Estimating Ensemble Flood Forecasts’ Uncertainty: A Novel “Peak-Box” Approach for Detecting Multiple Peak-Flow Events

Antonio Giordani,Massimiliano Zappa,Mathias W Rotach

doi:10.3390/atmos11010002

Antonio Giordani, Massimiliano Zappa + Show 1 more

Open Access

https://doi.org/10.3390/atmos11010002

Copy DOI

Abstract

Ensemble flood forecasts are an established tool to provide information about the uncertainty of runoff predictions. However, their interpretation may not be straightforward, especially when dealing with extreme events; therefore, the development of new tools to enhance their understanding and visualization is necessary. Recently, the so-called “peak-box” approach has been developed to help decision makers in the interpretation and verification of peak-flow forecasts, receiving positive feedbacks within the hydrological community. However, this method has proven to be limited when multiple peak-flow events occur within the forecast, being unable to separate close discharge peaks. The aim of this paper is then to develop a new algorithm designed to accomplish this task. To do so, we consider runoff probabilistic forecasts obtained with a coupled hydrometeorological flood forecasting system formed by the high resolution meteorological Ensemble model COSMO-E and the hydrological model PREVAH, for the small Verzasca basin, Switzerland, during October and November 2018. The application of this new method, despite the limitation given by the small sample size considered in this study, indicates a successful implementation: the new algorithm is able to distinguish among different events and to provide sharper and more skillful forecasts, and its verification yields slightly better timing estimations compared to the former approach.

Highlights

The prediction of hazardous floods triggered by severe precipitation events is an important issue (e.g., [1]); especially for the Alpine region where the most severe events in Europe usually take place [2], the forecasting value has been increasing
Ensemble flood forecasts are affected by many sources of uncertainty [11,12], and the most important one is related to the numerical weather prediction (NWP) forecasts’ input [13,14,15,16]
The flood forecasting system considered for this study is formed by:

Summary

Introduction

The prediction of hazardous floods triggered by severe precipitation events is an important issue (e.g., [1]); especially for the Alpine region where the most severe events in Europe usually take place [2], the forecasting value has been increasing. Atmosphere 2020, 11, 2 for hydrological purposes, HEPS) These models provide an ensemble of river flow predictions for the same forecast period, probabilistically assessing future river conditions [6], and are currently widely applied to obtain hydrological forecasts (e.g., [7,8,9]). Ensemble flood forecasts are affected by many sources of uncertainty [11,12], and the most important one is related to the numerical weather prediction (NWP) forecasts’ input (especially regarding precipitation estimations) [13,14,15,16]. With the high-resolution forcing meteorological model we used, having an horizontal grid spacing of 2.2 km, the uncertainty related to the meteorological forcing was found to be five times larger than the hydrological parameters’

Objectives

Methods

Results

Discussion

Conclusion