Q. Chen Hongqing Wang Nan Wang Kelin Hu William D. Capurso Lukasz M. Niemoczynski Gregg Snedden Ling Zhu 2023 <p><span>Living shorelines gain increasing attention because they stabilize shorelines and reduce erosion. This study leverages physics-based models and bagged regression tree (BRT)&nbsp;machine learning algorithm&nbsp;to simulate&nbsp;wave&nbsp;dynamics at a living shoreline composed of constructed oyster reefs (CORs) in upper Delaware Bay. The physics-based models consist of coupled Delft3D-FLOW and SWAN in four-level nested domains. The model accuracy converges with increasing&nbsp;mesh&nbsp;resolution. The simulated wave-induced current circulation substantiates the effectiveness of CORs in trapping sediments. The simulated yearly-averaged wave power correlates qualitatively with historical shoreline retreat rates. BRT is adopted to improve the model accuracy, identify key processes responsible for simulation errors in wave height (<i>H</i>₈</span><span>) and wave period (<i>T</i>_p</span><span>), and quantify their importance. In the CORs sheltered area, BRT reveals that simulation errors of wind seas mainly arise from wind forcing, wave breaking and wave triad interactions. Wave breaking is seven times more important than wind forcing for simulating <i>H</i>₈</span><span>, while wind forcing and triad interactions are of equal importance for simulating <i>T</i>_p</span><span>. Simulation errors of swells mostly stem from&nbsp;bottom friction&nbsp;and offshore wave boundary conditions. Results from this study can help the assessment and&nbsp;adaptive management&nbsp;of CORs-based living shoreline restoration projects under climate change.</span></p> application/pdf 10.1016/j.oceaneng.2023.115207 en Elsevier Modeling surface wave dynamics in upper Delaware Bay with living shorelines article