Enhancing mountainous permafrost mapping by leveraging rock glacier inventory

Permafrost is a key component of the cryosphere, which plays significant roles in surface energy, hydrological, and biogeochemical processes. Moreover, permafrost, a sensitive indicator of climate change, has experienced widespread degradation in recent decades. The Tibetan Plateau, hosting the largest mid-low latitude permafrost area, is particularly susceptible to these changes, warranting a deeper understanding of permafrost distribution and environmental interactions. However, permafrost mapping traditionally relies on empirical and physical models, each with its set of advantages and drawbacks. Empirical models, while user-friendly, introduce uncertainties due to data quality and scale issues. On the other hand, physical models, offering precision, demand high-quality data and face challenges in extensive simulations over large areas. With the advancement of artificial intelligence technologies, machine learning has rapidly formed many implementation algorithms and been applied in different fields. Permafrost mapping has been investigated with a variety of machine learning algorithms (i.e., neural networks, support vector machines, random forest, and gradient boosting), and demonstrated superior accuracy over traditional methods when applied to large areas, especially when there are abundant training data available.Despite these advancements, challenges persist, notably in mountainous areas characterized by scarce in situ data and complex topography. This study proposes a novel approach involving rock glaciers as valuable indicators for permafrost mapping. Intact and relict rock glaciers, representing the presence or absence of permafrost, offer crucial insights, particularly in mountainous regions where traditional methods fall short. The study focuses on the Qilian Mountains, a representative mountainous area on the Tibetan Plateau. Leveraging machine learning and rock glaciers, the research aims to simulate the Permafrost Zonation Index (PZI). Rigorous accuracy evaluations and comparisons with existing permafrost maps are conducted, promising a nuanced understanding of permafrost dynamics in this challenging terrain. The integration of technological advancements and innovative approaches holds the potential not only to advance permafrost research but also to inform conservation strategies and climate change assessments on a broader scale.