Verification of an Operational Gulf Stream Forecasting Model

A verification study for an operational ocean forecasting system that uses the quasi- geostrophic version of the Harvard Open Ocean Model as its dynamical model com- ponent is presented. The study is designed to test the ability of both the model and the system to perform 1-week duration forecasts in the Gulf Stream Meander and Ring region. The forecast system is evaluated using the results of an over 2-year duration real time operational forecasting exercise. The forecast model component is evaluated using six 1-week duration forecast simulations chosen from data rich time periods during the real time exercise. In each case, the dynamical model is initialized using Gulf Stream and ring feature models located in the positions de- termined from a composite data set consisting of satellite infrared imagery, Geosat altimetry, and in situ temperature profiles. Forecast sensitivities to three versions of the Gulf Stream feature model and to initial condition variations within the data error bars are examined. One-week duration .lorecasts are evaluated based on two measures of success: 1) the reproduction o( observed ring formation, interaction, or absorption events, and 2) the average offset between the forecast Gulf Stream axis and the verification data. The quasigeostrophic model is found to reproduce all observed events during the six 1-week forecast simulations and to generate an aver- age axis offset that, in the mean, is less than the average offset of persistence and close to the average error bars assigned to the verification analyses. During the real time exercise the forecast system correctly predicted 89% of the observed events, generated an average offset that was consistent with the idealized structure of the Gulf Stream feature model and produced an offset standard deviation consistent with the larger data error bars.