Spatially Resolved Greenhouse Gas Emissions from U.S. Dairy Products Authors: Rylie E. O. Pelton, Fabian Bernal, Tim Kurt Published: 2026 ABSTRACT As the world’s second-largest milk producer, the U.S. dairy industry aims to achieve net GHG neutrality for all dairy products by 2050. Meeting this target requires understanding emissions across the dairy supply chain, extending beyond farms to include full processing activities. While previous Life Cycle Assessments (LCAs) evaluated individual products like fluid milk and cheese, no comprehensive cradle-to-processing gate LCA exists for the industry. Moreover, prior LCAs often allocated facility-level emissions across multiple products, overlooking variations in energy and materials among coproducts. This study addresses these gaps, quantifying total cradle-to-processing gate emissions, emissions per kilogram of product, and product emissions per kilogram of fat and protein-corrected milk (FPCM) across 12 regions covering 99.96% of U.S. dairy production. In 2020, total emissions were 163.3 million tonnes CO2e, with raw milk production accounting for an average of 81%. Postfarm gate emissions varied regionally due to differences in product mix, energy use, electricity grid composition, and transportation distances. Results identify raw milk production as the largest emissions source across the evaluated uncertainty range, indicating that mitigation will require coordinated action across both farm and processing stages. Findings provide critical insights for policymakers and stakeholders to effectively prioritize strategies such as energy efficiency improvements and renewable energy adoption in future sustainability initiatives. Introduction The United States (U.S.) is the world’s second-largest producer of milk and dairy products with a complex supply chain that spans raw milk production, distribution, and processing, across diverse geographic regions. As part of broader efforts to address climate change, the U.S. dairy industry has set a goal of achieving net greenhouse gas (GHG) emissions neutrality across dairy products by 2050. Meeting this goal requires a robust, transparent, and spatially resolved understanding of GHG emissions across the dairy supply chain, extending beyond the farm gate to encompass the entire dairy production process. Establishing such a baseline is a necessary prerequisite for evaluating mitigation strategies, tracking progress over time, and ensuring consistency in emissions accounting across products and regions. While previous Life Cycle Assessments (LCAs) have been conducted for specific dairy products such as fluid milk, cheese, and butter, − there has been no comprehensive LCA estimating the cradle-to-processing gate GHG emissions for the entire U.S. dairy products industry. Furthermore, many existing LCAs assess emissions at the facility level, allocating impacts across products without considering differences in processing. This approach may misrepresent actual emissions, as energy consumption varies significantly across different dairy products and steps of production, and facility-level allocation fails to capture these differences accurately. Therefore, this study has three primary objectives. First, it aims to establish an internally consistent, unified cradle-to-processing gate baseline of U.S. dairy industry GHG emissions for the year 2020, following ISO LCA principles, across 12 distinct dairy product processing regions (Figures and S1 in Supporting Information–SI). While prior studies have produced robust estimates for individual dairy products or segments of the supply chain, no previous assessment has provided a unified, product-resolved baseline designed to support industry-wide tracking of progress toward net GHG neutrality targets. The analysis encompasses nearly all dairy products produced in the U.S., accounting for 99.96% of total production (as-sold by mass). To improve product-level attribution and comparability, life cycle inventories are delineated at the unit-process level rather than allocated at the facility level, enabling more precise representation of energy and material inputs across coproducts. Second, the study seeks to estimate GHG emissions per kilogram of product and per kilogram of fat and protein corrected milk (FPCM), providing complementary metrics for downstream use and cross-product comparisons. Finally, this research aims to identify GHG emission hotspots across the supply chain and specific dairy products, helping to prioritize climate action strategies within the industry. System boundary for U.S. Dairy Cradle-to-Processing gate LCA (see SI Figures S2–S19). Abbreviations: WPC = whey protein concentrate, WPI = whey protein isolate, MPC = milk protein concentrate, WPPC = whey protein phospholipid concentrate. This study builds on recent advances in quantifying emissions from enteric fermentation, manure management, and feed production across 12 U.S. dairy production regions, integrating these upstream inventories into a harmonized, spatially explicit cradle-to-processing gate framework. By extending the system boundary beyond the farm gate to include raw milk transport, primary processing, secondary transport, secondary processing, and primary packaging (see Figure ), this analysis links milk production emissions to product- and region-specific processing pathways. The resulting framework provides a consistent baseline of GHG emissions across nearly all U.S. dairy products, designed to support industry benchmarking, policy evaluation, and assessment of mitigation opportunities toward the U.S. dairy sector’s net neutral GHG target by 2050. This study advances prior approaches by systematically integrating three methodological refinements that, to our knowledge, have not previously been implemented together at industrial scale. First, we implement a unit-process LCA across all dairy products, delineating emissions by specific production steps rather than allocating facility-level impacts, enabling precise attribution of energy and material inputs to individual coproducts. Second, we adopt IDF (2022) conditional allocation guidance that applies economic allocation to feed-directed coproducts while using milk-solids allocation for food and industrial applications, weighted by market-specific feed shares. This approach better reflects the distinct value propositions and end-use markets of dairy coproducts. Third, we integrate spatially explicit supply chain modeling with regional energy grids and processing infrastructure, capturing geographic heterogeneity in emissions that national-level assessments overlook. These methodological refinements yield regionalized product-level emission estimates that improve product-level attribution and consistency, and are more actionable for industry sustainability planning. Materials and Methods Building on the cradle-to-farm gate assessment, the scope of this analysis encompasses cradle-to-processing gate, with a focus on the postfarm gate, including raw milk transport, primary processing, secondary transport, secondary processing, and primary packaging. Life cycle emissions from the production of material and energy inputs, including additional ingredients used in product formulation, such a sugar, eggs, starch, salt, etc., and the transportation of such inputs are also included in the system boundary. Additives, however, such as pectin, fruit, honey, essential oils, extracts, are excluded from the analysis. These inputs constitute a minor fraction of total material flows and vary widely across and within dairy product categories, and data describing their use are limited. Consistent with ISO 14044 cutoff principles and prior dairy LCAs, such minor inputs are therefore excluded, as they are expected to contribute only a negligible share of total product GHG emissions relative to milk inputs and processing energy. Given the goals of this study, we focus on characterizing emissions from the cradle-to-processing gate, constituting around 90% of the total cradle-to-grave emissions for most dairy products, , and embodying the phases of dairy products’ life cycle most directly influenced by the dairy industry. The geographic scope covers 12 distinct regions of dairy product production, mirroring those analyzed in the cradle-to-farm gate study. The analysis thus captures regional variability in production practices, transport distances, and processing methods, ensuring a comprehensive understanding of emissions across the U.S. dairy sector. We evaluate 25 unique dairy products across 36 subcategories (Table ) using a unit-process approach that assigns burdens directly to processes whenever feasible, minimizing the need for allocation. Where allocation is required, it is applied at the unit-process levelthat is, at each individual coproduct separation point within the processing chainrather than at the facility level (see Figures S2–S19). For unit processes yielding multiple coproducts, allocation follows IDF (2022) guidance which is conditional on end use. Specifically, for coproduct sets in which any output is destined for animal feed, we apply economic allocation to that set; for coproducts used exclusively for nonfeed purposes (e.g., food, nutraceutical, industrial), we apply milk-solids allocation based on the relative milk solids content of each product (SI Table S1). The share of each product category that is routed to animal feed is taken from the American Dairy Products Institute (ADPI) market statistics. The conditional allocation approach is consistent with broader international guidance for livestock supply chain assessment, including the FAO LEAP framework, particularly with respect to the treatment of animal feed. When dairy coproducts are directed to animal feed, applying economic allocation ensures conceptual alignment with how feed-related burdens are accounted for in upstream cradle-to-farm gate LCAs, where relative economic value is recommended to allocate impacts across feed ingredients. In contrast, for dairy coproducts intended for food, nutraceutical, or industrial applications, the IDF recommends milk-solids allocation, reflecting a preference for allocation based on stable physical properties rather than market prices which can vary substantially across regions and over time. Using a physical basis for allocation is also consistent with the allocation hierarchy defined in ISO 14044, and enhances the stability and comparability of product-level emission estimates, thereby facilitating benchmarking against past and future studies while minimizing sensitivity to short-term price fluctuations. For coproduct systems subject to conditional allocation, product-level emission intensities are calculated as a feed-share-weighted average of economic and milk-solids allocation results (eqs S1–S2). Figure S23 illustrates the allocation approach for a cheese-whey coproduct system for the economic and milk-solids allocation approaches. 