Spatiotemporal Variations of Global Human-Perceived Heatwave Risks and their Driving Factors Based on Machine Learning

Abstract

With ongoing global warming, heatwave-related disasters are on the rise, exerting a multifaceted impact on both the natural ecosystem and human society. While temperature has been extensively studied in the effects of extreme heat on human health, humidity has often been ignored. It is crucial to consider the combined influence of temperature and humidity when assessing heatwave risks and safeguarding human well-being. This study, leveraging remote sensing products and reanalysis data, presented the first analysis of the spatiotemporal variations in global human-perceived heatwaves on a seasonal scale from 2000 to 2020, and further employed the Random Forest (RF) regression model to quantitatively assess the explanatory power of seven driving factors. The study found that since the 21st century (1) changes in Heat Index (HI) have varied significantly worldwide, with the majority of regions witnessing an increase, particularly at higher latitudes. The largest HI-increasing area was observed in the second quarter (S2), while the overall HI increase peaked in the third quarter (S3); (2) except for the decreasing area of none-risk regions, the regions under all other risk levels expanded (the proportion of high-risk areas in the world increased from 2.97% to 3.69% in S2, and from 0.04% to 0.35% in the fourth quarter (S4)); (3) aspect demonstrated the greatest explanatory power for the global heatwave distribution (0.69–0.76), followed by land-use coverage (LUCC, 0.48–0.55) and precipitation (0.16–0.43), while the explanatory power of slope and nighttime light (NTL) was rather low; (4) over the years, the explanatory power of each factor for heatwave distribution underwent a minor decrease without significant trend, but exhibited seasonal periodicity. Climatic factors and LUCC were most impactful in the first quarter (S1), while DEM and other human factors dominated in S2; and (5) interaction factors showed no significant trends over the years, but the explanatory power of DEM and slope increased notably when interacting with climate factor, aspect, and LUCC, respectively. The interactions between the aspect and LUCC with precipitation yielded the highest explanatory power (above 0.85) across all interactions. To effectively tackle heatwave risks, it is suggested to concentrate on high-latitude regions, reinforce land use and urban planning with eco-friendly strategies, scrutinize the interplay between precipitation and heatwaves, capitalize on topographic data for devising well-informed prevention measures, and tailor response strategies to accommodate seasonal fluctuations. This study offers valuable insights for enhancing climate change adaptation, disaster prevention, and mitigation strategies, ultimately contributing to the alleviation of extreme heatwaves and risk reduction.