A Variational Assimilation System for Nearshore Wave Modeling

A variational data assimilation system is developed for the stationary, homogeneous portion of the wave model Simulating Waves Nearshore (SWAN). The system is based on a numerical adjoint constructed for the discrete forward SWAN code; its performance is compared to that of an earlier system based on a discretized analytical adjoint (Walker; Veeramony et al.). This paper describes the development and validation of individual numerical adjoint subroutines, followed by the testing and evaluation of the assimilation system as a whole with an idealized twin experiment and with data from Duck, North Carolina. In the twin experiment, the present system performs on par with that of Walker. Estimates of wave spectra and spectral statistics also compare well to measured spectral data at Duck, North Carolina. The error in these estimates is partly due to the exclusion of nonlinear source and sink terms from the adjoint and partly due to different spectral processing techniques used for different types of instruments.