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

Overview

This study analyzed public discourse on Sina Weibo during Typhoon Muifa (2022) to characterize the temporal dynamics and thematic structure of social media engagement with a multi-landfall typhoon event in China. The research examined 19,417 microblog posts generated across the typhoon's lifecycle, treating social media as a quantifiable source of real-time public sentiment and attentional patterns correlated with meteorological intensity.

Methods and approach

The analysis employed three complementary computational approaches. Latent Dirichlet Allocation topic modeling was applied to extract dominant thematic categories from the corpus. Sentiment analysis classified posts according to emotional valence. Correlation statistics, including Pearson's R and q-value significance testing, quantified relationships between daily precipitation totals and post volume across affected provinces (Zhejiang, Shanghai, Shandong, Liaoning). The dataset was stratified by account type (personal versus official) to differentiate discourse patterns by source category.

Results

Four primary topic clusters emerged from the corpus: typhoon impacts, weather conditions, meteorological information dissemination, and disaster response activities. Personal accounts concentrated on impact-related and weather condition discussions, whereas official accounts dominated information and response narratives. Daily precipitation exhibited a strong positive correlation with daily post counts (R squared = 0.84, q < 0.001), with heightened effects in forecasted landfall provinces (q < 0.05). Negative sentiment demonstrated high correlation with precipitation increases, predominantly attributable to posts within the typhoon impact category. The correlation pattern indicates systematic attentional and affective responsiveness to meteorological intensity gradients.

Implications

The findings establish social media activity as a quantifiable proxy for localized public sentiment and attention intensity during acute meteorological disasters. The differential thematic contributions by personal and official accounts indicate distinct communicative roles: personal accounts function as sentiment indicators and impact reporters, while official accounts serve information distribution functions. These differentiated patterns suggest opportunities for tailored risk communication strategies that leverage account-type-specific affordances. The strong precipitation-sentiment correlation provides an empirical basis for real-time sentiment monitoring to assess public distress levels and identify geographic areas requiring enhanced communication or resource deployment.