Quantifying Light Response of Leaf-Scale Water-Use Efficiency and Its Interrelationships With Photosynthesis and Stomatal Conductance in C3 and C4 Species

Light intensity (I) is the most dynamic and significant environmental variable affecting photosynthesis (An), stomatal conductance (gs), transpiration (Tr), and water-use efficiency (WUE). Currently, studies characterizing leaf-scale WUE–I responses are rare and key questions have not been answered. In particular, (1) What shape does the response function take? (2) Are there maximum intrinsic (WUEi; WUEi–max) and instantaneous WUE (WUEinst; WUEinst–max) at the corresponding saturation irradiances (Ii–sat and Iinst–sat)? This study developed WUEi–I and WUEinst–I models sharing the same non-asymptotic function with previously published An–I and gs–I models. Observation-modeling intercomparison was conducted for field-grown plants of soybean (C3) and grain amaranth (C4) to assess the robustness of our models versus the non-rectangular hyperbola models (NH models). Both types of models can reproduce WUE–I curves well over light-limited range. However, at light-saturated range, NH models overestimated WUEi–max and WUEinst–max and cannot return Ii–sat and Iinst–sat due to its asymptotic function. Moreover, NH models cannot describe the down-regulation of WUE induced by high light, on which our models described well. The results showed that WUEi and WUEinst increased rapidly within low range of I, driven by uncoupled photosynthesis and stomatal responsiveness. Initial response rapidity of WUEi was higher than WUEinst because the greatest increase of An and Tr occurred at low gs. C4 species showed higher WUEi–max and WUEinst–max than C3 species—at similar Ii–sat and Iinst–sat. Our intercomparison highlighted larger discrepancy between WUEi–I and WUEinst–I responses in C3 than C4 species, quantitatively characterizing an important advantage of C4 photosynthetic pathway—higher An gain but lower Tr cost per unit of gs change. Our models can accurately return the wealth of key quantities defining species-specific WUE–I responses—besides An–I and gs–I responses. The key advantage is its robustness in characterizing these entangled responses over a wide I range from light-limited to light-inhibitory light intensities, through adopting the same analytical framework and the explicit and consistent definitions on these responses. Our models are of significance for physiologists and modelers—and also for breeders screening for genotypes concurrently achieving maximized photosynthesis and optimized WUE.

