The barley lines used in this screening experiment were from a diversity set assembled to represent the breadth of European barley breeding during the 20^th century (Xu et al., 2018). The set studied here comprised 81 spring barley lines, 72 two-row and 9 six-row, from 14 countries. A subset of 18 varieties were selected from the 81 as exemplars of their water use strategy, as described below. The subset comprised 12 two-row and 6 six-row lines from 7 countries. The 18 chosen lines are distributed across the diversity space ( Supplementary Figure S1 ), as revealed by 864 gene-based single nucleotide polymorphism (SNP) markers (Tondelli et al., 2013). Details for 81- and 18-line sets, including their release year, country of origin, pedigree and breeders, are given respectively in Supplementary Tables S1 and S2 .

Seeds were sown into 50 ml cones filled with peat soil, on trays, one seed per cone, and the trays covered with plastic wrap and aluminum foil for two weeks at 4°C as a means to break dormancy and to enhance germination (Galkin et al., 2018). The trays were then placed in a controlled glasshouse under short-day conditions (8/16 hr light/dark, 16/10°C day/night). Following emergence, the seedlings were grown for either 8 weeks (81-line experiment) or 12 weeks (18-line experiment) in total. Seedling roots were washed and the seedlings transplanted into potting soil (Dalal et al., 2020; Paul et al., 2023) in 4L pots and placed into a semi-controlled greenhouse. After two weeks of acclimatization, the pots were mounted on the lysimeter system ( Supplementary Figure S2 ). To maximize homogeneous exposure to the ambient conditions on the lysimeter platform, all the pots were placed in random order. Temperature and relative humidity (RH) were respectively around 24–33°C and 30–60% for the 81-line experiment, and 20–35°C and 20–80% for the 18-line experiment. The PlantArray lysimeter platform (Plant-Ditech Ltd., Israel; Dalal et al., 2019) includes a meteorological station, continuously records the physiological conditions of the experiment and is equipped with an automated irrigation system. During the 81-line and 18-line experiments, the system recorded minimum and maximum mid-day daily light intensities ranging from 116 to 493 mmol s^-1 m^-2 and 74 to 514 mmol s^-1 m^-2 respectively ( Figures 1A ; Supplementary Figure S3A ), and vapor pressure deficits (VPD) of 1.5 to 4.3 kPa (average 2.6 kPa) and 0.6 to 4.8 kPa (average 2.2 kPa) respectively.

Drought experiments were carried out as previously on the PlantArray platform, which is a high-throughput physiological diagnostic system consisting of a highly sensitive, temperature compensated multi-lysimeter array (Halperin et al., 2017; Dalal et al., 2019; Appiah et al., 2023; Paul et al., 2023) and comprised three phases: pre-treatment, drought, and rewatering. In pre-treatment, the plants were maintained on the lysimeter platform and were well watered. Under drought, the plants were exposed to dehydration by limiting the water content in each pot until they reached a specific SWC. Irrigation was carried out as described in Halperin et al. (2017). Drought was imposed as described in Paul et al. (2023), but only one drought phase. During rewatering, plants were irrigated as in the pre-treatment phase.

Altogether 81 lines were screened without replication for drought stress response and yield-related QPTs. Of this set, 18 were then chosen for more intensive investigation. For the 18 lines, 5 to 8 biological replications were made, creating a total of 139 plants. For both sets of experiments, a well-watered period, with irrigation at night until each pot reached its full capacity, was followed by a drought treatment in which irrigation was minimized. The QPTs were monitored in real time on a multi-lysimeter platform, with data collected every few minutes by SPAC analytical software throughout the experimental period. The 81-line experiment lasted a total of 29 days: 9 days of pre-treatment, 11 days of drought, and 9 days of rewatering. For the subsequent 18-line in-depth evaluation, the set was given 11 days of pre-treatment ( Supplementary Figure S2B ), followed by 15 days of drought ( Supplementary Figure S2C ) to attain a similar SWC as in the first set. The drought period was terminated for each line individually as soon as it fell below 15% SWC. All lines were then given 29 days of rewatering for recovery ( Supplementary Figure S2D ). Due to the second experiment’s average atmospheric VPD being lower than that of the first, the drought treatment was longer. The total elapsed time of the second experiment was 55 days.

