One way of explaining how plants cope with environmental conditions, and how they coexist, is to investigate their morphological, reproductive, physiological and phenological traits. Although derived from species characteristics, traits are defined, based on a tested or an assumed role they play on plant growth, survival and reproduction. In reality, no species can have traits to be competitive in every environment as most adaptations also come at a cost, causing trade-offs, i.e., the value of one trait cannot increase without the decrease of that of another. One example is the number of seeds a plant can produce. For a given amount of energy, a plant can make many small seeds but much fewer big seeds, known as the seed size - seed number tradeoff¹. Another well-known trade-off, the “worldwide leaf economics spectrum”^2,3 captures various leaf traits, running “from quick to slow return on investments of nutrients and dry mass in leaves, and operates largely independently of growth form, plant functional type or biome”². A similar worldwide spectrum has been described for wood traits⁴ and more recently for roots⁵.

The leaf-economics spectrum² is tightly related to resource capture and use^3,6. Species that are specialised for resource rich environments, such as fertile soils, generally possess a high specific leaf area (SLA) which increases light capture per unit leaf biomass, and high phosphorus (P) and nitrogen (N) concentrations which increases metabolic rates, photosynthetic capacity, carbon gain and growth^2,7,8,9. In contrast, species that are specialised for resource poor environments, such as infertile soils, produce thick, dense, and structurally well-defended leaves. Additionally, they have high leaf carbon concentrations and C:N ratios reflecting investments in structural and chemical defences such as lignin, tannins and soluble phenolics¹⁰. In combination, these traits increase leaf lifespan and nutrient conservation, while enhancing the length of the photosynthetic revenue from leaves^2,3. Hence, in wet forests on infertile soils an evergreen leaf habit is important for nutrient conservation.

Another aboveground spectrum, known as the stature−recruitment trade-off, runs from small to tall species with increasing stem tissue density, leaf size, and seed mass^3,6,11,12,13. Wood density relates to biomechanical resistance against biophysical hazards such as wind and pathogens, and can relate to increased resistance to drought-induced embolism¹⁴. Therefore wood density enhances a tree species’ lifespan⁴, and is a conservative trait that is beneficial on infertile soils, as it increases the biomass residence time¹⁵. Wood density has been shown to increase towards the northeast in Amazonia, presumably linked to poor soils^16,17. Seed mass is a key functional trait that influences a plant’s ecological strategy, dispersal, and establishment success¹⁶ and decreases with soil fertility. Large seeds contain more carbohydrate and nutrient reserves, and produce larger, more robust seedlings which enhances seedling establishment and survival in low resource environments (such as nutrient-poor soils and the shaded wet forest understory)^{16,18,19,20,21}. Seeds with a low mass (i.e. small seeds) often have wind-, bird- or bat-based dispersal, an ability to colonise larger areas, but lower success in individual seedling establishment^22,23, unless frequent disturbances make light available. As with wood density, seed mass has been shown to be much higher in eastern than western Amazonia¹⁶.

More recently a below ground economics spectrum was described, with two dimensions based on six fine root traits: a collaboration gradient running from a do-it-yourself strategy for acquiring resources to an outsourcing one relying on fungi, and a conservation gradient opposing fast-slow fine roots⁵. We have no root morphological data for the vast number of species in Amazonia but do have information on three important root traits (mycorrhiza, nitrogen-fixation, and aluminium accumulation). Most plant species are associated with vesicular arbuscular mycorrhiza^5,24, and distributed quite evenly across what was called the root economic space⁵, while species associated with ectomycorrhiza were more found in the non-collaborative space⁵. Ectomycorrhizae (EM) are important for monodominant forests on very poor soils in Central Africa²⁵ and in Guyana^26,27,28 but are otherwise very rare in Amazonia^24,29. Nitrogen fixation is thought to be an important trait on nutrient poor soils. However, earlier work has shown that non-nitrogen fixing ‘caesalpinioid’ Fabaceae dominate poor, acidic soil regions in Amazonia^16,29, perhaps because of a lack of trace elements necessary for nodulation. An additional reason might be that symbiotic nitrogen fixers prefer non-acidic, somewhat drier soils²⁴ and early successional forests^30,31, rather than shaded old growth forest because of the high energy requirements to maintain the rhizobial symbiotic bacteria. Aluminium accumulation occurs in only 18 tree families³². Some plant species accumulate aluminium that enters their roots into their leaves on soils with high aluminium content, to protect the toxic effect of aluminium on root tips. Aluminium accumulation is restricted to a number of families^33,34,35. Aluminium accumulators are found mainly on aluminium rich soils, such as those of the Cerrado in Brazil^36,37 but no information is yet available on the distribution of aluminium accumulation in Amazonia.

At species level the leaf economics spectrum produces trade-offs as well: that of plant species being either slow growing and shade tolerant or fast growing but light demanding^38,39,40,41. A similar trade-off is found among species growing on nutrient rich or nutrient poor soils^42,43. In both cases an increasingly limiting resource (light, nutrients) leads to the need for conserving plant tissues, as they cannot easily be replaced under strongly limiting conditions. Thus, under favourable conditions, plants will show traits that allow them to grow fast at the cost of being less defended. Their fast growth, however, allows them to compensate for loss of their tissue to herbivores^44,45,46. The slow/fast division in plant strategies has been observed much earlier and led to the much-used pioneer-climax tree division, with pioneers having cheap, large, thin, short-lived leaves, soft wood, small seeds, subsequent fast growth in high light but high mortality in shade. Climax species tend to have thick long-lived leaves, hard wood, slow growth but superior survival under shaded conditions^47,48. Thus, “species with large seeds, long-lived leaves, or dense wood have slow life histories, with mean fitness (i.e., population growth rates) more strongly influenced by survival than by growth or fecundity, compared with fast life history species with small seeds, short-lived leaves, or soft wood”⁴⁹. The link between the leaf economic spectrum and wood economic spectrum, however, is not that clear⁴, and in a large study with 668 neotropical tree species and 16 leaf and wood traits it was shown that the traits of leaf-economic spectrum and those of the wood-economic spectrum were orthogonal rather than correlated. This suggests that the “trade-offs operate independently at the leaf and at the stem levels”⁵⁰, which is consistent across Amazonia⁵¹.

Recently information became available on breeding systems of Amazonian trees and the production of nectar producing taxa world-wide⁵². We added these traits to our list to investigate their relationship with the above traits. While there is strong support for the leaf-economic-spectrum and stature-recruitment trade-off, it is less clear if different vegetative organs (leaf, stem and roots), reproductive strategies (breeding system, fleshy fruits, seed size), and symbioses (nitrogen fixing bacteria, mycorrhizae) align along these two main identified spectra (leaf-economic-spectrum and stature-recruitment trade-off) or represent independent, novel axes of strategy variation, thereby expanding the opportunities for niche differentiation and species coexistence.

Because plant species assemble to make up local communities, it is often assumed that these strategy spectra at the species level are filtered and translated into community-level functional composition^53,54. However, in spite of ongoing studies, it remains unclear whether this is the case for different tropical forests and how this translates into ecosystem functioning (carbon storage and sequestration) at the regional scale, which at the smaller scale is known to be affected by soil fertility and forest dynamics⁵⁵. Addressing these questions is especially critical for tropical ecosystems, as they provide by far the majority of the planet’s species diversity and terrestrial ecosystem function⁵⁶.

Here, we investigate the relationship between 13 tree functional traits from 812 Amazonian tree genera, including above- and below-ground traits, and reproductive traits for the world’s largest and most diverse tropical forest. We scale up from genus to community level across the entire Amazon region by calculating community-weighted mean (CWM) trait data for over 2000 tree-inventory plots in seven forest/soil combinations, mapping each trait and the result of the main strategy axis. We investigate potential drivers (climate, soil) of community trait values, and assess the implications of community trait values for forest functioning in terms of carbon storage and sequestration. We also test how life-histories, such as short-lived or (early) pioneers, long-lived pioneers, and old growth species – often used in models¹², and one life-history characteristic (observed maximum diameter) are related to the strategy axes of both genus-level and community-level analyses. Because it was recently shown that pre-Colombian people left a lasting imprint in some Amazonian forests^57,58,59, we also included information on the abundance of domesticated species⁶⁰ and the probability of finding evidence of human occupation (geoglyphs = human constructed earth works)⁵⁸ as factors potentially affecting forest traits composition and function.

Our main research questions are: (1) What plant-strategy spectra are found among Amazonian tree genera? (2) Are community-level strategy spectra similar to genus-level strategy spectra? (3) What are the main spatial gradients in functional composition across Amazonian forest types, and how does this relate to climate, large-scale disturbance, soil, and pre-Columbian human legacies? (4) How does community functional composition affect carbon storage and carbon sequestration?

If ecological filters are not modifying the relationships between traits from genera to communities, we may expect similar trait spectra at the genus and community levels. If the wood and leaf economic spectra, uncoupled at species level^50,51, react to soil fertility across Amazonia in a similar way, we expect convergence among traits within communities^9,61, and divergence among traits between communities driven by local and regional differences in soil⁶². We expect forest communities with, on average, slow traits to have high community mean wood density^16,17 but the relationship with forest productivity is less clear^55,63.

See Supplementary Box 1 for a description of the traits used and a discussion on their importance for tree ecology.

Traits at genus level

To understand what plant-strategy spectra are found among Amazonian tree genera (question one), we started by identifying plant strategies at genus level. A principal component analysis was carried out on a dataset from all 2253 plots (Supplementary Fig. 1), including all 13 traits, using only the 535 genera with data for all traits, representing 90.8% of all individuals. Four leaf traits (N, P, SLA, C:N, Fig. 1, Table 1) were strongly related with Principal component 1 (PC1, Eigenvalue = 2.98), which explains 22.97% of variance in the data (Supplementary Table 1a). PC1 therefore represents the ‘leaf economics spectrum’ (LES) running from ‘fast’ productive leaves with high SLA and leaf nutrient concentrations to the left, to ‘slow’ well-defended leaves with high C:N to the right. Traits related to tree size and reproduction were mostly related to PC2 (Eigenvalue = 1.75), which explains 13.52% of the variance in the data and represents the “stature−recruitment trade-off’, running from small species with fleshy fruits (FF) at the bottom to large species with somewhat large maximum diameter (Max), high seed mass (SM), high wood density (WD), a hermaphroditic breeding system (Her), and nectar (Supplementary Box 1). Life-history classification was based on seed mass and wood density^29,64. Life histories were therefore closely linked with the stature-recruitment trade-off, being defined by three strategies, as follows: short-lived pioneers (SLP) with small seeds and soft wood occupying the small side, and old growth species (OGS) with large seeds and dense wood occupying the tall side of the spectrum. Long lived pioneers (LLP) are intermediate with light wood and relatively large seeds (Fig. 1, Supplementary Table 1c, Supplementary Fig. 2). Total functional richness was 87.7, SLP and LLP had partial functional richness of 71.6 and 69.1 respectively (Supplementary Fig. 2), while OGS had a partial functional richness of 60.6. We separately tested the contribution of Fabaceae, the most abundant and species-rich family in Amazonia (16% of all species and individuals). Fabaceae has remarkably high functional richness for a single family (60), compared to all other families combined (68.5; Supplementary Fig. 2). Fabaceae occupies the top of the trait space (Supplementary Fig. 2); it is almost entirely hermaphroditic (Her; 757 of 814 species), contains all Amazonian nitrogen-fixing species (Nfix) in our data (with the exception ofTrema, Ulmaceae), and has on average relatively large seeds (SM)¹⁶. Aluminium accumulation was related to the third axis (Table 1), seed mass and leaf carbon content were also related to this axis. Ecto-mycorrhizal symbiosis was also related to the third axis but only explained 2.4% of the first three axes (Table 1).

