Surface wildfires in central Amazonia: short-term impact on forest structure and carbon loss

Surface wildfires in central Amazonia: short-term impacton forest structure and carbon loss

Torbjørn Haugaasen*, Jos Barlow, Carlos A. PeresCentre for Ecology, Evolution and Conservation, School of Environmental Sciences,

University of East Anglia, Norwich NR4 7TJ, UK

Received 8 November 2001; received in revised form 15 July 2002; accepted 7 October 2002

Abstract

Changes in forest structure were examined 10–15 months after an unprecedented understorey wildfire burnt previously

undisturbed primary forest in central Brazilian Amazonia, following the severe 1997–1998 El Nino dry season. On the basis of

20 0.25 ha plots (10 m � 250 m) in both burnt and unburnt forest, we found marked differences in the overall live biomass,

canopy openness and understorey vegetation. On average, 36% of all trees equal to or greater than 10 cm DBH were found to be

dead in the burnt forest, and there was also a near-complete mortality in all pre-burn saplings. Using an allometric equation to

predict biomass mortality we estimate that the tree mortality rates found would commit an additional 25.5 t C/ha to be released

from these BFs. The dramatic increase of aboveground dead biomass in BF is of major global concern because of the increased

flux of CO2 to the atmosphere, which has a role in enhancing the greenhouse effect and promoting climate change.

# 2002 Published by Elsevier Science B.V.

Keywords: Aboveground biomass; Carbon emissions; El Nino; Fire disturbance; Global warming; Surface fires

1. Introduction

Tropical forests are widely known to harbour the

highest species diversities in the world, but are cur-

rently under enormous pressure from rising human

populations (Pahari and Murai, 1999; Cincotta et al.,

2000), growing agricultural practices (Dale and Pear-

son, 1997), logging (Nepstad et al., 1999a,b; Putz et al.,

2001) and other disturbances such as hunting (Robin-

son and Bennett, 2000; Peres, 2000). In the Brazilian

Amazonia alone, a forest area the size of France has

been cleared in the last three decades (Fearnside,

1999), affecting an area nearly twice as large through

edge effects and forest fragmentation (Skole and

Tucker, 1993). In these areas, synergistic interactions

between structural and non-structural forms of dis-

turbance continue to erode biodiversity (Peres, 2001).

Since the mid 1980s, however, tropical forests have

been haunted by yet another threat which has steadily

increased in importance as greater areas of forest

become affected each year—surface wildfires.

Tropical rain forests are normally thought to be fire-

resistant environments because of high atmospheric

and soil moisture levels, and major wildfires have been

historically rare events (Turcq et al., 1998). Charcoal

dating evidence suggests that catastrophic fires have

occurred in the Amazon only four times in the past two

millennia, at 1500, 1000, 700 and 400 BP (Sanford

Forest Ecology and Management 6154 (2002) 1–11

* Corresponding author. Tel.: þ44-1603-591099;

fax: þ44-1603-507719.

E-mail address: [email protected] (T. Haugaasen).

0378-1127/02/$ – see front matter # 2002 Published by Elsevier Science B.V.

PII: S 0 3 7 8 - 1 1 2 7 ( 0 2 ) 0 0 5 4 8 - 0

et al., 1985) whilst the pollen record suggests that

there was a period of fires lasting from 7000 to

4000 years BP (Turcq et al., 1998). However, as a

result of recent climatic oscillations and large-scale

disturbances such as logging and fragmentation, fire is

becoming an increasingly common pantropical phe-

nomenon (Goldhammer, 1999). In Amazonia, surface

wildfires threaten a much greater area of forest than

does deforestation per se (Nepstad et al., 1999a;

Cochrane, 2001), even though current deforestation

rates in the Amazon are the fastest anywhere (Whit-

more, 1997).

In contrast to the far more detectable canopy fires,

surface fires rarely exceed 10–30 cm in height under

normal fuel and humidity conditions, burning the fine

and course litter on the forest floor (Holdsworth and

Uhl, 1997; Cochrane and Schulze, 1999; Nepstad et al.,

1999a). However, these apparently innocuous fires

have been shown to have serious detrimental effects

on both the forest structure (Peres, 1999; Barbosa and

Fearnside, 1999) and the vertebrate fauna (Kinnaird

and O’Brien, 1998; Haugaasen, 2000; Barlow et al.,

2002; Peres et al., in press), because of their extremely

rare occurrence in evolutionary time (Uhl and Kauff-

man, 1990).

By killing a substantial amount of the forest bio-

mass, the fires increase the carbon flux to the atmo-

sphere (Barbosa and Fearnside, 1999), enhancing the

greenhouse effect and promoting climate change. In

particular, this source of carbon release has been

severely underestimated or disregarded in estimates

of the carbon contribution to the atmosphere asso-

ciated with land cover changes in Amazonia (Fearn-

side, 1997a; Laurance et al., 1998).

In this paper we present tree mortality data and

subsequent changes in forest structure following a

large surface fire in central Brazilian Amazonia. We

provide estimates of the aboveground dead biomass

(AGDB) resulting from this unprecedented wildfire,

and estimate the amount of carbon that will eventually

be released from these fires (committed carbon emis-

sions). We also discuss the degradation and impover-

ishment of tropical forests in light of potentially

irreversible ecosystem transitions given a scenario

of more severe or more frequent El Nino events.

