Influence of vapour pressure deficit and CO2 on the thermal sensitivity of stomatal function in tropical trees Authors: Alexander W. Cheesman, Peter Cox, Simon Jones, Kali B. Middleby, Lucas A. Cernusak, Peter J. Franks Published: 2026 ABSTRACT Summary Stomatal conductance to water vapour (g s) is typically considered to operate in coordination with photosynthesis (A) to balance carbon gain and water loss in an optimal manner. However, at high temperatures that suppress A, g s has been seen to increase or remain high – reducing the predictive accuracy of some stomatal conductance models.We investigated the temperature sensitivity of leaf‐level gas exchange and evaluated a temperature‐dependent modification of a widely used stomatal conductance model across a temperature gradient (30 to 40°C) in three tropical tree species (Mallotus philippensis, Ficus congesta and Elaeocarpus grandis) grown and measured under near‐ambient (420 ppm) or elevated (820 ppm) CO2, and under either constant or increasing vapour pressure deficit (VPD).Measured A and g s, as well as the model term g 1, all exhibited a temperature sensitivity that was impacted by [CO2] and VPD regime. Importantly, under both constant and increasing VPD, g 1 increased with leaf temperature.These findings demonstrate that the coupling between A and g s is not static but varies with temperature and environmental context, and that incorporating a representation of this thermal sensitivity provides a tractable improvement to models of plant carbon–water exchange under future climate conditions. Introduction In terrestrial plants, stomatal pores represent the dynamic interface of carbon and water fluxes between the leaf and the atmosphere (Berry et al., 2010). As such, understanding and modelling how stomatal behaviour will be impacted by a rapidly changing climate is fundamental to predicting changes in terrestrial productivity and water use (Cox et al., 1999; Damour et al., 2010; Buckley & Mott, 2013; Rogers et al., 2017). Generally, stomata have evolved to be highly responsive to a range of environmental cues (Clark et al., 2022), including plant water status (Buckley, 2019), light intensity and quality (Vialet‐Chabrand et al., 2021), atmospheric CO2 concentration (Xu et al., 2016), leaf temperature (Pankasem et al., 2024) and humidity – or more directly vapour pressure deficit (VPD; Grossiord et al., 2020). Attempts to capture and simulate their dynamic nature have resulted in a plethora of stomatal models (Damour et al., 2010). However, the prevailing theory of stomatal function in higher plants is that their responsiveness has evolved to minimise the integrated sum of E – λA, where E is transpiration rate (mol H2O m−2 s−1), A is CO2 assimilation rate (μmol CO2 m−2 s−1), and λ (mol H2O mol−1 CO2) represents the Lagrangian multiplier or marginal water cost of plant carbon gain (Cowan & Farquhar, 1977). This stomatal optimisation theory has been widely explored (Buckley et al., 2017), but its mathematical complexity presents practical limitations, requiring either simplifying assumptions or novel computational methods (Jones et al., 2022; Potkay et al., 2025a). The Medlyn et al. (2011) model was an attempt to reconcile the complexities of optimisation theory with the utility of empirical stomatal conductance models, such as those of Ball et al. (1987) and Leuning (1995), formulating stomatal conductance to water vapour (g s, mol H2O m−2 s−1) as per Eqn 1: (Eqn 1) gs≈g0+1.61+g1VPDLACa where VPDL is vapour pressure deficit of the leaf relative to the air (kPa), C a is the external CO2 concentration (μmol mol−1), g 0 (mol H2O m−2 s−1) is water loss in the absence of photosynthesis, and g 1 (kPa0.5) is a parameter that characterises the nature of the relationship between g s and A. The parameter g 1 reflects both physiological and hydraulic constraints on stomatal regulation and, when calibrated for different species, encapsulates interspecific variation in water‐use strategies (Buckley, 2019). Models such as Eqn 1, which imply a coupling between A and g s such that dgsdA is always positive, have successfully explained stomatal behaviour across diverse plant types and environmental conditions (Lin et al., 2015). Their simplicity and predictive capability have led to their widespread integration within the land component of Earth System Models (Lawrence et al., 2019; Oliver et al., 2022) and dynamic vegetation demography models (Li et al., 2022). However, empirical evidence suggests that this predictable relationship becomes less reliable under high temperatures. Specifically, it has been observed that while A decreases above a thermal optimum (T opt) due to the thermal sensitivity of component biochemical processes (Lin et al., 2012), g s may continue to rise (Urban et al., 2017; Huang et al., 2021; Feng et al., 2023; Marchin et al., 2023; Diao et al., 2024). Such thermal ‘decoupling’ likely reflects a direct stomatal response to temperature (Mott & Peak, 2010; Mills et al., 2024; Pankasem et al., 2024) even though increasing VPD – which typically covaries with temperature – would be expected to reduce g s (Slot et al., 2024). Although empirical modifications of coupled photosynthesis‐stomatal conductance models have been proposed to improve predictions under such conditions (Lamour et al., 2022), these modifications tend to reduce mechanistic clarity and limit applicability under novel scenarios – a critical consideration for the utility of such models in robust climate‐change projections. Within the Medlyn et al. (2011) framework, the parameter g 1 is predicted to vary systematically with water‐use efficiency (WUE) and plant functional type (Lin et al., 2015; Miner et al., 2017; Franks et al., 2018). However, although commonly derived and applied as a fixed factor, g 1 is also likely to be impacted by local environmental conditions (Stefanski et al., 2023; Potkay et al., 2025b). In Medlyn et al.'s (2011) formulation, the term g 1 (Eqn 2) is considered equal to the square root of a term containing the CO2 compensation point of A in the absence of day respiration (Γ*, Pa) pressure (P, Pa), and the marginal water cost of carbon gain (λ, mol H2O mol−1 CO2). (Eqn 2) g1≈3Γ*Pλ1.6 Due to differences in the temperature sensitivity of CO2 fixation and photorespiration by the photosynthetic enzyme RuBiSCO, we know that Γ* varies substantially over a biologically relevant temperature range. This temperature dependence (Supporting Information Fig. S1) is generally assumed to be similar for all C3 species (Bernacchi et al., 2001) and can be modelled by Eqn 3 as per Sharkey et al. (2007), (Eqn 3) ΓT*=ec−ΔHaRT where c (11.187) is a scaling constant to result in a Γ* at 25°C Γ25* of 3.743 Pa, ΔH a is enthalpy of activation (24.46 kJ mol−1), and we assume R = 8.314 J mol−1 K−1 and temperature (T) is expressed in kelvin where (0°C = 273.15 K). Furthermore, in