Tradeoffs between savanna woody plant diversity and carbon storage in the Brazilian Cerrado

Article (PDF Available)inGlobal Change Biology 22(10) · February 2016with 346 Reads
DOI: 10.1111/gcb.13259
Incentivizing carbon storage can be a win-win pathway to conserving biodiversity and mitigating climate change. In savannas, however, the situation is more complex. Promoting carbon storage through woody encroachment may reduce endemic plant diversity, even as the diversity of encroaching forest species increases. This trade-off has important implications for the management of biodiversity and carbon in savanna habitats, but has rarely been evaluated empirically. We quantified the nature of diversity-carbon relationships in the Brazilian Cerrado by analyzing how woody plant species richness changed with carbon storage in 206 sites across the 2.2 million km(2) region at two spatial scales. We show that total woody plant species diversity increases with carbon storage, as expected, but that the richness of endemic savanna woody plant species declines with carbon storage both at the local scale, as woody biomass accumulates within plots, and at the landscape scale, as forest replaces savanna. The sharpest tradeoffs between carbon storage and savanna diversity occurred at the early stages of carbon accumulation at the local scale but the final stages of forest encroachment at the landscape scale. Furthermore, the loss of savanna species quickens in the final stages of forest encroachment, and beyond a point, savanna species losses outpace forest species gains with increasing carbon accumulation. Our results suggest that although woody encroachment in savanna ecosystems may provide substantial carbon benefits, it comes at the rapidly accruing cost of woody plant species adapted to the open savanna environment. Moreover, the dependence of carbon-diversity tradeoffs on the amount of savanna area remaining requires land managers to carefully consider local conditions. Widespread woody encroachment in both Australian and African savannas and grasslands may present similar threats to biodiversity. This article is protected by copyright. All rights reserved.
Figures - uploaded by Adam F. A. Pellegrini
Author content
All content in this area was uploaded by Adam F. A. Pellegrini
Trade-offs between savanna woody plant diversity and
carbon storage in the Brazilian Cerrado
Department of Ecology and Evolutionary Biology, Princeton University, Princeton, NJ 08544, USA,
School of Aquatic and
Fishery Sciences, University of Washington, Seattle WA 98105, USA
Incentivizing carbon storage can be a win-win pathway to conserving biodiversity and mitigating climate change. In
savannas, however, the situation is more complex. Promoting carbon storage through woody encroachment may
reduce plant diversity of savanna endemics, even as the diversity of encroaching forest species increases. This trade-
off has important implications for the management of biodiversity and carbon in savanna habitats, but has rarely
been evaluated empirically. We quantified the nature of carbon-diversity relationships in the Brazilian Cerrado by
analyzing how woody plant species richness changed with carbon storage in 206 sites across the 2.2 million km
region at two spatial scales. We show that total woody plant species diversity increases with carbon storage, as
expected, but that the richness of endemic savanna woody plant species declines with carbon storage both at the local
scale, as woody biomass accumulates within plots, and at the landscape scale, as forest replaces savanna. The shar-
pest trade-offs between carbon storage and savanna diversity occurred at the early stages of carbon accumulation at
the local scale but the final stages of forest encroachment at the landscape scale. Furthermore, the loss of savanna spe-
cies quickens in the final stages of forest encroachment, and beyond a point, savanna species losses outpace forest
species gains with increasing carbon accumulation. Our results suggest that although woody encroachment in
savanna ecosystems may provide substantial carbon benefits, it comes at the rapidly accruing cost of woody plant
species adapted to the open savanna environment. Moreover, the dependence of carbon-diversity trade-offs on the
amount of savanna area remaining requires land managers to carefully consider local conditions. Widespread woody
encroachment in both Australian and African savannas and grasslands may present similar threats to biodiversity.
Keywords: biodiversity, carbon-biodiversity cobenefits, carbon sequestration, cerrado, forest, savanna, woody encroachment
Received 30 September 2015 and accepted 20 February 2016
Savannas are responsible for globally significant carbon
fluxes, and the balance between carbon sequestration
via plant biomass accumulation and emission via bio-
mass burning influences the trajectory of the global
land carbon sink (Van Der Werf et al., 2010; Liu et al.,
2015). The great capacity for carbon storage in savannas
(Scurlock & Hall, 1998) as well as the perception of
savannas as degraded landscapes (Sasaki & Putz, 2009)
has raised the issue of their possible management for
carbon sequestration (Midgley & Bond, 2015; Veldman
et al., 2015a,b).
In some ecosystems (e.g., tropical forests), promoting
carbon storage may lead to greater biodiversity, thus
bringing conservation-climate mitigation cobenefits
(Venter et al., 2009; Gilroy et al., 2014). In the savanna
ecosystem, which is defined as having sparse tree cover
and a continuous herbaceous layer (Scholes & Archer,
1997; Bond, 2008), the relationship between diversity
and carbon is more complex. Across savannas globally,
fire suppression generally leads to woody thickening
and eventually forest encroachment in areas with suffi-
cient rainfall (Higgins et al., 2007; Bradstock & Wil-
liams, 2009; Staver et al., 2011); during this process,
there is substantial carbon accumulation in above- and
belowground pools (Grace et al., 2006; Heckbert et al.,
2012; Pellegrini et al., 2014) and a rapid colonization by
tree species typical of forests (Moreira, 2000; Rodrigues-
Souza et al., 2015). However, the plant species adapted
to the open savanna habitat tend to be excluded from
the ecosystem during forest formation (Durigan & Rat-
ter, 2006; Hoffmann et al., 2012; Rodrigues-Souza et al.,
Management strategies that promote carbon seques-
tration could have significant consequences for savanna
plant species in the Brazilian Cerradoa biodiversity
hotspot spanning >2 million km
with specially
adapted plant and animal communities, many of which
are found nowhere else on Earth (Myers et al., 2000;
Ratter et al., 2006). Loss of plant species from the Cer-
rado region due to land-use change has previously
been reported (Klink & Machado, 2005), but recent fire
suppression and subsequent forest encroachment are
Correspondence: Adam F.A. Pellegrini, tel. +1 612 669 5952, fax +1
609 258 1712, e-mail:
3373©2016 John Wiley & Sons Ltd
Global Change Biology (2016) 22, 3373–3382, doi: 10.1111/gcb.13259
increasingly pressing threats (Durigan & Ratter, 2006,
2015; Pinheiro & Monteiro, 2010). Governmental fire
prevention measures, with encouragement from non-
governmental organizations (Veldman et al., 2015b),
risk triggering widespread declines of plant species that
are adapted to the open savanna habitat and thus
dependent on disturbance by fire (Parr et al., 2014;
Durigan & Ratter, 2016; Veldman et al., 2015a,b). At the
same time, fire in the Cerrado region contributes sub-
stantially to annual greenhouse gas emissions (Van Der
Werf et al., 2010), and carbon sequestration from fire
suppression may constitute a large sink offsetting emis-
sions (Grace et al., 2006; Pellegrini et al., 2014). The
simultaneous loss of savanna species, gain in forest spe-
cies, and gain in ecosystem carbon stocks resulting
from woody thickening and forest encroachment dur-
ing fire suppression produces a conservation dilemma.
The decision to manage the Cerrado and numerous
other savanna landscapes globally (see Veldman et al.,
2015b) for either carbon or endemic savanna species
diversity has large implications for both climate change
and the preservation of thousands of plant species;
however, the severity and nature of the trade-off
between savanna endemic diversity and carbon remain
unknown. Quantitative evaluation of this trade-off is
critical because (i) the magnitude of savanna endemic
diversity losses will determine the absolute diversity
cost of gaining carbon and (ii) the nature of the relation-
ship between carbon and savanna species biodiversity
will determine the stage of carbon accumulation (e.g.,
initial vs. final) where losses are most severe.
To address the issue, we quantified how the diversity
of woody plants change with increasing carbon storage
using a dataset of >18 000 observations of 1797 species
censused at 206 sites (totaling 209 ha in area sampled)
distributed across savannas and forests spanning the
2.2 million km
Cerrado region (Fig. S1). We utilize a
space-for-time substitution by sampling plots spanning
a range of woody plant biomass (open savanna to
dense forest) under the assumption that in terms of
their carbon storage and species composition, the plots
mimic the outcome of different approaches to fire man-
agement on plots of Cerrado that are currently savanna
(e.g., Moreira, 2000; Henriques & Hay, 2002; Pellegrini
et al., 2014). We consider changes in diversity and even-
ness of woody plant species adapted to savannas or for-
ests (i.e. savanna species diversity vs. forest species
diversity) as well as total species diversity (i.e., all
woody plant species) using multiple diversity indices.
We hypothesize that (i) increases in total species diver-
sity with carbon gains will mask large losses of species
endemic to savannas and (ii) the rate of change in
diversity will depend on the standing amount of
woody biomass and the remaining area of savanna.
To test our hypotheses, we analyzed the impact of
carbon accumulation on different components of diver-
sity at two spatial scales. At the local scale, we mea-
sured how the richness of the woody plant species
adapted to savanna or forests (i.e., savanna or forest
guild) or total woody plant diversity changes with
increasing carbon stocks at the plot level (i.e., due to
woody thickening). At the regional scale, we measured
how the richness of woody plant species in savanna
and forest landscapes changed with increasing carbon
due to converting savanna area into forest (i.e., due to
woody or forest encroachment).
Materials and methods
Quantifying biodiversity
The Cerrado region in Brazil (Fig. S1) spans a large climatic
gradient (ranging from semi-arid to mesic environments),
with the core region receiving ~14001600 mm mean annual
precipitation and a pronounced dry season from May to
September (Oliveira & Marquis, 2002). The Cerrado region
contains a number of vegetation formations that differ in
structure and species compositions, the most prominent being
savannas and forests, which we here refer to as biomes.
