Nonparametric prediction of the onset of regional floods: floods in north-western Bohemia, Czech Republic, 2010

Currently, ongoing global climate change brings, among other things, an increasing frequency of extreme weather events such as heavy rains, which can cause flash floods. Responsible authorities have tried to develop systems for early warning of such events. Such systems already exist in the US and in some European countries. They often rely on the prediction of extreme rainfall, possibly with the use of weather radar data, as well as rainfall-runoff models. The weakness in these systems, which limits their global usage, is based on the precise use of rainfall-runoff models and the attempt to quantify the impacts of extreme rainfall in the affected area. Therefore, we have developed a methodology based on the simplified data inputs (data from weather radar) that release a warning for potentially vulnerable areas in the longest time possible before extreme rainfall effects are due to occur. Our ambition is not to quantify these effects. Due to the short time interval between downpours and flash floods caused by them, we do not consider this information to be significant. We decided to test our methodology inter alia on a case of regional flooding, which was the result of regional precipitations combined with extreme local rains. The results presented in this paper show that, even in this situation, the proposed methodology allows us to provide an early warning for the population to take refuge in a safe site.