Content uploaded by Sarayudh Bunyavejchewin
Author content
All content in this area was uploaded by Sarayudh Bunyavejchewin
Content may be subject to copyright.
CSIRO PUBLISHING
International Journal ofWildland Fire 2008, 17, 782–792 www.publish.csiro.au/journals/ijwf
The impacts of large-scale, low-intensity fires on the forests
of continental South-east Asia
Patrick J. Baker
A,D
, Sarayudh Bunyavejchewin
B
and Andrew P. Robinson
C
A
Australian Centre for Biodiversity and School of Biological Sciences, Monash University,
VIC 3800, Australia.
B
Research Office, National Parks, Wildlife and Plant Conservation Department, Chatuchak,
Bangkok 10900, Thailand.
C
Department of Mathematics and Statistics, University of Melbourne, Parkville, VIC 3010, Australia.
D
Corresponding author. Email: patrick.baker@sci.monash.edu.au
Abstract. South-eastAsia’s tropical forests harbour high levelsofspecies richness and endemism. In continental South-
east Asia strong rainfall seasonality driven by the Asian monsoon lead to ground-fires during the dry season in most
years. How these fires influence the region’s landscape mosaic of evergreen and deciduous forests and the biodiversity
they support is poorly understood. In this paper we report on the impacts of the El Niño–Southern Oscillation-induced
1997–98 fires that burned across much of westernThailand. We compare fire effects in the three common regional forest
types – seasonal evergreen (SEG); mixed deciduous (MDF); and deciduous dipterocarp – and use data from a 50-ha study
plot to evaluate the impacts of fire on these forests. We found few differences among the forest types. The fires created
more large gaps in MDFthan the other foresttypes.The SEGexperienced greater fire mortality in thesmallestsize classes,
abundant resprouting, and showed some evidence of lagged mortality among larger trees. The resilience of the SEG to
fire and lack of major differences in fire effects among the forest types suggest that infrequent landscape-scale fires may
have little effect on biodiversity in the landscape mosaic of seasonal tropical forests of continental South-east Asia.
Introduction
In the 14th and 15th centuries, as Europeans were exploring the
far reaches of their known world, map-makers would mark the
end of ‘discovered’landswithspecific phrases to warn sailors of
the uncertainty and dangers that might lie beyond. On one early
representation of the known world, the Lenox Globe, a small
copper sphere made in 1503, the phrase ‘Hic sunt dracones’–
Here be dragons – is inscribed over the region that we now know
as South-east Asia. This phrase may have originated from the
legends of elemental dragons, common to many Asian cultures,
that had begun to trickle back toEurope on spice-laden caravans
and merchant ships. In today’s world, there are few places that
remain unexplored; however, ‘Here be dragons’is an apt phrase
to describe the current understanding of largefires in South-east
Asia, for this is in many ways unexplored territory. Much of our
currentunderstanding oftherole offiresin thetropicalforests of
South-east Asia is based on anecdote and conjecture. Few data
are available on the impacts of fire in general in these ecosys-
tems. Even less is known of large, landscape-scale fires in the
region.Indeed,fundamental questionsabout theimpacts offires,
large and small, on the structure, composition, and dynamics of
tropical forests in South-east Asia remain largely unanswered
and unexplored. The lack of empirical data on landscape-scale
fires severely restricts the development of consistent operational
guidelines and strategic policies for managing fire in the region.
The tropical forests of continental and insular South-east
Asia harbor some of the highest levels of species richness and
endemism in the world (Sodhi et al. 2004). Within the two
broadbiogeographicrealmsof theregionwestofWallace’s Line,
the Indo-Burmese and Sundaland regions, there are hundreds
of species of mammals, amphibians, and reptiles, more than a
thousand species of birds, tens of thousands of plant species
and untold numbers of insects. Species-level endemism varies
across the different biogeographic regions of South-east Asia,
but may exceed 50% across a range of phyla in some areas (e.g.
Sodhiet al.2004). Becauseof thisgreatbiologicalrichness,both
known andunknown, potential threatstotropical forests, suchas
unrestricted logging and hunting, fragmentation, andconversion
to agriculture, represent potential threats to thousands of plant
and animal species.
Forest fires are not typically considered a threat to tropical
forests. However, in recent decades a growing number of stud-
ies have documented the occurrence of fires in tropical forests
and their impacts on the resident flora and fauna (Sanford et al.
1985; Leighton andWirawan 1986; Kinnaird and O’Brien 1998;
Cochrane and Schulze 1999; Goldammer 1999; Peres 1999;
Barlow et al. 2003a, 2003b; Cochrane 2003; Haugaasen et al.
2003; Slik and Eichhorn 2003).Although fires have occurred in
continentaland insularSouth-eastAsiathroughout theHolocene
(Haberle et al. 2001; Penny 2001; Maxwell 2004; Hope et al.
2005), their geographical extent was not fully appreciated until
the massive Bornean fires of 1982–83 and 1997–98. In both
cases, an intense El Niño–Southern Oscillation (ENSO) event
generated extremedrought conditions overcontinental and insu-
lar South-east Asia that were accompanied by extensive fires
throughouttheregion.InKalimantanalone,thefiresarebelieved
to have burned at least 5 million hectares each time, although
estimates of the total area burned vary (Malingreau et al. 1985;
© IAWF 2008 10.1071/WF07147 1049-8001/08/060782
Impacts of large fires in continental South-eastAsia Int. J. Wildland Fire 783
Siegert et al. 2001). Such extensive fires, and the impacts they
had on both human populations and biodiversity within the
region, forced tropical forest scientists, policy makers, and the
general public to re-evaluate their understanding of the role of
fire in tropical forests and, in particular, the role of extreme
climatic conditions in generating landscape-scale wildfires that
could burn through even the wettest rainforests.
Although most of the attention during the 1997–98 fires was
on Malaysia and Indonesia, the ENSO-induced drought led to
extensive fires elsewhere in the region. Unlike the aseasonal
forestsofinsularSouth-eastAsia,theseasonaltropicalforestson
the continent experiencefire relatively frequently. Each year, the
Asian monsoon system generates a dry season of 2–6 months in
whichlittle to no rain falls across a regionstretching from south-
ern India to eastern Vietnam. The landscapes of this region are
dominated by a patchwork mosaic of three well-described forest
types –seasonaldryevergreen,mixed deciduous, anddeciduous
dipterocarp – that differ markedly in stand structure and species
composition (Champion and Seth 1968).The seasonal evergreen
forestcontainsthemostspecies,hasacanopy>50mtall,andhas
a relatively small proportion of deciduous canopy trees, which
in most years allows the relative humidity in the understorey
to remain high even during the dry season. Mixed deciduous
forests have fewer species, a shorter main canopy (30–40 m tall)
and a majority of canopy trees that are deciduous during some
or all of the dry season. The predominance of deciduous trees
means that during this period considerable sunlight penetrates
through to the understorey and ground layer, lowering relative
humidity at ground level and increasing drying of fine fuels.
