High-resolution precipitation prediction in Bangladesh via ensemble learning

Abstract

As a developing agricultural country, Bangladesh is vulnerable to the effects of climate change, so accurate precipitation prediction is of great value to Bangladesh in achieving sustainable development. Traditional climate simulation models and prediction tools find it challenging to meet the growing needs on high spatial resolution. In this paper, we developed a XGBoost-based spatio-temporal precipitation prediction model and then generated high-resolution precipitation distribution maps in Bangladesh from 2025 to 2035, where the spatial resolution can reach 0.1° latitude and longitude. Finally, the EOF analysis reveals three leading modes in high-resolution precipitation evolution during 2025–2035.