A high-resolution, unstructured Simulating WAves Nearshore (SWAN) model with a resolution of 200 m within 20 km of the coast was developed to provide a reliable setting for a long-term wave energy characterization of the Gulf of Mexico, Puerto Rico, and the U.S. Virgin Islands. A thorough model parameter sensitivity analysis, as well as a calibration process for selecting the whitecapping dissipation formulation, were conducted. Sensitivity analyses for the simulation timestep and number of iterations highlighted the less-studied interplay between these parameters in SWAN, which can substantially affect simulation accuracy and cost and is vital for the next step long-term simulation of the wave energy resources. For the present study, a 3 min timestep and three iterations are optimum. The “Garden Sprinkling” effect and the cut-off frequency were also investigated. Using 36 directional bins and a larger cut-off frequency (1.0 Hz) enable the best agreement between the model simulation and the in situ wave observations. This subsequently leads to improved model skill and performance in resolving the observed International Electrotechnical Commission (IEC) wave energy resource parameters that are highly non-linear functions of wave spectral moments.