Page 236 - 《高原气象》2026年第4期
P. 236
高 原 气 象 45 卷
1166
Downscaling study based on the mechanism of
spatial distribution of precipitation
Wei Dibo , Dong Xiaohua , Ma Yaoming 1, 3, 4, 5, 6, 7, 8 , Hu Xue’Er ,
1, 2
1, 2
1, 2
1, 2
Jiang Runhua , Wei Chong , Yu Dan 1, 2
1, 2
(1. College of Hydraulic and Environmental Engineering, China Three Gorges University, Yichang 443002, Hubei, China;
2. Comprehensive Law Engineering Research Center of Eco-environment in Three Gergs Reservoir Region,
Ministry of Education, Yichang 443002, Hubei, China;
3. Land-Atmosphere Interaction and Its Climatic Effects Group, State Key Laboratory of Tibetan Plateau Earth System, Environment
and Resources(TPESER), Institute of Tibetan Plateau Research, Chinese Academy of Sciences, Beijing 100101, China;
4. College of Earth and Planetary Sciences, University of Chinese Academy of Sciences, Beijing 100101, China;
5. College of Atmospheric Science, Lanzhou University, Lanzhou 730000, Gansu, China;
6. National Observation and Research Station for Qomolongma Special Atmospheric Processes
and Environmental Changes, Dingri 858200, Tibet, China;
7. Kathmandu Center of Research and Education, Chinese Academy of Sciences, Beijing 100101, China;
8. China-Pakistan Joint Research Center on Earth Sciences, Chinese Academy of Sciences, Islamabad 45320, Pakistan)
Abstract: High-precision precipitation data are crucial for watershed water resource management, ecological
conservation, and hydrological modeling. However, existing satellite remote sensing precipitation products have
relatively low spatial resolution in complex terrain regions, making it difficult to accurately capture the spatial
heterogeneity of precipitation. Therefore, this study takes the Three-River Source Region as the research area
and, based on the widespread applicability of Tropical Rainfall Measuring Mission (TRMM) products, first em‐
ploys Geodetector to analyze the explanatory power of ten environmental factors on precipitation spatial distribu‐
tion, including Normalized Difference Vegetation Index (NDVI), Actual Evapotranspiration (AET), Land Sur‐
face Temperature (LST), Wind Speed (WS), Total Cloud Cover (TCC), Digital Elevation Model (DEM),
Slope, Aspect, longitude (Lon), and latitude (Lat). The study also explores the enhancing or weakening effects
of factor interactions to provide scientific basis for data input in downscaling models. Subsequently, Geographi‐
cally Weighted Regression (GWR) and Multiscale Geographically Weighted Regression (MGWR) models are
constructed to enhance the spatial resolution of TRMM precipitation data from 0. 25°×0. 25° to 1 km×1 km, and
the downscaling accuracy of the two models is compared. The research results indicate: 1) Based on Geodetector
results at monthly, seasonal, and annual scales, seven environmental factors with strong influence on precipita‐
tion spatial distribution in the Three-River Source Region are selected as input variables for MGWR and GWR
models, namely NDVI, WS, AET, Slope, TCC, LST, and DEM. 2) Downscaling result evaluation based on
data from 12 stations shows that TRMM precipitation products have high applicability in the Three-River Source
Region. The MGWR model outperforms the GWR model in terms of R², BIAS, and RMSE indicators, particu‐
larly showing higher downscaling accuracy in the comparison between dry and wet years. The MGWR model per‐
forms better than the GWR model across monthly, seasonal, and annual scales, reducing errors in TRMM pre‐
cipitation products to some extent. Especially in high precipitation areas and topographically complex regions,
the MGWR model can more precisely capture precipitation spatial details, and its bandwidth optimization strate‐
gy reveals the scale differences of various environmental factors. 3) In different topographic and geomorphologi‐
cal regions (Lancang River Source, Yellow River Source, Yangtze River Source), the overall goodness of fit of
MGWR downscaling results exceeds 0. 85 in all three source regions, demonstrating the applicability and accura‐
cy of this method under diverse terrain conditions and different precipitation levels. The Geodetector-MGWR
downscaling framework constructed in this study can effectively enhance precipitation spatial resolution in com‐
plex terrain regions.
Key words: Sanjiangyuan Basin; Geodetector; MGWR; GWR; TRMM; downscaling

