Multi-Source Precipitation Data Merging for Heavy Rainfall Events Based on Cokriging and Machine Learning Methods

Gridded precipitation data with a high spatiotemporal resolution are of great importance for studies in hydrology, meteorology, and agronomy. Observational from meteorological stations cannot accurately reflect the distribution variations over large area. Meanwhile, radar-derived restricted by low accuracy areas complex terrain satellite-based spatial resolution. Therefore, hourly models were employed to merge stations, Radar, satellites; used five machine learning algorithms (XGBoost, gradient boosting decision tree, random forests (RF), LightGBM, multiple linear regression (MLR)), as well CoKriging method. In north Guangdong Province, four heavy rainfall events 2018 processed geographic obtain merged data. The method secured best prediction accumulated precipitation, followed tree-based (ML) algorithms, significantly, MLR deviated actual pattern. All methods showed poor performances timepoints little during events. ML performance at some when was over-related latitude, longitude, distance coast.