![Sample our Geography journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5937/TOC-Subject-Sample-2021-Geography.jpg"})
ABSTRACT
Our understanding of permafrost distribution is still limited, particularly in mountainous areas where highly heterogeneous environments and a lack of reliable field data tend to prevail. The extensive distribution of rock glaciers in the Qilian Mountains, located in the northeastern Tibetan Plateau, offers the opportunity to develop a novel approach for permafrost mapping in mountainous regions. In this study, a total of 1,530 rock glacier records were combined with in situ data to drive machine learning models for estimating permafrost presence. Three machine learning algorithms were adopted, and their accuracies were assessed in both mountains and plains by comparing the mapped permafrost to reserved field data as well as other published permafrost datasets. Among the algorithms tested, the CatBoost model presented the highest accuracy, with an overall accuracy of 83.3%. The model was thus chosen to produce a 250-m resolution permafrost zonation index (PZI) map, which identified a total area of 73.1 × 103 km2 permafrost in the Qilian Mountains, accounting for 39.1% of the area. The map also presented higher accuracy than other published permafrost maps. This study demonstrated that rock glacier records coupled with gradient-boosting machine-learning algorithms can help improve permafrost mapping, especially in the most challenging mountainous permafrost areas.
Acknowledgments
The authors thank the reviewers for the helpful comments.
Data availability
The PZI of Qilian Mountains can be obtained from https://doi.org/10.11888/Cryos.tpdc.300667.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Correction Statement
This article has been corrected with minor changes. These changes do not impact the academic content of the article.