Estimating and conserving patterns of invertebrate diversity: a test case of New Zealand land snails

Abstract

Aim Using New Zealand land snails as a case study, we evaluated recent spatial modelling approaches for the analysis of diversity in species-rich invertebrate groups. Applications and prospects for improved conservation assessment were investigated.Location New Zealand.Methods The study used a spatially extensive and taxonomically comprehensive, plot-based dataset on community structure in New Zealand land snails. Generalized regression analysis and spatial prediction (GRASP) was used to model and predict species richness as a function of environmental variables (including aspects of climate, soils and vegetation). Generalized dissimilarity modelling (GDM) was used to model turnover in species composition in relation to environmental and geographical distances, and to assess community similarity and the representativeness of the reserve network.Results Observed land snail richness in 20 · 20 m plots ranged from 1 to 74 (mean 17.5). The GRASP model explained a modest 27% of the variation in richness. The GDM model explained 57% of the variation in species turnover and indicated approximately equal amounts related to environmental (Cody's beta diversity) and geographical distance (Cody's gamma diversity). Temperature and moisture were the most important environmental variables. Results indicate that snail distributions are not only sorted by environment but are also strongly influenced by historical effects consistent with those expected of poorly dispersing taxa that have persisted in refugia during past climatic change. The GDM model enabled spatial classifications of snail communities, highlighting diverse communities in heterogeneous regions, such as the South Island mountains, and also enabled continuous depictions of community similarity and adequacy of New Zealand's protected natural areas network.Main conclusions The GRASP and GDM analyses allowed us to model and depict spatial patterns of diversity in land snail communities involving 845 species, and produce community classifications and estimates of community similarity. These tools advance conservation assessment in species-rich groups, but require further conceptual and methodological development.

Compositional patterns in terrestrial fauna and wetland flora and fauna across the Pilbara biogeographic region of Western Australia and the representativeness of its conservation reserve system

“…This predictive capacity of GDM provides a foundation for various subsequent spatial analyses such as an appraisal of the representativeness of conservation reserve networks (Ferrier et al 2004;Overton et al 2009;Thomassen et al 2011), estimating biodiversity loss (Allnutt et al 2008), survey gap analyses (Ferrier 2002;Funk et al 2005), conservation prioritisation (Arponen et al 2008;Thomassen et al 2010) and climate-change impacts (Fitzpatrick et al 2011;Mackey et al 2012;Prober et al 2012).…”

Using Generalised Dissimilarity Models and many small samples to improve the efficiency of regional and landscape scale invertebrate sampling

“…Each of these four predictors was selected in at least five of the seven models, and their large influence on the results was further confirmed by the high ranges on the partial responses graphs (Overton et al, 2009; Table 1). Mean annual temperature (parameter 1), minimum temperature of the coolest period (parameter 6), precipitation seasonality (parameter 15) and maximum temperature of the warmest period (parameter 5) had a large influence in one or two models, but were excluded altogether from four or five of the seven models.…”

“…GDM software for the R statistical software environment can be downloaded from http://www.biomaps.net.au/gdm/, although we used an alternative command line executable provided by Glenn Manion and Simon Ferrier. GDM software calculated the Bray-Curtis dissimilarities (d ij ) in species composition between sampled sites (in paired combinations of i and j), and then derived monotonically increasing functions (f k ) for each of p environmental factors using a Generalised Linear Model (GLM) and an exponential link function using the following equation (Ferrier et al, 2007;Overton et al, 2009):…”

“…Two sites with identical environmental conditions have a dissimilarity that is determined by the intercept b (Overton et al, 2009). We then used GDM software to estimate the dissimilarities between other geographic locations in the Pilbara (cells in a raster image) based on their environmental conditions, and used multi-dimensional scaling techniques to classify the study area into landscapes that were relatively homogeneous in environmental conditions and species composition (Ferrier et al, 2007).…”

“…GDM software automatically removed environmental factors that did not significantly affect the turnover in species composition. The influence of each factor was assessed based on the range of the partial response graph (Overton et al, 2009). …”