Você está na página 1de 12

OIKOS 106: 27 /38, 2004

Comparative dynamics of small mammal populations in treefall gaps


and surrounding understorey within Amazonian rainforest

Harald Beck, Michael S. Gaines, James E. Hines and James D. Nichols

Beck, H., Gaines, M. S., Hines, J. E. and Nichols, J. D. 2004. Comparative dynamics of
small mammal populations in treefall gaps and surrounding understorey within
Amazonian rainforest. / Oikos 106: 27 /38.

Variation in food resource availability can have profound effects on habitat selection
and dynamics of populations. Previous studies reported higher food resource
availability and fruit removal in treefall gaps than in the understorey. Therefore, gaps
have been considered ‘‘keystone habitat’’ for Neotropical frugivore birds. Here we test if
this prediction would also hold for terrestrial small mammals. In the Amazon, we
quantified food resource availability in eleven treefall gaps and paired understorey
habitats and used feeding experiments to test if two common terrestrial rodents
(Oryzomys megacephalus and Proechimys spp.) would perceive differences between
habitats. We live-trapped small mammals in eleven gaps and understorey sites for two
years, and compared abundance, fitness components (survival and per capita
recruitment) and dispersal of these two rodent species across gaps and understorey
and seasons (rainy and dry). Our data indicated no differences in resource availability
and consumption rate between habitats. Treefall gaps may represent a sink habitat for
Oryzomys where individuals had lower fitness, apparently because of habitat-specific
ant predation on early life stages, than in the understorey, the source habitat.
Conversely, gaps may be source habitat for Proechimys where individuals had higher
fitness, than in the understorey, the sink habitat. Our results suggest the presence of
source-sink dynamics in a tropical gap-understorey landscape, where two rodent species
perceive habitats differently. This may be a mechanism for their coexistence in a
heterogeneous and species-diverse system.

H. Beck and M. S. Gaines, Dept of Biology, Univ. of Miami, P. O. Box 249118, Coral
Gables, FL 33124-0421, USA / HB also at Center for Tropical Conservation, Duke
Univ., 3705-C Erwin Rd, Durham, NC 27705, USA (harald@duke.edu). / J. E. Hines,
and J. D. Nichols, USGS Patuxent Wildlife Research Center, 11510 American Holly Dr.
201, Laurel, MD 20708-4017, USA.

One of the main challenges in ecology is to understand 1992, Abrams 2000, Delibes et al. 2001). Habitat
the spatial and temporal distribution of individuals in selection and dispersal patterns of tropical animals are
heterogeneous landscapes (Pusenius et al. 2000, Lin and largely unknown. MacArthur et al. (1966) suggested that
Batzli 2001, Calsson-Granér and Thrall 2002). The because of higher structural heterogeneity in tropical
concept of habitat selection (MacArthur and Pianka forests than in the temperate areas, tropical individuals
1966) has stimulated an increasing body of theoretical should have higher habitat specificity than individuals in
and empirical research. Both habitat selection models the temperate zone.
and optimal foraging theories assume that an individual Treefall gaps are the most frequently occurring
attempts to maximize its survival, reproductive success, perturbation in tropical forests and represent structu-
and the net rate of energy intake (McPeek and Holt rally heterogeneous patches within the undisturbed

Accepted 7 November 2003


Copyright # OIKOS 2004
ISSN 0030-1299

OIKOS 106:1 (2004) 27


understorey matrix (Denslow 1995, Molino and Sabatier Because previous studies found higher recourse avail-
2001, Beck 2002, 2003). In general, a given forest ability in gaps than in the understorey (Bullock 1980,
consists of gaps or patches at different stages of a cycle Piñero et al. 1984, Sarukhán et al. 1984, Denslow et al.
that includes new treefall openings, as well as regenerat- 1986, Levey 1988, 1990), we predict that resource
ing and closed-canopy phases (Hubbell et al. 1999, densities should drive habitat-specific population dy-
Molino and Sabatier 2001). An extensive body of namics of the two most common species of small
literature has demonstrated abiotic differences between terrestrial mammals, the rice rat (Oryzomys megacepha-
undisturbed forest and treefall gaps (Lee 1978, Chazdon lus, Fischer 1814) and the spiny rat (Proechimys spp.,
and Fetcher 1984). Other differences include higher rates Allen 1899). Because spatio-temporal variation in re-
of seed predation (Alvarez-Buylla and Martinez-Ramos source availability can lead to different carrying capa-
1990), herbivory (Swaine 1996), and fruit removal cities between habitats, we first tested for differences in
(Murray 1987) in gaps than in understorey. Additionally, resource availability and consumption by both rodent
many plants in gaps produce more fruits than when in species across habitats. Based on the work of Levey
(1988, 1990) we predicted higher resource availability in
the understorey (Bullock 1980, Piñero et al. 1984,
gaps than understorey. We also predicted higher resource
Sarukhán et al. 1984, Denslow et al. 1986, Clark and
availability during the rainy season than during the dry
Clark 1987, De Steven et al. 1987, Marquis 1988,
season (Foster 1985a,b, Levey 1988, Janson and Em-
Martinez-Ramos et al. 1988, Levey 1990). Although
mons 1990, Lugo and Frangi 1993). We then quantified
many ecological studies have been concerned with the
how variation in resource availability in gaps and
flora of gaps, their fauna remains largely ignored.
understorey affect population dynamics by comparing
Schemske and Brokaw (1981) found that the species abundance, fitness (survival and per capita recruitment)
richness of insectivorous birds was higher in gaps than in and dispersal patterns of both rodent species across
the understorey. Similar results were reported in other habitats and seasons. We predicted greater abundance,
studies (Hoppes 1987, Levey 1988). Levey (1988) exam- survival, per capita recruitment and immigration in the
ined the distribution and abundance of fruiting plants habitat with greater resource availability. Similarly, we
and birds in gaps and the understorey. He found a expected abundance, survival, recruitment and immigra-
distinct gap community of frugivorous birds. Because of tion to show a positive relationship with seasonal
the continuously higher resource availability in gaps variation in resource availability. Finally, we discuss
compared to understorey, Levey (1990) categorized gaps results in terms of ecological theory, including the
as keystone habitats for frugivores. Crome and Richards source-sink model.
(1988) studied insectivorous bats and found that certain
species selected gap habitats in preference to the under-
storey. The above studies found that volant species select
gaps as foraging sites. Within tropical rainforests the
Material and methods
area of treefall gaps depends primarily on the number of Study area and trapping regime
trees that were knocked down. For example, in Costa
Rica, Sanford et al. (1986) found that 25% of 394 treefall We conducted this study at Cocha Cashu Biological
Station (CC), which is located within the 1.8 million
gaps were larger than 200 m2. Previous studies on small
hectare World Biosphere Park Manu in Peru, (118 54? S,
mammals in treefall gaps did not provide data on gap
718 22? W). CC is one of the few research sites within the
size or age, but found species-specific responses. In
Neotropics, which offers a completely undisturbed
Vietnam, Adler et al. (1999) trapped small mammals in
ecosystem with a full complement of both top predators
different forest habitats including treefall gaps. They
and their prey (Terborgh 1990). Average annual rainfall
found one species mainly in gaps, and suggested that this
during the two-year study period was 2,424 mm, with
species is a gap specialist. In Africa, Lwanga (1994) and 81% occurring during the rainy season (November to
Struhsaker (1998) found that several terrestrial rodent May, Fig. 1). Mean annual temperature for the same
species had higher densities in gaps while others had time period was 22.98C (range 98C /348C, Beck 2002).
higher densities in the understorey. They suggested that In September 1994, a storm created hundreds of
dense groundcover provided ideal habitat for many treefall gaps within Manu, and more than 40 gaps within
rodents. In Panama, Lambert and Adler (2000) found the 12 km2 study area of CC (Foster and Terborgh 1998).
that Proechimys semispinosus was significantly asso- All gaps were permanently marked and are part of a
ciated with treefall gap habitats. In Cocha Cashu, Peru, long-term study on forest dynamics (Beck, unpubl.).
several small mammal species such as the Neacomys This storm event provided a unique opportunity for
spinosus were almost exclusively found in gaps (Beck research because it offered many similar sized gaps of the
2002, 2003). At the same study site, Emmons (1982) same age and successional status. Gaps of comparable
radio tracked three Proechimys species and reported that age and size are similar in abiotic factors such as light
individuals preferred treefall gaps to the understorey. and temperature (Chazdon and Fetcher 1984) and biotic

