A machine learning approach to predict the heatwave in Bangladesh region due to global warming

Abstract

Since heatwaves are becoming more frequent and intense all over the world due to global warming, accurate heat wave predictions become essential for ensuring global safety. According to the Ministry of Environment, Forest and Climate Change, Bangladesh is one of the vulnerable countries that is being affected by global warming. To the best of our knowledge, no remarkable machine learning based research has been conducted yet on the actual dataset from Bangladesh. In Bangladesh, heatwaves typically occur during the pre-monsoon season. It has been observed in various seasons in recent years. The heatwave increased the health risk of humans as well as the death rate. Accurate early warning of heatwaves will be beneficial to keep human life safe. This study employs multiple machine learning approaches used to predict the division-wise heatwave based on meteorological data recorded by the Bangladesh Meteorological Department (BMD) from 35 stations between 2000 and 2024. Heatwave events are calculated with the operational definition of heatwaves used by BMD based on three or more consecutive days with maximum temperatures of 36°C or higher. This research explores ML binary classification models for classifying heatwaves, multi-regressor and hybrid deep learning models to provide early heatwave alerts using BMD recommended post-thresholding on forecasted temperature. The core objective of this research is to contribute to Bangladesh’s heatwave forecasting system, which will help with upcoming environmental risk management initiatives.