ARTICLE https://doi.org/10.1038/s41467-021-22378-8 OPEN Non-structural carbohydrates mediate seasonal water stress across Amazon forests Caroline Signori-Müller 1,2 ✉, Rafael S. Oliveira 3, Fernanda de Vasconcellos Barros4,5, Julia Valentim Tavares 2, Martin Gilpin2, Francisco Carvalho Diniz2, Manuel J. Marca Zevallos6,7, Carlos A. Salas Yupayccana7, Martin Acosta8, Jean Bacca6, Rudi S. Cruz Chino 7, Gina M. Aramayo Cuellar9, Edwin R. M. Cumapa6, Franklin Martinez9, Flor M. Pérez Mullisaca 6, Alex Nina 7, Jesus M. Bañon Sanchez6, Leticia Fernandes da Silva 8, Ligia Tello6, José Sanchez Tintaya6, Maira T. Martinez Ugarteche9, 1234567890():,; Timothy R. Baker2, Paulo R. L. Bittencourt4,5, Laura S. Borma 10, Mauro Brum11,5, Wendeson Castro 8, Eurídice N. Honorio Coronado 12, Eric G. Cosio13, Ted R. Feldpausch 4, Letícia d’Agosto Miguel Fonseca10, Emanuel Gloor2, Gerardo Flores Llampazo14, Yadvinder Malhi 15, Abel Monteagudo Mendoza6, Victor Chama Moscoso6, Alejandro Araujo-Murakami9, Oliver L. Phillips 2, Norma Salinas15,13, Marcos Silveira 8, Joey Talbot16, Rodolfo Vasquez17, Maurizio Mencuccini18,19 & David Galbraith 2 Non-structural carbohydrates (NSC) are major substrates for plant metabolism and have been implicated in mediating drought-induced tree mortality. Despite their significance, NSC dynamics in tropical forests remain little studied. We present leaf and branch NSC data for 82 Amazon canopy tree species in six sites spanning a broad precipitation gradient. During the wet season, total NSC (NSCT) concentrations in both organs were remarkably similar across communities. However, NSCT and its soluble sugar (SS) and starch components varied much more across sites during the dry season. Notably, the proportion of leaf NSCT in the form of SS (SS:NSCT) increased greatly in the dry season in almost all species in the driest sites, implying an important role of SS in mediating water stress in these sites. This adjustment of leaf NSC balance was not observed in tree species less-adapted to water deficit, even under exceptionally dry conditions. Thus, leaf carbon metabolism may help to explain floristic sorting across water availability gradients in Amazonia and enable better prediction of forest responses to future climate change. A full list of author affiliations appears at the end of the paper. NATURE COMMUNICATIONS | (2021)12:2310 | https://doi.org/10.1038/s41467-021-22378-8 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-22378-8 P lants rely on both newly assimilated carbon and stored ≤100 mm month−1 (Supplementary Fig. 1). We focus not only on reserves of non-structural carbohydrates (NSC) for growth total NSC concentrations (NSCT) but also on the partitioning of and other physiological functions such as respiration, NSC into its two major components: soluble sugars (i.e., oligo- osmotic regulation and defence1,2. As NSC stores reflect the saccharides such as glucose, sucrose, fructose, etc.) and starch, as balance of carbon supply via photosynthesis and its utilisation for these fulfil distinct roles in plants26. Soluble sugars (SS) provide an plant metabolism, they are highly dynamic in time3. NSC stored immediate energy substrate for respiration, defence, plant stress during periods when supply exceeds demand are thought to signalling, phloem transport and osmoregulation1. Starch repre- constitute an important buffer during periods of environmental sents a transient or long-term energy store that plants can convert stress when carbon demand outstrips supply2,4. As a result of this, to SS for use when C demand exceeds supply29. To characterise considerable attention has been paid to the potential role of plant water status at the time of NSC sampling, we measure stored NSC in mediating tree tolerance and survival under midday leaf water potential (ΨMD) in all sites during the dry drought5,6, during which stomatal conductance and assimilation season, and in the two sites without a climatological dry season rates are reduced to prevent water loss7. Experimental studies on (Alp and Suc). Community-level (mean value of all species in each both temperate8 and tropical9 seedlings have indicated an site) ΨMD ranges from −0.62 ± 0.05 MPa (mean ± SE) in the ever- important role of NSC in the physiological mechanism of mor- wet Suc and Alp sites to −2.18 ± 0.30 MPa in the ecotonal Ken site tality, as plants with higher NSC content had higher survivorship with the longest dry season. under drought. However, the extent to which NSC metabolism We use this multi-site dataset to gain insights into how NSC and moderates tolerance to water deficit in adult trees over large water deficit responses are related across Amazonian forests. We geographical domains remains unclear10. evaluate how leaf and branch NSC vary with water availability, both An understanding of the functional role of NSC in response to in space and seasonally, and also the relative roles of taxonomy and water deficit is of particular importance for the Amazon rain- environment in determining NSC concentrations. Based on eco- forest, the Earth’s largest tropical forest region and a major ter- system modelling results16 and observations suggesting prioritisa- restrial carbon sink, responsible for absorbing 5–10% of global tion of aboveground growth under drought in Amazonia4, we anthropogenic CO2 emissions11. Over the last 15 years, the hypothesise that the drier sites would experience more seasonality Amazon has been subject to three large-scale drought events12–14 in NSC stores and would also have greater NSC stores in the wet and climate models project an intensification of drought risk over season. Our analyses reveal an important role of non-structural large parts of the Basin in the future15. Recent modelling results carbohydrates, and soluble sugars in particular, in mediating suggest that NSC play an important role in regulating the impacts responses to seasonal water stress in Amazonian forests (Abstract of drought on carbon fluxes in the Amazon by maintaining available in Portuguese and Spanish, Supplementary Notes 1, 2). growth under water deficit16. Moreover, observations of sustained stand-scale net primary productivity during the 2010 drought have also led to suggestions that Amazon trees deplete their NSC Results and discussion reserves during periods of water stress to prioritise growth4. Wet season (baseline) NSC: role of environment vs. taxonomy. However, empirical studies of NSC dynamics in tropical forests Despite the wide range of species sampled and the differences in are rare, being limited to a small number of