Estimation of Habitat Quality and Summer Population Density for Muskrats on a Watershed Basis

Abstract

A model for assessing habitat quality for muskrats (Ondatra zibethicus) was field tested in central Massachusetts during 1978-79 and in northcentral Pennsylvania during 1981-82. The habitat model predicted favorable habitat and ranked watersheds with respect to potential muskrat abundance using topographic maps. Local population attributes must be combined with habitat analyses to obtain reliable quantitative estimates of muskrat numbers. Summer density estimates of 23 muskrats/km and 48 muskrats/km were obtained for an unglaciated and a glaciated river, respectively. J. WILDL. MANAGE. 50(2):269-273 To fill the need for an economically feasible method of inventorying riverine muskrats, we developed a multivariate model of habitat selection based on principal components regression (PCR) with data from 5 watersheds in Massachusetts (Brooks 1980, Brooks and Dodge 1981, Allen and Hoffman 1984). When applied to 3 additional rivers in Massachusetts this model successfully identified favorable muskrat habitat. However, when quantitative predictions from the model were compared to actual censuses, the numbers of burrows were underestimated ( +? SE = 55 ? 4%, range = 32-71%). A curvilinear equation was developed to correct for this underestimation (Brooks and Dodge 1981). The importance of validating habitat models before implementation has been discussed by many authors (see Lancia et al. 1982). Therefore, the objective of this study was to determine the reliability of the corrected habitat model for qualitative assessments of muskrat habitat and quantitative estimates of muskrat abundance. Appreciation is extended to L. Bors-Koefoed for his assistance in the field and to S. D. Miller, J. M. Payne, D. R. Progulske, A. J. Steiner, and R. H. Yahner for earlier reviews of the manuscript. This study was cooperatively supported by the Mass. Coop. Wildl. Res. Unit (U.S. Fish and Wildl. Serv., Mass. Div. Fish. and Wildl., Univ. Massachusetts, Amherst, and Wildl. Manage. Inst.), Pennsylvania State Univ. (DuBois Campus and School For. Resour.), and SigmaXi. This is J. Ser. 6963 of the Pa. Agric. Exp. Stn. STUDY AREA AND METHODS The habitat model was developed on data collected on 8 watersheds in Massachusetts (Brooks 1980). The main study area, the Ware River, is in the central plateau region and flows southwest about 90 km through upland forests and wooded wetlands, agricultural land and pasture, and mill towns into the Connecticut River Basin. Fifty 300-m sections were selected randomly on the Ware River from May to October 1978 for study. Supplemental data were collected from 24 sections that were chosen randomly on 4 other rivers during July-August 1979: the Quaboag and Chicopee rivers of the Chicopee Basin, the Manhan of the Connecticut River Basin, and the Nashua which flows north to the Merrimack River Basin. A model of habitat selection by muskrats was developed from data collected on these 5 rivers. To refine the predictive capability of the habitat model, we studied 16 sections on 3 additional rivers in September 1979: the north-flowing ConcordAssabet system of the Merrimack River Basin and the Massachusetts portion of the southflowing Housatonic River (Brooks and Dodge 1981). The adjusted habitat model was tested on data from Sandy Lick Creek in northcentral Pennsylvania during the summers of 1981 and 1982. Sandy Lick Creek flows 48 km from the eastern continental divide west to the Ohio River Basin. Information on hydrology, soils, vegetation, land use, bank structure and composition, burrow and house abundance, and muskrat signs was