Leaf energy budgets are governed in part by the absorbance of incoming solar radiation and exchange of latent and sensible heat energy (Fauset et al., 2018;Lambers et al., 2008;Michaletz et al., 2015). Environmental conditions within plant canopies such as sunlight, air temperature, humidity and wind speed influence leaf radiant heating and heat transfer between leaves and the surrounding microclimate (Gutschick, 2016;Jones, 2014;Michaletz et al., 2016). Leaf size and conductance to water vapour alter leaf temperature by governing thickness of leaf boundary layers and how much heat loss occurs through sensible and latent heat flux per unit surface area (Dong et al., 2017;Leigh et al., 2017;Michaletz et al., 2016). However, the extent to which plants in natural environments are adapted to regulate leaf temperature in response to thermal stress is largely unknown.
Leaf carbon budgets are tightly coupled to leaf energy budgets because increases in leaf temperature (T leaf ) (see Table 1 for definitions and abbreviations) above an optimal temperature reduces Plant Cell Environ. 2022;1-18.
wileyonlinelibrary.com/journal/pce © 2022 John Wiley & Sons Ltd.
| 1 photosynthetic rates while increasing rates of respiration (Teskey et al., 2015). Likewise, leaf temperature affects the solubility of CO 2 in the liquid phase, kinetics of Rubisco, electron transport efficiency, and mesophyll conductance (Cen & Sage, 2005;Lambers et al., 2008;Yamori et al., 2006). In particular, high leaf temperatures increase rates of photorespiration and subsequently negatively affect net photosynthesis (Atkin et al., 2006;Berry & Bjorkman, 1980;Lambers et al., 2008;Wahid et al., 2007). Exposure to extreme heat waves can also damage photosynthetic processes as high temperatures disrupt cell membranes and metabolism (Hazel, 1995). Therefore, traits that facilitate the maintenance of leaf temperatures close to the optima for photosynthesis should be highly favoured by selection (Helliker & Richter, 2008;Michaletz et al., 2015;2016;Slot & Winter, 2016).
Leaves exhibiting morpho-physiological traits that modify thermal fluxes can display substantial differences between T leaf and air temperature (T air ) (Blasini et al., 2020;Leigh et al., 2017;Michaletz et al., 2015;2016). Leaf size, width, shape, orientation, reflectance and stomatal density can all modify T leaf relative to T air (Beerling et al., 2001;Leigh et al., 2017;Michaletz et al., 2015;O'sullivan et al., 2017;Wright et al., 2017). For example, under hot air temperature and high irradiance conditions, larger leaves are particularly susceptible to experience damaging leaf temperatures because they form thicker boundary layers that slow sensible and latent heat loss (Farquhar & Sharkey, 1982;Lambers et al., 2008;Martin et al., 1999;Wright et al., 2017). Consequently, larger leaves (i.e., large surface area) tend to display larger leaf-to-air temperature differences than smaller leaves (Leigh et al., 2017;Wright et al., 2017). On the other hand, because the high latent heat vaporization of water, stomatal regulation and subsequent leaf evaporative cooling caused by transpiration is arguably the most effective mechanism for regulating leaf temperature in extreme hot environments (Curtis et al., 2012;Drake et al., 2018;Hetherington & Woodward, 2003;Radin et al., 1994;Upchurch & Mahan, 1988).
Recent evidence suggests that some plant taxa adapted to extreme hot environments display an alternative water-use strategy that prioritizes leaf evaporative cooling over immediate returns on water loss in the form of carbon acquisition (Aparecido et al., 2020;Urban et al., 2017). However, an inevitable tradeoff with maintaining high transpiration rates in hot and dry conditions runs the risk of operating with leaf water potentials at or near the turgor loss point and hydraulic failure. Thus, fine-tuning stomatal regulation of leaf water potential to balance midday leaf cooling with hydraulic failure avoidance may be a critically important trait in heat-adapted plants.
Here, we examine stomatal regulation of leaf water potential relative to midday leaf cooling in Populus fremontii, Sarg. (Fremont cottonwood), an obligate riparian phreatophytic tree species that inhabits arid regions in the southwest United States and northern Mexico. This species is an ideal candidate for studying genotypic variation in traits related to leaf thermoregulation because it is found across extremely broad elevational (0-2000 m.a.s.l) and climate gradients that encompass subfreezing to extreme hot temperatures (>40°C). Recent common garden experiments have found that P. fremontii displays large intraspecific variation in productivity (Grady et al., 2011), phenology (Cooper et al., 2019), and functional trait coordination (Blasini et al., 2020) Fluxes and conductance (Ikeda et al., 2017). Previous studies have identified genotypes sourced from populations adapted to cooler temperatures in the Colorado Plateau region that display a distinct combination of shorter growing seasons (i.e., later leaf flush and earlier fall senescence) with more conservative trait expression, while warm-adapted genotypes from the Sonoran Desert region exhibit longer growing seasons with more acquisitive trait expression at multiorgan levels (Blasini et al., 2020;Cooper et al., 2019). Here we define genotypes from populations with higher mean maximum summer temperatures than the common garden location as a warmadapted ecotype, and a cool-adapted ecotype as genotypes from populations at or below mean maximum summer temperatures in relation to the common garden location.
