Bias and balance in the Ensemble Kalman Filter

Principal Investigator

Dr Olwijn Leeuwenburgh
Institute for Marine and Atmospheric Research
University Utrecht
P O Box 80005
3508 TA Utrecht
The Netherlands

o.leeuwenburgh@phys.uu.nl

Other researchers: Dr Peter Jan van Leeuwen

Project description

The proposed project is a follow-up of the ENACT project which finished in January 2005, and has links to the ENSEMBLES project, in which ECMWF has a strong involvement.

An ocean-data assimilation system based on the Ensemble Kalman Filter (EnKF) has been developed, tested and run on the HPCD computer system. With this system, a large ocean GCM ensemble can be run forward in parallel, after which an analysis step combines the model ensemble with in-situ measurements of temperature and salinity, and with satellite measurements of sea level. During the ENACT project is was shown that this method has considerable potential as part of future versions of the seasonal forecast system as used by ECMWF; the time-dependent errors were significantly reduced with respect to a control run.

For the proposed project we suggest adding additional features to the system. In particular, a bias correction method has already been implemented. Bias appears in ocean assimilation systems for two primary reasons; 1) assimilation of sea level anomalies requires a reference mean sea level surface which is generally unknown and may change slowly over time, 2) the use of relatively poor-quality surface forcing products in combination with high-quality ocean data results in an in-balance between the near-surface pressure gradient and the surface wind stress, especially in equatorial regions, leading to spurious circulations.

While one paper has already appeared on the implementation in the EnKF of a correction method for the mean sea level, this method has not been used for correction of the surface win stress, which is especially straightforward with the EnKF, at least when compared to the currently used IO methods. Furthermore, we propose to do a twin experiment in which the benefits of the bias correction method can be demonstrated explicitly rather than having to rely on innovation statistics. A third new aspect of the proposed project is the determination of balance/inbalance in the ocean states resulting from the EnKF when local analysis is applied. This has so far only been investigated in an atmospheric application.

ll software and data files are already installed on the HPCD or on ECFS. This means that the project can produce results very quickly. We only need computing resources to run the ensembles and the analyses.

For more details, please refer to the latest progress report.

Additional information

Project period 1997-2007