Light intensity (I) is the most dynamic and significant environmental variable affecting photosynthesis (A n ), stomatal conductance (g s ), transpiration (T r ), and water-use efficiency (WUE). Currently, studies characterizing leaf-scale WUE-I responses are rare and key questions have not been answered. In particular, (1) What shape does the response function take? (2) Are there maximum intrinsic (WUE i ; WUE i−max ) and instantaneous WUE (WUE inst ; WUE inst−max ) at the corresponding saturation irradiances (I i−sat and I inst−sat )? This study developed WUE i -I and WUE inst -I models sharing the same non-asymptotic function with previously published A n -I and g s -I models. Observation-modeling intercomparison was conducted for field-grown plants of soybean (C 3 ) and grain amaranth (C 4 ) to assess the robustness of our models versus the non-rectangular hyperbola models (NH models). Both types of models can reproduce WUE-I curves well over light-limited range. However, at light-saturated range, NH models overestimated WUE i−max and WUE inst−max and cannot return I i−sat and I inst−sat due to its asymptotic function. Moreover, NH models cannot describe the downregulation of WUE induced by high light, on which our models described well. The results showed that WUE i and WUE inst increased rapidly within low range of I, driven by uncoupled photosynthesis and stomatal responsiveness. Initial response rapidity of WUE i was higher than WUE inst because the greatest increase of A n and T r occurred at low g s . C 4 species showed higher WUE i−max and WUE inst−max than C 3 species-at similar I i−sat and I inst−sat . Our intercomparison highlighted larger discrepancy between WUE i -I and WUE inst -I responses in C 3 than C 4 species, quantitatively characterizing an INTRODUCTION Stomata control the balance between carbon flux driven by photosynthesis and water flux dominated by transpiration, which is characterized by water-use efficiency (WUE) at various scales (Sinclair et al., 1984;Gilbert et al., 2011;Eamus et al., 2016;Medlyn et al., 2017). WUE can thus indicate the natural selection on the balance between these fluxes (Hetherington and Woodward, 2003). Characterizing the environmental impacts on WUE among plant species and/or plant function types can advance our knowledge on differential plant adaptation strategies, and improve our prediction on consequences of environmental challenges (Avola et al., 2008;Egea et al., 2011;Zhou et al., 2014Zhou et al., , 2016De Kauwe et al., 2015;Köhler et al., 2016;Ahrar et al., 2017). For instance, plant species with the highest WUE would show the greatest fitness in dry habitats (Dudley, 1996;Zhou et al., 2019). WUE is also an important metric in crop breeding and genotype selection, especially for irrigated crops whose water use significantly affects crop productivity and profitability (Duursma et al., 2013;Flexas et al., 2013;Bota et al., 2016;Webster et al., 2016). WUE can be estimated using different techniques, based on observations of leaf gas exchange, stable isotope discrimination, and eddy covariance fluxes (Medlyn et al., 2017). Among these techniques, WUE is most commonly estimated by measuring leaf gas exchange, facilitated by portable photosynthesis system allowing simultaneous measurement of leaf-scale carbon and water fluxes (Medrano et al., 2015). WUE derived from leaf gas exchange measurement is usually defined as the ratio of net CO 2 assimilation rate (A n ) to stomatal conductance for water vapor (g s )-intrinsic water-use efficiency (WUE i ;von Caemmerer and Farquhar, 1981), or the ratio of A n to transpiration rate (T r )instantaneous water-use efficiency (WUE inst ; Fischer and Turner, 1978) (see Table 1 for a summary of parameters and units). WUE i can be used to compare photosynthetic characteristics independently of evaporative demand (Linares and Camarero, 2012). WUE inst is a key determinant of whole-plant WUE as it summarizes plant dry mass production per unit of water loss (Sinclair et al., 1984;Duursma et al., 2013; but see Medrano et al., 2015 for constraints). WUE i and WUE inst have been widely used as an index of plant and vegetation performances in response to various environmental changes, such as changed water or light availabilities, vapor pressure deficit (VPD), temperature and CO 2 concentration (Aranda et al., 2007;Avola et al., 2008;Linares and Camarero, 2012;Duursma et al., 2013;Bota et al., 2016).
Light is often viewed as the most significant environmental variable affecting photosynthesis, stomatal behavior and WUE (Knapp and Smith, 1987;Aranda et al., 2007;McAusland et al., 2016). Plants in most ecosystems experience rapid short-term variability in light resource , which can cause continual transition of A n , g s , T r , WUE i , and WUE inst throughout the growing season (Knapp and Smith, 1990;Knapp, 1993). However, studies characterizing the light response of WUE are rare (McAusland et al., 2016). It is largely unknown whether there is a maximum WUE i or WUE inst -and the corresponding saturation irradiance-for plants under dynamic irradiance conditions, or how plant species or plant function types (PFTs) would differ in their light responses of WUE i and WUE inst .
Characterization of the interrelationships among light responses of A n , g s , T r , WUE i and WUE inst -which can be simultaneously measured-will be fundamental to the scaling-up modeling of WUE-I responses at the whole-plant and ecosystem scale. The foremost step toward this direction calls for a robust model, with which (1) the WUE i and WUE inst responses to a gradient of irradiance intensity (I) levels (WUE i -I and WUE inst -I response curve, respectively) can be characterized, and (2) the key quantities defining the response curves-such as the initial slope of the response curve, the maximum WUE and the corresponding saturation irradiance-can be quantified. Ideally, the model can accurately represent the differential WUE i and WUE inst responses among plant species or PFTs, such as that reported between C 3 and C 4 species with contrasting light responses of photosynthesis, stomatal functioning, and WUE (Pearcy and Ehleringer, 1984;Knapp, 1993). For a given A n , g s and T r are higher in C 3 than C 4 plants, leading to higher WUE i and WUE inst in C 4 plants, which has higher utilization efficiency of CO 2 at relatively lower intercellular CO 2 concentration (Pearcy and Ehleringer, 1984). The objectives of this study were to develop a leaf-scale WUE-I model and assess the model performance against experimental field observations of C 3 and C 4 species in order to answer key questions of how best to model the light response of WUE i and WUE inst . In particular: (1) What shape does the leaf-scale WUE-I response function (2) Can the model well represent the differential WUE i -I and/or WUE inst -I response characteristics between C 3 and C 4 species? By integrating the published A n -I (Ye, 2007;Ye et al., 2013) and g s -I (Ye and Yu, 2008) response function, we developed an explicit WUE-I modeling framework and hypothesized that the species-specific light response curves of WUE i and WUE inst can be quantitatively characterized using the same non-asymptotic function. The hypothesis was tested using an observation-modeling intercomparison on WUE i -I and WUE inst -I responses for field-grown C 3 [soybean (Glycine max L.)] and C 4 species [grain amaranth (Amaranthus hypochondriacus L.)] under high I condition in the growing season. Model performance against that of the non-rectangular hyperbola model was also evaluated.