Bamboos are relatively common in the mixed deciduous forest
and occasional patches of grass are found throughout the forest.
Many of the bamboo species of continental South-east Asia are
monocarpicand flower gregariously.The synchronousmortality
over large areas provides a large, but infrequent, source of fuels
into the mixed deciduous forest systemsandmay influence local
and regional fire dynamics (Keeley and Bond 1999; Saha and
Howe 2001). The deciduous dipterocarp forest has the fewest
tree species, being dominated by one of 5–6 deciduous diptero-
carpspecies(e.g.Shoreasiamensis,Dipterocarpusobtusifolius),
the shortest main canopy (typically <25m tall), and is almost
completely deciduous for several months during the dry season.
Most importantly, however, the groundstorey of the deciduous
dipterocarp forest is dominated by grasses that dry during the
dry season when the canopy oftheforest isleafless (Troup 1921;
Williams et al. 2008).
Although fire is a relatively common feature of continental
South-east Asia, the nature of the fire regimes in the landscape
mosaic of forest types is poorly understood. Because of dif-
ferences in fuel composition and microclimatic conditions at
the forest floor, it is widely believed that deciduous diptero-
carp forests burn more frequently than mixed deciduous forests,
which in turn burn more frequently than seasonal evergreen
forests. Goldammer (1993) noted that deciduous dipterocarp
forests may burn annually, but that the frequent, low-intensity
ground-fire regime commonly associated with deciduous dipte-
rocarp forests was likely an anthropogenic artefact. Stott et al.
(1990) have suggested that the natural fire regime in these
landscapes has a ‘long-term cycle’ (i.e. infrequent fires). Such
assertions, however, may be misleading. First, there are no
empiricaldata onfire historiesfrom any forestswithin theregion
on which to base such assertions. Second, the limited anecdo-
tal data may not be representative of current or historical fire
regimes, particularly for the seasonal evergreen forests, which
are often more remote and less well known. Third, the complex
interdigitation of the forest types across these landscapes means
that the fire regimes are likely to be highly variable spatially and
temporally (as will be discussed in greater detail below).
The mosaic nature of these forested landscapes is of critical
importance to the conservation of biodiversity in continen-
tal South-east Asia. Among the terrestrial fauna of tropical
Asia, the larger species, especially the ungulates (e.g. elephant
(Elephasmaximus),gaur(Bosgaurus),banteng(Bosjavanicus))
and carnivores (e.g. tiger (Panthera tigris), leopard (Panthera
panthera)),areconcentratedintheselandscapemosaicsofdecid-
uous and evergreen forest. The grasses and bamboo thickets of
the deciduous forests supply the main grazing during the wet
season, whereas the understorey of the evergreen forest pro-
vides browse during the dry season. Importantly, the evergreen
forest provides cover throughout the year. As such, maintaining
an abundant andwidely distributed evergreen component within
the regional forest mosaic is seen as a key element of conserva-
tion plans for theterrestrial fauna of continental South-eastAsia
(Nakhasathien and Stewart-Cox 1990).
The occurrence of fires, therefore, raises grave concerns for
regional biodiversity. Of the three dominant forest types in the
region, the seasonal evergreen forest type is considered to be
the most sensitive to individual fires, with most tree species
believed to lack common adaptations to fire, such as thick bark
or the ability to resprout vigorously (Stott 1988; Rabinowitz
1990). Studies of fire effects in Borneo in the wake of the 1998
fires have shown that fire has dramatic impacts on the commu-
nity structure and composition of evergreen forests (Kinnaird
and O’Brien 1998; Slik and Eichhorn 2003). Because settle-
ment and development of formerly remote sites in continental
South-east Asia have increased substantially, the forest mosaic
hasbecome increasingly fragmentedinrecentdecades,exposing
more forests to more frequent ignition sources. Conservationists
are increasingly concerned that changes in fire regimes in these
landscapes will reduce the area of seasonal evergreen forest and
shift much of the landscape towards less evergreen forest and
more deciduous forest, resulting in a net loss of species diver-
sity as a consequence of ecosystem simplification (e.g. Ashton
1990; Rabinowitz 1990). However, this is largely unexplored
territory.
Because large fires are relatively uncommon, direct obser-
vations of their impacts on forest dynamics and local species
assemblages are rare, particularly for remote tropical forests.
Empirical data are sorely needed to underpintheongoing debate
about fire management in the seasonal tropical forests of con-
tinental South-east Asia. In the present paper, we report on the
impacts of a landscape-scale fire in deciduous and evergreen
foreststhatoccurredinwesternThailandattheHuaiKhaKhaeng
WildlifeSanctuary(HKK)duringthe1997–98ENSOevent.Our
study was an opportunistic one – such fires have only occurred
three times in the past 20 years at HKK – but took advantage of
a well-established research infrastructure within the sanctuary,
which included a long-term forest dynamics plot that had been
established 5 years before the fire. We focussed on two specific
784 Int. J. Wildland Fire P. J. Baker et al.
questions relevant to understanding the impacts of a large fire
on the forest mosaic at HKK. First, do fire effects differ among
the three major forest types found on the landscape? Second,
in the more diverse, and putatively more fire-sensitive, seasonal
evergreen forest, were certain species, functional groups, or size
classes more prone to fire-induced mortality?
Study area
The Huai Kha Khaeng Wildlife Sanctuary is one of Thailand’s
premierwildlife sanctuariesand isof great conservationvalue.It
is the second largest of 17 National Forests and Wildlife Sanctu-
ariesthattogetherconstituteThailand’sWesternForestComplex,
the largest area of contiguous protected forest in continental
South-eastAsia.The forests in theWestern Forest Complex sup-
port a widerangeof threatened andendangered plant andanimal
species including tigers, clouded leopards, elephants, banteng
and gaur (Nakhasathien and Stewart-Cox 1990). In recognition
of their importance for the conservation of the regionalflora and
fauna, HKK and the neighbouringThungYai-Naresuan Wildlife
Sanctuary were awarded UNESCO World Heritage Site status
in 1991.
