Sensitivity of Snow Simulations to Different Atmospheric Forcing Data Sets in the Land Surface Model CAS-LSM

Abstract

The quality of snow simulations in land surface models (LSMs) largely depends on the accuracy of the atmospheric forcing data, especially precipitation and air temperature. To investigate the sensitivities of snow simulations to atmosphere forcing, historical simulations from 1981–2010 were conducted using the Chinese Academy of Sciences land surface model (CAS-LSM) with four atmospheric forcing data sets: third Global Soil Wetness Projects (GSWP3), the Water and Global Change (WATCH) Forcing Data (WFD/WFDEI), the Climate Research Unit - National Centers for Environmental Prediction (CRU-NCEP), and Princeton. A sensitivity index (Ψ) is utilized to quantify the sensitivity of the simulated snow cover fraction (SCF), snow water equivalent (SWE), and snow depth (SDP) to the uncertainties in forcing data. By comparing the simulated results with satellite-based products and in situ observations, we find that CAS-LSM generally captured the spatial and seasonal variations of the SCF, SWE, and SDP. The simulation based on GSWP3 produced more reasonable estimates than the other three simulations, particularly for the SCF and SDP. The sensitivity analysis suggested that the SWE and SDP suffered the most from the uncertainties in the atmospheric forcing data sets. The sensitivities of the SCF, SWE, and SDP to precipitation uncertainties had various seasonal cycles depending on the climate regime. The highest sensitivity in boreal climates appeared during January–March, while in warm temperate climates the highest sensitivity existed during April–June. On average, the strongest precipitation sensitivity was found for temperate climates with cold, dry winters. The sensitivities to uncertainties in air temperature showed similar patterns; however, air temperature sensitivity was generally dominant over precipitation sensitivity in boreal climates, while in warm temperate climates both of them were high.