Assessing the Potential for Compound Storm Surge and Extreme River Discharge Events at the Catchment Scale with Statistical Models: Sensitivity Analysis and Recommendations for Best Practice

Two-sided extreme conditional sampling regularly is coupled with copula theory to assess the dependence between flood-risk drivers such as extreme precipitation or river discharge and storm surge. The approach involves many subjective choices, including sampling techniques used to identify extreme events [block maxima or peaks-over-threshold (POT)], whether to account for the fit of marginal distributions, and time-lags considered between the two drivers. In this study, estimates of the potential for compound events at three sites along the Texas Gulf Coast, where the USACE is undertaking coastal storm risk management (CSRM) projects, were shown to be highly sensitive to the setup of the statistical model. A pragmatic approach accounting for marginal fit in a POT framework is proposed and was shown to provide stable estimates of the compounding potential for high discharge and storm surge events. We also explored the effect of using precipitation as a proxy for discharge in the absence of sufficiently long discharge records.

Automated summary

Two-sided extreme conditional sampling combined with copula theory is used to assess the dependence between flood-risk drivers and storm surge. The setup of the statistical model greatly affects the estimation of compound events. A pragmatic approach that considers marginal fit in a POT framework is proposed and provides stable estimates of the compounding potential for high discharge and storm surge events. Using precipitation as a proxy for discharge is also explored when discharge records are not available.