Content uploaded by Brian J. Enquist
Author content
All content in this area was uploaded by Brian J. Enquist on Dec 10, 2017
Content may be subject to copyright.
PRIMARY RESEARCH ARTICLE
Fire effects and ecological recovery pathways of tropical
montane cloud forests along a time chronosequence
Imma Oliveras
1
|
Rosa M. Rom
an-Cuesta
2
|
Erickson Urquiaga-Flores
3
|
Jose A. Quintano Loayza
4
|
Jose Kala
4
|
Vicky Huam
an
4
|
Nohemi Liz
arraga
4
|
Guissela Sans
4
|
Katia Quispe
4
|
Efrain Lopez
4
|
David Lopez
4
|
Israel Cuba Torres
4
|
Brian J. Enquist
5
|
Yadvinder Malhi
1
1
Environmental Change Institute, School of
Geography and the Environment, University
of Oxford, Oxford, UK
2
Laboratory of Geo-Information and
Remote Sensing, Wageningen University
and Research, Wageningen, The
Netherlands
3
Institute of Systematic and Evolutionary
Botany, University of Zurich, Z€
urich,
Switzerland
4
Universidad de Santo Antonio Abad del
Cusco, Cusco, Per
u
5
Department of Ecology and Evolutionary
Biology, University of Arizona, Tucson,
Arizona
Correspondence
Imma Oliveras, Environmental Change
Institute, School of Geography and the
Environment, University of Oxford, Oxford,
UK.
Email: imma.oliveras@ouce.ox.ac.uk
Funding information
Natural Environment Research Council,
Grant/Award Number: NE/G006385/1;
National Science Foundation, Grant/Award
Number: 106586, DEB 1457812; Jackson
Foundation and a European Research Grant
Advanced Investigator Award GEM-TRAIT,
Grant/Award Number: ACRYQY00
Abstract
Tropical montane cloud forests (TMCFs) harbour high levels of biodiversity and large
carbon stocks. Their location at high elevations make them especially sensitive to cli-
mate change, because a warming climate is enhancing upslope species migration, but
human disturbance (especially fire) may in many cases be pushing the treeline downs-
lope. TMCFs are increasingly being affected by fire, and the long-term effects of fire
are still unknown. Here, we present a 28-year chronosequence to assess the effects
of fire and recovery pathways of burned TMCFs, with a detailed analysis of carbon
stocks, forest structure and diversity. We assessed rates of change of carbon (C) stock
pools, forest structure and tree-size distribution pathways and tested several
hypotheses regarding metabolic scaling theory (MST), C recovery and biodiversity.
We found four different C stock recovery pathways depending on the selected C pool
and time since last fire, with a recovery of total C stocks but not of aboveground C
stocks. In terms of forest structure, there was an increase in the number of small
stems in the burned forests up to 5–9 years after fire because of regeneration pat-
terns, but no differences on larger trees between burned and unburned plots in the
long term. In support of MST, after fire, forest structure appears to approximate
steady-state size distribution in less than 30 years. However, our results also provide
new evidence that the species recovery of TMCF after fire is idiosyncratic and follows
multiple pathways. While fire increased species richness, it also enhanced species dis-
similarity with geographical distance. This is the first study to report a long-term
chronosequence of recovery pathways to fire suggesting faster recovery rates than
previously reported, but at the expense of biodiversity and aboveground C stocks.
KEYWORDS
carbon allocation, forest structure, metabolic scaling theory, regeneration, species diversity
1
|
INTRODUCTION
Tropical montane cloud forests (TMCFs) represent around 14% of all
tropical forests, with an estimated area of 2.2 Mkm
2
between
23.5°N and 35°S (Mulligan, 2010). They are characterized by a
unique set of biological features including being one of the world’s
most diverse ecosystems with high levels of endemism and many
critically endangered species (Myers, Mittermeier, Mittermeier, da
Fonseca, & Kent, 2000). They are also carbon rich and store high
amounts of carbon in their forest biomass and soils (Girardin et al.,
Received: 3 July 2017
|
Accepted: 19 September 2017
DOI: 10.1111/gcb.13951
Glob Change Biol. 2017;1–15. wileyonlinelibrary.com/journal/gcb ©2017 John Wiley & Sons Ltd
|
1
2010, 2013; Zimmermann et al., 2010). In the tropical Andes, many
TMCFs sit on organic soils (ca. 360,000 km
2
) (Rom
an-Cuesta et al.,
2014) with carbon stocks over the range of 137–285 MgC ha
1
(Oliveras, Girardin, et al., 2014; Zimmermann et al., 2009). The upper
limit of the Andean TMCFs forms a treeline with high-altitude
grasslands (puna or p
aramo ecosystems) that is a zone of ecological
tension where climate and fires play opposite roles. While there is
evidence of TMCF species migrating upslope (Feeley et al., 2011)
due to increases in mean annual temperature (0.11°C per decade
since 1950s, (Urrutia & Vuille, 2009), the treeline has abundant pres-
ence of fires (Oliveras, Anderson, & Malhi, 2014; Sarmiento, 2006;
Young & Le
on, 2007). Treeline fires are increasingly penetrating into
the Andean TMCFs as they tend to burn downslope from the grass-
land ecosystems through the treeline (Oliveras, Girardin, et al., 2014;
Rom
an-Cuesta et al., 2011).
Forest fires are one of the most important drivers of tropical
forest degradation (Arag~
ao & Shimabukuro, 2010; van Marle et al.,
2016), and are currently one of the main sources of carbon emis-
sions in many tropical regions (de Andrade et al., 2017; Van Der
Werf et al., 2010). Fire-driven tropical forest degradation causes
major alterations to ecosystems composition, structure and function
(Barlow & Peres, 2008; Malhi, Gardner, Goldsmith, Silman, &
Zelazowski, 2014) and represents a major threat to global biodiver-
sity (e.g. Peres et al., 2010) and to ecosystem services, such carbon
storage and climate regulation (Berenguer et al., 2014; Malhi et al.,
2014).
Currently, most forest degradation studies combine several dri-
vers (e.g. logging and fire (Berenguer et al., 2014; Ewers et al., 2016;
de Andrade et al., 2017), implying that the responses of forests to
disturbance will be idiosyncratic (de Andrade et al., 2017; Chazdon
et al., 2016; Connell & Slatyer, 1977). Therefore, there is a pressing
need to better understand the mechanism and processes that lead
to fire-driven ecosystem degradation, as well as the recovery path-
ways after fire.
In the lowland Amazon, repeated burning causes irreversible shifts
in forest tree-size distribution (Balch et al., 2008; Barlow & Peres,
2008; Barlow, Peres, Lagan, & Haugaasen, 2003), and a reduction in
carbon stocks (Zarin et al., 2005). Fires are also a recognized driver of
forest degradation and emissions in TMCFs (Asbjornsen, Vel
azquez-
Rosas, Garc
ıa-Soriano, & Gallardo-Hern
andez, 2005; Asbjornsen &
Wickel, 2009; Crausbay et al., 2014; Oliveras, Girardin, et al., 2014;
Rom
an-Cuesta et al., 2011). Palaecological records suggest that fires
may have caused profound changes in the species diversity of forests
in the past, increasing the abundance of generalists (Bush et al.,
2015). More recently, in a study in the southern tropical Andes, Oliv-
eras, Malhi, et al., (2014) reported that around 45% of the species in
burned TMCFs were fire tolerant or fire thrivers (thrivers being spe-
cies that improve in performance after fire). However, there is still
very little information on the processes and mechanisms that lead to
shifts or differences in species composition, forest structure and car-
bon dynamics after fires in TMCFs. Most importantly, the long-term
ecological pathways of TMCF recovery after fire are still, to the best
of our knowledge, still largely undocumented.
This study provides an analysis of the short-, mid- and long-term
ecological pathways of TMCF recovery after fire in the southern Peru-
vian Andes. We present a 28-year chronosequence on changes in
diversity, carbon dynamics and forest structure after fire. Our objec-
tive is to quantify how TMCFS tree biodiversity, carbon stocks and
forest structure recover from fire and if they return to pre-burn condi-
tions. We address three specific questions and present several
hypotheses:
1. What are the short- and long-term effects of fire on carbon
stocks in TMCFs? We hypothesize that fire has a long-term
effect on C stocks, and that, on a thirty-year timescale, old-
burned forests do not recover C stocks to prefire conditions in
the aboveground biomass and soil (Barlow & Peres, 2008; Barlow
et al., 2003).
2. Do fire-affected TMCFs recover the forest structure of unburned
forests? We hypothesize significant changes over forest structure
after fire, with mortality of big stems shortly after fire, and an
increase on small stems in medium-term burned forests (Barlow
et al., 2003; Berenguer et al., 2014). We expect, however, that
old-burned forests will not have reached a mature forest tree-size
distribution within the time chronology of this study (28 years)
and therefore will deviate from predicted scaling exponents by
the Metabolic Scaling Theory (Enquist, West, & Brown, 2009).
