Biodiversity and Conservation 13: 16111632, 2004.

# 2004 Kluwer Academic Publishers. Printed in the Netherlands.

Population structure and dynamics of a neotropical

palm in fire-impacted fragments of the Brazilian
Atlantic Forest
ALEXANDRE F. SOUZA1 and FERNANDO R. MARTINS2,*
1

Programa de Pos-Graduacao em Ecologia, Universidade Estadual de Campinas, Caixa Postal 6109,

Campinas 13083-970, SP, Brazil; 2Departamento de Botanica, IB, Universidade Estadual de Campinas,
Caixa Postal 6109, Campinas 13083-970, SP, Brazil; *Author for correspondence (e-mail: fmartins@
unicamp.br; fax: 55-19-3289-4457)
Received 10 December 2002; accepted in revised form 6 June 2003

Key words: Attalea humilis, Matrix population models, Sensitivity analysis, Tropical forest
Abstract. The demography of the acaulescent palm Attalea humilis was studied from 1996 to 1999 in
three forest fragments (1.6, 6.4, and 9.9 ha) in southeastern Brazil (228300 228330 S, 428150 428190 W).
A human-induced fire damaged the populations in AugustSeptember 1997. Palm density was higher in
the medium-sized fragment, where it was concentrated in a single disturbed sector. Population ontogenetic stage structure was dependent on fragment and year, but in general later stages were more
abundant than initial ones. Population finite growth rates projected from linear, stage-structured demographic models indicated that the populations were increasing before fire. Population growth rate
in the smallest fragment was larger due to a single recruitment event, being most sensitive to changes
in survival and growth of seedlings, virgins and the immature, and relatively insensitive to changes in
shrinkage and fecundity. Observed growth rates for the entire study period showed that populations
remained in equilibrium after fire, but the population in the medium-sized fragment showed a declining
tendency. Population structure was resilient to fire, which caused a transitory concentration of individuals in the medium-sized stage. The self-sustainability of the population in the largest, most
preserved fragment suggests that habitat-protection measures may suffice to ensure the persistence of
populations in small fragments. However, population fluctuations and exclusion from hyper-disturbed
sectors in the smallest fragment suggest that reduced fragment area, habitat degradation and humaninduced fires can interact to threaten the persistence of even abundant, preadapted to fire, species like
A. humilis.

Introduction
One current view in conservation biology is that the conservation of any particular
species in nature will only be successful through the maintenance of its constituent
populations (Simberloff 1988; Primack 1995; Meffe and Carroll 1997). This view
emerges from the notion that the population is the ecologically and evolutionarily
functional unit (Pulliam and Dunning 1997), and that genetic variability, ecological
and evolutionary processes, and ecosystem functioning are dependent on the presence of self-maintaining natural populations through most of their original geographic range (Meffe and Carroll 1997). Accordingly, the extinction of a species
occurs through the successive extirpations of its constituent populations, turned
locally unsustainable after direct or indirect changes in habitat quality.

1612
One of the main threats to natural populations in terrestrial ecosystems is habitat
fragmentation. Fragmented sub-populations depend largely on migration of individuals between habitats to maintain populations in unfavourable habitats or
recolonise available habitat patches (Pulliam and Dunning 1997), a dependence
which will be stronger as the sub-populations become small and isolated (Case
2000). Large-bodied species with short-range dispersal systems are especially
susceptible to be fragmented into spatially isolated sub-populations, with little
migration between them, each so small that demographic accidents rapidly drive
the population to extinction (Pimm 1991). In the tropics, high deforestation rates
have converted formerly continuous rain forests to anthropogenic landscapes,
where forest habitats are reduced to scattered forest fragments of varying sizes and
quality (Noss and Csuti 1997; Whitmore 1997).
Forest fragments undergo a number of small-scale edge effects such as microclimatic changes (Camargo and Kapos 1995; Kapos et al. 1997; Turton and
Freiburger 1997; Didham and Lawton 1999), increased density and biomass of
pioneer trees and lianas (Kapos et al. 1997; Oliveira-Filho et al. 1997; Tabanez et al.
1997), and increased mortality (Laurance et al. 1998b) and recruitment (Laurance
et al. 1998a) of trees. Furthermore, large-scale edge effects like fire, originating in
the commonly aggressive anthropogenic matrix in which most forest fragments are
embedded, frequently affect even large fragments (> 1000 ha; Laurance 2000;
Cochrane and Laurance 2002), which may be completely destroyed by receding
edges responding to continuous external pressure (Gascon et al. 2000).
Despite the recognised conservation value of forest remnants (Janzen 1988; Turner
1996; Maycock and Fahselt 2001), and the well-established threats fragmentation
may bring to natural populations, surprisingly few studies have devoted attention to
the population ecology of individual long-lived plant species in tropical forest
fragments (but see Barrow et al. 1993; Ratsirarson et al. 1996; Silva Matos et al.
1999; Souza and Martins 2002). Among other reasons, such studies are needed
because long-lived, iteroparous plants may build up remnant populations under
conditions where the completion of the whole life cycle is not possible (Eriksson
1996). Such populations may respond favourably to habitat improvement under
management policies, but are condemned if unfavourable conditions, represented by
forest degradation, prolong for more than the lifetime of its individuals.
In this paper we present data on distinct aspects of the population structure and
dynamics of Attalea humilis, an endemic and abundant understorey palm of the
Brazilian Atlantic Forest in three fire-impacted forest fragments, using Lefkovitch,
stage-based matrix models. Previous spatial analyses of the same data set (Souza
and Martins 2002) revealed that the spatial distribution of the populations was
strongly influenced by large-scale spatial heterogeneity of the fragments, that is, the
occurrence of preserved versus disturbed sectors, with most individuals concentrated in moderately impacted areas. Here we ask whether the effects of habitat
variation on population spatial structure are associated with corresponding changes
in demographic rates and population structure. Our main objective was to describe
population growth in fragments of contrasting quality and evaluate the impact of
fire on population structure. Apart from the results of this project (Souza et al.

1613
Table 1. General characterisation of the study forest fragments in the National Biological Reserve of
Poc o das Antas, Rio de Janeiro State, southeast Brazil.
Fragment

Area
(ha)

Number of
10 m  10 m plots

Sampled
area (%)

Trees ha1
(DBH  10 cm)a

Mean tree
heighta (m)

Litterfallb
(t ha1 y 1)

I
II
III

1.6
6.4
9.9

29
61
70

18.1
9.5
7.1

426
514
797

7.6
9.4
9.3

4.1
5.3
5.7

Silva Matos et al. (1998a).

Silva Matos et al. (1998b).