28 OIKOS 106:1 (2004)


Fig. 1. Monthly rainfall and
mean (9/1 SE) resource density
based on 138 seed traps
(41.5m2) in nine gaps and
understorey sites in CC, Peru.
There was no difference
between habitats, but both had
a significant higher resource
density in the rainy seasons
(Jan /May, Sept /May) than in
the dry seasons (Jun /Aug,
Jun /Oct).

factors such as plant communities and species richness 4.7 km, and 427 m9/104 SE, respectively. In gaps, trap
(Hubbell et al. 1999). In July 1998, we measured the area lines began 1 to 5 m behind the root ball of the fallen tree
of 40 gaps following Brokaw (1982). We selected the four where disturbance was obvious and continued towards
largest treefall gaps and randomly selected seven small the fallen crown until the entire gap area was covered.
gaps for permanent trapping grids. We randomly Trap lines and stations were spaced 10 m apart, with
selected distance and compass direction from each gap each station comprising one Sherman trap (10 /12 /38
to determine the location for a same-sized control cm), one Tomahawk trap (13 /13 /30 cm) at the
trapping grid in the understorey (Fig. 2). The maximum ground, and one arboreal Tomahawk trap, attached
and mean distance between the experimental gaps was onto the vegetation up to 3 m high. We used the trapping

Fig. 2. Spatial configuration of


4 large (/X/981 m29/73.6 SE)
and 7 small (/X/295 m29/62.7
SE) treefall gaps and
understorey trapping locations
within the lowland rain forest
of CC, Peru. The maximum
and mean distance between the
gaps was 4.7 km, and 427 m9/
104 SE, respectively. Size
classes of gaps and understorey
are indicated by different
ellipse sizes. The CC study area
encompasses 12 km2, and is
served by over 50 km of trails.
An approximately 40 to 150
meter wide meander belt of the
white-water Manu River
borders the study area to the
South and West, while
continuous lowland and
upland forest extends to the
North and East.

OIKOS 106:1 (2004) 29


grid dimensions from each gap for its respective under- Thus, we limited identification of Proechimys to the
storey control-trapping grid. The mean area of the small genus level.
gaps was 295 m2 (n /79/62.71 SE), and on average we For the feeding experiments, we trapped rodents in the
used 41 traps (range: 39 /54) for each individual trapping forest near the CC field station. Individuals of both
grid. Large gaps had a mean of 981 m2 (n /49/73.68 species were kept temporarily in 50/15 /15 cm cages
SE), with a mean of 73 traps (range: 60 /84) for each with free access to water and food. During daily trials
individual grid. each animal was provided with an excess random sample
Each primary trapping period consisted of a five-week of at least four different fruit species. On the following
period in which all eleven gaps and understorey systems day, animals were removed from the cage and the fruit
were sampled for seven consecutive days. This sampling samples were examined. We recorded whether seeds and
schedule was repeated three more times during each fruit were consumed. Utilization was defined as a sample
rainy season (January to April) and dry season (July to being partially or completely consumed. For statistical
October) for two years. Traps were baited every morning comparisons of utilized resource density across habitats,
with a mixture of rice, peanut butter and vanilla extract, size classes and seasons, we included only data from the
and checked the following morning. We recorded the overall resource densities (obtained from the seed traps)
following data from captured animals: species, gender, for those plant species that were consumed by the rodent
trap location, ear tag number (if previously captured), species. Furthermore, we used the monthly mean weights
and reproductive condition. of the utilized plants as covariates in models to estimate
rodent capture and survivor probabilities (below).

