A Study on the Application of Google Earth Engine for Precipitation Monitoring

Abstract

Monitoring precipitation is an important basis for climate change studies and for the prevention and control of hydrological disasters. There are currently skewed distributions of ground-based meteorological observation stations, and satellite-based precipitation products still lack sufficient spatial resolution and systematic accuracy, directly influencing the accuracy of runoff models and early flood and drought warning systems. In this paper, the advancement in precipitation monitoring on the Google Earth Engine (GEE) platform was reviewed. It methodically evaluates the synthesis and use of mainstream precipitation products, including IMERG, CHIRPS, and ERA5-Land, as well as complementary information on topography and vegetation, and ground-truth validation data on the GEE cloud platform. The article focuses on the essence of technical workflow procedures for the precise spatial resolution of precipitation in complex terrain and urban environments by combining spatial-decay techniques with machine-learning algorithms, including Random Forest and XGBoost, on the platform. It also examines existing issues, such as inefficiencies in the parameter optimisation of prebuilt algorithms at GEE, the lack of deep learning functionality, and the inability to validate multi-source precipitation products under extreme precipitation conditions. Moreover, the paper outlines future development directions, including deep integration with external AI frameworks and the implementation of higher-resolution spatiotemporal observational data.