1Dairy Products and Co-Products Included within the LCA System Boundary Primary ProductsCoproductsFluid milkCreamWhole milk (3.25%)Reduced-fat milk (2%)Low-fat milk (1%)Skim milkDairy BeverageButtermilkN/APuddings/CustardsN/ACheeseWhey PowderWhey CreamCottage CheeseN/AYogurtWhey AcidSour CreamN/AWhey Protein Concentrate (WPC)Whey Permeate25–49.9% protein25–49.9% protein removed50–89.9% protein50–89.9% protein removed LactoseFrom WPC25–49.9% permeatesFrom WPC50–89.9% permeatesWhey Protein Isolate (WPI)Whey Protein Phospholipid Concentrate (WPPC)Reduced lactose whey (RLW) powderLactoseButterButtermilk powderEvaporated MilkN/ASkimWholeCondensed MilkN/ASkimWholeMilk PowderN/ASkimWholeMilk Protein Concentrate (MPC)Permeate SolidsIce CreamN/AaAllocation defaults to milk-solids allocation. For co-product sets in which one co-product is used as animal feed, as shown in bold, we apply economic allocation to the feed-directed share and retain milk-solids allocation for the share used in other applications (food, nutraceutical, industrial), consistent with IDF (2022) (see SI eqs S1 and S2, Figure S23, and Table S1). Allocation factors are implemented at the unit process level, as such, allocation factors vary across intermediate co-product sets and processing stage, and are represented in the corresponding process-flow diagrams (Figures S2–S19). Dairy Products and Co-Products Included within the LCA System Boundary Allocation defaults to milk-solids allocation. For co-product sets in which one co-product is used as animal feed, as shown in bold, we apply economic allocation to the feed-directed share and retain milk-solids allocation for the share used in other applications (food, nutraceutical, industrial), consistent with IDF (2022) (see SI eqs S1 and S2, Figure S23, and Table S1). Allocation factors are implemented at the unit process level, as such, allocation factors vary across intermediate co-product sets and processing stage, and are represented in the corresponding process-flow diagrams (Figures S2–S19). Figures S2–S19 in the Supporting Information (SI) show the different distinct unit processes considered across dairy products, the respective allocation factors across coproducts, and feed-share assumptions applied across product systems. Emissions are quantified on both a per kilogram of product basis and a per kilogram of fat and protein corrected milk (FPCM) input basis. This dual approach allows for the emissions intensity per kg of product to be used directly for characterizing impacts in downstream industries, while the emission intensity per kg FPCM enables comparison in impacts across dairy products on a nutritionally equivalent basis (Table S2). We estimate the emission intensities for the feed-bound coproduct sets under economic allocation and the emission intensities for all other uses under milk-solids allocation (Figures S2, S4, S6–S8, S16, and S19) and present the weighted average of these two results using the feed share. Emissions are based on the carbon dioxide equivalents (CO2e) considering the Intergovernmental Panel on Climate Change (IPCC) sixth assessment report (AR6), which represents the latest available scientific consensus on climate forcing. The AR6 assigns a 100-year GWP (GWP100) of 27.2 to biogenic methane, 29.8 to fossil methane, and 273 to nitrous oxide. The following sections provide details on total dairy product production across regions and estimation of GHG emissions across each stage in the farm-gate to processing-gate phases of the life cycle. Total Dairy Products Production The cradle-to-processing gate analysis includes 99.96% of all dairy products produced in the U.S. in 2020, measured on a wet mass (as-produced) basis. We exclude condensed sweet whey solids due to negligible production quantities. Figure illustrates the total national quantities produced for each dairy product. Among fluid milk products, sales distribution by mass indicates that 41% is whole milk, 38% reduced fat milk, 14% low-fat milk, and 7% skim milk. Production volumes for pudding and custard were estimated from the total number of packages sold and average package weights. We assume a 50:50 split between ready to eat (canned) pudding and instant pudding. For instant products, household preparation requires adding fluid milk; consistent with our system boundaries, emissions from the added milk are inventoried under the “fluid milk” category and would be attributed to the pudding product in its use phase. The production of cream, a significant coproduct of various dairy processes (e.g., fluid milk, milk powder, condensed milk, and evaporated milk), was calculated by subtracting the quantity of cream utilized in downstream products like cottage cheese, butter, ice cream, and sour cream from the total cream produced as coproducts of producing other dairy products. The remaining cream was attributed to direct cream sales, capturing the balance of cream production in the industry. Total million tonnes (MMT) of dairy products produced in 2020 for final consumption, with products arranged from left to right in descending order by total quantity sold (by weight). Fluid milk includes 41% whole milk, 38% reduced-fat (2%), 14% low-fat (1%), and 7% skim based on weight of product sold. All products estimated are based on USDA data. , Cream and pudding/custard are estimated; only portion of pudding/custards canned are included in quantities, see text for details. Note that condensed sweet whey solids are excluded from analysis, representing a total of 0.04% of total dairy products. The location of dairy product processing plays a significant role in determining total emissions, as regional differences in electricity grid mixes, transportation distances, and milk or cream sourcing practices directly influence a product’s environmental footprint. For example, electricity grid emission intensity across the 12 dairy processing regions ranges from approximately 270 to 564 kgCO2e/MWh, reflecting regional variation in generation mixes. To account for these regional variations, we estimated production quantities for each dairy product across each region in the U.S. This was based on a hybrid approach, combining USDA data, which identifies the primary states for producing various dairy products, with anonymized aggregated industry data voluntarily provided by dairy processors through the Innovation Center for U.S. Dairy [Processor] Stewardship Commitment, covering more than 75% of total U.S. dairy milk processing. This method allowed us to fill critical data gaps across both regions and product types, ensuring comprehensive coverage of all dairy products included in the analysis. However, for certain product categories, regional production data was unavailable from both USDA and industry sources. In these cases, proxy data was used to estimate regional distributions. For example, the regional distribution for fluid milk production served as a proxy for buttermilk, milk powders’ regional production was used for milk protein concentrates (MPC), liquid permeate products were used for permeate solids and whey permeate, and liquid whey concentrate was used for whey protein concentrates (WPC). Figure S25 presents the final estimated distribution of dairy product production across the 12 defined regions, providing a detailed breakdown of where each product is estimated to be predominantly processed. Raw Milk Transport and Sourcing Raw milk transport and sourcing are estimated with the FoodS linear optimization model which estimates the supply chain of commodity flows, − including milk transported from farms to primary processing facilities. The model links county locations of milk supplies to facility-scale milk demand from primary milk and cheese processors across the U.S., by minimizing the total system impedance. The model provides one-way origin-destination (OD) distances for each supply demand pair. For comparability, we report a round-trip equivalent distance (2× OD distance), where the supply weighted average is estimated at 340 km, with a range of 172–1,223 km across the 12 processing regions, with longer hauls in regions with sparse supply (e.g., the Northern Great Plains). Ulrich et al. report higher round-trip distances, with an average of 800 km, and range of 745–2,477 km for individual truck trips (representing about 17% of the market). However, because the Ulrich et al. analysis excludes load-sharing trips, a common industry practice, these estimates may be skewed toward the higher-end. For emissions, we assume milk is transported in insulated tank trucks and use LCA for Experts (LCA FE) freight data sets, assuming 47,000 lbs. payload. Because these kgkm emission factors already incorporate average load factors and empty/partially