Scholarlink

<p>To address the limitations of traditional urban flood monitoring methods, including insufficient spatio-temporal coverage, response delays, and potential failure during extreme weather. This study developed an integrated framework combining multimodal data collection, feature extraction, and spatial correlation analysis based on social media. The framework first employs natural language processing to extract flood event locations and contextual information from social media, then utilizes the Mask R-CNN model to identify water depths in user-submitted images. Subsequently, sentiment analysis and word cloud techniques are employed to characterize public emotional fluctuations and focal points during disasters. Finally, extracted flood features are spatially overlaid and statistically analyzed with geographic data (such as road network and land use types) on a GIS platform, systematically revealing the distribution patterns of inundation points and their driving mechanisms. Using Zhengzhou “7.20” torrential rainfall event as a case study, the research findings indicate: (1) Social media provides critical flood event information with high spatio-temporal resolution, offering real-time and broad coverage advantages unattainable by traditional monitoring methods. It effectively fills gaps in fixed monitoring networks while reflecting public sentiment dynamics. (2) Flood locations extracted from social media exhibit high accuracy. Within 50&#xa0;m buffer zone, social media data covered 85.9% of measured flood points, though coverage is constrained by public activity patterns and observation conditions. (3) Flood inundation points exhibit distinct non-random spatial clustering patterns closely linked to the built urban environment. The study confirms that, in scenarios where traditional monitoring data is limited, an analytical framework leveraging multi-source social media information can provide effective data sources for real-time perception, mechanism analysis, and decision support in urban flooding. This enhances the resilience and information assurance capabilities of urban disaster prevention systems.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Urban flood information extraction and spatial-temporal analysis of “7.20” Zhengzhou flood risk from social media perspective

  • 期刊论文

  • 发表日期：

    2026年2月26日

查看全文
  • Ronghu Miao,
  • Guoru Huang,
  • Zhiwei Chen,
  • Mingyu OuYang

摘要

To address the limitations of traditional urban flood monitoring methods, including insufficient spatio-temporal coverage, response delays, and potential failure during extreme weather. This study developed an integrated framework combining multimodal data collection, feature extraction, and spatial correlation analysis based on social media. The framework first employs natural language processing to extract flood event locations and contextual information from social media, then utilizes the Mask R-CNN model to identify water depths in user-submitted images. Subsequently, sentiment analysis and word cloud techniques are employed to characterize public emotional fluctuations and focal points during disasters. Finally, extracted flood features are spatially overlaid and statistically analyzed with geographic data (such as road network and land use types) on a GIS platform, systematically revealing the distribution patterns of inundation points and their driving mechanisms. Using Zhengzhou “7.20” torrential rainfall event as a case study, the research findings indicate: (1) Social media provides critical flood event information with high spatio-temporal resolution, offering real-time and broad coverage advantages unattainable by traditional monitoring methods. It effectively fills gaps in fixed monitoring networks while reflecting public sentiment dynamics. (2) Flood locations extracted from social media exhibit high accuracy. Within 50 m buffer zone, social media data covered 85.9% of measured flood points, though coverage is constrained by public activity patterns and observation conditions. (3) Flood inundation points exhibit distinct non-random spatial clustering patterns closely linked to the built urban environment. The study confirms that, in scenarios where traditional monitoring data is limited, an analytical framework leveraging multi-source social media information can provide effective data sources for real-time perception, mechanism analysis, and decision support in urban flooding. This enhances the resilience and information assurance capabilities of urban disaster prevention systems.