Resource density and feeding experiments


Model selection for capture and survival
We used seed traps to quantify seasonal differences in
probabilities
resource availability between gaps and undisturbed
understorey. From the eleven gaps and understorey Before estimating quantities of interest for each species,
trapping grids, we randomly selected three large gaps, we tested if the trap data within each seven-day sampling
their understorey controls, and six additional small gap- period would meet the assumptions of a closed popula-
understorey systems for seed traps. Depending on the tion (no additions via immigration or birth and no
total area of each gap-understorey system, we set up 7 to deletions through death or emigration) using the pro-
12 seed traps at random locations. Each plastic seed trap gram CAPTURE (Rexstad and Burnham 1991). We had
had a collecting area of 0.3 m2, a depth of 0.4 m and was initially hoped to use the robust design (Pollock et al.
suspended approximately 0.5 m above the ground. In the 1990) approach to estimation, taking advantage of the
nine gaps and understorey systems, we installed a total flexible modelling capability to estimate abundance
of 138 seed traps with a total collecting area of using closed population models (Otis et al. 1978).
approximately 41.5 m2. We collected seed samples at Because the test results indicated our data violated the
weekly intervals from January 1999 through October closure assumptions, we employed the open population
2000. Samples were identified to genus or species level, Jolly-Seber (JS) model to estimate capture probability
then air-dried for 24 hours and weighed to the nearest (pi), abundance (/N̂i ) and survival probabilities (fi)
0.1 gram. We estimated the monthly mean resource utilizing Program MARK version 1.8 (White 2001). We
availability in grams per square meter for each gap and thus pooled captures within each 7-day period to
understorey system for comparisons across habitats, size develop capture histories that simply reflected whether
classes and seasons. or not an animal was caught at least once in the period.
Although seed traps allow the quantification of overall We used the capture histories for each species by habitat
resource density, not every fruit species is utilized equally (i.e., combining all sample locations within each habitat
by every consumer species. Therefore, we performed by size class) and ran the most general (full time-
feeding experiments throughout the two-year study dependent) model for survival (fi) and capture prob-
period with individuals of both Oryzomys megacephalus ability (pi). The full model for each of the four species-
and Proechimys spp. and fruit samples obtained from habitat combinations was thus a 2-group JS model with
the seed traps. parameters dependent on both time (t) and plot size (g);
At least four different Proechimys species occur f(g /t) p(g /t). Following Lebreton et al. (1992), we
sympatrically in the lowland forest of Manu Park tested whether or not this general model fit the data by
(Voss and Emmons 1996). Because these species exhibit conducting goodness-of-fit-tests using the program Re-
overlapping morphological characteristics (Weksler et al. lease version 3.0 (Burnham et al. 1987) with data from
2001), species identification requires extensive collec- each habitat. Our data fit the general model for
tions or invasive procedures. Neither is allowed within Oryzomys in gaps: x20.05(30) /9.634, P/ 0.999; in the
the Manu Park nor is desirable in an ecological study. understorey: x20.05(21) /7.689, P / 0.996, and for Pro-

30 OIKOS 106:1 (2004)


Table 1. Comparison between the most parsimonious (AICc) question and the model with the lowest AICc) values
and several other models to estimate survival (fi) and capture
probabilities (pi) for Oryzomys megacephalus and Proechimys calculated for each model. We estimated monthly
spp. in gap and understorey habitats. Subscripts indicate how abundance (/N̂i ) for a given time period (t) by dividing
both parameters (fi and pi) were estimated: (g) size classes were the count statistic, the number of animals caught during
not combined, (t) monthly time dependency, (×/) time indepen- periods (ni), by the corresponding estimate (/p̂i ) of the
dent, (r) resource density used as a covariate.
capture probability (Seber 1982): N̂i ni =p̂i :
Model No. of parameters DAICc

a) Oryzomys in gaps
f(g/r) p(g/r) 8 0.00
f(g) p(g/t) 24 13.73 Fitness parameter estimation
f( ×/) p(×/) 2 16.39
f(g/t) p(g/t) 42 36.51 Because we were not able to measure fitness directly in
b) Oryzomys in understorey
the field, we used monthly survival probability estimated
f(g/r) p(t/r) 6 0.00 from capture-recapture models (obtained from the
f(g) p(g/t) 24 10.08 model selection procedure by Program MARK) and
f(g/t) p(g/t) 39 14.95
f( ×/) p(×/) 2 27.67 per capita recruitment as fitness surrogates. Because the
complements of survival probability estimates (1 /f)
c) Proechimys in gaps from capture-recapture models include permanent emi-
f( ×/) p(×/) 2 0.00
f(g/r) p(g/r) 4 0.40 gration (Pollock et al. 1990), these estimates are not ideal
f(g/t) p(t) 27 32.12 metrics reflecting fitness.
f(g/t) p(g/t) 32 41.93 We also estimated number of recruits per female.
d) Proechimys in understorey Given our intensive trapping within each habitat, we
f( ×/) p(×/) 2 0.00 assumed that all new recruits were born within the
f(t/r) p(g/r) 6 1.62
f(g/t) p(t) 28 46.47 habitat in which we caught them. We considered
f(g/t) p(g/t) 31 50.87 Oryzomys individuals with a body mass 5/28 g (max-
imum adult weight was 120 g), and all new Proechimys
individuals with a body mass 5/160 g (maximum adult
weight was 537 g) to be recent recruits. To test for
echimys in gaps: x20.05(13) /5.460, P/ 0.963, in the differences in per capita recruitment across habitats, size
understorey: x20.05(21) /7.689, P / 0.996. classes and seasons, we compared the mean number of
We derived additional, parsimonious models by con- recruits trapped for the first time divided by the capture
straining the number of parameters and their time probability (/p̂i ) to estimate the abundance of recruits and
dependency (Table 1). For example, in the model divided that number by the estimated abundance of
f(×/)p( ×/), survival (f) and capture probability (p) were females in reproductive condition trapped during that
constant (×/) with respect to time, hence only two given month. To examine whether habitats, size classes
parameters were estimated. In the model f(g /t) p( ×/), and/or seasons had an effect on the fitness of each
survival was estimated separately for each month (t) and rodent species, we tested for differences between the two
size class (g) separately, whereas (p) was constant over monthly fitness surrogates combined into seasons.
time and size class. We used the AICc (Akaike’s
information criterion) values to determine which model
best explained the variation in the data while using the
Movement into habitats
fewest parameters. AIC is an information /theoretical
approach that treats model selection as an optimisation We estimated the number of monthly new immigrants
problem rather than as a hypothesis-testing problem (Pollock et al. 1990) into a given habitat (animals not
(Burnham and Anderson 1998). Additional models in present in the habitat at time t but immigrating onto it
the set tested whether survival and capture probability between t and t/1 and still present at t/1) by first
varied with consumed resource density. We specified the estimating the abundance of adult individuals for month
monthly means of resource density as a covariate for the t/1. We defined an adult as an individual of a body
survival and capture probabilities within several models. mass equal to or greater than the body mass of
Many individuals survived the transition periods individuals reaching sexual maturity for each species.
between rainy and dry seasons (May /June, and From adult abundance, we subtracted the estimated
November /December when no trapping occurred) and number of surviving adults and juveniles (that survived
were captured in the following season. Thus, we used the and grew up to be adults) from the previous period
mean resource density over these transition periods as a computed as the product of the estimated abundance in
covariate for survivorship. We followed Anderson et al. month t and estimated survival probability between t
(1994) and selected the most parsimonious models based and t/1. Within each season, we then compared the
on the DAICc (difference between AICc for the model in mean number of immigrants to gaps from the under-

