Accurate soil thermal conductivity (STC) data and their spatiotemporal variability are critical for the accurate simulation of future changes in Arctic permafrost. However, in-situ measured STC data remain scarce in the Arctic permafrost region, and the STC parameterization schemes commonly used in current land surface process models (LSMs) fail to meet the actual needs of accurate simulation of hydrothermal processes in permafrost, leading to considerable errors in the simulation results of Arctic permafrost. This study used the XGBoost method to simulate the spatial-temporal variability of the STC in the upper 5 cm active layer of Arctic permafrost during thawing and freezing periods from 1980 to 2020. The findings indicated STC variations between the thawing and freezing periods across different years, with values ranging from-0.4 to 0.28 W & sdot;m-1 & sdot;K-1. The mean STC during the freezing period was higher than that during the thawing period. Tundra, forest, and barren land exhibited the greatest sensitivity of STC to freeze-thaw transitions. This is the first study to explore the long-term spatiotemporal variations of STC in Arctic permafrost, and these findings and datasets can provide useful support for future research on Arctic permafrost evolution simulations.

来源平台：COLD REGIONS SCIENCE AND TECHNOLOGY