The PlantArray platform continuously collects data and sends it to a central computer for additional analysis by SPAC (soil–plant–atmosphere continuum) analytical software of soil and atmosphere data alongside plant traits (Dalal et al., 2020). The dataset included daily data (one value per day for each plant) and momentary data acquired every 3 min (480 values per day for each plant). The processed data allowed us to determine the physiological parameters of single whole plants in individual pots simultaneously for the entire experimental time, including growth rate, transpiration, stomatal conductance, and WUE (Halperin et al., 2017; Dalal et al., 2019). The physiological parameters were calculated according to Halperin et al. (2017) with data analyses made using Matlab software (MathWorks, Natick MA, USA). We chose five consecutive days during each of the pre-treatment, drought, and rewatering stages, where the VPD and PAR were most stable, to analyze the quantitative physiological traits (QPTs) of the lines in detail. Over the course of the 18-line experiment, these periods comprised days 7 to 11 during pre-treatment, 22 to 26 during drought, and days 51 to 55 during recovery. Using SPSS software, means and standard errors were determined and the one-way ANOVA Tukey Post-Hoc test at a significance level of p=0.05 was carried out for multiple comparisons of the lines. In the reported analyses, lines without significant differences share superscript letters.

Daily transpiration: The VPD and PAR during the whole experimental period is shown in Figure 1A . During the well-watered pre-treatment, daily transpiration of all 81 barley lines increased in parallel but diverged as the plants grew at different rates ( Figure 1B ). During the drought, the lines responded differentially regarding the day and degree to which their daily transpiration (DT) dropped. Similarly, following rewatering on day 21, though all the plants’ DT grew substantially throughout the rewatering phase, the lines responded disparately to water availability and showed varying rates and degrees of recovery, which is a measure of resilience. Initial DT was between 50 and 100 g d^-1, a 2-fold spread, which during recovery enlarged to 100 to 350 g d^-1, a 3.5-fold variation. Within that range, the 81 barley lines could be divided into four groups by their transpirational behavior, from which we extracted 18 lines (Groups A, five lines; B, five lines; C, five lines; D,3 lines) as representative of the range ( Figure 1C ) for the second set of experiments.

Physiological drought point (θ_c): Theta-crit (θ_c) is the critical SWC which becomes a limiting threshold for supporting maximal transpiration values, leading the plant to respond by reducing its transpiration rate (Halperin et al., 2017; Dalal et al., 2019), shown in Figure 2A for the 18 chosen barley lines during the 81-line experiment. The four groups ( Figure 1C ) define four distinct patterns for combinations of θ_c and the maximum transpiration rate (TR_max) under well-watered conditions. Group A and B had similar θ_c, at about ~30% SWC, though at different DT; Group C had a still lower DT and a very low θ_c, at ~20% SWC; Group D had the highest θ_c, ranging from 30 to 35% SWC, but the lowest DT. Comparison between the TR_max and θ_c under drought ( Figure 2B ) shows a significant difference between each group and a correlation between the two measures, TR_max declining from Group A successively to B, C, and D under well-watered conditions. The groups showed distinct rates of decline in transpiration rate vs SWC, with Groups A and B declining slowly and Groups C and D precipitously ( Figure 2A ). Supplementary Table S3 contains TR_max and θ_c for each of the 18-line subset.

In the replicated and more extensively analyzed experiments with the 18-line subset, we regrouped the lines to take into account their behavior through drought and recovery. Group 1, with highest transpiration, had four lines (Hydrogen, Hankkija 673, Baronesse, Isaria), followed by Group 2 (Etu, Gorm, Gate, Frisia, Barke, Chanell, Favorit) and Group 3 (Arvo, Herse, Formula, Artturi, Eero, Binder, Freja) with seven lines each ( Figure 3 ). During pre-treatment, the overall transpiration of all barley lines ranged from 73 to 207 g day^-1 ( Figure 3A ), which dropped to 36 to 131 g day^-1 ( Figure 3B ) during drought stress and recovered to 174 to 380 g day^-1 ( Figure 3C ). A maximum DT of 207 g day^-1 during pre-treatment was observed in Hydrogen, while the lowest DT was observed in Freja at 73 g day^-1.

