ArticlePDF Available

Abstract and Figures

The latitudinal diversity gradient (LDG) is one of the most recognized global patterns of species richness exhibited across a wide range of taxa. Numerous hypotheses have been proposed in the past two centuries to explain LDG, but rigorous tests of the drivers of LDGs have been limited by a lack of high-quality global species richness data. Here we produce a high-resolution (0.025° × 0.025°) map of local tree species richness using a global forest inventory database with individual tree information and local biophysical characteristics from ~1.3 million sample plots. We then quantify drivers of local tree species richness patterns across latitudes. Generally, annual mean temperature was a dominant predictor of tree species richness, which is most consistent with the metabolic theory of biodiversity (MTB). However, MTB underestimated LDG in the tropics, where high species richness was also moderated by topographic, soil and anthropogenic factors operating at local scales. Given that local landscape variables operate synergistically with bioclimatic factors in shaping the global LDG pattern, we suggest that MTB be extended to account for co-limitation by subordinate drivers.
This content is subject to copyright. Terms and conditions apply.
Co-limitation towards lower latitudes shapes global forest
diversity gradients
Published in the Nature Ecology & Evolution, 2022. https://doi.org/10.1038/s41559-022-01831-x
Authors:
Jingjing Liang 1, Javier G. P. Gamarra 2, Nicolas Picard 3, Mo Zhou 1, Bryan Pijanowski 4, Douglass F.
Jacobs 4, Peter B. Reich 5,6,7, Thomas W. Crowther 8, Gert-Jan Nabuurs 9,10, Sergio de-Miguel
11,12, Jingyun Fang 13, Christopher W. Woodall 14, Jens-Christian Svenning 15,16, Tommaso Jucker
17, Jean-Francois Bastin 18, Susan K. Wiser 19, Ferry Slik 20, Bruno Hérault 21,22, Giorgio Alberti
23,24,25, Gunnar Keppel 26, Geerten M. Hengeveld 27,28, Pierre L. Ibisch 29, Carlos A. Silva 30, Hans
ter Steege 31, Pablo L. Peri 32, David A. Coomes 33, Eric B. Searle 34, Klaus von Gadow 35,36,37,
Bogdan Jaroszewicz 38, Akane O. Abbasi 1, Meinrad Abegg 39, Yves C. Adou Yao40, Jesús Aguirre-
Gutiérrez 41,42, Angelica M. Almeyda Zambrano43, Jan Altman 44,45, Esteban Alvarez-Dávila46,
Juan Gabriel Álvarez-González47, Luciana F. Alves 48, Bienvenu H. K. Amani 49, Christian A. Amani50,
Christian Ammer 51, Bhely Angoboy Ilondea52, Clara Antón-Fernández 53, Valerio Avitabile 54,
Gerardo A. Aymard55, Akomian F. Azihou 56, Johan A. Baard57, Timothy R. Baker 58, Radomir
Balazy59, Meredith L. Bastian 60,61, Rodrigue Batumike62, Marijn Bauters 63,64, Hans Beeckman 65,
Nithanel Mikael Hendrik Benu 66, Robert Bitariho 67, Pascal Boeckx 64, Jan Bogaert68, Frans Bongers
10, Olivier Bouriaud 69, Pedro H. S. Brancalion70, Susanne Brandl71, Francis Q. Brearley 72, Jaime
Briseno-Reyes 73, Eben N. Broadbent 30, Helge Bruelheide 74,75, Erwin Bulte 76, Ann Christine Catlin
77, Roberto Cazzolla Gatti 78, Ricardo G. César 70, Han Y. H. Chen 34, Chelsea Chisholm 79, Emil
Cienciala 80,81, Gabriel D. Colletta 82, José Javier Corral-Rivas 73, Anibal Cuchietti 83, Aida Cuni-
Sanchez 84,85, Javid A. Dar 86,87,88, Selvadurai Dayanandan89, Thales de Haulleville 65,68, Mathieu
Decuyper 10, Sylvain Delabye 90,91, Géraldine Derroire 92, Ben DeVries 93, John Diisi 94, Tran Van
Do 95, Jiri Dolezal 44,96, Aurélie Dourdain 92, Graham P. Durrheim 57, Nestor Laurier Engone Obiang
97, Corneille E. N. Ewango 98, Teresa J. Eyre 99, Tom M. Fayle 91,100, Lethicia Flavine N.
Feunang101, Leena Finér102, Markus Fischer 103, Jonas Fridman 104, Lorenzo Frizzera 105, André L.
de Gasper 106, Damiano Gianelle 105, Henry B. Glick 107, Maria Socorro Gonzalez-Elizondo 108, Lev
Gorenstein 77, Richard Habonayo 109, Olivier J. Hardy 110, David J. Harris 111, Andrew Hector 112,
Andreas Hemp 113, Martin Herold 114, Annika Hillers 115,116, Wannes Hubau 65,117, Thomas
Ibanez 118, Nobuo Imai119, Gerard Imani 120, Andrzej M. Jagodzinski 121,122, Stepan Janecek90,
Vivian Kvist Johannsen 123, Carlos A. Joly 124, Blaise Jumbam 125,126, Banoho L. P. R. Kabelong 101,
Goytom Abraha Kahsay 127, Viktor Karminov 128, Kuswata Kartawinata 129, Justin N. Kassi130,
Elizabeth Kearsley131, Deborah K. Kennard 132, Sebastian Kepfer-Rojas 123, Mohammed Latif Khan
133, John N. Kigomo134, Hyun Seok Kim 135,136,137,138, Carine Klauberg 30, Yannick Klomberg 90,
Henn Korjus 139, Subashree Kothandaraman 87,88, Florian Kraxner 140, Amit Kumar 141, Relawan
Kuswandi66, Mait Lang 139,142, Michael J. Lawes 143, Rodrigo V. Leite 144, Geoffrey Lentner 77,
Simon L. Lewis 58,145, Moses B. Libalah101,146, Janvier Lisingo 147, Pablito Marcelo López-
Serrano148, Huicui Lu149, Natalia V. Lukina150, Anne Mette Lykke151, Vincent Maicher 90,91,152,
Brian S. Maitner 153, Eric Marcon 92,154, Andrew R. Marshall 155,156,157, Emanuel H. Martin 158,
Olga Martynenko 128, Faustin M. Mbayu98, Musingo T. E. Mbuvi 159, Jorge A. Meave 160, Cory
Merow161, Stanislaw Miscicki162, Vanessa S. Moreno 70, Albert Morera 12, Sharif A. Mukul163, Jörg
C. Müller 164,165, Agustinus Murdjoko 166, Maria Guadalupe Nava-Miranda148, Litonga Elias
Ndive167, Victor J. Neldner99, Radovan V. Nevenic 168, Louis N. Nforbelie 101, Michael L.
Ngoh169,170, Anny E. N’Guessan 40, Michael R. Ngugi 99, Alain S. K. Ngute 163,171, Emile Narcisse
N. Njila101, Melanie C. Nyako 101, Thomas O. Ochuodho 172, Jacek Oleksyn 121, Alain Paquette 173,
Elena I. Parfenova 174, Minjee Park 4, Marc Parren 10, Narayanaswamy Parthasarathy88, Sebastian
Pfautsch 175, Oliver L. Phillips 58, Maria T. F. Piedade 176, Daniel Piotto 177, Martina Pollastrini 178,
Lourens Poorter 10, John R. Poulsen 152, Axel Dalberg Poulsen111, Hans Pretzsch 179, Mirco
Rodeghiero 105,180, Samir G. Rolim 177, Francesco Rovero181,182, Ervan Rutishauser183, Khosro
Sagheb-Talebi184, Purabi Saikia 185, Moses Nsanyi Sainge 169,186, Christian Salas-Eljatib
187,188,189, Antonello Salis2, Peter Schall 51, Dmitry Schepaschenko 140,174,190, Michael Scherer-
Lorenzen 191, Bernhard Schmid 192, Jochen Schöngart176, Vladimír Šebeň 193, Giacomo Sellan
72,194, Federico Selvi 178, Josep M. Serra-Diaz 195, Douglas Sheil 10,196, Anatoly Z. Shvidenko 140,
Plinio Sist 197, Alexandre F. Souza198, Krzysztof J. Stereńczak 59, Martin J. P. Sullivan 72, Somaiah
Sundarapandian 88, Miroslav Svoboda 45, Mike D. Swaine 199, Natalia Targhetta176, Nadja
Tchebakova 174, Liam A. Trethowan 200, Robert Tropek 90,91, John Tshibamba Mukendi 201, Peter
Mbanda Umunay 202, Vladimir A. Usoltsev 203, Gaia Vaglio Laurin204, Riccardo Valentini 204,
Fernando Valladares 205, Fons van der Plas 206, Daniel José Vega-Nieva73, Hans Verbeeck 131,
Helder Viana 207,208, Alexander C. Vibrans 209, Simone A. Vieira210, Jason Vleminckx 211,
Catherine E. Waite 212, Hua-Feng Wang 213, Eric Katembo Wasingya98, Chemuku Wekesa 214,
Bertil Westerlund 104, Florian Wittmann 215, Verginia Wortel216, Tomasz Zawiła-Niedźwiecki217,
Chunyu Zhang 218, Xiuhai Zhao218, Jun Zhu219, Xiao Zhu77, Zhi-Xin Zhu213, Irie C. Zo-Bi 220 and
Cang Hui 221,222
Abstract:
The latitudinal diversity gradient (LDG) is one of the most recognized global patterns of species
richness exhibited across a wide range of taxa. Numerous hypotheses have been proposed in the
past two centuries to explain LDG, but rigorous tests of the drivers of LDGs have been limited by a
lack of high-quality global species richness data. Here we produce a high-resolution (0.025° × 0.025°)
map of local tree species richness using a global forest inventory database with individual tree
information and local biophysical characteristics from ~1.3 million sample plots. We then quantify
drivers of local tree species richness patterns across latitudes. Generally, annual mean temperature
was a dominant predictor of tree species richness, which is most consistent with the metabolic
theory of biodiversity (MTB). However, MTB underestimated LDG in the tropics, where high species
richness was also moderated by topographic, soil and anthropogenic factors operating at local scales.
Given that local landscape variables operate synergistically with bioclimatic factors in shaping the
global LDG pattern, we suggest that MTB be extended to account for co-limitation by subordinate
drivers.
Main text
Identifying which mechanisms moderate global biodiversity patterns1,2 has perplexed the
scientific community for more than two centuries3,4. The most noticeable pattern, the latitudinal
diversity gradient (LDG), is a trend of declining local species richness (alpha diversity) from low to
high latitudes. This trend has been observed for many taxonomic groups and across land, freshwater
and marine environments5,6. More than 30 hypotheses have been proposed3,4,7,8 to explain LDG9,
but few can be reconciled with existing observational data for predicting biodiversity decline towards
the poles. To test these varied hypotheses, biodiversity data must be assembled that are global in
scope with sufficient sample coverage across all ecoregions and biomes.
In addition to biodiversity data, testing these varied hypotheses also requires data on a wide
spectrum of potential drivers that may moderate biodiversity at local scales9,10, such as climate, soil
and land features and anthropogenic factors. For instance, environmental temperature (that is,
ambient temperature of the air, represented by annual mean temperature) is largely responsible for
the generation and maintenance of biodiversity through the effects of solar radiation on
demographic rates (for example, growth and mortality), ecological interactions (for example,
predation and competition) and evolutionary rates of change (for example, speciation and
extinction)11,12. Soil and topographic heterogeneity facilitate niche partitioning via inducing
microclimatic variation, contributing to compositional variation13 and biodiversity
maintenance14,15. Furthermore, humans have a long history of reshaping biodiversity through the
selective use of natural resources and the modification of native species composition16. In addition,
multiple subordinate factors jointly affecting biodiversity could potentially increase the diversity of
niche opportunities, thereby resulting in species rich assemblages.
Here we quantified the relative contribution of a wide range of environmental factors across
space on local tree species richness in forested areas around the world. To accomplish this, we
standardized a global tree species richness (that is, as alpha diversity) database (Fig. 1) and quantified
the relative contribution of 47 explanatory variables including bioclimatic conditions (for example,
annual mean temperature), vegetation and survey attributes (for example, sample plot size),
topographic covariates (for example, terrain roughness), soil covariates (for example, bulk density)
and anthropogenic spatial features (for example, size of roadless areas) in an attempt to test
whether local co-limitation exists when multiple subordinate drivers co-dominate (Figs. 2 and 3). We
conducted a three-stage analysis (Fig. 1 and Methods in Supplementary Information for details)
based on two independent ground-sourced forest inventory datasets (Phase I and Phase II; Extended
Data Fig. 1). The main dataset (Phase I) consisted of 1,255,444 sample plots, while the validating
dataset (Phase II) consisted of 22,131 sample plots, most of which are located in unsampled and
under-sampled regions of the Phase I dataset. Together, our sample data covered 424 of the 435
(97%) forested ecoregions worldwide (Extended Data Fig. 1), with a total of ~55 million sample trees
representing more than 32,000 species.
Results and discussion
Global patterns of local tree species richness and LDG.