Finally, we consider whether accidental fires are likely

to become a permanent feature of many previously

undisturbed neotropical forest ecosystems.

2. Methods

2.1. Study area

The study was conducted on both banks of the Rio

Maro, a tributary of the Rio Arapiuns which flows into

the mouth of the Rio Tapajos, westernmost Para,

Brazil (028440S; 558410W; see Barlow et al. (2002)

or Peres et al. (in press) for a map of the study area and

sample plots). The work took place from October 1998

to March 1999, around 1 year after an accidental

wildfire swept through large parts of the forest on

both sides of the Rio Maro.

Average annual rainfall at the nearest meteorologi-

cal station (Santarem) is 2041 mm epr year (range ¼1287�2538 mm per year 1992–1997; INFRAERO,

1998), with a pronounced dry season typically lasting

3–5 months, normally from July to November. The

recent fires occurred in November–December of 1997,

at the end of the longest dry season in living memory.

A rainless period of 110 days, linked to the 1997–1998

El Nino Southern Oscillation (ENSO) event led to leaf

abscission and permitted sunlight to penetrate the

understorey, allowing normally fire-resistant forests

to become flammable. The soils in this region are

typically sandy (podzolic) and highly permeable,

which further increases the risk of breaching the forest

flammability threshold (Hammond and ter Steege,

1998).

2.2. Data collection

Data were obtained within 20 0.25 ha plots (10 m�250 m) located in burnt and unburnt forest across the

Maro river basin. Sixteen plots were placed >500 m

from the clearly distinguishable fireline, and were

equally split between burnt and unburnt forest on both

banks of the river. A further four plots were placed

perpendicular to, and cutting across the fireline so that

each half of these plots (0.125 ha) lay in either burnt or

unburnt forest. The four forest treatments are hereafter

referred to as BF (burnt forest), UF (unburnt forest),

BFI (burnt forest interface) and UFI (unburnt forest

interface). In order to compare the BF and UF plots

away from the fireline with those cutting across the

fireline interface, tree and basal area densities are

expressed on a per hectare basis. Biomass and carbon

values were converted likewise.

2 T. Haugaasen et al. / Forest Ecology and Management 6154 (2002) 1–11

Within each plot a number of habitat variables were

recorded (see Fig. 1 for a schematic diagram of the

sampling area). The diameter of all trees � 10 cm

DBH (diameter at breast height, or above tallest

buttress) with at least half of their basal stem falling

inside the plot were measured, and their mortality

(defined as the death of all aboveground tissue) was

assessed based on an examination of cambial damage

to the basal tree trunk and the drying and abscission of

leaves. All saplings taller than 1 m were measured

within a 1 m � 100 m (100 m2) subplot (or two

1 m � 50 m subplots in the interface plots), and were

categorised into 10 different size classes of 1 cm (up to

10 cm DBH). Sapling mortality was also defined as

above. Fallen trees and branches � 10 cm in diameter

that lay within the plot were also measured (diameter

and length within plot) and categorised as new or old

falls. Tree trunks and branches on the ground were

classified as recently fallen if they had obvious signs

of being relatively fresh (e.g. leaves still attached),

which in BF presumably fell in the aftermath of the

fires (post-burn).

Canopy openness was quantified with the use of a

convex spherical densiometer at 24 evenly spaced

points within each plot (with 4 readings taken per

point; see Lemmon, 1957). Regeneration of forest

floor vegetation was assessed within 2 m � 2 m quad-

rats laid to the left-hand side of each 250 m transect at

each point where a canopy cover reading was taken

(24 quadrats per plot; see Fig. 1). Within each quadrat

the total percentage cover of undergrowth vegetation

was estimated. Understorey openness was measured

using a 2.5 m graduated pole of 4 cm in diameter held

vertically 15 m to the side of the transect, which was

examined with a pair of 10 � 40 binoculars by an

observer stood on the transect (Fig. 1). A total of 40

readings were recorded within each plot correspond-

ing to the number of 10 cm pole sections (range: 0–25)

that were clearly visible.

2.3. Biomass and carbon estimates

Biomass estimates were calculated using two equa-

tions. We used the regression equation

Y ¼ expð�2:134 þ 2:530 � lnðDÞÞ

where Y is the biomass per tree in kg, and D the DBH

in cm (Brown, 1997). We also used an allometric

model for aboveground dry biomass (AGDB):

AGDB ¼ exp 3:323 þ 2:546 � lnDBH

100

� ��

� 600

derived from 319 destructively sampled trees ranging

from 5 to 120 cm DBH in a structurally similar

Amazonian terra firme (upland) forest located north

Fig. 1. A schematic diagram of the typical sampling protocol undertaken at each forest plot.

T. Haugaasen et al. / Forest Ecology and Management 6154 (2002) 1–11 3

of Manaus, Brazil (Santos, 1996). While there are

many potential problems with using DBH to estimate

forest biomass and carbon (e.g. Fearnside, 1997b), the

Santos equation formulated through the destructive

harvesting of 319 trees near Manaus is considered to

be the most representative of the Brazilian Amazon,

being further supported by results from another site in

eastern Amazonia (Araujo et al., 1997). We therefore

focus our discussion on the results from this equation,

although include the Brown equation for its compara-

tive value. Following Fearnside (1997a) and Brown

(1997), the carbon content of biomass was assumed to

be 50%.

