. short to 5.1 10.5 17.7 81 232 266 108 310 343 1 peak $50 $20 Avg. Annual OPEX (million real$) $18 $40 $16 CAPEX (million real$) $14 $30 $12 $10 $20 $8 $6 $10 $4 $2 $0 $0 Mod RF, Low RF High RF Mod RF, Mod RF, Low RF High RF Mod RF, long to peak short to long to peak short to Simulation Simulation peak peak CO2 Distribution CO2 Plant CO2 Purchase CO2 Recycling Well O&M Inj. Well Surface Equip. Prod. Well Surface Equip. Lifting Water Disp. Well P&A Well Rework - Prod. to Inj. Exist. Well Testing/ Rehab. G&A Figure 218. CAPEX (left) and average annual OPEX (right) for the 20-year project in the Napoleon Oifield. Note: Although the well plugging and abandonment is slated to occur at the end of the last year of the project, the cost is averaged over the entire project period. Battelle | May 2026 227 $900 $800 $700 $600 (million real$) $500 $400 $300 $200 $100 $0 Mod RF, long to peak Low RF High RF Mod RF, short to peak Figure 219. Total Taxes/Royalties (red) and net revenue without 45Q (light green) and with 45Q (dark green) at oil prices of $60 per STB (patterned fill) and $80 per STB (solid fill) for all simulations of a 20- year project in the Napoleon Oilfield. Discussion Cash Flow 12-Year Project in the Albion-Scipio: The number of years with positive cash flow, the year of the project where it becomes profitable, IRR, NPV, and CO2 break-even price are presented in Table 44 for all simulations in both simulation groups. Net cumulative cash flows are shown for all simulations when oil is $60 per STB and $80 per STB in Figure 220 and Figure 221, respectively. For Group 1, only the most optimistic case is profitable in any of the simulations. When the oil price is $60 per STB, 45Q tax credits are required to make the simulation profitable (NPV of $550 million). When the oil price is increased to $80 per STB, the project is profitable with or without 45Q, although the inclusion of the tax credits does improve the economics (NPV of $872 million without 45Q and $1,479 million with 45Q). These simulations all become profitable in year three or four of operations. The CO2 break-even cost exceeds the price of CO2 available from dilute sources (e.g., $70 per tonne) for each of these simulations. For the simulation where oil is $60 per STB and the 45Q tax credits are not claimed, the CO2 break- even price drops to $64.32 – a cost of capture that is easily achievable for purer sources, such as ethanol plants and nitrogen plants. For all other simulations, the CO2 break-even price is negative, meaning an entity would have to pay the operator for disposal of CO2, oil would need to be more than $80 per STB, and/or another source of income would need to be secured for the project to be profitable. Using existing infrastructure greatly improves the economics for the Albion-Scipio Field. Both the optimistic case and the moderate recovery factor with a short period to peak recovery were profitable with or without 45Q for scenarios where oil was $60 or $80 per STB, with NPV ranging from $19 million to $705 million and CO2 break-even prices ranging from $88.99 to $330.07 per tonne. The moderate recovery scenario with a long period to peak recovery was also profitable if oil was $80 per STB and 45Q tax credits were claimed. The CO2 break-even Battelle | May 2026 228 price for this simulation was $107.65 per tonne. These simulations all become profitable between years two to five of operations. Other scenarios for this simulation were profitable at a lower price of CO2. The CO2 break-even price is $43.05 when oil is $80 per STB but 45Q credits are not claimed and $55.00 when oil is $60 per STB and 45Q credits are claimed, both achievable costs for CO2 for relatively pure sources. The remaining scenario for this simulation (oil is $60 per STB and 45Q tax credits are not claimed) are not profitable and the break-even price of CO2 is negative, meaning additional income sources would be required. The pessimistic case is not profitable in any of the scenarios and the CO2 break-even price is negative. Table 44. Cash Flow, IRR, NPV, and Break-Even Price of a 12-year project in Albion-Scipio Oilfield for Group 1 (newly drilled and existing wells) and Group 2 (existing wells only). Years with Year Project is IRR (No 45Q) CO2 Break-Even Oil Positive Cash NPV ($mil) Profitable (%) (nominal) Price ($/tonne)1 Sim. Price Flow ($/bbl) No With No With No With No With With No 45Q 45Q 45Q 45Q 45Q 45Q 45Q 45Q 45Q 45Q Group 1 Mod RF, long to $60 5-12 5-12 NA NA -20.5 -8.3 -1405 -1161 Negative Negative peak $80 5-12 5-12 NA NA -6.1 -0.9 -1206 -963 Negative Negative $60 6-8 6-9 NA NA - - -1525 -1434 Negative Negative Low RF $80 6-10 6-11 NA NA - - -1447 -1356 Negative Negative $60 2-12 2-12 6 4 14.1 23.3 -56 550 64.32 125.68 High RF $80 2-12 2-12 4 3 27.0 35.0 872 1479 158.27 219.63 Mod RF, short to $60 2-11 2-12 NA NA - -7.5 -1029 -801 Negative Negative peak $80 2-12 2-12 10 8 1.5 6.6 -645 -418 Negative Negative Group 2 Mod RF, long to $60 5-12 5-12 5 5 -3.1 12.0 -82 -15 Negative 55.00 peak $80 5-12 5-12 5 5 10.7 21.0 -28 39 43.05 107.65 $60 6-8 6-9 6 6 - - -114 -89 Negative Negative Low RF $80 6-10 6-11 6 5 - - -93 -68 Negative Negative $60 2-12 2-12 3 2 50.6 71.6 284 450 174.76 236.12 High RF $80 2-12 2-12 2 2 73.5 93.4 538 705 268.71 330.07 Mod RF, short to $60 2-11 2-12 4 3 21.4 38.4 19 82 88.99 150.35 peak $80 2-12 2-12 3 2 43.0 57.0 115 177 183.20 244.56 Notes: 1. Text in italics indicates positive CO2 break-even price. Bold text indicates CO2 break even price above $70 per tonne, which is the price for which CO2 would be available from nearby sources. Battelle | May 2026 229 Figure 220. Net cumulative discounted cashflow for Group 1 (top) and Group 2 (bottom) Simulations without 45Q tax credits (left) and with 45Q tax credits (right) for 12-Year projects with oil price of $60 per STB. Simulations include the Low RF (black), Mod RF/long to peak (red), Mod RF/short to peak (orange), and High RF (blue). Battelle | May 2026 230 Figure 221. Net discounted cumulative cashflow for Group 1 (top) and Group 2 (bottom) Simulations without 45Q tax credits (left) and with 45Q tax credits (right) for 12-Year projects with oil price of $80 per STB. Simulations include the Low RF (black), Mod RF/long to peak (red), Mod RF/short to peak (orange), and High RF (blue). 20-Year Project in the Albion-Scipio: The number of years with positive cash flow, the year of the project where it becomes profitable, IRR, NPV, and CO2 break-even Price are presented in Table 45 for all simulations in both simulation groups. Net cumulative cash flows are shown for all simulations when oil is $60 per STB and $80 per STB in Figure 222 and Figure 223, respectively. For Group 1, only the most optimistic case is profitable in any of the simulations. When the oil price is $60 per STB, 45Q tax credits are required to make the simulation profitable (NPV of $844 million). When the oil price is increased to $80 per STB, the project is profitable with or without 45Q, although the inclusion of the tax credits does improve the economics (NPV of $522 million without 45Q and $1,450 million with 45Q). These simulations all become profitable in year three or four of operations. The CO2 break-even cost exceeds the price of CO2 available from dilute sources (e.g., $70 per tonne) for each of these simulations. For the simulation where oil is $60 per STB and the 45Q tax credits are not claimed, the CO2 break- even price drops to $61.44 – a cost of capture that is easily achievable for purer sources, such Battelle | May 2026 231 as ethanol plants and nitrogen plants. For all other simulations, the CO2 break-even price is negative, meaning an entity would have to pay the operator for disposal of CO2, oil would need to be more than $80 per STB, and/or another source of income would need to be secured for the project to be profitable. Table 45. Cash Flow, IRR, NPV, and Break-Even Price of a 20-year project in Albion-Scipio Oilfield for Group 1 (newly drilled and existing wells) and Group 2 (existing wells only). Years with Year Project IRR (No 45Q) CO2 Break-Even Oil Positive Cash NPV ($mil) Breaks Even (%) (Nominal) Price ($/tonne)1 Simulation Price Flow ($/bbl) No With No With No With No With With No 45Q 45Q 45Q 45Q 45Q 45Q 45Q 45Q 45Q 45Q Group 1 Mod RF, long to $60 5-14 5-14 NA NA - - -1428 -1185 Negative Negative peak $80 5-17 5-17 NA NA - -1.0 -1202 -959 Negative Negative $60 6-8 6-9 NA NA - - -1504 -1413 Negative Negative Low RF $80 6-10 5-11 NA NA - -19.9 -1426 -1335 Negative Negative $60 2-17 2-17 6 4 13.8 22.4 -85 522 61.44 122.80 High RF $80 2-19 2-19 4 3 26.0 33.7 844 1450 155.39 216.75 Mod RF, short to $60 2-11 2-12 NA NA -10.5 -2.9 -1018 -791 Negative Negative peak $80 2-14 2-14 NA 9 1.2 6.2 -669 -441 Negative Negative Group 2 Mod RF, long to $60 5-14 5-14 NA 8 - 10.5 -89 -22 Negative 50.15 peak $80 5-17 5-17 8 6 11.3 20.5 -27 40 45.77 106.21 $60 6-8 6-9 NA NA - - -1504 -1413 Negative Negative Low RF $80 6-10 5-11 NA NA - -19.9 -1426 -1335 Negative Negative $60 2-17 2-17 3 2 46.2 65.3 279 445 168.21 226.77 High RF $80 2-19 2-19 2 2 67.7 85.7 546 712 262.25 320.81 Mod RF, short to $60 2-11 2-12 4 3 7.9 35.2 <0.1 62 70.06 128.62 peak $80 2-14 2-14 3 3 40.2 53.5 100 163 164.36 222.92 Notes: 1. Text in italics indicates positive CO2 break-even price. Bold text indicates CO2 break even price above $70 per tonne, which is the price for which CO2 would be available from nearby sources. Like the 12-year project, using existing infrastructure greatly improves the economics for the Albion-Scipio Field. Both the optimistic case and the moderate recovery factor with a short period to peak recovery were profitable with or without 45Q for scenarios where oil was $60 or $80 per STB, with NPV ranging from less than $100,000 to $712 million and CO2 break-even prices ranging from $70.06 to $320.81 per tonne. The moderate recovery scenario with a long period to peak recovery was also profitable if oil was $80 per STB and 45Q tax credits were claimed. The CO2 break-even price for this simulation was $106.21 per tonne. These simulations all become profitable between years two to six of operations but are unprofitable after years 12 to 17. Other scenarios for this simulation were profitable at a lower price of CO2. The CO2 break-even price is $45.77 per tonne when oil is $80 per STB but 45Q credits are not claimed and $50.15 per tonne when oil is $60 per STB and 45Q credits are claimed, both achievable costs for CO2 for relatively pure sources. The remaining scenario for this simulation (oil is $60 per STB and 45Q tax credits are not claimed) is not profitable and the break-even price of CO2 is negative, meaning additional income sources would be required. The pessimistic case is not profitable in any of the scenarios and the CO2 break-even price is negative. Battelle | May 2026 232 Figure 222. Net discounted cumulative cashflow for Group 1 (top) and Group 2 (bottom) Simulations without 45Q tax credits (left) and with 45Q tax credits (right) for 12-Year projects with oil price of $60 per STB. Simulations include the Low RF (black), Mod RF/long to peak (red), Mod RF/short to peak (orange), and High RF (blue). Battelle | May 2026 233 Figure 223. Net discounted cumulative cashflow for Group 1 (top) and Group 2 (bottom) Simulations without 45Q tax credits (left) and with 45Q tax credits (right) for 12-Year projects with oil price of $80 per STB. Simulations include the Low RF (black), Mod RF/long to peak (red), Mod RF/short to peak (orange), and High RF (blue). 12-Year Project in the Napoleon: The number of years with positive cash flow, the year of the project where it becomes profitable, IRR, NPV, and CO2 break-even Price are presented in Table 46 for all simulations in both simulation groups. Net cumulative cash flows are shown for all simulations when oil is $60 per STB and $80 per STB in Figure 224 and Figure 225, respectively. Nearly all cases are profitable in the simulations regardless of whether the oil price is $60 per STB or $80 per STB or whether or not 45Q tax credits are claimed. The NPV for the profitable simulations ranges from $5 million to $263 million. IRR (nominal) for profitable simulations range from 35.4% to 105.5% without 45Q and 26.2% to 128.5% with 45Q. All profitable simulations have a positive cash flow and break even in year one to four. CO2 break- even prices for profitable simulations range from $146.57 per tonne without 45Q and $82.80 to Battelle | May 2026 234 $377.62 per tonne with 45Q. The only simulation with an NPV that is negative (-$3 million) is the pessimistic case with a $60 per STB oil price and no 45Q tax credits. This simulation has a CO2 break-even price of $17.58 per tonne, which is only likely for sources with high purity of CO2. Years one to ten have positive cash flows and the project breaks even in year five, suggesting that a 10-year project would be viable even under the most pessimistic simulation. Table 46. Cash Flow, IRR, NPV, and Break-Even Price of a 12-year project in Napoleon Oilfield using existing wells only. Years with Year Project IRR (No 45Q) CO2 Break-Even Oil Positive Breaks NPV ($mil) Cash Flow Even (%) (nominal) Price ($/tonne)1 Simulation Price ($/bbl) No With No With No With No With No With 45Q 45Q 45Q 45Q 45Q 45Q 45Q 45Q 45Q 45Q Mod RF, $60 1-12 1-12 4 3 35.4 50.1 21 38 119.07 184.29 long to peak $80 1-12 1-12 3 2 57.1 70.8 49 66 228.64 293.68 $60 1-10 1-12 5 4 - 26.2 -3 5 17.58 82.80 Low RF $80 1-12 1-12 3 2 41.6 51.4 15 22 164.04 229.25 $60 1-12 1-12 2 2 76.2 99.6 132 181 209.03 271.85 High RF $80 1-12 1-12 2 2 105.5 128.5 214 263 314.80 377.62 Mod RF, $60 1-12 1-12 2 2 48.8 66.1 34 54 146.57 209.39 short to $80 1-12 1-12 2 1 75.8 92.2 69 89 258.19 321.01 peak Notes: 1. Text in italics indicates positive CO2 break-even price. Bold text indicates CO2 break even price above $70 per tonne, which is the price for which CO2 would be available from nearby sources. Figure 224. Net discounted cumulative cashflow for the Napoleon Field Simulations without 45Q tax credits (left) and with 45Q tax credits (right) for 12-Year projects with oil price of $60 per STB. Simulations include the Low RF (black), Mod RF/long to peak (red), Mod RF/short to peak (orange), and High RF (blue). Battelle | May 2026 235 Figure 225. Net discounted cumulative cashflow for the Napoleon Field Simulations without 45Q tax credits (left) and with 45Q tax credits (right) for 12-Year projects with oil price of $80 per STB. Simulations include the Low RF (black), Mod RF/long to peak (red), Mod RF/short to peak (orange), and High RF (blue). 20-Year Project in the Napoleon: The number of years with positive cash flow, the year of the project where it becomes profitable, IRR, NPV, and CO2 break-even Price are presented in Table 47 for all simulations in both simulation groups. Net cumulative cash flows are shown for all simulations when oil is $60 per STB and $80 per STB in Figure 226 and Figure 227, respectively. Nearly all cases are profitable in the simulations regardless of whether the oil price is $60 per STB or $80 per STB or whether or not 45Q tax credits are claimed. The NPV for the profitable simulations ranges from $5 million to $263 million. IRR (nominal) for profitable simulations range from 35.4% to 105.5% without 45Q and 26.2% to 128.5% with 45Q. All profitable simulations have a positive cash flow and break even in year one to four. CO2 break- even prices for profitable simulations range from $146.57 per tonne without 45Q and $82.80 to $377.62 per tonne with 45Q. The only simulation with an NPV that is negative (-$3 million) is the pessimistic case with a $60 per STB oil price and no 45Q tax credits. This simulation has a CO2 break-even price of $17.58 per tonne, which is only likely for sources with high purity of CO2. Years one to ten have positive cash flows and the project breaks even in year five, suggesting that a 10-year project would be viable even under the most pessimistic simulation. Battelle | May 2026 236 Table 47. Cash Flow, IRR, NPV, and Break-Even Price of a 20-year project in Napoleon Oilfield using existing wells only. Years with Year Project IRR (No 45Q) CO2 Break-Even Oil Positive Cash NPV ($mil) Breaks Even (%) (nominal) Price ($/tonne)1 Simulation Price Flow ($/bbl) No With No With No With No With No With 45Q 45Q 45Q 45Q 45Q 45Q 45Q 45Q 45Q 45Q Mod RF, long $60 1-18 1-18 3 4 36.3 50.4 24 41 123.62 183.10 to peak $80 1-19 1-19 3 2 57.4 71.0 55 72 233.43 292.90 $60 1-10 1-12 5 4 - 5.0 -3 5 15.85 75.33 Low RF $80 1-13 1-13 3 2 39.2 48.2 15 23 159.45 218.92 $60 1-19 1-19 2 2 76.2 99.6 144 193 211.19 269.59 High RF $80 1-19 1-19 2 2 105.5 128.5 233 282 317.24 375.65 Mod RF, $60 1-17 1-17 3 2 44.4 60.3 36 56 131.45 189.86 short to peak $80 1-19 1-19 2 2 69.5 84.6 75 95 257.66 316.07 Notes: 1. Text in italics indicates positive CO2 break-even price. Bold text indicates CO2 break even price above $70 per tonne, which is the price for which CO2 would be available from nearby sources. Figure 226. Net discounted cashflow for the Napoleon Field Simulations without 45Q tax credits (left) and with 45Q tax credits (right) for 20-Year projects with oil price of $60 per STB. Simulations include the Low RF (black), Mod RF/long to peak (red), Mod RF/short to peak (orange), and High RF (blue). Battelle | May 2026 237 Figure 227. Net discounted cumulative cashflow for the Napoleon Field Simulations without 45Q tax credits (left) and with 45Q tax credits (right) for 20-Year projects with oil price of $80 per STB. Simulations include the Low RF (black), Mod RF/long to peak (red), Mod RF/short to peak (orange), and High RF (blue). Spurring Investment in CO2 Capture The scenarios described thus far have assumed that CO2 is available in the quantities needed at a given market price ($70 a tonne for the Albion-Scipio Field, satisfied by a less pure CO2 source, and $40 a tonne for the Napoleon, satisfied by a purer CO2 source). A more realistic scenario in the short term would be for the CO2-EOR operator to provide a guarantee to purchase a specific amount from a CO2 source so that the source is incentivized to build the capture plant. To simulate this, the pessimistic (Low RF) and optimistic (High RF) cases for the 12-year and 20-year projects in the Napoleon Oilfield were used. The altered simulations assume that all conditions are the same except that the amount of CO2 purchased per year is held constant at the maximum annual CO2 purchase modeled (17,358 for the Low RF scenario and 135,423 for the High RF scenario). Three different options were explored for the excess CO2 not injected by in the Napoleon Oifleid: (1) a $8 per tonne fee to simulate either a disposal fee paid to a third party for saline injection or the transfer of the CO2 to another oilfield via pipeline, (2) no cost for the excess CO2 to simulate and offtake agreement of excess CO2 for nearby Class VI or CO2 EOR, and (3) $8 per tonne income from the sale of the CO2 to another operator for saline injection or CO2-EOR. The results of this effort for 12-year and 20-year projects are provided in Table 48 and Table 49, respectively. While the requirement to obtain the maximum amount of CO2 impacts the economics of these projects, particularly for 20-year projects that must pay an $8 per tonne disposal fee, the impact is not enough to make a project unprofitable. The NPV is impacted by $1 million or less for 12-year or 20-year projects, and the IRR is largely unchanged. The CO2 break-even price is impacted, however, because additional CO2 must be purchased, particularly in later years of the project. This is most notable for the high recovery factor simulations, which decrease by more than 20% from the CO2 break-even prices of the original simulations. The economics may be improved by the resale of the excess CO2. The current model assumes that excess CO2 can be sold for a net of $8/tonne, which may be low because the purchase price is originally $40 per tonne. Battelle | May 2026 238 Table 48. Results of 12-year project simulations where the purchase of purchase of 17,358 tonnes per year for all 12 years for the Low RF simulations and 135,423 tonnes per year for all 12 years for the