Analyzing Deep Convective Cloud Properties for Philippines’ Monsoon Heavy Precipitation Using Self-Organizing Maps

Rainfall and precipitation influenced by synoptic systems such as monsoons play a major role in the Philippines’ climate variability and contribute to natural disasters like landslides and floods. The monsoon season brings heavy rainfall to much of the country, but modeling this rainfall remains challenging due to the complexity of deep convection processes. This study uses an unsupervised machine learning method, the Self-Organizing Map (SOM), to analyze deep convective properties during the monsoon season in the Philippines while accounting for key meteorological factors. The SOM is trained on 20 years (2000–2019) of data from the European Centre for Medium-Range Weather Forecasts Reanalysis 5, centered over the Philippines. Input variables include daily 0400 UTC geopotential height anomalies at 850 hPa. Additional variables—such as cloud fraction, water vapor, and other convective properties—are taken from the Moderate Resolution Imaging Spectroradiometer Level 3 global gridded data and the Clouds and the Earth’s Radiant Energy System Synoptic fluxes and cloud data. The properties derived from the datasets are projected onto each SOM regime. The results reveal a clear seasonal separation where dry season regimes are associated with low cloud fraction and water vapor, while wet season regimes exhibit high cloud fraction and water vapor. The patterns also correspond to reduced shortwave and enhanced longwave fluxes. The findings demonstrate that SOMS can effectively link large-scale atmospheric dynamics with local cloud and radiation properties, providing insight into the drivers of monsoons in the region.