Performance of a nowcast/forecast system for Prince William Sound,Alaska

A coastal ocean extended Prince William Sound nowcast/forecast system (EPWS/NFS) has been running semi-automatically for an extended domain of Prince William Sound (PWS), Alaska for 2 years. To determine the performance of this modeling system, an assessment is conducted. EPWS/NFS and PWS/NFS (viz., its predecessor) nowcasts are compared with observed time series of sea surface temperature (SST) and coastal sea level (CSL) at a few stations, and to velocity profiles from a moored ADCP. With the extension of the model domain to include the continental shelf outside PWS and forced by an operational global ocean model (Global-Navy Coastal Ocean Model (Global-NCOM)) and a 2D tidal model at the open boundary, EPWS/NFS has achieved significant improvement over PWS/NFS, which covered only PWS per se, for most of the predicted variables in this study. In both magnitude and phase, EPWS/NFS accurately predicts the coastal tide fluctuations, as well as M2 tidal currents in Central Sound, although significant errors in coastal tides exist during some spring and neap tide cycles. Other than for the tidal motions, EPWS/NFS generally produces less energetic CSL and velocity variations than those observed. In comparison, although PWS/NFS well predicts the coastal tides, it suffers from the absence of low-frequency CSL variations, as well as misprediction of M2 tidal currents in Central Sound. For 40 h low-passed PWS/NFS and EPWS/NFS velocities, significant phase error occurs during the model–date comparison period, while EPWS/NFS nowcasts generally produce less root-mean-square-error (rmse) and smaller correlations with the observations than PWS/NFS does. Both observations and EPWS/NFS have similar vertical profiles of baroclinic velocity standard deviations, but some substantial discrepancies occur in the velocity direction. Also, in the Central Sound, EPWS/NFS predicts well the SST seasonal cycle and a major cooling event during the summer 2005. However, for periods shorter than 1 week, both PWS/NFS and EPWS/NFS SST underestimated the observed fluctuations by an order of magnitude.