Application of a hybrid 3D-var data assimilation system in the Monterey Bay to study regional dynamics of the California Current System

Chudong Pan


A data assimilation system combining 3-dimensional variational scheme and Navy Coastal Ocean Model was applied to the Monterey Bay area to assimilate temperature and salinity glider data collected in August 2003. The hybrid background error covariance model in the present 3-dimensional variational system incorporates both the static and the flow-dependent background error covariance. To explore the impact of high temporal resolution on the overall skill of the assimilation system, the intended data assimilation interval was set to 1 hour in the present study. A Floating Temporal Window approach is designed to keep the computational efficiency of the scheme and to retain the flow-dependent. Both twin data experiments with "synthetic" data and real data experiments were performed in the present study. The performance of the traditional 3-dimensional variational scheme and the hybrid scheme in twin data experiments were evaluated with respect to the "true state." The hybrid scheme reduces both temperature and salinity errors during the twin data experiment, especially during the transition period from wind relaxation to upwelling events and vice versa. In real data experiments, the model performance is validated against independent mooring observations. In the first real data experiment, the model free run, data assimilation run with traditional scheme, and data assimilation run with hybrid scheme were compared. The assimilation skill was tested by calculating the normalized distance between the assimilation results and observations at the mooring locations. The performance of the hybrid scheme generally exceeded the performance of the traditional scheme. Although there was no velocity data assimilation in the experiments, the change of temperature and salinity fields originated from glider data assimilation had a positive impact on the velocity fields according to mooring velocity records. The second real data experiment compared the hybrid schemes with the full ensemble and the Floating Temporal Window ensemble. Results showed that the Floating Temporal Window scheme provided lower discrepancy between the values of temperature, salinity and velocity predicted by the model and observed at the moorings. The improvement became more clearly visible during the upwelling and relaxations events associated with intermittent wind forcing. The results from the Navy Coastal Ocean Model free run and the data assimilation run with the hybrid scheme were compared with historical data. Comparisons have shown that the Navy Coastal Ocean Model combining with the hybrid 3-dimensional variational assimilation system was capable of reproducing major dynamical features, including onshore-offshore translation of the Monterey Bay Eddy during upwelling and relaxation events, coastal upwelling and respective upwelling filaments, the appearance of the California Undercurrent, and the interaction between the California Undercurrent and California Current.