Modeling wind waves from deep to shallow waters in Lake Michigan using unstructured SWAN

Accurate wind-wave simulations are vital for evaluating the impact of waves on coastal dynamics, especially when wave observations are sparse. It has been demonstrated that structured-grid models have the ability to capture the wave dynamics of large-scale offshore domains, and the recent emergence of unstructured meshes provides an opportunity to better simulate shallow-water waves by resolving the complex geometry along islands and coastlines. For this study, wind waves in Lake Michigan were simulated using the unstructured-grid version of Simulating Waves Nearshore (un-SWAN) model with various types of wind forcing, and the model was calibrated using in situ wave observations. Sensitivity experiments were conducted to investigate the key factors that impact wave growth and dissipation processes. In particular, we considered (1) three wind field sources, (2) three formulations for wind input and whitecapping, (3) alternative formulations and coefficients for depth-induced breaking, and (4) various mesh types. We find that un-SWAN driven by Global Environmental Multiscale (GEM) wind data reproduces significant wave heights reasonably well using previously proposed formulations for wind input, recalibrated whitecapping parameters, and alternative formulations for depth-induced breaking. The results indicate that using GEM wind field data as input captures large waves in the midlake most accurately, while using the Natural Neighbor Method wind field reproduces shallow-water waves more accurately. Wind input affects the simulated wave evolution across the whole lake, whereas whitecapping primarily affects wave dynamics in deep water. In shallow water, the process of depth-induced breaking is dominant and highly dependent upon breaker indices and mesh types.