Coastal ocean models are used for a variety of applications, including modeling tides and hurricane storm surge. These models numerically solve the shallow water equations, which are derived by depth integrating the Navier-Stokes equations. The inherent uncertainties in coastal ocean models are a result of many factors, including unknown model parameters. Parameters of particular importance are those used to define the bottom friction of the physical domain. In this work we will estimate bottom friction coefficients using statistical data assimilation methods. Statistical data assimilation methods, such as the ensemble Kalman filter, are generally used for state estimation. However, the implementation of statistical data assimilation methods for parameter estimation is straightforward. The evolution of the model parameters is considered to be a stationary process; model parameters are “forecasted” by adding a small amount of random noise to the initial estimates. The forecasted parameters are then updated by data; they are first projected into an observation space using the numerical model, and then the residual between the forecasted model parameters and the observed data is weighted and used to update the original estimate. Statistical data assimilation for parameter estimation is a promising method for reducing the uncertainty in coastal ocean models.
This talk was given on 4-23-15.