Long‐term drought effects on the thermal sensitivity of Amazon forest trees

Abstract The continued functioning of tropical forests under climate change depends on their resilience to drought and heat. However, there is little understanding of how tropical forests will respond to combinations of these stresses, and no field studies to date have explicitly evaluated whether sustained drought alters sensitivity to temperature. We measured the temperature response of net photosynthesis, foliar respiration and the maximum quantum efficiency of photosystem II (F v/F m) of eight hyper‐dominant Amazonian tree species at the world's longest‐running tropical forest drought experiment, to investigate the effect of drought on forest thermal sensitivity. Despite a 0.6°C–2°C increase in canopy air temperatures following long‐term drought, no change in overall thermal sensitivity of net photosynthesis or respiration was observed. However, photosystem II tolerance to extreme‐heat damage (T 50) was reduced from 50.0 ± 0.3°C to 48.5 ± 0.3°C under drought. Our results suggest that long‐term reductions in precipitation, as projected across much of Amazonia by climate models, are unlikely to greatly alter the response of tropical forests to rising mean temperatures but may increase the risk of leaf thermal damage during heatwaves.

with longer and more severe droughts across much of Amazônia (Allen et al., 2015;Marengo et al., 2018;Toomey et al., 2011). The response of forests to these changes in climate will depend on their capacity to acclimate to changing baseline environmental conditions and resilience to extreme stress (Corlett, 2016;Galbraith et al., 2010;Geange et al., 2021;Smith & Dukes, 2013;Sterck et al., 2016).
Crucially, forest tree species may respond differently to heat and drought stress and this will likely influence changes in species composition, vegetation density and forest ability to sequester carbon (da Costa et al., 2010;Esquivel-Muelbert et al., 2019).
Currently, there are limited data available to aid understanding of tropical forest sensitivity to increasing temperatures. The few studies to date suggest that gas exchange processes and photosynthetic thermotolerance in tropical species are capable of some degree of thermal acclimation, however, considerable variation in both baseline thermal sensitivities and acclimation potential exists amongst species (Atkin & Tjoelker, 2003;Carter et al., 2020Carter et al., , 2021Drake et al., 2016Drake et al., , 2018Slot & Winter, 2017b, 2017cSlot et al., 2014;Tiwari et al., 2021). Very few studies have examined how drought might affect tropical forest thermal sensitivity (Geange et al., 2021;Sastry et al., 2018). This represents an important knowledge gap, as research on crop species suggests the effect of simultaneous heat and drought stress on plant productivity, tissue damage and mortality are not necessarily predictable based on sensitivity to drought or heat stress alone (Rizhsky et al., 2004;Zandalinas et al., 2018).
The thermal sensitivity of a plant is often described using properties (herein referred to as thermal traits) that confer information about the stability of gas exchange processes and/or the integrity of photosynthetic machinery under moderate to extreme heat. For example, T opt describes the optimum temperature for photosynthesis, whilst T span describes the breadth of temperature over which photosynthesis rates are sustained >80% of optimum rates, and T max , the high end of the temperature range within which a leaf is able to assimilate CO 2 (Slot & Winter, 2017b) ( Figure 1a, Table 1). The temperature sensitivity of respiration is usually inferred through comparing basal rates at a standard temperature (25°C, R 25 ), and the steepness of the instantaneous increase in respiration rates over a 10°C rise in temperature (Q 10 ) (Atkin & Tjoelker, 2003) (Figure 1b, Table 1). The above traits can be used to evaluate the capacity of leaves to maintain a positive carbon balance under rising temperatures, some of which form integral components of global vegetation models (Booth et al., 2012;Cox et al., 2000;Galbraith et al., 2010). Alternatively, the temperature at which the maximum quantum efficiency of photosystem II (F v /F m ) is reduced to 50% of its value under non-stressed conditions (T 50 ) ( Figure 1c, Table 1) is a measure of a plant's ability to protect the integrity of photosystem II (PSII) at high temperatures (Figueroa et al., 2003;Lípová et al., 2010;Zhang et al., 2012). T 50 provides thermotolerance estimates comparable to those gained from classical leaf necrosis tests with the same temperature exposure time (Krause et al., 2010). Thus, it is a relevant metric for considering the potential for short periods of extreme temperature stress to impact carbon assimilation.
