Applied Joint Probabilistic Modeling of Compound Coastal Flood Hazard: An Extension of the Joint Probability Method with Optimal Sampling

Abstract

Compound coastal flooding i.e. coastal flooding driven by storm surge, rainfall, and riverine dynamics poses a significant and complex hazard. We present a novel framework for statistical modeling of this hazard as applied in a preliminary pilot study in Louisiana. This framework extends the Joint Probability Modeling with Optimal Sampling (JPM-OS), previously used for purely surge and wave driven flooding, with a stochastic rainfall field generator to produce an empirical distribution of compound surge-rainfall events with pre-computed surge and wave behavior modeled via ADCIRC + SWAN and hydrologic behavior modeled via HEC-HMS. A clustering-based discretization scheme is then applied to the sampling distribution in order to reduce set of outcomes to a size which can be tractably simulated via HEC-RAS while minimizing the square error induced by discretization. While model improvements are ongoing, the clustering-based discretization scheme is highly generalizable, provides guaranteed convergence to local optima, and performs well in preliminary analysis.