2.4. Data analysis

One-way ANOVAs (with Tukey’s HSD post-hoc

test) were used to test for differences in the forest

structure between the four treatments (Table 1). Data

were checked for normality, and subsequently left

untransformed. In order to conduct the examination

of size-dependent mortality (Fig. 2), data from all

plots was pooled, and split into two categories accord-

ing to whether the forest had burnt or not. This pooling

of all forest plots minimised the stochastic variation in

tree densities inherent within relatively small 0.25 ha

plots, and was necessary to increase the number of

trees sampled in the larger (and hence rarer) DBH

classes. Data for saplings was treated similarly.

Paired-t-tests were used to compare the results of

the two biomass equations.

3. Results

3.1. Tree mortality

A total of 2643 standing trees were measured in the

20 forest plots, with 1341 being found in the UF

(2.5 ha) and 1302 in BF (2.5 ha). The abundance of

trees was strongly size-dependent, and both forest types

showed a reverse J distribution curve for the tree

assemblage as a whole (Fig. 2a). Trees smaller than

20 cm DBH accounted for 74% of all trees in all BF and

Table 1

Habitat data obtained from four groups of forest plots (*, y: subsets from Tukey’s HSD)a

Mean (S.E.) per plot in d.f. F P

UF ðn ¼ 8Þ UFI ðn ¼ 4Þ BFI ðn ¼ 4Þ BF ðn ¼ 8Þ

No. trees per hectare 533.5 19.2 548.0 21.8 528.0 64.7 519.0 20.8 3, 20 0.16 0.92

No. trees dead per hectare 24.0* 3.1 28.0* 4.0 152.0y 41.2 184.0y 9.2 3, 20 34.69 <0.001

Basal area per hectare 121.2 7.3 110.9 5.6 123.1 28.1 105.9 9.2 3, 20 0.49 0.69

Basal area dead per hectare 6.2* 1.2 6.5* 1.3 17.8y 2.1 25.7y 2.5 3, 20 24.53 <0.001

Brown equation

Total biomass per hectare 349.9 33.2 285.4 23.2 342.4 99.1 282.6 32.9 3, 20 0.67 0.58

Dead biomass per hectare 16.9* 4.5 15.1* 3.5 38.6*,y 6.6 593*,y 7.9 3, 20 11.18 <0.001

Potential Carbon loss per hectare 8.5* 2.3 7.6* 1.8 19.3*,y 3.3 29.7y 4 3, 20 11.18 <0.001

Santos equation

Total biomass per hectare 424.3 40.7 344.6 28.4 414.5 120.7 341.7 40.1 3, 20 0.67 0.58

Dead biomass per hectare 20.5* 5.6 18.2* 4.3 46.4*,y 8 71.4y 9.6 3, 20 10.93 <0.001

Potential Carbon loss per hectare 10.2* 2.8 9.1* 2.1 23.2*,y 4 35.7y 4.8 3, 20 10.93 <0.001

Other habitat data

All fallen wood per hectare (m3) 45.7 9.1 50.1 8.5 30.4 3.2 45.2 5.9 3, 20 0.81 0.51

Newly fallen wood per hectare (m3) 0.8* 0.4 2.3* 1.7 12.9y 3.3 10*,y 2.8 3, 20 6.58 0.003

% Saplings dead 4.4* 1.0 3.1* 0.7 63.1y 13.2 76y 5.7 3, 20 47.03 <0.001

% Canopy gap 4.7* 0.6 5.5* 0.5 19.5y 6.2 19.2y 2.1 3, 20 11.34 <0.001

% Ground vegetation cover 20.6 3.5 25.7 3.6 50.9 15.7 48.5 6.9 3, 20 4.59 0.01

Understorey vegetation openness 4.5 0.8 6.6 1.3 9.6 3 9.1 1.2 3, 20 2.98 0.06

a UF: unburnt forest >500 m from the fireline; UFI: unburnt forest plots adjoining and perpendicular to the fireline; BF: burnt forest

>500 m from the fireline; BFI: burnt forest plots adjoining and perpendicular to the fireline.


UF plots. There was no significant difference between

the mean overall number of trees per hectare or the

mean basal area per hectare in any forest type (Table 1).

As expected, however, tree mortality, was clearly

greater in BF plots; 36% of all trees in BF plots were

classified as dead, compared to only 4.5% of those in

UF plots. The number of dead trees per hectare and

the mean dead basal area per hectare were thus

significantly greater in the BF plots than in the

UF plots. Distance to the fireline did not appear

Fig. 2. (a) Frequency distribution of the size of standing trees expressed in terms of DBH in all BF and UF plots. Relationship between tree

DBH for UF and BF plots and tree survival (b), and (c) the estimated AGDB, using both the Brown (1997) and Santos (1996) equations.


to have a significant effect on tree mortality, and the

interface plots were not significantly different from

those >500 m from the fireline in either BF or UF

(Table 1).

In order to examine size-dependent mortality, all

trees were classed according to their DBH. Within the

BF, there was a strong positive relationship between

the DBH class of the tree and the percentage of live

trees (r2 ¼ 0:77, F1;10 ¼ 33:7, P ¼< 0:001), with

smaller stems succumbing to higher mortality

(Fig. 2b). No such relationship was apparent in UF

plots, where trees appeared to have a mortality peak

around 30–35 cm DBH (Fig. 2b).

