• Energy Research

Drought threatens tropical rainforests over seasonal to decadal timescales, but the drivers of tree mortality following drought remain poorly understood. It has been suggested that reduced availability of non-structural carbohydrates (NSC) critically increases mortality risk through insufficient carbon supply to metabolism ('carbon starvation'). However, little is known about how NSC stores are affected by drought, especially over the long term, and whether they are more important than hydraulic processes in determining drought-induced mortality. Using data from the world's longest-running experimental drought study in tropical rainforest (in the Brazilian Amazon), we test whether carbon starvation or deterioration of the water-conducting pathways from soil to leaf trigger tree mortality. Biomass loss from mortality in the experimentally droughted forest increased substantially after >10 years of reduced soil moisture availability. The mortality signal was dominated by the death of large trees, which were at a much greater risk of hydraulic deterioration than smaller trees. However, we find no evidence that the droughted trees suffered carbon starvation, as their NSC concentrations were similar to those of non-droughted trees, and growth rates did not decline in either living or dying trees. Our results indicate that hydraulics, rather than carbon starvation, triggers tree death from drought in tropical rainforest.

Increasing heatwaves are threatening forest ecosystems globally. Leaf thermal regulation and tolerance are important for plant survival during heatwaves, though the interaction between these processes and water availability is unclear. Genotypes of the widely distributed foundation tree species Populus fremontii were studied in a controlled common garden during a record summer heatwave—where air temperature exceeded 48 °C. When water was not limiting, all genotypes cooled leaves 2 to 5 °C below air temperatures. Homeothermic cooling was disrupted for weeks following a 72-h reduction in soil water, resulting in leaf temperatures rising 3 °C above air temperature and 1.3 °C above leaf thresholds for physiological damage, despite the water stress having little effect on leaf water potentials. Tradeoffs between leaf thermal safety and hydraulic safety emerged but, regardless of water use strategy, all genotypes experienced significant leaf mortality following water stress. Genotypes from warmer climates showed greater leaf cooling and less leaf mortality after water stress in comparison with genotypes from cooler climates. These results illustrate how brief soil water limitation disrupts leaf thermal regulation and potentially compromises plant survival during extreme heatwaves, thus providing insight into future scenarios in which ecosystems will be challenged with extreme heat and unreliable soil water access.

Feedbacks between land carbon pools and climate provide one of the largest sources of uncertainty in our predictions of global climate. Estimates of the sensitivity of the terrestrial carbon budget to climate anomalies in the tropics and the identification of the mechanisms responsible for feedback effects remain uncertain. The Amazon basin stores a vast amount of carbon, and has experienced increasingly higher temperatures and more frequent floods and droughts over the past two decades. Here we report seasonal and annual carbon balances across the Amazon basin, based on carbon dioxide and carbon monoxide measurements for the anomalously dry and wet years 2010 and 2011, respectively. We find that the Amazon basin lost 0.48 ± 0.18 petagrams of carbon per year (Pg C yr(-1)) during the dry year but was carbon neutral (0.06 ± 0.1 Pg C yr(-1)) during the wet year. Taking into account carbon losses from fire by using carbon monoxide measurements, we derived the basin net biome exchange (that is, the carbon flux between the non-burned forest and the atmosphere) revealing that during the dry year, vegetation was carbon neutral. During the wet year, vegetation was a net carbon sink of 0.25 ± 0.14 Pg C yr(-1), which is roughly consistent with the mean long-term intact-forest biomass sink of 0.39 ± 0.10 Pg C yr(-1) previously estimated from forest censuses. Observations from Amazonian forest plots suggest the suppression of photosynthesis during drought as the primary cause for the 2010 sink neutralization. Overall, our results suggest that moisture has an important role in determining the Amazonian carbon balance. If the recent trend of increasing precipitation extremes persists, the Amazon may become an increasing carbon source as a result of both emissions from fires and the suppression of net biome exchange by drought.

