Modeling the spatial distribution of subarctic forest in northern Manitoba using GIS-based terrain and climate data

Northern forested ecosystems are predicted to change dramatically in response to climate change during the next century. The purpose of this paper was to use logistic regression analysis to model the effects of climate and topography on the spatial distribution of the northern forest around Churchill, Manitoba, Canada. Climate maps were modeled using kriging interpolation of actual climate data collected from 34 long-term monitoring sites distributed throughout the study area, and topographic information was derived from commercially available digital elevation models. Five of the 18 independent variables contributed appreciably (p<0.15) to the final logistic regression model: distance from the Hudson Bay coast, summer soil temperature, snow density, slope, and snow water equivalent. Current forest distribution was predicted with 66% accuracy using the final model, and Kappa statistics indicated significant agreement between modeled and actual forest extents. Significant explanatory variables demonstrate important synergistic effects of Hudson Bay, wind, and snow in determining forest distribution. Modeled forest extents were further south than actual forest limits, which suggest that the treeline is not likely in equilibrium with the present climate.