ArticlePDF Available

Abstract and Figures

Global change has resulted in chronic shifts in fire regimes. Variability in the sensitivity of tree communities to multi-decadal changes in fire regimes is critical to anticipating shifts in ecosystem structure and function, yet remains poorly understood. Here, we address the overall effects of fire on tree communities and the factors controlling their sensitivity in 29 sites that experienced multi-decadal alterations in fire frequencies in savanna and forest ecosystems across tropical and temperate regions. Fire had a strong overall effect on tree communities, with an average fire frequency (one fire every three years) reducing stem density by 48% and basal area by 53% after 50 years, relative to unburned plots. The largest changes occurred in savanna ecosystems and in sites with strong wet seasons or strong dry seasons, pointing to fire characteristics and species composition as important. Analyses of functional traits highlighted the impact of fire-driven changes in soil nutrients because frequent burning favoured trees with low biomass nitrogen and phosphorus content, and with more efficient nitrogen acquisition through ectomycorrhizal symbioses. Taken together, the response of trees to altered fire frequencies depends both on climatic and vegetation determinants of fire behaviour and tree growth, and the coupling between fire-driven nutrient losses and plant traits.
This content is subject to copyright. Terms and conditions apply.
1Department of Earth System Science, Stanford University, Stanford, CA, USA. 2Department of Plant Sciences, University of Cambridge, Cambridge, UK.
3Department of Natural Resources & Environmental Science, University of Nevada, Reno, NV, USA. 4Department of Environmental Systems Science,
Institute of Integrative Biology, ETH Zürich, Zürich, Switzerland. 5Physical and Life Sciences Directorate, Lawrence Livermore National Laboratory,
Livermore, CA, USA. 6Department of Ecology and Evolutionary Biology, Yale University, New Haven, CT, USA. 7Southern Research Station, USDA Forest
Service, Auburn, AL, USA. 8National Parks Service, Sequoia & Kings Canyon National Parks, Three Rivers, CA, USA. 9Department of Forestry, Wildlife, and
Fisheries, University of Tennessee, Knoxville, TN, USA. 10Scientific Services, South African National Parks, Kruger National Park, Skukuza, South Africa.
11School of Natural Resource Management, Nelson Mandela University, Port Elizabeth, South Africa. 12Southern Research Station, USDA Forest Service,
Pineville, LA, USA. 13Department of Ecology, Evolution, and Behavior, University of Minnesota, St. Paul, MN, USA. 14Department of Plant and Microbial
Biology, North Carolina State University, Raleigh, NC, USA. 15School of Forestry & Wildlife Sciences, Auburn University, Auburn, AL, USA. 16Department of
Agriculture and Fisheries, Queensland Government, Brisbane, Queensland, Australia. 17Rocky Mountain Research Station, USDA Forest Service, Flagstaff,
AZ, USA. 18Forestry Program, Holdsworth Natural Resources Center, University of Massachusetts, Amherst, MA, USA. 19Department of Biology, Stanford
University, Stanford, CA, USA. 20Department of Forest Resources, University of Minnesota, St. Paul, MN, USA. 21Hawkesbury Institute for the Environment,
Western Sydney University, Penrith, New South Wales, Australia. 22School of Geosciences, University of Edinburgh, Edinburgh, UK. 23College of Natural
Resources, University of Wisconsin–Stevens Point, Stevens Point, WI, USA. 24Department of Geography, University of Colorado-Boulder, Boulder, CO,
USA. 25Division of Forestry and Natural Resources, West Virginia University, Morgantown, WV, USA. 26Northern Research Station, USDA Forest Service,
Houghton, MI, USA. 27Department of Integrative Biology, University of Wisconsin, Madison, WI, USA. 28Tall Timbers Research Station, Tallahassee, FL,
USA. 29Woods Institute for the Environment, Stanford University, Stanford, CA, USA. 30Precourt Institute for Energy, Stanford University, Stanford, CA,
USA. e-mail:
Ecosystem resilience to changing fire regimes13 will be a key
determinant of how terrestrial ecosystems respond to global
change3,4. Fire is a pervasive disturbance, burning ~5% of
global land area each year and releasing carbon stored in plant bio-
mass equivalent to ~20% of anthropogenic fossil fuel emissions5.
Historically, much of this carbon is re-sequestered through time as
plants recover and regrow, then lost again in the next fire. However,
in many systems, changes in climate and land use have shifted fire
frequencies, potentially changing the ability of plants, especially
trees, to regrow between fires1,68.
Although trees can be top-killed in fires, previous studies have
presented mixed results for the effect of burning frequency on tree
biomass913, highlighting the need to understand the factors lead-
ing to different fire effects. Climate extremes can moderate fire
effects on trees by influencing both tree growth and mortality as
well as fire intensity: sites with strong wet seasons have trees lack-
ing physiological adaptations to burning14, and sites with strong dry
seasons can have intense fires and droughts15. Furthermore, trees
in different ecosystems respond to changes in fire frequency dif-
ferently, partly owing to fuel load and composition but also their
physiology16. For example, traits conferring physiological protec-
tion from heating during fire and/or the capacity to colonize and
regrow rapidly could decrease tree biomass losses under frequent
burning13,17,18. Tests of these hypotheses using observational data are
Decadal changes in fire frequencies shift tree
communities and functional traits
Adam F. A. Pellegrini 1,2 ✉ , Tyler Refsland 3, Colin Averill4, César Terrer 1,5, A. Carla Staver 6,
Dale G. Brockway7, Anthony Caprio8, Wayne Clatterbuck9, Corli Coetsee10,11, James D. Haywood12,
Sarah E. Hobbie 13, William A. Hoffmann 14, John Kush15, Tom Lewis16, W. Keith Moser17,
Steven T. Overby17, William A. Patterson III18, Kabir G. Peay 19, Peter B. Reich 20,21, Casey Ryan 22,
Mary Anne S. Sayer12, Bryant C. Scharenbroch23, Tania Schoennagel24, Gabriel Reuben Smith 4,19,
Kirsten Stephan25, Chris Swanston 26, Monica G. Turner 27, J. Morgan Varner28 and
Robert B. Jackson 1,29,30
Global change has resulted in chronic shifts in fire regimes. Variability in the sensitivity of tree communities to multi-decadal
changes in fire regimes is critical to anticipating shifts in ecosystem structure and function, yet remains poorly understood.
Here, we address the overall effects of fire on tree communities and the factors controlling their sensitivity in 29 sites that expe-
rienced multi-decadal alterations in fire frequencies in savanna and forest ecosystems across tropical and temperate regions.
Fire had a strong overall effect on tree communities, with an average fire frequency (one fire every three years) reducing stem
density by 48% and basal area by 53% after 50 years, relative to unburned plots. The largest changes occurred in savanna eco-
systems and in sites with strong wet seasons or strong dry seasons, pointing to fire characteristics and species composition as
important. Analyses of functional traits highlighted the impact of fire-driven changes in soil nutrients because frequent burning
favoured trees with low biomass nitrogen and phosphorus content, and with more efficient nitrogen acquisition through ectomy-
corrhizal symbioses. Taken together, the response of trees to altered fire frequencies depends both on climatic and vegetation
determinants of fire behaviour and tree growth, and the coupling between fire-driven nutrient losses and plant traits.
Articles NATurE EcoloGy & EVoluTioN
limited by the collinearity between fire regime and environmental
variables at individual sites19,20; instead, tests require experimental
manipulations across broad geographic ranges that are long-term,
ideally multi-decadal, because fire-driven mortality and top-kill can
take decades to impact the tree community10,13,21.
In addition to external factors, changes within the tree commu-
nity can modify the effect of repeated fires through the increasing
abundance of species that are more tolerant of the direct and indi-
rect effects of fire. Fire effects may be buffered by the colonization
and growth of species with structural traits that help trees to avoid
heat-induced mortality, but also of species with different nutrient
use and acquisition traits that help plants optimize for fire-driven
changes in nutrient availability22,23. Such shifts in nutrient use and
acquisition strategies have ecosystem-level implications, as changes
in these traits significantly influence soil nutrient cycling24,25. For
example, plants that form symbioses with ectomycorrhizal (ECM)
fungi, arbuscular mycorrhizal (AM) fungi, or nitrogen-fixing bacte-
ria may be better equipped to access nutrients26 that can be depleted
under frequent burning22. The distinction between strategies is
key because ECM plants tend to slow nutrient cycling and pro-
ductivity, while arbuscular and nitrogen-fixing species accelerate
nutrient cycling and productivity2426. Tissue stoichiometry is also
relevant, with species containing lower nutrient concentrations bet-
ter equipped to tolerate low nutrient supply, but in turn contributing
to slower nutrient turnover and lower nutrient availability24,25. Thus,
how fire filters for species with different nutrient use and acquisi-
tion strategies can influence long-term site productivity.
Here, we evaluate how changes in fire frequency alter tree
communities, and how climate, vegetation composition and soils
influence the variability in the sensitivity of trees to changing fire
regimes. Furthermore, we quantify how fire changes the distribu-
tion of functional traits in the tree community to evaluate multiple
nutrient use traits that are indicative of tree responses to fire-driven
changes in nutrients. We combined published data with new sur-
veys on tree populations from 29 sites and 374 plots in four con-
tinents. At 27 of the sites (324 plots), surface fire frequency was
experimentally manipulated or varied naturally for 16–64 years
(mean of 30 years), and at two sites (50 plots), natural variation in
crown fire frequency presented a natural experiment. We focused
on tree responses because of their importance for carbon storage,
ecosystem productivity, and nutrient cycling23,27,28. The sites cover
locations in North America (coniferous, deciduous and mixed for-
ests, and broadleaf savanna), South America (tropical savanna and
temperate shrubland), Africa (broadleaf savanna) and Australia
(wet and dry eucalyptus forests), all of which are ecosystems that
experience frequent burning (Extended Data Fig. 1, Supplementary
Table 1 and Supplementary Information). For our main analysis, we
define ecosystems broadly based on dominant fuel type (grass ver-
sus tree litter) and tree functional composition (angiosperm broad-
leaf versus gymnosperm needleleaf trees), but test the robustness of
our conclusions to other classifications. Each surface fire site con-
tains replicate plots including an unburned treatment and differ-
ent prescribed burning frequencies (Extended Data Fig. 2), where
fire frequencies ranged from approximately one fire every decade
to one fire every year (Supplementary Table 1). We evaluated the
effects of fire alone and in combination with environmental covari-
ates using model selection (Supplementary Information) to test the
importance of climate, soil and species composition in modifying
the effect of fire.
Results and discussion
Effects of repeated burning compounded over multiple decades.
There were clear overall effects of fire treatments on tree population
size. Stem density (stems per hectare) and basal area were lower in
frequently burned plots relative to infrequently or unburned plots
(see Figs. 1a and 2a for effects in each site and Figs. 1b and 2b for
the response ratio across sites). A comparison between the most
extreme fire frequency treatments using response ratios illustrated
that density and total basal area were 44 ± 25% and 73 ± 45% lower,
respectively, in the most frequently burned plots compared with
unburned plots (Figs. 1b and 2b). The differences between fire treat-
ment effects were larger when fire frequencies contrasted more (for
example, the effect of intermediate versus high fire frequency was
52% and 28% lower than the effect of low versus high frequency
for both stem density and basal area, respectively; Figs. 1b and 2b,
and Supplementary Table 2, errors are 95% confidence intervals).
Duration of exposure to altered frequencies was also significant, as
sites with longer durations of altered fire frequencies had larger dif-
ferences between fire treatments, with the slope between duration
and community size growing more negative with more frequent
burning (density: F1,280 = 8.4, P = 0.004; basal area: F1,289 = 23.3,
P < 0.001; Figs. 1c and 2c). For example, relative to unburned plots,
plots with a three-year fire-return interval had 26% lower stem
density and 27% lower basal area after 30 years; the differences
increased to 48% lower stem density and 53% lower basal area after
50 years (Figs. 1c and 2c, and Supplementary Table 3; see Extended
Data Fig. 3 for non-transformed results). In annually burned plots,
the most extreme fire frequency, burned plots had 63% lower stem
density and 72% lower basal area than unburned plots after 50 years
(Figs. 1c and 2c, and Supplementary Table 3). Thus, both duration
of experiment and prescribed fire frequency help to reconcile the
variable effects of fire across studies. Accounting for these factors
illustrates that the effects of changing fire frequencies may take sev-
eral decades to become substantial, but the impact of the changes
will continue to increase for many decades.
