Combining social media data and meteorological sensors for urban flood detection: a statistical analysis in São Paulo City

Abstract

Floods are among the most frequent and costly natural disasters in urban areas, often resulting from intense precipitation. Leveraging geospatial data from social media and physical sensors offers a valuable opportunity for effective flood detection. This study conducts a statistical analysis employing Anderson-Darling and Shapiro-Wilk tests to assess the normality of the data distributions. Correlation analyses were conducted to evaluate the relationships between rainfall levels, river levels, and Twitter (currently X), while the Mann-Whitney U test was used to compare data from flood and non-flood events. Meteorological variables, such as rainfall data from rain gauges and radar, proved critical in establishing a link between precipitation levels and flooding events. River level data from the São Paulo Flood Alert System revealed a strong correlation between river levels and flood conditions, particularly during “Warning” and “Emergency” situations. Additionally, the analysis of social media data demonstrated a significant correlation between the frequency of flood-related keywords in tweets and the occurrence of actual flood events. This finding highlights the potential of Twitter data as an alternative source for urban flood detection. By leveraging real-time, user-generated content, this approach offers a novel methodology for early warning systems, enhancing situational awareness and improving flood monitoring capabilities. The findings underscore the effectiveness of integrating multiple data sources for comprehensive flood monitoring, offering practical insights for improving flood detection and management in urban environments.