Research on zoning multi-source fusion precipitation simulation method of the Yangtze River Basin based on random forest model

A multi-source fusion precipitation simulation method for the Yangtze River Basin based on the random forest model (FCM-RF algorithm) was proposed using three types of satellite precipitation products. Using the fuzzy C-means algorithm and ground observation station data, the precipitation regions in the Yangtze River Basin were divided. The precipitation ratio was introduced to characterize the spatial distribution of precipitation. The fusion results were further optimized using the ordinary Kriging interpolation method. A set of multi-source fusion precipitation products with a spatial resolution of 0. 25 毅伊0. 25 毅 in the Yangtze River Basin were obtained and evaluated. The results show that the FCM-RF algorithm has good performance in the Yangtze River Basin, which can effectively improve the capture ability of original satellite precipitation products for precipitation events. The correlation coefficient between simulated precipitation and measured precipitation at the validation stations can reach 0. 76. The FCM-RF algorithm has similar interannual variation characteristics and is highly sensitive to precipitation in spring and autumn. It performs poorly in summer due to heavy rainfall, and exhibits small errors and low correlation coefficients in winter due to sparse rainfall and solid precipitation. The FCM-RF algorithm has strong precipitation capture ability in the southeast region, but its accuracy is low in the Qinghai Tibet Plateau region. © 2024 Editorial Board of Water Resources Protection. All rights reserved.