Contrasting responses of stomatal conductance and photosynthetic capacity to warming and elevated CO2 in the tropical tree species Alchornea glandulosa under heatwave conditions

Abstract


Introduction
Global atmospheric CO2 concentrations are increasing, as are air temperatures, with both patterns expected to continue in the coming decades.Plants are a critical part of global biogeochemical cycles, at the interface of the atmosphere and the land surface, with forests storing 65% of terrestrial aboveground biomass (Liu et al., 2015).Plants respond to environmental stimuli, with long-term adaptation and short-term acclimation to changes in light, temperature and other conditions.Photosynthesis, evapotranspiration, and respiration are the primary functions of leaves.Our understanding of leaf-level physiology is used to drive vegetation and land surface models, and hence to project future climate.Experimental research on the responses of forests to elevated CO2 has been heavily focussed on temperate ecosystems (Leakey et al. 2012) despite tropical forests stocking more carbon than temperate and boreal forests combined (Pan et al. 2011).Similarly, there are very few studies of thermal acclimation on tropical species (Dusenge & Way 2017).Although temperature increases in the tropics are predicted to be smaller than in other regions (e.g.boreal zone, Collins et al. 2013), tropical forests experience much lower diurnal and seasonal variation in temperature than temperate or boreal forests, and over geological time have experienced a relatively stable climate, potentially reducing the acclimation potential of tropical tree species (Janzen 1967, Dusenge & Way 2017).
Investigating the responses of tropical tree species to temperature and CO2 is therefore a research priority.
Increasing air temperatures and atmospheric CO2 concentrations lead to changes in stomatal conductance (gs) over short and long timescales (Way et al., 2015).In the short-term (instantaneous responses), increasing air temperatures typically lead to a reduction in gs (Way et al., 2015, Slot & Winter 2017a) due to stomatal closure with increasing vapour pressure deficit (D), which prevents excessive water loss under high evaporative demand.At very high temperatures, gs may actually increase in order to avoid reaching dangerously high leaf temperatures (Slot et al., 2016, Slot & Winter 2017b, Urban et al., 2017, Drake et al., 2018).Evidence of acclimation of gs to higher temperatures in trees over the long-term is varied, however some species show declines (Way et al., 2015).The instantaneous response of gs to increased CO2 is to decrease, which reduces water loss while maintaining a high internal leaf CO2 concentration (ci) (Gaastra, 1959).Similarly, under long-term CO2 enrichment, gs reduces.Such declines in gs may increase leaf temperature (TL) through reduced evaporative cooling (under increased air temperatures, reduced evaporative cooling would also depend on the extent of increased D, Oren et al. 1999).Higher TL could push leaves beyond their photosynthetic temperature optima (Topt) (Doughty & Goulden 2008, Slot & Winter 2017c), and potentially above their physiological temperature tolerances (O'Sullivan et al., 2017) causing permanent leaf damage under extreme heat conditions (Warren et al., 2011).While the response of gs to combined elevated CO2 (eC) and temperature (eT) has rarely been tested (Way et al., 2015, Becklin et al., 2017), experiments on eucalyptus (Ghannoum et al., 2010), douglas-fir (Lewis et al., 2002) and loblolly pine (Wertin et al. 2010) showed little interactive effect; if the two do interact and lead to even greater decreases in gs, this would increase TL further.
Long-term increasing air temperatures and CO2 concentrations are also predicted to induce changes in net photosynthesis, both directly by impacting biochemical processes and indirectly through changes in gs.Increases in TL either directly from increased air temperatures or indirectly from a long-term reduction in gs could shift the leaf beyond Topt, leading to reductions in photosynthesis.Some experimental studies have shown partial photosynthetic acclimation to increasing temperatures through increases in Topt (Yamori et al., 2014, Slot & Winter 2017b), which could occur due to alterations in membrane fluidity, expression of heat shock proteins, and production of greater quantities of Rubisco activase or a heat-stable Rubisco activase (Yamori et al. 2014).These changes would lead to altered temperature responses of the photosynthetic capacities Vcmax (maximum rate of carboxylation) and Jmax (maximum rate of electron transport).A recent study of four tropical tree species showed that gs rather than Vcmax or Jmax limited net photosynthesis beyond Topt (Slot & Winter 2017a), and hence a change to the temperature (or D) response of gs could also be important for shifts in Topt.Photosynthetic capacities are also influenced by growth CO2 concentrations.Under high CO2, Rubisco concentrations typically reduce and hence Vcmax declines (Way et al. 2015).Decreases in gs (as a consequence of increased air temperature or CO2) lead to reduced ci which can reduce assimilation.Under high CO2 concentrations, this effect could be limited if ci remains above the Rubisco limited portion of the A-ci curve, however the downregulation of Vcmax commonly observed results in plants still being Rubisco limited even at high CO2 (Ainsworth & Rogers 2007) and hence reduced gs could still reduce assimilation (Way et al., 2015).
The effect of decreased conductance on TL is well understood biophysically (Jones 1992) and is expected to influence TL under elevated CO2 (Drake et al. 1997), as has been shown in a small number of experiments (e.g.Siebke et al. 2002, Sigut et al. 2015).However, this effect has not been investigated in any tropical species.
Furthermore, because TL and, to a lesser extent, gs show high temporal variation with changing microclimate (e.g.Fauset et al., 2018), to fully investigate the effect of altered gs as a response to elevated temperature and CO2 it is necessary to measure TL and microclimate with a high temporal resolution.
In this study, we address the following questions using a factorial eT x eC open top chamber experiment with juveniles of tropical tree species Alchornea glandulosa (Poepp.& Endl) (Euphorbiaceae): i) To what extent does gs reduce with elevated temperature (eT) and elevated CO2 (eC) treatments?; ii) Is there an interactive effect of eT and eC on gs?; iii) Does reduced gs as a result of eT and/or eC cause an increase in TL?; iv) Do the photosynthetic temperature optima (Topt) and temperature response of photosynthetic capacities (Vcmax, Jmax) shift with higher leaf temperatures?A.
glandulosa is a pioneer species often found, but not restricted to, riverine environments (Pascotto 2006), distributed in the Atlantic forest, western Amazon/Andes and central America (GBIF Secretariat 2017), with over 100,000,000 individual trees estimated to occur in the Amazon (ter Steege et al., 2013).It is utilized as a timber species, produces medicinal compounds and is used for reforestation in the Atlantic forest region.The fruits of this tree are an important food source for birds (Pascotto 2006).This species was also selected because leaf temperature and stomatal conductance field data for congeneric species Alchornea triplinervia were available from the Atlantic forest (Fauset et al., 2018).

