Near-surface permafrost distribution mapping using logistic regression and remote sensing in interior Alaska

Abstract

A combination of binary logistic regression (BLR) and remote sensing techniques was used to generate a high-resolution spatially continuous near-surface (< 1.6 m) permafrost map. The BLR model was used to establish the relationship between vegetation type, aspect-slope, and permafrost presence; it predicted permafrost presence with an accuracy of 88%. Near-surface permafrost occupies 45% of the total vegetated area and 37% of the total study area. As less than 50% of the study area is underlain by near-surface permafrost, this distribution is characterized as "sporadic" for the study area.