AI-based Image & Text Analysis and Citizen Sensing to Improve Warnings for Extreme Weather Events and their Impacts

Important results from the project

The project aimed to evaluate the potential of combining AI-based image processing and text analysis with the impact-based weather warning system. In collaboration between experts at SMHI, the County Board in Östergötland and Linköping University, an AI-based pipeline with several components was developed. These were tested with regional and national actors, and the project identified how the AI4CA pipeline could be integrated with existing systems. We also evaluated the possibilities for integration of climate adaptation aspects with the weather warning system.

Expected long term effects

Project results include: · An AI-based system for the identification of flood events through analysis of social media posts and news media. · A visual interface that integrates information about flood events, their spatiotemporal characteristics and consequences. · An application for mobile phones for the collection of annotated image & text reporting in case of floods. · An analysis of different areas of use for the system, as well as how the system can be integrated into existing structures and bridge the gap between crisis preparedness and long-term climate adaptation in the future.

Approach and implementation

The project developed and tested a system integrating AI-based techniques for text and image classification to identify flood events through analysis of social media posts and news media to study how these can support SMHI´s impact-based warning system. The project also developed and tested an application for smartphones to collect annotated data on flood events. We conducted several interviews, workshops and user tests with various actors who contribute to or are affected by the new system.

External links

Webbsidan för projektet på www.liu.se