A survey of public perception of heat waves across China: A social media-based geospatial modeling approach

The widespread occurrence of heat waves is exacerbating public health risks. Understanding the region where the heatwave occurs, the population it affects, and the extent of its social attention are essential for heatwave preparedness and response. Previous studies have recognized heatwaves as a major health risk, but limited research on the public's perception of heatwaves using location-based social media data. This crowdsourcing approach remains a significant asset in understanding the public's perception of natural hazards, such as heatwaves, and in assessing the potential effects of heatwaves on local communities. Centering on examining the association between public attention and heatwaves, the present study collected social media (Weibo) data, weather data, and other geographical data from open-source platforms, then adopted a BERT-based language model to extract heat-related Weibo tweets from May to September 2024. Using a cumulative heatwave intensity index (HCI), this study investigates the spatiotemporal patterns of public attention to heatwaves and assesses the influence of meteorological, demographic, and socioeconomic variables. Results show that the number of heatwave-related tweets (NHT) peaked at 14:00 and 22:00, supporting the notion that heatwaves exacerbate the impact on human activities during nighttime. High NHT was primarily concentrated in densely populated regions of eastern and southern China, with fewer in the western regions. Clustering analysis using MGWR and K-means++ revealed regional variations in public sensitivity to heat waves. NHT is primarily influenced by socioeconomic and demographic attributes such as GDP per capita, population density, and penetration rate. In contrast, the heat-related tweets (RHTs) ratio is more dependent on environmental factors (e.g., HCI, precipitation, dew point temperature) and population density. Further research is needed to enhance understanding of heatwave perception among different populations. This study has the potential to provide insights into improving population resilience during heat waves using location-based social media data.