Twitter - a new citizen science solution for urban flood database #urban floods #flood database

Abstract

Abstract High spatial heterogeneous behavior of urban floods offers challenges in its monitoring, modeling and assessment, and lack of flood database further increases complexity. This study proposes a twitter-based framework to develop a flood database. Three flood events differing in rainfall characteristics are selected, and event-specific flood inventories are developed using the proposed framework. The inventory was validated with newspaper-based information. Analysis of tweets imply a broad agreement among both twitter- and newspaper- based flooding instances, however, highlight the need to include identifiable location information in tweets. Further, the results suggest floating and using of hashtags, and importantly active participation of individuals and various agencies aids in the development of reliable and accurate flood inventory of decent size. Integration of artificial intelligence and machine learning techniques allows to develop cost-effective flood inventories of finer space-time resolutions with minimized manual efforts for many cities around the world; the developed flood inventories make many urban areas data rich; thus it decreases complexity of urban flood relevant challenges at least by one important dimension, thereby plays a key role in modeling and mitigation efforts.