Wave hydrodynamics and attenuation in idealized mangrove forest: Large-scale physical and numerical modeling

Mangroves play a crucial role in mitigating coastal flooding, protecting shorelines, and enhancing coastal hazard resilience. While numerous studies have investigated the protective performance of mangroves against tsunami-like waves, the effectiveness of Rhizophora mangrove forests, characterized by their complex prop-root systems, has not been thoroughly quantified. Thus, a series of prototype-scale experiments were conducted to evaluate the protective performance of an idealized Rhizophora mangrove forest with a moderate cross-shore width against flooding scenarios. Additionally, a computational fluid dynamics (CFD) model using OpenFOAM was carried out and validated utilizing laboratory experiments to provide a detailed understanding of the hydrodynamic interactions between tsunami-like waves and mangrove elements. The study further examines the influence of wave parameters and mangrove properties on wave attenuation coefficients. The results indicate that increases in the water depths generally lead to a reduction in the wave attenuation coefficient. While increasing incident wave height resulted in an increase in the water attenuation coefficient in the shallow water depth, as the water depth increases, the impact of incident wave height on wave attenuation tends to be reduced. Moreover, mangrove configurations with higher stem density exhibit significantly greater wave attenuation, with higher stem density attenuation coefficients up to 3.5 times those found in lower density configurations. The study also investigates the relationship of dimensionless parameters derived from wave and mangrove parameters on wave attenuation coefficient, employing Multivariate Non-Linear Regression (MNLR) and prediction models based on machine learning, including Deep Neural Networks (DNN), Support Vector Regression (SVR), and eXtreme Gradient Boosting (XGBoost). While an empirical equation developed using the MNLR method showed a strong correlation (R² = 0.86), the DNN model outperformed the other prediction algorithms, demonstrating superior accuracy in predicting wave attenuation coefficients (R² = 0.97). Furthermore, the DNN model was used to evaluate the relative importance of each influencing parameter, revealing that the density and cross-shore width of the mangrove forest are the dominant variables, contributing 47% and 30%, respectively, to the wave attenuation coefficient.