Fire type was also important, with frequent crown fires affect-
ing tree populations to a greater degree than frequent surface fires.
Comparison of 50 plots in needleleaf forests that experienced natu-
ral variability in the frequency of stand-replacing crown fires (that
is, wildfires) illustrated that stands with shorter fire-return intervals
had significantly lower tree densities, especially when plots with
the shortest return intervals were considered (F1,26.5 = 5.2, P = 0.03
and F1,21 = 10.3, P = 0.004, respectively; Extended Data Fig. 4).
Experimental manipulation of surface fire frequency (that is, pre-
scribed fires) in needleleaf forests in North America showed that
stem densities were lower in more frequently burned plots, but less
so than differences caused by frequent crown fires (F1,47.1 = 17.2,
P = 0.001; Extended Data Fig. 4). The large effect of short-interval
crown fires on tree communities, supported by studies from other
regions29,30, highlights the importance of higher fire intensities hav-
ing more severe effects.
Fire sensitivities differed across biomes and climates. Although
fire frequency had a large overall effect on trees, there was substan-
tial variability in the sensitivity across sites. Part of the variability was
attributable to fire history prior to the establishment of the experi-
ment. In forest sites that burned regularly in the decades before the
onset of the experiment, fire exclusion resulted in stem density and
basal area being 47% (±40%) and 65% (±71%) higher, respectively,
than treatments that maintained historical burning frequencies (see
Supplementary Table 1 for site fire histories). In contrast, the rein-
troduction of prescribed fire into forests that had not burned for
several decades before the onset of the experiment had relatively
minimal effects. These results probably differ from those commonly
observed with wildfires, which can have larger effects in forests with
a history of fire exclusion due to high fuel accumulation31,32 because
prescribed surface fires are less intense. In savannas, where the fire
experiments were all initiated in landscapes that burned regularly
in the decades preceding the experiment, fire exclusion resulted
in basal area increasing by 41% (±18%), but increasing fire fre-
quency resulted in basal area declining by 48% (±22%), while stem
density remained unchanged, relative to an intermediate interval
NATurE EcoloGy & EVoluTioN
that maintained the pre-experiment frequency (statistics from log
response ratios ±95% confidence intervals, P < 0.001 for both basal
area comparisons; Supplementary Information). Taken together, the
largest effects of altered fire frequencies were owing to fire exclusion
in landscapes that had burned regularly for the past few decades.
Model selection illustrated that climate, vegetation type and con-
tinent played significant roles in explaining the variability in the sen-
sitivity of trees to fire. Fire effects were largest in areas that received
more rainfall in the wet season, less rainfall in the dry season and
had lower mean annual temperatures (F1,285.1 = 50.3, P < 0.001;
F1,285.6 = 7.5, P = 0.007; and F1,280.8 = 16.8, P < 0.001, respectively; Fig.
3a–c and Supplementary Table 4; see Supplementary Table 5 for
stem density). For example, plots that experienced more frequent
burning (two fires every three years, one standard deviation above
mean frequency) had 65% lower tree basal area than unburned plots
in sites with high wet season precipitation. In sites with average wet
season precipitation, the difference between the same treatments
was only 22% (Fig. 3a, Extended Data Fig. 5 and Supplementary
Table 4; see Supplementary Information for details on calculations).
Dry season precipitation had the opposite correlation with fire
effects: sites with lower precipitation in the dry season experi-
enced twice as large an effect of fire on basal area (frequent burning
resulted in 44% versus 22% lower tree basal area in sites with low
versus average dry season precipitation; Fig. 3c and Supplementary
Table 4). The contrasting response to precipitation in the wet versus
dry season is consistent with our understanding that fires are most
intense in areas with stronger wet seasons (leading to more fuel)
and more severe dry seasons (lower fuel moisture), thus contrib-
uting to potential losses with more frequent burning3335. Rainfall
in the dry season probably also influences fire effects by determin-
ing the water available for tree growth. Neither soil texture nor soil
carbon explained the sensitivity to changing fire frequencies across
sites (Supplementary Table 4). Thus, climate was a key determinant
of fire effects, with more moderately seasonal sites being the least
sensitive to changing fire regimes.
The effect of fire on tree basal area also differed across eco-
systems, with frequent burning having a larger effect on tree
basal area in savannas relative to broadleaf and needleleaf forests
10 29 48 67 67 6710 29
Years of repeated burning
48 10 29 48
Dry Peachester
Highland Rim
University Missouri
Chimney Spring
Kings Canyon
Tall Timbers
Cedar Creek
Kruger Mopani
Kruger Pretoriuskop
Kruger Satara
Kruger Skukuza
La Pampa
Broadleaf forest
Needleleaf forest
Fire frequency 0–0.21 yr–1 Fire frequency 0.22–0.66 yr–1 Fire frequency 0.67–1 yr–1
Low versus
Low versus
Mid versus
Treatment comparisons
Partial residual log(stem density) Stem density log(RR)
Stem density log(RR)
Fig. 1 | Tree stem density declines with both fire frequency and duration of fire regime. a,b, log response ratios (RR) of stem densities and the
surrounding 95% confidence intervals. Shown are comparisons between ln (burned/unburned) treatments within each individual site coloured by broad
ecosystem categorization with the burned treatment being the most extreme fire frequency treatments in a (a; n= 25; Supplementary Table 1, ecosystem
groups based on broad differences in fire fuel and tree composition; Supplementary Information). Also shown are averages across all sites comparing the
different levels of fire frequency in studies with 3 levels; less frequent treatment always in denominator (b; n= 23, n= 17, n= 18; Supplementary Table
2). c, Partial residuals plot from a mixed-effects model including fire frequency, the number of years of repeated burning and their interaction for ln stem
density (n= 303; Supplementary Table 3); site was used as a random intercept. Panels are centred on cross-section values of one fire every ten years, one
every three years and one every year, but encompass a range of fire frequencies within each panel. IBGE, Reserva Ecológica do IBGE; YellStonL, Yellowstone
low severity; YellStonH, Yellowstone high severity.
Articles NATurE EcoloGy & EVoluTioN
(F2,279.7 = 14.0, P < 0.001, model incorporating climate effects; Fig. 3d
and Supplementary Table 4). Relative to the unburned plots, basal
area in frequently burned plots was 5% lower in needleleaf forests
and 22% lower in broadleaf forests (Fig. 3d, burn frequency of two
fires every three years; Supplementary Information; continent dif-
ferences in Extended Data Fig. 6). In savannas, frequently burned
plots had 70% lower basal area relative to the unburned plots (Fig. 3d
and Supplementary Table 4). Interestingly, stem density responses
to fire frequency were qualitatively different between savannas and
forests (Supplementary Table 5). Stem densities increased with
more frequent burning in forests while basal area decreased, poten-
tially due to higher light availability and tree recruitment in the for-
ests. To assess the sensitivity of these findings to our classification
of ecosystem type, we also tried using a subdivided classification
by partitioning broadleaf forests into oak and eucalypt types, and
needleleaf forests into those that transitioned between oak and pine
dominated (Supplementary Table 1). When included in the final
model, the subdivided vegetation classification still had a significant
main effect (F4,16.8 = 11.1, P = 0.0001) and a significant interaction
with fire frequency (F4,282.5 = 7.1, P < 0.001; Extended Data Fig. 7),
with basal area in savannas responding the most to changes in fire
frequency (Extended Data Figs. 7 and 8).
Fire changed the functional composition of tree communities.
Fire-driven changes in basal area and stem density are important for
ecosystem function, but fire can also impact ecosystems by chang-
ing the functional composition of trees. To address this, we anal-
ysed functional composition in only the experiments from North
America (77 tree species, 16 sites, 181 plots) because trait data were
available there to (1) categorize species by nutrient acquisition strat-
egies and (2) assign wood, leaf and root traits related to growth, sur-
vival and nutrient use strategies.
We found that structural traits were important for explaining
cross-site variability in the sensitivity of tree communities to fire.
Across sites, frequent burning impacted basal area more where
tree species had thinner bark and denser wood (bark thickness:
Partial residual log(basal area)
Basal area log(RR)
Basal area log(RR)
Broadleaf forest
Needleleaf forest
Dry Peachester
Highland Rim
University Missouri
Wet Peachester
Chimney Spring
Kings Canyon
Tall Timbers
Cedar Creek
Kruger Mopani
Kruger Pretoriuskop
Kruger Satara
Kruger Skukuza
Low versus
Low versus
Mid versus
Treatment comparisons
Fire frequency 0–0.21 yr–1 Fire frequency 0.22–0.66 yr–1 Fire frequency 0.67–1 yr–1
10 29 48 67 67 6710 29
Years of repeated burning
48 10 29 48
Fig. 2 | Tree basal area declines with both fire frequency and duration of fire regime. a,b, log response ratios of basal area and the surrounding 95%
confidence intervals. Shown are comparisons within each individual site coloured by broad ecosystem categorization in the most extreme fire frequency
treatments ln (burned/unburned) (a; n= 23; Supplementary Table 1, ecosystem groups based on broad differences in fire behaviour and tree composition;
Supplementary Information). Also shown are averages across all sites comparing the different levels of fire frequency in studies with 3 levels; less
frequent treatment always in denominator (b; n= 23, n= 17, n= 17; Supplementary Table 2). c, Partial residuals plot from a mixed-effects model including
fire frequency, the number of years of repeated burning and their interaction for ln basal area (n= 309; Supplementary Table 3); site was used as a random
intercept. Panels are centred on cross-section values of one fire every ten years, one every three years and one every year, but encompass a range of fire
frequencies within each panel (ranges: unburned to one every five years; one every five years to two every three years; two every three years to every year).
NATurE EcoloGy & EVoluTioN
F1,164.3 = 9.2, P = 0.003; wood density: F1,165 = 10, P = 0.002; see
Supplementary Table 6 for both in the same model, and Fig. 4a,b).
However, within sites we found mixed evidence that fire filtered for
species with different structural trait values. Mean wood density
of the tree community tended to be lower in frequently burned
plots, potentially because of increasing dominance of gymnosperm
trees, which tend to have lower wood density. In contrast, we did
not observe any effect of fire on the mean bark investment of the
Partial residual
log(basal area)
Partial residual
log(basal area)
0.02 0.34
MAT (°C)
13–17 18–22
Precipitation wet quarter (mm):
302–450 451–737 12–79
Precipitation dry quarter (mm):
80–187 188–351
Broadleaf forest Needleleaf forest Savanna
0.66 0.34
Fire frequency (fires yr
) Fire frequency (fires yr
0.661 0.02 1 0.34 0.660.02 1 0.02 0.34 0.66 0.34 0.661 0.02 1 0.34 0.660.02 1
0.02 0.34 0.66 0.34
Fire frequency (fires yr
) Fire frequency (fires yr
0.661 0.02 1 0.34 0.660.02 1 0.02 0.34 0.66 0.34 0.661 0.02 1 0.34 0.660.02 1
Fig. 3 | Climate and ecosystem type modify effects of fire frequencies on tree basal area. ad, Partial residual plots of the mixed-effects model illustrating
the interactive effects between covariates (site as a random intercept, n= 309 plots, n= 25 sites). Panels centred on cross-sectional values from one
standard deviation around the median (1, 0, 1). MAT, mean annual temperature. Partial residual predictions account for the values of all other covariates
in the model. Comparisons of rainfall scenarios relative to the mean in the text used wet season precipitation of +1 standard deviation above the mean
(525 versus 375 mm yr1) and dry season precipitation of 1 standard deviation below the mean (25 versus 133 mm yr1). The duration of experiment held
at its mean of 28 years. All model fits are P < 0.05; statistics are in Supplementary Table 4.