Fires are a recurrent event at HKK. In the past two decades,
fires have been recorded somewhere within the sanctuary nearly
every year, although sanctuary-wide fires occur much less fre-
quently (1991, 1998, 2004). Increasing population densities and
agricultural activities adjacent to the buffer zone surrounding
HKK have led to a ready ignition source. Although extensive
landscape-scale fires are relatively rare, there is concern that
incursionsoffire intothe species-rich seasonalevergreenforests
are becoming more common and pose a serious threat to its
long-term presenceacrossthe landscape(Rabinowitz1990).The
1998 fires, whichburned through ∼1500km
2
of forest, were the
largest fires at HKK in at least the past 30 years.
Since 1991, forest scientists from the Royal Forest Depart-
ment ofThailand, the National Parks,Wildlifeand Plant Conser-
vation Department, and the Center for Tropical Forest Science
oftheSmithsonian Institutionhavebeen working together onthe
establishment and maintenance of a large-scale, permanent for-
estdynamicsstudyplot inseasonalevergreenforest atHKK.The
plot, whichcovers 50haandincludesevery tree >1cmdiameter
at breast height (DBH), includes 80000 trees from 291 species
and has been measured three times (in 1994, 1999, and 2004). In
February and March of 1998, the HKK fires burned through the
50-ha plot and adjacent areas of mixeddeciduous and deciduous
dipterocarp forest, providing a unique opportunity to describe
and compare the impacts ofa landscape-scale fire on each of the
major forest typesandtoexamine the potential role oflarge fires
in structuring and maintaining the landscape forest mosaic.
Methods
Fire intensity and fire-induced mortality among forest types
To determine if landscape-scale fires have different impacts on
the different forest types, we conducted an intensive survey of
tree mortality, as well as an extensive survey of fire-induced
gap formation in each of the three forest types. For our inten-
sive survey of fire effects in each forest type, we established two
20×125mtransects<1weekafter fire(March 1998)in areasof
seasonal evergreen, mixed deciduous and deciduousdipterocarp
forest that had been burnt by the fires. Each transect was divided
into 100 5×5m quadrats. To characterise the relative intensity
of the fire within each quadrat, we recorded whether fire had
occurred in each quadrat, the percentage of the quadrat area that
had beenburned, andtheaverage crownscorch height within the
quadrat based on loss or scorching of leaves. Within each tran-
sect, we also used a nested sampling design to characterise tree
abundance and mortality in different size classes. We identified
and measured all trees (>10cm DBH) within 10m of the tran-
sect, all poles (4.5–10 cm DBH) within 5 m of the transect, and
all saplings(1–4.5cm DBH) within1m of thetransect.The total
samplearea foreachtransectwas2500m
2
,1250 m
2
,and 250m
2
for trees, poles, and saplings, respectively. In addition, we tal-
lied, but did not identify, seedlings (<1cm DBH) in 1-m
2
plots
every 5m along the transect.The total sample area for seedlings
in each transect was 25 m
2
. For all trees, poles, and saplings,
the following information was recorded at the time the transect
was established: species, DBH, crown scorch, and whether the
individual was dead or alive. Crown scorch was scored from
1to5(1=80–100%, 2=60–80%, 3 =40–60%, 4=20–40%,
and 5 <20%crown scorch); deciduousspecieswerenoted.Each
transect was revisited at the end of the rainy season (November
1998, ∼8 months after fire) and the following information was
recorded for each individual: dead or alive, presence or absence
of sprouting, and the presence or absence of basal scarring.
To characterise the stand-scale impacts of the fires, we esti-
mated the number and size of canopy gaps that were created by
fire-killed trees in each forest type. For the seasonal evergreen
forest, a detailed assessment of damage was conducted in the
50-ha plot. Using the 20×20m grid system within the plot, 26
parallel 1000-m transects were established, beginning from the
northern edge of the plot. The size of each canopy gap created
by a fallen tree or group of trees was measured in two perpen-
dicular directions and calculated as the area of an ellipse. The
mode of death for each fallen tree was recorded as burnt-out
above or below 2m height on the stem, or as incidental mor-
tality (knocked over by a falling tree). Large permanent study
plots were not available for the mixed deciduous and decidu-
ous dipterocarp forests types. Instead, within each forest type,
we selected a large area of relatively homogeneous forest where
we established series of parallel transects ∼30 m apart and of
varying length.All fire-killed trees>10cm DBH that had fallen
within 15m of the transect were identified and measured. Total
areas sampled for the mixed deciduous and deciduous diptero-
carp forests were 12.5 and 12.1ha, respectively. All three forest
types were sampled 4–6 weeks after the fires had occurred. The
total area of canopy gaps created by the fires and the size distri-
bution of the canopy gaps were calculated for each forest type.
We compared gap size distributions among forest types using
non-parametric Kruskal–Wallis ANOVA.
Fire-induced mortality in the seasonal evergreen forest
To assess whether specific elements of the seasonal evergreen
forestcommunityare particularlysusceptibleto fire,weused the
50-ha plot database to compare species-specific mortality and
abundance patterns for seasonal evergreen forest tree species
during 1994–99, the period in which the 1998 fires occurred,
and 1999–2004, when no fires entered the plot.The 1999census
Impacts of large fires in continental South-eastAsia Int. J. Wildland Fire 785
was initiated in May 1998, 2 months after the fires occurred,
and completed ∼10 months later. We fitted locally weighted
smoothing (Loess) splines and estimated confidence intervals
for mortality as a function of tree size (DBH) for each species
separately and for all species pooled for the fire interval and
the non-fire interval. The Loess smoothing splines are fitted to
a variable probability sample of size 4000 from the measured
trees of each species, selected with probability proportional to
tree size. If there were fewer than 4000 representatives from the
species, then all trees were used. We used a sample size of 4000
to work around memory limits of the available algorithms. The
Loess fitting algorithm used the sampling weights, if sampling
was done. We also examined whether tree species from differ-
ent functional groups (evergreen v. deciduous) or possessing
certain life-history traits (large size: 90th percentile DBH; fast
growth: 90th percentile growth rate) were more or less prone to
fire-induced mortality. To determine whether tree species in the
seasonal evergreen forest were capableof sprouting, we checked
the 50-ha plot database for records of basal sprouts from trees
in the post-fire census period.