AbstractUnderstanding the relationship between photosynthesis, net primary productivity and growth in forest ecosystems is key to understanding how these ecosystems will respond to global anthropogenic change, yet the linkages among these components are rarely explored in detail. We provide the first comprehensive description of the productivity, respiration and carbon allocation of contrasting lowland Amazonian forests spanning gradients in seasonal water deficit and soil fertility. Using the largest data set assembled to date, ten sites in three countries all studied with a standardized methodology, we find that (i) gross primary productivity (GPP) has a simple relationship with seasonal water deficit, but that (ii) site‐to‐site variations in GPP have little power in explaining site‐to‐site spatial variations in net primary productivity (NPP) or growth because of concomitant changes in carbon use efficiency (CUE), and conversely, the woody growth rate of a tropical forest is a very poor proxy for its productivity. Moreover, (iii) spatial patterns of biomass are much more driven by patterns of residence times (i.e. tree mortality rates) than by spatial variation in productivity or tree growth. Current theory and models of tropical forest carbon cycling under projected scenarios of global atmospheric change can benefit from advancing beyond a focus on GPP. By improving our understanding of poorly understood processes such as CUE, NPP allocation and biomass turnover times, we can provide more complete and mechanistic approaches to linking climate and tropical forest carbon cycling.

Tropical forest leaf albedo (reflectance) greatly impacts how much energy the planet absorbs; however; little is known about how it might be impacted by climate change. Here, we measure leaf traits and leaf albedo at ten 1-ha plots along a 3,200-m elevation gradient in Peru. Leaf mass per area (LMA) decreased with warmer temperatures along the elevation gradient; the distribution of LMA was positively skewed at all sites indicating a shift in LMA towards a warmer climate and future reduced tropical LMA. Reduced LMA was significantly (P < 0.0001) correlated with reduced leaf near-infrared (NIR) albedo; community-weighted mean NIR albedo significantly (P < 0.01) decreased as temperature increased. A potential future 2 °C increase in tropical temperatures could reduce lowland tropical leaf LMA by 6-7 g m-2 (5-6%) and reduce leaf NIR albedo by 0.0015-0.002 units. Reduced NIR albedo means that leaves are darker and absorb more of the Sun's energy. Climate simulations indicate this increased absorbed energy will warm tropical forests more at high CO2 conditions with proportionately more energy going towards heating and less towards evapotranspiration and cloud formation.

Field leaf trait and spectroscopy data – We used leaf trait and spectral data from an extensive field campaign along an elevation gradient (from 3500 m to 220 m elevation) in the Peruvian Amazon where leaf traits for 60-80% of basal area of trees >10cm DBH were measured within a well-studied 1 ha plot network from April – November 2013 (Enquist et al., 2017). In each one ha plot (N=10 plots), we sampled the most abundant species as determined through basal area weighting (enough species generally to cover ~80% of the plot’s basal area). For each species, we sampled the five (three in the lowlands) largest trees (based on diameter at breast height (DBH)) and sampled one sun and one shade branch. On each of these branches, leaf chemistry and leaf mass area (LMA) were measured with the methodology detailed in Asner et al. (2014). On five randomly selected leaves for each branch, we measured hemispherical reflectance with an ASD Fieldspec Handheld 2 with fiber optic cable, a contact probe that has its own calibrated light source, and a leaf clip (Analytical Spectral Devices High-Intensity Contact Probe and Leaf Clip, Boulder, Colorado, USA) following (Doughty et al., 2017). We measured leaf spectroscopy (400-1075 nm) on the same branches where the leaf traits were collected. Both LMA and Chlorophyll A had previously been shown with this dataset to have a correlation with leaf spectroscopy (Doughty et al., 2017). However, we had not previously tried to compare leaf spectral data with DBH directly. Plot data – Aboveground biomass - We used 2,102 of 19,160 total AGB field plots between +30° and -30° latitude classified as broadleaf evergreen trees by MODIS PFT using public data from Duncanson et al 2022 that was organized and publicly available through ORNL DAAC as an RDS (R data serialization) file. Distribution plots are shown in Fig S1 (AGB) and S2 (residuals). NPP and GPP - We also used 21, 1 ha plots where NPP and sometimes GPP were measured following the GEM protocol (Malhi et al., 2021). We focused on two regions: a Peruvian elevation transect with both NPP + GPP (n= 10, RAINFOR plot codes are ALP11, ALP30, SPD02, SPD01, TRU03, TRU08, TRU07, ESP01, WAY01, ACJ01(Malhi et al., 2017)) and a Bornean logging transect with only NPP (n= 11 RAINFOR plot codes are DAN-04, DAN-05, LAM-01, LAM-02, MLA-01, MLA-02, SAF-01, SAF-02, SAF-03, SAF-04, SAF-05 (Riutta et al., 2018). These plots were chosen because there are large changes in NPP/GPP across the elevation or logging gradient. GEDI data – We used the vertical forest structure (L2A and L2B, Version 2) and biomass (L4a) products from the GEDI instrument (R. Dubayah et al., 2020) between April 2019 to December 2022 for tropical forest regions (R. O. Dubayah et al., 2023). We used a quality filtering recipe developed in collaboration with GEDI Science Team members from the University of Maryland and NASA Goddard to identify the highest quality GEDI vegetation shots (R. Dubayah et al., 2022). A data layer that this iterative local outlier detection algorithm uses to exclude data is publicly available at R. O. Dubayah et al., (2023). For instance, some of the key data filters we applied were: included degrade flags of 0,3,8,10,13,18,20,23,28,30,33,38,40,43,48,60,63,68, L2A and L2B quality flags = 1 (only use highest quality data), sensitivity >= 0.98. With the GEDI data, we used canopy height, the height of median energy (HOME), and the number of canopy layers following Doughty et al 2023 (Doughty et al., 2023). Across all tropical forests, we created 300 by 300 m pixels containing all averaged (mean) GEDI data between 2019 and 2022. Using the centroid coordinates from each of the 2,102 plots, we found the 300 by 300 m averaged GEDI pixel that encompassed the plot. If the plot was not encompassed by the GEDI data, we searched a wider area by incrementally averaging a gradually increasing area of 1, 3, 5, and 10 pixels. In other words, if no 300 by 300 m pixel encompassed the plot, then we averaged all GEDI data an area one pixel out (4 by 4 = 1200 by 1200 m, 6 by 6 = 1800 by 1800m, 11 by 11 = 3300m by 3300m), gradually increasing the square until it encompassed an area with GEDI data. To compare with the NPP/GPP plots we compared RS trait and GEDI data for individual footprints within a 0.03 km radius of the plot coordinates. Remotely sensed leaf trait data – Based on a broader set of field campaigns, Aguirre-Gutiérrez et al., (2021) used Sentinel-2, climatic, topography, and soil data to create remotely sensed canopy trait maps for P=phosphorus % leaf concentration, WD = wood density g.cm-3, and LMA=Leaf mass area g m-2. Other data layers – We compared % one peak to several other climates, soils, leaf traits, and ecoregion maps listed below for the Amazon basin. Each dataset had its own resolution, which we standardized to 0.1 by 0.1 degrees. We used total cation exchange capacity (CEC) from soil grids (Batjes et al., 2020) from 0-5cm in units of mmol(c)/kg. We averaged TerraClimate (Abatzoglou et al., 2018) data between 2000 and 2018 for Vapor Pressure Deficit (VPD in kPa), Mean Monthly Precipitation (MMP) (mm/month), potential evapotranspiration (PET) and maximum and minimum temperature (°C). Statistical analysis – We used the Matlab (Matlab, MathWorks Inc., Natick, MA, USA) function “fitlm” to fit linear models to compare variables such as soil data, environmental data, leaf trait data (at 0.1° resolution) and GEDI structure data (300m and bigger resolution) to field biomass and NPP/GPP estimates. The P values listed are for the t-statistic of the two-sided hypothesis test. We used R to create a linear model to predict the best model ranked by AIC and parsimony using the dredge function from the MuMIn library (Bartoń, 2009). We also used the CAR package (Fox J & S, 2019) and the VIF command to test for multi-collinearity between variables. To account for spatial autocorrelation, we used Simultaneous Auto-Regressive (SARerr) models (F. Dormann et al., 2007) using the R library ‘spdep’ (Bivand, Hauke, & Kossowski, 2013). We tested different neighborhood distances from 10 km to 300 km and found that AIC was minimized at 80 km (Fig S3) and the corresponding correlogram showed reduced spatial autocorrelation (Fig S4). To predict leaf traits with the spectral information, we used the Partial Least Squares Regression (PLSR) (Geladi & Kowalski, 1986) using the PLSregress command in Matlab (Matlab, MathWorks Inc., Natick, MA, USA). To avoid over-fitting the number of latent