Spatially continuous assessment of satellite-based precipitation products using triple collocation approach and discrete gauge observations via geographically weighted regression

Abstract

Assessment of the accuracy of satellite-based precipitation products (SPPs) is essential prior to their use in formal applications. However, thorough assessments of SPPs by traditional approaches are typically not practical when gauge observations are sparse. The triple collocation (TC) approach can be used to perform spatially continuous assessments, but TC is affected by error and bias. In this study, a novel approach for the spatially continuous assessment of SPPs that integrates TC and the traditional discrete gauge-based approach via geographically weighted regression (GWR) was developed. This approach, referred to as TC-GWR, allows SPPs to be assessed in areas with limited gauge observations. The TC-GWR was tested comprehensively through using four widely-used SPPs over mainland China. The TC-GWR generates spatially continuous assessment results for SPPs and outperforms significantly the original TC approach in assessing the accuracy of SPPs quantitatively. The performance of TC-GWR was generally robust to variation in the density of gauge stations (50, 100, 200, 400, and 800 gauge stations) and showed acceptable performance when gauge data are relatively sparse, except for the extremely sparse gauge cases (only 10 stations over mainland China). Overall, TC-GWR integrates the advantages of the traditional approach (higher reliability) and the TC approach (spatial continuity) and provides a valuable alternative for assessing SPPs in areas with limited gauge observations, such as remote mountainous areas and developing countries.