Three sets of linear regression models relating vegetation and hydrographic variables to observed beaver colony densities were developed for 3 study areas in Quebec where the effects of trapping on beaver densities were considered negligible. The first two study areas were located in mature northern hardwood forest. This research focused principally on the first study area, Papineau-Labelle Wildlife Reserve, where beaver colony density and habitat variables were studied at a 1:20,000 scale on 20 4-sq-km sample plots. Over 700 habitat variables were collected and generated from forest inventory maps, a computerized forest stand database, and base maps using a digitizer, microcomputer, mainframe and SAS programs. From the set of 32 models presented, the model explaining the most variation (95%) showed that beaver colony densities increased with density of stream length between 2% and 6% gradient and amount of wetland broadleaf cover forest area and decreased with forest stand area between 60 and 100 years of age. A set of decision rules was developed to help managers select the best model for determining beaver colony density according to availability, effort needed to measure and generalizability of the habitat information, and according to desired precision of the estimation. The model explaining the most variation (62%) in beaver colony numbers with the least measurement effort used total stream length measured from maps with the line intercept method. In the second study area, Gatineau Park, hydrographic and shoreline vegetation variables in 63 4-sq-km quadrats were examined. The model explaining the most variation (35%) in beaver colony numbers showed that beaver colony densities increased with density of stream length in the 1 % to 10% gradient class. The third study area was located in the boreal forest of the St. Lawrence North Shore area and included 49 trapping units in which the effects of trapping on colony densities was considered negligible. Beaver density on the traplines was related to % cover as portrayed on 1: 50,000 scale MER summary forest inventory maps. Of the set of 4 models retained, the model explaining the most variation (61 %) in beaver colony numbers showed that beaver colony densities increased with the area of young broadleaf cover and the burnt area. In the 2 mature northern hardwood study areas, hydrographic characteristics seemed to explain beaver colony density variations, whereas in the boreal forest where clearcutting and fires disturbed the forest, a significant portion of beaver colony density variation seemed to be explained by food abundance differences. The sets of models in Papineau-Labelle and the St. Lawrence North Shore are considered of potential use to furbearer managers in Quebec because they provide a relatively low cost means to determine potential beaver colony densities. / Master of Science