Machine learning approach to predict mortality reduction from summer heatstroke and heart disease through urban modification scenarios for climate change adaptation

A hot summer or heatwave event induces heat stress-related human deaths. Reducing the number of those deaths is stated on one of social issues for urban resilience and sustainability. This study aims to evaluate a change in the deaths if energy-saving or temperature-decreasing measures in urban modifications would be installed as the climate change adaptation in the whole urban area, from a novel approach combined machine learning (ML) techniques with meteorological model simulations. In this study, the WRF-CMBEM (the WRF, the urban multi-layer, and building energy models) was used to simulate spatiotemporally urban meteorological conditions, while the ML was applied to perform and predict daily heat stress-related deaths in an urban area.

We covered the most populated Tokyo's 23 wards in Japan and demonstrated the extremely hot summer of 2018. As expected urban modification scenarios, the cases of a ground surface greening (GRN), no anthropogenic heat from buildings to atmosphere (noAH), rooftop photovoltanic (PV), and cool roofs (CRF) were evaluated in this study.

The ML performed well for heat-related daily deaths from heatstroke (HS) and ischaemic heart disease (IHD) by detecting important meteorological factors. After meteorological changes from a control case to four urban modification scenarios were predicted by the WRF-CMBEM, potential reductions in heat-related deaths were estimated using previous successful ML-trained models. As a result in July–August 2018, the GRN case performed the most effective decreases of 0.28 °C (50%ile), 0.37 °C (90%ile), and 0.56 °C (Max) in the outdoor surface air temperature of all grids resolved at 1 km. The temperature changes reduced HS deaths by 43% and IHD deaths by 18% during a peak period of the deaths in the summer 2018. The second effective modification was the CRF case, which showed temperature decreases of 0.23 °C (50%ile), 0.31 °C (90%ile), and 0.36 °C (Max), and a 14% reduction in HS deaths and a 13% reduction in IHD deaths. These demonstrations suggest that the combined implementation of urban modifications is more effective in reducing heat stress-related deaths, especially during heatwaves and extreme hot summers.