3. What are the effects of fire over tree species alpha and beta-
diversity, and species turnover? Our working hypothesis is that
alpha diversity will be reduced immediately after fire, but diversity
will increase with time since last fire, with old-burned forests
being more diverse than unburnt forests (Oliveras, Malhi, et al.,
2014). However, beta-diversity will not be affected by fire, as spe-
cies turnover because of fire will override fire effects on species.
2
|
MATERIALS AND METHODS
The study area was located in the Cusco Department of Peru
(13°530S, 70°80W). A total number of 48 plots (30 930 m) were
sampled at altitudes ranging from 2,180 m to 3,552 m a.s.l.
(Table S1). Plots were located near the forest-grassland treeline, and
distributed inside or across the buffer area of several parks: the Manu
National Park and the Historical Sanctuary of Machu Picchu (Fig-
ure 1). We explicitly searched for areas with low human impact to
avoid mixed and frequent disturbances (e.g. grazing, fire, logging).
Plots were therefore established in remote locations (e.g. up to
2 days’walk). The plots encompassed undisturbed primary forests,
and a gradient of primary forests that have been exposed to different
levels of fire. Fieldwork was conducted during 2010, 2011 and 2012.
2.1
|
Experimental design and fire chronology
We assessed the fire impacts on TMCFs’(1) carbon recovery path-
ways, (2) structural recovery and (3) biodiversity. We selected a
chronology of representative fire-affected, fire-unaffected and
2
|
OLIVERAS ET AL.
carbon-similar forests ecosystems, in what is a classical space-for-
time substitution as an alternative to long-term studies (Benscoter &
Vitt, 2008; Blois, Williams, Fitzpatrick, Jackson, & Ferrier, 2013;
Dunne, Harte, & Taylor, 2003; Pickett, 1989). We followed the tem-
poral recovery responses through a chronology of burned forests
that presented different times since last fire (TLF) subdivided into
Recent Burns (RB: 1–4 years since last fire, n=9), Middle-term
Burns (MB: 5–9 years since last fire, n=8) and Old Burns (OB:
10–28 years since last fire, n=7).
We obtained the last burning date of the burned forests through
local information (local forest rangers, land owners), previous efforts
(Rom
an-Cuesta et al., 2011) and satellite records (Oliveras, Ander-
son, et al., 2014) (Table S1). We selected areas that had suffered
crown fires with flame lengths equalling at least the trees’heights
(e.g. up to 15 m) and intensities and residence times that allowed
organic soil removal (e.g. Oliveras, Anderson, et al., 2014; Rom
an-
Cuesta et al., 2011).
To assess fire-specific responses, we followed a matched, or
paired-plot design, appropriate for after-only studies that controls
for spatial heterogeneity (Van Butsic, Lewis, Radeloff, Baumann, &
Kuemmerle, 2017). For every burnt plot, an adjacent unburnt plot
was sampled (paired). A total number of 24 paired burned-unburned
plots (48 plots), each of 30 930 m, were included in the study. We
identified unburnt plots in nearby forests (but at least 50 m from
the burned forests) that showed little fire disturbances, were in the
same slope (or at least the same aspect and elevation), and were
structurally similar.
Due to a myriad of factors (soil, endemism, human influences, cli-
mate, land use legacy, etc.), Andean forests show high heterogeneity
in forest structure and composition. In spite of it, we expected to see
emerging trends from our chronology due to wildfire and explicitly
assumed that both unburnt plots and burnt plots in the three chrono-
logical groups were structurally similar (similar tree-size distributions)
and showed similar carbon allocations among pools (i.e. similar carbon
stocks in the aboveground biomass, CWD, soils, etc.) before the fire.
The unburnt plots were explicitly selected to maximize homogeneity
among these factors, and we further tested this assumption for each
carbon pool, with Kruskal–Wallis tests, using the three chronologies
(RB, MB, OB) as the grouping factor. The burned plots, because of the
strict selection criteria and matched experimental design, were also
assumed to be very similar before the fire (Van Butsic et al., 2017).
2.2
|
Carbon pools and forest inventories
In each plot, we measured different pools of forest carbon: above-
ground biomass (live standing trees), coarse woody debris (dead
FIGURE 1 Map of the study area in south-eastern Peru (Cusco district) Triangles represent the sampled paired-plot locations selected for
the study
OLIVERAS ET AL.
|
3
standing trees and fallen wood material >2 mm in diameter), litter
and belowground biomass (fine <2 mm and coarse roots >2mmin
diameter) and soil organic carbon.
2.2.1
|
Standing live and dead trees
We recorded all live and dead standing trees ≥10 cm in diameter at
breast height (DBH =1.3 m). Tree height was estimated as the aver-
age value of five independent visual estimates using a clinometer.
We recorded all small live and dead stems with DBH ≥2–10 cm in a
subplot of 300 m
2
(30 910 m). Biomass for each tree was esti-
mated by applying Chave’s allometric equation for wet forest stands
that includes DBH and height (Chave et al., 2005). Lianas were sam-
pled but were not included in the study as their abundance was very
low even in burned forests (Table S1). Trees were identified to spe-
cies level, but 28% could not be identified to the species level and
were identified to genus or family level. We extracted wood samples
with a wood corer of 2.5 mm diameter for each tree ≥10 cm DBH
and measured its wood density in the laboratory through water dis-
placement methods (Table S2). We used a carbon ratio of 0.5 g C/g
biomass.
2.2.2
|
Litter
Litter was collected at nine fixed points in each plot, in
25 925 cm
2
squares. Material was taken to the laboratory and
oven dried at 72°C until constant weight. Components of litterfall
included leaves; reproductive organs (fruits and flowers); twigs; epi-
phytes and bromeliads (all parts combined) and fine debris (unidenti-
fiable particles that pass through 2 mm mesh). We used a carbon
ratio of 0.45 g C/g biomass (Girardin et al., 2010).
2.2.3
|
Coarse woody debris
Coarse woody debris was assessed using a modification of the linear
intercept method (Brown, 1974) along a 30 m transect in the middle
of the plot. Fallen dead wood was separated into five diameter
classes: (I) <0.6 cm, (II) 0.61–2.54 cm, (III) 2.55–7.5 cm, (IV) 7.51–
12 cm and (V) >12 cm. Pieces of fallen dead wood on categories I
and II were tallied for the first 10 m along each transect in all plots.
Pieces of fallen dead wood in categories III, IV and V were tallied
along the entire 30-m transect in all plots. Decomposition class was
noted for each piece of fallen wood and was classified into five
states, from solid to fully decomposed (Baker et al., 2007). Wood
samples were oven dried at 72°C until constant weight and their val-
ues divided by water-displaced volumes to obtain wood densities.
(Table S3). We used a carbon ratio of 0.5 g C/g biomass.
2.2.4
|
Roots
Roots were divided into fine (<2 mm diameter) and coarse (>2 mm).
We extracted the fine roots for the first 20 cm of soil at four fixed
points per plot, using a soil corer. We measured coarse roots by
digging a 1 91 m pit at the centre of the plot and extracting all
roots at 10 cm-depth intervals until 50 cm, and then 50–100 cm-
depth. Coarse roots could not be sampled in all plots, and therefore
they are reported for C pools but not accounted for total C stocks.
All samples were taken to the laboratory, rinsed and oven dried at
70°C until constant weight. We used a carbon ratio of 0.475 g C/g
biomass. Coarse roots were only measured for n=8(n=3 for RB,
MB, and n=2 for OB).
2.2.5
|
Soils
Soil samples were extracted at four fixed points per plot using hori-
zon depths Oi, Oe and Oa (Zimmermann et al., 2010). The Oi hori-
zon was characterized by entire recognizable leaves at early stages
of decomposition. The Oe horizon was a laminated mixing of leaf
fragments and small twigs at a further stage of decomposition, also
containing large amounts of small roots. Oa was a compacted dark
coloured organo-mineral material, with fewer small roots and fre-
quent presence of charcoal. At each point, we measured the depth
of the horizon and we extracted a soil sample of known volume of
20 cm
3
. Soil carbon content was analysed for one composite soil
sample per horizon and per plot, with a Carlo Erba Elemental Analy-
ser (Milano, Italy) at the University of Saint Andrews (UK). Soil bulk
densities were determined from undisturbed samples collected with
stainless steel rings of 166 cm
3
. Bulk density was calculated as the
soil dry weight after subtracting the stone and fine root dry weights
(Table S4). Final carbon stock values per horizon were obtained by
multiplying the horizon depths by the carbon contents and by the
bulk densities, and appropriately transformed to MgC ha
1
.
Total carbon stocks were calculated as the sum of all the forest
pools. To assess carbon allocation per chronology and fire category
(burned/unburned), we first estimated the medians for the different
chronologies, for the burned and unburned plots separately. We then
estimated the contribution of each pool, for each chronology, as the
percentage of their stocks in relation to the median. We ran a statisti-
cal test to assess carbon recovery pathways and resilience (i.e. restor-
ing the original conditions): (1) Kruskal–Wallis tests to check for
carbon stock differences along time in the burned and unburned plots
separately, and (2) Wilcoxon Signed-Rank paired tests to assess car-
bon responses between the control and the burned plots, along time.
