The current spatial atmospheric forcing data cannot accurately depict the actual conditions of the Qinghai-Tibet Plateau (QTP), where monitoring stations are scarce and unevenly distributed. This deficiency in atmospheric data hinders accurate simulation of plateau permafrost changes on the plateau. In this study, we develop a new approach to evaluate regional permafrost changes, which does not rely on spatially distributed meteorological data but instead uses the regional climate change processes or temperature change rates. Centred on a transient heat conduction permafrost model, this approach was applied to the Qinghai Hoh Xil National Nature Reserve (referred to as Hoh Xil) within the QTP from 1960 to 2015, using the rate of air temperature change provided by the Wudaoliang Meteorological Station, the only national station in Hoh Xil. Simulation results showed that the difference between the simulated and observed change rates of mean annual ground temperature (MAGT) was less than 0.04 degrees C per decade from 2001 to 2015 at five long-term monitoring sites. The simulated ground temperature profiles in four boreholes from various permafrost zones revealed an error of less than 0.7 degrees C below 5 m in depth. Model validation demonstrates the reliability of this approach for predicting long-term permafrost changes. Future regional permafrost changes were further simulated based on the latest warming scenarios (BCC-CSM2-MR) from the Coupled Model Intercomparison Project Phase 6. Predictions revealed significant differences in the regional permafrost degradation rate under different climate warming scenarios. Under the most severe warming scenario (SSP58.5), permafrost in the study area is projected to still cover 72.2% of the total area by 2100, with most of the Hoh Xil's permafrost becoming warm (MAGT > 1 degrees C) permafrost. This approach not only facilitates the simulation of frozen ground changes in areas with few meteorological monitoring stations but also provides a new perspective for using coarse-resolution palaeoclimate data to investigate permafrost formation and evolution over long time scales.

Numerical simulation is of great importance to the investigation of changes in frozen ground on large spatial and long temporal scales. Previous studies have focused on the impacts of improvements in the model for the simulation of frozen ground. Here the sensitivities of permafrost simulation to different atmospheric forcing data sets are examined using the Community Land Model, version 4.5 (CLM4.5), in combination with three sets of newly developed and reanalysis-based atmospheric forcing data sets (NOAA Climate Forecast System Reanalysis (CFSR), European Centre for Medium-Range Weather Forecasts Re-Analysis Interim (ERA-I), and NASA Modern Era Retrospective-Analysis for Research and Applications (MERRA)). All three simulations were run from 1979 to 2009 at a resolution of 0.5 degrees x 0.5 degrees and validated with what is considered to be the best available permafrost observations (soil temperature, active layer thickness, and permafrost extent). Results show that the use of reanalysis-based atmospheric forcing data set reproduces the variations in soil temperature and active layer thickness but produces evident biases in their climatologies. Overall, the simulations based on the CFSR and ERA-I data sets give more reasonable results than the simulation based on the MERRA data set, particularly for the present-day permafrost extent and the change in active layer thickness. The three simulations produce ranges for the present-day climatology (permafrost area: 11.31-13.57 x 10(6) km(2); active layer thickness: 1.10-1.26 m) and for recent changes (permafrost area: -5.8% to -9.0%; active layer thickness: 9.9%-20.2%). The differences in air temperature increase, snow depth, and permafrost thermal conditions in these simulations contribute to the differences in simulated results.