sites and species17–19. species composition across our study sites, we found little variation While a study in Panama found that NSC concentrations in leaf and branch NSCT and its components across sites in the wet increased in the dry season18, the only detailed community-level season (monthly precipitation > 100 mm month−1). In fact, wet study in lowland Amazon forests, in a throughfall exclusion season NSCT and SS in both leaves and branches did not differ experiment in Eastern Amazonia, found that NSC reserves in significantly across sites (Fig. 2a, e; Supplementary Fig. 2; Supple- trees subjected to long-term drought did not differ from those in mentary Table 2). Overall, our results support the vegetation unstressed trees20. Yet Amazon forests vary greatly in climate21, modelling assumption of spatially invariant baseline community- soils22 and plant life history strategies23–25, of which could level NSCT16, thus potentially simplifying modelling of NSC potentially influence NSC dynamics and forest response to cli- dynamics in Amazonia. Only leaf starch exhibited significant dif- mate change1. The scarcity of empirical data in Amazonia ferences across sites in the wet period, being markedly lower in the impedes understanding of the significance of NSC in modulating moderately seasonal sites (Man and Tam; p = 0.001, Supplementary forest responses to water stress and thus limits current vegetation Fig. 2) than in the driest Ken site and the two wettest sites (Suc and model development efforts to simulate drought impacts on tro- Alp). The higher wet season leaf starch concentrations in these sites pical forests16,26. may represent important strategies for maintaining function under To address this significant data gap, we conduct a large-scale an extended period of depleted water availability29 in the case of sampling of NSC across Amazon forests, using fully standardised Ken or light limitation30 in the case of Suc and Alp. field and laboratory protocols, performing all NSC analyses in the We find that nested family–genus–species identity is very same lab (see Methods). Such standardisation is critical as differ- important, explaining much more of the overall variation in wet ences in sampling and laboratory extraction protocols can yield season NSCT and SS than sampling site for both leaves and substantial variation in NSC estimates, obstructing meta-analysis branches (Fig. 2b, f). Taxonomy was a particularly important and comparisons across studies27,28. We analyse the concentration control for leaf NSC, explaining 67% and 72% of the wet season of NSC in leaves and branches of 82 canopy tree species in six sites variation in leaf NSCT and SS (Fig. 2b). Within individual sites, across the Amazon Basin (Fig. 1; Table 1; Supplementary Table 1) there were clear species-level differences in NSCT and SS, which effectively spanning the entire Amazon gradient in mean annual varied by a factor of between 4 and 10 across species precipitation (1167–3155 mm year−1; Fig. 1) and seasonality (Supplementary Table 3, Supplementary Fig. 3 & 4). Moreover, (0–7 months with rainfall ≤100 mm month−1) and including one when species occurred across more than one site, they largely site (Man) that experienced an atypically strong drought event14. maintained similar wet season levels of NSC and its fractions We collect the plant material for NSC analyses in all sites during across sites (Supplementary Figs. 5–8). Despite the large amount wet months (precipitation >100 mm month−1), hereafter referred of variance explained by taxonomy, NSCT and its constituent to as the wet season. For the four sites with more marked sea- fractions are generally not related to plant traits that are sonality (Ken, Fec, Man and Tam), we also collect plant material indicative of life history strategies25, such as potential tree size, in the peak of the dry season, where monthly precipitation is mean growth, mortality rates and wood density19 (Supplementary 2 NATURE COMMUNICATIONS | (2021)12:2310 | https://doi.org/10.1038/s41467-021-22378-8 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-22378-8 ARTICLE 400 Suc Monthly precipitation (mm) 300 200 100 0 May Sep Dec Nov Mar Aug Feb Jan Jun Apr Oct Jul 400 Alp Monthly precipitation (mm) 300 200 100 0 May Sep Dec Nov Mar Aug Feb Jan Jun Apr Oct Jul 400 Man Monthly precipitation (mm) 300 200 100 0 400 Tam 400 Fec 400 Ken Monthly precipitation (mm) Monthly precipitation (mm) Monthly precipitation (mm) May Sep Dec Nov Mar Aug Feb Jan Jun Apr Oct Jul 300 300 300 200 200 200 100 100 100 0 0 0 May May May Sep Dec Sep Dec Sep Dec Nov Nov Nov Mar Aug Mar Aug Mar Aug Feb Feb Feb Jan Jun Jan Jun Jan Jun Apr Oct Apr Oct Apr Oct Jul Jul Jul Fig. 1 Location of sampled sites and monthly precipitation. The map depicts mean annual precipitation in South America (mm year−1). The Amazon basin is bounded by the black outline. Bar plots show the monthly precipitation for each site, blue represents precipitation >100 mm month−1 and red ≤100 mm month−1. Data for the map and bar plots are from WorldClim v2 (1970–2010, 30 s resolution)51. Table 1 Site information. Site name Short code Site location MAP DSLb ΨMD (MPa)c n Collection period (mm)a (months) Kenia Ken Ascensión de Guarayos, Santa 1167 7 −2.18 ± 0.81 ~47% b.ad Dry: Jun 2017 Cruz, Bolivia 9 species Wet: Mar 2017 (16°1’S, 62°43’W) 25 trees Catuaba Experimental Farm Fec Senador Guiomard, 1658 4 −2.05 ± 0.51 ~40% b.a Dry: Jul 2017 Acre, Brazil 14 species Wet: Apr 2017 (10°4’S, 67°37’W) 40 trees Cuieras Biological Reserve Man Manaus, Amazonas, Brazil 2420 1 −1.58 ± 0.40 ~12% b.a Dry: Oct 2015 (2°36’S, 60°12’W) 13 species Wet: Jun 2016 34 trees Tambopata National Reserve Tam Puerto Maldonado, Madre de 2451 3 −1.10 ± 0.36 ~40% b.a Dry: Jul 2017 Dios, Peru 21 species Wet: Feb 2017 (12°49’S, 69°16’W) 58 trees Allpahuayo Mishana Alp Iquitos, Maynas, Peru 2660 0 −0.64 ± 0.26 ~26% b.a Oct 2017 National Reserve (3°56’S, 73°25’W) 27 species 85 trees Sucusari Ecological Reserve Suc Iquitos, Maynas, Peru 3155 0 −0.62 ± 0.33 ~21% b.a Nov 2017 (3°15’S, 72°54’W) 31 species 93 trees aMAP mean annual precipitation (mm) from WorldClim Bioclimatic variables version 2, 30-s resolution51. bDSL dry season length, number of months with precipitation ≤100 mm, data extracted from WorldClim Version 2, 30 s resolution51. cΨMD = midday community leaf water potential in the sampled day during dry season ± SD (except in the sites where there is no meteorological dry season, ALP and SUC). db.a. = basal area, referring to the sampled percentage. NATURE COMMUNICATIONS | (2021)12:2310 | https://doi.org/10.1038/s41467-021-22378-8 | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-22378-8 Fig. 2 Species-mean total NSC (NSCT) and variance partitioning into site and taxonomic components. Concentrations of NSCT and variance partitioning results are displayed separately for leaves (panels a–d) and branches (panels e–h) during the wet and dry seasons. Left hand panels: Each box encompasses the 25th to 75th percentiles; the median is indicated by the horizontal line with each box while external horizontal lines indicate the 10th and 90th percentiles; dots indicate outliers. Sites are ordered and colour-coded from left to right from driest to wettest; red to yellow boxes represent the seasonal sites and two blue boxes the aseasonal sites; n indicates the number of species sampled in each site. Differences among sites were tested using Kruskal–Wallis. Sites with different letters are statistically distinguishable (p < 0.05, post hoc Mann–Whitney–Wilcoxon Rank Sum test using Bonferroni correction is indicated by small letters). Right hand panels: Partitioning of total variance of NSCT, starch and soluble sugars into genetic (family/genus/ species), environmental (site) and error (residual) components; for the variance partitioning analysis, values were log1p-transformed prior to analysis. Figs. 9 and 10). Although our data on plant phenological osmoregulatory and defence demands than other plant organs32. strategies were limited, we do find that evergreen species have Thus, differences across taxa in photosynthetic and respiration higher total leaf NSCT (p = 0.023) and SS (p = 0.012) in the wet rates as well as osmoregulatory and defence mechanisms, season than semideciduous/deciduous species (Supplementary although little studied, may help to explain the strong taxonomic Figs. 11 and 12; Supplementary Table 4). The strong taxonomic signatures we find. influence on leaf NSCT and SS likely further relates to physiological attributes for which we have little current informa- tion. Leaves have intense metabolic requirements due to their NSC seasonality. NSCT exhibited greater differences across sites roles in photosynthesis and phloem loading31 and face greater in the dry season compared to the wet season (Fig. 2c, g). This 4 NATURE COMMUNICATIONS | (2021)12:2310 | https://doi.org/10.1038/s41467-021-22378-8 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-22378-8 ARTICLE was especially the case for leaf NSC, where NSCT, SS and starch site (Supplementary Fig. 15)33. In the two less water-limited sites exhibited significant differences across sites, while in branches, for which we had data in both seasons (Man and Tam), we only SS varied across sites (Fig. 2 and Supplementary Fig. 2). In interpret the seasonal dynamics of NSCT and its fractions to be line with this, our variance partitioning analysis also showed that driven mainly by growth. In the Man site, productivity is site accounted for more of the overall variation in NSC metrics in maintained at high levels during the dry season34 (Supplementary the dry season than it did in the wet season (Fig. 2d, h). Fig. 5), and the observed depletion of branch starch (60% Seasonal patterns of NSC across sites diverged markedly (Fig. 3; reduction; p < 0.001) accompanied by an increase in SS Supplementary Table 5; Supplementary Figs. 13 and 14). In the concentrations (30% increase; p = 0.001) may be associated with two driest sites in our network, Ken and Fec, we find strong enhanced branch growth35. In Tam, however, the dominant evidence of mobilisation of starch reserves to SS. In Ken, the pattern was one of greater leaf NSCT and SS in the wet season driest site evaluated, leaf starch reserves declined by 81% (p < (Fig. 3), which we attribute to higher productivity in the wet 0.001) in the dry season while leaf SS concentrations remained season, as observed by in situ NPP measurements36 and also seen unchanged (p = 1), despite a 43% reduction in leaf NSCT (Fig. 3, in MODIS-derived Enhanced Vegetation Index (EVI) values p = 0.019). In Fec, the second driest site evaluated, leaf starch (Supplementary Figs. 15 and 16). concentrations also decreased markedly in the dry season (72% reduction; p < 0.001). However, in this site, significant increases in leaf (32% increase; p < 0.001) and branch (48% increase; p < 0.001; Relationship between leaf NSC and water potential. Despite the Fig. 2) SS were observed, with an overall increase in leaf NSCT in widely varying seasonal patterns across sites, we find a strong the dry season. The reduction of leaf NSCT in Ken but not in Fec relationship between ΨMD measured in the driest period of the may be attributed to stronger source limitation in the driest Ken year and the proportion of leaf NSCT allocated to SS (SS:NSCT) Fig. 3 Seasonal variation of total NSC (NSCT), starch and soluble sugars (SS) across Amazon forests. Sites are ordered from left to right, from driest to wettest site. Panels a, c and e represent leaves and panels b, d and f represent branches. Red boxes denote the dry season and blue boxes denote the wet season. Each box encompasses the 25th to 75th percentiles; the median is indicated by the horizontal line with each box while external horizontal lines indicate the 10th and 90th percentiles; dots indicate outliers. Number of species sampled in each site are the same in the dry and wet season and are as follows: Ken = 9, Fec = 14, Man = 13, Tam = 21. To test for differences between season within site we used paired sample Wilcoxon tests. NATURE COMMUNICATIONS | (2021)12:2310 | https://doi.org/10.1038/s41467-021-22378-8 | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-22378-8 Fig. 4 Relationship between leaf NSC and water status, and distributions of species-level seasonal shifts in leaf NSC allocation. Relationship between a leaf SS:NSCT (proportion of leaf NSCT in the form of soluble sugars) and midday leaf water potential (ΨMD); c leaf starch and ΨMD. Distributions of species-level seasonal shifts in b leaf SS:NSCT allocation (SS:NSCT Wet − SS:NSCT Dry) and d starch allocation (starch wet − starch dry). In the panel a and c the SS:NSCT, starch and ΨMD represent mean of all species sampled in each site in the dry season, except in the two sites where there is no dry season (monthly precipitation ≤ 100 mm month−1). Vertical and horizontal bars denote one standard error of the mean. The relationship between SS:NSCT, starch and ΨMD was fitted using standardised major axis (SMA) regression. In the panel b and d sites are ordered from top to bottom from drier to wettest. Long vertical