Global Characteristics of Heavy Rainfall from Harmonized Geostationary Satellite Observations

Heavy rainfall is one of the most impactful hydrometeorological extremes, frequently causing floods, landslides, and severe socioeconomic damage worldwide. Continuous, high-temporal-resolution monitoring of heavy rainfall is essential for disaster risk reduction and early warning. Recent advances in satellite remote sensing and artificial intelligence (AI) have opened new possibilities for global-scale observation and analysis of extreme precipitation by integrating multi-platform satellite data within a unified framework. In this study, we develop a harmonized global geostationary satellite dataset by integrating observations from multiple operational platforms, including the GEO-KOMPSAT-2A (GK2A), Meteosat Second Generation (MSG), and the Geostationary Operational Environmental Satellite (GOES). To address differences in temporal sampling and radiometric characteristics among these satellites, we apply a deep learning–based video frame interpolation (VFI) technique. This approach enables temporally consistent interpolation across overlapping satellite domains and facilitates the construction of seamless global cloud maps with high temporal continuity. Heavy rainfall characteristics are analyzed by linking the harmonized geostationary cloud-top observations with satellite-derived precipitation estimates produced using AI-based retrieval algorithms. These AI-driven precipitation products are designed to capture nonlinear relationships between cloud properties and surface rainfall, providing enhanced sensitivity to intense precipitation events. To assess their robustness and physical consistency, the AI-based precipitation estimates are systematically compared with conventional satellite precipitation products derived from traditional physically based or empirically calibrated retrieval methods. This comparison allows us to evaluate the added value of AI-based precipitation retrievals in representing heavy rainfall intensity and occurrence at the global scale. The analysis focuses on identifying global and regional characteristics of heavy rainfall in relation to cloud-top temperature, emphasizing climatic contrasts across tropical, subtropical, and midlatitude regimes, as well as land–ocean differences. This study demonstrates that the synergy between harmonized multi-geostationary satellite observations and AI-based precipitation retrievals provides a powerful framework for global heavy rainfall analysis. The physically interpretable relationships identified between cloud-top signals and heavy rainfall establish a solid observational basis for future AI-driven or hybrid early warning systems. By combining continuous geostationary monitoring with advanced AI methodologies, this work contributes to improved global assessment of heavy rainfall risk and supports the development of more reliable hydrometeorological early warning capabilities.