Experimental setup
The study was carried out at the University of São Paulo from February to March 2017 (23.56° S, 46.73° W, elevation 760 m).Alchornea glandulosa seedlings were sourced from a local plant nursery where they were germinated in shade houses before growing for 12 months outside.
The seedlings were moved to the glasshouse in September 2016 and in November transferred into containers (4l PVC pots with one plant per plot).Hoagland fertilizer solution was added every 2 weeks.The experiment was conducted using four polycarbonate open top chambers (OTCs) with modifications (Aidar et al., 2002) located within the glass house.The four treatments were: i) control (aTaC), ii) elevated CO2 (ambient temperature, 800 ppm CO2, aTeC), iii) elevated temperature (temperature 1.5°C above ambient, ambient CO2, eTaC), and iv) elevated CO2 and elevated temperature (temperature 1.5°C above ambient, 800 ppm CO2, eTeC).Each chamber had an air inlet at the base with a fan, and a spiral heater and/or CO2 gas inlet was present depending on the treatment (Figure S1).Temperature within the chamber was thermostatically controlled using RICS software (Remote Integrated Control System) with the heater switched on or off to maintain a higher temperature than the unheated chambers.No attempt was made to control for differences in D due to temperature treatments as increases in temperature would be associated with increases in D under future conditions assuming no change in relative humidity.CO2 was passively added to the eC treatments through the use of pressurized CO2 cylinders.
The CO2 concentrations of the eC chambers was monitored daily and the flow into the chambers altered at a valve if the concentration decreased.Further details of the experimental design can be found in Aidar et al., 2002 andde Souza et al. (2008).Ten seedlings were placed into each chamber on 1 February 2017 and allowed to acclimate for one month before measurements began.Vertical height of each seedling was recorded prior to placement in the OTCs, and placement of seedlings into OTCs was stratified to ensure an even spread of vertical heights.

Microclimate measurements
Within each OTC air temperature (TA), relative humidity (h) and CO2 concentration were measured at 5 min intervals (Testo 535, Testo Inc., Flanders, NJ, USA).An additional TA sensor (107 thermistor, Campbell Scientific) recorded air temperature every 10 s inside each chamber.

