Florida freshwater and tidal stream fish distribution mapping: Panhandle region

The United States is home to 10% of freshwater fish species, 30% of freshwater mussels, and 61% of freshwater crayfish that have been described worldwide.  However, this diversity is disappearing at an alarming rate (Sowa et al. 1994, Jelks et al. 2008). Modeling potential species habitat is an increasingly important tool for ecology, conservation, invasive species management, and climate change research (Pearson 2007).  I compared two commonly used presence-only species distribution modeling methods, Maxent and the Genetic Algorithm for Rule-Set Prediction (GARP), creating predictive species distribution models for 55 fish documented to occur in the freshwater streams of Florida's Panhandle.  I used GARP and Maxent to search for non-random associations between 20 environmental characteristics of fish presence locations versus those of the entire study region.  Overall, both modeling methods were able to predict the location of the validation occurrence points significantly better than random in both binomial tests of omission and receiver operating characteristic (ROC) analysis. The area under the ROC curve (AUC) were almost always higher for Maxent; however, sensitivity scores were almost always higher for GARP.  These results can be used as a guide for habitat modeling for freshwater species.