Climatic controls of decomposition drive the global
biogeography of forest-tree symbioses
B. S. Steidinger1,15, T. W. Crowther2,15*, J. Liang3,4,15*, M. E. Van Nuland1, G. D. A. Werner5, P. B. Reich6,7, G. Nabuurs8,
S. de-Miguel9,10, M. Zhou3, N. Picard11, B. Herault12,13, X. Zhao4, C. Zhang4, D. Routh2, GFBI consortium14 & K. G. Peay1*
The identity of the dominant root-associated microbial symbionts
in a forest determines the ability of trees to access limiting
nutrients from atmospheric or soil pools1,2, sequester carbon3,4
and withstand the effects of climate change5,6. Characterizing the
global distribution of these symbioses and identifying the factors
that control this distribution are thus integral to understanding
the present and future functioning of forest ecosystems. Here we
generate a spatially explicit global map of the symbiotic status of
forests, using a database of over 1.1million forest inventory plots
that collectively contain over 28,000 tree species. Our analyses
indicate that climate variables—in particular, climatically
controlled variation in the rate of decomposition—are the
primary drivers of the global distribution of major symbioses.
We estimate that ectomycorrhizal trees, which represent only 2%
of all plant species7, constitute approximately 60% of tree stems
on Earth. Ectomycorrhizal symbiosis dominates forests in which
seasonally cold and dry climates inhibit decomposition, and is the
predominant form of symbiosis at high latitudes and elevation. By
contrast, arbuscular mycorrhizal trees dominate in aseasonal, warm
tropical forests, and occur with ectomycorrhizal trees in temperate
biomes in which seasonally warm-and-wet climates enhance
decomposition. Continental transitions between forests dominated
by ectomycorrhizal or arbuscular mycorrhizal trees occur relatively
abruptly along climate-driven decomposition gradients; these
transitions are probably caused by positive feedback effects between
plants and microorganisms. Symbiotic nitrogen fixers—which
are insensitive to climatic controls on decomposition (compared
with mycorrhizal fungi)—are most abundant in arid biomes with
alkaline soils and high maximum temperatures. The climatically
driven global symbiosis gradient that we document provides a
spatially explicit quantitative understanding of microbial symbioses
at the global scale, and demonstrates the critical role of microbial
mutualisms in shaping the distribution of plant species.
Microbial symbionts strongly influence the functioning of forest eco-
systems. Root-associated microorganisms exploit inorganic, organic
and/or atmospheric forms of nutrients that enable plant growth
mine how trees respond to increased concentrations
the respiratory activity of soil microorganisms
and affect plant spe-
cies diversity by altering the strength of conspecific negative density
. Despite the growing recognition of the importance of
root symbioses for forest functioning1,6,10 and the potential to inte-
grate symbiotic status into Earth system models that predict functional
changes to the terrestrial biosphere10, we lack spatially explicit quanti-
tative maps of root symbioses at the global scale. Quantitative maps of
tree symbiotic states would link the biogeography of functional traits
of belowground microbial symbionts with their 3.1trillion host trees
which arespread across Earth’s forests, woodlands and savannahs.
The dominant guilds of tree root symbionts—arbuscular mycorrhizal
fungi, ectomycorrhizal fungi, ericoid mycorrhizal fungi and nitro-
gen-fixing bacteria (N-fixers)—are all based on the exchange of plant
photosynthate for limiting macronutrients. Arbuscular mycorrhizal
symbiosis evolved nearly 500million years ago, and ectomycorrhizal,
ericoid mycorrhizal and N-fixer plant taxa have evolved multiple times
from an arbuscular-mycorrhizal basal state. Plants that are involved in
arbuscular mycorrhizal symbiosis comprise nearly 80% of all terres-
trial plant species; these plants principally rely on arbuscular mycor
rhizal fungi for enhancing mineral phosphorus uptake
. In contrast
to arbuscular mycorrhizal fungi, ectomycorrhizal fungi evolved from
multiple lineages of saprotrophic ancestors and, as a result, some ecto-
mycorrhizal fungi are capable of directly mobilizing organic sources
of soil nutrients (particularly nitrogen)
. Associations with ectomycor-
rhizal fungi—but not arbuscular mycorrhizal fungi—have previously
been shown to enable trees to accelerate photosynthesis in response
to increased concentrations of atmospheric CO
when soil nitrogen
is limiting6, and to inhibit soil respiration by decomposer microor-
ganisms3,8. Because increased plant photosynthesis and decreased soil
respiration both reduce atmospheric CO2 concentrations, the ecto-
mycorrhizal symbiosis is associated with buffering the Earth’s climate
against anthropogenic change.
In contrast to mycorrhizal fungi, which extract nutrients from the
soil, symbiotic N-fixers (Rhizobia and Actinobacteria) convert atmos-
to plant-usable forms. Symbiotic N-fixers are responsible for
a large fraction of biological soil-nitrogen inputs, which can increase
nitrogen availability in forests in which N-fixers are locally abundant13.
Symbioses with either N-fixers or ectomycorrhizal fungi often demand
more plant photosynthate than does arbuscular mycorrhizal symbio-
sis12,14,15. Because tree growth and reproduction are limited by access
to inorganic, organic and atmospheric sources of nitrogen, the distri-
bution of root symbioses is likely to reflect environmental conditions
that maximize the cost:benefit ratio of symbiotic exchange as well as
physiological constraints on the different symbionts.
One of the earliest efforts16 to understand the functional bio-
geography of plant root symbioses categorically classified biomes by
their perceived dominant mycorrhizal type, and hypothesized that
seasonal climates favour hosts that associate with ectomycorrhizal fungi
(owing to the ability of these hosts to compete directly for organic nitro-
gen). By contrast, it has more recently been proposed that sensitivity
to low temperatures has prevented N-fixers from dominating outside
of the tropics, despite the potential for nitrogen fixation to alleviate
nitrogen limitation in boreal forests
. However, global-scale tests of
1Department of Biology, Stanford University, Stanford, CA, USA. 2Department of Environmental Systems Science, ETH Zürich, Zürich, Switzerland. 3Department of Forestry and Natural Resources,
Purdue University, West Lafayette, IN, USA. 4Research Center of Forest Management Engineering of State Forestry and Grassland Administration, Beijing Forestry University, Beijing, China.
5Department of Zoology, University of Oxford, Oxford, UK. 6Department of Forest Resources, University of Minnesota, St Paul, MN, USA. 7Hawkesbury Institute for the Environment, Western Sydney
University, Penrith, New South Wales, Australia. 8Wageningen University and Research, Wageningen, The Netherlands. 9Department of Crop and Forest Sciences - Agrotecnio Center
(UdL-Agrotecnio), Universitat de Lleida, Lleida, Spain. 10Forest Science and Technology Centre of Catalonia (CTFC), Solsona, Spain. 11Food and Agriculture Organization of the United Nations,
Rome, Italy. 12Cirad, UPR Forêts et Sociétés, University of Montpellier, Montpellier, France. 13Department of Forestry and Environment, National Polytechnic Institute (INP-HB), Yamoussoukro,
Côte d’Ivoire. 14A list of participants and their affiliations appears in the online version of the paper. 15These authors contributed equally: B. S. Steidinger, T. W. Crowther, J. Liang.
*e-mail: firstname.lastname@example.org; email@example.com; firstname.lastname@example.org
404 | NATURE | VOL 569 | 16 MAY 2019
these proposed biogeographical patterns and their climate drivers are
lacking. To address this, we compiled a global ground-sourced survey
database to reveal the numerical abundances of each type of symbiosis
across the globe. Such a database is essential for identifying the poten-
tial mechanisms that underlie transitions in forest symbiotic state along
We determined the abundance of tree symbioses using an extension
of the plot-based Global Forest Biodiversity (GFB) database that we
term the GFBi; this extended database contains over 1.1million forest
inventory plots of individual-based measurement records, from which
we derive abundance information for entire tree communities (Fig.1).
Using published literature on the evolutionary histories of mycorrhizal
and N-fixer symbioses, we assigned plant species from the GFBi to
one of five root-associated symbiotic guilds: arbuscular mycorrhizal,
ectomycorrhizal, ericoid mycorrhizal, N-fixer and weakly arbuscular
or non-mycorrhizal. We then used the random-forest algorithm with
K-fold cross-validation to determine the importance and influence of
variables related to climate, soil chemistry, vegetation and topography
on the relative abundance of each tree symbiotic guild (Fig.2). Because
decomposition is the dominant process by which soil nutrients become
available to plants, we calculated annual and quarterly decomposition
coefficients according to the Yasso07 model
, which describes how
temperature and precipitation gradients influence mass-loss rates
of different chemical pools of leaf litter (with parameters fit using a
previous global study of leaf decomposition) (Fig.3, Supplementary
Fig.5). Finally, we projected our predictive models across the globe
over the extent of global biomes that fell within the multivariate distri-
bution of our model training data (Fig.4, Supplementary Figs.14, 15;
seeMethods for full description).
Our analysis shows that each one of the three most-numerically
abundant guilds of tree symbiosis has a reliable environmental signa-
ture, with the four most-important predictors accounting for 81, 79 and
52% of the total variability in relative basal area for ectomycorrhizal,
arbuscular mycorrhizal and N-fixer symbioses, respectively. Given the
relative rarity of ericoid mycorrhizal and weakly arbuscular or non-
mycorrhizal symbiotic states among trees, models for these symbioses
lack strong predictive power—although the raw data do identify
some local abundance hotspots for ericoid mycorrhizal symbiosis
(Supplementary Fig.1). As a result, we focus on the three major tree
symbiotic states (ectomycorrhizal, arbuscular mycorrhizal and N-fixer).
