Social media services such as Twitter have become an important channel for reporting real-world events. For example, they can describe the current situation during a disaster. The decisions in crises management are based on detailed on-site information such as what is happening, where and when an event is happening, and who is involved. Thus, in real applications, monitoring the events over social media will enable to analyse the current overall situation. In this paper, the authors introduce a prototype for real-time Twitter-based natural disaster detection and monitoring. The detection approach is multilingual and calculates a statistical based probability for a potential disaster event. For an automatic geo-referencing of the disaster, the approach applies spatial gridding. On this basis the grid cells are subject to a spatial-thematic clustering which uses a method similar to region growing. The application’s output is an automatically generated email alert, containing specific information on the disaster.