Files
Abstract
Urban flooding threatens infrastructure, public safety, and economic stability, with increasing frequency due to climate change and urbanization. Traditional monitoring methods - sensors, models, and remote sensing - are effective but limited by cost, time delays, and low spatial resolution. This thesis explores Twitter as a complementary data source for urban flood monitoring. A framework was developed to collect, filter, and analyze flood-related tweets using natural language processing, machine learning, sentiment analysis, and geocoding. Social media data was then integrated with rainfall data to generate near real-time flood maps. Additionally, a rain-on-mesh simulation using HEC-RAS incorporated terrain, land cover, and soil data to validate results. Findings show that approximately 75% of flood-affected zones identified via Twitter matched those from model-generated inundation maps. This demonstrates that social media can enhance situational awareness and support rapid flood response, making it a valuable tool for supporting traditional urban flood monitoring systems.