Partial residual log(basal area)
10.5 mm 11.6 mm 16.0 mm 0.38 g cm
0.54 g cm
0.59 g cm
0.50 0.75
Fire frequency (fires yr
Arbuscular mycorrhizal Ectomycorrhizal
Fire frequency (fires yr
Years of repeated burning
Relative BA (arc sine)
0.02 0.34 0.66 1 0.02
Fire frequency (fires yr
0.34 0.66 1 10 29 48 67 10 29 48 67
Green leaf N (mg g
Root N (mg g
0.25 0.50 0.75 1.000 0.25 0.50 0.75 1.000 0.25 0.50 0.75 1.000 0.25 0.50 0.75 1.000 0.25 0.50 0.75 1.000
c d
Partial residual log(basal area)
Fig. 4 | Functional composition of tree communities both responds to and modifies the effects of frequent burning. a,b, Partial regression plots from
mixed-effects models with community-weighted means of bark investment (BI) scaled to a 10 cm stem size (a) and wood density (WD) as modifying
variables (b; n= 172 plots, n= 16 sites). Bark investment and wood density were included in the same model and were negatively correlated σ=0.86.
Statistics are in Supplementary Table 6. Basal area is ln transformed. c, Relative basal area of nutrient acquisition strategies (arcsine transformed) of AM
versus ECM trees as a function of fire frequency. d, Community-weighted means of nutrient use strategies, showing green leaf and live root N as a function
of study duration. Statistics are in Supplementary Tables 7–9 (n= 172 plots and n= 16 sites), which also show data on litter N and resorption. Phosphorus
content data are shown in Extended Data Fig. 7.
Articles NATurE EcoloGy & EVoluTioN
tree community (Supplementary Table 6), demonstrating that bark
investment at the community scale does not appear to change in
response to fire. Nevertheless, bark investment helped to predict
basal area loss patterns across broad biogeographic scales.
Frequent fire favoured conservative nutrient strategies. We found
evidence for compensatory responses in the nutrient use and acqui-
sition strategies of the tree communities under different fire fre-
quencies. On average, plots burned frequently for longer periods of
time were dominated by tree species with low nitrogen (N) concen-
trations in green and senesced leaves and roots, and that resorbed
a greater proportion of N before leaf senescence (P < 0.001 for all
variables; Fig. 4d and Supplementary Table 7). This is consistent
with the hypothesis that fire-driven soil N losses22 filter for species
with adaptations to low N conditions. Tissue phosphorus (P) con-
centrations also declined with frequent burning in leaves and litter
but not in roots (Extended Data Fig. 9 and Supplementary Table 7).
Given that P is less prone to being lost than N22,36,37, this result was
surprising and raises questions about how fire may change the P
economy of trees in an opposite way to changes in soil P availability.
Taken together, fire filters for species with more conservative nutri-
ent use strategies in both their leaves and fine roots across several
North American savannas and forests.
Fire also affected the relative abundance of nutrient acquisi-
tion strategies, evaluated by trees’ abilities to form symbioses38.
Trees that formed symbioses with ECM and AM fungi were the
most abundant nutrient acquisition strategies across our plots; eri-
coid and nitrogen-fixing trees were absent from most sites (Fig.
4c; categorization based on tree species’ taxonomy38). ECM trees,
which associate with fungal symbionts capable of acquiring N from
organic matter39, tended to be more successful in frequently burned
plots. The relative abundance of ECM trees increased from 85% in
unburned plots to nearly 100% in annually burned plots (Fig. 4c
and Supplementary Table 8). ECM trees were also more common
in warmer climates and on soils with low carbon concentrations
(Supplementary Table 8). ECM trees typically have lower con-
centrations of N and P in leaves, litter and roots than AM trees38
(Supplementary Table 9), suggesting the turnover in symbiont
composition may be driving the shift in stoichiometry of the tree
community. As repeated fire tends to decrease inorganic N and N
mineralization23, it makes sense that frequent fire causes the tree
community to shift towards species capable of obtaining N from
soil organic matter.
The tendency for frequently burned plots to have tree communi-
ties dominated by ECM species with low N and P content in leaves,
roots and litter indicates that frequent burning favours conserva-
tive nutrient use and acquisition strategies. This trend is probably a
result of fire-driven soil nutrient losses, which should favour species
better equipped to cope with low nutrient environments; however,
other factors such as phylogenetic conservatism of traits may be
at play38. These trait shifts may themselves reinforce an important
fire–nutrient feedback if N losses cause a decline in productivity
that limits the ability of trees to regrow between fire events, fur-
ther decreasing ecosystem N turnover. Our results unpack only one
part of this feedback loop (fire effects on traits), and further study
is needed to connect changes in strategies with the observed differ-
ences in nutrient availability and plant growth.
Our analysis also highlights several areas for future work. For
one, an improved representation of fire experiments in different
ecosystem types across continents in our dataset (for example,
tropical forests in Africa) would help further unpack the variabil-
ity across ecosystems and continents. Second, longitudinal data on
how fire effects emerge through time could assist with better under-
standing how the turnover in tree species composition influences
changes in total tree cover within experiments10,18,40. Third, consid-
ering other plant groups (for example, herbaceous plants) will help
obtain a more comprehensive picture of how shifting fire regimes
change ecosystem function. Finally, the extent of fire effects on plant
strategies across temperate ecosystems highlights the need for more
studies of plant strategies to consider fire. For example, the effect of
fire on fungal symbiosis strategies should be integrated into theories
seeking to explain their biogeographic distribution, which generally
rely on climatic factors alone41.
Our observation that changing fire frequencies shift both tree
basal area and density as well as trait composition is important
because it identifies two means by which fire can indirectly alter
carbon and nutrient storage in other ecosystem pools, such as
soil organic matter. For one, the observed multi-decadal decline
in tree populations mirrors multi-decadal shifts in soil carbon
and nitrogen22,23, both in timescale and across ecosystems. The
greater rates of change in tree population sizes in savannas and
broadleaf forests are consistent with the higher losses of mineral
soil carbon and nitrogen reported in those ecosystems relative to
needleleaf forests22,42. Furthermore, a shift towards conservative
nutrient use and acquisition traits is consistent with lower soil N
mineralization23,42. Consequently, our findings support hypotheses
that fire-driven changes in tree biomass inputs and the turnover
of plant traits may both contribute to changes in soil carbon and
nutrient pools4245.
Widespread changes in fire regimes are likely to shift both the popu-
lation size and functional composition of tree communities, with
both factors affecting the storage and cycling of carbon and nutri-
ents. The effects are heterogeneous, however, with certain ecosys-
tems being especially sensitive, such as savannas experiencing rapid
encroachment of tree cover when fire is excluded. Climatic factors
were also key, as regions with extreme precipitation amounts in the
wet and dry seasons (high and low, respectively) changed more than
areas with more moderate precipitation. The effects of fire were
not limited to tree population sizes, but also extended to the func-
tional composition of the community. The convergent response of
frequent burning promoting conservative nutrient use strategies
indicates that fire impacts nutrient cycling not only over decadal
timescales, but also suggests that fire probably influences the evo-
lution of these plant strategies. Consequently, climatically sensitive
shifts in fire frequency, even when relatively low intensity, will alter
the structure and functioning of ecosystems through multiple direct
and indirect pathways.
Site descriptions and experimental designs. e majority of sites sampled are
in ecosystems that experience surface res (from re manipulation experiments,
n = 27). Our main analyses are based on the surface re experiments, but we
compare these data with a network of plots across n = 2 sites with natural
variability in the frequency of stand-replacing crown res to evaluate the eect
of re regime. We describe the sites briey in Supplementary Table 1 and present
detailed descriptions of site history in Supplementary Data 3. Most of the data were
obtained from existing studies, but we complemented these data with unpublished
surveys in the Sequoia and Kings Canyon sites. We identied these sites rst by
using a list of sites from a previous meta-analysis of multi-decadal changes in re
frequencies22. Not all of the sites in the other study contained vegetation surveys
and in some cases the authors of the other studies did not share data. We then
complemented these studies with other long-term re manipulation experiments
using a literature search and conference presentations. Finally, data from some of
the sites were collected specically for this study.
The surface fire experiments are mostly experimental prescribed burn plots.
The managers generally try to burn in a broad seasonal window (for example, a
spring fire in North America may occur anytime from March to May) to optimize
burn timing for the local fire conditions most suitable for their planned fire
intensities. Treatments were not applied in a uniform fashion across sites, which
is one motivation for using mixed-effects models (see description of our statistical
approach below).
The sites contained different land use histories before the establishment of
the experiment, which was not always documented in detail, but we describe key
factors in Supplementary Table 1 and Supplementary Data 3. We describe how we
NATurE EcoloGy & EVoluTioN
evaluated the potential role of land use history in ‘Testing role of fire and land use
history’ in the Supplementary Information.
Environmental data. Soil chemistry. We collected and analysed soil data
using several methods. First, we determined the dominant soil type using
either author descriptions or reported soil texture analysis. Second, we used
the highest resolution soil data possible (for example, soil samples from each
replicate plot within a re treatment), but some sites contained only site-level
soil properties. Consequently, we analysed overall eects of re on all sites
without any covariates, followed by a model that uses model selection to
account for collinearities among variables when testing for factors that modify
re eects. To extend data on soils across plots, we sampled soils (top 0–5 cm
of the mineral horizon) in 24 plots across four sites: Kings Canyon, Sequoia,
Limestone Flats and Chimney Springs. Each site contained three replicate plots
of an unburned treatment and a high re frequency treatment. We collected n = 5
pseudo-replicates within the true replicate plot, analysed the soils for carbon and
texture, and averaged within each plot. For chemical analyses, soils were sieved
to <2 mm, dried to constant weight and ground on a ball mill. Subsets were then
analysed using combustion on an elemental analyser at Stanford University;
duplicates were run every 10 samples (analytical accuracy was >90%). Texture
was analysed using the hydrometer method (adopted from the Kellogg Biological
Station, Long-Term Ecological Research, on
sieved soils.
Climate. To obtain long-term climate averages at each site, we used WorldClim46.
WorldClim integrates data from 1970–2000 across 9,000–60,000 weather
stations and spatially extrapolates the values by integrating other covariates from
topography maps and satellite data (described in detail in ref. 46). The timespan of
the climate data overlaps the duration of the experiments reasonably well (mean
study initiation = 1983; mean survey year = 2006). Managers timed burning to
coincide with consistent weather conditions over the course of the experiment,
therefore we did not obtain high-resolution interannual variability in climate.
We focused on several climate variables based on ecologically relevant a priori
hypotheses: (1) precipitation partitioned into the driest and wettest quarters of the
year because precipitation influences fuel accumulation (primarily in the wettest
quarter) and fire conditions (primarily in the driest quarter); and (2) mean annual
temperature because of its large effect on a variety of biogeochemical processes.
Precipitation in wet and dry quarters are not as correlated with one another
but are highly correlated with mean annual precipitation and temperature
(Supplementary Table 10).
Ecosystem type. The vegetation composition at each site differs substantially,
ranging from diverse tropical savannas with dozens of tree species (for example,
Kruger sites) to monodominant coniferous forests (for example, Limestone
Flats and Chimney Springs). Classifying the sites into broad categories was
done methodologically, by balancing the need to maintain parsimony (and thus
statistical power) with accurately capturing how plant composition may modify
fire effects. Consequently, we performed two levels of classification: (1) a coarse
categorization separating savannas versus forests, and within forests treating
broadleaf and needleleaf forests separately, which we refer to as a vegetation type;
and (2) accounting for variability within forest types by partitioning broadleaf
forests into Myrtaceae (eucalypt) versus Fagaceae (oak) dominated, and needleleaf
forests into forests that are near completely dominated by needleleaf trees versus a
mixed forest containing both needleleaf and broadleaf trees, which we refer to as a
sub-vegetation type.