High RF simulations are required. Year Years with CO2 Break- Project IRR (No 45Q) Oil Positive NPV ($mil) Even Price Breaks (%) (nominal) Simulation Price Cash Flow ($/tonne)1 Even ($/bbl) No With No With No With No With No With 45Q 45Q 45Q 45Q 45Q 45Q 45Q 45Q 45Q 45Q $60 1-10 1-12 5 4 - 26.2 -3 5 17.58 82.80 Low RF $80 1-12 1-12 3 2 41.6 51.4 15 22 164.04 229.25 Low RF – $60 1-10 1-11 5 4 - 25.7 -3 5 15.78 76.72 Fee for $80 1-11 1-11 3 2 41.5 51.3 14 22 152.64 213.59 disposal Low RF – No $60 1-10 1-11 5 4 - 25.8 -3 5 16.43 77.38 cost for $80 1-11 1-11 3 2 41.5 51.3 14 22 153.29 214.24 excess Low RF – $60 1-10 1-11 5 4 - 25.9 -3 5 17.08 78.03 Sale of $80 1-11 1-11 3 2 41.6 51.3 22 153.95 214.89 14 excess $60 1-12 1-12 2 2 76.2 99.6 132 181 209.03 271.85 High RF $80 1-12 1-12 2 2 105.5 128.5 214 263 314.80 377.62 High RF – $60 1-11 1-11 2 2 75.4 99.1 122 171 163.58 213.29 Fee for $80 1-12 1-12 2 2 105.1 128.1 204 253 247.27 296.98 disposal High RF – No $60 1-11 1-12 2 2 75.6 99.2 124 173 165.41 215.11 cost for $80 1-12 1-12 2 2 105.2 128.2 206 255 249.10 298.81 excess High RF – $60 1-11 1-12 2 2 75.8 99.4 125 174 167.23 216.94 Sale of $80 1-12 1-12 2 2 105.3 128.3 208 257 250.93 300.64 excess Table 49. Results of 20-year project simulations where the purchase of purchase of 17,358 tonnes per year for all 20 years for the Low RF simulations and 135,423 tonnes per year for all 20 years for the High RF simulations are required. Year Years with CO2 Break- Project IRR (No 45Q) Oil Positive NPV ($mil) Even Price Breaks (%) (nominal) Simulation Price Cash Flow ($/tonne)1 Even ($/bbl) No With No With No With No With No With 45Q 45Q 45Q 45Q 45Q 45Q 45Q 45Q 45Q 45Q Low RF – $60 1-10 1-12 5 4 - 5.0 -3 5 15.85 75.33 Original $80 1-13 1-13 3 2 39.2 48.2 15 23 159.45 218.92 model Low RF – $60 1-10 1-11 5 4 - 6.9 -4 3 11.99 61.96 Fee for $80 1-12 1-12 3 2 42.0 51.6 14 22 132.65 182.62 disposal Low RF – No $60 1-10 1-11 4 4 - 6.5 -4 4 13.32 63.29 cost for $80 1-13 1-13 3 2 42.1 51.6 14 22 133.98 183.95 excess Low RF – $60 1-10 1-11 4 4 - 6.3 -4 4 14.08 64.62 Sale of $80 1-13 1-13 3 2 42.1 51.6 22 134.74 185.28 15 excess High RF – $60 1-19 1-19 2 2 76.2 99.6 144 193 211.19 269.59 Original $80 1-19 1-19 2 2 105.5 128.5 282 317.24 375.65 233 model High RF – $60 1-17 1-17 2 2 75.4 99.1 126 175 144.92 185.68 Fee for $80 1-19 1-19 2 2 105.1 128.1 215 264 218.93 259.68 disposal High RF – No $60 1-17 1-17 2 2 75.6 99.2 129 178 147.37 188.13 cost for $80 1-19 1-19 2 2 105.2 128.2 218 267 221.38 262.14 excess High RF – $60 1-18 1-18 2 2 75.8 99.4 131 181 149.16 190.58 Sale of $80 1-19 1-19 2 2 105.3 128.3 220 270 223.17 264.59 excess Battelle | May 2026 239 9.2 Development Plan Discussion The economic model suggests that a project is viable at the Napoleon Oilfield in all but the most conservative simulation (Low RF) at the current oil price of around $60 per STB and claiming 45Q tax credits could make this simulation economically viable as well. These results depend on the availability of CO2 at $40 per tonne, which is well within the per tonne capture cost at an ethanol plant (DOE/NETL, 2023). A requirement to purchase CO2 at same volume for the entire project does not impact the economics of these simulations drastically because of the nature of CO2 demand at for CO2-EOR in the field. In the early years of a project, CO2 is injected but has not reached the producing wells. As the project progresses, CO2 is produced and recycled at an increasing rate. Thus, the simulations in which CO2 must be purchased at a constant rate are similar to the simulations in which CO2 is purchased as needed in the early years of the project when the impact to NPV and IRR are greatest and only diverge as the project progresses. The use of existing infrastructure was found to have the most profound impact on project economics. Developing the entire Albion-Scipio Oilfield is uneconomic in all but the most optimistic scenarios because only a little over a quarter of the wells that originally produced this field are still active. The economics are improved when considering only the existing infrastructure because the single largest capital expenditure (new well drilling and completion) is negated; however, these simulations assume that the wells are placed in a way that would allow for the creation of injection/producing well patterns conducive to CO2-EOR operations. In addition, the CO2 demand, particularly for full field development, would necessitate capture from a larger, more dilute industrial source. This would increase the cost of CO2 purchased during operations. The Napoleon Oilfield does not have these same limitations because many of the wells that originally produced the field are active. In addition, the lower CO2 demand means the CO2 can be obtained by a purer CO2 source like a nearby ethanol plant. Other sources of CO2 could be considered but are likely not as economically advantageous. There are no nearby geologic sources of CO2 and their use would require high capital costs for transport via pipeline or high operating costs for transport via train. In addition, the CO2 obtained from a nearby industrial source would be more cost effective than purchasing industrially manufactured CO2, which was purchased at cost of $150 to $165 per tonne during this project, exceeding the per tonne capture costs from industrial facilities (DOE/NETL, 2023) by two to four times. The immediate next steps to support CO2-EOR operations in the oilfield are to continue studying the field to determine if the assumptions the model makes are correct. This can be done by introducing additional data and layers of complexity to the model assumptions. Limitations of the economic models described in this report include a simplified development schedule. The models assume that CAPEX is incurred entirely within the first year of the project and that operations occur within the entire oilfield or all existing wells (in the case of the Group 2 simulations of the Albion-Scipio Oilfield) at the same time. Staged field development may impact the NPV and IRR of the project. This staged development may help to limit the size of the CO2 recycling plant, particularly if the CO2 demand is higher than anticipated. The economics of these projects may also be improved through saline storage of excess CO2 acquired from purchase agreements. This would provide an outlet for excess CO2 that could derive additional revenue from 45Q tax credits. In addition to field development strategies, additional modeling that considers the heterogeneity of the geologic and reservoir properties of the field must be done to ensure an accurate representation potential production of from CO2-EOR of the oilfields of interest. Battelle | May 2026 237 Finally, an effort to characterize the existing infrastructure in the oilfields of interest and the additional infrastructure required for CO2-EOR is also needed. This effort should ensure the mechanical integrity of the existing wells, evaluate the feasibility of converting a subset of these wells to CO2 injections wells, and ensure that the wells spacing is correct for CO2-EOR operations. Additional infrastructure needed to obtain and transport CO2 to the injection wells, separate and recycle produced CO2, and store and transport produced oil and water for refining and disposal, respectively, must also be sited and acquired. Subtask 8.2. Development Strategy Plan The Development Strategy Plan is intended to guide project development in the TBR in southern Michigan. The plan describes how methodologies developed during the project will be used to accelerate CO2-EOR adoption in TBR fields. The plan is geared for use by individual operators because production in the region is conducted by small- to medium-sized independent producers. To this end, the plan covers the requirements of building out a CO2- EOR project in the area while recognizing the level of access to resources and capital these operators face. The plan helps to identify, assess, and develop solutions to technical, regulatory, and economic challenges to enable wide-scale field development. Specifically, the plan covers (1) geologic characterization of the reservoir; (2) required oil properties to implement CO2-EOR; (3) required surface infrastructure; (4) meeting the greenhouse gas reporting requirements of the federal government and state of Michigan; and (5) developing a monitoring, reporting, and verification plan. 1. Geologic Characterization This study has thoroughly characterized the geologic conditions of the Trenton-Black River trend using the available data. Trend-wide characterization was completed in Task 2. However, the significant heterogeneity of the vuggy porosity and fracture systems in this trend continue to present a challenge both during primary production and EOR operations. These unknowns present themselves as major risks for drilling success, reservoir presence, injection and production strategies, infrastructure utilization, and other aspects of field development, all of which can strongly impact project economics. Given this remaining uncertainty, the team makes the following recommendations to continue characterizing the TBR pore system, drive down uncertainty, and improve recovery factors. Improve understanding of pore system using borehole geophysics Geophysical techniques used for finding HTD fields are well established- 3D surface seismic with high-quality processing and thorough attribute analysis is effective for identifying HTD reservoirs (Hart et al., 2009). But these techniques fall short when it comes to seeing inside the reservoirs, specifically in identifying fracture and vug density and porosity distribution within the targeted reservoir. Because this study does not deal with exploration for new TBR HTD fields, it does not address improvements in reservoir identification and will instead focus on reservoir characterization. The key limiting factor for 3D surface seismic assessment of these reservoirs is the relatively thin nature of the vuggy primary porosity and the discontinuous nature of the fracture system. Vug trains can range from less than a foot thick to several tens of feet, far too small to be imaged directly by most surface geophysical techniques. As a result, vug presence should be inferred via the presence of other features (faults, folds, sags) that are thought to be involved in Battelle | May 2026 238 hydrothermal dolomite formation, and seismic attributes like attenuation or anisotropy that give us clues as to vug or fracture presence but do not provide direct evidence as to porosity type, quality, or continuity or whether fractures are open or closed. Thus, it is very difficult to directly or indirectly identify vuggy flow pathways in these highly complex reservoirs. Similarly, fractures within the reservoirs can vary in scale from feet to miles in length, may be singular or in swarms, and can range from tightly cemented and impermeable to open with darcy-scale permeability. Moreover, a single fracture or fracture swarm can exhibit all of these characteristics at once, resulting in seismic responses that may vary wildly across a single feature, all at scales that are far below the resolution of today’s most cutting-edge land-based 3D surveys. This necessitates the use of borehole geophysical techniques such as VSPs, image logs, cross- well seismic tomography, and other techniques. Image logs have been utilized extensively across the play, but because of their depth of investigation, they provide limited information about the character and extent of highly variable vug and fracture systems. Single- and multi-well downhole seismic acquisitions have been shown to be effective both for identification of fractures farther away from wellbores and for identification of CO2 location within reservoirs (Gritto et al., 2004). Where well spacing is sufficiently close (less than 1000’ between wells), cross-well seismic is recommended, as it provides the highest resolution data available. These opportunities are limited: only 17 active wells in the trend are located within 250 ft of another active well, an additional six wells are located with 500 ft of their nearest neighbor, and 63 wells are located between 500-1,000 ft of their nearest neighbor (Figure 228). These techniques may be able to illuminate key vug or fracture pathways, allowing for improved reservoir modeling and identification of preferential flow pathways. Acquisition campaigns should be undertaken prior to CO2 injection to identify key flow pathways, allowing for better planning of producer-injector arrangement, and during CO2 injection to understand real-time CO2 migration and to improve conceptual, static, and dynamic models of TBR reservoirs. Data gathered during injection can also be useful for monitoring, reporting, and verification, providing strong evidence of containment. Battelle | May 2026 239 Figure 228. Active TBR wells colored by distance to nearest neighbor. High-quality cross-well seismic can be acquired in wells <250 ft apart, moderate quality in wells <500 ft apart, and low quality in wells 500-1,000 ft apart. Increase scope of lab testing to better understand impact of fractures and large vugs In this study, lab testing investigated CO2-oil interactions, surfactant interactions with oil, water, and the rock face, and tested the ability of surfactant-enhanced CO2 flood to improve oil recovery in TBR reservoirs. It also investigated the porosity, permeability, and character of vuggy pore networks and matrix pore networks in core samples taken from TBR reservoirs and compared oil chemistry from primary production to oil chemistry after surfactant-enhanced CO2 stimulation. The main conclusions from this work support the idea that surfactant-enhanced CO2 stimulation should improve oil recovery within the TBR trend, and should be significantly more effective than water flooding. Field tests support these conclusions, with one notable caveat: Lab testing indicated that the CETAC-30 surfactant used in the field test produces a durable foam under reservoir conditions, and that the foam could be useful in diverting CO2 from high-transmissibility pathways like large vugs and open fractures toward lower transmissibility pathways like smaller vug systems and matrix porosity. This behavior was not observed in the field test, where pressure and temperature data at offset wells indicates that CO2 moved more or less freely across the reservoir. It is believed that because there are vugs and fractures within the reservoir that are too large to be effectively sampled at core scale, laboratory testing protocols did not fully characterize the Battelle | May 2026 240 full range of permeability and transmissivity, and the impact that the presence of exceptionally transmissive features would have on behavior of oil, gas, water, and CO2 within the reservoir. The maximum vug size identified in the core analysis was 11.1 mm in diameter (Figure 229), orders of magnitude smaller than the largest vugs that exist within the field. Moreover, no samples used to assess oil recovery contained large, open fractures. This likely caused an overprediction of production uplift, with core floods showing incremental recovery of up to 61% of OOIP. Repeating core flooding testing on larger core samples or manufactured media that contain large, open fractures and vugs to better assess incremental recovery is recommended. Figure 229. CT scan of the largest pore identified in laboratory-tested core samples (sample #6-44). Repeat core flooding & oil recovery tests at a broader range of pressures All core flooding tests and surfactant tests were conducted at 1,200 psi, a high pressure where CO2 is in the liquid or supercritical phase and oil and CO2 behave as pseudo-miscible to miscible fluids. The team does not recommend undertaking a water repressurization operation prior to CO2 stimulation, as this would require a vast quantity of water, damage CO2 storage capacity and recovery efficiency, and extend project timelines. This means that operators should inject CO2 into ambient reservoir conditions, where CO2 would behave as a gas until sufficiently high pressure is established to cause a phase change. Core flooding tests that assess the behavior and effectiveness of CO2 and surfactant at pressures that represent the entire range of potential reservoir conditions (<100 psi to 1,300 psi) are recommended so field operators can understand the behavior of CO2 and oil in this lower pressure regime, allowing for optimized early-time CO2 injection and oil production strategies. Investigate the benefits of other conformance technologies, such as polymer flooding Our lab testing indicates that foam flooding in the core samples that were used provided at least some CO2 mobility control. Given the lack of CO2 mobility control observed during field testing, it Battelle | May 2026 241 is not believed that CETAC or other surfactants provide sufficient mobility on their own in reservoirs where the dominant flow pathways are large vug systems and open fractures. It is recommended to investigate polymer flooding as a means for diverting injected CO2 and/or water from highly permeable vugs and fractures toward smaller, lower-permeability features such as small vugs and matrix pores. Studies by Zhang et al. (2011), Wu et al. (2023), Prakesh et al. (2024) and others indicate that polymer floods, and polymer-alternating-gas (PAG) floods can be effective for improving mobility control and increasing volumetric sweep for both miscible and immiscible CO2 floods, and in the presence or absence of surfactants. Polymer behavior can be highly sensitive to surfactant chemistry, water chemistry and salinity (Lee et al., 2019; Wu et al., 2023), so care should be taken to analyze candidate polymers and surfactants at reservoir conditions, in the presence of reservoir oil, and in the presence of both reservoir water and any water that may be used as an injectate. Use larger well, seismic, production and laboratory datasets to generate high definition geomodels The data set for the Lee 26 (Rice Creek) field that was used as this projects field test site was relatively limited: Four wells, a moderate- to high-quality 3D seismic volume, and incomplete production data. Because the wells were effectively in a straight-line, three-dimensional distribution of reservoir characteristics within the 3D static geomodel was highly uncertain away from the plain of the wells. Because of the relatively thin reservoir compared to depth, seismic porosity inversions were insufficiently high-resolution to provide meaningful control on porosity distribution. And because the production data set had very limited pressure and production volume data for some parts of the field life (and no samples or analysis of reservoir fluid composition were available at any point during field life), generating a history-matched reservoir depletion model was challenging. These challenges resulted in a reservoir model that did not fully predict actual reservoir performance, as described in the Task 7 section of the final report. The project team recommends that operators planning to implement CO2-EOR in the TBR trend utilize the much larger datasets available to them in the Napoleon, Stoney Point, and Albion- Scipio trends to generate higher-resolution geomodels and more robust reservoir history matches. This will allow for improved prediction of reservoir behavior and give operators a better understanding of how to plan CO2-EOR campaigns in the trend. 