Although the herbaceous layer contains a wealth of species
(Durigan & Ratter, 2016), we focus on woody plants (shrubs
and trees, generally >5 cm in basal diameter) in savannas and
forests because they are critical carbon storage reservoirs for
the land carbon sink (Dixon et al., 1994) and one of the essen-
tial criteria for defining the Cerrado region as a biodiversity
hotspot (Myers et al., 2000).
In order to analyze how woody plant diversity changes
with woody biomass, we compiled and analyzed a plant bio-
diversity database containing complete floristic surveys from
savanna, woodland (referred to as cerrad~
ao in Brazil, which
contains intermediate tree cover and biomass between
savanna and forest), and forests from 206 sites with minimal
anthropogenic disturbance (e.g. located in protected areas or
in sites where the authors did not note cultivation and/or
deforestation). These plots were distributed across the Cer-
rado region and into the adjacent Atlantic Forest region with
similar environments (Fig. S1; references listed in Table S1).
Forests spanned a range of types including riparian, semi-
deciduous, and dry forests (montane and rainforests were
excluded). These plots span a woody biomass gradient from
open savanna to dense forest.
Plant species in forests tend to be shade-tolerant but fire-
sensitive, while savanna species are relatively fire-tolerant but
shade-intolerant (Hoffmann et al., 2012). As a result of these
differences, distinct plant communities occupy forest and
savanna biomes, with few generalist species capable of thriv-
ing in both (Hoffmann et al., 2012). Due to the potential biases
and inconsistencies in the ‘expert opinion’ approach to classifi-
cation, we classified species as belonging to savanna or forest
functional guilds based on their relative occurrences in
savanna or forest vegetation formations using a species
©2016 John Wiley & Sons Ltd, Global Change Biology,22, 3373–3382
3374 A. F. A. PELLEGRINI et al.
association index (C
aceres & Legendre, 2009). The association
index statistically evaluates whether species occur in a particu-
lar habitat disproportionately more than would be expected by
chance. The species classifications were then used to quantify
plant species in each site as savanna species or forest species.
We were unable to classify species that were generally rare
in either savanna or forest plots due to a lack of statistical
power (e.g., ~70% of the unclassified species occurred less
than three times). For the more abundant unclassified species,
we assessed the potential of incorporating them into guilds
based on a plant diversity database (Flora do Brasil, 2015). We
discuss this potential classification procedure in detail in the
Supplementary Information; but briefly, of the unclassified
species that occurred >5 times, ~40% of the species were classi-
fied as occurring in a ‘dense’ vegetation in Flora do Brasil,
which can be interpreted as a dense savanna woodland, an
ambiguous definition to be included in an analysis of species
preferences. Due to the variety of different environmental con-
ditions that this classification may be referring to, we maintain
the conservative assumption that the unclassified species can-
not robustly be categorized into either savanna or forest
Importantly, although the number of unclassified species
limits our ability to quantify exact changes in species richness,
there were substantially fewer unclassified species than classi-
fied species in both savanna plots (20% unclassified vs. 53%
classified as savanna-guild species) and forest plots (21%
unclassified vs. 68% as forest guild species).
Quantifying carbon in aboveground woody biomass
For each site in the database, we quantified aboveground car-
bon in woody biomass using allometric relationships. We
employed a general allometric equation for tropical savannas
that has been used to calculate woody biomass stocks in
savannas (Supplementary Information; Lehmann et al., 2014).
Although savanna and forest species can differ in their foliar
biomass allometries (i.e., leaf biomass per-unit stem diameter),
we know of no evidence that they differ in woody biomass.
As such, we used the same allometric equation for forest spe-
cies for consistency. This equation also expresses woody bio-
mass as a function of total basal area in a plot, thereby
allowing us to quantify biomass in studies that do not present
stem diameter measurements of individuals.
Here, we focus on woody biomass as the primary changing
carbon pool for two reasons. First, shifts in carbon storage in
the woody biomass pool occur on timescales most relevant for
mitigating carbon emissions (Pan et al., 2011). Second, other
vegetation carbon pools such as the herbaceous layer are small
in comparison to changes in woody biomass carbon (e.g.,
woody biomass can increase from 10 to 100 MgC ha
herbaceous biomass losses only amount to ~3 MgC ha
; Pel-
legrini et al., 2014). Finally, soil pools can increase with forest
development (a previous study found gains of ~30 MgC ha
across a gradient in forest growth [Pellegrini et al., 2014]), but
the generality of these changes needs to be verified before
incorporated into a full analysis (Silva et al., 2013). We discuss
the implications below.
Calculating species richness at the local scale
To quantify woody plant species richness at the local (plot)
scale, we first standardized measures of local species richness
to a common area of 0.5 ha because the inventories at each
plot varied in the total area surveyed (median of 1 ha; inter-
quartile range 0.7–1.2). A spatial scale of 0.5 ha was chosen
because (i) smaller plots tend to inaccurately capture both true
species diversity and carbon storage (Wagner et al., 2010) and
(ii) a large number of the plots compiled from the literature
exceed this size threshold (n=169 of 206). To calculate stan-
dardized measures of species richness at the 0.5 ha plot scale,
we removed plots smaller than 0.5 ha, resulting in 169 plots
with a median area of 1 ha. For all plots larger than 0.5 ha, we
rarefied individuals out of the plots to create a set of resam-
pled half-hectare plots. In order to calculate stem densities in
plots rarefied to 0.5 ha, we determined the number of stems
expected within a half-hectare for each plot by rescaling the
original density of stems within each plot to a half-hectare,
removing individuals at random until the density was
Because we were interested in quantifying species richness
of both the savanna and forest guilds as well as total species
richness (i.e., all woody plants in a plot), we performed this
analysis three times in separate rarefactions to determine the
species richness of (i) savanna guild species, (ii) forest guild
species, and (iii) all woody plant species. To explore whether
other indicators of biodiversity and community composition
change with carbon, we also calculated Shannon’s diversity
index and Pielou’s evenness index.
We used maximum likelihood methods to model the rela-
tionship between carbon storage and the richness of the
savanna guild, forest guild, and total species at the plot scale.
Because we were concerned with accurately capturing the
precise shape of the relationship between carbon and species
richness, we analyzed multiple models with different func-
tional forms to compare both linear and nonlinear changes.
We used five functional forms that are observed during spe-
cies turnover through succession: saturating, linear, quadra-
tic, logarithmic, and sigmoidal (see Supplementary
Information for exact functional forms). We used the Akaike
information criterion (AIC) to compare the performance of
these five models. All analyses were performed in R using
the anneal function in the likelihood package (R Development
Core Team, 2010).
To evaluate possible spatial autocorrelation of the model
fits, we calculated the Moran’s I on model residuals. Follow-
ing a significant signature of spatial autocorrelation, we
inspected a correlogram to determine the minimum neces-
sary separation between sites to eliminate spatial autocorre-
lations (Hua et al., 2015). We then used this minimum
distance to resample groups of plots sufficiently far apart,
which we then fit with the appropriate selected model (see
Supplementary Information for a full explanation). We iter-
ated this process 100 times to generate means and confi-
dence intervals on the fitted parameters and goodness of fits
on the full model and compared the resampled model to the
full dataset model to evaluate whether our conclusions were
robust to autocorrelation.
©2016 John Wiley & Sons Ltd, Global Change Biology,22, 3373–3382
Calculating species richness at the landscape scale
To quantify how species richness changed with increasing car-
bon storage at the landscape scale, we assumed the landscape
area could be broadly defined as either savanna or forest (Sta-
ver et al., 2011). We do not explicitly account for the interme-
diate formation of woodland vegetation, which is relatively
limited in extent compared to the distribution of savanna and
forest (Hirota et al., 2010, 2011). To quantify the number of
species in savanna, we generated an individual-based species
accumulation curve for savanna tree species across all plots
within the savanna biome, excluding species classified as for-
est guild species. We then repeated these analyses by quanti-
fying the number of species in forest in a similar fashion,
excluding species classified as belonging to the savanna guild
(Supplementary Information; Fig. S2).
In contrast to our local-scale species richness analyses, here
we included unclassified species present in either the savanna
or forest plots in analyses of landscape-scale richness because
these unclassified rare species can make large contributions to
landscape-scale biodiversity (Bridgewater et al., 2004). Overall,
the unclassified species represent 47% of total species found in
savanna and 52% of total species found in forest. Importantly,
75% of the unclassified species only occurred in either savan-
nas or forests (i.e., not in both), thereby minimizing double
counting of species that may occur in both forest and savanna.
To quantify the number of species throughout the land-
scape, savanna and forest species accumulation curves were
fit with a power-law function, as predicted from theoretical
expectations of species-area curves (Tjørve, 2009 and see Sup-
plementary Information), using a nonlinear least-squares
model to estimate parameters. These species accumulation
curves were then used to extrapolate the number of species in
a landscape based on the expected number of individuals in
the landscape, determined by multiplying an average density
of individuals (calculated for savanna and forest from the
empirical data) by the total area of either savanna or forest. As
a robustness check to this extrapolation approach, we per-
formed a swapping rarefaction of the actual plots (see Supple-
mentary Information for a detailed description) to verify the
nonlinearity while preserving local abundance distributions
and allowing that savanna species might persist in forests.
To obtain estimates of species richness at the landscape scale
relevant to the current conservation efforts in the Cerrado
region, we examined the IUCN-designated protected areas
(PA; categories I and II, n=74) in the Cerrado region. Within
each PA, we determined the area of savanna and forest (using
50% tree cover as a break point, sensu Hansen et al., 2013) using
a high-resolution (~30 930 m) LANDSAT satellite product of
tree cover (Hansen et al., 2013). We then determined species
richness in savanna and forest in each PA using the method
described above. These estimates were used to quantify the
potential trade-offs between savanna species richness and car-
bon accumulation and explore the variability across PAs.
