基于GWR模型的长江流域TRMM数据降尺度

(1.桂林理工大学 测绘地理信息学院, 广西 桂林 541006; 2.黑龙江省农垦科学院, 哈尔滨 150038)

TRMM 3B43; GWR; 降尺度; 植被指数; 长江流域

TRMM Downscaling Data of Yangtze Based on GWR Model
ZHANG Hanbo1, XU Yong1, DOU Shiqing1, JING Juanli1, ZHANG Nan1, ZHANG Weidong2

(1.College of Geomatics and Geoinformation, Guilin University of Technology Guilin, Guangxi 541006, China; 2.Heilongjiang Academy of Agricultural Reclamation, Harbin 150038, China)

TRMM 3B43; GWR; scale down; vegetation index; Yangtze River Basin

备注

遥感降水数据应用广泛,但其空间分辨率未能满足区域空间精度要求。以2001—2019年热带降雨测量卫星降水产品TRMM 3B43为数据源,通过构建地理加权回归模型(GWR),结合数字高程模型(DEM)、增强型植被指数(EVI)、归一化植被指数(NDVI)、坡向,分别以目前主流的植被指数(NDVI,EVI)为解释变量对长江流域TRMM数据进行降尺度研究,并用研究区内147个气象站点观测值对降尺度前后的TRMM数据进行精度验证。研究结果表明:(1)TRMM数据与气象站点实测数据在空间和时间上表现出较好的一致性和适用性,各站点相关系数为0.46~0.97。(2)降尺度数据在保证数据精度的前提下,空间分辨率有较大提升(由0.25°提升至1 km),且NDVI_降尺度TRMM数据优于EVI_降尺度TRMM数据,能更真实反映研究区域内的降水特征。(3)相比原始TRMM数据,降尺度数据降水区间范围扩大,且湿润年份降尺度结果优于干旱年份。通过对TRMM数据进行降尺度有助于推进中小尺度区域降水的时空变异特征研究,具有一定研究意义。
Remote sensing precipitation data is widely used, but its spatial resolution fails to meet the regional spatial accuracy requirements. Based on the Tropical Rainfall Measuring Mission(TRMM)precipitation product TRMM 3B43 from 2001 to 2019, combining with the digital elevation model(DEM), enhanced vegetation index(EVI), normalized difference vegetation index(NDVI), and aspect data, a geographically weighted regression model(GWR)was construct. The current mainstream vegetation indices(NDVI, EVI)were selected as explanatory variables to downscale the TRMM data of the Yangtze River Basin. We verified the accuracies of the TRMM data before and after downscaling based on the observation values of 147 meteorological stations located in the study area. The results showed that:(1)the TRMM data and in situ observation data showed good consistency and applicability in space and time; the correlation coefficient of each station ranged from 0.46 to 0.97;(2)the spatial resolution of downscaling data was greatly improved(from 0.25° to 1 km)under the premise of ensuring the accuracy of the data; the NDVI downscaling TRMM data were better than the EVI downscaling TRMM data, which could more truthfully reflect the characteristics of precipitation in the study area;(3)compared with the original TRMM data, the precipitation range of the downscaling data was extended; the downscaling results of precipitation of wet years were better than those of dry years. Downscaling TRMM data are of great importance to advance the research on the temporal and spatial variability of regional precipitation, which has the certain research significance.