Quantifying environmental effects. Several methods exist to calculate variable
importance, with no clear optimal method47. We chose to use the regression
coefficients in the model to understand the sensitivity of basal area and stem
density to changes in relative values of each variable. Importantly, the models were
fitted to rescaled data by subtracting each value by the mean and dividing by the
standard deviation of the variable. Consequently, the product between the mean
value of a variable and its coefficient is always zero. Thus, we can compare the
relative impact of variables by comparing the magnitude of the fitted coefficients
because they reflect the potential change in basal area for a one standard deviation
change in a variable value.
Wet season precipitation. Wet season precipitation varied one standard deviation
above the mean versus at the mean (525 versus 375 mm yr1, respectively). Fire
frequency varied from unburned to one standard deviation above the mean (two
fires every three years). Wet season precipiation is shown in Fig. 3a.
Dry season precipitation. Dry season precipitation was one standard deviation
below the mean versus at the mean (25 versus 133 mm yr1, respectively). Fire
frequency varied from unburned to one standard deviation above the mean (two
fires every three years). Dry season precipiation is shown in Fig. 3c.
Vegetation type. Fire frequency effects were made using two levels of comparison.
Unburned plots versus burning at the mean frequency (one fire every three years)
and unburned plots versus burning at one standard deviation above the mean
frequency (two fires every three years). Vegetation type is shown in Fig. 3d.
Statistics. Fire eect calculations. We rst evaluated the overall eect of re
frequency and duration that frequency was altered on tree basal area and stem
density without considering any potential modifying role of covariates. To
accomplish this we analysed (1) a mixed-eects model containing re frequency,
re period and their interaction; and (2) log response ratios of stem density and
basal area relativized within each site. We excluded the 50 crown re plots for this
initial analysis. We tted the mixed-eects models with site as a random intercept.
e statistical design is nested because each site has several replicate plots receiving
dierent re treatments. In all cases of mixed-eects models, we tested for model
signicance using Satterthwaite’s approximation for degrees of freedom and a type
III analysis of variance48. In the event of an insignicant main eect but signicant
interaction, we tested whether the main eect could be dropped from the model
using a change in Akaike information criterion with a threshold of two.
Fire–environment interactions. For the plots with surface fires, we performed model
selection by incorporating covariates of climate, soil and plant composition into
mixed-effects models to test for pairwise interactions and possible collinearities
(see discussion below of collinearities). Finally, we constructed a full model
containing fire, climate, soil and composition variables based on our hypotheses
that these factors will interact with fire frequency as well as information gained
from the pairwise tests. There were several insignificant effects in the final model,
which we tested for removal using model selection with a threshold Akaike
information criterion of two. All variables were rescaled by subtracting the mean
and dividing by their standard deviation.
Traits. Bark thickness data were collected from a dataset in the Fire and Fuels
Extension to the Forest Vegetation Simulator (
fvs/docs/gtr/FFEaddendum.pdf). Although broad syntheses of bark investment
exist for many tree species in North America, not all species contained data from
empirical measurements, thus we used the data from the Fire and Fuels Extension.
Bark thickness was assumed to scale linearly with stem diameter, which is generally
valid for smaller stems, but it is known that bark saturates with increasing stem
diameter49. The ability of bark investment to predict fire effects will probably
improve with better consideration of the non-linear relationship between bark
and stem diameter. We evaluate the relative bark investment, not absolute bark
thickness, which is based on bark investment as well as stem size.
Wood density was compiled from the literature using a global wood density
database50, supplemented with additional data51,52. We assigned a genus-level
average for 19 species lacking data. Plant tissue stoichiometry and mycorrhizal type
were determined using both trait data and phylogenetic trait estimates calibrated
to trait data used in a previous global analysis of plant mycorrhizal traits38.
Full data selection criteria are presented in ref. 38, but we describe them in the
Supplementary Information.
To test for fire effects on the relative abundance of symbiotic strategies, we
calculated the relative basal area of the different strategies (ECM, AM and the less
abundant ericoid mycorrhizal, non-mycorrhizal and nitrogen-fixing tree species).
Given the low occurrences of ericoid, non-mycorrhizal and nitrogen-fixing species,
we analysed the relative abundance of only AM and ECM species. We then fitted
mixed-effects models with relative basal area as the dependent variable and fire,
climate, broad vegetation type (broadleaf, needleleaf, savanna) and soil conditions
as the independent variables, each modified by a symbiont term. Relative basal area
was arcsine transformed. This analysis was conducted in the North American plots.
To test how fire influenced the trait composition of the community we
fit mixed-effects models to test the effect of fire and environmental factors in
explaining the community-weighted mean trait values. We do not include an
independent effect of either wood density or bark thickness because we are
primarily concerned with how they may modify fire effects. We also tested for
whether the symbiotic strategies differed in their traits. To do so, we assigned
symbiotic strategies and the dominant ecosystem in which they occurred to
different species. We then analysed linear models incorporating symbiotic strategy
and ecosystem type as additive effects.
Reporting Summary. Further information on research design is available in the
Nature Research Reporting Summary linked to this article.
Data availability
Wood density data are from refs. 5052, plant tissue stoichiometry and mycorrhizal
type data are from ref. 38, and bark thickness data are from the Fire and Fuels
Extension to the Forest Vegetation Simulator (
docs/gtr/FFEaddendum.pdf). Supplementary Data 1 and 2 contain the woody
population size data and Supplementary Data 3 contains land use data. Refs. 5376
describe the fire experiments.
Received: 10 July 2020; Accepted: 25 January 2021;
Published: xx xx xxxx
Articles NATurE EcoloGy & EVoluTioN
1. Andela, N. et al. A human-driven decline in global burned area. Science 356,
1356–1362 (2017).
2. Westerling, A. L., Hidalgo, H. G., Cayan, D. R. & Swetnam, T. W. Warming
and earlier spring increase western US forest wildre activity. Science 313,
940–943 (2006).
3. Turner, M. G. Disturbance and landscape dynamics in a changing world.
Ecology 91, 2833–2849 (2010).
4. Higgins, S. I. & Scheiter, S. Atmospheric CO2 forces abrupt vegetation shis
locally, but not globally. Nature 488, 209–212 (2012).
5. van der Werf, G. R. G. R. et al. Global re emissions estimates during
1997–2016. Earth Syst. Sci. Data 9, 697–720 (2017).
6. Schoennagel, T. et al. Adapt to more wildre in western North American
forests as climate changes. Proc. Natl Acad. Sci. USA 114, 4582–4590 (2017).
7. Westerling, A. L., Turner, M. G., Smithwick, E. A. H., Romme, W. H. & Ryan,
M. G. Continued warming could transform Greater Yellowstone re regimes
by mid-21st century. Proc. Natl Acad. Sci. USA 108, 13165–13170 (2011).
8. Johnstone, J. F. et al. Changing disturbance regimes, ecological memory, and
forest resilience. Front. Ecol. Environ. 14, 369–378 (2016).
9. Lewis, T. Very frequent burning encourages tree growth in sub-tropical
Australian eucalypt forest. Forest Ecol. Manag. 459, 117842 (2020).
10. Peterson, D. W. & Reich, P. B. Prescribed re in oak savanna: re frequency
eects on stand structure and dynamics. Ecol. Appl. 11, 914–927 (2001).
11. Tilman, D. et al. Fire suppression and ecosystem carbon storage. Ecology 81,
2680–2685 (2000).
12. Pellegrini, A. F. A., Hedin, L. O., Staver, A. C. & Govender, N. Fire alters
ecosystem carbon and nutrients but not plant nutrient stoichiometry or
composition in tropical savanna. Ecology 96, 1275–1285 (2015).
13. Russell-Smith, J., Whitehead, P. J., Cook, G. D. & Hoare, J. L. Response of
eucalyptus-dominated savanna to frequent res: lessons from Munmarlary,
1973–1996. Ecol. Monogr. 73, 349–375 (2003).
14. Uhl, C. & Kauman, J. B. Deforestation, re susceptibility, and potential tree
responses to re in the eastern Amazon. Ecology 71, 437–449 (1990).
15. Case, M. F., WigleyCoetsee, C., Nzima, N., Scogings, P. F. & Staver, A. C.
Severe drought limits trees in a semiarid savanna. Ecology 100, e02842 (2019).
16. Keeley, J. E., Pausas, J. G., Rundel, P. W., Bond, W. J. & Bradstock, R. A. Fire
as an evolutionary pressure shaping plant traits. Trends Plant Sci. 16,
406–411 (2011).
17. Schoennagel, T., Turner, M. G. & Romme, W. H. e inuence of re interval
and serotiny on postre lodgepole pine density in Yellowstone National Park.
Ecology 84, 2967–2978 (2003).
18. Higgins, S. I. et al. Which traits determine shis in the abundance of tree
species in a re-prone savanna? J. Ecol. 100, 1400–1410 (2012).
19. Lehmann, C. E. R. et al. Savanna vegetation–re–climate relationships dier
among continents. Science 343, 548–552 (2014).
20. Staver, A. C., Archibald, S. & Levin, S. A. e global extent and determinants
of savanna and forest as alternative biome states. Science 334, 230–232 (2011).
21. Higgins, S. I., Bond, J. I. & Trollope, W. S. Fire, resprouting and variability: a
recipe for grass–tree coexistence in savanna. J. Ecol. 88, 213–229 (2000).
22. Pellegrini, A. F. A. et al. Fire frequency drives decadal changes in soil carbon
and nitrogen and ecosystem productivity. Nature 553, 194–198 (2018).
23. Reich, P. B., Peterson, D. W., Wedin, D. A. & Wrage, K. Fire and vegetation
eects on productivity and nitrogen cycling across a forest–grassland
continuum. Ecology 82, 1703–1719 (2001).
24. Phillips, R., Brzostek, E. & Midgley, M. e mycorrhizalassociated nutrient
economy: a new framework for predicting carbon–nutrient couplings in
temperate forests. New Phytol. 99, 41–51 (2013).
25. Hobbie, S. E. Plant species eects on nutrient cycling: revisiting litter
feedbacks. Trends Ecol. Evol. 30, 357–363 (2015).
26. Read, D. J. & PerezMoreno, J. Mycorrhizas and nutrient cycling in
ecosystems – a journey towards relevance? New Phytol. 157, 475–492 (2003).
27. Dixon, R. K. et al. Carbon pools and ux of global forest ecosystems. Science
263, 185–190 (1994).
28. Jackson, R. B. et al. Trading water for carbon with biological carbon
sequestration. Science 310, 1944–1947 (2005).
29. Whitman, E., Parisien, M. A., ompson, D. K. & Flannigan, M. D.
Short-interval wildre and drought overwhelm boreal forest resilience.
Sci. Rep. 9, 18796 (2019).
30. Hart, S. J. et al. Examining forest resilience to changing re frequency in a
re-prone region of boreal forest. Glob. Change Biol. 25, 869–884 (2019).
31. Stephens, S. L. et al. Managing forests and re in changing climates. Science
342, 41–42 (2013).
32. Steel, Z. L., Saord, H. D. & Viers, J. H. e re frequency–severity
relationship and the legacy of re suppression in California forests. Ecosphere
6, 1–23 (2015).
33. Scott, J. & Burgan, R. Standard Fire Behavior Fuel Models: A Comprehensive
Set for Use with Rothermel’s Surface Fire Spread Model General Technical
Report RMRS-GTR-153 (USDA, Forest Service and Rocky Mountain
Research Station, 2005).
34. Liu, Y. Y. et al. Recent reversal in loss of global terrestrial biomass. Nat. Clim.
Change 5, 470–474 (2015).
35. Brandt, M. et al. Satellite passive microwaves reveal recent climate-induced
carbon losses in African drylands. Nat. Ecol. Evol. 2, 827–835 (2018).
36. Butler, O. M., Elser, J. J., Lewis, T., Mackey, B. & Chen, C. e
phosphorus-rich signature of re in the soil–plant system: a global
meta-analysis. Ecol. Lett. 21, 335–344 (2018).