their original formulation of the optimal stomatal theory, Cowan & Farquhar (1977) explicitly recognised that λ was unlikely to remain constant and would be impacted by the biophysical context of the studied plant. This dynamic view (Cowan, 1978) has often been superseded by the assumption that variation in λ arises primarily from (a) fixed species‐level differences in stomatal behaviour and (b) longer‐term responses to factors, such as low soil moisture availability (Manzoni et al., 2011; Lin et al., 2015). Nonetheless, short‐term variation in λ may also arise due to temperature‐driven changes in either the physical properties of water and therefore changes to the cost of water movement through the Soil–Plant–Atmosphere Continuum (SPAC) or through additional temperature dependencies of carbon gain. For example, the viscosity (η, Pa·s) of water (inverse of fluidity) can be assumed to be independent of pressure at the earth surface but does vary with temperature over a biologically relevant range (Roderick & Berry, 2001) with viscosity at a given temperature ηT given by Eqn 4 (Fig. S1): (Eqn 4) ηT=1.95×1014Tw7 where T w is the temperature of the liquid water in K. This means that, as temperature increases, water becomes less viscous – that is more fluid – thereby increasing hydraulic conductance and reducing the hydraulic pressure gradient from roots to leaves for a given transpiration‐driven flow rate through the xylem. This reduced viscosity of xylem sap essentially reduces the depression in leaf water potential resulting from transpiration and, with leaves maintaining higher hydration, stomata potentially maintain higher λ – that is, they will allow more water loss from leaves per unit carbon gained from photosynthesis. Note that λ here refers to the marginal water cost dEdA following Cowan & Farquhar (1977), in contrast to formulations like Prentice et al. (2014) where λ is defined as dAdE. At the same time, the solubility of CO2 in the aqueous phase (CO2(aq)) is known to be temperature‐dependant, with Henry's Law coefficient at a given temperature (HTs, mol m−3 Pa−1) varying across a biologically relevant range as per Eqn 5: (Eqn 5) HTs=H25s×eB1Tw−1298.15 where H25s is 0.00034 mol m−3 Pa−1, and B is 2300 for CO2 in fresh water (Sander, 2023). This means that as temperature increases, the solubility of CO2 in water decreases (Fig. S1). Given stomata respond to the substomatal CO2 concentrations (C i) via carbonic anhydrase enzymes functioning in stomatal guard cells (Hu et al., 2010; Engineer et al., 2016), a lower CO2 solubility means that higher C i is required to maintain the same CO2(aq), and this necessitates an increase in g s, which in turn leads to higher values of λ and g 1. If we assume a standard thermal sensitivity of Γ* and postulate that λ changes proportionally with the inverse of change in water viscosity and the solubility of CO2, we can expect a significant impact of temperature on theoretical g 1 (Fig. 1). Assuming all three processes contribute to an observed thermal sensitivity, this increase in g 1 would be as much as 50% between 30 and 40°C (Fig. 1). Incorporating this thermal sensitivity into predicted stomatal behaviour offers a mechanistic basis for the observed behaviour of A and g s under elevated temperatures, wherein g s is often found to be underestimated at higher temperatures when assuming a constant g 1 (Marchin et al., 2023). Doing so could also provide a simple pathway to improve the predictive accuracy of terrestrial biosphere models employing a coupled A and g s framework, such as the Medlyn et al. (2011) model, under elevated temperature (Rogers et al., 2017). Examining the potential thermal sensitivity of the model parameter g 1 when accounting for the additive nature of underlying components, including CO2 compensation point (Г*) and the marginal water cost of plant carbon gain (λ), when assuming λ is dependent upon changes in the viscosity (inverse of fluidity) of water (η, Pa·s) and the solubility of CO2 (HTS, mol m−3 Pa−1). Data normalised for g 1 at 30°C using Eqn 8. In this study, we sought to develop a cohesive data set that would allow us to test the coupling between A and g s across a range of elevated leaf temperature (T L) and investigate how this relationship is impacted by both VPD and CO2 concentrations [CO2]. Specifically, we examined leaf‐level gas exchange in three tropical tree species grown under low‐VPD conditions and either near‐ambient control (i.e. 420 ppm) or elevated (i.e. 820 ppm) CO2. Leaf‐level gas exchange was then determined with VPDL either held constant (c. 1.6 kPa) or allowed to rise in line with air temperature by maintaining a constant dew point of the air. We subsequently examined whether incorporating the postulated thermal sensitivity of g 1 into the Medlyn et al. (2011) model could allow it to more accurately capture the observed response of g s. Materials and Methods Study species and experimental material We examined stomatal behaviour in three broadly distributed tropical tree species found within the Australian Wet Tropics: Mallotus philippensis (Lam.) Mull.Arg. (Euphorbiaceae), Ficus congesta Roxb. var. congesta (Moraceae) and Elaeocarpus grandis F.Muell. (Elaeocarpaceae; Fig. S2). All plant material was obtained from a local nursery, and for each species, the material originated from a single seed source in the lowlands. All three species show a broad geographical distribution and are typically found within the Australian Wet Tropics between sea level and 1100 m. Seedlings were potted into 145‐mm square pots (4 l) containing a high organic matter potting mix with ¼ (by volume) of a local volcanic stone, Quincan (Northside Raw Materials & Timber Supplies, Cairns, Australia), to improve drainage. Before treatments, seedlings were established in a shade house (75% solar transmission) for 1 month. At the start of the experiment, six plants per species were placed into one of two glasshouse chambers (each 2.8 × 5 m) at James Cook University's Environmental Research Complex (https://www.jcu.edu.au/environmental‐research‐complex). The chambers are conditioned by individual air handling units and controlled by a building management system able to provide independent control of temperature, CO2 concentration and humidity (Middleby et al., 2024). Both chambers tracked external air temperatures, and had ultrasonic humidifiers set to turn on if VPD of the air (VPDa) rose above 1.5 kPa. The two chambers differed in the target CO2 concentration, set to be either a near‐ambient control (420 ppm) or elevated (i.e. 820 ppm). Each chamber was fitted with a combined temperature/ relative humidity probe (HMP60; Vaisala, Vantaa, Finland) and data logger (CR300; Campbell Scientific, Logan, UT, USA) to monitor conditions independently of the building management system. To avoid unintended bias between and within