Climate projections predict that the North American region in which P. fremontii occurs will become warmer and more arid over the remainder of the century (Breshears et al., 2013;Garfin et al., 2013;Seager et al., 2014). During the first decade of the 21st century, the region experienced higher daily average temperatures and more recurrent heat waves than in the previous 100 years (Garfin et al., 2013). As a consequence of episodic drought and heatwaves, P. fremontii has experienced recent mortality surges across its geographical range (Whitham et al., 2020).
We examined the overarching hypothesis that genotypes from the warm-adapted ecotype prioritize leaf cooling over hydraulic safety compared to genotypes from the cool-adapted ecotype. To test this hypothesis, we measured leaf temperature, leaf morphology and sap-flux-scaled canopy transpiration (E) and stomatal conductance (G s ) in P. fremontii genotypes sourced from seven populations representing the warm-and cool-adapted ecotypes and growing together in a common garden located near the mid-point of the species climate distribution (Cooper et al., 2019;Hultine et al., 2020a). We evaluated three primary modes of canopy thermal 1). In addition to the local Agua Fria National Monument population, three populations, respectively, from sites with higher and lower mean maximum summer temperatures than the common garden location were selected. The three populations from the lower Sonoran Desert were defined as a warmadapted ecotype because the extreme mean maximum summer temperatures (>40°C) they experience at their source sites is above that of the common garden location (Figure 2). The three populations from higher elevation provenances in the Sonoran Desert and Colorado Plateau were sourced from locations with similar or lower mean maximum summer temperatures than the common garden location and therefore were categorized as a cool-adapted ecotype (Figure 2). 2), defined as the MAT and MMST of the source population location subtracted from the MAT and MMST of the common garden location (Grady et al., 2011).
A micrometeorological station installed at the garden measured relative humidity, air temperature, photosynthetically active radiation
In 2016, we randomly collected fully expanded leaves from midheight and south-facing canopy of each genotype (n = 56) to determine stomatal density, length, width and area. We followed the nail polish impression method (Hilu & Randall, 1984) ( )
where d w is the diffusivity of water in air (2.43 × 10 -5 m 2 /s), v is the molar volume of air (0.024 m 3 /mol; Jones, 2014), D stom is the stomatal density, a max is the maximum area of the open stomatal pore, approximated as π(p/2) 2 , where p is stomatal pore length (µm), assumed to be stomatal length divided by two (Franks & Farquhar, 2007).
Climatic variables of the seven provenances at the Agua Fria National Monument common garden
Climatic characteristics include mean annual temperature (MAT), mean maximum summer temperature (MMST), mean annual precipitation (MAP), and 30-year annual average freezing days .
Transfer distances for MAT (MAT of the garden minus MAT of the provenance) and MMST. The population 988 m is located near the common garden and thus has a transfer distance of zero. Results of mean ± SE (n = 56) comparison of whole-tree stem areas and heights in a common garden in central Arizona.
TRADEOFFS BETWEEN LEAF COOLING AND HYDRAULIC SAFETY | 5
2.3.2 | Leaf traits SLA (cm 2 g -1 ) was calculated as the one-sided area of a fresh leaf, divided by its oven-dry mass (Wright & Westoby, 2002). SLA was measured in June, July and September 2017. A subset of 12-20 fully expanded leaves from the mid and south facing section of the canopy were collected per genotype and scanned with a high-resolution computer scanner, and one-sided leaf area was measured with Image J. The scanned leaves were then oven-dried for 72 h at 75°C and weighed to calculate SLA. Leaf size (S l , cm 2 ) and leaf width (w l , cm)
were derived from the average leaf size from these measurements (Ackerly et al., 2002).