Only genera with complete trait data were used (353 genera, of the 812 in our plots). PC1 has an Eigenvalue of 2.98 and represents the leaf economic spectrum (SLA, N, P, C:N). PC2 (Eigenvalue 1.62) represents the stature-recruitment trade-off (WD, SM) and is strongly linked to short lived pioneers (SLP, negatively) and old-growth species and maximum diameter (OGS, Max, positively).Legend: Colours indicate the probability of trait combinations in the trait space defined by the PCA (red = high probability; yellow = low probability). Contour lines indicate 0.99, 0.50, and 0.25 quantiles of the probability distribution. N leaf nitrogen concentration, C:N ratio of leaf carbon to leaf nitrogen, SLA specific leaf area, SM seed mass, P leaf phosphorus concentration, AA aluminium accumulation, Nfix atmospheric N-fixation, WD wood density (overlapping with OGS), C leaf carbon content, FF fleshy fruit, EM ectomycorrhiza, Her hermaphroditic, Nectar nectar producing. Life histories (dark green): OGS old-growth species, LLP long-lived pioneer, SLP short-lived pioneer⁶⁴; Max, maximum diameter¹⁶⁵. For description of the traits and units, see Supplementary box 1.

Full size image

Using only traits generally included in plant functional analyses (wood density, specific leaf area, leaf carbon content, leaf N content, leaf P content, leaf C:N ratio)^2,3,50, resulted in much higher explained variance for PC1 and PC2 of 41% and 19% respectively. As most other traits have eigenvalues close to one or much lower (Supplementary Table 1a) - they are either uncoupled from these two spectra or explain little variance in the data (Table 1).

Traits at tree community level

To identify tree community-level strategies (question 2), a second PCA was carried out using community weighted means (CWM) of 13 traits of the 2054 forest communities with species level identification (Fig. 2). Compared to the genus PCA, the trait loadings of the community PCA appear rotated to the right by 20–40 degrees. The CMW related to the leaf-economic-spectrum are still mostly associated with community PC1, similar to genus-level analysis. However, size and reproductive traits that were mostly associated with genus PC2, the stature-recruitment trade-off, in the analysis of the genera are now mostly weighing on community PC1 (PC1, eigenvalue 4.39, 33.8%, Supplementary Table 2a). Nitrogen fixation changed from a positive to a negative association with community PC2. Other traits mainly linked to PC2 are fleshy fruit, hermaphrodism, and nectar.

PC1 has an Eigenvalue of 4.39 (explained variance 33.3%), and appears to be related to the ‘leaf economic spectrum’ (SLA, N, P, C:N) but also WD, SMC, and hermaphroditism contribute to this axis. Life-history forms SLP and LLP are also positively correlated with PC1. Environmental factors sum of bases and pH are strongly positively correlated to this axis. PC2 is linked to nodulation of Fabaceae and fleshy fruits and poorly correlated to the climatic factors used. Legend: Colours indicate the probability of trait combinations in the trait space defined by the PCA (red = high probability; yellow = low probability). Contour lines indicate 0.99, 0.50, and 0.25 quantiles of the probability distribution. N leaf nitrogen concentration, C:N ratio of leaf carbon to leaf nitrogen, SLA specific leaf area, SM seed mass,P leaf phosphorus concentration, AA aluminium accumulation, Nfix atmospheric N-fixation, WD wood density, C leaf carbon content, FF fleshy fruit, EM ectomycorrhiza, Her hermaphroditic, Nectar nectar producing. Life histories (dark green): OGS old-growth species, LLP long-lived pioneer, SLP short-lived pioneer⁶⁴; Environmental variables: Annual, Annual precipitation (Bioclim12)¹⁶⁶; CWD cumulative water deficit, CAPE Convective atmospheric potential energy⁶⁵, WTC Windthrow count⁶⁵; PZ podzol, white-sand forest, FL flooded (swamp forest; várzea; igapó); pH, soil acidity; SB, log(sum of bases)¹⁵⁴; G.prob, geoglyph probability⁵⁸; DSpp, domesticated species⁵⁷. Note that SLA, N and P are overlapping, as are DSpp, G.prob, pH and SB. For description of the traits and units, see Supplementary box 1.

Full size image

Community PC1 runs from forest communities with ‘fast’, acquisitive traits (SLA, N, P), to the left, to forest communities with ‘slow’, conservative’ traits (C:N, wood density, seed mass) and high percentage of hermaphroditism to the right (Fig. 2, Table 2). Communities with fleshy fruits and nectar have the highest loading on community PC2 (Eigenvalue 2.27, 17.4%). We will use the term ‘fast-slow forest spectrum’ for PC1 of this second PCA of the forest communities. PC2 could be considered and axis of breeding system.

Comparing the trait associations at genus and community level

Question 2

asked if genus-level strategy spectra are similar to community-level strategy spectra. To assess if genera and forest communities show similar trait associations (Figs. 1,2; Supplementary Fig. 3a, b), a Mantel test was carried out over distances of traits in PC space. The Mantel R (0.78,p = 0.001) suggests that a higher distance in genus-level trait values was also associated with a higher distance in plot CMW. This was mainly caused by the scores on the PC1’s of each ordination, however, which were strongly correlated (Supplementary Fig. 3c) and less by the scores on the PC2’s, which were not significantly correlated (Supplementary Fig. 3d).

We evaluated how environmental factors (soil pH, sum of bases, annual rainfall, cumulative water deficit, windthrow count, convective atmospheric potential energy), past human disturbances and management (domesticated species⁵⁷ and geoglyph-probability⁵⁸), as well as life history strategies are associated with community functional composition by a-posteriori plotting them in the PC trait space (question three). Soil explanatory variables (sum of bases, pH) and human impact (domesticated species, geoglyph probability) were mainly related to the ‘fast-slow forest spectrum’, although their explained variance is generally low (Supplementary Table 2c). Soil sum of bases explained most of the variance (33%) of the ‘fast-slow forest spectrum’, while pH explained 21.4% (Supplementary Figs. 4 a, b,5). Domesticated species and geoglyph probability (Supplementary Fig. 4e, f) were also mainly aligned to the ‘fast-slow forest spectrum’ but explained <6% of the variance (Supplementary Table 2c). Climatic factors (annual precipitation, cumulative water deficit, windthrow count, convective atmospheric potential energy, flooding) are the best predictors of PC2, although all have an R² less than 7% (Supplementary Fig. 4c, d, Supplementary Table 2c).

Forest types differ in their functional composition, as indicated by the plot scores for the ‘fast-slow forest spectrum’. Igapó, terra firme from the Guyana Shield, and especially the white sand podzol forests are, on average, ‘slow’ forests (TFBS, IG, TFGS, PZ) with positive scores for the ‘fast-slow forest spectrum’ (Fig. 3a). In contrast, terra firme on the Pebas formation and várzea (TFPB, VA) are, on average, ‘fast’ forests with negative scores for the ‘fast-slow forest spectrum’. Forest types with low nutrient status (TFBS, IG, TFGS, PZ) are also positioned in the right part of the trait space (positive scores for the ‘fast-slow forest spectrum’; Supplementary Figs. 5b,6). Total functional richness of the complete dataset was 130.4. Forests on white sands and swamps had low functional richness (53.2, 55.4), terra firme ranged from 74.3 to 77.9, while the two flooded forest types várzea and igapó both had relatively high functional richness (95.1 and 86.7, respectively). Regions are also ranked from those with generally high soil fertility to those with low soil fertility (Fig. 3b, SWA > GS) and positioned from right (positive scores on the ‘fast-slow forest spectrum’; low soil fertility) to left (negative scores on the ‘fast-slow forest spectrum’; high soil fertility) in trait space (Supplementary Fig. 7). Most regions had identical functional richness ranging from 76.5–79.8, with Northwestern and Central Amazonia having a somewhat higher functional richness (Supplementary Fig. 7).

‘The fast-slow forest spectrum’ is associated mostly with the economic spectra, and the order of forest types appears determined by general soil fertility (see Supplementary Fig. 29a). Note the very high value of the poorest soils in Amazonia (lowest sum of bases (Supplementary Fig. 29a), white sand podzol (PZ).b‘The fast-slow forest spectrum’ as determined by Amazonian region. The order of regions also appears follow general soil fertility (Supplementary Fig 29b). From rich to poor: TFPB terra firme Pebas Formation, VA várzea, SW swamp forest, TFBS terra firme Brazilian Shield, IG igapó, TFGS terra firme Guiana Shield, PZ white sand forest, SWA south west Amazonia, NWA northwest Amazonia, SA southern Amazonia, EA eastern Amazonia, CA central Amazonia, GS Guiana Shield. Colours follow the major forest type (SWA, NWA: TFPB; SA: TFBS; CA, GS: TFGS; EA: mix of TFBS, TFGS). Red dotted line: mean of all data.

Full size image

For the tree communities, the 13 community-weighted mean (CWM) traits showed similar spatial patterns across Amazonia, with values linked to the fast-soft, acquisitive end of the leaf economics spectrum both in the regions of north-western and south-western Amazonia and forest types (TFPB, VA) where relatively high soil-fertility and plant productivity are expected. Values linked to the slow-tough, conservative end were found in the regions (Central Amazonia, Guiana Shield, Southern Amazonia) and forest types (PZ, IG, TFGS, TFBS) with expected low soil fertility and productivity. Each trait is discussed in more detail in the Supplementary text and figures (Supplementary Figs. 8–25).

