Multi-model ensembles for assessment of flood losses and associated uncertainty

Abstract

Abstract. Flood loss modelling is a crucial part of risk assessments. However, it is subject to large uncertainty that is often neglected. Most models available in the literature are deterministic, providing only single point estimates of flood loss, and large disparities tend to exist among them. Adopting any one such model in a risk assessment context is likely to lead to inaccurate loss estimates and sub-optimal decision-making. In this paper, we propose the use of multi-model ensembles to address these issues. This approach, which has been applied successfully in other scientific fields, is based on the combination of different model outputs with the aim of improving the skill and usefulness of predictions. We first propose a model rating framework to support ensemble construction, based on a probability tree of model properties, which establishes relative degrees of belief between candidate models. Using 20 flood loss models in two test cases, we then construct numerous multi-model ensembles, based both on the rating framework and on a stochastic method, differing in terms of participating members, ensemble size and model weights. We evaluate the performance of ensemble means, as well as their probabilistic skill and reliability. Our results demonstrate that well-designed multi-model ensembles represent a pragmatic approach to consistently obtain more accurate flood loss estimates and reliable probability distributions of model uncertainty.

Quantifying Flood Vulnerability Reduction via Private Precaution

“…They have the advantage of inherently quantifying uncertainty associated with the loss estimation. Thus, a variety of models with varying complexities and working concepts are currently available, and it is not trivial to decide which one to use for a specific application (Apel et al, 2009;de Moel et al, 2015;Figueiredo et al, 2018). Several studies have tested and compared various flood loss models in respect to their predictive accuracy and reliability (e.g., Cammerer et al, 2013;Gerl et al, 2016;Hasanzadeh Nafari et al, 2016;Jongman et al, 2012).…”

Seamless Estimation of Hydrometeorological Risk Across Spatial Scales

“…Despite the progress in damage estimation in recent years (Figueiredo et al, 2018;Kreibich et al, 2016;Schröter et al, 2014;Sieg et al, 2017;Wagenaar et al, 2017), the available damage models are not able to describe the damage process reliably, and the process is stochastic to a substantial extent. Therefore, as implemented in this approach, the distribution of the damage estimate should be assessed at every spatial scale in order to report the lack of knowledge about the damage estimates.…”

From Hazard to Consequences: Evaluation of Direct and Indirect Impacts of Flooding Along the Emilia-Romagna Coastline, Italy

“…to the final damage estimations. It was demonstrated that the uncertainty related to this component represents a dominant factor affecting the overall uncertainty of damage assessments (De Moel and Aerts, 2011), especially when damage calculations are applied in deterministic ways (Figueiredo et al, 2018). Additionally, damage models are generally site-dependent.…”

“…The type of flood hazard and related damage models are also important, as marine floods are different from fluvial ones in terms of consequences to buildings (e.g., salinity, short-period waves). These aspects contribute to lower the predictive skills of damage models (Figueiredo et al, 2018). The damage curve used in this study (adapted from Scorzini and Frank, 2017) was developed for residential buildings, for a fluvial flood occurred in Veneto in 2010.…”

“…Additionally, the curve is also applied to commercial activities, since a specific model was not available. As deterministic approaches based on flood damage curves showed limitations and low predictive skills, recent approaches are moving toward probabilistic (e.g., Dottori et al, 2016) or model-ensemble approaches (Figueiredo et al, 2018). The damage is, therefore, calculated using different damage models in order to produce a distribution of possible damages and, at the same time, to quantify the uncertainty of the outcomes.…”