基于GWR模型的典型区域GPM数据降尺度研究——以浙江省为例

为提高研究区域的降尺度效果,基于地理加权回归法(Geographically Weighted Regression,GWR),选取全球降水计划(Global Precipitation Measurement,GPM)3IMERGM产品,以数字高程模型(Digital Elevation Model,DEM)作为控制解释变量,将其分别与解释变量水汽通量散度、气温构建两个降尺度模型、与解释变量归一化植被指数(Normalized Difference Vegetation Index,NDVI)构建对照降尺度模型,对浙江省进行降尺度研究。利用研究区域内气象站点的实测数据,对由不同解释变量构建的3个降尺度模型的降尺度结果进行对比分析及精度验证。结果表明:构建的3个降尺度模型中,引入解释变量水汽通量散度构建的降尺度模型的综合效果优于其余两种模型,水汽通量散度较NDVI、气温更适合作为解释变量。构建的降尺度模型有效地提高了GPM数据的空间分辨率(由0.1°提升至1 km),降尺度数据维持了精度且能够更真实反映研究区域内的降水量分布情况。

Downscaling research of GPM data in typical region based on GWR model: a case study of Zhejiang Province

In order to explore precipitation explanatory variables which better conform to regional typical characters, on the basis of GPM 3IMERGM precipitation product in Zhejiang Province, three Geographically Weighted Regression (GWR) models for downscaling were constructed taking advantage of Digital Elevation Model (DEM). Compared with Normalized Difference Vegetation Index (NDVI) which was commonly used as explanatory variable in downscaling research, divergence of moisture flux and air temperature were innovatively introduced in models respectively. In light of measured data from Zhejiang meteorological stations, precision verification and comparison between the models were taken for further analysis. The results show that the downscaling model with divergence of moisture flux as explanatory variable has better fitting effect and precision than others among the three models constructed. Divergence of moisture flux is more suitable than NDVI and air temperature as explanatory variable. Meanwhile, the constructed downscaling model effectively improves the spatial resolution of GPM data (from 0.1° to 1 km). The downscaling data maintains the precision and reflects the precipitation distribution more accurately in study region.