Evaluation of sap extraction methods for nutrient diagnostics in lettuce, cucumber, and tomato in controlled environments Authors: Husnain Rauf, Daniel Jackson, Jason Lessl, Timothy Coolong, Zilfina Rubio Ames, Cristian Toma, Rhuanito Soranz Ferrarezi Published: 2026 ABSTRACT Controlled Environment Agriculture (CEA) enables the cultivation of high-value crops under controlled environmental conditions, requiring precise nutrient monitoring to optimize fertigation. Standard practices to monitor plant nutrient status include nutrient leaf analysis (NLA), the reference method, but it does not reflect real-time status. Plant sap analysis offers real-time nutrient diagnostics; however, there is no standardized protocol for sap extraction methods across crops and individual nutrients. The objective of this study was to evaluate and compare multiple sap extraction methods for their reliability, practicality, and cost-effectiveness in diagnosing plant nutrient status in key CEA crops. Five sap extraction methods: ammonium acetate (AA), dry freezing (DF), extraction with ethyl ether (EE), extraction with potassium chloride (KCl), and tissue crushing (TC) were tested against NLA for assessing macro- and micronutrients in tomato (Solanum lycopersicum), cucumber (Cucumis sativus), and lettuce (Lactuca sativa) grown under CEA. For nitrate (NO₃⁻), KCL method was used as a reference and compared with EE and TC. The results indicated correlations between sap extraction methods and NLA were weak to moderate and varied across crops and nutrient types. The strongest correlations were obtained in lettuce using TC, with very high values for K (r = 0.99), Mg (r = 0.86), and Ca (r = 0.68). Other methods often yielded weak or negative correlations (e.g., tomato Ca with AA: r = − 0.36; cucumber Mg with KCl: r = − 0.15; lettuce Zn with EE: r = − 0.50). Sap NO₃⁻ correlations were variable, ranging from weakly positive (cucumber EE vs. KCl: r = 0.26) to negative (tomato TC vs. KCl: r = − 0.38), underscoring the complexity of NO₃⁻ dynamics and their dependence on crop physiology and developmental stage. A cost analysis showed that TC was the most affordable option, although it only worked for certain crops. These results indicate that, whereas some sap methods of extraction can produce indicative evidence of plant nutrient status in certain crops, none can accurately serve as a direct substitute for NLA. Introduction Controlled Environment Agriculture (CEA) is rapidly increasing as a production system for high-value crops like lettuce (Lactuca sativa), cucumber (Cucumis sativus), and tomato (Solanum lycopersicum) and CEA enables year-round cultivation by precisely managing environmental conditions, such as light, temperature, and humidity (Cowan et al., 2022). In 2024, the United States (US) tomato production was estimated to be worth $1.97 billion. Approximately 12.05 million hundredweight (cwt) were grown for the fresh market and 11.09 million tons for processing, with Florida and California being the top tomato producers. Lettuce is an important crop for fresh food with 16.1 million cwt grown in 2024. Lettuce is mostly grown in California and Arizona, the industry totals around $1.27 billion. In 2024, the US produced over $225.5 million worth of cucumbers, with Florida, Michigan, and Georgia being the top three states1. Tomato ranks as the most-produced vegetable worldwide in terms of weight, with a harvest of 186.11 million metric tons in 2022, followed by lettuce and cucumber as the next most cultivated vegetables2. CEA systems facilitate intensive production cycles, but plant development and productivity can be impacted by inadequate nutrient management, in these rapidly growing crops, therefore, require real-time monitoring of crop nutrition. Nutrient leaf analysis (NLA) or dry tissue analysis has long been the accepted technique for determining the nutritional status of plants3,4. This process, although accurate, involves drying, grinding, and laboratory digestion of samples and typically requires at least 24–48 h to obtain results. Moreover, because NLA reflects nutrient accumulation over previous days or weeks rather than real-time plant status, it often fails to capture current nutrient imbalances. In a CEA production system, fertigation adjustments need to be managed promptly an in real-time, thus quick nutrient analysis techniques are necessary. Plant sap analysis has become an alternative technique, which provides quicker insight into the nutritional status of plants. Sap analysis provides growers and production managers with information about the nutrients that are actively moving through the plant in the ionic forms of mobile nutrients found in the xylem and phloem5. Recent advances, such as portable ion-selective electrodes (ISEs), have further enhanced sap analysis by allowing rapid, on-site measurement of key nutrients like nitrate (NO₃⁻), potassium (K⁺), and calcium (Ca2⁺). These tools provide growers with near-immediate results at a lower cost, enabling more timely fertigation decisions5–9. Because the results of sap analysis and NLA often differ, they reveal different aspects of plant nutritional status. NLA measures the total amount of nutrients accumulated within plant tissues, and sap analysis determines the concentrations of transportable nutrients in the plant’s vascular system. Depending on the species, developmental stage, and nutrient type, the relationship between the two measurements may differ10. For instance,11 found that the correlation between total nitrogen (TN) and petiole sap NO₃⁻ in cotton (Gossypium hirsutum L.) leaf tissue changed with growth stage, with coefficients of determination (R2) peaking at mid-bloom and ranging from 0.42 to 0.71. Similarly, during the vegetative growth of pak choi (Brassica rapa var. Chinensis group ‘Bonsai’) cultivated in a Hawaiian greenhouse,12 reported strong correlations (r = 0.80–0.99) between petiole sap K⁺ and leaf tissue K concentrations at 14 weeks after transplantation. However, in contrast13, observed no significant correlation between petiole sap NO3⁻ and TN in tomato, highlighting the fact that sap tissue nutrient interactions can be vary depending on species, nutrient mobility, and growth stage. The extraction technique is a critical and often overlooked aspect that affects the sap analysis results. In the literature, several extraction methods have been reported, each with different advantages and limitations. Mechanical pressing is the most widely used method, typically using garlic or hydraulic presses, especially on crops that have fleshy petioles such as tomatoes8,14. The freeze–thaw method can enhance nutrient release by rupturing cell membranes7 while chemical methods such as potassium chloride (KCL) or ammonium acetate (AA) can improve solubilization of ionic nutrients15,16. Additional techniques, including ethyl ether pre-treatment and cryogenic grinding with liquid nitrogen, have been proposed to maximize nutrient recovery and minimize oxidation17,18. However, no standardized methodology for sap extraction currently exists across crops in CEA systems. Differences in leaf anatomy, sap composition, and nutrient partitioning among species can significantly affect the efficiency and representativeness of a given extraction method. Thus, the lack of standardization presents a major challenge for growers and researchers, as it may lead to inconsistent results and hinder accurate interpretation for fertigation decision-making5,17. The objective of this study was to systematically evaluate and compare sap extraction methods in lettuce, cucumbers, and tomatoes grown under commercial CEA conditions. Six different extraction methodologies were assessed, including NLA as a control, extraction with AA, extraction with ethyl ether (EE), extraction with KCL, tissue crushing (TC), and dry freezing (DF) with liquid nitrogen (LN2) followed by extraction using AA. These all-extraction methods were tested across cultivars and developmental stages to identify which methods provide reliable, crop-specific results. By analyzing the correlations of each method with the NLA method, we assessed which extraction technique can serve as a practical, rapid indicator for checking nutrient