3.2. Sapling mortality

Mortality of saplings averaged 76% in the eight BF

plots and 63.1% in the four BFI plots. Both of these

means were significantly different from the 3 to 4%

background mortality rate found in the UF treatments

(Table 1). The proportion of dead saplings in each

DBH class, however, was not significantly related to

stem diameter in either BF or UF plots (P < 0:5 in

both the cases).

3.3. Fallen dead wood

There was no significant difference between the

total volume of fallen wood per hectare in the four plot

categories (Table 1), though the mean values were

variable, ranging from 30.4 to 50.1 m3 per hectare.

However, the BFIs had a significantly greater number

of recently fallen trees than either unburnt forest

treatment (UF and UFI).

3.4. Canopy openness and understorey

regeneration

Canopy openness in all BF plots was significantly

greater than that in UF (Table 1), this being largely

explained by the high density of the leafless crowns of

dead canopy trees left standing in the BF. Although

there was significant differences in the percentage

ground vegetation cover in the understorey, the Tukey’s

post-hoc test showed that no treatment was signifi-

cantly different from another. However, understorey

vegetation cover was found to be significantly related

to percentage canopy openness (Fig. 3). The regen-

eration of the BF understorey was highly variable in

Fig. 3. Relationship between canopy openness and understorey vegetation cover for forest plots >500 m from the fireline. Error bars represent

standard errors. Regression line was fitted to mean values.


terms of structure and species composition, though

Cecropia spp., Palicourea guianensis (Rubiaceae),

and Aparisthmium cordatum (Euphorbiaceae) were

the most common pioneer tree species.

3.5. Forest biomass and potential carbon loss

Using either the Brown or Santos equations to

calculate tree biomass, there were no significant dif-

ferences between the mean total biomass density

across any of the forest treatment types (Table 1).

However, there were significant differences between

the mean dead biomass per hectare, with both equa-

tions displaying similar trends. The eight BF plots

contained a significantly greater dead biomass per

hectare than the UF and UFI, while the BFI was

intermediate and did not differ significantly from

any of the other forest treatments. Fig. 2c shows

how the total AGDB is distributed among each

DBH class comparing the two equations.

The Santos equation added considerable tonnage to

the biomass estimates, and the total estimates for both

UF and BF from each equation were significantly

different (mean Santos UF S:E: ¼ 424:3 40:7ðn ¼ 8Þ; mean Brown UF S:E: ¼ 349:9 33:2ðn ¼ 8Þ; paired-t ¼ 9:6, d:f: ¼ 7, P ¼< 0:001; mean

Santos BF S:E: ¼ 341:7 40:1 ðn ¼ 8Þ; mean

Brown BF S:E: ¼ 282:6 32:9 ðn ¼ 8Þ; paired-

t ¼ 8:2, d:f: ¼ 7, P ¼< 0:001).

The total potential carbon loss (due to the increase

in dead AGDB) in BF plots was 3.5 times than that

found in the equivalent UF plots, an increase of

25.5 t C/ha using the Santos equation. The difference

across the interface was less dramatic, though the total

potential loss in the BFI was still 2.5 times than that of

the UFI, an increase of 14.1 t C/ha using the Santos

equation.

4. Discussion

4.1. Tree mortality and biomass

The mortality of 36% of all trees � 10 cm DBH in

the BF plots was 4.5 times greater than that reported

for seven post-burn plots of 750 m2 at three forest sites

in Roraima (Barbosa and Fearnside, 1999) but closely

matches that of previously logged eastern Amazonian

forests near Paragominas (Holdsworth and Uhl, 1997;

Cochrane and Schulze, 1999; Table 2). This study thus

shows that seasonal central Amazonian forests such as

around the Maro can be severely affected by surface

wildfires during strong El Nino events, even if their

history of logging disturbance has been negligible.

Mortality and AGDB estimates in this study may also

have been underestimated. The study by Peres (1999)

carried out in the same region immediately after the

fires estimated a standing dead biomass at 43.3 t/ha, a

figure almost 40% lower than our estimates, suggest-

ing that there had been an increase in mortality in the

following year. We may expect this increase in mor-

tality to continue following exposure to sublethal

thermal stress; other studies in both Amazonia (Holds-

worth and Uhl, 1997) and Sumatra (Sunarto, 2000)

have shown mortality to increase for up to 2 years after

the initial fire disturbance as injured trees gradually

succumb to infections of fungi or other pathogens, and

eventually die whether standing or fallen. Further-

more, the occurrence of both visually and acoustically

detected tree falls plots recorded during the time of the

study was five times higher in the BF than UF (J.

Barlow, unpublished data), further opening up the

forest canopy and knocking over trees that had sur-

vived the initial fire. Another cause of underestimation

comes from the exclusion of woody lianas and vines in

Table 2

Mortality of individuals �10 DBH (n/ha) and aboveground

biomass (t/ha) from five different studies in Amazonian Brazil

Mortality

(n/ha)

Aboveground

biomass (t/ha)

Study

Dead 147 71 Cochrane and

Schulze (1999)% Dead 38 35.5

Totals 384 200

Dead 161 Holdsworth and

Uhl (1997)% Dead 44

Totals 367

Dead 46 17.4 Barbosa and

Fearnside (1999)% Dead 7.9 7.9

Totals 585 219.7

Dead 68 16.1 Santos et al.