OIKOS 106:1 (2004) 31


storey, as well as the number of immigrants to the Feeding experiments
understory area.
Overall, we tested 53 individual Oryzomys in the feeding
experiments using 32 identified plant species. Oryzomys
consumed 83.3% fruit biomass of the identified plants
Statistical analyses obtained from gaps and 84.3% from the understorey.
There was no difference in fruit consumption by
We analyzed the data using two-way repeated-measures Oryzomys between habitats (F1,33 /0.001, P/0.978)
analyses of variance (rm ANOVAs: Type III sums of or size classes (F1,33 /2.884, P/0.099). However,
squares, Zar 1996). We specified the two-way rm Oryzomys consumed more fruits during the rainy season
ANOVAs with habitat (gaps and understorey) and size than the dry seasons (F1,33 /10.965, P/0.002).
class (large and small) as between-subject factors and We tested 59 individual Proechimys and 39 identified
season (rainy and dry) as a within-subject factor. If a plant species during the feeding experiments. Proechimys
two-way rm ANOVA revealed a significant interaction, consumed 95.6% fruit biomass of the identified plants
we used one-way rm ANOVAs to test for simple effects obtained from gaps and 96.1% from the understorey.
(Stevens 1996). We compared survival for Oryzomys There was no difference in fruit consumption by
across habitats and size classes within each season using Proechimys between habitats (F1,33 /0.083, P/0.775)
two-way rm ANOVAs. Here season (e.g. dry season of or size classes (F1,33 /0.000, P/0.986). However,
1999 and dry season 2000) was the within-subject factor. Proechimys consumed more fruits during the rainy
Because MARK indicated no difference in survival season than the dry season (F1,33 /18.989, P B/0.0001).
across size classes, we compared survival of Proechimys
only across habitats within each season using a one-way
rm ANOVA. Prior to any rm ANOVA, we tested whether
the data met the assumption of normality using the
Model selection for capture and survival
Shapiro-Wilk statistic. We log10 (x/1)-transformed re-
source density-, and feeding experiment data, and
probabilities
square-root arcsine transformed survival data. Homo- For Oryzomys the low-AICc model for gap habitats
geneity of variance was tested by Levene’s test. Equality included variation in survival and capture probability
of sphericity, which is used to assess the circularity of a among months and also between large and small gaps.
variance-covariance matrix (Ende 2001), was estimated Therefore, both survival and capture probability varied
using Mauchly’s test. If the assumption of sphericity was by size class. Because the capture /recapture survival
violated, we used the Greenhouse-Geisser correction to probability estimates include permanent emigration, our
adjust the degrees of freedom (Zar 1996). All analyses result of a lower survival in small understorey is expected
were performed using SPSS version 10.1.0 (SPSS 2000). because there is a higher probability of emigration of
individuals from small than large understorey grids. This
expectation is a simple consequence of the large peri-
meter to area ratio in small grids. In contrast, in the
Results understorey, capture probability varied among month
but not by size classes. Both models included linear-
Resource densities
logistic models for both parameters as functions of the
Over the course of 22 months, we collected a total of monthly mean consumed resource from each habitat as
6,176 g (dry mass) of fruits and seeds from the seed traps covariates (Table 1a, b). For gaps the slope (beta
installed in nine gaps. We were able to identify 39 plant estimate) of the linear-logistic model was 0.119/0.06
species. For the understorey, we obtained a total of 8,963 SE, while for the understorey the slope was 1.279/0.25
g and identified 31 plant species. The 39 species from the SE. Both slopes are positive indicating a positive
gaps represented 55.2% of the overall seed weight, and relationship between survival and mean consumed
the 31 species from the understorey represented 58.7% of resource.
the total. We attempted to test all identified plant species For Proechimys in gaps, the low-AICc model per-
in feeding experiments, depending on the availability of mitted variation in survival and capture probability
the rodent species (below). monthly but contained no variation across size classes.
There was no difference in resource density (F1,33 / The best model for the understorey permitted monthly
0.107, P/0.746) between gaps (/X/7.0 g/m29/1.20) and variation as a function of the monthly mean of
understorey (/X/9.2 g/m29/1.46) or between size classes consumed resource for both survival and capture prob-
(F1,33 /0.125, P/0.726, Fig. 1). However, more re- ability, but the latter parameter was kept constant across
sources were available during the rainy season (/X/11.4 size classes (Table 1c, d). The slope of the linear-logistic
g/m29/1.46) than during the dry season (/X/5.4 g/m29/ model for gaps was 0.099/0.08 SE, and for the under-
1.17; F1,33 /22.625, PB/0.0001, Fig. 1). storey the slope was 1.019/0.93 SE. Again, both slopes

32 OIKOS 106:1 (2004)


indicate a positive relationship between survival and Proechimys abundance did not differ seasonally
mean consumed resource. (F1,40 /1.369, P /0.249); however, it was higher in
both large (/X/9.329/1.5) and small (/X/6.49/2.36)
gaps than in the large (/X/4.09/1.13) and small (/X/
1.449/0.37) understorey (F1,40 /18.794, PB/0.0001, Fig.
Abundance estimates 3b).

In large gaps, Oryzomys had a higher monthly abun-


dance during the rainy season (/X/19.199/6.7) than the
dry season (/X/8.679/1.9, F1,7 /5.221, P /0.056, Fitness parameters
Fig. 3a). In contrast, in small gaps, there was no seasonal
difference (F1,13 /1.168, P/0.299). In large under- Survival
storey, Oryzomys had a higher abundance during the During the rainy seasons, there was no difference in
dry season (/X/15.329/2.3) than the rainy season (/X/ Oryzomys survival by habitat (F1,8 /2.371, P/0.162)
6.059/2.3, F1,7 /21.406, P/0.002). Similarly, in small or size class (F1,8 /3.123, P/0.115). However, during
understorey, Oryzomys had a higher abundance during the dry seasons, Oryzomys had a higher survival in the
the dry season (/X/9.629/1.6) than during the rainy understorey than in gaps (F1,8 /166.981, PB/0.0001,
season (/X/3.599/0.6, F1,13 /18.364, P/0.001, Fig. Fig. 4a), and was higher in large system than small ones
3a). During the rainy season, Oryzomys was more (F1,8 /449.290, P B/0.0001).
abundant in gaps (/X/10.739/2.8) than in the under- Proechimys had a higher survival (F1,10 /7.055, P/
storey (/X/4.499/0.6, F1,18 /10.086, P/0.005, Fig. 0.024, Fig. 4b) in gaps than in the understorey during the
3a). During the dry season, however, there were no rainy seasons. However, there was no difference in
differences between habitats (F1,18 /2.908, P /0.105) or survival during the dry seasons between habitats
size classes (F1,18 /2.101, P /0.164, Fig. 3a). (F1,6 /1.107, P / 0.333).

