ABSTRACT The spatial distribution of permafrost in mountainous regions is influenced by various factors such as topography, climate, vegetation, and substrate. Despite the existence of comprehensive permafrost maps at national and global levels, they fail to accurately represent the patch‐scale (e.g., < 25 m) permafrost distribution in characteristic landforms of alpine zones, such as talus slopes, moraines, and rock glaciers. This study aims to improve the understanding of permafrost distribution in these environments, focusing on patch‐scale variability and the influence of sediment size. By using data‐driven techniques (i.e., logistic regression, support vector machines, and random forests), the spatial distribution of permafrost in six alpine basins within the Canadian Rockies was examined. The results indicate that the enhanced vegetation index, sediment size, and slope angle are the most important variables for predicting permafrost at the patch scale. However, the influence of the predicting variables strongly varied across different sites. Although the models trained with data from all sites effectively capture site‐specific features and provide accurate representations of permafrost distribution, their applicability is limited in areas where predictor values fall outside the training domain. Nevertheless, this study contributes to understanding the factors influencing permafrost distribution at different scales, emphasizing the importance of including sediment size in predictive models.
Zegers et al. (Wed,) studied this question.