At the local scale, the total number of woody plant spe-
cies increased with plot biomass. The best fit curve was
a quadratic function (r
=0.32), illustrating that initial
gains in total richness were rapid but began to saturate
at high biomass values (Fig. 1a). It is unlikely that the
slight decline in the fitted quadratic at high biomass
values is biologically meaningful given the high diver-
sity of mature forests in this region (Ratter et al., 2006).
A logarithmic function with saturating total richness at
high biomass values explained similar levels of vari-
ance (r
=0.30) but was not chosen because it had a
higher AIC (Table S2).
The increase in overall plot biodiversity concealed
large losses in the savanna guild. Species richness of
the savanna guild declined monotonically and nonlin-
early as carbon stocks increased (r
=0.36, Figs 1b and
S3). Initially, the decline in savanna species richness
with increasing carbon storage was rapid, but deceler-
ated as carbon stocks continued to increase and species
became rarer (Fig. 1b, c). In contrast, the richness of the
forest guild increased rapidly with carbon storage at
low carbon density, but saturated with increasing car-
bon (r
=0.44, Figs 1b, c and S3).
Analyses of the Shannon diversity index produced
similar patterns: a nonlinear trade-off between gains in
total species diversity and losses of savanna-guild
diversity (Supplementary Information,Fig. S4). Even-
ness did not significantly change across the gradient for
the total woody plant community, species in the
savanna guild, or species in the forest guild (Supple-
mentary Information, Fig. S5). Given the consistencies
between the Shannon diversity index and total richness,
we focus on total richness for further analysis.
We next evaluated whether spatial autocorrelation
may be influencing the richness-carbon relationship.
Moran’s I indicated significant autocorrelation in the
residuals of the model between species richness and
carbon (Supplementary Information). Inspection of the
autocorrelation using a correlogram illustrated that a
minimum distance of 300 km between plots was
required for residuals to be independent. Repeated
random resampling of independent plots (i.e., plots
separated by >300 km) and fitting with maximum
likelihood indicated that spatial autocorrelation did
not substantially affect fits for the forest guild,
savanna guild, and total species richness (Fig. S6;
Table S4); the full dataset model fell within the confi-
dence intervals of the resampled model ensemble
(Fig. S6) and differences in the fitted coefficient values
were minor (Tables S3 and S4). Moreover, the mean
goodness of fit (r
) values of the resampled models
were similar to those obtained from the full dataset
model (Table S4). Consequently, the resampled mod-
els were consistent with the full model and we con-
clude our model results are robust to possible spatial
©2016 John Wiley & Sons Ltd, Global Change Biology,22, 3373–3382
3376 A. F. A. PELLEGRINI et al.
We next considered how diversity of species in a
savanna landscape may change if savanna area was
lost due to forest encroachment. From the rarefaction
analysis, we found that savanna species richness rose
rapidly with increases in the abundance of individuals
but that the rates of accumulation declined with
increasing abundance of individuals (Fig. 2a). Corre-
spondingly, there was an exponential decline in
savanna species richness losses per unit carbon gain
and increasing total savanna area within a PA
(Fig. 2b). Small PAs (e.g., <100 hectares) showed dra-
matic losses of savanna species with carbon gains
(Fig. 2b), illustrating the sensitivity of particular PAs to
carbon gains. On the other hand, large PAs displayed
much less pronounced reductions in savanna species
richness with increasing carbon, resulting in substan-
tial variability in the sensitivity of savanna species rich-
ness to a fixed absolute amount of forest encroachment
across PAs (Figs 2c and S7). We confirmed the nonlin-
ear nature of these declines in savanna species richness
using an additional approach that accounted for the
potential for savanna species to exist in forest plots
(Supplementary Information).
The loss in savanna species richness was mirrored
by strong gains in forest species richness with
increasing forest area. Importantly, however, the ratio
of forest species gains vs. savanna species losses was
dependent on the total size of the area being consid-
ered. When the combined area of savanna and forest
was small, gains in forest species richness with
increasing forest area were greater than the losses of
savanna species richness (Fig. 3a). In contrast, when
the combined area of savanna and forest was large,
gains in forest species richness with increasing forest
area were less than losses of savanna species richness
(Fig. 3b). Consequently, in locations with relatively
small areas of forest and savanna, the number of for-
est species gained exceeds the number of savanna
species lost per unit of carbon gained; however, in
locations with large areas of both savanna and forest,
the number of savanna species lost exceeds the num-
ber of forest species gained per unit of carbon
To illustrate the variety in the potential ratio of forest
species gained vs. savanna species loss per unit of car-
bon gained, we determined the ratio across the PAs,
which vary in size and proportional areas of savanna
and forest (Fig. 3c). Our ability to quantify the ratio in
each PA is limited by the extrapolation that we can
make with our species accumulation curves, but, once
again, the qualitative nature of these functional forms
will produce this trade-off function. For the PAs that
contained estimated individuals within the bounds of
our species accumulation curves (Fig. S2), the calcu-
lated ratios ranged from 0.2 to 8.1 (Fig. 3c), with values
<1 indicating reserves where savanna species losses
exceeded forest species gains with increasing carbon
(calculated to be 39 PAs, Fig. 3c).
Plot C (MgC/0.5 ha)
Plot C (MgC/0.5 ha)
Number of species
per-C species change
Plot C (MgC/0.5 ha)
Number of species
Savanna guild
Forest guild
All species
Savanna guild
Forest guild
Fig. 1 (a) Species richness of total woody plant community
with fitted quadratic (r
=0.32, n=151); (b) changes in species
richness for savanna and forest guilds as a function of plot car-
bon with fitted quadratics (savanna: r
=0.36, n=151; forest:
=0.44, n=151). Richness was determined by subsampling
plots down to a common plot size of 0.5 ha (Supplementary
Information). Curves fitted using maximum likelihood with
model selection. (c) local-scale carbon-diversity trade-offs using
model estimates of the change in number of savanna and forest
guild species relative to change in carbon. Dots represent each
predicted value from a plot, colored by species guild.
©2016 John Wiley & Sons Ltd, Global Change Biology,22, 3373–3382
Taken together, these results illustrate that managing
for carbon and biodiversity in tropical savannas pro-
duces an important conservation dilemma. Overall,
and as expected, there is a general trend for total
woody plant biodiversity to increase with carbon stor-
age. The gains in total diversity were driven by
increases of woody plant species adapted to forests,
which were substantial enough in magnitude to over-
compensate for the losses of woody plant species
adapted to savannas. The loss of plant species that
exclusively occur in the open savanna biome presents a
key carbon-biodiversity trade-off. While savanna-guild
species always decline during afforestation by defini-
tion, our results reveal important nonlinearities in how
savanna species decline with increasing carbon accu-
Rates of savanna species losses at the plot scale are
steepest during the initial stages of woody thickening,
while rates of savanna species losses at the landscape
scale are steepest at the final stages of forest encroach-
ment (Figs 1b, c and 2b). Accordingly, where carbon
sequestration is a priority, it is best implemented by
maximizing sequestration at the plot scale in localized
areas while minimizing the total area encroached.
At the landscape scale, the saturating relationship
between forest species diversity and carbon results in
diminishing returns of carbon gains and forest species
gains as the amount of forested area increases. On the
other hand, declines in savanna species richness only
continue to accelerate. As a result, the number of forest
species gained becomes equal to the number of savanna
species lost. This represents an important transition
point, past which managing for carbon will likely result
in larger savanna species losses than gains in forest spe-
cies per unit of carbon stored.
The transition point thus signifies a threshold sepa-
rating contrasting outcomes of (i) maximizing carbon
while realizing modest forest diversity gains and con-
siderably greater losses of savanna species, or (ii)
restricting forest regrowth and carbon storage potential
while retaining a large amount of savanna species
diversity. Land managers seeking to balance the multi-
ple objectives of carbon storage and maintenance of
savanna species diversity should prioritize this second
outcome, managing the landscape in reserves below
this threshold. Although our sample size limits our
ability to provide an exact quantitative assessment of
each PA, as long as the exponent of the fitted species
accumulation curve in savannas exceeds that of forests,
this threshold will occur. In turn, quantifying this tran-
0 40 000 80 000 120 000
1 100 10 000 1 000 000
Area currently savanna in a PA (ha)
Species lost
per-MgC gained
Number of individuals
(a) (b)
10 000
100 000
1 000 000
Reserve size (ha)
per- detseroffaah
Individual protected areas
spp. loss
Reserve size
Fig. 2 (a) Sample-based rarefaction of total species richness in savanna sites, expressed as a function of individuals. Red dots indicate
the estimated richness of Cerrado protected areas given total savanna area and corresponding estimated abundance of individuals in
each PA (n=72, Supplementary Information). (b) Relationship between the number of species lost per-MgC gained within a landscape
as a function of total savanna area within Cerrado PAs. (c) Comparison of the distribution of species lost per-ha converted from
savanna to forest (gray) with park area (red) across the Cerrado PAs. The loss rate is a function of the total savanna area within a PA.
The total size of the PA sets a limit on the potential losses, but within PAs of equivalent sizes there can be large variability due to the
actual amount of savanna area (in this case relative to forest area).
©2016 John Wiley & Sons Ltd, Global Change Biology,22, 3373–3382
3378 A. F. A. PELLEGRINI et al.
sition point based on surveys within specific reserves
undergoing forest encroachment (e.g., Durigan &
Ratter, 2006) would be a useful next step to identify the
relative biodiversity costs and benefits by promoting
carbon storage.