In July 2017, we used allometric relationships between whole-tree stem diameter and leaf area through a branch summation approach to estimate whole-tree canopy leaf area (A l ) and sapwood area (A s ). The diameter of all leaf-bearing branches from the main stem in each of the 56 genotypes were measured with a digital caliper. To calculate whole canopy leaf area, a subset of the collected leaves per genotype was scanned with a high-resolution computer scanner, and one-sided leaf area was measured with Image J. Then, we generated a regression of branch diameter to leaf area from a subset of branches per genotype. Scanned leaves were oven-dried for 72 h at 75°C within each subset of branches, and then their weight was multiplied by SLA to determine total leaf area of the branch (see Section 2.3.2). Wholetree height (H), canopy diameters (4-8 measurements per genotype)
and their respective canopy areas were measured five times during the 2017 growing season with a telescoping measuring pole. Canopy area (A c ) was determined using the ellipse equation, πab, where a is the mean radius of longest canopy axis and b is the radius of two perpendicular canopy axes (Ansley et al., 2012).
We installed heat dissipation sensors (Granier, 1987) that measured stem sap flux density (J s , g H 2 O m -2 sapwood s -1 ) on all 56 genotypes from June 2nd (day 153) to October 2nd (day 275) 2017. Each sensor consisted of a pair of 20 mm long, 2 mm diameter stainless steel probes inserted approximately 15 cm apart along the axis of the hydro-active xylem. The sap flux density was calculated from the differences in temperature between the heated and unheated reference probes. Sap flux density, J s (g cm -2 s -1 ), was calculated as
For diffuse porous tree species (e.g., P. fremontii), q is the prefactor coefficient (0.0119), p is the scaling exponent (1.23) and k is related to the temperature difference between the two probes (Bush et al., 2010;Granier, 1987):
where ΔT is the difference in temperature between the heated and unheated probes and ΔT 0 is the temperature difference during hydrostatic conditions (data provided in repository). We assumed that hydrostatic conditions only occurred during evening periods when vpd was at or near zero. Thus, in some cases a single value for ΔT 0 was used to calculate k over several days.
We calculated canopy transpiration (E, g m -2 s -1 ) using the total sap flux density and sapwood area to leaf area ratio (A s : A l ) according to:
From the sap flux measurements, we also calculated canopy conductance (G c , mmol m -2 s -1 ) using a simplified version of the Penman-Monteith equation (Campbell & Norman, 1998;Hultine et al., 2013;Monteith & Unsworth, 2013):
where A s is the conducting sapwood area (m 2 ), A l is the total leaf area (m 2 ), y is the psychrometric constant (kPa K -1 ), λ is the latent heat of vaporization (J kg -1 ), ρ is the density of moist air (kg m -3 ), and c p is the specific heat of air at constant pressure (J kg -1 K -1 ).
Internal bark diameter and the depth of hydro-active xylem was estimated to obtain A s . Because of the young age of the trees (2.5 years), and because Populus trees tend to have large active sapwood depths with uniform sap velocities (Lambs & Muller, 2002), we assumed the active sapwood included the entire cross-sectional area beneath the bark. Measurements of leaf area index (LAI), the projected leaf area per unit of ground area (Bréda, 2003;Chapin et al., 2011;Watson, 1947), provided a way to estimate the physical boundaries between the whole-tree canopy and the surrounding atmosphere (Bréda, 2003). Therefore, whole tree leaf area (A l ) and canopy area (A c ) were used to calculate intraspecific differences in LAI and therefore canopy boundary layer resistances, where LAI is given by
Canopy stomatal conductance (G s ) was extracted from measurements of G c by evaluating leaf boundary layer conductance (G bl ; mmol m -2 s -1 ), which can be small enough in broadleaf plants to decouple plant canopies from atmospheric conditions. We therefore
We calculated a dimensionless decoupling coefficient (Ω) (Hultine et al., 2013;Martin, 1989) to evaluate the sensitivity of transpiration to changes in boundary layer conductance.
ε
where ε is the change of latent heat to the change in sensible heat of saturated air and G r is the long-wave radiative transfer conductance.
Ω is expected to reach its upper limit (1.0) as the influence of stomatal resistance over transpiration decline.
From June to September 2017, leaf water potentials (Ψ) were measured every month at predawn (Ψ pd ; 0300-0500 h local time) and midday (Ψ md ; 1100-1300 h) on each genotype that was instrumented with sap flux probes using the Scholander pressure chamber (PMS Instruments; Scholander et al., 1965;Turner, 1988).
To take these measurements, a single shoot tip from each of the 56 genotypes was cut with a sharp razor blade at mid-height and southfacing canopy. Differences between Ψ pd and Ψ md (ΔΨ) were calculated for each genotype, population, and ecotype over each measurement period, to provide an index of the transpiration-induced changes in water potential gradients from the roots to the leaves.
We measured leaf temperature on 17 warm-and 24 cool-adapted genotypes (total n = 41) instrumented with the sap flux probes in the common garden between 13:00 and 15:00 h of 28th August (day 240) and 1st September (day 244) of 2017: two of the warmest days during the study. We evaluated leaf temperature on three to four separate leaves in each individual genotype using a thermal imaging camera ThermaCam (Flir One, Flir Systems). This handheld device integrates a thermal and visual sensor of 80 × 60 and 1440 × 1080 pixels, respectively, to a smartphone, with a typical accuracy range of ±3°C or ±5% (https://www.flir.com/products/flir-one-gen-3/).
