Title: A multi-scale three-dimensional variational data assimilation scheme and its application to coastal oceans

Authors: Zhijin Li (Jet Propulsion Laboratory)
Yi Chao (Jet Propulsion Laboratory)
James C. McWilliams (UCLA)
Kayo Ide (UCLA)

A multi-scale three-dimensional variational data assimilation scheme (MS-3DVAR) for high resolution ocean models that encompass a wide range of spatial scales is presented. An equivalence between the standard 3DVAR cost function and a set of cost funtions for spatially distinct scales is derived. This equivalence provides a pathway for developing the MS-3DVAR scheme using partitioned cost functions and thus background error covariance of multi-decorrelation length scales. MS-3DVAR improves the effectiveness of assimilation of both very sparse and high resolution observations by using the background error covariance of multi-decorrelation length scales and reducing the inherent observational representativeness errors. In the implementation presented, the cost function set consists of two componenets for large and small scales, and 3DVAR is then implemented sequentially from large to small scales. In coastal ocean applications, MS-3DVAR is effective in assimilating two of the most common types of ocean observations - sparse vertical profiles and high-resolution surface measurements - simultaneously. A set of identical twin experiments and one month of results from an operational implementation are used to demonstrate the advantagges of MS-3DVAR over a conventional 3DVAR scheme.

GMAO Head: Michele Rienecker Global Modeling and Assimilation Office NASA Goddard Space Flight Center

Curator: Nikki Privé Last Updated: May 27 2011