Determining spider species richness in fragmented coastal dune habitats by extrapolation and estimation (Araneae)

In this contribution we review which species richness estimators can be used if spiders are sampled with pitfall traps or other relative sampling methods. Due to the inherent bias typical for activity-based traps, only sample-based estimators were useful (Chao2, ICE and two jackknifing estimates). We estimated species richness for spiders from our four fragmented coastal dune habitats: dense and short grasslands, moss dunes and marram dunes. As extrapolation of the collector curve is only appropiate if the curve reaches a ceiling (asymptotic, such as parabolic and hyperbolic models), total species richness could not be determined by extrapolation in either of the investigated habitats. Because of the log-linear nature of the curves, non-parametric methods were applied. An absolute estimate of total habitat species richness was thus difficult to calculate; differences in total regional species richness could be analysed by comparing the different non-parametric estimator curves. If only habitat specific species were taken into account (we analysed this for grey dune habitats), extrapolation of the collector curve was appropiate (hyperbolic models). A relatively low number of traps were already enough for sampling more than 95% of the specific species. Estimates of the richness of specific species from grey dune patches (in contrast to species richness of the habitat) could already be derived from five traps. Our data revealed that grassland habitats were characterised by higher spider richness than moss-dominated grey dunes and that the latter were more diverse than Marram dunes. The number of specific species was slightly, although non-significantly, larger in dune grasslands than in moss-dominated dunes.