2000; Souza and Martins 2002), there has been no ecological study done on this
species.
Specifically, we asked the following questions: (1) what are the probabilities of
survival, growth and shrinkage for each ontogenetic stage?; (2) how do these descriptors differ between forest fragments? This question is based on the assumptions, strengthened by some initial evidence (see Methods: Study site and species),
that fragmentation altered the microclimate and structure of the fragments studied
in a number of ways, and that each species responds to these changes through its
demographic descriptors; (3) how does the stage structure of the populations vary
through space and time?; and (4) what is the effect of fire on both population
growth and stage structure in fragments differing in size, shape and structure?

Methods
Study site and species
The study was carried out in the National Biological Reserve of Poc o das Antas
(228300 228330 S, 428150 428190 W), municipality of Silva Jardim, state of Rio de
Janeiro, southeastern Brazil. The Reserve covers ca. 5000 ha of lowland Atlantic
Rainforest (floresta pluvial baixo-montana, according to Rizzini 1979), with
maximum elevation of 205 m a.s.l., and comprises a mosaic of vegetation patches at
different successional phases (IBDF/FBCN 1981). The Reserve is surrounded by
pastures, cultures and secondary forests. The climate is classified as Walter and
Lieths Equatorial type (Walter 1971), with mean annual rainfall of ca. 2100 mm
(19871997 data). There is no distinct dry season and, despite large variation
between years, a drier period, in which episodic drought occurs, normally extends
from May to August.
Field data were collected in three forest fragments, which are part of the complex
locally known as the Barbados archipelago, a mosaic of eight forest fragments
located on small hills near the eastern margin of the Sa o Joa o river. Distances
between adjacent fragments ranged from ca. 33 to 190 m. The studied fragments
were 1.6, 6.4 and 9.9 ha, and showed structural and physical differences (Table 1).
The smallest fragment is subjected to higher air and soil temperatures and lower air

1614
relative humidity and soil moisture content, as well as larger wind penetration
distances (Silva Matos et al. 1998b). It is almost entirely disturbed, and probably
has no remaining core habitat, with its eastern sector dominated by dense stands
of Trema micrantha (L.) Blum. (Ulmaceae). The matrix surrounding the fragments,
originally represented by a low-stature, waterlogged forest, was drained in 1981
after the alteration of the water regime of the Sa o Joa o River by the building of the
Juturnaba dam. The fragments are now embedded in a vegetation matrix dominated by grasses, patches of the bracken fern (Pteridium aquilinum (L.) Kuhn.,
Polypodiaceae) and isolated stands of Cecropia sp. (Cecropiaceae) and Trema
micrantha. Matrix soils are classified as Histosols (USDA 1990), a type of organic
soil originating from peat. In the fragments, soils form an indiscriminate grouping
of alic Oxisols/Ultisols, with moderate A horizon and clayey to very clayey texture
(Takizawa 1995). Since the drainage of the matrix in 1981, the area suffered recurrent, human-induced fires in 1982, 1990, 1993 and 1997. These fires have taken
place during the driest months and frequently invaded the fragments, causing
variable damage.
Attalea humilis Mart. ex. Spreng., hereafter Attalea, is a monoecious, solitary
palm with a short, subterranean stem (saxophone-axis type; Tomlinson 1990). The
apical meristem, protected by the surrounding leaf sheaths, is also subterranean.
The aerial portion is composed by the crown, which has 516 pinnatifid, spineless,
up to 7 m long leaves (see Souza et al. (2000) for a morphological description). The
species is endemic to the central-southeastern Atlantic Forest, ranging from the
state of Bahia to the state of Sa o Paulo (Henderson et al. 1995; Lorenzi et al. 1996).
Through most of its range, it occupies mainly the crests or slopes of hills, but rarely
flooded sites. In the study area, Attalea is absent from the matrix surrounding the
forest fragments. Based on morphological and morphometrical analysis, we distinguished five ontogenetic stages in Attalea (Souza et al. 2000): seedlings have
entire, distichous leaves; juveniles have bifid, incompletely segmented ones; immature individuals produce completely segmented leaf blades, being distinguished
from virgins by having fewer leaves and smaller but highly variable youngest leaf
blade length; and reproductive palms are recognised by the production of flowers
and fruits.

Data collection
Palms were censused along two perpendicular 10 m wide transects that cut across
each fragment. The transects were divided into 10 m  10 m contiguous plots, and
covered 718% of each fragment area (Table 1). All individuals of Attalea were
tagged with a numbered plastic label from July 1996 to January 1997 (1996 field
survey), and the fates of the plants were subsequently monitored in JulyAugust
1997 (1997 field survey), when all new plants were also tagged. At the end of the
1997 drier period (AugustSeptember), a fire damaged ca. 1000 ha of the Reserve,
including the study area. Each of the three fragments studied was burned differently
by the fire. The largest fragment (III) was burnt mainly in the edge. The medium-

1615
sized fragment (II) was burnt in the western sector and on the eastern edge; while
the smallest one (I) was extensively burnt. Fire prevented the conclusion of the
1997 field survey, and data could not be collected in fragment II. Furthermore, fire
destroyed most plastic tags in all fragments studied, thus preventing monitoring the
fates of individual plants in subsequent years. To assess the effects of fire on the
structure of populations under study, field surveys were carried out in April 1998,
after reconstruction of the destroyed plots, and in June 1999. During April and May
1997, the number of fruits produced was recorded for all reproductive individuals in
all fragments. Infructescences with signs of fruit removal or abortion were excluded
from calculations of individual fruit set.

Data analysis: abundance and population structure

To test whether observed population sizes changed significantly from one year to
the next, we compared population size distribution at successive pairs of years (ti,
ti 1) with uniform distributions considering ti values. Population density was
compared non-parametrically between fragments with KruskalWallis ANOVA. To
determine whether the stage structure of the populations varied through space and
time, we performed a log-linear analysis considering stage S, year Y, and fragment
F (three fragments), by building a three-way contingency table (Sokal and Rohlf
1995). We tested the null hypothesis of three-factor interaction by fitting the fully
saturated model (SYF term present) (Sokal and Rohlf 1995).
Previous spatial pattern analysis (Souza and Martins 2002) showed that the
population in fragment II was mostly concentrated in the western, disturbed sector
of the fragment. We investigated such internal heterogeneity by testing the null
hypothesis that the stage structure was independent of fragment sector and year
(log-linear analysis considering stage S, year Y, and sector D).

