基于地理加权回归的河南省遥感降水产品降尺度估计

Downscaling Estimation of Satellite-Based Precipitation Products in Henan Province Based on Geographically Weighted Regression

摘要: TRMM降水数据应用广泛，多适用于水文、气象、灾害预警等方面的研究，但0.25˚ × 0.25˚的空间分辨率太低过于粗糙，难以满足当前研究的需求。因此本文通过归一化植被指数、数字高程、和经纬度等地理环境因子构建地理加权回归模型以河南省为研究区域，对2001~2025年河南省TRMM降水数据进行降尺度处理，使降水数据集的分辨率提高。通过决定系数(R²)、相对偏差(Bias)、均方根误差(BMSE)等评价指标，使用气象站点的实测降水数据作为“真实值”对降尺度后的降水数据进行精度验证。研究发现：经过降尺度操作后，降水数据的数据精度整体有所提高，数据的分辨率得到明显的提升，降水空间分布更加细腻，能够更好地表现河南省降水的空间分布和时空变化。

Abstract: TRMM precipitation data are widely used in studies related to hydrology, meteorology, and disaster early warning. However, the spatial resolution of 0.25˚ × 0.25˚ is relatively coarse and insufficient to meet the requirements of fine-scale research. Therefore, this study constructs a geographically weighted regression (GWR) model using geographic environmental factors, including the normalized difference vegetation index (NDVI), digital elevation model (DEM), and latitude-longitude information. Taking Henan Province as the study area, the TRMM precipitation data from 2001 to 2025 were downscaled to improve the spatial resolution of the precipitation dataset. The downscaled precipitation data were evaluated with statistical indicators including the coefficient of determination (R²), relative bias (Bias), and root mean square error (RMSE). Observed precipitation data from meteorological stations were used as “ground truth” for accuracy validation of the downscaled precipitation data. The results indicate that the downscaling procedure significantly improved the overall accuracy of the precipitation data. In addition, the spatial resolution was markedly enhanced, resulting in a more refined representation of precipitation patterns and more precise characterization of spatial details.