DdZ/> KDWh^dK
1.1.4. MATERIALES DE REFUERZO
1.1.4.2. FIBRA DE ABACÁ
F-LARGE 0.7632 0.2307 -0.0677 0.0791 F-SMALL 0.4165 -0.6945 -0.4820 -0.1398 PALMS -0.8854 -0.2460 -0.2240 -0.0945 HUMUS -0.8230 0.2619 -0.1168 0.4211 PH 0.7605 0.4280 0.3173 -0.1996
small wood was, however, somewhat weakly correlated with axis 2, being also
negatively correlated with soil humus depth.
The results showed that palm density, soil humus depth, and the strongly
negatively correlated soil pH, were the most important factors determining variation in
species composition of the termite assemblage across the 13 sites. Greater density of
palms and soil humus depth were associated with the species compositions of two
large island transects (T8 and T2) and three medium sized island transects (T9, T5 and
TIO), whereas low palm density and soil humus depth were associated with the
species composition of transects from two mainland (T1 and T13) and two small
island sites (T6 and T il). Logio (distance to forest edge), as represented by the second
axis, was still shown to have a significant effect on species composition, with the
result that the two mainland transects furthest from the forest edge (T1 and T13)
showed greatest compositional differences with the four smallest islands (T3, T7, T6
and T il).
The partial RDA for soil feeding species only, showed almost identical results
to those for the complete assemblage, because the latter was in any case dominated by
soil feeding species. For this reason, only the resulting RDA bi-plot is presented
of the bi-plot was divided into quarters, each corresponding (approximately) to the
zone in which species vectors are most closely positively associated with one of the
four main environmental conditions: high palm density, low soil pH; low palm
density, high soil pH; large edge distance (and fragment area); and short edge distance
(and small fragment area). This revealed a clear difference in the environmental
conditions most favourable to two taxonomic sub-divisions of the soil feeder
assemblage. The great majority of Apicotermitinae (which make up 47% of all soil
feeding species found) were negatively associated with palm density (15 species)
rather than positively associated (three species). The remaining soil feeding species
(belonging to two further termite subfamilies, the Termitinae and Nasutitermitinae)
showed a slightly smaller imbalance, but tended to favour areas of high palm density
(15 species vs. nine species). The dichotomy in the division of species associations
between larger and smaller fragment size, and edge distance, was much less apparent.
The only difference was that more species of Apicotermitinae were associated with
smaller rather than larger fragment size and edge distance (nine species vs. five
species).
The partial RDA for wood and leaf litter feeders, however, showed that
environmental variables associated most closely with fragmentation, rather than other
factors, were having the greatest influence on species composition. Specifically,
forward selection showed that volume of standing dead wood had greatest marginal
significance in explaining variation in species composition, followed by soil moisture
and soil pH (Table 4.6). Again, the eigenvalues for the partial RDA (0.16, 0.12, 0.07,
and 0.13, for axes 1-4, respectively) were not markedly lower than those for the partial
LO L G E D G E JL L arger f r a g m e n ts Ap: 8 spp. N on-A p: 5sp p . P H ' S "N Ap: 15 sp p > N on-A p: 9 spp. P A L M S Ap: 3 sp p . N on-A p: 15 sp p . S m a ll e r f r a g m e n ts Ap: 8 sp p . N o n -A p : Sspp. LT) 1 . 5 + 1 . 5
Figure 4.13. Ordination bi-plot for partial RDA o f soil feeder species data and environmental variables, with season (wet and dry) entered as nominal categorical covariables. Centroids for sites have been removed for clarity. Axes 1 and 2 ordination space is divided into quadrants (by thick dashed line), corresponding to zones o f association with parts o f environmental gradients. Numbers o f species o f Apicotermitinae (Ap) and Non-Apicotermitinae (Non-Ap) associated with each quadrant are indicated. Species arrows are coded as follows: solid = Apicotermitinae species; dashed = Non-Apicotermitinae species (i.e. Termitinae,
Nasutitermitinae, see Appendix 4.2). (Environmental variable abbreviations are as for Figure 4.7).
Table 4.6. Results of partial RDA forward selection procedure, showing environmental variables explaining significant variance in termite wood and leaf litter species composition across 13 transect sites. À] = eigenvalue (fit) for each variable on its own; = increase in eigenvalue (additional fit); £Aa = cumulative total of eigenvalues Aa',P = significance level of effect, as obtained with a Monte Carlo permutation test under the null model with 999 random permutations conditioned on the covariables.
