A Feasibility Study of Dynamical Assimilation of Tide Gauge Data in the Chesapeake Bay

The feasibility of dynamical assimilation of surface elevation from tide gauges is investigated to estimate the bottom drag coefficient and surface stress as a first step in improving modeled tidal and wind-driven circulation in the Chesapeake Bay. A two-dimensional shallow water model and an adjoint variational method with a limited memory quasi-Newton optimization algorithm are used to achieve this goal. Assimilation of tide gauge observations from ten permanent stations in the Bay and use of a two-dimensional model adequately estimate the bottom drag coefficient, wind stress and surface elevation at the Bay mouth. Subsequent use of these estimates in the circulation model considerably improves the modeled surface elevation in the entire Bay. Assimilation of predicted tidal elevations yields a drag coefficient, defined in the hydraulic way, varying between 2.5 x 10-4 and 3.1 x 10-3. The bottom drag coefficient displays a periodicity corresponding to the spring-neap tide cycle. From assimilation of actual tide gauge observations, it is found that the fortnightly modulation is altered during frontal passage. Furthermore, the response of the sea surface to the wind forcing is found to be more important in the lower Bay than in the upper Bay, where the barometric pressure effect could be more important. In addition, identical twin experiments with model generated data show that a penalty term has to be added to the simple cost function defined as the distance between modeled and observed surface elevation in order to assure smoothness of the surface elevation field at the Bay mouth. Classical scaling of the parameters to bring them to the same order of magnitude was not effective in accelerating the convergence during the assimilation procedure and yielded larger errors in the estimated parameters