37. Raison, R. J., Khanna, P. K. & Woods, P. V. Transfer of elements to the
atmosphere during low-intensity prescribed res in three Australian subalpine
eucalypt forests. Can. J. Forest Res. 15, 657–664 (1985).
38. Averill, C., Bhatnagar, J. M., Dietze, M. C., Pearse, W. D. & Kivlin, S. N.
Global imprint of mycorrhizal fungi on whole-plant nutrient
economics. Proc. Natl. Acad. Sci. USA
pnas.1906655116 (2019).
39. Shah, F. et al. Ectomycorrhizal fungi decompose soil organic matter using
oxidative mechanisms adapted from saprotrophic ancestors. New Phytol. 209,
1705–1719 (2016).
40. Woinarski, J. C. Z., Risler, J. & Kean, L. Response of vegetation and vertebrate
fauna to 23 years of re exclusion in a tropical eucalyptus open forest,
Northern Territory, Australia. Austral Ecol. 29, 156–176 (2004).
41. Steidinger, B. S. et al. Climatic controls of decomposition drive the global
biogeography of forest–tree symbioses. Nature 569, 404–408 (2019).
42. Pellegrini, A. F. A. et al. Repeated re shis carbon and nitrogen cycling by
changing plant inputs and soil decomposition across ecosystems. Ecol.
Monogr. 90, e01409 (2020).
43. Newland, J. A. & DeLuca, T. H. Inuence of re on native nitrogen-xing
plants and soil nitrogen status in ponderosa pine - Douglas-r forests in
western Montana. Can. J. Forest Res. 30, 274–282 (2000).
44. Johnson, D. W. & Curtis, P. S. Eects of forest management on soil C and N
storage: meta analysis. Forest Ecol. Manag. 140, 227–238 (2001).
45. Pellegrini, A. F. A. Nutrient limitation in tropical savannas across multiple
scales and mechanisms. Ecology 97, 313–324 (2016).
46. Hijmans, R. J., Cameron, S. E., Parra, J. L., Jones, P. G. & Jarvis, A. Very high
resolution interpolated climate surfaces for global land areas. Int. J. Climatol.
25, 1965–1978 (2005).
47. Harrison, X. A. et al. A brief introduction to mixed eects modelling and
multi-model inference in ecology. PeerJ 2018, e4794 (2018).
48. Bates, D., Mächler, M., Bolker, B. & Walker, S. Fitting linear mixed-eects
models using lme4. J. Stat. Sow. 67, 1–48 (2015).
49. Jackson, J. F., Adams, D. C. & Jackson, U. B. Allometry of constitutive
defense: a model and a comparative test with tree bark and re regime. Am.
Nat. 153, 614–632 (1999).
50. Chave, J. et al. Towards a worldwide wood economics spectrum. Ecol. Lett.
12, 351–366 (2009).
51. Homann, W. A., Marchin, R. M., Abit, P. & Lau, O. L. Hydraulic failure and
tree dieback are associated with high wood density in a temperate forest
under extreme drought. Glob. Change Biol. 17, 2731–2742 (2011).
52. Harmon, M. E. Decomposition of standing dead trees in the southern
Appalachian Mountains. Oecologia 52, 214–215 (1982).
53. Hedges, L. V., Gurevitch, J. & Curtis, P. S. e meta-analysis of response
ratios in experimental ecology. Ecology 80, 1150–1156 (1999).
54. Gurevitch, J., Morrow, L. L., Wallace, A. & Walsh, J. S. A meta-analysis of
competition in eld experiments. Am. Nat. 140, 539–572 (1992).
55. Zanne, A. E. et al. ree keys to the radiation of angiosperms into freezing
environments. Nature 506, 89–92 (2014).
56. Pearse, W. D. et al. pez: phylogenetics for the environmental sciences.
Bioinformatics 31, 2888–2890 (2015).
57. Kembel, S. W. et al. Picante: R tools for integrating phylogenies and ecology.
Bioinformatics 26, 1463–1464 (2010).
58. Brockway, D. G. & Lewis, C. E. Long-term eects of dormant-season
prescribed re on plant community diversity, structure and productivity
in a longleaf pine wiregrass ecosystem. Forest Ecol. Manag. 96,
167–183 (1997).
59. Lewis, T. & Debuse, V. J. Resilience of a eucalypt forest woody understorey to
long-term (34–55 years) repeated burning in subtropical Australia. Int. J.
Wildl. Fire 21, 980–991 (2012).
60. Scudieri, C. A., Sieg, C. H., Haase, S. M., ode, A. E. & Sackett, S. S.
Understory vegetation response aer 30 years of interval prescribed burning
in two ponderosa pine sites in northern Arizona, USA. Forest Ecol. Manag.
260, 2134–2142 (2010).
61. Lewis, T., Reif, M., Prendergast, E. & Tran, C. e eect of long-term
repeated burning and re exclusion on above- and below-ground blackbutt
(Eucalyptus pilularis) forest vegetation assemblages. Austral Ecol. 37,
767–778 (2012).
62. Stratton, R. Eects of Long-Term Late Winter Prescribed Fire on Forest Stand
Dynamics, Small Mammal Populations, and Habitat Demographics in a
Tennessee Oak Barrens. MSc thesis, Univ. Tennessee (2007).
63. Wade, D. D. Long-Term Site Responses to Season and Interval of Underburns
on the Georgia Piedmont (Forest Service Research Data Archive, 2016).
NATurE EcoloGy & EVoluTioN
64. Pellegrini, A. F. A., Homann, W. A. & Franco, A. C. Carbon accumulation
and nitrogen pool recovery during transitions from savanna to forest in
central Brazil. Ecology 95, 342–352 (2014).
65. Nesmith, C. B., Caprio, A. C., Pfa, A. H., McGinnis, T. W. & Keeley, J. E. A
comparison of eects from prescribed res and wildres managed for
resource objectives in Sequoia and Kings Canyon National Parks. Forest Ecol.
Manag. 261, 1275–1282 (2011).
66. Haywood, J. D., Harris, F. L., Grelen, H. E. & Pearson, H. A. Vegetative
response to 37 years of seasonal burning on a Louisiana longleaf pine site.
South. J. Appl. For. 25, 122–130 (2001).
67. Higgins, S. I. et al. Eects of four decades of re manipulation on woody
vegetation structure in savanna. Ecology 88, 1119–1125 (2007).
68. Gignoux, J., Lahoreau, G., Julliard, R. & Barot, S. Establishment and early
persistence of tree seedlings in an annually burned savanna. J. Ecol. 97,
484–495 (2009).
69. Tizon, F. R., Pelaez, D. V. & Elia, O. R. e inuence of controlled
res on a plant community in the south of the Caldenal and its
relationship with a regional state and transition model. Int. J. Exp. Bot. 79,
141–146 (2010).
70. Neill, C., Patterson, W. A. & Crary, D. W. Responses of soil carbon, nitrogen
and cations to the frequency and seasonality of prescribed burning in a Cape
Cod oak–pine forest. Forest Ecol. Manag. 250, 234–243 (2007).
71. Ryan, C. M., Williams, M. & Grace, J. Above and belowground carbon
stocks in a miombo woodland landscape of Mozambique. Biotropica 43,
423–432 (2011).
72. Scharenbroch, B. C., Nix, B., Jacobs, K. A. & Bowles, M. L. Two decades of
low-severity prescribed re increases soil nutrient availability in a
midwestern, USA oak (Quercus) forest. Geoderma 183–184,
80–91 (2012).
73. Burton, J. A., Hallgren, S. W., Fuhlendorf, S. D. & Leslie, D. M. Jr. Understory
response to varying re frequencies aer 20 years of prescribed burning in an
upland oak forest. Plant Ecol. 212, 1513–1525 (2011).
74. Stewart, J. F., Will, R. E., Robertson, K. M. & Nelson, C. D. Frequent re
protects shortleaf pine (Pinus echinata) from introgression by loblolly pine
(P. taeda). Conserv. Genet. 16, 491–495 (2015).
75. Knapp, B. O., Stephan, K. & Hubbart, J. A. Structure and composition of an
oak–hickory forest aer over 60 years of repeated prescribed burning in
Missouri, U.S.A. Forest Ecol. Manag. 344, 95–109 (2015).
76. Olson, M. G. Tree regeneration in oak–pine stands with and without
prescribed re in the New Jersey Pine Barrens: management implications.
North. J. Appl. For. 28, 47–49 (2011).
A.F.A.P. was supported by a NOAA Climate and Global Change postdoctoral
fellowship programme and the USDA National Institute of Food and Agriculture
grant 20186701228077. R.B.J. received support from the Gordon and Betty Moore
Foundation. The experiments at the sites were organized and funded through the Cedar
Creek Long Term Ecological Research programme (DEB 1234162, 0620652, 1831944
and DBI 2021898), the National Park Service and Sequoia Parks Conservancy, and South
African National Parks. C.T. was supported by a Lawrence Fellow award supported by
the LLNL-LDRD Program under Project No. 20-ERD-055.
Author contributions
A.F.A.P. and R.B.J. conceived and designed the overall study. T.R., C.A. and C.T. helped with
data acquisition and provided feedback on statistical analyses. D.G.B., A.C.S., W.C., C.C.,
J.D.H., S.H.E., W.A.H., J.K., T.L., W.K.M., S.T.O., W.A.P., K.G.P., P.B.R., C.R., M.A.S.S., B.C.S.,
T.S., G.R.S., K.S., C.S., M.G.T. and J.M.V. provided data and/or assisted with interpreting the
field data from experiments. All authors contributed to the writing of the manuscript.
Competing interests
The authors declare no competing interests.
Additional information
Extended data is available for this paper at
Supplementary information The online version contains supplementary material
available at
Correspondence and requests for materials should be addressed to A.F.A.P.
Peer review information Nature Ecology & Evolution thanks Donald Falk and the other,
anonymous, reviewer(s) for their contribution to the peer review of this work.
Reprints and permissions information is available at
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
© The Author(s), under exclusive licence to Springer Nature Limited 2021
Articles NATurE EcoloGy & EVoluTioN
Articles NATurE EcoloGy & EVoluTioN
Extended Data Fig. 1 | Distribution of sites. a, Map displaying the distribution of sites (dots) with the surface fire sites filled with black and the crown
fire sites filled with white. The coloration indicates the average fire frequency within a gridcell using 1. The sample size of plots is written adjacent to
the continent. b, distribution of sites in climate space overlying Whittaker’s biome distribution 77. (1 = tundra, 2=boreal forest, 3=woodland/shrubland,
4=temperate grassland/desert, 5=temperate forest, 6=temperate rainforest, 7=subtropical desert, 8=tropical forest and savanna, 9=tropical rainforest).
Dots colored according to broad vegetation type category. Plots span a mean annual temperate range from 5.2–27.3 °C and a mean annual precipitation
range from 408–2378 mm yr1. c, aerial picture of two different fire treatment plots from Cedar Creek, a temperate oak savanna, where different fire
frequencies have created a stark biome boundary between forests in unburned plots and savannas in biennial burn plots.
NATurE EcoloGy & EVoluTioN Articles
NATurE EcoloGy & EVoluTioN
Extended Data Fig. 2 | Example of the experimental layout of a fire manipulation experiment. Example of the experimental layout of a fire manipulation
experiment taken from Cedar Creek (a temperate savanna in Minnesota, USA), where fires have been manipulated since 1964. Aerial imagery (taken in
2017) from the National Agriculture Imagery Program from the Farm Service Agency. Plots are outlined with a color corresponding to their fire frequencies
expressed in terms of number of fires per year (for example 0.33 is one fire every 3 years).