We used IBM SPSS statistics version 23. Significance levels were
assessed at the 0.05 (significant) and 0.1 (marginally significant) levels.
2.3
|
Forest structure
Changes in forest structure were assessed through variations in dia-
metrical tree-size density and relative frequency distributions binned
at 1 cm DBH size among plots. We explored the difference in the
proportion of grouped tree-size categories in burnt and control plots.
In each plot, based on the tree-size frequency distributions, we
grouped trees in DBH bins: 2–5, 5–10, 10–20, 20–40 cm, and bigger
than 40 cm and calculated the proportion of number of trees within
each bin.
4
|
OLIVERAS ET AL.
We tested differences between burned and mature forest plot
structure, by assessing predictions from metabolic scaling theory or
MST (Enquist & Niklas, 2001). Metabolic scaling theory predicts that
the structure of a forest, as measured by the sizes and numbers of
plants in a community will tend to follow an inverse power function
(e.g. self-thinning, Yoda, Kira, Ogawa, & Hozumi, 1963; Weller, 1987)
with characteristic scaling exponents. Enquist and Niklas (2001),
Enquist et al. (2009) proposed a model to explain how the inverse
size density relationship could emerge from shared individual plant
allometries. They assumed, first, that all individuals share a common
allometry of resource use, which is proportional to metabolic rate (B),
and Bis proportional to M, mass raised to the 3/4 power (M
3/4
); and,
second, that all individuals in the community compete for limiting
resources such that, in steady state, the rate of resource use approxi-
mates that of resource supply (R). Thus, the maximum number of indi-
viduals per size class, N
max
, that can be supported per unit area is
related to the average size of an individual as N
max
/R(M
3/4
). As
plant biomass is generally proportional to the 8/3 power of diameter
(West, Brown, & Enquist, 1999), the size distribution under approxi-
mate resource use steady state, in terms of stem diameter, is pre-
dicted to follow an inverse square rule where N/D
2
.
Metabolic scaling theory predicts that under approximate
resource use steady state, the N/D
2
relationship also indicates
the forest is in approximate demographic steady state where recruit-
ment and growth are offset by mortality. Thus, many ecological
dynamics that influence demography (such as competitive size hier-
archies, episodic recruitment and disturbance) will produce system-
atic deviations away from the MST predicted 2 size distribution
(e.g. Coomes, Duncan, Allen, & Truscott, 2003; Kerkhoff & Enquist,
2006, 2007). Deviations away from the 2 exponent then are diag-
nostic of community reorganization and are useful for understanding
(and managing for) ecosystem resilience (Enquist et al., 2009; Ker-
khoff & Enquist, 2007). If correct, the MST predictions offer a base-
line by which to quantify the relative magnitude of how differing
drivers and time size last burn influence forest structure and ecosys-
tem dynamics within and across TMCFs. To test these predictions in
our burned plots, we fitted their tree-size frequency distributions
using the package powerRlaw (Gillespie, 2015), and compared on
whether their scaling exponents differed across time since last fire.
As maximum tree size appears to reach a limit, we fit a truncated
Pareto distribution to the data, setting a maximum tree size of
300 cm DBH. Exponents were estimated using the Maximum Likeli-
hood Estimate (MLE, White, Enquist, & On, 2008).
2.4
|
Diversity: Species richness and turnover
We compared three gradients of dissimilarity to predict shift in spe-
cies richness and composition: (1) fire chronosequence, (2) burned/
unburned status and (3) geographical distance. We estimated alpha
diversity using the Fisher’s alpha metric based on a species abun-
dance matrix per burned/unburned forests and fire chronosequence,
using the function fisher.alpha() of the Rpackage vegan (Oksanen
et al., 2017). We explored turnover and beta-diversity using the
Sorensen’s index (bSOR), Simpson’s dissimilarity index (bSIM) and
species nestedness (bSNE) using the Rpackage betapart (Baselga,
2010; Baselga & Orme, 2012). We calculated multisite SOR, SIM and
SNE with the function beta.multi(), pairwised SOR, SIM and SNE
along spatial (geographic location) and disturbance gradient (burned/
unburned, fire chronosequence) with the function beta.pair(), and the
SOR, SIM SNE between matched paired control-burned plots using
the function beta.temp(). We used distance matrices for all pairwise
analysis of beta-diversity. The resulting dissimilarity matrices which
we then compared with a distance matrix of the main quantifiable
parameters (fire chronosequence, geographical distance) using a
Mantel permutation test, based on 9,999 permutations. Regressions
parameter statistics and 95% confidence intervals were obtained
using bootstrap resampling with the boot() function from the boot R
package (Canty & Ripley, 2017). We transformed the latitude/longi-
tude coordinates of the plots into geographical distances using the
spDists() function (package sp, Pebesma & Bivand, 2005; Bivand &
Pebesma, 2013).
3
|
RESULTS
3.1
|
Effects of fire on TMCF carbon dynamics
3.1.1
|
Responses of carbon stocks along time in
burned and control plots
The carbon stocks of unburned plots along time were not signifi-
cantly different for most of the pools, therefore confirming our
assumption that all sampled plots were similar (Figure 2).
The exception was coarse roots C stock, which was significantly
different along time in these unburned (Table 1): 19.6 MgC ha
1
in
the unburned plot from RB category vs. 3.4 MgC ha
1
in the
unburned plot from OB category. Soils were marginally significantly
different (96.7 vs. 158.2 MgC ha
1
). Contrarily to the unburned
TMCFs, we expected significant differences along time among the
carbon stocks of several pools in the burned TMCFs, as an indication
of postfire carbon stocks recovery. However, only three pools
(CWD, soils and fine roots) showed significant responses (Figure 2,
Table 1). Soil was the most dynamic carbon variable in the burned
TMCFs, with a ca. twofold increase in soil carbon, in the time-span
of 28 years: RB: 48.5, MB: 59.4, OB: 90.3 MgC ha
1
(Table 1,
Figure 3). Coarse woody debris and fine roots significantly increased
too (93 and 93.5 respectively), but had accumulation peaks in the
middle of the chronology: RB 1.5, MB: 14.5, OB: 5 MgC ha
1
, and
RB: 1.8, MB: 9.6, and OB: 6.3 MgC ha
1
for CWD and fine roots
respectively). Contrary to these carbon accumulations, the above-
ground biomass (standing live trees) and the standing dead trees of
burned TMCFs showed nonsignificant differences with time,
although dead standing trees increased their carbon stocks (Table 1).
The total carbon stocks of burned plots showed remarkable carbon
accumulation: RB: 116.6, MB: 161.1, OB: 295.3 MgC ha
1
.Asit
was the case for the unburned plots, soils also led the increases in
total carbon stocks along time (Table 1, Figure 3).
OLIVERAS ET AL.
|
5
FIGURE 2 Carbon allocation expressed
as the percent of total carbon stock in
each pool, for the burned (upper panel)
and unburned (lower panel) TMCFs. Stars
express significant results from Kruskal–
Wallis test on the effects on C stocks
(over medians) for the burned and
unburned TMCFs at *p<.1, **p<.05
6
|
OLIVERAS ET AL.
3.1.2
|
Responses of carbon stocks along time:
Burned vs. unburned plots
Fire in TMCFs had a different effect in the carbon stock for the dif-
ferent carbon pools (Table 1, Figure 3), and therefore we identified
four possible carbon recovery pathways along the fire chronose-
quence (Table 1, Figure 3):
1. Short-term carbon responses (significant differences in RB only).
This response was found for fine roots and total carbon stocks.
2. Time-lagged carbon responses (no significant difference on car-
bon stocks immediately after fire, but it becomes significant 5–
9 years after fire, that is, RB not significant and MB and OB sig-
nificant). These responses were found for standing live trees and
CWD.
3. Medium-term carbon responses (significant effects up to 9 years
after fire, that is, RB and MB significant but OB not significant):
this was the case for standing dead trees and soils.
4. No carbon responses (no significant differences on carbon stocks
along the 28 years of this chronology): seedlings, large trees
(DBH ≥40 cm), coarse roots and litter.
Therefore, our results showed that the fire effects in TMCFs C
stocks were surprisingly short-lived. At the end of the chronology, all
carbon variables showed nonsignificant differences against their
unburned paired plots except for standing live trees, whose lack of
carbon recovery persisted after 28 years, mainly affecting small-
medium trees (DBH 10–40 cm) (Figures 3 and 4). There were no sig-
nificant differences on soil carbon and on total carbon (sum of all
carbon pools) stocks between unburnt-burnt plots at the MB and
OB categories (Figure 3).
3.1.3
|
Responses in carbon allocation
Relative to their total carbon stocks, both the burned and unburned
TMCFs consistently held the most carbon in the soil and above-
ground biomass (standing live trees) along time (Figure 2). However,
in the unburned plots soils and standing live stocks consistently rep-
resented ≥70% of the total carbon, while in the burned plots the
contribution to total carbon was lower and more time-dependent
(ca. 60% for RB and MB, and 45% for OB). The pool with least car-
bon allocation varied between the burned and unburned forests:
standing dead trees in the case of the unburned plots (≤2.5% in all
the chronologies), while the burned plots showed that CWD was the
lowest for the RB and OB plots. The burned forests showed stand-
ing dead trees as the third highest allocation along the chronology
(5%–7.5%), followed by litterfall (5%–2.5%), whose contribution
decreased with time and peaked in the RB plots (Figure 2).