Analytical Models
A non-asymptotic model has been previously developed and tested to well characterize the light response of photosynthesis (Ye, 2007;Ye et al., 2013), with its simplified form as follows: where α is the initial slope of light response curve of photosynthesis, I is the irradiance, and β and γ are the photoinhibition coefficient and saturation coefficient, respectively, and R d is the dark respiratory rate. The key model parameters are listed in Table 1.
The saturation irradiance (I sat ) corresponding to the lightsaturated photosynthetic rate (A nmax ) can be calculated as follows: Eq. 1 has been widely used to characterize photosynthetic light response curves of various plant species under different environmental conditions, highlighting its better performance than that of rectangular (Baly, 1935) and non-rectangular hyperbolic models (Thornley, 1976;Wargent et al., 2011;Xu et al., 2012a,b;Song et al., 2015;Chen et al., 2016). The rectangular and the non-rectangular hyperbolic models have been reported to overestimate A nmax (dos Santos et al., 2013), and cannot quantify I sat (Gomes et al., 2006;dos Santos et al., 2013;Song et al., 2015;Chen et al., 2016). Meanwhile, a model of the same non-asymptotic form as Eq. 1 has been developed and tested to well characterize the light response of stomatal conductance (Ye and Yu, 2008), as follows: where α 0 is the initial slope of light response curve of stomatal conductance, g s0 is the residual stomatal conductance, and β 0 and γ 0 are two coefficients that are independent of I (Ye and Yu, 2008). Most existing stomatal conductance models cannot quantify the g s−max or the corresponding I g−sat under changing irradiance conditions (Dewar, 2002;Buckley et al., 2003;Buckley and Mott, 2013;Flexas et al., 2013).
The g s -I model developed by Ye and Yu (2008) can well characterize the g s -I response, from which key parameters defining the g s -I response-such as g s−max and I g−sat -can be easily obtained. The saturation irradiance (I g−sat ) corresponding to the lightsaturated stomatal conductance (g s−max ) can be calculated as follows: Here, we hypothesize that the light response of WUE i can be characterized using the same non-asymptotic form as that of the A n -I (Eq. 1) and g s -I (Eq. 4) response functions, as follows: where α 1 represents the initial slope of light response curve of WUE i , β 1 , and γ 1 are coefficients that are independent of I, and K i is the residual intrinsic water-use efficiency. The saturation irradiance (I i−sat ) corresponding to the maximum WUE i (WUE i−max ) can be calculated as follows: , solid lines were fitted using Eq. 1 and dashed lines were fitted using the non-rectangular hyperbola model (Eq. S1). In plots (C) and (D), solid lines were fitted using Eq. 4. Data are the mean ± SE (n = 4).
Eq. 1 Obs. The parameters are initial slope of the A n -I curve (α), the maximum net photosynthetic rate (A nmax ) and the corresponding saturation irradiance (I sat ), photoinhibition coefficient (β), saturation coefficient (γ), and dark respiration rate (R d ). All values are the means ± SE (n = 4). Different letters denote statistically significant differences (P ≤ 0.05) between soybean and grain amaranth within each column of fitted (Eq. 1) or measured (Obs.) values. There is no statistically significant difference between fitted (Eq. 1) and measured (Obs.) values for each parameter. See Table 1 for definitions of abbreviations.

