Effective multi-satellite precipitation fusion procedure conditioned by gauge background fields over the Chinese mainland

The accurate estimation of precipitation has a basis for atmospheric and hydrological applications. However, large uncertainties exist in the estimation of regional precipitation derived solely from remote sensing data. Herein, we develop an effective method to fuse multiple satellite-based precipitation products to capitalize on their strengths and maximize their applicability at the 0.25 degrees resolution scale. Specifically, we describe an approach based on bias correction in individual satellite-based products and their merging with up-scaled (1.0 degrees) gauge-based interpolated data to produce finer-resolution (0.25 degrees) data outputs. By focusing on seven regions to generate the estimates of merged precipitation across the Chinese mainland (MPCM), we performed a full evaluation based on a range of error metrics. The results showed that the bias correction procedure can reduce the discrepancies in the individual satellite-based precipitation products. In addition, the mean absolute error values are also considerably reduced (20%-45%) with respect to the original data over the Chinese mainland. Error statistical metrics demonstrated that the MPCM generated considerably better estimates on a daily scale, and performed better than the satellite-based precipitation products across different precipitation thresholds. The intercomparison results clearly states the superiority of the MPCM against Multi-Source Weighted-Ensemble Precipitation version 2 global product. Nonetheless, we also suggest that the large uncertainties in satellite-based products should be paid more attention over mountainous areas where rain gauge data are insufficient.

Automated summary

This study develops an effective method to fuse multiple satellite-based precipitation products and improve their accuracy. The results demonstrate that the fused precipitation estimates outperform individual satellite-based products and perform better across different precipitation thresholds.