DdZ/> KDWh^dK
1.1.4. MATERIALES DE REFUERZO
1.1.4.3. FIBRA DE VIDRIO [14]
5.1 Introduction
Changes in community trophic structure accompanying biodiversity loss are a
commonly observed outcome of tropical forest fragmentation. These changes are
important in influencing trajectories of ecosystem function as well as ecosystem
collapse of forest fragments.
A number of factors are thought to drive trophic structural changes in habitat
fragments generally. Area and isolation effects may drive differential species loss
from higher trophic levels resulting in shortening of food chain lengths in small habitat
fragments (Schoener 1989). Models developed by Holt (1996) mirror much empirical
observation in predicting that the slope of species-area relations should increase with
trophic rank. Biotic and abiotic edge effects of fragmentation can also influence
trophic structuring of rain forest communities via the sum of differential species
responses to microclimatic and other gradients. The type of matrix habitat surrounding
forest fragments may also have a profound influence on trophic structuring. In the case
of rain forest fragments surrounded by a terrestrial matrix, influx of disturbed habitat
species may be important. However, where the matrix habitat is aquatic, trophic
distortions affecting communities of tropical forest islands are known to be more
on species dispersion and mobility and their greater negative impact on persistence of
higher trophic levels.
Studies of changes in trophic structuring resulting from rain forest
fragmentation have largely focused on above- rather than below-ground food webs,
and upon vertebrates rather than invertebrates. Hence, the responses of decomposer
communities have been largely neglected. Yet decomposers of plant-derived organic
material are likely to be strongly directly affected by forest fragmentation due to the
phenomenon of biomass collapse.
An 18-year study of tree community structure and dynamics at the BDFFP has
shown that hotter, drier microclimatic conditions and greater exposure to wind, with
decreasing distance to forest edges, are responsible for increased tree mortality, and
consequent tree recruitment (Laurance et al. 1998). Laurance et a l (1997) have shown
that BDFFP forest fragments have undergone a net loss in biomass of around 10%
within two to four years of isolation, the time interval during which the rate of tree
die-off may be maximal. We might expect that the increased availability of dead plant
matter would present a resource bonanza for decomposer organisms that are able to
tolerate the negative influences of forest fragmentation. Changes in diversity and
abundance of decomposers surviving in forest fragments over time should have a
direct bearing on rates of breakdown of dead plant biomass and greenhouse gas
production. Moreover, we would expect these changes to have an impact on above
ground food webs inside forest fragments.
This study presents data on the early effects of forest fragmentation on
functional group composition of the termite assemblage across the SEFP (for further
presented in order to establish the degree to which biotic and abiotic edge effects had
become established on islands at the time of termite sampling. In particular, the
evidence for tree-die off and consequent changes in canopy cover and availability of
dead wood are presented. The findings are compared with those of the BDFFP study
(De Souza and Brown 1994), particularly in the context of the profound differences in
the matrix habitats surrounding fragments at the two sites. Possible hypotheses to
explain the observed patterns are reviewed, and the implications for changes in
ecosystem function and carbon flux discussed.
5.2 Methods
TERMITE FEEDING GROUPS.
Following Donovan et a l (in press) we placed each species in the St. Eugène
assemblage into one of four feeding groups according to the morphological characters
of the workers indicating their position along the humification gradient of organic
matter decomposition. Using this classification, group I wood feeders feed on the least
humified substrates and group IV soil feeders feed on the most humified ones.
Additionally, a more fundamental level of feeding group distinction used was that
between wood and leaf-litter feeders (groups I and II combined) and soil feeders
ANALYSES
(i) Relationships between environmental variables across sites.
Logistic regression was used to analyse the apportionment of the total numbers of
large standing tree trunks (of >10 cm diameter above breast height) per transect,
between standing dead and live trunks combined, in relation to logio (distance to forest
edge) and logio (area), and to analyse the relationship between percentage canopy
cover and the same two fragmentation variables.
In order to control for any possible influence of the ghost forest termite
assemblage upon smaller island assemblages, all ghost forest species also found in
terrestrial transect samples were excluded from datasets, both of termite relative
encounters and species richness across sites.
(ii) Termite relative encounters across feeding groups in relation to environmental
variables.
Termite relative encounters were analysed at the levels of feeding group (I-IV) using
ordination methods. PCA was the indirect gradient method used, and RDA the direct
gradient method used to investigate the relative encounters of termites across feeding
groups in relation to environmental variable data collected at each of the 13 sites (see
Chapter 4, methods section). As in ordinations of species level data (Chapter 4),
partialling out of any effects of season on relative encounter data was achieved by
were partial analyses. Data for number of encounters within each feeding group were
logio (x + 1) - transformed prior to analyses. In the RDA, forward selection was used
to rank environmental variables in order of their importance in explaining variation in
relative encounters across feeding groups. Marginal eigenvalues were computed for
each environmental variable, and significance at each stage (i.e. for each variable
selected) was tested using a Monte Carlo permutation test with 999 random
permutations.
(iii) Relative encounters o f wood and soil feeding termites, and the whole assemblage,
in relation to environmental variables.
Linear regression was used to test the relationship between both main fragmentation
variables, logio (edge distance) and logic (fragment area), and logic (encounters) for
wood and leaf-litter feeders (groups I + II) and soil feeders (groups in + IV), and for
the total combined assemblage. Seasonality is known to influence abundance and
diversity of termite samples (Dibog et al. 1998). As termite sampling was carried out
in the rainy season for nine transects, and during the dry season for four transects, we
tested for significant effects of season upon species richness and logic (encounters).
Where season was shown to be significant, all regression analyses were performed
with season partialled out.
Forward stepwise multiple regression was used to determine which of the 17
environmental variables measured were significant predictors of logic (encounters)
and species richness for wood and leaf-litter feeders, soil feeders, and for the whole
used simply to see whether fragmentation variables, or uncorrelated environmental
gradients, were having the greatest influence on the distribution of logio (encounters)
and species richness between functional groups, and for the whole assemblage, across
sites.
5.3 Results
(i) Environmental variables.
Linear regressions against the two fragmentation variables produced significant results
for only three environmental variables in the case of logic (distance to forest edge),
with only two of these showing significance with logic (area) (Table 5.1). As logic
Table 5.1. Significant results of linear regressions of log] o (distance to forest edge) and logio (area) against 15 environmental variables, across 13 standardised transects (see also Figure 5.1).
Dependent variable: Independent variable: F d P Slope F-SMALL LGEDGE 13.42 1, 1 0.004 0.550 -v e F-SMALL LGAREA 6.70 1, 1 0.025 0.378 -v e V-STAND LGEDGE 5.06 1, 1 0.046 0.315 -v e MOISTURE LGEDGE 9.21 1, 1 0.011 0.456 +ve MOISTURE LGAREA 5.85 1, 1 0.034 0.347 +ve
(distance to forest edge) showed consistently stronger correlations with the
environmental variables than logic (area), it is the relationships with the former that
are presented graphically (Figure 5.1). The most significant correlation was with
frequency of small wood which showed a strongly negative, and approximately linear,
140 O 120 S 100 ST 40 3.0 0.6 1.0 1.4 1.8 2.2 2.6 3.4