Leaf temperature and leaf surface PAR
The eight healthiest of the ten seedlings in each chamber were selected for measurement of leaf temperature.On each selected seedling, one fully expanded healthy leaf was chosen (typically the fourth or fifth newest leaf).These leaves were formed inside the glass house but prior to movement of the seedling into the OTCs.
Prior to selection, we verified that the leaves were photosynthetically active.A twojunction thermocouple (copper-constantan, type T) that measured leaf-to-air temperature difference (ΔTL) was attached to the abaxial surface of each sample leaf using a piece of breathable tape (Transpore, 3M, St. Paul MN) following the protocol of Fauset et al. (2018).One thermocouple was used per leaf.Absolute leaf temperatures (TL) were calculated from ΔTL and TA in each chamber measured by the thermistors.A photosynthetically active radiation (PAR) sensor built to the specification of Fielder & Comeau (2000) was positioned adjacent to each sample leaf at the same angle and orientation.PAR sensors were calibrated against a quantum sensor (LightScout, Spectrum Technologies, Aurora, Illinois).ΔTL and leaf surface PAR were monitored continuously at 10 s measuring frequency between 24 February -15 March 2017 using two CR800 data loggers and two AM16/32 multiplexers (Campbell Scientific).Measurements of some leaves were terminated between 10 and 15 March.See Fauset et al. (2017Fauset et al. ( , 2018) ) for further details of these sensors.

Stomatal Conductance
Stomatal conductance (gs) of each leaf temperature sample leaf was measured under growth conditions inside the chambers on 19 occasions over six days (including four days where gs of each leaf was measured at least four times, 28 February -7 March 2017) using an SC-1 porometer (Decagon).For each time point, two measurements of gs were recorded, one from either side of the midrib, and the mean value was used for analysis.

Photosynthetic measurements
The temperature response of photosynthesis was measured using a LI-COR 6400XT portable photosynthesis measurement system (LI-COR, Nebraska).Data were collected from 10 -18 March 2018.Light response curves on 3 leaves showed saturating photosynthesis at 800 μmol m -2 s -1 PAR (Figure S2), hence all measurements were taken at 800 μmol m -2 s -1 PAR using the standard red-blue LED light source.Note that the glasshouse roof was made of a diffusing plastic which reduced the incoming PAR by c. 60 % compared with the outside, and leaf level PAR reached c. 800 μmol m -2 s -1 , varying with leaf angle and orientation.Three seedlings from each OTC were selected for photosynthesis measurements and the leaf measurements were performed on the same leaf as leaf temperature monitoring.Two sets of measurements were made, net photosynthesis-temperature curves (A-TL curves where net photosynthesis at saturating light intensity is measured at different temperatures), and A-ci curves (where net photosynthesis at saturating light intensity is measured at different CO2 concentrations) at three different temperatures.A-T curves were run with the CO2 concentration of the relevant OTC (either 400 of 800 ppm CO2) and assimilation was measured at leaf chamber temperatures of 20,25,27,29,31,33,35 and 40 °C, with 5 measurements recorded at each temperature after the photosynthetic rate and gs had stabilized.Measurements at 20, 30 and 35 °C were supplemented using the relevant measurements from the A-ci curves.A-ci curves used the following sequence of CO2 concentrations (ppm); 400,200,100,50,400,600,800,1200,1500,2000.A-ci curves were performed at three temperatures, 20, 30, and 35 °C, and each curve was performed twice for each leaf on either side of the midrib.
For all measurements, h was maintained as close as possible to 50 % using a combination of desiccant and adjusting the air flow rate; it was difficult to maintain this h at leaf temperatures above 37 °C (on average 46 %, minimum values were 40 %).The temperature of the chamber was mostly controlled using the inbuilt temperature control system.In addition, for most of the measurements the sensor head was placed inside a specially designed temperature control chamber to enable better control of the chamber temperature (Yepes Mayorga 2010).The temperature control box was switched off during measurements but was used to aid the change of chamber temperature between measurements.Measurements were made at an atmospheric pressure in the greenhouse of 92.6 kPa.

Plant growth
Vertical height (from soil surface) and number of leaves of each seedling was measured three times (1 and 21 February, and 16 March).On the latter two measurement days, the length of the seedling from the soil surface to the end of the longest branch was also recorded, and on 16 March the total plant length including all branches was recorded.