Data analysis: life cycle and matrix models

Data collected in the plots were used to build Lefkovitch, stage-based matrix
models with five ontogenetic stages. Matrix models and their inherent assumptions
are discussed in Caswell (1989). This approach relies on using information on the
transition probabilities between life cycle categories from one time period to the
next, in order to describe and project population dynamics through a square matrix
(A) and a vector (v). The following parameters are required for construction of A:
the probability that an individual in stage i will survive and remain within stage i
(Pi); the probability that an individual in stage i will survive and grow into stage
i 1 (Gi); and reproductive output or fecundity (Fi). The v vector defines the
number of individuals in the population for each stage at time t. When the matrix A
is multiplied by the vector v one can obtain the population vector at time t 1.
The data summarised in Figure 1 were put in two projection matrices corresponding to fragments I (smallest) and III (largest), since fire prevented identification

1616

Figure 1. Life cycle diagram for A. humilis in two forest fragments of the Atlantic Forest. Values along
arrows represent transition probabilities between stages (from one circle to another), or stasis probabilities
(from a circle to the same one). S seedling, J juvenile, I immature, V virgin and R reproductive.

of individuals in fragment II. Following the ontogenetic stages identified in Souza

et al. (2000), we constructed matrix models with only one category of reproductive
plants. Entries of matrix A represent transition probabilities in a one-year period
(19961997). The data in the first row represent the contribution of each stage i to the
seedling stage, which was estimated by dividing the number of new seedlings
emerged at time t 1 by the number of reproductive palms observed at time t. Data
on fruit production were pooled from the three fragments, due to small sample sizes
when considering only integer infructescences by site (NI 1, NII 14, NIII 4). In
the model, the fruit category was omitted because from one year to the following, all
seeds either germinate or die.
Because both seedlings and juveniles were absent from fragment I in 1996, the
transition from seedling to juvenile in this fragment was estimated from the stage

1617
distribution by dividing the mean number of juveniles in the period 19971999 by
the mean number of seedlings in the same period (Pinard 1993). We opted for using
mean values instead of values from the observed stage structure from any one year
due to remarkable year-to-year variation in the stage structure (see Results). The
same was not possible for the transition from the juvenile to the immature stage,
because immature individuals outnumbered juveniles in all fragments and years. In
this case we attributed a probability of 0.999 to this transition. This probably is not
unrealistic in face of the transitional character of the juvenile stage (Souza et al.
2000). We assigned a probability value of dying of 0.001 to those stages in which
no individuals died (Pin ero et al. 1984). We did the same for the transition probabilities from the virginal to the reproductive stage in fragment I, and from the
immature to the virginal stage in fragment III. These transitions were not observed,
because the time period considered in the model is short compared to the timing of
natural events in the life cycle of a long-lived perennial like Attalea. Both transitions were observed in one of the fragments, and biologically meaningless developmental sequences would be assumed if zero values were attributed to them (e.g.,
in fragment III palms would have to reproduce before entering the virginal stage).
We estimated confidence intervals for l by drawing 1000 bootstrap samples from
the original data, building a life table and a matrix for each, and solving it for l
(Alvarez-Buylla and Slatkin 1991).
To determine whether transition probabilities varied between fragments, we
performed a log-linear analysis (Caswell 1989; Sokal and Rohlf 1995), considering
stage (five stages, as above) at time t, fragment (fragments I and III, as above), and
fate (stage at time t 1) as the categorical variables. This analysis was based in a
three-way contingency table defined by the factors stage S, fragment P, and fate F.
We tested the null hypothesis of conditional independence of fate and fragment,
fitting the model SF, SP (Caswell 1989). All log-linear analyses were done after
adding 0.5 to each cell (Sokal and Rohlf 1995). The finite rates of population
increase (l) were calculated from the dominant eigenvalues using the power
method (Caswell 1989), as well as the corresponding right eigenvector (i.e., the
stable-stage distribution). The sensitivity is an index that measures how changes in
individual life history parameters influence population growth rate (Caswell 1989).
Although many recent papers have reported elasticities (defined in De Kroon et al.
1986) instead of sensitivities, we preferred to report sensitivities because elasticity
analysis has mathematical constraints derived from the relationship between the
elasticities of growth and fecundity (Matos and Silva Matos 1998).
The population finite rate of increase (l) and the stable distribution are
asymptotic properties of the populations under the initial environmental conditions
(Caswell 1989), in this case the 19961997 period. To examine the transient dynamics of the populations, we projected the stage-structure and total population size
in fragments I and III in 1997, 1998 and 1999 by multiplying the transition matrix
of each fragment by the initial (1996) population vector of each fragment (i.e., a
column vector whose elements ni are the number of individuals in each stage). To
examine whether observed populations were significantly different from the stablestage and projected distributions, we did G tests comparing observed and expected

1618
distributions for each fragment and year. To determine how similar two distributions were, we calculated theP
PS, the proportional similarity index, as in Horvitz
i1
and Schemske (1995): PS
n minai ; bi  100; where there are n stages and
ai is the proportion of individuals in the ith stage of the stable-stage distribution and
bi is the proportion of individuals in the ith stage of the observed stage distribution.
Analyses were performed with the SYSTAT software (SYSTAT 1998) at the 5%
significance level.

Results
Abundance and population structure
Mean density of Attalea in the fragments ranged from 1.2 to 9.0 palms per 100 m2
plot (i.e., 120900 palms ha1). There was more variation among fragments than
between years in the same fragment (Figure 2). In all years, palm density in
fragment II (750900 palms ha1) was significantly higher and had a broader range
than in the other two fragments, where density remained consistently lower (130
200 palms ha1 in fragment I and 120150 palms ha1 in fragment III). Fragment II
had a projected initial population size of 5758 palms (based on the observed
number of individuals and the percentage of the total area sampled), while the other
two had smaller projected populations (199 and 1394 palms in fragments I and III,
respectively). Before fire (19961997 period), the population in fragment I was
growing (observed finite rate of increase significantly greater than unity), but not in
fragment III (Table 2). In the period comprising the occurrence of fire and after it,
all populations remained in equilibrium (growth rates not significantly different
from unity). Considering the entire study period, population sizes in fragments I
and III remained at equilibrium, but the population in fragment II decreased.
The stage structure of the populations (i.e., the relative densities of the ontogenetic stages) was dependent on fragment and year (three-way log-linear model,
G 76.8, df 20, P < 0.0001), varying in a complex way across both space and
time (Figure 3). Generally, later stages were more abundant than initial ones;
seedlings represented only 2.833.9% of individuals, and in most cases outnumbered only juveniles. The juvenile was the least abundant stage, and was absent
from all fragments in 1996. The population in fragment I had the most dynamic
stage structure, while the one in fragment III had the most stable one. Seedlings
were absent from fragment I in the first year, but in the second year both seedlings
and juveniles were present, the former at relatively high numbers. Fragments II and
III had a similar stage structure, with prevalence of the immature, virgin and
reproductive stages.
Fire had a striking effect on the structure of all populations. Six months after the
burning (1998 data), the immature was the prevalent stage in all fragments, accounting for 62.484.7% of the individuals. The juvenile, virgin and reproductive
stages disappeared from fragment I, but the number of seedlings remained nearly
unchanged. The population in fragment II showed the strongest concentration of