Fig. 3. Mean (9/1 SE) seasonal abundance in large (n/4) and


small (n/7) gaps and understorey habitats based on two years
of trapping in CC, Peru. (a) Oryzomys had a significantly higher Fig. 4. Mean (9/1 SE) seasonal survival probability in large
abundance during the rainy season in gaps than in understorey. (n/4) and small (n/7) gaps and understorey habitats based
There was no difference in abundance during the dry season on two years of trapping in CC, Peru. (a) During the dry season,
between habitats or size classes. (b) Proechimys abundance did Oryzomys had a significant higher survival in the understorey
not differ seasonally, but was significantly higher in both large than in gaps. (b) During the rainy season, Proechimys had a
and small gaps than in the large and small understorey. significant higher survival in gaps than in understorey.

OIKOS 106:1 (2004) 33


Per capita recruitment There was no difference in the number of Oryzomys
immigrants by season (F1,41 /0.071, P /0.791) or by
There was no difference in Oryzomys per capita recruit-
size class (F1,41 /3.475, P /0.069). But more indivi-
ment between seasons (F1,20 /0.052, P/0.823), or size
duals immigrated into gaps than into the understorey
classes (F1,20 /1.100, P /0.307). However, Oryzomys
(F1,41 /8.884, P/0.005, Fig. 6a).
had a higher seasonal per capita recruitment (F1,20 /
Because of low sample size for Proechimys, we pooled
6.993, P /0.016, Fig. 5a) in the understorey (/X/
the data across seasons and size class and tested for
3.179/0.48) than in gaps (/X/1.679/0.48).
differences between habitats using a Mann /Whitney U-
For Proechimys there was no difference in per capita
recruitment between seasons (F1,19 /0.007, P/0.936), test. More individuals immigrated into the understorey
or size classes (F1,19 /0.074, P /0.789). However, than into the gaps (U/91.00, P/0.003, Fig. 6b).
Proechimys had a higher seasonal recruitment rate
(F1,19 /12.932, P/0.002, Fig. 5b) in gaps (/X/7.89/
0.59) than in the understorey (/X/3.89/0.51).
Discussion
Resource density
Movement into habitats The higher resource availability in the understorey
during the rainy season than during the dry season
We never encountered an individual that was captured in
the same habitat type at different locations. Therefore, was consistent with previous studies (Foster 1985a,b,
the distance between individual gaps or understorey Levey 1988, Janson and Emmons 1990, Lugo and Frangi
plots was large enough for the plots to be considered 1993). However, the similarity in resource availability
independent sampling units. between the two habitats is inconsistent with previous

Fig. 5. Mean (9/1 SE) per capita recruitment in large (n /4) Fig. 6. Mean (9/1 SE) number of immigrants in large (n/4)
and small (n/7) gaps and understorey habitats based on two and small (n/7) gaps and understorey habitats based on two
years of trapping in CC, Peru. (a) Oryzomys had a significant years of trapping in CC, Peru. (a) Significantly more Oryzomys
higher seasonal per capita recruitment in the understorey than immigrated into gaps than into the understorey, but there was
in gaps. There was no difference between size class and season. no difference between size class and season. (b) Because of low
(b) Proechimys had a significant higher per capita recruitment sample size for Proechimys data were pooled across seasons and
in gaps than in the understorey. There was no difference size classes. Significantly more individuals immigrated into the
between size class and season. understorey than into gap habitats.

34 OIKOS 106:1 (2004)


studies (Piñero et al. 1984, Sarukhán et al. 1984, Barrett 1996, Struhsaker 1998, Batzli et al. 1999, Beck
Denslow et al. 1986, Levey 1988, 1990). 2002) are two major limiting factors for mammals.
The salient question is whether different methodolo- Numerous studies suggest that many frugivorous and
gical approaches used among studies to measure re- herbivorous mammals are controlled in a bottom-up
source availability led to different conclusions. Blake et fashion by seasonal shortages in food resources (Foster
al. (1990) evaluated different methods of quantifying 1985b, Janson and Emmons 1990, Milton 1990).
resource availability. They pointed out several potential Furthermore, Terborgh (1986) suggested that food
biases when using seed traps. For example, fruits might shortages during the dry season cause frugivorous
not fall to the ground and therefore, are not available for mammals in Cocha Cashu to turn to alternative food
terrestrial animals. However, they can be harvested by resources that include insects and leaves. Despite our
arboreal frugivores. If the objective is to quantify data indicating low resource availability in both habitats
resource availability for terrestrial animals, as in this during the dry seasons, one cannot necessarily conclude
study, however, then seed traps are an appropriate and that animals experience famine. For example, Janos et al.
effective method. This assumes that traps are regularly (1995) quantified the consumption of vesicular-
checked and maintained to minimize biases (e.g. due to arbuscular mycorrhizal fungi by measuring spores in
decay processes). faecal samples of Oryzomys and Proechimys at Cocha
Most studies have quantified fruit phenology using Cashu. They found that faecal samples from Proechimys
either direct counts of fruits on plants or area-based contained a significantly higher number of Glomus
sample approaches. For example, Levey (1988, 1990) spores during the dry season than during the rainy
assessed resource availability monthly along transects by season.
recording all fruiting plants and their fruit abundance. Based on our findings of comparable resource avail-
He found that fruiting plants were more abundant in ability in gaps and understorey and similar consumption
gaps and carried more fruits over a longer time period rate of both rodent species, we conclude that resource
than conspecifics in the understorey. Additionally, Levey availability is not driving habitat-specific population
(1990) found a lower seasonal variance in fruit produc- dynamics.
tion in gaps than in the understorey. The higher resource We specified the monthly means of resource density as
availability in the canopy of gaps led to a higher covariates for modelling survival. As indicated by the
abundance of frugivorous birds and consequently higher positive slopes of the linear-logistic models, survival was
fruit removal rates than in the understorey (Alvarez- a function of resource availability. Despite the wide-
Buylla and Martinez-Ramos 1990). The continuously spread belief that animal population dynamics are
closely tied to resource abundance, evidence of a
higher resource availability in gaps compared to under-
relationship between resources and survival is relatively
storey led Levey (1990) to categorize gaps as keystone
rare and represents an important result of this study.
habitats for frugivores.
We are aware of only two other studies that employed
seed traps to quantify resource availability simulta-
neously in treefall gaps and understorey. Augspurger
and Franson (1988) quantified the number of seeds Population dynamics
dispersing into 43 treefall gaps and understorey during A question is whether non-volant mammal species could
an 11-week period in the dry season in Panama. They maintain a local population within gaps. How small in
found that more wind-dispersed seeds arrived in gaps area can a habitat patch be to encompass all ecological
than in the understorey, whereas a higher number of requirements needed to maintain a local terrestrial
non-wind-dispersed seeds arrived in the understorey population? In Panama, Lambert and Adler (2000)
than in gaps. For total seeds, however, there was no found that Proechimy s preferred gap habitats to the
significant difference between habitats. Loiselle et al. understorey. In CC, Emmons (1982) radio tracked three
(1996) established seed traps directly at ground level in Proechimys species and reported that individuals pre-
five treefall gaps and understorey locations in Costa ferred treefall gaps to the understorey. Their daily home
Rica. They concluded that there was no difference in range ranged from 0.023 to 0.174 ha. Taking together,
resource availability between gaps and understorey. The these studies demonstrate that local populations can
results of these two studies agree with our findings on persist and reproduce within habitat patches or home
resource availability. Although it is possible that gaps are ranges that are smaller than the gaps used in this study
‘‘keystone habitat’’ for arboreal frugivores as suggested (overall X/545 m2, 9/95 SE, range 173 /1143 m2).
by Levey (1990), our data suggest that this is not the case Within a heterogeneous landscape where individuals
for terrestrial species including terrestrial arthropods are distributed among patches of different qualities,
and mammals (Beck 2002). alternative models have been proposed to describe
Food resources (Foster 1985b, Terborgh 1986, Milton population dynamics and dispersal. For example,
1990) and/or cover as refuge from predation (Peles and based on different assumptions, authors have suggested