Articles NATurE EcoloGy & EVoluTioN
Articles NATurE EcoloGy & EVoluTioN
Extended Data Fig. 3 | Untransformed data on stem density and basal area. Untransformed data on stem density (a-b) and basal area (c-d) as a function
of the duration that plots have been exposed to burning in the experiment (0 = unburned plots). Each dot represents a site and the dashed lines connect
treatments within sites. Columns represent two sets of fire frequency contrasts comparing unburned vs. the intermediate frequency in a and c, and
unburned vs. the high frequency in b and d (levels defined based on treatments within sites). Dots and bars based on mean and standard error calculated
across the replicate plots within a fire treatment in a site. Note y-axis is on a log10 scale.
NATurE EcoloGy & EVoluTioN Articles
NATurE EcoloGy & EVoluTioN
Extended Data Fig. 4 | Comparison between fire types. Comparison between fire types (surface in a, F1,94.3= 50.6, p < 0.001, n = 9 sites and n = 104 plots;
and crown in b, F1,21= 10.3, p = 0.004, n = 24 plots) in needleleaf forests with fire expressed in terms of return period (crown fire plots are all 12 years
postfire, data subset to include short-interval burn plots). c, illustrates the mean response ratios ± standard error for the fire types with crown fires split
into high (>2,400 m) and low (<2,400) elevation sites (Crown 1 and Crown 2, respectively; n = 25 plots for each elevation category). Analyses were
robust to considering surface fires in only Western US needleleaf forests: F1, 47.1 = 17.2, p = 0.001. Response ratios were split into long and short fire return
interval plots (Crown 1 and 2, respectively), with the justification for definition of interval in 17.
Articles NATurE EcoloGy & EVoluTioN
Articles NATurE EcoloGy & EVoluTioN
Extended Data Fig. 5 | Fire frequency effects across precipitation in the wet quarter bins. Partial residual plot displaying the relationship between loge
basal area and precipitation in the wettest quarter cross-sectioned based on fire frequency. This plot is based on the same mixed-effects model presented
in Fig. 3 and Supplementary Table 4, just re-arranged to emphasize how precipitation-basal area relationship changes with more frequent burning. n = 25
sites and n = 309 plots.
NATurE EcoloGy & EVoluTioN Articles
NATurE EcoloGy & EVoluTioN
Extended Data Fig. 6 | Differences in fire effects across continents. Partial residual plot between the length of time plots were exposed to frequent
burning and the log basal area (a, n = 25 sites and n = 309 plots) and stem density (b, n = 25 sites and n = 303 plots) in the different continents (from the
main mixed-effects model with site as a random intercept in Supplementary Tables 4-5).
Articles NATurE EcoloGy & EVoluTioN
Articles NATurE EcoloGy & EVoluTioN
Extended Data Fig. 7 | Fire effects in sub-vegetation classifications. Partial residual plot between the length of time plots were exposed to frequent
burning and the log basal area in the different sub-vegetation types (from the main mixed-effects model, presented in Supplementary Table 4 but
substituting the broad vegetation effect with the more detailed classification. We found no evidence that accounting for the finer-scale variability in
ecosystem classification increased the accuracy of the model or changed our conclusions. n = 25 sites and n = 309 plots.
NATurE EcoloGy & EVoluTioN Articles
NATurE EcoloGy & EVoluTioN
Extended Data Fig. 8 | Responses of stem densities to fire across environmental covariates. Partial residual plots of the mixed-effects model for
stem densities illustrating how fire frequency effects changed according to wet-season precipitation, mean annual temperature, and ecosystem type.
Panels structured by standard deviations around the median to visualize the spread (1, 0,1), PWQ: precipitation in the wet quarter, MAT: mean annual
temperature. All model fits are p < 0.05 and specific results can be found in Supplementary Table 5. The predictor variables are mean-centered and
standard deviations are scaled to facilitate comparisons of variable influence. In needleleaf and broadleaf forests, stem densities actually increased with
more frequent burning initially, but declined with increasing experiment duration, potentially because of increased light availability initially stimulating
recruitment of small trees (Extended Data Fig. 6, Supplementary Table 5). Stem density in African sites changed little through time (Extended Data Fig. 6).
The trends in density may reflect the ability of many of the tree species to re-sprout in between fire events78. n = 25 sites and n = 303 plots.
Articles NATurE EcoloGy & EVoluTioN
Articles NATurE EcoloGy & EVoluTioN
Extended Data Fig. 9 | Effect of fire on phosphorus stoichiometry. Partial residual plots of the phosphorus (P) stoichiometry of community weighted
means as a function of years of repeated burning. Taken from mixed-effects models presented in Supplementary Table 7. The models include a vegetation
type effect. Tissue P is rescaled by subtracting the mean and dividing by the standard deviation. n = 16 sites and n = 172 plots.
nature research | reporting summary April 2020
Corresponding author(s): Adam F. A. Pellegrini
Last updated by author(s): 12/14/2020
Reporting Summary
Nature Research wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency
in reporting. For further information on Nature Research policies, see our Editorial Policies and the Editorial Policy Checklist.
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.
n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement
A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly
The statistical test(s) used AND whether they are one- or two-sided
Only common tests should be described solely by name; describe more complex techniques in the Methods section.
A description of all covariates tested
A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons
A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient)
AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals)
For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted
Give P values as exact values whenever suitable.
For Bayesian analysis, information on the choice of priors and Markov chain Monte Carlo settings
For hierarchical and complex designs, identification of the appropriate level for tests and full reporting of outcomes
Estimates of effect sizes (e.g. Cohen's d, Pearson's r), indicating how they were calculated
Our web collection on statistics for biologists contains articles on many of the points above.
Software and code
Policy information about availability of computer code
Data collection We did not develop code to collect data, we used field survey data and existing databases.
Data analysis We conducted all analyses in the programming language R with separate packages (all published) for specific analyses not included in the base
statistics package. GitHub will be used to deposit the final code upon publication.
For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors and
reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Research guidelines for submitting code & software for further information.
Policy information about availability of data
All manuscripts must include a data availability statement. This statement should provide the following information, where applicable:
- Accession codes, unique identifiers, or web links for publicly available datasets
- A list of figures that have associated raw data
- A description of any restrictions on data availability
Data will be published as a dataset alongside the manuscript. All data used to create the figures will be freely available.
nature research | reporting summary April 2020
Field-specific reporting
Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.
Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see
Ecological, evolutionary & environmental sciences study design
All studies must disclose on these points even when the disclosure is negative.
Study description We quantified the effects of fire frequency on tree cover across broad biogeographic and climatic scales, incorporating additional
factors that may explain variability in fire effects on tree communities. We used original survey data collected across 29 repeated
burn experiments within a large collaborative network. We also compiled data on plant functional traits within North American fire
manipulation experiments to test how traits could predict the effects of fire as well as how fire changed the functional composition
of the plant community.
Research sample We combined published data with new surveys on tree populations from 29 sites and 374 plots in four continents. At 27 of the sites
(324 plots), surface fire frequency was experimentally manipulated for 16-64 years (mean of 30 years), and at two sites (50 plots),
natural variation in crown fire frequency presented a natural experiment. We focused on tree responses because of their importance
for carbon storage, ecosystem productivity and nutrient cycling23,27,28. The sites cover locations in North America (coniferous,
deciduous, and mixed forests, broadleaf savanna), South America (tropical savanna and temperate shrubland), Africa (broadleaf
savanna), and Australia (wet and dry eucalyptus forests) all of which are ecosystems that experience frequent burning (Figure S1,
Table S1, Supplemental Information, SI). For our main analysis, we define ecosystems broadly based on dominant fuel type (grass vs.
litter) and tree functional composition (angiosperm broadleaf vs. gymnosperm needleleaf trees) but test the robustness of our
conclusions to other classifications. Each surface fire site contains replicate plots including an unburned treatment and different
prescribed burning frequencies (Figure S2), where fire frequencies ranged from approximately one fire every decade to one fire
every year (Table S1). We evaluated the effects of fire alone and in combination with environmental covariates using model selection
(SI) to test the importance of climate, soil, and species composition in modifying the effect of fire.
Sampling strategy Our study contains two levels of relevant sampling strategies. First, the level of choosing the sites included in our analysis. This paper
was a result of a large collaborative group that has communicated through a series of meetings with the 1st author Adam Pellegrini.
This study is not an exhaustive meta-analysis of the literature. Clearly, not every single fire manipulation experiment in the world is
included in this study, partly because in some cases, organizers of fire manipulation experiments declined to be a part of the study. In
other cases, organizers never responded or did not have raw data necessary for our analysis.
The second level of sampling included collecting data on the tree communities and soils. Transects to survey trees and collect soils
were randomly established in the plots. Survey area was generally designed to capture the ecological heterogeneity within a
landscape (e.g., larger surveys in areas with patchy tree cover such as savannas). All functional trait data were acquired from the
most extensive trait databases currently freely accessible.
Data collection We provide information on the experimental designs in Table S1 because each site varied (we also include the corresponding citation
that describes each experiment in greater detail). This table includes a list of sites with key meta-data such as the continent,
vegetation type present in broad categories and the families of the dominant tree species. Number of plots is the total within the
entire site. Duration is the number of years over which fire frequencies have differed across plots. We also describe the prior
conditions of the ecosystem at the beginning of the experiment (e.g. whether the site experienced regular burning prior to the
Soil data were collected using previously published values, unpublished values, or our own sampling and analyses. The types of data
for each site are described in the section "Soil chemistry data".
Plant functional trait data were compiled from published datasets, with the appropriate citations given in the text. The section
describing these data is "Species classifications and functional traits".
Timing and spatial scale Sampling was conducted over the course of several decades because the timing that the fire manipulation experiments were
surveyed differed across sites. The plots were relatively large (>0.5 hectares, and in some cases up to 7 hectares in size). The tree
community surveys attempted to account for ecological heterogeneity within the plots.
Data exclusions In certain cases, experiments contained multiple surveys through time and treatments in different seasons. These are described in
the section "Choice of plots within 27 sites with surface fires" in our Supplemental Information section. Our decision to exclude plots
was conservative because we wanted to avoid pseudoreplication arising from repeated measures on the same fire treatment in
different years within a site. We avoided using different seasons of burns within the same site because the range of fire conditions in
the seasons varied considerably in some sites (e.g., burned anytime between March - June).
Reproducibility These experiments are large-scale fire manipulations at the ecosystem level for multiple decades. The reproducibility of the
experimental findings from a single study is partly captured by comparing the effects of fire from multiple independent experiments.
Randomization Replicate plots within each experiment were randomized at the onset of the experiment.
Blinding Blinding was not consistently used during data collection but we did randomize the location of the transects within a plot.
nature research | reporting summary April 2020
Did the study involve field work? Yes No
Field work, collection and transport
Field conditions Fieldwork was generally conducted in the growing season, which allowed for the proper identification of tree species. Weather
conditions during the field season varied considerably across sites given their broad range in climates.
Location All experimental locations are provided in Table S1. We provide a broad description of the sites as well as the original references that
describe the experimental design of each site.
Access & import/export We did not conduct any importing or exporting of samples requiring permits.
Disturbance There was no disturbance of the sites. Soil cores are minimal and not considered destructive.
Reporting for specific materials, systems and methods
We require information from authors about some types of materials, experimental systems and methods used in many studies. Here, indicate whether each material,
system or method listed is relevant to your study. If you are not sure if a list item applies to your research, read the appropriate section before selecting a response.
Materials & experimental systems
n/a Involved in the study
Eukaryotic cell lines
Palaeontology and archaeology
Animals and other organisms
Human research participants
Clinical data
Dual use research of concern
n/a Involved in the study
Flow cytometry
MRI-based neuroimaging
... However, the impact of encroachment will vary among species due to their functional traits and relative sensitivity to changing environmental conditions (Ding et al., 2020;Eldridge & Ding, 2021). Hence, understanding how encroachment relates to ecosystem function is contingent on understanding how functional diversity and demographic change are interrelated (Pellegrini et al., 2021). ...