Despite the fact that data from North America and South America
constitute 65% of the training data (at the 1°-by-1° grid scale), our
models accurately predict the proportional abundances of the three
major symbioses across all major geographical regions (Supplementary
Fig.10). The high performance of our models—which is robust to
K-fold cross-validation and to rarefying samples such that all conti-
nents are represented with equal depth (Supplementary Figs.11, 12)—
suggests that regional variations in climate (including indirect effects
on decomposition) and soil pH (for N-fixers) are the primary factors
that influence the relative dominance of each guild at the global scale;
geographical origin explained only approximately 2–5% of the
variability in residual relative abundance (Supplementary Table8,
Whereas a recent global analysis of root traits concluded that plant
evolution has favoured a reduced dependence on mycorrhizal fungi
we find that trees that associate with the relatively more carbon-
demanding and recently derived ectomycorrhizal fungi
the dominant tree symbiosis. By taking the average proportion of ecto-
mycorrhizal trees, weighted by spatially explicit global predictions for
tree stem density11, we estimate that approximately 60% of tree stems on
earth are ectomycorrhizal—despite the fact that only 2% of overall plant
species associate with ectomycorrhizal fungi (versus nearly 80% that
associate with arbuscular mycorrhizal fungi)
. Outside of the tropics,
the estimate for the relative abundance of ectomycorrhizal symbiosis
increases to approximately 80% of trees.
Turnover among the major symbiotic guilds results in a tri-modal
latitudinal abundance gradient, in which the proportion of ectomyc-
orrhizal trees increases (and the proportion of arbuscular mycorrhizal
trees decreases) with distance from the equator and the upper quan-
tiles of nitrogen-fixing trees reach a peak in abundance in the arid
zone at around 30°N or S (Figs.3a and 4). These trends are driven by
abrupt transitional regions along continental climatic gradients (Fig.2),
which skew the distribution of symbioses among biomes (Fig.3a) and
drive strong patterns across geographical and topographic features that
influence climate. Moving north or south from the equator, the first
transitional zone separates warm (aseasonal) tropical broadleaf forests
dominated by arbuscular mycorrhizal symbiosis (>75% median basal
area versus 8% for ectomycorrhizal trees) from the rest of the world forest
system, which is dominated by ectomycorrhizal symbiosis (Figs.2a, b
and 3a). The transition zone occurs across the globe at around 25° N
and S, just beyond the dry tropical broadleaf forests (which have 25% of
their basal area consisting of ectomycorrhizal trees) (Fig.3a) in which
average monthly temperature variation reaches 3–5°C (temperature
seasonality) (Fig.2a, b).
Moving further north or south, the second transitional climate zone
separates regions in which decomposition coefficients during the
warmest quarter of the year are less than two (Fig.3b gives the asso-
ciated temperature and precipitation ranges). In North America and
China, this transition zone occurs around 50°N, and separates the
mixed arbuscular mycorrhizal and ectomycorrhizal temperate forests
Fig. 1 | The global distribution of GFBi training data. The global map
has n=2,768 grid cells at a resolution of 1°×1° latitude and longitude.
Cells are coloured in the red, green and blue spectrum according to
the percentage of total tree basal area occupied by N-fixer, arbuscular
mycorrhizal (AM) and ectomycorrhizal (EM) tree symbiotic guilds, as
indicated by the ternary plot.
16 MAY 2019 | VOL 569 | NATURE | 405
from their neighbouring ectomycorrhizal-dominated boreal forests
(75 and 100% of their basal area, respectively, consisting of ectomy-
corrhizal trees) (Fig.3a). This transitional decomposition zone is not
present in western Europe, which has a temperature seasonality of
>5°C but lacks sufficiently wet summers to accelerate decomposition
coefficients beyond the values that are associated with mixed arbuscular
mycorrhizal and ectomycorrhizal forests. The latitudinal transitions in
symbiotic state observed among biomes are mirrored by within-biome
transitions along elevation gradients. For example, in tropical Mexico
decomposition coefficients of less than two during the warmest and
wetter quarters of the year occur along the slopes of the Sierra Madre,
where a mixture of arbuscular-mycorrhizal and N-fixer woodlands
in arid climates transition to ectomycorrhizal-dominated tropical
coniferous forests (75% basal area) (Figs.3a and 4a–c, Supplementary
Figs.16–18). The Southern Hemisphere—which lacks the landmass
to support extensive boreal forests—experiences a similar latitudinal
transition in decomposition rates along the ecotone that separates its
tropical and temperate biomes, at around 28°S.
The abrupt transitions that we detected between forest symbiotic
states along environmental gradients suggest that positive feedback
effects may exist between climatic and biological controls of decom-
position10,20. In contrast to arbuscular mycorrhizal fungi, some
Fig. 2 | A small number of environmental variables predict the majority
of global turnover in forest symbiotic status. a–c, Partial feature
contributions of different environmental variables to forest symbiotic
state. Each row plots the shape of the contribution of the four most-
important predictors of the proportion of tree basal area that belongs
to the ectomycorrhizal (a), arbuscular mycorrhizal (b) and N-fixer (c)
symbiotic guilds (n=2,768). Variables are listed in declining importance
from left to right, as determined by the increase in node purity (inc. node
purity), and with points coloured with a red to green to blue gradient
according to their position on the x axis of the most-important variable
(left-most panels for each guild), allowing cross-visualization between
predictors. Each panel lists two measures of variable importance; inc. node
purity (used for sorting) and percentage increase in mean square error (%
inc. MSE) (see Supplementary Information). The abundance of each type
of symbiont transitions sharply along climatic gradients, which suggests
that sites near the threshold are particularly vulnerable to switching their
dominant symbiont guild as climate changes. Warmest and wettest quarter,
the warmest and wettest quarters of the year, respectively.
Fig. 3 | The distribution of forest symbiotic status across biomes
is related to climatic controls over decomposition. a, Biome level
summaries of the median±1 quartile of the predicted percentage of
tree basal area per biome for ectomycorrhizal, arbuscular mycorrhizal
and N-fixer symbiotic guilds (n=100 random samples per biome).
b, The dependency of decomposition coefficients (k, solid and dotted
lines; in the region between the solid lines, the model transitions abruptly
between dominant symbiotic status) on temperature and precipitation
during the warmest quarter with respect to predicted dominance of
mycorrhizal symbiosis. The transition from arbuscular mycorrhizal forests
to ectomycorrhizal forests between k=1 and k=2 is abrupt, which is
consistent with positive feedback between climatic and biological controls
406 | NATURE | VOL 569 | 16 MAY 2019
ectomycorrhizal fungi can use oxidative enzymes to mineralize organic
nutrients from leaf litter and convert nutrients to plant-usable forms
Relative to arbuscular mycorrhizal trees, the leaf litter of ectomycor-
rhizal trees is also chemically more resistant to decomposition, and
has higher C:N ratios and higher concentrations of decomposition-
inhibiting secondary compounds10. Thus, ectomycorrhizal leaf
litter can exacerbate climatic barriers to decomposition and promote
conditions in which ectomycorrhizal fungi have nutrient-acquiring
abilities that are superior to those of arbuscular mycorrhizal fungi
recent game-theoretical model has shown that positive feedbackeffects
between plants and soil nutrients can lead to local bistability in myc-
orrhizal symbiosis22. Such positive feedbackeffects are also known
to cause abrupt ecosystem transitions along smooth environmental
gradients between woodlands and grasses: trees suppress fires (which
promotes seedling recruitment), whereas grass fuels fires that kill
. The existence of abrupt transitions also suggests that
forests in transitional regions along decomposition gradients should
be susceptible to marked turnover in symbiotic state with future
To illustrate the sensitivity of global patterns of tree symbiosis to
climate change, we use the relationships thatwe observed for current
climates to project potential changes in the symbiotic status of forests
in the future. Relative to our global predictions that use the most-recent
climate data, model predictions that use the projected climates for 2070
suggest that the abundance of ectomycorrhizal trees will decline by as
much as 10% (using a relative concentration pathway of 8.5W per m
(Supplementary Fig.24). Our models predict that the largest declines
in ectomycorrhizal abundance will occur along the boreal–temperate
ecotone, where small increases in climatic decomposition coefficients
cause abrupt transitions to arbuscular mycorrhizal forests (Fig.2a, b).
Although our model does not estimate the time lag between climate
change and forest community responses, the predicted decline in ecto-
mycorrhizal trees corroborates the results of common garden transfer
and simulated warming experiments, which have demonstrated that
some important ectomycorrhizal hosts will decline at the boreal–
temperate ecotone under altered climate conditions24.
The change in dominant nutrient-exchange symbioses along
climate gradients highlights the interconnection between atmospheric
and soil compartments of the biosphere. The transition from arbus-
cular mycorrhizal to ectomycorrhizal dominance corresponds with
a shift from phosphorus to nitrogen limitation of plant growth with
. Including published global projections of total
soil nitrogen or phosphorus, microbial nitrogen or soil phosphorus
fractions (labile, occluded, organic and apatite) did not increase the
amount of variation explained by the model, or alter the variables iden-
tified as most important; we therefore dropped these projections from
our analysis. However, our finding that climatic controls of decom-
position are the best predictors of dominant mycorrhizal associations
provides a mechanistic link between symbiont physiology and climatic
controls on the release of soil nutrients from leaf litter. These findings
are consistent with Read’s hypothesis16 that slow decomposition at
high latitudes favours ectomycorrhizal fungi owing to their increased
capacity to liberate organic nutrients
. Thus, although more experi-
ments are necessary to understand the specific mechanism by which
nutrient competition favours the dominance of arbuscular mycorrhizal
or ectomycorrhizal symbioses
, we propose that the latitudinal and
elevational transitions from arbuscular-mycorrhizal-dominated to
ectomycorrhizal-dominated forests be named ‘Read’s rule’.
