the bathymetric difference could be as large as 30 percent. Therefore, finer grid resolution and better accuracy of model bathymetry are critical for the improvement of model predictions.
According to the materials provided to the committee (Georgia Tech Research Corporation, 2011), the model tends to reproduce observed tidal elevations well. This is essential for the accurate prediction of the currents, but it may not be sufficient. It is possible for a model to reproduce tidal elevations well but still to have incorrect current patterns. Comparisons between predicted and observed currents indicated that errors associated with predicted currents may be 30 percent or more (ORNL, 2011). One of the main concerns surrounding the model calibration and validation efforts is the limited number of current observation stations used in the study—24 stations for model calibration and 15 for model validation (five stations are excluded because modeled and measured depths differ more than 30 percent), which means many of the 52 submodel domains do not contain any current data. Thus, the comparisons are more akin to spot-checking than actual validation, and comparisons are often poorest in the regions of most interest.
For example, at the site of the Snohomish Public Utility District pilot project in Admiralty Inlet, field data from the Northwest National Marine Renewable Energy Center shows a mean power density of 2 kW/m2, which can be compared to the mean power density of 0.8 kW/m2 given by the tidal resource assessment database. Field data also show a significant ebb dominance and directional asymmetry, in contrast to flood dominance and directional symmetry given by the resource maps.
The committee feels that efforts should have been focused on obtaining more observational data in the validation study rather than on producing a large metric of error statistics between model results and observations. It could be useful to consider more conventional model evaluation skill metrics used in the ocean modeling field (Warner et al., 2005; Patchen, 2007; NOAA, 2011a). Because power is related to the cube of current speed, errors of 100 percent or more occur in the prediction of tidal power density in many model regions. It is unclear whether model calibration through the adjustment of the single friction coefficient is more appropriate than adjustment or improvement of other factors, such as model bathymetry, grid resolution, or offshore boundary conditions. As noted by the tidal resource assessment group, errors in currents may be a consequence of inadequate model resolution rather than of an erroneous friction coefficient or uncertain forcing from the open boundary (ORNL, 2011).