During drought, we observed Group 1 transition from the highest to the set with the lowest DT and Group 3 shift from lowest to highest, with the exception of Eero, which remained in the center of the range. The majority of the lines in all three groups reverted to their pre-treatment positions throughout the rewatering phase, even if the DT increased from a maximum of 200 g d^-1 to 350 g d^-1. Hankkija 673, Arvo, and Formula, on the other hand, responded differently during rewatering, remaining at a transpiration level similar to that during drought, not returning to the pre-drought level.

Except for a few outliers, most of the lines with intermediate pre-treatment DT (Group 2) remained in the center of the distribution during both drought and recovery. When the lines were compared between the treatments, we found a strong negative correlation between pre-treatment and drought, with r = -0.83 ( Supplementary Figure S4A ), which is driven by the opposing behaviors of Groups 1 and 3. There was a weak positive correlation, r = 0.58, between pre-treatment and rewatering ( Supplementary Figure S4B ), and a weak negative correlation for drought versus rewatering, r = -0.59 ( Supplementary Figure S4C ), likewise driven by differential behaviors by Groups 1 and 3. The daily transpiration of the 18 barley lines with all biological replicates (139 plants) for 55 consecutive days during pre-treatment, drought, and rewatering ( Supplementary Figure S3B ), with daily fluctuation in VPD and PAR ( Supplementary Figure S3A ), shows similar variations as found in the first, 81-line screening ( Figures 1B, C ).

Transition to drought response at θ_c was highly correlated (r=0.89) with whole plant transpiration levels (E_max) ( Supplementary Figure S5B ); three lines from Group 1 (Hydrogen, Baronesse, Hankkija 673) had both the highest SWC at θ_c and the highest E_max ( Figure 4 ). Freja had both the lowest transpiration (E_max =8.8 ± 0.8 mmol s^-1g^-1) and θ_c (26.7 ± 1.2%). All the lines fell within a transpiration range of 8 to 15 mmol s^-1g^-1, with all θ_c values between 26 to 42% SWC ( Figure 4 ; Supplementary Table S4 ). Hydrogen had not only the highest transpiration (E_max = 15.1 ± 1.3 mmol s^-1g^-1) but also the highest θ_c (42.6 ± 2.9%).

Canopy Stomatal conductance (GSc): During pre-treatment, Hydrogen (816 mmol s^-1g^-1) had the highest GSc, whereas Freja (428 mmol s^-1g^-1) had the lowest ( Figure 5A ). The DT and GSc are highly (r=0.94) correlated, with Groups 1,2, and 3 showing differential response ( Supplementary Figure S5A ). When compared by DT group during pre-treatment ( Figure 3A ), the three lines from Group 1 with the greatest GSc (Hydrogen, Hankkija 673, Baronesse) also had the highest DT, except Isaria, which was in the middle. Similarly, the six lines from Group 3 (Freja, Artturi, Binder, Herse, Eero, and Formula) had the lowest GSc ( Figure 5A ) and DT ( Figure 3A ) values, apart from Arvo in the middle. Nevertheless, the majority of the lines in GSc pre-treatment Group 2 (filled bars) are in the same location as the DT pre-treatment.

During drought, however, Freja (highest) had 7.2X greater GSc than Baronesse (lowest) ( Figure 5B ), although the difference between the highest and lowest GSc was only 2X during pre-treatment and rewatering ( Figures 5A, C ). Like for DT, we observed a shift in Group 3 from lowest to highest GSc under drought, and a flip from highest to lowest GSc in Group 1 ( Figure 5B ). This was also evident when GSc during pre -treatment was compared to GSc during drought using Pearson correlation (r), which revealed a negative correlation of 0.95 ( Supplementary Figure S6A ).