empty return behavior, with impacts expressed per unit of goods transported, we calculated transport emissions as milk mass multiplied by the one-way OD distance and the emission factor. This avoids double counting return-leg emissions that are already embedded in the factor. The distribution of sourcing regions associated with each milk processing region is shown in SI Figure S20. This distribution of sourcing regions is combined with cradle-to-farm gate milk production emission factors to estimate the weighted average emission factors from sourcing raw milk in each processing region. Primary Processing Primary milk processing involves a sequence of steps to convert raw milk into cream and various fluid milk products such as whole milk, reduced-fat milk (2%), low-fat milk (1%), and skim milk (Figure ). The process begins with separation, where centrifugation is used to isolate cream from skim milk. The separated cream (containing approximately 40% milk fat) may either be pasteurized and sent for secondary processing to produce products like butter or may be packaged directly for sale. Simultaneously, a portion of the cream may be reincorporated with the skim milk during the standardization step to achieve the desired milk fat content for the final products: 3.25% for whole milk, 2% for reduced-fat milk, and 1% for low-fat milk. For instance, we estimate approximately 82% is retained in the production of whole milk 48% of the cream is retained in production of 2% milk, and 23% is retained in production of 1% milk, with the remaining cream in each case either sent to downstream secondary dairy product processing or packaged for direct consumption. The standardized milk is then homogenized to ensure uniform consistency, followed by pasteurization to eliminate pathogens. After cooling, the milk is either forwarded to secondary processing for other dairy products or packaged and refrigerated for downstream distribution, retail and direct consumer sale. Primary processing system boundary producing Milk and Cream coproducts.A CIP = clean-in-place A94% allocation to whole milk, 6% allocation to cream; 83% allocation to reduced-fat milk (2%), 17% allocation to cream; 75% allocation to low-fat milk (1%), 25% allocation to cream; 67% allocation to skim milk, 33% allocation to cream. To ensure compliance with food safety regulations, a clean-in-place (CIP) process is also used, which is an automated system used to clean the interior surfaces of dairy processing equipment, such as tanks, pipes, pasteurizers, and filling machines. The CIP process in primary processing involves high temperature water, alkaline detergents and acids to eliminate mineral deposits, followed by onsite wastewater treatment. Material and energy inputs and outputs at each step, such as electricity and thermal energy requirements, are estimated using established process-engineering models, including those by Tomasula et al. for milk and cream processing, and is complemented by data from Konstantas and Bush for HVAC and lighting processes. , To capture regional variation in utilities, we scale the average HVAC thermal (heating) energy by the ratio of each region’s 2020 heating degree days to the 2020 national average, and we scale HVAC electrical (cooling) energy by the analogous ratio for cooling degree days. For CIP, electricity use was estimated for cleaning filling machines, tanks, and pasteurizers used for the respective milk and cream product, while the consumption of water, cleaning chemicals and thermal energy was attributed between cream and milk products based on their respective volumes processed. Figure shows the estimated weighted average energy requirements distributed between electrical and natural gas energy use by primary processing step, weighted by the total whole milk, reduced-fat milk (2%), low-fat milk (1%), and skim milk produced in the U.S. For the primary milk production process, electricity represents around 74% of the energy use, due primarily to refrigeration (58%), HVAC/lighting utilities (19%), and packaging/filling operations (11%). Thermal energy requirements represent around 26% of the energy used in primary processing, with the majority used in clean-in-place (CIP) processes, followed by pasteurization (18%) and HVAC utilities (13%). Electric and thermal energy use distribution by process step in primary processing. Represents weighted average across whole milk, reduced-fat (2%), low-fat (1%), and skim milk product sales. See Figure S25 for uncertainty. Secondary Transport Secondary transport involves the movement of intermediary dairy products (e.g., skim milk, cream) from primary processing facilities to secondary processing locations, or from secondary processing facilities to other facilities for additional processing (e.g., evaporated milk), a key logistical step in the production of many secondary dairy products. We assume the secondary transport stage is applicable to all secondary dairy products except cheese and whey powder coproducts, which have a streamlined supply chain by utilizing raw milk directly and applying a truncated set of primary processing steps that often exclude separation, standardization, and homogenization processes, and are often colocated with the secondary cheese-making processes. For all other secondary dairy products, we assume a standard transport distance of 100 km between primary and secondary processing facilities due to data limitations. Although national data are unavailable, data from the Pacific Northwest indicate a weighted average range of 37 to 129 km for these distances. The 100 km assumption reflects the widespread distribution of dairy processing facilities across the U.S. and the economic incentive to locate secondary processing facilities near supply sources to minimize transportation costs. Given the limited availability of national data on secondary transport distances, this assumed value is treated as an uncertain parameter and is varied over a wide range in the uncertainty analysis to assess its influence on emissions (see Section ). For perishable intermediary products such as pasteurized milk and cream, insulated refrigerated transport is assumed to ensure product quality and safety during transit to secondary processing facilities. Conversely, for nonperishable intermediary products like evaporated milk and sweet whey, standard combination trucks are assumed sufficient for transport, as these products do not require temperature control. Secondary Processing Secondary processing of dairy products involves transforming intermediate products from primary processingsuch as cream, whole milk, or skim milkinto a diverse range of dairy products. These include yogurt, butter, cheese, whey protein concentrates (WPC), lactose, and various powdered milk products. This phase incorporates multiple unit processes tailored to specific product outcomes. For example, yogurt production involves fermentation, while butter production necessitates churning and separation processes. Several secondary processing unit operations are inherently energy-intensive (e.g., evaporation, spray drying), and, prior to quantification, were expected to contribute meaningfully to postfarm gate emissions, which motivated the use of a unit-process modeling approach to ensure these operations were explicitly represented and appropriately attributed to coproducts. To construct life cycle inventories (LCIs) for secondary processing, we rely on data from a variety of sources, including peer-reviewed literature, patents, equipment manufacturer specifications, industry white papers, and engineering principles (SI Table S3). These sources enabled detailed modeling of energy and material flows for each dairy product. Parameters such as moisture content of raw and intermediate materials and finished products, specific heat capacities, and temperature differentials are integrated into engineering equations to calculate energy consumption and emissions. Each dairy product undergoes distinct unit processes, which are depicted in SI Figures S2–S19, illustrating the sequence of operations considered for each product. Clean-in-place energy and chemical inputs assumptions are product-specific and reflect the set of unit operations involved in producing each dairy product that requires cleaning (e.g., pasteurization, evaporation, spray drying). These inputs represent industry-average practices and are applied consistently across regions. While multiple production pathways exist for many dairy products, we focus on the most commonly used processing methods. For example, in primary processing although milk can be produced using standard pasteurization or ultrahigh temperature pasteurization (UHT), we assume traditional pasteurization processes. Similarly, for secondary processing, although yogurt can be produced using different pathways (e.g., centrifugation, fortification, or ultrafiltration), we assume the most prevalent pathway in the industry (i.e., centrifugation). Further, variations in processing can result in different product types, such as the numerous varieties of cheese. Although future work may incorporate these process variations to better reflect product-specific differences, we use the cheddar production process as a conservative proxy for all cheese, as it shares key steps with most cheese varieties. Figure shows the final estimated distribution of energy used in the postfarm gate (including both primary and secondary processing) across each process, with approximately 60% of total energy requirements derived from thermal energy and the remaining 40% from electricity. The distribution of energy use varies significantly across unit operations and products. SI Figures S21, S22, and Table S4 highlight the partitioning of energy use among processes and products. Spray drying and evaporation