Leaf temperatures were taken 30 cm away from a full expanded leaf at three to four different locations in the canopy. Only leaves located in the middle (1.5-2.0 m) and sun-lit areas of each tree canopies were used to measure leaf temperature. This resulted in a total of 81 and 92 leaf temperature measurements for Day 240 and Day 244, respectively (total n = 173 measurements, 99 measurements for cooladapted genotype and 74 for warm-adapted genotypes on both days). We used air temperature data collected from the on-site micrometeorological station to calculate the difference between air and leaf temperatures (ΔT).
All statistical analyses were conducted in R version 4.0.2 (R Development Core Team 2011). Before analysing the data, we examined whether each variable met the assumptions of normality and homogeneity of variance, using a Shapiro and Barlett test.
Morphological trait comparisons between cool-and warm-adapted ecotypes, including all measurements of leaf morphology, leaf area to sapwood area ratios, and mean daily sap-flux-scaled canopy fluxes were conducted using a standard Student's t-test.
Because riparian tree species are found exclusively in places with abundant water available, stomatal conductance, and wholetree water use in this species are intrinsically influenced by atmospheric characteristics like irradiance, atmospheric CO 2 concentrations, and atmospheric vapour pressure deficit (Landsberg et al., 2017). Specifically, increases in vpd have been found to correlate with decreases in stomatal conductance while the stomatal sensitivity to changes in vpd has been described to be proportional to the stomatal conductance at low vpd levels (<1 kPa). This sensitivity of stomatal conductance to changes in vpd can be estimated from (Domec et al., 2009;Hultine et al., 2013;Oren et al., 1999): Differences in Ψ pd , Ψ md and ΔΨ between cool-and warmadapted genotypes were analysed by individual mixed-effects repeated measures ANOVA (type III with Satterthwaite's method) using the 'lmer' R package (Bates et al., 2015;Kuznetsova et al., 2017). In this test, individual traits were represented as response variables while the group (cool-and warm-adapted ecotypes) and month were treated as categorical fixed effects with two and three levels, respectively. Individual genotype nested within ecotypes was incorporated as a random effect.
A leaf energy balance model was executed in R (R Core Team, 2018)
through the "tealeaves" package (Muir, 2019) (Muir, 2019;Okajima et al., 2012).
A sensitivity analysis was performed on three and two environmental and morphophysiological variables, respectively, to determine their overall effect on leaf temperature resulted from the leaf energy balance model.
These variables were air temperature, relative humidity, wind speed (environmental) and stomatal conductance and leaf size (morphophysiological). We used the 'konfound' package in R to run the sensitivity analysis (Frank et al., 2013) to determine the influence that environmental variables (relative humidity, wind speed and air temperature) and morphophysiological traits (leaf size and stomatal conductance) have on the modelled leaf temperatures in P. fremontii.
We performed a principal component analysis (PCA) to analyse the relationship between seven morphophysiological traits (S l , A s :A l , D stom , G s , Ψ md , SLA and S stom ) and ΔT at the population level using the 'factoextra' and 'FactoMineR' packages (Kassambara & Mundt, 2017;Lê et al., 2008). We determined the number of meaningful PCA axes using the Kaiser criterion and the Broken Stick Model in the 'vegan' Subsequently, we performed ANOVA Tukey's HSD tests to assess significant differences in PC axes scores at the population level.
As hypothesized, under well-watered conditions, the warm-adapted ecotype displayed cooler midday leaf temperatures than the cool-adapted ecotype during the hottest time of the day (13:00 to 15:00). Average leaf temperature in the warm-adapted ecotype was 2.80°C below the common garden ambient temperature while the cool-adapted ecotype exhibited a mean leaf temperature of 0.98°C above air temperature (t = 3.84, df = 50, p < 0.001) (Figure 3a and Table 3). The warm-adapted ecotype displayed 39% smaller leaves (t = 4.15, df = 68, p < 0.001) (Figure 3b) with 35% greater stomatal densities (t = -5.95, df = 67.39, p < 0.001) (Figure 3c) and 13% higher SLA (t = 2.85, df = 40, p < 0.01, Table 3) than the cool-adapted ecotype. Additionally, the warm-adapted ecotype exhibited 28% narrower leaves (t = -4.68, df = 66, p < 0.001) with shorter 17% stomata (t = 6.56, df = 53.09, p < 0.001) and slightly greater (8%) maximum theoretical stomatal conductance (G smax ) than the cool-adapted ecotype (Table 3).
