Effects of different soil thermal conductivity schemes on the simulation of permafrost on the Tibetan Plateau

Soil thermal conductivity (STC) is an essential parameter for soil temperature and soil heat flux prediction in the land surface model. In recent years, various STC schemes have been developed and evaluated based on direct laboratory measurements. Appropriate selection of STC scheme is the key to applying land surface model in permafrost regions. In this study, we compared the estimation from nine typical STC schemes with laboratory measurements obtained on undisturbed soil samples, and then incorporated these schemes into the latest version of the Community Land Model (CLM5.0) to evaluate their performance in simulating soil temperature and permafrost extent on the Tibetan Plateau. Statistical analysis shows that among the nine schemes, Lu (L2014) and Nikoosokhan (N2015) schemes provide good STC estimation for undisturbed soil samples with varying soil saturation. The performance of the default scheme of CLM5.0 ranked after the two schemes with a root mean squared error of 0.33 W m−1 K−1. The single-point simulation results show that L2014 scheme is significantly better than the other schemes for incorporating into CLM to simulate soil temperature. The scheme that performs best in soil samples STC estimation do not necessarily perform best in soil temperature simulation due to bias in soil moisture simulation. The permafrost simulation results show that the L2014 scheme with higher simulated temperatures gives the smallest overestimated fraction of the simulated permafrost extent, while the N2015 scheme gives the smallest underestimated fraction. According to the evaluation analysis, we found that different STC schemes have important effects on the simulation of permafrost dynamic. In addition, uncertainties in land surface model such as atmospheric forcing, soil moisture and soil texture also have non-negligible effects on the accuracy of permafrost simulations.