The invention discloses a dynamic precipitation downscaling method considering local variables and global variables, comprising the following steps: 1, obtaining local variables and global variables of a watershed, wherein the local variables include DEM, slope, aspect, surface roughness and distance to a coastline, and the global variables include vegetation cover index and wind speed; 2, obtaining the original satellite precipitation data of the watershed; and 3, constructing a mixed geographically weighted regression model, establishing a relationship equation between the original satelliteprecipitation and the variables, estimating coefficients of Rkm by a weight function, and obtaining spatial precipitation of Rkm according to the estimated coefficients of Rkm. The invention considers both local variables and global variables, realizes the spatial downscaling of satellite precipitation, further obtains the spatial precipitation with high spatial-temporal resolution, and can moreaccurately reflect the spatial distribution of precipitation.