Human-Altered Tree-Based Habitats and Their Value in Conserving Bird and Bat Diversity in Northeast Luzon, The Philippines - Smallholder Tree Growing for Rural Development and Environmental Services - page 342

(bats: all captures per mist-net-line is one sample) for each habitat type. To com-
pensate for systematic differences in the mean number of individuals observed or
captured per sample, we rescaled sample-based species accumulation curves by
individuals (Gotelli and Colwell 2001). Species accumulation curves are asymp-
totic: with infinite sampling effort, species richness eventually no longer increases
(Diaz-Frances and Soberon 2005). The slope of smoothed species accumulation
curves thus provides information about the reliability of species richness assess-
ments, with flattening curves indicating most species are detected and unsaturated
curves indicating sampling effort is insufficient for a reliable species richness
appropriation. For birds, we calculated species accumulation curves for all resident
species and a subset of forest species. To compare bird densities between habitat
types we plotted smoothed individual accumulation curves versus sample size
(point counts), for all resident and forest species. For bats, we calculated species
accumulation curves for all species and a subset of fruit bats. We also calculated
individual accumulation curves here, plotted versus the number of mist-net lines.
In addition we calculated nine non-parametric species richness estimators
(EstimateS 7.5, Colwell 2005). These estimators are based on species-abundance rela-
tionships (ACE, Chao1), species-incidence relationships (ICE, Chao2, Jackknife1,
Jackknife2, Bootstrap) or extrapolation of asymptotic species accumulation curves
(MMMeans, MMRuns). Non-parametric species richness estimators are useful for
comparative purposes when sample sizes or sampling strategies differ, and are espe-
cially precise at small sampling unit size (Hortal et al. 2006). Because the perform-
ances of different species richness estimators vary between data and cases (Walter and
Moore 2005), we calculated the average of the nine estimators (following, e.g. Sodhi
et al. 2005; Posa and Sodhi 2006) to assess the completeness of our surveys and pro-
vide a comparison of patterns in species richness differentiation between habitat types.
For birds, we calculated the average of the nine estimators for all resident species and
for a subset of forest birds; for bats, we calculated the average for all species and a
subset of fruit bats.
We calculated Sørensen similarity indices between pairs of habitat types to
assess the proportion of shared species. The Sørensen similarity index is calculated
by first determining the number of shared species in two habitat types. This figure
is then multiplied by two and divided by the sum of all species in habitat type one
and all species in habitat type two.
Point counts, mist-net lines and habitat characterisation plots within one locality
are spatial pseudo-replicates, meaning these samples are not statistically independ-
ent with common locality factors affecting all variables measured in the locality.
The smallest independent experimental unit to which statistical analyses are to be
applied (Hurlbert 1984) is, in our study, the locality. Hence the averages for all
measured variables (point count results birds, mist-net results bats, habitat charac-
terisation variables) were calculated for each habitat type per locality. Tests for
homogeneity of variances (Levene test) and normal distributions (Shapiro-Wilk)
showed that not all data sets were normally distributed and had equal variances.
Hence, non-parametric two-related-samples (Wilcoxon signed rank) tests were
used to compare the locality averages of variable scores between habitat types.