Our analyses confirmed, with a high level of accuracy, one general spatial trend in local tree
species richness worldwide that has led us to three conclusions regarding the mechanisms underlying
patterns of tree species richness. We found that LDG for tree species richness was consistent with
that of most other groups of organisms, with a decline from the tropics to the poles (Figs. 2 and 4). In
the Northern Hemisphere, tree species richness dropped sharply from the equator (98 species per
ha) to 10 N with an average rate of decline of 6 species per ha per 1 increase in latitude, after
which the decline diminished and stabilized at 4 species per ha at 50 N. In the Southern Hemisphere,
tree species richness declined from the equator to 25 S on average by 3 species per ha per 1
increase in latitude, after which tree species richness fluctuated before another steep drop from 25
species per ha (43 S) to 4 species per ha (50 S). We were able to detect and map regional patterns
and global peaks of tree species diversity, with a high spatial resolution (0.025 × 0.025). The
Amazonian, Southeast Asian and Melanesian rainforests are the regions with the greatest local tree
species richness worldwide, containing >200 tree species per ha above the 5 cm diameter-at-breast-
height (DBH) threshold, confirming previous findings17,18. Tropical African rainforests generally
contain 50% fewer tree species per hectare than Amazonian rainforests. In the temperate forests of
the Northern Hemisphere, the Changbai Mountains in Northeast Asia (up to ~28 species per ha) and
the Central Appalachian forests in the Eastern United States (up to ~20 species per ha) display high
local species richness. In the Southern Hemisphere, the sclerophyllous and Nothofagus-dominated
forests in south-central Chile are among the most species-rich temperate communities (up to 50
species per ha). Boreal forest communities are consistently low in local tree species richness, with
typically five or fewer tree species per hectare.
The above LDG pattern of tree species richness was generally consistent with the metabolic
theory of biodiversity (MTB)19,20, except at low latitudes (Fig. 5). According to MTB, environmental
temperature is largely responsible for the generation and maintenance of biodiversity12,21,22, and
the natural logarithm of species richness is linearly associated with 1,000 T1, where T is the absolute
environmental temperature in Kelvin (mean annual temperature +273.15 K), with a slope ranging
from 7.5 K to 9.0 K. Our global tree species richness gradient was largely consistent with MTB, with
a slope of 8.0 K (P < 0.001) and a coefficient of determination of 0.82 (Metabolic Theory of
Biodiversity in Methods), indicating that environmental temperature is generally a good predictor of
LDG. However, at low latitudes, MTB substantially underestimated LDG. In fact, near the equator
where the actual LDG peaked (98 species per ha), observed tree species richness was almost twice as
high as predicted by MTB (56 species per ha) (Fig. 4a). Our results suggest that within this low
latitudinal range, other factors are also important to the maintenance of biodiversity.
The under-estimation of local tree species richness by MTB at low latitudes is attributable, in
part, to the lack of a definite dominant environmental factor, suggesting a co-limitation of multiple
subordinate drivers at low latitudes (Fig. 5). In general, bioclimatic factors predominantly determined
species richness in 82.6% of the forested areas, while co-limitation (that is, absence of any
dominating factor) occurred in 11.7% of forested areas globally. However, in the low-latitude range
between 5 N and 15 S, the percentage area of co-limitation increased to 37.1%, more than three
times the global average. Furthermore, forested areas under co-limitation contained on average 81.1
± 0.1 species per ha, much higher than the average local tree species richness of forested areas
predominantly determined by topographic (43.9 ± 0.1), anthropogenic (35.6 ± 0.2), soil (33.9 ± 0.2)
and bioclimatic (19.4 ± 0.02) factors (Fig. 5b). This suggests that the pattern of co-limitation is
pervasive in species-rich tropical forests. In South America, transitional areas between Amazonia and
savanna formations nearby are subject to co-limitation that is partly attributable to a dynamic
equilibrium between closed forest and savanna23, edaphic conditions and natural fire regimes24. In
Africa, anthropogenic influences such as selective timber extraction and fuelwood collection,
together with large-scale degradation25 affect local tree species richness (Fig. 5 and Extended Data
Fig. 7). In Central Africa, the evolution of anthropogenic influences from prehistoric to present times
has imposed a substantial effect on species diversity26 and resulted in the development of a complex
system of mixes with light-demanding and old-growth tree species.
Bioclimatic dominance and co-limitation.
In addition to an overall positive response of local tree species richness to the rise of annual
mean temperature (partial dependence plot of C1 in Fig. 3 and Extended Data Fig. 3), the importance
of environmental temperature (2.7%) was topped by the total annual precipitation (C12, 7.6%) (Fig.
3). Our findings are consistent with previous discoveries of a joint role of water and
temperature/energyas a proxy for net primary productivity27on plant species richness, with
water dominating particularly at warmer, lower latitudes22,28. Predicted tree species richness
accelerated exponentially with temperature and rainfall, although independently, as shown in the
colddry quadrant and the convex contours of the 2D partial dependence plot (Extended Data Fig. 3),
until each has reached its respective threshold (1,500 mm for total annual precipitation and 10 C for
annual mean temperature). Beyond one of these thresholds, species richness is limited only by the
predictor below its threshold (that is, by annual mean temperature in the coldwet quadrant or by
annual precipitation in the hotdry quadrant). When both predictors have reached their thresholds,
that is, in the hotwet quadrant, co-limitation predominates in most tropical forests. Net primary
productivity in the tropics, thus, requires co-limitation of other factors besides only temperature and
rainfall29. As the response of carbon flux mirror the low-latitude co-limitation pattern for tree
species diversity, the matching determinants for both diversity and productivity may explain the
similar latitudinal gradient in productivity and the positive diversityproductivity relationship30,31.
Our findings also indicate that under climate change, intensified droughts coupled with increased
annual mean temperature32 can potentially trigger declines of tree species richness, although
possible increases in water-use efficiency from elevated CO2 and the dominance of highly contingent
co-limiting factors may partially buffer this effect in the tropics33.
Here we articulate evidence for co-limitation in LDG. Resource co-limitation is a common
concept in ecology (for example, refs. 34,35), often used to describe how the synergistic interactions
of two or more factors limit ecological productivity36. Our use of the term co-limitation emphasizes
the reduced effect of bioclimate on tree species richness at low latitudes, although bioclimate is the
globally predominant driver of species richness, recognizing that several local subordinate factors
synergistically contribute to increased tree species richness in this latitudinal range. We thus argue
that the inclusion of co-limitation could substantially improve the explanatory power of biodiversity
models in estimating alpha diversity by considering multiple subordinate factors where single-factor
dominance is lacking, especially in the tropics. At high latitudes, bioclimatic conditions, particularly
environmental temperature, are the major limiting factors and thus the dominant drivers of tree
species diversity. As the latitude declines, the influence of bioclimatic conditions dwindles and the
maintenance of tree species richness is moderated by many interacting drivers without a clear
dominance, which is especially well expressed between 5 N and 15 S (Fig. 5). This prevalence of co-
limiting factors is thus not a mere coincidence as to why the observed LDG at low latitudes is almost
double that predicted by MTB (Fig. 2). While each of the existing hypotheses underpinning LDG
addresses a certain process10,12 (for example, selection, drift, dispersal or speciation), the evidence
of co-limitation highlights synergistic interactions of local processes across the latitudinal gradient.
Concluding remarks.
More research is needed to fully elucidate patterns of LDG driven by climatic and other
influences, especially those outlined in competing hypotheses. First, our analyses lack explicit
consideration of some evolutionary, ecological and historical factors. These include mid-domain
stochastic effects37, the legacies of the poleward expansion of tree species after the Last Glacial
Maximum38,39 and recent human land use/management. Alternative hypotheses, such as niche
conservatism or climatic history, are more difficult to test due to data limitations. In addition, long-
term effects at geological and millennial time scales also play a role, but it is difficult to disentangle
these effects due to collinearity40. A major source of uncertainty in our results (Fig. 4b) came from
an uneven sample coverage between developed and developing countries (Extended Data Fig. 1). To
address this gap, we argue that there needs to be a shared responsibility among forestry agencies at
various levels of government, scientists, indigenous communities and other biodiversity monitoring
groups to improve sample coverage of forest inventories in developing countries. Innovative
biodiversity funding mechanisms, for example, forest inventories funded by carbon initiatives such as
REDD+, should be incorporated into a comprehensive global forest biodiversity database.
Meanwhile, the severe shortage of experts and database management infrastructures, especially in
developing countries, poses another major challenge to address this gap41. The education and
training of new generations of forest scientists, taxonomists and foresters can bring tangible benefits
to biodiversity monitoring while improving local economies as well.
Considering co-limitation in addition to MTB enables a refined description of the biogeographic
distribution of biodiversity and mechanisms underlying LDG. Our analysis has resulted in the
production of a high-resolution map of tree species richness across the global forest range, along
with visuals of those factors responsible for the moderation of local tree species richness. Such tools
are necessary for conservation management which requires assessments of factors responsible for
biodiversity patterns at multiple scales that matterfrom local and regional to global scales. Patterns
of local tree species richness and associated drivers may provide insights into how and why the
diversity of other forest flora, fauna and microbes42,43 vary across space and time. Furthermore, the
high-resolution map of local tree species richness presented here provides a benchmark for
evaluating the impact of biodiversity loss on the productivity and functioning of forest
ecosystems31,44. Finally, aligned with current international calls for spatially explicit monitoring of
ecosystem attributes45, this study delivers detailed biogeographic information to support
international endeavours46 focused on valuing natural capital and advancing global conservation.
Methods
As illustrated in Fig. 1, we conducted data analyses and modelling in three stages.
Stage 1. For this study, we compiled individual in situ tree data from all the regional and
national GFBi forest inventory datasets (Supplementary Table 2) into a standardized GFBi dataframe,
that is, the GFBi tree list. In this standardized GFBi dataframe, each row represents an individual tree,
and columns represent nine key tree- and plot-level attributes. These attributes are tree ID (FID), a
unique number assigned to each individual tree; plot ID (PLT), a unique string assigned to each plot;
plot coordinates (LAT and LON); tree species name (SPCD); DBH or above buttress; year of
measurement; and dataset name (DSN), a unique number assigned to each forest inventory dataset
(Supplementary Table 2). With a total of 56 million trees surveyed, GFBi individual-based dataframe
represents 1/50,000 of the approximately 2.7 trillion trees47 worldwide. Because all trees in each
sample plot were identified and measured, GFBi data make it possible to quantify forest community
structure, composition and species distribution. To ensure consistency and maximum accuracy in
species names, we standardized observations from different forest inventory datasets with the
following protocol. First, all multi-stem trees were divided so that each stem represents an individual
tree. The scientific names were extracted from original datasets, keeping only the genus and species
(authority names were removed). Next, all the species names were compiled into five general species
lists, one for each continent. We verified individual species names against 23 online taxonomic
databases or web application programming interfaces (API) using the gnr_resolve() function from the
‘taxize’ package48 of R49. We then manually verified and corrected all the names that did not match
with the majority of the online taxonomic databases, that is, the names with a matching score lower
than 0.9. For individuals denoted by morphospecies, we assigned each a unique name comprising the
genus name and a unique species code. The unique species code consisted of the string ‘spp’, plus
the dataset name followed by a unique number denoting if two individuals belong to the same
species. For example, ‘Aidia sppCDi1’ and ‘Aidia sppCDi2’ represented two different species under
the genus ‘Aidia’, and both species have been observed in a forest inventory of the Democratic
Republic of the Congo named ‘CDi’. To maximize our species coverage, a tree was defined in this
study as a perennial plant with an elongated woody stem that supports branches and leaves,
including woody angiosperms, gymnosperms and taller palms (Arecaceae). Tree ferns (Cyatheales)
and bamboos (Bambusoideae) were excluded from our analysis.
From the GFBi individual tree-level dataframe, we derived a global species abundance matrix.
The global species abundance matrix consisted of the number of individuals by species (column
vectors) within individual sample plots (row vectors). The global species abundance matrix consisted
of two complementary datasets: Phase I dataset contained 1,255,444 sample plots and Phase II
dataset contained 22,131 sample plots, most of which are located in unsampled and under-sampled
regions of Phase I dataset. Phase I sample plots cover 394 ecoregions across the world, and Phase II
sample plots cover an additional 30 ecoregions in Africa, South America, Southeast Asia, Mexico,
India and Japan. Together, our ground-based forest sample plots cover 424 of 435 (97.5%) forested
ecoregions across the world. The global species abundance matrix contains ~1.3 million rows (plots)
by 32,608 columns (species). Key plot-level information was added to the matrix, including PLT, DSN,
plot coordinates, basal area (B), the total cross-sectional areas (m2) of living trees per ha calculated
from DBH and TPH (expansion factor) and the year of measurement. TPH denotes the number of
trees per hectare represented by each sampled individual. It ranged from 1 to 5,244 across the GFBi
data, with a mean of 48 trees per ha.
We quantified for each sample plot tree species richness (S), which is the total number of tree
species in a community. Due to the difference in plot size (standard deviation = 0.09 ha) and
threshold DBH values (standard deviation = 2.52 cm) across GFBi sample plots, we developed
machine learning models to standardize tree species richness for a common basis of 1 ha in area and
5 cm in threshold DBH. The models incorporated both plot area (A) and threshold DBH (D) as
predictors to account for the underlying species-area relationship5052 and species-individual size
distribution53 in a rarefaction-based approach54. This standardization approach justifies compiling
direct tree species diversity estimates from GFBi in situ data of different sources and sampling
protocols5557, an issue highlighted in earlier large-scalealthough less extensiveforest
biodiversity studies57,58. To evaluate the accuracy of this standardization approach, we tested the
machine learning models using cross-sample validation and compared our global maps of estimated
tree species diversity against other standardization approaches based on sample completeness
(Model Evaluation below).
The machine learning models employed 47 environmental covariates to predict tree species
richness. These covariates, derived from satellite-based remote sensing and ground-based survey
data, can be summarized into five general categories: bioclimatic (for example, annual mean
temperature, total annual precipitation, potential evapotranspiration and indexed annual aridity);
soil (bulk density, pH, electrical conductivity, C/N ratio and total nitrogen); topographic, including
elevation, slope, aspect and terrain features; vegetation and survey attributes (plot size, basal area,
threshold diameter and percent forest canopy cover); and anthropogenic variables (human footprint,
roadless areas and size of protected areas) (Supplementary Table 1). We extracted all geospatial
covariate values from raster datasets to point locations of GFBi plots using ArcMap 10.3 (ref. 59) and
R 3.4.1 (ref. 49), to build a standardized plot-level dataframe.