(1998)% Dead 16

Totals 425

Dead 184 71.4a This study

% Dead 35.5 20.9a

Totals 519 341.7a

a Numbers derived with the Santos equation.


this study. Despite being extremely susceptible to fire

damage (Cochrane and Schulze, 1999), large (�10 cm

DBH) lianas were fairly rare in this region and were

not considered here.

Tree mortality in the BF plots appeared to be

strongly size-dependent, with small stems (10–

20 cm DBH) in BF suffering higher mortality. Uhl

and Kauffman (1990) suggest that this size-dependent

mortality may be a consequence of differences in bark

thickness, as thicker bark is better able to protect the

cambium layer from heat stress. This is consistent with

a recent study in the Arapiuns basin where bark

thickness appeared to be an important morphological

correlate of tree survival (Barlow et al., in press). In

contrast, tree mortality in UF plots peaked at 30–

40 cm DBH, which was presumably due to other

causes.

4.2. Edge and interface effects

Initially we predicted that burn intensity (and hence

tree mortality) may have been lower close to the

fireline, as interviews with local people indicated that

the fires had been reduced in intensity by the rains that

eventually extinguished them. Furthermore, as edges

are associated with an increase in tree mortality in

tropical forest fragments (Mesquita et al., 1999) we

also expected a higher mortality and treefall rate in the

UFI close to the fireline. While we consider it note-

worthy that these predictions were supported numeri-

cally (Table 1), this must be qualified by the small

sample sizes at the interfaces, and the lack of statistical

support.

4.3. Sapling mortality and understorey regeneration

As expected from the size-dependent mortality of

trees, mortality was very high amongst the pre-burn

saplings in all BF plots matching other reports on post-

fire sapling mortality (Holdsworth and Uhl, 1997).

Following the fires, the greatly increased light envir-

onment in the understorey seemed to favour the rapid

growth of early successional tree species, although in

other areas that had apparently been severely burned

the growth of aggressive bamboo and sedges appeared

to inhibit seedling regeneration, which is confirmed by

findings elsewhere in the Amazon (Nepstad et al.,

1991).

4.4. Positive feedback mechanisms

One of the most alarming features of these fires is

their potential to trigger a positive feedback system,

potentially leading to the progressive impoverishment

and degradation of vast expanses of tropical forest

(Cochrane et al., 1999; Nepstad et al., 1999b). The

high post-burn tree mortality significantly reduced

canopy cover, and will therefore create a hotter and

drier microclimate as more solar radiation passes

through to the understorey, increasing the rate of fuel

drying (Uhl and Buschbacher, 1985; Uhl and Kauff-

man, 1990; Holdsworth and Uhl, 1997; Nepstad et al.,

1999b). Furthermore, the dead and dying trees that fall

to the forest floor add to the combustible fuel layer. We

calculate that nearly 14 m3 /ha of coarse fuel has been

added to the forest floor in BF just 1 year after the fire,

and the greatly augmented treefall rate suggests that

much more will be added in the near future. Further-

more, given the standing and fallen dead trees, sap-

lings and lianas, fuel continuity will be greatly

enhanced from the forest floor up to the canopy.

Combined with another prolonged dry season, these

factors could increase the risk of a further fire in these

forests. The resulting second burn is likely to be much

more severe than the first, and Cochrane et al. (1999)

expect up to 98% of all remaining trees to be suscep-

tible to recurrent fires. Moreover, pioneer saplings

may be no more likely to survive a second burn than

non-pioneers (Cochrane and Schulze, 1999) meaning

that if fire return rates are frequent enough, new

regeneration will effectively be prevented from reach-

ing maturity and these forest ecosystems will shift

towards scrub-savannahs where arborescent plants

will be far less prevalent. Indeed, our experience in

forest areas that local people report to have burnt twice

support these predictions, as they are heavily domi-

nated by bamboo (Guadua spp.), with very few stand-

ing trees remaining (J. Barlow, unpublished data).

Without drastic fire prevention measures aimed at

excluding fire from local agricultural systems, it seems

unlikely that these forests will ever recover in time to

prevent a subsequent fire. Many years of regrowth are

necessary for BFs to recover the fire resistance of

primary forests (Cochrane and Schulze, 1998) because

tall trees are needed to establish the full shade and

moist microclimate typical of the primary forest inter-

ior. Selectively logged forests have been found to take


5–6 years (Mason, 1996) and 7–12 years (Johns, 1989)

to return to approximately pre-disturbance conditions

and we could expect a similar or greater time-span

after fire disturbance. Therefore, the evidence that El

Nino events are increasing in frequency and severity in

recent years (Trenberth and Hoar, 1996; Timmermann

et al., 1999) does not bode well for the future of these

forests.