... These data could lead to improvements in ecosystem management and public policies, providing insights into which species should be removed or maintained to reach the aimed ecosystem service in restoration and conservation plans (Ding et al., 2020;Eldridge & Ding, 2021). Surprisingly, most temporal analyses of woody encroachment on vegetation focus on structure and taxonomic diversity (Moreira, 2000;Abreu et al., 2017;Maracahipes-Santos et al., 2018), and few studies look at the changes in functional traits and diversity over time (Pellegrini et al., 2021). The demographic processes of such functional change should be driven by the growth of resident tree species and the recruitment of new individuals with different ensembles of traits (Brudvig et al., 2011;Passos et al., 2018). ...
Woody encroachment in savannas has been associated with changing taxonomic composition and ecosystem function. Interestingly, there is little understanding of how encroachment impacts plant functional diversity and how those changes relate to plant demography, a crucial mediator between taxonomic composition and ecosystem function. Southeastern Brazil. Using a landscape scale fire suppression experiment in a diverse Brazilian savanna, we quantify how change in species composition over seven years impacted vegetative and reproductive tree functional diversity as determined by new recruits, dead and surviving trees. Over seven years, tree aboveground biomass increased by 15%, while total species richness did not change. Despite minor changes, species composition remained overall similar (82%), with few species contributing significantly to plot dissimilarity over time. There were small changes in vegetative traits, where the community weighted mean increased in maximum tree height (↑ 2.1%) and specific leaf area (↑ 5.3%), and decreased in wood density (↓ 1.3%) and bark thickness (↓ 9.4%). Changes in reproductive traits were larger than in vegetative traits, with an increase in the prevalence of monoecy (↑ 32.6%), dioecy (↑ 44.2%), large seeds (↑ 20.3%), animal mediated seed dispersal (↑ 4.9%) and pollination by very small insects (↑ 45.5%), and a decrease in the prevalence of hermaphroditism (↓ 9%), small seeds (6.8%) and pollination by small insects (12.5%). The overall decrease in bark thickness and increase in monoecy and dioecy were mainly driven by characters of the new recruits, while the overall increase in SLA and decrease in small seeds appeared largely determined by the loss of trees possessing those traits. Encroachment leads to changes that are likely increasing ecosystem vulnerability to fire and drought. Further, the compositional changes observed appear to drive marked change in reproductive traits, indicating increasing dependence on animals for dispersal and reproduction. Understanding post‐hoc encroachment impacts in an era of widespread pervasive encroachment is fundamental to reconciling ecosystems functions such as nutrient cycling and pollination services as there is a loss of species with open ecosystem life history strategies. Among savannas, there remains an urgent need to understand relationships between woody cover and ecosystem function to determine thresholds in woody cover promoting resilient savanna ecosystems.
... However, predictable shifts in traits following increased disturbance frequency and/or intensity are not always observed (Cavender-Bares & Reich, 2012;Pellegrini et al., 2021;Reich et al., 1995). The broader literature on plant traits proposes several processes that mediate the magnitude of trait responses to environmental gradients: (i) the magnitude of environmental change determines the magnitude of trait changes, (ii) the plant lineages present are constrained in their functional variation, and (iii) physiological trade-offs between traits that constrain the ability to optimise to new environmental conditions (Ackerly, 2004;Ackerly et al., 2000;Cavender-Bares et al., 2004;Cornwell & Ackerly, 2009). ...
... Fire in savannas has two key effects on plant resources-increasing light availability by reducing tree cover and decreasing N availability via N volatilisation during combustion-that could change plant SLA and leaf N (Dantas et al., 2013;Hoffmann et al., 2012;Pellegrini et al., 2018Pellegrini et al., , 2021Reich et al., 2001). However, trait responses to fire that are driven by species turnover and/or replacement versus those driven by within-species responses such as phenotypic plasticity may either reinforce or counteract each other. ...
Full-text available
The functional response of plant communities to disturbance is hypothesised to be controlled by changes in environmental conditions and evolutionary history of species within the community. However, separating these influences using direct manipulations of repeated disturbances within ecosystems is rare. We evaluated how 41 years of manipulated fire affected plant leaf economics by sampling 89 plant species across a savanna-forest ecotone. Greater fire frequencies created a high-light and low-nitrogen environment, with more diverse communities that contained denser leaves and lower foliar nitrogen content. Strong trait–fire coupling resulted from the combination of significant intraspecific trait–fire correlations being in the same direction as interspecific trait differences arising through the turnover in functional composition along the fire-frequency gradient. Turnover among specific clades helped explain trait–fire trends, but traits were relatively labile. Overall, repeated burning led to reinforcing selective pressures that produced diverse plant communities dominated by conservative resource-use strategies and slow soil nitrogen cycling.
... In drier and warmer ecosystems, which dominate global burned area, most SOC is in the mineral horizon where heat rapidly dissipates 18 and little direct combustion of SOC occurs 16,19,20 . In these drier sites, re-driven shifts in plant biomass inputs, especially from trees [21][22][23] , are thought to determine changes in SOC stored in the mineral horizon [24][25][26] . ...
Full-text available
Widespread changes in the intensity and frequency of fires across the globe are altering the terrestrial carbon (C) sink 1–4 . Although the changes in ecosystem C have been reasonably well quantified for plant biomass pools 5–7 , an understanding of the determinants of fire-driven changes in soil organic C (SOC) across broad environmental gradients remains unclear, especially in global drylands 3,4,7–9 . Here, we combined multiple datasets and original field sampling of fire manipulation experiments to evaluate where and why fire changes SOC the most, built a statistical model to estimate historical changes in SOC, and compared these estimates to simulations from ecosystem models. We found that drier ecosystems experienced larger relative changes in SOC than humid ecosystems—in some cases exceeding losses from plant biomass pools—primarily explained by high fire-driven declines in tree biomass inputs in dry ecosystems. Ecosystem models provided more mixed insight into potential SOC changes because many models underestimated the SOC changes in drier ecosystems. Upscaling our statistical model predicted that soils in 1.57 million km 2 savanna-grassland regions experiencing declines in burned area over the past ca. two decades may have 23% more SOC, equating to 1.78 PgC in topsoils. Consequently, ongoing declines in fire frequencies have likely created an extensive carbon sink in the soils of global drylands that may have been underestimated by ecosystem models.
... ground-, air-, and space-borne lidars, radars, [hyperspectral] spectrometers, and [multispectral] radiometers) would enable measurements to help characterize surface and atmospheric structure and chemistry and better understand human land cover and land use in conjunction with fire impacts on air and water quality, ecosystems, and energy balance. We must use molecular techniques to capture the direct and indirect effects of soil heating on soil organic matter composition (132), belowground biological communities (133,134), organism physiology (135), and ecosystem function processes (136). Finally, laboratory work can help better understand the mechanisms of heat transfer (137,138), firebrand ember generation, behavior and transport (139,140), atmospheric emissions (141), and transformation of fire plumes (115). ...
Fire is an integral component of ecosystems globally and a tool that humans have harnessed for millennia. Altered fire regimes are a fundamental cause and consequence of global change, impacting people and the biophysical systems on which they depend. As part of the newly emerging Anthropocene, marked by human-caused climate change and radical changes to ecosystems, fire danger is increasing, and fires are having increasingly devastating impacts on human health, infrastructure, and ecosystem services. Increasing fire danger is a vexing problem that requires deep transdisciplinary, trans-sector, and inclusive partnerships to address. Here, we outline barriers and opportunities in the next generation of fire science and provide guidance for investment in future research. We synthesize insights needed to better address the long-standing challenges of innovation across disciplines to (i) promote coordinated research efforts; (ii) embrace different ways of knowing and knowledge generation; (iii) promote exploration of fundamental science; (iv) capitalize on the "firehose" of data for societal benefit; and (v) integrate human and natural systems into models across multiple scales. Fire science is thus at a critical transitional moment. We need to shift from observation and modeled representations of varying components of climate, people, vegetation, and fire to more integrative and predictive approaches that support pathways toward mitigating and adapting to our increasingly flammable world, including the utilization of fire for human safety and benefit. Only through overcoming institutional silos and accessing knowledge across diverse communities can we effectively undertake research that improves outcomes in our more fiery future.
... A single stressor-such as fire used to clear land-can rapidly alter the stable state of forest to that of savanna. Although both savanna and seasonal forest plants can cope with drought, differences in fire regimes between these two ecosystems mean they support fundamentally different species (Pellegrini et al. 2021). Seasonal forest plants tend to be highly intolerant to fire, meaning that any switches to savanna are less reversible, posing a great risk in the highly threatened seasonal forests of Nusa Tenggara. ...
Full-text available
Wallacea—the meeting point between the Asian and Australian fauna—is one of the world's largest centers of endemism. Twenty-three million years of complex geological history have given rise to a living laboratory for the study of evolution and biodiversity, highly vulnerable to anthropogenic pressures. In the present article, we review the historic and contemporary processes shaping Wallacea's biodiversity and explore ways to conserve its unique ecosystems. Although remoteness has spared many Wallacean islands from the severe overexploitation that characterizes many tropical regions, industrial-scale expansion of agriculture, mining, aquaculture and fisheries is damaging terrestrial and aquatic ecosystems, denuding endemics from communities, and threatening a long-term legacy of impoverished human populations. An impending biodiversity catastrophe demands collaborative actions to improve community-based management, minimize environmental impacts, monitor threatened species, and reduce wildlife trade. Securing a positive future for Wallacea's imperiled ecosystems requires a fundamental shift away from managing marine and terrestrial realms independently.
Forests influence climate and mitigate global change through the storage of carbon in soils. In turn, these complex ecosystems face important challenges, including increases in carbon dioxide, warming, drought and fire, pest outbreaks and nitrogen deposition. The response of forests to these changes is largely mediated by microorganisms, especially fungi and bacteria. The effects of global change differ among boreal, temperate and tropical forests. The future of forests depends mostly on the performance and balance of fungal symbiotic guilds, saprotrophic fungi and bacteria, and fungal plant pathogens. Drought severely weakens forest resilience, as it triggers adverse processes such as pathogen outbreaks and fires that impact the microbial and forest performance for carbon storage and nutrient turnover. Nitrogen deposition also substantially affects forest microbial processes, with a pronounced effect in the temperate zone. Considering plant-microorganism interactions would help predict the future of forests and identify management strategies to increase ecosystem stability and alleviate climate change effects. In this Review, we describe the impact of global change on the forest ecosystem and its microbiome across different climatic zones. We propose potential approaches to control the adverse effects of global change on forest stability, and present future research directions to understand the changes ahead.
Full-text available
Predicting how belowground carbon storage reflects changes in aboveground vegetation biomass is an unresolved challenge in most ecosystems. This is especially true for fire-prone savannas, where frequent fires shape the fraction of carbon allocated to root traits for post-fire vegetation recovery. Here I review evidence on how root traits may respond to frequent fires and propose to leverage root traits to infer belowground carbon dynamics in fire-prone savannas. Evidently, we still lack an understanding of trade-offs in root acquisitive vs. conservative traits in response to frequent fires, nor have we determined which root traits are functionally important to mediate belowground carbon dynamics in a frequently burned environment. Focusing research efforts along these topics should improve our understanding of savanna carbon cycling under future changes in fire regimes.
Full-text available
Although the importance of natural habitats to pollinator diversity is widely recognized, the value of forests to pollinating insects has been largely overlooked in many parts of the world. In this review, we (i) establish the importance of forests to global pollinator diversity, (ii) explore the relationship between forest cover and pollinator diversity in mixed-use landscapes, and (iii) highlight the contributions of forest-associated pollinators to pollination in adjacent crops. The literature shows unambiguously that native forests support a large number of forest-dependent species and are thus critically important to global pollinator diversity. Many pollinator taxa require or benefit greatly from resources that are restricted to forests, such as floral resources provided by forest plants (including wind-pollinated trees), dead wood for nesting, tree resins, and various non-floral sugar sources (e.g. honeydew). Although landscape-scale studies generally support the conclusion that forests enhance pollinator diversity, findings are often complicated by spatial scale, focal taxa, landscape context, temporal context, forest type, disturbance history, and external stressors. While some forest loss can be beneficial to pollinators by enhancing habitat complementarity, too much can result in the near-elimination of forest-associated species. There is strong evidence from studies of multiple crop types that forest cover can substantially increase yields in adjacent habitats, at least within the foraging ranges of the pollinators involved. The literature also suggests that forests may have enhanced importance to pollinators in the future given their role in mitigating the negative effects of pesticides and climate change. Many questions remain about the amount and configuration of forest cover required to promote the diversity of forest-associated pollinators and their services within forests and in neighbouring habitats. However, it is clear from the current body of knowledge that any effort to preserve native woody habitats, including the protection of individual trees, will benefit pollinating insects and help maintain the critical services they provide.