Our analyses focus on prediction at large spatial scales that are appro-
priate to the available data, but our findings with respect to Read’s rule
also provide insight into how soil factors structure the fine-scale distri-
butions of tree symbioses within our grid cells. For example, at a coarse
scale, we find that ectomycorrhizal trees are relatively rare in many wet
tropical forests; however, individual tropical sites in our raw data span
the full range from 0 to 100% basal area dominated by ectomycorrhizal
trees. In much of the wet tropics, these ectomycorrhizal-dominated
sites exist as outliers within a matrix of predominantly arbuscular
mycorrhizal trees. In an apparent exception that proves Read’s rule, in
aseasonal, warm neotropical climates—which accelerate leaf decom-
position and promote the regional dominance of arbuscular mycor-
rhizal symbiosis (Fig.3)—ectomycorrhizal-dominated tree stands can
develop in sites in which poor soils and recalcitrant litter slow the rates
of decomposition and nitrogen mineralization18,27. Landscape-scale
variation in the relative abundance of symbiotic states also changes
along climate gradients: variability is highest in xeric and temperate
biomes (Supplementary Figs.3, 4), which suggests that the potential of
local nutrient variability to favour particular symbioses is contingent
Whereas ectomycorrhizal trees are associated with ecosystems in
which plant growth is thought to be primarily nitrogen-limited, N-fixer
trees are not. Our results highlight the global extent of the apparent
‘nitrogen cycling paradox’ in which some metrics suggest that nitrogen
limitation is greater in the temperate zone25,26 and yet nitrogen-fixing
trees are relatively more common in the tropics15,28 (Fig.3a). We find
that N-fixers—which we estimate represent 7% of all trees—dominate
forests with annual maximum temperatures >35°C and alkaline soils,
particularly in North America and Africa (Fig.2c). N-fixers have the
highest relative abundance in xeric shrublands (24%), tropical savan-
nahs (21%) and dry broadleaf forest biomes (20%), but are nearly absent
from boreal forests (<1%) (Figs.3a and 4). The decline in N-fixer tree
EM tree basal area (%)
AM tree basal area (%)
048 36 40
N-xer tree basal area (%)
Fig. 4 | Global maps of predicted forest-tree symbiotic state.
a–c, Maps (left) and latitudinal gradients (right; solid line indicates
median; coloured ribbon spans the range between the 5% and 95%
quantiles) of the percentage of tree basal area for ectomycorrhizal
(a), arbuscular mycorrhizal (b) and N-fixer (c) symbiotic guilds. All
projections are displayed on a 0.5°-by-0.5° latitude and longitude scale.
n=28,454 grid cells.
16 MAY 2019 | VOL 569 | NATURE | 407
abundance with increasing latitude that we observed is also associ-
ated with a previously documented latitudinal shift in the identity of
nitrogen-fixing microorganisms, from facultative rhizobial N-fixers in
tropical forests to obligate actinorhizal N-fixers in temperate forests
Our data are not capable of fully disentangling the several hypotheses
that have previously been proposed to reconcile the nitrogen cycling
paradox15. However, our results are consistent with the model
prediction17 and regional empirical evidence19,29,30 that nitrogen-fixing
trees are particularly important in arid biomes. Based primarily on
the observed positive nonlinear association of the relative abundance
of N-fixers with the mean temperature of the hottest month (Fig.2c),
our models predict a twofold increase in relative abundance of N-fixers
when transitioning from humid to dry tropical forest biomes (Fig.3a).
Although soil microorganisms are a dominant component of forests
in terms of both diversity and ecosystem functioning5,6,10, identifying
global-scale microbial biogeographical patterns remains an ongoing
research priority. Our analyses confirm that Read’s rule—which is
one of the first proposed biogeographical rules specific to microbial
symbioses—successfully describes global transitions between myc-
orrhizal guilds. More generally, climate driven turnover among the
major symbioses between plants and microorganisms represents a
fundamental biological pattern in the Earth system, as forests tran-
sition from low-latitude arbuscular mycorrhizal through N-fixer to
high-latitude ectomycorrhizal ecosystems. The predictions of our
model (available in the Supplementary Data as global raster layers)
can now be used to represent these critical ecosystem variations in
global biogeochemical models that are used to predict climate–
biogeochemical feedbackeffects within and between trees, soils and the
atmosphere. Additionally, the raster layer that contains the proportion
of nitrogen-fixing trees can be used to map potential symbiotic nitro-
gen fixation, which links atmospheric pools of carbon and nitrogen.
Future work can extend our findings to incorporate multiple plant
growth forms and non-forested biomes (in which similar patterns are
likely to exist) to generate a complete global perspective. Our predictive
maps leverage a comprehensive global forest dataset to generate a
quantitative global map of forest tree symbioses, and demonstrate how
nutritional mutualisms are coupled with the global distribution of plant
Any methods, additional references, Nature Research reporting summaries, source
data, statements of data availability and associated accession codes are available at
Received: 27 April 2018; Accepted: 21 March 2019;
Published online 15 May 2019.
1. Batterman, S. A. et al. Key role of symbiotic dinitrogen xation in tropical forest
secondary succession. Nature 502, 224–227 (2013).
2. Shah, F. et al. Ectomycorrhizal fungi decompose soil organic matter using
oxidative mechanisms adapted from saprotrophic ancestors. New Phytol. 209,
3. Averill, C., Turner, B. L. & Finzi, A. C. Mycorrhiza-mediated competition between
plants and decomposers drives soil carbon storage. Nature 505, 543–545 (2014).
4. Clemmensen, K. E. et al. Roots and associated fungi drive long-term carbon
sequestration in boreal forest. Science 339, 1615–1618 (2013).
5. Cheeke, T. E. et al. Dominant mycorrhizal association of trees alters carbon and
nutrient cycling by selecting for microbial groups with distinct enzyme function.
New Phytol. 214, 432–442 (2017).
6. Terrer, C., Vicca, S., Hungate, B. A., Phillips, R. P. & Prentice, I. C. Mycorrhizal
association as a primary control of the CO2 fertilization eect. Science 353,
7. Brundrett, M. C. & Tedersoo, L. Evolutionary history of mycorrhizal symbioses
and global host plant diversity. New Phytol. 220, 1108–1115 (2018).
8. Averill, C. & Hawkes, C. V. Ectomycorrhizal fungi slow soil carbon cycling.
Ecol. Lett. 19, 937–947 (2016).
9. Bennett, J. A. et al. Plant–soil feedbacks and mycorrhizal type inuence
temperate forest population dynamics. Science 355, 181–184 (2017).
10. Phillips, R. P., Brzostek, E. & Midgley, M. G. The mycorrhizal-associated nutrient
economy: a new framework for predicting carbon-nutrient couplings in
temperate forests. New Phytol. 199, 41–51 (2013).
11. Crowther, T. W. et al. Mapping tree density at a global scale. Nature 525,
12. van der Heijden, M. G., Martin, F. M., Selosse, M. A. & Sanders, I. R. Mycorrhizal
ecology and evolution: the past, the present, and the future. New Phytol. 205,
13. Binkley, D., Sollins, P., Bell, R., Sachs, D. & Myrold, D. Biogeochemistry of
adjacent conifer and alder-conifer stands. Ecology 73, 2022–2033
14. Leake, J. et al. Networks of power and inuence: the role of mycorrhizal
mycelium in controlling plant communities and agroecosystem functioning.
Can. J. Bot. 82, 1016–1045 (2004).
15. Hedin, L. O., Brookshire, E. N. J., Menge, D. N. L. & Barron, A. R. The nitrogen
paradox in tropical rainforest ecosystems. Annu. Rev. Ecol. Evol. Syst. 40,
16. Read, D. J. Mycorrhizas in ecosystems. Experientia 47, 376–391 (1991).
17. Houlton, B. Z., Wang, Y.-P., Vitousek, P. M. & Field, C. B. A unifying framework
for dinitrogen xation in the terrestrial biosphere. Nature 454, 327–330
18. Peay, K. G. The mutualistic niche: mycorrhizal symbiosis and community
dynamics. Annu. Rev. Ecol. Evol. Syst. 47, 143–164 (2016).
19. Pellegrini, A. F. A., Staver, A. C., Hedin, L. O., Charles-Dominique, T. & Tourgee, A.
Aridity, not re, favors nitrogen-xing plants across tropical savanna and forest
biomes. Ecology 97, 2177–2183 (2016).
20. Tuomi, M. et al. Leaf litter decomposition—estimates of global variability based
on Yasso07 model. Ecol. Modell. 220, 3362–3371 (2009).
21. Ma, Z. et al. Evolutionary history resolves global organization of root functional
traits. Nature 555, 94–97 (2018).
22. Lu, M. & Hedin, L. O. Global plant–symbiont organization and emergence of
biogeochemical cycles resolved by evolution-based trait modelling. Nat. Ecol.
Evol. 3, 239–250 (2019).
23. Scheer, M., Carpenter, S., Foley, J. A., Folke, C. & Walker, B. Catastrophic shifts in
ecosystems. Nature 413, 591–596 (2001).
24. Reich, P. B. et al. Geographic range predicts photosynthetic and growth
response to warming in co-occurring tree species. Nat. Clim. Change5, 148
25. McGroddy, M. E., Daufresne, T. & Hedin, L. O. Scaling of C:N:P stoichiometry in
forests worldwide: implications of terrestrial redeld-type ratios. Ecology 85,
26. Reich, P. B. & Oleksyn, J. Global patterns of plant leaf N and P in relation to
temperature and latitude. Proc. Natl Acad. Sci. USA 101, 11001–11006
27. Corrales, A., Mangan, S. A., Turner, B. L. & Dalling, J. W. An ectomycorrhizal
nitrogen economy facilitates monodominance in a neotropical forest. Ecol. Lett.
19, 383–392 (2016).
28. Menge, D. N., Lichstein, J. W. & Angeles-Pérez, G. Nitrogen xation strategies
can explain the latitudinal shift in nitrogen-xing tree abundance. Ecology 95,
29. Liao, W., Menge, D. N. L., Lichstein, J. W. & Ángeles-Pérez, G. Global climate
change will increase the abundance of symbiotic nitrogen-xing trees in
much of North America. Glob. Change Biol. 23, 4777–4787 (2017).
30. Gei, M. et al. Legume abundance along successional and rainfall gradients in
neotropical forests. Nat. Ecol. Evol. 2, 1104–1111 (2018).