1619

Figure 2. Density of A. humilis in three fragments of the Atlantic Forest. Central horizontal bars
represent median values; upper and lower vertical bars (whiskers) represent the 4th and 1st percentiles,
respectively; notches surrounding the median represent median 95% confidence intervals; vertical bars
connect the minimum and maximum values measured; extreme value;
outlier. Non-overlapping
of notches indicates significant difference at the 95% confidence level. The results of the KruskalWallis
ANOVA (H) or the MannWhitney test (U), indicating whether there is a significant difference between
the densities of the three fragments, are given; df for each case in parentheses.

plants in the immature stage, while the one in fragment III showed the smallest
structural change. In 1999, despite the fact that the immature was still the most
abundant stage in all three populations, its prevalence was reduced. Juvenile and
reproductive palms were observed again in fragment I, and the relative abundance
of virgin and reproductive palms increased in the other fragments.
The densest population (fragment II) was marked by strong internal heterogeneity. Log-linear analysis revealed that stage structure differed between sectors
(G 46.5, df 14, P < 0.0001) (Figure 4) and was dependent on year (G 396.0,

1620
Table 2. Observed finite rates of increase (l Nt1/Nt) and the significance of w2 values
testing the hypothesis that l were constant from year to year.
Fragment

Time period
19961997

I
II
III

1.639

1.071

14.7***

0.5 ns

19971998a
2

0.848

0.877

1.4 nsb

1.6 ns

19981999
2

0.880
0.925
0.924

0.7 ns
2.8 ns
0.5 ns

19961999
l

w2

1.222
0.838
0.869

1.8 ns
14.5***
1.7 ns

Time period comprising the fire.

Non-significant.
***P < 0.001.

Figure 3. Frequency distributions of stages for the three populations of A. humilis. * Indicates the
occurrence of fire. S seedling, J juvenile, I immature, V virgin and R reproductive.

df 12, P < 0.0001). In the year before fire, the stage structure had the highest
similarity (PS 67.4%) between the western sector and the others pooled together,
but the proportion of reproductive plants was higher in the western sector. After
fire, the concentration of individuals in the immature stage was higher and juveniles

1621

Figure 4. Frequency distributions of stages for the sub-populations of A. humilis in the western,
disturbed sector of a fragment of the Atlantic Forest, and in the other sectors pooled. * Indicates the
occurrence of fire. S seedling, J juvenile, I immature, V virgin and R reproductive.

were absent in the western sector, leading to reduced similarity with the distribution
in the other sectors (PS 61.0%). In 1999 the proportion of seedlings and juveniles
was lower in the western sector.

Life cycle and matrix models

Life cycles of the populations of Attalea in fragments I and III are illustrated in
Figure 1. In a one-year period, palms could survive without a change in stage, or
they could shrink by one stage or grow by one or two stages. The proportion of
individuals that changed from one stage to the next after a one-year period varied
widely, ranging from 0.1% for the virgins in fragment I and the immature in
fragment III to 99.0% for the juveniles in both fragments. Seedling survival was
remarkably high (79.0% in fragment III). In general, the probability of surviving
and staying in the same stage showed the highest values, along with the transition
from the juvenile to the immature stage. The transition from seedling to the immature stage, observed in fragment III, was absent from fragment I. The average
fruit production of 19 reproductive palms was 75.3  29.3 fruits individual1
  SD.
year1 X

1622
Table 3. Transition matrices for A. humilis populations in two fragments of the
Atlantic Forest. The corresponding mean value of l (95% confidence interval in
parentheses) for each matrix is also given. Missing matrix elements all have a
value of zeroa.
Stage at time t 1

Stage at time t
S

Fragment I; l 1.229(1.2181.240)
S
0.807
J
0.192
I
0.999
V
R
Fragment III; l 1.066(1.0591.073)
S
0.556
J
0.056
I
0.167
0.999
V
R

R
4.000

0.444
0.333
0.111

0.077
0.921
0.001

0.200
0.799
0.632

0.909
0.001
0.030

0.241
0.654
0.103

0.105
0.894

S seedling, J juvenile, I immature, V virgin, R reproductive.

Table 4. Sensitivity matrices for observed transitions of A. humilis populations in

two fragments of the Atlantic Foresta.
Stage at time t 1

Stage at time t
S

Fragment I
S
J
I
V
R
Fragment III
S
J
I
V
R

0.31
0.56

0.03
0.11

0.08
0.16
0.16

0.18
0.06
0.99

0.31
0.11
1.65

0.02
0.31
0.06

0.01

0.40
0.67
1.46

0.04
0.06
0.13

0.21
0.45

S seedling, J juvenile, I immature, V virgin, R reproductive.

The transition coefficients of fragments I and III were marginally different

(log-linear analysis, G 31.762; df 20; P 0.046). Besides the absence of the
transition from seedling to the immature stage in fragment I, the probabilities of
surviving and staying in the same stage for seedlings and virgins were lower in
fragment III than in fragment I, while the opposite was true for the immature.
Therefore, the data from each fragment were used to build two separate matrices

1623

Figure 5. The stable-stage distribution for fragments I (solid bars) and III (hatched bars). The G
statistic, indicating whether there is a significant difference between the two distributions, is given.
S seedling, J juvenile, I immature, V virgin and R reproductive.

Table 5. Statistical results of comparisons between actual and stable-stage

distributions predicted by the models for two fragments of the Atlantic Forest. The
proportional similarity index (PS), indicating the magnitude of the difference
between the two distributions, is givena.
Year

Fragment
I

1996
1997
1998
1999

III

PS (%)

PS (%)

110.7
8.0
88.5
71.8

<0.0001
0.093 ns
<0.0001
<0.0001

41.8
82.5
42.0
49.5

26.2
13.4
15.6
16.9

<0.0001
0.01
0.004
0.002

73.6
77.4
81.2
82.9

df 4 in all cases.

(Table 3). Population finite rates of increase were greater than unity in both fragments, but the population in fragment I was growing at a higher annual rate (21.8
24.0%) than the population in fragment III (5.97.3%). Sensitivity values varied
over two orders of magnitude between ontogenetic stages and type of transition
(Table 4). Population growth rate was most sensitive to rapid growth of the immature and survivorship (stasis) of reproductive palms. However, there were conspicuous differences between the populations. In fragment III, the behaviour of the

1624

Figure 6. Transient analysis for A. humilis by projection of the stage structure in two fragments of the
Atlantic Forest. The G statistic, indicating whether there is a significant difference between the projected
and the observed distribution, and PS, indicating the magnitude of the difference between the two
distributions, are given. df 4 in all cases. **P < 0.01, ***P < 0.001.

immature had the largest impact on population growth compared to other stages,
while in fragment I it was the growth of both pre-reproductive stages followed by
the growth of seedlings that had the largest sensitivities. The population in fragment
III was much more dependent on stasis than the population in fragment I, which
was dependent on the pathway seedling ? juvenile ? immature ? reproductive to
a greater extent.
The models predicted different stable stage distributions for the two fragments
(Figure 5). Seedlings made up the highest proportion of the population in fragment
I (49.9%), followed by the virgins (23.4%), while the distribution in fragment III
was dominated by the immature (54.7%), followed by seedlings (22.1%) and reproductive palms (17.0%). All but one (fragment I in 1997) of the observed stage

1625
distributions differed significantly from the expected stable stage distribution
(Table 5). Nevertheless, the proportional similarities indicated that observed dis = 78.8%, SD = 4.1) were much closer to the stabletributions in fragment III (X
 = 54.0%, SD = 19.4). The stage structure
stage distributions than in fragment I (X
projected for both populations in 1998 and 1999 (predicted transient dynamics) was
significantly different from the observed distributions in these post-fire years
(Figure 6). However, projected stage distributions in fragment III were much closer
to the actual distributions than in fragment I (Figure 3).