chambers, pots were periodically rotated within chambers and treatments rotated between chambers every c. 2 months over the 6‐month growing period (Fig. S3). Gas‐exchange equipment Leaf‐level gas‐exchange measurements were conducted in a third climate‐controlled chamber using one of three LI‐6400/XT portable photosynthesis systems (LI‐COR Inc., Lincoln, NE, USA) equipped with a 2 × 3 cm leaf cuvette with a red‐blue LED light source (6400‐02B; LI‐COR). Given the inability of standard LI‐6400/XT's environmental controls to provide continuous and stable humidity and CO2 control, we pretreated inlet air to the LI‐6400/XTs and bypassed the local sodalime and drierite scrubbers. Specifically, inlet air to the LI‐6400/XT machines was first scrubbed of CO2 with an external soda lime cartridge (c. 2 l) and pump, and then passed through one of three dew point generators (LI610; LI‐COR Inc.) before being delivered at excess to T‐junctions from which the LI‐6400/XT could sample CO2 free air of a predetermined dew point. The CO2 cylinder in each individual LI‐6400/XTs was then used to deliver an inlet air stream with stable, user‐defined CO2 and H2O concentrations. The LI‐6400/XTs were networked to a computer external to the growth chamber running LI6400XTerm v.3.4.2 while the dewpoint generators were controlled by independent 0–5 V control signals supplied by an Arbitrary Waveform Generator (AG1022; OWON, Helmond, the Netherlands). To avoid issues of condensation, air lines delivering conditioned air to the LI‐6400/XTs were made of PTFE and the growth chamber temperature was generally set 2°C above the LI‐6400/XTs cuvette setpoint. Nonetheless, care was still needed to stabilise the temperature profile of all components and so they were shielded from direct solar radiation with reflective panels. Temperature ramping protocol On days of data collection, three well‐watered plants grown under low‐VPDa conditions and either control or elevated CO2 concentrations were moved to the third chamber where the CO2 concentration was matched to growth conditions. The LI‐6400/XT cuvettes were then affixed to leaves initiated and matured under those treatment CO2 concentrations with a standard block temperature of 30°C, photosynthetic photon flux density (PPFD) of 1000 μmol m−2 s−1, and CO2 concentration of incoming reference air at either 420 or 820 ppm depending on growth CO2. The dew point of incoming air was then adjusted to establish an initial VPDL of between 1.5 and 1.8 kPa. After leaf‐level gas exchange had stabilised to these initial conditions (typically c. 1 h), baseline gas‐exchange measurements were collected in three successive logging periods (20 s logging for 3 min per period; total duration 9 min). Temperature ramping was then initiated by increasing both the glasshouse air temperature and the LI‐6400/XT block temperature in 1°C increments. During temperature ramping, one of two VPD regimes was applied: (a) a constant‐VPDL treatment, in which the dew point of inlet air was increased at each temperature step to maintain VPDL approximately constant (mean curve‐average VPDL = 1.6 kPa; maximum curve VPDL = 2.2 kPa); and (b) an increasing‐VPDL treatment, in which the dew point of inlet air was held constant, allowing VPDL to rise inline with temperature, reaching c. 5 kPa at a T L of 42.5°C (Fig. S4). In both cases, gas exchange was allowed to stabilise before the next incremental increase with data collected as per initial conditions every 2°C. Temperature ramping typically took 3.5 h and ran from a block temperature of 30 to 40°C resulting in T L ranging, on average, from 32.4 to 42.5°C. After ramping experiments, plants were returned to their growth chambers and were not worked on again for at least 3 d. Response curves were collected on all three species in plants grown and measured under both control and elevated CO2, with a constant VPDL. In the control CO2 plants only, a second set of curves was also collected with the increasing‐VPDL treatment. For each species, approximately six (five to seven) high‐quality curves were obtained per scenario tested. Leaf‐level functional traits To provide species‐level context for observed leaf‐level gas exchange, at the conclusion of temperature ramping experiments leaf‐level functional traits were determined on at least one leaf per replicate plant (i.e. minimum n = 6 per species, per growth CO2). This included performing photosynthetic CO2 response curves (A–C i) at a cuvette temperature of 30°C to obtain estimates of the maximum rates of RuBisCO carboxylation (V cmax) and the rates of electron transport at 1000 μmol PPFD m−2 s−1 (J 1000; Buckley & Diaz‐Espejo, 2015). The A–C i curves were analysed using the ‘fitacis’ function in the plantecophys package in R (Duursma, 2015). To account for the control over block temperature instead of T L (which differed from block temperature by up to 2°C), the fitted parameters V cmax and J 1000 were normalised to the typically reported 25°C using temperature response parameters from Kelly (2014). Leaf mass per unit area (LMA) was determined by scanning leaves for area calculations and then drying at 70°C until constant mass. Data analysis All statistical analyses and visualisations were performed using the R Statistical Software, v.4.5.1 (R Core Team, 2025). Logged data points for each curve and setpoint were used to determine WUE calculated as A/E. The g 1 parameter was calculated using the inversion‐approach set out by Potkay et al. (2025b). In this approach, g 1 can be calculated directly using a rearrangement of Eqn 1 by assuming that g 0 ≈ 0, as in Eqn 6, and g c, the stomatal conductance to CO2, is equal to (g s/1.6). Given the controlled nature of the data collection protocol and stability of observed gas exchange, this approach was found to be equivalent to g 1 determined at each curve's individual setpoints via the typical regression method using the R plantecophys package (Duursma, 2015; Fig. S5). (Eqn 6) g1=VPDLgcCaA−1 Parameters (e.g. T L, VPDL, A, g s, E, WUE and g 1) were then averaged for each leaf at each cuvette setpoint before further analysis. To examine whether plants had acclimated leaf‐level functional traits to growth CO2 concentration, we initially used Welch's two sample t‐test (which does not assume equal variances between groups) to compare LMA, V cmax and J 1000 in leaves grown under both control and elevated CO2. We subsequently used Welch's two sample t‐test to examine gas‐exchange data (i.e. A, g s, C i/C a, WUE and g 1) of leaves under the initial conditions (i.e. cuvette temperature 30°C) used in the ramping experiments. Given the unbalanced experimental design, t‐tests were initially applied to gas‐exchange data of plants grown under control CO2 at the beginning of either constant or progressively increasing (hereafter referred to as increasing) VPD curves. We subsequently examined the