processes, critical for producing milk powders and whey protein concentrates (WPC), account for 25% and 18%, respectively, of thermal energy use. Conversely, refrigeration and freezing require the majority of electrical energy use, representing 44% of the total, primarily from ice cream, milk, and cheese products. Additional contributors to energy demand include utilities (23% of electrical use) and clean-in-place (CIP) processes, which contribute 29% and 10% of thermal and electrical energy use, respectively, and are used in all dairy product processing. Electric and thermal energy use distribution by process step in postfarm gate processing (primary + secondary) across all dairy products. See Figure S26 for uncertainty. Cradle-to-gate emissions from material inputs used as ingredients, such as sugar, eggs, salt, lactic acid bacteria, are included in the analysis, along with emissions from primary packaging materials, including polypropylene (PP), high density polyethylene (HDPE), and polyethylene (PE) plastics and resins, paperboard, and tin cans. Packaging is based on industry average estimates for each product category, as detailed in each respective study (SI Table S3). Emissions from both material and energy inputs, including direct and indirect emissions, are derived from the LCA FE and GREET databases, which reflect U.S. market conditions. , For refrigeration, the U.S. dairy industry predominantly uses ammonia-based refrigerants, which have a global warming potential (GWP) of zero for leaked emissions, , unlike high-GWP alternatives such as R134a. Therefore, this analysis includes both embedded emissions from ammonia production and the indirect emissions from ammonia volatilization and subsequent N2O formation, though these emissions are relatively minor compared to the fugitive emissions associated with high-GWP refrigerants. Additionally, all materials are assumed to be transported an average distance of 500 km, reflecting the assumption that some ingredients are processed and sourced from greater distant locations, which is examined in the uncertainty assessment (Section ). Uncertainty Analysis and Sensitivity to GWP Characterization and Allocation To evaluate the robustness of GHG emissions estimates in dairy product processing, we conducted two complementary analyses: (i) an uncertainty analysis addressing variability in key upstream and postfarm gate processes, and (ii) a discrete sensitivity analysis examining the influence of alternative IPCC Global Warming Potential (GWP) characterization frameworks and allocation approaches on reported CO2e emissions. These analyses serve distinct purposes: the former characterizes variability in system parameters, while the latter evaluates the sensitivity of results to methodological conventions commonly used in the literature and in greenhouse gas accounting. Uncertainty in total emissions estimates arises from both upstream processes (e.g., raw milk production) and postfarm gate activities (e.g., transportation, processing). To characterize this uncertainty, we integrated uncertainty ranges from Pelton et al., which quantified variability in raw milk production emissions incorporating standard error in key processes such as enteric fermentation, manure volatile solids and nitrogen excretion models, and feed-related emissions. Building on this, we extended the analysis to capture variability in postfarm gate processes, including transport distances, sourcing regions, processing efficiencies and related energy inputs. For raw milk transportation, we accounted for regional sourcing uncertainty by using the standard deviation of transport distances, which varied ±92% from the average distance traveled. Due to data limitations for secondary transport, we assume ±92% similarly represents the variation in secondary transport and materials transport distances across regions. Figure S20 shows that several processing regions draw from multiple sourcing regions, implying that differences in sourcing mix can shift regional total emissions. To reflect uncertainty associated with milk sourcing region assumptions, we evaluated a bounded set of alternative milk sourcing scenarios. For each processing region, 25% of the milk supplied by the dominant sourcing region was reallocated either to the lowest-emission-intensity or highest-emission-intensity sourcing region observed for that processing region, holding all other sourcing shares constant. This approach illustrates how reasonable variation in sourcing patterns could influence regionally weighted milk production emission factors, without suggesting that such reallocations are feasible at the national scale. In primary and secondary processing, where energy inputs such as electricity and natural gas are the dominant sources of emissions, we incorporated ±25% variability in baseline energy inputs to reflect potential differences in processing equipment efficiencies and facility-specific practices. Using this bounded error propagation method, we combined uncertainty from upstream and postfarm gate processes to estimate plausible emission ranges, applying either all minimum or all maximum parameter values. Although this does not produce a full probability distribution, it transparently illustrates the potential variability in total cradle-to-processing gate emissions, providing a more comprehensive assessment of potential variability in industry-wide emission estimates if all examined uncertainty parameters were to simultaneously skew toward their upper and lower bounds. However, structural model uncertainties, such as the choice of enteric methane prediction equations, feed emission factor databases, and LUC characterization approaches, are not captured by the bounded analysis and could shift absolute emission estimates beyond the reported range. The implications of these structural uncertainties for the interpretation of results are discussed in Section . In addition, because allocation choices can materially influence product-level impacts in multioutput dairy processing systems, we examined sensitivity to allocation approach by comparing the two methods embedded in the IDF (2022) conditional allocation guidance: milk-solids allocation (applied to coproduct sets used for food, nutraceutical, or industrial applications) and economic allocation (applied to coproduct sets with animal-feed outputs). For all product systems affected by conditional allocation (e.g., cheese and whey coproducts and downstream whey derived ingredients), we recalculated cradle-to-processing gate emission intensities under (i) 100% milk-solids allocation and (ii) 100% economic allocation, holding all upstream inventories constant. We report the resulting percent difference in emission intensity between allocation approaches for each affected product (Table S15), allowing differentiation between results driven by physical and process-level differences (Table S11) and those driven by allocation conventions. The sensitivity analysis of GWP characterization factors evaluates how reported CO2e emissions vary when applying different Intergovernmental Panel on Climate Change (IPCC) assessment report frameworks (AR6, AR5 with and without climate-carbon feedbacks (AR5-cc, AR5), and AR4). This analysis is not intended to represent uncertainty in the underlying climate impacts of greenhouse gas emissions, as AR6 values represent the current scientific consensus on radiative forcing and atmospheric lifetimes of GHGs. Rather, it is designed to illustrate the magnitude of variation that arises from the continued use of different GWP conventions across studies, data sets, and corporate reporting tools, helping stakeholders contextualize results relative to prior studies that may have relied on earlier IPCC frameworks (e.g., AR4 or AR5). Result and Discussion Cradle-to-Processing Gate Greenhouse Gas Emissions This study provides a comprehensive baseline assessment of greenhouse gas (GHG) emissions across the U.S. dairy product supply chain, incorporating 12 distinct processing regions and accounting for regional variations in milk sourcing, product mix, and grid energy mixes. In 2020, total cradle-to-processing gate emissions from U.S. dairy product production were estimated at 163.3 million tonnes of CO2e (MMT CO2e) per year (Figure a). The raw milk production phase (cradle-to-farmgate) is the largest source of emissions, contributing on average 81% of total cradle-to-processing gate emissions across all dairy products (131.97 MMT CO2e/year). When pairing upper-bound upstream uncertainty with lower-bound postfarm gate uncertainty, the farm-gate contribution rises to approximately 89%; conversely, under lower-bound upstream and upper-bound postfarm gate assumptions, the contribution decreases to approximately 70% (SI Tables S9–S10). The relative contribution of milk production varies by product type, ranging from approximately 35% to 94%, reflecting differences in milk input requirements, processing intensity and regional sourcing mix. The farm-gate share reported here is contextualized relative to prior U.S. dairy LCAs in Section ; when LUC emissions are excluded, the average industry-wide farm-gate contribution decreases to approximately 72%. Secondary processing together accounts for 10.1 MMT CO2e/year (6% of emissions) across dairy products, followed by emissions from clean-in-place (CIP) and wastewater treatment processes at 7.0 MMT CO2e/year (4% of emissions), and milk transportation at 3.9 MMT CO2e/year (2% of emissions). (A) Cradle-to-processing gate total greenhouse gas emissions (MMT CO2e) from U.S. dairy products production in 2020 across major life cycle stages with uncertainty range.A (B) Postfarm gate greenhouse gas emissions (MMT CO2e) from U.S. dairy product production in 2020 disaggregated by unit process and uncertainty range.A Panels A and