Role of Social Media in Real-Time Flood Early Warning Systems: Data Regression using Data Mining Models

Flood early warning systems, predominantly foundered on data-driven models using techniques such as image processing, deep learning, or recurrent neural networks, are gaining popularity and applicability 1. This trend is particularly evident due to advancements in computational capabilities and the increased availability of quantified data. However, these systems heavily rely on hydrological and hydraulic data, which can contain uncertainties and may not always be accessible in real-time 2. To address these limitations, there is a growing recognition of the importance of incorporating a new qualitative data source within these frameworks especially stakeholders' opinions. These opinions are often expressed and documented on social media platforms, news websites, and various online forums 3. This becomes especially crucial when other sources of real-time data may not be fully accessible or available within the scope of the disaster monitoring system. Furthermore, integrating social media and online platforms into flood early warning systems may serve as a valuable supplement, offering insights and information that may not be captured through conventional hydrological and hydraulic data sources 4. In the present study, relationship between qualitative social media databases and flood disaster announcements is investigated to reveal the potential applications of these new data sources. To achieve this, a case study focusing on the Waikato region in New Zealand is conducted in which news and tweets from X platform spanning the last decade are collected with centring content of occurrences of floods and their impacts. Through the application of data mining models, the correlation between reported flooding incidents, encompassing coastal, pluvial, and fluvial flooding, and the information derived from social media are scrutinised. Research findings revealed a significant correlation, indicating a robust relationship between the information gathered from social media and the incidence of floods. This underscores the vital role of social media in facilitating information dissemination, serving as crowdsourced data, enhancing community engagement, contributing to public awareness, and assisting in the verification of information. While these results offer valuable insights for the development of a community-based early warning systems, it's essential to recognise that this represents an early stage in data-driven modelling. Considered as a foundation, this research sets the stage for future directions, paving the way for more sophisticated and nuanced community-centric early warning systems. References 1 Piadeh, F., Behzadian, K., Chen, A.S., Kapelan, Z., Rizzuto, J., Campos, L.C. (2023). Enhancing urban flood forecasting in drainage systems using dynamic ensemble-based data mining. Water Research, 247, p.120791. 2 Piadeh, F., Behzadian, K., Chen, A.S., Campos, L.C., Rizzuto, J., Kapelan, Z. (2023). Event-based decision support algorithm for real-time flood forecasting in urban drainage systems using machine learning modelling. Environmental Modelling Software, 167, p.105772. 3 Odiase, O., Wilkinson, S., Neef, A. (2020). South African community in Auckland: Natural hazards and the resilience to a potential disaster event. Disaster Prevention and Management: An International Journal, 29(3), pp. 390404. 4 Fathianpour, A., Wilkinson, S., Jelodar, M.. Evans, B. (2023). Reducing the vulnerability of tourists to tsunami: challenges for decision-makers. Natural Hazards, 118(2), pp. 13151339.