2. Required Oil Properties Based on the analysis of the mobilized oil in the Lee field, and comparison with analog oil for OFIP from the Hoffman field, it appears that the primary components mobilized are in the C12- C23 range, with some slightly heavier components also being mobilized. Lighter hydrocarbons (C1-C3) are essentially non-existent because of the depleted nature of the field, and it appears that the heavier hydrocarbons (C30+) are also left behind in the pore spaces of the reservoir. Paraffin buildup has been observed in this and other fields, indicating the presence of paraffinic compounds in the in-situ fluids. Based on the observation that C30+ concentration is much lower in the mobilized oil, it appears that the mobilization of these longer hydrocarbon chains is being inhibited within the reservoir. This could either be due to pore-throat restrictions in the matrix porosity systems or paraffin buildup within the depleted reservoir. Given these observations, it appears that an optimal reservoir fluid for surfactant-enhanced CO2-EOR in the TBR trend is a lighter oil, containing more medium-weight hydrocarbons and Battelle | May 2026 242 minimal heavy hydrocarbon molecules (paraffins/asphaltenes) that are resistant to mobilization by surfactant/CO2. Most of the oils in this trend are lighter (API of 38+), likely containing many of these medium-weight hydrocarbons that appear to be preferentially mobilized. If a field containing a fluid that exhibited minimal paraffin issues during primary production could be identified, this would likely present an optimal candidate for a surfactant-enhanced CO2 flood. Given that some amount of oil was mobilized in the Lee field, a history of paraffin buildup should not preclude the viability of a reservoir from this style of EOR treatment. It must be noted that data is limited on Lee reservoir fluid from field testing due to minimal production of in-situ fluids during the test. While conclusions can be drawn from the single analysis performed, additional data (particularly time-series compositional data) would provide a much stronger basis for recommended oil properties. The lack of baseline oil composition from the Lee field does create challenges in interpreting the optimal fluid characteristics for a flood of this nature. However, the Hoffman oil appears to provide a solid analog to Lee field discovery fluids based on all available data, increasing confidence that the compositional trends observed between the two oil samples are indeed reflective of fluid interactions occurring within the reservoir. 3. Required Surface Infrastructure Surface infrastructure required for a CO2-EOR project includes pipelines to transport CO2 to the field, distribution lines to deliver CO2 to the individual injection wells, wellhead equipment for injection and productions wells, equipment for artificial lift of fluids from the well, distribution lines to deliver produced fluids to separators, CO2 recycling equipment and compressors, storage tanks for produced oil for transport to refineries, and storage tanks for produced water for transport to disposal sites. The generalized block-flow diagram in Figure 230 shows how these equipment interconnect at a typical CO2-EOR site. Generally, the most economic method to transport large quantities of CO2 is via pipeline (Carbon Storage Leadership Forum, 2021; Svensson et al., 2004). Several CO2-emitting facilities exist close the oilfields of interest in this study (Table 50). These include industrial facilities with relatively pure streams of CO2 (ethanol) and industrial and electric power facilities with more dilute streams of CO2 (hydrogen plants and power plants). The distance from these nearby sources to the fields of interest is also shown. These sources were used, in conjunction with forecasted CO2 demand, to develop a price of capture for the economic analyses completed for the project. Battelle | May 2026 243 Figure 230. Generalized block flow diagram showing typical CO2-EOR production facility setup. Table 50. Electric power and industrial facilities and the distance to the center of the Lee, Napoleon, and Albion-Scipio Oilfields. Capturable Straight-Line Distance to Field Facility Transport (miles) No. Facility CO2 Type Mech.1 (MMt/yr) Lee Napoleon Albion-Scipio 1 Anderson’s Marathon Ethanol 0.186 Pipeline 7.8 31.2 13.3 2 Carbon Green Ethanol 0.068 Truck 31.1 62.4 51.9 3 POET Ethanol 0.090 Truck 107.0 100.3 114.8 4 Air Products Hydrogen 0.363 Pipeline 87.1 54.5 78.5 5 Dan E. Karn Power Power Plant 2.978 Pipeline 103.8 105.9 115.9 Plant 6 Midland Generation Power Plant 3.253 Pipeline 91.1 100.5 105.9 7 Dearborn Industrial Power Plant 2.713 Pipeline 87.1 54.5 78.5 Notes: 1. Transport mechanism selected using the process developed by Battelle (2025). Once the CO2 arrives at the oil field, the CO2 must be delivered to individual injection wells using a distribution network. For the purposes of this project, it is assumed that this distribution network delivers CO2 from the surface facilities (CO2 pipeline inlet and the CO2 recycling plant) to the injection wells via pipeline. For the purposes of economic and field development Battelle | May 2026 244 modeling, it is assumed a single injection well is required for every four production wells. Existing production wells can be converted into injection wells or new injection wells can be drilled as necessary between production wells. After CO2 is injected and fluids are produced, a network of gathering lines delivers the fluids to separators. For the purposes of this study, it was assumed that the separators are co-located with the pipeline inlet and CO2 distribution system. The separators route oil and water to storage containers for off-site refining and disposal, respectively. Separated CO2 is routed to the CO2 recycling plant where it is compressed and distributed back to the injection wells where the cycle repeats. Siting project infrastructure requires an analysis of existing rights-of-way and potential sensitive areas (Figure 230). Battelle (2024) provided an accounting of high consequence areas and other potential project obstacles and developed a routing philosophy to improve safety and limit project delays related to permitting pitfalls or public opposition. Every project has unique on-the- ground conditions, permitting potential, and tolerance for the risk of public opposition; however, Table 51 shows the philosophy proposed by Battelle (2024) utilizing the protective protocol of the features shown in Figure 231. Figure 231. Existing rights-of-way (ROWs) (left) and sensitive areas (right) for the project area (U.S. EPA, 2024; United States Census Bureau, 2023; Pipeline and Hazardous Materials Safety Administration (PHMSA), 2019, 2018, 2017; Energy Information Administration (EIA), 2022, 2020; Federal Railroad Administration [FRA], 1995; U.S. Census Bureau, 2017; EGLE, 2022; USGS, 2022, 2016; Department of Energy (DOE), 2015). Battelle | May 2026 245 Table 51. Sensitive features to consider when siting project infrastructure. The protective protocol and rationale are examples from Battelle (2024). High Additional Protective Feature Rationale Consequence Permitting or Protocol Area Coordination Mapped in Figure Must coordinate with Department of Transportation Interstate/Roads Moderate No Yes (DOT) or state or local authorities Rail Lines Moderate No Yes Avoid areas with critical infrastructure unless Existing Pipelines Moderate existing rights-of-way (ROWs) can be used to No Yes expedite siting Transmission Lines Moderate No Yes Commercially Avoid areas where commercial activities could be Navigable High impacted Yes Yes Waterways Waterbodies could impact logistics of infrastructure Waterbodies Moderate No Maybe development Wellhead Protection Wellhead protection areas as a proxy for High Yes (Proxy) Yes Areas High Populated High Yes No Areas Avoid to limit population exposure and public Other Populated opposition High Yes No Areas Not Mapped in Figure Sole source karst N/A for study aquifer recharge Not applicable study area Yes No area area Schools/day cares High (Buffer) Avoid vulnerable populations No No Hospitals/elder care High (Buffer) Avoid vulnerable populations No No facilities Ensure pipelines do not impact sensitive Airports Moderate No Yes operations/structures Ensure pipelines do not impact sensitive Dams Moderate No Yes operations/structures Military Installations Moderate Must coordinate with military No Yes Must coordinate with local, state, or federal Emergency Services Moderate No Yes authorities Existing electrical power or industrial Moderate Avoid critical infrastructure, where possible No Yes facilities Avoid critical infrastructure and buffer operational Mineral Extraction Moderate No Yes footprint Critically imperiled species or ecological community; multi-species assemblage areas; Environmental migratory bird concentration areas; Areas High Yes Maybe HCAs containing imperiled aquatic, aquatic dependent, or limited range species or ecological communities; and sensitive species with EORANK of A or B Beaches of Certain Coast Waters. Siting projects Coastal Beaches High Yes Yes here would be complicated by public opposition. Certain Coastal Includes the Great Lakes. Siting projects here High Yes Yes Waters would be complicated by public opposition. Battelle | May 2026 246 Table 51 (continued). Sensitive features to consider when siting project infrastructure. High Additional Protective Feature Rationale Consequence Permitting or Protocol Area Coordination Protected Areas Dataset of the These are federal, state, local, or private protected United States (PAD- High lands. Gap Status 1 and 2 prevent development of No Yes US) – Gap Status #1 these areas. and #2 Gap Status 3 lands are allowed for multi-use PAD-US – Gap extraction; however, public opposition may still High No Yes Status #3 exist and the parcel plan may not allow pipeline development PAD-US – Gap Gap Status 4 lands have no protection mandate Moderate No Yes Status #4 but may still be areas of interest to the public United States Fish These areas are critical habitats to endangered and Wildlife Service High and protected species. Developing them may lead No No (US FWS) Critical to public opposition Habitats While developing wetlands is possible, it can require additional permitting and coordination with Wetlands Moderate No Yes the US Army Corp of Engineers (USACE) if they are considered waters of the US. National Registry of This data includes areas with historical or cultural Historic Places High No Yes significance. (NHRP) Structures meant for human occupancy could present an asphyxiation hazard if a pipeline accident were to occur. Structures should be Structures Moderate/High1 avoided where possible, and careful CO2 No Yes dispersion modeling, emergency management, and notification systems should be developed for any structure near a pipeline. Notes: 1. Not considered by Battelle (2024). 4. Underground Injection Control (UIC) Permitting and Regulation Michigan received primacy for the UIC Class II, the class of wells that includes CO2 injections wells for CO2-EOR, in 2022 (U.S. EPA, 2022). The program is administered by the Michigan Department of Environment, Great Lakes, and Energy (EGLE). Michigan Administrative Code R. 324.802 now describes the requirements operators must adhere to in order to drill, convert, and/or operate an injection well. The full requirements are available in the Michigan Administrative Code and summarized below: 1. Notify oil, gas, and injection well permittees and landowners within 1,320 feet of the proposed well (herein referred to as the area of interest). 2. Complete a plat with wells, ownership information, landowners, and nearby water supplies within the area of interest. 3. Submit the completion report and geologic records, per R 324.418(a), and borehole records, per R. 324.419(1), for an existing well that is to be converted to a Class II well. This well must meet the requirements outlined in R 324.201(1) and R 324.201(2). 4. Provide plugging records for abandoned wells and construction records for existing active wells within the area of interest, including a corrective action plan for dealing these penetrations that might become a conduit for fluid migration Battelle | May 2026 247 from the injection zone to underground sources of drinking water (USDWs). 5. Create a wellbore diagram for the proposed well or existing well that is to be converted to a Class II well and includes the names and top and bottom depths of geological formations penetrated that are USDWs. 6. Provide proof that injection of fluids will not exceed the fracture pressure gradient. 7. Provide a plugging and abandonment plan. The application also provides a plan for operations that details the maximum daily injection rate and pressure, information about injected fluids, representative fluid samples, and geologic name of injection interval(s) and its separation from the deepest USDW. The application also requires that the operator identify faults, major structural features, and other geological and operational considerations (e.g., karst, mines, lost circulation zones) that may impact reservoir containment or well integrity. The operator is also required to provide a plan to mitigate any impacts to containment or well integrity. Finally, the operator is required to submit evidence that the well contraction is protective of USDWs. The timeline for the Class II application follows the steps outlined in Table 52. Landowners and well leaseholders within the Area of Interest must be notified of the application. The timeline for permit approval could theoretically be completed within 30 to 60 days of the submission of an application if no valid comments or objections are submitted to EGLE by land or leaseholders. If valid comments are received, then the application may take up to 120 days to complete, a public hearing may be required, and the applicant will have to show compliance with the regulations and demonstrate that the operations are not a risk to public safety or USDWs. Table 52. Potential timeline for a Class II application in Michigan, including individual permitting steps, and the timeline in instances where no valid comments or objections are made and instances where comments or objections are deemed valid. Scenario Steps Total Time1 Day 0 – Application Submitted Application submitted Up to Day 10 – Notice of application made public by EGLE (within 10 days of and there are no application)2 comments/objections or Up to Day 40 – Last day of public comment period (within 30 days of public notice) comments/ objections Up to Day 50 – Comments assessed for validity (within 10 days of the public 30 to 60 days are not considered comment period) valid by EGLE Up to Day 60 – Permit approved (no more than 20 days after public comment period) Day 0 – Application Submitted Up to Day 10 – Notice of application made public by EGLE (within 10 days of application)2 Application submitted Up to Day 40 – Last day of public comment period (within 30 days of public and EGLE receives notice). Valid comment/objection received; commentor/objector requests public comments/objections hearing. deemed valid by EGLE Up to Day 50 – Comments assessed for validity (within 10 days of the public 50 days to 120 and commentor/ comment period) days objector has requested Up to Day 60 – Public Hearing notice issued/scheduled (within 20 days of public a public hearing comment period) Up to Day 90 – Public hearing held (no less than 20 days but no more than 30 days of the notice of the hearing) Up to Day 120 – Permit approved if applicant meets requirements after the public comment period (within 30 days of public hearing) Notes: 1. Timelines show the minimum and maximum possible timeframes for each scenario. 2. This notification, completed by the EGLE supervisor within 10 days of the receipt of the application, will include project information that identifies the applicant, location, and depth of the Class II well; notes the injection interval depth and formation name; details the maximum injection pressures and rates; and solicits comments or objections to be submitted to EGLE within 30 days of the notice. Battelle | May 2026 248 Class II injection well construction requirements are promulgated in R. 324.804. These requirements outline that the well must be constructed to prevent fluid migrations (e.g., with casing, cement, tubing, packers, etc.) and outlines acceptable testing methods to demonstrate external mechanical integrity and fluid migration prevention. The rules also state that all casing annuli must be accessible at the surface, operators must ensure fluids are confined to the injection zone, and that the injection interval must be at least 500 feet below the lowest USDW in karst areas. Prior to receiving authority to inject, a testing program, outlined in R 324.806, must be undertaken. This program requires mechanical integrity testing. The permittee must inform EGLE at least five (5) days prior to conducting the test and provide a report and form within 14 days of the testing completion. At this point, the permittee may receive verbal and then written authorization to inject from the EGLE supervisor. Operational requirements listed in the regulations include the maximum injection pressure (calculated using the equation in R 324.807); operational testing requirements (R 324.808); weekly, monthly, quarterly, and annual monitoring and record keeping and filing (R 324.810); notification procedures in the even of a loss of mechanical integrity (R 324.811); the process for requesting and completing temporary abandonment of the Class II well (R 324.812); the authorization of the EGLE supervisor to require the cessation of Class II injection operations in the instance of public safety or health or protection of USDWs (R 324.813); and the process for Class II permit modifications (R 324.815). Meeting greenhouse gas reporting requirements and qualifying for 45Q tax credits Greenhouse gas reporting requirements have recently changed due to the recission of Subpart RR of the U.S. EPA Greenhouse Gas Reporting Program (GHGRP). As a result, the reporting requirements for a CO2 storage facility and the way in which the 45Q tax credits are claimed have changed. Subpart RR established mandatory monitoring, reporting, and verification requirements for facilities that inject CO₂ underground for long‑term geologic sequestration, and applied all wells that are used to inject CO2 for long term or permanent storage in geological reservoirs. This includes all Class VI wells, which are not unlikely to be a part of any development plan for the TBR trend, but it also applied to EOR wells that expect to permanently store any volume of CO2. CO2 volumes that could be verified as permanently stored using agreed upon MRV plans would be eligible to receive 45Q tax credits. Since most economically viable CO2-EOR scenarios for this trend require the operator to seek 45Q tax credits, this would have applied to most CO2 injection wells in the trend. The United States Internal Revenue Service (IRS) has indicated that it will provide new regulations for 45Q tax credits, but in the interim, the IRS has provided guidance on an alternative method for quantifying and documenting permanent EOR-related CO2 storage via the International Organization for Standardization (ISO) provision ISO 27916:2019. This document lays out a framework for quantifying, documenting and verifying permanent CO₂ storage via CO₂‑EOR operations, and provides guidance for verification of the portion of CO₂ that becomes geologically stored. It also allows operators to differentiate between storage of CO2 from anthropogenic sources (which would qualify for tax credits) and CO2 from natural sources (which would not qualify for tax credits). Battelle | May 2026 249 It is recommended that all operators who intend to undertake CO2-EOR in the TBR trend build an operations management plan that is consistent with ISO 27916:2019, regardless of whether they intend to seek 45Q credits. The state of Michigan does not have separate reporting requirements for greenhouse gas emissions. Thus, if the changes to GHGRP are finalized, oil and gas producers in Michigan will no longer be required to report their emissions to the federal or state government. 