Spatially based methods to assess the ecological status of riverine fish assemblages in European ecoregions

The objective was to develop spatially based (type-specific) methods to assess the ecological status of European rivers according to the EU Water Framework Directive. Some 15 000 samples from about 8000 sites were pre-classified within a five-tiered classification system based on hydromorphological and physico-chemical pressures. The pre-classification was used to identify reference conditions and to calibrate the assessment methods. Clustering reference sites based on relative species composition resulted in 60 fish assemblage types within 11 of the ecoregions under study. Discriminant function analyses (DFAs) were employed to identify environmental parameters characterising fish assemblage types; altitude, river slope, wetted width, mean air temperature and distance from source were the principal predictors. These environmental parameters were used to assign impacted sites with altered fish assemblage composition to the reference fish assemblage type. Metrics (fish assemblage descriptors) responding to human pressures were selected based on correlation and DFAs. Assessment methods were developed for 43 fish assemblage types. Metrics based on individual sentinel species were more often used in type-specific methods than metrics related to reproduction, habitat and feeding. Metrics based on long-distance migrants and potamodromous species were more sensitive to human pressures than overall composition metrics, e.g. total number of species. Only some of the tested metrics showed pressure-specific responses, i.e. reacted to one type of pressure but not to others. Insectivorous, intolerant and lithophilic species exclusively responded (decreased) to chemical and hydromorphological pressures in 14¿19%. Omnivorous species was the only metric type that showed a consistent reaction (increase) to continuum disruptions in 25% of the cases. Accuracy of methods based on cross-validation with pre-classification varied between 47% and 98% (mean 81%) when contrasting calibration data set (class 1 and 2) with degraded sites (class 3, 4 and 5)