black line denotate the mean ΔSS:NSCT and starch, each dot represents one species. Values to the left of the blue line denote species that increased SS:NSCT and starch in the dry season while those to the right of the line increased SS:NSCT and starch in the wet season. (Fig. 4a; p = 0.02, R2 = 0.77). This relationship also exists when on a small number of shrubs40 and trees41 have shown that foliar taking predawn leaf water potential (ΨPD) as a measure of water SS can be very important for osmotic adjustment, actively status and at the species level as well as the community-level accumulating in response to declining water potentials and thus (Supplementary Fig. 17) and is underpinned by a strong decline helping plants to avoid dessication42. Our study suggests this across sites in leaf starch concentrations with increasing water phenomenon is widespread and that leaf SS contributes to the deficit (Fig. 4c; p = 0.001, R2 = 0.93). This relationship points to maintenance of hydraulic function in adult tropical trees across a an notable increase of SS relative to starch in leaves during per- broad range of taxa. A key future focus of research should be the iods of water stress and is reinforced by the fact that in the two identification of the specific sugars involved in osmoregulation in driest sites we found that almost all species increased SS:NSCT in tropical plants. Studies on herbaceous species and temperate tree leaves during the dry season (Ken: p = 0.007, Fec: p < 0.001; species suggest an important osmoregulatory role of oligosac- Fig. 4b; Supplementary Figs. 18 and 19). Indeed, in several species charides such as raffinose and pinitol43–45, but their role in in Ken and Fec, dry season leaf starch reserves were effectively osmoregulation in tropical forest trees is unknown. exhausted (zero or very close to zero) while leaf SS levels were The ability of species to rapidly mobilise leaf starch into SS unchanged (Ken) or increased (Fec). Furthermore, species co- under water deficit is likely an important mechanism for occurring in both Fec and in the more mesic Tam site almost all tolerating water stress46. While our study shows that species had higher dry season foliar SS and lower foliar starch content in found in drier forests of the Amazon almost all have this capacity, the drier site (Supplementary Figs. 20 and 21). However, we do it is far from clear that species in less seasonal regions of the not find evidence of enhanced conversion of foliar starch to SS in Amazon are able to adjust their SS balance to the same extent. Man (Fig. 4b), a mesic site where species are less-adapted to Indeed, the lack of seasonal adjustment in leaf NSC allocation to prolonged water deficit, despite sampling during one of the most SS under exceptional water stress in the mesic Man site suggests intense droughts on record at that site (Supplementary Fig. 1). that tree species that are less adapted to strong seasonal drought The increase of foliar SS relative to starch in the driest sites may not have the capacity to rapidly adjust foliar SS under water (Fig. 4b) and its strong relationship with community-level water stress. This capacity may ultimately be an important determinant status (Fig. 4a) suggest an important role of foliar SS in mediating of future Amazon forest composition under continued climate responses to water deficit in Amazonian forests. The increased change. allocation of leaf NSC to SS in these sites is further independent of the seasonal behaviour of NSCT (Supplementary Fig. 22) and thus does not simply result from changes in source-sink Methods Sites and species. Plant material for NSC analysis was collected in six Amazonian dynamics. Studies mainly on agricultural crop systems37–39 and sites (Fig. 1 and Table 1). These sites were selected from the RAINFOR network of 6 NATURE COMMUNICATIONS | (2021)12:2310 | https://doi.org/10.1038/s41467-021-22378-8 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-22378-8 ARTICLE permanent and well-identified forest plots12,47,48 as representing the Amazon-wide Reflectance Distribution function, at a spatial resolution of 1 km and aggregated to gradient in mean annual precipitation (MAP) and seasonality (Supplementary biweekly (16-day) composites using the median values in this product, before EVI Fig. 22 and 23). The study sites also encompass a broad range of soil types and calculation54,55. The EVI was calculated using Eq. 1: forest plant communities22,49. Our sites included aseasonal forests with no cli- ρNIR  ρRed matological dry season (no months with rainfall ≤100 mm50; Alp, Suc), forests with EVI ¼ 2:5*  ð1Þ a moderate dry season (1–4 months with rainfall ≤100 mm; Fec, Man, Tam) and a ρNIR þ 6*ρRed  7:5*ρBlue þ 1 transitional forest in the most seasonal site at the southern margin of the Amazon where ρNIR is infrared reflectance, ρRed is red reflectance, and ρBlue is blue biome (7 months with rainfall <100 mm; Ken)51. Sampling took place during what reflectance. The constants (6, 7.5, 1, and 2.5) in the divisor represent the aerosol were climatologically normal years in most sites, with the exception of Man, where coefficient adjustment of the atmosphere for the red and blue band, the adjustment sampling took place during the strong 2015 El Niño drought event (Supplementary factor for the soil and the gain factor, respectively55,56. Fig. 1). In total we sampled 82 canopy tree species, from 63 genera and 29 families The composites were retrieved considering only cloud-free and low atmospheric (Supplementary Table 1). In each site, sampling was focused on the most dominant turbidity according to MAIAC quality flags. Further information on image canopy species in terms of basal area, with the total number of species sampled at processing and correction are described in Dalagnol et al.54. MODIS pixel values each site ranging from 9 to 31 (Table 1). Species-level identification of all trees were extracted considering the coordinate system of each site using raster57 and sampled is based on botanical vouchers previously collected and deposited in rgdal58 R packages. Amazon state herbaria (AMAZ, CUZ, HOXA, INPA, UFACPZ, USZ) by RAIN- FOR partners. All plot trees are tagged and identifications were obtained from the ForestPlots.net database48 (https://www.forestplots.net/; Ken, Fec, Tam, Alp and Statistical analysis. We performed all statistical analysis in R (R Core Team 2018, Suc) and from collaborator databases for Man24. All branches collected had fully version 3.6.2)59. Preliminary tests included: analysis of normality (Shapiro–Wilk), expanded leaves, no evidence of liana infestation or injuries and were not shaded. and homogeneity of variances (Fligner–Killeen) for each NSC fraction (NSCT, SS Species collected in the wet season that did not keep their leaves during the dry and starch) in each organ (branches and leaves). As NSCT, starch and SS were not season were excluded from all analysis to avoid potential biases due to different normally distributed, these parameters were log1p-transformed60 prior to variance phenological strategies1. partitioning. For comparison of means across sites and seasons, data were not log1p-transformed and non-parametric tests were used. To evaluate differences across sites in NSCT, SS and starch concentrations, a Non-structural carbohydrate (NSC) sampling and analysis. Two to six indivi- Kruskal–Wallis test was used (R base function). We conducted statistical analyses duals were sampled per species, with all individuals being >20 cm in diameter at separately for each plant organ (leaves and branches) and season (wet and dry). breast height (DBH). To minimise effects of diurnal changes in NSC concentra- When a significant site effect was found, a post hoc Mann–Whitney–Wilcoxon tions, samples were obtained before sunrise in all sites, except Man. In Man, Rank Sum test using Bonferroni correction (from “Agricolae” package in R61) was branches and leaves were obtained just after sunrise and always before 8 a.m. Leaf performed to evaluate which sites were significantly different. To evaluate whether and branch samples were obtained by a tree climber from first order fully there were significant differences in NSCT, starch and SS between seasons, we sunlit branches with fully expanded leaves and kept in ice during sampling and performed paired sample Wilcoxon tests (R base package) for each site and plant transported to the laboratory. Upon arrival at the laboratory, samples were organ separately. The ALP and SUC sites were excluded from the seasonal analysis, microwaved for 90 s at 700 W to stop enzymatic activity that would otherwise affect as these sites were only sampled in one point in time. Figures were constructed NSC levels, and oven-dried at ~60 °C for at least 48 h or until they were completely using the “ggplot2” package62 and to display the p-values in the figures we used the dry (no >72 h). All NSC sample preparation and analyses were performed at “ggpubr” package (stat_compare_means function)63. the University of Campinas, in the laboratory of plant ecophysiology. Prior to We conducted standardised major axis (SMA) regression, using the “smatr” NSC quantification, samples were ground to a fine powder (Geno/Grinder® package in R64 to assess relationships between NSCT, SS and starch and site level SPEX SamplePrep mill). Branch samples had their bark removed before being water-status (Ψmin and ΨPD) at both the community-level (mean species value per ground. site) and species level. To account for differences across sites in seasonal source Non-structural carbohydrates (NSC) are defined here as free, low molecular limitation we also tested for relationships between the ratio of SS and NSCT (SS: weight sugars (i.e., oligosaccharides such as glucose, fructose, sucrose, etc.) plus NSCT) and ΨMD and ΨPD. SMA regressions were conducted for each organ starch. NSC was analysed as described in Hoch et al.52 with minor modifications. separately, using NSC values corresponding to the same season Ψ was measured. Two replicates of each sample were analysed and the mean of the two replicates We further tested for bivariate relationships between species-mean NSC metrics used as the sample NSC value. First, we diluted ~15 mg of the ground plant and plant attributes indicative of species life history strategies, including branch material with 1.6 mL of distilled water and then incubated in a water bath at wood density (Tavares et al. in prep and Barros et al. unpublished data), mean 90–100 °C for 60 min to solubilise sugars. Then we took an aliquot of 700 μL from growth rate, potential maximum size and mean mortality rate25,65 using the lm each sample (700 μL). We use the remaining aliquot volumes (900 μL) to determine function (R base package). the SS concentration using invertase from Saccharomyces cerevisiae (Sigma- To determine the relative importance of taxonomy vs. measurement site in Aldrich, St. Louis, MO, USA) to break down sucrose and fructose to glucose20. determining NSCT and its fractions, we undertook a variance partitioning analysis Additionally, for both reaction routines, we used GAHK (Glucose Assay as described by Fyllas et al.23, where a multilevel model was first fitted for each NSC Hexokinase Kit - Sigma-Aldrich, St. Louis, MO, USA) together with fraction, organ and season according to Eq. 2: phosphoglucose isomerase from Saccharomyces cerevisiae (Sigma-Aldrich, St. T ¼ μ þ p þ f =g=s þ ε ð2Þ Louis, MO, USA). The concentration of free glucose was measured photometrically in a 96-well microplate spectrophotometer at 340 μm (EPOCH-Biotek Instruments where µ is the overall mean species value of each NSC fraction (T), p is the random INC-Winooski, VT-USA). The aliquot that we initially separated was incubated site effect, i.e., the effect of the location at which each individual was found (soil overnight to react with Amyloglucosidase from Aspergillus niger (Sigma-Aldrich, and climate), f/g/s represents the random effect caused by the genetic structure of St. Louis, MO, USA) to breakdown the total NSC to glucose. Thereafter total the data, i.e., that each individual belongs to a species (s), nested in a genus (g), glucose (corresponding to NSC) was determined as described above and starch was nested in a family (f), and ε is the residual term, which includes both the within- calculated as total NSC minus soluble sugars. All NSC values are expressed in mg/g species variability not explained by site, as well as any measurement error. All dry mass. parameters were estimated by the Residual or Restricted Maximum Likelihood (REML) method with the “lme4” package in R66. Predawn (ΨPD) and midday (ΨMD) leaf water potential. In situ predawn and midday leaf water potential (ΨPD and ΨMD respectively) measurements were made Reporting summary. Further information on research design is available in the Nature in the same day on the same trees for which we obtained samples for determination Research Reporting Summary linked to this article. of NSC concentrations. We sampled 2–6 trees per species and measured the ΨPD and ΨMD in 2–5 canopy fully expanded leaves using a pressure chamber (PMS Data availability 1505D and PMS 1000, PMS instruments) and the values were then averaged per The non-structural carbohydrates concentration data are available at www.forestplots.net/ individual. ΨPD measurements were taken before sunrise from 3:30–5:30 and ΨMD data-packages/Signori-Muller-et-al-2021 (https://doi.org/10.5521/forestplots.net/ from 11:00 am –2:30 pm. Water potential data collection took place in what is 2021_3)67, all recorded species, genus and family names were checked and standardized typically the driest time of the year in each sampling plot except in the Alp and using the Taxonomic Name Resolution Service (tnrs.biendata.org)68. The mean growth Suc sites where there is no climatological dry season (months with precipitation rate, potential tree size and mortality rate are available at www.forestplots.net/data- ≤100 mm) and little seasonality in rainfall. Owing to logistical limitations, we did packages/coelho-de-souza-et-al-2016 (https://doi.org/10.5521/FORESTPLOTS.NET/ not measure ΨPD in the Man site. 2016_4)66. The climatic data are available at www.worldclim.org/data/index.html51. Enhanced vegetation index data are available at www.zenodo.org/record/3159488#. Enhanced vegetation index (EVI). To gain further insights into how seasonality YBW_u3f7Tlw (https://doi.org/10.5281/ZENODO.3159488)53,54. Leaf water potential in canopy productivity might affect our observations, we extracted Enhanced data are available upon reasonable request to the correspondence author. Branch wood Vegetation Index (EVI) values for each site derived from the MODIS-MAIAC density data are from Tavares et al. (in prep) and Vasconcelos Barros (unpublished data). product, using data from 2003 to 201853. The surface reflectance data were nor- The inventory data for species selection are from the RAINFOR network available upon malised to nadir target and 45-degree solar zenith angle through the Bidirectional request at www.forestplots.net47,48. NATURE COMMUNICATIONS | (2021)12:2310 | https://doi.org/10.1038/s41467-021-22378-8 | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-22378-8 Received: 26 July 2020; Accepted: 19 February 2021; 30. Poorter, L. & Kitajima, K. Carbohydrate storage and light requirements of tropical moist and dry forest tree species. Ecology 88, 1000–1011 (2007). 31. Nikinmaa, E. et al. Assimilate transport in phloem sets conditions for leaf gas exchange. Plant Cell Environ. 36, 655–669 (2013). 32. Tyree, M. T. & Ewers, F. W. The hydraulic architecture of trees and other woody plants. N. Phytol. 119, 345–360 (1991). References 33. Guan, K. et al. Photosynthetic seasonality of global tropical forests constrained 1. Martínez-Vilalta, J. et al. Dynamics of non-structural carbohydrates in by hydroclimate. Nat. Geosci. 8, 284–289 (2015). terrestrial plants: A global synthesis. Ecol. Monogr. 86, 495–516 (2016). 34. Restrepo-Coupe, N. et al. What drives the seasonality of photosynthesis across 2. Hartmann, H. & Trumbore, S. Understanding the roles of nonstructural the Amazon basin? A cross-site analysis of eddy flux tower measurements carbohydrates in forest trees–from what we can measure to what we want to from the Brasil flux network. Agric. Meteorol. 182–183, 128–144 (2013). know. N. Phytol. 211, 386–403 (2016). 35. AbdElgawad, H. et al. Starch biosynthesis contributes to the maintenance of 3. Richardson, A. D. et al. Seasonal dynamics and age of stemwood nonstructural photosynthesis and leaf growth under drought stress in maize. Plant. Cell carbohydrates in temperate forest trees. N. Phytol. 197, 850–861 (2013). Environ. https://doi.org/10.1111/pce.13813 (2020). 4. Doughty, C. E. et al. Source and sink carbon dynamics and carbon allocation 36. Malhi, Y. et al. The productivity, metabolism and carbon cycle of two lowland in the Amazon basin. Glob. Biogeochem. Cycles 29, 645–655 (2015). tropical forest plots in south-western Amazonia, Peru. Plant Ecol. Divers. 7, 5. Mcdowell, N. et al. Mechanisms of plant survival and mortality during 85–105 (2014). drought: why do some plants survive while others succumb to drought? N. 37. Sánchez, F. J., Manzanares, M., De Andres, E. F., Tenorio, J. L. & Ayerbe, L. Phytol. 178, 719–739 (2008). Turgor maintenance, osmotic adjustment and soluble sugar and proline 6. Sala, A., Piper, F. & Hoch, G. Physiological mechanisms of drought-induced accumulation in 49 pea cultivars in response to water stress. F. Crop. Res. 59, tree mortality are far from being resolved. N. Phytol. 186, 274–281 (2010). 225–235 (1998). 7. Farquhar, G. D. & Sharkey, T. D. Stomatal conductance and photosynthesis. 38. Morgan, J. M. Osmoregulation and water stress in higher plants. Annu. Rev. Annu. Rev. Plant Physiol. 33, 317–345 (1982). Plant Physiol. 35, 299–319 (1984). 8. Adams, H. D. et al. Nonstructural leaf carbohydrate dynamics of Pinus edulis 39. Thalmann, M. & Santelia, D. Starch as a determinant of plant fitness under during drought-induced tree mortality reveal role for carbon metabolism in abiotic stress. N. Phytol. 214, 943–951 (2017). mortality mechanism. N. Phytol. 197, 1142–1151 (2013). 40. Guo, J. S., Gear, L., Hultine, K. R., Koch, G. W. & Ogle, K. Non-structural 9. O’Brien, M. J., Leuzinger, S., Philipson, C. D., Tay, J. & Hector, A. Drought carbohydrate dynamics associated with antecedent stem water potential and survival of tropical tree seedlings enhanced by non-structural carbohydrate air temperature in a dominant desert shrub. Plant Cell Environ. https://doi. levels. Nat. Clim. Chang. 4, 710–714 (2014). org/10.1111/pce.13749 (2020). 10. McDowell, N. G. Mechanisms linking drought, hydraulics, carbon metabolism, 41. Kuang, Y., Xu, Y., Zhang, L., Hou, E. & Shen, W. Dominant trees in a and vegetation mortality. Plant Physiol. 155, 1051–1059 (2011). subtropical forest respond to drought mainly via adjusting tissue soluble sugar 11. Brienen, R. J. W. et al. Long-term decline of the Amazon carbon sink. Nature and proline content. Front. Plant Sci. 8, 1–13 (2017). 519, 344–348 (2015). 42. Turner, N. C. Turgor maintenance by osmotic adjustment: 40 years of 12. Phillips, O. L. et al. Drought Sensitivity of the Amazon Rainforest. Science (80- progress. J. Exp. Bot. 69, 3223–3233 (2018). .) 323, 1344–1347 (2009). 43. Kandler, O. & Hopf, H. in Carbohydrates: Structure and Function. Vol. 3, 13. Lewis, S. L., Brando, P. M., Phillips, O. L., van der Heijden, G. M. F. & 221–270 (Elsevier, 1980). Nepstad, D. The 2010 amazon drought. Science (80-.) 331, 554–554 (2011). 44. Deslauriers, A. et al. Impact of warming and drought on carbon balance 14. Jiménez-Muñoz, J. C. et al. Record-breaking warming and extreme drought in related to wood formation in black spruce. Ann. Bot. 114, 335–345 (2014). the Amazon rainforest during the course of El Niño 2015-2016. Sci. Rep. 6, 45. Ford, C. W. Accumulation of low molecular weight solutes in water-stressed 1–7 (2016). tropical legumes. Phytochemistry 23, 1007–1015 (1984). 15. Duffy, P. B., Brando, P., Asner, G. P. & Field, C. B. Projections of future 46. Mitchell, P. J., O’Grady, A. P., Tissue, D. T., Worledge, D. & Pinkard, E. A. meteorological drought and wet periods in the Amazon. Proc. Natl Acad. Sci. Co-ordination of growth, gas exchange and hydraulics define the carbon safety U.S.A. 112, 13172–13177 (2015). margin in tree species with contrasting drought strategies. Tree Physiol. 34, 16. Jones, S. et al. The impact of a simple representation of non-structural 443–458 (2014). carbohydrates on the simulated response of tropical forests to drought. 47. Malhi, Y. et al. An international network to monitor the structure, Biogeosciences https://doi.org/10.5194/bg-2019-452 (2019). composition and dynamics of Amazonian forests (RAINFOR). J. Veg. Sci. 13, 17. Dünisch, O. & Puls, J. Changes in content of reserve materials in an evergreen, 439–450 (2002). a semi-deciduous, and a deciduous Meliaceae species from the Amazon. J. 48. Lopez-Gonzalez, G., Lewis, S. L., Burkitt, M. & Phillips, O. L. ForestPlots.net: a Appl. Bot. 77, 10–16 (2003). web application and research tool to manage and analyse tropical forest plot 18. Würth, M. K. R., Peláez-Riedl, S., Wright, S. J. & Körner, C. Non-structural data. J. Veg. Sci. 22, 610–613 (2011). carbohydrate pools in a tropical forest. Oecologia 143, 11–24 (2005). 49. Sakschewski, B. et al. Resilience of Amazon forests emerges from plant trait 19. Dickman, L. T. et al. Homoeostatic maintenance of nonstructural diversity. Nat. Clim. Chang. 1, 1–5 (2016). carbohydrates during the 2015–2016 El Niño drought across a tropical forest 50. Sombroek, W. Spatial and temporal patterns of amazon rainfall. AMBIO A J. precipitation gradient. Plant Cell Environ. 42, 1705–1714 (2019). Hum. Environ. 30, 388–396 (2001). 20. Rowland, L. et al. Death from drought in tropical forests is triggered by 51. Fick, S. E. & Hijmans, R. J. WorldClim 2: new 1-km spatial resolution climate hydraulics not carbon starvation. Nature 528, 119–122 (2015). surfaces for global land areas. Int. J. Climatol. 37, 4302–4315 (2017). 21. Malhi, Y. et al. Spatial patterns and recent trends in the climate of tropical 52. Hoch, G., Popp, M. & Korner, C. Altitudinal increase of mobile carbon pools rainforest regions. Philos. Trans. R. Soc. Lond. B. Biol. Sci. 359, 311–329 (2004). in Pinus cembra suggests sink limitation of growth at the Swiss treeline. Oikos 22. Quesada, C. A. et al. Variations in chemical and physical properties of Amazon 98, 361–374 (2002). forest soils in relation to their genesis. Biogeosciences 7, 1515–1541 (2010). 53. Dalagnol, R., Wagner, F. H., Galvão, L. S. & Aragão, L. E. O. C. The MANVI 23. Fyllas, N. M. et al. Basin-wide variations in foliar properties of Amazonian product: MODIS (MAIAC) nadir-solar adjusted vegetation indices (EVI and forest: phylogeny, soils and climate. Biogeosciences 6, 2677–2708 (2009). NDVI) for South America. Zenodo https://doi.org/10.5281/ZENODO.3159488 24. de Barros, F. V. et al. Hydraulic traits explain differential responses of (2019). Amazonian forests to the 2015 El Niño-induced drought. N. Phytol. 223, 54. Dalagnol, R., Wagner, F. H., Galvão, L. S., Nelson, B. W. & De Aragão, L. E. O. 1253–1266 (2019). E. C. Life cycle of bamboo in the southwestern Amazon and its relation to fire 25. Coelho de Souza, F. et al. Evolutionary heritage influences Amazon tree events. Biogeosciences 15, 6087–6104 (2018). ecology. Proc. R. Soc. B Biol. Sci. 283, 20161587 (2016). 55. Fonseca, L. D. M. et al. Phenology and seasonal ecosystem productivity in an 26. Dietze, M. C. et al. Nonstructural carbon in woody plants. Annu. Rev. Plant Amazonian floodplain forest. Remote Sens. 11, 1–17 (2019). Biol. 65, 667–687 (2014). 56. Huete, A. et al. Overview of the radiometric and biophysical performance of 27. Tixier, A., Orozco, J., Amico Roxas, A., Earles, J. M. & Zwieniecki, M. A. the MODIS vegetation indices. Remote Sens. Environ. 83, 195–213 (2002). Diurnal variation in non-structural carbohydrate storage in trees: 57. Hijmans, R. J. et al. Raster: Geographic Data Analysis And Modeling. (R remobilization and vertical mixing. Plant Physiol. 178, 1602–1613 (2018). package, 2020). 28. Landhäusser, S. M. et al. Standardized protocols and procedures can precisely 58. Bivand, R. et al. Rgdal: Bindings for the ‘Geospatial’ Data Abstraction Library. and accurately quantify non-structural carbohydrates. Tree Physiol. 38, (R package, 2020). 1764–1778 (2018). 59. R Core Team. R: A Language And Environment For Statistical Computing. 29. MacNeill, G. J. et al. Starch as a source, starch as a sink: the bifunctional role of URL https://www.R-project.org/. (R Foundation for Statistical Computing, starch in carbon allocation. J. Exp. Bot. 68, 4433–4453 (2017). 2018). 8 NATURE COMMUNICATIONS | (2021)12:2310 | https://doi.org/10.1038/s41467-021-22378-8 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-22378-8 ARTICLE 60. Hull, T. E., Fairgrieve, T. F. & Tang, P. T. P. Implementing complex Integral (AIDER) for field campaign support in Peru; Ezequiel Chavez and Noel Kempff elementary functions using exception handling. ACM Trans. Math. Softw. 20, Natural History Museum for field campaign support in Bolivia; and Hugo Ninantay and 215–244 (1994). Alex Ninantay for sample collection. 61. De Mendiburu, F. Agricolae: Statistical Procedures For Agricultural Research (R package version 1.1, 2014). 62. Wickham, H. ggplot2: Elegant Graphics for Data Analysis (Springer-Verlag Author contributions D.G., C.S.-M. and R.S.O. designed the study. C.S-M. led the data analysis with inputs New York, 2016). from D.G. and M.B. C.S.-M and D.G. wrote the manuscript with significant inputs from 63. Kassambara, A. ggpubr: ‘ggplot2’ Based Publication Ready Plots. (R package R.S.O. and M.M. J.V.T. and C.S.-M. led the field sampling in all sites, except MAN, version 0.1.8, 2020). which was led by F.V.B. C.S-M., F.V.B. and M.G. performed the NSC analysis. C.S.-M., 64. Warton, D. I., Duursma, R. A., Falster, D. S. & Taskinen, S. smatr 3- an R J.V.T., M.G., F.C.D., M.J.M.Z., C.A.S.Y., M.A., J.B., R.S.C.C., G.M.A.C., E.R.M.C., F.M., package for estimation and inference about allometric lines. Methods Ecol. F.M.P.M., A.N, J.M.B.S., L.F.d.S., L.T., J.S.T. and M.T.M.U. collected the samples from all Evol. 3, 257–259 (2012). sites except for MAN where it was sampled by F.V.B., P.R.L.B., R.O. and L.S.B. T.R.B, 65. Coelho de Souza, F. et al. Trait data from: ‘Evolutionary heritage influences W.C., E.N.H.C., E.G.C., T.R.F., E.G., G.F.L., Y.M., A.M.M., V.C.M., A.A.