Stage 2. We trained random forests (RF)60, an ensemble learning method that detects general
trends present in the data using a multitude of decision trees, to estimate standardized community-
level tree species diversity. The RF algorithm applies the general technique of bootstrap aggregating
(bagging) with a modified tree learning algorithm that selects at each candidate split in the learning
process a random subset of the features (that is, feature bagging). Because a random subset of
variables is chosen for each tree, the RF algorithm based on bagged tree ensembles avoids
overfitting60 and mitigates the multicollinearity issue61 posed by high correlations between some of
the predictors variables (Fig. 3). Using subsamples of GFBi data as the training set (that is, training
dataframe) with response S, bagging repeatedly for B times selects a random sample with
replacement of the training set and trains a regression tree fb. After training, RF can predict for
unseen samples Xʹ with the response variable S being tree species richness per ha:
=
(󰆒)
 (1)
For rigorous model evaluation, we employed three very different cross validation approaches:
randomized cross validation (RCV), spatial cross validation (SCV) and post-sample validation (PSV). In
RCV, a model was trained for each continent with a random subsample that accounted for 90% of the
training data from that continent, and the remaining 10% of the training data were used as the
testing set. This process was repeated 20 times with sample replacement to examine the accuracy of
estimated tree species diversity values. In SCV, all sample data from an ecoregion62 were reserved
for testing the model that was trained with the remaining samples from the larger continent within
which the ecoregion is situated. We decided to use ecoregions as spatial blocks because (1) unlike
political units such as countries and provinces, ecoregions are delineated based on ecological and
bioclimatic conditions; and (2) with a total of ~700 terrestrial ecoregions across the world, each
ecoregion encompasses 1,800 sample plots on average, which is a large enough sample size for
training RF models. This process was repeated until all the forested ecoregions across the world had
been tested. SCV was more rigorous than RCV because samples from an entire ecoregion rather than
random samples were withheld for validation. PSV was the most rigorous among the three validation
processes. For PSV, we have collated an independent sample dataset from 22,131 forest sample
plots, which we named Phase II sample plots to highlight their independence from the original GFBi
dataset (that is, Phase I sample plots). In PSV, we used Phase II data as the testing set to evaluate the
accuracy of the predictive models that were trained for each continent with the Phase I data.
Using these three cross validation processes, we also evaluated the performance of the RF
model against two other predictive models, including multiple regression with ordinary least squares
(OLS) and Extreme Gradient Boosting (XGBoost). For each model, we derived predicted values of tree
species richness of the testing sets and compared these predicted values against observed data using
mean absolute error, root-mean-squared error (RMSE), and coefficient of determination (R2) (ref.
63). The process was repeated 20 times to select the best model for each continent.
The OLS model estimated values of standardized point diversity for non-sampled point location
s, based on spatially explicit values of covariates:
()= ()+() (2)
where Y(s) is tree species richness at location s; X a design matrix for the predictor variables at
location s; α is a vector of coefficients; and e is a random vector following a Gaussian probability
density function, with an expected value of zero and variance of σ2. Spatial autocorrelation64 was
not accounted for here due to computational limitations. GFBi data collected from sample plots of
various sizes were harmonized to represent local forest community populations per ha using the
expansion factor65, and we used the standardized species richness per ha values for the response
variables. We fit a model (2) for each continent. To mitigate the multicollinearity issue66, we
selected for the OLS model the best subset of predictor variables for each continent from the
predictor variables used in the RF models, using step-wise regression and Akaike information
criterion67.
XGBoost is a scalable machine learning system68 that implements the gradient boosting
decision tree algorithm69. With this ensemble technique, an initial model was trained, with new
models added sequentially to correct for errors made by each existing model until no further
improvements could be made. Then, new and initial models were merged to make a final prediction
that minimized errors. With its algorithm engineered for efficiency in computing time and memory
resources, XGBoost is widely used by data scientists to achieve state-of-the-art results on a number
of machine learning challenges68. In this study, the XGBoost model estimated tree species diversity
values in three steps. First, an initial model F0 was defined to predict the target variable Y. This
model was associated with a residual (Y – F0). Second, a new model h1 was fit to the residuals from
the previous step, and F0 and h1 were combined to form the boosted model F1:
()=()+() (3)
of which the mean squared error was lower than that from F0. Finally, to improve the performance
of F1, we modelled after the residuals of F1 to create a new model, F2, and repeated it for m
iterations until the mean squared error converged:
()=()+() (4)
Before training RF and XGBoost models, we fine tuned four key hyper-parameters, two for
each model. Using 20 bootstrapping iterations on random training sets consisting of 90% of the
samples, we first evaluated the sensitivity of RMSE of the testing sets (consisting of the remaining
10% of the samples) to the number of trees to grow and the number of variables randomly sampled
as candidates at each split for the RF model and selected the optimal hyper-parameter values
(Extended Data Fig. 5). Similarly, we selected the optimal values of the maximum number of boosting
iterations (that is, number of rounds) and the maximum depth of a tree for the XGBoost model
(Extended Data Fig. 6). As a result, we obtained a preliminary RF model.
Because the RF model emerged as the most accurate model from all three cross validation
processes (Extended Data Fig. 2), we selected the RF as the final model, and re-calibrated the final RF
model using all the sample data (Phase I and Phase II data).
Stage 3. Global map of local tree species richness. To map community-level tree species richness over
the global forest range, we first derived the global forest range map from version 1.3 of the Global
Forest Change database70 (years 20002015). To ensure consistency with the definition of forest71
by the Food and Agriculture Organization of the United Nations (FAO), the global forest range in this
study was defined as forested areas with 10% tree crown coverage per unit area. The tiled
‘treecover2000’, ‘loss’ and ‘gain’ datasets were integrated to obtain current forest cover estimates
for the year 2015. To minimize processing artefacts, the ~1 arcsecond spatial resolution tiles were
spatially aggregated to an even multiple of their native resolution that approximated the resolution
of our covariates. The datasets were then converted to vector point files before being reconverted to
raster format with the exact resolution and origin of our covariates. After mosaicking each set of
tiles, we computed ‘tree cover’ (scaled) ‘loss’ + ‘gain’ to obtain the 2015 global forest cover,
represented as percent forest cover per ~30 arcsecond pixel. Artefacts in the original data led to
0.08% of all terrestrial pixels having forest cover estimates greater than 100% and 1.9% of terrestrial
pixels having estimates less than 0%. These values were truncated to 100% and inflated to 0%,
respectively. Finally, the global forest range consisted of those pixels with a percent forest cover
10% in 2015. In total, each map consisted of 9,944,908 pixels of 0.025° × 0.025° (hereafter, the
pixel) of forested areas. This range is rather conservative and potentially underestimates many
remnant forests in drylands and grasslands72.
We then estimated tree species richness at a 1 ha scale for all pixels within a continent based
on the final RF model trained for that continent, using both Phase I and Phase II data. Spatially
explicit local environmental covariate data across the global forest range were used for the
imputation, except that plot size and threshold DBH were set as 1 ha and 5 cm, respectively. For
ecoregions with extremely low sample coverage, we further fine tuned the RF model using samples
of similar environment characteristics from other continents. More specifically, we first identified
two ecoregions of extremely low sample coverage, that is, the temperate forests in South America
and the tropical forests in Oceania, as there were fewer than 1,000 sample plots for the entire biome
on those continents. We then trained a new RF model for each ecoregion, using all the sample data
from the same biome across the world and fine tuned the mapping data for that ecoregion using the
biome-specific RF model.
We computed and mapped the width of the 95% confidence interval for our local estimates of
tree species richness per ha across the global forest range. To this end, we employed a rigorous
spatial-block approach, analogous to the spatial cross validation, to derive the 95% confidence
interval. More specifically, we computed the width of the 95% confidence interval for each 0.025° ×
0.025° mapping pixel by ecoregion. For a pixel p in ecoregion e, we trained 20 RF models using
random subsamples that accounted for 90% of the training data from the same continent, which
included all samples except those from ecoregion e. We then derived the standard error and the
width of the 95% confidence interval for this pixel p in ecoregion e from the predictions of the 20 RF
models trained for this ecoregion. This process was repeated until all the forested ecoregions across
the world had been assessed and mapped.
Uncertainty in our global diversity estimates was caused by two types of error. The first was
measurement error from in situ forest inventories. We mitigated this type of error by implementing
stringent species name check and data standardization protocols (Stage 1 Data Standardization). The
second arose from the imputation process to map tree species diversity. We minimized this type of
error using the three cross validation approaches introduced in Stage 2.
Metabolic theory of biodiversity. Using the global standardized tree species richness values
predicted from the final RF models, we quantified the global LDG of tree species richness and tested
the effect of environmental temperature based on the MTB19:
ln()=,
 + (5)
where S represents species richness and Tenv here represents absolute environmental temperature
(mean annual temperature +273.15 K); α and β represent coefficients to be estimated by ordinary
least squares. According to both the original and extended MTB19,20, the slope α is expected to
range between −7.5 K and −9.0 K, under the assumption that tree community abundance per area
does not vary with latitude.
Variance partitioning. We used variance partitioning73 based on the sample data from ~1.3 million
plots to quantify the unique and joint fractions of spatial variance in tree species richness explained
by environmental factors and latitude. Due to the correlation between species and environment and
between the spatially explicit environmental factors, the variance partitioning approach mitigates
type I error inflated by spatial autocorrelation74. With variance partitioning, we tested the
significance of environmental effects on tree species richness in a series of nested RF models. (A) The
full model (Extended Data Fig. 4a) consisted of latitude and 47 environmental variables (including 21
bioclimatic ones). (B) The reduced model I (Extended Data Fig. 4b) consisted of all but the 21
bioclimatic variables. (C) The reduced model II (Extended Data Fig. 4c) consisted of only a zero
constant. The overall significance of all environmental factors plus latitude was tested in an one-
tailed F-test by comparing the residual sum of squares of error (RSS) of model (A) and model (C):
=




, (6)
where n − nA and nC − nA stand for the degree of freedom for the full model and the difference in the
degrees of freedom between the full model and the reduced model II, respectively.
The significance of bioclimatic factors, with the effect of latitude being controlled, was tested
in an one-tailed F-test by comparing RSS of model (A) and model (B):
=




, (7)
where nB − nA = 21 stands for the difference in the degrees of freedom between the full model and
the reduced model I.
We partitioned the spatial variance in observed species richness into four components: a
represents the fraction of variance uniquely explained by environmental factors (that is, bioclimatic,
topographic, anthropogenic and soil variables), after latitudinal effects have been taken into account;
b represents the fraction of variance jointly explained by environmental factors and latitudinal
effects; c represents the fraction of variance explained by latitudinal effects after removing
environmental effects; and d represents the fraction of variance not explained by the full RF model.
Then, the total fraction of variance explained by both environmental factors and latitude was a + b +
c, the fraction of variance explained by environmental factors was a + b, and the fraction of variance
explained by latitude was b + c. Components a + b + c, a + b and b + c were estimated by the R2
statistics from the RF models trained for each continent using all factors, environmental factors and
latitude, respectively (Stage 2 Model Training and Evaluation). Components a, b and c were
computed from the previous components using arithmetic relationships that ensure that
a + b + c + d = 100%.
Model sensitivity. Based on the final RF models and sample data from ~1.3 million plots, we
mapped the dominant drivers of tree species richness with a 0.025° × 0.025° resolution (that is,
global map of co-limitation), following a standard procedure for model sensitivity analysis75:
Step 1: using the full RF model, and the values of environmental factors X(s) specific to a
0.025°-pixel s, we had already estimated local tree species richness Sfull(s):
() = ( ()), (8)
where f() represents the RF model, and X(s) environmental factors in four categories, namely E1:
bioclimatic, E2: topographic, E3: anthropogenic and E4: soil.
Step 2: for the above-mentioned pixel, we estimated a new local tree species richness value
S−E1 (s), using a reduced RF model in which all E1 (bioclimatic) variables were removed:
() =  (( 1) ()) , (9)
where f−E1() represents the RF model trained with all but 21 bioclimatic variables, and (X − E1) (s)
encompassed environmental factors in three categories, namely E2: topographic, E3: anthropogenic
and E4: soil.
Step 3: for a given pixel, we calculated the relative sensitivity of predicted species richness to
E1:
(1) = ()  () / () . (10)
Step 4: we repeated steps 2 and 3 to calculate, for a given pixel, the relative sensitivity of each
of the following categories (that is, E2: topographic, E3: anthropogenic and E4: soil), respectively. The
dominant driver (that is, limiting factor) for this pixel was then the category with the highest relative
sensitivity, provided that this relative sensitivity was greater than or equal to 0.2.
Step 5: if the relative sensitivities were less than 0.2 for all categories, we considered that this
was a scenario of joint effects of multiple categories of factors (that is, co-limitation), rather than
dominance of a single category. Where clear dominance of a single category was lacking, we denoted
the dominant driver of this pixel as ‘E5: co-limitation.’
Step 6: we repeated the steps above to calculate, for all the remaining pixels of the global grid,
the relative sensitivity of each of the five categories of environmental factors, namely E1: bioclimatic,
E2: topographic, E3: anthropogenic, E4: soil and E5: co-limitation. On the basis of these values, we
created a wall-to-wall map of dominant drivers of tree species richness across the global forest range
by labelling the category with the highest relative sensitivity for each pixel (Fig. 5a).
Step 7: on the basis of the relative sensitivity obtained from the steps 1–6, we computed
percent prevalence (0100%) of bioclimatic, topographic, anthropogenic and soil factors and a lack of
dominance (co-limitation) in all the forested pixels along each latitudinal band.
