Multi-Scale Three-Dimensional Variational Data Assimilation System for Coastal Ocean Prediction

Zhijin Li,Yi Chao,P. Peggy Li

2012-01-01

Abstract:A multi-scale three-dimensional variational data assimilation system (MS-3DVAR) has been formulated and the associated software system has been developed for improving high-resolution coastal ocean prediction. This system helps improve coastal ocean prediction skill, and has been used in support of operational coastal ocean forecasting systems and field experiments. The system has been developed to improve the capability of data assimilation for assimilating, simultaneously and effectively, sparse vertical profiles and high-resolution remote sensing surface measurements into coastal ocean models, as well as constraining model biases. In this system, the cost function is decomposed into two separate units for the large- and small-scale components, respectively. As such, data assimilation is implemented sequentially from large to small scales, the background error covariance is constructed to be scale-dependent, and a scale-dependent dynamic balance is incorporated. This scheme then allows effective constraining large scales and model bias through assimilating sparse vertical profiles, and small scales through assimilating high-resolution surface measurements. This MS-3DVAR enhances the capability of the traditional 3DVAR for assimilating highly heterogeneously distributed observations, such as along-track satellite altimetry data, and particularly maximizing the extraction of information from limited numbers of vertical profile observations.