status. Our goal is to improve the consistency, accuracy, and practical application of sap analysis for real-time nutrient monitoring in CEA systems. Results Correlations between nutrient concentrations obtained using each sap extraction method and those obtained with the standard NLA are presented for tomato, cucumber, and lettuce. We first describe results for each extraction method (AA, DF, EE, KCl, TC) against NLA for the suite of macro- and micronutrients. NO₃⁻ results are presented separately, comparing the EE and TC methods to the KCl extraction (since NLA did not provide NO₃⁻). Interpretation of r values, when r values close to + 1 indicates that the sap extraction method closely positively related with NLA values (when one is high, the other is high, etc.), whereas r close to -1 means the method consistently gives an inverse or negative relationship to NLA (when one is high, the other is low) and values near 0 means there is no linear correspondence between the methods. Ammonium acetate (AA) AA extracted sap exhibited weak agreement with NLA, with negative correlations, especially in tomato (Table 1). In tomato, mostly nutrients had negative correlations, e.g. Ca had r = − 0.36, Mg had r = − 0.27, and Sr had r = − 0.49 (all negative correlations, meaning the AA extract values tended to be higher when NLA values were lower, or vice versa). Micronutrients in tomato via AA were also poorly correlated (Mn r = − 0.66, B r = − 0.41). K and P in tomato showed near 0 or very slight positive correlation (r = 0.06 for K, − 0.07 for P). In cucumber, AA correlations near zero for most nutrients, P was the only one with a modest positive correlation (r = 0.31), while K, Ca, Mg, S, and micronutrients all had r values < 0.2 (essentially no meaningful correlation, some slightly positive, some slightly negative). In lettuce, AA performed a bit better for a few nutrients, notably K had a very high correlation r = 0.94 (suggesting AA extractable K in lettuce sap mirrors total K in tissue quite well), and P and Mg showed modest positive correlations (r = 0.24 for P, 0.23 for Mg). However, even in lettuce, some nutrients were showing a negative correlation with NLA, for example, Zn r = − 0.66 and Mn r = − 0.40.Table 1Pearson correlation coefficients (r) between nutrient concentrations measured by the ammonium acetate (AA) extraction method and nutrient leaf analysis (NLA) in tomato, cucumber, and lettuce.CropPKCaMgSBCuMnZnFeTomato− 0.070.06− 0.36− 0.27− 0.49− 0.41− 0.12− 0.660.06− 0.00Cucumber0.310.17− 0.17− 0.07− 0.05− 0.06− 0.13− 0.140.01− 0.00Lettuce0.240.940.110.230.26− 0.070.12− 0.40− 0.66− 0.00 Table 1Pearson correlation coefficients (r) between nutrient concentrations measured by the ammonium acetate (AA) extraction method and nutrient leaf analysis (NLA) in tomato, cucumber, and lettuce.CropPKCaMgSBCuMnZnFeTomato− 0.070.06− 0.36− 0.27− 0.49− 0.41− 0.12− 0.660.06− 0.00Cucumber0.310.17− 0.17− 0.07− 0.05− 0.06− 0.13− 0.140.01− 0.00Lettuce0.240.940.110.230.26− 0.070.12− 0.40− 0.66− 0.00 Pearson correlation coefficients (r) between nutrient concentrations measured by the ammonium acetate (AA) extraction method and nutrient leaf analysis (NLA) in tomato, cucumber, and lettuce. Dry freezing (DF) The DF method (freeze in LN₂, then AA extraction) yielded variable and generally weak correlations, as illustrated in the heatmap of r values in Fig. 1. In tomato, DF had near-zero correlations for several nutrients (e.g., Mg r ≈ 0.00, indicating no linear relationship). Some nutrients in tomato even showed a slight negative correlation with NLA (S and Mn were slightly negative). Cucumber showed a similar trend where r was close to 0, though P had r = 0.27 and Mg r = 0.17 (small positive trends). Lettuce again showed relatively better alignment, P had a moderate positive r = 0.48 under DF, and Mg around 0.29; K and Ca in lettuce DF extracts were weakly positive as well (r = 0.2 to 0.3). However, lettuce Zn via DF was negatively correlated (around -0.40).Fig. 1Heatmap of Pearson correlation coefficients (r) between nutrient concentrations measured by dry freezing (DF) sap extraction and nutrient leaf analysis (NLA) in tomato, cucumber, and lettuce. Shading represents the strength and direction of correlations, where pink indicates positive correlations, blue indicates negative correlations, and white represents values near zero. Numeric r-values are displayed within cells for clarity. Heatmap of Pearson correlation coefficients (r) between nutrient concentrations measured by dry freezing (DF) sap extraction and nutrient leaf analysis (NLA) in tomato, cucumber, and lettuce. Shading represents the strength and direction of correlations, where pink indicates positive correlations, blue indicates negative correlations, and white represents values near zero. Numeric r-values are displayed within cells for clarity. Ethyl ether (EE) The EE method (ether extraction) generally showed weak correspondence to NLA, as shown in Table 2. Tomato nutrients mostly had slightly negative r values, Ca r = − 0.12, Mg = − 0.08, S = − 0.16. Tomato P and Zn were essentially zero (-0.03 and -0.02, respectively). An exception was tomato Cu, which had r = 0.38 (a moderate positive correlation). This suggests that EE might have captured the Cu proportionally, although it is unclear why Cu alone stood out. Cucumber showed a slight positive correlation with P r = 0.29, but K, Ca, Mg, and S were all near − 0.05 to − 0.12 (no meaningful correlation), and Cu was − 0.29 (weak negative). Lettuce exhibited a strong correlation in K r = 0.80, indicating that EE in lettuce released K roughly in line with total K. However, other lettuce nutrients via EE showed poor agreement, for example, S r = − 0.45, Zn r = − 0.50, Cu r = − 0.35 (all moderately negative). Lettuce P and Mg with EE were near zero (P = 0.01, Mg = 0.05). Thus, apart from K in lettuce, the EE method did not correlate with NLA, and in several cases, it gave inverse indications (especially for S, Zn, Cu in lettuce and for some micronutrients in tomato and cucumber).Table 2Pearson correlation coefficients (r) between nutrient concentrations measured by the ethyl ether (EE) extraction method and nutrient leaf analysis (NLA) in tomato, cucumber, and lettuce.CropPKCaMgSBCuMnZnFeTomato− 0.03− 0.25− 0.12− 0.08− 0.16− 0.020.38− 0.19− 0.02− 0.00Cucumber0.29− 0.05− 0.06− 0.12− 0.12− 0.02− 0.290.06− 0.090.00Lettuce0.010.80− 0.090.05− 0.45− 0.24− 0.35− 0.08− 0.50− 0.02 Table 2Pearson correlation coefficients (r) between nutrient concentrations measured by the ethyl ether (EE) extraction method and nutrient leaf analysis (NLA) in tomato, cucumber, and lettuce.CropPKCaMgSBCuMnZnFeTomato− 0.03− 0.25− 0.12− 0.08− 0.16− 0.020.38− 0.19− 0.02− 0.00Cucumber0.29− 0.05− 0.06− 0.12− 0.12− 0.02− 0.290.06− 0.090.00Lettuce0.010.80− 0.090.05− 0.45− 0.24− 0.35− 0.08− 0.50− 0.02 Pearson correlation coefficients (r) between nutrient concentrations measured by the ethyl ether (EE) extraction method and nutrient leaf analysis (NLA) in tomato, cucumber, and lettuce. Potassium chloride (KCl) The KCl extraction method showed a weak correlation with NLA (Table 3). Most importantly, there is one complication in the KCl extract that adds K, so K values are not reported. In tomatoes, KCl-extracted Ca and Mg had weak positive correlations (Ca r = 0.21, Mg r = 0.14). Tomato P and S were slightly negative (P = − 0.09, S = − 0.11). Most micronutrients in tomatoes via KCl showed a correlation of around zero (e.g., Mn r = 0.22, Cu r = 0.01, Zn r = − 0.14). Cucumber KCl correlations were nearly all zero or negative, P r = 0.01, Ca r = − 0.06, Mg r = − 0.15, S r = − 0.10. Lettuce showed one strong result, with K having a correlation coefficient of r = 0.88 (again, because K was in the extractant; we do not interpret this result). Excluding K, lettuce Mg was modestly correlated, r = 0.21, and Mn, r = 0.29. Lettuce P, Ca, Fe were near 0 (− 0.04 to 0.04), and S and Zn were negatively correlated (S r = − 0.31, Zn r = − 0.50).Table 3Pearson correlation coefficients (r) between nutrient concentrations obtained by the potassium chloride (KCl) extraction method and nutrient leaf analysis (NLA) in tomato, cucumber, and lettuce. Potassium (K) values are reported but were not interpreted