Soybean
Since g s controls leaf T r at a given VPD (Duursma et al., 2013), we hypothesize that the light response of WUE inst can also be characterized using the same non-asymptotic function as that of WUE i -I response function (Eq. 7), as follows: where α 2 represents the initial slope of light response curve of WUE inst , β 2 and γ 2 are coefficients that are independent of I, and K inst is the residual instantaneous water-use efficiency. The saturation irradiance (I inst−sat ) corresponding to the maximum WUE inst (WUE inst−max ) can be calculated as follows: In this study, we tested if Eqs. 7 and 10 can well characterize the species-specific WUE-I response characteristics against modeloriented field observations and the simulations using the nonrectangular hyperbola model-in terms of the initial slope of light response curve of WUE (α 1 and α 2 , respectively), the maximum WUE inst (WUE i and WUE inst−max , respectively), and the saturation irradiance (I i−sat and I inst−sat , respectively).

Study Site and Plant Material
The field observations on one C 3 species-soybean (Glycine max L.) and one C 4 species-grain amaranth (A. hypochondriacus L.) were conducted at the Yucheng Comprehensive Experiment Station of the Chinese Academy of Sciences, located at the irrigation district of the Yellow River Basin in the North China Plain. This region is dominated by the warm-temperate semi-humid monsoon climate and is suitable for planting soybean and grain amaranth with high yields. This region has ample energy resource, and the light intensity in the growing season usually reaches ∼2000 µmol m −2 s −1 in sunny days. Soybean and grain amaranth were planted in field on May 3rd and June 15th 2012, respectively. All plants were kept under moist condition throughout the experiment.

Light Response Curve Measurement
The leaf gas exchange measurements were conducted after 45 days of growth in field-June 16th for soybean and July 29th for grain amaranth. Fully expanded sun-exposed leaves of four plants for each species were measured using a portable photosynthesis system (LI-6400, Li-Cor Inc., Lincoln, NE, United States). Before each measurement, the leaf was acclimated in the chamber to achieve stable gas exchange, with reference CO 2 concentration maintained at 380 µmol CO 2 mol −1 , irradiance intensity maintained at 2000 µmol photon m −2 s −1 , and leaf temperature maintained at 35 • C. After the leaf acclimated to the cuvette environment, the photosynthetic light response curve measurements were conducted with a descending gradient of irradiance intensity Eq. 4 Obs.
Eq. 4 Obs. The parameters are initial slope of the g s -I curve (α 0 ), coefficients β 0 and γ 0 , the maximum stomatal conductance (g s-max ) and the corresponding saturation irradiance (I g-sat ), and the residual stomatal conductance (g s0 ). All values are the means ± SE (n = 4). Different letters denote statistically significant differences (P ≤ 0.05) between soybean and grain amaranth within each column of fitted (Eq. 4) or measured (Obs.) values.

