A Comprehensive 4D-Var Data Assimilation and Analysis System Applied to the California Current System using ROMS

Andy Moore, Hernan Arango, Gregoire Broquet, Chris Edwards, Milena Veneziani, Brian Powell, Dave Foley and Jim Doyle

ROMS is one of the few community ocean models for which a 4The ROMS 4D-Var data assimilation platforms have been applied extensively to the California Current System (CCS) using ROMS, and have provided a wealth of valuable experience with regard to the challenges of data assimilation in a mesoscale ocean circulation environment. ROMS currently supports three 4D-Var data assimilation systems, and a suite of powerful diagnostic tools. In this talk we will present results from a 2.5 year period during which 4D-Var was applied sequentially to the CCS. During this period, a variety of data were assimilated from various platforms. Example analyses will be presented, including estimates of the expected analysis error, and quantification of the level of redundancy and efficiency of the observing array. The impact of correcting for uncertainties in the initial conditions, surface forcing, and open boundary conditions on various different aspects of the CCS circulation estimates will also be presented. In addition, the impact of each observation platform on circulation estimates during the analysis and forecast cycles will also be explored.


ROMS 4D-Var: The Secrets Revealed

ROMS is one of the few community ocean models for which a
4-dimensional variational (4D-Var) data assimilation capability has been developed. The ROMS 4D-Var system, however, is by far the most comprehensive of all those currently available. ROMS 4D-Var comprises three different 4D-Var platforms: one that searches for the best estimate of the ocean circulation in the full space spanned by the model, and two that search in the sub-space spanned by the observations. In all cases, the best estimate circulation can be identified by adjusting the model initial conditions, surface forcing, and open boundary conditions. In the case of 4D-Var in observation space, model errors can also be admitted, leading to the so-called weak-constraint problem. Several powerful 4D-Var post-processing algorithms have also been develop for ROMS, including an observation sensitivity driver. This talk will be a follow-on to the talk entitled "A Comprehensive 4D-Var Data Assimilation and Analysis System Applied to the California Current System using ROMS," and the focus will be a review of the fundamental ideas that underpin the entire 4D-Var package for ROMS.