2.3.2 Distribution of ant assemblages along vertical and horizontal gradients
To understand how ant richness and abundance changed vertically and horizontally, we used linear regression models. The abundance (ln( x+1) transformed) and richness of ants at each sampling point were used as response variables with vertical height as a continuous explanatory variable and horizontal position of vertical transects as a categorical explanatory variable.
To explore vertical and horizontal differences in assemblage composition, we used NMDS ordinations. These used both abundance-based (ln (x+1) transformed) assemblage data and species presence/absence data. We generated non-metric multi-dimensional scaling (NMDS) plots using Bray-Curtis distance index for the abundance-based community and Jaccard distance index for the presence/absence data of ant assemblages at each sampling point grouped by vertical level and transect. For all ant assemblage composition analyses analysis (NMDS and MRMs, see below), as results from measures using abundance-based and presence-absence based dissimilarity were similar, we only present results using abundance-based dissimilarity in the main text (see supplementary materials for results of presence-absence analyses). To increase the sample size within groups for the vertical analyses, and hence to increase statistical power, we assigned all sampling points within ten-metre bins to the same groups (i.e. 0-10 m, 10-20 m, etc). We tested for differences in assemblage composition between these groupings using PERMANOVA analyses (adonis function in the vegan package, 999 permutations) (Oksanen et.al. 2013, R Core Team, 2013).
To explore relationships between beta diversity and spatial distance we conducted multiple regressions on distance matrices (MRMs). We conducted two different analyses: one looking at beta diversity horizontally by summing assemblage data from entire transects across all heights, and a second looking at beta diversity vertically by summing assemblage data across all assemblages of the same height between different transects. We then calculated the pairwise assemblage dissimilarity across these summed data using the beta.pair.abund function (for abundance-based dissimilarity) and beta.pair function (for presence-absence data) in the betapart package (Baselga and Orme 2012). We then tested effects of horizontal/vertical distance on pairwise assemblage dissimilarity between transects/strata by conducting multiple regressions on these distance matrices (MRM function in R ecodist package) (Golsee and Urban 2007). To test whether degree of turnover differed horizontally and vertically, we conducted replicated MRM analysis using data from individual transects (to assess vertical turnover), and from individual heights (to assess horizontal turnover). We then took the slopes and intercepts from the fitted MRMs, and used linear models (function lm in R base package) to test for differences in these parameters horizontally and vertically. Slopes represent the strength of the distance-similarity decay relationship, with a more positive slope indicating a more rapid increase in dissimilarity. Intercepts represent the turnover at very small spatial scales (technically when distance = 0 m).