Modelling the presence and identifying the determinant factors of dominant macroinvertebrate taxa in a karst river.

Modelling the macroinvertebrate community is important for evaluating the status of aquatic ecosystem health. Alternative to physical-based approaches, this study proposed two data-driven methods, support vector machine (SVM) and artificial neural network (ANN), to model the presence of macroinvertebrate species in rivers based on abiotic features. A famous karst river, Lijiang River, in Southwest China was selected as the study case. A total of 300 records containing data on 11 physicochemical parameters were collected from the upstream, midstream and downstream reaches of the river over a 2-year period (2009-2010) and were used for model construction and verification. Ten dominant macroinvertebrate taxa in the study area were modelled. In addition, the performance of the two methods was compared, and the relative importance of the independent variables was identified. The obtained results mapped abiotic factors to the species presence and could be used in combination with a two-dimensional hydro-environmental model to assess the impacts of flow regulation on macroinvertebrate dynamics. Furthermore, the SVM model performed slightly better than the ANN model in the studied case.