Assessing Heat Wave Vulnerability in India Using Machine Learning and Climate Model Insights

This study investigates the spatiotemporal characteristics of maximum temperatures and heat wave (HW) vulnerability across India under the context of global warming. Using high-resolution gridded surface air temperature (SAT) data (1951–2022) from the India Meteorological Department (IMD), three regions of maximum temperatures and distinct heat wave zones were identified, highlighting their divergence. Local radiative heating and anomalous wind flows from maximum temperature zones were identified as primary drivers of heat waves, with a notable increase in HW occurrences in southeast India post-1970, attributed to global warming. Machine Learning (ML) models, including Artificial Neural Networks (ANN), multiple linear regression, and support vector machines, were employed alongside CMIP6 climate models to predict maximum SAT for India (1981–2022). ANN outperformed other ML models with minimal biases and high accuracy, showcasing its capability to enhance HW predictability. Future projections (2023–2050) reveal a gradual rise on SAT during March–May, indicating heightened HW risks. Additionally, HW intensification during El Niño decay years was linked to anomalous anticyclonic circulations, reduced cloud cover, and enhanced shortwave radiation. This caused a rise in discomfort indices and extreme temperature hours, particularly in northwest and central India. Findings emphasize the critical role of ML techniques in improving HW forecasts and guiding adaptation strategies. These insights are vital for agriculture, health, urban planning, and disaster mitigation, equipping stakeholders to address escalating climate risks and societal impacts effectively

Keywords: Heat Waves (HW); Maximum Temperatures; Machine Learning (ML); Climate Change; Vulnerability Analysis