A priori typology-based prediction of benthic macroinvertebrate fauna for ecological classification of rivers.

We evaluated a simple bioassessment method based on a priori river typology to predict benthic macroinvertebrate fauna in riffle sites of rivers in the absence of human influence. Our approach predicted taxon lists specific to four river types differing in catchment area with a method analogous to the site-specific RIVPACS-type models. The reference sites grouped in accordance with their type in NMS ordination, indicating that the typology efficiently accounted for natural variation in macroinvertebrate assemblages. Compared with a null model, typology greatly increased the precision of prediction and sensitivity to detect human impairment and strengthened the correlation of the ratio of observed-to-expected number of predicted taxa (O/E) with the measured stressor variables. The performance of the typology-based approach was equal to that of a RIVPACS-type predictive model that we developed. Exclusion of rarest taxa with low occurrence probabilities improved the performance of both approaches by all criteria. With an increasing inclusion threshold of occurrence probability, especially the predictive model sensitivity first increased but then decreased. Many common taxa with intermediate type-specific occurrence probabilities were consistently missing from impacted sites, a result suggesting that these taxa may be especially important in detecting human disturbances. We conclude that if a typology-based approach such as that suggested by the European Union's Water Framework Directive is required, the O/E ratio of type-specific taxa can be a useful metric for assessment of the status of riffle macroinvertebrate communities. Successful application of the approach, however, requires biologically meaningful river types with a sufficient pool of reference sites for each type.