Recent Updates to SWANFAR (registered trademark), a 5DVAR Data Assimilation System for SWAN

2016
Abstract : SWANFAR(registered name), a five-dimensional variational (5DVar) data assimilationsystem, was built from the architecture of the spectral wave modelSWAN. Initially designed to solve only the stationary, homogeneous wave action equation, the system has been updated to include adjoint and tangent-linear subroutinesfor nonlinear source/sink components of SWAN version 40.81, to perform nonstationary assimilations, and to apply covariance multipliers. The initial development of the stationary system, excluding source and sink terms, is described in Orzech et al. (2013). In the present report, we first discuss the building and testing of the nonstationary system with source and sink terms, detailing the status of all related subroutines. We then briefly summarize the implementation of radial, spectral, and temporal covariance expressions, which are more fully detailed in Veeramony et al. (2016). Results from twin experiments and field tests of the system at different stages are presented, followed by discussion of the advantages and limitations of spectral data assimilationbuilt upon the SWAN architecture. Guidelines for performing an assimilation with SWANFAR(registered name), along with samples of relevant scripts and files, are provided in several appendices.
    • Correction
    • Source
    • Cite
    • Save
    0
    References
    0
    Citations
    NaN
    KQI
    []
    Baidu
    map