3.2
|
Effects of fire on TMCFs forest structure
The tree-size distributions were similar for all plot types and fire
chronosequences. We observed a bimodal distribution of tree sizes
TABLE 1 Carbon stocks (median [min–max]) for all the pools: aboveground (standing live trees, standing dead trees), belowground (fine and coarse roots), coarse woody debris (CWD), litter,
soil organic carbon, and derived total carbon, for the burned and unburned plots for the Recent Burns (1–4 years since last fire) (n=9); Mid-term Burns (5–9 years since last fire) (n=7); and
Old Burns (10–28 years since last fire) (n=7). pvalues (columns p) indicate significance of paired burned-control Wilcoxon Signed-Rank tests. Type of carbon response column indicates overall
carbon response based on the paired burned-unburned analysis along the chronosequence
Carbon pools MgC ha
1
RB MB OB Type of carbon
response
Burned Unburned p
RB
Burned Unburned p
MB
Burned Unburned p
OB
(p
RB
+p
MB
+p
OB
)
Small stems <10 cm 4.1 (0.9–6.5) 8.2 (2.2–24.1) ns 4.7 (2.5–10.4) 4.4 (1–13.4) ns 5.2 (0.1–7) 5.4 (0.8–15.4) ns No response
Standing live trees >10 cm 26.1 (7.1–88.3) 52.4 (8.5–163.3) ns 43.2 (10.3–55.5) 61.9 (37.4–123.5) ** 36.6 (5.7–132.3) 113.4 (35.6–192.1) ** Time lagged
Standing dead trees 6.2 (1–64.6) 2 (0.6–11.8) * 8.4 (0.7–41.4) 2.5 (1.3–7.9) * 21.1 (1.1–140.8) 1.9 (0.9–5.6) ns Medium term
CWD 1.5 (0–4.4) 1.6 (0–17.7) ns 14.5 (0.1–56.7) 8.3 (0–30.2) * 5 (0.3–52.8) 3.7 (0.8–15.5) ns Time lagged
Litter 7.2 (2.7–8.0) 8 (1.7–14.3) ns 6.0 (5.2–11.0) 7.6 (3.5–9.6) ns 5.0 (5.0–5.0) 10.4 (8.0–12.7) ns No response
Soil 48.5 (8.8–150) 96.7 (66.9–315.7) ** 59.4 (26.4–109.6) 62.4 (44.7–116.2) * 90.3 (66.1–303.7) 158.2 (60.3–360.4) ns Medium term
Fine roots 1.8 (0–6.1) 4.1 (1.3–15.6) ** 9.6 (0.8–19.8) 4 (0.5–6.4) ns 6.3 (0.4–10.2) 10.1 (1.7–18.5) ns Short term
Coarse roots
a
7 (0.5–9.8) 3.1 (1.1–4.0) ns 7.2 (3.2–11.2) 4.8 (1.8–5.8) ns 4.6 (3.6–5.5) 19.6 (5.4–20.4) ns No response
Total C 116.6 (51.7–79.8) 236.5 (111.2–441.2) ** 161.1 (94.4–214.0) 154.3 (136.7–254.8) ns 295.3 (238.2–408.5) 341.5 (136.9–494.9) ns Short term
a
For coarse roots n=8(n=3 for RB, n=3 for MB, and n=2 for OB).
*p≤.1, **p≤.05.
OLIVERAS ET AL.
|
7
with a higher relative frequency of small stems (smaller than 10 cm
DBH) and of trees between 10 and 15 cm DBH (Fig. S1). However,
the burned plots had consistently a higher relative frequency of the
smallest stems (2–5 cm DBH) than the unburned plots (Fig. S1), sug-
gesting regeneration through either recruitment or resprouting after
fire. Indeed, in the RB and MD fire categories, the proportion of
smallest stems (2–5 cm DBH) was significantly higher in the burned
plots compared to their paired unburned plots (Figure 4). Compared
to the proportion of trees in the unburned TMCFs, the proportion of
5–10 cm diameter trees was significantly lower in burned plots in
RB (probably due postfire mortality), nonsignificantly different in the
MB category, and significantly higher in the burned plots in the OB
category—suggesting postfire recovery. Interestingly, there were no
significant differences in the proportion of larger trees between
recently burned plots and their paired unburned plots, while there
was a significantly less proportion of trees 10–40 cm in the burned
plots in the MB category, suggesting medium-term responses after
fire in these tree-size categories (Figure 4).
The MLE exponents of plot-specific tree-size frequency power-
law distributions increased with time since last disturbance, converg-
ing to the exponent of 2 in old burned plots, confirming that these
forests experienced a directed succession towards primary forests
(Figure 5). Our results suggested a wide variation of disturbance
state within the first 10 years after disturbance and a convergence
towards a size distribution consistent with a steady-state mature for-
est at about 15 years after fire.
3.3
|
Effects of fire over TMCFs species richness
and turnover
The unburned TMCFs in the RB (seven out of nine sites) and MB
(six out of eight) fire chronosequences usually presented a higher
number of species than the burned plots (Table S1), but the differ-
ence was not significant when expressed as a diversity measure
(Fisher’s alpha, RB: F=0.77, p=.390; MB: F=2.00, p=.109,
Figure 6). For the OB sites, however, four of seven burned plots had
more species than their paired unburned plots, with a significantly
higher Fisher’s alpha coefficient than the control plots (F=3.62,
p=.041, Figure 6).
Total species dissimilarity among plots showed a high species turn-
over between plots (bSOR =0.973, bSIM =0.965), which was simi-
larly partitioned between control and burned plots (Table S5). When
looking at dissimilarity within fire chronosequences (Table S5), the
bSOR values between burned and unburned plots were usually slightly
higher in the burned plots, showing higher nestedness in the burned
plots, especially in the RB and OB chronosequences (Table S5).
Paired-plot dissimilarity (i.e. that of matched paired burned-
unburned plots) ranged from site to site from bSOR =0.95 (WAY2
site) to bSOR =0.51 (PLGD site) (Table 2). For all sites but ALF,
bSIM was higher than bSNE, revealing that changes in species were
more associated to species turnover than to nestedness. Dissimilarity
partitioning also revealed that usually those sites with higher paired
bSOR had a very small contribution of bSNE, i.e. that most of the
differences in species composition were associated to species turn-
over and little to nestedness, and that sites with more shared spe-
cies (i.e. lower values of bSOR) showed also more nestedness
(Table 2). However, this was not always the case (e.g. PAIT and
PLGRD showed bSOR, 0.6 and bSNE <0.10).
When species dissimilarity was explored in terms of geographical
distance, Mantel permutation tests revealed no increases in species
dissimilarity with distance for the unburned plots in any of the three
time categories (RB: r=.170, p=.12; MB: r=.145, p=.838; OB:
r=.432, p=.060, Figure 6), but the species dissimilarity was signifi-
cant for burned plots on the RB and OB categories (RB: r=.336,
p=.021, MB: r=.148, p=.232; OB: r=.660, p=.003, Figure 7).
4
|
DISCUSSION
The long-term effects of fire on tropical forests are still largely
unknown, and to the best of our knowledge this is the first study
that utilizes a long-term chronosequence to assess the effects of fire
and recovery pathways of TMCFs to fires, with a detailed analysis
on carbon stocks, forest structure and diversity recovery pathways.
FIGURE 3 Panels show relative differences in C stocks for each C pool compared to the paired unburned plots. Significant differences are
highlighted in bold, and arrows indicate significantly higher (upper arrow) or lower (lower arrow) C stocks in the burned plots compared to the
unburned. On the right side, absolute difference in carbon stocks along the chronosequence for burned plots (red) and unburned plots (green).
For absolute values, refer to Table1
8
|
OLIVERAS ET AL.
4.1
|
Effects of fire on forest carbon stocks
Our study shows that carbon stocks recovery depends on the
selected pools (i.e. total carbon vs. aboveground carbon), and on the
time-allowed to recover. Our results showed four possible recovery
pathways (short-term, middle-term, no response and time-lagged,
Table 1). Thus, if we look at the total carbon stocks, TMCFs are
highly carbon resilient as there was a surprisingly short-lived signifi-
cant effect on carbon stocks for all forest pools and for total carbon,
with the exception of the standing live tree pool. However, if we
only looked at aboveground biomass (standing live trees), then we
could conclude that TMCFs are highly carbon nonresilient with time-
lagged carbon responses (appearing after 5–9 years) leading to sig-
nificantly reduced carbon stocks after 28 years. These results are
supported by the time-lagged effects on CWD and standing dead
tree biomass, suggesting that fire-associated mortality shows a med-
ium-term response—some trees die charred from the fire, but a
number of them might not immediately die from fire, but result
highly damaged and be unable to recover or be more prone to herbi-
vore or pathogen attacks.