45Q Tax Credits are the primary way to monetize CO2 saline storage and can provide an additional source of revenue for CO2-EOR. Originally valued at $20/tonne for saline storage and $10/tonne for CO2 utilization, subsequent provisions in the 2017 Tax Cuts and Jobs Act, the Infrastructure Investment and Jobs Act (also known as the Bipartisan Infrastructure Law), and the One Big Beautiful Bill Act have incrementally raised the value of the tax credit to $85/tonne for both permanent saline storage and CO2-EOR for CO2 that is captured from an industrial facility. Qualification for the 45Q Tax credits has several steps. Obtaining the credit is contingent on the injection of CO2 into a well that is qualified under the UIC program administered by the EPA or the individual states that have obtained primacy (§1.45Q-3[b]). Class VI wells, which are those that inject CO2 for permanent geological storage, are administered by the U.S. EPA Regional Offices in most states, except for the six states with primacy over the UIC Class VI program: Arizona, Louisiana, North Dakota, Texas, West Virginia, and Wyoming. Class II wells, which include wells that inject CO2 for the purposes of EOR, are largely administered by the states, including the states of Kentucky and Ohio (U.S. EPA, 2021). Additional requirements related to the minimum amount of CO2 that must be injected annually, The next requirement to claim 45Q tax credits is a reporting mechanism. Initially, the law required the claimant to submit self-certified verification of their permanent storage of CO2 in accordance with the U.S. EPA GHGRP, Subpart RR (§1.45Q-3[d]). The U.S. EPA, however, has proposed that the current Subpart RR and other provisions of the GHGRP, be scrapped by June 2026 (U.S. EPA, 2025). This would potentially leave companies seeking to claim 45Q tax credits with no mechanism to report their storage to the IRS (Carbon Capture Coalition, 2025). In response, the IRS released a notice addressing the cancellation and providing an alternative approach for claiming the 45Q tax credits (Internal Revenue Service [IRS], 2025). The notice says that if the U.S. EPA has not launched the electronic Greenhouse Gas Reporting Tool (e- GGRT) for reporting year 2025, provisions set forth in Section 3 of the Guidance, “Safe Harbor for Calendar Year 2025 Secure Geological Storage”, under sections 3.02(2) and 3.02(3) of the guidance, will ensure that those wishing to claim the 45Q tax credits comply with reporting requirements promulgated under 45Q. The guidance states that the Annual Report that is normally submitted to e-GGRT must instead be submitted to an independent engineer or geologist for verification. The notice states that the evaluating engineer or geologist must be a registered professional engineer (PE) or professional geologist (PG) in any state and certify that the report is accurate and complete according to the reporting rules for 45Q. For CO2-EOR, it may be preferable to report verified storage according to the provisions set forth in ISO 27916:2019. This route is not affected by the cancellation of Subpart RR. Finally, the IIJA placed requirements for facilities that to be eligible for the tax credits, they must adhere to prevailing wage requirements set forth in §1.45-7 and the apprenticeship requirements set forth in §1.45-8. The wage determination sets forth that the taxpayer must ensure that “laborers and mechanics employed by the taxpayer or any contractor or Battelle | May 2026 250 subcontractor in the construction, alteration, or repair of the facility are paid [the prevailing wage]” (26CFR§1.45-7). Apprenticeship requirements state that qualified apprentices must be hired to perform at least 15% of the construction, alteration, or repair work conducted at the facility (26CFR§1.45-8). The reporting and recordkeeping requirements for the prevailing wage and apprenticeship requirements are set forth in §1.45-12. 5. Developing monitoring, reporting, and verification plans Based on the economic analysis completed in Subtask 8.1, the team believes that most operators attempting CO2-EOR in the TBR trend will likely benefit from claiming 45Q tax credits. Projects not claiming 45Q tax credits are subject only to the relatively straightforward MRV standards set forth by the EPA under 40 CFR Parts 144 and 146, as well as Michigan state- specific requirements. Since the State of Michigan has primacy for UIC Class 2 injection wells, MRV is handled through the MiSTAR injection reporting application, and must meet all reporting requirements as shown in Table 53 (MiSTAR, 2025). Table 53. Table of MiSTAR reporting requirements for UIC Class II injection wells in the State of Michigan (EGLE, 2026) For operators planning to apply for 45Q tax credits, additional MRV measures are needed. MRV plans are intended to verify the safe, permanent storage of CO2 and account for lifecycle emissions of greenhouse gases caused by project activities. Specific requirements for the MRV Plan include: Battelle | May 2026 251 • The maximum monitoring area (MMA) and active monitoring areas (AMA). The MMA typically includes the outer boundaries of any field or subsurface reservoir that will undergo CO2 injection, plus a 0.5 mile buffer. • Leakage pathway identification and monitoring, which must be able to identify any leakage of CO2 from the EOR complex through geological or well‑related leakage pathways. Potential pathways include but are not limited to:  Faults and fractures  Lateral migration out of the reservoir  Migration due to induced or natural seismicity  Diffuse leakage through trapping or sealing formations  Existing Wellbores (including abandoned wells within the MMA)  New wellbores  Injection and production equipment • Monitoring strategies to detect and quantify surface leakage as well as established baselines to help determine surface leakage. These may include but are not limited to:  Visual inspections  Injection well surveillance with pressure gauges, flow meters, temperature gauges, infrared CO2 monitoring, spectroscopy, and other strategies.  Production well surveillance with pressure gauges, flow meters, temperature gauges, infrared CO2 monitoring, and other strategies.  Mechanical Integrity Testing (MIT) • Variables for the mass balance equation for the lifecycle analysis (LCA) to account for equipment leaks, CO2 ventilation, and CO2 in produced fluids • Well identification numbers, well name, well type, and status for all wells within the MMA, and • Date to begin data collection for equation included in Subpart RR (Eq. 1) Note: 1. Only applicable for CO2-EOR projects Monitoring plans will be specific to each project and should be clearly thought out prior to project implementation. Although regulations for some aspects of CO2-EOR MRV planning have changed in recent years, the Northern Niagaran Pinnacle Reef Trend (NNPRT) CO2 MRV plan created by Battelle and Core Energy for DOE Project #DE-FC26-05NT42589 provides a comprehensive handrail for EPA-approved MRV strategies (Mawalker et al., 2020). Battelle | May 2026 252 10.0 Summary and Outlook on CO2-EOR for the Trenton-Black River Trend in Southern Michigan Over the course of six years of collaboration between Battelle and project partners, all stated objectives of the program have been completed, including full geological characterization of the TBR trend, laboratory and modeling experiments to determine the optimum composition and design of CO2-EOR operations in the TBR trend, execution of a field test of chemically- enhanced CO2 in a TBR well, and integration of the data and learnings gathered during these efforts into a full-trend development plan. Detailed reporting on these activities, their outcomes, and implications for trend-wide development is provided in the preceding sections and the accompanying appendices and attachments. This report and encompassed data will provide TBR field operators with detailed information on what worked, what did not work, and how to proceed with production optimization of their TBR assets using chemically-enhanced CO2-EOR. CO2-EOR is a relatively well understood and broadly implemented strategy for increasing incremental production across the oil and gas industry, but its application has been primarily focused on reservoirs with limited heterogeneity. The intention of this project was show first that the same physical mechanisms that improve recovery factors in homogeneous reservoirs (namely wettability alteration, viscosity alteration, oil swelling, and mobility control) are at play in heterogeneous reservoirs. This was proven by the project’s laboratory studies and dynamic simulations, with the potential exception of mobility control, which needs further study. The second intention was to demonstrate via direct field testing that CO2-EOR can work in a strongly heterogeneous reservoir. While the field test strategy implemented during this project did not succeed in producing oil, data gathered during the test sheds light on what may work for field operators who try chemically-enhanced CO2-EOR within their own reservoirs, significantly reducing the level of uncertainty carried by first-of-a-kind commercial efforts that could (and should) follow this test. Simultaneously, the project has identified several large-volume ethanol plants and other sources of CO2 emissions in the region and provided a handrail that CO2 emitters and field operators can leverage to capture, transport, and inject that CO2 into their fields. This project has also shown that, in many cases, the economics of CO2-EOR in the TBR are attractive. And finally, by completing a project of this scope in the southern Michigan Basin, the project has contributed to the knowledge base and operational experience of field operators, state regulatory agencies, local service companies, and state universities, with CO2-EOR projects which should allow follow-on projects to proceed safely and efficiently. The outlook for chemically-enhanced CO2-EOR in Trenton-Black River hydrothermal dolomite trend is good. Some challenges still remain, but opportunities for growth are clear and the pathway toward successful deployment is relatively straightforward. CO2-EOR can be a central part of regional CO2 utilization strategies over the coming years and can be a point of leadership for the State of Michigan in the regional and national CCUS landscape. Battelle | May 2026 253 11.0 References Advanced Resources International. 2006. Basin Oriented Studies for Enhanced Oil Recovery: Permian Basin. U.S. Department of Energy/Office of Fossil Energy. . Ahmed, T. (2019). Reservoir Engineering Handbook (Fifth Edition). Gulf Professional Publishing. ISBN: 978-0-12-813649-2. https://doi.org/10.1016/C2016-0-04718-6. Bhattacharya, S., et al., 2024, Building Realistic Time-Lapse Seismic Models for CO2 Plume Monitoring, Integrating Geology, Reservoir Simulation, and Rock Physics: Case Study from an Onshore CCS Site, presented at SEG Summer workshop on Geophysical Research for Gigatonnes CO2 Storage Program at Colorado School of Mines. Catacosinos, P.A., Daniels, P.A., Jr., and Harrison, W.B., III, 1990, Structure, stratigraphy, and petroleum geology of the Michigan Basin, in Leighton, M.W., Kolata, D.R., Oltz, D.F., and Eidel, J.J., eds., Interior cratonic basins: American Association of Petroleum Geologists Memoir 51, p. 561–601. Environmental Protection Agency. Website. Underground Injection Control: Classes of Well. Updated August, 2012. Accessed July, 2015. . Environmental Protection Agency. 2014. Greenhouse Gas Reporting Program 2010-2013 Datasets. Updated August 14, 2014. Accessed March 3, 2015. < http://www2.epa.gov/ghgreporting>. Frank J. Krist, David P. Lusch, Glacial history of Michigan, U.S.A.: A regional perspective, Editor(s): J. Ehlers, P.L. Gibbard, Developments in Quaternary Sciences, Elsevier, Volume 2, Part B, 2004, Pages 111-117, ISSN 1571-0866, ISBN 9780444515926, https://doi.org/10.1016/S1571-0866(04)80191-X. Hartono, K.F., Permadi, A.K., Siagian, U.W.R. et al. The impacts of CO2 flooding on crude oil stability and recovery performance. J Petrol Explor Prod Technol 14, 107–123 (2024). https://doi.org/10.1007/s13202-023-01699-y Haskin, H. K., and Alston, R. B. "An evaluation of CO2 huff'n'puff tests in Texas." Journal of petroleum technology 41(02), 1989. P. 177-184. Howell, Paul D., and Ben A. Van Der Pluijm. "Structural sequences and styles of subsidence in the Michigan basin." Geological Society of America Bulletin 111.7 (1999): 974-991. macrotrends. 2026. Crude Oil Prices (1946-2026). https://www.macrotrends.net/1369/crude-oil- price-history-chart Linde Gas https://www.linde-gas.com/industries/oil-and-gas-upstream/energized- solutions/enhanced-oil-recovery/co2-huff-and-puff Battelle | May 2026 254 McCoy, S., and Rubin, E. 2006. A Model of CO2-Flood Enhanced Oil Recovery with Applications to Oil Price Influence on CO2 Storage Costs. In: Proceedings of 8th International Conference on Greenhouse Gas Control Technologies (GHGT-8), Trondheim, Norway. June, 2006. McCoy, S. 2009. The Economics of CO2 Transport by Pipeline and Storage in Saline Aquifers and Oil Reservoirs. Department of Engineering and Public Policy Dissertation, Carnegie Mellon University. . McCoy, S., and Rubin, E. 2008. An engineering-economic model of pipeline transport of CO2 with application to carbon capture and storage. International Journal of Greenhouse Gas Control 2, 219–229. McCoy, S., and Rubin, E. 2009. The Effect of High Oil Prices on EOR Project Economics. Energy Procedia 1, 4143–4150. doi:10.1016/j.egypro.2009.02.223 Michigan Department of Treasury. 2014. Severance Tax. Updated December, 2014. Accessed September, 2015. < ttp://www.michigan.gov/taxes/0,4676,7-238-43542_43545---,00.html>. Michigan Department of Environment, Great Lakes, and Energy (MI EGLE). Michigan Oil, Gas and Minerals Division Data Explorer. Accessed 28 Jan, 2026. https://www.egle.state.mi.us/dataminer/. Miller, B. J., Bardon, C. P., and P. Corlay. "CO2 Huff 'n' Puff Field Case: Five-Year Program Update." Paper presented at the Permian Basin Oil and Gas Recovery Conference, Midland, Texas, March 1994. doi: https://doi.org/10.2118/27677-MS Mohammed-Singh, L., Singhal, A. K., and S. Sim. "Screening Criteria for Carbon Dioxide Huff ‘n’ Puff Operations." Paper presented at the SPE/DOE Symposium on Improved Oil Recovery, Tulsa, Oklahoma, USA, April 2006. doi: https://doi.org/10.2118/100044-MS Monson, C., Korose, C., and Frailey, S. 2014 Screening methodology for regional-scale CO2 EOR and storage using economic criteria. Energy Procedia 63, 7796–7808. Rubin, E. 2012. Understanding the pitfalls of CCS cost estimates. International Journal of Greenhouse Gas Control 10, 181–190. Rubin, E., Chen, C., and Rao, A. 2007.Cost and performance of fossil fuel power plants with CO2 capture and storage. Energy Policy 35, 4444–4454. Rubin, E., Davison, J., and Herzog, H. 2015. The cost of CO2 capture and storage. International Journal of Greenhouse Gas Control. http://dx.doi.org/10.1016/j.ijggc.2015.05.018 S&P Global. 2024. Upstream Costs and Supply Chain — Cost indexes. Standing, M. B. and Katz, D. L. (1942). Density of Natural Gases. Trans. 146: 140–149. https://doi.org/10.2118/942140-G Swezey, C.S., Hatch, J.R., East, J.A., Hayba, D.O., and Repetski, J.E., 2015, Total petroleum systems of the Michigan Basin—Petroleum geology and geochemitstry and assessment of Battelle | May 2026 255 undiscover resources, chap. 2 of U.S. Geological Survey Michigan Basin Province Assessment Team, Geologic assessment of undiscovered oil and gas resources of the U.S. portion of the Michigan Basin: U.S. Geological Survey Digital Data Series DDS–69–T, 162 p., http://dx.doi. org/10.3133/ds69T Thomas, G.A. “Cyclic CO2 for Recovery of Light Oils: Field Test Database Evaluation.” Master’s Thesis, Louisiana State University, Baton Rouge. August, 1989. U.S. Department of Energy, National Energy Technology Laboratory. 2011. Improving Domestic Energy Security and Lowering CO2 Emissions with “Next Generation” CO2-Enhanced Oil Recovery (CO2-EOR). Prepared by Advanced Resources International. DOE/NETL-2011/1504. . U.S. Department of Energy, National Energy Technology Laboratory. 2023a. Capture costs were determined based on published data from NETL (https://netl.doe.gov/projects/files/NETLsCostofCapturingCO2fromIndustrialSourcesandIndustria lCarbonCaptureRetrofitDatabase_012423.pdf) 2023b. (https://netl.doe.gov/projects/files/CostofCapturingCO2fromIndustrialSources_033123.pdf) 2022a. PowerPoint Presentation Vazquez, M. and Beggs, H.D. (1977). Correlations for Fluid Physical Property Prediction. Paper presented at the SPE Annual Fall Technical Conference and Exhibition, Denver, Colorado, October 1977. https://doi.org/10.2118/6719-MS Verma, M.K., ed., 2017, Three approaches for estimating recovery factors in carbon dioxide enhanced oil recovery: U.S. Geological Survey Scientific Investigations Report 2017–5062–A–E, variously paged, https://doi.org/10.3133/sir20175062. Warren, J.E. and Root, P.J. 1963. The Behavior of Naturally Fractured Reservoirs. SPE J. 3 (3): 245–255. SPE-426-PA. https://dx.doi.org/10.2118/426-PA. Yang, F., Deng, J., and Y. Xue. "Jiangsu Oilfield’s Carbon Dioxide Cyclic Stimulation Operations: Lessons Learned and Experiences Gained." Paper presented at the SPE International Conference on CO2 Capture, Storage, and Utilization, New Orleans, Louisiana, USA, November 2010. doi: https://doi.org/10.2118/139599-MS Battelle | May 2026 256 Appendix A: Supporting Data for Task 8 Economic Analysis Table D-1 lists the categorization and definitions of well cost parameters included in the cost model. D-1. Categorization and definitions of well cost parameters included in the cost model. Cost Category Definition Items Included Well Costs Drilling & Tangible and intangible costs site survey, leasehold and prospect costs, site Completion associated with on-site drilling preparation and reclamation, labor, supervision, and completion phases for surface and downhole drilling equipment, drilling production and injection wells. cost per foot, drilling cost per hour (day rate), logging costs, perforating, acid jobs, chemicals, drilling and completion fluid management, completion rig, all casing, cementing, plugging, facility costs Production Well Capital costs of tangible and, to electric services, all surface and subsurface Equipment a lesser extent, intangible items production well equipment costs: tubing, rods, down- associated with production well hole pump, pumping unit, motor, well head, equipment and installation hookups, and controller, separator, storage tanks, walkways gas sales meter, flowlines, gauges and valves Injection Well Capital costs of tangible and, to electrical services, all surface and subsurface Equipment a lesser extent, intangible items injection well equipment costs: unloading pad, associated with injection well storage tanks, oil separation equipment, filtering equipment and installation system, injection pump, injection lines, injection tubing, injection packer, buildings/shelters, well head, pressure monitoring system, flow meter/totalizer, gauges and valves Well Conversion Tangible and intangible capital labor, mechanical integrity testing, re-perforating, costs of updating or converting stimulation, acid or clean up treatment, cement bond existing injection and logging, fabrication services and materials, production wells to establish equipment replacement costs, pressure test lines, all the number of suitable injection other surface and subsurface well equipment and and production wells required rework costs: packer, tubing, well head/tubing head, for the well pattern/project fittings, regulator, monitoring equipment, gauges and valves CO2 Costs Battelle | May 2026 257 CO2 Transport & Fixed and variable capital costs high pressure trunkline pipe material, mileage, and Distribution associated with transporting diameter; similar to natural gas gathering systems the volume of CO2 required for (variable components), manifolds, values, flow injection from the regional controls, and distribution lines required for transport pipeline to the EOR site, as well to and from wells and the recycling facility as all manifolds and flow lines required for on site CO2 distribution CO2 Recycling Plant Capital costs of a recycling CO2/hydrocarbon & CO2/water separators, fluid facility of sufficient size to distribution lines, gas stream flow lines, CO2 facilitate CO2 separation from compression equipment, pumps, storage tanks produced fluids during peak Cost Category Definition Items Included production and recycling (crude and