Spatial downscaling of monthly precipitation based on TRMM satellite in Ya’an

Abstract

In the study of precipitation in ya’an with complex terrain, a statistical downscaling algorithm based on the monthly precipitation data of TRMM 3B43 from 2000 to 2008, digital elevation model (DEM) and normalized difference vegetation index (NDVI) was presented. The precipitation fields from TRMM 3B43 were downscaled to l km × l km resolution simulated precipitation by constructing multiple linear regression equation and using regression model and residuals method. Combined with the observation values obtained from 8 weather stations in the study area, the downscaled precipitation was tested in the two aspects of overall accuracy and singel station. On this basis, the spatial and temporal distribution characteristics of monthly precipitation in recent 9 years in the study area were analyzed. The results showed that: The correlation coefficients of TRMM before and after downscaling and observation values were 0.863 and 0.854, and the biases were -0.08 and -0.12. The correlation and accuracy of the two types of data were basically consistent on the whole. 75% of the stations had a high correlation with the downscaling results, and the absolute value of the bias of each station was no more than 0.35, which showed that the data accuracy was high, the data was suitable for the precipitation analysis in the research area. The distribution of monthly precipitation in 8 stations over the years showed that the precipitation in the middle of ya’an was concentrated in January and November to December. From February to April and July to August the north received more precipitation than the south, while the opposite was true from May to June and September to October, and the August in the main flood season reached the heaviest rainfall of the year.