Towards a combination of web-text and earth observation data for crisis and disaster risk reanalysis An explorative study on a flooding event

Kurzfassung

During and after natural disasters, all relevant information about a given situation, both in real time and in the post-analysis phase, is crucial for search and rescue operations, relief efforts, recovery and future mitigation and prevention measures. Systematically analyzing web text data from various sources such as news articles, social media and general websites can provide valuable insights for disaster monitoring and analysis when available in near real-time. However, it remains a challenge to capture large amounts of web text data in a timely manner and extract meaningful insights from this unstructured data to support disaster relief efforts. This study focuses on extracting information from large amounts of web text data relevant to crisis and disaster monitoring and synergistically analyzing it with other sources of Earth Observation data of the same event, especially satellite imagery. Satellite imagery has proven to be a valuable source of information for disaster monitoring and mitigation, e.g. in rapid mapping for disaster relief. However, Earth Observation data can have limitations, such as temporal availability, limitation of flight campaigns due to poor weather conditions, etc. The aim of this study is to investigate the potential of fusing web text data and satellite imagery to create a better awareness of the situation of large-scale flooding. To achieve this goal, we conduct an exploratory study focusing on large-scale flooding events for which Earth Observation data and web text data like online media and news reports are available. Specifically, it is about the reanalysis of a flood event that occurred in June 2021 in the Ahr valley in Germany and led to significant damage and human and economic losses. In this use case, information on a specific disaster situation is extracted from web text data such as the Open Web Index of the OpenWebSearch.eu project, carefully georeferenced and combined with satellite-based Earth Observation data for joint visualization and analysis. Using a combination of different natural language processing methods for disaster risk management related information is extracted, analyzed and filtered from the web-text data. In combination with fine-grained geoparsing approaches, the text-based information is further processed to link it with Earth Observation data in space and time. Earth Observation data used in this study includes flood maps, inundation maps or information on building damage. Once the systematic derivation of web-text information can be used to identify hotspots and aims to be operationalized in a routine and timely manner, e.g. through a publicly available European Open Web Index, such information extracted can be used for emergency response and e.g. for the planning of satellite or airborne emergency mapping campaigns. Beyond the acute phase of disaster response, crisis and disaster-related information derived from web texts can be used to enrich and contextualize satellite-based disaster management and mitigation information in other phases of a disaster risk management cycle. With continued research and development, this can lead to a more robust and efficient disaster management system that effectively utilizes both sources.