water), manifolds, valves, flow controls, operations processing equipment CO2 Recycling Expenses related to operation energy required for recycling facility operation over a (O&M) of the recycling facility given time period, including power for separation, corresponding to the rate of processing, CO2 compression to the MMP required production, recycling, and for injection injection simulated for a specific time period, with the energy required for CO2 surface management, treatment and compression generated on-site Periodic O&M Costs associated with operation periodic well workover, servicing, operating expenses and maintenance of a CO2-EOR for labor, personnel, electricity, consumables, surface site at the required capacity and subsurface equipment maintenance and over a given time period, replacement including periodic well workover, and other surface and subsurface repairs/remediation Liquid Lifting Costs Operational costs of pumping energy required for lifting liquids up the wellbore and (to surface), managing, and surface routing (re)distributing liquids produced during a given time period of CO2-EOR operation, with the required energy generated on- site General and Other expenses associated with staff salaries/wages, insurance, accounting, Administrative routine activities and processes consulting, management, accounting and record (G&A) required for effective operation keeping supplies of the overall project, but not directly related to the production of marketable goods and services Battelle | May 2026 258 Attachment A: Advanced Field Characterization and Machine Learning Based Data Integration, Task 2 Final Report Full Task 2 topical report included as separate Attachment A (https://doi.org/10.2172/3367193). The U.S. Department of Energy (DOE) has a long history of supporting research in advancing technology and methodologies to improve enhanced oil recovery (EOR). Although EOR is not uncommon, extensive studies stem from the Permian Basin, which may not be applicable to other major plays/basins. Additionally, there are extensive, historically produced plays that have not been explored using EOR or new techniques. There is a critical gap in applying new EOR methodologies to reservoirs/plays outside of the Permian Basin and in historical fields. This FOA addresses AOI 1B to DOE: “enhanced oil recovery technologies for conventional resources in un-tested reservoirs/plays.” The primary objective of the proposed research is to carry out a comprehensive laboratory experiment, computer modeling, and field testing-based evaluation of chemically enabled CO2- EOR in the Southern Michigan Basin (MB)—with a focus on the conventional Trenton/Black River (TBR) play. This work will be conducted in the complex multi-porosity (i.e., matrix, vuggy, and fracture porosities) and/or hydrothermally altered dolomite (HTD) reservoirs developed along fault systems. These reservoirs are especially challenging for EOR due to heterogeneities, compartmentalization, and presence of thief zones, and as a result, there is a heightened need to customize the CO2 injection-program for satisfactory incremental oil recovery. Injection of pure CO2 alone may not be effective in recovering stranded oil. These challenges will be addressed through advanced field characterization, integrated physics-based machine learning (ML) and data analytics, laboratory process development, and optimized field tests. The research concept involves the integration of multiple data types to evaluate fields in the study area to determine those with the lowest technical and environmental risk and optimal setting for EOR. Laboratory experiments will be used to optimize a CO2 flood specific to HTD rock properties and subsequently design and simulate injection scenarios that offer wettability alteration, foaming, and reduced surface tension. This approach will improve oil recovery from matrix porosity and mitigate the impact of fracture zones. Results will provide strategies to improve oil recovery in complex carbonate formations and development plans for historical plays, thus increasing the likelihood of the application of EOR methods for similar reservoirs and numerous operators across a wide range of reservoirs and fields. The Advanced Field Characterization and Machine Learning Based Data Integration task (Task 2) involved a systematic geologic characterization of the TBR play in the SMB which included the development of comprehensive datasets, advanced data analytics, risk assessment, and piggyback field characterization. These activities aimed to address the complex carbonate systems by evaluating the extent of fractures, dolomitization, facies, and reservoir properties with the primary objective of informing the static and dynamic modeling, field injection test, and providing input into the development strategy plan. The task was divided into four subtasks: • Subtask 2.1 – Data compilation, review, and analysis Battelle | May 2026 259 • Subtask 2.2 – Risk Assessment • Subtask 2.3 – Advanced Field Characterization • Subtask 2.4 – Integrated Physics-Based Machine Learning and Advanced Data Analytics Several activities were planned and executed under Task 2 to meet the project objectives. Figure A-1 illustrates the flow of results between subtasks, which will lead to the development of a static earth model (Task 4). Figure A-1. Workflow illustrating the flow of results from data compilation and analysis to machine learning and static earth modeling. Data Compilation, Review, and Analysis Data was compiled for the TBR play for three fields: Albion-Scipio (location of the field injection test), Stoney Point, and Napoleon. Together, this included information for more than 900 wells with more than 7,600 wireline logs curves and 48 wells with core data. The data ranged in age, vendor, and type creating challenges in integration and analysis. Workflows were established to digitize, quality check, correct, and normalize the data to generate a higher quality database. Additionally, advanced data was collected through the processing of existing image logs and high-resolution scanning of existing whole core to characterize important reservoir features. Results included: • Fractures and vugs were commonly observed in dolomitic samples • Porosity ranged from 0.1-20.7%, with dolostone having the highest average • Permeability ranged from 0.01-939 mD, with dolostone having the highest average Battelle | May 2026 260 The 3D seismic data over the Albion-Scipio field was analyzed by project partners which include well tie analysis, horizon mapping, attribute analysis, and fault and fracture identification. Three attributes were identified as potential attributes to link to petrophysical work, including RMS, instantaneous frequency, and variance. More than 50 faults were identified in the volume which included a few large faults with numerous smaller faults/fractures. The petrophysical analyses yielded a set of computed logs to represent the effective porosity and permeability of the reservoir. The data was used to compute a set of attributes for each well including average wireline log properties (porosity and permeability), facies thicknesses (dolomite, limestone, cap dolomite), and reservoir attributes (net thickness, porosity feet, and net to gross ratio). Results included: • Stoney Point field had the thickest cap dolomite while the Albion-Scipio field had the greatest summed occurrence of dolomite. • Log measured porosity and computed permeability was highest in the Napoleon field • Zone attributes (net thickness, porosity feet, and net to gross ratio) were highest in the Napoleon field, followed by the Albion-Scipio field The advanced data (image logs and CT scans) were used to develop a set of flags and indicators for vugs and fractures. The features were characterized to determine classification, dip, azimuth, and frequency of occurrence. Results included: • The samples were dominated by small to large mesopores (1/16mm- 4mm) with larger pores (megapores) due to secondary porosity (Vugs and fractures) • Numerous vugs and fractures were identified with up to 20% secondary porosity Advanced Field Characterization The McCann 1-20 well was drilled in late 2020 to collect modern data, characterize the injection site, and develop a production history. The wireline log data analysis identified three zones of interest due to the presence of enhanced porosity with vugs and fractures: • 3,840-3,845 ft- dolomitic with bedded vugs, but water saturated • 3,851-3,855 ft- dolomitic and highly fractured • 3,863-TD ft- dolomite with vugs and fractures, oil saturated Integrated Physics-Based Machine Learning and Advanced Data Analytics All the databases and types were integrated into a single dataset, which required the use of advanced data analysis techniques to match and merge appropriately. The purpose of the integrated database was to 1) identify electrofacies, 2) develop permeability prediction models, and 3) develop vug and fracture prediction models. Cluster analysis techniques were applied to the wireline log dataset to identify how the data naturally clustered into groups. The resulting clusters were compared with geologic interpretations to determine geologic meaning and importance of each cluster. Six clusters were identified which represented high porosity dolomite, low to medium porosity dolomite, non- reservoir limestone, and shale. Core measured porosity and permeability data, along with wireline log data were explored to develop permeability predictions. Due to the heterogeneity of the carbonate reservoir, a Battelle | May 2026 261 permeability model could not be accurately fit. Alternatively, a relationship was established between porosity and permeability for each electrofacies to provide a permeability estimate for modeling purposes. The vug and fracture flags were evaluated with the wireline log data to determine if a predictive model could be developed for the presence of the features. Unfortunately, the vugs and fractures could not be predicted with the wireline log database. The database yielded important information about feature frequency and characteristics which will be used to develop a 3D fracture model. Risk Assessment A series of risk assessments were designed to evaluate the planned injection site and surrounding areas which included an evaluation of existing wellbores, geohazards, caprock assessment, and a risk screening-based analysis and activity-based “what if” analysis. A systematic evaluation of the wellbores was used to identify any potential leakage risks based on construction records, plugging records, and cement bond logs. Overall, there were few wells that had insufficient information within the Albion-Scipio field, but none within proximity to the injection site. The evaluation of geohazards and caprocks showed there were no red flags in the area and system. Additionally, the other analyses showed the only risks were those that were typical with oil and gas operations. Overall, this task developed a comprehensive and integrated database which captured the variability of the complex carbonate reservoir. Moving forward, efforts will be focused on analyzing data in 3-dimensions to develop a representative static earth model, which includes changes in lithology, facies, and secondary porosity features. Additionally, key field and pool characteristics will be identified to be used in a comparative study to aid in the scale-up assessment and development strategy plan. Battelle | May 2026 262 Attachment B: UT Austin Laboratory Experiments to Optimize CO2-EOR Fluids for Complex Carbonate Reservoirs Full report of laboratory work is included as Attachment B to this final report. This work was completed by Dr. Kishore Mohanty and associated faculty and students at UT Austin. Task 3 of the project was conducted at The University of Texas at Austin. It consisted of core characterization, MMP determination, CO2 flood, foam flood and compositional simulation. CT scanning showed that the cores are heterogeneous and have a complex pore geometry. Mercury porosimetry showed a dual pore size distribution from nanometers to 100 microns. The relative permeability of oil can be fitted to a Corey expression, but not the water relative permeability. These cores are oil-wet. Slim tube experiments were conducted for reservoir oil with CO2. The MMP for CO2 with Michigan oil was 2100psi at 40°C. The MMP for CO2 with Napoleon oil was 1300psi at 40°C. The MMP for CO2 with McCann live oil was 1300psi at 40°C. Additives ME-1 and ME-2 cannot contribute to an improvement in miscibility performance between Michigan oil and CO2. Continuous CO2 flooding with 0.5 PV CO2 injection could improve 14.4% OOIP over water flooding. WAG injection could help control CO2 mobility and increase the oil recovery to 20.1% OOIP. Foam stability tests and contact angle tests were conducted to screen foam agents. And CO2 foam core flooding experiment was carried out to evaluate the effectiveness of foam flooding in reservoir core samples. CETAC-30 can generate better foam and can tolerate high salinity. Also, it can alter the wettability from oil-wet to water- wet. Soloterra 843 is not a good foaming agent at reservoir conditions and performed worse at high salinity. LMDO can generate the most stable foam among these three surfactants. However, it cannot alter the rock wettability. CO2-foam core flooding by using surfactant CETAC- 30 can increase oil recovery by 24% OOIP over the water flooding. This is 9.6% higher oil recovery compared with CO2 flooding and 4% additional oil recovery over the WAG flooding. Compositional simulations showed that introduction of 10%-20% rich gas can significantly decrease MMP between crude oil and CO2. 263 Chemically Enabled CO2-EOR in Multi-Porosity, Hydrothermally Altered Carbonates in the Southern Michigan Basin Final Report to Battelle Jieqiong Pang Kishore K. Mohanty The University of Texas at Austin December 15, 2023 Task 3 of the project was conducted at The University of Texas at Austin. It consisted of core characterization, MMP determination, CO2 flood, foam flood and compositional simulation. CT scanning showed that the cores are heterogeneous and have a complex pore geometry. Mercury porosimetry showed a dual pore size distribution from nanometers to 100 microns. The relative permeability of oil can be fitted to a Corey expression, but not the water relative permeability. These cores are oil-wet. Slim tube experiments were conducted for reservoir oil with CO2. The MMP for CO2 with Michigan oil was 2100psi at 40°C. The MMP for CO2 with Napoleon oil was 1300psi at 40°C. The MMP for CO2 with McCann live oil was 1300psi at 40°C. Additives ME-1 and ME-2 cannot contribute to an improvement in miscibility performance between Michigan oil and CO2. Continuous CO2 flooding with 0.5 PV CO2 injection could improve 14.4% OOIP over water flooding. WAG injection could help control CO2 mobility and increase the oil recovery to 20.1% OOIP. Foam stability tests and contact angle tests were conducted to screen foam agents. And CO2 foam core flooding experiment was carried out to evaluate the effectiveness of foam flooding in reservoir core samples. CETAC-30 can generate better foam and can tolerate high salinity. Also, it can alter the wettability from oil-wet to water-wet. Soloterra 843 is not a good foaming agent at reservoir conditions and performed worse at high salinity. LMDO can generate the most stable foam among these three surfactants. However, it cannot alter the rock wettability. CO2-foam core flooding by using surfactant CETAC-30 can increase oil recovery by 24% OOIP over the water flooding. This is 9.6% higher oil recovery compared with CO2 flooding and 4% additional oil recovery over the WAG flooding. Compositional simulations showed that introduction of 10%- 20% rich gas can significantly decrease MMP between crude oil and CO2. Contents Introduction ................................................................................................................................ 4 Task 3.1 Core Characterization .................................................................................................. 4 3.1.1 Core Samples .................................................................................................................... 4 3.1.2 CT-Scan ............................................................................................................................. 6 3.1.3 Mercury Porosimetry ........................................................................................................ 7 3.1.4 Oil/Water Relative Permeability Test.............................................................................. 10 3.1.5 Imbibition Tests............................................................................................................... 13 3.1.6 Conclusions ..................................................................................................................... 13 Task 3.2: MMP Determination................................................................................................. 14 3.2.1 Methodology ................................................................................................................... 14 Material and Conditions........................................................................................... 14 Experimental equipment .......................................................................................... 14 Experimental procedure ........................................................................................... 15 3.2.2 Results ............................................................................................................................. 15 MMP determination for reservoir oils ..................................................................... 15 The effects of additives ............................................................................................ 18 3.2.3 Conclusions ..................................................................................................................... 20 Task 3.3: CO2 Core Flood ........................................................................................................ 22 3.3.1 Materials ......................................................................................................................... 22 3.3.2 Methodology ................................................................................................................... 22 3.3.3 Results ............................................................................................................................. 23 Continuous CO2 flooding......................................................................................... 23 WAG flooding .......................................................................................................... 24 3.3.4 Conclusions ..................................................................................................................... 25 Task 3.4: CO2 Foam Flood....................................................................................................... 26 3.4.1 Materials ......................................................................................................................... 26 3.4.2 Methodology ................................................................................................................... 26 Foam stability test .................................................................................................... 26 Contact angle test ..................................................................................................... 27 Foam core flooding experiment ............................................................................... 27 3.4.3 Results ............................................................................................................................. 28 Foam stability........................................................................................................... 28 Contact angle ........................................................................................................... 