Real-time bias adjustment for satellite-based precipitation estimates over Mainland China

摘要

An improved cumulative distribution function (CDF)-based approach to reduce the systematic biases of multi-satellite precipitation estimates in real time is proposed and verified over Mainland China. Efforts are primarily focused on establishing the bias-adjusting model by adopting the CDF based on a Self-adaptive Moving Window (CSMW), which systematically integrates the China Gauge-based Daily Precipitation Analysis (CGDPA) into the real-time TRMM Multisatellite Precipitation Analysis (TMPA-RT). In our modelling experiments, the first 9-yr (2008-2016) precipitation data pairs were used to calibrate the CSMW model and establish a satellite-gauge relationship, which was then applied to the last 3 years of 2017-2019 as validation. Assessment results during the independent validation period show that the CSMW approach can significantly reduce the systematic positive bias of original TMPA-RT precipitation estimates in that the relative bias (RB) during the validation period decreases from 16.01% before adjustments to 0.29%, and the root-mean-square error (RMSE) also has a dramatic drop of 13%. The error component analysis indicates that the substantial improvement is mainly manifested in the hit events (observed rain was correctly detected by satellite) but it failed to reduce the miss bias (observed rain was not detected by satellite). This arises because a majority of missed precipitation is drizzle and falls below the rain/no-rain discriminant threshold, which is normally excluded from the CSMW algorithm. Additionally, the CSMW approach seems to have significantly improved the TMPA-RT estimates at the mediumhigh rain rates (>8 mm/day), but it also has a limitation in enhancing the correlation coefficient between satellite retrievals and ground observations. The major advantage of this approach is its applicability when real-time gauge data are not available, which could further facilitate the expansion of satellite-based precipitation estimates for real-time natural hazards forecasting.