4.5. Carbon loss

Substantial carbon loss is derived from surface

wildfires through the long-term decomposition of

unburnt dead biomass and the initial combustion of

dead leaves, twigs and branches on the forest floor,

with fine branches and the leaf litter making up as

much as 70% of all fuel on the forest floor (Chambers

et al., 2000). However, the small litter and rootmat on

the forest floor comprise only around 3.7–8.0% of the

aboveground biomass (Kauffman et al., 1995) and

saplings and small trees (<10 cm DBH) account for

only 6.2% of the total AGDB resulting from a surface

fire in our study region (Peres, 1999). We therefore

emphasise the potential (or committed) carbon loss

from the large tree mortality (�10 cm DBH), which is

likely to be far more important to the overall residual

fuel load. In the absence of a subsequent fire, carbon

will be released from the standing and fallen dead

wood through bacterial decomposition and termite

activity, which occurs largely over the first decade

following death (Melillo et al., 1996; Fearnside,

1997a). Although there was an almost 8-fold increase

in tree mortality in the BF plots, the size-dependent

mortality (which selected for smaller trees; see Fig. 2b)

meant that this only resulted in a 3.5-fold increase in

the amount of dead biomass and potential carbon loss.

Our AGDB estimates closely resemble those found by

Cochrane and Schulze (1999) in eastern Amazonia,

but far exceeds those from other studies such as

Barbosa and Fearnside (1999) and Santos et al.

(1998; Table 2).

Using the Santos equation the potential loss of an

additional 25.5 t C/ha in BF (above the background

level of 10.2 t C/ha in unburnt controls) dramatically

increases the amount of carbon released from Ama-

zonian forests. Deforestation activities in the Brazilian

Amazon emit approximately 250�350 � 106 t C

annually (Fearnside, 1999), or 4–5.5% of the annual

global flux of carbon to the atmosphere caused by

human activities, which in 1998 was an estimated

6318 � 106 t C (Worldwatch News Brief, 1999).

However, if our results (based on short-term estimates

of tree mortality) are extrapolated to the rest of the

Brazilian Amazon, the burning of all the primary

forests that were prone to surface wildfires after the

1998 dry season (270,000 km2; cf. Nepstad et al.,

1999b) would commit 6885 � 106 t C to be released,

supporting the assertion of Nepstad et al. (1999b) that

committed carbon estimates from the Brazilian Ama-

zon could be doubled during strong El Nino years such

as 1998. Furthermore, fire degrades both the forest

structure and species composition and will presum-

ably disrupt the important role Amazonian forest play

as a sink for global carbon emissions (Chambers et al.,

2001).

However, caution should be applied in making wide

extrapolations across Amazonia from a restricted net-

work of sampling sites. Soils in the study area were

characterised by sand fractions, and therefore may

support a lower standing biomass than other Amazo-

nian forests on clay soils (Laurance et al., 1999).

However, the lower standing biomass may be offset

by the increased root biomass for the acquisition of

soil nutrients (Wilson and Tilman, 1991), an important

point as total root and soil carbon represent almost half

the total carbon content in South American tropical

forests (Dixon et al., 1994), and little is known about

belowground carbon loss. Furthermore, much of the

forests at risk of burning will also be on sandy soils

because of their poor water retention capacity (Ham-

mond and ter Steege, 1998).

5. Conclusion

In brief, perhaps the most important ecological

effect of surface fires is that they increase the like-

lihood that fire will become a permanent feature of the

forest ecosystem, because of decreasing intervals

between consecutive El Nino events and the long

process of regrowth which is necessary for secondary

forest to recover the fire resistance of primary forests.

Consequently, the fires disrupt what is a well-balanced

carbon cycle between earth and atmosphere (Pielke

et al., 1998) as closed-canopy forests hold more

carbon per unit area in vegetation and soil than any


of the ecosystems replacing them (Melillo et al.,

1996).

Fearnside (1999) acknowledges that the avoidance

of ‘natural’ disasters represents a major factor in the

carbon balance of tropical forests, and one that should

be addressed in global warming mitigation strategies,

because the vast areas involved ensures that carbon

emissions are substantial. However, surface fires in

neotropical forests are often portrayed as fairly innoc-

uous, with low-impact fires resulting in a relatively

small percentage of dead trees (e.g. Fearnside, 1999).

In this paper we have shown that 36% of all midstorey

and canopy trees may die as a result of these fires,

representing a loss of an extra 21.6 t C/ha. Surface

fires are of huge global concern both in terms of

biodiversity conservation and ecosystem services,

and should be taken more seriously than ever before

because of the vast areas that will become prone to

fires at the end of future El Nino dry seasons.

Acknowledgements

This study was funded by the Centre for Applied

Biodiversity Science (CABS) of Conservation Inter-

national and the Josephine Bay and Michael Paul

Foundation. Jos Barlow’s fieldwork was funded by

a NERC Ph.D. studentship at the University of East

Anglia. We are especially thankful to the political

leadership of the Reserva Extrativista do Tapajos-

Arapiuns and the villagers of Cachoeira do Maro,

Sao Jose and Porto Rico for allowing us to conduct

this study, and to the local assistance of Nan, Arnei and

Torozinho. Both Rionaldo and Edith de Santos pro-

vided enormous logistical assistance during all stages

of the project.

References

Araujo, T.M., Higuchi, N., Junior, J.A.D., 1997. Comparacao de

Metodos para Determinar Biomassa na Regiao Amazonica.

Anais da Academia Brasileira de Ciencias 68 (1).

Barbosa, R.I., Fearnside, P.M., 1999. Incendios na Amazonia

Brasileira: estimativa da emissao de gases do efeito estufa pela

queima de diferentes ecossistemas de Roraima na passagem do

evento ‘‘El Nino’’ (1997/1998). Acta Amazonica 29, 513–534.