factors, we minimized the mean square error with K-fold cross-validation. We use 70% of our data to calibrate our model and then the remaining 30% to test the accuracy of our model using r2. We use adjusted r2 which penalizes for small sample sizes throughout the manuscript. # Satellite-derived trait data slightly improves tropical forest biomass, NPP, and GPP predictions [https://doi.org/10.5061/dryad.ttdz08m5n](https://doi.org/10.5061/dryad.ttdz08m5n) The dataset contains leaf trait and spectral data to create Figures 1 and 2. It contains plot biomass data and satellite-derived leaf trait and structure data to create Figures 3-6. It contains plot NPP, GPP, and satellite-derived leaf trait and structure data to create Figures 7-8. ## Description of the data and file structure, including the associated Code/Software The Matlab code Finalcode_GEDIbiomass_Doughty2024.m contains all the code and data to create all the figures in the paper. The code has several sections that can be run independently. The first section starting on line 1 uses the dataset traitcompare.mat to create Figure 1. This dataset contains two tables of leaf trait data called carnegiechem and merged. Units and column names are contained within the table. The second section starting on line 62 uses the dataset traitgedidat.mat to create Figures 3-5 and Figures S1 and S2. This dataset contains a table called agball with plot biomass and coordinates. It also has the trait and GEDI data for these plots in nested structures. Units and descriptions are given in the code. The third section starting on line 443 uses the datasets traitgedidat.mat and soilclimdata.mat to create Figure 6. The dataset traitgedidat.mat is the same as described above and soilclimdata.mat contains 0.1 by 0.1 degree gridded data for climate variables like Tmax (C) or VPD (Pa) or soil chemistry like CEC. Units and description are given in the code. The forth section starting on line 536 uses the dataset Tamtreeheight.mat to create Figures 7 and 8. The dataset gedivsplot.mat contains table data with plot data, and nearby trait and GEDI data for several GEM plots. Units and description are given in the code and the tables. The fifth section starting on line 791 uses the dataset CombspecDBH.mat to create Figure 2. The dataset has variables specallz which is the leaf spectral data from 350-1075 nm for each leaf and dbhz1 with is the corresponding tree dbh (cm). It also has LMAz which is the LMA data with datz as the corresponding spectral data. To estimate spatial autocorrelation and the best model by AIC to create Table 1 and Figures S3 and 4, we used the R code processgedidata.r and the dataset biomass_trait_GEDI.xlsx. This dataset contains a table with latitude, longitude, field biomass and remote sensed biomass (Mg Ha-1), and traits LMA (g m2), Phosphorus (%), tree height (m), HOME (m) and % one peak (unitless). ## Sharing/Access information Original GEDI data are available from the USGS. Improving tropical forest biomass predictions can accurately value tropical forests for their ecosystem services. Recently, the Global Ecosystem Dynamics Investigation (GEDI) lidar was activated on the international space station (ISS) to improve biomass predictions by providing detailed 3D forest structure and height data. However, there is still debate on how best to predict tropical forest biomass using GEDI data. Here we compare GEDI predicted biomass to 2,102 tropical forest biomass plots and find that adding a remotely sensed (RS) trait map of LMA (Leaf Mass per Area) significantly (P<0.001) improves field biomass predictions, but by only a small amount (r2=0.01). However, it may also help reduce the bias of the residuals because, for instance, there was a negative relationship between both LMA (r2 of 0.34) and %P (r2=0.31) and residuals. This improvement in predictability corresponds with measurements from 523 individual trees where LMA predicts Diameter at Breast height (DBH) (the critical measurement underlying plot biomass) with an r2=0.04, and spectroscopy (400-1075 nm) predicts DBH with an r2=0.01. Adding environmental datasets may offer further improvements and max temperature (Tmax) predicts Amazonian biomass residuals with an r2 of 0.76 (N=66). Finally, for a network of net primary production (NPP) and gross primary production (GPP) plots (N=21), RS traits are better at predicting fluxes than structure variables like tree height or Height Of Median Energy (HOME). Overall, trait maps, especially future improved ones produced by surface biology geology (SBG), may improve biomass and carbon flux predictions by a small but significant amount.