We mapped the ‘fast-slow forest spectrum’ (Fig. 4). Because the” ‘fast-slow forest spectrum’ is built up from the CWM’s of the 13 traits, many of which correlate well with this axis the patters of the traits are fairly similar to the ‘fast-slow forest spectrum’ (Supplementary Figs 8–25). ‘Slow’ forests (with a positive score on the ‘fast-slow forest spectrum’) make up ~40% of all plots (Supplementary Fig. 26) and are found in areas with low soil fertility, such as the Guiana Shield and central Amazonia (yellow-beige colours in Fig. 4), where also most of the white sand forests are located. ‘Fast’ forests (negative score on the ‘fast-slow forest spectrum’, blue-purple colours in Fig. 4) make up ~30% of all plots (Supplementary Fig. 26) and are found in western Amazonia and southern Amazonia but not the areas directly bordering the Cerrado savanna area (for the delimitation of zones in Amazonia see Supplementary Fig. 1). The pattern of large-scale disturbances is quite similar to rainfall patterns in Amazonia (see Fig. 1 of ref.⁶⁵), and has little effect on trait data and on the ‘fast-slow forest spectrum’ (Supplementary Table 2c). To show that all traits are aligned to the ‘fast-slow forest spectrum’, we carried out an a-posteriori test, dividing the spectrum in three classes (fast < -1.2; medium -1.2 - 0.65, slow > 0.65 [Fig. 4], which have 29%, 31%, and 40% of all plots, respectively, Supplementary Fig. 25) and provide a boxplot for each trait by class. Individual CWM traits follow the same continuum as the ‘fast-slow forest spectrum’, although with different explained variance (R² values, Supplementary Figs 27,28).

Forest with positive score on the ‘fast-slow forest spectrum’ (yellow, beige) are forests at the “slow”, tough side of economic spectra (high CN ratio, low SLA,N andP), high wood density, low numbers of fleshy fruit, high levels of hermaphroditism, high in nectar producing individuals, occurring mainly on low to very low nutrient soils. Forests with negative score on the ‘fast-slow forest spectrum’ (blue, purple) are the opposite in terms of trait values and occur mainly on nutrient rich soils. The isolines divide Amazonia into three regions, tough-slow (PC1 > 0.65, yellow-beige), soft-fast (PC1 < -1.2 blue-purple) and intermediate (green). Colouring the plots based on their PC1 scores shows that their colour mostly matches the area colour, except if they are white sand plots (PZ) in a green area, and várzea plots (blue dots) in green and yellow areas. Note that the legend has been truncated at 2 standard deviations. Red polygon: Amazonian Biome limit¹⁶⁷. Base map source (country.shp, rivers.shp), ESRI (http://www.esri.com/data/basemaps, © Esri, DeLorme Publishing Company).

Full size image

Community functional composition affects ecosystem functioning

Finaly, question four asked if functional composition had consequences for forest functioning. This is expected as the ‘fast-slow forest spectrum’ is strongly associated with soil fertility (Supplementary Fig. 4a,5, Supplementary Table 2c, R2 = 32%,P < < 0.001). Indeed, above ground woody biomass (AGB) is significantly positively related to the ‘fast-slow forest spectrum’ (Fig. 5a). Forests with ‘fast’ traits have low biomass and those with ‘slow’ traits have high biomass. Absolute aboveground woody productivity (AGWP) is not significantly related to the ‘fast-slow forest spectrum’, suggesting that all forests have a similar, though variable, absolute productivity (Fig. 5b). Consequently, forests with high biomass have low relative AGWP (AGWP / AGB, Fig. 5c), whereas forests with a low biomass have a high relatively AGWP. Thus, relative AGWP is higher for forests with ‘fast’ trait values. Relative AGWP also increases with soil fertility (sum of bases, Fig. 5d). While the direct contribution of the ‘fast-slow forest spectrum’ and sum of bases to the AGWP is 33% and 27% explained variance, respectively, their combined contribution is 34% explained variance. Thus, their contribution is largely coinciding, strengthening the notion that fertility may be an underlying driver of both trait composition (PC2, the fast-slow forest spectrum), and productivity.

a ‘Slow’ forests (positive value) have much higher above ground woody biomass (AWB) than ‘fast’ forests (negative values)b Absolute above ground woody productivity (AGWP) does not vary significantly with the‘fast-slow forest spectrum’.c Biomass produced per biomass standing (= Relative AGWP [100*AGWP/AGB]) is highest in ‘fast’ forests (negative values for slow-fast forest spectrum).d Relative AGWP is positively correlated with predicted sum of bases¹⁶. Red lines indicate 95% confidence intervals. Biomass data from sources^55,83,168. Colours: Red, terra firme Pebas formation; brown, terra firme Brazilian Shield; orange, terra firme Guiana Shield; yellow, white sand forest; purple, swamp forest; light blue, várzea.

Full size image

Amazonian trees show two main strategy spectra

Across the globe, two main plant strategy spectra are found related to (1) plant size¹² and (2) leaf economics³. For Amazonian tree genera, the same two strategy spectra are found, but the order is reversed; the leaf economics spectrum (LES) is the spectrum describing most of the trait variance, probably reflecting adaptations to the strong Amazonian soil fertility gradient (see below). The size spectrum is only secondary (Fig. 1), presumably as we focus here solely on the tree life form compared to global analyses^3,6 that included many small herbaceous plant life forms. Across Amazonian tree genera, the wood economic spectrum (WD) was uncoupled from the LES (Fig. 1), as previously shown for Amazonian tree species^50,51,66. This suggests that leaves and stems provide independent avenues for specialisation, potentially leading to more opportunities for niche differentiation and species coexistence.

We expand on previous analyses by showing that, even within trees, reproductive characteristics (breeding system, fleshy fruits, seed mass) are closely related to the size spectrum indicating that plant lifespan (tree size) and reproductive strategies are closely intertwined. Life history strategies were mainly related to the size-reproductive spectrum, in which small, short-lived pioneers produce many small animal-dispersed seeds to colonise ephemeral canopy gaps, whereas tall, long-lived old-growth species with durable wood (high wood density) produce large seeds which enables their seedlings to establish and survive successfully in the shade⁴⁷. Long-lived pioneers lie somewhere in between trait-wise. The second axis reflects therefore the stature−recruitment trade-off which is often found in closed-canopy forests^11,12,13, where taller species have better access to light and smaller species have relatively high seed production and fast life cycle. It should be noted that long-lived pioneers and especially old growth species that produce large seeds generally have higher total seed mass production per fruiting event⁶⁷. As they also live much longer they may thus have greater life-time seed production than short-lived pioneers⁶⁸.

Despite the global importance attributed to the LES, Amazonian pioneers and old growth species, surprisingly, do not differ much in their position on the LES. LES traits may be more important for the growth and survival of small seedlings and saplings that have a small total leaf area^41,69, compared to adult trees in which carbon gain is more determined by their large size and total leaf area than by leaf-level trait differences^2,70.

Plant strategies only partly translate into community strategies

Trait associations scale up from genera (Fig.1) to communities (Fig. 2) but not perfectly (Supplementary Fig. 3) and most traits are more strongly related to the first PCA axis in the communities, the ‘fast-slow forest spectrum’. At the community level, the LES traits, size and reproductive traits are all aligned with the first principal component (Supplementary Table 2c), resulting in one overall spectrum from ‘fast’ to ‘slow’ Amazonian forests, which closely parallels the soil fertility gradient (see below). For example, ‘slow’ forests on infertile soils tend to be tall, evergreen, densely shaded with low turnover dynamics and infrequent tree-fall gaps⁷¹. Under those conditions, high seed mass facilitates seedling establishment and survival^{16,20,61,72,73}. Nutrient-poor conditions may select for species with dry fruits that tend to have low nutrient concentrations, high seed toxicity, and for hermaphroditic species that maximise fitness¹. In low turn-over forests, tree species do not produce many small seeds but rather few large seeds, providing offspring with higher survival, a classic example of the “high growth in light vs. low mortality in shade trade off”^67,74. In sum, the two plant strategy axes converge into one main community strategy axis because of strong environmental filtering by soils. This may explain why the pioneer-climax dichotomy⁴⁷ has been so appealing for such a long time.

Nearly all of the dry-fruited trees in the Amazon are hermaphroditic and, because wind- or unassisted-dispersal is not favoured in the subcanopies of dense forests⁷⁵, which tend to be also tall. Heavy seed mass also tends to be associated with larger trees⁷⁶. Very little is known on the nectar producing species in Amazonia but it appears positively associated with infertile soils and hermaphroditism. The link with infertile soils is most likely due to the fact that under conditions of high solar energy and abundant moisture but low soil nutrients, production of carbohydrate-rich exudates is favoured⁷⁷. Flowers producing abundant nectar also tend to be large⁷⁸, rarely unisexual, but associated with hermaphroditic breeding systems. In contrast, wind-pollinated species produce large amounts of nitrogen-rich pollen⁷⁹, no nectar, and have mainly unisexual flowers.

Atmospheric N fixation was positively linked to the size-recruitment spectrum in the genus ordination (Fig. 1). Species in Fabaceae, the main N-fixers, are characterised by high wood density and large seeds. However, in the community ordination their position was reversed from a positive (Fig. 1) to negative relation (Fig. 2). Fabaceae, dominate the forests of the upper Rio Negro, Guyana and Suriname¹⁶, but the species that dominate there are ‘caesalpinoid’ legumes that generally do not form N-fixing root nodules¹⁶. N-fixation is mainly found in the genera occurring in western Amazonia, which also have smaller seeds¹⁶, which explains the reversal.

Areas along the Amazon main stem and other várzea rivers have negative scores for the ‘fast-slow forest spectrum’ and are known to be very fertile (see also Supplementary Fig. 29), having among the highest litter productivity of Amazonia⁸⁰. It should be noted that the most fertile soils are also associated with regions of greatest soil instability^81,82, seasonal flooding (várzea), and in southern Amazonia with incidence of storms⁸³, making it difficult to disentangle effects of disturbance and soil fertility. The intermediate disturbance theory^84,85 has long held that in Amazonia, higher soil fertility would lead to faster tree growth and turn-over, gap dynamics, and heterogenous forest structure, ultimately yielding higher plant diversity^16,61,86. Other studies have countered this conclusion⁸⁷. Our data suggests that tree species richness has no relationship with the ‘fast-slow forest spectrum’ and also explains very little variance of the trait distribution (Supplementary Table 2c). In Amazonia, even though large windfall disturbances (from 5 to over 2000 ha) are not uncommon, their return frequency is between 27.000 years in Western Amazonia and 90.000 years in Eastern Amazonia⁸⁸. Thus, it is unlikely that they contribute much to disturbance related species richness.