Data analysis
Differences in microclimate between OTCs (air temperature, CO2 concentration, h and D) were tested using ANOVA and Tukey post-hoc test.
The effects of the warming and the elevated CO2 treatments on gs (porometer measurements pooled from all times of day) were tested using two-way ANOVA with a mixed effects model with leaf as a random factor to account for multiple measurements of the same leaves (function 'lme' of the R package nlme, Pinheiro et al. 2017).To investigate the response of gs to microclimate variables and under different treatments, all possible models of PAR and leaf-to-air vapour pressure deficit DL (where leaf temperature was taken from thermocouple data), with interactions with CO2 treatment and warming treatment were compared using AIC to select the best model with the function 'dredge' in R package MuMIn (Bartoń 2017).
Again, a linear mixed effect model with leaf as a random factor was used to account for multiple measurements of the same leaf/seedling.A quadratic effect of time was also included in the model to account for diurnal changes in gs not directly linked to PAR, temperature or DL.R 2 for mixed-effects models are given using as the marginal pseudo R 2 that accounts for fixed factors only rather than the conditional pseudo R 2 which also accounts for random effects (Nakagawa & Schielzeth 2013) unless otherwise stated; R 2 values for mixed effects models were calculated using the function provided in the R package MuMIn.We also estimated the g1 parameter of the optimal stomatal conductance model (Medlyn et al. 2011, Lin et al. 2015) from the A-TL curve data collected with the LI-COR 6400.
where Ca is the atmospheric CO2 concentration in the leaf chamber.The model was fit for each leaf, and the g1 parameter was compared between chambers using ANOVA.
Because leaf temperatures are strongly influenced by microclimate (Jones 1993, Fauset et al. 2018), to assess the influence of treatment on TL it is necessary to compare TL within microclimatic envelopes.We subsetted the data into envelopes based on leaf-level PAR, chamber air temperature and D. The data was split into low (100 -200 μmol m -2 s -1 ), medium (400 -500 μmol m -2 s -1 ) and high (700 -800 μmol m -2 s -1 ) PAR, and low (28 -30 °C, 1 -2 kPa), medium (33 -35 °C, 2 -3 kPa), and high (38 -40 °C, 3 -4 kPa) air temperature and D. An unanticipated effect of the switching on and off of the heater in the warmed chambers was a cycle in leaf temperature.This was particularly clear at night, but also occurred during the day.
When the heater was switched on, the ΔTL became more negative as the air heated faster than the leaf (Figure S3).The ΔTL then rose to reach an equilibrium temperature.Because of this cycle in the ΔTL data, it was not possible to compare leaf temperatures directly between the ambient and heated chambers, and hence direct comparisons on ΔTL were only made between CO2 treatments within temperature treatments.
The temperature response of photosynthesis is typically modelled as a parabolic curve which provides a Topt parameter (e.g.Robakowski et al. 2012).However, as no evidence of a decline of A with increasing TL was found (see section 3.4), we could not use the parabolic curve to find Topt (Fig. S4) which was beyond the range of our measurements.Hence, a linear mixed effect model with leaf as a random factor was used to test the relationship between A and TL.As for stomatal conductance we selected the best model based on AIC from all possible models, here including TL as a continuous fixed effect and interactions with CO2 treatment and warming treatment.
Vcmax and Jmax were estimated for each leaf and each temperature from the A-ci curve using the Farquhar-von Caemmerer-Berry model using the R package plantecophys (Duursma 2015).For some curves (six for Jmax and one for Vcmax, all at 20 °C), the parameters could not be adequately estimated and estimates were not used.Of the remaining fits, the root mean square error ranged 0.18 -1.57μmol m -2 s -1 .The temperature responses of Vcmax and Jmax were modelled using the Arrhenius function (Medlyn et al. 2002) where k25 is the value of Vcmax or Jmax at 25 °C, Ea is the activation energy (kJ mol -1 ), Tk is the leaf temperature (°K) and R is the universal gas constant (8.314J mol -1 K -1 ) .
The parameters were fit using non-linear least squares (R function nls).This function was fit separately for each chamber, and significant differences in parameter estimates were tested by comparing the 95 % confidence intervals (following Varhammar et al.

2015)
. A peaked Arrhenius function was not used as the data did not show a decline in Vcmax or Jmax at high temperatures.

Microclimate over the study period
The experimental period coincided with an anomalously hot summer in São Paulo city including a 4 day heatwave (Figure 1).Using the definition of a heatwave from Russo et al. (2015) as ≥3 consecutive days where the maximum temperature exceeds the 90 th percentile of maximum temperatures from a monthly window for the period 1981-2010, and climate data for the Mirante de Santana weather station (INMET), a four day heatwave period occurred (maximum temperatures above 32.3 °C) in mid-February (Figure 1, Figure S5).The heatwave occurred during the acclimation period but before the initiation of data collection.During this time the maximum daily air temperatures within the OTCs exceeded 45 °C (Figure 1).
Mean daily temperatures within the OTCs over the acclimation and measurement periods were significantly different between chambers (F = 5.4,P = 0.001, ANOVA, Figure 2a).Temperatures were significantly lower in the aTeC treatment (28.8  2.3 °C mean  SD) than the eTeC treatment (30.4  2.3 °C), however the difference between aTaC (29.0  2.2 °C) and eTaC (30.3  2.4 °C) was marginally insignificant (P = 0.07, Tukey post-hoc test, Figure 1a).Mean daily CO2 concentration was significantly higher in the aTeC and eTeC treatments (829.9  71.6 ppm and 836.7  70.6 ppm, respectively, Figure 1b) than the ambient CO2 treatments, however the concentration in the eTaC chamber (399.0  8.9 ppm) was significantly lower than the aTaC chamber (459.2  12.2 ppm).Relative humidity also varied by treatment with lower values in the elevated CO2 treatments (Figure 2c), and D was higher in elevated temperature treatments, significantly so for eTeC (Figure 2d).