OIKOS 106:1 (2004) 35


a variety of source-sink models (Shmida and Ellner structural heterogeneity (i.e. more ground cover, more
1984, Holt 1985, Pulliam 1988, Pulliam and Danielson vines and lianas) than the understorey (Beck 2002,
1991). Per capita reproductive recruitment exceeds 2003). For Proechimys gaps may represent source
mortality for population in source habitats, with density- habitats where individuals have higher fitness compared
dependent dispersal. In the absence of movement, to individuals in the understorey.
populations in sink habitats experience an average In conclusion, our seed trap data clearly demonstrated
growth rate of less than one. Sink populations are only that treefall gaps did not receive greater resource input
able to persist because of continuous net immigration than the understorey, and both habitats exhibited
from a source population. seasonal variation in resource availability. Therefore,
Oryzomys in the understorey had higher survival and gaps are not superior habitats for terrestrial frugivore
per capita recruitment than in gaps. Furthermore, more taxa compared to the understorey with respect to fruit
individuals immigrated from the understorey into gap and seed resource availability or consumption rate.
than into understorey habitat. Given the absence of However, species-specific predation pressure by ants
resource differences between gaps and understorey, one may lead to unequal fitness across habitats. Gaps may
reason why Oryzomys apparently had fewer recruits in represent a sink habitat for Oryzomys populations,
gaps than in the understorey could be the frequent ant whereas the understorey may be a source habitat.
attacks in gaps. In both habitats, ants attacked trapped Conversely, gaps may represent a source habitat for
individuals. Free-ranging individuals undoubtedly are Proechimys, whereas, the understorey may be a sink
able to avoid potential attacks. However, Oryzomys habitat. Our findings of potential reciprocal source /sink
produce altricial young; their eyes open after a week dynamics for two rodent species may shed light on the
(Eisenberg and Redford 1999). These immobile new- community structure of mammals and provide a possible
borns are vulnerable to ant attacks (Janzen and Wilson mechanism for local co-existence within a species diverse
1983). If we assume that ant attacks on trapped Amazonian rainforest.
individuals reflect ant foraging activity in each habitat, Acknowledgements / We thanks the Peruvian Ministry of
we can test if predation pressure by ants differs between Agriculture, Directorate of Forestry and Fauna, for
habitats by comparing the frequency of individuals permission to work in the Manu National Park. For their
help in identifying innumerable seeds and fruits we thank J.
attacked in traps. We found that two species of ants Terborgh, P. Nuńez, M. Foster, R. King, and F. Bossuyt. D.
(Camponotus sericeiventris and C. femoratus ) attacked Davidson identified the ant species. M. Rosenzweig provided
more individuals in gaps (n/20) than in the understorey stimulating discussions. For assistance in the field we thank D.
(n /4, x20.05 (1) /9.481, P/ 0.01). Similarly, Adams et O’Dell, A. Helm, A. Hudson, A. Sorenson, G. Wells, P. Alvarez
and N. v. Vliet. Financial support to HB was provided by a NSF
al. (2001), Murphy (2001) and Rosenheim (2001) found Dissertation Improvement Grant (DEB-9901993), Sigma Xi
that predation on early life stages of several taxa led to a Grant-in-Aid, Tropical Fellowship through the University of
significant decline in reproduction and population Miami, Biology Department, and GAFAC grants from the
University of Miami. For review of the manuscript we thank D.
density and resulted in source-sink dynamics. Habitat- DeAngelis, T. Fleming, L. Hansson, C. Horvitz, D. Janos and
specific predation may reduce the fitness of Oryzomys in D. Wilson.
gaps. Therefore, gaps may represent sink habitats for
Oryzomys, and their populations may be maintained by
immigration from the understorey source habitat.
Proechimys in gaps had higher survival and per capita References
recruitment than in the understorey. A lower predation Abrams, P. A. 2000. The impact of habitat selection on the
rate, possibly because of greater structural heterogeneity spatial heterogeneity of resources in varying environments.
/ Ecology 81: 2902 /2913.
in gaps (Beck 2002, 2003), could be one mechanism Adams, A. A. Y., Skagen, S. K. and Adams, R. D. 2001.
explaining why survival may be higher in gaps than in Movements and survival of lark bunting fledglings.
understorey. Peles and Barrett (1996) also reported that / Condor 103: 643 /647.
Adler, G. H., Mangan, S. A. and Suntsov, V. 1999. Richness,
greater structural heterogeneity led to decrease in pre- abundance, and habitat relations of rodents in the Lang Bian
dation rates. Unlike Oryzomys, Proechimys has precocial mountains of southern Viet Nam. / J. Mammal. 80: 891 /
newborn; therefore, ant attacks should not affect their 898.
Alvarez-Buylla, E. R. and Martinez-Ramos, M. 1990. Seed
survival. Ants attacked only one trapped individual in a bank versus seed rain in the regeneration of a tropical
gap. Abundance in gaps was higher than in the under- pioneer tree. / Oecologia 84: 314 /325.
storey throughout the year. These results are consistent Anderson, D. R., Burnham, K. P. and White, G. C. 1994. AIC
with those of Emmons (1982), who radio-tracked several model selection in overdispersed capture-recapture data.
/ Ecology 75: 1780 /1793.
Proechimys spp. in Cocha Cashu and reported that they Augspurger, C. K. and Franson, S. E. 1988. Input of wind-
primarily utilized gap habitats. Similarly, in Panama, dispersed seeds into light-gaps and forest sites in a Neo-
Lambert and Adler (2000) reported that Proechimy s tropical forest. / J. Trop. Ecol. 4: 239 /252.
Batzli, G. O., Harper, S. J., Lin, Y. K. et al. 1999. Experimental
preferred gaps and areas with high densities of logs and analysis of population dynamics: scaling up to the land-
lianas within the forest. Treefall gaps have higher scape. / In: Barrett, G. W. and Peles, J. D. (eds), Ecology of

