Computation of spatially distributed rainfall by merging raingauge measurements, satellite observations and topographic information: a case study of the 21 July 2012 rainstorm in Beijing, China

Abstract

This paper develops a new method to estimate spatially distributed rainfall by merging the raingauge measurements (RGMs) and the satellite observations (SLOs), and considering the topographic effect on rainfall. The rainstorm occurred on 21 July 2012 in Beijing, China, was taken as a case to evaluate the performance of this method. First, three SLOs (i.e., TMPA 3B41RT, 3B42RT, and CMORPH) were compared against the RGMs. It revealed that among the three SLOs, the CMORPH dataset best matched the RGMs. Furthermore, as the CMORPH dataset also had the highest temporal and spatial resolution, it was selected for merging with the RGMs. Second, since the original grid size of the CMORPH dataset (8 km × 8 km) is large, each grid was separated into smaller grids (1 km × 1 km) by using the Digital Elevation Model (DEM) data, according to the relationship between elevation and rainfall. Third, for any interest point in the study area, only the raingauge stations within a calculated effective influence radius were used, and differences between the RGMs and the separated CMORPH dataset at those stations were calculated. Accordingly, the two rainfall datasets could be merged as a linear combination of the separated CMORPH dataset at the interest point and the weighted differences at those stations. Finally, by repeating the above steps for the other points in the study area, spatially distributed rainfall could be obtained. To validate this method, the merged rainfall data were compared with the two original rainfall datasets, which indicated that the merged rainfall data could describe the spatial distribution of rainfall more properly than the areal SLOs, and preserve the characteristic of the point RGMs well. Moreover, results of cross-validation for streamflow simulation by using the merged rainfall data based on the Digital Yellow River Integrated Model showed that the merged rainfall data had better performance than any of the two original rainfall datasets. The development of the new data merging method paves the way of tackling one of the most changeling problems in hydrological modeling, providing more accurate rainfall data to distributed hydrological models.