B are not intended to sum directly. Several unit processes contribute to multiple life cycle stages (e.g., cooling, pasteurization) but are summed in panel B. AConsiders uncertainty in regional milk production emissions (including in enteric methane formation, volatile solids and nitrogen excretion, feed dry matter intake, and feed emission), regional milk transportation, materials transport, and secondary transport distances, and uncertainty in primary processing and secondary processing energy use (electrical and thermal). AConsiders uncertainty in regional milk transportation and secondary transport distances, and uncertainty in primary processing and secondary processing energy use (electrical and thermal). For several unit processes, uncertainty ranges are smaller than the displayed numerical resolution (MMT CO2e) and therefore are not visually apparent, although they are fully included in the uncertainty analysis. Emissions from raw milk production primarily stem from three major sources including feed production, enteric fermentation and manure management. This study uses estimates from Pelton et al. which identified feed production and sourcing (concentrates, byproducts, forage) as the largest contributor, accounting for 40% of raw milk production, followed by enteric fermentation at 29% and manure management at 25%. These sources substantially influence the overall emissions profile of dairy products, emphasizing the need for targeted mitigation strategies at the farm level. Beyond the farm, the postfarm gate emissions exhibit considerable variation depending on the specific processes involved in dairy product manufacturing. While secondary processing in aggregate is the largest collective source of emissions with 32% of postfarm gate emissions, clean-in-place (CIP) processes, used to maintain hygienic standards across both primary and secondary processing facilities, represent the largest source of emissions from a single processing step, with 6.9 MMT CO2e, or 21% of postfarm gate emissions (Figure b and SI Figure S27). The transportation of raw milk inputs accounts for a total of 3.9 MMT CO2e, accounting for 12% of postfarm gate emissions (2% of cradle-to-processing gate emissions), reflecting the significance of logistics and supply chain structure in overall GHG estimates. Several specialized processes within dairy processing also contribute substantially to emissions. Fermentation, which is used in the production of yogurt, sour cream, cottage cheese, and buttermilk, generate 5.1 MMT CO2e per year, or approximately 16% of postfarm gate emissions, due in part to the emissions embedded in the production of lactic acid bacteria. Spray drying and evaporation, critical steps for producing whey protein concentrates, whey powders, milk powders, and other high-value, low moisture dairy products, each account for 1.2 MMT CO2e per year, or combined 8% of postfarm gate emissions, primarily due to the high thermal energy required for moisture reduction. Refrigeration and freezing required for processing and storage generate 2.8 MMT CO2e/year, contributing approximately 9% to postfarm gate emissions. Product-Specific Emissions Intensity The emissions intensity of dairy products, measured per kilogram of product, varies significantly across different dairy categories and regions (Figure , SI Figure S28 and Table S5). Among all dairy products analyzed, whey protein isolate (WPI) and whey phospholipid protein concentrate (WPPC) have the highest emissions per kg of product, at approximately 25 kg CO2e/kg, due to the intensive processing requirements and the large volume of milk required to produce these products. Other value-added dairy products, including whey permeate solids, WPC, and lactose, also have high emissions per kg of product, reflecting both the large volume of milk input to produce a kg of product and the energy intensive nature of drying and membrane separation processes. When emissions are expressed per kg of fat and protein corrected milk (FPCM) input, intensities for these products are relatively low, ranging between 0.1 to 0.4 kg CO2e/kg FPCM, as 1 kg of FPCM yields only a small amount of WPI, WPPC, WPC, permeate solids or lactose products. Emission intensity (A) per kg of product including primary packaging (relevant for product carbon footprint benchmarking and downstream reporting), and (B) per kg of FPCM input (enabling comparison across dairy products on a nutritionally equivalent basis); uncertainty in SI Table S7. Products are presented in descending order based on total mass sold for direct consumption (see Figure ), with the product at the top having the highest total quantity sold (by weight).A Emissions from cheese production are based on cheddar production process. Acid whey, a coproduct of yogurt production that is commonly fed to livestock, has the lowest emission intensity per kg of product at 0.07 kg CO2e/kg. This is due to its minimal processing requirements, low raw milk input demand (∼0.6 kg acid whey/kg FPCM), and the allocation of upstream impacts to yogurt production. Fluid milk, which undergoes comparatively limited processing, has the second lowest emissions intensity, at approximately 1.34 kg CO2e/kg of product. The weighted average emission intensity across all dairy products is approximately 1.60 kg CO2e/kg FPCM, corresponding to 3.76 kg CO2e/kg generic dairy product (SI Table S7). SI Table S6 shows the corresponding ranges in product emission intensity across regions. While emission intensities expressed per kilogram of product are primarily intended for benchmarking and product carbon footprint applications (e.g., Scope 3 accounting and downstream reporting), the per kg FPCM metric facilitates comparison of processing emissions across dairy products on a nutritionally equivalent basis. Total Emissions by Product and Region The total emissions contribution of each dairy product is strongly influenced by both production volume and product carbon footprint (Figure ). Cheese production, which is both emission-intensive and produced in high quantities, accounts for 50.9 MMT CO2e per year, representing 31% of total emissions. Fluid milk, despite its relatively low emission intensity, contributes 24.1 MMT CO2e per year, or 15% of total emissions, due to its high sales volume. Butter, which has a higher emission intensity per kg of product than ice cream, generates 13.9 MMT CO2e, 9% of total emissions, even though ice cream production volumes are higher. Milk powders, which have moderate production volumes and emissions intensities, contribute approximately 8% of total emissions. Total GHG cradle-to-processing gate emissions estimated across dairy products based on total production (and associated total milk inputs required for production) and emission intensity of each product, and contribution across (A) detailed production processes and (B) summary emission sources (MMT CO2e) (uncertainty in SI Tables S7 and S8, and complementary breakdown excluding milk production in SI Figure S30). Note that Feed includes concentrates, byproducts, forages, feed milling/mixing, feed transport. Total emissions by region likewise depend on total production of dairy products, the production mix of products produced in each region, and the associated emission intensity of the dairy products (Figure S29). With the Upper Midwest being the largest producer of dairy products across all regions, it generates the largest quantity at 40.3 MMT CO2e (25% of national total), driven by cheese, butter, ice cream, milk and milk powder production. The Southwest (25.5 MMT CO2e, 16% of total) and West (24.4 MMT CO2e, 15% of total) have the second and third highest emissions by region, respectively, due in large part to cheese, lactose, milk powders and yogurt. These results highlight key opportunities for targeted mitigation efforts, focusing on areas that have both high production and high relative emission intensity, particularly in regions where energy-intensive processes such as spray drying for whey protein concentrates, permeates, and milk powders are prevalent. Taken together, regional differences in total and product-level emissions are driven by the combined effects of regional electricity grid composition, specialization in energy and milk-intensive products (e.g., cheese, powders), milk sourcing and associated production emissions. Because cradle-to-farm gate milk production contributes the majority of total emissions, regional totals are strongly influenced by both the milk emission intensity of sourcing regions and the sourcing distributions used to derive processing region-specific weighted milk emission factors. Greenhouse Gas Contributions by Type GHG emissions across the dairy supply chain are predominantly driven by methane (CH4), carbon dioxide (CO2) and nitrous oxide (N2O), including direct and indirect (Figure ). Biogenic methane accounts for 39% of cradle-to-processing gate emissions on average (35–42% across uncertainty ranges), primarily originating from enteric fermentation and manure management. Carbon dioxide contributes 23% of total emissions (range from 22 to 24%), mainly from fossil fuel combustion in transportation and processing. Nitrous oxide (direct and indirect), which represents 9% of total emissions (ranging from 7 to 11%), is largely associated with fertilizer application in feed production and manure emissions. Land use change (LUC) emissions from feed production contributes an additional 8% to total emissions (ranging from 5 to 12%). Some feed-related emissions, which could not be disaggregated by gas type due to database limitations (e.g., GFLI), are classified as aggregate CO2e emissions. Contribution of total cradle-to-processing gate greenhouse gases to 2020 U.S. dairy emissions (in CO2e), delineated by cradle-to-farm gate and postfarm gateA. AConsiders