Data Analysis
All statistical tests were performed using the statistical package SPSS 18.5 statistical software (SPSS, Chicago, IL, United States). The analysis of variance (ANOVA) was used to assess species effects. Paired-sample t tests were conducted to test whether there were significant differences between fitted and measured values of quantitative traits (α, A nmax , I sat , α 0 , g s−max , I g−sat , α 1 , WUE i−max , I i−sat , α 2 , WUE inst−max , I inst−sat , etc.). Goodness of fit of the mathematical model to experimental observations was assessed using the coefficient of determination (r 2 = 1-SSE/SST, where SST is the total sum of squares and SSE is the error sum of squares).

RESULTS
Light Response Curves of A n , g s , and T r The increase of I led to a rapid initial increase of A n (Figures 1A,B), g s (Figures 1C,D), and T r (Figures 1E,F) for both C 3 and C 4 species. However, the initial increase rate of A n was 100-fold higher than that of g s for both species (Tables 2 and 3). The high coefficient of determination (r 2 ) values indicated that the species-specific A n -I response curves fitted by Eq. 1-and the g s -I response curves fitted by Eq. 4were highly representative of the observations for both species (Figure 1). Soybean exhibited a single-peaked pattern for both A n -I and g s -I responses, characterized by the increase of A n and g s with the increasing I until reaching the A nmax and g s−max at the corresponding I sat and I g−sat , respectively (Figures 1A,C  and Tables 2 and 3). Compared with Eq. 1, the nonrectangular hyperbola model (Supplementary Eq. S1) showed similarly high r 2 value in simulating A n -I response curves but significantly overestimated the A nmax (Figures 1A,B and Supplementary Table S1). Paired-sample t tests showed there were no significant differences between the fitted values and the measured values of A nmax , I sat , g s−max , and I g−sat for soybean (Tables 2 and 3). Grain amaranth kept increasing its A n and g s within the range of irradiance intensity applied during measurements (0-2000 µmol photon m −2 s −1 ), without showing an observational A nmax , I sat , g s−max , or I g−sat (Figures 1B,D  and Tables 2 and 3). Grain amaranth showed relatively higher (not significant) initial increase rate of g s , characterized by an initial slope of the light response curve of g s (α 0 ) (Figure 1 and Tables 2 and 3). , solid lines were fitted using Eq. 7 and dashed lines were fitted using the non-rectangular hyperbola model (Eq. S2). In plots (C) and (D), solid lines were fitted using Eq. 10 and dashed lines were fitted using the non-rectangular hyperbola model (Eq. S3). Data are the mean ± SE (n = 4).

Light Response Curves of WUE i and WUE inst
Within the low range of irradiance intensity, WUE i and WUE inst of both species increased almost linearly with the increasing I. Both soybean and grain amaranth exhibited a singlepeaked WUE i -I and WUE inst -I response pattern, respectively. In particular, both species showed an increase of WUE i and WUE inst with the increasing I until reaching the speciesspecific WUE i−max and WUE inst−max at the corresponding species-specific saturation irradiance levels (I i−sat and I inst−sat , respectively) (Figure 2 and Tables 4 and 5). However, soybean showed significantly lower observed and fitted WUE i−max and WUE inst−max (P ≤ 0.05) than grain amaranth (Figure 2 and Tables 4 and 5). The two species showed no significant difference in I i−sat , I inst−sat or the initial increase rate of WUE i or WUE instcharacterized by a maximal slope of the light response curves (α 1 and α 2 , respectively) (Figure 2 and Tables 4 and 5).
The high r 2 values indicated that WUE i -I response curves fitted by Eq. 7-and the WUE inst -I response curves fitted by Eq. 10-were highly representative of the observations of both species (Figure 2 and Tables 4 and 5). There were no significant differences between fitted and observed values in WUE i−max , WUE inst−max , I i−sat , I inst−sat , K i , or K inst (Tables 4 and 5). Compared with Eqs. 7 and 10, the nonrectangular hyperbola model (Supplementary Eqs. S2 and S3, respectively) showed similarly high r 2 values but significantly overestimated WUE i−max and WUE inst−max for the two species (Supplementary Tables S2 and S3).