During rewatering ( Figure 5C ), we saw a distinctive GSc pattern in which only three lines (Herse, Artturi, and Eero) out of seven in Group 3 returned to their pre-treatment behavior, while the other three (Binder, Formula, and Freja) did not shift, showing no resilient behavior. Interestingly, Arvo remained in the middle group through all three experimental phases. From Group 1, three lines (Hydrogen, Isaria, and Baronesse) moderately recovered GSc, but not Hankkija 673, which recovered poorly, as it did for DT. Correlation analysis also supports the GSc pattern, where only 2% (r = 0.02) of the lines recovered from stress after 29 days of rewatering ( Supplementary Figure S6B ), whereas 58% of the lines showed recovery in DT ( Supplementary Figure S4B ).

Calculated Plant Weight Gain (CPW): Once rewatering relieves drought-induced desiccation and biomass production begins again, calculated plant weight (CPW) gain per day indicates resumption of growth and recovery rather than simply rehydration. Like for DT and GSc, initial CPW for Hydrogen (22 g d^-1) is significantly higher than for Freja (6 g d^-1; Figure 6A ). Group shifts between experimental phases ( Figures 6A–C ) are similar to those for DT ( Figures 3A–C ). Pre-treatment and drought CPWs are negatively correlated (r =-0.68; Supplementary Figure S7A ), whereas pre-treatment and recovery show a positive correlation (r = 0.68; Supplementary Figure S7B ). Altogether 68% of the 18 lines recovered from stress. Among Group 1, only Hankkija 673 did not regain high CPW, as likewise seen with GSc. Comparing drought and recovery, we found that line ranking shifted in a similar way for DT and CPW with respect to GSc.

The line groupings (1, 2, 3) based on DT proved meaningful also for vigor, defined as average gain in CPW per day during pre-treatment, with the Group 1 lines showing the highest vigor ( Figure 7A ). Most of the lines in Group 3 had the lowest vigor, Freja being the poorest. However, during drought, Group 3 had the highest daily CPW ( Figure 6B ). If resilience is considered a return to a pre-drought CPW rate, then three Group 3 lines show the highest resilience, but three within Group 3 the lowest ( Figure 7B ). However, Group 3 lines, which have a high (or non-responsive) CPW during drought ( Figure 6B ), return to a low daily CPW in absolute terms ( Figure 6C ). The lines displayed as well differential recovery of DT during rewatering compared with drought conditions ( Figure 7C ). The high recovery set included: Hydrogen and Baronesse, comprising half of Group 1; Gorm, the only line from Group 2; Freja, Formula, and Binder of the seven lines in Group 3.

The quantity of biomass or grain produced per unit of water transpired gives an overall idea of the tradeoffs between carbon fixation and water transpiration made during the life cycle of the plant. First, we calculated the mean total biomass (combined dry weight of husk, seed, and shoot; Supplementary Figure S8A ), yield ( Supplementary Figure S8B ), and seed number ( Supplementary Figure S8C ) from each biological replicate for all barley lines. Biomass, yield, and seed number varied respectively from 31.7 to 78.3 g, 0.4 to 14.3 g, and 9.6 to 380.4 per replicate. Isaria (78.3 ± 3.8 g) produced the highest biomass but one of the lowest yields (0.61 ± 0.06 g) and seed numbers (39 ± 5.9), hence, lowest harvest index (0.69 ± 0.08%). Eero had the lowest biomass (31.7 ± 5.1 g), a low yield (1.78 ± 0.5 g) and moderate seed number (54.8 ± 20.2). Hydrogen had high total biomass (72.2 ± 3.5 g) but poor yield (1.3 ± 0.2 g) following drought; Freja produced low biomass (41.3 ± 5.6 g) and yield (0.4 ± 0.1 g) as well as the lowest seed number (9.6 ± 2.7). Notably, Hankkija 673, which is a six-row spring barley from Finland released in 1973, had a modest biomass output (60.3 ± 3.9 g) but one of the highest yields (10.2 ± 1.4 g) and seed numbers (380.4 ± 37.1), despite its poor resilience (which measures overall biomass production) and recovery. As a six-row variety, Hankkija had only a middling seed weight (TGW; Supplementary Figure S9 ). The high yield of Herse ( Supplementary Figure S8B ) appears to be a product of its high seed number ( Supplementary Figure S8C ), high harvest index ( Supplementary Figure S8D ), and relatively high TGW ( Supplementary Figure S9 ). The cumulative transpiration, taken as the sum of the daily transpiration of each plant throughout the experiment, enabled us to calculate the water use efficiency (WUE) for the 18 lines ( Figure 8 ), which ranged from 0.004 to 0.009 g g^-1, with Eero having the lowest (0.0045 ± 0.00032 g g^-1) and Chanell having the highest (0.009 ± 0.0003 g g^-1), twice that of Eero. The WUE of Freja and Hydrogen, which displayed during pre-treatment respectively the lowest and highest values for DT ( Figure 3A ), GSc ( Figure 5A ), and CPW ( Figure 6A ), are however, in the mid-range, with Hydrogen being more efficient (0.007 ± 0.0003 g g^-1) than Freja (0.006 ± 0.0003 g g^-1). Thus Baronesse, in addition to its dynamic transportational drought response, most efficiently makes the tradeoff between water lost and carbon gained.

