Previous Abstract | Back to Abstracts Page | Next Abstract
Ensemble Optimal Interpolation: multivariate properties in the Gulf of Mexico
Counillon Francois, Bertino Laurent
Mohn-Sverdrup center/NERSC
(Abstract received 07/13/2007 for session X)
ABSTRACT
High-resolution models are needed in the Gulf of Mexico (GOM) to accurately reproduce mesoscale dynamics and the variability of the area. These cannot provide accurate locations of currents without data assimilation. The computer time requirements of high-resolution models constrain the choice of the data assimilation method. Therefore the Ensemble Optimal Interpolation (EnOI), derived from the Ensemble Kalman Filter (EnKF), is used here. The covariance matrix is extracted from an historical ensemble. It is 3-dimensional, multivariate, and conserves the linear balance (e.g. geostrophy). The correlations between Sea Level Anomaly (SLA) and velocity, temperature, salinity and layer thickness show good agreement with the dynamics of the area at two characteristic locations studied: (1) in central part of the GOM where high SLA is strongly related to the presence of an anticyclonic circulation; (2) in coastal areas where high SLA is related to seasonal variability. The multivariate relations between variables are reasonably linear, as assumed by the EnOI. Our data assimilation set up is dynamically balanced, even with exaggerated parameter. However, during the assimilation process a combination of members (i.e. temperature and salinity), initiates a density perturbation in the isopycnal layers, or “artificial caballing”. This introduced small high frequency noise in the model. The accuracy of the data assimilation system is tested consecutive to an assimilation on the 5th up to the shedding of the highly dynamic Eddy Yankee event on the 19th of August 2006, and beats obvious predictor. To clearly illustrate the multivariate properties of the EnOI, only the SLA has been assimilated, but other variables can potentially be assimilated.
Previous Abstract | Back to Abstracts Page | Next Abstract
2007 LOM Workshop, Bergen, Norway, August 20-22, 2007