Discussion
Our results showed that the three populations in the studied fragments had important differences in ontogenetic stage structure, dynamics and density in pre-fire
years. Furthermore, fire had definite effects on population structure, and distinct
population responses to fire occurred in each fragment.
The three forest fragments were strikingly different and represented a range of
conditions from the smallest, more disturbed fragment to the largest, more preserved one (Table 1). Small tropical forest fragments are normally characterised by
high internal heterogeneity, with preserved sectors surrounded by disturbed sectors
and edges, where species diversity and canopy heights are generally lower and tree
densities, canopy openings, and abundance of pioneer and climbing species are
usually higher (Laurance 1997; Oliveira-Filho et al. 1997; Lykke and Goudiaby
1999; Tabanez and Viana 2000). In a detailed spatial analysis of the same data set
(Souza and Martins 2002), we found that seedlings and juveniles form discrete,
low-density clumps not associated with reproductive plants, which occur in more
diffuse aggregates. Furthermore, palm density was not consistently related to distance from fragment edge, but was concentrated in the most disturbed transect arms
in the two smaller fragments. This picture led to the hypothesis that the population
in the largest fragment would be in equilibrium, and its dynamics would be mostly
driven by interactions with natural forest dynamics. On the other hand, the populations in degraded, more open sectors of smaller fragments would be growing,
leading to the formation of the large aggregates observed there, while hyper-disturbed conditions like those found in the eastern sector of the smallest fragment
would lead to local extinction (Souza and Martins 2002). Many aspects of the
structure and dynamics of the Attalea populations agree with these predictions.

Population structure and dynamics in preserved fragments

Population growth rates in the largest fragment were equal to or slightly greater
than unity during the whole study period, thus indicating that it was favourable to
the persistence of Attalea (Horvitz and Schemske 1995). A similar result has been
found for other palm species in forest fragments larger than 100 ha (Ratsirarson
et al. 1996; Silva Matos et al. 1999), and reflects the ability of these species to

1626
persist in the forest environment present in these fragments. The population in the
largest fragment was the most stable of the three populations studied, since it
showed neither the fluctuation seen in fragment I nor the declining tendency seen in
fragment II after fire (Table 2). This picture is reinforced by the high similarity
between observed and stable-stage structures in this fragment. Due to its preserved
structural characteristics, the occurrence of open, resource-rich sites in the largest
fragment we studied is mostly restricted to natural treefall canopy openings. As
both growth and the probability of reproduction of understorey palms are greatly
enhanced under canopy gaps (Martnez-Ramos et al. 1988; Abrahamson 1999), it is
likely that both the fast growth of seedlings into the immature stage and the
transition of virginal and immature palms into the reproductive stage (see Figure 1)
take place predominantly in recently opened gaps.
The stage structure of Attalea populations was characterised by low frequencies
of seedlings and juveniles, although their relative frequencies varied in time and
space. Populations of the palms Iriartea deltoidea (Pinard 1993) and Phytelephas
seemannii (Bernal 1998) also showed a concentration of individuals in intermediate-sized classes. Reduced frequency of initial stages could result from lower
survival in these stages, differential growth of individuals in each stage, or low
population growth rates (Condit et al. 1998). The concentration of palms in the
immature and virgin stages is likely the outcome of the accumulation of individuals
from successive cohorts, as well as previously reproductive palms that moved back
to the virgin stage. The same process was found by Bernal (1998) for P. seemannii.
Population growth rate was mainly dependent on stasis, especially of seedlings and
non-reproductive large plants, a pattern found by Silvertown et al. (1993) for woody
plants in general and by Horvitz and Schemske (1995) for the understorey rhizomatous herb Calathea ovandensis. Thus, our study species behaved like rhizomatous plants and trees, in which survival is most important (Caswell 1989;
Silvertown et al. 1993).

Population structure and dynamics in disturbed fragments

Abundance differences between the populations in the fragments did not follow an
intuitive pattern related to fragment size. The mid-sized fragment had much higher
densities and a disproportionately larger population, while the other two had
smaller projected populations at similar low densities. This suggests recently increased growth of the sub-population in the large disturbed western sector of
fragment II, relative to the background density found in the more preserved conditions represented by the largest fragment. Many species of the genus Attalea have
the ability to persist and thrive in disturbed areas (Hogan 1988; Henderson et al.
1995). A. humilis is not an exception to this pattern, often forming dense aggregates
in more open or disturbed sites (Henderson et al. 1995; Lorenzi et al. 1996).
The population in the smallest and most degraded fragment had a background
density similar to that in the largest fragment and a higher pre-fire population
growth rate, and was absent from its hyper-disturbed eastern sector (Souza and

1627
Martins 2002). The main cause of the high population growth rates found in this
fragment was a large recruitment event in 1997, where 20 seedlings (33.4% of the
sampled population in that year; Figure 3) emerged, almost all of them in a single
aggregate in the Attalea-rich sector of the fragment (Souza and Martins 2002). No
new emerging seedlings were observed after the 1997 event (personal observation).
Random variations in birth and death rates (demographic stochasticity) and the
resulting fluctuations in the growth rate of small populations are among the main
causes of population extinction (Pimm 1991; Primack 1995; Case 2000). Additionally, the absence of Attalea individuals from the Trema-dominated sectors in
fragment I and the declining population size in fragment II (Table 2) suggest the
possibility of its competitive exclusion from hyper-disturbed patches following fire.
The Attalea-absent patches are dominated by monospecific, dense stands of the
pioneer tree Trema micrantha (Silva Matos et al. 1998b), with high occurrence of
vine and liana tangles, understorey bamboos, and a suite of sun-loving species
(personal observation). The open crown architecture, low stature and slow growth
of Attalea may make it a poor competitor in such dense milieu.