Acknowledgements This work was made possible by the Global Forest
Biodiversity Database, which represents the work of over 200 independent
investigators and their public and private funding agencies (see Supplementary
Reviewer information Nature thanks Martin Bidartondo, David Bohan and the
other anonymous reviewer(s) for their contribution to the peer review of this
Author contributions K.G.P. and T.W.C. conceived the study; T.W.C., J.L., P.B.R.,
G.N., S.d.-M., M.Z., N.P., B.H., X.Z. and C.Z. conceived and organized the GFBi
database; K.G.P., B.S.S., G.D.A.W. and M.E.V.N. compiled the symbiosis database;
B.S.S. carried out the primary data analysis; M.E.V.N. and D.R. contributed
to data compilation and analysis; B.S.S., T.W.C., M.E.V.N. and K.G.P. wrote the
initial manuscript; B.S.S., T.W.C., J.L., M.E.V.N., G.D.A.W., P.B.R., G.N., S.d.-M.,
M.Z., N.P., B.H., X.Z., C.Z. and K.G.P. made substantial revisions to all versions of
the manuscript; all other named authors provided forest inventory data and
commented on the manuscript.
Competing interests The authors declare no competing interests.
Supplementary information is available for this paper at https://doi.org/
Reprints and permissions information is available at http://www.nature.com/
Correspondence and requests for materials should be addressed to T.W.C., J.L.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.
© The Author(s), under exclusive licence to Springer Nature Limited 2019
408 | NATURE | VOL 569 | 16 MAY 2019
Meinrad Abegg16, C.Yves AdouYao17, Giorgio Alberti18,19, Angelica
AlmeydaZambrano20, Esteban Alvarez-Davila21, Patricia Alvarez-Loayza22,
Luciana F. Alves23, Christian Ammer24, Clara Antón-Fernández25, Alejandro
Araujo-Murakami26, Luzmila Arroyo26, Valerio Avitabile27, Gerardo Aymard28,
Timothy Baker29, Radomir Bałazy30, Olaf Banki31, Jorcely Barroso32, Meredith
Bastian33, Jean-Francois Bastin2, Luca Birigazzi11, Philippe Birnbaum162,
Robert Bitariho34, Pascal Boeckx35, Frans Bongers, Olivier Bouriaud36,
PedroH.S. Brancalion37, Susanne Brandl38, Francis Q. Brearley39, Roel
Brienen29, Eben Broadbent175, Helge Bruelheide40,41, Filippo Bussotti42,
Roberto CazzollaGatti43, Ricardo Cesar37, Goran Cesljar44, Robin Chazdon45,46,
Han Y. H. Chen47,48, Chelsea Chisholm49, Emil Cienciala50,51, Connie J.
Clark52, David Clark53, Gabriel Colletta174, Richard Condit55, David Coomes56,
Fernando CornejoValverde57, Jose J. Corral-Rivas58, Philip Crim59,60,
Jonathan Cumming60, Selvadurai Dayanandan61, André L. de Gasper62,
Mathieu Decuyper8, Géraldine Derroire63, Ben DeVries64, Ilija Djordjevic65,
Amaral Iêda66, Aurélie Dourdain63, Nestor Laurier Engone Obiang67, Brian
Enquist68,69, Teresa Eyre70, Adandé Belarmain Fandohan71, Tom M. Fayle72,
Ted R. Feldpausch73, Leena Finér74, Markus Fischer75, Christine Fletcher76,
Jonas Fridman77, Lorenzo Frizzera78, Javier G. P. Gamarra11, Damiano
Gianelle78, Henry B. Glick79, David Harris80, Andrew Hector81, Andreas
Hemp82, Geerten Hengeveld8, John Herbohn46, Martin Herold8, Annika
Hillers83,178, Eurídice N. Honorio Coronado84, Markus Huber16, Cang Hui85,86,
Hyunkook Cho87, Thomas Ibanez88, Ilbin Jung87, Nobuo Imai89, Andrzej M.
Jagodzinski90,91, Bogdan Jaroszewicz92, Vivian Johannsen93, Carlos A. Joly54,
Tommaso Jucker94, Viktor Karminov95, Kuswata Kartawinata22, Elizabeth
Kearsley96, David Kenfack97, Deborah Kennard98, Sebastian Kepfer-Rojas93,
Gunnar Keppel99, Mohammed Latif Khan100, Timothy Killeen26, Hyun Seok
Kim101,102,103,104, Kanehiro Kitayama105, Michael Köhl106, Henn Korjus107,
Florian Kraxner108, Diana Laarmann107, Mait Lang107, Simon Lewis29,109,
Huicui Lu110, Natalia Lukina111, Brian Maitner68, Yadvinder Malhi112, Eric
Marcon113, BeatrizSchwantes Marimon114, Ben Hur Marimon-Junior114,
Andrew Robert Marshall46,115,176, Emanuel Martin116, Olga Martynenko95,
Jorge A. Meave117, Omar Melo-Cruz118, Casimiro Mendoza119, Cory
Merow45, Abel MonteagudoMendoza120,121, Vanessa Moreno37, Sharif A.
Mukul46,122, Philip Mundhenk106, Maria G. Nava-Miranda123, David Neill124,
Victor Neldner70, Radovan Nevenic65, Michael Ngugi70, Pascal Niklaus125,
Jacek Oleksyn90, Petr Ontikov95, Edgar Ortiz-Malavasi126, Yude Pan127, Alain
Paquette128, Alexander Parada-Gutierrez26, Elena Parfenova129, Minjee
Park101, Marc Parren130, Narayanaswamy Parthasarathy131, Pablo L. Peri132,
Sebastian Pfautsch133, Oliver Phillips29, Maria Teresa Piedade134, Daniel
Piotto135, Nigel C. A. Pitman22, Irina Polo136, Lourens Poorter8, Axel Dalberg
Poulsen80, John R. Poulsen52, Hans Pretzsch137, Freddy RamirezArevalo138,
Zorayda Restrepo-Correa139, Mirco Rodeghiero78,177, Samir Rolim135,
Anand Roopsind140, Francesco Rovero141,142, Ervan Rutishauser55, Purabi
Saikia143, Philippe Saner125, Peter Schall24, Mart-Jan Schelhaas8, Dmitry
Schepaschenko108, Michael Scherer-Lorenzen144, Bernhard Schmid125, Jochen
Schöngart134, Eric Searle47, Vladimír Seben145, Josep M. Serra-Diaz146,147,
Christian Salas-Eljatib179,180, Douglas Sheil181, Anatoly Shvidenko108, Javier
Silva-Espejo148, Marcos Silveira149, James Singh150, Plinio Sist12, Ferry Slik151,
Bonaventure Sonké152, Alexandre F. Souza153, Krzysztof Stereńczak30, Jens-
Christian Svenning147,154, Miroslav Svoboda155, Natalia Targhetta134, Nadja
Tchebakova129, Hans ter Steege31,156, Raquel Thomas157, Elena Tikhonova111,
Peter Umunay79, Vladimir Usoltsev158, Fernando Valladares159, Fons van der
Plas160, Tran Van Do161, Rodolfo VasquezMartinez120, Hans Verbeeck96, Helder
Viana163,164, Simone Vieira165, Klaus von Gadow166, Hua-Feng Wang167, James
Watson168, Bertil Westerlund77, Susan Wiser169, Florian Wittmann170, Verginia
Wortel171, Roderick Zagt172, Tomasz Zawila-Niedzwiecki173, Zhi-Xin Zhu167 &
Irie Casimir Zo-Bi13
16Swiss Federal Institute for Forest, Snow and Landscape Research, WSL, Birmensdorf,
Switzerland. 17UFR Biosciences, University Félix Houphouët-Boigny, Abidjan, Côte d’Ivoire.
18Department of Agricultural, Food, Environmental and Animal Sciences, University of Udine,
Udine, Italy. 19Institute of Biometeorology, National Research Council (CNR-IBIMET), Florence,
Italy. 20Spatial Ecology and Conservation Laboratory, Department of Tourism, Recreation and
Sport Management, University of Florida, Gainesville, FL, USA. 21Fundacion ConVida,
Universidad Nacional Abierta y a Distancia, UNAD, Medellin, Colombia. 22Field Museum of
Natural History, Chicago, IL, USA. 23Center for Tropical Research, Institute of the Environment
and Sustainability, UCLA, Los Angeles, CA, USA. 24Silviculture and Forest Ecology of the
Temperate Zones, University of Göttingen, Göttingen, Germany. 25Division of Forest and Forest
Resources, Norwegian Institute of Bioeconomy Research (NIBIO), Ås, Norway. 26Museo de
Historia Natural Noel Kempff Mercado, Universidad Autonoma Gabriel Rene Moreno, Santa
Cruz de la Sierra, Bolivia. 27European Commission, Joint Research Centre, Ispra, Italy.
28UNELLEZ-Guanare, Programa de Ciencias del Agro y el Mar, Herbario Universitario (PORT),
Portuguesa, Venezuela. 29School of Geography, University of Leeds, Leeds, UK. 30Department of
Geomatics, Forest Research Institute, Raszyn, Poland. 31Naturalis Biodiversity Centre, Leiden,
The Netherlands. 32Centro Multidisciplinar, Universidade Federal do Acre, Rio Branco, Brazil.
33Smithsonian’s National Zoo and Conservation Biology Institute, Washington, DC, USA.
34Institute of Tropical Forest Conservation, Mbarara University of Sciences and Technology,
Mbarara, Uganda. 35Isotope Bioscience Laboratory - ISOFYS, Ghent University, Ghent, Belgium.
36Integrated Center for Research, Development and Innovation in Advanced Materials,
Nanotechnologies, and Distributed Systems for Fabrication and Control (MANSiD), Stefan cel
Mare University of Suceava, Suceava, Romania. 37Department of Forest Sciences, Luiz de
Queiroz College of Agriculture, University of São Paulo, Piracicaba, Brazil. 38Bavarian State
Institute of Forestry, Freising, Germany. 39Manchester Metropolitan University, Manchester, UK.