Marginal Effects Variable M P Conditional Effects Variable h. P EA,2 V-STAND 0.15 0.012 V-STAND 0.15 0.012 0.15 MOISTURE 0.14 0.002 PH 0.11 0.035 0.26 PH 0.12 0.038
environment correlations were high (0.978, 0.972, and 0.842, for axes 1-3,
repectively). A notable feature of the resulting bi-plot (Figure 4.14), in which only
significant marginal variables were included, was that the four smallest islands
showed the greatest differences in species composition, as indicated by the relatively
wide spacing of their centroids across axes 1 and 2 ordination space (T3, T6, T7 and
T il). In particular, the centroid for T6 from the smallest island, was most closely
associated with those for mainland and large island transects (Tl, T13 and T4).
Notable also, was the strong association between the centroids for transects (T il and
T7) from the second and third smallest islands, respectively, and species of
Cornitermes. This genus is known to inhabit hard epigeal soil mounds which are likely
to protect resident colonies from the microclimatic harshness that are a feature of drier
o Glypto A T4 T13 Het tenu IndKA Rugi A Ns guay MOISTURE
Ns oct Cor pug
V-STAND
Cyl parv
112 IndKB Cor web
N s J /j g Ns D
Ns gai Dol long
T11 Ac sub Ns suri 18 Vel beeb Ns sim Syn spin T3 Ns bank LO 1 . 0 + 2 . 0
Figure 4.14. Ordination bi-plot for partial RDA o f wood and leaf litter feeder species data and environmental variables, with season (wet and dry) entered as nominal categorical co variables. Circles representing centroids for transects are coded as: white = small islands (<1 ha); light grey = medium islands (1-10 ha); dark grey = large islands (10-100 ha); and black = mainland (1000 ha). Abbreviated species names are as follows: Ac sub, Acorhinotermes subjus ci ceps'.
Cor A, Cornitermes sp. A nr. cumulons'. Cor pug, Cornitermes pugnax'. Cor web, Cornitermes
weberi', Cyl parv, Cylindrotermes parvignathus', Dol long, Dolichorhinotermes longilabius',
Glypto A, Glyptotermes sp. A nr. guianensis', Het tenu, Heterotermes tenuis', IndK A ,
Indeterminate Kalotermitidae sp. A; IndK B, Indeterminate Kalotermitidae sp. B; Ns bank, Nasutitermes banksr, Ns gai, Nasutitermes gaigev, Ns guy, Nasutitermes guayanae', Ns oct, Nasutitermes octopilus', Ns sim, Nasutitermes similis', Ns suri, Nasutitermes surinamensis', Ns
D, Nasutitermessp. D; N sJ , Nasutitermessp. J; Rugi A, Rugitermessp. A; Syn spin,
Syntermes spinosus', Syn A, Syntermessp. A nr. longiceps', Vel beeb, Velocitermes beebei.
Ordinations to assess spatial autocorrelation.
The forward selection procedure for the partial RDA of the complete species datatset
across all 13 terrestrial sites, and the 9-term spatial matrix, indicated that three terms
(x, y, and xy) showed significant marginal effects (i.e. when each variable was tested
on its own) upon variation in species composition (Table 4.7) indicating significant
spatial autocorrelation in the species composition data. With these spatial terms
entered as covariables, together with season, a further partial RDA of the complete
species data and measured environmental variables revealed that only soil moisture
showed marginal significance in explaining the variance in species composition.
Nevertheless, soil moisture is strongly correlated with distance to the forest edge i.e.
forest fragmentation.
Table 4.7. Results of partial RDA forward selection procedure, showing spatial positioning terms explaining significant variance in termite species composition across 13 transect sites.
= eigenvalue (fit) for each variable on its own; = increase in eigenvalue (additional fit); IÀq, = cumulative total of eigenvalues ^ , P = significance level of effect, as obtained with a
Monte Carlo permutation test under the null model with 999 random permutations conditioned on the covariables. Marginal Effects Variable P Conditional Effects Variable ^2 P IÀ2 y 0.12 0.008 y 0.12 0.013 0.12 xy 0.12 0.022 X 0.12 0.028
Partitioning of the variation in species composition revealed that 32.6% of the
total variance in species composition was spatially structured, of which over a third
(9.1% of total variance) was spatially structured environmental variation, the
44.9% o f total variation w as due to non-spatial environm ental variation, leaving 22.6% as unexplained variation. Again, as soil feeders make up the majority o f sp ecies, the results o f an analysis for soil feeders alone w ere very sim ilar to th ose for the overall assem blage. A n alysis o f w oo d and le a f litter feed ers only, revealed a sm aller spatially structured com ponent o f variation (21.3% o f total) including considerably less spatially structured environm ental variation (just 3.0% o f total).