Resistance and resilience of populations to fire

Attalea population growth rates showed high resistance to fire. The most depressing
effect fire had was the reduction of the population growth rate in fragment I to
values indistinguishable from 1.0 (Table 2). On the other hand, the stage structure of
populations was resilient to fire. Although there had been a marked concentration
of individuals in the immature stage 6 months after the burning (Figures 3 and 6), this
change was mostly reversed one year later. Palm (Barrow et al. 1993; Ratsirarson
et al. 1996) and tree (Uhl and Kauffman 1990) species adapted to rain forest environments, where fire is not a natural part of the environment, typically experience
high adult mortality when impacted by fire. Our results concerning population
density and stage structure suggest that the majority of large individuals did not die
but shrank to the immature stage. To a large extent, this result derives from the fact
that the apical meristem of Attalea is buried and well insulated by leaf bases (Souza
et al. 2000), becoming protected from the high temperatures during a fire (McPherson
and Williams 1998). Many species of Attalea occur in savannah-like ecosystems,
where recurrent fires are an important factor, and have short or subterranean stems
(Henderson et al. 1995), so that the habit and resprout strategy of A. humilis could be
regarded as a preadaptation to fire (Whelan 1995).
However, the resilience of population structure to fire may largely rely on habitat
heterogeneity. Juveniles disappeared from and the number of seedlings was reduced
in the western sector of fragment II (which was more strongly affected by fire;
personal observation) after fire (Figure 4). Forest fires are normally patchy (Whelan
1995) and tend to occur predominantly at the outer boundaries of fragments
(Kellman and Meave 1997), in open, disturbed forest patches and in previously
burned forest sectors (Cochrane and Schulze 1999). Young plants are unlikely to
survive fire due to their undeveloped subterranean system and, hence, superficial

1628
meristem position (McPherson and Williams 1998). Therefore, seedling and juvenile persistence in the studied populations depends on seedling emergence in
more closed, protected sectors of the fragments (Souza and Martins 2002), thus
escaping the majority of fires.

Persistence of populations in fragments of the Atlantic Forest

Although population growth rates showed to be resistant to a single fire, and population structure was capable of recovering after a short period, the persistence of
A. humilis populations in fragments of the Atlantic Forest may be largely dependent
on habitat quality and fire frequency, as suggested by the analysis of the spatial
distribution of the populations between fragment sectors (Souza and Martins 2002).
The situation of the population in the 10 ha fragment shows that palm populations in tropical forest fragments are not necessarily remnant populations (Scariot
1999; sensu Eriksson 1996), but instead can be self-maintaining if forest structure
could be preserved in fragments. The persistence of populations in the absence of
any kind of protection of forest structure, however, is less certain, and could be
dependent on fire return interval and its effect on vegetation structure. Both theoretical (Iwasa and Kubo 1997) and empirical (Hoffmann 1999) studies have
suggested that periodic removal of aerial parts promotes success of low-stature,
long-lived resprouting species. Thus, considering that Attalea population growth
rate relies heavily on survivorship of seedlings, immature and virgin plants, it is
unlikely that fire threatens its persistence in the fragments in face of the resprouting
ability of established plants, as long as seed disperser (scatterhoarding in this case;
Souza et al. 2000) activity permits new seedlings to emerge in relatively protected
sites. Nevertheless, if fire frequency becomes too high, even populations of resprouting species may experience increased mortality of established plants and
suffer rapid changes (Whelan 1995). Thus, provided fires do not become too frequent, fire could promote A. humilis population growth through the combined
effects of increased fecundity after burning, growth under open canopies and increased nutrient availability, as observed for other species (Martnez-Ramos et al.
1988; Abrahamson 1999; Hoffmann 1999). This scenario is likely to be widespread
in face of the large-scale effects of fires in tropical fragments (Cochrane and
Laurance 2002), and would lead to a species-impoverished, palm-dominated system in forest fragments (McPherson and Williams 1998).
Based on the present and other analyses (Souza and Martins 2002), we expect
that in small, shrinking fragments of the Atlantic Forest, embedded in landscapes
with high matrix harshness (sensu Gascon et al. 2000), Attalea populations will first
increase, forming monospecific, high-density stands, and henceforth collapse under
the prevalence of secondary vegetation and lack of rodent seed dispersers, frequently extinct in such habitats (Chiarello 1999). The progression from palm
dominance to decline follows the collapse of the forest habitat in face of matrix
harshness and fire frequency (ca. one occurrence in each 5-year interval), which are
breaking down the process of forest regeneration along the margins of many if not

1629
most tropical lowland forest fragments (Laurance 1997; Oliveira-Filho et al. 1997;
Tabanez and Viana 2000), resulting in a gradually receding edge (Gascon et al.
2000, personal observation). In this case, habitat-directed management practices
carried out by both government and farmers, such as fire and poaching prevention,
formation of vegetation buffer zones around fragments and control of liana proliferation could suffice to ensure the persistence of Attalea in most Atlantic Forest
remnants, which are predominantly small (Ranta et al. 1998).
Although the proximity between the fragments in Poc o das Antas may permit
pollinator-mediated gene flow between populations in different fragments (Meffe
and Carroll 1997), the large size of Attalea fruits and its probable dependence on
scatterhoarding rodents for effective seed dispersal (Souza et al. 2000) make it
unlikely that migration between fragments, and hence the establishment of metapopulations, be frequent or even possible. Chance events, in this case represented
by demographic and environmental stochasticity, and natural or human-induced
catastrophes (Pulliam and Dunning 1997) can interact to draw small populations in
an extinction vortex (Gilpin and Soule 1986). Indeed, despite its apparently successful resprouting strategy, the population in the smallest fragment is already small
and possibly in danger of local extinction. The present results can be viewed as a
warning that reduced fragment area, habitat degradation and human-induced fires
can interact to threaten even the persistence of species that are abundant and
preadapted to such disturbances, like Attalea.

Acknowledgements
We are indebted to the Fundac a o de Amparo a` Pesquisa do Estado de Sa o Paulo
(FAPESP) for grant no. 98/01263-3, the Fundac a o O Botica rio de Protec a o a`
Natureza and the Fundac a o de Amparo a` Pesquisa do Estado do Rio de Janeiro
(FAPERJ) for field support. We are grateful to Dalva M. Silva Matos for coordination of fieldwork and lending us her unpublished data about the studied
fragments. Fieldwork was only possible due to valuable help from Flavio N.
Ramos, Mariana C. Torres, Jose F. Caluca, Ludmila P. Siqueira, Jandui Amorim,
Antonio M. Rosa and Luiz F.L. Duarte. We are indebted to Paulo F. de Souza and
Ebert for computer programming concerning resampling procedures. We also thank
Flavio A.M. dos Santos for valuable help in the analysis and to Elena R. AlvarezBuylla, Fabio R. Scarano, Flavio A.M. dos Santos, Gislene M.S. Ganade, Marcio Z.
Cardoso, Paulo F. de Souza and Tiago Quental for the stimulating discussions and
comments on the manuscript.
References
Abrahamson W.G. 1999. Episodic reproduction in two fire-prone palms, Serenoa repens and Sabal
etonia (Palmae). Ecology 80: 100115.
Alvarez-Buylla E.R. and Slatkin M. 1991. Finding confidence limits on population growth rates. Trends
in Ecology and Evolution 6: 221224.