Field Data - We estimate canopy temperature at the km 83 eddy covariance tower in the Tapajos region of Brazil 1–3 using a pyrgeometer (Kipp and Zonen, Delft, Netherlands) mounted at 64 m to measure upwelling longwave radiation (L↑ in W m-2) with an estimated radiative-flux footprint of 8,000 m2 4. Data were collected every 2 seconds and averaged over 30-minute intervals between August 2001 and March 2004. We estimated canopy temperature with the following equation: Eq 1 – Canopy temperature (°C) = (L↑/(E*5.67e-8))0.25-273.15 We chose an emissivity value (E) of 0.98 for the tower data, as this was the most common value used in the ECOSTRESS data (SDS_Emis1-5 (ECO2LSTE.001) and the broader literature for tropical forests 5. We compared canopy temperature derived from the pyrgeometer to eddy covariance derived latent heat fluxes (flux footprint ~1 km2), air temperature at 40 m, which is the approximate canopy height (model 076B, Met One, Oregon, USA; and model 107, Campbell Scientific, Logan, Utah, USA) and soil moisture at depths of 40 cm (model CS615, Campbell Scientific, Logan, Utah, USA). Further details on instrumentation and eddy covariance processing can be found in 1,3. This site was selectively logged, which had a minor overall impact on the forest 6, but did not affect any trees near the tower. Leaf thermocouple data - We measured canopy leaf temperature at a 30 m canopy walk-up tower between July to December of 2004 and July to December of 2005 at the same site. We initially placed 50 thermocouples on canopy-exposed leaves of Sextonia rubra, Micropholis sp., Lecythis lurida) (originally published in Doughty and Goulden 2008). Fine wire thermocouples (copper constantan 0.005 Omega, Stamford, CT) were attached to the underside of leaves by threading the wire through the leaf and inserting the end of the thermocouple into the abaxial surface. The thermocouples were wired into a multiplexer attached to a data logger (models AM25T and 23X, Campbell Scientific, Logan, UT, USA) and the data were recorded at 1 Hz. Additional upper-canopy leaf thermocouple data from Brazil7, Puerto Rico8, Panama9, Atlantic forest Brazil10 and Australia 11, were generally collected in a similar manner. Satellite data - ECOSTRESS data (ECO2LSTE.001) – The ECOsystem Spaceborne Thermal Radiometer Experiment on Space Station (ECOSTRESS) mission is a thermal infrared (TIR) multispectral scanner with five spectral bands at 8.28, 8.63, 9.07, 10.6, and 12.05 µm. The sensor has a native spatial resolution of 38 m x 68 m, resampled to 70 m x 70 m, and a swath width of 402 km (53°). Data are collected from an average altitude of 400 ± 25 km on the International Space Station (ISS). ECOSTRESS is an improvement over other thermal sensors because no other sensors provide TIR data with sufficient spatial, temporal, and spectral resolution to reliably estimate LST at the local-to-global scale for a diurnal cycle 12. To ensure the highest quality data, we used ECOSTRESS quality flag 3520, which identifies the best quality pixels (no cloud detected), a minimum-maximum difference (MMD) indicative of vegetation or water (Kealy and Hook 1993), and nominal atmospheric opacity. We accessed ECOSTRESS LST data through the AppEEARS website (https://lpdaac.usgs.gov/tools/appeears/) for the following products and periods: SDS_LST (ECO2LSTE.001) from a long longitudinal swath of the Amazon for 25 December 2018 to 20 July 2020 (SI Fig 1a red box) and then a larger area of the western Amazon for 18 September to 29 September 2019 (SI Fig 1a green box), Central Africa for 1 August to 30 August 2019 (SI Fig 1b), and SE Asia for 15 January to 30 February 2020 (SI Fig. 1c). The dates were chosen as all ECOSTRESS data available at the start of the study for the smaller regions and for warm periods with low soil moisture for the larger areas. We calculated “peak median,” which is defined as the average of the highest three medians of each granule (i.e., for the Amazon SI Fig. 1a, there were 934 granules) for each hour period. Comparison of LST data – We compared ECOSTRESS LST to VIIRS LST (VNP21A1D.001) and MODIS LST (MYD11A1.006). A more detailed comparison and