Our analysis indicated a strong correlation (r = 0.95) between vigor (CPW gain per day) and transpiration, those lines (Group 1) with the highest daily transpiration also having the highest vigor. Vigor is defined as the rate of gain in plant weight. Studies suggest that fractional changes in stomatal conductance lead to changes in transpiration (Drake et al., 2013; Zhang et al., 2019). We found a strong correlation (r = 0.94) between daily transpiration and stomatal conductance under well-watered conditions, confirming this ( Supplementary Figure S7B ). The correlation between higher transpiration and higher vigor therefore indicates that Group 1 lines have the capacity for the increased carbon fixation made possible by greater gas exchange, and moreover fix the carbon, especially Baronesse, with high WUE ( Figure 8 ). However, exceptions such as Isaria and Arvo were observed, where the correlation was weak ( Figures 3A , 4A ). This discrepancy may be related to variations in stomatal density, distribution, size, and number within the cultivated species; our data indicate that Arvo has the highest stomatal density of the lines we have examined (Pereira et al., in prep).

Across the analyzed lines, we saw wide and significantly different responses to drought. Group 1 lines, particularly Hydrogen, Hankkija 673 and Baronesse, which displayed the highest transpiration (E_max; Figure 7A ), total daily transpiration (DT; Figure 3 ), canopy stomatal conductance (GSc; Figure 5 ), daily CPW ( Figure 6 ), and WUE ( Figure 8 ) when well-watered, shift as a group from a water non-conserving to a water-conserving strategy during drought. Group 1 then displays the lowest DT, GSc, and daily CPW gain. Notably, most Group 1 members reach θ_c at a higher SWC than do other lines ( Figure 7A ), between day 16 to 18 (drought began day 12), aligning with their higher transpiration rates. Thus, Group 1 displays a dynamic or plastic response to drought, shifting from high transpiration and growth to low, water-conserving levels during drought. Gate, a Group 2 member by its behavior under well-watered conditions, clustered with Group 1 in having low DT, GSc, and daily CPW gain under drought.

The lines in Group 3 displayed the opposite behavior as those in Group 1, having low DT, GSc, and daily CPW gain under well-watered conditions but comparatively high values under drought. Among Group 3, cv Freja was the most extreme regarding transpiration under non-limiting and limiting water, and likewise reached θ_c at the lowest SWC, on the final day of the drought (day 26). Herse and Artturi, respectively having among the lowest DT under well-water conditions, had together with Freja the highest DT and GSc as well as CPW gain under drought. The three lines also transitioned to θ_c at the lowest SWC among the 18 examined in detail. Group 3 can be described as anisohydric, as they reach θ_c late. Lines in Group 2, except for Gate as described above, took a middle road regarding transpiration parameters both before and during drought, as well as reaching θ_c at intermediate SWC. Compared to Group 3, they can be considered isohydric.