Data availability
The global map of tree species richness is available under license CC BY 4.0, with the identifier:
10.6084/m9.figshare.17232491. This map can be downloaded in two formats. One is a geoTIFF file
(S_mean_raster.tif) containing the fully geo-referenced map of tree species richness worldwide at a
0.025° × 0.025° resolution. The other is a comma-separated file (S_mean_grid.csv) with the following
attributes: S is local average tree species richness per ha and x, y are centroid coordinates of all
0.025° × 0.025° pixels. The global map of co-limitation is available under license CC BY 4.0, with the
identifier: 10.6084/m9.figshare.17234339. The metadata of the entire training dataframeincluding
the characteristics and references of all the in situ Phase I and Phase II datasets and the definitions,
units and summary statistics of the environmental covariatesis available under license CC BY 4.0
with the identifier: 10.6084/m9.figshare.19733449.v1. The public version of the training dataframe,
including the plot-level species richness and all the covariates, which is needed to reproduce the
models and results presented here, is available at: https://doi.org/10.6084/m9.figshare.20055488.
The maps and dataframe are also available on the international web research platform: Science-i
(https://science-i.org/). Raw forest inventory data are commonly subject to a wide array of
confidentiality clauses in regard to open access policies. Despite recent efforts to make some of
these data fully open76,77, some governments and private data owners, especially those from the
developing countries generally have decided to keep their data confidential. This decision is based on
well-founded arguments to protect certain trees or forests (because of their large size or protected
taxonomic status) from illegal logging or trespassing and to protect landowners’ privacy against the
misuse of plot information such as the geographic coordinates. The sensitive information in the
training dataframe, including the plot coordinates and tree-level information, will be available from
the corresponding author (albeca.liang@gmail.com) upon a request via Science-i (https://science-
i.org/) or GFBI (https://www.gfbinitiative.org/) and an approval from data contributors.
Code availability
All the models in this study were constructed using command line applications written in the R
programming language, which processed and restructured the input data, trained the model and
performed cross validation. Due to the massive amount of data, we used Purdue University’s Brown
supercomputing cluster to accelerate the training process. The development of the GFBi database,
tabular data cleaning, creation of species abundance matrices, evaluation of diversity determinants
and geostatistical imputation were conducted in R49 (v.3.4.2) through the use of several Linux-based
high-performance computing resources at Purdue University and a custom HPC interface developed
using Amazon Web Services, each designed for batch processing, scalable resource distribution,
embarrassingly parallel computations and/or large RAM jobs. Compute nodes with up to 1 TB of RAM
and clusters of up to 64 nodes were employed in this study. Portions of the covariate preparation,
mapping and quality control assessment were conducted on Windows-based operating systems with
up to 128 GB of RAM. Final continental-level RF models and the R codes we developed to train the
models are available under license MIT with the identifier: 10.6084/m9.figshare.17234729.
Additional information
Extended data is available for this paper at https://doi.org/10.1038/s41559-022-01831-x.
Supplementary information The online version contains supplementary material available at
https://doi.org/10.1038/s41559-022-01831-x.
Figures
Fig. 1. A conceptual diagram of the three-stage process employed in the study.
Stage 1: Two independent global forest biodiversity individual-based (GFBi) datasets (Phase I and
Phase II; Extended Data Fig. 1) were standardized into a global tree-level dataframe and aggregated
into a global species abundance matrix. On the basis of plot locations, we merged the abundance
matrix with 47 explanatory variables (Fig. 3) into a standardized plot-level dataframe. Stage 2: We
compared three candidate models (RF, random forests; XGB, XGBoost; OLS, ordinary least squares)
trained from the Phase I plot-level dataframe, using random and spatial cross validation based on
Phase I data and post-sample validation based on Phase II data (Extended Data Fig. 2). The final
model was then selected and re-calibrated with both Phase I and Phase II data. Stage 3: Using the
final model, we standardized and mapped local tree species richness per hectare across the global
forest range. On the basis of this globally continuous map, we quantified the associated LDG (Fig. 4a)
and tested for the MTB (Fig. 2). We further developed the global map of co-limitation (Fig. 5a) based
on model sensitivity analysis and quantified the contribution of key factors to local species richness
patterns using variance partitioning (Fig. 6). Dotted boxes represent processes or models and dashed
ones represent data or results. Methods provide details.
Fig. 2. Latitudinal gradients of estimated tree species richness and co-limitation of drivers.
a, The LDG of tree species richness per hectare was first empirically derived for all 0.025° pixels
within the global forest range and aggregated by latitude (Methods). Data are presented as mean
values (solid lines) ± standard deviation (shaded areas) and then compared to LDG predicted by the
MTB based on local mean annual temperatures.
b, The co-limitation illustrated here was the product of LDG and the percentage prevalence of
dominant drivers by latitude (Fig. 5).
Fig. 3. A total of 47 explanatory variables in five categories were used in random forests models to
predict local tree species richness and quantify LDG.
According to standardized variable importance values (horizontal bar plots to the left), bioclimatic
variables contributed the most to LDG, followed by vegetation and survey, topographic,
anthropogenic and soil variables. The correlogram to the right illustrates correlations between any
two variables by the colour and size of a disk. The partial dependence plots to the left (see Extended
Data Fig. 3) show the effect of each predictor variable on the species richness, while all the other
predictors remained constant at their sample mean. The metadata of the training dataframe provide
a detailed description of the explanatory variables. Min., minimum; Temp., temperature; Precip.,
precipitation; t., temperature; Pot., potential; Max., maximum; dbh, diameter at breast height; part.
deriv., partial derivative; Topo., topographic; HF, human footprint; El., electrical.
Fig. 4. Estimated tree species richness per hectare in forested areas worldwide.
a, Tree species richness per hectare was first derived for the ~1.3 million GFBi plots across the world
and then imputed to the global forest extent. Top left, curves (solid lines representing the mean and
shaded areas representing the standard deviation of the mean) represent the observed LDG (black)
of tree species diversity in comparison with LDG (red) predicted by the MTB based on local mean
annual temperatures (Fig. 2a).
b, Width of the 95% confidence interval (C.I.) for the estimated tree species richness per hectare. All
map layers are displayed at a 0.025 × 0.025 resolution with an equirectangular projection (Plate
Carrée) for better illustration of the latitudinal gradients.
Fig. 5 | Dominant drivers of tree species richness in forested areas worldwide.
a, Driver dominance was derived for each pixel from four driver categories (that is, bioclimatic,
topographic, anthropogenic and soil) and co-limitation, which represents a lack of clear dominance
among the four foregoing categories. The pixel-level drivers were then aggregated by 0.5° latitudinal
bins to show the percentage prevalence of dominant drivers by latitude (top left).
b, The violin charts show the kernel probability density of tree species richness per ha for different
drivers. Inside boxes indicate the median (line in the centre) and interquartile range (bounds of
boxes). The numbers on top of the violin charts indicate the percentage of forested pixels globally
that corresponds to each driver category. The red line represents the mean and 95% confidence
interval of tree species richness per ha (81.1 ± 0.1) for all the 0.025° × 0.025° pixels of co-limitation.
The vertical axis is on a logarithmic scale for better illustration.
Fig. 6. Patterns and variance of local tree species richness per ha by continent.
The collage of maps shows the zoomed-in view of the distribution of predicted local tree species
richness per ha (Fig. 4a) by continent. Circular Venn diagrams (with the legend in the centre) show,
for each continent, the spatial variance in observed tree species richness partitioned as follows: a
(mean = 14.3%) represents the fraction of variance uniquely explained by environmental factors (that
is, bioclimatic, topographic, anthropogenic and soil variables) after latitudinal effects had been
accounted for. b (mean = 68.2%) stands for the fraction of variance jointly explained by
environmental factors and latitudinal effects. c (mean = 0.3%) represents the fraction of variance
explained by latitudinal effects after removing environmental effects. d (mean = 17.2%) represents
the fraction of unexplained variance in tree species richness. The fractions were based on contrasting
the amount of local richness variations in sample data from ~1.3 million plots explained by the R2
statistics from the continental-scale random forest models with the full set of factors versus those
with targeted factors removed.
References
1. Sutherland, W. J. et al. Identification of 100 fundamental ecological questions. J. Ecol. 101, 5867
(2013).
2. Gaston, K. J. Global patterns in biodiversity. Nature 405, 220227 (2000).
3. Pickering, C. The Geographical Distribution of Animals and Plants (Little, Brown & Co., 1854).
4. Humboldt, A. V. & Bonpland, A. Essai sur la Géographie des Plantes Accompagné d’un Tableau
Physique des Régions Équinoxiales (Levrault, Schoell & Co., 1805).
5. Hillebrand, H. On the generality of the latitudinal diversity gradient. Am. Nat. 163, 192211 (2004).
6. Crame, J. A. Taxonomic diversity gradients through geological time. Diversity Distrib. 7, 175189
(2001).
7. Gough, L. & Field, R. Latitudinal diversity gradients. eLS
https://doi.org/10.1002/9780470015902.a0003233.pub2 (2007).
8. Willig, M. R., Kaufman, D. M. & Stevens, R. D. Latitudinal gradients of biodiversity: pattern,
process, scale, and synthesis. Annu. Rev. Ecol., Evolution, Syst. 34, 273309 (2003).
9. Pontarp, M. et al. The latitudinal diversity gradient: novel understanding through mechanistic eco-
evolutionary models. Trends Ecol. Evol. 34, 211223 (2019).
10. Vellend, M. The Theory of Ecological Communities (MPB-57) Vol. 75 (Princeton Univ. Press, 2016).
11. Brown, J. H. Why are there so many species in the tropics? J. Biogeogr. 41, 8–22 (2014).
12. Currie, D. J. et al. Predictions and tests of climate-based hypotheses of broad-scale variation in
taxonomic richness. Ecol. Lett. 7, 11211134 (2004).
13. Baldeck, C. A. et al. Soil resources and topography shape local tree community structure in
tropical forests. Proc. R. Soc. B 280, 20122532 (2013).
14. Qian, H. & Ricklefs, R. E. Large-scale processes and the Asian bias in species diversity of
temperate plants. Nature 407, 180182 (2000).
15. Stein, A., Gerstner, K. & Kreft, H. Environmental heterogeneity as a universal driver of species
richness across taxa, biomes and spatial scales. Ecol. Lett. 17, 866880 (2014).
16. Sponsel, L. E. in Encyclopedia of Biodiversity (ed. Scheiner, S.M.) 137152 (Academic Press, 2013).
17. Sullivan, M. J. et al. Diversity and carbon storage across the tropical forest biome. Sci. Rep. 7,
39102 (2017).
18. Cazzolla Gatti, R. et al. The number of tree species on Earth. Proc. Natl Acad. Sci. USA 119,
e2115329119 (2022).
19. Allen, A. P., Brown, J. H. & Gillooly, J. F. Global biodiversity, biochemical kinetics, and the
energetic-equivalence rule. Science 297, 15451548 (2002).
20. Stegen, J. C., Enquist, B. J. & Ferriere, R. Advancing the metabolic theory of biodiversity. Ecol. Lett.
12, 10011015 (2009).
21. Wright, D. H. Species-energy theory: an extension of species-area theory. Oikos 41, 496506
(1983).
22. Hawkins, B. A. et al. Energy, water, and broad-scale geographic patterns of species richness.
Ecology 84, 31053117 (2003).
23. Staal, A., Dekker, S. C., Xu, C. & van Nes, E. H. Bistability, spatial interaction, and the distribution
of tropical forests and savannas. Ecosystems 19,
10801091 (2016).
24. de L. Dantas, V., Batalha, M. A. & Pausas, J. G. Fire drives functional thresholds on the savanna
forest transition. Ecology 94, 24542463 (2013).
25. Bodart, C. et al. Continental estimates of forest cover and forest cover changes in the dry
ecosystems of Africa between 1990 and 2000. J. Biogeogr. 40, 10361047 (2013).
26. Hubau, W. et al. The persistence of carbon in the African forest understory. Nat. Plants 5, 133
140 (2019).
27. Šímová, I. & Storch, D. The enigma of terrestrial primary productivity: measurements, models,
scales and the diversityproductivity relationship. Ecography 40, 239252 (2017).
28. Kreft, H. & Jetz, W. Global patterns and determinants of vascular plant diversity. Proc. Natl Acad.
Sci. USA 104, 59255930 (2007).
29. Clark, D. A. et al. Net primary production in tropical forests: an evaluation and synthesis of
existing field data. Ecol. Appl. 11, 371384 (2001).
30. Pärtel, M., Laanisto, L. & Zobel, M. Contrasting plant productivity–diversity relationships across
latitude: the role of evolutionary history. Ecology 88, 10911097 (2007).
31. Liang, J. et al. Positive biodiversityproductivity relationship predominant in global forests.
Science 354, aaf8957 (2016).
32. Hammond, W. M. et al. Global field observations of tree die-off reveal hotter-drought fingerprint
for Earth’s forests. Nat. Commun. 13, 1761 (2022).
33. Martens, C. et al. Large uncertainties in future biome changes in Africa call for flexible climate
adaptation strategies. Glob. Change Biol. 27, 340358 (2021).
34. Chapin, F. S., Bloom, A. J., Field, C. B. & Waring, R. H. Plant responses to multiple environmental
factors. BioScience 37, 4957 (1987).
35. Tilman, D. Resource Competition and Community Structure (MPB-17) Vol. 17 (Princeton Univ.
Press, 1982).
36. Harpole, W. S. et al. Nutrient co-limitation of primary producer communities. Ecol. Lett. 14, 852
862 (2011).
37. Colwell, R. K., Rahbek, C. & Gotelli, N. J. The mid-domain effect and species richness patterns:
what have we learned so far? Am. Nat.163, E1E23 (2004).
38. Feng, G. et al. Species and phylogenetic endemism in angiosperm trees across the Northern
Hemisphere are jointly shaped by modern climate and glacialinterglacial climate change. Glob. Ecol.
Biogeogr. 28, 13931402 (2019).
