Flood Alert System Using High-Resolution Radar Rainfall Data: Nowcasting Model Validation 1 , 2

Abstract

An analysis for the nowcasting requires a combination of meteorological and hydrological statistics, as this permits a better understanding of behavior of the spatial and temporal accuracy of storm prediction. A good nowcasting include accuracy of the spatial, as well as in the temporal level and accuracy of the predicted rainfall intensity. The accuracy of rainfall prediction of each pixel can be measured by decomposing the rainfall process into sequences of discrete and continuous random variables, that is, the presence or absence of rainfall events and rainfall intensity. Examples of quantitative parameters used in the research include: Contingency table, Mean square Error, Root Mean Square Error, Bias Ratio and Mean Absolute Error. The joint distribution of the forecast and observations has fundamental interest with respect to the verification of forecasts. In the most practical setting, both the forecast and observations are discrete variables.