Confronting existing knowledge on ecological preferences of stream macroinvertebrates with independent biomonitoring data using a Bayesian multi-species distribution model

• Published in: Freshwater science Vol. 40; no. 1; pp. 202 - 220
• Main Authors: Vermeiren, Peter, Reichert, Peter, Graf, Wolfram, Leitner, Patrick, Schmidt-Kloiber, Astrid, Schuwirth, Nele
• Format: Journal Article
• Language: English
• Published: Lawrence The University of Chicago Press 01.03.2021; University of Chicago Press
• Subjects: Anthropogenic factors; Bayesian analysis; Benthos; Bioindicators; Biomonitoring; Conditional probability; Data; Datasets; Distribution; Ecological effects; Environmental factors; Flow velocity; Fluvial morphology; Geographical distribution; Human influences; Identification; Indicator organisms; Indicator species; Insecticides; Knowledge; Knowledge bases (artificial intelligence); Macroinvertebrates; Morphology; Organic matter; Phylogeny; Preferences; Probability theory; Resolution; Rivers; Statistical inference; Stream pollution; Substrata; Taxa; Taxonomy; Temperature; Temperature preferences; Velocity; Zoobenthos; ecological niches; biomonitoring; Bayesian inference; habitat suitability; taxonomic resolution
• DOI: 10.1086/713175
• ISSN: 2161-9549

A wide knowledge base regarding the ecological preferences of benthic macroinvertebrates is synthesized in public databases. This knowledge can assist in disentangling the influence of multiple environmental factors on the probability of occurrence of macroinvertebrates and in identifying anthropogenic impacts on the macroinvertebrate assemblage. We aimed to examine and extend current knowledge on ecological preferences by confronting it with independent biomonitoring datasets and to assess how the taxonomic resolution of datasets and the prevalence of taxa affects our ability to do so. We used a habitat suitability-based multi-species distribution model (HS-MSDM) and applied Bayesian inference to confront current knowledge (formalized as prior probability distributions) against independent biomonitoring data across rivers in Switzerland. Shifts in the resulting posterior probability distributions relative to the priors indicate a disagreement with the current knowledge of ecological preferences. Ecological preferences for temperature and organic matter had the highest influence on the predicted occurrence of macroinvertebrates in the model, followed by flow velocity, insecticide pollution, and substratum. Three-fold cross-validation tests demonstrated that the HS-MSDM predicted the distribution of taxa with a relative frequency of occurrence between 0.2 and 0.8 considerably better than a model without consideration of environmental factors. However, it was less able to predict the distribution of taxa with a frequency of occurrence <0.1 or >0.9. Nine taxa with a frequency of occurrence between 0.4 and 0.8 were identified as potentially useful bioindicators, given their strong association with the environmental factors in the model. We also identified 29 taxa for which part of the ecological preference data, particularly temperature and flow-velocity preferences, should be re-examined. For river morphology, 18 sensitive and 10 insensitive taxa were identified, although direct and uniquely linked prior knowledge regarding morphology was lacking for all taxa. Phylogenetically derived information on ecological preferences could be integrated and updated to fill gaps in ecological preference databases. However, the taxonomic resolution of the biomonitoring and ecological preference data plays an important role, as we show by identifying families comprising species that respond differently to environmental factors. These results demonstrate the value of conducting biomonitoring at the most detailed taxonomic level possible.