because of interference from background K in the extractant.CropPCaMgSBCuMnZnFeTomato− 0.090.210.14− 0.110.090.010.22− 0.14− 0.00Cucumber0.01− 0.06− 0.15− 0.100.01− 0.02− 0.080.02− 0.00Lettuce− 0.040.040.21− 0.31− 0.240.160.29− 0.50− 0.04 Table 3Pearson correlation coefficients (r) between nutrient concentrations obtained by the potassium chloride (KCl) extraction method and nutrient leaf analysis (NLA) in tomato, cucumber, and lettuce. Potassium (K) values are reported but were not interpreted because of interference from background K in the extractant.CropPCaMgSBCuMnZnFeTomato− 0.090.210.14− 0.110.090.010.22− 0.14− 0.00Cucumber0.01− 0.06− 0.15− 0.100.01− 0.02− 0.080.02− 0.00Lettuce− 0.040.040.21− 0.31− 0.240.160.29− 0.50− 0.04 Pearson correlation coefficients (r) between nutrient concentrations obtained by the potassium chloride (KCl) extraction method and nutrient leaf analysis (NLA) in tomato, cucumber, and lettuce. Potassium (K) values are reported but were not interpreted because of interference from background K in the extractant. Tissue crushing (TC) The TC method (pressed sap) showed the strongest correlations with NLA among the sap methods, particularly in lettuce. Figure 2 provides a heatmap of TC vs. NLA correlations. In lettuce, TC had notably high positive correlations for several macronutrients, K r = 0.99 (essentially a one-to-one correspondence between sap K and K), Mg r = 0.86, Ca r = 0.68, and P r = 0.58. These indicate that for lettuce, the simple pressed sap values of these nutrients reflected the relative differences seen in the standard tissue analysis quite well. Lettuce sap S via TC was moderately correlated (r = 0.26, not as high as others but positive). A few nutrients in lettuce still showed negative correlations even with TC (Mn r = − 0.53, Zn r = − 0.49), suggesting that sufficiency vs. deficiency of Mn and Zn might not be discernible from sap alone in lettuce. In tomato, TC results were mixed; P and K in tomato sap had small positive correlations (around 0.2 to 0.3), and Ca was about 0.21. Mg in tomato via TC showed near zero correlation. Micronutrients in tomato sap had weak or negative correlations. Similarly, in cucumber, TC showed only modest positive correlation for some macronutrients (P = 0.3, Ca = 0.2, Mg = 0.1) and near-zero or negative for others. K in cucumber sap by TC was not strongly correlated. Many cucumber micronutrients, as measured via TC, exhibited slight negative correlations (e.g., Cu and Zn, around − 0.2). However, even TC did not correlate well for all nutrients or all crops.Fig. 2Heatmap of Pearson correlation coefficients (r) between nutrient concentrations measured by tissue crushing (TC), sap extraction, and nutrient leaf analysis (NLA) in tomato, cucumber, and lettuce. pink color indicates positive correlations, blue indicates negative correlations, white represents values near zero. Numeric r-values are displayed within cells for clarity. Heatmap of Pearson correlation coefficients (r) between nutrient concentrations measured by tissue crushing (TC), sap extraction, and nutrient leaf analysis (NLA) in tomato, cucumber, and lettuce. pink color indicates positive correlations, blue indicates negative correlations, white represents values near zero. Numeric r-values are displayed within cells for clarity. Sap NO₃⁻ method comparisons Because NO₃⁻ was only measurable in the KCl, EE, and TC methods, we compared those three methods for NO₃⁻. Table 4 shows Pearson correlation coefficients between KCl-extracted NO₃⁻ and the other two methods (EE and TC) for each crop. In cucumber, both EE and TC showed weakly positive correlations with KCl in NO₃⁻, with r values of 0.26 for EE vs. KCl and 0.24 for TC vs. KCl. This suggests that in cucumber, a high NO₃⁻ reading by KCl tended to coincide with a slightly higher reading by EE or TC, but the relationship was not strong (r = 0.25). In lettuce, a disagreement was observed: EE vs. KCl for NO₃⁻ yielded r = − 0.15 (slightly negative, indicating essentially no agreement), whereas TC vs. KCl for NO₃⁻ showed r = 0.72, a strong positive correlation. Thus, for lettuce, the pressed sap (TC) NO₃⁻ values followed the KCl-extracted NO₃⁻ reasonably well (which in turn should reflect actual sap NO₃⁻ content), but the ether method did not. Finally, in tomato, both alternative methods had poor agreement with KCl NO₃⁻: EE vs KCl r = − 0.18 (essentially no correlation, possibly slight inverse) and TC vs KCl r = -0.38 (a moderately negative correlation).Table 4Correlation coefficients (r) between nitrate (NO₃⁻) concentrations measured by ethyl ether (EE) and Tissue Crushing (TC) compared with KCL extraction for tomato, cucumber, and lettuce.CropMethodsCorrelation with KCL (r)CucumberEE0.26CucumberTC0.24LettuceEE− 0.15LettuceTC0.72TomatoEE− 0.18TomatoTC− 0.38 Table 4Correlation coefficients (r) between nitrate (NO₃⁻) concentrations measured by ethyl ether (EE) and Tissue Crushing (TC) compared with KCL extraction for tomato, cucumber, and lettuce.CropMethodsCorrelation with KCL (r)CucumberEE0.26CucumberTC0.24LettuceEE− 0.15LettuceTC0.72TomatoEE− 0.18TomatoTC− 0.38 Correlation coefficients (r) between nitrate (NO₃⁻) concentrations measured by ethyl ether (EE) and Tissue Crushing (TC) compared with KCL extraction for tomato, cucumber, and lettuce. Cost comparison Table 5 summarizes the estimated labor, time, and cost for each method, along with practical notes. The TC method was the fastest and most cost-effective for routine use. On average, preparing and analyzing a sample via TC (including collecting petioles, pressing sap, and performing a quick analysis, such as ISE or a portable meter) took ~ 20 min of active labor, and results could be obtained immediately (within 20 min). The consumable cost was negligible (no reagents were needed), so the main cost was labor ($7 per sample), and lab analysis was approximately $25, making the total approximately $32 per sample for TC. The process is simple, safe (no chemicals), and can be performed directly in the field or greenhouse, with results analyzed using ICP-OES. The EE method was the most labor-intensive and slow. Active labor per sample was ~ 57 min (because it requires freezing overnight, handling hazardous ether, phase separating, plus additional drying or cleaning steps). The turnaround time was about 24–26 h (since samples must freeze overnight and then thaw). The reagent cost (ether and additional supplies) was estimated at $1.40 per sample. Labor costs were approximately $18 per sample, and lab analysis costs were approximately $25/sample, for a total cost of roughly $44.40 per sample. EE requires a fume hood and careful handling of flammable solvents; it’s not practical on-site and mainly of research interest, where extra recovery of certain nutrients might be needed. The DF method was moderately laborious, which requires 40 min of labor per sample (due to grinding in liquid N₂ and careful handling), but has a relatively quick turnaround (you can process a sample in < 1 h if the equipment is ready). Reagent cost (liquid N₂ and supplies) was low ($0.15/sample). Labor cost is approximately $15, and lab analysis costs roughly $25, for a total of $40.15 per sample. Technically, DF is demanding (requiring protective gear for LN₂, a freezer mill, or mortar, etc.) and is not feasible without a laboratory setup. It may be useful for research when trying to maximize the extraction of all intracellular content. The AA extraction was moderate in labor (36 min of active work, like KCl, as both involve cutting, shaking, and filtering). Turnaround was < 1 h per sample once started. The reagent cost is negligible ($0.15 per sample for the acetate solution). Labor: ~ $13 and lab: ~ $25, total cost: ~ $38.15 per sample. AA involves safe reagents (non-hazardous) but still requires a lab or at least a benchtop setup to filter and, possibly, an ICP for analysis. Therefore, it is not on-site portable, but relatively easy and inexpensive for a routine lab. The KCl extraction was almost identical to AA in effort and took 36 min of labor, with a < 1-h turnaround. Reagent cost is low ($0.13/sample for KCl). Labor = $13 and lab ~ $25, total cost $38.13 per sample. KCl is also safe and easy; it can even be performed