description of these sensors can be found in Hulley et al 202113. Details for the sensors and quality flags used are given in Table S1. Broadly, G1 for ECOSTRESS and VIIRS is classified as vegetation (using emissivity) and of medium quality. G2 is classified as vegetation, but of the highest quality. MODIS landcover classifies this region as almost entirely broadleaf evergreen vegetation, but using MMD (emissivity) only 18% (VIIRS) and 12% (ECOSTRESS) of the data are classified as vegetation, rather than as soils and rocks (Table S2). Therefore, we use the vegetation classification (from MMD) as a very conservative estimate of complete forest canopy cover and not farms, urban, or degraded forest where rocks or soils are more likely to appear to satellites. SMAP data – To estimate pantropical soil moisture, we use the Soil Moisture Active Passive (SMAP) sensor and the product Geophysical_Data_sm_rootzone (SPL4SMGP.005). SMAP measurements provide remote sensing of soil moisture in the top 5 cm of the soil 14 and the L4 products combine SMAP observations and complementary information from a variety of sources. We accessed SMAP data from the AppEEARS website for the following products and periods: Amazon for 25 December 2018 to 20 July 2020 (SI Fig 1a), Central Africa for 25 December 2019 to 20 July 2020 (SI Fig 1b), and Borneo for 25 December 2018 to 20 July 2020 (SI Fig 1c). Warming experiments – For model validation, we used the results of three upper-canopy leaf and branch warming experiments of 2°C (Brazil), 3°C (Puerto Rico), and 4°C (Australia). The first experiment (Brazil), was 4 individual leaf-resistant heaters on each of 6 different upper-canopy species at the Floresta National (FLONA) do Tapajos as part of the Large-Scale Biosphere–Atmosphere Ecology Program (LBA-ECO) in Santarem, Brazil14. On the same six species, black plastic passively heated branches by an average ~2°C. Initially, heat balance sap flow sensors and the passive heaters were added to 40 branches, but we had confidence in the data from 9 heated and 4 control in the final analysis. The second experiment (Puerto Rico) had two species (Ocotea sintenisii (Mez) Alain and Guarea guidonia (L.) Sleumer where leaves were heated by 3 °C at the Tropical Responses to Altered Climate Experiment (TRACE) canopy tower site at Sabana Field Research Station, Luquillo, Puerto Rico8. The final experiment (Australia), which increased leaf temperatures by 4 °C, was conducted at Daintree Rainforest Observatory (DRO) in Cape Tribulation, Far North Queensland, Australia. Leaf heaters were installed using a pair of 30-gauge copper-constantan thermocouples, one reference leaf, and one heated with a target temperature differential of 4°C. There were two pairs in the upper canopy of each tree crown installed in 2–3 individuals across four species with the thermocouples installed on the underside of the leaves. Two absolute 36-gauge copper-constantan thermocouples were installed in each species to measure the leaf temperatures of the reference leaves. Thermocouple wires connected into an AM25T multiplexer from Campbell Scientific connected to a CR1000 Campbell datalogger. More details about the experiment and sensors can be found in 11. Model – We created a model of individual leaves on a tree (100 by 100 grid where each leaf is a pixel) to estimate the upper limit of tropical canopy temperatures with projected changes in climate. At the start of the simulation, we randomly applied the measured distribution (ambient Fig 1c) of canopy leaf temperatures >31.2 °C (chosen to give a mean canopy temperature of 33.2 ± 0.4 °C, matching the canopy average Fig 1b) to the entire grid. Each year we increased the mean air temperatures by 0.03°C to simulate a warming planet. As air temperatures reached +2, 3, and 4°C, we applied the leaf temperature distributions (but subtracted out the air temperature increases) from the different warming experiments (+2°C (Brazil), +3°C (Puerto Rico), and +4°C (Australia), respectively (Fig S7)). We ran the model at a daily time step with leaves flushing once a year (all dead leaves reset to living each year). In addition, to take into account the