Stomatal Conductance
Analysing gs data with measurements at all times of day pooled, gs was significantly lower under the elevated temperature treatments (P = 0.0001, mixed effects model with leaf as a random factor), with no significant effect of CO2 treatment (Figure 3).
Conductance was highest in the control treatment and similarly low in both elevated temperature treatments, with an intermediate gs in the aTeC treatment (Figure 3).
The best mixed effects model of gs accounting for microclimate and diurnal changes included time of day, PAR, DL, and interactions between DL, warming treatment and CO2 treatment (Figure 4, Table 1).The overall pseudo marginal R 2 of the model was 0.38.If the random effect of leaf is also accounted for, the pseudo conditional R 2 increases to 0.67 showing that there is high leaf-to-leaf variation in gs (Supplementary Figure 6).Interaction plots (Figure 4) of the model show that the relationship between gs and DL was weak (with no significant effect of DL alone, Table 1) and varied between treatments (interactions between heat treatment and DL, and heat treatment, CO2 treatment and DL were significant, Table 1).Under the aTeC and eTeC treatments gs was fairly invariant with DL, whilst under the eTaC treatment gs declined with DL and under the control aTaC treatment gs increased with DL.
However, there is large scatter in the data (Figure 4, Figure S6).
The parameter g1 (inversely proportional with the carbon cost of transpiration and hence low when a plant is conservative in its water use) estimated from the A-TL curves did not show any significant differences between chambers, despite a lower mean for the eTaC chamber (Figure 5).

Observed Leaf Temperatures
Diurnal patterns of average ΔTL, TL, PAR and D are shown for all chambers in Figure 6 based on the period 24 February -15 March 2017.There are differences in the patterns of average ΔTL for each chamber (Figure 6c,d), and these patterns are linked to the patterns of average PAR (Figure 6e,f).In order to properly compare the leaf temperatures between different leaves and chambers, the varying microclimate needs to be accounted for.
Mean ΔTL values were not significantly different between elevated and ambient CO2 within the warming treatment under any specified microclimate (Figure 7, eTeC versus eTaC).In contrast, under the majority of microclimates tested ΔTL values were significantly higher in the elevated CO2 treatment compared to the ambient CO2 treatment when under ambient temperatures (Figure 7, aTeC versus aTaC).The microclimate conditions under which no significant differences were found were both in the high PAR category where there were much fewer data points, and the pattern in the data was similar to other microclimates.The extent of the difference in ΔTL between aTaC and aTeC increased under increasing air temperature and increasing PAR, with a difference of 2.8 °C under high PAR and high air temperature.Analysing the data for TL rather than ΔTL produced the same results (data not shown).

Photosynthetic Temperature Response Curves
Despite measuring photosynthesis at leaf temperatures up to 40 °C, there was no evidence of reaching Topt as A continued to increase with TL for the majority of leaves (Figure 8).Consequently, estimation of Topt was not attempted and linear models were used to analyse the A-TL curves.There was no significant effect of temperature treatment, however TL, CO2 treatment and their interaction were included in the best model.A (measured at the growth CO2 concentration) was higher and the slope of the A-T relationship was steeper under the elevated CO2 treatments (Figure 8).The marginal pseudo-R 2 of the model was 0.53, and all model terms (TL, CO2 treatment and their interaction) were significant (Table 2).

Temperature Responses of Vcmax and Jmax
As for A, both Vcmax and Jmax increased with measurement temperature (Figure 9) and no optimum temperature was found within the measurement range (20 -35 °C).
Temperature treatment had no significant effect on either of the two variables, however Vcmax was lower and the temperature response of Vcmax was weaker (lower activation energy) under elevated CO2, with significant differences between eTaC and eTeC treatments (Figure 9, Table 3).The ratio of Jmax/Vcmax decreased with increasing temperature (30 -35 °C, not sufficient Jmax data at 20°C), and was significantly higher in the elevated CO2 treatment (Figure S7).