36 OIKOS 106:1 (2004)


small mammals at the landscape level: experimental ap- Hoppes, W. G. 1987. Pre-and post-foraging movement of
proaches. Springer Verlag, pp. 107 /127. frugivores birds in an eastern deciduous forest woodland,
Beck, H. 2002. Population dynamics and biodiversity of small USA. / Oikos 49: 281 /290.
mammals in treefall gaps within an Amazonian rainforest. Hubbell, S. P., Foster, R. B., O’Brien, S. T. et al. 1999. Light-gap
Ph.D. thesis / Univ. of Miami. disturbances, recruitment limitation, and tree diversity in a
Beck, H. 2004. Natural disturbances increases structural Neotropical forest. / Science 283: 554 /557.
heterogeneity and small mammal biodiversity in the Ama- Janson, C. H. and Emmons, L. H. 1990. Ecological structure of
zon. / Ecology, in press. the nonflying mammal community at Cocha Cashu Biolo-
Blake, J. G., Loiselle, B. A., Moermond, T. C. et al. 1990. gical Station, Manu National Park, Peru. / In: Gentry, A.
Quantifying abundance of fruits for birds in tropical H. (ed.), Four neotropical rainforests. Yale Univ. Press, pp.
habitats. / Studies Avian Biol. 13: 73 /79. 314 /338.
Brokaw, N. V. L. 1982. The definition of treefall gap and its Janos, D. P., Sahley, C. T. and Emmons, L. H. 1995. Rodent
effect on measures of the forest dynamics. / Biotropica 14: dispersal of vesicular-arbuscular mycorrhizal fungi in Ama-
158 /160. zonian Peru. / Ecology 76: 1852 /1858.
Bullock, S. H. 1980. Demography of an under-growth palm in Janzen, D. H. and Wilson, D. E. 1983. Mammals: Introduction.
littoral Cameroon. / Biotropica 12: 247 /255. / In: Janzen, D. H. (ed.), Costa Rican natural history. The
Burnham, K. P. and Anderson, D. R. 1998. Model selection and Univ. of Chicago Press, pp. 426 /440.
inference: a practical information theoretic approach. Lambert, T. D. and Adler, G. H. 2000. Microhabitat use by a
/ Springer Verlag. tropical forest rodent, Proechimys semispinosus, in Central
Burnham, K. P., Anderson, D. R., White, G. C. et al. 1987. Panama. / J. Mammal. 81: 70 /76.
Design and analysis methods for fish survival experiments Lebreton, J. D, Burnham, K. P., Clobert, J. et al. 1992.
based on release- recapture. / Am. Fish. Soc. Monogr. 5. Modeling survival and testing biological hypotheses using
Calsson-Granér, U. and Thrall, P. H. 2002. The spatial marked animals: a unified approach with case studies.
distribution of plant populations, disease dynamics and / Ecol. Monogr. 62: 67 /118.
evolution of resistance. / Oikos 97: 97 /110. Lee, R. 1978. Forest microclimatology. / Columbia Univ. Press.
Chazdon, R. L. and Fetcher, N. 1984. Photosynthetic light Levey, D. J. 1988. Tropical wet forest treefall gaps and
environments in a lowland tropical rain forest in Costa distributions of understory birds and plants. / Ecology 69:
Rica. / J. Ecol. 72: 553 /564. 1076 /1089.
Clark, D. B. and Clark., D. A. 1987. Population ecology and Levey, D. J. 1990. Habitat-depending fruiting behavior of an
microhabitat distribution of Dipterix panamensis, a neotro- understory tree, Miconia centrodesma , and tropical treefall
pical rain forest emergent tree. / Biotropica 19: 236 /244. gaps as keystone habitats for frugivores in Costa Rica. / J.
Crome, F. H. J. and Richards, G. C. 1988. Bats and gaps: Trop. Ecol. 6: 409 /420.
microchiropteran community structure in a Queensland rain Lin, Y. K. and Batzli, G. O. 2001. The influence of habitat
Forest. / Ecology 69: 1960 /1969. quality on dispersal, demography, and population dynamics
Delibes, M., Gaona, P. and Ferreras, P. 2001. Effects of an of voles. / Ecol. Monogr. 71: 245 /274.
Loiselle, B. A., Ribbens, E. and Vargas, O. 1996. Spatial and
attractive sink leading into maladaptive habitat selection.
temporal variation of seed rain in a tropical lowland wet
/ Am. Nat. 158: 277 /285.
forest. / Biotropica 28: 82 /95.
Denslow, J. S. 1995. Disturbance and diversity in tropical rain
Lugo, A. E. and Frangi, J. L. 1993. Fruit fall in the
forests: the density effect. / Ecol. Appl. 5: 962 /968.
Luquillo Experimental Forest, Puerto Rico. / Biotropica
Denslow, J. S., Moermond, T. C. and Levey, D. J. 1986. Spatial
25: 73 /84.
components of fruit display in understory trees and shrubs.
Lwanga, J. L. 1994. The role of seed and seedling predators and
/ In: Estrada, A. and Fleming, T. H. (eds), Frugivores and browsers in the regeneration of two forest canopy species
seed dispersal. W. Junk Publisher, pp. 37 /44. (Mimusops bagshawei and Strombosia scheffleri ) in Kibale
De Steven, D., Windsor, D. M., Putz, F. E. et al. 1987. Forest Reserve, Uganda. Ph.D. thesis, Univ. of Florida,
Vegetative and reproductive phenologies of a palm assem- Gainesville.
blages in Panama. / Biotropica 19: 342 /356. MacArthur, R. H. and Pianka, E. R. 1966. On optimal use of a
Eisenberg, J. F. and Redford, K. H. 1999. Mammals of the patchy environment. / Am. Nat. 100: 603 /609.
Neotropics. Volume III. The central Neotropics Ecuador, MacArthur, R. H., Recher, H. and Cody, M. 1966. On the
Peru, Bolivia, and Brazil. / Univ. of Chicago Press. relationship between habitat selection and species diversity.
Emmons, L. H. 1982. Ecology of Proechimys (Rodentia, / Am. Nat. 100: 319 /332.
Echimyidae) in south-eastern Peru. / Trop. Ecol. 23: 280 / McPeek, M. A. and Holt, R. D. 1992. The evolution of dispersal
290. in spatially and temporally varying environments. / Am.
Ende, v. C. N. 2001. Repeated-measures analysis. / In: Nat. 140: 1010 /1027.
Schneider, S. M. and Gurevitch, J. (eds), Design and analysis Marquis, R. J. 1988. Phenological variation in the neotropical
of ecological experiments, 2nd ed. Oxford Univ. Press, pp. understory shrub Piper arieianum : causes and conse-
134 /157. quences. / Ecology 69: 1552 /1565.
Foster, R. B. 1985a. The seasonal rhythm of fruitfall on Barro Martinez-Ramos, M., Sarukhán, J. and Piñero, D. 1988. The
Colorado Island. / In: Leigh, E. G., Rand, A. S. and demography of tropical trees in the context of forest gap
Windsor, D. M. (eds), The ecology of a tropical forest: dynamics: the case of Astrocaryum mexicanum at Los
seasonal rhythms and long term changes, 2nd ed. Smithso- Tuxtlas tropical rain forest. / In: Davy, D. J., Hutchings,
nian Institution Press, Washington D.C., pp. 151 /172. M. J. and Watkinson, A. R (eds), Plant population ecology.
Foster, R. B. 1985b. Famine on Barro Colorado Island. / In: Blackwell, pp. 293 /313.
Leigh, E. G., Rand, A. S. and Windsor, D. M. (eds), The Milton, K. 1990. Annual mortality patterns of a mammal
ecology of a tropical forest: seasonal rhythms and long term community in central Panama. / J. Trop. Ecol. 6:
changes, 2nd ed. Smithsonian Institution Press, Washington 493 /499.
D.C, pp. 201 /213. Molino, J. F. and Sabatier, D. 2001. Tree diversity in tropical
Foster, M. S. and Terborgh, J. 1998. Impact of a rare storm rain forests: a validation of the intermediate disturbance
event on an Amazonian Forest. / Biotropica 30: 470 /474. hypothesis. / Science 294: 1702 /1704.
Holt, R. D. 1985. Population dynamics in two-patch environ- Murphy, M. T. 2001. Habitat-specific demography of a long-
ments: some anomalous consequences of an optimal habitat distance, neotropical migrant bird, the eastern kingbird.
distribution. / Theoret. Popul. Biol. 28: 181 /208. / Ecology 82: 1304 /1318.

