Abstract This study proposes a novel approach combining machine learning (ML) techniques with meteorological model simulations to evaluate the heat-related mortality reduction potential of a climate change adaptation measure, namely, the installation of energy-saving or temperature-decreasing modifications in an urban area (e.g., greening, high-albedo paints, and photovoltaics). These methods have been used separately to assess the future urban health. The Weather Research and Forecasting–Canopy-Building Energy Model (WRF–CMBEM) was used to simulate spatiotemporal urban meteorological conditions, and ML was applied to predict daily heat-related deaths in the 23 wards of Tokyo during the extremely hot summer of 2018. The urban energy-saving and heat island mitigation scenarios evaluated in this study were ground surface greening, no anthropogenic heat from buildings to the atmosphere, rooftop photovoltaics, and cool roofs. ML accurately predicted heatstroke- and ischemic heart disease (IHD)-related daily deaths using important meteorological factors. After meteorological changes from the control case to four urban modification scenarios were predicted using the WRF–CMBEM, potential reductions in heat stress-related deaths were estimated using previously successful ML-trained models. The results showed that in July–August 2018, the ground surface greening case effectively decreased the outdoor surface air temperature by 0.28 °C (50-percentile), 0.37 °C (90-percentile), and 0.56 °C (Max) in all grids resolved at 1 km. Temperature changes reduced heatstroke deaths by 43% and IHD deaths by 18% during the peak period of deaths in summer 2018. Cool roofs resulted in temperature decreases of 0.23 °C (50-percentile), 0.31 °C (90-percentile), and 0.36 °C (Max) and 14% and 13% reductions in heatstroke and IHD deaths, respectively. The results suggest that the implementation of urban modifications can effectively reduce heat-related deaths, especially during heatwaves and extremely hot summers.
Read full abstract