40Institute of Biology, Geobotany and Botanical Garden, Martin Luther University Halle-
Wittenberg, Halle-Wittenberg, Germany. 41German Centre for Integrative Biodiversity Research
(iDiv) Halle-Jena-Leipzig, Leipzig, Germany. 42Department of Agriculture, Food, Environment
and Forest (DAGRI), University of Firenze, Florence, Italy. 43Biological Institute, Tomsk State
University, Tomsk, Russia. 44Department of Spatial Regulation, GIS and Forest Policy, Institute of
Forestry, Belgrade, Serbia. 45Department of Ecology and Evolutionary Biology, University of
Connecticut, Storrs, CT, USA. 46Tropical Forests and People Research Centre, University of the
Sunshine Coast, Maroochydore, Queensland, Australia. 47Faculty of Natural Resources
Management, Lakehead University, Thunder Bay, Ontario, Canada. 48Key Laboratory for Humid
Subtropical Eco-Geographical Processes of the Ministry of Education, Fujian Normal University,
Fuzhou, China. 49Institute of Integrative Biology, ETH Zürich, Zurich, Switzerland. 50IFER -
Institute of Forest Ecosystem Research, Jilove u Prahy, Czech Republic. 51Global Change
Research Institute CAS, Brno, Czech Republic. 52Nicholas School of the Environment, Duke
University, Durham, NC, USA. 53Department of Biology, University of Missouri-St Louis, St Louis,
MO, USA. 54Department of Plant Biology, Institute of Biology, University of Campinas, UNICAMP,
Campinas, Brazil. 55Smithsonian Tropical Research Institute, Balboa, Panama. 56Department of
Plant Sciences, University of Cambridge, Cambridge, UK. 57Andes to Amazon Biodiversity
Program, Madre de Dios, Peru. 58Facultad de Ciencias Forestales, Universidad Juárez del Estado
de Durango, Durango, Mexico. 59Department of Physical and Biological Sciences, The College of
Saint Rose, Albany, NY, USA. 60Department of Biology, West Virginia University, Morgantown,
WV, USA. 61Biology Department, Concordia University, Montreal, Quebec, Canada. 62Natural
Science Department, Universidade Regional de Blumenau, Blumenau, Brazil. 63Cirad, UMR
EcoFoG, Kourou, French Guiana. 64Department of Geographical Sciences, University of
Maryland, College Park, MD, USA. 65Institute of Forestry, Belgrade, Serbia. 66National Institute of
Amazonian Research, Manaus, Brazil. 67IRET, Herbier National du Gabon (CENAREST),
Libreville, Gabon. 68Department of Ecology and Evolutionary Biology, University of Arizona,
Tucson, AZ, USA. 69The Santa Fe Institute, Santa Fe, NM, USA. 70Department of Environment
and Science, Queensland Herbarium, Toowong, Queensland, Australia. 71Ecole de Foresterie et
Ingénierie du Bois, Université Nationale d’Agriculture, Ketou, Benin. 72Biology Centre of the
Czech Academy of Sciences, Institute of Entomology, Ceske Budejovice, Czech Republic.
73Geography, College of Life and Environmental Sciences, University of Exeter, Exeter, UK.
74Natural Resources Institute Finland (Luke), Joensuu, Finland. 75Institute of Plant Sciences,
University of Bern, Bern, Switzerland. 76Forest Research Institute Malaysia, Kuala Lumpur,
Malaysia. 77Department of Forest Resource Management, Swedish University of Agricultural
Sciences SLU, Umea, Sweden. 78Department of Sustainable Agro-Ecosystems and
Bioresources, Research and Innovation Center, Fondazione Edmund Mach, San Michele
all’Adige, Italy. 79School of Forestry and Environmental Studies, Yale University, New Haven, CT,
USA. 80Royal Botanic Garden Edinburgh, Edinburgh, UK. 81Department of Plant Sciences,
University of Oxford, Oxford, UK. 82Department of Plant Systematics, University of Bayreuth,
Bayreuth, Germany. 83Centre for Conservation Science, The Royal Society for the Protection of
Birds, Sandy, UK. 84Instituto de Investigaciones de la Amazonía Peruana, Iquitos, Peru. 85Centre
for Invasion Biology, Department of Mathematical Sciences, Stellenbosch University,
Stellenbosch, South Africa. 86Theoretical Ecology Unit, African Institute for Mathematical
Sciences, Cape Town, South Africa. 87Division of Forest Resources Information, Korea Forest
Promotion Institute, Seoul, South Korea. 88Institut Agronomique néo-Calédonien (IAC), Equipe
Sol & Végétation (SolVeg), Nouméa, New Caledonia. 89Department of Forest Science, Tokyo
University of Agriculture, Tokyo, Japan. 90Institute of Dendrology, Polish Academy of Sciences,
Kórnik, Poland. 91Pozna
University of Life Sciences, Department of Game Management and
Forest Protection, Pozna
, Poland. 92Faculty of Biology, Białowieża Geobotanical Station,
University of Warsaw, Białowieża, Poland. 93Department of Geosciences and Natural Resource
Management, University of Copenhagen, Copenhagen, Denmark. 94Centre for Environment and
Life Sciences, CSIRO Land and Water, Floreat, Western Australia, Australia. 95Forestry Faculty,
Bauman Moscow State Technical University, Mytischi, Russia. 96CAVElab – Computational and
Applied Vegetation Ecology, Department of Environment, Ghent University, Ghent, Belgium.
97CTFS-ForestGEO, Smithsonian Tropical Research Institute, Balboa, Panama. 98Department of
Physical and Environmental Sciences, Colorado Mesa University, Grand Junction, CO, USA.
99School of Natural and Built Environments and Future Industries Institute, University of South
Australia, Adelaide, South Australia, Australia. 100Department of Botany, Dr Harisingh Gour
Central University, Sagar, India. 101Department of Forest Sciences, Seoul National University,
Seoul, South Korea. 102Interdisciplinary Program in Agricultural and Forest Meteorology, Seoul
National University, Seoul, South Korea. 103National Center for Agro Meteorology, Seoul, South
Korea. 104Research Institute for Agriculture and Life Sciences, Seoul National University, Seoul,
South Korea. 105Graduate School of Agriculture, Kyoto University, Kyoto, Japan. 106Institute for
World Forestry, University of Hamburg, Hamburg, Germany. 107Institute of Forestry and Rural
Engineering, Estonian University of Life Sciences, Tartu, Estonia. 108Ecosystems Services and
Management, International Institute for Applied Systems Analysis, Laxenburg, Austria.
109Department of Geography, University College London, London, UK. 110Faculty of Forestry,
Qingdao Agricultural University, Qingdao, China. 111Center for Forest Ecology and Productivity,
Russian Academy of Sciences, Moscow, Russia. 112School of Geography, University of Oxford,
Oxford, UK. 113UMR EcoFoG, AgroParisTech, Kourou, France. 114Departamento de Ciências
Biológicas, Universidade do Estado de Mato Grosso, Nova Xavantina, Brazil. 115Department of
Environment & Geography, University of York, York, UK. 116Department of Wildlife Management,
College of African Wildlife Management, Mweka, Tanzania. 117Departamento de Ecología y
Recursos Naturales, Facultad de Ciencias, Universidad Nacional Autónoma de México, Mexico
City, Mexico. 118Universidad del Tolima, Ibagué, Colombia. 119Colegio de Profesionales
Forestales de Cochabamba, Cochabamba, Bolivia. 120Jardín Botánico de Missouri, Oxapampa,
Peru. 121Universidad Nacional de San Antonio Abad del Cusco, Cusco, Peru. 122Department of
Environmental Management, School of Environmental Science and Management, Independent
University Bangladesh, Dhaka, Bangladesh. 123Instituto de Silvicultura e Industria de la Madera,
Universidad Juárez del Estado de Durango, Durango, Mexico. 124Universidad Estatal
Amazónica, Puyo, Pastaza, Ecuador. 125Department of Evolutionary Biology and Environmental
Studies, University of Zürich, Zürich, Switzerland. 126Forestry School, Tecnológico de Costa Rica
TEC, Cartago, Costa Rica. 127Climate, Fire, and Carbon Cycle Sciences, USDA Forest Service,
Durham, NC, USA. 128Centre for Forest Research, Université du Québec à Montréal, Montréal,
Quebec, Canada. 129V. N. Sukachev Institute of Forest, FRC KSC, Siberian Branch of the Russian
Academy of Sciences, Krasnoyarsk, Russia. 130Department of Forestry, World Research Institute,
Washington, DC, USA. 131Department of Ecology and Environmental Sciences, Pondicherry
University, Puducherry, India. 132Instituto Nacional de Tecnología Agropecuaria (INTA),
Universidad Nacional de la Patagonia Austral (UNPA), Consejo Nacional de Investigaciones
Científicas y Tecnicas (CONICET), Rio Gallegos, Argentina. 133School of Social Sciences and
Psychology (Urban Studies), Western Sydney University, Penrith, New South Wales, Australia.
134Instituto Nacional de Pesquisas da Amazônia, Manaus, Brazil. 135Laboratório de Dendrologia
e Silvicultura Tropical, Centro de Formação em Ciências Agroflorestais, Universidade Federal do
Sul da Bahia, Itabuna, Brazil. 136Jardín Botánico de Medellín, Medellín, Colombia. 137Chair for
Forest Growth and Yield Science, TUM School for Life Sciences, Technical University of Munich,
Munich, Germany. 138Universidad Nacional de la Amazonía Peruana, Iquitos, Peru. 139Servicios
Ecosistémicos y Cambio Climático (SECC), Fundación Con Vida & Corporación COL-TREE,
Medellín, Colombia. 140Department of Biological Sciences, Boise State University, Boise, ID,
USA. 141Tropical Biodiversity Section, MUSE - Museo delle Scienze, Trento, Italy. 142Department
of Biology, University of Florence, Florence, Italy. 143Department of Environmental Sciences,
Central University of Jharkhand, Ranchi, India. 144Faculty of Biology, Geobotany, University of
Freiburg, Freiburg im Breisgau, Germany. 145National Forest Centre, Forest Research Institute
Zvolen, Zvolen, Slovakia. 146Université de Lorraine, AgroParisTech, Inra, Silva, Nancy, France.