39. Fraser, C. I., Nikula, R., Ruzzante, D. E. & Waters, J. M. Poleward bound: biological impacts of
Southern Hemisphere glaciation. Trends Ecol. Evol. 27, 462471 (2012).
40. Algar, A. C., Kerr, J. T. & Currie, D. J. A test of metabolic theory as the mechanism underlying
broad-scale species-richness gradients. Glob. Ecol. Biogeogr. 16, 170178 (2007).
41. Wilson, E. O. A global biodiversity map. Science 289, 2279–2279 (2000).
42. Castagneyrol, B. & Jactel, H. Unraveling plantanimal diversity relationships: a meta-regression
analysis. Ecology 93, 21152124 (2012).
43. Steidinger, B. S. et al. Climatic controls of decomposition drive the global biogeography of forest–
tree symbioses. Nature 569, 404–408 (2019).
44. Bongers, F. J. et al. Functional diversity effects on productivity increase with age in a forest
biodiversity experiment. Nat. Ecol. Evol. 5, 1594–1603 (2021).
45. UN Statistical Commission. System of EnvironmentalEconomic AccountingEcosystem
Accounting: Final Draft (United Nations, 2021).
46. Green Bond Impact Report (The World Bank Treasury, 2019).
47. Crowther, T. et al. Mapping tree density at a global scale. Nature 525, 201205 (2015).
48. Chamberlain, J. L., Prisley, S. & McGuffin, M. Understanding the relationships between American
ginseng harvest and hardwood forests inventory and timber harvest to improve co-management of
the forests of eastern United States. J. Sustainable For. 32, 605624 (2013).
49. R Core Team. R: A Language and Environment for Statistical Computing (R Foundation for
Statistical Computing, 2020).
50. Borda-de-Água, L., Hubbell, S. P. & McAllister, M. Species-area curves, diversity indices, and
species abundance distributions: a multifractal analysis. Am. Nat. 159, 138155 (2002).
51. Connor, E. F. & McCoy, E. D. The statistics and biology of the species-area relationship. Am. Nat.
113, 791833 (1979).
52. He, F. & Legendre, P. On species-area relations. Am. Nat. 148, 719737 (1996).
53. White, E. P., Ernest, S. K. M., Kerkhoff, A. J. & Enquist, B. J. Relationships between body size and
abundance in ecology. Trends Ecol. Evol. 22, 323330 (2007).
54. Gotelli, N. J. & Colwell, R. K. Quantifying biodiversity: procedures and pitfalls in the measurement
and comparison of species richness. Ecol. Lett. 4, 379391 (2001).
55. Chirici, G., Winter, S. & McRoberts, R. E. National Forest Inventories: Contributions to Forest
Biodiversity Assessments Vol. 20 (Springer Science & Business Media, 2011).
56. Tomppo, E. et al. National Forest Inventories: Pathways for Common Reporting (Springer, 2010).
57. Winter, S., Böck, A. & McRoberts, R. E. Estimating tree species diversity across geographic scales.
Eur. J. For. Res. 131, 441451 (2012).
58. McRoberts, R. E., Winter, S., Chirici, G. & LaPoint, E. Assessing forest naturalness. For. Sci. 58,
294309 (2012).
59. Release 10.3 of Desktop, ESRI ArcGIS (Environmental Systems Research Institute, 2014).
60. Breiman, L. Random forests. Mach. Learn. 45, 532 (2001).
61. Strobl, C., Boulesteix, A.-L., Kneib, T., Augustin, T. & Zeileis, A. Conditional variable importance for
random forests. BMC Bioinf. 9, 307 (2008).
62. Olson, D. M. & Dinerstein, E. The global 200: priority ecoregions for global conservation. Ann. Mo.
Bot. Gard. 89, 199224 (2002).
63. Hyndman, R. J. & Koehler, A. B. Another look at measures of forecast accuracy. Int. J. Forecast.
22, 679688 (2006).
64. Legendre, P. Spatial autocorrelation: trouble or new paradigm? Ecology 74, 16591673 (1993).
65. Husch, B., Beers, T. W. & Kershaw Jr, J. A. Forest Mensuration 4th edn (John Wiley & Sons, 2003).
66. Farrar, D. E. & Glauber, R. R. Multicollinearity in regression analysis: the problem revisited. Rev.
Econ. Stat. 49, 92107 (1967).
67. Venables, W. N. & Ripley, B. D. Modern Applied Statistics with S-PLUS (Springer Science &
Business Media, 2013).
68. Chen, T. & Guestrin, C. XGBoost: a scalable tree boosting system. In Proc. 22nd ACM SIGKDD
International Conference on Knowledge Discovery and Data Mining 785794 (Association for
Computing Machinery, 2016).
69. Friedman, J. H. Greedy function approximation: a gradient boosting machine. Ann. Stat. 29,
11891232 (2001).
70. Hansen, M. C. et al. High-resolution global maps of 21st-century forest cover change. Science 342,
850853 (2013).
71. Global Forest Resources Assessment 2015How are the World’s Forests Changing? (Food and
Agriculture Organization of the United Nations, 2015).
72. Bastin, J.-F. et al. The extent of forest in dryland biomes. Science 356, 635638 (2017).
73. Peres-Neto, P. R., Legendre, P., Dray, S. & Borcard, D. Variation partitioning of species data
matrices: estimation and comparison of fractions. Ecology 87, 26142625 (2006).
74. Peres-Neto, P. R. & Legendre, P. Estimating and controlling for spatial structure in the study of
ecological communities. Glob. Ecol. Biogeogr. 19, 174184 (2010).
75. Saltelli, A. et al. Global Sensitivity Analysis: the Primer (John Wiley & Sons, 2008).
76. Liang, J. & Gamarra, J. G. P. The importance of sharing global forest data in a world of crises. Sci.
Data 7, 424 (2020).
77. Towards Open and Transparent Forest Data for Climate ActionExperiences and Lessons Learned
(United Nations Food and Agriculture Organization, 2022).
Acknowledgements
The team collaboration and manuscript development are supported by the web-based team science platform: science-i.org,
with the project number 202205GFB2. We thank the following initiatives, agencies, teams and individuals for data
collection and other technical support: the Global Forest Biodiversity Initiative (GFBI) for establishing the data standards
and collaborative framework; United States Department of Agriculture, Forest Service, Forest Inventory and Analysis (FIA)
Program; University of Alaska Fairbanks; the SODEFOR, Ivory Coast; University Félix Houphouët-Boigny (UFHB, Ivory Coast);
the Queensland Herbarium and past Queensland Government Forestry and Natural Resource Management departments
and staff for data collection for over seven decades; and the National Forestry Commission of Mexico (CONAFOR). We
thank M. Baker (Carbon Tanzania), together with a team of field assistants (Valentine and Lawrence); all persons who made
the Third Spanish Forest Inventory possible, especially the main coordinator, J. A. Villanueva (IFN3); the French National
Forest Inventory (NFI campaigns (raw data 2005 and following annual surveys, were downloaded by GFBI at
https://inventaire-forestier.ign.fr/spip.php?rubrique159; site accessed on 1 January 2015)); the Italian Forest Inventory (NFI
campaigns raw data 2005 and following surveys were downloaded by GFBI at https://inventarioforestale.org/; site accessed
on 27 April 2019); Swiss National Forest Inventory, Swiss Federal Institute for Forest, Snow and Landscape Research WSL
and Federal Office for the Environment FOEN, Switzerland; the Swedish NFI, Department of Forest Resource Management,
Swedish University of Agricultural Sciences SLU; the National Research Foundation (NRF) of South Africa (89967 and
109244) and the South African Research Chair Initiative; the Danish National Forestry, Department of Geosciences and
Natural Resource Management, UCPH; Coordination for the Improvement of Higher Education Personnel of Brazil (CAPES,
grant number 88881.064976/2014-01); R. Ávila and S. van Tuylen, Instituto Nacional de Bosques (INAB), Guatemala, for
facilitating Guatemalan data; the National Focal Center for Forest condition monitoring of Serbia (NFC), Institute of
Forestry, Belgrade, Serbia; the Thünen Institute of Forest Ecosystems (Germany) for providing National Forest Inventory
data; the FAO and the United Nations High Commissioner for Refugees (UNHCR) for undertaking the SAFE (Safe Access to
Fuel and Energy) and CBIT-Forest projects; and the Amazon Forest Inventory Network (RAINFOR), the African Tropical
Rainforest Observation Network (AfriTRON) and the ForestPlots.net initiative for their contributions from Amazonian and
African forests. The Natural Forest plot data collected between January 2009 and March 2014 by the LUCAS programme for
the New Zealand Ministry for the Environment are provided by the New Zealand National Vegetation Survey Databank
https://nvs.landcareresearch.co.nz/. We thank the International Boreal Forest Research Association (IBFRA); the Forestry
Corporation of New South Wales, Australia; the National Forest Directory of the Ministry of Environment and Sustainable
Development of the Argentine Republic (MAyDS) for the plot data of the Second National Forest Inventory (INBN2); the
National Forestry Authority and Ministry of Water and Environment of Uganda for their National Biomass Survey (NBS)
dataset; and the Sabah Biodiversity Council and the staff from Sabah Forest Research Centre. All TEAM data are provided by
the Tropical Ecology Assessment and Monitoring (TEAM) Network, a collaboration between Conservation International, the
Missouri Botanical Garden, the Smithsonian Institution and the Wildlife Conservation Society, and partially funded by these
institutions, the Gordon and Betty Moore Foundation and other donors, with thanks to all current and previous TEAM site
manager and other collaborators that helped collect data. We thank the people of the Redidoti, Pierrekondre and Cassipora
village who were instrumental in assisting with the collection of data and sharing local knowledge of their forest and the
dedicated members of the field crew of Kabo 2012 census. We are also thankful to FAPESC, SFB, FAO and IMA/SC for
supporting the IFFSC. This research was supported in part through computational resources provided by Information
Technology at Purdue, West Lafayette, Indiana.This work is supported in part by the NASA grant number 12000401 ‘Multi-
sensor biodiversity framework developed from bioacoustic and space based sensor platforms’ (J. Liang, B.P.); the USDA
National Institute of Food and Agriculture McIntire Stennis projects 1017711 (J. Liang) and 1016676 (M.Z.); the US National
Science Foundation Biological Integration Institutes grant NSF-DBI-2021898 (P.B.R.); the funding by H2020 VERIFY (contract
776810) and H2020 Resonate (contract 101000574) (G.-J.N.); the TEAM project in Uganda supported by the Moore
foundation and Buffett Foundation through Conservation International (CI) and Wildlife Conservation Society (WCS); the
Danish Council for Independent Research | Natural Sciences (TREECHANGE, grant 6108-00078B) and VILLUM FONDEN grant
number 16549 (J.-C.S.); the Natural Environment Research Council of the UK (NERC) project NE/T011084/1 awarded to J.A.-
G. and NE/S011811/1; ERC Advanced Grant 291585 (‘T-FORCES’) and a Royal Society-Wolfson Research Merit Award
(O.L.P.); RAINFOR plots supported by the Gordon and Betty Moore Foundation and the UK Natural Environment Research
Council, notably NERC Consortium Grants ‘AMAZONICA’ (NE/F005806/1), ‘TROBIT’ (NE/D005590/1) and ‘BIO-RED’
(NE/N012542/1); CIFOR’s Global Comparative Study on REDD+ funded by the Norwegian Agency for Development
Cooperation, the Australian Department of Foreign Affairs and Trade, the European Union, the International Climate
Initiative (IKI) of the German Federal Ministry for the Environment, Nature Conservation, Building and Nuclear Safety and
the CGIAR Research Program on Forests, Trees and Agroforestry (CRP-FTA) and donors to the CGIAR Fund; AfriTRON
network plots funded by the local communities and NERC, ERC, European Union, Royal Society and Leverhume Trust; a
grant from the Royal Society and the Natural Environment Research Council, UK (S.L.L.); National Science Foundation CIF21
DIBBs: EI: number 1724728 (A.C.C.); National Natural Science Foundation of China (31800374) and Shandong Provincial
Natural Science Foundation (ZR2019BC083) (H.L.). UK NERC Independent Research Fellowship (grant code: NE/S01537X/1)
(T.J.); a Serra-Húnter Fellowship provided by the Government of Catalonia (Spain) (S.d.-M.); the Brazilian National Council
for Scientific and Technological Development (CNPq, grant 442640/2018-8, CNPq/Prevfogo-Ibama number 33/2018)
(C.A.S.); a grant from the Franklinia Foundation (D.A.C.); Russian Science Foundation project number 19-77-300-12 (R.V.);
the Takenaka Scholarship Foundation (A.O.A.); the German Research Foundation (DFG), grant number Am 149/16-4 (C.A.);
the Romania National Council for Higher Education Funding, CNFIS, project number CNFIS-FDI-2022-0259 (O.B.); Natural
Sciences and Engineering Research Council of Canada (RGPIN-2019-05109 and STPGP506284) and the Canadian Foundation
for Innovation (36014) (H.Y.H.C.); the project SustESAdaptation strategies for sustainable ecosystem services and food
security under adverse environmental conditions (CZ.02.1.01/0.0/0.0/16_019/0000797) (E.C.); Consejo de Ciencia y
Tecnología del estado de Durango (2019-01-155) (J.J.C.-R.); Science and Engineering Research Board (SERB), New Delhi,
Government of India (file number PDF/2015/000447)‘Assessing the carbon sequestration potential of different forest
types in Central India in response to climate change’ (J.A.D.); Investissement d’avenir grant of the ANR (CEBA: ANR-10-
LABEX-0025) (G.D.); National Foundation for Science & Technology Development of Vietnam, 106-NN.06-2013.01 (T.V.D.);
Queensland government, Department of Environment and Science (T.J.E.); a Czech Science Foundation Standard grant (19-
14620S) (T.M.F.); European Union Seventh Framework Program (FP7/20072013) under grant agreement number 265171
(L. Finer, M. Pollastrini, F. Selvi); grants from the Swedish National Forest Inventory, Swedish University of Agricultural
Sciences (J.F.); CNPq productivity grant number 311303/2020-0 (A.L.d.G.); DFG grant HE 2719/11-1,2,3; HE 2719/14-1 (A.
