Spatiotemporal Distribution of Cross-Platform Public Opinion in the 2023 Dezhou Earthquake: Implications for Disaster-Resilient Emergency Management

Social media platforms have emerged as a critical infrastructure for disaster communication and emergency management. However, how public opinion varies across platforms during earthquake events and how such differences can inform resilient disaster strategies remain underexplored. This study analyzes public opinion responses to the 2023 M5.5 Dezhou earthquake across three major Chinese social media platforms—Sina Weibo, Bilibili, and Douyin—based on 28,557 posts. By combining Latent Dirichlet Allocation (LDA), Word2Vec, and Convolutional Neural Networks (CNNs), we examine the temporal, spatial, thematic, and emotional patterns of public discourse. The results show (1) a bimodal public attention pattern within 24 h of the earthquake, with platform-specific response timings; (2) spatial clustering of public concern in the epicenter (Shandong) and historically high-risk regions (Sichuan–Chongqing); (3) differentiated topic preferences reflecting platform functions—emotional expression (Weibo), science popularization (Bilibili), and real-time impact sharing (Douyin); and (4) a predominance of positive/neutral sentiment, influenced by user demographics and algorithmic content curation. This study proposes a resilience-oriented public opinion analysis framework aligned with the disaster lifecycle and offers recommendations for platform-specific risk communication, psychological support, and policy planning. Findings contribute to digital disaster governance and the integration of social media analytics into sustainable emergency management.