Structure and timing of recirculation around Georges Bank: an observational and modeling study at the Great South Channel: Part I--ensemble smoother

An ensemble smoother is used to assimilate moored temperature, salinity, and velocity data into a local area primitive equation model. The overall goal of the analysis is to estimate variability of Georges Bank recirculation, i.e., northward flow through the Great South channel in support of the US Global Ocean Ecosystem Dynamics (GLOBEC) Georges Bank experiment. Here, identical twin experiments are carried out to test the ensemble smoother with a finite-element circulation model of the Great South Channel, based on a previous formulation designated QUODDY. The ensemble smoother utilizes a finite number of Monte Carlo model simulations to estimate model error covariance. The prior distribution from which the ensemble members are simulated is implicitly defined by the forward model by adding spatially correlated Gaussian random variables to the initial conditions, and time-dependent boundary elevations. Atmospheric forcing (wind stress) is derived from buoy measurements and is assumed to be known with certainty. The accuracy of the estimator depends on the state space variable being estimated and proximity to the data. In these twin experiments the domain-wide mean error variance of temperature, salinity, and velocity were reduced 96%, 93%, and 89%, respectively. The prediction statistics for the estimate are accurate throughout the domain. Non-linearity of the forward model and subsequent skewness of the posterior probability density function (pdf) are investigated. It is found that the posterior distribution is sufficiently Gaussian to use Gaussian confidence intervals. These results give confidence for using the numerical formulation and ensemble smoother to examine variability in circulation at Great South Channel with available data.