uncertainty in regional milk production emissions (including in enteric methane formation, volatile solids and nitrogen excretion, feed dry matter intake, and feed emission factors), regional milk transportation and secondary transport distances, and uncertainty in primary processing and secondary processing energy use (electrical and thermal). Uncertainty, Sensitivity, and Benchmark Analysis Given the uncertainty in milk production, sourcing, transport, and processing, cradle-to-processing gate emissions are estimated to range between 115.9 and 218.8 MMT CO2e per year (SI Tables S8–S9). Among evaluated sources, variability in milk production parameters has the greatest influence on total emissions, followed by uncertainty in milk transportation distances. To further characterize the sources of upstream uncertainty, Table S10 shows the disaggregated contribution of individual farm-gate parameters to total cradle-to-processing gate emissions variability. This indicates that uncertainties associated with feed production (with upper and lower bounds spanning ±12% from the mean) and feed intake (±8%), have a substantially greater influence on total emissions estimates than uncertainties in enteric methane formation (±4%) or manure emission models (±2%). The relatively wide uncertainty range for feed emission factors reflect in part the sensitivity to LUC characterization methods and data sources. Together, these results indicate that greater supply chain transparency and improved data collection on milk and feed sourcing and production (see Section ) could meaningfully narrow uncertainty in region and industry-wide estimates. Sensitivity to parameter variation also differs across products; for example, butter exhibits greater sensitivity to uncertainty in secondary transport distances due to its reliance on relatively heavy intermediate cream inputs (Table S11). In addition to parameter uncertainty, product-level emission intensities are influenced by methodological choices related to allocation in multioutput dairy processing systems. For coproduct systems subject to conditional allocation (e.g., cheese and whey systems and downstream whey-derived ingredients), applying 100% economic allocation versus 100% milk-solids allocation results in product-specific emission intensities that differ from −93% (whey acid) to +62% (milk protein concentrate) (Table S15), with the final emission intensity weighted based on the feed-use share. These differences arise from allocation conventions rather than physical process differences and highlight the importance of transparent and harmonized methodological assumptions when comparing product or organizational carbon footprints across studies and reporting frameworks. Reported CO2e emissions also vary as a function of the IPCC Global Warming Potential (GWP) characterization framework applied, reflecting differences in methodological conventions rather than inventory uncertainty. Using alternative GWP characterization factor conventions (AR4, AR5, AR5-cc, and AR6) results in cradle-to-processing gate emissions ranging from 158.8 MMT CO2e (AR4) and 180.2 MMT CO2e (AR5-cc), compared to 163.3 MMT CO2e with AR6 (SI Table S12). At the product level, this corresponds, for example, to fluid milk emission intensities ranging from 1.30 (AR4) to 1.49 (AR5-cc) kg CO2e/kg packaged fluid milk (compared to 1.34 kg CO2e/kg under AR6), and cheese ranging from 6.90 (AR4) to 7.99 (AR5-cc) kg CO2e/kg cheese (compared to 7.13 kgCO2e/kg cheese under AR6) (SI Tables S13 and S14). Notably, the variation in total emissions attributable to alternative GWP frameworks is smaller than the uncertainty associated with upstream milk production parameters, reinforcing that characterization choice primarily affects comparability across studies rather than the overall uncertainty of inventory results. While this study adopts the 100-year Global Warming Potential (GWP100), consistent with prevailing life cycle assessment practice and current U.S. dairy industry climate targets, alternative climate metrics such as the Global Temperature change Potential (GTP) and other time-dependent approaches may provide complementary perspectives on climate risk, particularly for short-lived climate forcers such as methane. Evaluation of such metrics could be valuable in future work but is beyond the scope of this analysis. Many of the resulting emission intensities align closely with prior U.S. dairy product LCAs, indicating that the enhanced spatial resolution and unit-process modeling improve internal consistency and product-level attribution without disrupting the broader empirical understanding in the literature. Importantly, rather than producing divergent estimates, the principal contribution of the present analysis lies in its ability to resolve greenhouse gas emissions by product, region, and processing pathway within a single, internally consistent cradle-to-processing gate framework. While many prior studies employ broadly similar system boundaries and functional units, fully harmonized comparisons across studies remain limited by differences in underlying inventory data sources, regionalization approaches, processing representations, and other methodological assumptions. Despite these limitations, the results of this study compare well with previous estimates and should be interpreted as contextual benchmarks rather than indicators of temporal trends. For fluid milk, cradle-to-postfarm gate emissions, which may include processing, packaging, distribution, and retail emissions, depending on the system boundaries assumed in each study, range from 0.99 to 2.05 kg CO2e/kg FPCM in the U.S. ,,, Thoma et al. estimated cradle-to-processing gate emissions for fluid milk at 1.43 kg CO2e/kg FPCM using AR4 GWP factors which is slightly higher than our AR4 estimate of 1.25 kg CO2e/kg FPCM (SI Table S14). Similarly, Henderson et al. estimated cradle-to-processing gate emissions under AR4 at 1.29 kg CO2e/kg FPCM, increasing to 1.55 kg CO2e/kg FPCM under AR5-cc which corresponds well to our estimated 1.43 kg CO2e/kg FPCM. Consistent with prior studies, global estimates for fluid milk emissions range widely, from 1.0 to 7.0 kg CO2e/kg product, , reflecting regional variations in production practices, efficiencies, feed sources, and energy systems. The relatively low placement of U.S. dairy production within this range is therefore expected, but remains important context for interpreting product-level benchmarks, particularly for international comparison. For milk transportation, Thoma et al. estimated emissions at 3% of total cradle-to-processing gate emissions for fluid milk, which corresponds well with our estimate of 2%. However, the total average roundtrip distance estimated between regions in our study was 340 km, substantially lower than the 829 km reported by Ulrich et al. and used in Thoma et al. This discrepancy may be partly explained by differences in our methodology. Our estimates are based on a least-cost optimization transport model, whereas the Ulrich et al. (2012) estimates rely on survey responses covering about 17% of the U.S. milk market in 2007. However, Ulrich et al. excluded load sharing trips, a common industry practice, which likely resulted in an overestimation of transport distances. In contrast, the lower average transport distances estimated in this study appear to better align with industry consensus, as informed by consultations and discussions with several major dairy processors. The cradle-to-farm-gate contribution to total cradle-to-processing gate emissions for fluid milk was estimated at 88% in this study, closely aligning with estimates from Thoma et al. and Henderson et al., which reported 86% and 89%, respectively. This alignment is notable because these earlier studies exclude LUC emissions from feed, which differs from the present study where LUC accounts for ∼8% of total cradle-to-processing gate emissions and would therefore raise the apparent farm-gate share relative to those earlier studies. However, this study also uses updated enteric methane and manure modeling approaches that yield lower upstream emissions than Ellis and ASAE models used in prior assessments. , Applying those earlier models in our analysis would increase farm-gate emission intensity from 1.30 to approximately 1.40 kgCO2e/kg FPCM, also approximately an 8% effect, thus offsetting the increase from LUC inclusion in absolute farm-gate emissions terms. Notably, because LUC contributes to both the cradle-to-farm-gate numerator and the cradle-to-processing gate denominator, its inclusion has a smaller effect on the farm-gate share than on absolute emissions, which is why our with LUC-share (88%, 84–91% across upper and lower bound parameter values) and without-LUC share (87%, 83–89% across bounds) differ by only one percentage point. This indicates that the identification of farm-gate emissions as the leading source of milk cradle-to-processing gate emissions is robust across both structural model and system boundary choices. For cheese, Kim et al. estimated cradle-to-processing gate emissions for cheddar cheese at 7.62 kg CO2e/kg product, which compares well with our estimate of 7.13 kg CO2e/kg cheese (6.6 for milk-solids allocation, and 10.6 kgCO2e/kg product for economic allocation). Heller& Keolian reported a broader range for cheese emissions, from 6.6 to 16.3 kg CO2e/kg product, with an average of 9.8 kg CO2e/kg product. Kim et al. also estimated that 89% of cradle-to-processing gate cheddar cheese emissions originate from cradle-to-farm gate processes, compared to 94% in our analysis (92–96% across upper and lower bound values). LUC inclusion has minimal effect on the cheese farm-gate share (93.5% without LUC vs 94.1% with LUC); the somewhat higher share in our analysis instead likely reflects differences in allocation