Mean J s measured over the growing season was largely similar between warm-versus cool-adapted ecotypes until about mid-August (~Day 230) after which J s was on average 12% higher in the warm-adapted ecotype (Figure S1). Mean J s varied dramatically over the course of the growing season from less than 5 g m -2 s -1 to over 60 g m -2 s -1 depending on vapour pressure deficit (Figure S1) and photosynthetic active radiation (Q, data not shown). The warm-adapted ecotype displayed a 36% greater A s :A l than the cool-adapted ecotype (t = 9.92, df = 54, p < 0.001)
(Figure 3d and Table 3). As a consequence, the warm-adapted ecotype exhibited 42.2% higher mean afternoon transpiration rates per unit leaf area (t = -7.49, df = 199.2, p < 0.001) over the course of the growing season, with contrasts between ecotypes becoming particularly large starting in mid-August around Day 230 (Figure 4a).
Similarly, the warm-adapted ecotype exhibited 38.8% higher canopy conductance (G c ) than the cool-adapted ecotype (Table 3). We found that cool and warm-adapted ecotypes displayed their lowest canopy stomatal conductance values of the season (2.57 and 3.03 mmol m -2 s -1 , respectively) on 19 June (Day 170), the day with the second highest recorded afternoon vpd (7.33 kPa) during the season (Figure 4b). On the other hand, both cool-and warm-adapted ecotypes displayed their greatest G c values (50.5 and 69.5 mmol m -2 s -1 , respectively) on the day with the lowest vpd (1.35 kPa) recorded in the season (Day 205, July 24) (Figure 4b).
Canopies of both warm-and cool-adapted ecotypes were well coupled to the atmosphere such that the mean canopy decoupling coefficient was never higher than 0.05 for either group (Table 4).
Therefore, cool-and warm-adapted ecotypes displayed G s values that mirrored G c . The warm-adapted ecotype exhibited 37.8% higher whole-season G s (20.4 mmol m -2 s -1 ) than the cool-adapted ecotype (14.8 mmol m -2 s -1 ) (Table 4). Log-scale vpd explained 53% (F = 136.7, p < 0.0001) and 62% (F = 198.4, p < 0.0001) of the variation in mean daytime G s of warm-and cool-adapted ecotypes, respectively (Figure 5a).
Analysis of covariance revealed that across both ecotypes, G s was highly correlated with vpd (F = 294.9, p < 0.0001). Increases in vpd resulted in decreases in G s (Figure 5a). However, the relationship between G s and vpd differed between warm-and cool-adapted ecotypes (F = 46.9, p < 0.0001). The interaction between vpd and ecotypes was significant (F = 9.40, p < 0.01), as mean G s converged at higher vpd values (Figure 5a).
There was a strong correlation between G s :G sref and vpd in both warm-(R 2 = 0.53, F = 2222, p < 0.0001) and cool-adapted (R 2 = 0.58, F = 2733, p < 0.0001) ecotypes (Figure 5b). At the reference value of vpd = 1 kPa, reference G s (i.e., G sref ) was 59% higher in the warmadapted ecotype relative to the cool-adapted ecotype (66.3 mmol m -2 s -1 vs. 41.7 mmol m -2 s -1 , Figure 5a). However, the slope (m) that relates G s with vpd was also 62% higher in warm-adapted ecotype than cool-adapted ecotype (-71.1 vs. 46.5 mmol m -2 s -1 , Figure 5a).
Thus, stomatal sensitivity (S) to vpd was nearly equal between cool-(1.10 ± 0.020) and warm-adapted ecotypes (1.12 ± 0.029) (Figure 5b). relatively well-watered, with the exception of occasional brief periods between irrigation events. We did not find significant differences in Ψ pd between warm-and cool-adapted ecotypes (Figure 6a). However, the warm-adapted ecotype operated with lower mid-day leaf water potentials (Ψ md ) than the cool-adapted ecotype throughout the summer (F = 11.63, df = 1, p < 0.01, Figure 6b and Table S1), supporting hypothesis 4. Differences were most pronounced later in the growing season paralleling increased differences in canopy E between ecotypes (i.e., Figure 4a). Specifically, differences in Ψ md between cool-and warm-adapted ecotypes were less than 0.1 MPa on Days 157 and 180, but increased to 0.23 MPa (t = 3.15, df = 40, p < 0.01) and 0.33 MPa (t = 3.36, df = 40, p < 0.01) on Days 208 and 236, respectively, with mean Ψ md in the warm-adapted ecotype falling below -2.0 MPa on Day 236 (Figure 6b). As a consequence of progressive differences in Ψ md , differences in ΔΨ between ecotypes also increased over the growing season (Figure 6c).