Studies documenting composition shifts during
woody thickening and biomass accumulation have
found inconsistent and sometimes contrasting changes
in the woody plant diversity of the savanna guild,
including increases, no change, or decreases (Moreira,
2000; Roitman et al., 2008; Rodrigues-Souza et al.,
2015). Disagreement among studies may be due to a
number of factors, such as the initial woody plant bio-
mass in a plot (and thus how quickly the canopy
closes and savanna species become excluded) and the
length of the study period (sufficient enough time to
allow for substantial shifts in vegetation structure and
subsequent turnover in community composition; Duri-
gan & Ratter, 2006). Here, using a space-for-time sub-
stitution, we find that the richness of woody plant
species endemic to the savanna biome does not bene-
fit from woody thickening and on average savanna
species richness declines quickly across plots increas-
ing in woody biomass. The replacement of savanna
species by forest species during carbon accumulation
in a plot reflects the effective competitive exclusion of
the slow-growing shade-intolerant savanna species by
fast-growing shade-tolerant forest species. This is
likely reflective of the faster growth and colonization
ability of forest species, whose recruitment can greatly
outpace savanna species (Roitman et al., 2008; Hoff-
mann et al., 2012).
Quantitative predictions of carbon-diversity relation-
ships at the landscape scale might be made more accu-
rate by accounting for spatially structured tree beta
diversity within the Cerrado (Ratter et al., 2003; Bridge-
water et al., 2004). Beta diversity in savannas in the Cer-
rado is estimated to be high, with previous studies
finding 35% of recorded species to occur in a single site
only (Ratter et al., 2003), and Sørensen similarity indices
to range from 0.37 to 0.625 in the core Cerrado region
(Bridgewater et al., 2004), illustrating high species turn-
over. While we do not calculate beta diversity here,
given the potential limitations arising from extrapolat-
ing across space in areas where plots are infrequent
(Fig. S1), we found a similar pattern of rare species con-
tributing a large amount to overall richness. Because
our sites spanned the entire Cerrado region, the overall
species accumulation curve from which our estimates
were drawn may be steeper than the species accumula-
0 1000 2000 3000 4000 5000
For spp. gain perC
Sav spp. loss perC
10 100 1000 10 000
per-C gain
Area of either savanna or forest biome (ha) Area of either savanna or forest biome (ha)
Ratio of forest sp. gains
/savanna sp. losses
10 000
100 000
1 000 000
Reserve size (ha)
(a) (b)
Reserve size
Individual protected areas
Fig. 3 (a) Gains vs. losses of species richness per-unit carbon gain for savanna and forest species at the landscape scale. Forest species
richness increases per-C gain, and is expressed as a function of forest biome area; savanna species richness decreases per-C gain, and is
expressed as a function of savanna biome area. (b) Ratio of per-carbon richness changes of forest vs. savanna species [the ratio of the
two curves in (a)] as a function of existing biome area. Red asterisk illustrates intersection at ~4440 ha. (c) Distribution of ratios (species
richness changes of forest vs. savanna species per-unit C gain) across the protected areas (black bars, primary axis) with the corre-
sponding size of the PA (red bars, secondary axis, n=72). Dashed line indicates a ratio of one.
©2016 John Wiley & Sons Ltd, Global Change Biology,22, 3373–3382
tion curve within each PA (owing to higher spatial
turnover of species across the Cerrado than within a
local PA; Bridgewater et al., 2004; Ratter et al., 2003).
Nevertheless, the main finding of nonlinear declines in
savanna-guild richness with carbon accumulation is
unlikely to change.
The large variability in the sensitivity of savanna-
guild diversity to carbon accumulation across PAs
points to the need to consider heterogeneity in potential
carbon-diversity relationships across PAs (Fig. S7). Spa-
tially explicit analyses of species diversity in the Cer-
rado savannas have found evidence for certain areas
being especially rich in woody plant diversity (Ratter
et al., 2010). Moreover, potential carbon storage in for-
ests also varies across the Cerrado (Pellegrini et al.,
2016). Combining maps of species richness with poten-
tial carbon storage would be one useful step to identify-
ing areas where the trade-offs are likely to be most
severe. Spatial analyses will also improve the evalua-
tion of potential species losses in individual PAs.
Comprehensive conservation planning in the Cer-
rado must also take into account other plant guilds
such as the highly diverse herbaceous species, as well
as animal groups. The savannas of the Cerrado are
exceptionally diverse in herbaceous species (Ratter
et al., 1997), which tend to be highly light sensitive and
thereby likely to be excluded during woody thickening
and forest encroachment. Recent work has identified
the afforestation of grasslands as a critical conservation
issue, with losses of herbaceous species being one
unequivocally negative consequence (Durigan & Ratter,
2016; Veldman et al., 2015b). Globally, grasslands have
diverse herbaceous layers (Veldman et al., 2015b) and
consideration of declines in herbaceous species richness
will be critical. Future studies should attempt to cap-
ture the complete biodiversity consequences of woody
thickening, which is likely to strengthen the relation-
ship between carbon accumulation and savanna ende-
mic declines and possibly negate the net-diversity gains
in total plant species with increasing plot carbon
Consideration of carbon pools in nonwoody vegeta-
tion will also help further refine the carbon-diversity
relationship. Carbon stored in soil layers can be sub-
stantial in grassy ecosystems (Jackson et al., 2002; Pelle-
grini et al., 2015), with the majority of total ecosystem
carbon in the savanna existing in belowground pools
(Grace et al., 2006). However, carbon in plant pools
tend to be the most responsive to changes in distur-
bance regime (Higgins et al., 2007) and shift on time-
scales relevant for mitigating carbon emissions (Pan
et al., 2011). These gains in carbon can offset the loss in
the herbaceous layer (~23 MgC ha
). Furthermore,
forests in general tend to have higher soil carbon than
savannas in the Cerrado region, which accumulates
during woody biomass accumulation and forest
encroachment. For example, a previous study estimated
gains of ~30 MgC ha
in the upper soil from open
savanna to closed-canopy transitional forest (Pellegrini
et al., 2014). Subsequently, while we might expect gains
in soil carbon to follow gains in woody biomass carbon,
the loss of belowground herbaceous biomass in deeper
soil layers can sometimes offset ecosystem carbon gains
(Jackson et al., 2002), requiring large-scale measurement
before generalities can be made.
The potential impact of fire on offsetting the carbon
gains during woody thickening and forest encroach-
ment needs to be considered as well. Woody plant spe-
cies adapted to savannas can accumulate thick bark
and grow quickly enough to survive fire (Hoffmann
et al., 2009) forming relatively fire-resistant stable car-
bon pools. In contrast, forest trees invest less biomass
in bark, and consequently are more sensitive to dying
during a fire event because their bark thickness is insuf-
ficient to protect their cambium from overheating
(Brando et al., 2012; Pellegrini et al., 2016). As a conse-
quence, the increase in carbon during woody thicken-
ing will be at high risk of being lost due to burning,
even once forests have formed. Forest fires in this
region occur regularly and are expected to increase
with shifting climates and greater droughts (Alencar
et al., 2011, 2015). When forests do burn, large amounts
of carbon can be lost (Kauffman et al., 1993; Morton
et al., 2013) and thus the expected gains in carbon may
be greatly diminished over the long term (Pellegrini
et al., 2016). This may result in significant reductions in
the effective number of forest species gained per unit
carbon stored over the long term.
Our space-for-time substitution allowed for the com-
parison of a large number of plots across a large spatial
scale, a method employed in other carbon-diversity
studies (e.g., Gilroy et al., 2014). This is specifically a
space-for-time substitution under the scenario of fire
exclusion and woody encroachment, which is a reason-
able expectation given current management (Durigan &
Ratter, 2006, 2016) and the trajectory of biomass accu-
mulation during exclusion (Moreira, 2000; Roitman
et al., 2008; Rodrigues-Souza et al., 2015). Consequently,
we argue that the different ‘spaces’ adequately reflect
different ‘times’ or levels of forest development, with
varying nonlinear consequences to savanna diversity.
Monitoring of fire exclusion experiments enabling a
truly temporal analysis would undoubtedly add useful
validity to the trends presented here.
The nature of carbon-diversity trade-offs is relevant
to a number of other savanna regions threatened by
woody encroachment due to either fire management or
other processes. In Africa, large areas are experiencing
©2016 John Wiley & Sons Ltd, Global Change Biology,22, 3373–3382
3380 A. F. A. PELLEGRINI et al.
woody thickening, attributed to a number of factors
such as increasing CO
and land management (Wigley
et al., 2010; Buitenwerf et al., 2012). Similar to Brazil,
fire suppression is one common candidate leading to
woody thickening, either through direct reduction of
fires or through increased grazing (Roques et al., 2001;
Higgins et al., 2007). However, African savannas tend
to be drier, which may limit the degree of woody
encroachment and possible carbon sequestration (Sta-
ver et al., 2011). In Australia, long-term changes in pre-
cipitation have been hypothesized to play a role in
woody encroachment (Fensham et al., 2005), but similar
to both Africa and South America, fire also influences
woody cover in Australia (Fensham et al., 2003). Conse-
quently, management of the fire regime is at the root of
limiting woody encroachment in savannas globally
(Bond et al., 2005). Future studies should carefully eval-
uate the diversity consequences of increased woody
biomass in African and Australian savannas.
In conclusion, our results illustrate that there is a
large carbon-diversity trade-off between maintaining
species endemic to savannas vs. promoting carbon
storage. When considering all species in a plot, biodi-
versity increases with carbon storage in an expected
manner, similar to previous studies (e.g., Gilroy et al.,
2014). In the Cerrado, this pattern is driven by the
fact that forests have more woody species than
savannas; the dramatic community replacement and
subsequent exclusion of savanna endemics that
occurs with carbon gains complicate the cobenefit
between carbon and total biodiversity. Importantly,
however, the relative carbon-diversity benefits for for-
est species vs. carbon-diversity trade-offs for savanna
species include a point of diminishing returns, where
increasing carbon will result in larger savanna spe-
cies losses than gains in forest species per unit of car-
bon stored. The nature of this trade-off must be
acknowledged in future management of this critical
biodiversity hotspot.
