Multivariate geospatial detection and quantification of ecological thresholds corresponding to dispersion of Castor canadensis Kuhl. (North American beaver) in central Massachusetts
Date
2013
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of Delaware
Abstract
Beaver (Castor canadensis Kul.) populations have increased greatly since the 1950s in Massachusetts, thereby amplifying beaver-human conflict. As such, it is important to understand the dispersal of beavers in relation to habitat characteristics. This is the main goal of this thesis. An annual census dataset of beaver habitat use since 1968 permitted two ecological models – a multiple logistic regression and a discriminant analysis – to be empirically developed and tested using multivariate statistical techniques within a geographical information system (GIS) to spatially evaluate ecological thresholds and interrelationships of beavers and their aquatic habitats within the Quabbin Reservation in Central Massachusetts. The beaver population has held relatively steady at the study area’s carrying capacity since approximately 1988. Input parameters for both methods include a variety of geomorphic and vegetative habitat variables that modulate the physical processes and biological responses of beavers, which subsequently influence the species’ geographic distribution and dispersal behavior. The multiple logistic regression (MLR) model produced probability maps depicting the likelihood of beaver occurrence over a continuous landscape, successfully predicting 72.2% of the habitats colonized by beavers in 2012 within the study area. Alternatively, the discriminant analysis (DA) model assigned individual habitats throughout the study area into three habitat quality groups (non-preferred, occasional, and preferred), successfully classifying 90% of the plots surveyed into their respective habitat quality groups. Results of both ecological models suggest that gradual stream slopes, areas that are far away from the reservoir shoreline, and the presence of woody shrub species are the most influential environmental variables in identifying preferred beaver habitat. Moreover, the MLR model found that sustainable beaver colonization, over the course of several decades, usually occurs in areas with favorable geomorphic attributes (e.g. gradual stream slopes, large areas of upstream watershed), and exhibit a distinct vegetative signature of disproportionately higher coniferous tree cover. The MLR and discriminant function models are most applicable for any riparian habitats which share similar geomorphology and hydrology characteristics and harbor beaver populations at or near carrying capacity. These two models can ultimately be made readily available to land planners and managers, through the medium of a GIS, as tools to help guide and modify beaver management strategies to be proactive and adaptive to the everevolving land-use dynamics of beavers.