and inventory development methodology across the product system. Beyond cheese, several product categories show systematic divergence from prior LCAs also driven by differences in allocation assumptions and inventory completeness. For yogurt, prior studies report 1.26 to 3.35 kg CO2e/kg product, with mean values of 1.8 to 2.02 kg CO2e/kg product, ,, compared to our estimate of 4.35 kg CO2e/kg. This discrepancy largely reflects our assumption that 100% of acid whey is used as animal feed, which triggers economic allocation. If, instead, acid whey is directed to food, nutraceutical or industrial applications, triggering milk-solids allocation, we estimate 2.38 kgCO2e/kg, aligning more closely with prior estimates. For cottage cheese, reported emissions range from 1.8 to 3.7 kg CO2e/kg product, , which is lower than our estimate of 7.7 kg CO2e/kg product, possibly due in part to the notable embedded emissions in lactic acid bacteria inputs which were excluded in previous assessments, and the assumption that 100% of the impacts are allocated to cottage cheese versus acid whey coproducts. Most prior studies similarly excluded LUC emissions from feed production; without-LUC emissions factors for all dairy products in this study are reported in SI Table S6 for transparent comparison. For dry whey powder, previous estimates range from 7.4 to 16.4 kg CO2e/kg product, ,,− closely aligning with our estimate of 8.4 kg CO2e/kg powder. For whey protein concentrate (WPC), emissions have been estimated between 16.4 and 17.4 kg CO2e/kg product, whereas our analysis produced a slightly higher range of 16.4 to 20.4 kg CO2e/kg WPC. For ice cream, Heller & Keolian estimated emissions at 3.1 kg CO2e/kg product, which is comparable to our estimate of 3.3 kg CO2e/kg ice cream. While few studies provide explicit uncertainty estimates for dairy product emissions, Kim et al. reported an uncertainty range for cheddar cheese emissions from 5.2 to 10.9 kg CO2e/kg product (cradle-to-processing gate), representing −32% to +42% relative to the average estimate. This range is quite similar to the evaluated uncertainty in our study, which places cheddar cheese emissions between 4.9 and 10.2 kg CO2e/kg product (−31% to +42% relative to the mean estimate). By quantifying uncertainty in both upstream and postfarm gate processes and across dairy products, we provide a comprehensive assessment of the factors contributing to variability in dairy product emissions (SI Tables S10 and S11). This approach ensures that the results are both scientifically robust and practically useful for decision-making in environmental policy and supply chain management. Limitations Although this study provides a comprehensive assessment of greenhouse gas (GHG) emissions associated with dairy processing, several limitations should be acknowledged. First, upstream emissions from feed production rely in part on existing data sets characterized using AR5-cc GWP100 values, reflecting the fact that some widely used feed inventories were developed prior to the release of AR6. Differences between AR5-cc and AR6 GWP values for relevant gases are on the order of 10% and are small relative to uncertainty associated with upstream milk production parameters and other inventory assumptions. Future updates to feed inventories using AR6 GWP values would further improve internal harmonization across life cycle stages. Second, while the unit-process approach enables more precise tracking and attribution of emissions to specific coproducts, the life cycle inventories used in this analysis rely in part on limited publicly available data, with some sources over a decade old. Over this period, dairy product formulations have evolved through expanded use of whey protein fractions (e.g., WPC/WPI), and greater reliance on filtration and enzymatic processes to manage lactose content and protein concentration. However, for most major dairy products, formulation changes over the past decade have been incremental rather than structural and are expected to be secondary drivers of product-level emission intensity relative to milk input requirements. Quantifying formulation-driven changes over time and implications for industry emissions remains an important area for future research. Future work should also explore the implications of updated production technologies and pathways, account for regional variation in technology adoption, and incorporate more comprehensive primary data collection at the unit-process level specific to each dairy product, as well as improved characterization of regional and product-specific transportation distances. Improvements in data collection for dairy processing should focus on the unit processes that have substantial influence on overall emissions, such as, for example, clean-in-place, spray drying, evaporation, refrigeration, and facility utilities (HVAC and lighting). Similarly, while multiple packaging options exist for each dairy product, baseline estimates here include only the most common packaging formats despite industry-wide variation. Given that packaging contributes a relatively small share of cradle-to-processing gate emissions, incorporating a broader distribution of packaging options is unlikely to materially affect overall results. However, packaging assumptions may have greater implications for mitigation analyses, particularly due to secondary effects influencing downstream emissions in retail and consumption. Third, while cheese production represents a significant contribution to the overall emissions footprint of dairy processing, this study primarily focuses on the cheddar cheese production process. Although cheddaring is a widely used process in the U.S., it does not account for variability in emissions across different cheese types, which differ in milk inputs, processing techniques, and energy demands. For instance, soft cheeses such as ricotta or cream cheese have higher moisture content and require less aging, whereas hard cheeses like Parmesan undergo extended aging and processing, leading to differences in energy use and emission intensity. A global meta-analysis by Gaillac and Marbach found that emissions from cheese types (other than cottage cheese) range from 4.5 to 17 kgCO2e/kg product, with cream cheese at the lower end and pecorino hard cheese at the higher end. Despite variability across cheese styles, cheddar and mozzarella, which collectively account for 62% of total U.S. cheese production, exhibit similar emission intensities and overlapping uncertainty ranges. Given the significant contribution of cheese production to dairy processing emissions, future research should focus on developing spatially explicit and process-specific models to better capture diversity of cheese production systems, as this may materially affect total regional emissions. Taken together, the limitations described above primarily affects the precision of the estimates rather than their magnitude. Many of the identified areas for future research correspond to parameters that were explicitly varied in the uncertainty analysis presented in this study. As a result, while future refinements are expected to reduce the range of uncertainty, the resulting estimates are expected to remain within the bounds reported here. Beyond GHG emissions, other environmental impact categories, such as water scarcity and eutrophication potential, are also important considerations for dairy sustainability and have been examined in prior studies for individual products and systems. Building on the unified, industry-wide modeling framework developed here, future research could extend this analysis to incorporate regionally explicit impact characterization methods across a broader set of environmental indicators. Such extensions would facilitate more comprehensive identification of shared hotspots across value-chain stages, products, geographic regions, and emission sources, and enable evaluation of trade-offs and cobenefits across mitigation strategies while maintaining consistency in system boundaries, allocation approaches, and data sources. Mitigating Emissions in the Dairy Industry Across products, regions, and processing, several consistent emissions hotspots emerge. Raw milk production is the largest contributor to total emissions across all dairy products, ranging from an average of 35 to 92% of cradle-to-processing gate emissions across products and regions, consistent with the central role of feed production, enteric methane, and manure management in shaping overall climate impact. Within postfarm gate operations, a small number of energy-intensive unit processes, particularly clean-in-place, fermentation inputs, evaporation, spray drying, and refrigeration, account for a disproportionate share of processing emissions. Products that concentrate milk solids or remove water (e.g., cheese, milk powders, whey ingredients) exhibit higher emissions per kg of product, whereas products with minimal processing show lower intensities. Regionally, hotspots align with both production volume and energy system characteristics, with regions specializing in cheese and powder production and those relying on higher-carbon electricity grids exhibiting elevated processing emissions. These patterns indicate that effective mitigation will require coordinated strategies addressing both farm-level emissions and targeted decarbonization of specific processing operations. Given that postfarm gate operations account for an average of 19% of total cradle-to-processing gate emissions (11% to 30% across the bounded uncertainty range), targeted strategies to reduce emissions from energy use in dairy processing can meaningfully reduce this portion of the footprint, though their influence on total industry-wide emissions is bounded by the relative contribution of postfarm gate activities. Under an upper