32 Foam flooding .......................................................................................................... 33 3.4.4 Conclusions ..................................................................................................................... 34 Task 3.5: Lab-Scale Compositional Modeling......................................................................... 35 3.5.1 Methodology ................................................................................................................... 35 3.5.2 Results ............................................................................................................................. 36 Slim tube simulation of live oil................................................................................ 36 Slim tube simulation of dead oil .............................................................................. 37 3.5.3 Conclusions ..................................................................................................................... 38 3.6: Overall Conclusions…………………………………………………………………..…39 3 Introduction Task 3 of the project was conducted at The University of Texas at Austin. It consisted of core characterization, MMP determination, CO2 flood, foam flood, and simulations. The results of these subtasks are described below. Task 3.1 Core Characterization The objective of this task is to describe characteristics of reservoir core samples by a series of laboratory tests. First, CT-scanning was applied to identify the presence of vugs and fractures within the core samples. Following this, two core samples were used to determine pore size distribution and capillary pressure by mercury porosimetry. After that, one composite core mount was used to determine the oil and water relative permeabilities by the unsteady state method. Then imbibition tests were performed to determine the wettability of these core samples. 3.1.1 Core Samples Thirteen core samples with a diameter of 1.5 inches were obtained from the Napoleon field. The basic information of these core samples is shown in Table 3.1.1. These core samples have a high-level of heterogeneity and complex multi-porosity regions, including many vugs and fractures. The air permeabilities of these core samples were measured under different pressure drops. The results are shown in Figure 3.1.1, which indicates that the permeability range of these core samples is very wide, from several millidarcy to thousands of millidarcy. Table 3.1.1 Reservoir core samples from Napoleon field Sample No. Depth (ft) Length (cm) Mass (g) Porosity (%) 1 1-5 4063.6 4.14 124.6 6.95 2 1-24 4082.95 4.17 103.0 -- 3 1-31 4089.1 4.04 114.1 13.57 4 1-40 4098.65 4.11 115.1 -- 5 1-45 4103.95 4.2 121.3 13.2 6 1-50 4108.7 4.12 110.9 -- 7 6-41 4300.2 4.02 116.4 -- 8 6-42 4301.2 4.05 117.3 8.08 9 6-43 4302.1 4.04 119.6 6.3 10 6-44 4303.8 4.05 104.6 17.82 11 6-48 4307.95 3.96 112.6 12.65 12 6-51 4310.15 4.00 115.6 8.07 13 6-54 4313.6 4.00 111.5 12.19 4 Figure 3.1.1 Air permeabilities of core samples under different pressure drops Sixteen core samples with a diameter of 1 inch were obtained from well Rowe A-2 and well Nerreter. The basic information of these core samples is shown in Table 3.1.2. These core samples also have complex multi-porosity regions, including low permeability matrix, vuggy regions, and fractures. The porosity of these core samples was roughly 7% to 10%. For air permeability, most of them ranged from 0.1mD to 0.3mD. These core samples were not used in core flooding experiments due to their extremely low permeability. Table 3.1.2 Reservoir core samples from well Rowe A-2 and well Nerreter Air permeability Depth Sample Length Porosity Well Mass (g) (mD) (ft) No. (mm) (%) (100psi) R11 58.6 82.0900 8.0 0.1 R12 65.3 90.8857 7.0 0.09 3721 R13 50.2 70.2186 -- -- R14 51.3 71.6856 0.1 ROWE A-2 R21 46.2 63.5356 7.9 0.14 R22 57.7 80.4414 -- 0.11 3867 R23 62.1 84.2654 -- -- R24 66.0 91.4618 -- 6.99 N11 54.2 76.2179 10.1 0.15 N12 73.9 102.6867 10.0 3.07 4431 N13 77.7 108.6033 8.2 0.4 N14 77.9 108.9542 -- 0.23 NERRETER N21 75.4 104.1868 10.3 0.18 N22 74.4 104.9275 7.6 0.21 4446.5 N23 73.7 103.2580 8.7 0.16 N24 76.7 106.4110 -- 0.23 5 3.1.2 CT-Scan Core sample #6-44 was CT-scanned to identify its vug and fracture content, utilizing Micro- CT technology with a resolution of 34.632μm. Figure 3.1.2 shows the CT image of the core cross section and Figure 3.1.3 displays a 3D CT image of the core sample. These images reveal numerous small and large vugs in the core sample, and no obvious presence of fractures. Notably, the largest pore size measures approximately 11.1mm, which is shown in Figure 3.1.4. Figure 3.1.2 CT image of core sample #6-44 Figure 3.1.3 3D CT image of core sample #6-44 6 11.102mm Figure 3.1.4 The largest pore present in the core sample #6-44 3.1.3 Mercury Porosimetry Mercury porosimetry was conducted on two core samples from Napoleon field, #1-40 and #6- 41. The average air permeability of #1-40 was 34.6mD and the average air permeability of #6- 41 was 3.7mD. Intrusion volume of mercury versus applied pressure were obtained, and the pressures were converted into pore throat sizes. Figure 3.1.5 shows the mercury porosimetry curves for core sample #1-40, including the cumulative intrusion and incremental intrusion versus pore throat diameter. It indicates that core sample 1-40 has a dominant pore throat size, 14.97μm. For the rest of pores, it has another dominant pore throat size, 80.87μm, which might be attributed to microfractures or vugs present in the core sample. Figure 3.1.6 shows the mercury porosimetry curves for core sample #6-41. It has one dominant pore throat size, 26.96μm, which means core sample #6-41 is more homogeneous than core sample #1-40. The largest pores connected to the surface of core sample #1-40 and # 6-41 are very close, which are 180.38μm and 180.54μm respectively. From mercury porosimetry results, the capillary pressure curve can be obtained. Figure 3.1.7 shows the capillary pressure versus mercury saturation for these two core samples. These two core samples have the same threshold pressure, which is 1.0 psi. The capillary pressure curve of #6-41 is flatter than that of #1-40, which means that core sample #6-41 has more uniform pores. 7 Figure 3.1.5 Mercury porosimetry curves for core sample #1-40 Figure 3.1.6 Mercury porosimetry curves for core sample #6-41 8 Figure 3.1.7 Capillary pressure curves for core sample #1-40 and #6-41 Mercury porosimetry was also conducted on two core samples from well Rowe A-2 and well Nerreter, #R11 and #N11, with the permeability of 0.1mD and 0.15mD respectively. Figure 3.1.8 shows the mercury porosimetry curves for core sample #R11, including the cumulative intrusion and incremental intrusion versus pore throat diameter. It indicates that core sample R11 has two dominant pore throat sizes, 23.26μm and 0.07μm, which were attributed to microfractures or vugs and matrix respectively. Figure 3.1.9 shows the mercury porosimetry curves for core sample #N11. It also has two dominant pore throat sizes, 105.49μm and 0.009μm, and has more dispersed pore throat size distribution. For these two core samples, the extremely low matrix pore throat sizes also reveal the ultra-low permeabilities. Figure 3.1.8 Mercury porosimetry curves for core sample #R11 9 Figure 3.1.9 Mercury porosimetry curves for core sample #N11 3.1.4 Oil/Water Relative Permeability Test One composite core mount was used to determine oil and water relative permeabilities by the unsteady state method. The core mount was composed of three core samples, #1-54, #1-45, and #6-48. The properties of this core mount are shown in Table 3.1.3. Its permeability was 53mD, which was close to the reservoir average permeability. The properties of crude oil and formation brine are shown in Table 3.1.4 and Table 3.1.5, respectively. In the unsteady state method, the core mount was first saturated with the crude oil at irreducible water saturation and aged at 100°C for two weeks. After that, crude oil was injected into the core mount at several different flow rates to calculate the oil relative permeability at irreducible water saturation. Then, formation brine was injected into the core mount at a constant flow rate of 0.15ml/min (5ft/day) to displace crude oil. The effluents coming from the core mount and the pressure difference between the inlet and outlet were monitored during the process. Finally, the water relative permeability at the residual oil saturation was calculated. The oil and water relative permeabilities were estimated from experimental data by the Johnson, Bossler, and Naumann (JBN) method. The calculated results are shown in Table 3.1.6. Figure 3.1.10 shows the oil-water relative permeability curves from the unsteady state experiment. It should be noted that the relative permeability curves are obtained over the limited saturation range of 0.458≤Sw≤0.64. The relative permeability curves between Swirr = 0.298 and Sw = 0.458 cannot be obtained from the unsteady state method. The Corey-correlation (Eq. (1) and Eq. (2)) is used to match the experimental results. The parameters in Corey-correlation are shown in Table 3.1.6 and the Corey type relative permeability curves are shown in Figure 3.1.8. The oil relative permeability curve matches very well; however, the water relative permeability curve match is worse. = krw krw0 S n= , kro kro 0 (1 − S ) m Eq. (1) S w − S wr Eq. (2) S= 1 − Sor − S wr 10 Table 3.1.3 The properties of the core mount1 Core Mount1: 6-54, 1-45, 6-48 Length (cm) 12.16 Diameter (cm) 38.1 Mass (g) 345.4 Absolute permeability (mD) 53 Porosity (%) 11 Initial oil saturation (%) 70.3 Table 3.1.4 The properties of crude oil Crude oil Napoleon oil Gravity (API°) 42.6 Viscosity (25°C, cp) 5.5 Viscosity (40°C, cp) 4.25 Density (25°C, g/ml) 0.805 Molecular weight (g/mol) 202.21 Table 3.1.5 The properties of formation brine Brine Density (25°C, g/ml) 1.16 pH (25°C) 6.38 Viscosity (25°C, cp) 1.66 Viscosity (40°C, cp) 1.26 TDS (mg/L) 247526.5 Cations (mg/L) + Na 65307.84 +2 Ca 20800 +2 Mg 6100 Anions (mg/L) - Cl 155245.66 - HCO3 73 Table 3.1.6 Calculated results for oil and water relative permeabilities PVI Vo (ml) Vw (ml) △P (psi) Sw kro krw 0.2980 0.83396 0.00000 0.36 0.30 1.20 5.21 0.4591 0.05191 0.13702 0.46 0.10 1.40 4.91 0.4643 0.04505 0.14893 0.55 0.10 1.40 4.87 0.4697 0.04010 0.15858 0.75 0.20 3.10 4.58 0.4811 0.03334 0.17375 1.14 0.30 5.70 3.76 0.4983 0.02561 0.19509 1.76 0.30 9.40 3.64 0.5154 0.01936 0.21736 2.35 0.10 8.80 3.56 0.5203 0.01599 0.23240 2.94 0.10 9.10 3.43 0.5256 0.01375 0.24413 11 3.54 0.10 8.80 3.3 0.5311 0.01214 0.25375 4.13 0.10 8.90 3.24 0.5368 0.01091 0.26190 4.72 0.10 8.90 3.12 0.5425 0.00994 0.26897 5.31 0.30 8.60 3.06 0.5615 0.00915 0.27520 5.90 0.20 8.80 3.05 0.5740 0.00849 0.28079 7.67 0.20 26.80 2.92 0.5856 0.00704 0.29466 9.44 0.20 26.80 2.84 0.5975 0.00606 0.30561 12.06 0.20 39.40 2.75 0.6092 0.00506 0.31847 13.83 0.10 26.90 2.64 0.6150 0.00457 0.32560 15.60 0.10 26.90 2.65 0.6208 0.00418 0.33183 16.49 0.01 13.40 2.57 0.6212 0.00401 0.33467 19.14 0.01 40.30 2.54 0.6210 0.00358 0.34232 20.92 0.02 27.00 2.53 0.6218 0.00335 0.34682 21.80 0.03 13.50 2.53 0.6235 0.00325 0.34891 22.69 0.04 13.50 2.53 0.6259 0.00315 0.35092 0.6400 0.00000 0.57358 Figure 3.1.10 Oil-water relative permeability curves Table 3.1.7 Parameters for Corey-correlation Corey-correlation ka (mD) 53 kw0 (mD) 30.4 ko0 (mD) 44.2 krw0 0.574 kro0 0.834 Swr (%) 0.298 Sor (%) 0.64 n 2.8 m 4.3 12 3.1.5 Imbibition Tests Imbibition tests were conducted using two core samples, #1-31 and #6-44. The properties of these two core samples are shown in Table 3.1.8. The procedure commenced by saturating the core samples with formation brine, followed by flushing with crude oil until irreducible water saturation was achieved. Subsequently, the core samples were aged at 100°C for two weeks. Following this, core samples were immersed in formation brine at reservoir temperature (40°C) and the volume of oil displaced was measured. Figure 3.1.11 shows the oil recovery with time. After 15 days, the oil recovery of core sample #1-31 was 3.06%, and the oil recovery of core sample #6-44 was 5.58%. These two core samples can be considered as oil-wet because of the low imbibition oil recovery. Also, the core sample with the higher permeability had a higher imbibition oil recovery compared to the core sample with the lower permeability. Table 3.1.8 Properties of core samples used to do the imbibition tests Core sample #1-31 #6-44 Length (cm) 4.11 4.05 Diameter (cm) 3.81 3.81 Absolute permeability (mD) 127.5 490.8 Porosity (%) 13.1 14.2 Initial oil saturation (%) 64 39.8 Figure 3.1.11 Imbibition oil recovery for core sample #1-31 and #6-44 3.1.6 Conclusions  CT scanning showed that the cores are heterogeneous and have a complex pore geometry.  Mercury porosimetry showed a dual pore size distribution from nanometers to 100 microns.  The relative permeability of oil can be fitted to a Corey expression, but not the water relative permeability.  The cores are oil-wet. 13 Task 3.2: MMP Determination Minimum miscibility pressure (MMP) is one of the most crucial parameters in any gas injection project. It is defined as the lowest pressure at which the injected gas and the reservoir oil become miscible through a multi-contact process at the reservoir temperature. CO2 floods are efficient if conducted at pressures at or above MMP where CO2 develops multi-contact miscibility with the oil. Slim tube experiments were conducted to determine MMP for various reservoir oils. The effects of CO2 additives were also studied by slim tube experiments. 3.2.1 Methodology Material and Conditions (1) Experimental fluids Displacing phase: pure CO2 Displaced phase: three different kinds of oils were used to determine MMP. The properties of crude oils are shown in Table 3.2.1. Michigan oil and Napoleon oil were tested as dead oils, while McCann oil was a live oil, having been pre-saturated with methane. Table 3.2.1 Properties of crude oils Crude oil Michigan oil Napoleon oil McCann live oil Gravity (API°) 37.5 42.6 40.8 Viscosity (25°C, cp) 14.7 5.5 8.5 Viscosity (40°C, cp) 8.3 4.25 6.05 Density (25°C, g/ml) 0.83 0.805 0.82 Molecular weight 201.3 202.21 215.75 (g/mol) GOR (scf/bbl) -- -- 220 (2) Slim tube: A 20 ft coiled slim tube was used for the experiments to determine MMP. The tube was made of stainless steel with an inner diameter of 0.26 in, thickness of 0.063 in, and was packed with 100-150 mesh sand. The porosity was 37.4%. (3) Temperature: 40°C (4) Pressure: 1000psi - 2300psi Experimental equipment Figure 3.2.1 is the schematic of the set up for slim tube experiment, which includes: (1) a 20 ft long slim tube (2) three ISCO pumps: two are used to displacing fluids and the third one is used for back pressure (3) two accumulators which contains oil and CO2 respectively (4) one pressure transducer is used to monitor pressure drop (5) one view cell is used to observe oil-CO2 phase behavior (6) one back pressure regulator (BPR) to control the experiment pressure (7) an effluent collector 14 (8) two cameras: one is used to monitor the view cell and the other one is used to monitor the collector. Pressure Transducer Coil CO2 Oil View Cell BPR ISCO PUMP ISCO PUMP Figure 3.2.1. Schematic of the set up for MMP determination apparatus using slim tube Experimental procedure (1) To guarantee that the porous medium is clean and dry before each run, 2 pore volumes of toluene is injected at a rate of 0.5 cm3/min followed by 2 PV of supercritical CO2, and then the coil is subjected to vacuum. During that stage, the absence of leaks is verified with the assistance of a vacuum gauge. (2) The temperature is set to the reservoir condition (40°C). (3) Crude oil is injected at a rate of 1 cm3/min while gradually increasing the back pressure until the desired test pressure is reached. After the stabilization of pressure and temperature, 1 PV of the dead crude oil is injected to guarantee the system is fully saturated. (4) 1.2 PV of CO2 is pumped at a rate of 0.018 cm3/min (6.5ft/day) to displace the oil. The oil recovery and pressure drop are monitored. (5) The test pressure is set to various values (1000psi - 2300psi) in separate experiments and the steps (1)-(4) are repeated. The resulting recovery factors are plotted as a function of pressure. Two regions are identified. Firstly, the ‘immiscible region’ is characterized by recovery factors that exhibit a strong dependency on pressure variation. Secondly, the ‘miscible region’, where the recovery factor either approaches or exceeds 95% and exhibits a weak function of pressure. The MMP is determined as the point of intersection between these two regions. 3.2.2 Results MMP determination for reservoir oils (1) Michigan oil Slim tube experiments were conducted with Michigan oil at seven different pressures and the oil recoveries at 1.2 PVI are shown in Table 3.2.5. The 1.2 PVI recovery factors are plotted as 15 a function of pressure in Figure 3.2.2. The oil recovery increased with pressure. The first five oil recoveries (1000psi-2000psi) were lower than 95%. However, the last two recoveries (2100psi and 2300psi) surpassed 95% with a slight difference. Therefore, MMP was determined as 2100psi. Table 3.2.5 Oil recovery at different pressures (Michigan oil) Pressure (psi) Oil recovery (%) 1000 66.4 1500 76.9 1750 86.9 1900 89.4 2000 93.9 2100 97.4 2300 97.9 Figure 3.2.2 The relationship between oil recovery and pressure (Michigan oil) (2) Napoleon oil Slim tube experiments were conducted with Napoleon oil at three different pressures, and the corresponding oi recovery factors are presented in Table 3.2.6 and Figure 3.2.3. At 1100psi, the oil recovery was 78.3%, which is below 95%, indicating an immiscible flooding. At 1300psi, the oil recovery notably improved to 93.4%. Increase of pressure to 1500psi resulted in an oil recovery of 94.6%, which was close to the previous one. Therefore, the MMP for Napoleon oil and CO2 is approximately 1300psi at 40°C. Table 3.2.6 Oil recovery at different pressures (Napoleon oil) Pressure (psi) Oil recovery (%) 1100 78.2 1300 93.4 1500 94.6 16 Figure 3.2.3 The relationship between oil recovery and pressure (Napoleon oil) (3) McCann live oil Slim tube experiments were performed using McCann live oil, composed of McCann dead oil combined with methane at a Gas-Oil Ratio (GOR) of 220 scf/bbl. These experiments were conducted at four pressures, and the resulting oil recovery factors are shown in Table 3.2.7 and graphically represented in Figure 3.2.4. At 1100psi, the oil recovery was 68%, indicating an immiscible flooding. At higher pressures (1300psi, 1500 psi, and 1700 psi), the oil recovery was higher than 95%, indicating a shift to a miscible flooding. Consequently, the MMP for McCann live il and CO2 was determined as 1300psi. Table 3.2.7 Oil recovery at different pressures (McCann live oil) Pressure (psi) Oil recovery (%) 1100 68.2 1300 96.4 1500 96.8 1700 98.9 17 Figure 3.2.4 The relationship between oil recovery and pressure (McCann live oil) The effects of additives For the target reservoir, the reservoir pressure is lower than the MMP. It has been proposed in the past that the miscibility of CO2 and crude oil can be improved by solubilizing an appropriate additive in the CO2. Two additives from Sasol Company, ME-1 and ME-2, were tested for lowering MMP. The properties of these two additives are listed in Table 3.2.8. They have high flash point, and the density is higher than 0.9 g/cm3. And ME-2 has a higher viscosity than ME-1. The slim tube experiments are conducted with Michigan oil and these two additives mixed with CO2 separately. Table 3.2.8 The properties of additives Flash Point Density @ 20ºC Viscosity @ 20ºC Additives (°C) (g/cm3) (mPa · s) ME-1 133 0.958 12 ME-2 145 0.915 32 Additive ME-1 Four separate slim tube experiments were conducted using ME-1, and their details are listed in Table 3.2.9. Firstly, a 2wt% concentration of ME-1 was combined with CO2 and injected together into the slim tube at 1500 psi. The resulting oil recovery, as indicated in Figure 3.2.5, was 76.4%. In comparison, the oil recovery for the previous experiment utilizing only CO2 at the same pressure was 76.9%, showcasing a minimal difference of merely 0.5%. Subsequently, the experimental pressure was increased to 1750psi. The result is shown in Figure 3.2.6. The oil recovery at 1.2PV was 83.3%, which was lower than the recovery of the case without additive, 86.9%. This suggested that the addition of 2wt% ME-1 did not enhance the miscibility performance at 1500psi and 1750psi. 18 In the next experiment, the ME-1 concentration was increased to 5wt% at 1750psi. The result is also shown in Figure 3.2.6. For this case, the oil recovery was 83.3% at 1.2PV injection, also lower than the previous result without additives. Moreover, exploring a different approach, where 5wt% ME-1 was co-injected with CO2 rather than pre-mixing. The result shows that the oil recovery was even lower at 75.7% compared to the pre-mixed additive experiment at 1750psi. Thess results suggest that the introduction of additive ME-1, weather pre-mixed or co-injected, did not yield an improvement in oil recovery, thus no improvement in miscibility performance between oil and CO2. Table 3.2.9 Slim tube experiments on ME-1 Concentration (wt%) Pressure (psi) Injection mode Mixed in 1 2 1500 advance Mixed in 2 2 1750 advance Mixed in 3 5 1750 advance 4 5 1750 Co-injection Figure3.2.5 The oil recovery at 1500psi with or without additive ME-1 19 Figure 3.2.6 The oil recovery at 1750psi at different situations Additive ME-2 A single slim tube experiment was performed using ME-2 as an additive. 