Predicting fish community properties within estuaries: Influence of habitat type and other environmental features

Abstract

Statistical models predicting species distributions are essential not only to increase knowledge on species but for their application in conservation and ecologically-based management. The variation of fish species richness and abundance in the most representative habitats (saltmarsh, mudflat and subtidal) in five estuaries along the Portuguese coast was analysed through seasonal sampling surveys in 2009. Generalized additive models (GAM) were developed to describe the variation of species richness and abundances with a set of geomorphologic, hydrologic and environmental characteristics from the sampled estuaries and habitats. GAM were chosen as the complex interactions dominating these ecosystems and species distribution are non-linear. Final models built for each estuary and for all estuaries together performed well during the calibration phase and also during the validation phase, where an unused data sub-set from each estuary was used. There was not a similar combination of variables retained by the models for the studied estuaries but factors such as the area of the habitat, the distance to estuary mouth, percentage of mud in the sediment and depth were commonly retained. The partial effect of these predictor variables on the variation of species richness and abundance in the estuaries varied markedly and the importance of preserving the heterogeneity of habitats within estuaries was highlighted. Models for each individual estuary performed better than models for estuaries combined. Predictive models could be useful as a preliminary tool to prepare long-term conservation plans at different scales.