Barlow, J., Haugaasen, T., Peres, C.A., 2002. Effects of ground fires

on understorey bird assemblages in Amazonian forests. Biol.

Conserv. 105, 157–169.

Barlow, J., Lagan, B.O., Peres, C.A., in press. Morphological

correlates of fire-induced tree mortality in a central Amazonian

forest. J. Trop. Ecol.

Brown, S., 1997. Estimating biomass and biomass change of

tropical forests. FAO Forestry Paper 134.

Chambers, J.Q., Higuchi, N., Schimel, J.P., Ferreira, L.V., Melack,

J.M., 2000. Decomposition and carbon cycling of dead trees in

tropical forests of the central Amazon. Oecologia 122, 380–

388.

Chambers, J.Q., Higuchi, N., Tribuzi, E.S., Trumbore, S.E., 2001.

Carbon sink for a century. Nature 410, 429.

Cincotta, R.P., Wisnewski, J., Engelman, R., 2000. Human

population in the biodiversity hotspots. Nature 404, 990–992.

Cochrane, M.A., 2001. In the line of fire—understanding the

impacts of tropical forest fires. Environment 43, 28–38.

Cochrane, M.A., Schulze, M.D., 1998. Forest fires in the Brazilian

Amazon. Conserv. Biol. 12, 948–950.

Cochrane, M.A., Schulze, M.D., 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.

Cochrane, M.A., Alencar, A., Schulze, M.D., Souza, C.M.,

Nepstad, D.C., Lefebvre, P., Davidson, E.A., 1999. Positive

feedbacks in the fire dynamic of closed canopy tropical forests.

Science 284, 1832–1835.

Dale, V.H., Pearson, S.M., 1997. Quantifying habitat fragmentation

due to land-use change in Amazonia. In: Laurance, W.F.,

Bierregaard, R.O. (Eds.), Tropical Forest Remnants: Ecology,

Management, and Conservation of Fragmented Communities.

University of Chicago Press, Chicago, pp. 400–409.

Dixon, R.K., Brown, S., Houghton, R.A., Solomon, A.M., Trexler,

M.C., Wisniewski, J., 1994. Carbon pools and flux of global

forest ecosystems. Science 263, 185–190.

Fearnside, P.M., 1997a. Greenhouse gases from deforestation in

Brazilian Amazonia: net committed emissions. Clim. Change

35, 321–360.

Fearnside, P.M., 1997b. Wood density for estimating forest biomass

in Brazilian Amazonia. For. Ecol. Manage. 90, 59–87.

Fearnside, P.M., 1999. Forests and global warming mitigation in

Brazil: opportunities in the Brazilian forest sector for responses

to global warming under the clean development mechanism.

Biomass Bioenergy 16, 171–189.

Goldhammer, J.G., 1999. Forests on fire. Science 284, 1782–

1783.

Hammond, D.S., ter Steege, H., 1998. Propensity of fire in Guianan

rainforests. Conserv. Biol. 12, 944–947.

Haugaasen, T., 2000. Effects of ground fires on understorey birds in

central Amazonia, Brazil. M.Sc. Thesis. University of East

Anglia, Norwich.

Holdsworth, A.R., Uhl, C., 1997. Fire in Amazonian selectively

logged rain forest and the potential for fire reduction. Ecol.

Appl. 7, 713–725.

INFRAERO, 1998. Infraero rainfall data. Superintendencia da

Infraero, Sala AIS/SBSN, Santarem Airport, Santarem.

Johns, A.D., 1989. Recovery of a peninsular Malaysian rainforest

avifauna following selective timber logging: the first twelve

years. Forktail 4, 89–105.


Kauffman, J.B., Cummings, D.L., Ward, D.E., Babbitt, R., 1995.

Fire in the Brazilian Amazon. 1. Biomass, nutrient pools, and

losses in slashed primary forests. Oecologia 104, 397–408.

Kinnaird, M.F., O’Brien, T.G., 1998. Ecological effects of wildfire

on lowland rainforest in Sumatra. Conserv. Biol. 12, 954–956.

Laurance, W.F., Laurance, S.G., Delamonica, P., 1998. Tropical

forest fragmentation and greenhouse gas emissions. For. Ecol.

Manage. 110, 173–180.

Laurance, W.F., Fearnside, P.M., Laurance, S.G., Delamonica, P.,

Lovejoy, T.E., Rankin-de Merona, J., Chambers, J.Q., Gascon,

C., 1999. Relationship between soils and Amazon forest biomass:

a landscape-scale study. For. Ecol. Manage. 118, 127–138.

Lemmon, P.E., 1957. A new instrument for measuring forest

overstory density. J. For. 55, 667–668.

Mason, D., 1996. Responses of Venezuelan understorey birds to

selective logging, enrichment strips and vine cutting. Biotro-

pica 28, 296–309.

Melillo, J.M., Houghton, R.A., Kicklighter, D.W., McGuire, A.D.,

1996. Tropical deforestation and the global carbon budget.

Ann. Rev. Energy Environ. 21, 293–310.

Mesquita, R.C.G., Delamonica, P., Laurance, W.F., 1999. Effect of

surrounding vegetation on edge-related tree mortality in

Amazonian forest fragments. Biol. Conserv. 91, 129–134.