Following the drought treatment, resumption of a full irrigation regime led to an increase in DT in all lines. Only two line – Hydrogen and Baronesse – displayed simultaneously high initial DT ( Figure 3A ) as well as high vigor and recovery and good resilience ( Supplementary Table S5 ; Figure 7A ), Only Hydrogen showed both low drought DT ( Figure 3B ) and high initial GSc ( Figure 5A ). A similar degree of resilience, which is the degree to which a plant resumes growth, could in principle be obtained with either low, medium or high vigor; the line need only return to its initial state. In fact, we observed that the most resilient lines ( Supplementary Table S5 ; Figure 7C ) were in fact all members of Group 3—Binder, Freja, and Formula—in terms or returning most fully to their initial growth rate, which was in any case low. Hence, as illustrated by Group 3, resilience per se is not necessarily a path to the highest post-drought yield. Freja, Formula and Binder also showed the best recovery of pre-drought DT, likewise initially low; among Group 1, Hydrogen and Baronesse recovered DT best. During rewatering, Isaria, Hydrogen, and Baronesse (all Group 1) returned to the highest rates of vigor ( Figure 3C ); Isaria showed the best resilience of Group 1.

We examined which of the physiological measures correlated with the final biomass and yield ( Supplementary Figures S8 , S9 ) in pots on the lysimeter. The top five for biomass included three from Group 1 and two from Group 2; while all seven from Group 3 (the anisohydric lines) were found at the bottom of the range. The top of the DT range (Group 1, Figure 3 ) during pre-treatment and rewatering, as well as for pre-treatment and rewatering CPW vigor ( Figures 6A, C ), showed the best correlation with harvest biomass (Hydrogen, Baronesse, and Isaria) or yield (Hankkija 673). The superior yield of Hankkija 673 despite its mediocre biomass is due to its first-rank harvest index, being a six-row variety, and above average grain weight ( Supplementary Figure S9 ). Yields in greenhouse pot experiments, with highly constricted soil volumes and root architectures, are not expected to be closely similar to those from field experiments. Nevertheless, on the lysimeter, carbon capture as indicated by CPW and as the tradeoff for DT still correlates with final biomass.

The set of lines examined here can be categorized by the two formally contrasting drought strategies, “isohydric”, or water-conserving, and “anisohydric”, or water-non-conserving (Sade et al., 2012; Moshelion et al., 2015; Dalal et al., 2017). Isohydric plants would be expected to limit transpiration and to transition to θ_c at a relatively high SWC in order to maintain constant leaf water potential. Anisohydric plants trade a constant hydricity for higher gas exchange and carbon fixation, which would put them at risk in prolonged droughts. Based on their physiological responses, we classify Freja, Artturi, Herse, and Binder as the most anisohydric (all in Group 3), or least drought-responsive, displaying the highest transpiration under drought and reaching the θ_c at low SWC ( Figure 7A ) on the last days of that phase. Group 2 is more isohydric than Group 3, having higher pre-drought DT and a higher θ_c ( Figure 7A ).

In earlier experiments, a single cultivar, RGT Planet, was identified as representing a third drought response type, which may be described as dynamic or plastic (Appiah et al., 2023). This variety (released 2010) is currently the most popular malting barley in Europe, in part due to its consistent and high yields under farm conditions. On the PlantArray system, RGT Planet displayed high transpiration under well-watered conditions, followed by a moderate transpiration decrease under drought, conferring high resilience and, in the pot experiments, high yields that were not significantly reduced by drought. Here, we have established by screening 81 varieties that the dynamic drought strategy is not unique to the relatively new RGT Planet but is found in older lines as well. Group 1, including the “star” lines Hydrogen and Baronesse, displays this dynamic response to drought, transitioning from high-transpiration, water non-conserving to water-conserving behavior, with θ_c at high SWC ( Figure 7A ). While two- and six-row barley generally derive from distinct breeding programs, resulting in genetic structure ( Supplementary Figure S1 ), the lines analyzed here did not divide categorically by row number into drought response strategies. Neither did drought strategy correlate with year of release: Group 1 spans 1924 (Isaria) to 1999 (Hydrogen).

