Skip to Main Content
This paper describes the implementation of a prototype operational system for providing near-real-time information on the ocean water property and circulation environment in the Gulf of Mexico (GOM)/Georges Bank (GB) region. This application of the Harvard Ocean Prediction System (HOPS) model to the Advanced Fisheries Management Information System (AFMIS) was developed and tested during a 52-week sequence of weekly nowcast/forecasts. The initial assimilation format for a single sea surface temperature (SST) image was expanded to accommodate the assimilation of a trio of SST images, which produced a more realistic thermocline. The model system was applied to the winter-like conditions of March 2002 and the summer-like conditions of August 2002, when the model temperature and velocity fields could be compared with moored time-series measurements at several locations. This data assimilation model system produced qualitatively correct ocean temperature and flow patterns. During late winter, the wind dominated the variability of the model ocean surface Ekman transport, GB north flank jet, and Maine Coastal Current. During late summer, the assimilation of very warm SST dominated the variability of the model coastal upwelling, a stronger and more stable GB north flank jet, and Maine Coastal Current. The model did reveal a persistent anticyclone in the western central GOM that has not been well documented. Quantitatively, the surface temperatures from the model converged with those measured by the Gulf of Maine Ocean Observing System (GoMOOS) within 4-6 d from the beginning of the model run. Further, while the variability of the model and observed temperatures generally tracked each other, there were cases in which differences between contemporary and climatological temperatures led to systematic offsets. The model/observation velocity comparisons, while in general not as good as those for temperature, did improve after 6-8 d and the assimilation of two SST images. Thus, fu- ture work is needed to better understand the quantitative model/observation differences revealed by intercomparisons and to improve the initialization and assimilation protocols.