difference between plants grown under control and elevated CO2 concentrations when VPDL was to be held constant. To assess the influence of VPD regime and CO2 treatment on the temperature response of leaf‐level gas exchange, we first normalised individual curves to the initial‐rates observed when the cuvette was set to 30°C. Subsequently a generalised additive model (GAM) was fitted using the R‐package mgcv (Wood, 2017) to the control CO2 subset of data, allowing us to isolate the interaction between T L and VPD regime (i.e. constant or increasing VPDL). The general GAM form applied to describe A, g s, WUE and C i/C a (Eqn 7): (Eqn 7) included a smooth term for T L separately for each VPD regime, along with a main effect of VPD regime. As we were interested in the generalisable nature of observed trends, species identity was treated as a random effect to account for potential interspecific variation, with the model fitted using restricted maximum likelihood (REML) estimation. Subsequent analyses examined the influence of growth CO2 using a similar GAM framework applied to the subset of data from both control and elevated CO2 plants in which VPDL was held constant. When examining the potential difference in g 1 with increasing T L across CO2 and VPD regimes, we were interested in the absolute change in g 1 and thus did not normalise individual curve values for those observed under the initial conditions; instead, an additional ‘curve identity’ random factor was nested within species in the GAM. After examining trends in relative leaf‐level gas exchange and calculated g 1, we assessed whether incorporating temperature sensitivity into the g 1 term improved predictions of observed g s across the full data set. For each data point, measured g s was compared with two modelled estimates calculated using Eqn 1 (assuming negligible g 0), the corresponding A and C a, and one of two values of g 1. Either (a) the initial g 1 determined for that leaf at a block temperature of 30°C (g 1(ref)), or (b) a g 1 incorporating a postulated temperature sensitivity of g 1 (Eqn 8): (Eqn 8) g1T=g1ref×ΓT*Γref*×ηrefηT×HrefsHTs where g 1(T) is the modified value of g 1 at the measured T L, calculated as a proportional adjustment to the reference value g 1(ref) based on temperature‐induced changes in Г*, η and H s given by Eqns (Eqn 3), (Eqn 4), (Eqn 5) with the reference temperature (ref) taken as T L under the initial conditions (i.e. leaf chamber of 30°C). Model performance was assessed using root mean square error (RMSE) and the coefficient of determination (R2) as the square of the Pearson correlation coefficient between observed and predicted values. Residuals were calculated for both models, and a paired t‐test was performed to determine whether the residuals differed significantly between the models. Results Acclimation of leaf‐level functional traits LMA varied between species (Table 1); however, there was no significant impact observed of growth CO2 on LMA (Table 1, t (9.21) = −1.65, t (9.75) = −0.07, t (7.7) = −0.70 in Elaeocarpus, Ficus and Mallotus, respectively, P > 0.01). Both V cmax and J 1000 were significantly, but differentially, impacted by growth CO2 across the three species examined (Table 1). In Elaeocarpus and Ficus, V cmax was significantly (P < 0.01) lower in plants grown under elevated CO2 (t (12.9) = 3.30 and t (12.3) = 3.16 in Elaeocaprus and Ficus, respectively), whereas J 1000 was not significantly different. By contrast, V cmax in Mallotus was found to not differ significantly while there was a significant increase in J 1000 in plants grown under elevated CO2 (t (13) = −3.7, P < 0.001). In all species, these changes resulted in a highly significant (P < 0.001) increase in the C i transition point between the Rubisco‐limited and electron transport‐limited phases of photosynthesis for plants grown under elevated CO2 (t (7.8) = −4.35, t (12.1) = −5.68, t (12.4) = −4.96, in Elaeocarpus, Ficus and Mallotus, respectively). Table 1Leaf‐level functional traits and initial gas exchange of three tropical tree species grown under low‐VPD conditions and either near‐ambient control (420 ppm) or elevated (820 ppm) CO2 concentrations.Growth [CO2] V cmax(25) (μmol CO2 m−2 s−1) J 1000(25) (μmol m−2 s−1)LMA (g m−2)VPDPhotosynthesis (μmol CO2 m−2 s−1)Conductance (mol H2O m−2 s−1) C i/C a Transpiration (mmol H2O m−2 s−1) WUE (μmol CO2 mmol−1 H2O) g 1 Elaocarpus grandis Control 38.4 ± 6.7 81.7 ± 15.975 ± 6.2Free11.4 ± 1.50.20 ± 0.060.73 ± 0.042.9 ± 0.84.1 ± 0.64.2 ± 1.1Held 10.5 ± 1.8 0.21 ± 0.040.76 ± 0.033.1 ± 0.6 3.5 ± 0.4 5.0 ± 0.9Elevated 28.1 ± 5.3 79.1 ± 9.680 ± 4.6Held 15.1 ± 1.5 0.17 ± 0.040.78 ± 0.042.3 ± 0.4 6.6 ± 1.1 5.4 ± 1.3 Ficus congesta Control 36.5 ± 3.9 96.4 ± 13.082 ± 11.2Free13.5 ± 0.40.36 ± 0.030.81 ± 0.024.9 ± 0.42.8 ± 0.37.0 ± 0.9Held 12.5 ± 1.6 0.36 ± 0.080.82 ± 0.044.9 ± 1.0 2.6 ± 0.5 7.7 ± 1.9Elevated 30.4 ± 3.6 97.7 ± 11.183 ± 13.2Held 19.5 ± 1.4 0.34 ± 0.040.85 ± 0.024.3 ± 0.7 4.7 ± 1.0 9.0 ± 2.1 Mallotus philippensis Control20.3 ± 3.7 46.7 ± 8.9 52 ± 9.3Free9.7 ± 2.00.20 ± 0.070.77 ± 0.052.9 ± 1.03.6 ± 0.95.2 ± 1.8Held8.4 ± 3.4 0.21 ± 0.07 0.81 ± 0.022.8 ± 0.7 2.9 ± 0.5 6.5 ± 0.8Elevated22.0 ± 2.3 62.4 ± 7.5 55 ± 5.0Held9.2 ± 2.1 0.11 ± 0.03 0.81 ± 0.041.7 ± 0.3 5.6 ± 1.3 6.6 ± 2.1Data represent stable gas exchange with cuvette temperature set to 30°C, and photosynthetic photon flux density (PPFD) = 1000 μmol m−2 s−1. Variables in bold indicate a significant difference at P < 0.05 between plants grown under near‐ambient and elevated CO2 concentrations. Bold indicate a significant difference at P < 0.05 between plant grown under near‐ambient and elevated CO2 concentrations. Leaf‐level functional traits and initial gas exchange of three tropical tree species grown under low‐VPD conditions and either near‐ambient control (420 ppm) or elevated (820 ppm) CO2 concentrations. Data represent stable gas exchange with cuvette temperature set to 30°C, and photosynthetic photon flux density (PPFD) = 1000 μmol m−2 s−1. Variables in bold indicate a significant difference at P < 0.05 between plants grown under near‐ambient and elevated CO2 concentrations. Bold indicate a significant difference at P < 0.05 between plant grown under near‐ambient and elevated CO2 concentrations. Initial gas‐exchange data indicated species‐level differences in gas‐exchange characteristics, with Ficus having the highest A and g s and lowest WUE, while Mallotus had the lowest A, g s and highest WUE (Table 1). Species‐level average g 1 ranged from 4.8 kPa0.5 in Elaeocarpus to 7.9 kPa0.5 in Ficus. Within species, we found no significant differences in initial gas‐exchange characteristics (before the temperature ramping) in plants grown under near‐ambient control CO2 and about to be measured under either constant or increasing VPDL. By contrast, we found highly significant differences in initial gas‐exchange characteristics