Predicting lethal humidity and associated excess mortality using machine learning and high-resolution datasets

Humid heat waves represent one of the most significant emerging hazards in a warming world, yet the excess mortality associated with humid heat remains poorly understood in data-sparse and vulnerable regions such as Southeast Asia and Africa, which are projected to be the most severely affected. This study will leverage machine learning and high-resolution climate datasets to address this critical gap, providing a novel approach to quantifying and predicting excess mortality associated with lethal humidity across the globe. By integrating detailed urban modelling with climate projections, this work aims to develop a method for predicting deadly heat waves in cities.  The methodology will involve a two-step machine learning framework to translate coarse-gridded temperature and humidity observations/model outputs into fine-scale estimates of human-experienced conditions within urban environments. This ensemble of high-resolution projections will incorporate detailed data on urban form, fabric, and function, capturing the synergies between urbanization and climate change impacts. Case studies from ten high-risk cities will provide insights into potential future human survivability, with a focus on vulnerable urban populations in Southeast Asia and Africa. Ultimately, this work will advance our understanding of humid heat risk and support the development of targeted early warning and mitigation strategies around the globe.