1630
Barrow P., Duff G., Liddle D. and Russell-Smith J. 1993. Threats to monsoon rainforest habitat in
northern Australia, the case of Ptychosperma bleeseri Burret (Arecaceae). Australian Journal of
Ecology 18: 463471.
Bernal R. 1998. Demography of the vegetable ivory palm Phytelephas seemannii in Colombia and the
impact of seed harvesting. Journal of Applied Ecology 35: 6474.
Bierzychudek P. 1999. Looking backwards, assessing the projections of a transition matrix model.
Ecological Applications 9: 12781287.
Brown K.S. and Brown G.G. 1992. Habitat alteration and species loss in Brazilian forests. In: Whitmore
T.C. and Sayer J.A. (eds) Tropical Deforestation and Species Extinction. Chapman & Hall, London,
pp. 119142.
Bullock S.H. 1980. Demography of an undergrowth palm in littoral Cameroon. Biotropica 12: 247255.
Caswell H. 1989. Matrix Population Models. Sinauer Associates, Sunderland, Massachusetts.
Camargo J.L.C. and Kapos V. 1995. Complex edge effects on soil moisture and microclimate in central
Amazonian forest. Journal of Tropical Ecology 11: 205221.
Case T.J. 2000. An Illustrated Guide to Theoretical Ecology. Oxford University Press, Oxford, UK.
Chiarello A.G. 1999. Effects of fragmentation of the Atlantic Forest on mammal communities in southeastern Brazil. Biological Conservation 89: 7182.
Cochrane M.A. and Laurance W.F. 2002. Fire as a large-scale edge effect in Amazonian Forests. Journal
of Tropical Ecology 18: 311325.
Cochrane M.A. and 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: 216.
Condit R., Sukumar R., Hubbell S.P. and Foster R.B. 1998. Predicting population trends from size
distributions, a direct test in a tropical tree community. The American Naturalist 152: 495509.
Crawley M.J. 1997. Plantherbivore dynamics. In: Crawley M.J. (ed) Plant Ecology. Blackwell Science,
Oxford, UK, pp. 401474.
De Kroon H., Plaisier A., Van Groenendael J. and Caswell H. 1986. Elasticity, the relative contribution of
demographic parameters to population growth rate. Ecology 67: 14271431.
Didham R.K. and Lawton J.H. 1999. Edge structure determines the magnitude of changes in microclimate and vegetation structure in tropical forest fragments. Biotropica 31: 1730.
Eriksson O. 1996. Regional dynamics of plants: a review of evidence for remnant, source-sink and
metapopulations. Oikos 77: 248258.
Gascon C., Williamson G.B. and Fonseca G.A.B. 2000. Receding forest edges and vanishing reserves.
Science 288: 13561358.
Gilpin M.E. and Soule M.E. 1986. Minimum viable populations, processes of species extinction. In:
Soule M.E. (ed) Conservation Biology: The Science of Scarcity and Diversity. Sinauer Associates,
Sunderland, Massachusetts, pp. 1934.
Henderson A., Galeano G. and Bernal R. 1995. Field Guide to the Palms of the Americas. Princeton
University Press, Princeton, New Jersey.
Hoffmann W.A. 1999. Fire and population dynamics of woody plants in a neotropical savanna, matrix
model projections. Ecology 80: 13541369.
Hogan K.P. 1988. Photosynthesis in two neotropical palm species. Functional Ecology 2: 371377.
Horvitz C.C. and Schemske D.W.Y. 1995. Spatiotemporal variation in demographic transitions of a
tropical understory herb, projection matrix analysis. Ecological Monographs 65: 155192.
IBDF/FBCN 1981. Plano de manejo da Reserva Biolo gica de Poc o das Antas. Ministe rio da Agricultura,
Braslia, Brazil.
Iwasa Y. and Kubo T. 1997. Optimal size of storage for recovery after unpredictable disturbances.
Evolutionary Ecology 11: 4165.
Janzen D.H. 1988. Management of habitat fragments in a tropical dry forest: growth. Annals of the
Missouri Botanical Garden 75: 105116.
Kapos V., Wandelli E., Camargo J.L. and Ganade G. 1997. Edge-related changes in environment and
plant responses due to forest fragmentation in Central Amazonia. In: Laurance W.F. and Bierregaard
R.O. (eds) Tropical Forest Remnants: Ecology, Management and Conservation of Fragmented
Communities. University of Chicago Press, Chicago, Illinois, pp. 3344.

