Improving flash flood forecasting using a frequentist approach to identify rainfall thresholds for flash flood occurrence

Flash Flood Guidance (FFG) is a rainfall threshold which initiates flooding in streams. It merely provides a binary output (yes or no) which has large uncertainties in forecasting. In this paper, we propose a new method by combining FFG with the Frequentist method to present the probability of flash flood occurrence based on historical rainfall events. We first calculated deviation from the log transform rainfall data leading to flash floods. Kernel Density Estimation (KDE) was used to describe the deviation. Normal Distribution Function (NDF) was chosen to fit the KDE output and to calculate probabilities of flooding as per the Frequentist FFG. In order to aid decision making, three probability thresholds (10, 20 and 60%) were used for defining four flood risk classes, namely very low, low, significant and high, and were colour coded respectively as green, yellow, orange and red. The proposed Frequentist FFG method was then applied to the Posina River basin in Italy. Comparison of forecasts from the conventional FFG (with probability 0 or 1) and Frequentist FFG for 94 6-hourly rainfall events, including 23 flood events, shows that the Frequentist FFG presented a probability of flooding varying from 0 to 100% and the corresponding risk class can be used to reduce false alarms while still reducing the disaster risk. The application of the developed approach to the Posina basin shows that decision making regarding flash forecasting is easier with the presented approach compared to the traditional FFG approach.

Automated summary

This paper proposes a new method that combines historical rainfall events with Flash Flood Guidance (FFG) to provide the probability of flash flood occurrence. The method can reduce false alarms while lowering disaster risk.