Forecast Model Bias Correction in Ocean Data Assimilation

Abstract

Numerical models of ocean circulation are subject to systematic errors resulting from errors in model physics, numerics, inaccurately specified initial conditions, and errors in surface forcing. In addition to a time-mean component, the systematic errors include components that are time varying, which could result, for example, from inaccuracies in the time-varying forcing. Despite their importance, most assimilation algorithms incorrectly assume that the forecast model is unbiased. In this paper the authors characterize the bias for a current assimilation scheme in the tropical Pacific. The characterization is used to show how relatively simple empirical bias forecast models may be used in a two-stage bias correction procedure to improve the quality of the analysis.