An Amazonian spectrum from slow to fast forests, driven by soil fertility

Amazonian forests show one major functional spectrum, running from ‘fast’ productive forest communities with high mean SLA, N, P, and fleshy fruits to ‘slow’ conservative forest communities with high C:N, wood density, seed mass, and high percentage of hermaphroditism (Fig. 2). This spectrum is best explained by soil fertility (sum of bases, light vs. lowmortality in shade pH; see Supplementary Figs. 4,5), as has been suggested before for forest species and trait composition^16,62, but surprisingly little by macroclimate (annual rainfall, climatic water deficit, and large-scale disturbance, Supplementary Table 2c, Supplementary Fig. 4). It was previously shown that species from communities of fertile soils have higher SLA and leaf nutrient concentrations than those from infertile soils and the sum of bases and pH explain respectively 30% and 18% of the trait variance⁹. A global study encompassing all biomes, ranging from grasslands to forests, found two main axes of community trait variance (i.e., plant stature and resource economics) that were only weakly associated with climate and soil conditions¹¹. Functional composition of Amazonian forests is not driven by precipitation, possibly because all forest sites receive sufficient rainfall (>1800 mm/yr). Instead, functional composition and resource economics are strongly driven by soil fertility, as there is a major soil gradient running from the old weathered extremely nutrient poor soils from the Guiana Shield and the Brazilian Shield in the east, to the young and fertile soils formed by more recent Andean sediments⁸⁹. This gradient drives strong assembly rules, sensu Keddy⁹⁰, arguably driven by soil characteristics in Amazonia⁸². We see strong convergence^{9,61,62,91,92} of almost all traits when comparing low-productivity communities on poor soils to those with higher productivity and higher soil fertility. While soil fertility (total soil phosphorus [strongly related to sum of bases⁹]) was shown to be a strong driver of productivity in Amazonia, soil physical properties appear more important for forest turn-over⁸², which is twice as high in western Amazonia compare to central and eastern Amazonia⁸¹.

Although we did not include deciduousness in our analyses, it has been recently shown that increases in abundance of deciduous species is tightly linked to soil fertility and water availability^93,94. Additionally, community leaf nutrients increase towards wetter forests on younger fertile soils in the western fringes of Amazonia^9,62,82,95 (Supplementary Figs. 11,12).

Studies comparing nutrient-poor igapó and nutrient-rich várzea forests showed that within genera similar results were found, with traits conferring a ‘fast’ lifestyle being more common in fertile várzea and those with a more ‘slow’ lifestyle were more common in infertile igapó^91,92. Comparing congeneric species between terra-firme forest on clay soils and white-sand forest⁹⁶, the same result was found. Thus, it is likely that if we could have measured actual trait expression everywhere, the large-scale gradients would be reinforced.

Human legacies

We assessed to what extent the current functional composition of the Amazon is influenced by human legacies. Communities with ‘fast’ traits are significantly associated with the abundance of domesticated species (explaining 5.6% of the trait variance, Supplementary Table 2c) and geoglyph probability (explaining 5%, Supplementary Fig. 4, Supplementary Table 2c). This suggests that indigenous people may have domesticated faster growing species, and that long-term human presence and disturbance (open areas) may still have left its mark on the current vegetation^57,58. The higher soil fertility (sum of bases) and access from the open Cerrado could be one of the reasons that pre-Columbian people settled the edges of Amazonia. At the Amazonian scale, areas with naturally higher soil fertility may have facilitated past human occupation by increasing productivity of agroforestry systems⁹⁷. For instance, most domesticated tree and palm species benefit from fertile soils, and by contributing to enrich soil fertility through soil management practices, pre-Columbian people allowed domesticated species to persist in the forest over centuries⁶⁰. Although the effect of anthropogenic soil enrichment on domesticated species likely plays a role at the landscape scale, depending on the extent of landscape transformations by pre-Columbian peoples, soil enrichment could potentially influence tree communities over broader scales. Therefore, part of the functional variation we observe across Amazonian tree communities could still be a legacy of pre-Columbian landscape domestication.

Community functional composition affects ecosystem functioning

Functional composition of Amazonian forests has consequences for ecosystem functioning. While the relationship between soil physical and chemical properties are not always clear⁸², above ground woody biomass (AGB) is significantly, positively related to the ‘fast-slow forest spectrum’ (Fig. 5a) – indicating that ‘slow’ forests with conservative trait values and high wood density have high aboveground biomass⁹⁸. Our map of the ‘fast-slow forest spectrum’ (Fig. 4) is indeed similar to an earlier ground-based biomass map⁹⁹. Forest productivity is influenced by tree traits, frequency of disturbance and soil fertility (Fig. 2, Supplementary Table 2c). However, absolute aboveground woody productivity (AGWP) is not significantly related to the ‘fast-slow forest spectrum’ (Fig. 5b). Forests with high biomass have low relative biomass productivity (Fig. 5a, c), probably because a large proportion of the biomass is locked up in unproductive stems¹⁰⁰, whereas forests with low biomass have a higher biomass productivity, probably because of a higher light availability within the stand, and because a larger proportion of the biomass is in photosynthesising leaves¹⁰⁰. Relative biomass productivity (aboveground woody productivity/aboveground woody biomass) is higher for forests with faster traits that produce a higher amount of woody biomass per standing biomass, and this effect is correlated with ‘fast’ trait values (Fig. 5), also increasing with soil fertility (sum of bases, Fig. 5d). Because soil fertility is a driver of both biomass productivity and the main explanatory variable for the ‘fast-slow forest spectrum’, soil fertility is likely the driver of forest productivity by both influencing the community traits and allowing higher growth rates directly. It has been predicted that a positive relationship exists between forest biomass and productivity¹⁰¹. However, forests with high productivity tend to have both high turn-over^55,63, and low wood density¹⁷, making this relationship more complex. We found no difference in net biomass productivity between the various forest types along the fast-slow forest spectrum but rather a high variability (Fig. 5b). Forests on poor soils tend to have high biomass but limited growth, while forests on rich soils have less biomass but higher relative growth. This does not lead to higher biomass, because of the lower wood density and much higher turn-over of the forests on rich soils⁵⁵ (see also Supplementary Fig. 30). The ‘slow-fast forest spectrum’ should also have consequences for other trophic levels. ‘Slow’ forests combine a slow growth with poor food quality as they have tough, well-defended, nutrient-poor leaves, few fleshy fruits, and large, often toxic seeds. Combined, this results in less food for animal life (e.g. less insects, insectivores, and frugivores). Conversely, ‘fast’ forests faster growth producing higher quality food sources (e.g. thinner leaves with lower C:N ratio, more fleshy fruit), resulting in a higher biomass of insects, mammals and birds¹⁰².

Three functionally different Amazonian forest types

Based on the ordination analysis of 13 tree traits, Amazonia can broadly be divided into three regions with a different functional composition (Fig. 4). The very poor soils on the sandy deposits of the Roraima table mountains and the poor soils of the Guiana Shield, and the forests on white sands across other regions of Amazonia form one group. Forests that are part of this group generally have low diversity tree communities, except for the areas in central Amazonia with very high diversity¹⁰³. This result strongly contrasts with our earlier notion that forest productivity/turnover and diversity are strongly positively linked^16,104. The ‘slow’ forests are composed of mainly hermaphroditic species with tough, low palatability, low nutrient leaves with high C:N ratio, dense wood, dry fruit, and high levels of endemism^103,105. Western and southern Amazonia are the ‘faster’ forest areas that select for the opposite trait characteristics than those mentioned above. Compared to the other two regions they are generally found on richer soils (Western Amazonia), drier areas (Southern Amazonia) and in várzea forests in the other two regions. They have high (Western Amazonia) to medium diversity (várzea forest)¹⁰³. They are also characterised by high productivity¹⁵ and high turnover^71,81.

Because the three forest functional types are based on tree traits with a strong influence on forest functioning, our map could be included in dynamic vegetation models¹⁰⁶ and earth system models¹⁰⁷, thus making better predictions on the role of Amazonia in global carbon and water cycling¹⁰⁸, the risk of tipping points¹⁰⁹, and the fate of the Amazon in the face of global change¹¹⁰. Because of the reliability of species identification, and lack of species-specific trait data, our current analysis and maps are based on average, genus-level data. When more data becomes available, the functional maps could be improved by including species-level trait values and hence accounting for interspecific (and perhaps intraspecific) trait variance.

Tree inventory data were taken from the May 2024 version of the Amazon Tree Diversity Network inventory data^{111,112,113,114}. ATDN20240517 contains 2253genus-level plots (with 1,198,408 individuals, 812 genera, 98.5% of all individuals identified at genus level), 2054 of which withspecies composition (thespecies-level plots, 1,010,524 individuals, 5211 species, 88% identified at species-level). Most of tree-inventories were for 1-ha size plots and sampled trees with a diameter at breast height (DBH, at 1.30 m or above tabular roots) over 10 cm (for plot metadata, see Appendix 1). Species synonymy was updated following ref.¹¹⁵, but harmonising names with the World Flora Online¹¹⁶, using the December 2023 snapshot theWorldFlora R package¹¹⁷, with some modifications after Molino et al.¹¹⁸.

Species with aconfer (cf.) identification were accepted as belonging to the named species, while those withaffinis (aff.) were accepted only at the genus level and therefore removed from the species analysis.

The 2253 genus-level plots (Supplementary Fig. 1) provided a total of 1,216,222 trees, of which 1,198,408 (98.5%) were identified at the genus level. Most plots (2153) had more than 90% of their individuals identified to genus (Supplementary Fig. 30). A total of 812 genera were recorded, of whichEschweilera (61,061 individuals),Protium (56,943),Pouteria (51.777),Inga (27,619), andOenocarpus (22,907) were the five most abundant genera across all plots. Thirty-five genera made up 50% of all individuals and could be considered hyperdominant Amazonian tree genera^111,112. A total of 149 genera had 10 individuals or less, while 42 genera had only one individual. The percentage of individuals with trait data ranged from 94-97% (leaf traits), through 99% (wood, seed) to 100% (root traits, fruit fleshiness, breeding system). For a list of all traits, their units and ecological information see Supplementary Box 1.

Most of our analyses were carried out at the genus level because over such a large and species rich region trees are more reliably identified at the genus level (Supplementary Fig. 31), and because for many species there is a lack of species-specific trait data. For several traits it has been shown that traits are phylogenetically conserved and most trait-level variance is found above the species level, as has been found for wood density^{62,66,119,120}, seed mass^121,122, and SLA⁹. We used the average of the trait data for all species within a genus, except for breeding system, which may vary largely within a genus and which was analysed at species level. Our analyses and maps do therefore not consider different species distributions within genera or variance of trait values within species due to plasticity and/or acclimation. For the traits included in our analysis, in Amazonia, SLA, N and C, are most determined by species identity, whereas leaf P is also strongly influenced by site growing conditions⁹.

