摘要
Abstract
Accurately obtaining high-resolution ground precipitation information is crucial for studies such as basin hydrological modeling and crop growth simulation.Addressing the low spatial resolution deficiency common in existing TRMM satellite products,this study proposes a new downscaling method based on the GBDT algorithm for TRMM satellite precipitation products within the Haihe River Basin.This method integrates TRMM data with multi-source auxiliary data(such as topography,land cover,etc.)to construct high-dimensional input vectors.The GBDT regression model is then used to predict precipitation amounts at each high-resolution grid point,thereby achieving downscaling.The research results show that:① The GBDT algorithm possesses excellent nonlinear mapping capabilities,significantly improving downscaling accuracy within the Haihe River Basin;② After downscaling processing,the spatial resolution of TRMM products was enhanced from 0.25° to 4km,with a coefficient of determination(R2)of up to 0.73 when compared to ground station observations;③ The high-resolution product accurately represents the spatial heterogeneity of precipitation in the Haihe River Basin,revealing the significant impact of geographic elements such as terrain and vegetation on precipitation distribution patterns.This study provides a new approach and technical route for generating high-quality satellite precipitation data.关键词
TRMM卫星/降水/降尺度/梯度提升决策树/海河流域Key words
TRMM satellite/precipitation/downscaling/gradient boosting decision tree/haihe river basin分类
地球科学