of plants grown and measured under control and elevated CO2 concentrations when measured under constant VPDL. In both Elaeocarpus and Ficus, photosynthesis (A) was significantly (P < 0.001) higher in plants grown and measured under elevated CO2 (t (7.7) = −4.37, t (6.9) = −6.93, respectively); by contrast, Mallotus showed no significant difference in A, but a significant reduction in g s (t (4.97) = 2.82, P < 0.05). As a result, within all species there was a highly significant (P < 0.001) increase in WUE of plants grown and measured under elevated CO2 concentrations (t (5.03) = −6.09, t (4.26) = −3.85, t (8.11) = −4.92 in Elaeocarpus, Ficus and Mallotus, respectively). Neither C i/C a nor g 1 showed any significant differences between plants grown and measured under control or elevated CO2 (Table 1). Impact of VPD on temperature response of leaf‐level gas exchange GAMs were used to assess the effects of leaf temperature (T L) and VPD treatment (i.e. constant or increasing) on key physiological traits, including relative A, g s, C i/C a and WUE (Fig. 2). Effects of leaf temperature (T L) on relative photosynthesis (a), stomatal conductance (b), water‐use efficiency (c) and C i/C a (d) under conditions of constant (c. 1.6 kPa) or progressively increasing leaf vapour pressure deficit (VPDL) and near‐ambient control (i.e. 420 ppm) CO2. Data are shown relative to initial measurements (Table 1) with partial effects trend based on generalised additive models with species as a random factor. Relative A was well explained by the GAM, which accounted for 90.2% of the observed deviance (R 2‐adj = 0.90; Fig. 2a), indicating a strong influence of both T L and VPD regime (P < 0.001). Plants measured under constant VPDL had significantly higher relative A across the temperature range than those under increasing VPDL (estimated treatment effect = 0.08 ± 0.012, P < 0.001). While the temperature response of relative A differed significantly between VPD regimes (P < 0.001), with stronger thermal sensitivity and a steeper decline under increasing VPDL (estimated degrees of freedom (edf) = 3.44, F = 190.8, P < 0.001) compared with constant VPDL (edf = 3.60, F = 111.0, P < 0.001). Relative g s was also well explained by the GAM, accounting for 73.2% of the deviance (R 2‐adj = 0.72, Fig. 2b). The effect of VPD regime was highly significant (estimated treatment effect = 0.31 ± 0.02, P < 0.001), with higher relative g s observed under constant‐VPDL conditions. Moreover, the temperature response of relative g s differed significantly between treatments: under increasing VPDL, the decline in relative g s was ostensibly linear with increasing T L and associated VPDL (edf = 1.77, F = 89.2, P < 0.001), whereas under constant VPDL, relative g s initially increased to an optimum at c. 38°C before a more gradual decline at very high T L (edf = 3.10, F = 7.62, P < 0.001). As a result of the change in A and g s alongside changing evaporative demand under increasing VPDL, relative WUE (i.e. A/E) exhibited a similarly strong significant (P < 0.001) response to both T L and VPD regime. The GAM (R 2‐adj = 0.90, deviance explained = 90%; Fig. 2c) showed plants under constant VPDL to have significantly higher relative WUE compared with those measured under increasing VPDL (estimated treatment effect = 0.16 ± 0.01, P < 0.001). The temperature response of relative WUE was significant under both VPD regimes, with a decline observed across the temperature range in each case. Under increasing VPDL, the decline was more pronounced (edf = 3.19, F = 235.9, P < 0.001), while under constant VPDL, the decline was less steep but still highly significant (edf = 1.67, F = 195.0, P < 0.001). Plants under constant VPDL exhibited higher relative‐C i/C a values than those under increasing VPDL (estimate = 0.071 ± 0.005, P < 0.001). The temperature response of relative C i/C a varied significantly between treatments (P < 0.001), with a more complex, nonlinear temperature dependence under increasing VPDL (edf = 4.56, F = 9.6, P < 0.001) compared with the more linear response under constant VPDL (edf = 1.06, F = 1.1, P < 0.001). Under constant VPDL, relative C i/C a generally increased with rising T L, whereas under increasing VPDL, it initially declined before increasing again at T L above c. 41°C. Although all species showed a similar trend with increasing T L, species identity was found to explain a significant amount of variation in the rate of change in relative A, g s and WUE across all applied models (P < 0.001). By contrast, the model for relative C i/C a (Fig. 2d; R 2‐adj = 0.60, 61.4% of the deviance) revealed significant effects of T L and VPD regime, but no significant effect of species identity. Impact of CO2 on temperature responses of leaf‐level gas exchange To investigate the impact of elevated CO2 on the thermal sensitivity of relative leaf‐level gas exchange under constant VPDL, we fitted a GAM with a smooth interaction between T L and CO2 treatment, a main effect of CO2 treatment and species as a random effect. The model for relative A explained 80% of the deviance (R 2‐adj = 0.79; Fig. 3a), indicating a strong fit to the data. Plants developed and exposed to elevated CO2 exhibited significantly higher relative A across the temperature range compared with those under control CO2 (estimated treatment effect = 0.072 ± 0.015, P < 0.001). Temperature responses of relative A were significant under both CO2 treatments, though their shapes only differed slightly (edf = 3.29 under control CO2 and edf = 3.48 under elevated CO2; both P < 0.001). The random model term species identity accounted for significant additional variation (P < 0.001). Effects of leaf temperature (T L) on relative photosynthesis (a), stomatal conductance (b), water‐use efficiency (c) and C i/C a (d) of leaves under conditions of constant – vapour pressure deficit (c. 1.6 kPa) and either near‐ambient control (i.e. 420 ppm) or elevated (820 ppm) CO2 reflecting long‐term growth conditions of plants. Data are shown relative to initial measurements conditions (see Table 1) with partial effects trend based on generalised additive models with species as a random factor. The GAM for change in relative g s explained 47.7% of the deviance (R 2‐adj = 0.45; Fig. 3b), suggesting moderate predictive power. While there was no significant main effect of CO2 treatment on relative g s (estimated treatment effect = −0.033 ± 0.026, P = 0.20), the temperature responses differed significantly between CO2 treatments. Both control and elevated CO2 data exhibited significant temperature‐dependent patterns (edf = 2.98 and 2.75, under control and elevated CO2, respectively; both P < 0.001), with both showing a general increase to an optimum in relative g s at c. 38°C before a decline at higher T L. In this analysis, species identity accounted for additional significant variation (edf = 1.96, P < 0.001), supporting there being interspecific differences in the change in relative g s with increasing temperature. Relative WUE was found to generally decline with increasing T L with the GAM explaining 61.6% of the deviance (R 2‐adj = 0.60; Fig. 3c), and with a significant higher relative WUE under elevated CO2 (estimated treatment effect = 0.077 ± 0.018, P < 0.001). The temperature dependence of relative WUE differed between treatments: under control CO2, relative WUE decreased linearly with temperature (edf = 1.00, F = 156.1, P < 0.001), while under elevated CO2 the response declined but was more complex (edf = 3.32, F = 28.7, P < 0.001), suggesting that elevated CO2 not only enhances relative WUE but also modifies its thermal response. Species identity also accounted for significant variation (edf = 1.6, P < 0.01). Relative C i/C a increased with increasing T L under constant VPDL and both control and elevated CO2 concentrations (Fig. 3d). The relative‐C i/C a GAM explained 44.5% of the deviance (R 2adj = 0.42) suggesting moderate predictive power. Elevated CO2 significantly reduced relative C i/C a (estimated treatment effect = −0.0251 ± 0.005, P < 0.001), likely due to the noted higher photosynthetic demand, while the smooth term for T L was still significant under both treatments (edf =1.36 and 3.37, under control and elevated CO2, respectively. P < 0.001) showing an increase in relative C i/C a with increasing T L. Species variation in relative C i/C a was present but less pronounced than in other parameters (edf = 1.45, P < 0.05). Temperature sensitivity of g 1 We examined variation in the stomatal model parameter g 1 across experimental treatments and a range in T L to provide insight into possible changes in stomatal coupling to photosynthesis (Fig. 4). Initially we examined the change in g 1 across increasing T L under near‐ambient control CO2 concentrations and either constant or increasing VPDL (Fig. 4a). The GAM explained 70.2% of the deviance (R 2‐adj = 0.64). Across the data set, there was a significant main effect of VPD regime, with higher g 1 values under constant‐VPDL conditions (estimated treatment effect = 2.06 ± 0.99, P = 0.040). Leaf temperature was found to have a significant and nonlinear influence on g 1 under both increasing and constant VPDL (edf = 2.75 and 2.71, respectively; both P < 0.001), with a stronger temperature response evident under constant VPDL (F = 30.16) compared with increasing VPDL (F = 14.22). The random effect of curve identity contributed significantly to model fit (P < 0.001), whereas the species‐level effect was not significant (P = 0.13), suggesting variation in g 1 was more strongly associated with individual leaf context than species identity. Observed response of stomatal model parameter g 1 (kPa0.5) as defined by Medlyn et al. (2011) to leaf temperature (T L) under near‐ambient control CO2 and either increasing or constant vapour pressure deficit (VPD) (a), or constant low‐VPD and either control or elevated CO2 (b). Observations shown for three species of tropical tree as well as trends predicted from a generalised additive model, including random effects of species and curve identity. When examining the trends in g 1 over T L between CO2 treatments (Fig. 4b), the GAM explained 71.3% of the deviance (adjusted R 2 = 0.65). No significant main effect of CO2 treatment was detected (estimated treatment effect = −0.73 ± 1.15, P = 0.531), indicating that elevated CO2 did not alter the overall magnitude of g 1. However, the shape of the temperature response of g 1 differed significantly between treatments. The smooth functions describing the relationship between T L and g 1 were significant under both control and elevated CO2 (edf = 2.60 and 2.79, respectively; both P < 0.001), indicating strong nonlinear temperature responses, but the temperature response was more pronounced under control CO2 (F = 27.29) compared with elevated CO2 (F = 12.72), suggesting a dampening of g 1 thermal sensitivity under elevated CO2. The random effect of curve identity nested within species contributed significantly to model fit (P = 0.003), while the species‐level random effect alone was not significant (P = 0.08). Incorporating temperature sensitivity of g 1 We compared observed g s with modelled values using two approaches: one using a constant g 1 determined at the initiation of each curve (Fig. 5a), and another incorporating the postulated temperature sensitivity of g 1 (Fig. 5b). The model that included temperature sensitivity performed better, with a lower RMSE (0.028 vs 0.045) and a higher coefficient of determination (R 2 = 0.94 vs R 2 = 0.91). Residual analysis confirmed a significant improvement in model fit when the temperature sensitivity of g 1 was included. A paired t‐test comparing residuals of the two models showed a highly significant difference (t = 17.13, df = 295, P < 0.001), with the temperature‐sensitive model exhibiting consistently lower residuals (mean difference = 0.027, CI = 0.024–0.030 mol m−2 s−1). Comparison between measured g s and g s predicted using (a) constant g 1 based on the initial data for each curve or (b) the initial g 1 modified to account for postulated temperature sensitivity. Dashed red line indicates 1 : 1. Discussion Understanding how stomatal behaviour responds to rising temperatures is essential for predicting plant carbon–water dynamics under future climates (Blonder et al., 2023). Here, we sought to assess how variation in leaf temperature (T L), CO2 concentration and temperature‐VPD regime (i.e. whether VPDL was constrained or covaried with T L) influences the coupling between photosynthesis (A) and stomatal conductance (g s) in well‐watered saplings of three diverse tropical tree species. Our results demonstrate that both VPD regime and CO2 can significantly modify the observed thermal responses of A, g s, WUE and the stomatal model parameter g 1. Critically, we also show that incorporating a mechanistically derived temperature‐sensitive formulation of the model parameter g 1 improves the predictive accuracy of the Medlyn et al. (2011) model at high temperatures. Elevated CO2 alters photosynthetic capacity but not intrinsic stomatal regulation Growth under elevated CO2 induced species‐specific adjustments in photosynthetic capacity, but without coordinated changes in leaf structural traits (i.e. LMA) or intrinsic stomatal behaviour. In Elaeocarpus and Ficus, elevated CO2 reduced V cmax without altering J 1000, whereas Mallotus apparently maintained V cmax but exhibited enhanced electron transport capacity. Despite these contrasting biochemical responses, all species showed an increase in the C i transition point and a consistent enhancement of WUE under elevated CO2 typical of plants grown under elevated CO2 (Cernusak et al., 2019; Ainsworth & Long, 2021). Neither C i/C a nor g 1 differed