Multihazard Scenarios for Analysis of Compound Extreme Events

Abstract

Compound extremes correspond to events with multiple concurrent or consecutive drivers (e.g., ocean and fluvial flooding, drought, and heat waves) leading to substantial impacts such as infrastructure failure. In many risk assessment and design applications, however, multihazard scenarios of extremes and compound events are ignored. In this paper, we review the existing multivariate design and hazard scenario concepts and introduce a novel copula‐based weighted average threshold scenario for an expected event with multiple drivers. The model can be used for obtaining multihazard design and risk assessment scenarios and their corresponding likelihoods. The proposed model offers uncertainty ranges of most likely compound hazards using Bayesian inference. We show that the uncertainty ranges of design quantiles might be large and may differ significantly from one copula model to the other. We also demonstrate that the choice of marginal and copula functions may profoundly impact the multihazard design values. A robust analysis should account for these uncertainties within and between multivariate models that translate into multihazard design quantiles.

Trends in Compound Flooding in Northwestern Europe During 1901–2014

“…Along the North Atlantic coast, the majority of high tide flooding occurs in response to both tidal forcing during spring tides and a meteorologically driven component, such as storm surges (Haigh et al, ; Sweet et al, ). A compound flood event is determined as a joint occurrence of hourly annual maxima coastal water level (as an indicator of HCWL; Bevacqua et al, ; Moftakhari et al, ; Sadegh et al, ; Sweet et al, ) and the lagged daily fluvial peak discharge within a 500‐km radius of TGs (Bartsch‐Winkler & Lynch, ; Hoitink & Jay, ) in a time range of ±7 days from the day of occurrence of the HCWL event. The lag time of the discharge is determined using the average response time of the catchment to a precipitation event (see section ).…”

“…Previous researchers have identified a moderate to strong correlation between high coastal water levels (HCWLs) or storm surges and floods (either pluvial or fluvial) globally , and regionally across Europe (Kew et al, 2013;Paprotny et al, 2018;Petroliagkis et al, 2016;Reeve et al, 2008;Svensson & Jones, 2001;Svensson & Jones, 2004), the United States (Moftakhari et al, 2017;Sadegh et al, 2018), Australia (Wu et al, 2018;Zheng et al, 2013Zheng et al, , 2014, and China (Lian et al, 2013;Tu et al, 2018). In addition, they have focused on the joint probability of compound floods Kew et al, 2013;Moftakhari et al, 2019;Moftakhari et al, 2017;Sadegh et al, 2018;Ward et al, 2018) considering individual drivers and the conditional probability (Bevacqua et al, 2017) of occurrence of river floods given moderate to extreme coastal water levels or surges. These metrics are valuable for risk management and infrastructure design in delta areas (Wu et al, 2018).…”

“…The lag time of the discharge is determined using the average response time of the catchment to a precipitation event (see section 2). A lag of a few days has been considered only in a limited number of studies (Klerk et al, 2015;Kew et al, 2013;Petroliagkis et al, 2016;Zheng et al, 2017) but has been neglected in several other assessments (Bevacqua et al, 2018;Moftakhari et al, 2017;Reeve et al, 2008;Sadegh et al, 2018;Svensson & Jones, 2004;Tu et al, 2018;Wahl et al, 2015).…”

Climate‐Induced Changes in the Risk of Hydrological Failure of Major Dams in California

“…The length of data and choice of distribution can impose uncertainty on the return period and associated flood level estimations (Sadegh et al, 2017). This interval for the projection period is relatively more confined, ranging between 4,580 and 4,895 m 3 /s, given the longer data (that could provide more information) to constrain the GEV model parameters (Sadegh et al, 2018b). Figures S2 and S4 display posterior distribution of GEV model parameters in the historical and projection periods, respectively, which in turn translate to 100-year flood level distributions in Figures S3 and S5 (Jeremiah et al, 2011;Smith et al, 2010).…”

“…The 95% confidence interval for the 100-year flood level in the historical period for the Oroville Dame ranges between 3,720 and 4,190 m 3 /s. This interval for the projection period is relatively more confined, ranging between 4,580 and 4,895 m 3 /s, given the longer data (that could provide more information) to constrain the GEV model parameters (Sadegh et al, 2018b). We repeat this analysis with two other models, namely, inverse Gaussian and loglogistic, to analyze the impacts of distribution choice on the flood levels.…”