The objective of this study was to evaluate the effects of wildfires on soil nitrogen (N) content and forms in a cold temperate coniferous forest in the Daxing’an Mountains (NE China) 9 years after the fire. In June, August, and October 2017, soil samples were collected from 0 to 5 cm layer under Ledum palustre-Larix gmelinii, Rhododendron dauricum-Larix gmelinii, and Pinus sylvestris var. mongolica forests. Then, the contents of both mineral N and available N were analyzed and used to calculate the different ratios as indicators of N forms. The results showed that the responses of the mineral N content in the three forest types to wildfire had been exhausted, while wildfire had a significant effect on the soil AAN content in the RL and PM forests, and there was no specific effect on this variable in the LL forest. During this period, the seasonal dynamics of the soil mineral N content and AAN content differed with vegetation, topography, climate, and other factors. Wildfire did not change the form of mineral N, which means that NH4+ was the predominant form in the soil of the three forest types; however, changes in the form of available N at key time points were driven by changes in AAN content after wildfire. Based on the availability of AAN in conifer forests, the content and forms of soil N appear to be favorable for vegetation restoration 9 years after wildfire; however, little is known about AAN in the soil immediately after wildfire, suggesting that additional research is urgently needed.
Full-text available
Frequent forest fires cause air pollution, threaten biodiversity and spoil forest ecosystems. Forest fire vulnerability assessment is a potential way to improve the ability of forests to resist climate disasters and help formulate appropriate forest management countermeasures. Here, we developed an automated hybrid machine learning algorithm by selecting the optimal model from 24 models to map potential forest fire vulnerability over China during the period 2001–2020. The results showed forest aboveground biomass (AGB) had a vulnerability of 26%, indicating that approximately 2.32 Gt C/year of forest AGB could be affected by fire disturbances. The spatiotemporal patterns of forest fire vulnerability were dominated by both forest characteristics and climate conditions. Hotspot regions for vulnerability were mainly located in arid areas in western China, mountainous areas in southwestern China, and edges of vegetation zones. The overall forest fire vulnerability across China was insignificant. The forest fire vulnerability of boreal and temperate coniferous forests and mixed forests showed obviously decreasing trends, and cultivated forests showed an increasing trend. The results of this study are expected to provide important support for the forest ecosystem management in China.
Full-text available
Fires shape the biogeochemistry and functioning of many ecosystems, and fire frequencies are changing across much of the globe. Frequent fires can change soil carbon (C) and nitrogen (N) storage by altering the quantity and chemistry of plant inputs through changes in plant biomass and composition as well as altering decomposition of soil organic matter. How decomposition rates change with shifting inputs remains uncertain because most studies focus on the effects of single fires, where transient changes may not reflect responses to decadal changes in burning frequencies. Here, we sampled seven sites exposed to different fire frequencies. In four of the sites, we intensively sampled both soils and plant communities across four ecosystems in North America and Africa spanning tropical savanna, temperate coniferous savanna, temperate broadleaf savanna, and temperate coniferous forest ecosystems. Each site contained multiple plots burned frequently for 33‐61 years and nearby plots that had remained unburned over the same period replicated at the landscape scale. Across all sites, repeatedly burned plots had 25‐185% lower bulk soil C and N concentrations but also 2‐10‐fold lower potential decomposition of organic matter compared to unburned sites. Soil C and N concentrations and extracellular enzyme activities declined with frequent fire because fire reduced both plant biomass inputs into soils and dampened the localized enrichment effect of tree canopies. Examination of soil extracellular enzyme activities revealed that fire decreased the potential turnover of organic matter in the forms of cellulose, starch, and chitin (p<0.0001) but not polyphenol and lignin (p=0.09), suggesting a shift in soil C and N cycling. Inclusion of δ13C data from three additional savanna sites (19‐60 years of altered fire frequencies) showed that soil C losses were largest in sites where estimated tree inputs into soils declined the most (r2=0.91, p<0.01). In conclusion, repeated burning reduced C and N storage, consistent with previous studies, but fire also reduced potential decomposition, likely contributing to slower C and N cycling. Trees were important in shaping soil carbon responses across sites, but the magnitude of tree effects differed and depended on how tree biomass inputs into soil responded to fire.
Full-text available
The size and frequency of large wildfires in western North America have increased in recent years, a trend climate change is likely to exacerbate. Due to fuel limitations, recently burned forests resist burning for upwards of 30 years; however, extreme fire-conducive weather enables reburning at shorter fire-free intervals than expected. This research quantifies the outcomes of short-interval reburns in upland and wetland environments of northwestern Canadian boreal forests and identifies an interactive effect of post-fire drought. Despite adaptations to wildfire amongst boreal plants, post-fire forests at paired short- and long-interval sites were significantly different, with short-interval sites having lower stem densities of trees due to reduced conifer recruitment, a higher proportion of broadleaf trees, less residual organic material, and reduced herbaceous vegetation cover. Drought reinforced changes in proportions of tree species and decreases in tree recruitment, reinforcing non-resilient responses to short-interval reburning. Drier and warmer weather will increase the incidence of short-interval reburning and amplify the ecological changes such events cause, as wildfire activity and post-fire drought increase synergistically. These interacting disturbances will accelerate climate-driven changes in boreal forest structure and composition. Our findings identify processes of ongoing and future change in a climate-sensitive biome.
Full-text available
Increasingly frequent and severe droughts under climate change are expected to have major impacts on vegetation worldwide. However, research to date has focused on tree vulnerability to drought in forests. Less is known about trees and drought in savannas, where a sparse tree layer coexists with grass. These tree-grass interactions (often mediated by fire and herbivory) shape savanna tree ecology, and confound predictions of how strongly drought might affect trees. On the one hand, drought is physiologically stressful, which could harm trees and be exacerbated by herbivore impacts; on the other hand, trees adapted to semiarid savannas might be relatively drought tolerant, and the considerable impacts of drought on grass could even benefit trees via reduced grass competition and fire risk, especially in the year following a drought. Here, we sought to understand the net effects of severe drought on the savanna tree layer, and how fire and herbivory mediate these effects. We monitored tree growth, mortality, and community structure for 2 yr within existing long-term fire and herbivory experiments across a drought-severity contrast, following a major drought in Kruger National Park, South Africa. Overall, severe drought was a major stressor for trees. Tree mortality rates in most species increased by an order of magnitude in the year following drought, and slower growth rates for some persisted for 2 yr. At the community level, this translated into substantial decreases in tree densities. Herbivory and fire did little either to mitigate or exacerbate drought effects on trees, and overall, drought swamped effects of herbivory and fire that have otherwise been observed. However, species differed in their responses to drought, with some dominant encroaching species especially vulnerable. We suggest that increasing drought frequency and severity could drastically alter savanna vegetation by repeatedly killing off trees.
Full-text available
A spatially explicit global map of tree symbioses with nitrogen-fixing bacteria and mycorrhizal fungi reveals that climate variables are the primary drivers of the distribution of different types of symbiosis.
Frequent fire often has a negative impact of tree recruitment and growth. Tree growth rates, density and recruitment were compared among treatments of annual burning since 1952, triennial burning since 1973 and no burning (1946 to 1996) or single wildfire (1996 to 2018), in a dry sclerophyll eucalypt forest, south-eastern Queensland, Australia. Tree diameter (at breast height, DBH) growth rates were greater in the annually burnt treatment than in the triennially burn and single wildfire treatments over the period from 1974 to 2018, and these differences were also apparent pre-wildfire (period from 1974 to 1996). In the period from 1996 to 2018, the annually burnt treatment had greater DBH growth relative to the single wildfire treatment, but the triennial treatment had intermediate growth rates. Competitive interactions between trees (assessed using plot basal area) also had a negative impact on individual tree growth rates. The impacts of different fire regimes at this site on tree crown health were not apparent (P > 0.05) and there was only limited evidence that differences in growth rates were due to differences in soil nutrients (marginally higher topsoil phosphorus in the frequently burnt treatments, P = 0.075). Greater tree growth rates in the annually burnt treatment may be related to the lower density of understorey woody plants in this treatment and potentially reduced competition for soil moisture. The density of trees (DBH ≥ 10 cm) in 2018 was surprisingly higher in the triennially burnt treatment (381 stems/ha) relative to both the annually burnt (192 stems/ha) and single wildfire (234 stems/ha) treatments. This was largely due a higher level of recruitment over time and a higher density of stems 10–20 cm DBH in triennially burnt plots. Concerns regarding the impacts of frequent prescribed fire on tree recruitment and growth may be unwarranted in these remarkably resilient dry eucalypt forests.
Mycorrhizal fungi are critical members of the plant microbiome, forming a symbiosis with the roots of most plants on Earth. Most plant species partner with either arbuscular or ectomycorrhizal fungi, and these symbioses are thought to represent plant adaptations to fast and slow soil nutrient cycling rates. This generates a second hypothesis, that arbuscular and ectomycorrhizal plant species traits complement and reinforce these fungal strategies, resulting in nutrient acquisitive vs. conservative plant trait profiles. Here we analyzed 17,764 species level trait observations from 2,940 woody plant species to show that mycorrhizal plants differ systematically in nitrogen and phosphorus economic traits. Differences were clearest in temperate latitudes, where ectomycorrhizal plant species are more nitrogen use- and phosphorus use-conservative than arbuscular mycorrhizal species. This difference is reflected in both aboveground and belowground plant traits and is robust to controlling for evolutionary history, nitrogen fixation ability, deciduousness, latitude, and species climate niche. Furthermore, mycorrhizal effects are large and frequently similar to or greater in magnitude than the influence of plant nitrogen fixation ability or deciduous vs. evergreen leaf habit. Ectomycorrhizal plants are also more nitrogen conservative than arbuscular plants in boreal and tropical ecosystems, although differences in phosphorus use are less apparent outside temperate latitudes. Our findings bolster current theories of ecosystems rooted in mycorrhizal ecology and support the hypothesis that plant mycorrhizal association is linked to the evolution of plant nutrient economic strategies.
Future changes in climate are widely anticipated to increase fire frequency, particularly in boreal forests where extreme warming is expected to occur. Feedbacks between vegetation and fire may modify the direct effects of warming on fire activity and shape ecological responses to changing fire frequency. We investigate these interactions using extensive field data from the Boreal Shield of Saskatchewan, Canada, a region where >40% of the forest has burned in the past 30 years. We use geospatial and field data to assess the resistance and resilience of eight common vegetation states to frequent fire by quantifying the occurrence of short‐interval fires and their effect on recovery to a similar vegetation state. These empirical relationships are combined with data from published literature to parameterize a spatially explicit, state‐and‐transition simulation model of fire and forest succession. We use this model to ask if and how: (1) feedbacks between vegetation and wildfire may modify fire activity on the landscape and (2) more frequent fire may affect landscape forest composition and age structure. Both field and GIS data suggest the probability of fire is low in the initial decades after fire, supporting the hypothesis that fuel accumulation may exert a negative feedback on fire frequency. Field observations of pre‐ and post‐fire composition indicate that switches in forest state are more likely in conifer stands that burn at a young age, supporting the hypothesis that resilience is lower in immature stands. Stands dominated by deciduous trees or jack pine were generally resilient to fire, while mixed conifer and well‐drained spruce forests were less resilient. However, simulation modeling suggests increased fire activity may result in large changes in forest age structure and composition, despite the feedbacks between vegetation‐fire likely to occur with increased fire activity. This article is protected by copyright. All rights reserved.