5wt% ME-2 was combined with CO2 and injected into the slim tube to displace oil. The resulting oil recovery, shown in Figure 3.2.7, was 84.7%. There is a 2.2% decrease compared to 86.9% recovery achieved by CO2 flooding without any additives. Hence, the introduction of ME-2 as an additive did not contribute to an increase in oil recovery at the tested pressure of 1750psi. Figure 3.2.7 The oil recovery at 1750psi with or without additive ME-2 3.2.3 Conclusions Slim tube experiments were conducted for reservoir oil with CO2. The following conclusions can be drawn from the results. • The MMP for CO2 with Michigan oil was 2100psi at 40°C. 20 • The MMP for CO2 with Napoleon oil was 1300psi at 40°C. • The MMP for CO2 with McCann live oil was 1300psi at 40°C. • Additives ME-1 and ME-2 cannot contribute to an improvement in miscibility performance between Michigan oil and CO2. 21 Task 3.3: CO2 Core Flood The objective of this task is to assess the efficiency of CO2 flooding in enhancing oil recovery. One composite core was mounted in Hassler sleeves horizontally, and tertiary CO2 floods was conducted at the reservoir temperature and pressure. Subsequently, a 1:1 WAG injection was conducted to determine any enhancement in oil recovery. 3.3.1 Materials For CO2 core flooding experiments, synthetic produced water (SPW) was used as water phase, which has a total salinity of 256,664.9 ppm, with its ion composition detailed in Table 3.3.1. And Napoleon oil was used as oil phase. The core mount comprised four small core samples: #1-5, #6-15, #6-43, and #6-42. Table 3.3.2 describes the properties of the core mount. It had an average permeability of 14 mD and a porosity of 8.6%. Table 3.3.1 Synthetic produced water ion composition Ions (mg/L) Brine 1 Na+ 51,802.1 K+ 5,286.3 Ca2+ 32,697.7 Mg2+ 5,898.2 Cl- 156,207.1 SO42- 4,763.5 TDS 256,664.9 Table 3.3.2 Properties of the core mount for CO2 core floods Properties Core mount: 1-5, 6-51, 6-43, 6-42 Length (cm) 16.23 Diameter (cm) 3.81 Permeability (mD) 14 Porosity (%) 8.6 3.3.2 Methodology To investigate the effectiveness of CO2 flooding for enhanced oil recovery, core flooding experiments were conducted at a pressure of 1200 psi and a temperature of 40°C using the setup shown in Figure 3.3.1. The core samples were horizontally oriented to account for gravity segregation. To prepare the core samples, they were first vacuumed and then saturated with SPW. Crude oil was then injected to establish the condition of irreducible water saturation, after which the core samples were aged at 90°C for two weeks. After aging, water flooding was carried out first by injecting at least 2 pore volume (PV) of SPW at a flow rate of 1 ft/day until no oil was produced from the core samples. This was followed by separate CO2 flooding and WAG flooding tests. For CO2 flooding, 0.5 PV of CO2 22 was injected at a flow rate of 1 ft/day, followed by a chase water flooding (CWF). For WAG flooding, 0.5 PV CO2 and 0.5 PV SPW were injected into the core samples alternatively, with a slug size of 0.05 PV, at a flow rate of 1 ft/day, then followed by a CWF. During the tests, the pressure difference between the two ends of the core samples was monitored, and the effluent from the core samples was collected and measured over time. Figure 3.3.1 Schematic of the set up for CO2 flooding experiments 3.3.3 Results Continuous CO2 flooding Figure 3.3.2 shows the oil recovery, water cut, and relative pressure drop data versus pore volume injected (PVI) for continuous CO2 flooding. The experimental procedure involved 4 PV of SPW, followed by 0.5 PV of CO2, and then another 1.5 PV of SPW injection. During the water flooding, breakthrough of water occurred at 0.16 PVI, and a total of 36.5% of the original oil in place (OOIP) was recovered at the end of water flooding. Subsequently, during the injection of CO2, breakthrough of CO2 was observed at 0.1 PVI of CO2 injection. At the end of experiment, the incremental oil recovery was 14.4% OOIP, which was attributed to the multi- contact miscible behavior of CO2 and oil. Not all the oil was recovered because of the viscous fingering and high-level heterogeneity of the core sample. The relative pressure drop was calculated based on the pressure drip generated by single-phase oil flow at the same flow rate of 1 ft/day. The relative pressure drop during CO2 flooding was found to be small, less than 1psi, indicating that the CO2 tended to bypass the smaller pores in the core samples. 23 Figure 3.3.2 Oil recovery, water cut, and relative pressure drop for CO2 flooding experiment WAG flooding Figure 3.3.3 shows the data on oil recovery, water cut, and relative pressure drop versus pore volume injected (PVI) during WAG flooding. The water flood was followed by a WAG flood with the same amount of CO2. The water flood recovered 42.5% OOIP, hither than that in water flood in CO2 flood. The tertiary WAG flood yielded an additional 20.1% oil, indicating that the WAG flood could improve oil recovery over the continuous CO2 flood in this core mount. Moreover, breakthrough of CO2 was observed at 0.25 PVI of WAG injection, and the relative pressure drop during WAG injection was higher than that of CO2 flooding, which indicated that WAG could help control CO2 mobility. Figure 3.3.3 Oil recovery, water cut, and relative pressure drop for WAG flooding experiment 24 3.3.4 Conclusions Continuous CO2 flooding with 0.5 PV CO2 injection could improve 14.4% OOIP over water flooding. WAG injection could help control CO2 mobility and increase the oil recovery to 20.1% OOIP. 25 Task 3.4: CO2 Foam Flood The objective of this task is to develop proper foam formulation, which can alter wettability of the rock to preferentially water-wet condition. Subsequently, to evaluate the performance of CO2 foam flooding in the reservoir core samples. To achieve this, foam stability tests and contact angle tests were carried out to screen foam agents. After that, CO2 foam flooding experiment was conducted in a core mount using the selected foam formulation. 3.4.1 Materials A nonionic surfactant, Soloterra 843, was provided by Sasol Chemicals. A cationic surfactant, CETAC-30, and a zwitterionic surfactant, LMDO, were obtained from Stepan Company. There are two brines from well SWD and NW1 could be used to make surfactant solution with total salinities of 30,412.6 ppm and 165,573.7 ppm respectively. The ion composition of these two brines is shown in Table 3.4.1. Napoleon oil was also used as oil phase in this task. The core mount used in this study comprised three small core samples: #6-54, #1-45, and #6-48. Table 3.4.2 describes the properties of the core mount. It had an average permeability of 50 mD and a porosity of 11%. Table 3.4.1 Ion composition of two brines used for surfactant solution Ions (mg/L) SWDB NW1B Na+ 6,400 35,000 K+ 520 3,100 Ca2+ 3,700 20,000 Mg2+ 740 3,900 Cl- 19,025.6 103,463.7 SO42- 27 110 TDS 30,412.6 165,573.7 Table 3.4.2 Properties of the core mount for CO2 foam core floods Properties Core mount: 6-54, 1-45, 6-48 Length (cm) 12.16 Diameter (cm) 3.81 Permeability (mD) 50 Porosity (%) 11 3.4.2 Methodology Foam stability test The foam static stability tests were performed at the reservoir pressure (1200psi) and temperature (40°C) to screen the surfactants in two brines. The surfactant concentration was 0.5wt% and the foam quality was 80%. Foam was generated by co-injecting surfactant solution and CO2 through a 7 μm filter at 2ml/min, as shown in Figure 3.4.1. A high-pressure view cell with a height of 13 cm was used to collect the foam and the height of foam was measured over time. Moreover, foam stability tests were carried out in the presence of oil by injecting 1-2 ml of crude oil into the view cell after the foam was collected. The foam stability was characterized 26 by the half-life of foam. Contact angle test Contact angle tests were conducted to identify if the surfactants could alter the wettability of the oil-wet rock to water-wet. The Texas Cream carbonate core chips were used in the tests. First, the core chips were aged in crude oil for two weeks to make them become oil wet. After that, the core chips were merged in brine and surfactant solutions separately. After 1 day and 2 days, the contact angles were measured by putting a drop of oil below the core chips. Foam core flooding experiment To investigate the effectiveness of CO2 foam flooding for enhanced oil recovery, core flooding experiment was conducted at reservoir conditions using the setup shown in Figure 3.4.1. The core samples were horizontally oriented to account for gravity segregation. A high-pressure view cell was used to ensure foam generation before injection into the core samples. The core samples were vacuumed first and saturated with SPW. Crude oil was injected to establish irreducible water saturation, after which the core samples were aged at 90°C for two weeks. Water flooding was conducted first by injecting at least 2 PV of SPW at a flow rate of 1 ft/day until no oil was produced from the core samples. For CO2 foam flooding, 0.625 PV of foam was injected at a flow rate of 1 ft/day, followed by a chase water flooding. The injected volume of CO2 was the same as CO2 and WAG flooding experiments. During the tests, the pressure difference between the two ends of the core mount was monitored, and the effluent from the core samples was collected and measured over time. Figure 3.4.1 Schematic of the set up for CO2 foam flooding experiment 27 3.4.3 Results Foam stability Foam stability without crude oil Three surfactants were used to generate CO2 foam in two brines, and the evaluation of foam over time was presented in Figure 3.4.2 and Figure 3.4.3, respectively. It can be observed that, in the lower salinity condition (SWDB), half-life of foam generated by 0.5wt% Soloterra 843 was 30 minutes (Figure 3.4.2 (a)). However, as the salinity increased (NW1B), the foam exhibited extreme instability, failing to fully occupy the view cell (Figure 3.4.3 (a)). For the foam generated by surfactant CETAC-30, the foam half-life in SWDB was 1.5 hours (Figure 3.4.2 (b)) and increased to 6 hours in NW1B (Figure 3.4.3 (b)). This observation suggests that CETAC-30 can tolerate high salinity, generating more stable foam in high salinity conditions. For foam generated by surfactant LMDO, the foam half-life in SWDB was 30 hours (Figure 3.4.2 (c)) and decreased to 17 hours in NW1B (Figure 3.4.3 (c)). The results indicate that LMDO can generate the strongest foam among these surfactants in absence of crude oil, regardless of brine conditions. Foam stability with crude oil Foam stability tests were performed in the presence of crude oil. The evaluation of foam over time was shown in Figure 3.4.4 and Figure 3.4.5. In the presence of oil, the half-life of foam generated by Soloterra 843 in SWDB was only 8 minutes (Figure 3.4.4 (a)). The case of NW1B was not tested because of unstable foam generated. For the foam generated by surfactant CETAC-30, the foam half-life in the presence of oil was 30 minutes in SWDB (Figure 3.4.4 (b)) and 22 minutes in NW1B (Figure 3.4.5 (a)). The foam generated by LMDO exhibited a similar performance, with a half-life of 30 minutes in SWDB (Figure 3.4.4 (c)) and 25 minutes in NW1B (Figure 3.4.5 (b)). These findings reveal that the presence of oil had a negative impact on foam stability. The difference in half-life time of these surfactants in the presence of oil was not significant. 28 (a) 0.5 wt% Soloterra 843 in SWDB (b) 0.5 wt% CETAC-30 in SWDB (c) 0.5 wt% LMDO in SWDB Figure 3.4.2 The evolution of foam generated in SWDB without oil 29 (a) 0.5 wt% Soloterra 843 in NW1B (b) 0.5 wt% CETAC-30 in NW1B (c) 0.5 wt% LMDO in NW1B Figure 3.4.3 The evolution of foam generated in NW1B without oil 30 (a) 0.5 wt% Soloterra 843 in SWDB (b) 0.5 wt% CETAC-30 in SWDB (c) 0.5 wt% LMDO in SWDB Figure 3.4.4 The evolution of foam generated in SWDB with oil 31 (a) 0.5 wt% CETAC-30 in NW1B (b) 0.5 wt% LMDO in NW1B Figure 3.4.5 The evolution of foam generated in NW1B with oil Contact angle The contact angle tests focused on surfactants CETAC-30 and LMDO, omitting Soloterra 843 due to its limited foam generation capability. Figure 3.4.6 shows the contact angle results observed after 1 day and 2 days. The core chip merged in produced water remained oil-wet even after 2 days, evident from the oil drop spreading on the rock surface. The chip merged in CETAC-30 solution demonstrated a transition: after 1 day, it exhibited an intermediate wet with a contact angle of 90°, evolving into a water-wet with a contact angle of 70° after 2 days. The chip merged in LMDO solution displayed a contact angle of 120° after 2 days, indicating a change in the contact angle but maintaining oil wet. 32 Figure 3.4.6 Contact angle tests results Foam flooding According to the results of foam stability tests and contact angle tests, CETAC-30 was selected as foam agent because of its satisfied foam stability and wettability alteration capability. The CO2 foam core flooding experiment was carried out using 0.5wt% CETAC-30 in SWDB. The experimental procedure involved injecting 4 PV of SPW, followed by 0.625 PV of foam injection, and then another 1 PV of chase water flooding. Figure 3.4.7 presents the oil recovery, water cut, and relative pressure drop data as a function of PVI. Initial water flooding recovered 39.1% OOIP. Subsequently, during the injection of foam, breakthrough of CO2 was observed at 0.38 PVI, which was later compared to CO2 flooding and WAG flooding. At the end of experiment, the incremental oil recovery achieved through foam flooding was 24% OOIP, higher than that of previous two experiments. During the foam flooding, the relative pressure drops exhibited fluctuations with the highest value of 9. This fluctuated pressure drip indicates that some CO2 was pushed into smaller pores and displaced oil. Overall, foam flooding with CETAC-30 can effectively control the CO2 mobility and transfer CO2 into smaller pores, leading to a later CO2 breakthrough and an increased oil recovery compared to CO2 flooding and WAG flooding. 33 Figure 3.4.7 Oil recovery, water cut, and relative pressure drop for CO2 foam flooding experiment 3.4.4 Conclusions Foam stability tests and contact angle tests were conducted to screen foam agents. And CO2 foam core flooding experiment was carried out to evaluate the effectiveness of foam flooding in reservoir core samples. The following conclusions can be drawn from the results: • CETAC-30 can generate better foam and can tolerate high salinity. Also, it can alter the wettability from oil-wet to water-wet. • Soloterra 843 is not a good foam agent at reservoir conditions and performed worse at high salinity. • LMDO can generate the most stable foam among these three surfactants. However, it cannot alter the rock wettability. • CO2- foam core flooding by using surfactant CETAC-30 can increase oil recovery by 24% OOIP over the water flooding. This is 9.6% higher oil recovery compared with CO2 flooding and 4% additional oil recovery over the WAG flooding. 34 Task 3.5: Lab-Scale Compositional Modeling 3.5.1 Methodology Slim tube experiments were simulated by the CMG-GEM simulator to investigate the effect of introducing rich gas in CO2 on MMP. The model parameters are shown in Table 3.5.1. The model size is the same as the slim tube size, and 40 grids were set in horizontal direction. The rich gas composition is shown in Table 3.5.2. Both live oil and dead oil were considered in the simulation and the oil compositions are shown in Table 3.5.3. Table 3.5.1 Slim tube model parameters Grid 40 × 1 × 1 Size (ft × ft × ft) 1 × 0.01345 × 0.01345 Porosity 0.439 Permeability (D) 6 Temperature (°C) 40 Table 3.5.2 Rich gas composition Component Rich gas (mol%) N2 0.0465 CO2 0.0121 CH4 2.4219 C2H6 10.8444 C3H8 28.4433 IC4 6.7034 NC4 21.6646 IC5 5.4296 NC5 10.1573 C6+ 14.2769 Table 3.5.3 Oil composition used in simulation Component Live oil (mol%) Dead oil (mol%) N2 2.802 0 CO2 0.092 0 CH4 24.239 0 C2H6 9.529 0 C3H8 8.491 0 IC4 1.460 0 NC4 4.529 0 IC5 1.335 1.625 NC5 2.789 3.863 C6 3.524 6.643 35 C7+ 41.209 87.87 3.5.2 Results Slim tube simulation of live oil Slim tube simulations were carried out using live oil and pure CO2 at several different pressures, and the resulting oil recovery factors are shown in Figure 3.5.1. From the simulation results, the MMP between live oil and CO2 was identified as 1444psi, the intersection where live oil and CO2 became miscible. To investigate the effects of introducing rich gas in CO2, separate simulations were conducted using 10mol% and 20mol% rich gas combined with CO2, and the corresponding resulting oil recovery factors are shown in Figure 3.5.2 and Figure 3.5.3. Introducing 10mol% rich gas led to an MMP decrease to 1200psi, while a higher concentration of 20mol% rich gas resulted in a further reduction to 977psi. These observations indicate that the introduction of rich gas in CO2 can help decrease the MMP of live oil and higher mole content of rich gas results in lower MMP. Figure 3.5.1 The slim tube simulation results for live oil and pure CO2 Figure 3.5.2 The slim tube simulation results for live oil and 10mol% rich gas in CO2 36 Figure 3.5.3 The slim tube simulation results for live oil and 20mol% rich gas in CO2 Slim tube simulation of dead oil Slim tube simulations were carried out using dead oil and pure CO2 at several different pressures, and the resulting oil recovery factors are shown in Figure 3.5.4. The MMP between dead oil and CO2 was identified as 1373psi. To investigate the effects of introducing rich gas in CO2, separate simulations were conducted using 10mol% and 20mol% rich gas combined with CO2, and the corresponding resulting oil recovery factors are shown in Figure 3.5.5 and Figure 3.5.6. Introducing 10mol% rich gas led to an MMP decrease to 1190psi, while a higher concentration of 20mol% rich gas resulted in a further reduction to 837psi. These observations indicate that the introduction of rich gas in CO2 can help decrease the MMP of dead oil and higher mole content of rich gas results in lower MMP. Figure 3.5.4 The slim tube simulation results for dead oil and pure CO2 37 Figure 3.5.5 The slim tube simulation results for dead oil and 10mol% rich gas in CO2 Figure 3.5.6 The slim tube simulation results for dead oil and 20mol% rich gas in CO2 3.5.3 Conclusions Compositional simulations showed that introduction of 10%-20% rich gas can significantly decrease MMP between crude oil and CO2. 