from regional or global ocean models. To select the best suited models, the group initially assessed the configurations and performance of a set of models, including HYCOM-GLOBAL (Hybrid Coordinate Ocean Model), HYCOM-GOM (Gulf of Mexico), JPL-ROMS (NASA Jet Propulsion Laboratory Regional Ocean Model System), and NCOM (Navy Coastal Ocean Model).7 HYCOM-GLOBAL is a real-time 1/12-degree global nowcast/ forecast ocean circulation model maintained by the Naval Research Laboratory8 and sponsored by the National Ocean Partnership Program, while HYCOM-GOM is a regional model with grid resolution of 1/25 degree nested within the HYCOM-GLOBAL model. HYCOM employs a hybrid coordinate system that combines (1) an isopycnal coordinate in the stratified ocean, (2) a terrain-following coordinate in shallow coastal regions, and (3) a z-level coordinate in the mixed layer and/or unstratified seas. This unique approach in vertical coordinates gives HYCOM the ability to simulate different physical processes in the ocean at different scales using a single model (Halliwell et al., 2000). Power density distributions and variability and confidence intervals of the probability distributions are calculated based on the daily HYCOM-GLOBAL and HYCOM-GOM results from 2004 to the present. Model selection for the ocean current energy assessment was based on a number of criteria, including model error statistics, model grid resolution, length of simulation period, and model output intervals. HYCOM-GLOBAL was selected for the U.S. West Coast, Alaska, and Hawaii coastal regions; a hybrid NCOM-HYCOM was selected for the East Coast; and HYCOM-GOM was selected for the Gulf of Mexico and the Florida Current. The assessment group created an ocean current database for coastal waters, up to 200 nautical miles offshore, using the selected models.
An independent validation of the database was conducted by the Oak Ridge National Laboratory (ORNL, 2012). The main challenge of validating the predicted ocean currents is the paucity of observational data. The most complete datasets are those from high-frequency (HF) radar and from a cable off Florida’s east coast (both of which were quite limited in time), and an additional stationary acoustic Doppler current profiler (ADCP) dataset from Florida Atlantic University was also used. In general, the model results had a reasonably good match to the ADCP data but a poor one to the HF radar. Predicted daily flow from the model matched the distribution pattern of the submarine cable data, although the cable data showed a higher percentage of occurrence in the maximum flow range.
7 See footnote 3.