Accuracy Assessment Of The Aquatic GAP Fish Distribution Models For The Upper Missouri River Basin
Steve E. Freeling, Department of Wildlife and Fisheries, South Dakota State University, Brookings, SD 57007; (605) 688-4787; FAX (605) 688-4515; sfreeling@hotmail.com
Ryan M. Sylvester, Dept. of Wildlife and Fisheries, South Dakota State University, Brookings, SD 57007
Steven S. Wall, Dept. of Wildlife and Fisheries, South Dakota State University, Brookings, SD 57007
Charles R. Berry Jr., US Geological Survey, South Dakota Cooperative Fish and Wildlife Research Unit, Brookings, SD 57007
Aquatic GAP fish distribution models were developed to predict presence and absence of fish species in the upper Missouri River basin. Basins that contain thousands of stream kilometers and diverse habitat make fish sampling very time consuming and expensive. Modeling objectively predicts fish distributions over a large spatial area. GIS layers used in the modeling process are based on ten abiotic habitat characteristics of the lotic environment (e.g. elevation, temperature, stream size). The purpose of this poster is to describe model accuracy using empirical data. One watershed was sampled from each state or providence in the upper Missouri River basin: Beaver (North Dakota), Elm (South Dakota), Nowood (Wyoming), Sweet Grass (Montana), and Frenchman (Saskatchewan). A total of 19,556 fish consisting of 41 species were identified and added to deficient ichthyofaunal databases. In the case of the Sweet Grass watershed, our data was the first to be documented. Models were developed using a decision tree program, CHAID, Chi-squared Automatic Interaction Detector. Accuracy was evaluated with Cohen's Kappa by calculating the correct presences, correct absences, omission errors, and commission errors for fish species in each watershed. Preliminary results show that the models are 30 to 40 percent more accurate than random chance. Modeling is an objective method to predict species presence or absence, but the validity of the predictions must be understood to accurately determine gaps between protected river reaches and locations of rare species needing protection.