Soil carbon stocks showed an unexpected and remarkable med-
ium-term recovery at 5–9 years after fire, which was confirmed by
the significant positive increase of soil carbon stocks along time for
both the burned and the unburned plots, separately (Table 1). Similar
to boreal peatlands (Turetsky & Wieder, 2001; Turetsky et al., 2015)
and several tropical soils (e.g. “terra-preta”in the Amazon, Glaser,
Balashov, Haumaier, Guggenberger, & Zech, 2000), Andean montane
cloud forests have high soil carbon densities (Gibbon et al., 2010;
Oliveras, Girardin, et al., 2014; Zimmermann et al., 2010) which
relate to low decomposition rates due to low temperatures that
allow the slow incorporation of labile carbon pools. In TMCFs, fine
roots and CWD and litter- are labile carbon pools that are behind
the temporal increases in soil carbon, and should result in a reason-
ably closed carbon balance. However, the observed rapid increase of
soil carbon stocks in the burned plots (DC
soil
=42 MgC ha
1
in
28 years), can only be partly explained by the contribution of these
labile carbon variables (DC
CWD
=11.5 MgC ha
1
and DC
fineroots
=3–
6 MgC ha
1
and DC
litter
=0–1.1 MgC ha
1
) (Table 1), suggesting
some external entrance of carbon material into the soil pool. A com-
ponent that may have influenced on the recovery of soil C stocks is
the Sphagnum moss. The genus Sphagnum is one of the most impor-
tant groups of plant species sequestrating carbon in temperate and
northern bog ecosystems (Berendse et al., 2001), and an important
genera for TMCFs, because of the low decomposability of the dead
material it produces (Berendse et al., 2001). The genus-specific p-
hydroxy-b-carboxymethyl-cinnamic-acid strongly retards the decay
of litter of both Sphagnum and other neighbouring plants (Clymo &
Hayward, 1982). Moreover, by creating anoxic and acid conditions,
Sphagnum strongly reduces microbial degradation of the litter of co-
occurring plant species. Therefore, carbon sequestration in peatlands
RB MB OB
0.0 0.1 0.2 0.3 0.4
2–5 cm
5–10 cm
10–20 cm
20–40 cm
>40 cm
2–5 cm
5–10 cm
10–20 cm
20–40 cm
>40 cm
2–5 cm
5–10 cm
10–20 cm
20–40 cm
>40 cm
Proportion
Tree size category
Plot
Burned
Unburned
* +
*
* +
*
*
+
+
*
FIGURE 4 Relative proportions (with regard to total number of
stems) of trees per tree-size diameter classes in burnt (B) and
unburnt (U) forests along the fire chronosequence. RB, Recent Burns
(1–4 years since last fire); MB, Mid-term Burns (5–10 years since
last fire); and OB, Old Burns (10–28 years since last fire). Error bars
represent standard deviations. Star symbols (*) indicate significant
differences in the proportion of trees, and sum symbols (+) indicate
significant differences in C stocks between U and B plots for that
tree-size category at p<.05 (Wilcoxon-rank tests)
FIGURE 5 Maximum likelihood estimate (MLE) for the power-
law fit of the tree-size distributions of each burned plot through
time since last fire
OLIVERAS ET AL.
|
9
strongly depends on Sphagnum mass growth (Berendse et al., 2001).
Sphagnum has also been recognized as a keystone genus for habitat
restoration of bogs and peatlands (Gorham & Rochefort, 2003;
Rochefort, 2000). Unfortunately, this study did not directly quantify
Sphagnum moss and therefore we cannot prove that Sphagnum was
the ultimate responsible for the reported soil C recovery, but our
results clearly point out for a future research priority in this matter.
4.2
|
Forest succession pathways
There were medium-term responses (5–9 years) on the proportion
of small stems, with an increase of stems <5 cm in the burnt plots
and a decrease of proportion of stems between 5 and 10 DBH in
the recently burned plots. This was reflected in lower C stocks on
the 5–10 cm DBH category, but not in the 2–5 cm DBH category.
The increase on the smallest size stems proportion is likely to be
due to regeneration patterns via recruitment or resprouting (Oliveras,
Malhi, et al., 2014), and the decrease of stems 5–10 cm is likely to
be caused by a high mortality of these stems after fire. Small stems
contribute little to total C and therefore the variation of tree sizes
within this smallest DBH category (e.g. more 4 cm trees in the
unburnt plots and more 2.5 cm trees in the burnt plots) may over-
ride absolute differences in C between paired plots. However, mor-
tality of 5–10 cm stems after fire was high and led to a thinning
process of trees within this category, reflected in lower proportion
of trees and C stocks.
In the mid-term (5–9 years), the lower proportion of 10–40 cm
trees in burned TMCF provides further support to the time-lagged
mortality and woody decomposition dynamics on these forests.
However, recovery is clearly reflected in the much larger proportion
of small stems (2–5 cm) in the burned TMCFs 5–9 years after fire,
which results in a larger proportion of 5–10 cm in the long term
(>10 years after fire). The lack of significant difference in the
proportion of trees, but significant difference in long-term C stocks
in 10–40 cm DBH trees may be attributed to different dynamics in
succession patterns—our fire chronosequence includes a range of
years, and therefore variability in the responses. This is also reflected
in the large range of scaling exponent when fitting power-law tree
size scaling laws to the tree-size distribution of the burned plots
(Figure 3). Successional trajectories are known to be largely uncer-
tain (Chazdon et al., 2016)—in a recent study of forest succession, it
was identified that plot identity explained over 60% of the total vari-
ance on stem density, overriding stand age (Norden et al., 2015).
Our results support the metabolic scaling theory that predicts a
scaling relationship associated to disturbance (Enquist et al., 2009).
The MST predicts that the parameters of tree-size distribution scal-
ing laws (their coefficients and exponents) are significantly correlated
with the stem density, and that the shape of the size distributions
(i.e. its exponents) is indicative of time since disturbance (Kerkhoff &
Enquist, 2007). According to this, we may predict that burned
TMCFs have recovered their ecosystem function and are structurally
similar to primary undisturbed forests from 14 years after fire, when
they asymptotically converge to a scaling exponent of 2 (Figure 3,
Kerkhoff & Enquist, 2007).
6
8
10
12
14
RB MB OB
Fire chronosequence
Fisher’s alpha
FIGURE 6 Fisher’s alpha index for burnt (black) and unburnt
(grey) TMCFs forest along the fire chronosequence. RB, Recent
Burns (1–4 years since last fire, n=9); MB, Mid-term Burns (5–
10 years since last fire, n=8); and OB, Old Burns (10–28 years
since last fire, n=5)
TABLE 2 Pairwise dissimilarity indices (bSOR, Sorensen’s; bSIM,
Simpson species turnover; bSNE, nestedness) between paired
burned-unburned plots. TLF, fire chronosequence category (see text
for details). Site: site code where a paired pair burned-unburned
plots were sampled (Table S1)
TLF Site bSIM bSNE bSOR
RB AHO 0.571 0.189 0.760
ALF 0.167 0.511 0.677
CUSI 0.731 0.028 0.759
QORI 0.333 0.238 0.571
SAC 0.462 0.170 0.632
SUNCH 0.875 0.014 0.889
WAY1 0.560 0.032 0.593
WAY2 0.905 0.004 0.909
WAY3 0.613 0.018 0.631
MB CHALL 0.615 0.128 0.743
JAP 0.706 0.031 0.737
LAG 0.333 0.187 0.520
PAIT 0.500 0.088 0.588
PLGRD 0.471 0.043 0.513
WACH 0.765 0.061 0.826
WAY2a 0.944 0.004 0.949
YASP 0.667 0.054 0.721
OB ACHI 0.867 0.025 0.892
INKA 0.545 0.029 0.574
QUIPE 0.429 0.275 0.704
ROMPE 0.538 0.128 0.667
WINAY 0.538 0.048 0.586
YAN 0.708 0.066 0.774
YAN2 0.647 0.134 0.782
10
|
OLIVERAS ET AL.
4.3
|
Effects of fire over diversity
All sites showed high levels of species richness, regardless of distur-
bance. Our results support the observations that fire enhances diver-
sity in the long term (>10 years), as previously reported by other
studies in the area (Oliveras, Malhi, et al., 2014) as well as in other
Neotropical lowland forests (Balch et al., 2011; Devisscher, Malhi,
Rojas Landiar, & Oliveras, 2016). Palaeoecological studies provide
evidence that fire has been present in TMCFs since the beginning of
the Holocene (e.g. approx. 11,500 years) (Urrego, Silman, Correa-
0.5
0.6
0.7
0.8
0.9
1.0
0 50 100
Sorensen index
Unburned, RB
0.5
0.6
0.7
0.8
0.9
1.0
0 50 100 150
Sorensen index
Unburned, MB
0.5
0.6
0.7
0.8
0.9
1.0
0 5 10 15
Distance (km)
Sorensen index
Unburned, OB
0.5
0.6
0.7
0.8
0.9
1.0
050100
Sorensen index
Burned, RB
0.5
0.6
0.7
0.8
0.9
1.0
050100150
Sorensen index
Burned, MB
0.5
0.6
0.7
0.8
0.9
1.0
0 5 10 15
Distance (km)
Sorensen index
Burned, OB
FIGURE 7 Relationship between Sorensen’s similarity index and geographical distance between the studied plots, grouped by the fire
(burned, unburned) and time since last fire chronosequence (RB, recent burn; MB, mid-term burn; OB, old burn)
OLIVERAS ET AL.
|
11
Metrio, & Bush, 2011; Urrego et al., 2013), which may have already
induced a species-composition shift with the disappearance of fire-
sensitive species (Bush et al., 2015).