0) u c re "C re > re o •*-> 0) O) re -4-J c re e re Q . 2 3.1% 4 3.5% 2 1 .2 % 23.5% 11.7% 4 4.9% 44.0% 35.1%
Total W ood Soil
I I U n d e t e r m i n e d # # S p a c e ■ ■ S p a c e + e n v
E n v i r o n m e n t
Figure 4.15. Variance partitioning o f the termite species data across 13 terrestrial forest transects for; the total assemblage (total); for wood and leaf litter feeders (wood); and for soil feeders (soil), calculated from the results o f partial RDAs following Borcard e( a l (1992). Percentage variances for each component o f total variation are given inside white boxes. Abbreviations for variance components are as follows: undetermined, undetermined variance; space, spatially structured variance; space + env, spatially structured environmental variance; environment, non-spatially structured environmental variance.
4.4. Discussion
Overview o f main environmental and spatial effects upon assemblage composition.
Four years after the original inundation events, the effects of forest fragmentation
upon the total assemblage, and upon soil feeders in particular, were subordinate to the
influence of forest understory palm density, and the closely associated gradients of soil
humus depth and soil pH. Moreover, changes in these three variables were found to be
uncorrelated with any fragmentation gradient, and to largely account for the
component of spatially structured environmental variation detected in the species
composition data. Nevertheless, forest fragmentation appeared to be having a
significant effect on changes in termite species composition across the St. Eugène
study site. The validity of this result was brought into question by the finding that the
fragmentation gradient was partially confounded by spatial positioning of sites, when
considering variation in the complete assemblage. However, the significant influence
of soil moisture upon the wood and leaf-litter feeding sub-assemblage (when
controlling for spatial position) suggested a real effect of fragmentation. Further
indirect support for this species level fragmentation effect came from a feeding group
level analysis of the same dataset (see Chapter 5) showing that fragmentation has also
resulted in a clear shift in the relative encounters between wood and soil feeding
termites. In addition to the influence of palm density and fragmentation upon species
composition, purely spatial variation was clearly also influencing /^-diversity changes
The role o f forest palms in structuring termite assemblage composition from soil to understory.
The spatial heterogeneity in distribution of forest palms almost certainly pre
dated the inundation events that created forest islands. This is indirectly supported by
the findings of Scariot (1999) from the BDFFP site in Brazil showing that, between 10
and 15 years following habitat fragmentation, species richness of adult and juvenile
palms (but not seedlings) was unaffected by fragmentation, and that there was no
evidence of differential adult mortality. Increasing evidence points to the ubiquity of
edaphically-related spatial heterogeneity in plant community structure in old-growth
tropical wet forests, across scales ranging from 0.5 -10^ ha (Clark et al. 1995). Soil
drainage, topography, and forest architecture are all considered to be important in
influencing palm distributions in Amazonian terre firme forests (Kahn and de
Granville 1992). Although the relationship between palm density and soil pH has not
been reported previously, a Costa Rican study showed that lowest palm species
richness occurred on the soil type (alluvium) with highest fertility and pH (Clark et al.
1995). In Amazonian palm communities of terre firme forests, most species are small
or medium in size with 99% of individuals being under 10-m tall, with the result that
palms dominate the understory plant community (Kahn and de Granville 1992). The
funnel-like form of many adult understory palms with large leaves leads to
concentration of falling canopy leaf-litter and small woody debris towards their stems
(Kahn and de Granville 1992). This canopy litter can remain suspended in the crowns
of the palms themselves or become caught on above-ground stems, particularly in
Figure 4.16. Canopy litter and humus accumulations, and termite nest, between the armed petioles o f an understory palm with a single erect stem.
species, debris often accumulates in conspicuous raised piles at the bases of palms
(Figure 4.17). This feature almost certainly explains the positive correlation between
palm density and soil humus depth observed in the present study.
The palm community of upper slopes and ridge crests at St. Eugène, hence that which
coincided with termite transect sampling, is dominated by the arborescent species
Jessenia bataua (Martius), and the understory species Bactris oligocarpa Barbosa
Figure 4.17. Understory palm with funnel-like crown o f large leaves showing conspicuous pile o f accumulated leaf-litter around the base.
