A Practitioner's Approach to Process-Driven Modeling of Compound Rainfall and Storm Surge Extremes for Coastal Texas

Coastal Texas is no stranger to riverine and coastal flooding and, while these events can be catastrophic on their own, they can have an even larger impact when they occur together, as in the case of compound extremes. While the emphasis has been on quantifying the compounding of these two hazards, much less is known about generating design rainfall events with a high space-time resolution and preserving the dependence with storm surge. This is particularly problematic in areas where there are limited records and multiple flood generating mechanisms. Here we focus on five watersheds in coastal Texas and develop an approach to generate design events for different annual exceedance probabilities (AEPs) subject to different compounding characteristics. We show that tropical cyclones (TCs) exhibit a stronger dependence between the nontidal residuals and precipitation than non-TC events, and we provide approaches to generate design events for both types of storms. Moreover, we propose to infill missing data in water levels to mitigate issues related to small samples. We also quantify the lag between rainfall and storm surge peaks. Our study provides a roadmap for practitioners who need to develop design events with high spatio-temporal resolution for different AEPs, and are dealing with compounding, mixed distributions, short record length, and lag between peak rainfall and storm surge.