Permafrost distribution in northeast China based on boosted regression tree

Abstract:

The permafrost in northeastern China constitutes an important component of the Xing’an-(trans)Baikal permafrost, exhibiting characteristics of both high latitude and high altitude permafrost. The presence and dynamics of permafrost directly impact the ecological environment of the regional cold zone, water-carbon cycles, cold region engineering design, and operations. Currently, empirical and semi-empirical models based on thermal boundary conditions tend to overestimate the areal extent of permafrost and insufficiently consider environmental factors beyond temperature. To more accurately delineate the regional distribution of permafrost, this paper obtained the spatial variation characteristics of zonal and non-zonal factors influencing regional permafrost distribution through regional surveys and data integration and employed the boosted regression tree model for simulation and analysis. The results indicate that among the zonal factors, latitude, longitude, and altitude contribute 45.3%, 42.4%, and 12.3%, respectively. Among the non-zonal factors, temperature (including freezing and thawing indices), precipitation, soil-water conditions, snow cover, and vegetation contribute 46.4%, 18.9%, 13.1%, 12.5%, and 9.1%, respectively. This analysis clarifies the contributions of environmental factors to the development and dynamics of Xing’an-(trans)Baikal permafrost. Compared with a classification and regression decision tree, the boosted regression tree model achieves an accuracy of 0.91. This study provides data support and reference for regional permafrost research and related fields.