Hemp); European Union’s Horizon Europe research project OpenEarthMonitor grant number 101059548, CGIAR Fund INIT-
32-Mitigation and Transformation Initiative for GHG reductions of Agrifood systems RelaTed Emissions (MITIGATE+) (M.H.);
General Directorate of the State Forests, Poland (1/07; OR-2717/3/11; OR.271.3.3.2017) and the National Centre for
Research and Development, Poland (BIOSTRATEG1/267755/4/NCBR/2015) (A.M.J.); Czech Science Foundation 18-10781 S
(S.J.); Danish of Ministry of Environment, the Danish Environmental Protection Agency, Integrated Forest Monitoring
ProgramNFI (V.K.J.); State of São Paulo Research Foundation/FAPESP as part of the BIOTA/FAPESP Program Project
Functional Gradient-PELD/BIOTA-ECOFOR 2003/12595-7 & 2012/51872-5 (C.A.J.); Danish Council for Independent
Researchsocial sciencesgrant DFF 6109–00296 (G.A.K.); Russian Science Foundation project 21-46-07002 for the plot
data collected in the Krasnoyarsk region (V.K.); BOLFOR (D.K.K.); Department of Biotechnology, New Delhi, Government of
India (grant number BT/PR7928/NDB/52/9/2006, dated 29 September 2006) (M.L.K.); grant from Kenya Coastal
Development Project (KCDP), which was funded by World Bank (J.N.K.); Korea Forest Service (2018113A00-1820-BB01,
2013069A00-1819-AA03, and 2020185D10-2022-AA02) and Seoul National University Big Data Institute through the Data
Science Research Project 2016 (H.S.K.); the Brazilian National Council for Scientific and Technological Development (CNPq,
grant 442640/2018-8, CNPq/Prevfogo-Ibama number 33/2018) (C.K.); CSIR, New Delhi, government of India (grant number
38(1318)12/EMR-II, dated: 3 April 2012) (S.K.); Department of Biotechnology, New Delhi, government of India (grant
number BT/ PR12899/ NDB/39/506/2015 dated 20 June 2017) (A.K.); Coordination for the Improvement of Higher
Education Personnel (CAPES) #88887.463733/2019-00 (R.V.L.); National Natural Science Foundation of China (31800374)
(H.L.); project of CEPF RAS ‘Methodological approaches to assessing the structural organization and functioning of forest
ecosystems’ (AAAA-A18-118052590019-7) funded by the Ministry of Science and Higher Education of Russia (N.V.L.);
Leverhulme Trust grant to Andrew Balmford, Simon Lewis and Jon Lovett (A.R.M.); Russian Science Foundation, project 19-
77-30015 for European Russia data processing (O.M.); grant from Kenya Coastal Development Project (KCDP), which was
funded by World Bank (M.T.E.M.); the National Centre for Research and Development, Poland
(BIOSTRATEG1/267755/4/NCBR/2015) (S.M.); the Secretariat for Universities and of the Ministry of Business and
Knowledge of the Government of Catalonia and the European Social Fund (A. Morera); Queensland government,
Department of Environment and Science (V.J.N.); Pinnacle Group Cameroon PLC (L.N.N.); Queensland government,
Department of Environment and Science (M.R.N.); the Natural Sciences and Engineering Research Council of Canada
(RGPIN-2018-05201) (A.P.); the Russian Foundation for Basic Research, project number 20-05-00540 (E.I.P.); European
Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement number
778322 (H.P.); Science and Engineering Research Board, New Delhi, government of India (grant number YSS/2015/000479,
dated 12 January 2016) (P.S.); the Chilean Government research grants Fondecyt number 1191816 and FONDEF number
ID19 10421 (C.S.-E.); the Deutsche Forschungsgemeinschaft (DFG) Priority Program 1374 Biodiversity Exploratories (P.S.);
European Space Agency projects IFBN (4000114425/15/NL/FF/gp) and CCI Biomass (4000123662/18/I-NB) (D.
Schepaschenko); FunDivEUROPE, European Union Seventh Framework Programme (FP7/20072013) under grant
agreement number 265171 (M.S.-L.); APVV 20-0168 from the Slovak Research and Development Agency (V.S.); Manchester
Metropolitan University’s Environmental Science Research Centre (G.S.); the project ‘LIFE+ ForBioSensing PL
Comprehensive monitoring of stand dynamics in Białowieża Forest supported with remote sensing techniques’ which is co-
funded by the EU Life Plus programme (contract number LIFE13 ENV/PL/000048) and the National Fund for Environmental
Protection and Water Management in Poland (contract number 485/2014/WN10/OP-NM-LF/D) (K.J.S.); Global Challenges
Research Fund (QR allocation, MMU) (M.J.P.S.); Czech Science Foundation project 21-27454S (M.S.); the Russian
Foundation for Basic Research, project number 20-05-00540 (N. Tchebakova); Botanical Research Fund, Coalbourn Trust,
Bentham Moxon Trust, Emily Holmes scholarship (L.A.T.); the programmes of the current scientific research of the Botanical
Garden of the Ural Branch of Russian Academy of Sciences (V.A.U.); FCTPortuguese Foundation for Science and
TechnologyProject UIDB/04033/2020. Inventário Florestal NacionalICNF (H. Viana); Grant from Kenya Coastal
Development Project (KCDP), which was funded by World Bank (C.W.); grants from the Swedish National Forest Inventory,
Swedish University of Agricultural Sciences (B.W.); ATTO project (grant number MCTI-FINEP 1759/10 and BMBF 01LB1001A,
01LK1602F) (F.W.); ReVaTene/PReSeD-CI 2 is funded by the Education and Research Ministry of Côte d’Ivoire, as part of the
Debt Reduction-Development Contracts (C2Ds) managed by IRD (I.C.Z.-B.); the National Research Foundation of South
Africa (NRF, grant 89967) (C.H.). The Tropical Plant Exploration Group 70 1 ha plots in Continental Cameroon Mountains are
supported by Rufford Small Grant Foundation, UK and 4 ha in Sierra Leone are supported by the Global Challenge Research
Fund through Manchester Metropolitan University, UK; the National Geographic Explorer Grant, NGS-53344R-18 (A.C.-S.);
University of KwaZulu-Natal Research Office grant (M.J.L.); Universidad Nacional Autónoma de México, Dirección General
de Asuntos de Personal Académico, Grant PAPIIT IN-217620 (J.A.M.). Czech Science Foundation project 21-24186M (R.T., S.
Delabye). Czech Science Foundation project 20-05840Y, the Czech Ministry of Education, Youth and Sports (LTAUSA19137)
and the long-term research development project of the Czech Academy of Sciences no. RVO 67985939 (J.A.). The American
Society of Primatologists, the Duke University Graduate School, the L.S.B. Leakey Foundation, the National Science
Foundation (grant number 0452995) and the Wenner-Gren Foundation for Anthropological Research (grant number 7330)
(M.B.). Research grants from Conselho Nacional de Desenvolvimento Científico e Tecnologico (CNPq, Brazil) (309764/2019;
311303/2020) (A.C.V., A.L.G.). The Project of Sanya Yazhou Bay Science and Technology City (grant number CKJ-JYRC-2022-
83) (H.-F.W.). The Ugandan NBS was supported with funds from the Forest Carbon Partnership Facility (FCPF), the Austrian
Development Agency (ADC) and FAO. FAO’s UN-REDD Program, together with the project on ‘Native Forests and
Community’ Loan BIRF number 8493-AR UNDP ARG/15/004 and the National Program for the Protection of Native Forests
under UNDP funded Argentina’s INBN2.
Affiliations
1 Forest Advanced Computing and Artificial Intelligence Laboratory (FACAI), Department of Forestry and
Natural Resources, Purdue University, West Lafayette, IN, USA.
2 Forestry Division, Food and Agriculture Organization of the United Nations, Rome, Italy.
3 GIP ECOFOR, Paris, France.
4 Department of Forestry and Natural Resources, Purdue University, West Lafayette, IN, USA.
5 Institute for Global Change Biology, School for Environment and Sustainability, University of Michigan,
Ann Arbor, MI, USA.
6 Department of Forest Resources, University of Minnesota, St. Paul, MN, USA.
7 Hawkesbury Institute for the Environment, Western Sydney University, Penrith, New South Wales,
Australia.
8 Crowther Lab, Department of Environmental Systems Science, Institute of Integrative Biology, ETH
Zürich, Zürich, Switzerland.
9 Wageningen Environmental Research, Wageningen University and Research, Wageningen,
Netherlands.
10 Forest Ecology and Forest Management Group, Wageningen University and Research, Wageningen,
Netherlands.
11 Department of Crop and Forest Sciences, University of Lleida, Lleida, Spain.
12 Joint Research Unit CTFCAgrotecnioCERCA, Solsona, Spain.
13Institute of Ecology and Key Laboratory for Earth Surface Processes of the Ministry of Education,
College of Urban and Evironmental Sciences, Peking University, Beijing, China.
14 Northern Research Station, USDA Forest Service, Durham, NH, USA.
15 Center for Biodiversity Dynamics in a Changing World (BIOCHANGE), Department of Biology, Aarhus
University, Aarhus C, Denmark.
16 Section for Ecoinformatics and Biodiversity, Department of Biology, Aarhus University, Aarhus C,
Denmark.
17 School of Biological Sciences, University of Bristol, Bristol, UK.
18 TERRA Teaching and Research Centre, Gembloux Agro Bio-Tech, University of Liege, Gembloux,
Belgium.
19 Manaaki Whenua Landcare Research, Lincoln, New Zealand.
20 Environmental and Life Sciences, Faculty of Science, Universiti Brunei Darussalam, Gadong, Brunei
Darussalam.
21 Centre de Coopération Internationale en Recherche Agronomique pour le Développement,
Montpellier, France.
22 INP-HB (Institut National Polytechnique Félix Houphouet-Boigny), University of Montpellier,
Yamoussoukro, Ivory Coast.
23 Department of Agricultural, Food, Environmental and Animal Sciences, University of Udine, Udine,
Italy.
24 Faculty of Science and Technology, Free University of Bolzano, Bolzano, Italy.
25 Institute of Bioeconomy, CNR, Sesto, Italy.
26 Natural and Built Environments Research Centre, School of Natural and Built Environments, University
of South Australia, Adelaide, South Australia, Australia.
27 Biometris, Wageningen University and Research, Wageningen, Netherlands.
28 Wageningen University & Research, Forest and Nature Conservation Policy Group, Wageningen,
Netherlands.
29 Centre for Econics and Ecosystem Management, Eberswalde University for Sustainable Development,
Eberswalde, Germany.
30 School of Forest, Fisheries, and Geomatics Sciences, Institute of Food & Agricultural Sciences,
University of Florida, Gainesville, FL, USA.
31 Naturalis Biodiversity Center, Leiden, Netherlands.
32 Instituto Nacional de Tecnología Agropecuaria (INTA), Santa Cruz, Argentina.
33 Department of Plant Sciences, University of Cambridge, Cambridge, UK.
34 Faculty of Natural Resources Management, Lakehead University, Thunder Bay, Ontario, Canada.
35 University of Göttingen, Göttingen, Germany.
36 Beijing Forestry University, Beijing, China.
37 University of Stellenbosch, Stellenbosch, South Africa.
38 Białowieża Geobotanical Station, Faculty of Biology, University of Warsaw, Białowieża, Poland.
39 Swiss National Forest Inventory/Swiss Federal Institute for Forest, Snow and Landscape Research
WSL, Birmensdorf, Switzerland.
40 UFR Biosciences, University Félix Houphouët-Boigny, Abidjan, Ivory Coast.
41 Environmental Change Institute, School of Geography and the Environment, University of Oxford,
Oxford, UK.
42 Biodiversity Dynamics, Naturalis Biodiversity Center, Leiden, Netherlands.
43 Center for Latin American Studies, University of Florida, Gainesville, FL, USA.
44 Institute of Botany, Academy of Sciences of the Czech Republic, Trebon, Czech Republic.
45 Faculty of Forestry and Wood Sciences, Czech University of Life Sciences in Prague, Praha-Suchdol,
Czech Republic.
46 Escuela ECAPMA, National Open University and Distance (Colombia) | UNAD, Bogotá, Colombia.
47 Departamento de Ingeniería Agroforestal, Universidad de Santiago de Compostela, Lugo, Spain.
48 Center for Tropical Research, Institute of the Environment and Sustainability, University of California,
Los Angeles, CA, USA.
49 Université Jean Lorougnon Guédé, Daloa, Ivory Coast.
50 Université Officielle de Bukavu, Bukavu, Democratic Republic of Congo.
51 Silviculture and Forest Ecology of the Temperate Zones, University of Göttingen, Goettingen,
Germany.
52 Institut National pour l’Etude et la Recherche Agronomiques, Kinshasa, Democratic Republic of Congo.
53 Norwegian Institute of Bioeconomy Research (NIBIO), Division of Forestry and Forest Resources, Ås,
Norway.
54 European Commission, Joint Research Centre, Ispra, Italy.
55 Compensation International Progress S.A., Bogotá, Colombia.
56 Laboratory of Applied Ecology, University of Abomey-Calavi, Cotonou, Benin.
57 Scientific Services, South African National Parks, Knysna, South Africa.
58 School of Geography, University of Leeds, Leeds, UK.
59 Department of Geomatics, Forest Research Institute, Sekocin Stary, Raszyn, Poland.
60 Proceedings of the National Academy of Sciences, Washington, DC, USA.
61 Department of Evolutionary Anthropology, Duke University, Durham, NC, USA.
62 Department of Environment, Universtité du Cinquantenaire de Lwiro, Bukavu, Democratic Republic of
Congo.
