Large Uncertainties in Precipitation Exert Considerable Impact on Land Surface Temperature Modeling Over the Tibetan Plateau

The performance of land surface temperature (LST) simulations in land surface models is subject to uncertainties in atmospheric forcing datasets. However, the contributions of forcing variables in different datasets to the simulated LST over the Tibetan Plateau (TP) remain unclear. To investigate the sensitivity of LST simulation to atmospheric forcing, we conducted a series of experiments using the Community Land Model (CLM5.0) with different forcing datasets. Results showed that the simulated LST displayed large spatiotemporal variations, and the simulation driven by the China Meteorological Forcing Data set produced more reasonable estimates than the other simulations. Sensitivity analysis suggests that the 1% increase in air temperature causes the largest variation in LST (0.54%), while the same percentage change in precipitation leads to the smallest (only 0.002%) LST variation. However, the actual uncertainty in precipitation is much larger than that in air temperature, and thus uncertainty in precipitation contributes 24% of the LST variability in the cold season, compared to 18% for air temperature. Large precipitation uncertainty leads to strong uncertainties in snow processes (snowfall and snow cover fraction), and hence strong uncertainties in surface albedo and ground absorbed shortwave radiation, further exerting the absolute impact on LST in cold season. Moreover, the snowfall is also influenced by different schemes of snowfall/rainfall partitioning. This may result in uncertainties in snowfall and consequently LST for the southeastern TP. Our study highlights the importance of high-quality precipitation data, a forcing variable usually ignored in previous studies of LST simulation on the TP.