OIKOS 106:1 (2004) 37


Murray, K. G. 1987. Selection for optimal fruit-crop size in Sarukhán, J. (eds), Perspectives in plant population ecology.
bird-dispersed plants. / Am. Nat. 129: 18 /31. Sinauer Ass, Massachusetts, pp. 141 /165.
Otis, D. L., Burnham, K. P., White, G. C. et al. 1978. Statistical Schemske, D. W. and Brokaw, N. 1981. Treefall and the
interference from capture data on closed animal popula- distribution of understory birds in a tropical forest.
tions. / Wildl. Monogr. 62: 1 /135. / Ecology 62: 938 /945.
Peles, J. D. and Barrett, G. W. 1996. Effects of vegetative cover Seber, G. A. F. 1982. The estimation of animal abundance and
on the population dynamics of meadow voles. / J. Mammal. related parameter. 2nd ed. / MacMillan.
77: 857 /869. Shmida, A. and Ellner, S. 1984. Coexistence of plant species
Piñero, D., Martinez-Ramos, M. and Sarukhán, J. 1984. A with similar niches. / Vegetatio 58: 29 /55.
population model for Astrocaryum mexicanum and a SPSS. 2000. SPSS 10.1.0 for Windows. / SPSS Inc.
sensitivity analysis of its finite rate of increase. / J. Ecol. Stevens, J. 1996. Applied multivariate statistic for the social
72: 977 /991. sciences. 3rd ed. / Lawrence Erlbaum Associates.
Pollock, K. H., Nichols, J. D., Brownie, C. et al. 1990. Statistical Struhsaker, T. T. 1998. Ecology of an African rain forest.
interference for capture-recapture experiments. / Wildl. / Univ. Press of Florida.
Monogr. 107. Swaine, M. D. 1996. Foreword. / In: Swaine, M. D. (ed.), The
Pulliam, H. R. 1988. Sources, sinks, and population regulation. ecology of tropical forest tree seedlings. Man and the
/ Am. Nat. 132: 652 /661. biosphere series. Vol. 17. The Parthenon Publisher Group,
Pulliam, H. R. and Danielson, B. J. 1991. Sources, sinks, and pp. 11 /17.
habitat selection: a landscape perspective on population Terborgh, J. 1986. The socio-ecology of primate groups.
dynamics. / Am. Nat.137: S50 /S66. / Annu. Rev. Ecol. Syst. 17: 111 /135.
Pusenius, J., Ostfeld, R. S. and Kessing, F. 2000. Patch selection Terborgh, J. 1990. An overview of research station Cocha Cashu
and tree-seedling predation by resident vs immigrant Biological Station. / In: Gentry, A. H. (ed.), Four neotro-
meadow voles. / Ecology 81: 2951 /2956. pical rainforests. Yale Univ. Press, pp. 48 /59.
Rexstad, E. and Burnham, K. P. 1991. User’s guide for Voss, R. S. and Emmons, L. H. 1996. Mammalian diversity in
interactive program Capture. / Colorado Cooperative Neotropical lowland rainforests: a preliminary assessment.
Fish and Wildlife Research Unit. / Bull. Am. Mus. Nat. Hist. 230.
Rosenheim, J. A. 2001. Source-sink dynamics for a generalist Weksler, M., Bonvicino, C. R., Otazu, I. B. et al. 2001. Status of
insect predator in habitats with strong higher-order preda- Proechimys roberti and P. oris (Rodentia: Echimyidae) from
tion. / Ecol. Monogr. 71: 93 /116. eastern Amazonia and central Brazil. / J. Mammal. 82:
Sanford, R. L. Jr., Braker, H. E. and Hartshorn, G. S. 1986. 109 /122.
Canopy openings in a primary neotropical lowland forest. White, G. C. 2001. Program MARK. Version 1.8 Colorado
/ J. Trop. Ecol. 2: 277 /282. State Univ. http://www.cnr.colostate.edu//gwhite/mark/
Sarukhán, J., Martinez-Ramos, M. and Piñero, D. 1984. The mark.html
analysis of demographic variability at the individual level Zar, H. J. 1996. Biostatistical analysis. 3rd ed. / Prentice-Hall
and its population consequences. / In: Dirzo, R. and Inc. Simon and Schuster.

38 OIKOS 106:1 (2004)

Você também pode gostar