Evaluation of Winter Precipitation Products over the Tibetan Plateau with the Sublimation derived from Remotely Sensed Snow Cover Data

Many model results showed obvious wet biases during winter while the simulation was good during summer over the Tibetan Plateau (TP). Low gauge density and the limited capacity of snowfall may introduce dry biases into the observation and then exaggerate the overestimation of winter precipitation. To evaluate the winter precipitation products over the TP, we compared six precipitation products, including TRMM, ERA5, ERA-Interim, GLDAS, HAR, and the observation provided by China Meteorological Administration (CMA), against a sublimation dataset derived from remotely sensed snow cover data. The Kuzmin formula constrained with IMS snow cover product and land surface temperature was used to estimate sublimation. To ensure the reliability of the sublimation value, the accuracy of the simulated sublimation value was verified by the sublimation value observed at the pass area of Dadongshu Station and the consistency of two snow cover products was verified by using MODIS daily cloud-free snow cover products over the Tibetan Plateau.

The comparison revealed that the average underestimated area ratio of CMA on the TP and the Inner TP respectively were about 60% and 90%. CMA has an obvious underestimation (80% region showed underestimation and precipitation underestimation ratio mostly more than 100%) in the west of TP where lack of observation site. However, there was not obvious underestimation in East TP because of the dense stations available. It implies that the observation data has considerable dry biases (~200%) in winter precipitation over the Western TP where more ground stations are needed to get a reliable precipitation observation. For other precipitation products, HAR showed the smallest underestimation with a 12% region of precipitation underestimation. ERA5 and ERA-Interim are close behind HAR, but the underestimation area ratio of ERA5 was about 15% smaller than ERA-Interim in each statistical area of TP. TRMM and GLDAS show comparable underestimation and both are more apparent than ERA-interim. The underestimation phenomenon of TRMM shows little difference in the western and eastern TP and the underestimated area ratio of TRMM was 64.68% on the TP.