Propagation of hydro-meteorological uncertainty in a model cascade framework to inundation prediction

Abstract

Abstract. This investigation aims to study the propagation of meteorological uncertainty within a cascade modelling approach to flood prediction. The methodology was comprised of a numerical weather prediction (NWP) model, a distributed rainfall-runoff model and a 2-D hydrodynamic model. The uncertainty evaluation was carried out at the meteorological and hydrological levels of the model chain, which enabled the investigation of how errors that originated in the rainfall prediction interact at a catchment level and propagate to an estimated inundation area and depth. For this, a hindcast scenario is utilised removing non-behavioural ensemble members at each stage, based on the fit with observed data. At the hydrodynamic level, an uncertainty assessment was not incorporated; instead, the model was setup following guidelines for the best possible representation of the case study. The selected extreme event corresponds to a flood that took place in the southeast of Mexico during November 2009, for which field data (e.g. rain gauges; discharge) and satellite imagery were available. Uncertainty in the meteorological model was estimated by means of a multiphysics ensemble technique, which is designed to represent errors from our limited knowledge of the processes generating precipitation. In the hydrological model, a multi-response validation was implemented through the definition of six sets of plausible parameters from past flood events. Precipitation fields from the meteorological model were employed as input in a distributed hydrological model, and resulting flood hydrographs were used as forcing conditions in the 2-D hydrodynamic model. The evolution of skill within the model cascade shows a complex aggregation of errors between models, suggesting that in valley-filling events hydro-meteorological uncertainty has a larger effect on inundation depths than that observed in estimated flood inundation extents.

Panta Rhei 2013–2015: global perspectives on hydrology, society and change

“…Water governance measures for flood risk reduction are typically designed to ensure both better flood management and an increase in infrastructure resilience. However, the assessment of hydro-meteorological risk must take into account uncertainty (Rodríguez-Rincón et al 2015). Numerical tools and models that represent reality in an incomplete manner incorporate errors that can interact and aggregate to compromise prediction reliability.…”

Challenges, Opportunities, and Pitfalls for Global Coupled Hydrologic‐Hydraulic Modeling of Floods

“…Such a rigorous assessment requires identification of all the sources of uncertainty within the modeling cascade and results in a large number of model runs with extremely high computational costs, especially when using distributed models (e.g., Pappenberger et al, 2005). Simplified approaches designed to reduce the computational burden of uncertainty estimation were applied, for instance, by Pappenberger et al (2005), McMillan and Brasington (2008), and Rodríguez-Rincón et al (2015) for event-based analysis; by Falter et al (2016) in a continuous modeling approach; and by Sampson et al (2014) and Zischg et al (2018) in a flood damage assessment model.…”

Simulación de avenidas mediante un modelo hidráulico/hidrológico distribuido en un tramo urbano del río Ginel (Fuentes de Ebro)

“…Los resultados presentados en esta sección no se pueden comparar con ningún evento real, ya que se carece de datos sobre avenidas de similares características en la región de estudio. Por otro lado, sería interesante realizar un análisis de incertidumbre de los resultados numéricos [30] para el cual, resulta especialmente útil la implementación del modelo en GPU. Esto permite ahorrar un tiempo de cálculo considerable en las numerosas repeticiones que se requieren en el análisis.…”