The Nacho Nyäk Tagé (Stewart River) watershed in the traditional territory of the First Nation of Na-Cho Nyäk Dun in the central Yukon is characterized by variable topography, land cover, permafrost conditions, surficial geology, climatic history, and land use patterns. Consequently, increasing air temperatures, wildfires, and changing precipitation patterns have a range of impacts on the permafrost landscape. Permafrost terrain disturbances (PTDs) were mapped in satellite imagery to understand the current spatial distribution of thaw-induced geohazards in the watershed. PTDs are indicators of sensitive permafrost terrain that is likely to respond strongly to climate change. Particularly, retrogressive thaw slumps (RTSs) are indicators of ice-rich permafrost and have increased in frequency and activity across the watershed. Along the banks of Nacho Nyäk Tagé, they are typically associated with ice-rich glaciolacustrine sediments or tills, as validated by field observations.
Terrain susceptibility towards PTDs was modelled using random forest machine learning and revealed distinct spatial patterns related to physiography and climatic history. Different presence-absence training point sampling strategies and cross validation techniques are compared and discussed. Tenfold cross-validation indicates high accuracy of model predictions. Model results indicate that RTSs are predominantly found on gentle north-facing slopes and riverbanks in east-west oriented valleys that are within the limits of the McConnell glaciation (maximum extent ca. 18 ka).
Mapping the current extent of regional permafrost landscape change and modelling terrain susceptibility are useful data in support community-driven regional land use planning that incorporates a holistic assessment of climate change impacts on the environment.