Results
Comparison of fire effects across the three forest types revealed
few differences. Nearly every study quadrat in each forest type
was burned. The proportion of each quadrat burned ranged
from 20 to 100%, but did not differ significantly among for-
est types. Mortality patterns were heterogeneous among size
classes and forest types. The greatest difference was between
sapling mortality in seasonal evergreen forest (24%) and mixed
deciduous forest (63%); however, most of the mixed deciduous
forestspeciesresproutedseveralmonthsafterthefire.Basalscar-
ring was greater in the seasonal evergreen forest (55%) than in
the mixed deciduous (32%) and deciduous dipterocarp forests
(30%). Scorch heights ranged from 0m in unburned quadrats
to 4m in a single quadrat in the seasonal dry evergreen forest.
The distribution of scorch heights did not differ significantly
among forest types, although the mixed deciduous forest had
a slightly lower mean scorch height (1.12 ±0.53m) than the
seasonal evergreen (1.46±0.67m) and deciduous dipterocarp
(1.48±0.48m) forests (Fig. 1).
In contrast, the stand-scale assessment of fire impacts on
canopy gap formation revealed substantial differences among
forest types. In the deciduous dipterocarp and seasonal dry ever-
green forest types,there was relatively little canopy disturbance.
In contrast, the mixed deciduous forest suffered substantial
canopy disturbance from the fire. Canopy gaps ranging from 25
to 1000 m
2
were created in the three forest types during the fires
and for ∼2 months afterward. In the 50-ha plot, the 1998 forest
fires created 85 new gaps, or 1.7 gaps ha
−1
and mean gap size
was ∼95m
2
(Fig. 2). The total area of fire-created gaps in the
seasonal evergreen forest was 1.61% of the total area surveyed.
In the 12.5ha of deciduous dipterocarp forest, only three small
(mean gap size =35m
2
) gaps occurred as a result of the fires.
The few fire-induced canopy gaps andtheirsmall size accounted
for only 0.09% of the deciduous dipterocarpforestcanopy being
disturbed by the 1998 fires at HKK. In the 12.1-ha study area
of mixed deciduous forest 5km to the east of the 50-ha plot, the
fires created 52 gaps (4.3ha
−1
) with a mean gap size of 105m
2
.
50
40
30
20
10
0
01234
Percentage of total quadrats sampled
Scorch height (m)
Deciduous dipterocarp forest
Mixed deciduous forest
Seasonal evergreen forest
Fig. 1. Distribution of scorch heights among the three forest types: sea-
sonalevergreenforest (SEG),mixeddeciduousforest(MDF), and deciduous
dipterocarp forest (DDF). In each forest type, scorch height was measured
in 100 contiguous quadrats within each of two transects. Scorch height was
determined based onthe averageheight of leafand crownscorch withineach
quadrat.
Although the mixed deciduous forest gaps were slightly larger
than those in the seasonal dry evergreen forest, the greater fre-
quencyof gapsinthe mixeddeciduous forestmeantthat thetotal
area of fire-created gaps (4.5%) was substantially greater than
in the other two forest types.
Within the species-rich seasonal evergreen forest, detailed
analysis of demographic data from the 50-ha plot revealed sev-
eral striking patterns. First, there were significant differences
in mortality patterns between the first (fire) and second (fire-
free) census intervals (Fig. 3). Overall mortality was higher in
the first interval (6.7%) than the second interval (4.2%). There
were also distinct patterns in size-dependent mortality. Mortal-
ity rates among small trees (<2 cm DBH) were nearly twice
as great in the first census (Fig. 3) owing to their shorter stature
(andconsequentlygreatercrown scorch)andthinnerbark (Baker
and Bunyavejchewin2006).Mortality rates during the fire inter-
val (i.e. intercensus period 1994–99 during which the 1998 fire
occurred) were also higher for trees 20–40-cm DBH, but were
not substantially different for trees >40 cm DBH. Examination
of mortality profiles for several abundant species demonstrated
that the plot-wide patterns masked considerable interspecific
heterogeneity in fire-induced mortality (Fig. 4). In most cases,
the smallest trees experienced higher mortality rates during
the fire interval than the non-fire interval – the two dominant
dipterocarps of the forest canopy, Hopea odorata and Dipte-
rocarpus alatus, being the exceptions. An interesting feature of
the species-specificcomparisonsof mortalityrates is that during
the fire-free interval, several species had higher mortality rates
in the largest size classes. Species exhibiting this pattern were
all large canopy species and included Hopea odorata, Diptero-
carpus alatus, Neolitsea obtusifolia, Tetrameles nudiflora, and
Saccopetalum lineatum. Observations in the years following the
fires suggest that among large trees, fire-induced mortality may
be subject to temporallagsasthe fire-scarred bases or weakened
786 Int. J. Wildland Fire P. J. Baker et al.
0 100 200 300 400 500 600 700 800 900 1000
Northing
0
100
200
300
400
500
Westing
Fig. 2. Location of all 85 gap-creating treefalls created by the 1998 El Niño–Southern Oscillation (ENSO)-associated
fires as they burned through the Huai Kha Khaeng 50-ha plot.
0 20 40 60 80 100
0.0
0.2
0.4
0.6
0.8
1.0
Diameter (DBH)
P(Dead)
94–99
99–04
Fig. 3. Loess-smoothed estimates of the probability (P) of mortality rates
for the two intercensus intervals (1994–99 and 1999–2004) as a function of
diameter at breast height (DBH, cm) for all trees on the Huai Kha Khaeng
50-ha plot. The grey lines around each smoothing represent one standard
error in each direction.
buttress roots increase a tree’s susceptibility to being toppled by
wind storms associated with the onset of the annual summer
monsoon. Haugaasen et al. (2003) have reported similar lagged
mortality among largetreessubjectto low-intensitygroundfires
in theAmazon.The only other large canopy tree,Vatica odorata,
showed no difference in large tree mortality during the fire-
and fire-free intervals. In contrast, the mid-storey trees Baccau-
rea ramiflora, Croton roxburghii, and Polyalthia viridis all had
greater mortality in the largesizeclassesduringthe fire interval.