on-site up to the filtration step, but the extracts typically then require lab analysis (e.g., using an ion meter or specialized lab instrument). Its weakness is not cost or speed, but the interpretability issues (added K and not fit for K results). Finally, NLA (dry tissue analysis) is the most labor intensive. Although active labor per sample might be ~ 25 min (collecting, drying, weighing, and digesting is mostly waiting time, but requires skilled labor to set up digestion), the turnaround is often the slowest, often overnight, just for drying and then digesting, and in practice, up to 2–3 days if sending to a lab. Reagent cost (primarily nitric acid and vessels) is approximately $0.60 per sample. Labor costs are roughly $20 (a skilled technician is required), and the lab costs are approximately $25, so the total is $45.60 per sample. NLA is not feasible on-site at all; it requires lab facilities (such as a microwave digestor and ICP-OES) but provides the most comprehensive and reliable data when time is not a factor.Table 5Estimated cost and practicality of six nutrient analysis methods (per sample). Labor cost assumed at $20/hour. Reagent costs are consumables per sample. Turnaround includes any waiting time for results.MethodKey StepsActive Labor (min)TurnaroundReagent $Labor $Lab running cost/sampleTotal $Practical NotesTCCut and press sap, analyze2020 min$0.00$7.00$25$32.00Very simple; no chemicals; on-site readyEEFreeze in ether (overnight), shake, phase separate5725 h (incl. freeze)$1.40$18.00$25$44.40Requires fume hood; flammable solvent; slowDFFreeze in LN₂, grind, extract, filter4040 min$0.15$15.00$25$40.15Needs liquid N₂ & grinder; lab onlyAAShake in AA, filter3636 min$0.15$13.00$25$38.15Moderate effort: lab needed (ICP analysis)KClShake in KCl, filter3636 min$0.13$13.00$25$38.13Easy and cheap; lab analysis neededNLADry, grind, acid digest, ICP25 (+ long wait)27 h (or more)$0.60$20.00$25$45.60Accurate but slow; lab only (not field) Table 5Estimated cost and practicality of six nutrient analysis methods (per sample). Labor cost assumed at $20/hour. Reagent costs are consumables per sample. Turnaround includes any waiting time for results.MethodKey StepsActive Labor (min)TurnaroundReagent $Labor $Lab running cost/sampleTotal $Practical NotesTCCut and press sap, analyze2020 min$0.00$7.00$25$32.00Very simple; no chemicals; on-site readyEEFreeze in ether (overnight), shake, phase separate5725 h (incl. freeze)$1.40$18.00$25$44.40Requires fume hood; flammable solvent; slowDFFreeze in LN₂, grind, extract, filter4040 min$0.15$15.00$25$40.15Needs liquid N₂ & grinder; lab onlyAAShake in AA, filter3636 min$0.15$13.00$25$38.15Moderate effort: lab needed (ICP analysis)KClShake in KCl, filter3636 min$0.13$13.00$25$38.13Easy and cheap; lab analysis neededNLADry, grind, acid digest, ICP25 (+ long wait)27 h (or more)$0.60$20.00$25$45.60Accurate but slow; lab only (not field) Estimated cost and practicality of six nutrient analysis methods (per sample). Labor cost assumed at $20/hour. Reagent costs are consumables per sample. Turnaround includes any waiting time for results. Discussion Our results demonstrate that the efficacy of sap analysis is highly dependent on both the extraction method used and the crop species and plant tissue. No single sap extraction method emerged as a universally reliable substitute for standard tissue analysis across all three crops (tomato, cucumber, lettuce) and all nutrients. However, certain methods showed promise for specific crop-nutrient combinations. Among chemical extractants, the AA method was notably inconsistent across crops. In tomato, AA-extracted sap showed negative correlations with NLA (Table 1) for Ca, Mg, S, and some microelements, indicating that it can be misleading (e.g., a high Ca in AA sap often corresponded to low total Ca in tissue). A likely reason is that AA is a strong ion-exchange medium that might leach out certain ions from cell walls or vacuoles non-selectively. Tomato petioles, being fibrous, might retain nutrients that AA can extract even if those nutrients are not proportionally available in the sap, hence the poor correspondence. By contrast, lettuce showed significantly positive correlation for K with AA (r = 0.94), suggesting that all K in lettuce tissue is readily exchangeable with the AA solution, which in turn reflects total K. This result aligns with older sap testing literature, as7 found that a quick petiole sap test for K in grapevines could reflect vine K status under certain conditions. The divergent tomato vs lettuce AA results highlight crop-specific extraction effects. AA likely “over-extracts” mobile cations from dense tissues, whereas in lettuce, it extracts what is in the sap. Similarly,19 reported that in conifer seedlings, different sap extraction techniques (including chemical extracts) gave widely varying absolute values that did not consistently match conventional foliar analysis, notifying that such methods need species-by-species validation. The DF method was designed to maximize nutrient release by physically breaking down cell structures before chemical extraction. However, our results indicate that this additional effort did not yield better correlations with NLA, as illustrated in Fig. 1. DF correlations were mostly near zero (no improvement over simple AA extraction). In some cases, DF even gave negative correlations where AA had positive (e.g., lettuce Zn became more negatively correlated in DF). A possible explanation is that the freeze–thaw and grinding process may release extracellular or organelle-bound nutrients in proportions that differ from those in the mobile sap fraction. The DF method essentially turns the sample into a fine powder and then extracts it, which is closer to a total digestion but still incomplete and selective. If one aims for total nutrient content, the nitric acid digest is the gold standard; partial chemical extractions may bring out variable fractions. This may explain why DF did not provide a consistent advantage. Another factor is analytical: freeze–thaw can alter the chemical forms (e.g., converting some NO₃⁻ to nitrite or ammonium, precipitating some phosphates with cations, etc.), thereby affecting detection. Our DF findings contrast with some successful sap quick tests in literature, for instance,8 reported R2 = 0.84 between sweet pepper petiole sap NO₃⁻ and the crop N status or nitrogen nutrition index (NNI) using direct sap measurements, but they did so with a consistent protocol (pressing petioles at specific growth stages). The DF method’s variability suggests that more processing is not always better for sap analysis; sometimes it introduces new variability. Similarly,20 found that while sap tests in broccoli could indicate trends, establishing universal sufficiency thresholds was complicated by environmental factors, implying that harsh extraction might not overcome underlying physiological fluctuations, and calibration remains necessary for extraction methods like DF. Our results showed that the EE method is generally unreliable for nutrient diagnostics, e.g., lettuce K had a correlation coefficient (r) of + 0.80 with NLA in Table 2, presumably because K⁺ (a water-soluble ion) mostly remained in the aqueous phase, and lettuce’s abundant sap allowed it to pass through. However, many nutrients exhibited poor or negative correlations, likely due to losses or partitioning issues. Ether is volatile; during handling, some ammonia or organic N compounds could be lost, affecting N readings. Additionally, the need to add water and separate phases can introduce errors and variability in the amount of aqueous fraction recovered. The fact that lettuce K was high (also lettuce had moderately good K in AA and KCl) suggests that lettuce K is easily extracted by any means, but other nutrients (especially micronutrients like Zn and Cu) might bind to pigments or precipitate during the extraction process (hence the negative correlations).5,21 both had success with NO₃⁻ sap tests using direct methods (pressing sap and then testing immediately). Our EE results reinforce the idea that simpler is better for sap NO₃⁻; the extra steps in EE likely just add more complication. In practice, EE extraction is too burdensome