Predicting the Occurrence and Success of Walleye Populations from Physical and Biological Features of Northern Wisconsin Lakes

Abstract

We developed a linear discriminant function (LDF) using surface area, mean depth, fetch, and three substrate descriptors to discriminate the presence and absence of walleyes Sander vitreus (formerly Stizostedion vitreum) in 120 northern Wisconsin lakes. The resulting LDF correctly classified 90% of walleye lakes and 85% of nonwalleye lakes. For model validation, the model classified 90% of walleye lakes and 93% of nonwalleye lakes in an independent set of 100 randomly selected lakes. In contrast, within walleye lakes, an LDF using the same variables correctly classified only 60% of 30 self-sustaining walleye lakes and 50% of 30 stocked walleye lakes. For model validation, the model classified 30% of self-sustaining walleye lakes and 73% of stocked walleye lakes in an independent set of 40 randomly selected lakes. To further assess the differences between walleye lakes with self-sustaining and stocked populations, the angler catch rates for largemouth bass Micropterus salmoides, smallmouth bass M. dolomieu, northern pike Esox lucius, and muskellunge E. masquinongy obtained from creel surveys during 1990-2001 were analyzed by means of a principal components analysis. The first principal component explained 29% of the variation and represented a linear contrast between the relative abundance of walleyes and muskellunge and the relative abundance of largemouth bass and northern pike. High largemouth bass and northern pike densities characterized lakes with stocked walleye populations, whereas high walleye and muskellunge densities characterized lakes with self-sustaining walleye populations. We conclude that physical features may be used to discriminate lakes where walleyes presently occur from lakes where walleyes do not presently occur, and that fish community differences exist between stocked and self-sustaining walleye populations in northern Wisconsin.