We would like to thank W. Hoffmann and G. Durigan for useful
discussion and four reviewers that greatly improved the quality
of the manuscript. We especially thank numerous authors and
botanists who collected the tree plot data and made it available
through their publications. AFAP was supported by a NSF-
GRFP, a National Geographic Society Young Explorer Grant,
and a Lassen Fellowship from the Program in Latin American
Studies, Princeton University.
Alencar A, Asner GP, Knapp D, Zarin D (2011) Temporal variability of forest fires in
eastern Amazonia. Ecological Applications,21, 23972412.
Alencar AA, Brando PM, Asner GP, Putz FE (2015) Landscape fragmentation, severe
drought, and the new Amazon forest fire regime. Ecological Applications,25, 1493
Bond WJ (2008) What limits trees in C4 grasslands and savannas? Annual Review of
Ecology, Evolution, and Systematics,39, 641659.
Bond W, Woodward F, Midgley G (2005) The global distribution of ecosystems in a
world without fire. New Phytologist,165, 525538.
Bradstock RA, Williams RJ (2009) Can Australian fire regimes be managed for carbon
benefits? New Phytologist,183, 931934.
Brando PM, Nepstad DC, Balch JK, Bolker B, Christman MC, Coe M, Putz FE (2012)
Fire-induced tree mortality in a neotropical forest: the roles of bark traits, tree size,
wood density and fire behavior. Global Change Biology,18, 630641.
Bridgewater S, Ratter JA, Ribeiro JF (2004) Biogeographic patterns, b-diversity and
dominance in the cerrado biome of Brazil. Biodiversity& Conservation,13,22952317.
Buitenwerf R, Bond W, Stevens N, Trollope W (2012) Increased tree densities in South
African savannas:>50 years of data suggests CO2 as a driver. Global Change Biol-
ogy,18, 675684.
aceres MD, Legendre P (2009) Associations between species and groups of sites:
indices and statistical inference. Ecology,90, 35663574.
Dixon RK, Solomon A, Brown S, Houghton R, Trexier M, Wisniewski J (1994) Carbon
pools and flux of global forest ecosystems. Science,263, 185190.
Durigan G, Ratter JA (2006) Successional changes in Cerrado and Cerrado/forest eco-
tonal vegetation in Western S~
ao Paulo State, Brazil, 1962-2000. Edinburgh Journal of
Botany,63, 119130.
Durigan G, Ratter JA (2016) The need for a consistent fire policy for Cerrado conser-
vation. Journal of Applied Ecology,53,1115.
Fensham R, Fairfax R, Butler D, Bowman D (2003) Effects of fire and drought in a
tropical eucalypt savanna colonized by rain forest. Journal of Biogeography,30,
Fensham R, Fairfax R, Archer S (2005) Rainfall, land use and woody vegetation cover
change in semi-arid Australian savanna. Journal of Ecology,93, 596606.
Gilroy JJ, Woodcock P, Edwards FA et al. (2014) Cheap carbon and biodiversity co-
benefits from forest regeneration in a hotspot of endemism. Nature Climate Change,
4, 503507.
Grace J, Jos
e JS, Meir P, Miranda HS, Montes RA (2006) Productivity and carbon
fluxes of tropical savannas. Journal of Biogeography,33, 387400.
Hansen MC, Potapov PV, Moore R et al. (2013) High-resolution global maps of 21st-
century forest cover change. Science,342, 850853.
Heckbert S, Russell-Smith J, Reeson A, Davies J, James G, Meyer C (2012) Spatially
explicit benefitcost analysis of fire management for greenhouse gas abatement.
Austral Ecology,37, 724732.
Henriques RP, Hay JD (2002) Patterns and dynamics of plant populations. In: The Cer-
rados of Brazil: Ecology and Natural History of a Neotropical Savanna (eds Oliveira P,
Marquiz R), pp. 140158. Columbia University Press, New York, NY.
Higgins SI, Bond WJ, February EC et al. (2007) Effects of four decades of fire manipu-
lation on woody vegetation structure in savanna. Ecology,88, 11191125.
Hirota M, Nobre C, Oyama MD, Bustamante MMC (2010) The climatic sensitivity of
the forest, savanna and forestsavanna transition in tropical South America. New
Phytologist,187, 707719.
Hirota M, Holmgren M, Van Nes EH, Scheffer M (2011) Global resilience of tropical
forest and savanna to critical transitions. Science,334, 232235.
Hoffmann WA, Adasme R, Haridasan M et al. (2009) Tree topkill, not mortality,
governs the dynamics of savanna-forest boundaries under frequent fire in central
Brazil. Ecology,90, 13261337.
Hoffmann WA, Geiger EL, Gotsch SG et al. (2012) Ecological thresholds at the
savanna-forest boundary: how plant traits, resources and fire govern the distribu-
tion of tropical biomes. Ecology Letters,15, 759768.
Hua F, Hu J, Liu Y et al. (2015) Community-wide changes in inter-taxonomic tempo-
ral co-occurrence resulting from phenological shifts. Global Change Biology,
Jackson RB, Banner JL, Jobb
agy EG, Pockman WT, Wall DH (2002) Ecosystem carbon
loss with woody plant invasion of grasslands. Nature,418, 623626.
Kauffman JB, Sanford RL Jr, Cummings DL, Salcedo IH, Sampaio EVSB (1993) Bio-
mass and nutrient dynamics associated with slash fires in neotropical dry forests.
Ecology,74, 140151.
Klink CA, Machado RB (2005) Conservation of the Brazilian cerrado. Conservation
Biology,19, 707713.
Lehmann CER, Anderson TM, Sankaran M et al. (2014) Savanna vegetation-fire-cli-
mate relationships differ among continents. Science,343, 548552.
List of Species of the Brazilian Flora. Rio de Janeiro Botanical Garden. Available at: (accessed 15 November 2015).
©2016 John Wiley & Sons Ltd, Global Change Biology,22, 3373–3382
Liu YY, Van Dijk AIJM, RaM DeJeu, Canadell JG, Mccabe MF, Evans JP, Wang G (2015)
Recent reversal in lossof global terrestrial biomass. Nature Climate Change,5,470474.
Midgley GF, Bond WJ (2015) Future of African terrestrial biodiversity and ecosystems
under anthropogenic climate change. Nature Climate Change,5, 823829.
Moreira AG (2000) Effects of fire protection on savanna structure in Central Brazil.
Journal of Biogeography,27, 10211029.
Morton DC, Le Page Y, Defries R, Collatz GJ, Hurtt GC (2013) Understorey fire fre-
quency and the fate of burned forests in southern Amazonia. Philosophical Transac-
tions of the Royal Society B: Biological Sciences,368, 20120163.
Myers N, Mittermeier RA, Mittermeier CG, Da Fonseca GA, Kent J (2000) Biodiversity
hotspots for conservation priorities. Nature,403, 853858.
Oliveira PS, Marquis RJ (2002) The Cerrados of Brazil: Ecology and Natural History of a
Neotropical Savanna. Columbia University Press, New York, NY.
Pan Y, Birdsey RA, Fang J et al. (2011) A large and persistent carbon sink in the
world’s forests. Science,333, 988993.
Parr CL, Lehmann CER, Bond WJ, Hoffmann WA, Andersen AN (2014) Tropical
grassy biomes: misunderstood, neglected, and under threat. Trends in Ecology &
Evolution,29, 205213.
Pellegrini AFA, Hoffmann WA, Franco AC (2014) Carbon accumulation and nitrogen
pool recovery during transitions from savanna to forest in central Brazil. Ecology,
95, 342352.
Pellegrini AFA, Hedin LO, Staver AC, Govender N (2015) Fire alters ecosystem car-
bon and nutrients but not plant nutrient stoichiometry or composition in tropical
savanna. Ecology,96, 12751285.
Pellegrini AFA, Franco AC, Hoffmann WA (2016) Shifts in functional traits elevate
risk of fire-driven tree dieback in tropical savanna and forest biomes. Global Change
Biology,22, 12351243.
Pinheiro M, Monteiro R (2010) Contribution to the discussions on the origin of the
cerrado biome: Brazilian savanna. Brazilian Journal of Biology,70,95102.
R Development Core Team (2010) R: A Language and Environment for Statistical Com-
puting. R Foundation for Statistical Computing, Vienna, Austria.
Ratter J, Ribeiro JF, Bridgewater S (1997) The Brazilian cerrado vegetation and threats
to its biodiversity. Annals of Botany,80, 223230.
Ratter J, Bridgewater S, Ribeiro J (2003) Analysis of the floristic composition of the
Brazilian cerrado vegetation III: comparison of the woody vegetation of 376 areas.
Edinburgh Journal of Botany,60,57109.
Ratter JA, Bridgewater S, Ribeiro JF (2006) Biodiversity patterns of the woody vegeta-
tion of the Brazilian Cerrado. Neotropical savannas and seasonally dry forests:
plant diversity, biogeography, and conservation. pp. 3166.
Ratter J, Bridgewater S, Ribeiro J et al. (2010) Analysis of the floristic composition of
the Brazilian cerrado vegetation IV: presentation of a Revised Data-Base of 367
Areas. Available at: http://www. cerrado. rbge. org. uk (accessed 15 November
Rodrigues-Souza J, Do Prado J
unior J, Do Vale V, Schiavini I, De Oliveira A, De
erio Arantes C (2015) Secondary forest expansion over a savanna domain at an
ecological reserve in the Southeastern Brazil after 15 years of monitoring. Brazilian
Journal of Botany,38, 311322.
Roitman I, Felfili JM, Rezende AV (2008) Tree dynamics of a fire-protected cerrado
sensu stricto surrounded by forest plantations, over a 13-year period (1991-2004)
in Bahia, Brazil. Plant Ecology,197, 255267.