Seedling growth
There were no significant effects of treatment on seedling size at any time point during the experiment (vertical height, total branch length, number of leaves, Figure S8).

Discussion
In this study we present a factorial elevated temperature and elevated CO2 experiment with juveniles of a tropical pioneer species.The study was performed under high temperature conditions including a heatwave during the acclimation period (Figure 1).
Our key findings are i) that the elevated temperature treatment had a stronger influence on gs than elevated CO2 (Figure 3, Table 1), ii) that reduced gs caused a change in leaf temperatures (Figure 7), iii) that net photosynthesis and photosynthetic capacities show very high temperature tolerances with no evidence for acclimation to the elevated temperature treatment (Figure 8), and iv) that there was no interactive effect of temperature and CO2 treatment on gs (Figure 3, Table 1)

Temperature and CO2 impacts on stomatal conductance
As expected, gs declined in the eC treatments compared with the control, as has been shown in many other studies.Here we find a 21.2 % reduction (95% CI 10.6 -30.2 % based on bootstrapping) in our aTeC treatment compared with the control (Figure 2).
In forest free air CO2 enrichment (FACE) experiments with CO2 elevated by 200 ppm gs declines on average by c. 20 % (Ainsworth & Rogers 2007), with stronger declines in angiosperm than gymnosperm species (Brodribb et al. 2009).Past chamber experiments performed on angiosperm trees with a doubling of CO2 show an average gs reduction of c. 18 % (from data in Saxe et al. 1998).Our data therefore shows consistency with species from other biomes, but with few tropical species included in existing studies.The literature on tropical species shows wide variation (Berryman et al. 1994, Goodfellow et al. 1997, Liang et al. 2001, Leakey et al. 2002, Khurana & Singh 2004, Cernusak et al. 2011, Dalling et al. 2016, Wahidah et al. 2017).Data from eight publications covering 22 tropical angiosperm species with CO2 enrichment in the range 300-400 ppm showed an average change in gs of 28.6 ± 18.4 % SD reduction.One species (Chrysophyllum cainito) showed a very small increase (Dalling et al. 2016), and the largest reduction of 61 % was shown by Inga punctata (Cernusak et al. 2011).Hence, the reduction we observed was below average but well within the range of observations of other tropical species in experiments.
A limitation of our experiment and its comparability with other studies is the short duration of exposure to the treatments.We measured the physiological responses on leaves formed before initiation of the experiment, which had been exposed to the treatments for c. 5 weeks.As stomatal properties (e.g.density) often differ on leaves formed in high CO2 environments (Saxe et al. 1998), there could potentially be greater changes than we observed, had new leaves formed.Whilst this is quite possible, the long-term response of gs to CO2 is typically similar to the short term response (Way et al. 2015), and hence while the mechanism of reduced gs may be different in short and long-term studies, the gs may be similar.However, a caveat to our results is that to truly observe the acclimation of leaves to the treatments, longer acclimation periods and production of new leaves is necessary.
The observed responses of gs to elevated temperature vary considerably in the few studies available (Way et al. 2015).Here we find strong reductions in gs in the temperature treatments with a 49.6 % (95% CI 42.2 -56.5 %) reduction under the eTaC treatment and 53.0 % (95% CI 52.9 -58.3 %) reduction in the combined eTeC treatment, although we did not find any significant difference between treatments for the g1 parameter value.This may be because the Medlyn et al. (2011) model incorporates the ambient CO2 concentration, and if the short-term and long term gs response to CO2 is the same there would not be a difference.The declines in gs are not driven purely by higher DL in the eT chambers as there are significant differences even when DL is controlled for (Table 1, Figure 4) or when gs is analysed within a narrow DL range (data not shown).This shows acclimation of gs due to higher air temperature and/or DL (both quantities strongly co-varied) which will reduce water loss from the plants.There were no significant differences in gs between the eTaC and eTeC treatments, hence the response to the temperature treatment (with significant differences) was stronger than the response to the CO2 treatment.The result is surprising given the very mixed results in the limited literature on elevated temperature impacts on gs, and even more so given that in this study the temperature treatment was fairly modest (+1.5 °C) compared to the CO2 treatment (+ 400 ppm), although the effect of eC on gs may have been limited by the lack of new leaf development (as stated above).This finding could also be because the ambient temperatures were very hot inside the chambers throughout the experiment and especially during the acclimation phase (Figure 1), which meant that a small increase in air temperature had a large impact, with stomata closing to reduce water loss.An experimental study of gas exchange of Solanum lycopersicum (cherry tomato) measured during and following a +14 °C heatwave showed reduced gs during the heatwave, which remained low when measured 5 days after the heatwave (Duan et al., 2016).Similarly, Duarte et al. (2016) found reduced gs of Pseudotsuga menziesii (Douglas fir) during +12 °C heatwaves which remained when measured one month later.This is somewhat in contrast with recent research suggesting stomata remain open under very high air temperatures for increased evaporative cooling (Slot et al., 2016, Slot & Winter 2017b, Urban et al., 2017, Drake et al, 2018).Responses are likely to be species specific, with an example of a late successional species reducing gs under heatwave conditions while a pioneer species showed increased gs (Vargas & Cordero 2013).However these studies are assessing the instantaneous response of gs to short-term warming rather than the long-term response.A field study reporting the impact of four months of experimentally elevated temperature on gs of existing leaves showed a c. 25 % reduction with 2 °C temperature increase averaged across six tropical species (Doughty 2011), lower than we observed.However, in contrast to our results for Alchornea glandulosa, Yepes Mayorga (2010) found that gs of Hymenea courbaril was more strongly controlled by elevated CO2 than elevated temperature in a similar study, as did Ameye et al. ( 2012) in a study of temperate species Quercus rubra and Pinus taeda in treatments of elevated by 320 ppm and TA elevated by 3 °C or with heat waves.Two studies of subtropical/temperate Eucalytpus spp.found no difference in gs of under treatments of CO2 elevated by 240 ppm and TA elevated by 3 °C or 4 °C after 15 and 7 months of acclimation respectively (Quentin et al. 2013, Duan et al. 2018).While more studies are needed to see if there is a general pattern for tropical broadleaf species, the results of this study suggest that there could be larger implications of rising temperature than rising CO2 for water use of at least some species of tropical tree, and even implications of modest temperature rises such as the ambitious aims of the Paris Agreement (UNFCCC 2015).
The gs dataset also showed a weak relationship with respect to DL, which varied with treatment (Table 1, Figure 3).Other studies with a congeneric species show that gs of A. triplinervia is more weakly linked to DL than other measured species (García-Núñez et al. 1995, Fauset et al. 2018).A weaker relationship between gs and DL is expected for low wood density pioneer species compared to species with higher wood density (Lin et al. 2015).In addition, as the species is commonly found in riparian areas (and therefore with access to a good water supply), its lack of stomatal control is not surprising.Our results show that despite a weak instantaneous response of gs to microclimate, A. glandulosa still showed acclimation and reduction in gs in response to long-term microclimate change.Hence, the short-term response of gs does not provide information on the long-term response.

