A hierarchical generalized linear model with variable selection: studying the response of a representative fish assemblage for large European rivers in a multi-pressure context

摘要

Global change calls for an understanding of how temperature and flow regimes influence aquatic ecosystems. Fish assemblages are a major component of river ecosystems and are thought to exhibit more integrative informative responses than single species to environmental variations, whether rare and sudden or gradual and continuous. The use of long-term datasets is thus of primary importance, allied to statistical modeling. For each of three previously identified species clusters, we performed Bayesian variable selection and inference within a hierarchical log Poisson Generalized Linear Model using a spike and slab normal prior to pinpoint which subset of environmental variables is of importance for each fish assemblage. Fish counts from electrofishing experiments are known to provide overdispersed data and, not surprisingly, the contribution of recorded environmental effects is found to be weak compared with those of other intra-assemblage sources of variation. The posterior distribution of the regression parameters is in coherence with what was expected from biological knowledge of the three species clusters. In particular, thermophilic species tend to benefit from warmer waters, whereas the recruitment of cold water species decreases due to global warming effects. Our study provides an example of the advantages of hierarchical modeling for quantifying interspecies ecological effects and selecting common environmental variables of importance.