between growth CO2 treatments, indicating that elevated CO2 did not alter intrinsic stomatal regulation, consistent with Gardner et al. (2023). Observed g 1 under initial conditions (i.e. Li6400/XT cuvette at 30°C), across the three species averaged 6.3 kPa0.5 (range 4.9 to 7.9 kPa0.5). This is substantially larger than that originally reported for rainforest trees by Lin et al. (2015; i.e. 3.77 kPa0.5) but which is comparable to the value used by more recent analysis by Oliver et al. (2022) when calibrating the land surface model JULES vn 5.6 (i.e. 5.31 kPa0.5). VPD shapes the thermal responses of leaf‐level gas exchange In this study, measurements of plants under contrasting VPD regimes (i.e. constant or increasing) demonstrated that evaporative demand strongly shapes the temperature responses of leaf‐level gas exchange in tropical plants grown under near‐ambient CO2. When VPDL was allowed to rise with increasing T L, both relative A and relative g s were lower than values measured at comparable T L made under constant (and low) VPDL (Fig. 2). These responses are consistent with previous work showing that high VPDL suppresses g s and can lower the temperature at which rates of A begin to decline (Grossiord et al., 2020; Slot et al., 2024). Indeed, the two VPD regimes clearly diverged in how g s constrained C i (and thereby A). Under constant‐VPDL, g s did not immediately decline with increasing T L and initially continued to rise, resulting in an increase in C i/C a despite declining A at higher temperatures. By contrast, when VPDL increased with T L, g s declined progressively, reducing C i/C a and imposing a stomatal limitation on photosynthesis. Only at very high temperatures (T L > 41°C), where A had declined to very low rates, did C i/C a increase again under rising VPDL, reflecting the collapse of biochemical demand for CO2 rather than stomatal reopening. WUE declined under both VPD regimes, though more rapidly under increasing VPDL, underscoring that changes in transpiration across T L reflect both stomatal regulation and the direct influence of VPDL on evaporative flux. Our results indicate that VPDL modulates the apparent thermal sensitivity of g 1, with a stronger temperature dependence observed when VPDL was held constant (Fig. 4a). At face value, this might suggest that the impact of VPDL on g s is incompletely captured by the inverse, square root term in Eqn 1. However, caution is warranted in interpreting this pattern. The calculation of g s from leaf‐level gas‐exchange data assumes that the air within the substomatal cavity is fully saturated with water vapour. In reality, this assumption can fail to be met, especially in leaves being measured under high VPDL (Cernusak et al., 2024). When leaves are unable to maintain saturation of the internal air space, g s may be underestimated as a result of additional ‘unsaturation resistance’ to water movement being wrongly assigned to stomatal resistance (Wong et al., 2022). When the potential impact of unsaturation on calculated g s has been examined in detail, it has been found to be as much as 30% under moderate‐to‐high VPDL (Wong et al., 2022; Cernusak et al., 2024). It is therefore possible that the observed difference in g 1 and C i/C a as a result of increasing‐VPD treatment (Fig. 4a) reflects, at least in part, the influence of substomatal unsaturation leading to an increasing underestimation of g s as VPDL increases. Impacts of elevated CO2 on stomatal behaviour Measurement of leaf‐level gas exchange under elevated CO2 demonstrated substantial acclimation in leaf function (Table 1). While all three species saw an increase in the C i transition point between RuBP carboxylation and RuBP regeneration limitation, Ficus and Elaeocarpus achieved this through a reduction in V cmax, while Mallotus showed no change in V cmax but rather an increase in J 1000. All species saw a significant increase in WUE – consistent with the well‐established CO2 fertilisation effect (Cernusak et al., 2019). In two of the species (Ficus and Elaeocarpus), this was predominantly due to an increase in A, while in Mallotus it was due to a reduction in g s. This difference may reflect a difference in the species' ecological niche, with Mallotus being typically found in monsoonal and drier rainforest where water availability is often a limiting factor, possibly favouring selection for more conservative water use (Zich et al., 2020). Across all three species, there was no significant change observed in g 1 across growth CO2 (Table 1), suggesting that the coordination between photosynthetic capacity and stomatal function is at least partially maintained under elevated CO2 (Fig. 3; Gardner et al., 2023). However, we did observe significant shifts in the temperature sensitivity of relative gas exchange under elevated CO2 (Figs 3, 4b). Although our data do not include T L low enough to allow for the reliable determination of T opt (i.e. thermal optimum for A), the apparent increase in T opt (Fig. 3a) under elevated CO2 is consistent with the known direct modification of photosynthetic biochemistry under elevated CO2 (Long, 1991; Smith & Keenan, 2020) as well as the coordinated acclimation of stomatal and mesophyll function (Hu et al., 2026). It is this increase in T opt and therefore the rate and shape of temperature‐induced decline in A that may directly impact the observed difference in temperature response of g 1 under elevated CO2. Incorporating temperature sensitivity into stomatal models The theoretical foundation of coupled photosynthesis‐stomata models such as Medlyn et al. (2011) suggests that g 1 scales with both the CO2 compensation point (Γ*) and the marginal water cost of carbon gain (λ), both of which are likely temperature‐sensitive (Sharkey et al., 2007; Prentice et al., 2014). By explicitly incorporating temperature dependencies of Γ*, water viscosity (η) and solubility of CO2 in water (Hs) into a predictive framework for g 1, we were able to mechanistically estimate how stomatal sensitivity should change with T L (Fig. 1). Incorporating this dynamic temperature‐sensitive g 1 into the Medlyn et al. (2011) model resulted in model behaviour that qualitatively matched that observed within our data set, including the relative trends in A, g s and C i/C a with increasing temperature, and under both elevated CO2 and different VPD treatments (See Data availability statement and Figs S6–S8). Moreover, using this temperature‐sensitive g 1 significantly improved predictions of g s at elevated temperatures reducing RMSE by 37% across our diverse leaf‐level gas‐exchange data (Fig. 5). While previous studies have adjusted g 1 empirically to improve model fit under heat stress (Lamour et al., 2022), it is clear that there exists a temperature dependence in the stomatal relationship to photosynthesis in at least well‐watered C3 plants (Pankasem et al., 