Abstract
Air quality is vital for human health and environmental stability, particularly in tropical regions such as equatorial Africa where extreme weather and limited monitoring networks pose major challenges. This study examines how heat waves, heavy rainfall, and stagnant wind conditions influence fine particulate matter (PM₂.₅) in the Democratic Republic of Congo from 2014 to 2024. Satellite observations (TROPOMI, MODIS, GPM) combined with ERA5 reanalysis reveal that stagnant winds strongly enhance PM₂.₅ accumulation due to reduced atmospheric dispersion, while intense rainfall effectively decreases concentrations through wet deposition. High temperatures also intensify photochemical processes that increase surface ozone formation. To address the non-linear interactions among meteorological factors, the study develops deep learning prediction models (LSTM and CNN-LSTM) capable of forecasting PM₂.₅ levels up to 48 hours ahead with high accuracy. These models offer strong potential for integration into early warning systems to support public health protection and climate adaptation, especially in data-sparse regions. The results demonstrate that an interdisciplinary framework uniting hydrometeorology, remote sensing, and artificial intelligence provides a powerful approach for understanding air quality dynamics in complex tropical environments.