M-WH-6
Predictive Modeling of Muskellunge Spawning Habitat in US Waters of the Upper Niagara River

Monday, September 9, 2013: 3:20 PM
White Oak (The Marriott Little Rock)
Derek Crane , Environmental and Forest Biology, State University of New York - College of Environmental Science and Forestry, Syracuse, NY
John M. Farrell , Environmental and Forest Biology, State University of New York - College of Environmental Science and Forestry, Syracuse, NY
Kevin Kapuscinski , Environmental and Forest Biology, State University of New York - College of Environmental Science and Forestry, Syracuse, NY
Conserving and restoring Muskellunge spawning habitat is essential for maintaining self-sustaining populations.  We developed a Maxent model based on presence-only data to investigate the relationship between habitat features and the occurrence of spawning Muskellunge in the upper Niagara River.  Muskellunge spawning points (n = 11) were determined by direct observation of spawning pairs.  Model inputs were based on microhabitat habitat features collected at each spawning point and a sample of (n = 120) background habitat points.  The full model was pruned to a three variable model to remove uninformative variables and limit overfitting and redundancy.  Training gain and a jacknife procedure developed for Maxent modeling were used to evaluate model performance.  The percent rank of total aquatic macrophyte/algae coverage, sediment class, and dominant aquatic macrophyte/algae taxa at muskellunge spawning points were the most informative habitat variables and retained in the final model.  The model demonstrated good predictive performance, and was able to assign presence to test samples in 7 of 11 jacknife model replicates.  Mean test gain (0.67; SD = 1.93) of the jackknife replicated models indicated that the average likelihood of Muskellunge spawning points was nearly two times greater than that of a randomly selected background point.  The jackknife test of predictive ability indicated that predictive performance was significantly better than random (n = 11; P = 0.0056).  Results from this research provide scientists and managers with a better understanding of Muskellunge reproductive ecology and information to guide habitat conservation and restoration in the upper Niagara River.