Traits were obtained from a number of sources. Wood density was mainly taken from^4,119. Leaf traits were mainly from four large TRY datasets^{9,14,50,51,66,123,124,125,126,127}, with additional data from^{128,129,130,131,132,133,134}. Seed mass was taken from^22,135,136 and various floras and tree guides^{137,138,139,140,141}. Because seed mass varies over several orders of magnitude, we used logarithmic classes for seed mass^22,61. For EM association we checked the most recent literature for confirmed EM tree species¹⁴². For nodulation we used^143,144. For aluminium accumulation we used^32,145 and references therein. We considered a genus EM positive, nodulating or Al-accumulating if more than 50% of the species in that genus reported were positive for that trait. Nectar production was taken from⁵² and mapped as a percentage by taxon. We first scored the percentage of species by genus and, if not available, we used the information by family. Breeding system may vary considerable in some genera and was taken at species level from¹⁴⁶ and descriptions from floras and monographs (in particular,Flora e Funga do Brasil). Jardim Botânico do Rio de Janeiro (http://floradobrasil.jbrj.gov.br/) issues ofFlora Neotropica (https://www.springer.com/series/16365); and the Springer book seriesThe Families and Genera of Vascular Plants^147,148,149 and other published revisions. We did not include adult tree height in our data, due to a lack of data for almost all genera.

We performed a principal component analysis (PCA) on the average trait values for all genera that had data for all traits (353 genera), scaling all data to a mean of zero and standard deviation of 1. While this is less than 50% of all genera, these 353 genera amounted to 90.8% of all individuals in our plots. While for several genera data is missing for particular traits, the percentage of individuals with trait data ranged from 94-97% (leaf traits), through 99% (wood, seed) to 100% (breeding system, root traits, fleshiness of fruits). Because of these high percentages we did not conduct data imputation. For all plots (communities) we calculated the community weighted mean of each trait, by calculating the average over all individuals of known taxonomy, thus using data of all genera. For discrete yes/no traits we used the percentage of individuals, rather than the mean.

The forest plots are subdivided in those that occur on floodplains (várzea (VA) and igapó (IG)), white sand podzols (PZ), terra firme (TF) and swamps (SW). For these four categories we constructed a separate spatial model of each trait across Amazonia with inverse distance weighting¹⁰³. As an example, for all white sand plots and wood density we made a spatial interpolation. This interpolation was then used to predict the mean trait value for each pixel on the soil map that was considered a white sand area (Supplementary Fig. 32b, yellow pixels). The same was done for all plots in várzea and igapó combined, all plots established on terra firme and finally for swamp forests. The forest map (0.1 degree resolution, Supplementary Fig. 32a) was based on the Amazon lowland forest^112,150, divided in the major soils corresponding to the forest-soil combinations used^111,151 (Supplementary Fig. 31b). While the soil grid was based on the major soil type, the soil type of the plots was determined independently of this map and based on the field observation of the person that established the plot. It is thus possible that a plot on white sand is located in a grid cell classified as terra firme. Even so, it is used in the white sand spatial model (see ref.¹⁰³ for a more detailed explanation). For all maps we truncated the legend and its colours to values between the mean ± 2 times the standard deviation, to avoid that outliers in 5% the data would influence the visible pattern too much.

We calculated the percentage of variance explained by the model by combining the observed and predicted community weighted mean of all four spatial models, using a simple linear model¹⁰³.

Annual rainfall was extracted by plot location from the grid data from Worldclim 2¹⁵². The cumulative water deficit (CWD) was calculated as¹⁵³ and can be considered a parameter of the strength of the dry season. Soil fertility (sum of bases, SB) was extracted from the latest Amazonia wide map¹⁵⁴. We used SB rather than the often-used CEC (cation exchange capacity), as the latter includes the full exchange complex, which on acid tropical soils often includes a large portion of Al³⁺ and H⁺, which are in fact toxic for most species. Although we used the most recent soil-fertility map¹⁵⁴, the overwhelming predominance of soil data from terra firme sites resulted in an artificially high interpolated SB for white sand forests and low SB for Várzea forest (Supplementary Fig. 29). We may thus expect stronger relationship between functional composition, SB, and other soil variables when improved soil maps become available. Soil acidity (pH) is also an often-used index of soil fertility (a low pH being infertile). We extracted pH data from Soterlac¹⁵⁵, ISRIC wise¹⁵⁶, RAINFOR sites^95,151, and refs.^157,158,159. For pH, we created a loess interpolation model, based on all data available. We then estimated pH for each plot based on the loess interpolation, sensu¹⁰³. Interpolated maps of SB and pH and boxplots for SB and pH based on plot data (sources as above) can be found in Supplementary Fig. 33. Large-scale disturbance was assessed in two ways: the density of large wind throws (5 – 2,223 ha; mapped at 0.25 degree resolution) caused by convective storms found on satellite images^65,88, and a map of convective available potential energy (CAPE), which is a strong driver of convective blowdowns⁶⁵.

We also calculated by plot four life-history characteristics: the fraction of short-lived pioneers (SLP); long-lived pioneers (LLP); old growth species (OGS)(Forestplots.net), and maximum observed diameter. Pioneers are defined after⁶⁴, by combining low wood density and low seed mass (wood density < 0.7 g/cm³), where SLP have seed mass < 0.1 g and LLP have a seed mass >= 0.1 g, and OGS species have a wood density > 0.7 g/cm³.

Domesticated species (Dsp) were taken from⁵⁷, we used the percentage of domesticated species per plot as a proxy of pre-Columbian legacy on the forest. Similarly, we used the probability of finding geoglyphs⁵⁸ as a second proxy of pre-Columbian influence on the forest.

Species richness/ha was calculated as in¹⁰³.

All analyses were carried out in the R programming environment, with custom made R¹⁶⁰ scripts, using the librariesFunspace¹⁶¹ (for PCA and functional space analyses and images),Vegan¹⁶² (Mantel test), andRaster¹⁶³.

Statistics and reproducibility

Statistic used are as described above.P-values for regression (Fig. 5) and ANOVA (Fig. 3) are calculated with standard linear models. Reproducibility was maintained by use of versioned scripts.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

All data necessary for producing the results reported here have been deposited on Figshare¹⁶⁴. At Figshare we also provide a spatial model (at the scale of 0.1 degree) for each trait, a high-resolution map of the slow-fast-forest spectrum (Fig.4), and plot-based community weighted averages for further research. Correspondence and requests for other materials, which is available upon reasonable request and following a ATDN data sharing agreement, should be addressed to Hans ter Steege.

R code (version 4.3.1) and data to produce the figures and tables have been deposited on Figshare¹⁶⁴.

This paper is the result of the work of hundreds of different scientists and research institutions in the Amazon over the past 80 years. Without their hard work this analysis would have been impossible. We especially thank here Dairon Cárdenas, Cid Ferreira, and Nállarett Dávila, who passed away during the preparation of this work. We thank Charles Zartman and Joost Duivenvoorden for the use of plots from Jutai and Araraquara, respectively. HtS, VFG, and RS were supported by grant 407232/2013-3 - PVE - MEC/MCTI/CAPES/CNPq/FAPs; PIP had support for this work from CNPq (productivity grant 310885/2017-5) and FAPESP (research grant #09/53413-5); CB was supported by grant FAPESP 95/3058-0 - CRS 068/96 WWF Brasil - The Body Shop; CF, DS, JFM, JE, PP and JC benefited from an “Investissement d’Avenir” grant managed by the Agence Nationale de la Recherche (CEBA: ANR-10-LABX-25-01); HLQ/MAP/JLLM received financial supported by MCT/CNPq/CT-INFRA/GEOMA #550373/2010-1 and # 457515/2012-0, and JLLM were supported by grant CAPES/PDSE # 88881.135761/2016-01 and CAPES/Fapespa #1530801; The Brazilian National Research Council (CNPq) provided a productivity grant to EMV (Grant 308040/2017-1); Floristic identification in plots in the RAINFOR forest monitoring network has been supported by the Natural Environment Research Council (grants NE/B503384/1, NE/ D01025X/1, NE/I02982X/1, NE/F005806/1, NE/D005590/1 and NE/I028122/1) and the Gordon and Betty Moore Foundation; BMF is funded by FAPESP grant 2016/25086-3. BSM, BHMJ and OLP were supported by grants CNPq/CAPES/FAPS/BC-Newton Fund #441244/2016-5 and FAPEMAT/0589267/2016; TWH was funded by National Science Foundation grant DEB-1556338. BGL was supported by FAPESP, grant #2015/24554–0, and #2019/03379-4. WEM: Plots in the PPBio system were financed by the INCT for Amazonian Biodiversity (CENBAM); the Program for Biodiversity Research in Western Amazonia (PPBio-AmOc) and a Productivity Grant (PQ - 301873/2016-0). JAG was funded by the Natural Environment Research Council (NERC; NE/T011084/1) and the Oxford University John Fell Fund (10667). This project has been supported by ForestPlots.net Project 187. Biogeography of the Amazonian Tree Flora. We thank Rodrigo Bergamin for help with the R funspace library and Sylvia Mota de Oliveira for final proofreading and valuable comments.

Authors and Affiliations

1. Naturalis Biodiversity Center, PO Box 9517, Leiden, 2300 RA, The Netherlands

   Hans ter Steege & Tinde R. van Andel

2. Quantitative Biodiversity Dynamics, Utrecht University, Padualaan 8, Utrecht, 3584 CH, The Netherlands

   Hans ter Steege & Edwin Pos

3. Forest Ecology and Forest Management Group, Wageningen University & Research, Droevendaalsesteeg 3, Wageningen, P.O. Box 47, 6700 AA, The Netherlands

   Lourens Poorter & Masha van der Sande

4. Environmental Change Institute, School of Geography and the Environment, University of Oxford, South Parks Road, Oxford, OX1 3QY, UK

   Jesús Aguirre-Gutiérrez, Erika Berenguer & Yadvinder Malhi

5. Leverhulme Centre for Nature Recovery, University of Oxford, Oxford, OX13QY, UK

   Jesús Aguirre-Gutiérrez

6. AMAP (botAnique et Modélisation de l’Architecture des Plantes et des végétations), Université de Montpellier, CIRAD, CNRS, INRAE, IRD, Montpellier, F-34398, France

   Claire Fortunel, Daniel Sabatier, Jean-François Molino, Julien Engel & Émile Fonty

7. Coordenação de Pesquisas em Ecologia, Instituto Nacional de Pesquisas da Amazônia - INPA, Av. André Araújo, 2936, Petrópolis, Manaus, AM, 69067-375, Brazil

   William E. Magnusson, Flávia R. C. Costa, Alberto Vicentini, Fernanda Antunes Carvalho & Fernanda Coelho de Souza

8. School of Geography, University of Leeds, Woodhouse Lane, Leeds, LS2 9JT, UK

   Oliver L. Phillips, Tim R. Baker, Ted R. Feldpausch, Roel Brienen, Fernanda Coelho de Souza, David Galbraith, Aurora Levesley & Georgia Pickavance

9. Utrecht University Botanic Gardens, P.O. Box 80162, Utrecht, 3508 TD, The Netherlands

   Edwin Pos

10. Departamento de Biologia Vegetal, Instituto de Biologia, Universidade Estadual de Campinas – UNICAMP, CP 6109, Campinas, SP, 13083-970, Brazil