Roques K, O’Connor T, Watkinson A (2001) Dynamics of shrub encroachment in an
African savanna: relative influences of fire, herbivory, rainfall and density depen-
dence. Journal of Applied Ecology,38, 268280.
Sasaki N, Putz FE (2009) Critical need for new definitions of “forest” and “forest
degradation” in global climate change agreements. Conservation Letters,2, 226232.
Scholes R, Archer S (1997) Tree-grass interactions in savannas. Annual Review of
Ecology and Systematics,28, 517544.
Scurlock JMO, Hall DO (1998) The global carbon sink: a grassland perspective. Global
Change Biology,4, 229233.
Silva LC, Hoffmann WA, Rossatto DR, Haridasan M, Franco AC, Horwath WR (2013)
Can savannas become forests? A coupled analysis of nutrient stocks and fire
thresholds in central Brazil. Plant and Soil,373, 829842.
Staver AC, Archibald S, Levin SA (2011) The global extent and determinants of
savanna and forest as alternative biome states. Science,334, 230232.
Tjørve E (2009) Shapes and functions of speciesarea curves (II): a review of new
models and parameterizations. Journal of Biogeography,36, 14351445.
Van Der Werf GR, Randerson JT, Giglio L et al. (2010) Global fire emissions and the
contribution of deforestation, savanna, forest, agricultural, and peat fires (1997
2009). Atmospheric Chemistry and Physics,10, 1170711735.
Veldman JW, Buisson E, Durigan G et al. (2015a) Toward an old-growth concept for
grasslands, savannas, and woodlands. Frontiers in Ecology and the Environment,13,
Veldman JW, Overbeck GE, Negreiros D et al. (2015b) Where tree planting and forest
expansion are bad for biodiversity and ecosystem services. BioScience,65, 1011
Venter O, Laurance WF, Iwamura T, Wilson KA, Fuller RA, Possingham HP (2009)
Harnessing carbon payments to protect biodiversity. Science,326, 1368.
Wagner F, Rutishauser E, Blanc L, Herault B (2010) Effects of plot size and
census interval on descriptors of forest structure and dynamics. Biotropica,42,
Wigley BJ, Bond WJ, Hoffman MT (2010) Thicket expansion in a South African
savanna under divergent land use: local vs. global drivers? Global Change Biology,
16, 964976.
Supporting Information
Additional Supporting Information may be found in the
online version of this article:
Table S1. List of sites, their classification, plant species rich-
ness and total biomass.
Table S2. Model comparison results.
Table S3. Best fit model coefficients.
Table S4. Results from repeated resampling analysis.
Figure S1. Map of study sites.
Figure S2. Species accumulation curves.
Figure S3. Species richness separated by plots categorized
into different biomes.
Figure S4. Shannon diversity index.
Figure S5. Pielou’s evenness metric.
Figure S6. Plot of results from repeated resampling analysis.
Figure S7. Distribution of selected protected areas and
potential species losses.
Figure S8. Swapping rarefaction approach.
Figure S9. Distribution of tree cover in select protected
©2016 John Wiley & Sons Ltd, Global Change Biology,22, 3373–3382
3382 A. F. A. PELLEGRINI et al.

Supplementary resources

  • ... (2008) in North American savannah where an increase in woody canopy cover was accompanied by high species richness. However, our results contrast with those of Pellegrini et al 2016 in South American savannah. We attributed these contrasting observations between our study and that of Pellegrini et al. (2016) to different climatic conditions prevailing in these savannahs. ...
    ... However, our results contrast with those of Pellegrini et al 2016 in South American savannah. We attributed these contrasting observations between our study and that of Pellegrini et al. (2016) to different climatic conditions prevailing in these savannahs. South American savannah is characterized by mesic conditions with mean annual precipitation of 2500mm which is 750mm above Africa's wettest savannahs ( Lehmann et al. 2011;Pellegrini et al. 2016). ...
    ... We attributed these contrasting observations between our study and that of Pellegrini et al. (2016) to different climatic conditions prevailing in these savannahs. South American savannah is characterized by mesic conditions with mean annual precipitation of 2500mm which is 750mm above Africa's wettest savannahs ( Lehmann et al. 2011;Pellegrini et al. 2016). Such mesic conditions combined with the absence of disturbance regime in South American savannah might be responsible for the transition to closed forest with increasing woody cover displacing savannah endemics ( Pellegrini et al. 2016). ...
    Full-text available
    Remote sensing applications in biodiversity research often rely on the establishment of relationships between spectral information from the image and tree species diversity measured in the field. Most studies have used normalized difference vegetation index (NDVI) to estimate tree species diversity on the basis that it is sensitive to primary productivity which defines spatial variation in plant diversity. The NDVI signal is influenced by photosynthetically active vegetation which, in the savannah, includes woody canopy foliage and grasses. The question is whether the relationship between NDVI and tree species diversity in the savanna depends on the woody cover percentage. This study explored the relationship between woody canopy cover (WCC) and tree species diversity in the savannah woodland of southern Africa and also investigated whether there is a significant interaction between seasonal NDVI and WCC in the factorial model when estimating tree species diversity. To fulfil our aim, we followed stratified random sampling approach and surveyed tree species in 68 plots of 90 m × 90 m across the study area. Within each plot, all trees with diameter at breast height of >10 cm were sampled and Shannon index − a common measure of species diversity which considers both species richness and abundance − was used to quantify tree species diversity. We then extracted WCC in each plot from existing fractional woody cover product produced from Synthetic Aperture Radar (SAR) data. Factorial regression model was used to determine the interaction effect between NDVI and WCC when estimating tree species diversity. Results from regression analysis showed that (i) WCC has a highly significant relationship with tree species diversity (r2 = 0.21; p < 0.01), (ii) the interaction between the NDVI and WCC is not significant, however, the factorial model significantly reduced the error of prediction (RMSE = 0.47, p < 0.05) compared to NDVI (RMSE = 0.49) or WCC (RMSE = 0.49) model during the senescence period. The result justifies our assertion that combining NDVI with WCC will be optimal for biodiversity estimation during the senescence period.
  • ... For instance, protecting a carbon-dense forest may reallocate human pressure to unprotected areas with lower carbon density, but high biodiversity (Di Marco et al., 2018). Also, shifting from natural veg- etation to tree plantations to maximize carbon stock leads to biodi- versity loss (Pichancourt et al., 2014), especially where natural grasslands or savannahs are afforested (Bremer & Farley, 2010;Burrascano et al., 2016;Pellegrini, Socolar, Elsen, & Giam, 2016). ...
    Policies to mitigate climate change and biodiversity loss often assume that protecting carbon-rich forests provides co-benefits in terms of biodiversity, due to the spatial congruence of carbon stocks and biodiversity at biogeographic scales. However, it remains unclear whether this holds at the scales relevant for management , and particularly large knowledge gaps exist for temperate forests and for taxa other than trees. We built a comprehensive dataset of Central European temperate forest structure and multi-taxonomic diversity (beetles, birds, bryo-phytes, fungi, lichens, and plants) across 352 plots. We used Boosted Regression Trees (BRTs) to assess the relationship between above-ground live carbon stocks and (a) taxon-specific richness, (b) a unified multidiversity index. We used Threshold Indicator Taxa ANalysis to explore individual species' responses to changing above-ground carbon stocks and to detect change-points in species composition along the carbon-stock gradient. Our results reveal an overall weak and highly variable relationship between richness and carbon stock at the stand scale, both for individual taxonomic groups and for multidiversity. Similarly, the proportion of win-win and trade-off species (i.e., species favored or disadvantaged by increasing carbon stock, respectively) varied substantially across taxa. Win-win species gradually replaced trade-off species with increasing carbon, without clear thresholds along the above-ground carbon gradient, suggesting that community-level surrogates (e.g., richness) might fail to detect critical changes in biodiversity. Collectively, our analyses highlight that leveraging co-benefits between carbon and biodiversity in temperate forest may require stand-scale management that prioritizes either biodiversity or carbon in order to maximize co-benefits at broader scales. Importantly, this contrasts with tropical forests, where climate and biodiversity objectives can be integrated at the stand scale, thus highlighting the need for context-specificity when managing for multiple objectives. Accounting for critical change-points of target taxa can help to deal
  • ... India is rich in phytodiversity, with about 45,000 plant species from the Western Ghats to Eastern Ghats along with the North-Eastern region and from the Greater Himalayas to the plain of Ganga. The forests of Odisha form a major part of the Eastern Ghats (Majumadar and Datta, 2015; Pellegrini et al., 2016). The state is also blessed with a biosphere reserve, Similipal Biosphere Reserve (SBR), which covers major part of Eastern Ghats (Figure 1). ...
    Full-text available
    A number of wild crops remain unexplored in this world and among them some have excellent medicinal and nutritional properties. India is a harbor of biodiversity in general and phytodiversity in particular. The plant diversity is distributed from the Western Ghats to Eastern Ghats, along with the North-Eastern region and from the Greater Himalayas to the plain of Ganga. Among these distributed floral regions of the country, the Eastern Ghats are important due to their rich floral diversity. The forests of Odisha form a major part of Eastern Ghats in general and the Similipal Biosphere Reserve (SBR) in particular. The SBR is inhabited by many local communities. The food and medicinal habits of these communities are not fully explored even today. They are dependent on the forests of SBR for their food and medicine. Among their collections from forests, root and tuberous plants play a significant role. The local communities of SBR use about 89 types of tuberous plants for various purposes. Dioscorea is one such tuber, having maximum use among the local of SBR. However, less documentation and no specific reports are available on the food and medicinal values of the species available in this part of the World. Dioscorea species, popularly known as Yam worldwide and as Ban Aalu in Odisha, India, is a prime staple medicinal-food substitute for the majority of rural and local people of the state of India. Of the 13 Dioscorea species available in SBR, 10 species are known to be bitter in taste and unpalatable when taken raw. Since less documentation is available on the Dioscorea species of SBR and their traditional uses, the present study was focused on the ethnobotany, nutritional and pharmacological values of these species along its nutraceutical importance.