The way in which combinations of drought and heat affect plant physiological processes is poorly understood, with most knowledge gained from crop species (Geange et al., 2021). However, there is a consensus that drought exacerbates temperature stress by restricting evaporative cooling and increasing leaf temperatures (Suzuki et al., 2014). Short-term coupled heat and drought events, such as those experienced during hot, dry El Niño years, have been linked to reduced tropical forest productivity, contributing to a weaker forest carbon sink (Cavaleri et al., 2017). However, long-term (multi-year to decadal) reductions in precipitation might be expected to interact differently with thermal stress, as plants have more time to employ structural and/or metabolic rate adjustments to sustained water limitation. For instance, net photosynthesis, respiration and leaf Hypothetical examples of the temperature response of net photosynthesis (A net ) (a), foliar respiration in the dark (R net ) (b), and maximum quantum efficiency of photosystem II (F v /F m ) (c), in the control (blue) and the TFE (red), if long-term drought were to induce thermal acclimation in these processes. TFE, through-fall exclusion. thermotolerance might thermally acclimate (i.e., adjust to maintain a positive carbon balance and thermal safety margin) to a sustained drought-induced rise in leaf temperatures ( Figure 1) (Atkin & Tjoelker, 2003;Berry & Bjorkman, 1980;Way & Yamori, 2014).
Thermal acclimation has been observed in leaf respiration (downregulation of R 25 and Q 10 ) and thermotolerance (upregulation of F v /F m at 47.5°C) metrics after relatively short (weeks to months) experimental droughts on tropical saplings (Gauthier et al., 2014;Sastry et al., 2018). Although not previously evaluated in relation to drought, and inconsistent across species, photosynthetic (upregulation of T opt , T span and T max ) and respiratory acclimation has been observed in warming studies on tropical saplings and in some understory shrubs (Carter et al., 2020;Slot & Winter, 2017c;Slot et al., 2014;Mujawamariya et al., 2021). Similarly, an in situ +3°C leaf warming study on two mature tropical tree species observed some photosynthetic and respiratory acclimation, though inconsistent between species and leaf canopy position (Carter et al., 2021). If thermal acclimation were widespread and sustained over long timescales, this might buffer the negative effects of heat and drought combinations on forest productivity. It has also been proposed that a partial overlap in protective mechanisms, such as altering chloroplast membrane compositions (Ladjal et al., 2000), or upregulation of antioxidant scavenging (Gill & Tuteja, 2010), could facilitate crossprotection between drought and warming and that exposure to drought could effectively prime physiological processes for heat exposure (Havaux et al., 1988). However, metabolic profiling studies on crop species, have observed little overlap in cellular responses to drought and heat (Rizhsky et al., 2004;Zandalinas et al., 2018), and therefore prolonged drought might not alter thermal sensitivity. On the other hand, sustained drought may restrict plant capacity to deal with stress-by-products (Gill & Tuteja, 2010), or cellular maintenance and repair. This would weaken plant ability to cope with additional heat stress, due to limited scope for additional upregulation, or because substrate reserves that would otherwise build-up under nonstress conditions become depleted (Shaar-Moshe et al., 2019).
Furthermore, sustained tissue damage under long-term droughtfor example, damage of the plant water transport system, may also constrain physiological processes so that they are less able to adjust to compounding adverse conditions, such as leaf transpirational cooling in response to heat (Rehschuh et al., 2020;Skelton et al., 2017).
In tropical forests where biodiversity is high, both drought and thermal sensitivity (and their acclimation potentials) have been shown to vary markedly amongst co-occurring species (Bittencourt et al., 2020;da Costa et al., 2010;Perez & Feeley, 2020;Rowland, Lobo-do-Vale, et al., 2015;Sastry & Barua, 2017;Slot & Kitajima, 2015;Slot & Winter, 2017b). Owing to the interaction between drought and heat, the effect of sustained drought on species thermal sensitivities will likely also vary. There is some evidence that physiological responses to drought and temperature stress might be coordinated. For instance, short-term drought tolerance (measured via leaf wilting), has been positively related to heat tolerance in saplings of 12 seasonally dry tropical forest tree species (Sastry et al., 2018). Understanding the extent to which this is true for adult tropical forest trees would help identify sensitive taxa and potential alterations in community composition, therefore advancing predictive insights of future ecosystem functioning.
In this study, we use a longstanding (17-year) rainfall exclusion experiment to evaluate the effect of long-term reduction in soil water availability on thermal sensitivity traits in an old-growth tropical forest in eastern Amazonia. Additionally, we compare thermal traits across tree species that have previously been classified as droughttolerant or intolerant, based on their mortality response to the same experimental drought (da Costa et al., 2010;Rowland, Lobo-do-Vale, et al., 2015). We test whether the direction and magnitude of thermal trait adjustment in response to sustained drought treatment is influenced by drought tolerance, and if there is coordination, independent of treatment, between drought and thermal sensitivity.