1631
Kellman M. and Meave J. 1997. Fire in the tropical gallery forests of Belize. Journal of Biogeography
24: 2334.
Laurance W.F. 1997. Hyper-disturbed parks, edge effects and the ecology of isolated rainforest reserves
in tropical Australia. In: Laurance W.F. and Bierregaard R.O. (eds) Tropical Forest Remnants:
Ecology, Management and Conservation of Fragmented Communities. University of Chicago Press,
Chicago, Illinois, pp. 7183.
Laurance W.F., Ferreira L.V., Rankin de Merona J.M., Laurance S.G., Hutchings R.W. and Lovejoy T.E.
1998a. Effects of forest fragmentation on recruitment patterns in Amazonian tree communities.
Conservation biology 12: 460464.
Laurance W.F., Ferreira L.V., Rankin de Merona J.M. and Laurance S.G. 1998b. Rain forest fragmentation and the dynamics of Amazonian tree communities. Ecology 79: 20322040.
Laurance W.F. 2000. Do edge effects occur over large spatial scales? Trends in Ecology and Evolution
15: 134135.
Lorenzi H., Souza H.M., Medeiros-Costa J.T., Cerqueira L.S.C. and Von Behr N. 1996. Palmeiras no
Brasil. edn. Plantarum, Nova Odessa, Brazil.
Lykke A.M. and Goudiaby A. 1999. Structure, floristic composition, and conservation potentials of a
remnant gallery forest at Mare du Dragon, Senegal. Nordic Journal of Botany 19: 561571.
Martnez-Ramos M., Sarukha n J. and Pin ero D. 1988. The demography of tropical trees in the context of
forest gap dynamics, the case of Astrocaryum mexicanum at Los Tuxtlas tropical rain forest. In: Davy
A.J., Hutchings M.J. and Watkinson A.R. (eds) Plant Population Ecology. Blackwell Scientific,
Oxford, UK, pp. 293313.
Matos M.B. and Silva Matos D.M. 1998. Mathematical constraints on transition matrix elasticity
analysis. Journal of Ecology 86: 706708.
Maycock P.F. and Fahselt D. 2001. A remarkable remnant mesic deciduous forest stand: are composition
and structure of old growth retained? Community Ecology 2: 121132.
McPherson K. and Williams K. 1998. Fire resistance of cabbage palms (Sabal palmetto) in the southeastern USA. Forest Ecology and Management 109: 197207.
Meffe G.K. and Carroll C.R. 1997. Genetics, conservation of diversity within species. In: Meffe G.K. and
Carroll C.R. (eds) Principles of Conservation Biology. 2nd edn. Sinauer Associates, Sunderland,
Massachusetts, pp. 269304.
Noss R.F. and Csuti B. 1997. Habitat fragmentation. In: Meffe G.K. and Carroll C.R. (eds) Principles of
Conservation Biology. 2nd edn. Sinauer Associates, Sunderland, Massachusetts, pp. 269304.
Oliveira-Filho A.T., Mello J.M. and Scolforo J.R.S. 1997. Effects of past disturbance and edges on tree
community structure and dynamics within a fragment of tropical semideciduous forest in south-eastern
Brazil over a five-year period (19871992). Plant Ecology 131: 4566.
Pimm S.L. 1991. The Balance of Nature? Ecological Issues in the Conservation of Species and Communities. University of Chicago Press, Chicago, Illinois.
Pinard M. 1993. Impacts of stem harvesting on populations of Iriartea deltoidea (Palmae) in an extractive reserve in Acre, Brazil. Biotropica 25: 214.
Pin ero D., Marinez-Ramos M. and Sarukha n J. 1984. A population model of Astrocaryum mexicanum
and a sensitivity analysis of its finite rate of increase. Journal of Ecology 72: 977991.
Primack R.B. 1995. A Primer of Conservation Biology. Sinauer Associates, Sunderland, Massachusetts.
Pulliam H.R. and Dunning J.B. 1997. Demographic processes, population dynamics on heterogeneous
landscapes. In: Meffe G.K. and Carroll C.R. (eds) Principles of Conservation Biology. 2nd edn.
Sinauer Associates, Sunderland, Massachusetts, pp. 203232.
Ranta P., Blom T., Niemela J., Joensuu E. and Siitonen M. 1998. The fragmented Atlantic rain forest of
Brazil, size, shape and distribution of forest fragments. Biodiversity and Conservation 7: 385403.
Ratsirarson J., Silander J.A. and Richard A.F. 1996. Conservation and management of a threatened
Madagascar palm species, Neodypsis decaryi, Jumelle. Conservation Biology 10: 4052.
Rizzini C.T. 1979. Tratado de fitogeografia do Brasil, aspectos sociolo gicos e florsticos. HUCITEC/Ed.
da Universidade de Sa o Paulo, Sa o Paulo, Brazil.
Scariot A. 1999. Forest fragmentation effects on palm diversity in central Amazonia. Journal of Ecology
87: 6676.

1632
Silva Matos D.M., Caluca J.F. and Souza A.F. 1998a. Conseque ncias da fragmentac a o florestal sobre a
densidade e tamanho de indivduos arbo reos na REBIO de Poc o das Antas, RJ. Anais do IV Simpo sio
de Ecossistemas Brasileiros II, Aguas de Lindoia, Brazil, pp. 120125.
Silva Matos D.M., Ramos F.N., Torres M.C., Souza A.F., Fonseca G.D.F.M., Siqueira L.P., Braz M.I.G.,
Lima L.S. and Portela R.C.Q. 1998b. A fragmentac a o florestal na Reserva Biolo gica de Poc o das
Antas (RJ), uma visa o fitoce ntrica. Anais do IV Simpo sio de Ecossistemas Brasileiros Volume V.
Publicac a o Aciesp no. 104, pp. 190195.
Silva Matos D.M., Freckleton R.P. and Watkinson A.R. 1999. The role of density dependence in the
population dynamics of a tropical palm. Ecology 80: 26352650.
Silvertown J., Franco M., Pisanty I. and Mendoza A. 1993. Comparative plant demography relative
importance of life-cycle components to the finite rate of increase in woody and herbaceous perennials.
Journal of Ecology 81: 465476
Simberloff D. 1988. The contribution of population and community biology to conservation science.
Annual Review of Ecology and Systematics 19: 473511
Sokal R.R. and Rohlf F.J. 1995. Biometry: The Principles and Practice of Statistics in Biological
Research. 3rd edn. Freeman, New York.
Souza A.F. and Martins F.R. 2002. Spatial distribution of the palm Attalea humilis in fire-prone fragments of the Brazilian Atlantic Forest. Plant Ecology 164: 141155.
Souza A.F., Martins F.R. and Silva Matos D.M. 2000. Detecting ontogenetic stages of the palm Attalea
humilis Mart. ex Spreng. in fragments of the Brazilian Atlantic Forest. Canadian Journal of Botany 78:
12271237.
SYSTAT 1998. Standard Version, Version 9. SPSS Inc., Chicago, Illinois.
Tabanez A.A.J., Viana V.M. and Dias A.S. 1997. Conseque ncias da fragmentac a o e do efeito de borda
sobre a estrutura, diversidade e sustentabilidada de um fragmento de floresta de planalto de Piracicaba,
SP. Revista Brasileira de Biologia 57: 4760.
Tabanez A.A.J. and Viana V.M. 2000. Patch structure within Brazilian Atlantic Forest fragments and
implications for conservation. Biotropica 32: 925933.
Takizawa F. 1995. Levantamento Pedolo gico e Zoneamento Ambiental da Reserva Biolo gica de Poc o das
Antas. Monografia da Disciplina de Reside ncia Florestal, Escola Superior de Agricultura Luiz de
Queiroz, Piracicaba, Brazil.
Tomlinson P.B. 1990. The Structural Biology of Palms. Clarendon Press, Oxford, UK.
Turner I.M. 1996. Species loss in fragments of tropical rain forest, a review of the evidence. Journal of
Applied Ecology 33: 200209.
Turton S.M. and Freiburger H.J. 1997. Edge and aspect effects on the microclimate of a small tropical
forest remnant on the Atherton Tableland, northeastern Australia. In: Laurance W.F. and Bierregaard
R.O. (eds) Tropical Forest Remnants: Ecology, Management and Conservation of Fragmented
Communities. University of Chicago Press, Chicago, Illinois, pp. 4554.
Uhl C. and Kauffman J.B. 1990. Deforestation, fire susceptibility, and potential tree responses to fire in
the eastern Amazon. Ecology 71: 437449.
USDA 1990. Keys to Soil Taxonomy by Soil Survey Staff. 4th edn. Soil Environmental Sciences
Department, Polytechnic Institute and State University, Blacksburg, Virginia.
Walter H. 1971. Ecology of Tropical and Subtropical Vegetation. Oliver and Boyd, Edinburgh, UK.
Whelan R.J. 1995. The Ecology of Fire. Cambridge University Press, Cambridge, UK.
Whitmore T.C. 1997. Tropical forest disturbance, disappearance, and species loss. In: Laurance W.F. and
Bierregaard R.O. (eds) Tropical Forest Remnants: Ecology, Management and Conservation of Fragmented Communities. University of Chicago Press, Chicago, Illinois, pp. 312.