147Center for Biodiversity Dynamics in a Changing World (BIOCHANGE), Department of
Bioscience, Aarhus University, Aarhus, Denmark. 148Departamento de Biología, Universidad de
la Serena, La Serena, Chile. 149Centro de Ciências Biológicas e da Natureza, Universidade
Federal do Acre, Rio Branco, Acre, Brazil. 150Guyana Forestry Commission, Georgetown, French
Guiana. 151Faculty of Science, Universiti Brunei Darussalam, Bandar Seri Begawan, Brunei
Darussalam. 152Plant Systematic and Ecology Laboratory, Department of Biology, Higher
Teachers’ Training College, University of Yaoundé, Yaoundé, Cameroon. 153Departamento de
Ecologia, Universidade Federal do Rio Grande do Norte, Natal, Brazil. 154Section for
Ecoinformatics & Biodiversity, Department of Bioscience, Aarhus University, Aarhus, Denmark.
155Faculty of Forestry and Wood Sciences, Czech University of Life Sciences, Prague, Czech
Republic. 156Systems Ecology, Free University Amsterdam, Amsterdam, The Netherlands.
157Iwokrama International Centre for Rainforest Conservation and Development (IIC),
Georgetown, French Guiana. 158Botanical Garden of Ural Branch of Russian Academy of
Sciences, Ural State Forest Engineering University, Ekaterinburg, Russia. 159LINCGlobal, Museo
Nacional de Ciencias Naturales, CSIC, Madrid, Spain. 160Systematic Botany and Functional
Biodiversity, Institute of Biology, Leipzig University, Leipzig, Germany. 161Silviculture Research
Institute, Vietnamese Academy of Forest Sciences, Hanoi, Vietnam. 162Cirad, UMR-AMAP,
CNRS, INRA, IRD, Université de Montpellier, Montpellier, France. 163Centre for the Research
and Technology of Agro-Environmental and Biological Sciences, CITAB, University of Trás-os-
Montes and Alto Douro, UTAD, Vila Real, Portugal. 164Agricultural High School, Polytechnic
Institute of Viseu, Viseu, Portugal. 165Environmental Studies and Research Center, University
of Campinas, UNICAMP, Campinas, Brazil. 166Department of Forest and Wood Science,
University of Stellenbosch, Stellenbosch, South Africa. 167Key Laboratory of Tropical
Biological Resources, Ministry of Education, School of Life and Pharmaceutical Sciences,
Hainan University, Haikou, China. 168Division of Forestry and Natural Resources, West Virginia
University, Morgantown, WV, USA. 169Manaaki Whenua–Landcare Research, Lincoln, New
Zealand. 170Department of Wetland Ecology, Institute for Geography and Geoecology,
Karlsruhe Institute for Technology, Karlsruhe, Germany. 171Centre for Agricultural Research in
Suriname (CELOS), Paramaribo, Suriname. 172Tropenbios International, Wageningen, The
Netherlands. 173Polish State Forests, Coordination Center for Environmental Projects, Warsaw,
Poland. 174Programa de Pós-graduação em Biologia Vegetal, Instituto de Biologia,
Universidade Estadual de Campinas, Campinas, Brazil. 175Spatial Ecology and Conservation
Laboratory, School of Forest Resources and Conservation, University of Florida, Gainesville,
FL, USA. 176Flamingo Land Ltd, Kirby Misperton, UK. 177Centro Agricoltura, Alimenti,
Ambiente, University of Trento, San Michele all’Adige, Italy. 178Wild Chimpanzee Foundation,
Liberia Office, Monrovia, Liberia. 179Centro de Modelación y Monitoreo de Ecosistemas,
Universidad Mayor, Santiago, Chile. 180Laboratorio de Biometria, Universidad de La Frontera,
Temuco, Chile. 181Faculty of Environmental Sciences and Natural Resource Management,
Norwegian University of Life Sciences, Ås, Norway.
We quantified the relative abundance of tree symbiotic guilds across >1.1 million
forest census plots combined in the GFBi database, an extension of the plot-
based GFB database31. The GFBi database consists of individual-based data that
we compiled from all the regional and national GFBi forest-inventory datasets.
The standardized GFBi data frame (that is, tree list) comprises tree identifier (ID)
(a unique number assigned to each individual tree); plot ID (a unique string
assigned to each plot); plot coordinates, in decimal degrees of the WGS84 datum;
tree size, in diameter-at-breast-height; trees-per-hectare expansion factor; year
of measurement; dataset name (a unique name assigned to each forest inventory
dataset); and binomial species names of trees.
We checked all species names from different forest inventory datasets for
errors in three steps. First, we extracted scientific names from original datasets,
and kept only the names of genus and species (authority names are removed).
Next, we compiled all the species names into five general species lists (one
for each continent). Finally, we verified individual species names against 23
online taxonomic databases using the ‘taxize’ package of the R programming
. We assigned each morphospecies a unique name that comprised
the genus, the string ‘spp’, followed by the dataset name and a unique number
for that species. For example, ‘Picea sppCNi1’ and ‘Picea sppCNi2’ represent
two different species under the genus Picea, observed in the first Chinese
We derived plot-level abundance information in terms of species-abundance
matrices. Each species-abundance matrix consisted of the number of individu-
als by species (column vectors) within individual sample plots (row vectors). In
addition, key plot-level information was also added to the matrices, including
plot ID, dataset name, plot coordinates, the year of measurement and basal area
(that is, the total cross-sectional areas (in m
) of living trees per one hectare of
Tree genera were assigned to a plant family using a plant taxonomy lookup table
generated by W. Cornwell (hosted on Github, https://github.com/traitecoevo/tax-
onlookup), which uses the accepted taxonomy from ‘The Plant List’ (http://www.
theplantlist.org/). The majority (96.5%) of genera of the species in the GFBi were
successfully matched to family; for those that could not be assigned, we manually
checked the genus and species in the GFBi against synonyms from The Plant List.
Of the 1,038 mismatches that remained after automated assignment to families,
an additional 440 genera were assigned to family either by updating older genera
and species names with their more-recent synonyms or by correcting obvious
misspellings. The remaining 598 entries that could not be matched to family were
excluded from further analysis.
We used a taxonomically informed approach to assign symbiotic states to plant
species from the GFBi. Plant species were assigned to one of five symbiotic guilds;
ectomycorrhizal, arbuscular mycorrhizal, ericoid mycorrhizal, weakly arbuscular
mycorrhizal or non-mycorrhizal (AMNM) or N-fixer (Supplementary Table1).
Although we did not model the relative abundance of ericoid mycorrhizal trees
(owing to their rarity), we have included a map of their relative abundance
from our grid (Supplementary Fig.1). We also include the full species list as
Supplementary Data; this list includes the columns used to assign species to guilds.
We also include a list of families and genera assigned to all guilds except the
arbuscular mycorrhizal guild (Supplementary Tables2–5), with notes for cases
of species of individual genera that were assigned to two guilds simultaneously
(for example, Alnus is an N-fixer and ectomycorrhizal) or for cases in which
species from individual genera were split between two different guilds (for exam-
ple, some Pisonia sp. are AMNM and some are ectomycorrhizal). An arbuscular
mycorrhizal summary table is excluded from the Supplementary Tables for length
considerations; this information is available as Supplementary Data (file name
The taxonomy of species in our inventory was compared with recently published
literature on the evolutionary history of mycorrhizal symbiosis
. For most species, symbiotic status could be reliably assigned at the
genus (for example, Dicymbe) or family level (for example, Pinaceae). For the few
groups for which status was unreliable or variable within a genus (for example,
Pisonia), we conducted additional literature searches.
We assigned species to the ectomycorrhizal category in three stages: first, at
the family level (for example, Pinaceae); then, at the genus level (for example,
Dicymbe); and, finally, by using literature searches for genera for which the sta-
tus was unclear (for example, in the genus Pisonia some species are arbuscular
mycorrhizal and others are ectomycorrhizal). We used a published list
species into the appropriate guild. For the genus Acacia, we followed previous
work7 by assuming that only endemic Australian species associate with ectomy-
corrhizal fungi (we sorted Acacia species according to provenance using http://
The AMNM category grouped all genera of terrestrial, non-epiphytic plants
that either lack arbuscular mycorrhizal fungi or have low or inconsistent records
of arbuscular mycorrhizal fungi colonization of roots. For example, although
there are some published records of arbuscular mycorrhizal fungi colonization
in the roots of plants of the Proteaceae family, these records are inconsistent and
colonization is generally low. Further, as Proteaceae are associated with a non-
mycorrhizal root morphology (the cluster or proteoid root system) that allows
them to access otherwise unavailable forms of soil nutrients
, we placed the entire
family within AMNM. The family Urticaceae (which we also characterized as
AMNM) was problematic—early successional species from tropical forests, such
as those in the genus Cecropia, have records of both low and absent arbuscular
mycorrhizal fungi colonization40. Our approach was to use the most broadly inclusive
categorization for AMNM plants.
N-fixer status was assigned at the genus level, using previously compiled
databases of global symbiotic N2 fixation34–37. Given that symbiotic N2 fixation
with rhizobial or Frankia bacteria has evolved only in four orders (Rosales,
Cucurbitales, Fabales and Fagales)41, all species outside of this nitrogen-fixing
clade were assigned non-fixing status. Some species could not be assigned
an N-fixer status because they were typed to a higher taxonomic level (for
example, family) that is ambiguous from the perspective of N-fixer status.
We recorded when our assignment of N-fixer status was based on phyloge-
netic criteria, but where symbiotic nitrogen fixation is evolutionarily labile.
Because these cases are more likely to be mis-assigned, we excluded them from
the nitrogen-fixation category. The N-fixer group contains species that are
colonized by arbuscular mycorrhizal fungi (for example, most genera from
Leguminosae) and others that are colonized by ectomycorrhizal fungi (for
example, Alnus sp.).
Most plant species form arbuscular mycorrhizal symbioses, the basal symbiotic
state relative to the later-derived ectomycorrhizal and nitrogen-fixing symbioses.