CO2 impacts on leaf temperature
The lower gs as a result of elevated CO2 caused increases in leaf temperatures (Figure 5).The differences in ΔTL increased with increasing PAR at the leaf surface, and to a lesser extent with increasing air temperature and D. This shows that the differences in leaf temperatures due to CO2-altered gs are more apparent under high thermal stress conditions (high PAR and high air temperature), and therefore that this impact is likely to be stronger under heat waves, which are expected to increase in frequency during the 21 st century (Coumou & Robinson 2013).When at high air temperatures, differences in ΔTL due to reduced gs could have significant consequences, as seen in observations of premature leaf senescence during a heatwave in a temperate FACE experiment (Warren et al. 2011).While the average differences in ΔTL between aTaC and aTeC reached 2.8 °C under high light and air temperature, the light conditions were limited by the greenhouse environment which reached only 1000 mol m -2 s -1 .
Under field conditions where incoming PAR can reach over 2500 mol m -2 s -1 the impact of reduced gs on ΔTL could be much higher.Unfortunately due to ΔTL fluctuations induced by heating the air (Figure S3) it was not possible to assess the impact of the high temperature treatment compared to the control.Within the two high temperature treatments there were no significant differences in ΔTL under any microclimate between the elevated and ambient CO2 treatments, which is expected as they did not show any significant differences in gs.