    Bruno Garcia Luize

11. International Center for Tropical Botany (ICTB) Department of Biological Sciences, Florida International University, 11200 SW 8th Street, OE 243, Miami, FL, 33199, USA

    Chris Baraloto & Julien Engel

12. Grupo de Investigación en Ecología y Evolución en los Trópicos-EETrop, Universidad de las Américas, Quito, 170124, Ecuador

    Juan Ernesto Guevara & María-José Endara

13. Department of Ecology and Evolutionary Biology, University of Michigan, Ann Arbor, MI, 48109, USA

    Maria Natalia Umaña

14. Coordenação de Biodiversidade, Instituto Nacional de Pesquisas da Amazônia - INPA, Av. André Araújo, 2936, Petrópolis, Manaus, AM, 69067-375, Brazil

    Maihyra Marina Pombo, Iêda Leão do Amaral, Luiz de Souza Coelho, Francisca Dionízia de Almeida Matos, Diógenes de Andrade Lima Filho, Juan David Cardenas Revilla, Mariana Victória Irume, Maria Pires Martins, José Ferreira Ramos, Juan Carlos Montero, Charles Eugene Zartman, Lorena Maniguaje Rincón, Juliana Schietti, Henrique Eduardo Mendonça Nascimento, Rogerio Gribel, Marcelo Petratti Pansonato, Marcelo Petratti Pansonato, Edelcilio Marques Barbosa, Luiz Carlos de Matos Bonates & Ires Paula de Andrade Miranda

15. Manaaki Whenua – Landcare Research, PO Box 69040, Lincoln, 7640, New Zealand

    Matt McGlone

16. Department of Geography and Planning, University of Liverpool, Liverpool, L69 3BX, UK

    Freddie C. Draper

17. Wetland Department, Institute of Geography and Geoecology, Karlsruhe Institute of Technology - KIT, Josefstr.1, Rastatt, D-76437, Germany

    Florian Wittmann, Adriano Costa Quaresma, Flávia Machado Durgante & John Ethan Householder

18. Ecology, Monitoring and Sustainable Use of Wetlands (MAUA), Instituto Nacional de Pesquisas da Amazônia - INPA, Av. André Araújo, 2936, Petrópolis, Manaus, AM, 69067-375, Brazil

    Florian Wittmann, Maria Teresa Fernandez Piedade, Layon O. Demarchi, Jochen Schöngart, Adriano Costa Quaresma, Flávia Machado Durgante, Gisele Biem Mori, Aline Lopes, Bianca Weiss Albuquerque & Maira Rocha

19. Programa Professor Visitante Nacional Sênior na Amazônia - CAPES, Universidade Federal Rural da Amazônia, Av. Perimetral, s/n, Belém, PA, Brazil

    Rafael P. Salomão

20. Coordenação de Botânica, Museu Paraense Emílio Goeldi, Av. Magalhães Barata 376, C.P. 399, Belém, PA, 66040-170, Brazil

    Rafael P. Salomão, Dário Dantas do Amaral, Leandro Valle Ferreira & Ima Célia Guimarães Vieira

21. Centro de Pesquisa Agroflorestal de Roraima, Embrapa Roraima, BR 174, km 8 – Distrito Industrial, Boa Vista, RR, 69301-970, Brazil

    Carolina V. Castilho

22. Departamento de Botânica, Instituto de Pesquisas Científicas e Tecnológicas do Amapá - IEPA, Rodovia JK, Km 10, Campus do IEPA da Fazendinha, Macapá, AP, 68901-025, Brazil

    Marcelo de Jesus Veiga Carim

23. Amcel Amapá Florestal e Celulose S.A, Rua Claudio Lucio - S/N, Novo Horizonte, Santana, AP, 68927-003, Brazil

    José Renan da Silva Guimarães

24. Catalogue of Life, Darwinweg 2, Leiden, 2333 CR, The Netherlands

    Olaf S. Bánki

25. Collections, Conservation and Research, The Field Museum, 1400 S. Lake Shore Drive, Chicago, IL, 60605-2496, USA

    Nigel C. A. Pitman, Luis Torres Montenegro & Corine Vriesendorp

26. School of Environmental Sciences, University of East Anglia, Norwich, NR4 7TJ, UK

    Carlos A. Peres

27. ICNHS, Federal University of Mato Grosso, Av. Alexandre Ferronato 1200, Setor Industrial, Sinop, MT, 78.557-267, Brazil

    Domingos de Jesus Rodrigues, Flávia Rodrigues Barbosa, Rainiellen de Sá Carpanedo & Janaína Costa Noronha

28. Institute of Science and Environment, University of Cumbria, Ambleside, Cumbria, LA22 9BB, UK

    Joseph E. Hawes

29. ICNHS, Universidade Federal de Mato Grosso, Av. Alexandre Ferronato, 1200, Sinop, MT, 78557-267, Brazil

    Everton José Almeida, Luciane Ferreira Barbosa, Larissa Cavalheiro & Márcia Cléia Vilela dos Santos

30. Divisao de Sensoriamento Remoto – DSR, Instituto Nacional de Pesquisas Espaciais – INPE, Av. dos Astronautas, 1758, Jardim da Granja, São José dos Campos, SP, 12227-010, Brazil

    Evlyn Márcia Moraes de Leão Novo

31. Herbario Vargas, Universidad Nacional de San Antonio Abad del Cusco, Avenida de la Cultura, Nro 733, Cusco, Cuzco, Peru

    Percy Núñez Vargas, Abel Monteagudo Mendoza, William Nauray Huari & William Farfan-Rios

32. Biological and Environmental Sciences, University of Stirling, Stirling, FK9 4LA, UK

    Thiago Sanna Freire Silva

33. Centro de Biociências, Departamento de Ecologia, Universidade Federal do Rio Grande do Norte, Av. Senador Salgado Filho, 3000, Natal, RN, 59072-970, Brazil

    Eduardo Martins Venticinque

34. Departamento de Biologia, Universidade Federal de Rondônia, Rodovia BR 364 s/n Km 9,5 - Sentido Acre, Unir, Porto Velho, RO, 76.824-027, Brazil

    Angelo Gilberto Manzatto

35. Programa de Pós- Graduação em Biodiversidade e Biotecnologia PPG- Bionorte, Universidade Federal de Rondônia, Campus Porto Velho Km 9, 5 bairro Rural, Porto Velho, RO, 76.824-027, Brazil

    Neidiane Farias Costa Reis, Katia Regina Casula, Susamar Pansini & Adeilza Felipe Sampaio

36. Department of Biology and Florida Museum of Natural History, University of Florida, Gainesville, FL, 32611, USA

    John Terborgh

37. Centre for Tropical Environmental and Sustainability Science and College of Science and Engineering, James Cook University, Cairns, Queensland, 4870, Australia

    John Terborgh, Susan G. W. Laurance & William F. Laurance

38. Department for Accelerated Taxonomy, Royal Botanic Gardens, Kew, Richmond, Surrey, TW9 3AE, UK

    Euridice N. Honorio Coronado & Bente Klitgaard

39. Jardín Botánico de Missouri, Oxapampa, Pasco, Peru

    Abel Monteagudo Mendoza, Rodolfo Vasquez & Luis Valenzuela Gamarra

40. Instituto Boliviano de Investigacion Forestal, Av. 6 de agosto #28, Km. 14, Doble via La Guardia, Casilla, 6204, Santa Cruz, Santa Cruz, Bolivia

    Juan Carlos Montero & Juan Carlos Licona

41. Embrapa Amazônia Ocidental, Manaus, AM, Brazil

    Cintia Rodrigues De Souza & Marcus Vinicio Neves de Oliveira

42. Geography, College of Life and Environmental Sciences, University of Exeter, Rennes Drive, Exeter, EX4 4RJ, UK

    Ted R. Feldpausch, Toby Pennington & Luciana de Oliveira Pereira

43. Herbario Amazónico Colombiano, Instituto SINCHI, Calle 20 No 5-44, Bogotá, DC, Colombia

    Nicolás Castaño Arboleda

44. Agteca-Amazonica, Santa Cruz, Bolivia

    Timothy J. Killeen

45. Programa de Pós-Graduação em Ecologia e Conservação, Universidade do Estado de Mato Grosso, Nova Xavantina, MT, Brazil

    Beatriz S. Marimon, Ben Hur Marimon-Junior & Edmar Almeida de Oliveira

46. Facultad de Ciencias Agrícolas, Universidad Autónoma Gabriel René Moreno, Santa Cruz, Santa Cruz, Bolivia

    Bonifacio Mostacedo

47. Biodiversity and Ecosystem Services, Instituto Tecnológico Vale, Belém, Pará, 66055-090, Brazil

    Rafael L. Assis

48. Centro de Investigaciones Ecológicas de Guayana, Universidad Nacional Experimental de Guayana, Calle Chile, urbaniz Chilemex, Puerto Ordaz, Bolivar, Venezuela

    Hernán Castellanos & Lionel Hernandez

49. Embrapa Recursos Genéticos e Biotecnologia, Parque Estação Biológica, Prédio da Botânica e Ecologia, Brasilia, DF, 70770-917, Brazil

    Marcelo Brilhante de Medeiros & Marcelo Fragomeni Simon

50. Projeto Dinâmica Biológica de Fragmentos Florestais, Instituto Nacional de Pesquisas da Amazônia - INPA, Av. André Araújo, 2936, Petrópolis, Manaus, AM, 69067-375, Brazil

    Ana Andrade & José Luís Camargo

51. Universidade do Estado de Mato Grosso, Nova Xavantina, Nova Xavantina, MT, Brazil

    Gisele Biem Mori

52. Programa de Pós-Graduação em Ecologia, Instituto Nacional de Pesquisas da Amazônia - INPA, Av. André Araújo, 2936, Petrópolis, Manaus, AM, 69067-375, Brazil

    Thaiane R. Sousa

53. Laboratório de Ecologia de Doenças Transmissíveis da Amazônia (EDTA), Instituto Leônidas e Maria Deane, Fiocruz, Rua Terezina, 476, Adrianópolis, Manaus, AM, 69060-001, Brazil

    Emanuelle de Sousa Farias

54. Instituto de Ciências Biológicas, Universidade Federal do Pará, Av. Augusto Corrêa 01, Belém, PA, 66075-110, Brazil

    Maria Aparecida Lopes

55. Programa de Pós-Graduação em Ecologia, Universidade Federal do Pará, Av. Augusto Corrêa 01, Belém, PA, 66075-110, Brazil

    José Leonardo Lima Magalhães

56. Empresa Brasileira de Pesquisa Agropecuária, Embrapa Amazônia Oriental, Trav. Dr. Enéas Pinheiro s/n°, Belém, PA, 66095-903, Brazil

