A system for effectively forecasting flash floods of the Arno River (Tuscany, Italy) should provide a flood warning with 10-12 h of lead time, primarily in order to evacuate the city of Florence. This goal may be achieved by acquiring and processing meteorological and hydrological data in realtime and, accordingly, by releasing alarms at different levels of reliability and concern. Through the application of both procedural language and expert system techniques, a prototype was developed which can readily handle a variety of relevant information and make predictions on flood hazard in Florence. The system was fairly successfully tested by processing simple meteorological data which enable a 24 hour forewarning to be released.

No abstract