63 Department of Environment, Ghent University, Ghent, Belgium.
64 Department of Green Chemistry and Technology, Ghent University, Ghent, Belgium.
65 Service of Wood Biology, Royal Museum for Central Africa, Tervuren, Belgium.
66 Balai Penelitian dan Pengembangan Lingkungan Hidup dan Kehutanan, Manokwari, Indonesia.
67 Institute of Tropical Forest Conservation, Mbarara University of Science and Technology, Mbarara,
Uganda.
68 Université de Liège, Gembloux Agro-Bio Tech, Gembloux, Belgium.
69 Integrated Center for Research, Development and Innovation in Advanced Materials,
Nanotechnologies, and Distributed Systems for Fabrication and Control (MANSiD), University Stefan
cel Mare of Suceava, Suceava, Romania.
70 Department of Forestry Sciences, ‘Luiz de Queiroz’ College of Agriculture, University of São Paulo,
Piracicaba, Brazil.
71 Bavarian State Institute of Forestry, Freising, Germany.
72 Department of Natural Sciences, Manchester Metropolitan University, Manchester, UK.
73 Facultad de Ciencias Forestales, Universidad Juárez del Estado de Durango, Durango, Mexico.
74 Institute of Biology and Botanical Garden, Martin Luther University Halle-Wittenberg, Halle (Saale),
Germany.
75 German Centre for Integrative Biodiversity Research (iDiv) Halle-Jena-Leipzig, Leipzig, Germany.
76 Development Economics Group, Wageningen University, Wageningen, Netherlands.
77 Rosen Center for Advanced Computing (RCAC), Purdue University, West Lafayette, IN, USA.
78 Department of Biological, Geological and Environmental Sciences (BiGeA), University of Bologna,
Bologna, Italy.
79 Institute of Integrative Biology, ETH Zürich, Zürich, Switzerland.
80 IFER – Institute of Forest Ecosystem Research, Jilove u Prahy, Czech Republic.
81 Global Change Research Institute of the CAS, Brno, Czech Republic.
82 Programa de Pós-graduação em Biologia Vegetal, Instituto de Biologia, Universidade Estadual de
Campinas, Campinas CEP, Biologia, Brazil.
83 Dirección Nacional de Bosques (DNB), Ministerio de Ambiente y Desarrollo Sostenible (MAyDS),
Ciudad Autónoma de Buenos Aires, Buenos Aires, Argentina.
84 Department of International Environment and Development Studies (Noragric), Faculty of Landscape
and Society, Norwegian University of Life Sciences (NMBU), Ås, Norway.
85 Department of Environment and Geography, University of York, York, UK.
86 Department of Environmental Science, School of Engineering and Sciences, SRM University-AP,
Guntur, India.
87 Department of Botany, Dr. Harisingh Gour Vishwavidyalaya (A Central University), Madhya Pradesh,
India.
88 Department of Ecology and Environmental Sciences, Pondicherry University, Puducherry, India.
89 Centre for Structural and Functional Genomics & Quebec Centre for Biodiversity Science, Biology
Department, Concordia University, Montreal, Quebec, Canada.
90 Department of Ecology, Faculty of Science, Charles University, Prague, Czech Republic.
91 Biology Centre of the Czech Academy of Sciences, Institute of Entomology, Ceske Budejovice, Czech
Republic.
92 Cirad, UMR EcoFoG (AgroParistech, CNRS, Inrae, Université des Antilles, Université de la Guyane),
Campus Agronomique, Kourou, French Guiana.
93 Department of Geography, Environment and Geomatics, University of Guelph, Guelph, Ontario,
Canada.
94 National Forest Authority, Kampala, Uganda.
95 Department of Silviculture Foundation, Silviculture Research Institute, Vietnamese Academy of Forest
Sciences, Hanoi, Vietnam.
96 Department of Botany, Faculty of Science, University of South Bohemia, Bohemia, Czech Republic.
97 IPHAMETRA, IRET, CENAREST, Libreville, Gabon.
98 Faculté de Gestion de Ressources Naturelles Renouvelables, Université de Kisangani, Kisangani,
Democratic Republic of Congo.
99 Queensland Herbarium, Department of Environment and Science, Toowong, Queensland, Australia.
100 School of Biological and Behavioural Sciences, Queen Mary University of London, London, UK.
101 Department of Plant Biology, Faculty of Science, University of Yaoundé I, Yaoundé, Cameroon.
102 Natural Resources Institute Finland, Joensuu, Finland.
103 Institute of Plant Sciences, University of Bern, Bern, Switzerland.
104 Department of Forest Resource Management, Swedish University of Agricultural Sciences, Umea,
Sweden.
105 Research and Innovation Centre, Fondazione Edmund Mach, San Michele all’Adige, Italy.
106 Herbário Dr. Roberto Miguel Klein, Universidade Regional de Blumenau, Blumenau, Brazil.
107 Glick Designs, LLC, Hadley, MA, USA.
108 CIIDIR Durango, Instituto Politécnico Nacional, Durango, Mexico.
109 Département des Sciences et Technologies de l’Environnement, Université du Burundi, Bujumbura,
Burundi.
110 Faculté des Sciences, Evolutionary Biology and Ecology Unit, Université Libre de Bruxelles, Brussels,
Belgium.
111 Royal Botanic Garden Edinburgh, Edinburgh, UK.
112 Department of Plant Sciences, University of Oxford, Oxford, UK.
113 Department of Plant Systematics, Bayreuth University, Bayreuth, Germany.
114 Helmholtz GFZ German Research Centre for Geosciences, Section 1 .4 Remote Sensing and
Geoinformatics, Potsdam, Germany.
115 Wild Chimpanzee Foundation, Liberia Representation, Monrovia, Liberia.
116 Centre for Conservation Science, The Royal Society for the Protection of Birds, Sandy, UK.
117 Department of Environment, Laboratory for Wood Technology (UGent-Woodlab), Ghent University,
Ghent, Belgium.
118 AMAP, University of Montpellier, CIRAD, CNRS, INRAE, IRD, Montpellier, France.
119 Department of Forest Science, Tokyo University of Agriculture, Tokyo, Japan.
120 Biology Department, Université Officielle de Bukavu, Bukavu, Democratic Republic of Congo.
121 Institute of Dendrology, Polish Academy of Sciences, Kórnik, Poland.
122 Poznan University of Life Sciences, Faculty of Forestry and Wood Technology, Department of Game
Management and Forest Protection, Poznan, Poland.
123 Department of Geosciences and Natural Resource Management, University of Copenhagen,
Copenhagen, Denmark.
124 Plant Biology Department, Biology Institute, University of Campinas (UNICAMP), Campinas, Brazil.
125 Department of Botany and Plant Pathology, Purdue University, West Lafayette, IN, USA.
126 Institute of Agricultural Research for Development (IRAD), Nkolbisson, Ministry of Scientific Research
and Innovation, Yaounde, Cameroon.
127 Department of Food and Resource Economics, University of Copenhagen, Copenhagen, Denmark.
128 Forestry Faculty, Bauman Moscow State Technical University, Mytischi, Russia.
129 Integrative Research Center, The Field Museum, Chicago, IL, USA.
130 Labo Botanique, Université Félix Houphouët-Boigny, Abidjan, Ivory Coast.
131 Computational and Applied Vegetation Ecology Lab, Ghent University, Ghent, Belgium.
132 Department of Physical and Environmental Sciences, Colorado Mesa University, Grand Junction, CO,
USA.
133 Department of Botany, Dr. Harisingh Gour Vishwavidalaya (A Central University), Sagar, India.
134 Kenya Forestry Research Institute, Department of Forest Resource Assessment, Nairobi, Kenya.
135 Department of Forest Sciences, Seoul National University, Seoul, Republic of Korea.
136 Interdisciplinary Program in Agricultural and Forest Meteorology, Seoul National University, Seoul,
Republic of Korea.
137 National Center for Agro Meteorology, Seoul, Republic of Korea.
138 Research Institute for Agriculture and Life Sciences, Seoul National University, Seoul, Republic of
Korea.
139 Institute of Forestry and Engineering, Estonian University of Life Sciences, Tartu, Estonia.
140 International Institute for Applied Systems Analysis, Laxenburg, Austria.
141 Department of Geoinformatics, Central University of Jharkhand, Ranchi, India.
142 Tartu Observatory, University of Tartu, Tõravere, Estonia.
143 School of Life Sciences, University of KwaZulu-Natal, Pietermaritzburg, South Africa.
144 Department of Forest Engineering, Federal University of Viçosa (UFV), Viçosa, Brazil.
145 Department of Geography, University College London, London, UK.
146 Plant Systematics and Ecology Laboratory (LaBosystE), Higher Teacher’s Training College, University
of Yaoundé I, Yaoundé, Cameroon.
147 Laboratoire d’Écologie et Aménagement Forestier, Département d’Ecologie et de Gestion des
Ressources Végétales, Université de Kisangani, Kisangani, Democratic Republic of Congo.
148 Instituto de Silvicultura e Industria de la Madera, Universidad Juarez del Estado de Durango,
Durango, Mexico.
149 Faculty of Forestry, Qingdao Agricultural University, Qingdao, China.
150 Center for Forest Ecology and Productivity RAS (CEPF RAS), Moscow, Russia.
151 Department of Ecoscience, Aarhus University, Silkeborg, Denmark.
152 Nicholas School of the Environment, Duke University, Durham, NC, USA.
153 Department of Ecology and Evolutionary Biology, University of Arizona, Tucson, AZ, USA.
154 AgroParisTech, UMR AMAP, University of Montpellier, CIRAD, CNRS, INRAE, IRD, Montpellier, France.
155 University of the Sunshine Coast, Sippy Downs, Queensland, Australia.
156 University of York, York, UK.
157 Flamingo Land Ltd., North Yorkshire, UK.
158 Department of Wildlife Management, College of African Wildlife Management, Mweka, Tanzania.
159 Kenya Forestry Research Institute, Headquarters, Nairobi, Kenya.
160 Departamento de Ecología y Recursos Naturales, Facultad de Ciencias, Universidad Nacional
Autónoma de México, Mexico City, Mexico.
161 Ecology and Evolutionary Biology, University of Connecticut, Storrs, CT, USA.
162 Department of Forest Management and Forest Economics, Warsaw University of Life Sciences,
Warsaw, Poland.
163 Tropical Forests and People Research Centre, University of the Sunshine Coast, Maroochydore DC,
Queensland, Australia.
164 Fieldstation Fabrikschleichach, Julius-Maximilians University Würzburg, Würzburg, Germany.
165 Bavarian Forest Nationalpark, Grafenau, Germany.
166 Fakultas Kehutanan, Universitas Papua, Jalan Gunung Salju Amban, Manokwari Papua Barat,
Indonesia.
167 Limbe Botanic Garden, Limbe, Cameroon.
168 Institute of Forestry, Belgrade, Serbia.
169 Tropical Plant Exploration Group (TroPEG), Buea, Cameroon.
170 Department of Ecology and Evolutionary Biology, University of Tennessee, Knoxville, TN, USA.
171 Applied Biology and Ecology Research Unit, University of Dschang, Dschang, Cameroon.
172 Department of Forestry and Natural Resources, University of Kentucky, Lexington, KY, USA.
173 UQAM, Centre for Forest Research, Montreal, Quebec, Canada.
174 V.N. Sukachev Forest Institute of FRC KSC SB RAS, Krasnoyarsk, Russia.
175 Urban Management and Planning, School of Social Sciences, Western Sydney University, Penrith,
New South Wales, Australia.
176 Instituto Nacional de Pesquisas da AmazôniaINPA, Grupo Ecologia. Monitoramento e Uso
Sustentável de Áreas Úmidas MAUA, Manaus, Brazil.
177 Centro de Formação em Ciências Agroflorestais, Universidade Federal do Sul da Bahia, Ilhéus, Brazil.
178 Department of Agriculture, Food, Environment and Forestry, University of Firenze, Firenze, Italy.
179 Technical University of Munich, School of Life Sciences Weihenstephan, Chair of Forest Growth and
Yield Science, Munich, Germany.
180 Centro Agricoltura, Alimenti, Ambiente, University of Trento, San Michele all’Adige, Italy.
181 Department of Biology, University of Florence, Sesto Fiorentino, Italy.
182 MUSEMuseo delle Scienze, Trento, Italy.
183 Infoflora c/o Botanical Garden of Geneva, Geneva, Switzerland.
184 Agricultural Research, Education and Extension Organization (AREEO), Research Institute of Forests
and Rangelands (RIFR), Tehran, Iran.
185 Department of Environmental Sciences, Central University of Jharkhand, Ranchi, India.
186 Institute of International Education Scholar Rescue Fund (IIE-SRF), One World Trade Center, New
York, NY, USA.
187 Centro de Modelación y Monitoreo de Ecosistemas, Facultad de Ciencias, Universidad Mayor,
Santiago, Chile.
188 Vicerrectoría de Investigación y Postgrado, Universidad de La Frontera, Temuco, Chile.
189 Departamento de Silvicultura y Conservación de la Naturaleza, Universidad de
Chile, Santiago, Chile.
190 Рeoples Friendship University of Russia (RUDN University), Moscow, Russia.
191 University of Freiburg, Faculty of Biology, Freiburg,
Germany.
192 Institution with City, Department of Geography, University of Zurich, Zurich, Switzerland.
193 National Forest Centre, Zvolen, Slovak Republic.
194 CNRS-UMR LEEISA, Campus Agronomique, Kourou, French Guiana.
195 Universite de Lorraine, AgroParisTech, INRA, Nancy, France.
196 Center for International Forestry Research (CIFOR), Situ Gede, Bogor Barat, Indonesia.
197 Cirad, University of Montpellier, Montpellier, France.
198 Universidade Federal do Rio Grande do Norte, Departamento de Ecologia, Natal, Brazil.
199 School of Biological Sciences, University of Aberdeen, Aberdeen, UK.
200 Herbarium Kew, Royal Botanic Gardens Kew, London, UK.
201 Faculté des Sciences Appliquées, Université de Mbujimayi, Mbujimayi, Democratic Republic of
Congo.