Ultimately, from the breeder’s and farmer’s perspectives, the optimal drought response strategy is the one that produces the highest sustainable yield and a tolerable annual variation, commensurate with yearly fluctuations in the growth environment – drought, in this case – and a particular level of inputs. A high WUE may favor effective growth and ultimately yield when water is limiting; two Group 2 (isohydric) lines, Chanell and Gorm, together with Baronesse of Group 1 (dynamic), showed the highest ( Figure 7B ). Yield stability despite an early-season drought is expected to benefit from recovery of carbon fixation and thereby growth, as reflected in our experiments by DT and vigor (CPW). For the varieties, growth conditions, and drought treatment here, membership in Group 1, with a dynamic strategy, tended to favor higher biomass. Although they reached θ_c at a high SWC, Group 1 members nevertheless lost weight (i.e. dehydrated) most rapidly ( Figure 6B ) during the drought treatment while displaying a low DT during drought ( Figure 3B ). Isaria showed the highest rewatering DT and yet reached θ_c at a lower SWC, while Hankkija 673 responded at a high SWC, but recovered to a low DT.

We observed a correlation between transpiration (DT, Figure 3 ; GSc, Figure 5 ; E_max, Figure 7A ) and SWC at θ_c ( Figure 7A ), with the dynamic lines (Group 1) having both the transpiration measures and θ_c high and the anisohydric both low (Group 3). This raises the questions of how and why the high DT, high vigor lines respond quickly to decreases in SWC, but the low DT lines reach θ_c only at low SWC, i.e. what controls stomatal aperture as SWC falls, and what the consequences of these water-use behaviors are. Enhanced GSc (such as in Group 1) is correlated with higher demand for root and stem hydraulic conductivity (Kudoyarova et al., 2011); increased demand under restricted hydraulic conductance from drying soil drives down water potential in the water column between root and leaf (Carminati and Javaux, 2020; Abdalla et al., 2022; Yang et al., 2023). Abundant evidence connects stomatal closure to drying signals from roots, with ABA and likely other factors implicated in the signaling (Saradadevi et al., 2017); ABA plays a central role in stomatal closure (Assmann and Jegla, 2016; Hsu et al., 2021). The overall picture therefore suggests that high DT, GSc, E_max lines, which are mostly found in Group 1, place high demand on hydraulic conductivity, which becomes limiting, setting off a drought signal to the stomatal guard cells at even moderately reduced SWC. Low conductivity lines, mostly in Group 3, in this model, put more limited demands on conductivity; SWC falls greatly without triggering stomatal closure, a behavior which has been characterized as non-conserving or anisohydric.

The rapid response to falling SWC among Group 1 was associated, particularly for Hydrogen and Baronesse, with the highest degree of recovery. Even though their weight fell rapidly during drought, Isaria, Hydrogen, and Baronesse were the quickest to gain weight during rewatering ( Figure 6C ; Supplementary Table S4 ), while displaying a high DT ( Figure 3 ). We earlier analyzed the gene networks upregulated and downregulated by drought and recovery in the anisohydric Golden Promise (Paul et al., 2023); autophagy was downregulated during recovery and thereby implicated in drought response. The connection of autophagy to differing degrees of recovery requires further investigation. The results presented here suggest that a dynamic drought response combined with rapid recovery, such as displayed by Group 1 here and by RGT Planet (Appiah et al., 2023), may offer a good ideotype by which to achieve yield stability under droughts of up to two weeks (conditions here). There have been many efforts to model both stomatal conductance (Buckley, 2005; Damour et al., 2010; Buckley and Mott, 2013) and the optimization of conductance vs. carbon gain (Sperry et al., 2017; Joshi et al., 2022).

While the literature touching on drought in barley is extensive (over 600 articles in PubMed), there has been a paucity of studies examining germplasm sets (10s or 100s of lines), rather than pairs of lines, with precision phenotyping such as is possible on the PlantArray platform. GWA, quantitative trait (QTL), and genomic prediction (GP) studies for yield and biomass or their components in pot- or field-grown barley populations under drought (Moualeu-Ngangué et al., 2020; Li et al., 2022; Ogrodowicz et al., 2023; Abdelghany et al., 2024), or on traits connected to drought tolerance or yield (Honsdorf et al., 2014; Dhanagond et al., 2019; Luo et al., 2020; Fusi et al., 2022; Puglisi et al., 2022) are more extensive, however. We anticipate that current collaborative work applying nested association-mapping populations, field and precision phenotyping, ideotyping, crop modeling, GWA, and multi-omic approaches will allow us to integrate the results reported here, based on the aforementioned background, into a practical framework for breeding improved drought tolerance and resilience into barley and other crops.