Leaf temperature derived from the 'tealeaves' leaf energy balance model predicted leaves from the cool-adapted ecotype to be consistently hotter than the warm-adapted ecotype under identical air temperature scenarios (Figure 7). The average of all modelled leaf temperatures was 1.09°C hotter in the cool-adapted ecotype than the warm-adapted ecotype. Differences in leaf temperature between ecotypes were largely independent of air temperature, reflecting the similarity in stomatal sensitivity to vpd between ecotypes (Figure 5b). Sensitivity analysis showed that of the five variables explored (air temperature, relative humidity, wind speed, stomatal conductance and leaf size), air temperature had the greatest effect on modelled leaf temperatures (Figure S2a).
However, when air temperature was held at a fixed value, leaf temperature was constrained primarily by canopy stomatal conductance (Figure S2b).
To estimate the collective influence that multiple morpho- S3). Specifically, the two highest elevation populations with the most positive MAT and MMST transfer distances formed their own group (Group "a"), while the third highest elevation population (elevation 1219 m) was its own group (Group "b"). Interestingly, Group "c" was made up of three populations, the two lowest elevation populations with the most negative MAT and MMST transfer distances and the local population of the garden (elevation, 988 m). The last Group "d" was formed by the third lowest elevation population (elevation, 666 m) and the lowest elevation population which was also found in Group "c". We examined the overarching hypothesis that warm-adapted genotypes of the foundation tree species P. fremontii prioritize leaf cooling over hydraulic safety compared to cool-adapted genotypes. Using an experimental common garden, we assessed whether warm-adapted genotypes maintained cooler mid-summer leaf temperatures than cool-adapted genotypes and whether cooler leaf temperatures were correlated with a higher mean midday stomatal conductance, smaller leaf size or both. Mid-summer, mid-afternoon leaf temperatures of genotypes sourced from the warm-adapted ecotype were on average 3.8°C cooler than genotypes from the cool-adapted ecotype, supporting our first subhypothesis. Contrasts in leaf temperatures between these two ecotypes corresponded with contrasts in leaf morphological traits including leaf size, leaf width, SLA and stomatal density, supporting our second subhypothesis. Genotypes of the warm-adapted ecotype also expressed a higher mean stomatal canopy conductance and associated transpiration rates, supporting our third subhypothesis, although stomatal sensitivity to vpd (S) was similar between the two ecotypes. Finally, the higher stomatal conductance in warm-adapted ecotype was coupled with lower mid-day leaf water potentials (Ψ md ) compared to the cool-adapted ecotype, supporting our fourth subhypothesis. Taken together, these results
indicate that the warm-adapted ecotype maintains cooler midsummer leaf temperatures that may be critical for maintaining leaf carbon budgets and avoiding leaf thermal damage under a warming climate in the southwestern US. However, increased leaf thermal regulation in the warm-adapted ecotype appears to correspond with enhanced hydraulic "risk taking" that could result in greater susceptibility to water deficits that are also predicted to increase in frequency and intensity in the southwest in the near future.
Genotypes on the warm edge of P. fremontii's distribution experience some of the most extreme summer heat waves in North America.
Cooler leaf temperatures exhibited by genotypes sourced from the warm-adapted ecotype likely reflect extreme selection pressures to cope with chronic thermal stress that may be induced by air temperatures that often approach 50°C. To optimize canopy thermal regulation, the warm-adapted ecotype displayed a suite of morphophysiological traits and hydraulic strategies that simultaneously The combination of adaptive traits in the warm-adaptive ecotype is expected to reduce leaf photorespiration and maintenance respiration rates (Lloyd & Farquhar, 2008;O'Sullivan et al., 2013;Slot et al., 2016), while avoiding irreversible damage to chloroplasts and subsequently electron transport capacity of photosystem II (Kozaki & Takeba, 1996;Osmond & Björkman, 1972). The critical temperatures that affect photosystem II activity are generally species-specific or related to previous high-temperature exposure (Knight & Ackerly, 2003;O'sullivan et al., 2017;Teskey et al., 2015;Yordanov, 1992).
However, high-temperature exposure has consistently been correlated with loss of chloroplast thermostability and decline of photosystem II quantum yield (Berry & Bjorkman, 1980;Hüve et al., 2011).
Similarly, thylakoid membranes increase their fluidity, leakiness, and partial dissociation of light-harvesting complexes of photosystem II at extreme hot temperatures (Armond et al., 1980;Briantais et al., 1996;Hüve et al., 2011), underscoring the importance of leaf thermal regulation in hot environments.
