Aquatic invasive species pose major ecological and economic threats to waterways worldwide through the displacement of native species and alteration of hydrologic cycles. Modeling the potential spread of alien aquatics through spatially explicit mapping is an increasingly important tool for risk analysis and rapid response. Habitat modeling provides the opportunity for identifying the key environmental variables influencing species distributions. We compared three presence-only modeling methods to predict the potential US distributions of Channa argus, Cyprinella lutrensis, Neogobius melanstoma, and Hypophthalmichthys molitrix using maximum entropy (Maxent), the genetic algorithm for rule set prediction (GARP), and support vector machines (SVM). We used inventory records from the USGS nonindigenous aquatic species database and a geographic information system of 70 raster climatic and environmental variables to produce spatially explicit maps for each species.
Results/Conclusions
Maxent and SVM produced higher accuracy predictions than GARP. Aquifer permeability, baseflow index, elevation, and mean annual precipitation were the key variables influencing fish distribution patterns. Results from this study provide insights into which locations and environmental conditions may promote the future spread of invasive fish in the US.