Optimal observations for variational data assimilation:
Determining the overturning in the North Atlantic

Armin Koehl

An important aspect of ocean state estimation is the design of an
observing system that allows to efficiently study climate aspects 
in the ocean.  A solution of a design problem is presented
here in terms of optimal observations that emerge as nondimensionalized
singular vectors of the data resolution matrix. It is demonstrated 
that such optimal observations when applied to determining the strength 
of the overturning stream-function in the North Atlantic at 30 N in 900 m
perform significantly better than section data in a 1 degree model. 
Preferred locations for hydrographic observations are primarily located 
along the western boundary north of 30N and along the eastern boundary 
south of 30 N and in the Iberian Sea. 

By revealing the main processes that accompany anomalies of the overturning, 
optimal observations are ideally suited for studying causes and effects of 
anomalies.  The response on short time-scales is mainly wind driven and 
accounts for about two thirds of the anomaly, it includes Eckman transport 
and coastal up- and down-welling.  A small part of the response is buoyancy 
driven and a slow response to primarily winter time anomalies in the Labrador 
Sea. Temperature and salinity anomalies in the initial conditions manifests 
induce alterations of the western boundary currents. The overall picture for
year 1997 is an east west density contrast that manifests primarily along 
the boundaries in the upper 1000 m.