Nepstad, D.C., Uhl, C., Serrao, E.A.S., 1991. Recuperation of a

degraded Amazonian landscape—forest recovery and agricul-

tural restoration. Ambio 20, 248–255.

Nepstad, D.C., Moreira, A.G., Alencar, A.A., 1999a. Flames in the

Rain Forest: Origins, Impacts and Alternatives to Amazonian

Fires. The Pilot Program to Conserve the Brazilian Rain Forest,

Brasilia, Brazil.

Nepstad, D.C., Verıssimo, A., Alencar, A., Nobre, C., Lima, E.,

Lefebvre, P., Schlesinger, P., Potter, C., Moutinho, P., Mendoza,

E., Cochrane, M., Brooks, V., 1999b. Large-scale impoverish-

ment of Amazonian forests by logging and fire. Nature 398,

505–508.

Pahari, K., Murai, S., 1999. Modelling for prediction of global

deforestation based on the growth of human population. ISPRS

J. Photogramm. Remote Sens. 54, 317–324.

Peres, C.A., 1999. Ground fires as agents of mortality in a central

Amazonian forest. J. Trop. Ecol. 15, 535–541.

Peres, C.A., 2000. Effects of subsistence hunting on vertebrate

community structure in Amazonian forests. Conserv. Biol. 14,

240–253.

Peres, C.A., 2001. Synergistic effects of subsistence hunting and

habitat fragmentation on Amazonian forest vertebrates. Con-

serv. Biol. 15, 1490–1505.

Peres, C.A., Barlow, J., Haugaasen, T., in press. Vertebrate

assemblage responses to wildfire disturbance in a central

Amazonian forest. Oryx.

Pielke, R.A., Avissar, R., Raupach, M., Dolman, A.J., Zeng, X.,

Denning, A.S., 1998. Interactions between the atmosphere and

terrestrial ecosystems: influence on weather and climate.

Global Change Biol. 4, 461–475.

Putz, F.E., Blate, G.M., Redford, K.H., Fimbel, R., Robinson, J.,

2001. Tropical forest management and conservation of

biodiversity: an overview. Conserv. Biol. 15, 7–20.

Robinson, J.G., Bennett, E.L. (Eds.), 2000. Hunting for Sustain-

ability in Tropical Forests. Columbia University Press, New

York.

Sanford, R.L., Saldarriaga, J., Clark, K.E., Uhl, C., Herrera, R.,

1985. Amazon rain-forest fires. Science 227, 53–55.

Santos, J., 1996. Analise de modelos de regressao para estimar a

fitomassa de floresta tropical umida de terra-firme da Amazonia

Central. Ph.D. Thesis. Universidade Federal de Vicosa, Minas

Gerais, Brazil.

Santos, J.R., Pardi Lacruz, M.S., Araujo, L.S., Xaud, H.A.M.,

1998. El proceso de queima de biomassa de bosque tropical y

de sabanas en la Amazonia Brasileira: experiencias de

monitioreo com dados opticos y de microondas. Revista Serie

Geografica 7, 97–108.

Skole, D., Tucker, C., 1993. Tropical deforestation and habitat

fragmentation in the Amazon: satellite data from 1978–1988.

Science 260, 1905–1910.

Sunarto, S., 2000. Survival of plant communities and dynamics of

forest ecosystems in response to fire: a long-term study from

Sumatra. M.Sc. Thesis. University of East Anglia, Norwich.

Timmermann, A., Oberhuber, J., Bacher, A., Esch, M., Latif, M.,

Roeckner, E., 1999. Increased El Nino frequency in a climate

model forced by future greenhouse warming. Nature 398, 694–

697.

Trenberth, K.E., Hoar, T.J., 1996. The 1990–1995 El Nino-southern

oscillation event: longest on record. Geophys. Res. Lett. 23,

57–60.

Turcq, B., Sifeddine, A., Martin, L., Absy, M.L., Soubies, F.,

Suguio, K., Volkmer-Ribeiro, C., 1998. Amazonia rainforest

fires: a lacustrine record of 7000 years. Ambio 27, 139–142.

Uhl, C., Buschbacher, R., 1985. A disturbing synergism between

cattle ranch burning practices and selective tree harvesting in

the eastern Amazon. Biotropica 14, 265–268.

Uhl, C., Kauffman, J.B., 1990. Deforestation, fire susceptibility,

and potential tree responses to fire in the eastern Amazon.

Ecology 71, 37–449.

Whitmore, T.C., 1997. Tropical forest disturbance, disappearance,

and species loss. In: Laurance, W.F., Bierregaard, R.O. (Eds.),

Tropical Forest Remnants: Ecology, Management, and Con-

servation of Fragmented Communities. University of Chicago

Press, Chicago, pp. 3–12.

Wilson, S.D., Tilman, D., 1991. Components of plant competitions

along an experimental gradient of nitrogen availability. Ecology

72, 1050–1065.

Worldwatch News Brief, 1999. World carbon emissions fall. http://

www.worldwatch.org/alerts/990727.html.


HTTP://WWW.WORLDWATCH.ORG/ALERTS/990727.HTML

HTTP://WWW.WORLDWATCH.ORG/ALERTS/990727.HTML

Surface wildfires in central Amazonia: short-term impact on forest structure and carbon loss

Documents

Transcript of Surface wildfires in central Amazonia: short-term impact on forest structure and carbon loss