Understanding the impact of precipitation and model uncertainties on extreme flood estimates

Dealing with large uncertainties associated with estimates of extreme floods is a major challenge for risk assessment and mitigation. It is important to understand and quantify the potential sources of these uncertainties to reduce risk and support cost-effective and safe infrastructure design.

In this study, we employ a framework based on a hydrometeorological modeling chain with long continuous simulations to estimate extreme floods (Viviroli et al., 2022). The first element of the modeling chain is the multi-site stochastic weather generator GWEX, which focuses on intense precipitation events. GWEX generates long scenarios that force a bucket-type hydrological model (HBV), which simulates discharge time series. Lastly, a hydrologic routing model (RS Minerve) implements simplified representations of river channel hydraulics, floodplain inundations and regulated lakes.

The main objective of this contribution is to quantify the uncertainty arising from the weather generator and the hydrological model at different return levels, as these two factors are highly relevant for hydrological extremes. To this end, we employ two weather generator parameterizations: the first one is the default parameterization, which serves as a benchmark, whereas specific parameters are conditioned on weather types in the second one. Then, two hydrological model configurations with different response functions are utilized. Varying these elements of the modeling chain allows us to understand their impact on the extreme flood estimates by interpreting the resulting variability as uncertainty. We run our simulations for three representative HBV-model parameter sets to account for model parameter uncertainty. This modeling framework is applied to nine large catchments (> 450 km²) located in different regions of Switzerland to consider the influence of catchment characteristics. The last step of our methodology includes the decomposition of uncertainty in extreme flood estimates using an analysis of variance (ANOVA).

Our results suggest that the contributions of different sources of uncertainty vary between the catchments. The dominant source of uncertainty may vary for different return periods ranging from 1 to 1000 years. These results highlight the challenge of generalizing a priori about the importance of the selected components contributing to the total uncertainty at the catchment scale, as physiographic catchment characteristics play a key role. Overall, this study sheds light on the role of uncertainties in a hydrometeorological modeling chain and will serve as a basis for follow-up studies related to hazard assessment, safety planning, and hydraulic engineering projects.

 

Reference:

Viviroli D, Sikorska-Senoner AE, Evin G, Staudinger M, Kauzlaric M, Chardon J, Favre AC, Hingray B, Nicolet G, Raynaud D, Seibert J, Weingartner R, Whealton C, 2022. Comprehensive space-time hydrometeorological simulations for estimating very rare floods at multiple sites in a large river basin. Natural Hazards and Earth System Sciences, 22(9), 2891–2920, doi:10.5194/nhess-22-2891-2022