1996). The last two species are known to accumulate canopy leaf-litter (de Granville
1977). Their structural form, their potential as a food source, and their associated
accumulations of canopy litter, make palms important habitats for a wide diversity of
vertebrate and invertebrate forest fauna. The present study demonstrates the
importance of understory forest palms and their associated humus accumulations as
key micro-sites for termites. Moreover, this is the first study to have shown the direct
levels in Amazonian tropical forests. In addition to those found on the above-ground
structures of understory palms, suspended litter and humus accumulations also occur
high in forest canopies, especially in association with epiphytes. However, we predict
that termite assemblages associated with canopy soils will share few species in
common with (or be a depauperate subset of) those of understory palms, due to the
much harsher micro-climatic conditions of the canopy. For a particular subset of the
termite assemblage, understory palms represent suspended humus and litter
accumulations that are well-drained, and yet in a microclimatically favourable stratum
of the forest, making them ideal sites for foraging and nesting that are not as
abundantly replicated by other components of above-ground forest architecture.
The results of ordinations analyses revealed that the primary influence of palm
density and associated environmental gradients was upon soil feeders rather than wood
feeders. This is partly due to the rich accumulations of humus, both suspended and at
ground-level, that are associated with many palms. However, there was a lower
diversity of soil feeding termites associated with high, as opposed to low, palm
density, suggesting that below-ground soil conditions in the vicinity of palms are less
favourable to soil feeders. This effect is likely to be at least as important as the
positive effect of palms as termite microsites. The reason for it remains uncertain, but
may be linked to the lower soil pH associated with high palm densities. Termites
actively increase the pH inside their guts to facilitate digestion of food substrates
(Bignell and Eggleton 1995). Palms may themselves be directly contributing to the
acidity of their associated soils in some way, or the palm community overall may
thrive in more acid soils and hence achieve higher densities under such conditions.
and soil feeder diversity. The most likely reason for the taxonomic bias in species that
are associated with palms, and those that are not, concerns the level of humification of
food substrates. Most species of Apicotermitinae are feeding on more humified
substrates in the soil, than the majority of soil feeding Nasutitermitinae or Termitinae.
Forest fragmentation effects and the influence o f the ghost forest/canopy assemblage.
Species composition of wood and leaf-litter feeders was shown to be under greater
influence from fi’agmentation-associated environmental variables (volume of standing
dead wood and soil moisture) than from others, even after removing any influence of
ghost forest species. In most studies where invertebrate abundance and diversity have
been found to increase close to forest edges or in smaller fragments, the most plausible
explanation has been an increase in abundance of gap- or disturbance-adapted species
and/or species influx from surrounding matrix habitat (Didham 1997). However, in the
present study, this appeared not to be so and the nearest savannas or open, human-
impacted terrestrial areas were at least 50 km away. Instead, it appears that species
that were most abundant in the ghost forest and littoral zone were relatively rare at or
near ground-level in terrestrial forest sites and were probably a relict canopy-adapted
assemblage. Other invertebrate studies have shown that communities encountered near
ground-level at forest edges showed more in common with those of forest canopies
than of forest interiors (Toda 1992, Malcolm 1997). However, only one additional
species which was known to occur in the canopy, but was not known from the ghost
forest, was found on terrestrial transects. This soil feeding species, Anoplotermes
and edge distance, and was the only soil feeder to be found during limited casual
sampling from epiphyte-associated humus suspended high in mainland forest canopy
at the study site.
Species richness, similarity and evenness.
The results of species compositional similarity comparisons within and between
categories of fragments of different sizes (Figure 4.10) resembled those of a study on
beetle responses to tropical forest fragmentation at the BDFFP (Didham et a l 1998) in
which there was no indication that highly disturbed sites shared a common beetle
fauna. Moreover, in the present study, mean similarity among small island transects
was not only lower than mean similarity among mainland forest transects, but was also
lower than that between small island and mainland transects. Together with the finding
of insignificant differences in evenness across fragment size categories, the latter
result is in accordance with the conclusion that small islands have not been colonized
by generalist disturbance-adapted species from more open habitats elsewhere.
An alternative explanation for the observed similarities between small islands
is that transects were sampling from a moderately large pool of disturbance-adapted
species that was present within the original forest assemblage, although with
apparently rather low occurrence at or near ground-level in mainland forest.
Specifically, these disturbance-adapted species are likely to be canopy-, forest edge-,
or treefall-gap-adapted. The inherent patchiness of edge-, canopy- or treefall-gap-
associated termite food resources in a relatively undisturbed tropical forest, could be