Nevertheless, our study shows how fire enhances species dissim-
ilarity with geographical distance in recently burned and old burned
plots. In the mid-term burned plots, species succession and turnover
may override overall effects in species dissimilarity. Our beta-diver-
sity analyses highlight the high species turnover in the region, due to
the high biodiversity values that make TMCFs one of the biodiver-
sity hotspots of the world (Myers et al., 2000).
4.4
|
Overall ecological responses
Time-lagged responses have been identified in tropical lowland forests
as an ecological response of trees that are not adapted to fire, with a
preferential mortality from large trees (Barlow et al., 2003; Berenguer
et al., 2014). Interestingly, high-elevation TMCFs show time-lagged
responses after 5–9 years that persist along the rest of the chronose-
quence, but these carbon reductions were led by the mortality of small
and medium tree sizes (i.e. DBH 10–20 cm and 20–40 cm). Thus,
while time-lagged responses in TMCFs may suggest nonfire adapta-
tion, the lack of fire effects on large trees suggests some fire toler-
ance. However, time-lagged responses may not only relate to fire
susceptibility but rather to slow forest dynamics in these cold and
cloudy ecosystems, with particularly slow stem productivities as sug-
gested by Malhi et al. (2017) in their comparative estimates of forest
NPP along an elevational gradient that cover our study area (i.e.
NPPstems of ca. 1.1 0.11 MgC ha
1
yr
1
in ≥3,020 m a.s.l., com-
pared to NPPstems of ca. 2.7 0.33 MgC ha
1
yr
1
in
≤210 m a.s.l.). Fire tolerance has long been described in TMCFs
through resprouting mechanisms which promotes the survival and
generalized expansion of fire-resistant TMCFs tree genera (e.g. Cle-
thra,Oreopanax,Weinmannia) with the detriment of fire-sensitive spe-
cies that only thrive in isolated fire protected patches (i.e. Polylepis,
Clusia) (Oliveras, Malhi, et al., 2014; Rom
an-Cuesta et al., 2011).
4.5
|
Challenges in monitoring fire disturbance
Assessing changes in forest structure, diversity and carbon stocks
following fire disturbance is a major challenge that includes the chal-
lenge of monitoring before natural burning events, and having to rely
on experimental approaches with obvious caveats and limitations.
The natural high forest heterogeneity of TMCFs ecosystems,
together with noncontrolled treatment conditions and the statisti-
cally low number of plots in our research, can lead to underestima-
tions in the significance tests (Type II errors). In practice, this means
that nonsignificant responses as those observed in Table 1 may be
the result of large data variability. Nonetheless, the strict sampling
design and careful data analysis allows to infer processes such as
successional pathways, supported with new ecological theory,
ensures the validity of the presented results, and this study still
responds to the largest and more ambitious field work done on
recovery pathways of TMCFs.
4.6
|
Implications for conservation
This study provides a detailed description of the pathways to recov-
ery of TMCF after fire disturbance, which has important implications
for their conservation. Our results show the immediate losses in car-
bon, and forest structure after fire, and a gradual recovery over time
of about 15 years. The good news is that our results show a relative
rapid recovery on soil carbon–in comparison, forest recovery of trop-
ical lowland forests is around100 years. However, this comes at a
price, and our results indicate that the highest price might biodiver-
sity and aboveground C stocks: the most fire-sensitive species may
disappear, and other more fire-tolerant species may replace them.
Therefore, burned TMCFS may recover total C stocks (although not
standing live carbon stocks) and tend towards a forest structure of a
primary forest within three decades after fires, but their species
composition will be different from undisturbed forests. Therefore,
investments in avoiding further disturbance can ensure the recovery
and high carbon storage large carbon in TMCFs, hence constituting a
great conservation opportunity.
In this century, the Andean region and most montane areas are
experiencing rises in temperature, elevation of the lower cloud level
(Malhi et al., 2010), and increasing droughts (Rom
an-Cuesta et al.,
2014). Severe droughts in the region enhance the occurrence of fires
and with this the chance of recurrent burning (Oliveras, Girardin,
et al., 2014; Rom
an-Cuesta et al., 2014). The tropical Andes host the
largest fraction of the world’s TMCFs, and given the climatic,
edaphic and biological similarities among TMCFs, we anticipate that
any results of this study will probably apply for many other fire-
threatened TMCFs in the world. Therefore, if we are to protect
these hugely important systems in terms of ecosystem services, and
climate change mitigation strategies, postburn protection of these
forests is imperative.
ACKNOWLEDGEMENTS
This paper is a product of the Andes Biodiversity and Ecosystems
Research Group (ABERG) research consortium. The authors wish
to thank to the research assistants for field assistance, and to
MNP forest rangers for allowing us to use their facilities. We also
thank the Amazon Basin Conservation Association for institutional
support. This material is based upon work supported by the UK
Natural Environment Research Council under grant NE/G006385/
1. BJE was supported by National Science Foundation awards
106586 and DEB 1457812. M is supported by the Jackson Foun-
dation and a European Research Grant Advanced Investigator
Award GEM-TRAIT (ACRYQY00). We warmly thank and remember
our beloved Flor Zamora and Rommel, whom shared many of the
field adventures while data collection, but today are not among us
anymore.
ORCID
Imma Oliveras http://orcid.org/0000-0001-5345-2236
12
|
OLIVERAS ET AL.
REFERENCES
de Andrade, R. B., Balch, J. K., Parsons, A. L., Armenteras, D., Roman-
Cuesta, R. M., & Bulkan, J. (2017). Scenarios in tropical forest degra-
dation: Carbon stock trajectories for REDD+.Carbon Balance and
Management,12,1–7.
Arag~
ao, L. E., & Shimabukuro, Y. E. (2010). The incidence of fire in Ama-
zonian forests with implications for REDD. Science (New York, N.Y.),
328, 1275–1278. https://doi.org/10.1126/science.1186925
Asbjornsen, H., Vel
azquez-Rosas, N., Garc
ıa-Soriano, R., & Gallardo-
Hern
andez, C. (2005). Deep ground fires cause massive above- and
below-ground biomass losses in tropical montane cloud forests in
Oaxaca, Mexico. Journal of Tropical Ecology,21, 427–434. https://doi.
org/10.1017/S0266467405002373
Asbjornsen, H., & Wickel, B. (2009). Changing fire regimes in tropical
montane cloud forests: A global synthesis. In M. A. Cochrane (Ed.),
Tropical fire ecology: Climate change, land use and ecosystem dynamics
(pp. 607–626). (Springer-Praxis books in environmental sciences). Ber-
lin, Germany: Springer; Chichester, UK: Praxis. https://doi.org/10.
1007/978-3-540-77381-8
Baker, T. R., Honorio Coronado, E. N., Phillips, O. L., Martin, J., Van Der
Heijden, G. M. F., Garcia, M., & Silva Espejo, J. (2007). Low stocks of
coarse woody debris in a southwest Amazonian forest. Oecologia,
152, 495–504. https://doi.org/10.1007/s00442-007-0667-5
Balch, J. K., Nepstad, D. C., Brando, P. M., Curran, L. M., Portela, O., De
Carvalho, O., & Lefebvre, P. (2008). Negative fire feedback in a transi-
tional forest of southeastern Amazonia. Global Change Biology,14,
2276–2287. https://doi.org/10.1111/gcb.2008.14.issue-10
Balch, J. K., Nepstad, D. C., Curran, L. M., Brando, P. M., Portela, O., Guil-
herme, P., ... de Carvalho, O. (2011). Size, species, and fire behavior
predict tree and liana mortality from experimental burns in the Brazil-
ian Amazon. Forest Ecology and Management,261,68–77. https://doi.
org/10.1016/j.foreco.2010.09.029
Barlow, J., & Peres, C. A. (2008). Fire-mediated dieback and composi-
tional cascade in an Amazonian forest. Philosophical Transactions of
the Royal Society of London. Series B, Biological sciences,363, 1787–
1794. https://doi.org/10.1098/rstb.2007.0013
Barlow, J., Peres, C. A., Lagan, B. O., & Haugaasen, T. (2003). Large tree
mortality and the decline of forest biomass following Amazonian
wildfires. Ecology Letters,6,6–8.