and hazardous to justify routine use, given that it did not clearly outperform easier methods. Using KCl extraction for sap is consistent with its regular use for measuring NO₃⁻ concentrations in petioles of several crops, including strawberry (Fragaria × ananassa) and cotton (Gossypium hirsutum). Because of this established use in soil analysis, we selected KCl-extracted sap NO₃⁻ as our reference method for comparison with the EE and TC methods, assuming it would provide the most reliable estimate of available NO₃⁻ in plant tissues. However, for other nutrients, KCl extracts did not correlate well with all tissue nutrient content, as shown in Table 3. One obvious issue is the K interference with our data, as K is not usable from KCl extracts. That alone is a major limitation because K is a key nutrient to monitor. This indicates that simply immersing tissue in a salt solution does not reliably extract nutrients in proportion to their total content. Some nutrients might be tightly bound in forms that KCl doesn’t extract, or the presence of high K⁺ and Cl⁻ ions might suppress the solubility of other ions. An interesting comparison is with the findings of22, who demonstrated that using portable ISE meters on diluted petiole sap can measure NO₃⁻ accurately (R2 = 0.89–0.93 in greenhouse conditions) when calibrated, without the need for such extractants. If KCl were to be useful, it would likely require specific calibration curves for each nutrient in each crop, which is impractical. KCl may still be a valid extractant for NO₃⁻ (as it is in soil tests), but even there, our tomato results showed TC vs KCl NO₃⁻ disagreement, implying that the tomato’s internal NO₃⁻ distribution (perhaps between petiole and leaflet, xylem vs vacuole storage) can confound even a straightforward KCl extract. This is physiologically plausible because NO₃⁻ is stored in vacuoles while reduction occurs mainly in the cytosol, so compartmentation alone can weaken direct relationships between sap-extracted NO₃⁻ and tissue-based N measurements23,24. In the TC method (expressed sap), there were relatively strong correlations with NLA (Fig. 2) for lettuce macronutrients. Lettuce sap from TC showed a nearly 1:1 correlation for K (r = 0.99) and positive correlations for Mg (r = 0.86), Ca (r = 0.68), and P (r = 0.58). Such strong correlations approach the level of agreement seen in studies where sap testing has been calibrated as a diagnostic tool. For instance,5 reported an average r = 0.85 between petiole sap NO₃⁻-N and the NNI, with sap tests correctly indicating N status for roughly two-thirds of the crop cycle in tomatoes in field conditions. Similarly,21 found the presence of NO₃⁻ in tomato or muskmelon could predict the NNI with R2 values ranging from 0.88 to 0.95. Our finding of r = 0.99 for lettuce K with TC suggests that, at least for that nutrient in lettuce, a quick sap test could be as diagnostically powerful as a full tissue analysis. The most likely explanation is that lettuce leaves are succulent and contain high proportions of water and readily exchangeable ions, so pressing the sap yields a solution that is compositionally similar to the total tissue nutrient content (especially for highly mobile nutrients like K⁺ that reside largely in cell sap/vacuoles). By contrast, in crops with more fibrous or woody tissues like tomato (thicker petioles, more structural binding of nutrients), even TC sap did not align well with total nutrients (tomato Ca r = + 0.21, Mg near 0). The structural differences mean that pressing a tomato petiole yields a small volume of sap that might disproportionately come from xylem or phloem sap and some nutrients stored in structural or cellular compartments, whereas in lettuce, the whole leaf tissue is soft, and the sap may be more representative. This discrepancy also resembles the findings by25 in raspberry (Rubus idaeus), they also observed inconsistent correlations between petiole sap and leaf tissue nutrients in raspberry, likely due to the heterogeneous distribution of nutrients in woody canes vs leaves. They also noted that to get a representative sap reading in such crops, many subsamples were needed. In fact,6 determined that due to spatial and temporal variability in greenhouse tomato, between 25 and 113 petiole subsamples might be required to reliably capture the true average sap NO₃⁻ or K of a crop at a given time. Our use of pooled samples aimed to mitigate some variability, but it highlights that crop physiology and anatomy play a decisive role in sap analysis reliability, for example, in lettuce (a herbaceous, high-water-content crop), which may differ from that of tomato (characterized by more lignified tissue). The NO₃⁻ comparisons in sap underscored how differently each crop reacts. Lettuce showed a strong correlation between TC and KCl extract NO₃⁻ (r = 0.72), as shown in Table 4, indicating that a quick sap test (such as a handheld meter on lettuce petiole juice) could be quite reliable for assessing NO₃⁻ status in lettuce. This makes sense given that lettuce is harvested at a vegetative stage when NO₃⁻ is often stored in tissues; sap NO₃⁻ can be directly indicative of how much NO₃⁻ is taken up relative to growth. The most likely physiological explanation is that sap NO₃⁻ reflects a more immediate and mobile nitrate pool, whereas total tissue N represents a larger accumulated pool that includes not only NO₃⁻, but also reduced and assimilated forms such as amino acids and proteins24. In addition, NO₃⁻ can accumulate substantially in vacuoles, while the NO₃⁻ reduction occurs primarily in the cytosol, so the relationship between sap NO₃⁻ and total tissue N may be weakened by intracellular compartmentation. This interpretation is further supported by evidence that vacuolar NO₃⁻ can be released back into the cytosol through tonoplast-localized transporters, contributing to NO₃⁻ remobilization and root-to-shoot reallocation rather than remaining in a static storage pool23. These distinctions are especially relevant when comparing leafy crops with fruiting crops, because NO₃⁻ uptake, storage, remobilization, and assimilation vary strongly with crop type and developmental stage. For tomatoes, however, our data agree with13 who found that petiole sap NO₃⁻ is not well correlated with total N in tissues or yield at different developmental stages. Tomato actively assimilates NO₃⁻ into organic N (amino acids, proteins), especially once fruiting begins, so the sap NO₃⁻ can drop even if the plant’s overall N is sufficient (the NO₃⁻ is being rapidly used). We observed a negative correlation (r = − 0.38) between TC and KCl NO₃⁻ in tomato, possibly indicating that at times, high tissue N (from protein) coincided with low sap NO₃⁻ because the plant had already converted it. Moreover, NO₃⁻ uptake itself is under strong regulatory control, and internal N status can feed back to suppress further nitrate uptake and transport, which may further decouple sap NO₃⁻ from total tissue N under certain conditions26.The cucumber was intermediate, showing slight positive correlations around 0.25, but certainly not strong enough for a confident test. These results emphasize that sap NO₃⁻ needs crop-specific interpretation.27 found that different cultivars of cucurbits, such as cucumber and melon (Cucumis melo L.), as well as solanaceous crops like pepper (Capsicum annuum), contained baseline sap NO₃⁻ levels varied by several hundred ppm under the same fertigation regime. For fruiting crops such as tomato, more sophisticated approaches may be required to interpret sap data reliably. From a practicality and cost standpoint, one of the clearest outcomes of our evaluation is that the TC (pressed sap) method is not only among the more reliable methods (especially for a crop like lettuce, and reasonably well correlated to NLA for some nutrients in tomato), but also the simplest and most cost-effective option by a substantial margin. This has real implications for growers and crop consultants. A method that costs approximately $7 in labor (perhaps ~ $32 total, including lab analysis) and gives immediate results could be adopted for frequent monitoring, whereas methods that require lab processing at $15-$20 labor plus days of waiting (like DF, EE, or