effect of climate inter-annual variation - specifically drought, these mean canopy temperatures were further increased or decreased by deviations from mean maximum air temperatures at 40 m pulled each day from the Tapajos eddy covariance tower1–3 and soil moisture at 40 cm depth (m3 m-3) which controlled canopy temperatures following equation 2 (Fig S6). Eq 2 – Canopy temperature (°C) = 46.5-33.6*soil moisture (m3 m-3) For example, in a non-drought year, on a day when max air temperatures were 0.1 °C higher than average and soil moisture was 0.01 m3 m-3 lower than average (which would add 0.3 °C to canopy temperatures (Eq 2)), we would add 0.4 °C to the grid canopy temperature that day. Every year, there was a 10% random probability of either a minor (80% probability) drought which reduced soil moisture by 0.1 m3 m-3 and increased air temperatures by 0.5 °C or severe drought (20% probability), which reduced soil moisture by 0.2 m3 m-3 and increased air temperatures by 1 °C. This is similar to the Amazon-wide temperature increases during the last El Niño 15. If an individual leaf temperature increases to above 46.7 °C (Tcrit) the leaf died, following Slot et al. (2021). Prior research has suggested that irreversible damage could begin at 45 °C 16 and T50 for tropical species is 49.9 °C 17, and we use these values in a sensitivity study. We further explore the impact of duration of Tcrit on mortality in a sensitivity study (ranging between needing a single exposure to four exposures to Tcrit to die). Over the season, if a leaf died, then it did not contribute towards canopy evapotranspiration. We ran simulations as a 3D canopy with an LAI of 5 where if the top leaf died, then it was replaced by a shade-adapted leaf with a Tcrit 1 °C lower 18. If each of the 5 LAIs died, then all leaves in that grid cell were dead and canopy evaporative cooling decreased by that percentage. Several lines of evidence suggest that under normal hydraulic conditions, when radiation load increases from ~350 to 1100 W m-2 (e.g. between shady and sunny conditions) average canopy temperature increases by ~3 °C and therefore, evaporative cooling for a full 1100 W m-2 is ~4.4°C4,19 (we vary this in a sensitivity study between 3.7 and 5.1°C). For example, if, over a year, 1000 leaves (10% of all leaves) surpass Tcrit and die, evaporative cooling for all leaves in the grid will be reduced by 10% (1000/(100 by 100 grid)) or 0.44 °C and 0.44 °C will be added to mean canopy temperature. Therefore, mean canopy temperature could heat up by a maximum of 4.4°C either due to a reduction of soil moisture or from an increase in dead leaves. We ran each simulation until the point where all leaves were dead and repeated this 30 times. We assumed loss of tree function following the death of all leaves, but we discuss this further in the discussion. We then ran sensitivity studies for several of the key variables (bold indicates the standard model parameter) including: drought (0.05, 0.1, to 0.2 m3 m-3 decrease in soil moisture), change in Tcrit (Tcrit: 45, 46.7, 49.9 °C), Tcrit range (100 by 100 grid =random distribution of 46.7±2, 100 by 100 grid =46.7±0), Max evaporative cooling (3.7, 4.4°C), (Tcrit duration (exceed Tcrit once, exceed Tcrit more than 3 times) and soil moisture coefficient (-33.6 -38.2; i.e. change the slope from Fig S6 by ± 1 sd). Methods References Miller, S. D. et al. Biometric and micrometeorological measurements of tropical forest carbon balance. Ecol. Appl. 14, 114–126 (2004). da Rocha, H. R. et al. Seasonality of water and heat fluxes over a tropical forest in eastern Amazonia. Ecol. Appl. 14, 22–32 (2004). Goulden, M. L. et al. Diel and seasonal patterns of tropical forest co2 exchange. Ecol. Appl. 14, 42–54 (2004). Kivalov, S. N. & Fitzjarrald, D. R. Observing the Whole-Canopy Short-Term Dynamic Response to Natural Step Changes in Incident Light: Characteristics of Tropical and Temperate Forests. Boundary-Layer Meteorol. 