Multiplecensusesforsuchalargenumberoftreesandspecies
at the HKK 50-ha plot also allowed us to determine the effects
of the fire on species loss and species immigration. Our data
show that neither species richness nor abundance was substan-
tially affected by the fire. Most species changed relatively little
in abundance (Fig. 5). Nine species were lost and four species
gained between the 1994 and 2004 censuses. All of the species
that were lost between 1994 and 2004 were rare – none had >40
individualsin 1994. However, among individual species, the fire
produced some clear winners and losers. Croton roxburghii, the
most abundant species on the plot, increased in abundance by
250% (11411 stems to 28778 stems). In contrast, Solanum eri-
anthum, an understorey treelet, lost most of its population (1504
stems to 14 stems). The 50-ha plot data also demonstrate that
many of the species in the seasonal evergreen forest do resprout
after fires. During the 1999 recensus, which was conducted in
the months that followed the 1998 fire, we recorded for every
stem whether basal sprouts were present.We recorded sprouting
in 163 species from the50-ha plot.Although sproutingoccurred
intreesranging from 1to 50cmDBH, sprouting was much more
common and prolific in the smallest stems. In some instances,
as manyas 25 sprouts were recorded on a single tree; the median
number of sprouts per tree was two (Fig. 6).
Comparison of functional groups and life-history traits
revealedlittleevidenceofdifferentialmortalityasaconsequence
of the fire. Pre- and post-fire abundances were similar for decid-
uous and evergreen species (Fig. 5). Fire-associated mortality
was correlated with tree stature, measured as the 90th percentile
DBH of each tree species, but not with growth rate, measured
as the 90th percentile DBH growth rate of the species (Fig. 7).
The relationship between mortality and tree stature was strongly
negativebetween1and15cm,afterwhichtherewaslittlechange
in the probability of mortality with increasing stature.
Discussion
Landscape-scale fires in continental and insular South-eastAsia
are widely believed to be detrimental to biodiversity within the
region (e.g. Rabinowitz 1990; Taylor et al. 1999; Sodhi et al.
2004). Short-term studies and anecdotal observations suggest
that wildlife is impacted directly through fire-induced mortality
(e.g. Stott 1986; Kinnaird and O’Brien 1998) or indirectly as
forest trees, which provide habitat and serve as food sources, are
Impacts of large fires in continental South-eastAsia Int. J. Wildland Fire 787
0 20406080100
0.0
0.2
0.4
0.6
0.8
1.0
CROTRO diameter (DBH)
P(Dead)
94–99
99–04
0 20406080100
0.0
0.2
0.4
0.6
0.8
1.0
VATIOD diameter (DBH)
P(Dead)
94–99
99–04
0 20406080100
0.0
0.2
0.4
0.6
0.8
1.0
POLYVI diameter (DBH)
P(Dead)
94–99
99–04
0 20406080100
0.0
0.2
0.4
0.6
0.8
1.0
HOPEOD diameter (DBH)
P(Dead)
94–99
99–04
0 20406080100
0.0
0.2
0.4
0.6
0.8
1.0
DIPTAL diameter (DBH)
P(Dead)
94–99
99–04
0 20406080100
0.0
0.2
0.4
0.6
0.8
1.0
NEOLOB diameter (DBH)
P(Dead)
94–99
99–04
0 20406080100
0.0
0.2
0.4
0.6
0.8
1.0
TETRNU diameter (DBH)
P(Dead)
94–99
99–04
0 20406080100
0.0
0.2
0.4
0.6
0.8
1.0
GARCSP diameter (DBH)
P(Dead)
94–99
99–04
0 20406080100
0.0
0.2
0.4
0.6
0.8
1.0
SACCLI diameter (DBH)
P(Dead)
94–99
99–04
0 20406080100
0.0
0.2
0.4
0.6
0.8
1.0
BACCRA diameter (DBH)
P(Dead)
94–99
99–04
Fig.4. Loess-smoothedestimates of the probability (P)ofmortality rates asafunctionof diameter at breast height (DBH,cm)for10
common species on the Huai Kha Khaeng 50-ha plot. Species abbreviations are Baccaurea ramiflora (BACCRA), Croton roxburghii
(CROTRO), Dipterocarpus alatus (DIPTAL), Garcinia speciosa (GARCSP), Hopea odorata (HOPEOD), Neolitsea obtusifolia
(NEOLOB), Polyalthia viridis (POLYVI), Saccopetalum lineatum (SACCLI),Tetrameles nudiflora (TETRNU), and Vatica odorata
(VATIOD). The grey lines around each smoothing represent one standard error in each direction.
788 Int. J. Wildland Fire P. J. Baker et al.
Species presence by sph 1994
Species presence by sph 2004
HYPTST
MELIAZ
MELILU
MURRPA
SOLAER
CASEGE
COLOJA
DALBAS
GLYCPE
MALLBA
MANGCA
POLYSU
CROTRO
0 0.1 1 10 100
0
0.1
1
10
100
Fig. 5. Comparisons of species-specific abundances between the first census interval, during which the 1998 fires
occurred, and the second census interval, which was fire-free. Values are in stems per hectare (sph). Evergreen species
are denoted with an empty circle and deciduous species with a cross. For some species, mostly those of low abundance,
the phenology is not known; these species are represented by a full grey circle. A line depicting a 1: 1 relationship is
drawn as a guide. Species that did not change in abundance between census intervals are located on the line. Species
that increased in abundance are above the line; species that decreased in abundance are below the line. Species that
were lost from the plot after the fire are on y =0; species that appeared in the plot following the fire are on x =0. The
species identified in the graph are those with the greatest difference in population density between intercensus intervals.
Species abbreviations are Caseariagrewiifolia (CASEGE),Croton roxburghii (CROTRO),Colona javanica(COLOJA),
Dalbergia assamica (DALBAS), Glycosmis pentaphylla (GLYCPE), Hyptianthera stricta (HYPTST), Mallotus barba-
tus (MALLBA), Mangifera caloneura (MANGCA), Melia azederach (MELIAZ), Melicope lunaankenda (MELILU),
Murraya paniculata (MURRPA), Polyalthia suberosa (POLYSU), Solanum erianthum (SOLAER).
killed by the fires (e.g. Rabinowitz 1990; Kinnaird and O’Brien
1998). However, empirical evidence of catastrophic impacts by
fire in South-east Asian forests and their associated flora and
fauna is limited. This may be because there are few studies
that have documented the long-term impacts of fire on these
forestsor,alternatively,the fireshaveless impactthanpreviously
believed.
