Bayesian rainfall thresholds for flash flood guidance

The aim of this work is to determine the FFG rainfall depth and compute FFG-values based on the minimization of a Bayesian Loss Function of the discharge conditional upon the state of saturation of the catchment. Rainfall thresholds are here defined as the cumulated volume of rainfall during a storm event which can generate a critical water stage (or discharge) at a specific river section. When the rainfall threshold value is exceeded, the likelihood that the critical river level (or discharge) will be reached is high and consequently it becomes appropriate to issue a flood alert; alternatively, no flood alert is going to be issued when the threshold level is not reached. In other words the rainfall thresholds must incorporate a “convenient” dependence between the cumulated rainfall volume during the storm duration and the possible consequences on the water level or discharge in a river section. The term “convenient” is here used according to the meaning of the decision theory under uncertainty conditions, namely the decision which corresponds to the minimum (or the maximum) expected value of a Bayesian cost utility function. There are described two possible approaches for the same methodology: (a) using the Monte-Carlo simulations or (2) using the Normal Quantile Transform. The main difference of the two is the requirements in terms of data, i.e. the timeseries of rainfall and discharge. Application of the methodology and comparison with other methodologies are provided for the Posina catchment in Italy.