Baselga, A. (2010). Partitioning the turnover and nestedness components
of beta diversity. Global Ecology and Biogeography,19, 134–143.
https://doi.org/10.1111/j.1466-8238.2009.00490.x
Baselga, A., & Orme, C. D. L. (2012). Betapart: An R package for the
study of beta diversity. Methods in Ecology and Evolution,3, 808–812.
https://doi.org/10.1111/j.2041-210X.2012.00224.x
Benscoter, B., & Vitt, D. (2008). Spatial patterns and temporal trajectories
of the bog ground layer along a post-fire chronosequence. Ecosystems,
11, 1054–1064. https://doi.org/10.1007/s10021-008-9178-4
Berendse, F., Van Breemen, N., Rydin, H., Buttler, A., Heijmans, M.,
Hoosbeek, M. R., ... Wall
en, B. (2001). Raised atmospheric CO
2
levels and increased N deposition cause shifts in plant species
composition and production in Sphagnum bogs. Global Change
Biology,7, 591–598. https://doi.org/10.1046/j.1365-2486.2001.
00433.x
Berenguer, E., Ferreira, J., Gardner, T. A., Arag~
ao, L. E., De Camargo, P.
B., Cerri, C. E., ... Barlow, J. (2014). A large-scale field assessment of
carbon stocks in human-modified tropical forests. Global Change Biol-
ogy,2005,1–14.
Bivand, R. S., & Pebesma, E. J. (2013). Applied spatial data analysis with R,
2nd ed. New York, NY: Springer. https://doi.org/10.1007/978-1-
4614-7618-4
Blois, J., Williams, J., Fitzpatrick, M., Jackson, S., & Ferrier, S. (2013).
Space can substitute for time in predicting climate-change effects on
biodiversity. Proceedings of the National Academy of Sciences of the
United States of America,110, 9374–9379. https://doi.org/10.1073/
pnas.1220228110
Brown, J. (1974). Handbook for inventorying downed woody material.
USDA Forest Service General Technical Report INT-19. Ogden, UT:
Intermountain Forest and Range Experiment Station.
Bush, M. B., Alfonso-reynolds, A. M., Urrego, D. H., Valencia, B. G., Correa-
metrio, Y. A., Zimmermann, M., & Silman, M. R. (2015). Fire and climate:
Contrasting pressures on tropical Andean timberline species. Journal of
Biogeography,42, 938–950. https://doi.org/10.1111/jbi.12470
Canty, A., & Ripley, B. (2017). boot: Bootstrap R (S-Plots) Functions.R
package version 1.3-19.
Chave, J., Andalo, C., Brown, S., Cairns, M. A., Chambers, J. Q., Eamus,
D., ... Lescure, J. P. (2005). Tree allometry and improved estimation
of carbon stocks and balance in tropical forests. Oecologia,145,87–
99. https://doi.org/10.1007/s00442-005-0100-x
Chazdon, R. L., Broadbent, E. N., Rozendaal, D. M. A., Bongers, F., Zam-
brano, A. M., Aide, T. M., ... Craven, D. (2016). Carbon sequestration
potential of second-growth forest regeneration in the Latin American
tropics. Science Advances,2, e1501639. https://doi.org/10.1126/sc
iadv.1501639
Clymo, R. S., & Hayward, P. M. (1982). The ecology of sphagnum. In A. J.
E. Smith (Ed.), Bryophyte ecology (pp. 229–289). Netherlands, Dor-
drecht: Springer. https://doi.org/10.1007/978-94-009-5891-3
Connell, J. H., & Slatyer, R. O. (1977). Mechanisms of succession in natu-
ral communities and their role in community stability and organisa-
tion. The American Naturalist,111, 1119–1144. https://doi.org/10.
1086/283241
Coomes, D. A., Duncan, R. P., Allen, R. B., & Truscott, J. (2003). Distur-
bances prevent stem size-density distributions in natural forests from
following scaling relationships. Ecology Letters,6, 980–989. https://d
oi.org/10.1046/j.1461-0248.2003.00520.x
Crausbay, S., Genderjahn, S., Hotchkiss, S., Sachse, D., Kahmen, A., &
Arndt, S. K. (2014). Vegetation dynamics at the upper reaches of a
tropical montane forest are driven by disturbance over the past
7300 years. Arctic, Antarctic, and Alpine Research,46, 787–799.
https://doi.org/10.1657/1938-4246-46.4.787
Devisscher, T., Malhi, Y., Rojas Landiar, V. D., Oliveras, I. (2016). Under-
standing ecological transitions under recurrent wordfire: A case study
in the seasonally dry tropical forests of the Chiquitania, Bolivia. Forest
Ecology and Management,360, 273–286.
Dunne, J. A., Harte, J., & Taylor, K. J. (2003). Subalpine meadow flower-
ing phenology responses to climate change: Integrating experimental
and gradient methods. Ecological Monographs,73,69–86. https://doi.
org/10.1890/0012-9615(2003)073[0069:SMFPRT]2.0.CO;2
Enquist, B. J., & Niklas, K. J. (2001). Invariant scaling relations across
tree-dominated communities. Nature,410, 655–660. https://doi.org/
10.1038/35070500
Enquist, B. J., West, G. B., & Brown, J. H. (2009). Extensions and evalua-
tions of a general quantitative theory of forest structure and dynam-
ics. Proceedings of the National Academy of Sciences of the United
States of America,106, 7046–7051. https://doi.org/10.1073/pnas.
0812303106
Ewers, R. M., Andrade, A., Laurance, S. G., Camargo, J. L., Lovejoy, T.
E., & Laurance, W. F. (2016). Predicted trajectories of tree commu-
nity change in Amazonian rainforest fragments. Ecography,40,26–
35.
Feeley, K. J., Silman, M. R., Bush, M. B., Farfan, W., Cabrera, K. G., Malhi,
Y., ... Saatchi, S. (2011). Upslope migration of Andean trees. Journal
of Biogeography,38, 783–791. https://doi.org/10.1111/jbi.2011.38.is
sue-4
Gibbon, A., Silman, M. R., Malhi, Y., Fisher, J. B., Meir, P., Zimmermann,
M., ... Garcia, K. C. (2010). Ecosystem carbon storage across the
grassland-forest transition in the high Andes of Manu National Park,
Peru. Ecosystems,13, 1097–1111. https://doi.org/10.1007/s10021-
010-9376-8
OLIVERAS ET AL.
|
13
Gillespie, C. S. (2015). Fitting heavy tailed distributions: The poweRlaw
package. Journal of Statistical Software,2,1–16.
Girardin, C. A. J., Espejob, J. E. S., Doughty, C. E., Huasco, W. H., Met-
calfe, D. B., Durand-Baca, L., ... Halladay, K. (2013). Productivity and
carbon allocation in a tropical montane cloud forest in the Peruvian
Andes. Plant Ecology and Diversity,7,1–17.
Girardin, C. A. J., Malhi, Y., Arag~
ao, L. E. O. C., Mamani, M., Huaraca-
Huasco, W., Durand, L., ... Salinas, N. (2010). Net primary productiv-
ity allocation and cycling of carbon along a tropical forest elevational
transect in the Peruvian Andes. Global Change Biology,16, 3176–
3192. https://doi.org/10.1111/j.1365-2486.2010.02235.x
Glaser, B., Balashov, E., Haumaier, L., Guggenberger, G., & Zech, W.
(2000). Black carbon in density fractions of anthropogenic soils of
the Brazilian Amazon region. Organic Geochemistry,31, 669–678.
https://doi.org/10.1016/S0146-6380(00)00044-9
Gorham, E., & Rochefort, L. (2003). Peatland restoration: A brief assessment
with special reference to Sphagnum bogs. Wetlands Ecology and Manage-
ment,11,109–119. https://doi.org/10.1023/A:1022065723511
Kerkhoff, A. J., & Enquist, B. J. (2006). Ecosystem allometry: The scaling
of nutrient stocks and primary productivity across plant communities.
Ecology Letters,9, 419–427. https://doi.org/10.1111/j.1461-0248.
2006.00888.x
Kerkhoff, A. J., & Enquist, B. J. (2007). The implications of scaling
approaches for understanding resilience and reorganization in ecosys-
tems. BioScience,57, 489. https://doi.org/10.1641/B570606
Malhi, Y., Gardner, T., Goldsmith, G. R., Silman, M. R., & Zelazowski, P.
(2014). Tropical forests in the Anthropocene. Annual Review of Envi-
ronment and Resources,39, 125–159. https://doi.org/10.1146/annure
v-environ-030713-155141
Malhi, Y., Girardin, C. A. J., Goldsmith, G. R., Doughty, C. E., Salinas, N.,
Metcalfe, D. B., ... Arag~
ao, L. E. (2017). The variation of productivity
and its allocation along a tropical elevation gradient: A whole carbon
budget perspective. New Phytologist,214, 1019–1032. https://doi.
org/10.1111/nph.2017.214.issue-3
Malhi, Y., Silman, M., Salinas, N., Bush, M., Meir, P., & Saatchi, S.
(2010). Introduction: Elevation gradients in the tropics: Laboratories
for ecosystem ecology and global change research. Global Change
Biology,16, 3171–3175. https://doi.org/10.1111/j.1365-2486.2010.
