GULF-STREAM AND RING FEATURE ANALYSES FOR FORECAST MODEL VALIDATION

A series of Gulf Stream forecast model test casts were developed for t he Data Assimilation and Model Evaluation Experiment (DAMEE). The mode l initialization and verification procedure relies heavily on a series of accurate synoptic snapshots of the Gulf Stream north wall and ring locations. Satellite infrared imagery, Geosat altimetry, and numerous in situ temperature profiles were combined using a geographic informa tion system to construct Gulf Stream and ring location analyses at app roximately weekly intervals during a 6-week, data-rich time period. To improve the accuracy of the feature analyses, a new image compositing technique called patching was developed to decrease the spatial smear ing experienced with standard warmest pixel composites. During the 6-w eek test period, three ring formations, two ring absorptions, and one ring merger event were observed. The average difference between the we ekly north wall positions ranged from 20 to 34 km (average 27 km). Whe n the DAMEE GSR north wall positions were compared to two independent analyses for the same time period, the average offsets were found to v ary from 14 to 53 km (average 29 km) for the first comparison set and 14 to 30 km (average 22 km) for the second. These differences, which a re similar in magnitude to the observed weekly evolution, are attribut ed to differences in the data treatment for cloudy regions and the sub jectivity of the image analysts when dealing with incomplete data.