On the generation of high‐resolution probabilistic design events capturing the joint occurrence of rainfall and storm surge in coastal basins

Abstract

Coastal areas are subject to the joint risk associated with rainfall-driven flooding and storm surge hazards. To capture this dependency and the compound nature of these hazards, bivariate modelling represents a straightforward and easy-toimplement approach that relies on observational records. Most existing applications focus on a single tide gauge-rain gauge/streamgauge combination, limiting the applicability of bivariate modelling to develop high-resolution space-time design events that can be used to quantify the dynamic, that is, varying in space and time, compound flood hazard in coastal basins. Moreover, there is a need to recognize that not all extreme events always come from a single population, but can reflect a mixture of different generating mechanisms. Therefore, this paper describes an empirical approach to develop design storms with high-resolution in space and time (i.e., ~5 km and hourly) for different joint annual exceedance probabilities. We also stratify extreme rainfall and storm surge events depending on whether they were caused by tropical cyclones (TCs) or not. We find that there are significant differences between the TC and non-TC populations, with very different dependence structures that are missed if we treat all the events as coming from a single population. While we apply this methodology to one basin near Houston, Texas, our approach is general enough to make it applicable for any coastal basin exposed to compounding flood hazards from storm surge and rainfall-induced flooding.

A Hybrid Framework for Rapidly Locating Transition Zones: A Comparison of Event‐ and Response‐Based Return Water Levels in the Suwannee River FL

“…Fitting separate models to the different populations of events, by adopting seasonality as a proxy for the hydrometerological conditions, will potentially increase the robustness of the modeling (Villarini & Smith, 2010). Partitioning the data in this way has revealed stronger correlations in the conditional samples (Kim et al., 2022). The rarity of conditions such as tropical cyclones may necessitate the inclusion of water levels from synthetic events propagated through a storm surge model to complement the existing in situ records (Dullaart et al., 2021).…”

On the compounding of nitrate loads and discharge

“…We apply a two‐sided conditional sampling method that can address an asymmetric problem as extreme discharges and extreme N loads do not have to happen concurrently. Accordingly, we have two types of POT events to select, depending on whether N loads or discharge are the conditioning variables (Bender et al., 2016; Jane et al., 2020; Kim et al., 2022). For the conditioning variable, we first select the POT events by setting a threshold that returns five events per year on average and allows only one peak in a ± 5‐day window to ensure independence of the POT events (Lang et al., 1999).…”

“…Early applications of copulas in hydrology primarily concerned flood frequency analysis (e.g., De Michele & Salvadori, 2003; Favre et al., 2004). Since then, bivariate copulas have gone on to account for spatial correlations of extremes at neighboring sites (e.g., Bender et al., 2016), been used in stochastic design storm generators (e.g., Kim et al., 2022; Vandenberghe et al., 2010) among other applications encompassing a variety of hydroclimatic variables (Tootoonchi et al., 2022). A flurry of recent studies chose copulas to model the joint probabilities of co‐occurring high freshwater fluxes (rainfall/river discharge) and high sea levels (e.g., Wahl et al., 2015); this is motivated by impactful events such as Hurricane Sandy (2012) and Hurricane Harvey (2017), where the interaction of these drivers likely exacerbated flooding.…”

“…conditioning variables (Bender et al, 2016;Jane et al, 2020;Kim et al, 2022). For the conditioning variable, we first select the POT events by setting a threshold that returns five events per year on average and allows only one peak in a ± 5-day window to ensure independence of the POT events (Lang et al, 1999).…”

Projecting Compound Flood Hazard Under Climate Change With Physical Models and Joint Probability Methods

“…For example, Hurricane Florence (2018) caused a roughly 100‐year storm tide in the Cape Fear River, NC while simultaneously dropping rainfall totals that exceeded the 500‐year level (Stewart and Berg, 2018). Moreover, the dependence structure between rainfall and storm surges may be different for TCs compared to ETCs (Kim et al., 2022). Therefore, to adequately characterize coastal flood risk, it is crucial to specifically account for the dependence between TC rainfall and storm surges and jointly model their flood impacts in coastal catchments.…”