Temperature and CO2 impacts on photosynthesis
The elevated temperature treatment had no discernible effect on A or photosynthetic capacity and their responses to elevated temperatures.The high temperature tolerance of both A and photosynthetic capacity was marked, with no decline in A found even at 40 °C.Consequently, we were not able to assess shifts in Topt with treatment as Topt tolerance of PSII over very short timescales (days) in response to high temperatures (Drake et al. 2018).Slot et al. (2017c) found that Topt measured in the field in Panama was around the mean maximum daily temperature (30-32 °C) for all 42 species measured, and that, for a smaller sample of four species, it was gs rather than Rubisco activase, Jmax, Vcmax or light respiration that limited the photosynthetic rates at high temperatures (Slot & Winter 2017a).In another study, Topt was higher than daily maximum air temperature in moist and wet tropical forest sites in Puerto Rico (Mau et al. 2018).In the case of the A. glandulosa seedlings measured here, the mean maximum daily temperature over the acclimation and measurement period was 40 -42 °C (varying by treatment, Figure 1), matching the minimum potential Topt of 40 °C, and showing tolerance to the high temperatures to which they were exposed.
Measurement under higher temperatures would be necessary to find the Topt for these plants.Over the measured temperature range, gs, Vcmax and Jmax did not decline.Yet, it should also be noted that the rates of A, Vcmax and Jmax were fairly low (c.3, 40 and 35 mol m -2 s -1 respectively, at the highest values and under ambient CO2).For example, these are lower than A of 12 -16 mol m -2 s -1 across 42 Panamanian species (Slot & Winter 2017c) and 5 -12 mol m -2 s -1 for four species in Puerto Rico (Mau et al. 2018), and Vcmax of 70 -300 mol m -2 s -1 and Jmax of 80 -220 mol m -2 s -1 across four Panamanian species (Slot & Winter 2017a), all at their optimum temperatures.The measured rates are also lower than plants in other high temperature environments e.g.
five desert species with A ranging 19 -35 mol m -2 s -1 (Mooney et al. 1981), and Mediterranean cork oak with Vcmax and Jmax both over 150 mol m -2 s -1 (Ghouil et al. 2003).Thus, high temperature tolerance of photosynthetic machinery in A. glandulosa may come at a cost of lower photosynthetic rates.An alternative explanation for the low photosynthetic rates is the low light conditions within the greenhouse, with maximum leaf surface PAR of 800-1000 mol m -2 s -1 .In the field, maximum PAR is likely to be much higher (> 2000 mol m -2 s -1 ), and leaves may achieve higher photosynthetic rates.The low light conditions in the greenhouse also have implications for the high temperature tolerance observed.Because at high temperatures photosynthetic biochemistry is under greater stress, there is a greater need for photoprotection from high incoming radiation.Perhaps under the higher light conditions found in the field, very high temperature tolerance of photosynthesis may be more difficult to achieve.Field studies under high temperature conditions are needed to establish whether the high tolerance we find here also occurs under natural conditions.
As for gs, the impacts of increased CO2 followed expectations from previous studies with increased net photosynthesis when measured at growth CO2, a steeper slope of A in response to temperature, and downregulation of photosynthetic capacity (Figure 8, 9).The steeper slope is due to the reduction in oxygenation of Rubisco due to higher ci under elevated CO2, which otherwise increases with temperature due to the reduced affinity of Rubisco for CO2 with higher temperature (Long 1991).The effect of the downregulation can be seen when the temperature response of A is plotted with added points taken from the A-ci curves at 400 and 800 ppm CO2 for the elevated and ambient CO2 treatments respectively, showing that without the downregulation of photosynthetic capacity A would have been higher in the elevated CO2 treatment (Figure S9).

Conclusions
This study demonstrates that the tropical tree species Alchornea glandulosa shows strong responses of stomatal conductance to elevated temperature and of photosynthetic parameters to elevated CO2.While a very high temperature tolerance of photosynthesis was observed in this species, photosynthetic rates were low under the high growth temperatures.These results show that this species will be able to cope with the predicted atmospheric changes over the coming century.Therefore, it is an appropriate species for reforestation activities, which are planned and ongoing in the Atlantic forest (Rodrigues et al. 2009).More studies of other species are required to determine whether similar results occur in other forest trees.
Tables Table 1.ANOVA results for stomatal conductance linear mixed effects model.3.
photosynthetic functions (e.g.Rubisco activase activity is strongly temperature Figure 1.Time series of daily maximum temperatures in São Paulo (Mirante de

Figure 2 .
Figure 2. Differences in microclimate variables between chambers a) mean air

Figure 3 .Figure 4 .
Figure 3.Effect of treatment on stomatal conductance where measurements from all

Figure 9 .
Figure 9. Temperature response of Vcmax and Jmax fit with Arrhenius functions.For

Table 2 .
ANOVA results for A-TL linear mixed effects model.CO2 refers to treatment.

Table 3 .
Parameter estimates of Arrhenius functions of the temperature sensitivity of Vcmax and Jmax.Standard errors are given in brackets.Significance of between treatment effects are shown: * P < 0.05,  P < 0.1, ns not significant.Letters denote differences between treatments.