基于地理加权回归模型的省级TRMM降水数据降尺度研究

Downscaling Research of TRMM Precipitation Data Based on GWR Model in Zhejiang Province

以2011—2013年浙江省地面监测站点降水数据为因变量,TRMM降水数据为自变量进行TRMM数据精度验证,结果表明两者存在比较明显的线性相关性.使用NDVI数据为空间因素,基于GWR模型进行TRMM数据的降尺度操作,三个典型年的降尺度结果与地面数据间的相关系数稳定在0.8左右,呈现良好的线性相关性,且降尺度后相关系数R提升,相对误差BIAS、均方根误差RMSE都有所降低.基于GWR模型的降尺度牺牲一部分数据精度为代价,可以提高TRMM数据的空间分辨率,同时使其能更好地表现变量的空间非平稳性.

The precipitation data of ground monitoring site and TRMM in Zhejiang Province from 2011 to 2013 are used to verify the accuracy of TRMM data. Results show that there is a significant linear correlation between observed precipitation data and TRMM data. Using NDVI data as the spatial factor,the downscaling operation of TRMM data based on GWR model is carried out. The correlation coefficient between the three typical years and the ground data is stable at about 0. 8,showing good linear correlation. After downscaling operation,R value increases,BIAS and RMSE values reduce. Downscaling operation based on GWR models sacrifices part of the data accuracy. However,it can improve spatial resolution of TRMM data,and make TRMM better to show the spatial nonstationary of variables.