Remote sensing spatiotemporal patterns of frozen soil and the environmental controls over the Tibetan Plateau during 2002-2016

The changing climate is affecting the frozen soil at an unprecedented rate across the Northern Hemisphere. However, due to sparse ground measurements, the changes of frozen soil and the environmental controls over the vast cryosphere are still unclear, such as in the Tibetan Plateau (TP). In this study, a process-based model solely driven by satellite remote sensing data is employed to investigate the spatiotemporal changes of seasonally frozen ground and permafrost over the entire TP (similar to 3 million km(2)) during 2002-2016 at a spatial resolution of 1 km. Comprehensive validations against in situ observations of frozen ground types, mean annual ground temperature, active layer thickness, soil temperature, and frozen depth at 608 boreholes and 109 meteorological stations demonstrate an overall satisfactory performance of the model in reproducing the spatial pattern and temporal evolution of the frozen soil in the region. Spatially, land surface temperature (LST; both in-season and off-season) primarily controls the frozen ground types and frozen depth, with seasonally frozen ground and permafrost covering similar to 56% and similar to 37% of the plateau, respectively. The estimated spatial-averaged annual maximum soil freeze depth (SFD) is similar to 1.29 m, and the annual maximum active layer thickness (ALT) of permafrost is similar to 1.85 m. Temporally, ALT shows an overall increasing trend at an average rate of +3.17 cm yr(-1), while SFD exhibits both decreasing (at similar to 62% areas) and increasing (at similar to 38% areas) trends in the region. Again, LST is found to be the dominant factor that controls the temporal changes in both SFD and ALT while precipitation (i.e., both rainfall and snowfall) plays an important (especially in more arid areas and regions near the lower limit of permafrost) but secondary role. Our results demonstrate the advantages of the satellite-based method in frozen soil simulations over large scales with complex topography and landscape and highlight the important roles of both temperature and precipitation in shaping the frozen soil patterns on the TP or other cold, dry regions.