The Inverse Ocean Modeling System. Part I: Implementation

Abstract

The Inverse Ocean Modeling (IOM) system constructs and runs weak-constraint, four-dimensional variational data assimilation (W4DVAR) for any dynamical model and any observing array. The dynamics and the observing algorithms may be nonlinear but must be functionally smooth. The user need only provide the model and the observing algorithms, together with an interpolation scheme that relates the model numerics to the observer?s coordinates. All other model-dependent elements of the Inverse Ocean Modeling assimilation algorithm (see both Chua and Bennett), including adjoint generators and Monte Carlo estimates of posteriors, have been derived and coded as templates in Parametric FORTRAN (Erwig et al.). This language has been developed for the IOM but has wider application in scientific programming. Guided by the Parametric FORTRAN templates, and by model information entered via a graphical user interface (GUI), the IOM generates conventional FORTRAN code for each of the many algorithm elements, customized to the user?s model. The IOM also runs the various W4DVAR assimilations, which are monitored by the GUI. The system is supported by a Web site that includes interactive tutorials for the assimilation algorithm.