| Plant material
Measurements were performed at the beginning of the dry season,   Figure S1).
Individual temperature response curves for photosynthesis, respiration and thermotolerance took on average 3 h to complete.

| Temperature response of net photosynthesis
Photosynthesis temperature (A-T) response curves were constructed between 09:00 and 14:00 h using three infra-red greenhouse gas analysers (two LI-6400XT and one LI-6800), with either LED (6400-02B) or fluorometer (6400-40) chamber heads (LI-COR). LI-COR machine use was distributed evenly across sampling (i.e., all species were measured with both LI-6400XT and LI-6800 machines). This enabled confirmation that the LI-COR model used did not result in inherent biases in extracted A-T traits (Supporting Information: b Species was not included in any analysis for T opt , A opt , T span , g sTopt and g sdiff as the shape of A-T curves in individuals from the control plot meant that it was not possible to extract these parameters (Supporting Information: Figure S2, Supporting Information: Methods S1). c Only thermotolerance measurements were collected for this species.
where A opt is the value of A net at the optimum temperature for photosynthesis (T opt ), and Ω is the temperature difference between T opt and the temperature at which A net drops to e -1 (37%) of A opt .
Equation 2 assumes that the slope of the A-T response asymptotes as A net approaches zero, such that A net never passes through zero, which is not the case in nature. To account for this, we also used an adjusted version of the June et al. (2004) where c is a constant that allows A net to pass through zero.
Equations 1-3 all assume a symmetrical A-T response around T opt , which was not the case for all species in this study. Accordingly, for those species with an asymmetrical A-T response, the data were also fitted using the model of Cunningham and Read (2002).
where T min and T max are the low and high-temperature CO 2 compensation points respectively and b and c are fitting coefficients.
The best-fitting equation was determined for each A-T curve using Akaike's Information Criterion (AIC). The equation with the lowest AIC value was then used to extract T opt , A opt , T span (comparing Equations 1-3) and T max (comparing Equations 1, 3 and 4) for each A-T curve. T span was calculated as the temperature range over which A net rates were >80% of A opt (Figure 1a). Most A-T curves had one equation that clearly fit best, however, for A-T curves where multiple equations fit equally well (i.e., within two AIC units of the most parsimonious equation), we compared extracted parameters to confirm they provided similar values (Supporting Information: Figure S5). Furthermore, all best-fitting equations were visually inspected to ensure that extracted parameter values were realistic.
The temperature response of g s (g s -T), derived from water vapour flux in the LI-COR, measured in conjunction with A net , was also fitted using Equations 1-4, replacing A net with g s in all equations. The g s -T equation with the lowest AIC value was used to extract g s rates at the T opt of each corresponding A-T curve (g sTopt ), representing g s rates at optimum temperatures for photosynthesis, and T leaf = 46°C (g sTL46 ), representing g s rates at high leaf temperatures (i.e., temperatures approaching T max for most species). The mean T max of all A-T response curves (T leaf = 46°C) was used as a standard 'adversely high' T leaf , rather than T max values from individual curves, to avoid excluding a few A-T response curves for which T max was not able to be extrapolated (Supporting Information: Figures S2 and S6, Supporting Information: Methods S1). Negative g s rates were considered biologically unrealistic, thus five g sTL46 values (representing 12% of all g sTL46 values) that were extracted from fitted g s -T curves that passed below 0 before T leaf = 46°C were replaced by 0.
Notably, g s -T curves did not always follow a typical bell-shape, with some species departing from optimal stomatal behaviour (Medlyn et al., 2011), by increasing g s as T leaf rose, likely to facilitate leaf cooling (Supporting Information: Figures S3 and S6, Supporting Information: Methods S2). Therefore, to distinguish species across a spectrum, from those that showed a strong downregulation in g s at high relative to optimum temperatures, to those that upregulated g s despite adversely rising temperatures, we also calculated g sdiff , as the difference between g sTL46 and g sTopt . Accordingly, a higher g sdiff denotes a greater downregulation in g s at high temperature, relative to T opt , whilst a negative g sdiff indicates an upregulation in g s at high relative to optimum temperatures.
As with A net and g s , ETR temperature (ETR-T) response curves were fit using Equations 1-3 (Supporting Information: Figure S7), and the optimum temperature for ETR (T optETR ), representing the temperature above which ETR becomes limiting for photosynthesis, was extracted from the ETR-T equation with the lowest AIC value. All fitting was performed using either using the linear 'lm' or non-linear least-squares 'nls' functions in the 'stats' package in R version 4.0.0 (R Core Team, 2020).

| Temperature response of dark respiration
Dark respiration temperature (R-T) response curves were measured between 13:00 and 19:00 h using the LI-6800 with a 2 cm 2 leaf aperture (LI-COR). Dark-adapted leaves (see 'Plant material' section for details) were clipped into leaf chambers with the chamber light source and fluorometer measurement lights switched off, and allowed to stabilize at reference CO 2 , RH and ambient air temperature for at least 15 min before R-T curves were initiated. R-T curves were constructed using the same temperature increments, reference CO 2 concentrations and RH controls as A-T curves (Supporting Information: Figure S8). Respiration rates were <0.1 µmol m -2 s -1 in a few species at the lowest temperatures. Since the precision of the LI-6800 IRGA at 400 µmol mol -1 is ≤0.1 µmol mol -1 , we considered any respiration values <0.1µmol m -2 s -1 could be a result of IRGA signal noise and thus they were removed before analysis. Respiration rates at T leaf~2 5°C (R~2 5 ),~30°C (R~3 0 ) and~45°C (R~4 5 ) were extracted by pooling the three data points closest to the respective leaf temperatures and calculating the mean T leaf and R net . The removal of respiration rates <0.1 µmol m -2 s -1 had no effect on parameter extraction other than preventing extraction of R~2 5 for three out of 42 individuals due to the low rates mentioned above.
Q 10 values were then calculated from R~3 0 , R~4 5 and their respective T leaf values as: where R~3 0 is respiration rate at T leaf~30 and R~4 5 is respiration rate at T leaf~45 . R 25 and R 45 were then extrapolated to exactly 25°C and 45°C respectively using R~2 5 and R~4 5 and their corresponding T leaf values as: where R T is respiration at temperature T, R Tleaf is measured respiration rate at T leaf , and Q 10 is the value estimated using Equation 5 for that individual R-T curve. Whilst R 25 was considered appropriate to represent basal R rates and maintain consistency with previous studies, R~3 0 as opposed to R~2 5 was used with R~4 5 to calculate all Q 10 values to avoid excluding the individuals for which R 25 was not able to be extracted.  Tiwari et al., 2021). For each F v /F m temperature response assay, a three-parameter logistic curve was fitted as:

| Thermotolerance of PSII
leaf 50 (7) where F v /F m.max is the upper asymptote, b is the slope of the decrease in F v /F m with rising temperature, T leaf is the leaf treatment temperature, and T 50 is the inflection point or temperature at which F v /F m drops to 50% of F v /F m.max (Supporting Information: Figure S9). Curves were fitted using 'nlsLM', a modified non-linear least squares function that incorporates the Levenberg Marquardt type fitting algorithm (Moré, 1978), in the 'minpack.lm' package (Elzhov et al., 2016) in R.

| Statistical analysis
Mixed effect model analysis, using the 'lme4' package (Bates et al., 2015) in R version 4.0.0 (R Core Team, 2020), was conducted to test for differences between control and TFE plots, and between drought-tolerant and intolerant species, for all thermal traits separately (see Dataset S1 for the full data set used in this study).
Mixed models were constructed with treatment (plot) and droughttolerance status (tolerant/intolerant) as fixed effects and species as random effects. To test whether thermal traits of drought-tolerant and intolerant species were differently affected by long-term drought, the same analysis was performed on the data separated by drought-tolerance status with treatment (plot) as the fixed affect and species as the random effect. These models were run separately rather than including their interaction in the main model as our relatively low sample size prevented robust detection of interaction effects (Leon & Heo, 2009

| TFE effects on thermal traits
Overall, there were slight shifts in the shape of the average temperature response of A net , and R net in TFE relative to the control in the direction of a thermal acclimation (i.e., T opt displacement to the right and downward shift in R net at high temperatures [ Figure 3]).