202 Yale School of Forestry and Environmental Studies, New Haven, CT, USA.
203 Ural State Forest Engineering University, Botanical Garden, Ural Branch of the Russian Academy of
Sciences, Yekaterinburg, Russia.
204 DIBAF Department, Tuscia University, Viterbo, Italy.
205 LINCGlobal, MNCN, CSIC, Madrid, Spain.
206 Plant Ecology and Nature Conservation Group, Wageningen University, AA Wageningen,
Netherlands.
207 Agricultural High School, ESAV, Polytechnic Institute of Viseu, IPV, Viseu, Portugal.
208 Centre for the Research and Technology of Agro-Environmental and Biological Sciences, CITAB,
UTAD, Quinta de Prados, Vila Real, Portugal.
209 Department of Forest Engineering, Universidade Regional de Blumenau, Blumenau, Brazil.
210 Nucleo de Estudos e Pesquisas Ambientais, Universidade Estadual de Campinas, Campinas
(UNICAMP), SP, Campinas, Brazil.
211 International Center for Tropical Botany, Department of Biological Sciences, Florida International
University, Miami, FL, USA.
212 Forest Research Institute, University of the Sunshine Coast, Sippy Downs, Queensland, Australia.
213 Sanya Nanfan Research Institute, Hainan Yazhou Bay Seed Laboratory, Hainan University, Sanya,
China.
214 Kenya Forestry Research Institute, Taita Taveta Research Centre, Wundanyi, Kenya.
215 Department of Wetland Ecology, Institute for Geography and Geoecology, Karlsruhe Institute for
Technology, Rastatt, Germany.
216 Department of Forest Management, Centre for Agricultural Research in Suriname, Paramaribo,
Suriname.
217 Polish State Forests-Coordination Centre for Environmental Projects, Warsaw, Poland.
218 Research Center of Forest Management Engineering of State Forestry and Grassland Administration,
Beijing Forestry University, Beijing, China.
219 Department of Statistics, University of WisconsinMadison, Madison, WI, USA.
220 Institut National Polytechnique Félix Houphouët-Boigny, DFR Eaux, Forêts et Environnement, BP,
Yamoussoukro, Ivory Coast.
221 Centre for Invasion Biology, Department of Mathematical Sciences, Stellenbosch University,
Matieland, South Africa.
222 African Institute for Mathematical Sciences, Muizenberg, South Africa.
... While, a very less tree species richness (10 spp.) with a comparable total tree density of 255 ind. ha −1 and slightly high dominance (10.11 m 2 ha −1 ) were reported in high fire zones in tropical deciduous forests of Bhoramdeo Wildlife Sanctuary, Chhattisgarh (Jhariya et al., 2014) and are primarily driven by latitudinal gradients (Liang et al., 2022). Altered species composition, diversity, and reduced seedling density were also been reported in areas of dry deciduous forest with the shortest fire return intervals compared to forest patches with lower fire frequency in the moist deciduous forests of Nilgiri Biosphere Reserve, Western Ghats (Kodandapani et al., 2009). ...
Article
Full-text available
The present study highlights the forest fire hotspots based on daily forest fire occurrence during 2000–2020 and its subsequent long-term impacts on different vegetation types in Central India. The MODIS (MCD14DL) based study exhibited the recurrence of fire incidences in central and southern parts of forests and affected substantial parts of the deciduous broadleaf forest (31.56% of select forest type) and deciduous needleleaf forest (30.23%), shrubland (7.52%), mixed forest (9.80%), with the major forest fire peaks during March and April, while negligible fire incidences were observed in plantation and grassland. 72% of field transects (157 out of the 218 field transects) were observed under high anthropogenic influences as evidenced by a large number of cut stumps. Six species (out of 106 studies tree species) including Tectona grandis L.f. (occurrence in 98% of fire hotspots), Butea monosperma (Lam.) Taubert (66%), were observed to be highly fire-resistant, recorded in >50% of the forest fire hotspots regions. The major forest fires hotspots (p < 0.01) were observed in the Malwa plateau, Vindhyan ranges, and Satpura range during 2000–2020, as observed with high fire intensity. While the forest in the northern parts attributed a cold spot with reference to forest fires (p < 0.01). The study elucidated the high susceptibility of forest fire in deciduous forests in Central India and necessitated proper management of forest fires by encouraging fire-resistant species, and prioritization of regions under fire hazards through the adoption of preparedness and response strategies to minimize the forest fire impacts.
Preprint
Full-text available
Ecosystem stability reveals how ecosystems respond to global change over time. Yet, the focus of past research on small spatial grains and extents overlooks scale dependence and how broad-scale environmental gradients shape stability. Here, we use forest inventory data covering a broad latitudinal gradient from the temperate to the tropical zone to examine cross-scale variation in stability of aboveground biomass and underlying stabilizing mechanisms. While stability did not shift systematically with latitude at either spatial grain, we found evidence that species asynchrony increased towards the tropics at the small spatial grain while species stability decreased at both spatial grains. Moreover, latitude stabilized forest communities via its effects on both stabilizing mechanisms, which compensated for the weak and destabilizing effects of species richness. Yet, the trade-off in the relative importance of species stability and species asynchrony for stability was not mediated by latitude, suggesting that context-dependent factors - to a greater extent than macroecological ones - underlie large-scale patterns of stability. Our results highlight the crucial role of species asynchrony and species stability in determining ecosystem stability across broad-scale environmental gradients, suggesting that conserving biodiversity alone may not be sufficient for stabilizing naturally-assembled forest ecosystems.
Technical Report
Full-text available
Our data shows that the protection of 80% of the Amazon is necessary and possible, but above all, urgent. If the current trend of deforestation continues, the Amazon as we know it today would not reach 2025. This report presents ten compelling conclusions resulting from the comparison of the status of the Priority Areas by territorial management regimes: protected areas, indigenous territories and undesignated areas. It also proposes a comparison by country.
Article
Distribution data from the North American Breeding Bird Survey shows an increasing mismatch between species distribution and climate.
Article
Full-text available
Earth's forests face grave challenges in the Anthropocene, including hotter droughts increasingly associated with widespread forest die-off events. But despite the vital importance of forests to global ecosystem services, their fates in a warming world remain highly uncertain. Lacking is quantitative determination of commonality in climate anomalies associated with pulses of tree mortality-from published, field-documented mortality events-required for understanding the role of extreme climate events in overall global tree die-off patterns. Here we established a geo-referenced global database documenting climate-induced mortality events spanning all tree-supporting biomes and continents, from 154 peer-reviewed studies since 1970. Our analysis quantifies a global "hotter-drought fingerprint" from these tree-mortality sites-effectively a hotter and drier climate signal for tree mortality-across 675 locations encompassing 1,303 plots. Frequency of these observed mortality-year climate conditions strongly increases nonlinearly under projected warming. Our database also provides initial footing for further community-developed, quantitative, ground-based monitoring of global tree mortality.
Article
Full-text available
One of the most fundamental questions in ecology is how many species inhabit the Earth. However, due to massive logistical and financial challenges and taxonomic difficulties connected to the species concept definition, the global numbers of species, including those of important and well-studied life forms such as trees, still remain largely unknown. Here, based on global ground-sourced data, we estimate the total tree species richness at global, continental, and biome levels. Our results indicate that there are ∼73,000 tree species globally, among which ∼9,000 tree species are yet to be discovered. Roughly 40% of undiscovered tree species are in South America. Moreover, almost one-third of all tree species to be discovered may be rare, with very low populations and limited spatial distribution (likely in remote tropical lowlands and mountains). These findings highlight the vulnerability of global forest biodiversity to anthropogenic changes in land use and climate, which disproportionately threaten rare species and thus, global tree richness.
Article
Full-text available
Significance Tree diversity is fundamental for forest ecosystem stability and services. However, because of limited available data, estimates of tree diversity at large geographic domains still rely heavily on published lists of species descriptions that are geographically uneven in coverage. These limitations have precluded efforts to generate a global perspective. Here, based on a ground-sourced global database, we estimate the number of tree species at biome, continental, and global scales. We estimated a global tree richness (≈73,300) that is ≈14% higher than numbers known today, with most undiscovered species being rare, continentally endemic, and tropical or subtropical. These results highlight the vulnerability of global tree species diversity to anthropogenic changes.
Article
Full-text available
Forest restoration increases global forest area and ecosystem services such as primary productivity and carbon storage. How tree species functional composition impacts the provisioning of these services as forests develop is sparsely studied. We used 10-year data from 478 plots with 191,200 trees in a forest biodiversity experiment in subtropical China to assess the relationship between community productivity and community-weighted mean (CWM) or functional diversity (FD) values of 38 functional traits. We found that effects of FD values on productivity became larger than effects of CWM values after 7 years of forest development and that the FD values also became more reliable predictors of productivity than the CWM values. In contrast to CWM, FD values consistently increased productivity across ten different species-pool subsets. Our results imply that to promote productivity in the long term it is imperative for forest restoration projects to plant multispecies communities with large functional diversity.
Article
Full-text available
Open access to global forest data, especially ground-measured (in situ) records, is critical for saving the world's forest systems. Integrated approaches to achieve sustainable data openness will involve legal assurances, shared ethics, innovative funding schemes and capacity development.
Article
Full-text available
Anthropogenic climate change is expected to impact ecosystem structure, biodiversity and ecosystem services in Africa profoundly. We used the adaptive Dynamic Global Vegetation Model (aDGVM), which was originally developed and tested for Africa, to quantify sources of uncertainties in simulated African potential natural vegetation towards the end of the 21st century. We forced the aDGVM with regionally downscaled high‐resolution climate scenarios based on an ensemble of six general circulation models (GCMs) under two representative concentration pathways (RCPs 4.5 and 8.5). Our study assessed the direct effects of climate change and elevated CO2 on vegetation change and its plant‐physiological drivers. Total increase in carbon in aboveground biomass in Africa until the end of the century was between 18% to 43% (RCP4.5) and 37% to 61% (RCP8.5) and was associated with woody encroachment into grasslands and increased woody cover in savannas. When direct effects of CO2 on plants were omitted, woody encroachment was muted and carbon in aboveground vegetation changed between –8 to 11% (RCP 4.5) and –22 to –6% (RCP8.5). Simulated biome changes lacked consistent large‐scale geographical patterns of change across scenarios. In Ethiopia and the Sahara/Sahel transition zone, the biome changes forecast by the aDGVM were consistent across GCMs and RCPs. Direct effects from elevated CO2 were associated with substantial increases in water use efficiency, primarily driven by photosynthesis enhancement, which may relieve soil moisture limitations to plant productivity. At the ecosystem level, interactions between fire and woody plant demography further promoted woody encroachment. We conclude that substantial future biome changes due to climate and CO2 changes are likely across Africa. Because of the large uncertainties in future projections, adaptation strategies must be highly flexible. Focused research on CO2 effects, and improved model representations of these effects will be necessary to reduce these uncertainties. Climate change and elevated CO2 are expected to drive vegetation changes in Africa. We used an ensemble of dynamic vegetation model simulations to assess the impacts of these drivers on carbon stocks and biomes until 2099. Climate change and elevated CO2 led to an 18% to 61% increase in carbon stocks, which was primarily driven by CO2 fertilization. Associated biome changes are likely across Africa, especially changes from savanna to forest. Disabling CO2 fertilization resulted in a −22% to +11% change in carbons stocks. These large uncertainties in future projections imply that adaptation strategies need to be flexible.
Article
Full-text available
A spatially explicit global map of tree symbioses with nitrogen-fixing bacteria and mycorrhizal fungi reveals that climate variables are the primary drivers of the distribution of different types of symbiosis.
Article
Phylogenetic endemism describes the extent to which unique phylogenetic lineages are constrained to restricted geographic areas. Previous studies indicate that species endemism is related to both past and modern climate, but studies of phylogenetic endemism are relatively rare and mainly focused on smaller regions. Here, we provide the first assessment of the patterns of species and phylogenetic endemism in angiosperm trees across the Northern Hemisphere as well as the relative importance of modern climate and glacial–interglacial climate change as drivers of these patterns. Northern Hemisphere. Angiosperm trees. Using tree assemblages at the scale of 100 km × 100 km grid cells and simultaneous autoregressive (SAR) models, we assessed the relationships between species endemism, phylogenetic endemism and modern climate variables, Last Glacial Maximum (LGM) to present temperature velocity. Species and phylogenetic endemism were associated with both modern climate and glacial–interglacial climate change, with higher values in areas with stable historical climate and warmer and wetter modern conditions. Notably, the multivariate SAR analyses showed that the combinations of variables with highest Akaike’s information criterion (AIC) weight always included both LGM–present climate instability and modern climate, that is, modern precipitation and temperature. Our results show that high phylogenetic endemism is partially dependent on long‐term climate stability, highlighting the threat posed by future climate changes to the preservation of rare, phylogenetically distinct lineages of trees.
Article
The latitudinal diversity gradient (LDG) is one of the most widely studied patterns in ecology, yet no consensus has been reached about its underlying causes. We argue that the reasons for this are the verbal nature of existing hypotheses, the failure to mechanistically link interacting ecological and evolutionary processes to the LDG, and the fact that empirical patterns are often consistent with multiple explanations. To address this issue, we synthesize current LDG hypotheses, uncovering their eco-evolutionary mechanisms, hidden assumptions, and commonalities. Furthermore, we propose mechanistic eco-evolutionary modeling and an inferential approach that makes use of geographic, phylogenetic, and trait-based patterns to assess the relative importance of different processes for generating the LDG.