Extracting public opinion on typhoon disasters in China: a sina weibo case study of landfalling typhoon Muifa (2022)

Abstract

Abstract This study investigates public opinion dynamics on Sina Weibo during Typhoon Muifa (2022), which made four landfalls in China. Using a dataset of 19,417 microblog posts, we employed Latent Dirichlet Allocation (LDA) topic modeling, sentiment analysis, and correlation statistics to characterize the evolution of public attention and discourse alongside the typhoon’s activity. Results identified four dominant discussion topic categories: typhoon impact, weather conditions, meteorological information, and disaster response. Personal accounts predominantly contributed to the first two categories, while official accounts dominated discussions on the latter two. A strong positive correlation emerged between daily total precipitation and the number of microblog counts (R 2  = 0.84, q < 0.001), which was particularly pronounced in forecasted landfall provinces Zhejiang, Shanghai, Shandong, and Liaoning (q < 0.05). Negative sentiment was highly correlated with rising precipitation, a trend largely driven by discussions within the typhoon impact topic category. Our findings underscore the potential of social media as a real-time indicator of localized public sentiment during disasters, with official risk narratives playing a key role in shaping attention. This study offers insights that may inform targeted risk communication and emergency management strategies.