Discussion
Our WUE i -I and WUE inst -I models represented cultivarspecific response curves over a wide range of light intensities extremely well (r 2 ≥ 0.996), including the decline of WUE i and WUE inst beyond the saturation irradiances which the NH models cannot represent due to the asymptotic function. Our models can also return values for WUE i−max , WUE inst−max , I i−sat , and I inst−sat , which were in very close agreement with the measured values. The NH models cannot characterize the decline in WUE i and WUE inst induced by high light, leading to overestimations of WUE i−max and WUE inst−max (Supplementary Tables S2 and S3).
Interrelationships of Light Responses of Photosynthesis, Stomatal Conductance, and Water-Use Efficiency WUE i and WUE inst increased rapidly within low range of I, mainly driven by the uncoupled rapidity of photosynthetic and stomatal responses (Figures 1, 2 and Tables 4 and 5; Knapp and Smith, 1990;McAusland et al., 2016). In this study, both C 3 and TABLE 4 | Fitted (Eq. 7) and measured (Obs.) values of parameters defining the light response curve of intrinsic water-use efficiency for C 3 (soybean) and C 4 species (grain amaranth).
Eq. 7 Obs. The parameters are initial slope of the WUE i -I curve (α 1 ), the maximum intrinsic water-use efficiency (WUE i-max ) and the corresponding saturation irradiance (I i-sat ), and coefficients β 1 and γ 1 and K i . All values are the means ± SE (n = 4). Different letters denote statistically significant differences (P ≤ 0.05) between soybean and grain amaranth within each column of fitted (Eq. 7) or measured (Obs.) values. There is no statistically significant difference between fitted (Eq. 7) and measured (Obs.) values for each parameter. See Table 1 for definitions of abbreviations.

Soybean
C 4 species showed 100-fold higher initial increase rate of A n (α) than that of g s (α 0 ) (Tables 2 and 3). The rapid initial increase of WUE i and WUE inst -characterized by α 1 and α 2 , respectivelyoccurred at low g s (and at low I), when small increase in g s exerted the greatest impacts on A n and T r (Hetherington and Woodward, 2003). The occurrence of the greatest A n and T r increase at low g s also determined that α 1 would be much higher than α 2 for a given species (Figure 2 and Tables 4 and 5).
With the increasing I (from 0 to ∼800 µmol m −2 s −1 ), faster photosynthesis response than stomatal response led to the decline of intercellular CO 2 concentration (C i ) (Supplementary Figure S1; McAusland et al., 2016), causing further opening of stomatal pores (Mott, 1988) which allowed for diffusion of ambient CO 2 into the leaf (Hetherington and Woodward, 2003). Further increase of g s -beyond the low g s range-led to minimal increase of A n and T r (Hetherington and Woodward, 2003), such that WUE i and WUE inst flattened quickly after reaching the WUE i−max and WUE inst−max (Figure 2). Further increase of I beyond I i−sat and I inst−sat led to a decrease in WUE i and WUE inst . To reach A nmax , both soybean and grain amaranth would have to show a decrease of WUE i (or WUE inst ) from WUE i−max (or WUE inst−max ) (Figures 1, 2).

