Case-Based Reasoning for the Prediction of Flash Flood

Abstract

The term flash-flood refers to the sudden raise in the water levels in a basin due to an abrupt change in the weather conditions. Early detection of flash-floods reduces the harm that they can produce in the infrastructure or even in preventing human losses. Up to now, the studies focus on the dynamics of the basins, determining how the water levels would be in a considered scenario. However, nothing have been done concerning the online prediction of flash-floods. This research focuses on this topic, proposing a Case-Based Reasoning tool to cope with the estimation of the water levels on a basin based on the current basin conditions and the weather forecast. Furthermore, this CBR tool has been designed to work in different basins provided enough data is available, either from past experiences or from simulation. This research is being designed, developed on two real basins, one from Spain and one from France; however, the experimentation has only been addressed with realistic data from the Venero-Claro basin in Spain. Expectancy is that the performance of the CBR tool will perfectly mimic the decision making of the public safety experts.