The HKK Wildlife Sanctuary is one of the few sites in
continental South-east Asia where research on the impacts of
landscape-scale fires have been conducted in an area within
which the forests are very well studied. The ENSO of 1997–98
generated extremely dry conditions atHKK that allowed several
small fires on the periphery of the sanctuary in January 1998
to become a much larger landscape-scale fire front that over the
next 3 months burned hundreds of square kilometres of forest.
The status of HKK in Thailand as a national conservation icon
meant that the 1998 fires were heavily covered in the national
media – on several occasions appearing on the front page of
every national newspaper simultaneously. Inevitably, the media
reported the fires as ecological catastrophes (e.g. ‘When Nature
Goes on a Fiery Rampage’, Bangkok Post, 18 February 1998).
Our study of fire impacts across the landscape mosaic of for-
est types at HKK suggested that the impacts of these fires on
the abundance and composition of tree species were relatively
limited, even in the putatively fire-sensitive seasonal evergreen
forest. We had anticipated that the substantial differences in for-
eststructureandcomposition amongthe three majorforest types
at HKK would lead to large differences in the amount of damage
Impacts of large fires in continental South-eastAsia Int. J. Wildland Fire 789
0 1020304050
DBH (cm)
5
0
10
15
20
25
No. of stump sprouts
Fig. 6. Relationship between number of basal sprouts on a tree stem and the diameter at breast height (DBH) of the tree. Owing
to the large number of tree stems with sprouts (n=8504) in the 50-ha plot database, individual points have been jittered by 20%
and overlain on a bivariate density plot to illustrate the relative abundance of number of stems as a function of DBH.
0 1020304050
0.0
0.2
0.4
0.6
0.8
1.0
90th percentile of diameter growth (cm per decade)
Mortality rate
0 50 100
150
0.0
0.2
0.4
0.6
0.8
1.0
90th percentile of diameter (cm)
Mortality rate
Fig. 7. Probability of mortality as a function of two life history traits: tree stature (left panel) and tree growth rate (right panel). Each point represents
one species and is based on the 90th percentile diameter at breast height (DBH) (left panel) or 90th percentile DBH growth rate (right panel) calculated
from the 50-ha plot data. A Loess smoothing spline indicates the locally weighted mean values as a function of either tree stature or tree growth rate.
that they incurred in the fires. Instead, we found that among
the forest types, there were relatively few differences in fire
extent,relative fire intensity (estimated from scorch height), and
mortality patterns across a range of size classes.
Despite the wide areal extent of the fires and the extreme
climatic conditions, the fires were predominantly low-intensity
surface fires with flame lengths ranging from 5 to 50 cm. The
fires never entered the canopy in any of the forest types and, in
790 Int. J. Wildland Fire P. J. Baker et al.
general, appeared to affect only trees <2m tall. Other studies
have shown that fires in tropical forests are often low-intensity
surface fires and that the initial impact of the fire is to kill the
aboveground portions of the smallest trees (e.g. Cochrane and
Schulze 1999; Peres 1999). One reason that fire impacts may
vary little among forest types is that fire intensity is consis-
tently low in each forest type. Despite differences in fine fuel
composition on the forest floor, scorch heights were not differ-
ent among the forest types. This is most likely a consequence
of the interaction between flame length and the speed of the
fire front (Agee 1993). In the deciduous dipterocarp forest, the
groundstoreyvegetationisdominatedbygrassesandherbaceous
vegetation. Flame lengths in these areas can be relatively high
(50–150cm),buttheopenunderstoreyallowsthefireto progress
quickly through the stand. In contrast, the seasonal evergreen
forest understorey is dominated by tree seedlings and leaf litter.
Flame lengths are shorter (5–30 cm), but the fire moves much
more slowly through the forest.
Within the seasonal evergreen forest, the 1998 fire led to
widespread aboveground mortality, despite the fire intensity
being relatively low. Low-intensity surface fires are typical in
tropical forests, particularly those in less seasonal environments
or in relatively moist edaphic conditions (e.g. Cochrane and
Schulze 1999; Peres 1999). Although the impacts of these fires
on wildlife may be limited, as many can move out of the way
of the slowly advancing fireline, the fires may influence com-
munity composition. Kinnaird and O’Brien (1998) noted that
species richness and abundance of the bird community in the
wake of the 1998 fires in evergreen forests in Sumatra did not
change. However, they found that the mortality of canopy trees
led to a decrease in frugivores and an increase in insectivores,
such as woodpeckers, which feed on wood-boring beetles that
attack injured or recently killed trees. What the long-term con-
sequences of the 1998 fireswill be on bird communities or other
species assemblages at HKK is unknown. Long-term studies
that document changes in community composition of the flora
and fauna in the wake of landscape-scale fires are sorely needed
across the region.
We found that in the immediate aftermath of the 1998 fire,
most mortality was in the smallest size classes (<5cm DBH).
Not surprisingly, the stature of a species as measured by its 90th
percentile DBH wasstrongly correlatedwithits susceptibility to
fire-induced mortality. This suggests that those small-statured
tree species that lack adaptations to fire, such as resprouting or
a persistent soil seedbank, will be most sensitive to changes
in population size induced by large fires. In contrast, small-
statured species that are well adapted to fire, such as Polyalthia
suberosa, which has particularly thick, corky bark, may benefit
from the fires (Fig. 5). Manystudiesinboth temperate and tropi-
calregionshavedemonstratedthatfire-inducedmortalityisoften
size-dependent for trees because larger trees have thicker bark
and are better able to protect the vascular cambium from catas-
trophic heating (e.g. Gill andAshton 1968; Hengst and Dawson
1994; Pinard and Huffman1997).Slik and Eichhorn (2003) also
found similar patterns of size-dependent mortality among ever-
green rainforest trees in East Kalimantan following the 1998
fires. Mortality was not restricted to the smallest trees, however.
We found evidence of a distinct delay in fire-induced mortal-
ity among large trees of several abundant canopy tree species.
Haugaasen et al. (2003) found a similar pattern of delayed mor-
tality in large trees after low-intensity surface fires in tropical
forest in the Central Amazon.
