基于地理加权回归模型和不同植被特征参数的TRMM 3B43降尺度研究——以云南省为例

The suitability of downscaling of TRMM 3B43 based on a geographically weighted regression model and different vegetation characteristic parameters:a case study of Yunnan Province

为了将空间分辨率约为27 km×27 km的热带降水测量计划卫星(TRMM)3B43数据降尺度为1 km×1 km,并对比不同植被参数下TRMM 3B43降尺度效果,以云南省为研究区,TRMM 3B43卫星降水数据、MOD13 A3归一化植被指数(NDVI)和增强型植被指数(EVI)数据、MOD17A2H GPP数据、气象站点月降水数据等为数据源,基于地理加权回归模型,开展不同植被特征参数的TRMM3B43降水数据降尺度研究,采用线性相关系数、偏离率和均方根误差验证云南省整体、不同气候区及单气象站点TRMM数据的降尺度精度.结果表明,不同植被特征参数中,以植被总初级生产力、NDVI数据为基础的降尺度结果为佳,利用EVI进行降尺度的结果较差;各时间尺度下,以月尺度的降水数据降尺度结果最佳,其中降雨量较多的月份相关性更高,其次为季节尺度,其中以秋季降尺度最佳,年尺度下的降尺度综合精度评价结果较差;不同气候带下,边缘热带区域内TRMM降尺度效果最佳,其次为南亚热带和中亚热带,高原气候区TRMM降尺度结果稍差;单气象站点中,江城、丽江等站点处的TRMM降尺度效果最佳,贡山站点降尺度效果最差,与立体气候特征显著有联系.

Yunnan Province was used as the research area and satellite precipitation data of tropical rainfall measurement mission(TRMM)3 B43,normalized difference vegetation index(NDVI)and enhanced vegetation index(EVI)of the third level of moderate-resolution imaging spectroradiometer No.13,gross primary productivity(GPP)of the second level of moderate-resolution imaging spectroradiometer No.17,and monthly precipitation data from meteorological stations were used as data sources.Based on the geographically weighted regression model,a TRMM 3 B43 downscaling study of different vegetation characteristic parameters was carried out,linear correlation coefficient,deviation rate and root mean square error used to verify the downscaling accuracy of TRMM data for Yunnan Province as a whole,regarding different climate regions and single meteorological stations.The purpose was to downscale the TRMM data with a spatial resolution of about 27 km×27 km to 1 km×1 km,and compare the TRMM downscaling effects of different vegetation parameters.The results showed that,of the different vegetation characteristic parameters,the downscaling results based on GPP and NDVI data were the best,while the results based on EVI were poor.In all time scales,the monthly downscaling results of precipitation data were the best,and the months with more rainfall were more correlated.The second was the seasonal scale,about which the fall was the best season for downscaling.The comprehensive accuracy evaluation result of downscaling under the annual scale was poor.Under different climate belts,the TRMM downscaling effect was the best in the marginal tropical region,followed by the south subtropical region and the middle subtropical region,and the TRMM downscaling effect was slightly worse in the plateau climate region.Among the single meteorological stations,the downscaling effect of TRMM at Jiangcheng and Lijiang was the best,and the downscaling effect at Gongshan station the worst,which was significantly related to the three-dimensional climate characteristics.