Differential Light Responses of Water-Use Efficiency Between C 3 and C 4 Species
The observation-modeling intercomparison in this study highlighted the differential single-peaked WUE i -I and WUE inst -I responses-besides differential A n -I and g s -I responses-between C 3 and C 4 species (Figure 2 and Tables 4 and 5). C 4 species (grain amaranth) showed higher WUE i and WUE inst than C 3 species (soybean), suggesting its better leaf-scale optimization of carbon uptake versus water loss than C 3 species (Figures 1, 2 and Tables 2, 4, and 5). This may be due to higher photosynthetic capacity and rapidity of stomatal response (α 0 ) in C 4 species under changing irradiance conditions (Figure 1 and Tables 2 and 3), which facilitate relatively closer coupling between A n and g s in C 4 species than C 3 species (McAusland et al., 2016).
Moreover, this study identifies greater interspecific difference in WUE inst than that in WUE i -at high I range when WUE i and WUE inst flatten (Figure 2 and Tables 4 and 5). C 3 species (soybean) showed larger discrepancy between its WUE i -I and WUE inst -I responses than that of C 4 species (grain amaranth). This may be due to differential water use strategies between C 3 and C 4 species-C 4 species holds smaller T r change per unit of g s change in relative to C 3 species (Knapp, 1993). These results quantitatively demonstrate that the differential WUE i -I responses between C 3 and C 4 species would not necessarily mirror their differential WUE inst -I responses (Figure 2).
These results support previous studies reporting that water conservation-in terms of high WUE-is an important consequence of the C 4 photosynthetic pathway (besides high carbon gain rate) at different scales including single leaf, whole plant, and even whole communities (Ludlow and Wilson, 1972), TABLE 5 | Fitted (Eq. 10) and measured (Obs.) values of parameters defining the light response curve of instantaneous water-use efficiency for C 3 (soybean) and C 4 species (grain amaranth).

Model Significance
By providing (1) analytical models characterizing the singlepeaked light responses of WUE i and WUE inst and (2) key quantitative traits defining WUE i -I and WUE inst -I response differences between C 3 and C 4 species, this study provides a practical and robust modeling approach-in a form potentially applicable to WUE-I models at whole-plant and/or ecosystem scale. In particular, the key quantitative traits-the initial increase rates of WUE i (α 1 ) and WUE inst (α 2 ) besides that of A n (α) and g s (α 0 ), the maximum WUE i (WUE i−max ) and WUE inst (WUE inst−max ) besides that of A n (A nmax ) and g s (g s−max ), and the corresponding saturation irradiances-will directly help physiologists and modelers investigate the interrelationships among photosynthesis, stomatal behavior, and WUE under changing irradiance conditions. Meanwhile, the above quantitative traits allow for easier and more extensive evaluation of light-intensity consequences on carbon and water relations among different species and/or PFTs. Such quantitative information, gathered on a wider range of species and/or PFTs, could allow (1) a deeper understanding of interspecific variation in light response strategies (Knapp, 1993;Hetherington and Woodward, 2003;McAusland et al., 2016), and (2) a realistic representation of adaptive WUE-I response differences among PFTs into ecosystem modeling.
The explicit models developed in this study can be viewed as an initial step toward filling the gap between investigating the trends of interspecific variation in short-term leaf-scale WUE-I responses and translating the variation into improved process representation in models of plant and ecosystem scales. The findings in this study remain to be validated (1) with species of different growth form and PFT membership (e.g., slower-growing woody species), which could hold different light response strategies , (2) with daily and seasonal integrals and/or whole-plant estimates of WUE that sometimes could show a low correlation with short-term leafscale WUE observations (Medrano et al., 2015), and (3) when leaf gas exchange is subjected to compound effects of other climatic conditions in current and future climate change scenarios.

CONCLUSION
The newly developed models (Eqs. 7 and 10, respectively) allow robust reproduction of the differential single-peaked WUE i -I and WUE inst -I trends between C 3 and C 4 species and easy parameterization of key traits defining the trends (α 1 , I i−sat , K i and WUE i−max , α 2 , I inst−sat , K inst , and WUE inst−max ). The models can be employed for fast and accurate assessment of plant WUE i and WUE inst responses-besides that of photosynthetic and stomatal responses using a consistent modeling frameworkacross all light-limited, light-saturated, and photoinhibitory light intensities. These findings are useful (1) for breeders screening for ideal genotypes target with maximized photosynthesis capacity and optimized WUE, (2) for plant physiologists quantifying intra-and/or inter-specific variation in leaf-scale WUE-I responses, and (3) for modelers working on better representation of the coupling between carbon and water processes under dynamic irradiance conditions.

DATA AVAILABILITY STATEMENT
The datasets generated for this study are available on request to the corresponding author.