NLA) will only be used on a limited scale. Our cost analysis aligns with the notion that rapid tests can significantly cut costs and enable more frequent nutrient monitoring.25 quantified this in a multi-year trial by performing sap analysis with portable ion meters in a raspberry (Rubus idaeus) production system, which was estimated to cost about $3,530 over 6 years, versus ~ $6,870 for the equivalent number of conventional lab leaf analyses, which was nearly a $3,340 savings (~ 49% reduction). Our analysis similarly shows that aside from initial setup costs (e.g., purchasing a sap press and handheld meters), the ongoing expenses of sap testing are minimal, largely just labor. TC essentially requires only a simple press (even a garlic press can suffice) and a portable meter or test strips for the nutrients of interest.28 noted that any monitoring tool for high-input systems like hydroponics needs to combine analytical reliability with operational simplicity to be scalable, and a criterion that direct sap tests can meet. By contrast, the more complex methods (EE, DF) not only cost more per sample but also involve safety hazards and cannot be performed without a full lab, making them unsuitable for routine testing in CEA. It is worth noting that emerging technologies and approaches may help to bridge this gap where sap extraction alone is insufficient. For example, integration of sap analysis with other diagnostic tools can improve strength of nutrient assessment.29 reported that the apple (Malus domestica) orchard nutrition with both leaf analysis and petiole sap (using a reflectometer for sap), finding poor direct correlations between leaf and sap for most nutrients, but strong agreement (R2 = 0.70–0.78) between sap NO₃⁻, K, Ca, Mg measured by a portable reflectometer kit and the same nutrients measured by standard lab methods. This suggests that if you use a properly calibrated device (RQflex reflectometer strips) and a standardized sampling procedure, you could acquire sap readings that closely approximate lab values for those nutrients. Similarly,30 used a machine-learning model that took SPAD chlorophyll meter readings, weather data, and cultivar info to predict potato (Solanum tuberosum L.) petiole sap NO₃⁻ and derived a vine N status index. Their model achieved a validation R2 = 0.80 and about 75% accuracy in classifying N status. In other words, rather than relying on sap NO₃⁻ alone (which can be variable in potato), they used sap NO₃⁻ plus other easily measured factors to get a reliable indicator. This type of data could be very useful in CEA, where environmental data is already being tracked. Another study carried out by31 in spring oat (Avena sativa L.) established critical sap NO₃⁻ thresholds corresponding to optimal yield ~ 650 mg/L at the early jointing stage and ~ 540 mg/L at the booting stage. They combined sap and SPAD readings to guide N fertilization in cereals. While oat in open fields is far from CEA lettuce, the concept of having quantitative sap nutrient guidelines and using them similarly with the traditional tissue sufficiency ranges is the main objective. Determining these guidelines for CEA crops will require many trials to correlate sap values with plant performance (yield, quality) under various conditions, and likely the thresholds will need to be crop and cultivar specific. This study should also be interpreted within the boundaries of its experimental scope. The evaluation was limited to three CEA crops, selected cultivars, specific developmental stages, and the extraction methods tested under the environmental and production conditions included in this work. Because sap composition is strongly influenced by crop type, tissue sampled, genotype, plant developmental stage, and environmental conditions, the observed relationships with NLA may not be directly transferable to other crops, cultivars, nutrients, or production systems without further calibration. In addition, the correlation-based comparisons used here describe associations among methods within this experimental framework, but they do not establish analytical equivalence, universal diagnostic thresholds, or direct method interchangeability across all conditions. Therefore, broader validation across crops, environments, and management scenarios is still required before sap analysis can be generalized as a routine substitute for conventional tissue analysis. While none of the tested sap extraction methods could consistently replace traditional tissue analysis for all nutrients and crops, we identified TC as the most broadly useful (and cheapest) method, with reliability in lettuce. Chemical extractants (AA, KCl) and freeze/solvent methods provided inconsistent results and would require careful crop-specific calibration for accurate use. The path forward for sap analysis in CEA likely involves refining these methods (e.g., improving TC sap collection volume and consistency, or developing a universal simple extractant) and pairing them with robust calibration data and complementary diagnostic tools. Materials and methods Study location This study was performed at three different commercial CEA production locations in Georgia (GA), USA, with different cultivation techniques. The first location, Pure Flavor Southern Flavor Farms in Fort Valley (32.5540°N, -83.8122°W), is a high-wire greenhouse facility specializing in the cultivation of tomatoes and cucumbers. The second location, Kalera in Ellenwood, GA (33.6572°N, -84.3179°W), utilizes a multi-tier vertical farming technique for cultivating lettuce. The third location, Local Bounti greenhouses in Byron, GA (32.6469°N, -83.7171°W), is a commercial greenhouse dedicated to lettuce cultivation. Plant material and environmental conditions Tomato and cucumber (pure flavor) Grafted transplants of tomato cultivars ‘Grandice’, ‘Macxize’, and ‘Endeavour’ and cucumber cultivars ‘Georgia’, ‘Seiber’, and ‘Keystone’ were transplanted at approximately 30–40 days after seeding on rockwool slabs (100 × 15 × 10 cm) (Grodan, Roermond, The Netherlands). The planting densities were set at 2.5 plants·m⁻2 for tomato and 2.0 plants·m⁻2 for cucumber. The environmental conditions were precisely controlled; night temperatures were maintained at 13 °C, gradually increasing to 16 °C at dawn. Daytime temperatures were kept below 28 °C for tomatoes and ≤ 30 °C for cucumbers, with mean 24-h temperatures of 23 °C for tomatoes and 21 °C to 22 °C for cucumbers even during summer months. Supplemental illumination was supplied using high-pressure sodium (HPS) lamps, with photoperiods established at 18 h for tomatoes and 20 h for cucumbers. Carbon dioxide (CO₂) levels were elevated yet remained below 1800 mg·L⁻1 for tomatoes and 1500 mg·L⁻1 for cucumbers. Drip fertigation provided a nutrient solution with an electrical conductivity (EC) of 3.2 to 3.5 dS·m⁻1 for tomatoes and 2.2 to 2.5 dS·m⁻1 for cucumbers, with a targeted pH of 5.2 for both crops. Vertical-farm lettuce (Kalera) Seeds of the lettuce cultivars ‘Rouxai’, ‘Fairly’, and ‘Lalique’ were pre-germinated at 24 °C and 100% relative humidity until radicle emergence (24 h) and subsequently transplanted into 2.5 × 2.5 cm peat-based grow media plugs (Quickplugs™, Grow-Tech LLC, South Portland, ME, USA). Seedlings were propagated for 12 days in a controlled growth room under light-emitting diode (LED) lighting (Signify, Eindhoven, The Netherlands) with a far-red/red/deep red/blue low-blue spectrum, providing a photosynthetic photon flux density (PPFD) of 200 ± 5 µmol·m⁻2·s⁻1 and an 18-h photoperiod. system using a modified Hoagland’s nutrient solution. After propagation, seedlings were mechanically transplanted at a density of 40 plants·m⁻2 and grown to harvest over 27 d in a dedicated grow room. Lighting conditions remained at 200 ± 5 µmol·m⁻2·s⁻1 PPFD but with a reduced photoperiod of 12 h, provided by broad-spectrum LED lamps (Signify, Eindhoven, The Netherlands) with a red/deep red/blue low-blue spectrum, consistent with recommendations for lettuce production under vertical farming conditions. During the production phase, plants were sub irrigated by using a shallow water culture system with a continuously supplied modified Hoagland’s nutrient solution, electrical