Similar to field data collected from leaf thermal imagery, the 'tealeaves' leaf energy balance model predicted cooler leaf temperatures in the warm-adapted ecotype relative to the cool-adapted ecotype, under the same environmental conditions. However, the leaf energy balance model yielded a much smaller difference of 1.1°C instead of the 3.8°C found in our leaf thermal measurements. This difference could be explained by the fact that stomatal sensitivity to vpd was similar between ecotypes. According to the model, T leaf was strongly governed by G s (Figure S2b), and at relatively low vpd (i.e., vpd < 3 kPa), mean G s of the warm-adapted ecotype was 55%-60% higher than the cool-adapted ecotype (Figure 5a). However, under warmer and drier conditions, G s between ecotypes converged reflecting their similar sensitivities to vpd (Figure 5b). Therefore, differences in mid-summer, afternoon leaf temperatures detected by the model were largely a function of the cool-adapted ecotype having a larger and wider mean leaf size than the warmadapted ecotype (Figure 3b), and not differences in evaporative cooling. Likewise, we cannot rule out an alternative hypothesis that maintaining a higher midday stomatal conductance is not as much an adaptive trait for leaf cooling but more simply a necessity for warmadapted plants to maintain photosynthesis and growth over the growing season. Nevertheless, given the 36% higher mean A s :A l (Figure 3d), and the equal to or higher mean J s of the warm-adapted ecotype relative to the cool-adapted ecotype (Figure S1), it is highly plausible that afternoon leaf evaporative cooling was significantly more pronounced in the warm-versus cool-adapted ecotypes.
Minor changes in leaf temperature can have a significant impact of leaf carbon budgets because both mitochondrial respiration and photorespiration increase exponentially with tissue temperature (O'sullivan et al., 2017;Slot et al., 2016). For example, Q 10 values (a proportional change in respiration with a 10°C increase in temperature) in the tissues of species in the genus Populus have been reported between 1.28 and 5.89 (Gielen et al., 2003;Griffin et al., 2001). Thus, a leaf temperature increases of just 1°C in these species would result in a 12.8%-58.9% increase in leaf respiration, although short-term (i.e., hours to days) thermal acclimation to warmer temperature could reduce plant Q 10 values (Atkin & Tjoelker, 2003;Tjoelker et al., 2001). Likewise, photorespiration rates in C3 plants can increase substantially at leaf temperatures above 25°C, with a corresponding decrease in leaf carboxylation rates that can substantially reduce photosynthesis (Busch et al., 2013). Although the effect of higher temperatures on leaf photorespiration can vary by a species' optimum growth temperature (Cavanagh & Kubien, 2014;Galmés et al., 2016), photorespiration increases are generally related to the effect of high temperature on Rubisco specificity and the differences in solubility between CO 2 and O 2 (von Caemmerer & Quick, 2000).
Hydraulic "risk taking" as an adaptive strategy to maintain cooler canopies rhizosphere remains moist (Brodribb et al., 2017;Grossiord et al., 2020;Sperry et al., 2002). In the present study, Ψ md was 0.3 MPa lower in the warm-relative to cool-adapted ecotypes by the end of the warm-season, corresponding with increased differences in G s between ecotypes. For drought-intolerant species such as P. fremontii, small contrasts in Ψ md may reflect important contrasts in plant water status between warm-and cool-adapted ecotypes. In fact, by late summer, mean Ψ md in the warmadapted ecotype fell below the xylem pressure (-1.88 MPa) at which near complete hydraulic failure has been reported to occur in P. fremontiidefined as the water potential that leads to 88% loss of xylem conductivity (Ψ 88 : Choat et al., 2012). Conversely, mean Ψ md in the cooladapted ecotype never fell below -1.76, a level that is slightly above the reported threshold for hydraulic failure in P. fremontii. Whether contrasts in Ψ md reflect differences in Ψ 88 or other plant xylem traits among
ecotypes is an open question, as previous studies have also reported midday water potentials in P. fremontii approaching -2.0 MPa (Leffler et al., 2000;Williams & Cooper, 2005). A similar study conducted at the same common garden as the present study reported that P. fremontii genotypes belonging to the relatively warm Sonoran Desert ecoregion had lower wood densities and xylem vessels with higher hydraulic mean diameters than genotypes sourced from the cooler Mogollon Rim ecoregion (Blasini et al., 2020): traits that could portend greater risk of hydraulic failure in the Sonoran Desert genotypes. In the present study, the warm-adapted ecotype had higher stomatal densities and a higher mean theoretical maximum stomatal conductance compared to the cooladapted ecotype. Taken together, the lower Ψ md values indicate that the warm-adapted ecotype may be adapted to operate with lower hydraulic safety margins to maintain cooler leaves. That, in turn, may limit the hydrological niche of the warm-adapted ecotype to locations with high perennial soil moisture availability (Hultine et al., 2020a).