| Differing TFE effects on drought-tolerant and intolerant species thermal traits
After separating drought-tolerant and drought-intolerant species, there was still no TFE effect on gas exchange traits for droughttolerant species, but drought-intolerant species showed significant reductions in R 45 (p = 0.03) and Q 10 (p = 0.008) in the TFE relative to the control. Drought-intolerant species R 45 was 31% lower in the TFE , we find no evidence that long-term soil water stress (Bittencourt et al., 2020), and an accompanying drought-induced rise in canopy air temperatures (Figure 2), alters the average thermal sensitivity of A net or R net across the species measured in this study.
However, we do observe a moderate weakening in leaf tolerance to extreme-heat damage, evidenced by a 1.5 ± 0.4°C reduction in mean plots , and expectations based on leaf energy balance considerations (Fauset et al., 2018) all imply higher TFE leaf temperatures. Thus, we expect that the second (limited ability to thermally acclimate) or third (drought-prevented thermal acclimation) possibilities are more likely to explain our results. The lack of observed acclimation in gas exchange contradicts previous studies that have shown strong thermal acclimation in tropical saplings (Mujawamariya et al., 2021;Slot & Winter, 2017c;Slot et al., 2014). However, it provides some support to studies observing limited capacity of tropical species to acclimate to warming (Carter et al., 2020(Carter et al., , 2021. For example, after~1 month of continuous +3°C leaf warming, Carter et al. (2021) observed no photosynthetic acclimation in upper canopy leaves and respiratory acclimation in only one of two adult tropical tree species studied.
The weakened thermotolerance in the TFE relative to the control also contrasts with findings from short-term drought studies (Sastry et al., 2018). Temperature-induced F v /F m decline is associated with a breakdown of the integrity of PSII, which can occur because of a build-up of excess heat energy and/or stress by-products (e.g., reactive oxygen species), that interfere with thylakoid membrane stability, causing disruptions and eventual disassembly of the light-harvesting antenna complex from the core of PSII (Figueroa et al., 2003;Lípová et al., 2010;Zhang et al., 2012).
Short-term drought stress can stimulate antioxidant production to deal with stress by-products (Gill & Tuteja, 2010), which might explain why short-term drought has been found to benefit thermotolerance (Sastry et al., 2018). Conversely, sustained drought, as investigated in this study, might deplete antioxidants and their substrates, therefore reducing capacities to deal with heat stress. Similarly, there is evidence that isoprene emission, a thermoprotective trait held by several species in this study (Jardine et al., 2020;Taylor et al., 2018), can be upregulated to lessen oxidative damage to photosynthetic machinery during short-term drought (Ryan et al., 2014;Tattini et al., 2015;Velikova et al., 2016).
However, isoprene synthesis is carbon intensive (Fang et al., 1996;Sharkey & Loreto, 1993;Tattini et al., 2015), and so may be less advantageous under regular (Taylor et al., 2018), or sustained water stress when CO 2 assimilation is already strained; potentially impeding the ability to protect photosynthetic machinery during subsequent extreme heat (Fortunati et al., 2008). Additionally, some proteins that provide protective functions during leaf desiccation (e.g., LEA), might interfere with the function of heat-shock proteins that otherwise help maintain membrane stability under high temperatures (Rizhsky et al., 2004;Soulages et al., 2002).

| Coordination of drought tolerance and thermal sensitivity
Independent of treatment, the T span of drought-tolerant species was on average, 6.0 ± 1.7°C wider than drought-intolerant species ( Figure 4c). This indicates that drought-tolerant species are more able to maintain high rates of photosynthesis over wide temperature ranges, and therefore, any rise in mean air temperatures would result in a smaller proportional reduction in A net for drought-tolerant, compared to drought-intolerant species. Similarly, g s at high temperatures (g sTL46 ) and downregulation in g s at high temperature relative to optimum leaf temperatures (g sdiff ) were c. four times lower and five times greater in droughtintolerant relative to drought-tolerant species respectively ( Figure 4f, However, whilst a lower g sdiff will influence T leaf by improving transpirational cooling at high temperatures (Suzuki et al., 2014), g s only provides a small contribution, relative to other leaf traits, to the thermal time constant used to infer overall leaf thermal stability (Michaletz et al., 2016). Moreover, a reduction in g s at high temperatures has been recognized to negatively influence T span in other studies (Slot & Winter, 2017a, 2017b, attributed to its direct influence on A net by constraining CO 2 diffusion into the leaf. Therefore, it is plausible that g s and thermal time constant have opposing relationships with T span . The breadth of T span is a direct consequence of the combined temperature sensitivity of stomatal and biochemical processes which limit photosynthesis above T opt (Slot & Winter, 2017a). Whilst drought-tolerant species had significantly lower g sdiff , T optETR (indicative of the thermal optimum of biochemical processes) did not differ between drought-tolerant and drought-intolerant species. Therefore, the narrower T span observed in drought-intolerant species appears more likely a consequence of a more conservative stomatal strategy, as opposed to a greater thermal sensitivity of biochemical processes. Interestingly, species-mean T opt also did not differ between drought-

