Testing Three Species Distribution Modelling Strategies to Define Fish Assemblage Reference Conditions for Stream Bioassessment and Related Applications

摘要: Species distribution models are widely used for stream bioassessment, estimating changes in habitat suitability and identifying conservation priorities. We tested the accuracy of three modelling strategies (single species ensemble, multi-species response community classification models) to predict fish assemblages at reference segments coastal subtropical Australia. aimed evaluate each strategy consistency predictor variable selection; determine which is most suitable bioassessment using indicators; appraise best match other management applications. Five models, one single two were calibrated presence-absence data from 103 sites. Models evaluated generality transferability through space time four external site datasets. Elevation catchment slope consistently identified as key correlates assemblage composition among models. The had high omission error rates contributed fewer taxa ‘expected’ component taxonomic completeness (O/E50) index than strategies. This potentially decreases model sensitivity impact assessment. ensemble accurately precisely modelled O/E50 training data, but produced biased predictions afforded relatively precision coupled with low bias across datasets lower They inherently included rare, predictable while excluding that poorly all suggest suited freshwater our study area. At level, exhibited without reductions specificity, relative this well non-bioassessment applications, e.g., priority areas conservation.