Not surprisingly given their lower Ψ md values, the reference midday G s (i.e., G sref ) was almost 60% higher in the warm-versus cool-adapted ecotype. However, contrasts between ecotypes were not driven by differences in J s , per se, but from the warm-adapted ecotype having a near 40% higher A s :A l . Leaf area to sapwood area ratios (the inverse of A s :A l ) strongly decrease with aridity in angiosperm tree taxa (Gleason et al., 2013;Togashi et al., 2015). A relatively high A s :A l can buffer plants from steep gradients in xylem water potential by maximizing the supply of water to individual leaves relative to demand that in turn can maximize leaf evaporative cooling under well-watered conditions. On the other hand, mean stomatal sensitivity to vpd was similar between ecotypes indicating that stomatal responses to atmospheric demand or dryness is a fixed trait among P. fremontii populations given similar exposure to soil water conditions.
A primary advantage of common gardens in ecological studies is that they provide opportunities for potential evaluation of the coordination among traits-that is, adaptive trait syndromes-and resource fluxes within and among plants (Freschet et al., 2010;Reich, 2014).
For example, a previous study conducted on P. fremontii at the same common garden as the present study revealed that traits were not only coordinated across multiple organs and scales, but also identified two clearly defined adaptive trait syndromes (Blasini et al., 2020).
Genotypes belonging to the relatively high-elevation Mogollon Rim ecoregion expressed a suite of conservative traits including spring leaf flush, leaf economic traits and wood economic traits relative to genotypes belonging to the lower elevation Sonoran Desert ecoregion (Blasini et al., 2020). Importantly, all five source populations that comprised the Sonoran Desert ecoregion shared similar mean trait values with one another, while all three populations that comprised the Mogollon Rim ecoregion also shared similar mean trait values with one another. These results indicate that all populations studied from the Sonoran Desert ecoregion shared a similar "acquisitive" adaptive trait syndrome that appeared to arise from selection to avoid leaf thermal damage (Blasini et al., 2020). Similarly, all populations from the Mogollon Rim ecoregion shared a similar "conservative" adaptive trait syndrome that appeared to arise from selection to avoid frost damage (Blasini et al., 2020).
In the present study, we defined a warm-adapted population as one where mean maximum summer temperature rose above 40°C.
Indeed, results detected clear trait contrasts between the three warm-adapted populations, and the three coolest populations (Figure 8) that mirrored contrasts previously reported between the Sonoran Desert and Mogollon Rim ecoregions (Blasini et al., 2020).
However, in contrast to our overarching hypothesis, genotypes from the mid-elevation population expressed traits that more closely paralleled genotypes sourced from the three warmest populations than the three cooler populations (Figure 8). One possible explanation is that selection pressures to cope with leaf thermal stress are expressed in genotypes from locations with lower mean maximum temperatures than 40°C. Alternatively, the mean trait expression of the mid-elevation population may reflect a certain "home field advantage" over genotypes from other locations given the 0°C transfer distance to the location of the common garden. Nevertheless, the clear adaptive trait syndromes identified here show that locally adapted P. fremontii populations may become maladapted under rapidly changing climate conditions across its geographical distribution.
The resiliency of groundwater-dependent forest ecosystems to environmental change depends largely on resiliency of phreatophytic vegetation such as P. fremontii to alterations in groundwater availability coupled with rising temperatures. The dynamics of fluvial hydrology and groundwater availability undoubtedly act as strong agents of selection in P. fremontii in terms of hydraulic architecture, xylem anatomy and stomatal regulation (Blasini et al., 2020;Hultine et al., 2020a;2020b). P. fremontii like other groundwater-dependent taxa in the southwestern United States occur in locations where groundwater is shallow enough for roots to maintain contact for most of the year. Shallow groundwater, therefore, allows P. fremontii trees to maximize productivity and resource uptake over hydraulic safety. However, hydraulic "risk taking" may be amplified in P. fremontii ecotypes occurring on the warm edge of its distribution to cope with growing season temperatures that often approach 50°C. In the present study, warm-adapted genotypes, including those sourced along the extreme warm-edge of P. fremontii's distribution, displayed a higher maximum stomatal conductance and lower mid-day leaf water potentials that corresponded with lower daytime leaf temperatures than genotypes sourced from relatively cool locations. There is a growing body of evidence that some warm-adapted species forego hydraulic safety to optimize leaf temperature via transpirational cooling when exposed to hot conditions (Aparecido et al., 2020;Drake et al., 2018;Slot et al., 2016).
Prioritizing leaf cooling over hydraulic safety would presumably limit the hydraulic niche of warm-adapted P. fremontii genotypes to locations where groundwater is not only shallow but is largely absent of daily or