| Differing TFE effects on drought-tolerant and drought-intolerant species thermal traits
After separating species by drought tolerance, there remained no TFE effect on any thermal traits in drought-tolerant species, aside from T 50 (Figure 4l). However, R 45 and Q 10 of drought-intolerant species were lower in the TFE relative to the control by 31% and 29% respectively (Figure 4j,k). Whilst only marginally significant (p = 0.06; Figure 4e, Supporting Information: Figure S11, Supporting Information: Table S3) g sTopt of drought-intolerant species tended to downregulate in the TFE compared to the control, indicating that even at optimal temperatures for photosynthesis, drought-intolerant species are tending towards a more water conservative stomatal strategy in the TFE. This will likely result in more pronounced increases in leaf temperatures (Fauset et al., 2018;Fauset et al., 2019) in the TFE for drought-intolerant compared to drought-tolerant species that showed no indication of downregulating g sTopt . This potential exposure to higher leaf temperatures might explain why drought-intolerant species exhibited some acclimation of physiological processes whilst drought-tolerant species did not, for example, R 45 and Q 10 were reduced only in drought-intolerant species.
Alternatively, the fact that acclimation was only observed in drought-intolerant species might suggest that they are generally more plastic in their response to stress compared to drought-tolerant species.
In contrast to gas exchange traits that either did not change or indicated a slight thermal acclimation (in the case of R 45 and Q 10 in drought-intolerant species), leaf thermotolerance was slightly weakened in both drought-tolerant and intolerant species.
Surprisingly, it was drought-tolerant species that showed greater reductions in T 50 , both in terms of magnitude and the proportion of species with T 50 reductions (Supporting Information: Figure S11).
Leaf thermotolerances have been found to relate to maximum recorded leaf temperatures in tropical trees (Perez & Feeley, 2020).
If drought-intolerant species are more likely to experience critically high leaf temperatures due to a more conservative stomatal strategy, as our data suggest, then investment in maintenance of high thermotolerance thresholds is likely a higher priority for these species than for drought-tolerant species.
However, the underlying mechanism behind these differences remains unclear.

| Wider context and conclusions
Our results suggest that, unlike short-term drought that might precondition plants for higher temperatures (Gauthier et al., 2014;Ghouil et al., 2003;Havaux et al., 1988;Ladjal et al., 2000;Sastry et al., 2018), sustained drought does not alter thermal sensitivity within moderate temperature ranges, but instead weakens trees' ability to protect photosynthetic machinery under extreme temperatures. Whether or not this is a concern will depend on the frequency with which critical leaf temperatures are reached. Currently, maximum annual air temperature at this site is 33.8°C, so a reduction in T 50 from 50 ± 0.3°C in the control to 48.5 ± 0.3°C under TFE conditions may seem irrelevant. However, leaf temperatures are known to exceed air temperatures by as much as 10°C-18°C (Doughty & Goulden, 2008;Fauset et al., 2018;Rey-Sánchez et al., 2016). Accordingly, current maximum annual leaf temperatures may already approximate thermal thresholds during the hottest part of the year. Thus, even without any climate warming, the 1.5°C reduction in T 50 of TFE trees could be sufficient to increase their risk of thermal damage. Whilst drought-tolerant species appear to have a stronger weakening in thermotolerance compared to droughtintolerant species in the TFE, it is important to contextualize this in terms of their thermal safety margins (the difference between T 50 and maximum leaf temperatures), which may not be that different if drought-tolerant species' ability to maintain g s rates at high temperatures translates to smaller leaf-to-air temperature differences. Indeed, it has been shown recently, on dryland plants, that high thermotolerance does not necessarily imply greater thermal safety but can signify greater hydraulic vulnerability and more acute exposure to heat stress . Whilst logistically challenging, continuous multi-canopy measurements of in situ leaf temperatures would help resolve these complexities and elucidate the extent to which long-term drought might increase the risk of leaf thermal damage in tropical forests.