02323.x
van Marle, M. J. E., Field, R. D., van der Werf, G. R., Estrada de Wagt, I.
A., Houghton, R. A., Rizzo, L. V., ... Tsigaridis, K. (2016). Fire and
deforestation dynamics in Amazonia (1973-2014). Global Biogeochem-
ical Cycles,31,24–38.
Mulligan, M. (2010). Modelling the tropics-wide extent and distribution
of cloud forests and cloud forest loss with implications for their con-
servation priority. In L. A. Bruijnzeel, F. N. Scatena, & L. S. Hamilton
(Eds.), Tropical montane cloud forests: Science for conservation and
management (pp. 14–38). Cambridge, UK and New York, NY: Cam-
bridge University Press.
Myers, N., Mittermeier, R. A., Mittermeier, C. G., da Fonseca, G. A. B., &
Kent, J. (2000). Biodiversity hotspots for conservation priorities. Nat-
ure,403, 853–858. https://doi.org/10.1038/35002501
Norden, N., Angarita, H. A., Bongers, F., Mart
ınez-Ramos, M., Granzow-
de la Cerda, I., Van Breugel, M., ... Finegan, B. (2015). Successional
dynamics in Neotropical forests are as uncertain as they are pre-
dictable. Proceedings of the National Academy of Sciences of the United
States of America,112, 8013–8018. https://doi.org/10.1073/pnas.
1500403112
Oksanen, J., Blanchet, F. G., Friendly, M., Kindt, R., Legendre, P., McGlinn,
D., ... Wagner, H. (2017). vegan: Community Ecology Package version
2.4-3. Retrieved from https://CRAN.R-project.org/package=vegan
Oliveras, I., Anderson, L. O., & Malhi, Y. S. (2014). Application of remote
sensing to understanding fire regime and biomass burning emissions
of the tropical Andes. Global Biogeochemical Cycles,28, 480–496.
https://doi.org/10.1002/gbc.v28.4
Oliveras, I., Girardin, C., Doughty, C. E., Cahuana, N., Arenas, C. E., Oliver,
V., ... Malhi, Y. (2014). Andean grasslands are as productive as tropi-
cal montane cloud forests. Environmental Research Letters,9, 115011.
https://doi.org/10.1088/1748-9326/9/11/115011
Oliveras, I., Malhi, Y., Salinas, N., Huaman, V., Urquiaga-Flores, E., Kala-
Mamani, J., ... Rom
an-Cuesta, R. M. (2014). Changes in forest struc-
ture and composition after fire in tropical montane cloud forests near
the Andean treeline. Plant Ecology and Diversity,7, 329–340.
https://doi.org/10.1080/17550874.2013.816800
Pebesma, E. J., & Bivand, R. S. (2005). Classes and methods for spatial
data in R. R News,5,9–13.
Peres, C. A., Gardner, T. A., Barlow, J., Zuanon, J., Michalski, F., Lees, A.
C., ... Feeley, K. J. (2010). Biodiversity conservation in human-modi-
fied Amazonian forest landscapes. Biological Conservation,143, 2314–
2327. https://doi.org/10.1016/j.biocon.2010.01.021
Pickett, S. T. A. (1989). Space-for-time substitution as an alternative to
long-term studies in long term studies in ecology. In G. E. Likens
(Ed.), Long-term studies in ecology: Approaches and alternatives (pp.
110–135). New York, NY: Springer-Verlag. https://doi.org/10.1007/
978-1-4615-7358-6
Rochefort, L. (2000). Sphagnum—A keystone genus in habitat restora-
tion. The Bryologist,103, 503–508. https://doi.org/10.1639/0007-
2745(2000)103[0503:SAKGIH]2.0.CO;2
Rom
an-Cuesta, R. M., Carmona-Moreno, C., Lizcano, G., New, M., Silman,
M., Knoke, T., ... Vuille, M. (2014). Synchronous fire activity in the
tropical high Andes: An indication of regional climate forcing.
Global Change Biology,20, 1929–1942. https://doi.org/10.1111/gcb.
12538
Rom
an-Cuesta, R. M., Salinas, N., Asbjornsen, H., Oliveras, I., Huaman, V.,
Guti
errez, Y., ... Astete, R. (2011). Implications of fires on carbon
budgets in Andean cloud montane forest: The importance of peat
soils and tree resprouting. Forest Ecology and Management,261,
1987–1997. https://doi.org/10.1016/j.foreco.2011.02.025
Sarmiento, L. (2006). Grazing impact on vegetation structure and plant
species richness in an old-field succession of the Venezuelan Para-
mos. In E. M. Spehn, M. Liberman, & C. K€
orner (Eds.), Land use
change and mountain biodiversity (pp. 119–135). Basel, Switzerland:
CRC Press. https://doi.org/10.1201/9781420002874
Turetsky, M. R., Benscoter, B., Page, S., Rein, G., van der Werf, G. R., &
Watts, A. (2015). Global vulnerability of peatlands to fire and carbon
loss. Nature Geoscience,8,11–14.
Turetsky, M. R., & Wieder, R. K. (2001). A direct approach to quantifying
organic matter lost as a result of peatland wildfire. Canadian Journal
of Forest Research,31, 363–366. https://doi.org/10.1139/x00-170
Urrego, D. H., Bush, M. B., Silman, M. R., Niccum, B. A., La Rosa, P.,
McMichael, C. H., ... Palace, M. (2013). Holocene fires, forest stabil-
ity and human occupation in south-western Amazonia. Journal of Bio-
geography,40, 521–533. https://doi.org/10.1111/jbi.12016
Urrego, D. H., Silman, M. R., Correa-Metrio, A., & Bush, M. B. (2011).
Pollen-vegetation relationships along steep climatic gradients in west-
ern Amazonia. Journal of Vegetation Science,22, 795–806. https://doi.
org/10.1111/jvs.2011.22.issue-5
Urrutia, R., & Vuille, M. (2009). Climate change projections for the tropi-
cal Andes using a regional climate model: Temperature and precipita-
tion simulations for the end of the 21st century. Journal of
Geophysical Research,114,1–15.
Van Butsic, A., Lewis, D. J., Radeloff, V. C., Baumann, M., & Kuemmerle,
T. (2017). Quasi-experimental methods enable stronger inferences
from observational data in ecology. Basic and Applied Ecology,19,1–
10. https://doi.org/10.1016/j.baae.2017.01.005
Van Der Werf, G. R., Randerson, J. T., Giglio, L., Collatz, G. J., Mu, M.,
Kasibhatla, P. S., ... van Leeuwen, T. T. (2010). Global fire emissions
and the contribution of deforestation, savanna, forest, agricultural,
and peat fires (1997-2009). Atmospheric Chemistry and Physics,10,
11707–11735. https://doi.org/10.5194/acp-10-11707-2010
14
|
OLIVERAS ET AL.
Weller, D. E. (1987). A re-evaluation of the -3/2 power rule of plant self-
thinning. Ecological Monographs,57,23–43. https://doi.org/10.2307/
1942637
West, G. B., Brown, J. H., & Enquist, B. J. (1999). A general model for
the structure and allometry of plant vascular systems. Nature,400,
664–667.
White, E. P., Enquist, B. J., & On, J. L. G. (2008). Ecological archives. Ecol-
ogy,12,1–3.
Yoda, K., Kira, T., Ogawa, H., & Hozumi, K. (1963). Self-thinning in over-
crowded pure stands under cultivated and natural conditions. Journal
of Biology, Osaka City University,14, 107–129.
Young, K. R., & Le
on, B. (2007). Tree-line changes along the Andes: Impli-
cations of spatial patterns and dynamics. Philosophical Transactions of
the Royal Society of London. Series B, Biological sciences,362, 263–
272. https://doi.org/10.1098/rstb.2006.1986
Zarin, D. J., Davidson, E. A., Brondizio, E., Vieira, I. C., S
a, T., Feldpausch,
T., ... Denich, M. (2005). Legacy of fire slows carbon accumulation in
Amazonian forest regrowth. Ecology,3, 365–369.
Zimmermann, M., Meir, P., Silman, M., Fedders, A., Gibbon, A., Malhi, Y.,
... Zamora, F. (2009). No differences in soil carbon stocks across the
tree line in the Peruvian Andes. Ecosystems,13,62–74.
Zimmermann, M., Meir, P., Silman, M. R., Fedders, A., Gibbon, A., Malhi,
Y., ... Zamora, F. (2010). No differences in soil carbon stocks across
the tree line in the Peruvian Andes. Ecosystems,13,62–74. https://d
oi.org/10.1007/s10021-009-9300-2
SUPPORTING INFORMATION
Additional Supporting Information may be found online in the sup-
porting information tab for this article.
How to cite this article: Oliveras I, Rom
an-Cuesta RM,
Urquiaga-Flores E, et al. Fire effects and ecological recovery
pathways of tropical montane cloud forests along a time
chronosequence. Glob Change Biol. 2017;00:1–15.
https://doi.org/10.1111/gcb.13951
OLIVERAS ET AL.
|
15