Furthermore, many ectomycorrhizal and nitrogen-fixing plants maintain the
ability to form arbuscular mycorrhizal symbioses. Thus, a tree species is most
likely to be arbuscular mycorrhizal if it does not form associations with another
symbiotic guild (or forgoes root symbiosis entirely), as evidenced by its inclusion
in exhaustive databases of plant symbiotic state
. In keeping with other
large-scale studies in the field33, we assigned tree species from the GFBi data-
base an arbuscular-mycorrhizal-exclusive state if they belonged to taxa that were
not matched to ectomycorrhizal, ericoid mycorrhizal, AMNM or N-fixer sym-
bioses. Thus, the arbuscular mycorrhizal and N-fixer groups in our dataset are
non-overlapping, despite the fact that most N-fixers also associate with arbuscular
The proportions of tree basal area and tree individuals were aggregated to
a 1°-by-1° grid by taking the weighted average of the plot-level proportions
(Supplementary Table6). This resulted in a total of 2,768 grid cells, each with a
score for the proportional abundance of ectomycorrhizal, arbuscular mycorrhizal,
N-fixer, ericoid mycorrhizal and AMNM trees. We calculated two measures of
relative abundance for each symbiotic guild: the proportion of tree stems and the
proportion of tree basal area. Because the measurements are highly correlated with
one another (Supplementary Fig.2), we chose to model only the proportion of
total tree basal area, which should scale more closely to proportion of tree biomass
as it accounts for differences in size among individual stems. Additionally, we
quantified variability among plots within each grid cell by calculating the weighted
standard deviation across the grid (Supplementary Information, Supplementary
To identify the key factors that structure symbiotic distributions, we assem-
bled 70 global predictor layers: 19 climaticindices (relating to annual, monthly
and quarterly temperature and precipitation variables), 14 soil chemicalindices
(relating to total soil nitrogen density, microbial nitrogen, C:N ratios and soil
phosphorus fractions, pH and cation exchange capacity), 5 soil physicalindi
ces (relating to soil texture and bulk density), 26 vegetative indices (relating
to leaf area index, total stem density, enhanced vegetation index means and
variances) and 5 topographic variables (relating to elevation and hillshade)
(Supplementary Table7). Because decomposition is the dominant process by
which soil nutrients become available to plants, we generated five additional
layers that estimate climatic control of decomposition. We parameterized decom-
position coefficients according to the Yasso07 model
, using the following
)×(1−exp[−1.21×P]), in which
P and T are precipitation and mean temperature (either quarterly or annually)
of a grid cell, and the constants 0.095, 0.00014 and −1.21 are parameters that
were fit using a previous global study of leaf litter massloss
. Although local
decomposition rates can vary considerably based on litter quality or microbial
community composition43, climate is the primary control at the global scale20.
Decomposition coefficients describe how fast different chemical pools of leaf
litter lose mass over time, relative to a parameter (α) that accounts for leaf chem-
istry. Decomposition coefficients (k) with values of 0.5 and 2 indicate a halving
and doubling of decomposition rates, respectively, relative to α (Supplementary
Information, Supplementary Fig.5).
We implemented the random-forest algorithm using the ‘randomForest’ pack-
age in R. Random-forest models average over multiple regression trees, each of
which uses a random subset of all the model variables to predict a response. We
first determined the influence and relationship of all 75 predictor layers on forest
symbiotic state, and then optimized our models using a stepwise reduction in
variables from least to most important. Variable importance was measured in two
ways: increase in node purity and percentage increase in MSE (with values reported
in Fig.2). The increase in node purity of variable x considers the decrease in the
residual sum of squares that results from splitting regression trees using variable x.
The percentage increase in MSE quantifies the increase in model error as a result of
randomly shuffling the order of values in the vector x. We chose to rank variables
according to the increase in node purity because we found that higher increases
in node purities were associated with larger effect sizes, whereas larger percent-
age increases in MSE were associated with more-linear responses with smaller
effect sizes. Whereas our inspection of partial feature contributions is derived
from univariate random-forest models, we additionally ran multivariate random
forests that predict the proportional abundance of ectomycorrhizal, arbuscular
mycorrhizal and N-fixer trees for each pixel. The multivariate models were run
using 50 regression trees each, with the unique set of the best 4 predictor variables
for each symbiotic guild in the univariate models (Fig.2, Supplementary Table7).
Despite strong negative correlations between the proportions of ectomycorrhizal
and arbuscular mycorrhizal basal area (Supplementary Fig.22), the results from
multivariate and univariate random forests are strongly correlated with one another
Using model selection based on eliminating variables with a low increase in
node purity, we removed most soil nutrient, vegetative and topographic varia-
bles from our models (Supplementary Figs.6, 7). Our final models include the
remaining 34 predictor layers with climate, decomposition and some soil physical
and chemical information (Supplementary Fig.8). To determine the parsimony
of our models, we compared the coefficient of determination in models run with
a stepwise reduction in the number of variables (starting with those with the
lowest increase in node purity). Based on performance of the ratio of coeffi-
cient of determination in models with 4 versus 34 variables, we determined that
the 4 most-important variables accounted for >85% of the explained variability
(Supplementary Fig.9). We also compared model performance visually with
plots of actual versus predicted proportions of each tree symbiotic guild among
continents and geographical subregions (Supplementary Fig.10). We used the
‘forestFloor’ package in R to plot the partial variable response of tree symbiotic
guilds to each predictor variable (Fig.2a–c, see Supplementary Figs.19–21 for
partial plots of the partial feature contributions of all 34 variables).
To test the sensitivity of model performance and predictions, we performed
cross-validation in R using the ‘rfUtilities’ package
. K-fold cross-validation tests
the sensitivity of model predictions to losing random subsets from the train-
ing data. For ectomycorrhizal, arbuscular mycorrhizal and N-fixer models, we
ran 99 iterations that withheld 10% of the model training data. We assessed the
decrease in model performance in the 99 iterations by manually calculating the
coefficient of determination, which uses the following formula: 1−Σ(actual per-
centage basal area–predicted percentage basal area)
/Σ(actual percentage basal
area−mean actual percentage basal area)2. For all symbiotic guilds, withholding
10% of the training data resulted in a mean loss in variance explained of less than
1% (Supplementary Fig.11). This shows that our training data have sufficient
redundancy to ensure that our model conclusions are robust. Similarly, to deter-
mine whether our random-forest models would make similar predictions if data
were equally distributed among continents, we rarefied our aggregated grid of
symbiotic states and predictor layers to an even depth. Specifically, we sub-sampled
all continents—North America (including Central America and the Caribbean),
South America, Europe, Asia and Oceania—to match the number of grid pixels
from Africa (n=50). This is a much more aggressive reduction of training data
than is typically used in K-fold cross-validations, as it involves dropping ~90% of
training data rather than retaining the same amount. We performed 99 iterations
of rarefaction each for the three symbiotic guilds. On average, models run with
the rarefied data explained about 10% less variance over the full training data (the
entire predictor/response grid) than did models run with all of the training data
(Supplementary Figs.12, 13).
To avoid projecting our random-forest models outside the ranges of their
training data (for example, grid cells with higher mean annual temperatures
than the maximum used to fit the models), we subset a global grid of predictor
layers depending on whether (1) the grid cell fell within the top 60% of land
surface with respect to tree stem density
and either (2) fell within the univar-
iate distribution of all the predictor layers from our training data and/or (3) fell
within an 8-dimensional hypervolume defined by the unique set of the 4 best
predictors of the relative abundance of each guild (Supplementary Fig.14). We
then projected our models across only those grid cells that met these criteria,
which constitutes 46% of the global land surface and 88% of global tree stems
(Fig.1, Supplementary Fig.15). Model projections were made at two resolutions:
1°-by-1° and 0.5°-by-0.5° (Fig.4). Although model validation indicates that our
projections are robust, additional studies to ground-truth these predictions and
identify any discrepancies would be valuable. If such discrepancies exist, they can
help to fine-tune climate–symbiosis models, or identify areas in which climate
might favour invasion by symbioses that have not yet evolved in or dispersed to
a particular biogeographical region.
We used the following equation to estimate the percentage of global tree stems
that belong to each tree symbiotic guild: Σi((predicted proportion of trees of
guild g in pixel i)×(total number of tree stems in pixel i))/Σi(total number
of tree stems in pixel i). The proportion of tree stems and the proportion of
tree basal area in each guild are highly correlated throughout the training data
(Supplementary Fig.4). The figures cited in the main text for each guild were
calculated using model projections across all pixels, even those that did not meet
the criteria for model projection because they fell outside the multivariate dis
tribution of the predictor layers or had insufficient stem density. However, our
estimates for the global percentage of trees occupied by each tree symbiotic guild
change by <1% when using only those pixels that met our criteria for model
In the main text, we state that sharp transitions between dominant sym-
biotic states with climate variables could lead to declines in ectomycorrhizal
trees, particularly in thesouthernrange limit of the northern boreal forests.
To determine this, we projected our random-forest models for each symbi-
otic guild using climate-change projections over our 19 bioclimatic variables
(Supplementary Table7), including the decomposition coefficients that use tem-
perature and precipitation values. Specifically, we considered the 2070 scenario
with a relative concentration pathway of 8.5 W per m2, which predicts an
increase of greenhouse gas emissions throughout the twenty-first century
We plot the difference in the proportion of forest basal area between the pro-
jections for 2070 and projections that use current climate data (Supplementary
Table7, Supplementary Fig.24). We qualify this prediction with the note that
vegetative changes to forests are constrained by rates of mortality, recruitment
After training and cross-validating our models with GFBi data exclusively,
we additionally tested whether our models accurately predicted the previously
symbiotic state of Eurasian forests. We assigned symbiotic status to
all of the trees in this previous publication, and aggregated plot-level data to a
1°-by-1° grid using the same methods as with the GFBi dataset (Supplementary
Fig.25). We found that—on average—our models predicted the symbiotic state
in the regional dataset within 13.6% of the value of this previously published data-
set (Supplementary Fig.26). For projected maps in Fig.4a–c, we included the
data with the GFBi training data to increase geographical
coverage throughout Eurasia.
Reporting summary. Further information on research design is available in the
Nature Research Reporting Summary linked to this paper.