38 3.6 Overall Conclusions Task 3 of the project was conducted at The University of Texas at Austin. It consisted of core characterization, MMP determination, CO2 flood, foam flood and compositional simulation. • CT scanning showed that the cores are heterogeneous and have a complex pore geometry. Mercury porosimetry showed a dual pore size distribution from nanometers to 100 microns. The relative permeability of oil can be fitted to a Corey expression, but not the water relative permeability. These cores are oil-wet. • Slim tube experiments were conducted for reservoir oil with CO2. The MMP for CO2 with Michigan oil was 2100psi at 40°C. The MMP for CO2 with Napoleon oil was 1300psi at 40°C. The MMP for CO2 with McCann live oil was 1300psi at 40°C. Additives ME-1 and ME-2 cannot contribute to an improvement in miscibility performance between Michigan oil and CO2. • Continuous CO2 flooding with 0.5 PV CO2 injection could improve 14.4% OOIP over water flooding. WAG injection could help control CO2 mobility and increase the oil recovery to 20.1% OOIP. • Foam stability tests and contact angle tests were conducted to screen foam agents. And CO2 foam core flooding experiment was carried out to evaluate the effectiveness of foam flooding in reservoir core samples. CETAC-30 can generate better foam and can tolerate high salinity. Also, it can alter the wettability from oil-wet to water-wet. Soloterra 843 is not a good foaming agent at reservoir conditions and performed worse at high salinity. LMDO can generate the most stable foam among these three surfactants. However, it cannot alter the rock wettability. • CO2-foam core flooding by using surfactant CETAC-30 can increase oil recovery by 24% OOIP over the water flooding. This is 9.6% higher oil recovery compared with CO2 flooding and 4% additional oil recovery over the WAG flooding. • Compositional simulations showed that introduction of 10%-20% rich gas can significantly decrease MMP between crude oil and CO2. 39 Attachment C: UT Austin Bureau of Economic Geology Seismic Interpretations of Fractured Trenton-Black River Carbonate in Southern Michigan Full report of seismic analysis and interpretation is included as Attachment C to this final report. This work was completed by Dr. Shuvajit Bhattacharya at the University of Texas Bureau of Economic Geology. The primary objective of this project was to improve the understanding of geologic complexity and reservoir properties in the Trenton and Black River (TBR) formations in the Michigan Basin. The team’s methodology centered on integrated geophysical analysis using 3D seismic datasets, well logs, and seismic inversion techniques. 3D seismic datasets from the Albion- Scipio field and Lee 26 field were interpreted to analyze fault distribution, fracture styles, and reservoir geometry. Conventional well logs were used to generate synthetic seismograms for horizon recognition, and post-stack model-based seismic inversion was applied to estimate acoustic impedance and porosity. The study also included a reconnaissance core analysis at the Michigan Geological Repository for Research and Education (MGRRE), and filtering techniques were used to correct porosity estimates in the Utica Shale and the TBR section. Our results demonstrate that 3D seismic may be able to improve property mapping accuracy in complex and compartmentalized reservoirs, as 2D seismic cannot capture lateral variability or localized fault and fracture distribution. Porosity inversion revealed that reservoir-quality HTDs in the TBR have lower acoustic impedance and higher porosity near fault and fracture zones, confirming the match between inversion-based and well-log-based measurements. The team concluded that precise seismic processing and migration are necessary to identify key geologic features, and that integrated geophysical approaches provide valuable insights for CO2 utilization and storage in Trenton-Black River hydrothermal dolomite reservoirs. Battelle | May 2026 264 Final report on the project DE- FE0031792 on tasks related to “Seismic interpretations of fractured Trenton-Black River carbonate in Southern Michigan” Shuvajit Bhattacharya Bureau of Economic Geology The University of Texas at Austin Introduction Structurally controlled hydrothermal dolomite (HTD) oftentimes have a good reservoir quality, which makes them a potential target for CCUS. HTD reservoirs present interesting technical challenges and opportunities for exploration and development. These reservoirs are commonly found in Michigan, Texas, western Canada, and other parts of the world (Hart et al., 2009; Saller, 2013; Hurley and Budros, 1990; Jiang et al., 2017). There are a few published studies on the TBR HTDs in the Michigan Basin (Gregg and Sibley, 1984; Hurley and Budros, 1990; Budai and Wilson, 1991; Grammer and Harrison, 2013; Dunseith et al., 2018; Bhattacharya et al., 2022; Haagsma et al., 2022; Bhattacharya et al., 2023). Sags associated with HTD reservoirs were recognized in the 1930s, for the Ordovician Dover (Trenton) field in southwestern Ontario (Burgess, 1960, reported in Hurley and Budros, 1990; Davies and Smith, 2006). The Dover field is considered as a structurally controlled HTD reservoir, with high correlation of productive wells to sags at the top Trenton level (Davies and Smith, 2006). Although structural sags as well as amplitude dimming associated with structurally controlled HTD reservoirs can be identified on 2D seismic sections, 3D seismic technology can greatly improve defining and mapping the full geometry of these complex and compartmentalized reservoirs. We present the results from integrated 3D seismic-assisted geophysical interpretations of the Trenton and Black River (TBR) formations from two oil fields (Albion-Scipio and Lee) in the Michigan Basin, United States (Figure 1). As the injection well drilled within the Albion-Scipio field was not suitable for the project, the Lee field was selected as the next field site for further experiments and interpretations. We interpret two 3D seismic datasets from each field to analyze fault distribution and styles, and integrate it with the image log and deep shear log-based analysis and acoustic impedance from seismic inversion. 1 Figure 1: Approximate locations of the TBR fields in Southern Michigan (modified after Haagsma et al., 2022). Lessons learned • 2D seismic cannot capture the lateral variability of complex HTD reservoirs, localized fracture, and fault distribution. Therefore, 3D seismic is critical to better understand the geologic complexity and map the reservoirs. • Seismic data needs to be processed and migrated accurately near sags associated with negative flower structures, carbonate dissolution and diagenesis. Otherwise, these features would be missed on the stack image. • Sparsely-distributed wells with vintage logs are helpful at the exploration stage; however, significant processing needs to be done and compare them with modern logging data for accurate reservoir characterization. • Acquisition, comparison, and integration of core, image logs, and far-field multi- component sonic logs (e.g., deep shear) in the same borehole are helpful to interpret, validate, and map sub-seismic-scale fractures. Many fractures (excluding tensile fractures) present at the core scale are not always observed on both image logs and deep shear logs and vice versa. • Operational challenges may hinder collecting good-quality data, for example, some of the multi-component acoustic tools (e.g., deep shear) require the borehole to be filled with fluid, which may not always be possible and adds to the operational costs and time. • Access to both pre-stack and post-stack seismic is critical for further investigation of reservoir complexities. This also helps if there is a need for reprocessing the dataset to remove scattered noise from glacial boulders, unwanted noise and multiples from carbonates, etc. • Machine learning models utilizing seismic data for rock property predictions do not always work in in complex geology, example, fractured carbonates with little velocity 2 variations. Better quality input data (e.g., 3D seismic) with high signal-to-noise ratio and reprocessing of vintage seismic are needed before deploying machine learning tools. In such cases, physics-based seismic inversion is the recommended approach. 3D seismic dataset features We used two 3D seismic datasets from the Albion-Scipio and Lee fields. The general 3D acquisition features are shown in Table 1. Table 1: Features of two 3D seismic surveys used in the study Features 3D seismic in Albion- 3D seismic in Lee field Scipio field Area ~12 sq. miles ~ 8 sq. miles Number of Inlines 687 240 Number of 579 288 crosslines Inline interval 55 ft 82.5 ft Crossline interval 55 ft 82.5 ft 3D seismic attributes and seismic inversion We used several 3D seismic attributes and seismic inversion process to better understand the reservoir complexity of the TBR interval. Seismic attributes are mathematical quantities derived from seismic data using a set of equations. There are several types of seismic attributes, including complex trace attributes and geometric attributes. RMS amplitude, instantaneous phase, and frequency attributes are complex trace attributes. These attributes are useful to visualize the change in reflection characteristics in the study area. The changes could be related to lithology, porosity, etc. The geometric and waveform attributes, such as curvature and variance can reveal the structural (e.g., faults) and stratigraphic complexity (e.g., channels). In addition, ant track can be computed to automatically extract faults from seismic data. The definitions and applications of a few widely used attributes are below: RMS amplitude attribute: The root mean square amplitude (RMS) is a commonly used technique to display amplitude values in a window of interest. With this attribute amplitude, hydrocarbon indicators can be mapped by measuring reflectivity. Instantaneous frequency attribute: It is the time rate of change of the instantaneous phase. This attribute is a good indicator of low impedance beds, fracture zones, hydrocarbon accumulations, etc. Variance attribute: It measures the similarity of waveforms or traces adjacent over given lateral and/or vertical windows. Therefore, it can image discontinuity of seismic data related faulting or stratigraphy. For more information on seismic attributes, please see Chopra and Marfurt (2007). 3 Apart from seismic attributes, we performed seismic inversion. Seismic inversion is a process of inverting the seismic amplitude traces to acoustic impedance. Acoustic impedance is more characteristic of the changes in rock properties than regular seismic amplitude. Essentially, it is the multiplication of acoustic velocity and density of the rocks. Inversion process requires multiple iterations to reduce the error between the original data and synthetic data. In terms of seismic inversion, we used a model-based post-stack inversion process. Workflows for such inversion are shown in Figure 2 and 3. Figure 2: Basics of seismic inversion process (after Maurya and Singh, 2018) Figure 3: Workflow on impedance estimates from seismic inversion (Karim et al., 2016) 4 Albion-Scipion dataset analysis We use a 3D seismic dataset and well log data from four wells in the Albion-Scipio Field. Several well logs, including triple-combo, dipole sonic, image logs, and deep shear wave logs, were collected from one new well, which was drilled as a part of the DOE project. We also conducted a reconnaissance Trenton-Black River core study at the Michigan Geological Repository for Research and Education (MGRRE). We use one of the wells for synthetic seismogram and recognizing formations on seismic data (Figure 4). Conventional well logs from a well were used to generate a synthetic seismogram to recognize the geologic horizons on seismic data. Ormsby filter (10/15-50/70-70/90 Hz) was designed and convoluted with reflectivity derived from the density and sonic logs to generate the synthetic seismogram. In this 3D seismic dataset, no phase rotation was applied. The synthetic seismogram shows that the Trenton has the strongest positive peak, which is the same on the seismic data. Trenton reflector is the marker here, which was used as a reference to pick the overlying Utica and underlying Black River and Glenwood. We interpret seismic amplitude sections (Figure 5) and map Utica, Trenton, and Black River surfaces and faults. Figure 3 shows complex fault styles affecting the TBR section. Sags are visible near the top of the Trenton formed due to a combination of various factors, including the negative-flower structures, carbonate dissolution, and calcite-to-dolomite diagenesis. After conducting dip-steering of the seismic cube, we compute seismic attributes, such as RMS amplitude, instantaneous frequency, coherence, and ant track to automatically extract faults from 3D seismic (Figure 6 and 7). We utilize regional geologic knowledge about fault distribution to guide the ant track process for geologically meaningful automated fault mapping (Bhattacharya et al., 2022). We convert the ant-tracked volume to fault patches and fault planes. Figure 4: An example of synthetic seismogram to recognize formations on seismic data. Trenton top has a peak, and Utica top has a trough. 5 Figure 5: Interpreted seismic section showing deep-seated faults (Bhattacharya et al., 2022). Note the Trenton Formation top has a strong peak on seismic reflection (in white), whereas the Utica Formation top has generally negative peak (trough) characteristic (in black). Figure 6: An example of computed seismic attributes 6 Figure 7: Map view of ant track showing the distribution of faults in en-echelon and Riedel shear pattern (Bhattacharya et al., 2022) Our current structural analysis of the Albion-Scipio field shows the presence of several basement- rooted deep sub-vertical faults, some of which also affected the TBR section (Bhattacharya et al., 2022). Many of these faults are en-echelon in nature, oriented along NW-SE, punctuated by breaks (e.g., Pulaski break), forming compartmentalized reservoirs (Figure 7). In such cases, CO2 may not go over the barrier. To validate seismic interpretations, we also corroborated complementary information from image logs and core. The interpreted image logs show many fractures oriented along NW-SE, with second-order fractures oriented along W-E (Bhattacharya et al., 2023). Our core observations at the MGRRE facility validated the presence of some of these fractures in the HTD TBR play (Figure 8). 7 Figure 8: An interpreted core and image log showing sub-vertical fractures in the TBR section (Bhattacharya et al., 2023) We also analyzed the available deep shear image log data (Figure 9). The log was acquired in an open-hole condition. Interpretations of the log indicate that some high-angle features (potential fractures and faults) are present around 120 ft away from the borehole. We computed attributes, such as coherence and amplitude-to-frequency ratio from the data, showing fracture branching out and complexities. Figure 9: An example of interpreted deep shear image logs, showing high-angle features around 120 ft away from the borehole (Bhattacharya et al., 2023). Dark high-angle bands are indicative of fractures. We conduct a post-stack model-based seismic inversion. Seismic inversion results show that the reservoir-quality HTDs in the TBR have lower acoustic impedance than non-reservoir limestones 8 (Figure 10). Porosity is higher towards the fault and fracture zones, where HTDs are present. The top portion of the Trenton Formation commonly has very low porosity, but reservoir-quality HTD can occur over the entire Trenton-Black River section, especially where negative flower structures and fracture zones are present. Figure 10: P-impedance (a) and porosity (b) from seismic inversion (Bhattacharya et al., 2022). There is a well in the cross-section that was used as blind test. Inversion-based impedance and porosity match with well-log-based impedance and porosity. Impedance curves in the well and seismic have the same scale and color pattern for visual comparison. Figure 11: Comparison between well-log-based impedance and seismic inversion-based impedance, showing overall good match. Lee field dataset analysis 9 We use another 3D seismic dataset and two wells in the Lee Field for seismic interpretation and inversion work. The Lee field seismic had poorer resolution and only post-stack data available, compared to the Albion-Scipion field. There are some small-scale faults present, but many are sub-seismic scale. We interpreted the Utica, Trenton, Black River, and Glenwood surfaces on the seismic section (Figure 12) and performed QA/QC using various interpolation approaches (Figure 13). This allowed us to fill up the holes in the interpretations, especially near complex areas. Figure 12: Interpreted seismic section showing interpreted surfaces. Note the Trenton Formation top has a strong peak on seismic reflection (in black), whereas the Utica Formation top has a generally negative peak (trough) characteristic (in white). 10 Figure 13: An example of the Utica surface interpolation and clean up. The first image on the left shows the map before processing, and the one on the right shows the map after processing. We used the QA/QC’ed surfaces and well data for generating a low-frequency model of the subsurface. Then, we performed a post-stack seismic inversion to derive P-impedance (Figure 14 and 15). Figure 16 shows the relationship between inversion-based impedance and log-based impedance. We used one well for training and another for blind test in the inversion process. We used the relationship between well-log-based porosity and P-impedance to convert the seismic- derived impedance to porosity (Figure 17). 11 Figure 14: An example of a single well-based inversion. The blue curve in the 1st track shows original impedance, and the red curve shows the seismic-inverted impedance. The error, the difference between the seismic trace and synthetic, is minimal (~0.07). We applied a few filters to produce the final porosity volume estimated from seismic inversion in the Lee field. Because the porosity log curves provided did not correct for clay boundwater effect in the Utica Shale, it was showing high porosity of >10-16 p.u. at places (which is wrong), whereas Trenton shows negligible porosity at places. Most importantly, we parameterized the TBR section in the seismic inversion process, not the Utica and overlying section, which is out of scope for this study. Therefore, we applied a filter in the Utica and above to address its apparently high porosities relative to the TBR section. This will help better visualize reservoir property variation in the TBR section. 12 Figure 15: P-impedance from seismic inversion in the Lee field. There are two wells inserted in the cross-section. Inversion-based impedance matches with well-log-based impedance. Impedance curves in both wells and seismic have the same scale and color pattern for visual comparison. Figure 16: Comparison between well-log-based impedance and seismic inversion-based impedance, showing overall good match. 13 Figure 17: Estimated porosity from seismic inversion in the Lee field. We also explored the possibility of porosity prediction from multiple seismic attributes using a probabilistic neural network (PNN). We used a combination of multiple seismic attributes as inputs to the PNN model, such as average frequency, coherence, and seismic data decomposed in different frequency bands, as well as inversion-based P-impedance to predict porosity (target). However, the results from PNN-based porosity prediction were poor after several iterations, with a correlation coefficient was 0.63. Our subsurface characterization workflow revealed the structural and reservoir complexities in the TBR in both fields, which are critical to map sweetspots for CCUS. Seismic inversion-based impedance highlighted the potential reservoirs. We anticipate that our workflow and lessons learned will lay the groundwork for future oil and gas and CCUS-related work in these HTD reservoirs in Michigan. In addition to this report, P-impedance and porosity volumes for the Lee field have been shared with Battelle in .segy format.