bound scenario in which the majority of postfarm gate energy use is decarbonized through renewable electricity, low carbon thermal energy, and efficiency improvements, postfarm gate emissions could plausibly be reduced by up to 70–80%. This would correspond to an approximate 10–15% reduction in total cradle-to-processing gate emissions across the U.S. dairy industry. Realized reductions, however, would depend on product mix, regional energy systems, and technology-specific adoption rates. Since thermal energy use accounts for 60% of the total energy consumption, with electricity comprising the remaining 40%, mitigation efforts should be prioritized toward key energy intensive processes such as spray drying, evaporation, CIP, pasteurization, and packaging systems which drive postfarm gate energy demands. For example, reducing moisture contents using membrane processes, centrifugation, and multieffect vacuum or vapor recompression evaporators prior to spray drying can greatly reduce energy demands, with approximately 0.04, 0.11, 0.65, and 4.3 MJ per kg of water removed across technologies, respectively. The regional variability in energy systems also presents an opportunity for location-specific mitigation strategies. For example, regions rich in bioenergy resources could transition to biogas systems, such as renewable natural gas (RNG) for boilers and to high efficiency energy systems to substantially reduce thermal energy emissions. Conversely, regions with high-carbon grid electricity could prioritize the electrification of thermal processes while simultaneously investing in renewable electricity sources to reduce indirect emissions. The Upper Midwest, for example, has abundant renewable energy resources but also has relatively high grid electricity emissions. Transitioning to both bioenergy and renewable electricity in this region could substantially reduce dairy processing emissions while also incentivizing upstream emission reductions. For example, greater demand for anaerobic digestion systems in manure management could reduce methane emissions from milk production, while simultaneously providing low impact energy for dairy processing facilities. However, due to high upfront capital costs, permitting requirements, and dependence on supportive policy and market conditions (e.g., renewable fuel credits, utility interconnection), their feasibility and adoption potential vary by region, farm size and regulatory context. In electricity-intensive processes such as refrigeration and freezing, increasing reliance on grid-based renewable electricity or on-site renewable generation (e.g., rooftop solar with battery storage) could ensure a stable and low-emission energy supply for cold chain operations. Additionally, energy efficiency upgrades in refrigeration systems could further reduce electricity demand. These strategies highlight the importance of tailored regional approaches that leverage local resources to maximize sustainability gains. Exploring the full mitigation potential of these strategies remains a critical area for future research. As an initial assessment, we evaluated two processing-stage decarbonization levers, upgrading on-site boilers from 80% to 96% efficiency and procuring 100% renewable electricity to illustrate the importance of different energy-related levers within the current structure of the dairy supply chain. The boiler upgrade reduces U.S. postfarm gate emissions by 2.6% (1.0%–4% across regions), corresponding to a 0.5% decrease at the cradle-to-processing gate level. Renewable electricity yields a larger 17% reduction in postfarm gate emissions (6%–35% across regions), corresponding to a 3% decrease at the cradle-to-processing gate level. Mitigation magnitudes vary with each product’s electricity-to-thermal energy profile. For ice cream, where freezing, hardening, and cold storage drive electricity use, renewable electricity lowers cradle-to-processing gate emissions by 13%, whereas for dairy powders (e.g., whey powder, permeate solids), the boiler upgrade delivers more than a 2% reduction. Widespread adoption of these and other strategies across the U.S. dairy industry, such as large-scale deployment of biogas systems, renewable electricity procurement, or electrification of thermal processes, could give rise to secondary and tertiary effects beyond the system boundary considered here. For example, increased demand for bioenergy could influence regional energy markets, feedstock availability, or competing uses of biomass, with large-scale shifts in electricity demand affecting grid composition, infrastructure investment, and marginal generation mixes. Evaluating such system-wide feedbacks would require consequential LCA or economic-wide modeling frameworks that explicitly account for market responses and cross-sector interactions. While such analyses are beyond the scope of the present study, they represent an important area for future research aimed at assessing the broader sustainability implications of industry-wide decarbonization pathways. Despite the opportunities for postfarm gate emissions reductions, raw milk production remains the largest source of GHG emissions across all U.S. dairy products, primarily driven by feed production, enteric fermentation, and manure management. This holds across the bounded parameter uncertainty range, with raw milk production accounting for 70–89% of total cradle-to-processing gate industry emissions. While additional sources of variation beyond those evaluated here may further affect absolute estimates, the prioritization of farm-gate emissions as the primary mitigation target appears robust to the parameter uncertainty quantified in this analysis. The absolute magnitudes of mitigation potential, however, are contingent on the inventory framework adopted; under alternative structural choices for enteric methane and manure modeling evaluated in Pelton et al., farm-gate intensities vary by approximately ±12% from the mean. As such, across both the parameter uncertainty range evaluated here and the structural uncertainty range reported in Pelton et al., farm-gate emissions remain the largest source of cradle-to-processing gate emissions, thus appearing robust to both forms of uncertainty. Reducing emissions from milk production is therefore necessary for the U.S. dairy industry to meet its net GHG neutrality targets by 2050, alongside meaningful contributions from processing-stage decarbonization. Potential mitigation strategies include adopting conservation tillage and fertilization practices, increasing herd productivity, modifying cow diets, and implementing anaerobic digester systems and other manure management technologies. However, these strategies require substantial coordination with upstream dairy producers and may involve significant financial investments, potentially shared by stakeholders throughout the value chain. LUC associated with feed production is an additional challenge. Emissions from land use change are included in the cradle-to-farm gate emission factors for all feedstuffs in this analysis, though the spatial resolution and characterization methods vary across feed ingredients and data sources. For the dominant row-crop contributors to dairy feed supplycorn, soy, and wheatprior work has quantified the share of total cropland expansion attributable to U.S. dairy feed demand, estimating approximately 9%, 1%, and 5% of total land conversion for these crops, respectively, corresponding to approximately 20,000 ha per year. However, a comparable industry-attribution analysisquantifying what portion of total LUC area from other feed ingredients (e.g., alfalfa, canola meal, animal-derived meals, and imported feed components such as palm-based products) is specifically driven by dairy demand versus other industry drivershas not been conducted and would require extending the modeling framework beyond the scope of the present study. While modest in magnitude relative to total agricultural land use change, such emissions remain difficult to abate and, like enteric methane, may require a combination of avoided conversion, yield improvements, and offsetting measures to achieve net neutrality targets. Further, this study does not capture downstream emissions associated with product distribution, retail, storage, preparation, consumption and end-of-life management, which together may account for approximately 10% of total cradle-to-grave GHG emissions for some dairy products. , Although these downstream stages are often outside the direct control of dairy producers and processors, they remain relevant for product-level climate impact assessments, particularly for products with energy-intensive storage or preparation requirements. Integrating downstream life cycle stages would enable evaluation of mitigation opportunities beyond the processing gate. Future research should assess the regional feasibility and cost-effectiveness of various mitigation strategies as well as estimate the life cycle emissions reduction potential of different mitigation options to ensure both sustainability and competitiveness in the U.S. dairy sector. To enable meaningful comparison, it is essential to ensure consistency in system boundaries and assumptions when evaluating the relative effectiveness and preferability of different approaches. A comprehensive, regionally specific framework that captures the heterogeneity in management practices across all stages of the supply chain, such as the approach deployed in this study, can help identify high-priority mitigation opportunities. By leveraging spatially explicit data, policymakers and industry stakeholders can track emissions more precisely and target sustainability investments that maximize emissions reduction potential while maintaining economic viability within the dairy sector. Supplementary Material