Information regarding symbiotic guild assignments, model selection (including
global rasters of our model projections for ectomycorrhizal, arbuscular myc-
orrhizal and N-fixer proportion of tree basal area) and analyses is available as
Supplementary Data. The GFBi database is available upon written request at
https://www.gfbinitiative.org/datarequest. Any other relevant data are available
from the corresponding authors upon reasonable request.
31. Liang, J. et al. Positive biodiversity–productivity relationship predominant in
global forests. Science 354, aaf8957 (2016).
32. Chamberlain, S. A. & Szöcs, E. taxize: taxonomic search and retrieval in R.
F1000Res. 2, 191 (2013).
33. Brundrett, M. C. Mycorrhizal associations and other means of nutrition of
vascular plants: understanding the global diversity of host plants by resolving
conicting information and developing reliable means of diagnosis. Plant Soil
320, 37–77 (2009).
34. Werner, G. D., Cornwell, W. K., Sprent, J. I., Kattge, J. & Kiers, E. T. A single
evolutionary innovation drives the deep evolution of symbiotic N2-xation in
angiosperms. Nat. Commun. 5, 4087 (2014).
35. Werner, G. D., Cornwell, W. K., Cornelissen, J. H. & Kiers, E. T. Evolutionary signals
of symbiotic persistence in the legume–rhizobia mutualism. Proc. Natl Acad.
Sci. USA 112, 10262–10269 (2015).
36. Afkhami, M. E. et al. Symbioses with nitrogen-xing bacteria: nodulation and
phylogenetic data across legume genera. Ecology 99, 502 (2018).
37. Tedersoo, L. et al. Global database of plants with root-symbiotic nitrogen
xation: Nod DB. J. Veg. Sci. 29, 560–568 (2018).
38. Hayward, J. & Hynson, N. A. New evidence of ectomycorrhizal fungi in the
Hawaiian Islands associated with the endemic host Pisonia sandwicensis
(Nyctaginaceae). Fungal Ecol. 12, 62–69 (2014).
39. Lambers, H., Martinoia, E. & Renton, M. Plant adaptations to severely
phosphorus-impoverished soils. Curr. Opin. Plant Biol. 25, 23–31 (2015).
40. Wang, B. & Qiu, Y.-L. Phylogenetic distribution and evolution of mycorrhizas in
land plants. Mycorrhiza 16, 299–363 (2006).
41. Soltis, D. E. et al. Chloroplast gene sequence data suggest a single origin of the
predisposition for symbiotic nitrogen xation in angiosperms. Proc. Natl Acad.
Sci. USA 92, 2647–2651 (1995).
42. Palosuo, T., Liski, J., Trofymow, J. & Titus, B. Litter decomposition aected by
climate and litter quality—testing the Yasso model with litterbag data from the
Canadian intersite decomposition experiment. Ecol. Modell. 189, 183–198
43. Bradford, M. A. et al. Climate fails to predict wood decomposition at regional
scales. Nat. Clim. Change4, 625 (2014).
44. Evans, J. S. & Murphy, M. A.rfUtilities: random forests model selection and
performance evaluation. R package version 2.1-3 https://cran.r-project.org/
45. Core Writing Team et al. (eds) Climate Change 2014 Synthesis Report (IPCC,
46. Schepaschenko, D. et al. A dataset of forest biomass structure for Eurasia.
Sci. Data 4, 170070 (2017).
nature research | reporting summary October 2018
Corresponding author(s): Brian S Steidinger
Last updated by author(s): Apr 4, 2019
Nature Research wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency
in reporting. For further information on Nature Research policies, see Authors & Referees and the Editorial Policy Checklist.
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement
A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly
The statistical test(s) used AND whether they are one- or two-sided
Only common tests should be described solely by name; describe more complex techniques in the Methods section.
A description of all covariates tested
A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons
A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient)
AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals)
For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted
Give P values as exact values whenever suitable.
For Bayesian analysis, information on the choice of priors and Markov chain Monte Carlo settings
For hierarchical and complex designs, identification of the appropriate level for tests and full reporting of outcomes
Estimates of effect sizes (e.g. Cohen's d, Pearson's r), indicating how they were calculated
Our web collection on statistics for biologists contains articles on many of the points above.
Software and code
Policy information about availability of computer code
Data collection Data were derived from the publicly available Global Forest Biodiversity Initiative (GFBi)
Data analysis All analyses were performed in R Studio (version 1.0.136), using the packages "raster", "randomForest", "forestFloor", "rgdal", "rgeos",
and "hypervolume", "rUtilities", and "taxize" (using the Tree genera were assigned to a plant family using a plant taxonomy lookup table
generated by Will Cornwell (hosted on Github https://github.com/traitecoevo/taxonlookup), which uses the accepted taxonomy from
“The Plant List.”) Raster images were exported from the Panoply (version 4.10.3). Multipanel figures were assembled in Inkscape
For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors/reviewers.
We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Research guidelines for submitting code & software for further information.
Policy information about availability of data
All manuscripts must include a data availability statement. This statement should provide the following information, where applicable:
- Accession codes, unique identifiers, or web links for publicly available datasets
- A list of figures that have associated raw data
- A description of any restrictions on data availability
The GFBi database is available upon written request at https://www.gfbinitiative.org/datarequest. Additionally, the symbiotic state assigned to tree species as a
supplementary file, as are global rasters of our model projections for EM, AM, and N-fixer proportion of tree basal area.
nature research | reporting summary October 2018
Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.
Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf
Ecological, evolutionary & environmental sciences study design
All studies must disclose on these points even when the disclosure is negative.
Study description ll tree individuals from >1.1 million forest inventory plots were assigned to symbiotic state using published data-bases of plant
taxonomy vs. root-symbioses. Plots were aggregated to a 1 by 1 degree grid along with a total of 75 climatic, soil, topographic, and
vegetative predictor layers. The random forest algorithm was used to determine the importance and influence of individual variables.
We simplified our models by removing variables with low inc node purity, a measurement of variable of importance. We then
present and interpret the environmental determinants of symbiotic state.
Research sample The GFBi database consists of individual-based data that we compiled from all the regional and national GFBi forest inventory data
sets. The standardized GFBi data frame, i.e. tree list, comprises tree ID, a unique number assigned to each individual tree; plot ID, a
unique string assigned to each plot; plot coordinates, in decimal degrees of WGS84 datum; tree size, in diameter-at-breast-height;
trees-per-hectare expansion factor; year of measurement; data set name, a unique number assigned to each forest inventory data
set; and binomial scientific tree species names.
Sampling strategy Plot-level data was aggregated to a 1 by 1 degree (lat / long) grid.
Data collection GFBI plots cover a significant portion of the global forest extent, including some of the most unique forest conditions: (a) the
northernmost (73°N, Central Siberia, Russia), (b) southernmost (52°S, Patagonia, Argentina), (c) coldest (-17°C annual mean
temperature, Oimyakon, Russia), (d) warmest (28°C annual mean temperature, Palau, USA) plots, and (e) most diverse (405 tree
species on the 1-ha plot, Bahia, Brazil). Plots in war-torn regions were assigned fuzzed coordinates to protect the identity of the plots
Timing and spatial scale GFBI plots cover a significant portion of the global forest extent, including some of the most unique forest conditions: (a) the
northernmost (73°N, Central Siberia, Russia), (b) southernmost (52°S, Patagonia, Argentina), (c) coldest (-17°C annual mean
temperature, Oimyakon, Russia), (d) warmest (28°C annual mean temperature, Palau, USA) plots, and (e) most diverse (405 tree
species on the 1-ha plot, Bahia, Brazil). Plots in war-torn regions were assigned fuzzed coordinates to protect the identity of the plots
Data exclusions N/A
Did the study involve field work? Yes No
Reporting for specific materials, systems and methods
We require information from authors about some types of materials, experimental systems and methods used in many studies. Here, indicate whether each material,
system or method listed is relevant to your study. If you are not sure if a list item applies to your research, read the appropriate section before selecting a response.
Materials & experimental systems
n/a Involved in the study
Eukaryotic cell lines
Animals and other organisms
Human research participants
n/a Involved in the study
Nature Research, brought to you courtesy of Springer Nature Limited (“Nature Research”)
Terms and Conditions
Nature Research supports a reasonable amount of sharing of content by authors, subscribers and authorised or authenticated users (“Users”), for
small-scale personal, non-commercial use provided that you respect and maintain all copyright, trade and service marks and other proprietary
considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the terms, a site licence or a personal subscription (to the
extent of the conflict or ambiguity only). By sharing, or receiving the content from a shared source, Users agree to be bound by these Terms.
We collect and use personal data to provide access to the nature content. ResearchGate may also use these personal data internally within
ResearchGate and share it with Nature Research, in an anonymised way, for purposes of tracking, analysis and reporting. Nature Research will
Users and the recipients of the nature content may not:
use the nature content for the purpose of providing other users with access to content on a regular or large scale basis or as a means to
circumvent access control;
use the nature content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability,
or is otherwise unlawful;
falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by either Nature
Research or ResearchGate in writing;
use bots or other automated methods to access the nature content or redirect messages; or
override any security feature or exclusionary protocol.
remove it or features or functionality at our sole discretion, at any time with or without notice. We may revoke this licence to you at any time and
remove access to any copies of the shared content which have been saved.
Sharing of the nature content may not be done in order to create substitute for our own products or services or a systematic database of our
content. Furthermore, we do not allow the creation of a product or service that creates revenue, royalties, rent or income from our content or its
inclusion as part of a paid for service or for other commercial gain. Nature content cannot be used for inter-library loans and librarians may not
upload nature content on a large scale into their, or any other, institutional repository.
To the fullest extent permitted by law Nature Research makes no warranties, representations or guarantees to Users, either express or implied
with respect to the nature content and all parties disclaim and waive any implied warranties or warranties imposed by law, including
merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Nature Research that we license from
If you intend to distribute our content to a wider audience on a regular basis or in any other manner not expressly permitted by these Terms
please contact us at
The Nature trademark is a registered trademark of Springer Nature Limited.