Simulation of Changes in the Near‐Surface Soil Freeze/Thaw Cycle Using CLM4.5 With Four Atmospheric Forcing Data Sets

Abstract

AbstractChange in the near‐surface soil freeze/thaw cycle is critical for assessments of hydrological activity, ecosystems, and climate change. Previous studies investigated the near‐surface soil freeze/thaw cycle change mostly based on in situ observations and satellite monitoring. Here numerical simulation method is tested to estimate the long‐term change in the near‐surface soil freeze/thaw cycle in response to recent climate warming for its application to predictions. Four simulations are performed at 0.5° × 0.5° resolution from 1979 to 2009 using the Community Land Model version 4.5, each driven by one of the four atmospheric forcing data sets (i.e., one default Climate Research Unit‐National Centers for Environmental Prediction [CRUNCEP] and three newly developed Modern Era Retrospective‐Analysis for Research and Applications, Climate Forecast System Reanalysis, and European Centre for Medium‐Range Weather Forecasts Reanalysis Interim). The observations from 299 weather stations in both Russia and China are employed to validate the simulated results. The results show that all simulations reasonably reproduce the observed variations in the ground temperature, the freeze start and end dates, and the freeze duration (the correlation coefficients range from 0.47 to 0.99, and the Nash‐Sutcliffe efficiencies range from 0.19 to 0.98). Part of the simulations also exactly simulate the trends of the ground temperature, the freeze start and end dates, and the freeze duration. Of the four simulations, the results from the simulation using the CRUNCEP data set show the best overall agreement with the in situ observations, indicating that the CRUNCEP data set could be preferentially considered as the basic atmospheric forcing data set for future prediction. The simulated area‐averaged annual freeze duration shortened by 8.03 days on average from 1979 to 2009, with an uncertainty (one standard deviation) of 0.67 days caused by the different atmospheric forcing data sets. These results address the performance of numerical model in simulating the long‐term changes in the near‐surface soil freeze/thaw cycle and the role of different atmospheric forcing data sets in the simulation, which are useful for the prediction of future freeze/thaw dynamics.