Ensemble nowcasting system for flash floods alerts based on radar data

In several regions of the world, flash floods are responsible for causing major catastrophes with high socio-economic damages. These events are mainly caused by atmospheric phenomena of convective origin. These systems are characterized by high spatial-temporal variability of precipitation. Nowcasting alert systems are essential to anticipate these events and reduce their impacts. These systems are based on nowcasting precipitation, which are generated, mainly, from precipitation observation. In this case, the radar data can help to develop these predictions using the information of spatial-time variability of the rainfall. The warnings provided by a forecasting system have several uncertainties. To consider these uncertainties an often used method is the generation of rainfall ensembles. The aim of this research is to develop a system able to generate probabilistic precipitation nowcasting using a geostatistical method based on conditional simulation approach. These predictions are applied in a hydrological system to generate flash flood alerts. The method is implemented in the region of Campinas city, Brazil, considering the information of the radar recently installed in Campinas. The results of this research contributed to develop an ensemble nowcasting system that presents good accuracy, reliability and events discrimination. This approach can be helpful to enhance flash floods alerts systems.