173, 1–52 (2019). Jin, M. & Liang, S. An Improved Land Surface Emissivity Parameter for Land Surface Models Using Global Remote Sensing Observations. J. Clim. 19, (2006). Miller, S. D. et al. Reduced impact logging minimally alters tropical rainforest carbon and energy exchange. Proc. Natl. Acad. Sci. 108, 19431 LP – 19435 (2011). Doughty, C. E. An In Situ Leaf and Branch Warming Experiment in the Amazon. Biotropica 43, 658–665 (2011). Carter, K. R., Wood, T. E., Reed, S. C., Butts, K. M. & Cavaleri, M. A. Experimental warming across a tropical forest canopy height gradient reveals minimal photosynthetic and respiratory acclimation. Plant. Cell Environ. 44, 2879–2897 (2021). Rey-Sanchez, A. C., Slot, M., Posada, J. & Kitajima, K. Spatial and seasonal variation of leaf temperature within the canopy of a tropical forest. Clim. Res. 71, 75–89 (2016). Fauset, S. et al. Differences in leaf thermoregulation and water use strategies between three co-occurring Atlantic forest tree species. Plant. Cell Environ. 41, 1618–1631 (2018). Crous K Y, A W Cheesman, K Middleby, Rogers Eie, A Wujeska-Klause, A Y M Bouet, D S Ellsworth, M J Liddell, L A Cernusak, C V M Barton, Similar patterns of leaf temperatures and thermal acclimation to warming in temperate and tropical tree canopies., Tree Physiology, 2023;, tpad054, https://doi.org/10.1093/treephys/tpad054. Xiao, J., Fisher, J. B., Hashimoto, H., Ichii, K. & Parazoo, N. C. Emerging satellite observations for diurnal cycling of ecosystem processes. Nat. Plants 7, 877–887 (2021). Hulley, G. C. et al. Validation and Quality Assessment of the ECOSTRESS Level-2 Land Surface Temperature and Emissivity Product. IEEE Trans. Geosci. Remote Sens. 60, 1–23 (2022). Reichle, R., Lannoy, G. De, Koster, R. D., Crow, W. T. & 2017., J. S. K. SMAP L4 9 km EASE-Grid Surface and Root Zone Soil Moisture Geophysical Data, Version 3. Boulder, Color. USA. NASA Natl. Snow Ice Data Cent. Distrib. Act. Arch. Center. doi https//doi.org/10.5067/B59DT1D5UMB4. (2017). Jiménez-Muñoz, J. C. et al. Record-breaking warming and extreme drought in the Amazon rainforest during the course of El Niño 2015–2016. Sci. Rep. 6, 33130 (2016). Berry, J. & Bjorkman, O. Photosynthetic Response and Adaptation to Temperature in Higher Plants. Annu. Rev. Plant Physiol. 31, 491–543 (1980). Slot, M. et al. Leaf heat tolerance of 147 tropical forest species varies with elevation and leaf functional traits, but not with phylogeny. Plant. Cell Environ. 44, (2021). Slot, M., Krause, G. H., Krause, B., Hernández, G. G. & Winter, K. Photosynthetic heat tolerance of shade and sun leaves of three tropical tree species. Photosynth. Res. 141, 119–130 (2019). Doughty, C. E. & Goulden, M. L. Are tropical forests near a high temperature threshold? J. Geophys. Res. Biogeosciences (2009) doi:10.1029/2007JG000632. The critical temperature beyond which photosynthetic machinery in tropical trees begins to fail averages ~46.7°C (Tcrit) 1. However, it remains unclear whether leaf temperatures experienced by tropical vegetation approach this threshold or soon will under climate change. We found that pantropical canopy temperatures independently triangulated from individual leaf thermocouples, pyrgeometers, and remote sensing (ECOSTRESS) have midday-peak temperatures of ~34°C during dry periods, with a long high-temperature tail that can exceed 40°C. Leaf thermocouple data from multiple sites across the tropics suggest that even within pixels of moderate temperatures, upper-canopy leaves exceed Tcrit 0.01% of the time. Further, upper-canopy leaf warming experiments (+2, 3, and 4°C in Brazil, Puerto Rico, and Australia) increased leaf temperatures non-linearly with peak leaf temperatures exceeding Tcrit 1.3% of the time (11% >43.5°C, 0.3% >49.9°C). Using an empirical model incorporating these dynamics (validated with warming experiment data), we found that tropical forests can withstand up to a 3.9 ± 0.5 °C increase in air temperatures before a potential collapse in metabolic function, but the remaining uncertainty in our understanding of Tcrit could reduce this to 2.6 ± 0.6°C. The 4.0°C estimate is within the “worst case scenario” (RCP-8.5) of climate change predictions2 for tropical forests and therefore it is still within our power to decide (e.g., by not taking the RCP 8.5 route) the fate of these critical realms of carbon, water, and biodiversity 3,4.