conductivity (EC) 1–4 – 1.6 mS/cm, pH 5.7–5.9, dissolved oxygen (DO) 18–25 ppm, oxidation reduction potential (ORP) 350–450 mV. The environmental conditions during the propagation and production stages were maintained at 21–23 °C, with a relative humidity of 65% to 80%, Vapor Pressure Deficit 0.8–0.9 kPa, and a CO₂ concentration of 900 µmol·mol⁻1, Greenhouse lettuce (local bounti) Lettuce cultivars ‘Deangila’, ‘Rafael’, and ‘Rosalyn’ were used for production at the Local Bounti greenhouse facility. Specific environmental conditions and fertigation parameters at this site were not disclosed due to a non-disclosure agreement (NDA). Experimental design and sampling Tomatoes and cucumbers The study was arranged in a split-plot design with repeated measures for tomatoes and cucumbers for this experiment, where cultivar was treated as the whole-plot factor, developmental stage and tissue type (leaf or petiole) as repeated measures factors, and sap extraction method as the split-plot factor. For each crop, three cultivars were evaluated (tomato: ‘Grandice’, ‘Macxize’, ‘Endeavour’; cucumber: ‘Georgia’, ‘Seiber’, ‘Keystone’), with three replicates per cultivar. Each replicate consisted of 20 plants. Sampling was conducted repeatedly at four developmental stages, including first fully expanded leaves at 60 days after transplantation (DAT), 1/3 developmental stage (140 DAT), 2/3 developmental stage (190 DAT), and final harvest (251 DAT) (Fig. 3). At each sampling time, one representative leaf was collected from each plant within the replicate. To control for potential differences in nutrient partitioning, the leaves were pooled by replicate into a single sheaf, subdivided into six equal units, and randomly assigned to six sap extraction methods. After sap was extracted, nutrient concentrations were measured for each sample. Similarly, petioles from the same plants were collected at each sampling, pooled by replicate, subdivided into six units, randomized across the six extraction methods, and analyzed similarly for nutrient concentrations. This design resulted in 9 experimental units (three replicates nested within each of three cultivars), measured repeatedly at four developmental stages. At each stage, two tissue types (leaf and petiole) were sampled, and within each tissue type, six subsamples were analyzed using the six sap extraction methods.Fig. 3Sampling workflow for tomato and cucumber at Pure Flavor. Tomato cultivars were ‘Grandice’, ‘Macxize’, and ‘Endeavour’, while cucumber cultivars were ‘Georgia’, ‘Seiber’, and ‘Keystone’. For each crop, 104 samples were collected, each consisting of one representative leaf from 20 plants. Sampling was carried out at four developmental stages: fully expanded leaf, one-third developmental stage, two-thirds developmental stage, and final harvest stage. Sampling workflow for tomato and cucumber at Pure Flavor. Tomato cultivars were ‘Grandice’, ‘Macxize’, and ‘Endeavour’, while cucumber cultivars were ‘Georgia’, ‘Seiber’, and ‘Keystone’. For each crop, 104 samples were collected, each consisting of one representative leaf from 20 plants. Sampling was carried out at four developmental stages: fully expanded leaf, one-third developmental stage, two-thirds developmental stage, and final harvest stage. Lettuce The study was arranged in a split-plot design with repeated measures, where cultivar represented the whole-plot factor, developmental stage served as the repeated measures factor, and sap extraction method was the split-plot factor for lettuce. Data were collected from three lettuce cultivars grown in the greenhouse (‘Deangila’, ‘Rafael’, and ‘Rosalyn’) and three cultivars from the vertical farm (‘Rouxai’, ‘Fairly’, and ‘Lalique’), with three replicates for each cultivar at both locations. Each replicate consisted of 20 plants. Sampling was conducted at two developmental stages: half development 12 DAT and final harvest at 24 DAT (Fig. 4). Because preliminary analyses indicated no significant differences or interactions in correlations between location, data from both environments were pooled for subsequent analysis to provide a single, system-independent estimate of sap extraction method performance. At each sampling time, one representative leaf from each plant was collected, pooled by replicate, assembled into a single sheaf, and then subdivided into six equal portions, which were randomly assigned to the six sap extraction methods. After sap extraction, nutrient concentrations were measured for each sample. This design resulted in nine experimental units (three replicates nested within each of three cultivars per location), measured repeatedly at two developmental stages. At each stage, six subsamples were analyzed using the six sap extraction methods.Fig. 4Sampling workflow for lettuce from Local Bounti (greenhouse) and Kalera (vertical farm). Greenhouse cultivars included ‘Deangila’, ‘Rafael’, and ‘Rosalyn’, while vertical farm cultivars included ‘Rouxai’, ‘Fairly’, and ‘Lalique’. For each production system, 54 samples were collected, each consisting of one representative leaf from 20 plants. Sampling was conducted at two developmental stages: half-developmental stage and final harvest stage. Sampling workflow for lettuce from Local Bounti (greenhouse) and Kalera (vertical farm). Greenhouse cultivars included ‘Deangila’, ‘Rafael’, and ‘Rosalyn’, while vertical farm cultivars included ‘Rouxai’, ‘Fairly’, and ‘Lalique’. For each production system, 54 samples were collected, each consisting of one representative leaf from 20 plants. Sampling was conducted at two developmental stages: half-developmental stage and final harvest stage. Sap extraction methods Six sap extraction methods were evaluated for their ability to release and allow for nutrient quantification from plant tissues of three different crops: tomato, cucumber, and lettuce. NLA method was utilized as a control, with AA, DF, EE, KCl, and TC. Due to the chemical composition of these extractants, nutrient analysis varied among methods. K was excluded from the KCl method because it was already present in the solution. Similarly, NO3⁻ was not measured in NLA, AA, or DF extracts because the digestion and extraction reagents contain N. Therefore, NO₃⁻ concentrations were only quantified from the TC, EE, and KCl methods. NLA (control) Tissue samples were oven-dried at 65 °C for approximately 24 h until a constant weight was achieved, ground in a Wiley mill (Thomas Scientific, Swedesboro, NJ, USA), and then passed through a 20-mesh stainless-steel screen. Subsamples ~ 0.5 g were digested according to the U.S. Environmental Protection Agency Method 305232. Each sample was weighed into fluorocarbon polymer microwave vessels, and 10 mL of concentrated nitric acid (HNO₃) was added. Vessels were sealed and processed in a microwave digestion system (Mars 6 Microwave; CEM Corp., Matthews, NC, USA) at 200 °C for 30 min. Digests were quantitatively transferred into volumetric flasks and diluted to a final volume of 50 mL with deionized water before analysis. Plant tissue digests were analyzed for P, K, Ca, Mg, S (macronutrients) and B, Cu, Fe, Mn, and Zn (micronutrients) following EPA Method 200.833. Elemental concentrations were determined using ICP-OES (Agilent 5110; Agilent Technologies, Santa Clara, CA, USA). Calibration standards were checked, and independent laboratory performance checks confirmed that recoveries were within 100% ± 5% of certified reference values. Because the digestion procedure contains HNO₃, the NLA method was not used for NO₃⁻ analysis. AA The AA extraction was performed following the procedure described by34. Plant tissues were cut into 1–2 cm pieces, weighed, and extracted in 20 mL of 0.1 N ammonium acetate solution (CH₃COONH₄, pH 7.0). Samples were shaken at 250 rpm for 15 min using a reciprocal shaker (SH-200 series; Cole-Parmer, Stone, Staffordshire, UK), filtered through folded Whatman No. 1 filter paper of 11 cm diameter (Cytiva, Marlborough, MA, USA). DF The DF extraction was adapted from35 but with the modifications for plant tissue. Fresh tissues were cut into 1–2 cm pieces, immediately frozen in LN₂, and then pulverized into a fine powder using a pre-chilled mortar and pestle stored at -80 °C. During