2024; Chen et al., 2025). Reanalysis of previously published data on stomatal ‘decoupling’ in temperate tree species by Diao et al. (2024) provides further evidence for this (Fig. S9). By applying Potkay et al.'s (2025b) inversion method to examine g 1, it is clear that there is highly significant temperature sensitivity of g 1 under constant VPDL, between 5 and 35°C in the three angiosperms Fagus sylvatica, Quercus petraea and Tilia cordata – although not in the gymnosperm Picea abies, that also had a lower g 1. In the angiosperms which showed a temperature sensitivity – values rose from a minimum observed g 1 at c. 20 to 25°C corresponding with an apparent minimum in C i/C a (Diao et al., 2024). At very low temperatures (i.e. 5°C) g 1 was also found to be higher; however, it is unclear whether this represents a true change in the coupling between g s and A, or an artefact that arises due to changes in the signal to noise ratio at very low rates of A and g s (Fig. S9). Decoupled but mechanistically linked Our results (Fig. 2b), together with those of Diao et al. (2024) and in Mills et al. (2024), demonstrate that under constant evaporative demand, g s does not always increase monotonically with T L. Instead, g s may stabilise or decline at high T L. This behaviour results from g s being simultaneously responsive to multiple, potentially antagonistic, environmental cues. At very high T L, biochemical limitations constrain A resulting in a response in g s via an increase in C i that may counter any direct temperature stomatal response (Mills et al., 2024; Takahashi et al., 2025). Furthermore, our incorporation of a temperature‐sensitive g 1 term within the Medlyn et al. (2011) model shows that, within a coupled framework, g s may still increase under conditions where A declines. Similar outcomes also arise in optimisation‐based models that incorporate the benefits of transpirational cooling at high temperatures (Friend, 1995; Sabot et al., 2022). This is consistent with strong selective pressures favouring elevated g s to mitigate thermal stress during temperature extremes (Drake et al., 2018). In this context, it should be noted that, even when dgsdA < 0, A and g s may remain mechanistically linked. Other sources of temperature sensitivity An often‐implicit assumption when employing stomatal models such as Medlyn et al. (2011) is that water loss in the absence of photosynthesis is minimal and can be discounted when considering changes in leaf‐level gas exchange in illuminated leaves (i.e. g 0 ≈ 0). This leads to the prediction that when A declines to zero due to thermal constraints g s also tends to zero (Fig. S6). However, minimum conductance (g min) due to incomplete stomatal closure or cuticular leakiness, while typically low, is not zero (Duursma et al., 2019). Moreover, emerging evidence suggests a thermal sensitivity in g min (Slot et al., 2021) and that this behaviour occurs through multiple physiological pathways (Garen & Michaletz, 2025). While Slot et al. (2021) showed minimal change in g min of tropical tree species between 30 and 40°C, in situations where measured water loss declines substantially, g min and its thermal sensitivity may play a more prominent role – leading to a breakdown in the assumption that g 0 is negligible. Incorporating temperature‐dependent formulations of g min, or its mathematical operator equivalent g 0, could therefore improve stomatal models, particularly under supra‐optimal thermal or VPD stress where this residual conductance could play a larger proportional role in overall water loss. Conclusion and outlook Most vegetation models that employ a photosynthesis‐conductance coupled formulation, such as Medlyn et al. (2011) apply static values for parameters like g 1 (Bonan et al., 2014; Oliver et al., 2022). However, our results, together with emerging evidence from other systems (Chen et al., 2025), demonstrate that the coupling between A and g s in well‐watered plants is dynamically modulated by temperature and growth environment. In our data, incorporating a generalised thermal sensitivity of g 1 improved predictions of g s across all experimental treatments tested here, with the largest gains in accuracy occurring at supra‐optimal temperatures for photosynthesis where models assuming a constant g 1 systematically underestimate g s. Importantly, the value of this approach is expected to be greatest when extrapolating stomatal behaviour beyond the temperature range over which g 1 has been parameterised, such as under future climate scenarios – characterised by more frequent and intense temperature extremes. While the formulation presented here does not yet account for second‐order interactions and possible shifts in the thermal sensitivity of g 1 with VPD regime or CO2 concentration, it offers a mechanistically grounded and computationally tractable pathway for improving projections of plant water use and carbon–water coupling in Earth system and vegetation models operating under projected future climate change. Competing interests None declared. Author contributions AWC, LC, PJF and PC planned and designed the research. AWC and KBM performed experiments and analysed data with interpretation from all authors. AWC, LAC and KBM wrote the manuscript draft, with all authors contributing to the final edit. Disclaimer The New Phytologist Foundation remains neutral with regard to jurisdictional claims in maps and in any institutional affiliations. Supporting information Fig. S1 Temperature dependence of biophysical parameters underlying Medlyn et al. (2011) model parameter g 1. Fig. S2 Images of study species used in this study. Fig. S3 Diel pattern of environmental conditions within growth chambers across the experimental growing period. Fig. S4 Comparison between leaf temperature (T L) and vapour pressure deficit of the leaf (VPDL) during gas‐exchange measurements. Fig. S5 Comparison of g 1 estimates made using regression‐based estimate and the inversion method. Fig. S6 Simulation of leaf‐level gas exchange using Photosyn function in the R‐package Plantecophys (Duursma, 2015), assuming either constant g 1 or temperature‐sensitive g 1. Fig. S7 Simulation of leaf‐level gas exchange using Photosyn function in the R‐package Plantecophys (Duursma, 2015), assuming temperature‐sensitive g 1 under either near‐ambient control (420 ppm) or elevated (820 ppm) [CO2]. Fig. S8 Simulation of leaf‐level gas exchange using Photosyn function in the R‐package Plantecophys (Duursma, 2015), assuming temperature‐sensitive g 1 under either constant VPDL (1.6 kPa) or with VPDL increasing in line with T L. Fig. S9 Impact of temperature on leaf‐level gas exchange in four temperate tree species, data from Diao et al. (2024). Please note: Wiley is not responsible for the content or functionality of any Supporting Information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office.