    José Leonardo Lima Magalhães, Joice Ferreira & Ademir R. Ruschel

57. Diretoria Técnico-Científica, Instituto de Desenvolvimento Sustentável Mamirauá, Estrada do Bexiga, 2584, Tefé, AM, 69470-000, Brazil

    Helder Lima de Queiroz

58. Programa de Pós-Graduação em Biologia (Botânica), Instituto Nacional de Pesquisas da Amazônia - INPA, Av. André Araújo, 2936, Petrópolis, Manaus, AM, 69067-375, Brazil

    Caroline C. Vasconcelos & Yuri Oliveira Feitosa

59. Programa de Ciencias del Agro y el Mar, Herbario Universitario (PORT), UNELLEZ-Guanare, Guanare, Portuguesa, 3350, Venezuela

    Gerardo A. Aymard C

60. Grupo de Pesquisa em Ecologia Florestal, Instituto de Desenvolvimento Sustentável Mamirauá, Estrada do Bexiga, 2584, Tefé, AM, 69553-225, Brazil

    Pâmella Leite de Sousa Assis, Darlene Gris & Karoline Aparecida Felix Ribeiro

61. Laboratorio de Ecología de Bosques Tropicales y Primatología, Universidad de los Andes, Carrera 1 # 18a- 10, Bogotá, DC, 111711, Colombia

    Pablo R. Stevenson, Angela Cano, Diego F. Correa, Sasha Cárdenas & Luisa Fernanda Casas

62. Museo de Historia Natural Noel Kempff Mercado, Universidad Autónoma Gabriel Rene Moreno, Avenida Irala 565 Casilla Post al, 2489, Santa Cruz, Santa Cruz, Bolivia

    Alejandro Araujo-Murakami, Luzmila Arroyo, Germaine Alexander Parada & Daniel Villarroel

63. School of Geography, Earth and Environmental Sciences, University of Birmingham, Edgbaston, Birmingham, B15 2TT, UK

    Bruno Barçante Ladvocat Cintra

64. Birmingham Institute for Forest Research, University of Birmingham, Edgbaston, Birmingham, B15 2TT, UK

    Bruno Barçante Ladvocat Cintra

65. Endangered Species Coalition, 8530 Geren Rd, Silver Spring, MD, 20901, USA

    Hugo F. Mogollón

66. Biology Department and Center for Energy, Environment and Sustainability, Wake Forest University, 1834 Wake Forest Rd, Winston Salem, NC, 27106, USA

    Miles R. Silman, William Farfan-Rios & Karina Garcia-Cabrera

67. Facultad de Ciencias Forestales y Ambientales, Instituto de Investigaciones para el Desarrollo Forestal, Universidad de los Andes, Via Chorros de Milla, 5101, Mérida, Mérida, Venezuela

    José Rafael Lozada

68. Inventory and Monitoring Program, National Park Service, 120 Chatham Lane, Fredericksburg, VA, 22405, USA

    James A. Comiskey

69. Center for Conservation and Sustainability, Smithsonian Conservation Biology Institute, 1100 Jefferson Dr. SW, Suite 3123, Washington, DC, 20560-0705, USA

    James A. Comiskey, Alfonso Alonso & Reynaldo Linares-Palomino

70. Universidade Federal do Amapá, Ciências Ambientais, Rod. Juscelino Kubitschek km2, Macapá, AP, 68902-280, Brazil

    José Julio de Toledo, Wegliane Campelo & Renato Richard Hilário

71. Gothenburg Global Biodiversity Centre, University of Gothenburg, Carl Skottbergs gata 22b, Gothenburg, 413 19, Sweden

    Gabriel Damasco

72. Centro para la Restauración y Bioeconomía Sostenible - CREBIOS, Lima, 15088, Peru

    Roosevelt García-Villacorta

73. Peruvian Center for Biodiversity and Conservation (PCBC), Iquitos, Loreto, Peru

    Roosevelt García-Villacorta

74. Postgraduate Program in Clean Technologies, UniCesumar and Cesumar Institute of Science, Technology, and Innovation (ICETI), UniCesumar, Av. Guedner, 1610 - Jardim Aclimação, Maringá, PR, 87050-900, Brazil

    Aline Lopes

75. Servicios de Biodiversidad EIRL, Jr. Independencia 405, Iquitos, Loreto, Peru

    Marcos Rios Paredes, Hilda Paulette Dávila Doza, George Pepe Gallardo Gonzales & Linder Felipe Mozombite Pinto

76. Andes to Amazon Biodiversity Program, Madre de Dios, Madre de Dios, Peru

    Fernando Cornejo Valverde

77. Center for Conservation and Sustainability, Smithsonian’s National Zoo & Conservation Biology Institute, National Zoological Park, 3001 Connecticut Ave, Washington, DC, 20008, USA

    Francisco Dallmeier

78. Nature and Sustainability Solutions LLC, 2710 Isles of St. Marys Way, St. Marys, GA, 31558, USA

    Francisco Dallmeier

79. Department of Biology, University of Turku, Turku, 20014, Finland

    Vitor H. F. Gomes

80. Environmental Science Program, Geosciences Department, Universidade Federal do Pará, Rua Augusto Corrêa 01, Belém, PA, 66075-110, Brazil

    Vitor H. F. Gomes

81. Universidad Estatal Amazónica, Puyo, Pastaza, Ecuador

    David Neill

82. Universidad Regional Amazónica IKIAM, Km 7 via Muyuna, Tena, Napo, Ecuador

    Maria Cristina Peñuela Mora

83. Procuradoria-Geral de Justiça, Ministério Público do Estado do Amazonas, Av. Coronel Teixeira, 7995, Manaus, AM, 69037-473, Brazil

    Daniel P. P. de Aguiar

84. Coordenação de Dinâmica Ambiental, Instituto Nacional de Pesquisas da Amazônia - INPA, Av. André Araújo, 2936, Petrópolis, Manaus, AM, 69067-375, Brazil

    Daniel P. P. de Aguiar

85. Norwegian University of Life Sciences (NMBU), Faculty of Environmental Sciences and Natural Resource Management, P.O. Box 5003 NMBU, Aas, 1432 Aas, Trondheim, Norway

    Yennie K. Bredin & Torbjørn Haugaasen

86. Norwegian Institute for Nature Research (NINA), Sognsveien 68, Oslo, 0855, Oslo, Norway

    Yennie K. Bredin

87. Universidade Federal de Minas Gerais, Instituto de Ciências Biológicas, Departamento de Genética, Ecologia e Evolução, Av. Antônio Carlos, 6627 Pampulha, Belo Horizonte, MG, 31270-901, Brazil

    Fernanda Antunes Carvalho

88. Department of Biology, University of Miami, Coral Gables, FL, 33146, USA

    Kenneth J. Feeley & Riley P. Fortier

89. Fairchild Tropical Botanic Garden, Coral Gables, FL, 33156, USA

    Kenneth J. Feeley

90. Instituto de Biociências - Dept. Ecologia, Universidade de Sao Paulo - USP, Rua do Matão, Trav. 14, no. 321, Cidade Universitária, São Paulo, SP, 05508-090, Brazil

    Alexandre A. Oliveira & Cláudia Baider

91. Dept. Biological Sciences, Florida Atlantic University, Boca Raton, FL, 33431, USA

    John J. Pipoly III

92. Broward County Parks and Recreation, Oakland Park, FL, 33309, USA

    John J. Pipoly III

93. Lancaster Environment Centre, Lancaster University, Lancaster, Lancashire, LA1 4YQ, UK

    Jos Barlow & Erika Berenguer

94. Postgraduate program in Biodiversity and Biotechnology – Bionorte, Federal University of Acre, Rodovia 364, km 4.5, Distrito industrial, Rio Branco, AC, 69900-000, Brazil

    Izaias Brasil da Silva

95. Scientific research program, Juruá Institute, Rua Ajuricaba, 359, Aleixo, Manaus, AM, 69083-000, Brazil

    Maria Julia Ferreira

96. Department of Integrative Biology, University of California, Berkeley, CA, 94720-3140, USA

    Paul V. A. Fine & Italo Mesones

97. Empresa Brasileira de Pesquisa Agropecuária, Embrapa Amapá, Rod. Juscelino Kubitschek km 5, Macapá, AP, 68903-419, Brazil

    Marcelino Carneiro Guedes & Janaina Barbosa Pedrosa Costa

98. Graduate Program in Ecology, Federal University of Santa Catarina (UFSC), Campus Universitário - Córrego Grande, Florianópolis, SC, 88040-900, Brazil

    Carolina Levis & Bernardo Monteiro Flores

99. Direccíon de Evaluación Forestal y de Fauna Silvestre, Av. Javier Praod Oeste 693, Magdalena del Mar, Peru

    Boris Eduardo Villa Zegarra

100. Instituto de Investigaciones Forestales de la Amazonía, Universidad Autónoma del Beni José Ballivián, Campus Universitario Final, Av. Ejercito, Riberalta, Beni, Bolivia

     Vincent Antoine Vos & Guido Pardo Molina

101. Escuela de Biología Herbario Alfredo Paredes, Universidad Central, Ap. Postal 17.01.2177, Quito, Pichincha, Ecuador

     Carlos Cerón

102. Direction régionale de la Guyane, Office national des forêts, Cayenne, F-97300, French Guiana

     Émile Fonty

103. Department of Biological Sciences, California State Polytechnic University, 1 Harpst Street, Arcata, CA, 95521, USA

     Terry W. Henkel

104. Herbario HAG, Universidad Nacional Amazónica de Madre de Dios (UNAMAD), Av. Jorge Chávez, 1160, Puerto Maldonado, Madre de Dios, Peru

     Isau Huamantupa-Chuquimaco

105. Centro de Ciências Biológicas e da Natureza, Universidade Federal do Acre, Rodovia BR 364, Km 4, s/n, Distrito Industrial, Rio Branco, AC, 69915-559, Brazil

     Marcos Silveira

106. Museo Nacional de Ciencias Naturales (MNCN-CSIC), C. de José Gutiérrez Abascal 2, Madrid, 28006, Spain

     Juliana Stropp

107. Iwokrama International Centre for Rain Forest Conservation and Development, Georgetown, Guyana

     Raquel Thomas

108. New York Botanical Garden, 2900 Southern Blvd, Bronx, New York, NY, 10458-5126, USA

     Doug Daly

109. Department of Life Sciences and Systems Biology, University of Turin, Turin, Italy

     Kyle G. Dexter

110. Department for Ecosystem Stewardship, Royal Botanic Gardens, Kew, Richmond, Surrey, TW9 3AE, UK

     William Milliken

111. Tropical Diversity Section, Royal Botanic Garden Edinburgh, 20a Inverleith Row, Edinburgh, Scotland, EH3 5LR, UK

     Toby Pennington

112. Latin America Department, Missouri Botanical Garden, 4344 Shaw Blvd, St. Louis, MO, 63110, USA

     Alfredo Fuentes Claros & J. Sebastián Tello

113. Herbario Nacional de Bolivia, Instituto de Ecologia, Universidad Mayor de San Andres, Carrera de Biologia, La Paz, Bolivia