-M., O.L.P., N.S., Amazon tree ecology’. ForestPlots.net . https://doi.org/10.5521/ M.S., J.T. and R.V. led the forestplot.net field expeditions for data collection. L.S.B. and FORESTPLOTS.NET/2016_4 (2016). L.d’A.M.F. provided the enhanced vegetation index data. All authors critically revised the 66. Bates, D., Sarkar, D., Bates, M. D. & Matrix, L. The lme4 Package. October 2, manuscript and gave final approval for publication. 1–6 (2007). 67. Signori-Müller, C. et al. Trait data from: ‘Non-structural carbohydrates mediate seasonal water stress across Amazon forests’. ForestPlots.net 5521 Competing interests https://doi.org/10.5521/forestplots.net/2021_3 (2021). The authors declare no competing interests. 68. Boyle, B. et al. The taxonomic name resolution service: an online tool for automated standardization of plant names. BMC Bioinformatics 14, 16 (2013). Additional information 69. Esquivel-Muelbert, A. et al. Tree mode of death and mortality risk factors Supplementary information The online version contains supplementary material across Amazon forests. Nat. Commun. 11, 5515 (2020). available at https://doi.org/10.1038/s41467-021-22378-8. Correspondence and requests for materials should be addressed to C.S-M. Acknowledgements Peer review information Nature Communications thanks Kaoru Kitajima, Henry Amazon fieldwork and laboratory analyses for this publication were supported by the UK Adams, and the other, anonymous reviewer(s) for their contribution to the peer review of Natural Environment Research Council project TREMOR (NE/N004655/1) to D.G. This this work. Peer review reports are available. paper is an outcome of C.S.-M.’s doctoral thesis in the Plant Biology Programme at University of Campinas. C.S.-M., J.V.T., F.B.V., M.B., P.R.L.B., L.d’.A.F., L.S.B. and R.S.O. Reprints and permission information is available at http://www.nature.com/reprints. were financed by Coordenação de Aperfeiçoamento de Pessoal de Nível Superior-Brasil (CAPES Finance Code 001). C.S.-M. received a scholarship from the Brazilian National Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in Council for Scientific and Technological Development (CNPq 140353/2017-8) and published maps and institutional affiliations. CAPES (science without borders 88881.135316/2016-01). D.G. acknowledges further support from the NERC-funded ARBOLES project (NE/S011811/1). J.V.T. acknowledges CAPES for its science without borders scholarship (99999.001293/2015-00). C.S.-M. and R.S.O. acknowledge the São Paulo Research Foundation (FAPESP-Microsoft 11/52072- Open Access This article is licensed under a Creative Commons 0). M.M. acknowledges support from MINECO FUN2FUN (CGL2013-46808-R) and Attribution 4.0 International License, which permits use, sharing, DRESS (CGL2017-89149-C2-1-R). R.S.O., F.V.B., L.S.B. and P.R.L.B. thank the U.S. adaptation, distribution and reproduction in any medium or format, as long as you give Department of Energy, project GoAmazon (FAPESP 2013/50531-2). P.R.L.B. acknowl- appropriate credit to the original author(s) and the source, provide a link to the Creative edges The Royal Society for its Newton International Fellowship (NF170370). A number Commons license, and indicate if changes were made. The images or other third party of the field sites (KEN, TAM, ALP) are part of the Global Ecosystems Monitoring (GEM) material in this article are included in the article’s Creative Commons license, unless network (gem.tropicalforests.ox.ac.uk) and were supported by the Gordon and Betty indicated otherwise in a credit line to the material. If material is not included in the Moore Foundation and by ERC Advanced Investigator Grant (GEM-TRAITS, 321131) to article’s Creative Commons license and your intended use is not permitted by statutory Y.M. The Amazon forest plots in the RAINFOR network analysed here were established, regulation or exceeds the permitted use, you will need to obtain permission directly from identified and measured with support from many colleagues and grants mentioned the copyright holder. To view a copy of this license, visit http://creativecommons.org/ elsewhere11,69. This study was carried out as a collaborative effort of the ForestPlots.net licenses/by/4.0/. meta-network, a cyber-infrastructure initiative developed at the University of Leeds that unites contributing scientists and their permanent plot records from the world’s tropical forests. This paper is an outcome of ForestPlots.net approved Research Project #18. We © The Author(s) 2021 additionally thank: Vanessa Hilares and the Asociación para la Investigación y Desarrollo 1 Department of Plant Biology, Institute of Biology, Programa de Pós Graduação em Biologia Vegetal, University of Campinas, Campinas, Brazil. 2 School of Geography, University of Leeds, Leeds, UK. 3Department of Plant Biology, Institute of Biology, University of Campinas, Campinas, Brazil . 4Geography, College of Life and Environmental Sciences, University of Exeter, Exeter, UK. 5Department of Plant Biology, Institute of Biology, Programa de Pós Graduação em Ecologia, University of Campinas, Campinas, Brazil. 6Universidad Nacional de San Antonio Abad del Cusco, Cusco, Peru. 7Pontificia Universidad Católica del Perú, Lima, Perú. 8Programa de Pós-Graduação em Ecologia e Manejo de Recursos Naturais, Universidade Federal do Acre, Rio Branco, Brazil. 9Museo de Historia Natural Noel Kempff Mercado, Universidad Autonoma Gabriel Rene Moreno, Santa Cruz, Bolivia. 10Earth System Science Centre, National Institute for Space Research, São José dos Campos, Brazil. 11Department of Ecology and Evolutionary Biology, University of Arizona, Tucson, AZ, USA. 12Instituto de Investigaciones de la Amazonia Peruana, Iquitos, Peru. 13 Sección Química, Pontificia Universidad Católica del Perú, Lima, Peru. 14Universidad Nacional Jorge Basadre de Grohmann, Tacna, Peru. 15 Environmental Change Institute, School of Geography and the Environment, University of Oxford, Oxford, UK. 16Institute for Transport Studies, University of Leeds, Leeds, UK. 17Jardín Botánico de Missouri, Pasco, Peru. 18CREAF, Universidad Autonoma de Barcelona, Barcelona, Spain. 19 ICREA, Barcelona, Spain. ✉email: carol.signori@gmail.com NATURE COMMUNICATIONS | (2021)12:2310 | https://doi.org/10.1038/s41467-021-22378-8 | www.nature.com/naturecommunications 9