Comparing statistical and physical methods for compound hazard assessment: The case of compound flooding during tropical cyclones

Many researchers have examined compound flooding by either statistically characterizing the joint occurrence of river flows and storm tides, or physically modeling selected storm scenarios within high-fidelity computational models. However, there has been less work on developing methods that quantify risk from multiple sources of flooding in terms of return period flood heights, which are crucial components of coastal design. In this study, we compare statistically driven versus physically driven approaches for compound flood height assessment in order to understand if computationally efficient statistical methods can accurately represent joint flood hazard. We utilize storm tides and river flows from 941 synthetic tropical cyclone (TC) events passing near Town Creek, NC. We first apply a widely used statistically driven approach that estimates the joint peak flow (Q) and storm tide (S) distribution using a bivariate copula and develops a limited set of scenarios for flood mapping. We compare this computationally efficient approach to a physically driven approach that simulates the full set of events within a hydrodynamic model and then estimates return period flood heights based on the modeled maximum depths. By comparing the two methods, we find that the statistically driven approach can capture the different flood zones present along the catchment and can estimate maximum water levels well compared to the physics-driven approach for low-return periods. However, for high-return periods the statistics-driven approach significantly underestimates water levels in the midstream portion of the catchment.