A Discretized Adjoint Model for SWAN

The SWAN wave model solves the spectral action balance equation and is widely used to produce forecasts and event-specific wave climatologies. Data assimilation could significantly improve the quality of SWAN forecasts, but it is presently not available in the operational model. Walker (2006) achieved limited success in tests with a simplified analytical adjoint to SWAN. However, its application to data assimilation was limited to stationary conditions and did not account for waves generated by winds or non-linear interactions (e.g., Veeramony et al., 2010). To overcome these deficiencies, we are developing a numerical adjoint that will directly reflect the properties of the discretized forward SWAN model, including non-linear interactions, depth-limited breaking, and non-stationary conditions. This presentation describes the first stages of the project, focusing on the creation and evaluation of a simplified numerical adjoint framework limited to stationary conditions and small nearshore domains. The basic adjoint will be tested with a twin experiment, in which it is initialized with spectral wave data from SWAN and used to estimate the original boundary conditions. The results will be compared to those from the Walker (2006) analytical adjoint. References: Veeramony, J., D. Walker, and L. Hsu (2010). A variational data assimilation system for nearshore applications of SWAN. Ocean Modelling, doi:10.1016/j.ocemod.2010.07.008, in press. Walker, D.T. (2006). Assimilation of SAR imagery in a nearshore spectral wave model. Tech. Rep. 200236, GDAIS.