Modeling the spatial distribution of fish communities at the watershed scale

The spatial distribution of fish species in streams is often used as a major index of the health of aquatic ecosystems, which reflects the location of suitable habitats within the watershed. This information is important for watershed ecological restoration, engineering system operation and water allocation in the spatial domain of river basins. This paper attempts to construct an ecological niche model by integrating genetic programming (GP) and logistic regression with various data sources, including 1) fish distribution, 2) water quality, 3) flow discharge and 4) geomorphology, The "Receiver Operating Characteristic" (ROC) curve and "Area Under the ROC Curve" (AUC) analysis were used to evaluate the performance of the niche model. The model with higher AUC scores will provide a more accurate presence probability of target species through the entire watershed. The locations with higher presence probability can be categorized as being associated with suitable habitats. Suitable habitats might migrate due to environmental changes cause by human or natural factors. The niche model can help researchers to determine the "presence probability" of target fish species under different environmental conditions. The Fox River in Illinois, USA was used as an example, two different species (with different spatial distribution patterns and densities): Quillback (Carpiodes cyprinis) and Largemouth Bass (Micropterus salmoides) were selected to demonstrate how the ecological niche model was constructed and how its performance was evaluated. The result of the model was displayed in maps, which provide the spatial information for possible fish communities' movement within the watershed context.