Abstract:   Precipitation exerts an important role in the exchange of matter and energy in the global water cycle, affecting soil moisture, vegetation growth, and surface runoff. By employing existing station data, the spatial distribution of precipitation obtained by the meteorological observation method was obtained by interpolation. However, the interpolation results of a small number of meteorological stations are challenging to accurately analyze the spatial variation characteristics of actual precipitation. Launched on 28 November 1997, the Tropical Rainfall Measuring Mission (TRMM) was jointly developed by the United States National Aeronautics and Space Administration (NASA) and the Japan Aerospace Exploration Agency (JAXA), thus providing long time series and covering most regions of the world with precipitation rate data. Nevertheless, TRMM was obtained by indirect precipitation measurement. Meanwhile, there were related errors and uncertainties. As a result, TRMM accuracy evaluation was the primary work of regional precipitation research. The terrain of Yunnan Province is complex, and the altitude difference is large. In addition, it is affected by the southwest monsoon and southeast monsoon. Complex factors such as the uneven distribution of precipitation may affect the detection capabilities of TRMM satellites. However, the current evaluation of the applicability of TRMM data in Yunnan Province is only a simple analysis of the coefficient of determination. Besides, the research on the factors affecting the accuracy of TRMM precipitation data and errors such as deviation rate is still lacking. In the present study, the accuracy of TRMM 3B43 precipitation data was evaluated in Yunnan, where the terrain was complex, aiming to provide reliable precipitation product data for regional precipitation research and hydrological forecasting. Monthly precipitation from 25 meteorological stations in Yunnan province from 2009 to 2018 provided by the China Meteorological Data Service Center was used to analyze the characteristics of TRMM 3B43 precipitation data. The correlation coefficient(R), BIAS, root mean square error (RMSE) and mean absolute error (MAE) were used to analyze the applicability between TRMM 3B43 monthly precipitation data and meteorological station data. Then, the relationship between the TRMM 3B43 precipitation data corresponding to each station and the elevation, slope, and aspect was discussed in this study. Finally, the data was downscaled to the seasonal and annual scales. At the same time, its applicability was evaluated. Some results the current study showed that: (1) the mean difference between TRMM 3B43 precipitation data and measured precipitation data was small, especially at Luxi station, where the difference was only 0.02mm. The TRMM 3B43 precipitation data was basically in consistence with the measured precipitation data, showing that there were more in the west and south and less in the east and north. It was roughly distributed step by step with the altitude. However, the difference in the spatial distribution of precipitation caused by the difference in altitude and latitude of each station also revealed the difference between TRMM 3B43 and the measured precipitation. Therefore, it is of much necessity to evaluate its applicability before using TRMM data.(2) With an R as high as 0.9392, BIAS close to zero, RMSE as low as 32.9776mm, and MAE as low as 20.5730mm, TRMM 3B43 displayed an extremely significant correlation between monthly precipitation and measured precipitation. In the range of TRMM 3B43 monthly precipitation less than 250mm, the accuracy of the fitting was relatively higher. The TRMM 3B43 precipitation data exceeded 0.735 at 25 stations, which passed the significance test at the 0.01 level, and the deviation and error of most stations were small with high overall accuracy. However, due to the different geographical locations of the stations, the deviation and error of the data presented certain differences. (3) The accuracy of TRMM 3B43 data was more affected by the slope than the altitude and aspect. The correlation coefficients of slope and R, RMSE and MAE were 0.8167, 0.7071 and 0.6865 respectively, showing strong quadratic function relationship characteristics. Except for the TRMM 3B43 precipitation data at Gongshan station and Weixi station, the accuracy of most stations at different altitudes, slopes and aspect was higher, having stronger applicability. Particularly, the data applicability was stronger for sites located at an altitude of 1000-2000m, slope less than 4°, and slope direction of 160°-240°. (4) The correlation coefficient of TRMM 3B43 data after time downscaling was slightly reduced. The error was slightly larger, especially in the winter and the annual scale slightly remained less suitable. With the largest error in summer, the RMSE and MAE of the TRMM 3B43 precipitation in Yunnan Province and the measured precipitation in each season were both less than 97mm and 78mm, respectively. The transmissibility of errors caused the RMSE and MAE of the annual scale TRMM data to become larger, and the applicability was the worst compared to other time scales. Therefore, the TRMM 3B43 monthly precipitation data had high accuracy in Yunnan region, which could thus provide effective supplement to the surface precipitation data.

Key words:  TRMM 3B43, Precipitation data, Applicability evaluation, Topographic factor, Yunnan Province