Application of non-linear quantile regression to macrozoobenthic species distribution modelling: Comparing two contrasting basins

The occurrence and distribution of macrozoobenthos in estuaries are strongly related to sediment grain-size characteristics. However, statistical prediction of the distribution of benthic populations as a response to a single environmental gradient has proven to be difficult, because the focal variable may set upper limits to the abundance, but other (partly uncorrelated) variables may cause considerable deviation from the maximum. A multi-quantile regression approach is better suited to characterize biota-environment relationships than a single (average or boundary) estimation, because it shows the variation in responses and quantifies the relative importance of other unmeasured factors. Here, a univariate application of non-linear quantile regression is proposed to account for heteroskedasticity and non-linearity in the biological response to sediment grain size. The analysis was applied to a large macrozoobenthic dataset from the SW Delta area (The Netherlands) to compare the relationships between sediment granulometry and macrozoobenthos in 2 neighboring but differing temperate coastal ecosystems (Oosterschelde and Westerschelde). Preference of individual species for grain size was consistent between both systems, although in general, a slightly higher median grain size (ca. +60% in grain diameter) was preferred in the Oosterschelde than in the Westerschelde. The major difference in the community was, however, that mud-preferring species dominated the assemblage in the Westerschelde, and sand-preferring species dominated the Oosterschelde. Although the prevalence of muddy and sandy sediments in both systems is similar, in the Westerschelde, strong hydrodynamic stress is correlated with sandy habitats, causing impoverishment of assemblages at sandy sites. In the Oosterschelde, sandy sediments are usually associated with much more benign conditions and have the richest species assemblage.