Spatial Downscaling Of TRMM Precipitation Data

Precipitation as the important component of the global water cycle, is the key parameters in hydrology, meteorology and climate. Conventional precipitation data from meteorological stations using interpolation method, which cannot reflect the spatial variation due to the limitation of the number of stations, especially where distribution of meteorological station is spare. In recent decades, with the development of remote sensing and meteorological satellite technology, using satellite remote sensing to detect rainfall information, has become an important source of spatial precipitation data. TRMM precipitation data with advantages in high spatial resolution and wide coverage, is popular at home and abroad. But when applied to the small regional scale, it still cannot meet the need, the downscaling is an effective way to solve the problem.In this study, we verify the applicability of TRMM precipitation data in the Greater Khingan Mountains with observed data from meteorological stations, and according to the correlation between TRMM precipitation data and NDVI data to build downscaling model based on the GWR model. In this way, it increases the spatial resolution of TRMM precipitation data to 1 km, and then uses the GDA calibration method to adjust date. Finally use the data to analyse the precipitation distribution regularity of the Greater Khingan Mountains. Main conclusions are as follows:(1) TRMM precipitation data is applicable to the Greater Khingan Mountatins. Except January,February and winter, the correlation coefficient between TRMM data and observed data has reaches above 0.85 in the allowable range, and the relative error is less than 20% from April to October, spring, summer, autumn. From the perspective of the spatial distribution of data error, the error of TRMM data is between-10% and 10% in the most of area of the Greater Khingan Mountatins, and it is overvalued.(2) Using TRMM precipitation data and NDVI data to establish GWR model to downscaling is feasible, and GWR model is more suitable for downscaling by comparison to the global OLS model. The downscaling results show that, for the downscaling based on GWR model, the correlation coefficient of annual mean, 2001, 2012 increases 0.02, 0.05, 0.15, the relative error decreases 0.06, 0.04, 0.02, the RMSE value decreases 24.54, 24.54, 11.09 respectively; as for the result based on OLS shows that the correlation coefficient only increases in 2012, the relative error decreases 0.11, 0.12, 0.07, the RMSE value decreases 18.7, 10.37, 28.25, respectively.(3) GDA method is available for calibrating downscaling results.Using GDA method to calibrating data and compare with the downscaling results, we find that the correlation coefficient of annual mean, 2001, 2012 increases 0.09, 0.04, 0.01; except 2012, the relative error of annual mean, 2001 decreases 0.10, 0.04, the RMSE value decreases 9.96, 9.24, respectively.(4) The Fractions derive from the GDA results can be used to disaggregate 1 km annual precipitation to 1 km monthly precipitation. The result shows that the correlation coefficient of the disaggregated 1 km monthly precipitation increases 0.002, the relative error decreases 15% and the RMSE value decreases 1.94.(5) The spatial variation of precipitation has following characteristics in Greater Khingan Mountatins. Firstly, the precipitation distribution of the Greater Khingan Mountatins is inhomogeneity.From January to April and from October to December, the precipitation is scarce and below 50 mm, among them February is the least and less than 10 mm, July is the largest and the maximum precipitation is 164 mm. Secondly, the precipitation mainly concentrated in the summer and the precipitation in winter is the least, its seasonal precipitation is obvious. Among them, spring accounts for 10% to 20% of annual precipitation, summer accounts for the proportion of annual precipitation is the largest, up to 70% to 80%, autumn is 10% to 20% in most area, and the precipitation of winter is the least, whose percentage is below 10%. Thirdly, the most precipitation in Greater Khingan Mountatins is located in the the east of middle part, near the Xiaoergou site, which annual rainfall is above 500 mm; the minimum value in the west flank of the north part, which annual rainfall is about 223 mm. On the whole, the spatial distribution of the annual precipitation in Greater Khingan Mountatins follows a decreasing trend from east to west, and the precipitation in the east of the middle part, the south of the south part is larger than the middle of the south part.