The size-dependent patterns of aboveground mortality and
dieback are counterbalanced by size-dependent patterns of
resprouting. Smaller trees were more susceptible to fire-induced
mortality, butwerealsomorelikely toresprout, oftenvigorously,
within months of the fire. For the seasonal evergreen forest,
which has long been considered the most sensitive of the main
forest types to fire, the outcome of these opposing dynamics
is a forest little changed by the 1998 ENSO fires. We found
little impact of the fires on community composition and pop-
ulation size structures within the seasonal evergreen forest. A
recent study has shown that there is no relationship between
bark thickness, a common surrogate for heat resistance (and
therefore fire tolerance), and various measures of population
size structures for 10 of the dominant tree species on the 50-ha
plot (Baker and Bunyavejchewin 2006).Our results are in direct
contrast to those from the aseasonal forests of South-east Asia
that experienced significant changes in structure and composi-
tion after the 1998 fires (e.g. Slik and Eichhorn 2003; Cleary
et al. 2006). Although the lack of long-term studies of forest
compositionchangefollowinglarge,landscape-scale fires limits
our ability to generalise our results, which derive from a single
extreme fire event from a single landscape within continental
South-east Asia, they do suggest that the ability to withstand
individual fire events may be a fundamental character that dis-
tinguishes the lowland evergreen dipterocarp forests of seasonal
South-eastAsian environments, located primarily in continental
sites, from their aseasonal analogues across the region (Baker
and Bunyavejchewin, in press).
In continental South-east Asia, there is concern that an
increase in fire frequency would allow the less diverse decid-
uous dipterocarp or mixed deciduous forest types to invade and
gradually replace the more species-rich seasonal evergreen for-
est(Stott 1988;Rabinowitz1990).Previousstudies atHKK have
shown that seedlings of common tree species from the seasonal
evergreen and mixed deciduous forests grow equally well when
planted in gaps in either forest type (Baker 1997). In a follow-
up survey 10 years after the seedlings were planted, we found
that the only species to survive in the fire-prone mixed decidu-
ous forest was Hopea odorata, the dominant canopytree species
in the seasonal evergreen forest, despite having been exposed
to three fires in a decade. More extensive studies of how tree
species from each of the forest types regenerate and how fires
impactthisregeneration areneeded before itispossibletoassess
potential shifts in the relative abundance of the evergreen and
deciduous forest types.
An overarching theme of the present review has been the
lack of data on the impacts of large-scale fires on biodiversity
in South-east Asia. Detailed data on tree species mortality and
abundance from the HKK 50-ha plot and surrounding stands
of mixed deciduous and deciduous dipterocarp forests in the
wake of the 1998 ENSO-associated fires provide much needed
empirical data on the impacts of landscape-scale fires in the
seasonal forests of mainland South-east Asia. Although the
results suggest that landscape-scale fires in the forest mosaic of
continental South-east Asia may not be particularly disastrous
for tree species, there remain many unanswered questions. For
Impacts of large fires in continental South-eastAsia Int. J. Wildland Fire 791
example, can the effects of a single, landscape-scale fire burn-
ing at low intensity have long-term consequences for variability
in demography? Can variability in demography, combined with
widespread heterogeneity in local fire intensity and canopy gap
formation, provide a mechanism for the tree species and forest
types to coexist in mosaic fashion across the landscape? Despite
our concerns for the remarkable diversity associated with the
tropical forests of South-eastAsia, we will not make progress in
understanding how landscape-scale fires influence the structure
and composition of the forest and the distribution and abun-
dance of its inhabitants across a range of spatial and temporal
scales without substantial increases in fire ecology research in
South-east Asia. Hic sunt dracones.
Acknowledgements
We would liketo thank the National ResearchCouncil ofThailand, the Royal
Forest Department and the staff ofthe Huai Kha KhaengWildlife Sanctuary
for enabling us to conduct this research at HKK.We would alsolike to thank
the National ScienceFoundation for support through grant DEB-0075334 to
PeterAshton and Stuart Davies, the USDA Forest Service through a grant to
Chad Oliver, and SigmaXi for a Grant-in-Aid-of-Resaerch to P.J. Baker.We
would also like to thank Dick Williams, Ross Bradstock, and an anonymous
reviewer for insightful comments on the manuscript. This is publication 182
of the Australian Centre for Biodiversity.
References
Agee JK (1993) ‘Fire ecology of Pacific Northwest forests.’ (Island Press:
Washington, DC)
Ashton PS (1990) Thailand: biodiversity center for the tropics of Indo-
Burma. Journal of the Science Society of Thailand 16, 107–116.
doi:10.2306/SCIENCEASIA1513-1874.1990.16.107
Baker PJ (1997)Seedling establishment and growth acrossforest types in an
evergreen/deciduous forest mosaic inwesternThailand. Natural History
Bulletin of the Siam Society 45, 17–41.
Baker PJ, Bunyavejchewin S (2006) Bark thickness and the influence of
fire on tree population structures in a seasonal evergreen tropical forest.
Natural History Bulletin of the Siam Society 54, 215–225.
Baker PJ, Bunyavejchewin S, Fire behavior and fire effects across the for-
est landscape of continental South-eastAsia. In ‘Tropical Fire Ecology:
Climate Change, Land Use, and Ecosystem Dynamics’. (Ed. M
Cochrane) (Springer-Praxis: Heidelberg), in press.
Barlow J, Peres CA, Lagan BO, Haugaasen T (2003a) Large tree mortal-
ity and the decline of forest biomass following Amazonian wildfires.
Ecology Letters 6, 6–8. doi:10.1046/J.1461-0248.2003.00394.X
Barlow J, Lagan BO, Peres CA (2003b) Morphological correlates of fire-
induced tree mortality in acentralAmazonian forest. Journal ofTropical
Ecology 19, 291–299. doi:10.1017/S0266467403003328
ChampionHG,SethSK (1968)‘Arevisedsurveyof theforesttypesofIndia.’
(Natraj Publishers: Dehra Dun, India)
Cleary DFR, Priadjat A, Suryokusumo BK, Menken SBJ (2006) Butterfly,
seedling, sapling and tree diversity and composition in a fire-affected
Bornean rainforest. Austral Ecology 31, 46–57. doi:10.1111/J.1442-
9993.2006.01542.X
Cochrane MA (2003) Fire science for rainforests. Nature 421, 913–919.
doi:10.1038/NATURE01437
Cochrane MA, Schulze MD (1999) Fire as a recurrent event in tropical
forests of the eastern Amazon: effects on forest structure, biomass, and
species composition. Biotropica 31, 2–16.