  • ... Looking across a rainfall gradient and land-use types in South Africa, Stevens et al. [23] report large increases in woody cover in just a few decades providing support for a global driver [23], while also noting the interaction with megaherbivores (elephants). Woody encroachment may provide carbon benefits, but will undoubtedly come at a biodiversity cost [64]. TGBs are characterized by seasonally dry and hot climates [30]. ...
    Full-text available
    Tropical grassy biomes (TGBs) are changing rapidly the world over through a coalescence of high rates of land-use change, global change and altered disturbance regimes that maintain the ecosystem structure and function of these biomes. Our theme issue brings together the latest research examining the characterization, complex ecology, drivers of change, and human use and ecosystem services of TGBs. Recent advances in ecology and evolution have facilitated a new perspective on these biomes. However, there continues to be controversies over their classification and state dynamics that demonstrate critical data and knowledge gaps in our quantitative understanding of these geographically dispersed regions. We highlight an urgent need to improve ecological understanding in order to effectively predict the sensitivity and resilience of TGBs under future scenarios of global change. With human reliance on TGBs increasing and their propensity for change, ecological and evolutionary understanding of these biomes is central to the dual goals of sustaining their ecological integrity and the diverse services these landscapes provide to millions of people. This article is part of the themed issue ‘Tropical grassy biomes: linking ecology, human use and conservation’. © 2016 The Author(s) Published by the Royal Society. All rights reserved.
  • Thesis
    Die Brasilianische Savanne, auch bekannt als der Cerrado, bedeckt ca. 24% der Landoberfläche Brasiliens. Der Cerrado ist von einer einzigartigen Biodiversität und einem starken Gradienten in der Vegetationsstruktur gekennzeichnet. Großflächige Landnutzungsveränderungen haben dazu geführt, dass annähernd die Hälfte der Cerrado in bewirtschaftetes Land umgewandelt wurde. Die Kartierung ökologischer Prozesse ist nützlich, um naturschutzpolitische Entscheidungen auf räumlich explizite Informationen zu stützen, sowie um das Verständnis der Ökosystemdynamik zu verbessern. Neue Erdbeobachtungssensoren, frei verfügbare Daten, sowie Fortschritte in der Datenverarbeitung ermöglichen erstmalig die großflächige Erfassung saisonaler Vegetationsdynamiken mit hohem räumlichen Detail. In dieser Arbeit wird der Mehrwert von Landsat-basierten Landoberflächenphänologischen (LSP) Metriken, für die Charakterisierung der Cerrado-Vegetation, hinsichtlich ihrer strukturellen und phänologischen Diversität, sowie zur Schätzung des oberirdischen Kohlenstoffgehaltes (AGC), analysiert. Die Ergebnisse zeigen, dass LSP-Metriken die saisonale Vegetatiosdynamik erfassen und für die Kartierung von Vegetationsphysiognomien nützlich sind, wobei hier die Grenzen der Einteilung von Vegetationsgradienten in diskrete Klassen erreicht wurden. Basierend auf Ähnlichkeiten in LSP wurden LSP Archetypen definiert, welche die Erfassung und Darstellung der phänologischen Diversität im gesamten Cerrado ermöglichten und somit zur Optimierung aktueller Kartierungskonzepte beitragen können. LSP-Metriken ermöglichten die räumlich explizite Quantifizierung von AGC in drei Untersuchungsgebieten und sollten bei zukünftigen Kohlenstoffschätzungen berücksichtigt werden. Die Erkenntnisse dieser Dissertation zeigen die Vorteile und Nutzungsmöglichkeiten von LSP Metriken im Bereich der Ökosystemüberwachung und haben demnach direkte Implikationen für die Entwicklung und Bewertung nachhaltiger Landnutzungsstrategien.
  • Article
    Full-text available
    In the “Cerrado”–Amazon ecotone in central Brazil, recent studies suggest some encroachment of forest into savanna, but how, where, and why this might be occurring is unclear. To better understand this phenomenon, we assessed changes in the structure and dynamics of tree species in three vegetation types at the “Cerrado”–Amazon ecotone that are potentially susceptible to encroachment: open “cerrado” (OC), typical “cerrado” (TC) and dense woodland (DW). We estimated changes in density, basal area and aboveground biomass of trees with diameter ≥ 10 cm over four inventories carried out between 2008 and 2015 and classified the species according to their preferred habitat (savanna, generalist, or forest). There was an increase in all structural parameters assessed in all vegetation types, with recruitment and gains in basal area and biomass greater than mortality and losses. Thus, there were net gains between the first and final inventories in density (OC: 3.4–22.9%; TC: 1.8–12.6%; DW: 0.2–8.3%), in basal area (OC: 8.3–18.2%; TC: 2–12.7%; DW: 2.3–8.9%), and in biomass (OC: 10.6–16.4%; TC: 1–12%; DW: 5.2–18.7%). Furthermore, all vegetation types also experienced net gains in forest and generalist species relative to savanna species. A decline in recruitment of savanna species was a likely consequence of vegetation encroachment and environmental changes. Our results indicate, for the first time based on quantitative and standardized multi-site temporal data, that concerted structural changes caused by vegetation encroachment are occurring at the ecotone between the two largest biomes in Brazil.
  • Article
    Full-text available
    The application of remote sensing in biodiversity estimation has largely relied on the Normalized Difference Vegetation Index (NDVI). The NDVI exploits spectral information from red and near infrared bands of Landsat images and it does not consider canopy background conditions hence it is affected by soil brightness which lowers its sensitivity to vegetation. As such NDVI may be insufficient in explaining tree species diversity. Meanwhile, the Landsat program also collects essential spectral information in the shortwave infrared (SWIR) region which is related to plant properties. The study was intended to: (i) explore the utility of spectral information across Landsat-8 spectrum using the Principal Component Analysis (PCA) and estimate alpha diversity (α-diversity) in the savannah woodland in southern Africa, and (ii) define the species diversity index (Shannon (H′), Simpson (D2) and species richness (S) – defined as number of species in a community) that best relates to spectral variability on the Landsat-8 Operational Land Imager dataset. We designed 90 m × 90 m field plots (n = 71) and identified all trees with a diameter at breast height (DbH) above 10 cm. H′, D2 and S were used to quantify tree species diversity within each plot and the corresponding spectral information on all Landsat-8 bands were extracted from each field plot. A stepwise linear regression was applied to determine the relationship between species diversity indices (H′, D2 and S) and Principal Components (PCs), vegetation indices and Gray Level Co-occurrence Matrix (GLCM) texture layers with calibration (n = 46) and test (n = 23) datasets. The results of regression analysis showed that the Simple Ratio Index derivative had a higher relationship with H′, D2 and S (r²= 0.36; r²= 0.41; r²= 0.24 respectively) compared to NDVI, EVI, SAVI or their derivatives. Moreover the Landsat-8 derived PCs also had a higher relationship with H′ and D2 (r² of 0.36 and 0.35 respectively) than the frequently used NDVI, and this was attributed to the utilization of the entire spectral content of Landsat-8 data. Our results indicate that: (i) the measurement scales of vegetation indices impact their sensitivity to vegetation characteristics and their ability to explain tree species diversity; (ii) principal components enhance the utility of Landsat-8 spectral data for estimating tree species diversity and (iii) species diversity indices that consider both species richness and abundance (H′ and D2) relates better with Landsat-8 spectral variables. © 2017 International Society for Photogrammetry and Remote Sensing, Inc. (ISPRS)
  • Article
    Full-text available
    Tropical savannas have been increasingly viewed as an opportunity for carbon sequestration through fire suppression and afforestation, but insufficient attention has been given to the consequences for biodiversity. To evaluate the biodiversity costs of increasing carbon sequestration, we quantified changes in ecosystem carbon stocks and the associated changes in communities of plants and ants resulting from fire suppression in savannas of the Brazilian Cerrado, a global biodiversity hotspot. Fire suppression resulted in increased carbon stocks of 1.2 Mg ha⁻¹ year⁻¹ since 1986 but was associated with acute species loss. In sites fully encroached by forest, plant species richness declined by 27%, and ant richness declined by 35%. Richness of savanna specialists, the species most at risk of local extinction due to forest encroachment, declined by 67% for plants and 86% for ants. This loss highlights the important role of fire in maintaining biodiversity in tropical savannas, a role that is not reflected in current policies of fire suppression throughout the Brazilian Cerrado. In tropical grasslands and savannas throughout the tropics, carbon mitigation programs that promote forest cover cannot be assumed to provide net benefits for conservation.
  • Article
    Full-text available
    For decades, there has been enormous scientific interest in tropical savannahs and grasslands, fuelled by the recognition that they are a dynamic and potentially unstable biome, requiring periodic disturbance for their maintenance. However, that scientific interest has not translated into widespread appreciation of, and concern about threats to, their biodiversity. In terms of biodiversity, grassy biomes are considered poor cousins of the other dominant biome of the tropics—forests. Simple notions of grassy biomes being species-poor cannot be supported; for some key taxa, such as vascular plants, this may be valid, but for others it is not. Here, we use an analysis of existing data to demonstrate that high-rainfall tropical grassy biomes (TGBs) have vertebrate species richness comparable with that of forests, despite having lower plant diversity. The Neotropics stand out in terms of both overall vertebrate species richness and number of range-restricted vertebrate species in TGBs. Given high rates of land-cover conversion in Neotropical grassy biomes, they should be a high priority for conservation and greater inclusion in protected areas. Fire needs to be actively maintained in these systems, and in many cases re-introduced after decades of inappropriate fire exclusion. The relative intactness of TGBs in Africa and Australia make them the least vulnerable to biodiversity loss in the immediate future. We argue that, like forests, TGBs should be recognized as a critical—but increasingly threatened—store of global biodiversity. This article is part of the themed issue ‘Tropical grassy biomes: linking ecology, human use and conservation’. © 2016 The Author(s) Published by the Royal Society. All rights reserved.
  • Article
    Full-text available
    Global climate change is known to affect the assembly of ecological communities by altering species' spatial distribution patterns, but little is known about how climate change may affect community assembly by changing species' temporal co-occurrence patterns, which is highly likely given the widely observed phenological shifts associated with climate change. Here we analyzed a 29-year phenological data set comprising community-level information on the timing and span of temporal occurrence in 11 seasonally occurring animal taxon groups from 329 local meteorological observatories across China. We show that widespread shifts in phenology have resulted in community-wide changes in the temporal overlap between taxa that are dominated by extensions, and that these changes are largely due to taxa's altered span of temporal occurrence rather than the degree of synchrony in phenological shifts. Importantly, our findings also suggest that climate change may have led to less phenological mismatch than generally presumed, and that the context under which to discuss the ecological consequences of phenological shifts should be expanded beyond asynchronous shifts. This article is protected by copyright. All rights reserved.
  • Article
    The Cerrado is a fire-dependent savanna requiring a clear and urgent fire management policy. The extensive misuse of fire for deforestation or pasture management in Brazil has created an overall perception that its use is always deleterious. This view, reinforced by threats of global warming and climatic change, has lead to current policies of fire suppression. Cerrado ecosystems depend on the historical fire regime to maintain their structure, biodiversity and functioning. The suppression of fire has transformed savanna vegetation into forests, causing biodiversity losses and drastic changes in ecological processes. Policy implications. The National Fire Policy required by law must be urgently implemented in Brazil, including use of fire for Cerrado conservation in public and private lands on the basis of existing knowledge of indigenous people and scientists. Objective regulations on prescribed burning, land manager training, incentives for fire research and experimentation and a broad campaign to disseminate the benefits of fire for Cerrado conservation should be the cornerstones of the policy. If implemented, the policy can give the biodiversity of the Cerrado a future that has previously been severely threatened by fire suppression.
  • Article
    Full-text available
    Numerous predictions indicate rising CO2 will accelerate the expansion of forests into savannas. Although encroaching forests can sequester carbon over the short-term, increased fires and drought-fire interactions could offset carbon gains, which may be amplified by the shift towards forest plant communities more susceptible to fire-driven dieback. We quantify how bark thickness determines the ability of individual tree species to tolerate fire and subsequently determine the fire sensitivity of ecosystem carbon across 180 plots in savannas and forests throughout the 2.2-million km2 Cerrado region in Brazil. We find that not accounting for variation in bark thickness across tree species underestimated carbon losses in forests by ~50%, totaling 0.22PgC across the Cerrado region. The lower bark thicknesses of plant species in forests decreased fire tolerance to such an extent that a third of carbon gains during forest encroachment may be at risk of dieback if burned. These results illustrate that consideration of trait-based differences in fire tolerance is critical for determining the climate-carbon-fire feedback in tropical savanna and forest biomes.
  • Article
    Full-text available
    Vegetation change plays a critical role in the Earth's carbon (C) budget and its associated radiative forcing in response to anthropogenic and natural climate change. Existing global estimates of aboveground biomass carbon (ABC) based on field survey data provide brief snapshots that are mainly limited to forest ecosystems. Here we use an entirely new remote sensing approach to derive global ABC estimates for both forest and non-forest biomes during the past two decades from satellite passive microwave observations. We estimate a global average ABC of 362 PgC over the period 1998-2002, of which 65% is in forests and 17% in savannahs. Over the period 1993-2012, an estimated '0.07 PgC yr '1 ABC was lost globally, mostly resulting from the loss of tropical forests ('0.26 PgC yr '1) and net gains in mixed forests over boreal and temperate regions (+0.13 PgC yr '1) and tropical savannahs and shrublands (+0.05 PgC yr '1). Interannual ABC patterns are greatly influenced by the strong response of water-limited ecosystems to rainfall variability, particularly savannahs. From 2003 onwards, forest in Russia and China expanded and tropical deforestation declined. Increased ABC associated with wetter conditions in the savannahs of northern Australia and southern Africa reversed global ABC loss, leading to an overall gain, consistent with trends in the global carbon sink reported in recent studies.
  • Article
    Full-text available
    Misperceptions about the world's grassy biomes contribute to their alarming rates of loss due to conversion for agriculture and tree plantations, as well as to forest encroachment. To illustrate the causes and consequences of these misperceptions, we show that the World Resources Institute and the International Union for Conservation of Nature misidentified 9 million square kilometers of ancient grassy biomes as providing “opportunities” for forest restoration. Establishment of forests in these grasslands, savannas, and open-canopy woodlands would devastate biodiversity and ecosystem services. Such undesired outcomes are avoidable if the distinct ecologies and conservation needs of forest and grassy biomes become better integrated into science and policy. To start with, scientists should create maps that accurately depict grassy biomes at global and landscape scales. It is also crucial that international environmental agreements (e.g., the United Nations Framework Convention on Climate Change) formally recognize grassy biomes and their environmental values.
  • Article
    Projections of ecosystem and biodiversity change for Africa under climate change diverge widely. More than other continents, Africa has disturbance-driven ecosystems that diversified under low Neogene CO2 levels, in which flammable fire-dependent C4 grasses suppress trees, and mega-herbivore action alters vegetation significantly. An important consequence is metastability of vegetation state, with rapid vegetation switches occurring, some driven by anthropogenic CO2-stimulated release of trees from disturbance control. These have conflicting implications for biodiversity and carbon sequestration relevant for policymakers and land managers. Biodiversity and ecosystem change projections need to account for both disturbance control and direct climate control of vegetation structure and function.
  • Article
    Full-text available
    Studies report the forest expansion process toward open savanna areas; however, most of them were conducted by means of satellite images and aerial photographs. This study has investigated the forest expansion process through community dynamics over 15 years of permanent plots monitoring. The study was conducted at a forest continuum of three distinct phytophysiognomies (riparian forest, dry forest, and woodland savanna), sampling the trees with diameter at breast height ≥4.77 cm, distributed into 211 plots (10 m × 10 m). Density and basal area were compared using a paired t test, and Shannon–Wiener diversity was determined through Hutcheson’s t test. The number of dead and recruit trees, and basal area increment and decrement were compared among the phytophysiognomies using variance analysis, followed by Tukey’s test. Dynamics rates were calculated for the interval of 15 years, as well as Jaccard and Morisita-Horn’s similarity indices among phytophysiognomies. In woodland savanna, density and basal area increased and richness and diversity decreased, whereas in dry forest density decreased and richness and diversity increased. No changes in floristic parameters were observed for riparian forest. The similarity among phytophysiognomies increased over time due to advance of typical dry forests species toward woodland savanna and riparian forest; and local decrease in richness of typical savanna species in woodland savanna (decline of 13 species) and riparian species in the riparian forest (three species). From floristic analysis, our results support the dry forests expansion process toward woodland savanna and riparian forest, which seems to be strongly related to the control of fire and decrease in river flow, respectively.
  • Article
    Full-text available
    New burned area datasets and top-down constraints from atmospheric concentration measurements of pyrogenic gases have decreased the large uncertainty in fire emissions estimates. However, significant gaps remain in our understanding of the contribution of deforestation, savanna, forest, agricultural waste, and peat fires to total global fire emissions. Here we used a revised version of the Carnegie-Ames-Stanford-Approach (CASA) biogeochemical model and improved satellite-derived estimates of area burned, fire activity, and plant productivity to calculate fire emissions for the 1997–2009 period on a 0.5° spatial resolution with a monthly time step. For November 2000 onwards, estimates were based on burned area, active fire detections, and plant productivity from the Moderate Resolution Imaging Spectroradiometer (MODIS) sensor. For the partitioning we focused on the MODIS era. We used burned area estimates based on Tropical Rainfall Measuring Mission (TRMM) Visible and Infrared Scanner (VIRS) and Along-Track Scanning Radiometer (ATSR) active fire data prior to MODIS (1997–2000) and Advanced Very High Resolution Radiometer (AVHRR) derived estimates of plant productivity during the same period. Average global fire carbon emissions were 2.0 Pg yr<sup>−1</sup> with significant interannual variability during 1997–2001 (2.8 Pg yr<sup>−1</sup> in 1998 and 1.6 Pg yr<sup>−1</sup> in 2001). Emissions during 2002–2007 were relatively constant (around 2.1 Pg yr<sup>−1</sup>) before declining in 2008 (1.7 Pg yr<sup>−1</sup>) and 2009 (1.5 Pg yr<sup>−1</sup>) partly due to lower deforestation fire emissions in South America and tropical Asia. During 2002–2007, emissions were highly variable from year-to-year in many regions, including in boreal Asia, South America, and Indonesia, but these regional differences cancelled out at a global level. During the MODIS era (2001–2009), most fire carbon emissions were from fires in grasslands and savannas (44%) with smaller contributions from tropical deforestation and degradation fires (20%), woodland fires (mostly confined to the tropics, 16%), forest fires (mostly in the extratropics, 15%), agricultural waste burning (3%), and tropical peat fires (3%). The contribution from agricultural waste fires was likely a lower bound because our approach for measuring burned area could not detect all of these relatively small fires. For reduced trace gases such as CO and CH<sub>4</sub>, deforestation, degradation, and peat fires were more important contributors because of higher emissions of reduced trace gases per unit carbon combusted compared to savanna fires. Carbon emissions from tropical deforestation, degradation, and peatland fires were on average 0.5 Pg C yr<sup>−1</sup>. The carbon emissions from these fires may not be balanced by regrowth following fire. Our results provide the first global assessment of the contribution of different sources to total global fire emissions for the past decade, and supply the community with an improved 13-year fire emissions time series.