Predicting permafrost probability in a variable boreal environment utilizing a multiple logistic regression model, Whatì, NT, Canada
University of Lethbridge. Faculty of Arts and Science
Lethbridge, Alta. : University of Lethbridge, Dept. of Geography & Environment
For remote communities, access to permafrost information for hazard assessment is a considerable challenge. This study applies analytical methods illustrating a time- and cost-efficient method for conducting community-scale permafrost mapping in the community of Whatì, NT. A binary logistic regression model was created using a combination of field data, digital elevation model-derived variables and remotely sensed products. Independent variables included categorical inputs such as vegetation, topographic position index and elevation breaks. The dependent variable is sourced from 139 physical checks of permafrost presence/absence. Vegetation was shown to be the strongest predictor of permafrost. The model predicts 50.0 % of the vegetated area is underlain by permafrost with a model accuracy of 91.4 % and spatial agreement of 72.8 % when compared to ground-truth pits. Compared to existing permafrost products this value is on the lowest edge of Whatì’s current classification (extensive discontinuous) illustrating there could be less permafrost than presumed.
Forest fire forecasting -- Northwest Territories , Logistic regression analysis , Permafrost -- Northwest Territories , Permafrost -- Remote sensing , Taigas -- Ecology , Taigas -- Northwest Territories , Dissertations, Academic