ABSTRACT The Nacho Nyäk Tagé (Stewart River) watershed in central Yukon (Canada) is characterized by discontinuous permafrost that is locally highly sensitive to thaw. This study aims to map the spatial distribution of permafrost terrain disturbances (PTDs) in the watershed and model thaw susceptibility to support community‐led land‐use planning by the First Nation of Nacho Nyäk Dun. A total of 246 PTDs, including active and stabilized retrogressive thaw slumps (RTSs) and minor thermoerosional features, were manually delineated from satellite imagery. PTD susceptibility was modeled using random forest machine learning, based on predictor variables reflecting the underlying topography, permafrost probability, and glacial history. The resulting susceptibility maps identified areas with high permafrost sensitivity. RTS susceptibility is highest along northwest‐ to northeast‐facing slopes and riverbanks in east–west striking valleys within the limits of the McConnell glaciation. Field visits to riverbank RTSs revealed the importance of fine‐grained and ice‐rich glaciolacustrine deposits for RTS occurrence in the study area. Model validation indicated strong predictive performance (average AUROC across tenfold cross‐validation = 0.93) of the RTS susceptibility model. We detected a significant positive trend in the number of average annual RTS initiations over the past four decades that is consistent with intensification trends observed across other permafrost landscapes. These findings provide valuable insights into the sensitivity of discontinuous permafrost in the Yukon Plateau. Areas of moderate to very high RTS susceptibility are predicted to respond strongly to anthropogenic disturbances, climate warming, changes in precipitation patterns, and wildfires.
Brieger et al. (Mon,) studied this question.