Estimating Compound Flooding Potential at the Catchment Level: Recommendations for Best Practice.

Abstract:

In low-lying coastal areas flooding drivers such as surge and precipitation, a driver of increased discharge, often exhibit significant dependence. Storm surge and extreme discharge interact often nonlinearly to exacerbate flooding, an example of a compound event. It is essential to account for the interaction between the two processes if robust flood risk estimates are to be obtained. This is often achieved by extending fluvial hydraulic models to include an ocean boundary condition, at the expense of increasing their complexity and computational burden. Recent global and national scale analyses demonstrate that although regional patterns can be discerned, exposure to compound flooding is highly dependent on localized factors such as catchment size. A prudent first step in estimating flood risk in low-lying coastal areas is therefore to assess the propensity for storm surge and extreme discharge to co-occur at the catchment scale.

A well-established approach for assessing compound flooding potential involves applying two-sided conditional sampling to identify extreme surge and discharge combinations, before fitting copulas to quantify the shape and asymptotic properties of any dependence. The approach requires subjective parameter choices during several steps in the model setup: time-lags considered between the drivers in the conditional sampling, as well as the detrending and sampling techniques employed including any relevant thresholds. The parameter choices should reflect the nature of the physical processes, statistical modeling assumptions, and data availability. To date the sensitivity of the results to the subjective parameter choices is often ignored or carried out to a limited extent. In this work the sensitivity of the copula and bivariate design event(s) to the parameter choices is explored for two sites along the Gulf coast where hurricane flood protection systems are being upgraded or installed. Based on the analysis an objective approach for setting up the model is proposed.