Abstract
The aspects of the navy's multivariate optimum interpolation (MVOI) for atmospheric analysis are presented. Included is an overview of how the MVOI is used in the navy's data-assimilation system, and the basic design of the MVOI. The specific features described include: a cursory presentation of the optimization method and some of its deficiencies, the application of the volume method in data selection and program design, the structure models used to represent the prediction errors, and the quality control checks applied to the observations.
Validation experiments that illustrate some of the features of the navy's analysis system are presented. Experiments showing the exactness of the geostrophic constraint, the effect of correlated observation error, the advantage of the geostrophic constraint, and the impact of satellite temperatures on the analysis are presented. These experiments were necessary to catch minor design and programming errors in the analysis system that are too small to be detected through casual inspection, yet which degrade the quality of the analysis. It has been shown that implementation of the volume method has given more precise geostrophic coupling over the gridpoint method, and that inputting satellite temperatures as pressure thicknesses rather than pressure-level heights produces results that agree with the satellite-derived layer thickness values, while it ties the analysis to observations of pressure heights. Finally, the validation experiments were shown to be highly effective at removing subtle errors in the analysis system, which led to rapid implementation and an extended error-free operational lifetime.