     Alfredo Fuentes Claros

114. Laboratorio de Plantas Vasculares y Herbario ISV, Universidad Nacional de Jaén, Carretera Jaén San Ignacio Km 23, Jaén, Cajamarca, 06801, Peru

     José Luis Marcelo Pena

115. Laboratoire Evolution et Diversité Biologique, CNRS and Université Paul Sabatier, UMR 5174 EDB, Toulouse, 31000, France

     Jerome Chave

116. Department of Anthropology, University of Texas at Austin, SAC 5.150, 2201 Speedway Stop C3200, Austin, TX, 78712, USA

     Anthony Di Fiore

117. Estación de Biodiversidad Tiputini, Colegio de Ciencias Biológicas y Ambientales, Universidad San Francisco de Quito-USFQ, Quito, Pichincha, Ecuador

     Anthony Di Fiore & Gonzalo Rivas-Torres

118. Fundación Puerto Rastrojo, Cra 10 No. 24-76 Oficina 1201, Bogotá, DC, 110311, Colombia

     Juan Fernando Phillips

119. Department of Wildlife Ecology and Conservation, University of Florida, 110 Newins-Ziegler Hall, Gainesville, FL, 32611, USA

     Gonzalo Rivas-Torres

120. Biosystematics group, Wageningen University, Droevendaalsesteeg 1, Wageningen, 6708 PB, The Netherlands

     Tinde R. van Andel

121. Fundación Estación de Biología, Cra 10 No. 24-76 Oficina, 1201, Bogotá, DC, Colombia

     Patricio von Hildebrand

122. Department of Anthropology, Tulane University, 101 Dinwiddie Hall, 6823 St. Charles Avenue, New Orleans, LA, 70118, USA

     William Balee

123. PROTERRA, Instituto de Investigaciones de la Amazonía Peruana (IIAP), Av. A. Quiñones km 2,5, Iquitos, Loreto, 784, Peru

     Ricardo Zárate Gómez

124. ACEER Foundation, Jirón Cusco N° 370, Puerto Maldonado, Madre de Dios, Peru

     Therany Gonzales

125. Amazon Conservation Team, 4211 North Fairfax Drive, Arlington, VA, 22203, USA

     Bruce Hoffman

126. Institut de Ciència i Tecnologia Ambientals, Universitat Autònoma de Barcelona, 08193, Bellaterra, Barcelona, Spain

     André Braga Junqueira

127. Instituto de Ciencias Naturales, Universidad Nacional de Colombia, Apartado, 7945, Bogotá, DC, Colombia

     Adriana Prieto & Agustín Rudas

128. Instituto de Ciência Agrárias, Universidade Federal Rural da Amazônia, Av. Presidente Tancredo Neves 2501, Belém, PA, 66.077-830, Brazil

     Natalino Silva

129. Escuela Profesional de Ingeniería Forestal, Universidad Nacional de San Antonio Abad del Cusco, Jirón San Martín 451, Puerto Maldonado, Madre de Dios, Peru

     César I. A. Vela

130. Laboratory of Human Ecology, Instituto Venezolano de Investigaciones Científicas - IVIC, Ado 20632, Caracas, DC, 1020A, Venezuela

     Egleé L. Zent & Stanford Zent

131. Cambridge University Botanic Garden, Cambridge University, 1 Brookside, Cambridge, CB2 1JE, UK

     Angela Cano

132. Programa de Maestria de Manejo de Bosques, Universidad de los Andes, Via Chorros de Milla, 5101, Mérida, Mérida, Venezuela

     Yrma Andreina Carrero Márquez

133. Centre for Biodiversity and Conservation Science CBCS, The University of Queensland, Brisbane, QLD, 4072, Australia

     Diego F. Correa

134. Resource Ecology Group, Wageningen University & Research, Droevendaalsesteeg 3a, Lumen, building number 100, Wageningen, Gelderland, 6708 PB, The Netherlands

     Milena Holmgren

135. Unique land use GmbH, Schnewlinstraße 10, Freiburg im Breisgau, 79098, Germany

     Michelle Kalamandeen

136. Núcleo de Estudos e Pesquisas Ambientais, Universidade Estadual de Campinas – UNICAMP, CP 6109, Campinas, SP, 13083-867, Brazil

     Guilherme Lobo

137. Facultad de Biologia, Universidad Nacional de la Amazonia Peruana, Pevas 5ta cdra, Iquitos, Loreto, Peru

     Tony Mori Vargas, Freddy Ramirez Arevalo & Elvis H. Valderrama Sandoval

138. Laboratório de Ciências Ambientais, Universidade Estadual do Norte Fluminense, Av. Alberto Lamego 2000, Campos dos Goytacazes, RJ, 28013-620, Brazil

     Marcelo Trindade Nascimento

139. Instituto de Investigaciones para el Desarrollo Forestal (INDEFOR), Universidad de los Andes, Conjunto Forestal, 5101, Mérida, Mérida, Venezuela

     Hirma Ramirez-Angulo, Emilio Vilanova Torre & Armando Torres-Lezama

140. Departamento de Biologia, Universidade Federal do Amazonas - UFAM – Instituto de Ciências Biológicas – ICB1, Av General Rodrigo Octavio 6200, Manaus, AM, 69080-900, Brazil

     Veridiana Vizoni Scudeller

141. Faculty of Social Sciences, University of Nottingham, University Park, Nottingham, NG7 2RD, UK

     Geertje van der Heijden

142. Wildlife Conservation Society (WCS), 2300 Southern Boulevard, Bronx, New York, NY, 10460, USA

     Emilio Vilanova Torre

143. The Mauritius Herbarium, Agricultural Services, Ministry of Agro-Industry and Food Security, Reduit, 80835, Mauritius

     Cláudia Baider

144. Department of Biology, Aarhus University, Building 1540, Aarhus C, Aarhus, 8000, Denmark

     Henrik Balslev

145. Escuela de Ciencias Forestales (ESFOR), Universidad Mayor de San Simon (UMSS), Sacta, Cochabamba, Bolivia

     Casimiro Mendoza

146. FOMABO, Manejo Forestal en las Tierras Tropicales de Bolivia, Sacta, Cochabamba, Bolivia

     Casimiro Mendoza

147. Fundación Amigos de la Naturaleza (FAN), Km. 7 1/2 Doble Vía La Guardia, Santa Cruz, Bolivia

     Daniel Villarroel

148. Tropenbos International, Horaplantsoen 12, Ede, 6717 LT, The Netherlands

     Roderick Zagt

149. School of Anthropology and Conservation, University of Kent, Marlowe Building, Canterbury, Kent, CT2 7NR, UK

     Miguel N. Alexiades

150. Herbario Nacional del Ecuador, Universidad Técnica del Norte, Quito, Pichincha, Ecuador

     Walter Palacios Cuenca

151. Instituto de Biodiversidade e Florestas, Universidade Federal do Oeste do Pará, Rua Vera Paz, Campus Tapajós, Santarém, PA, 68015-110, Brazil

     Daniela Pauletto

152. Department of Biology, University of Missouri, St. Louis, MO, 63121, USA

     Elvis H. Valderrama Sandoval

Contributions

H.t.S. & C.B. conceived the study. H.t.S. performed the analyses. H.t.S. wrote the first manuscript version, later with significant input from L.P., J.A.G., CF, W.M., O.L.P., E.P., B.G.L., J.E.G., M.J.E., T.R.B., M.N.U., M.v.d.S., M.M.P., M.M.G., F.C.D.. H.t.S. curated the A.T.D.N. data. A.L.e., G.P. curated the Forestplots data and approved the manuscript. L.P., M.v.S., H.t.S. provided trait data. I.A., L.S.C., F.W., F.D.A.M., D.L., R.P.S., H.t.S., C.V.C., J.E.G., M.C., O.L.P., M.T.P., W.E.M., D.S., J.F.M., L.D., J.D.C.R., J.o.S., M.I., M.P.M., J.R.S.G., J.F.R., O.B., A.Q., N.P., C.P., D.J.R., J.H., E.A., L.B., L.C., M.C.V.S., B.G.L., E.N., P.N., T.S.S., E.M.V., A.G.M., N.R., J.T., K.C., E.H., A.M., J.C.M., C.S., M.O., F.C., J.E., T.F., C.B., N.C., F.D.M., C.Z., T.K., B.S.M., B.H.M., R.V., B.M., R.A., D.D.A., H.C., J.E.H., M.B.M., M.F.S., A.S.A., J.L.C., S.L., W.L., L.M., G.B.M., J.S., T.R.S., E.F., M.A.L., J.L.L.M., H.E.M., H.L.Q., C.C.V., G.A., R.B., P.A., D.G.r, K.R., P.R.S., T.B., A.A.M., B.C., Y.F., H.F.M., M.R.S., L.F., J.R.L., F.C.D., J.A.C., J.J.T., G.D., R.G.V., A.L., M.R.P., A.V., I.V., F.C.V., A.A., L.A., F.D., V.F.G., W.N., D.N., M.C.P., D.A., F.B., Y.B., R.C., F.A.C., F.C.S., K.F., R.G., T.H., J., M.P.P, J.J.P, J.B., E.B., I.d., J.F., M.F., P.F., M.G., C.L., J.C.L, B.V., V.V., E.P., C.C., É.F., T.W.H., I.H., M.S., J.S.C., R.T.C., D.D., K.D., W.M., G.M., R.P., B.A., W.C., A.F., B.K., J.L.M.P., L.T.M., J.S.T., C.V., J.C., A.d.F., R.H., L.O.P., J.F.P., G.R., T.v.A., P.H., M.E., W.B., E.M.B., L.C.M.B., H.D., R.Z.G., G.G., T.G., B.H., A.J., Y.M., I.P.A.M., L.F.M.P., A.P., A.R., A.R.R., N.S., C.I.A.V., E.Z., S.Z., M.U., M.M.P., A.C., Y.C., D.C., J.B.C., B.M.F., D.g.a., M.H., M.K., G.L., T.M.V., M.N., A.O., H.R., M.R., V.S., G.v., E.V.T., C.b.a., H.e.B., S.C., L.F.C., W.F., R.L., C.M., I.M., G.P., A.T., D.V., R.Z., M.N.A., E.d., R.P.F., K.G., L.H., W.P., S.P.S., D.P., F.R., A.S., E.H.V., L.V. provided plot data, reviewed the manuscript, had the opportunity to comment on the manuscript and approved it.

Corresponding author

Correspondence toHans ter Steege.

Competing interests

The authors declare no competing interests.

Peer review information

Communications Biology thanks the anonymous reviewers for their contribution to the peer review of this work. Primary Handling Editor: Michele Repetto. A peer review file is available.

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visithttp://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions