REGULAR ARTICLE

Can savannas become forests? A coupled analysis of nutrient

stocks and fire thresholds in central Brazil
Lucas C. R. Silva & William A. Hoffmann & Davi R. Rossatto &
Mundayatan Haridasan & Augusto C. Franco & William R. Horwath
Received: 18 March 2013 / Accepted: 17 June 2013 / Published online: 27 July 2013
#Springer Science+Business Media Dordrecht 2013
Abstract
Aims The effects of fire ensure that large areas of the
seasonal tropics are maintained as savannas. The ad-
vance of forests into these areas depends on shifts in
species composition and the presence of sufficient
nutrients. Predicting such transitions, however, is dif-
ficult due to a poor understanding of the nutrient
stocks required for different combinations of species
to resist and suppress fires.
Methods We compare the amounts of nutrients re-
quired by congeneric savanna and forest trees to reach
two thresholds of establishment and maintenance: that
of fire resistance, after which individual trees are large
enough to survive fires, and that of fire suppression,
after which the collective tree canopy is dense enough
to minimize understory growth, thereby arresting the
spread of fire. We further calculate the arboreal and
soil nutrient stocks of savannas, to determine if these
are sufficient to support the expansion of forests fol-
lowing initial establishment.
Results Forest species require a larger nutrient supply
to resist fires than savanna species, which are better
able to reach a fire-resistant size under nutrient limi-
tation. However, forest species require a lower nutrient
supply to attain closed canopies and suppress fires;
therefore, the ingression of forest trees into savannas
facilitates the transition to forest. Savannas have suf-
ficient N, K, and Mg, but require additional P and Ca
to build high-biomass forests and allow full forest
expansion following establishment.
Conclusions Tradeoffs between nutrient requirements
and adaptations to fire reinforce savanna and forest as
alternate stable states, explaining the long-term persis-
tence of vegetation mosaics in the seasonal tropics.
Low-fertility limits the advance of forests into savan-
nas, but the ingression of forest species favors the
formation of non-flammable states, increasing fertility
and promoting forest expansion.
Plant Soil (2013) 373:829842
DOI 10.1007/s11104-013-1822-x
Responsible Editor: Michael Denis Cramer.
Electronic supplementary material The online version of this
article (doi:10.1007/s11104-013-1822-x) contains
supplementary material, which is available to authorized users.
L. C. R. Silva (*)
:
W. R. Horwath
Department of Land, Air and Water Resources,
University of California,
Davis, CA 95616, USA
e-mail: lcsilva@ucdavis.edu
W. A. Hoffmann
Department of Plant Biology,
North Carolina State University,
Raleigh, NC 27695, USA
D. R. Rossatto
Departamento de Biologia Aplicada, FCAV, Universidade
Estadual Paulista Jlio de Mesquita Filho-UNESP,
14884-900 Jaboticabal, SP, Brazil
M. Haridasan
Departamento de Ecologia, Universidade de Braslia,
70910-900 Braslia, DF, Brazil
A. C. Franco
Departamento de Botnica, Universidade de Braslia,
70910-900 Braslia, DF, Brazil
Keywords Cerrado
.
Ecosystem dynamics
.
Fire
.
Forest expansion
.
Nutrient cycling
.
Soil-plant
interactions
.
Succession
.
Tradeoffs
.
Tropics
Introduction
Over the past few thousand years, expansion of forests
into savannas and grasslands has occurred in many
parts of South America, a phenomenon that has been
generally attributed to climate fluctuations (Behling et
al. 2005; Silva et al. 2008, 2011; Dumig et al. 2008;
Silva and Anand 2011). Savannas, however, continue
to persist over large areas where climatic conditions
are adequate to support forests (Lehmann et al. 2011).
In some of these areas, transitions to forest depend
upon shifts in tree species composition (Geiger et al.
2011; Ratnam et al. 2011; Silva and Anand 2011), but
predicting these is complicated by interacting effects
of fire and soil nutrients on forest and savanna species.
In dry tropical regions, climate alone prevents the
development of forests (Sankaran et al. 2005), but
throughout mesic savannas (i.e., regions where annual
precipitation >800 mm), gradual forest expansion may
occur upon fire suppression (Hopkins 1965; Durigan
and Ratter 2006; Pinheiro et al. 2010), and vegetation
mosaics commonly coincide with edaphic gradients
(Furley et al. 1992; Silva et al. 2008, 2010a, b). In
central Brazil, for example, savanna vegetation is typ-
ically found over nutrient-poor oxisols, dominating
the landscape, while disjunct forests are found in areas
of high soil fertility, such as riverbeds and limestone
outcrops (Guido et al. 2002; Gottsberger and
Silberbauer-Gottsberger 2006; Furley 2007). Despite
these associations, the extent to which soil nutrients
and fire limit forest development is not well resolved.
Nutrient-poor soils may support savanna because
available pools of essential nutrients are simply insuf-
ficient to construct high tree biomass or, alternatively,
nutrient pools may be sufficient to construct forest
biomass (Bond 2010), but savannas persists because
growth rates are too low to overcome rates of biomass
loss caused by fire (Lehmann et al. 2011; Hoffmann et
al. 2012a, b).
In either case, to determine how fire interacts with
nutrient availability to govern the distribution of forest
and savanna, it is important to consider two thresh-
olds. The first threshold marks the point at which an
individual tree becomes large enough to reliably avoid
fire damage and topkill (i.e., stem death), which we
refer to as the fire-resistance threshold. The second
threshold marks the point beyond which the vegeta-
tion achieves sufficient foliage to exclude shade-
intolerant understory vegetation (Hoffmann et al.
2012a). The absence of savanna understory vegetation
(mostly grasses) greatly reduces ecosystem flamma-
bility (Hoffmann et al. 2012b), so we refer to this as
the fire-suppression threshold. The ability of trees to
reach these thresholds is regulated by their growth
rate, which is at least in part dependent on resource
availability. Moreover, the accumulation of biomass
necessary to surpass these thresholds requires substan-
tial uptake of nutrients from soils and if the required
stocks are large relative to availabilities in the envi-
ronment, transitions to forest would be unlikely.
It has long been hypothesized that vegetation of
dystrophic soils are inherently more fire-prone than
vegetation of fertile soils, because of the slow rates
on which trees establish under limiting conditions
(Kellman 1984). To determine whether this hypothesis
holds, we must consider savanna and forest tree spe-
cies as separate functional types, evaluating whether
nutrient stocks needed to resist or suppress fires de-
pend on intrinsic species traits. Savanna species pro-
duce thick bark, which allows stems to become fire
resistant at a smaller size (Hoffmann et al. 2012a).
Forest species, on the other hand, produce greater leaf
area (Rossatto et al. 2009), and are more effective at
generating a closed fire-suppressing canopy. If these
contrasting traits translate into differences in nutrients
requirements for reaching the fire-resistance and fire-
suppression thresholds, interacting effects of species
composition and soil fertility would be the key to
predict forest expansion/contraction.
To test for such differences, here we: (i) compare
the nutritional requirement of forest and savanna trees
growing under similar (nutrient-poor) conditions at
fire-resistance and -suppression thresholds; and (ii)
estimate total arboreal (stem and crown) and soil nu-
trient stocks of savannas and four different types of
interspersed forests that co-exist in central Brazil, to
determine whether nutrient stocks in savanna ecosys-
tems are sufficient to support a transition to forest. We
address these objectives describing how nutrient
requirements and adaptations to fire disturbance inter-
act, elaborating on whether nutrient availability impo-
ses a definite constraint on forest distribution in mesic
savanna regions.
830 Plant Soil (2013) 373:829842
Material and methods
Study sites
The present study combines some of our previously
published and unpublished data from forests and sav-
annas of central Brazil. Our sites are located within 20
km from the urban limits of Brasilia, in the Brazilian
Federal District at altitudes of ca. 1,100 m above the sea
level. At all sites, the average annual temperature and
rainfall are about 22.5 C and 1,400 mm, with most of
the precipitation occurring during the (southern) sum-
mer. Despite similar climatic conditions, our study sites
encompass vegetation gradients fromopen grasslands to
dense forests. We focused on typical savannas and ad-
jacent forests, delimited by sharp (few meters wide)
boundaries, where structural parameters (Fig. 1) and
species composition have been well described.
Our datasets were obtained at fire-protected areas that
include three savanna and four forest sites. The selected
savannas represent the regionally dominant (cerrado
sensu stricto) vegetation, with 2030 % tree cover and
a continuous grass layer, established over deep dystro-
phic oxisols (Furley 1999). These savannas have per-
sisted for several thousands of years in close proximity
to forests (Silva et al. 2008, 2010a) with no apparent
limitation imposed by soil physical properties (e.g. tex-
ture, density, depth of water table, etc.). The selected
forest sites represent two riparian, one xeromorphic and
one deciduous forest. Riparian forests, adjacent to two
of the savanna sites, are located at the Ecological
Reserve of the Institute of Geography and Statistics
RECOR-IBGE (1556S and 4756W). These forests
will be referred to as dry and wet riparian forests, as
they occur along a well-drained and a seasonally
flooded riverbed respectively. The xeromorphic forest
(locally known as cerrado), adjacent to our third sa-
vanna site, is located at the EMBRAPA Research
Reserve (1536S and 4742 W). The deciduous forest
is located at the preservation area of CIPLAN Mining
Company (1533S and 4751W). Savannas are present
at this last site but are associated with unusually shallow
soils and abundant calcareous concretions and, for this
reason, were not included in our analysis.
Sampling design
To compare the nutritional requirements of savanna and
forest trees under the same environmental conditions,
we sampled 18 pairs of tree species, each consisting of
one forest and one savanna species of the same genus.
To classify each species into forest or savanna types, we
followed the same criteria used by Hoffmann et al.
Fig. 1 Leaf area index (Silva et al. 2008, 2010a), tree density,
number of tree species and total basal area (Fonseca and Silva
Junior 2004; Silva Jnior 2004, 2005; Ribeiro and Haridasan
1984; Haidar 2008) at each studied site. These parameters
represent ecosystem-wide estimates and therefore have no asso-
ciated errors
Plant Soil (2013) 373:829842 831
(2005) and Rossatto et al. (2013), based on morpholog-
ical attributes, herbarium comparisons and previously
published floristic surveys (Mendona et al. 2008).
Generally, the evolutionary history of each species in-
fluence trait variability (Hendry et al. 2011) and phylo-
genetic differences often appear correlated with resource
use strategies and biomass allocation (e.g. Burns and
Strauss 2012) in plant communities. Here, all 18 conge-
neric pairs (Supplementary information) were sampled
under nutrient-poor (savanna) conditions. Each of these
pairs belongs to a different botanical family, ensuring
phylogenetic independence when comparing species
attributes and nutrient stocks in relation to their ecosys-
tem of origin (forest or savanna).
Additionally, we sampled species that occur in both
forest and savanna environments, to compare nutrient
requirements of these species across ecosystems. Most
of these species were the same forest species sampled
in savannas for the congeneric comparison described
above, but we also sampled five additional species that
could be found in both ecosystem types, namely:
Agonandra brasiliensis, Copaifera langsdorfii, Qualea
grandiflora, Kielmeyera coriacea, and Myrcia tomen-
tosa; sampling a total of 17 species common to forests
and savannas (Supplementary information) for this
comparison.
These two sets of data permitted the analysis of
differences between savanna and forest trees growing
in a common environment (savanna), as well as bet-
ween different forests and savanna environments
using tree species common to both ecosystems.
Plant collection and analysis
During May and June of 2007, we collected leaf
and wood samples from three to five individual
trees per species at each site where a species was
present. All sampled trees had stem diameters great-
er than 10 cm. From every individual we collected
fully expanded mature leaves from the outer (sunlit)
portion of the crown and wood cores at approxi-
mately 1.3 m above the ground level. We deter-
mined the specific leaf area (SLA), wood density,
leaf and wood N, P, K, Ca, Mg concentrations, of
all sampled trees. Prior to analyses, plant samples
were washed with distilled water, dried at 80 C,
and milled to a particle size of <1.5 mm. N con-
centration was determined by the micro-Kjeldahl
method (Allen et al. 1974). We used a mixture of
sulphuric, nitric, and perchloric acids (1:10:2) to
digest plant samples, and determined P content
colorimetrically, whereas K, Ca, and Mg were de-
termined by atomic absorption spectrophotometry
(Allen et al. 1974). To calculate SLA, leaf disks
from at least five leaves per individual tree were
measured and dried to constant mass at 60 C. To
determine wood density, the ratio of dry mass to
fresh volume of radial sections was calculated for
each wood core sampled.
Comparisons between savanna and forest trees
We combined information on SLA, allometry, and
nutrient concentration in the leaf and wood tissue
to estimate total above-ground nutrient stocks of
savanna and forest tree species at their fire-
suppression (Fig. 2a) and fire-resistance (Fig. 2b)
thresholds.
Following a compilation of field data we deter-
mined that the fire-resistance threshold is achieved at
bark thickness of 5.9 mm (Hoffmann et al. 2012a).
This bark thickness ensures a 50 % chance of trees
avoiding topkill in a low-intensity fire characteristic of
Brazilian savannas. For a typical savanna tree species,
this corresponds to a stem diameter of 4.7 cm, while
for forest species this corresponds to a stem diameter
of 10.1 cm (Fig. 2b). Stem biomass at these diameters
was estimated using an allometric equation for forest
trees (Chambers 2001). Although allometric relation-
ships have been published for savanna trees, we use
the same relationship to yield conservative compari-
sons between functional types. Leaf areas at these
threshold sizes were estimated from the relationships
in Fig. 2a. Crown biomass, here represented by foliage
mass, was estimated by dividing leaf area by the SLA
of each tree. Total nutrient content of the above-
ground biomass (stem and crown) for savanna and
forest species was calculated at the individual level,
as the product of estimated biomass and measured
nutrient concentrations.
We used an analogous approach to compare the
above-ground nutrient stocks needed by savanna or
forest tree species to attain a closed-canopy state
corresponding to the fire-suppression threshold. We
previously showed that grasses are eliminated when
overstory leaf area index (LAI) exceeds 3.0, so we
used this value as a reference point for an ecosystem
that has reached a relatively non-flammable state
832 Plant Soil (2013) 373:829842
(Hoffmann et al. 2012a, b). We estimated the vege-
tation nutrient pools at this threshold using the arbi-
trary condition in which all trees have a stem diam-
eter of 10 cm. According to intrinsic differences in
stem vs. crown biomass allocations between forest
and savanna trees (Fig. 2a), the corresponding leaf
areas (5.57 m
2
for savanna and 16.61 m
2
for forest)
determine the estimated tree density (5,400 and
1,800 ha
1
) necessary to attain the fire-suppression
threshold (LAI of 3 or 30,000 m
2
ha
1
).
Comparisons of savanna and forest environments
For each of our study sites, we estimated total
above-ground nutrient stocks, scaling from individ-
ual trees to the ecosystem level, by combining
estimates of wood and canopy biomass. Due to
inherent variations in the structure of predominant
trees, different allometric equations are typically
used to quantify wood biomass of forest and savan-
na ecosystems. Forest-calibrated equations generally
underestimate savanna biomass, while savanna-
calibrated equations tend to overestimate forest bio-
mass. To address this issue and yield comparable
values, we used two different allometric equations,
one calibrated to savanna and another to forest trees
(Brown et al. 1989; Delitti et al. 2006), reporting
the average values of these calculations:
WBexp 3:114 0:972ln D
2
H

WB28:77D
2
H=1000
where WB (kg ha
1
) represents wood biomass, giv-
en the average stand height H (m) and diameter D
(cm), calculated from total basal area and tree den-
sity at each ecosystem (Fig. 1).
Ecosystem level canopy biomass and nutrient
stocks were estimated using measurements of total leaf
area index (LAI) at each site (Fig. 1). Because LAI is a
dimensionless ratio of leaf area covering a unit of
ground area (m
2
/m
2
), we used SLA measurements to
estimate total leaf biomass. Since SLA represents total
area per mass of leaf (cm
2
g
1
), the ratio between LAI
and SLA provides an integrated estimate of total leaf
mass per unit of ground area. Based on biomass esti-
mates and measurements of leaf nutrient concentration
we calculated whole canopy nutrient stocks, presen-
ting results as average values obtained using data from
all studied species.
Wood and canopy biomass are expressed as total
mass of carbon per hectare, accounting for variations
in wood density and SLA, and assuming carbon con-
tent to be 50 % of the total dry mass. We did not use
herbaceous strata in these calculations.
Soil collection and analysis
We used 15 composite soil samples (five subsamples
each), collected at the top 020 cm depth, to charac-
terize the surface soil fertility at each site (three savan-
nas and four forest sites). We also characterized deep
soils by pressing 100 cm
3
tubes into 10 cm layers of
sequentially dug soil pits, up to 100 cm depth, in soil
profiles evenly distributed across forest-savanna
Fig. 2 Allometric relationships between tree diameter leaf area
and bark thickness, measured in forest and savanna species
naturally occurring in savannas. a Fire-suppression threshold,
defined by total leaf area, determined using data compiled from
seedlings saplings, and adults (Gotsch et al. 2010) of forest and
savanna species. b Fire-resistance threshold, defined by
diameters at which bark thickness prevents tree mortality (Hoff-
mann et al. 2009, 2012a) at high and low fire intensity scenarios.
In both panels each point represents a single individual tree and
regression lines represent significant relationships (Pearsons
correlation; p<0.01)
Plant Soil (2013) 373:829842 833
transitions. Tubes were inserted so that pressure was
only exerted on the tube walls, thereby preventing soil
compaction and overestimation of soil bulk density.
Ten such profiles were sampled at each transition,
totaling 15 profiles in savannas and five at each forest
ecosystem.
We analyzed soil samples for density using the dry
mass of soil samples, and estimated total stocks of
carbon and nutrient in mass per hectare. We deter-
mined total N concentration in soils using the
Kjeldahl method (Bremner and Mulvaney 1982).
Levels of P, K, Ca and Mg were determined by atomic
absorption spectrophotometry using extraction meth-
ods (Mehlich and KCl) that allow the measurement of
concentration of nutrients available to plants. Total
organic carbon content was analyzed by wet oxidation
(Walkley and Black 1934). Carbon levels and nutrient
stocks are expressed on a mass basis, after accounting
for variation in soil density, in the surface soil layer
(020 cm depth) and integrated to 100 cm depth in soil
profiles.
The amount of a given nutrient existing (or miss-
ing) for building a forest was determined by the dif-
ference between savanna biomass and soil stocks up to
100 cm depth, and in the above-ground biomass of
each forest ecosystems. Nutrients contained in the
understory vegetation or root biomass were not inclu-
ded in our calculations, but their potential relevance is
discussed.
Results
Savanna: Leaf and wood nutrients as determined
by species origin
When growing in the same environment (savanna),
leaves of forest tree species have significantly higher
concentrations of N (23 %), P (25 %), and K (47 %),
but lower levels of Ca (18 %), than savanna species
(Table 1). Leaf levels of Mg do not differ significantly
according to species origin. Forest trees have signifi-
cantly greater SLA (27 %), but wood density does not
differ in relation to species origin. Similarly, most
wood nutrients (P, K, Ca and Mg) show no significant
differences between forest and savanna trees.
Significant differences only occur in wood N concen-
trations, which are higher in savanna than in forest
trees (Table 1).
Savanna: Nutrient stocks at fire-resistance
and -suppression thresholds
Our calculations show that, forest species require a
much greater nutrient supply than savanna species to
reach a size at which they can resist fires. Stem stocks
of N, P, K, Ca, and Mg are approximately 5-fold
greater, while crown (foliar) stocks are about one order
of magnitude higher in forest than in savanna species
at the fire-resistance threshold (Fig. 3). To reach the
canopy cover necessary to suppress fires, however,
stands composed entirely of savanna species require
greater nutrient stocks than those composed of forest
species. Due to intrinsically higher allocation to stem
biomass relative to leaves (Gotsch et al. 2010), wood
stocks of N, P, K, Ca, and Mg would be much higher
in savanna than in forest trees at the fire-suppression
threshold. At this threshold, crown stocks would be
similar between forests and savannas, with the excep-
tions of Ca levels, which are higher in savanna trees
(Fig. 3). In short, compared to forest species, savanna
trees require significantly lower nutrient stocks to
become fire resistant, but need significantly higher
nutrient stock to create a closed canopy that can sup-
press fires. These effects are quite large despite sub-
stantial interspecific variation.
Forests vs. Savanna: Leaf and wood nutrients
of individual trees
When the same species are analyzed across different
ecosystems, leaf N, P, K, Ca and Mg concentrations
vary significantly, with the lowest and highest leaf nu-
trient levels (and SLA) observed in savannas and ripar-
ian forests, respectively (Table 2). Nutrient concentra-
tions are less variable in wood than in leaves. Wood Ca
and Mg content, as well as wood density, do not vary
significantly among ecosystems. Wood N and P con-
centrations, on the other hand, are higher in savanna
than in riparian forest trees, while K concentration is
highest in the dry riparian forest and no differences are
observed among the other ecosystems (Table 2).
Forests vs. Savanna: Total biomass and nutrient stocks
at the ecosystem level
Above-ground biomass varies widely across eco-
systems. Total carbon stocks in the biomass of xero-
morphic and deciduous forests (>24 Mg ha
1
) is more
834 Plant Soil (2013) 373:829842
than 4 times greater than in savannas (<6 Mg ha
1
)
(Fig. 4). However, soils in these forests have approx-
imately half of the carbon stored in savanna soils (170
and 210 vs. 380 Mg ha
1
). Among all studied ecosys-
tems, the largest carbon stocks are observed in riparian
forests, where the high and low ends of the spectrum
are represented by wet and dry forests, with between
78 and 92 Mg C ha
1
stored above-ground and 600 to
1,600 Mg C ha
1
in soils (100 cm depth). In all
ecosystems, the amount of carbon stored in soils is
much higher than in the vegetation.
In forests and savannas, above-ground N stocks
match changes in biomass and increase progressively
across the spectrum: savanna < xeromorphic < deci-
duous < riparian forests. Total N and available K and
Mg stocks are always higher in soils than in the above-
ground biomass of savannas and forests (Fig. 4). Stocks
of P, on the other hand, are higher in the above-ground
biomass of all ecosystems, with the greatest difference
observed in the xeromorphic forest, where the amount
of P stored in the vegetation is about 7 times higher than
available levels measured in soils. Similarly, Ca stocks
are higher in the above-ground biomass than in soils of
savannas and most forest ecosystems. The only excep-
tion is the deciduous forest, where available soil pools
are over 10 times higher than total Ca measured in the
vegetation (Fig. 4).
Savanna soils have sufficient nutrients (Fig. 4)
to support a transition to a non-flammable state
(Fig. 3b), when comparing total stocks, we find
that savannas have sufficient N, K, and Mg, but
require additional P and Ca, to support a full
transition to forest (Table 3). The amount of addi-
tional P required to build a xeromorphic forest
over savannas is relatively small (3 kg ha
1
).
However, Ca deficits to build this same relatively
low-biomass forest are substantial (46 kg ha
1
),
representing two thirds of the existing soil stock
in savannas. To build a deciduous forest, this
deficit would increase about five times, while to
Table 1 Average values ( standard deviation) of specific leaf
area (SLA), wood density, and macronutrient concentrations, in
leaves and wood of savanna and forest trees co-occurring in
savanna environments. Between 3 and 5 individual trees of 36
species (18 congeneric pairs; Supplementary information) were
analyzed and p-values represent differences (t-test) between
means according to species origin
Savanna species Forest species pvalue
Leaf
N (mg g
1
) 11.690.62 14.450.73 <0.01
P (mg g
1
) 0.690.04 0.860.04 <0.01
K (mg g
1
) 4.600.37 6.770.58 <0.001
Ca (mg g
1
) 7.460.51 6.050.39 0.03
Mg (mg g
1
) 1.760.14 1.880.14 0.5
SLA (cm
2
g
1
) 54.641.83 69.481.88 <0.001
Wood
N (mg g
1
) 3.610.20 2.820.20 <0.01
P (mg g
1
) 0.550.05 0.550.02 0.9
K (mg g
1
) 1.800.17 1.740.14 0.7
Ca (mg g
1
) 6.240.69 5.750.49 0.5
Mg (mg g
1
) 0.570.07 0.570.07 0.9
Density (g cm
3
) 0.740.02 0.760.02 0.5
Fig. 3 Box plots of total crown (foliar) and stem nutrient stocks
measured averaged over 36 forest and savanna tree species
(Table 1) at the fire resistance and suppression thresholds (Fig.
2). The boxes correspond to the 25th and 75th percentiles, the
line inside the box represents the median and error bars show the
minimum and maximum values. The asterisk (*) shows signif-
icant differences (ANOVA, p<0.05) between forest and savanna
species
Plant Soil (2013) 373:829842 835
develop high-biomass riparian forests additional P
and Ca inputs (>40 and 600 kg ha
1
respectively),
would have to be several fold higher than what is
currently found in savannas.
Discussion
Savanna: Nutrient requirements at fire-resistance
and -suppression thresholds
Our results show that differences in nutritional
requirements affect the ability of forest and savanna
trees to resist and suppress fires. Confirming previous
studies (Hoffmann et al. 2005; Rossatto et al. 2013),
we found that forest species have higher nutrient
content per mass of leaf than their congeneric sa-
vanna counterparts, even when growing under
nutrient-limiting conditions (Table 1). Stem nutrient
concentrations are similar between forest and savan-
na trees, but savanna species allocate more biomass
to stem relative to leaves, requiring a lower nutrient
supply to reach a fire resistant size (Fig. 3).
However, stands composed entirely of savanna spe-
cies would require over three times more nutrient
capital than those composed of forest species to
produce sufficient leaf area to exclude grasses and
suppress fires. Forest species are able to generate a
closed canopy with lower nutrient capital than sa-
vanna species, because of a greater leaf area per
unit of stem biomass (Gotsch et al. 2010).
Consequently, the establishment of forest species
into savannas would favor canopy closure, facilita-
ting a transition to a non-flammable state. This
result corroborates observations that succession of
mesic savanna to forest is associated with the pre-
sence of forest species (Geiger et al. 2011; Ratnam
et al. 2011; Silva and Anand 2011), a pattern that
has been previously explained by the higher shade
tolerance of forest trees (Rossatto et al. 2009), but
here receives an alternative explanation.
While the ingression of forest trees may permit
savannas to attain a non-flammable state with lower
nutrient capital than would be required by savanna
trees, reaching this state would also require long inter-
vals without fire. Since fire is common in savannas,
transition to forest is typically interrupted while still in
early stages of succession, when tree cover is suscep-
tible to fire because of high mortality of small stems.
To avoid severe fire damage, trees must reach a fire-
resistant size, which is more easily attained by savanna
species not only because of their relatively lower nutri-
ent requirements, but also because they have thicker
bark than forest species and become fire-resistant at a
smaller stem size (Hoffmann et al. 2009, 2012a).
Differences in species nutrient requirements and
Table 2 Average values of specific leaf area (SLA), wood
density, and macronutrient levels in leaves and wood of tree
species (3 to 5 individuals of 17 species; Supplementary
information) sampled in savannas and at least one forest eco-
system. Letters represent significant differences (Tukeys HSD,
p<0.05) between ecosystems
Savanna Xeromorphic Deciduous Dry riparian Wet riparian
Forest Forest Forest Forest
Leaf
N (mg g
1
) 13.41 B 16.14 A 15.72 AB 13.10 B 14.35 AB
P (mg g
1
) 0.81 B 1.15 A 0.97 AB 0.83 B 1.19 A
K (mg g
1
) 6.61 C 11.87 A 10.13 AB 6.30 C 7.03 B
Ca (mg g
1
) 6.81 C 7.60 BC 19.57 A 7.61 BC 10.89 B
Mg (mg g
1
) 1.91 B 2.70 AB 2.56 B 3.66 A 2.74 AB
SLA (cm
2
g
1
) 69.97 C 96.57 AB 101.60 AB 90.65 B 111.13 A
Wood
N (mg g
1
) 2.66 AB 2.89 A 3.36 A 2.69 AB 1.92 B
P (mg g
1
) 0.53 A 0.45 B 0.51 A 0.43 B 0.45 B
K (mg g
1
) 1.62 B 1.61 B 1.33 B 2.30 A 0.88 B
Ca (mg g
1
) 6.38 A 5.15 A 6.01 A 5.21 A 4.48 A
Mg (mg g
1
) 2.29 A 0.73 A 0.45 A 0.96 A 0.51 A
Density (g cm
3
) 0.79 A 0.83 A 0.84 A 0.76 A 0.76 A
836 Plant Soil (2013) 373:829842
Fig. 4 Carbon and nutrient
stocks at each studied eco-
system. Above-ground esti-
mates were based on vege-
tation structure (Fig. 1) and
allometric equations and ac-
count for interspecific varia-
tion in nutrient concentra-
tion, wood density, and spe-
cific leaf area (Table 2). Soil
stocks account for variation
in density, carbon and nutri-
ent content, of superficial
(020 cm) and deep (20100
cm) soil layers. Bars repre-
sent average values and er-
ror bars show standard de-
viation of the average
Plant Soil (2013) 373:829842 837
biomass allocation probably reflect evolutionary adap-
tations to resource availability and disturbance regimes
characteristic of their original environments (Haridasan
2008; Hendry et al. 2011). Here, these differences define
savanna and forest trees as separate functional types,
supporting the notion that trait divergence decreases
phylogenetic signals (interspecific differences) when
competition for resources regulates community assem-
bly (Burns and Strauss 2012).
After becoming able to resist or suppress fires, the
ability of trees to modify soil properties may be the key
to explain the very slow, but ongoing, process of forest
expansion identified in some ecotonal regions of South
America (Silva et al 2008, 2011; Silva and Anand 2011).
In dystrophic savannas and grasslands, colonizing trees
alter patterns of nutrient cycling and accumulation, trig-
gering biogeochemical feedback loops that increase fer-
tility and lead to further changes in ecosystem structure
and composition (Dahlgren et al. 2003; Rice et al. 1996;
Saha et al. 2009; Silva et al. 2013). Under fire suppres-
sion, this effect should be maximized, as forest species
grow faster than their savanna counterparts (Hoffmann
et al. 2005; Rossatto et al. 2009, 2013). Furthermore,
forest species enhancing soil fertility by means of increas-
ing litter quality and quantity (Silva and Anand 2011).
Thus, despite large demand for nutrients, soil fertility
would likely increase following the establishment of forest
communities. In addition, the exclusion of the grass layer
would simultaneously reduce competition for limiting
nutrients and loss from burning. Savanna fires result in
significant losses of nutrients by volatilization or particle
transport, even though part of it returns as dry and wet
deposition (Kauffman et al. 1994).
Forests vs. Savanna: Nutrient stocks and limitations
to forest expansion
Based on what we have seen, differences in soil fertility
commonly found across forest-savanna transitions
(Furley et al. 1992; Silva et al. 2008, 2010a, b), could
be a result of biotic processes that create fertility gra-
dients and facilitate establishment of forest trees, rather
than determined solely by pre-existing edaphic condi-
tions. When the same species are compared across vege-
tation types, wood density and nutrient concentrations
are fairly consistent, but leaf nutrient concentration and
SLA are always higher in forest sites (Table 2). While
differences in leaf nutrient concentration contribute to
the higher nutrient stocks in forest habitats, this effect is
not large because stem biomass accounts for most (>70
%) of the above-ground stocks of all nutrients (Fig. 4).
At the ecosystem level, the larger stock of nutrients in
forests compared to savannas is, therefore, mostly due to
wood biomass and we found no relationships between
Table 3 Comparison of savanna nutrient stocks with the
amount needed to build each of the forest types. Top: Measured
stocks in soils (0100 cm depth) and above-ground biomass of
savannas, and above-ground biomass of four forest ecosystems.
Bottom: The difference between total savanna pools and forest
aboveground pools, which represents the excess (or deficit) of
nutrients after building each forest type on savanna sites
Savanna Forest biomass
Soils Biomass Xeromorphic Deciduous Dry riparian Wet riparian
Total stock (kg.ha
1
)
C 383320 5420 24506 33049 78387 94196
N 15696 62 172 248 463 400
P 9.3 10 23 35 68 86
K 165.4 36 103 106 269 186
Ca 77.8 121 245 426 826 863
Mg 72.7 43 39 34 161 102
Stock difference (kg.ha
1
) Savanna (soil + biomass) Excess after building forests (savanna-forest)
N 15758 15586 15510 15295 15358
P 20 3 15 48 67
K 201 98 95 68 15
Ca 199 46 227 627 664
Mg 116 76 81 46 13
838 Plant Soil (2013) 373:829842
soil fertility and leaf or wood nutrient concentrations.
These results emphasize the role of intrinsic species
traits and vegetation structure in generating and rein-
forcing gradients of resource accumulation, as a result of
a slow redistribution from deep to surface soils through
organic matter deposition.
Our calculations show that savanna soils have
enough nutrients to allow tree communities to attain
fire-resistant and -suppression thresholds (Figs. 3 and
4). However, savannas require additional nutrients to
allow a full forest expansion following initial estab-
lishment. Comparing soils and vegetation stocks, we
find that total N in savanna soils is orders of magni-
tude larger than the amount needed to develop a forest,
supporting the notion that nutrients other than N limit
the distribution of tropical forests (Reich and Oleksyn
2004; Cleveland et al. 2011). Even though this assess-
ment is based on total soil N rather than available
forms, given the high potential for organic N mineral-
ization and symbiotic fixation of tropical soils
(Cleveland et al. 2011), we conclude that soil N would
not impose an absolute constraint on forest develop-
ment at our study sites. Savannas also have sufficient
K and Mg to develop low-biomass xeromorphic for-
ests. However, additional P and Ca inputs would be
required to support a transition to any forested state.
The amount of additional P required to build a
xeromorphic forest is relatively small and could
perhaps be attained if understory vegetation stocks
were also considered. Previous studies (Castro and
Kauffman 1998) have shown that the above-ground
biomass in open grasslands near our study sites is
about 5.5 Mg ha
1
(ca. 2.8 Mg C ha
1
), which repre-
sents roughly half of the arboreal biomass of savannas.
The concentration of P in native and exotic grasses
that occur in the region, range from 0.04 to 0.07 %
(Silva and Haridasan 2007), which falls within the
range found in savanna trees (Table 1). At these con-
centrations, grasses could provide up to 3.9 kg P ha
1
,
which would only be sufficient to build a xeromorphic
forest (Table 3). The average concentration of Ca in
grasses (0.2 %; Silva and Haridasan 2007) is, howev-
er, much lower than that found in the wood of savanna
trees (>0.6 %; Table 1) and is not sufficient to cover
the deficit of Ca to build a low-biomass forest. Further
Ca inputs, of over one third of the existing stock in the
arboreal biomass of savannas, would be necessary
from atmospheric and erosional deposition, or soil
depths greater than 100 cm, to build a xeromorphic
forest. These inputs would, have to be, however, over
one order of magnitude larger to attain high-biomass
states, such as riparian forests.
Stocks of Ca should pose a strong constraint par-
ticularly because mineral sources of this element are
present in negligible amounts in the acidic soils that
predominate in the region. Based on our calculations,
only a landscape-level process of Ca redistribution
could explain the ongoing expansion of riparian for-
ests over savannas (Silva et al. 2008). Among other
essential nutrients, Ca occupies a unique position in
regulating growth from individual trees to entire eco-
systems, controlling cell wall synthesis, biomass pro-
duction and structure of woody tissues (Lautner and
Fromm 2010). Savannas and most forest ecosystems
studied here have larger Ca stocks in the above-ground
biomass than in soil pools, indicating that any inter-
ruption on Ca recycling would favor the formation of
less productive (i.e., open) stable states. Stocks of P
show similar distribution and, in contrast with less
limiting nutrients, Ca and P are both found in higher
concentrations at the surface (020 cm) than in deeper
soil layers (Fig. 4), suggesting a continuous and rapid
recycling process following litter deposition. The decid-
uous forest is the only ecosystem where we observed
greater nutrient stocks in soils than in the above-ground
biomass. Its unusual soil fertility originates from the
weathering of shallow parent materials (entisols) and
calcareous concretions often found within less than 0.5
m from the soil surface. Such soils are rare in central
Brazil, where its occurrence is typically characterized by
the presence of deciduous trees (Moura et al. 2010;
Haidar 2008), which under high fertility conditions are
able to outcompete evergreen species that dominate
other forest ecosystems.
Final considerations
We show that interactions between species composi-
tion, disturbance, and resource availability favor the
long-term persistence of forest-savanna mosaics. At
our study sites, and probably elsewhere, savanna and
forest trees represent different functional types.
Tradeoffs between nutrient requirements and adapta-
tions to resist fires reinforce savanna and forest as
alternate stable states, with low fertility hindering forest
expansion. Forest species require a larger amount of
nutrients to become fire-resistant, so the combined
Plant Soil (2013) 373:829842 839
effect of fire and nutrient limitation should favor the
persistence of savanna communities.When only savan-
na species are present, the transition to a forested state
requires a larger nutrient supply than is available in the
soil. Canopy closure in this case depends upon the
ingression of forest species, which can facilitate forest
expansion by increasing soil fertility.
The above-ground biomass of savannas and forests at
our sites is lower than measured in other tropical ecosys-
tems. Above-ground carbon stocks of forests in the
Amazon region, for example, range from 155 to 217
Mg C ha
1
(Kauffman et al. 1995), whereas the largest
value we observed in any of our forest sites was 95 Mg C
ha
1
(ca. 20 times more carbon than stored in savannas).
Across all ecosystems, soil carbon stocks are larger than
in the above-ground biomass (Fig. 4), illustrating the
potential for long-term carbon storage in the absence of
fire. Developing non-flammable forested states is depen-
dent upon adequate nutrient availability and even though
savannas may occur under high fertility conditions (e.g.
Moura et al. 2010; Haidar 2008; Silva et al. 2010b;), our
results show that the indigenous fertility typical of
Brazilian savannas will not allow widespread forest ex-
pansion unless external inputs of P and Ca occur.
Roots were not included in our calculations, but we
know that root biomass increases over three-fold (ca.
8 to 26 Mg C ha
1
) from open grasslands to tree-
dominated savannas near our study sites (Castro and
Kauffman 1998). The exclusion of root nutrient pools
in our study, thus, probably underestimates nutritional
limitations to develop forests. The next phase of un-
derstanding will come from manipulations of limiting
resources in fertilization experiments using traceable
nutrients, coupled with quantitative assessments of
changes in vegetation structure and composition.
Differences in plant-soil water relations should also
be assessed under multiple nutritional constraints.
Variations in tree structure and phenology (Rossatto
et al. 2009, 2012), physiological adjustments to
changes in water availability, evaporative demand
(Bucci et al. 2008), and alternative water sources
(e.g., shoot uptake; Oliveira et al. 2005), have been
indicated as important factors controlling vegetation
distribution in mesic savannas, but their interactions
with fire disturbance and nutrient limitation remain to
be investigated.
Acknowledgments We thank the staff of RECOR-IBGE,
CIPLAN, and Embrapa Cerrados, for the research infrastructure
and logistic support. We also thank Ricardo Haidar, Gabriel
Damasco, Daniel Marra, Gabriel Ribeiro, and Artur Paiva, for
help with field work and species identification, and Timothy
Doane and three anonimous reviewers for valuable comments
on the manuscript. This research is based upon work supported
by the National Science Foundation Grant No. DEB-0542912
(W. H.), AW Mellon Foundation (W. H.), National Science
Foundation Grant No. EAR-BE-332051 (L. S.,M. H., F. M.-
W., A. F.), and the J. G. Boswell Endowed Chair in Soil Science.
References
Allen S, Grimshaw H, Parkinson J, Quarmby C (1974)
Chemical analysis of ecological materials. Blackwell
Scientific, Oxford
Behling H, Pillar V, Bauermann S (2005) Late Quaternary
grassland (Campos), gallery forest, fire and climate
dynamics, studied by pollen, charcoal and multivariate
analysis of the Sao Francisco de Assis core in western
Rio Grande do Sul (southern Brazil). Rev Palaeobot
Palynol 133:235248
Bond W (2010) Do nutrient-poor soils inhibit development of
forests? A nutrient stock analysis. Plant Soil 334:334347
Bremner JM, Mulvaney CS (1982) Nitrogen total. In: Page AL,
Miller RH, Keeney DR (eds) Methods of soil analysis:
Chemical and microbiological properties, Vol. 2, 2nd edn.
American Society of Agronomy, Madison, pp 595624
Brown S, Gillespie AJR, Lugo AE (1989) Biomass estimation
methods for tropical forests with applications to forest
inventory data. For Sci 35:881902
Bucci SJ, Scholz FG, Goldstein G, Meinzer FC, Franco AC,
Zhang Y, Hao G-Y (2008) Water relations and hydraulic
architecture in Cerrado trees: adjustments to seasonal
changes in water availability and evaporative demand.
Braz J Plant Physiol 20:233245
Burns JH, Strauss SY (2012) Effects of competition on phylo-
genetic signal and phenotypic plasticity in plant functional
traits. Ecology 93:126137
Castro EA, Kauffman JB (1998) Ecosystem structure in the
Brazilian Cerrado: a vegetation gradient of aboveground
biomass, root mass and consumption by fire. J Trop Ecol
14:263283
Chambers JQ (2001) Tree damage, allometric relationships, and
above-ground net primary production in central Amazon
forest. Forest Ecol Manage 152:7384
Cleveland CC, Townsend AR, Taylor PS et al (2011)
Relationships among net primary productivity, nutrients
and climate in tropical rain forest: a pan-tropical analysis.
Ecol Lett 14:939947
Dahlgren RA, Horwath WR, Tate KW, Camping TJ (2003) Blue
oak enhance soil quality in California oak woodlands. Calif
Agric 57:4247
Delitti WBC, Meguro M, Pausas JG (2006) Biomass and mine-
ral mass estimates in a cerrado ecosystem. Rev Bras Bot
29:531540
Dumig A, Schad P, Rumpel C, Dignac M, Kogelknabner I
(2008) Araucaria forest expansion on grassland in the
southern Brazilian highlands as revealed by
14
C and
13
C
studies. Geoderma 145:143157
840 Plant Soil (2013) 373:829842
Durigan G, Ratter JA (2006) Successional changes in cerrado
and cerrado/forest ecotonal vegetation in Western So
Paulo State, Brazil, 19622000. Edinburgh Journal of
Botany 63:119. Cambridge University Press
Fonseca MS, Silva Junior MC (2004) Fitossociologia e similar-
idade florstica entre trechos de Cerrado sentido restrito em
interflvio e em vale no Jardim Botanico de Braslia, DF.
Acta Bot Bras 18(1):1929
Furley PA (1999) The nature and diversity of neotropical savan-
na vegetation with particular reference to the Brazilian
cerrados. Glob Ecol Biogeogr 8:223241
Furley PA (2007) Tropical savannas and associated forests:
vegetation and plant ecology. Prog Phys Geogr 31:203211
Furley PA, Proctor J, Ratter JA (1992) Nature and dynamics of
forest-savanna boundaries. Chapman & Hall, London, p
616
Geiger EL, Gotsch SG, Damasco G, Haridasan M, Franco AC,
Hoffmann WA (2011) Distinct roles of savanna and forest
tree species in regeneration under fire suppression in a
Brazilian savanna. J Veg Sci 22:312321
Gotsch SG, Geiger EL, Franco AC, Goldstein G, Meinzer FC,
Hoffmann WA (2010) Allocation to leaf area and sapwood
area affects water relations of co-occurring savanna and
forest trees. Oecologia 163:291301
Gottsberger G, Silberbauer-Gottsberger I (2006) Other vegeta-
tion types in the Cerrado region and transitional forms. In:
Gottsberger G, Silberbauer-Gottsberger I (eds) Life in the
cerrado: A South American tropical seasonal ecosystem.
Vol I Origin, structure, dynamics and plant use. Reta
Verlag, Ulm, pp 6275
Guido P, Ruggiero C, Batalha MA, Pivello VR (2002) Soil-
vegetation relationships in cerrado ( Brazilian savanna )
and semideciduous forest, Southeastern Brazil. Plant Ecol
160:116
Haidar RF (2008) Fitiossociologia, diversidade e sua relao
com variveis ambientais em florestas estacionais do
bioma Cerrado no planalto central e nordeste do Brasil.
Dissertation, Universidade de Brasilia
Haridasan M (2008) Nutritional adaptations of native plants of
the cerrado biome in acid soils. Braz J Plant Physiol
20:183195
Hendry AP, Kinnison MT, Heino M, Day T, Smith TB, Fitt G,
Bergstrom CT, Oakeshott J, Jrgensen PS, Zalucki MP,
Gilchrist G, Southerton S, Sih A, Strauss S, Denison RF,
Carroll SP (2011) Evolutionary principles and their practi-
cal application. Evol Appl 4:159183
Hoffmann WA, Franco AC, Moreira MZ, Haridasan M (2005)
Specific leaf area explains differences in leaf traits between
congeneric savanna and forest trees. Funct Ecol 19:932
940
Hoffmann WA, Adasme R, Haridasan M, Carvalho MT, Geiger
EL, Gotsch SG, Franco AC(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, Rossatto DR, Silva
LCR, Lau OL, Haridasan M, Franco AC (2012a)
Ecological thresholds at the savanna-forest boundary:
how plant traits, resources and fire govern the distribution
of tropical biomes. Ecol Lett 15:759768
Hoffmann WA, Jaconis SY, Mckinley KL, Geiger EL, Gotsch
SG, Franco AC (2012b) Fuels or microclimate?
Understanding the drivers of fire feedbacks at savanna-
forest boundaries. Aust Ecol 37:634643
Hopkins B (1965) Vegetation of the Olokemeji Forest Reserve,
Nigeria: II. The climate with special reference to its sea-
sonal changes. J Ecol 50:559598
Kauffman JB, Cummings DL, Ward DE (1994) Relationships of
fire, biomass and nutrient dynamics along a vegetation
gradient in the Brazilian Cerrado. J Ecol 82:519531
Kauffman JB, Cummings DL, Ward DE, Babbitt R (1995) Fire
in the Brazilian Amazon: biomass, nutrient pools, and
losses in slashed primary forests. Oecologia 104:397408
Kellman M (1984) Synergistic relationships between fire and
low soil fertility in Neotropical Savannas: A hypothesis.
Oecologia 16:158160
Lautner S, Fromm J (2010) Calcium-dependent physiological
processes in trees. Plant Biol Stuttg Germany 12:268
274
Lehmann CER, Archibald AS, Hoffmann WA, Bond WJ (2011)
Deciphering the distribution of the savanna biome. New
Phytol 191:197209
Mendona RC, Felfili JM, Walter BMT, Silva-Jnior MC,
Rezende AB, Filgueiras TS, Nogueira PE, Fagg CW
(2008) Flora Vascular do Bioma Cerrado: Checklist com
12.356 espcies. In: Sano SM, Almeida SP, Ribeiro JF
(eds) Cerrado: Ecologia e Flora, vol 2. Braslia, Embrapa
Cerrados, pp 213228
Moura IOD, Gomes-Klein VL, Felfili JM, Ferreira HD (2010)
Diversity and community structure of cerrado sensu stricto
on rocky outcrops of the Pireneus State Park, Goias. Rev
Bras Bot 33:455467
Oliveira RS, Dawson TE, Burgess SSO (2005) Evidence for
direct water absorption by the shoot of the desiccation-
tolerant plant Vellozia flavicans in the savannas of central
Brazil. J Trop Ecol 21:585588
Pinheiro MHO, De Azevedo TS, Monteiro R (2010) Spatial-
temporal distribution of fire-protected savanna physiogno-
mies in Southeastern Brazil. An Acad Bras Cincias
82:379395
Ratnam J, Bond WJ, Fensham RJ, Hoffmann WA, Archibald S,
Lehmann CER, Anderson MT, Higgins SI, Sankaran M
(2011) When is a forest a savanna, and why does it
matter? Glob Ecol Biogeogr 20:653660
Reich PB, Oleksyn J (2004) Global patterns of plant leaf N and
P in relation to temperature and latitude. PNAS
101:1100111006
Ribeiro JF, Haridasan M (1984) Comparao fitossociolgica de
um cerrado denso e um cerrado em solos distrficos no
Distrito Federal. In: Congresso nacional de botanica 35,
Manaus, AM, Brazil
Rice C, Moorman T, Beare M (1996) Role of microbial biomass
carbon and nitrogen in soil quality. SSSA Spec Publ
49:203216
Rossatto DR, Hoffmann WA, Franco AC (2009) Differences in
growth patterns between co-occurring forest and savanna
trees affect the forest-savanna boundary. Funct Ecol
23:689698
Rossatto DR, Silva LCR, Villalobos-Vega R, Sternberg LDSL,
Franco AC (2012) Depth of water uptake in woody plants
relates to groundwater level and vegetation structure along
a topographic gradient in a neotropical savanna. Environ
Exp Bot 77:259266
Plant Soil (2013) 373:829842 841
Rossatto DR, Hoffmann WA, Silva LCR, Haridasan M,
Sternberg LSL, and Franco AC (2013) Seasonal variation
in leaf traits between congeneric savanna and forest trees in
Central Brazil: implications for forest expansion into sa-
vanna. Trees. doi:10.1007/s00468-013-0864-2
Saha A, Sternberg LSL, Miralles-Wilhelm F (2009) Linking
water sources with foliar nutrient status in the Everglades
National Park, USA. Ecohydrology 2:4254
Sankaran M, Hanan NP, Scholes RJ, Ratnam J, Augustine DJ,
Cade BS, Gignoux J, Higgins SI, Le Roux X, Ludwig F,
Ardo J, Banyikwa F, Bronn A, Bucini G, Caylor KK,
Coughenour MB, Diouf A, Ekaya W, Feral CJ, February
EC, Frost PGH, Hiernaux P, Hrabar H, Metzger KL, Prins
HHT, Ringrose S, Sea W, Tews J, Worden J, Zambatis N
(2005) Determinants of woody cover in African savannas.
Nature 438:846849
Silva Jnior MC (2004) Fitossociologia e estrutura diametrica
da mata de galleria do Taquara, na Reserva Ecologica do
IBGE, DF. Rev rvore 28:419428
Silva Jnior MC (2005) Fitossociologia e estrutura diamtrica
na mata de galleria do Pitoco, na Reserva Ecologica do
IBGE, DF. Cerne 11:147158
Silva LCR, Anand M (2011) Mechanisms of Araucaria
(Atlantic) forest expansion into Southern Brazilian
Grasslands. Ecosystems 14:13541371
Silva JSO, Haridasan M (2007) Acmulo de biomassa area e
concentrao de nutrientes em Melinis minutiflora P. Beauv.
e gramneas nativas do cerrado. Rev Bras Bot 30:337344
Silva LCR, Sternberg LSL, Haridasan M, Hoffmann WA,
Miralles-Wilhelm F, Franco AC (2008) Expansion of gal-
lery forests into central Brazilian savannas. Glob Chang
Biol 14:21082118
Silva LCR, Haridasan M, Sternberg LSL, Franco AC, Hoffmann
WA (2010a) Not all forests are expanding over central
Brazilian savannas. Plant Soil 333:431442
Silva LCR, Vale GD, Haidar RF, Sternberg LSL (2010b)
Deciphering earth mound origins in central Brazil. Plant
Soil 336:314
Silva LCR, Giorgis MA, Anand M, Enrico L, Prez-Harguindeguy
N, Falczuk V, Tieszen LL, Cabido M(2011) Evidence of shift
in C4 species range in central Argentina during the late
Holocene. Plant Soil 349:261279
Silva LCR, Corra RS, Doane T, Pereira E, Horwath WR (2013)
Unprecedented carbon accumulation in mined soils: the
synergistic effect of resource input and plant species inva-
sion. Ecol Appl. doi:10.1890/12-1957.1
Walkley A, Black IA (1934) An examination of the Degtjareff
method for determining soil organic matter and a proposed
modification of the chromic acid tritation method. Soil Sci
37:2938
842 Plant Soil (2013) 373:829842