Gill AM, Ashton DH (1968) The role of bark type in relative tolerance to
fire of three central Victorian eucalypts. Australian Journal of Botany
16, 491–498. doi:10.1071/BT9680491
GoldammerJG(1999)Forestsonfire.Science284,1782–1783.doi:10.1126/
SCIENCE.284.5421.1782A
Goldammer JG (1993) Historical biogeography of fire: tropical and sub-
tropical. In ‘Fire in the Environment: The Ecological, Atmospheric,
and Climatic Importance of Vegetation Fires’. (Eds PJ Crutzen,
JG Goldammer) pp. 297–314. (Wiley: NewYork)
HaberleSG,HopeGS,vander Kaars S (2001) Biomass burningin Indonesia
and Papua New Guinea: natural and human induced fire events in the
fossil record. Palaeogeography, Palaeoclimatology, Palaeoecology 171,
259–268. doi:10.1016/S0031-0182(01)00248-6
Haugaasen T, Barlow J, Peres CA (2003) Surface wildfires in central
Amazonia: short-term impact on forest structure and carbon loss. Forest
Ecology and Management 179, 321–331. doi:10.1016/S0378-1127(02)
00548-0
Hengst GA, Dawson JO (1994) Bark properties and fire resistance of
selectedtree species fromthecentralhardwoodregionof NorthAmerica.
Canadian Journal of Forest Research 24, 688–696. doi:10.1139/
X94-092
Hope G, Chokkalingam U,Anwar S (2005)The stratigraphyand fire history
of the Kutai Peatlands, Kalimantan, Indonesia.Quaternary Research64,
407–417. doi:10.1016/J.YQRES.2005.08.009
Keeley JE, Bond WJ (1999) Mast flowering and semelparity in bamboos:
the bamboo fire cycle hypothesis. American Naturalist 154, 383–391.
doi:10.1086/303243
Kinnaird MF, O’Brien TG (1998) Ecological effects of wildfire on lowland
rainforest in Sumatra. Conservation Biology 12, 954–956.
LeightonM,WirawanN(1986)CatastrophicdroughtandfireinBorneotrop-
icalrainforestassociatedwiththe1982–83ElNiño–SouthernOscillation
event. In ‘Tropical Forests and theWorld Atmosphere’. (Ed. GT Prance)
AAAS Selected Symposium 101, pp. 75–102. (American Association
for the Advancement of Science: Washington, DC)
Malingreau JP, Stephens G,Fellows L (1985)Remote sensing offorest fires:
Kalimantan and North Borneo in 1982–83.Ambio 14, 314–321.
Maxwell AL (2004) Fire regimes in north-eastern Cambodian monsoonal
forests, with a 9300-year sediment charcoal record. Journal of Biogeo-
graphy 31, 225–239.
Nakhasathien S, Stewart-Cox B (1990) Nomination of the Thung Yai/Huai
Kha Khaeng Wildlife Sanctuary as a UNESCO World Heritage Site.
Royal Forest Department Publication. (Bangkok, Thailand)
Penny DP (2001) A 40 000 year palynological record from north-east
Thailand; implications for biogeography and palaeo-environmental
reconstruction. Palaeogeography, Palaeoclimatology, Palaeoecology
171, 97–128. doi:10.1016/S0031-0182(01)00242-5
Peres CA (1999) Ground fires as agents of mortality in a central Ama-
zonian forest. Journal of Tropical Ecology 15, 535–541. doi:10.1017/
S0266467499000991
Pinard MA, Huffman J (1997) Fire resistance and bark properties of trees in
a seasonally dry forest in eastern Bolivia. Journal of Tropical Ecology
13, 727–740.
RabinowitzA (1990) Fire, dry dipterocarp forest, and the carnivore commu-
nity in Huai Kha Khaeng Wildlife Sanctuary, Thailand. Natural History
Bulletin of the Siam Society 38, 99–115.
Saha S, Howe HE (2001) The bamboo fire cycle hypothesis: a comment.
American Naturalist 158, 659–663. doi:10.1086/323593
Sanford RL, Saldarriaga J, ClarkKE, Uhl C, Herrera R (1985)Amazon rain-
forest fires. Science 227, 53–55. doi:10.1126/SCIENCE.227.4682.53
Siegert F, Ruecker G, Hinrichs A, Hoffmann AA (2001) Increased damage
from fires in logged forests during droughts caused by El Niño. Nature
414, 437–440. doi:10.1038/35106547
Slik JWF, Eichhorn KAO(2003) Fire survival of lowlandtropical rain forest
trees in relation to stem diameter and topographic position. Oecologia
137, 446–455. doi:10.1007/S00442-003-1359-4
SodhiNS,KohLP,Brook BW,NgPKL(2004)South-eastAsianbiodiversity:
an impending disaster. Trends in Ecology & Evolution 19, 654–660.
doi:10.1016/J.TREE.2004.09.006
Stott P (1986)The spatialpatternof dryseason fires in thesavanna forestsof
Thailand. Journal of Biogeography 13, 345–358. doi:10.2307/2845018
792 Int. J. Wildland Fire P. J. Baker et al.
Stott P (1988) The forest as phoenix: towards a biogeography of fire in
mainland South-east Asia. The Geographical Journal 154, 337–350.
doi:10.2307/634607
Stott PA, Goldammer JG, Werner WL (1990) The role of fire in the trop-
ical lowland deciduous forests of Asia. In ‘Fire in the Tropical Biota:
Ecosystem Processes and Challenges’. (Ed. JG Goldammer) pp. 32–44.
(Springer: NewYork)
TaylorD, Saksena P, Sanderson PG, Kucera K(1999) Environmental change
and rain forests on the Sunda shelf of Southeast Asia: drought, fire
and the biological cooling of biodiversity hotspots. Biodiversity and
Conservation 8, 1159–1177. doi:10.1023/A:1008952428475
http://www.publish.csiro.au/journals/ijwf
Troup RS (1921) ‘The Silviculture of Indian Trees. Vol. I Dilleniaceae to
Leguminosae (Papilionaceae).’(Clarendon Press: Oxford, UK)
Williams LJ, Bunyavejchewin S, Baker PJ (2008) Deciduousness in a sea-
sonal tropical forest in western Thailand: interannual and intraspecific
variation in timing, duration, and environmental cues. Oecologia 155,
571–582. doi:10.1007/S00442-007-0938-1
Manuscript received 15 October 2007, accepted 24 June 2008