Application of a Semiempirical Analytical Model to Predict Erosion of a Texas Coastal Defense Dune System during Storms under Climate Change

This study assesses the feasibility of a beach and dune system as flood defense against storm surge along the coastlines of the Houston–Galveston area, proposed as a part of the Coastal Texas Project. We apply a semiempirical analytical model to predict dune erosion in a dual-dune system under changing climate conditions. Synthetic storms were simulated using validated hydrodynamic, wave, and hurricane models to produce input data (storm surge, wave height, and period) for the dune erosion model, reflecting both present day and future climate scenarios that incorporate projected sea level rise (SLR). Bias-correction techniques were applied to climate model output using historical observations of storm surge and wave data. An alternative sampling approach was also developed to stochastically predict dune erosion by integrating synthetic data into a copula-based framework. Results indicate that the annual-average dune erosion is approximately 8%–10% of system volume in the present scenario and increases to 33%–40% in future scenarios with higher SLR, leading to estimated dune rehabilitation cycles of 8–10 and 2–2.5 years, respectively. These findings suggest that, although the proposed beach and dune system is likely to be effective for storm surge protection under the present climate condition, significant adjustments will be desirable to maintain its resilience in the face of evolving climate and sea level rise. Importantly, bias correction of input data yielded substantial reductions in predicted storm surge and significant wave height, resulting in more accurate dune erosion predictions. This demonstrates the necessity of bias correction of hydrodynamic and wave parameters derived from global climate simulations for reliable coastal risk assessment and future planning. The copula sampling approach produced results comparable to the original results, which considered storms with extremely low or high occurrence probabilities, while providing lower sensitivity to bias-correction methods and copula generator types.