Ensemble flood forecasting considering dominant runoff processes: I. Setup and application to nested basins (Emme, Switzerland)

Abstract

Abstract. Flash floods (FFs) evolve rapidly during and after heavy precipitation events and represent a risk for society. To predict the timing and magnitude of a peak runoff, it is common to couple meteorological and hydrological models in a forecasting chain. However, hydrological models rely on strong simplifying assumptions and hence need to be calibrated. This makes their application difficult in catchments where no direct observation of runoff is available.To address this gap, a FF forecasting chain is presented based on: (i) a nowcasting product which combines radar and 5 rain gauge rainfall data (CombiPrecip), (ii) meteorological data from state-of-the-art numerical weather prediction models (COSMO-1, COSMO-E), (iii) operationally available soil moisture estimations from the PREVAH hydrological model, and (iv) a process-based runoff generation module with no need for calibration (RGM-PRO). This last component uses information on the spatial distribution of dominant runoff processes from the so-called maps of runoff types (RTs), which can be derived with different mapping approaches with increasing involvement of expert knowledge. RGM-PRO is then parametrised a priori 10 based on the results of sprinkling experiments.This prediction chain has been evaluated using data from April to September 2016 in the Emme catchment, a medium-size FF prone basin in the Swiss Prealps. Two novel forecasting chains were set up with two different maps of RTs, which allowed sensitivity of the forecast performance on the mapping approaches to be analysed. Furthermore, special emphasis was placed on the predictive power of the new forecasting chains in nested subcatchments when compared with a prediction chain including 15 a conventional hydrological model relying on calibration.Results showed a low sensitivity of the predictive power on the amount of expert knowledge included for the mapping approach. The forecasting chain including a map of RTs with high involvement of expert knowledge did not guarantee more skill. In the larger basins of the Emme region, process-based forecasting chains revealed comparable skill as a prediction system including a conventional hydrological model. In the small nested subcatchments, the process-based forecasting chains 20 outperformed the conventional system, however, no forecasting chain showed satisfying skill.The outcomes of this study show that operational FF predictions in ungauged basins can benefit from the use of information on runoff processes, as no long-term runoff measurements are needed for calibration.

Ensemble flood forecasting considering dominant runoff processes: II. Benchmark against a state-of-the-art model-chain (Verzasca, Switzerland)

“…Quantitative FF forecasts are needed and can be provided by coupling meteorological and hydrological models. In order to expand FF prediction to ungauged catchments, introduced RGM-PRO, a new hydrological module with process-based runoff generation and no need for calibration (Antonetti et al, 2018). In this study, FF forecasting chains are set up using high-end model components, namely the RGM-PRO runoff generation module, atmospheric models COSMO-1 and COSMO-E and CombiPrecip nowcasting product for the Verzasca region in southern Switzerland.…”

“…RGM-PRO requires hourly gridded precipitation input and runs at a spatial resolution of 500 m. A more detailed description of the hydrological models mentioned here can be found in Viviroli et al (2009), in the companion paper Antonetti et al (2018) or in .…”

“…Following the setup presented in the companion paper (Antonetti et al, 2018), CombiPrecip (Sideris et al, 2014) has been used as high resolution hourly gridded precipitation product. CombiPrecip was used here only for forcing RGM-PRO.…”

“…Furthermore, events earlier than May are usually affected by meltwater and therefore could not be simulated with RGM-PRO. Please note that the availability of runoff data was different for the Verzasca and the Emme region (the latter is investigated in the companion paper (Antonetti et al, 2018)). At each alert date, which are explained in Sect.…”

“…In the companion paper (Antonetti et al, 2018), the potential of RGM-PRO in FF forecasting was assessed for the Emme basin in the Swiss Prealps. Furthermore, the sensitivity of the predictive power to various approaches for including knowledge on runoff processes into the hydrological model, i.e.…”