Correction of Gridded Precipitation Data Based on Machine Learning

Gridded precipitation data play a crucial role in hydrological modeling and climate research, but biases between these data and actual ground observations limit their accuracy and reliability. Correcting these biases is essential to improve the accuracy of gridded precipitation data for practical applications. In this study, we developed the MR-interaction model based on single-day and multi-day gridded precipitation data and compared its performance with five machine learning models to achieve optimal predictive accuracy. By applying the RM-CONS correction method to the gridded data, we significantly reduced accuracy errors, resulting in corrected gridded data that more closely match the observed data distribution.