homogenization, LN₂ was continuously added to the mortar to maintain cryogenic conditions, and tissues were kept submerged in LN₂ between grinding steps to prevent thawing. After pulverization, extraction was performed following the procedure described by34 using 20 mL of 0.1 N ammonium acetate solution (CH₃COONH₄, pH 7.0). The mixture was shaken at 250 rpm for 15 min using a reciprocal shaker (SH-200 series; Cole-Parmer, Stone, Staffordshire, UK) and then filtered through folded Whatman No. 1 filter paper with 11 cm diameter (Cytiva, Marlborough, MA, USA). EE Organic solvent-based extraction with ethyl ether was conducted following36. Plant tissues were cut into 1–2 cm pieces and submerged in 20 mL of analytical grade ethyl ether in centrifuge tubes. Samples were frozen at -15 °C overnight to induce cell rupture, then thawed and shaken at 250 rpm for 15 min using a reciprocal shaker (SH-200 series; Cole-Parmer, Stone, Staffordshire, UK). The mixture was filtered through folded Whatman No. 1 filter paper of 11 cm diameter (Cytiva, Marlborough, MA, USA) to separate the pigment-rich ether phase from the aqueous sap fraction. Due to immiscibility, the aqueous sap settled at the bottom of the tube, while pigments remained in the ether layer. To enhance sap recovery, 20 mL of deionized water was added before phase separation. Because freezing ruptured cellular membranes, the extract represented a mixture of true sap and cytoplasmic contents. KCL Nutrients were extracted by using KCl, following the method of16. Fresh leaves and petioles were cut into 1–2 cm, weighed, and transferred into centrifuge tubes containing 20 mL of 0.1 N KCl solution (7.456 g KCl L⁻1, pH 7.0). Tubes were shaken at 250 rpm for 15 min using a reciprocal shaker (SH-200 series; Cole-Parmer, Stone, Staffordshire, UK) to release nutrients. The suspensions were then filtered through folded Whatman No. 1 filter paper (11 cm diameter; Cytiva, Marlborough, MA, USA), and the extracts were collected for analysis. TC Tissue crushing is among the earliest and most widely used methods for plant sap analysis, relying on the direct mechanical extraction of sap from petioles or leaves. To avoid contamination with trace metals, crushers or presses constructed from PVC, stainless steel, or nylon are preferred18. Petioles were cut into 5–10 mm pieces and pressed with a stainless-steel garlic press to obtain sap8,14. Washing petioles after cutting has been suggested to reduce sap concentrations of N and K5, and therefore, samples were not washed prior to extraction. Hydraulic presses have been employed to improve extraction consistency and extract larger volumes6,19. Nutrient quantification All macronutrients and micronutrients (P, Ca, Mg, S, Fe, Mn, Zn, Cu, B, and K, where applicable) were quantified using inductively coupled plasma optical emission spectroscopy (ICP-OES) (Agilent 5110; Agilent Technologies, Santa Clara, CA, USA). NO₃⁻ concentrations were determined only in TC, EE, and KCl extracts. NO₃⁻ was analyzed using a gas-diffusion conductimetric method37,38 in which a Zn-reduction cartridge reduces NO₃⁻ and NO₂⁻ to NH₄⁺, which is converted to NH₃ at alkaline pH (11–13). The NH₃ diffuses through a hydrophobic membrane into a boric acid solution, where an increase in electrical conductivity allows conductimetric detection. NO₃⁻ measurements were conducted using an ammonia/nitrate analyzer (TL-2900; Timberline Instrument, Boulder, CO). Comparison and cost estimation of sap extraction methods We conducted a simplified cost analysis of the practical aspects of each method, including labor time, required supplies or reagents, safety considerations, and projected cost per sample to assist with decision-making. The authors note that a more detailed assessment should be made, considering additional costs that were not available to us. To ensure a fair comparison, we adopted a standard sampling intensity required for CEA monitoring. The labor requirements (in minutes) for each method encompassed sample handling from collection to the preparation of an extract for analysis. We estimated a labor cost of $20 per hour for our estimates. The costs of reagents were determined based on standard pricing (e.g., chemicals per sample, filter paper, etc.) in September 2025. Instrument expenses (such as ICP or NO₃⁻ meters) were excluded from the per-sample cost; however, for portable methods like TC, we presumed negligible capital expenditure. The turnaround time (total elapsed time from sample collection to results) was noted to provide context for the speed. The complete assessment (Table 5) was an estimated cost per sample for each method and notes on field applicability (e.g., whether it can be done on-site or requires a lab). Statistical analysis Correlations between each pair of sap extraction methods were estimated separately for each crop and nutrient, averaged over plant part, developmental stage, and cultivar. A linear mixed-effects model was fitted, including fixed main effects and all two-, three-, and four-way interactions for plant part, cultivar, developmental stage, and extraction method. To account for correlations among repeated measures, the model included random effects for each experimental unit and for each experimental unit by developmental stage combination. An unstructured covariance structure was specified for each experimental unit, defined by the combination of developmental stage and plant part, from which inter-method correlations were estimated. For lettuce, only one plant part (leaf) was evaluated; therefore, an unstructured covariance structure was specified for each experimental unit by developmental stage combination, and random effects for experimental unit by developmental stage were omitted. Random effects with estimated variance components equal to zero were excluded from the final models to facilitate model convergence. All models were fitted using restricted maximum likelihood estimation using a statistical software (SAS version 9.4; SAS Institute, Cary, NC, USA). Conclusion This study systematically compared five sap extraction methods with standard nutrient leaf analysis (NLA) across three key CEA crops (tomato, cucumber, and lettuce) to evaluate their practicality and diagnostic nutrient reliability. Overall, correlations between sap-based methods and NLA were generally weak and inconsistent across crops, nutrients, and developmental stages. Some partial patterns were observed, such as relatively stronger correlations for macronutrients in lettuce using the TC method; however, these results were not consistent across tomato and cucumber. Solvent and reagent-based methods (AA, DF, EE, KCl) showed limited diagnostic value, with frequent low or negative correlations, further underscoring the challenges of standardizing sap analysis. Sap NO₃⁻ determination was particularly variable across crops and methods, reflecting both methodological limitations and physiological dynamics of NO₃⁻ assimilation. Cost analysis highlighted TC as the most economical and field-feasible method due to its minimal labor requirements and absence of hazardous reagents. However, its diagnostic utility remains crop-dependent rather than universally reliable. These findings emphasize that none of the evaluated sap extraction methods consistently matched NLA across all crops, nutrients, and sampling conditions examined in this study. Therefore, sap analysis should not be considered a universal substitute for NLA under the conditions tested here. Moving forward, crop-specific calibration, standardized sampling protocols (e.g., Number of leaves, leaf age, growth stage, diurnal timing), and integration with complementary diagnostic approaches are needed to improve the accuracy and robustness of sap analysis in CEA systems. Future research should focus on refining, sap-based interpretive frameworks and crop-specific sufficiency ranges, accounting for genotypic and temporal variability, and testing methods under different production and stress conditions to advance precision fertigation strategies. In addition, the findings should be interpreted within the scope of this study, which was limited to three CEA crops, selected cultivars, and specific growing environments and extraction protocols.