A downscaling-merging method for high-resolution daily precipitation estimation

Abstract

Due to the strong variation of precipitation in time and space, it is necessary to generate accurate precipitation data with high spatiotemporal resolution. This is essential for many applications such as hydrological and meteorological analysis. However, achieving such data over a large area remains a tremendous challenge. In this study, we presented a downscaling-merging method to solve this problem. First, an integrated “temporal upscaling – spatial downscaling – temporal downscaling” strategy is introduced to spatially downscale satellite-based daily precipitation data to a high spatial resolution of 0.01°. The downscaled daily precipitation estimates are then merged with gauge observations in a multivariate geostatistical framework. To avoid the adverse effects of sparse gauges, the semivariogram model of daily precipitation is inferred from the downscaled results. The method was applied to the Tropical Rainfall Measuring Mission (TRMM) daily precipitation product covering central China’s Henan Province for the period 1 January 2015–31 December 2016. The experimental results show that: (1) the downscaling strategy can effectively spatially downscale the TRMM daily precipitation product, maintaining not only the accuracy of the original TRMM data, but also providing greater spatial detail; (2) a seasonal timescale is appropriate for downscaling analysis; and (3) the merging scheme greatly enhances the accuracy of downscaled daily precipitation estimates, reducing the mean absolute error and root mean square error by 57% and 42%, respectively, and increasing the critical success index by 0.17. The proposed method provided an efficient solution for producing accurate daily precipitation data with high spatial resolution over a large area.