A study on downscaling GPM precipitation data based on optimal interpolation and geographically weighted regression: A case study of Ganzi prefecture, Sichuan Province

Satellite precipitation can provide important precipitation information and is currently an important reference for precipitation. However, satellite precipitation data is limited by resolution and accuracy, resulting in shortcomings in its application. In this study, based on the optimal interpolation and geographically weighted regression methods, the GPM satellite precipitation data in Ganzi Prefecture were downscaled and fused using the normalized difference vegetation index (NDVI) and digital elevation model (DEM) as the dependent variables. The satellite precipitation data was downscaled from 0.1°×0.1° to 1 km×1 km. The quality of the downscaled data was evaluated using multiple parameters, and the spatiotemporal differences between the downscaled data and station data were compared. The following findings were observed. (1) Both GPM and OIGPM precipitation data can reflect the spatial distribution of precipitation, but GPM tends to underestimate precipitation while OIGPM is more accurate. (2) The fusion downscaled precipitation using GWR can improve the spatial resolution of precipitation and enhance its accuracy. (3) Quantitative calculations of multiple evaluation parameters show that the downscaled precipitation data obtained through the optimal interpolation and GWR methods are significantly better than the original data. (4) The Optimal Interpolated GPM precipitation data (GOIGPM) without residual correction has a higher Correlation Coefficient and lower Root Mean Square Error with rainfall gauge data in the Ganzi region, making it more applicable.