Geoscience Reference
In-Depth Information
cross-track geostrophic velocities computed using the LOESS filter for one day of
along-track altimeter data (10 January 2012). It is readily apparent that a tremendous
amount of mesoscale oceanographic information is contained in the geostrophic
velocities derived from the along-track altimeter data.
Once the altimeter SSH along-track geostrophic currents are calculated the
model equivalents are determined. Cross-track geostrophic velocity relative to a
deep level of no motion (2,000 m) is computed from the model using dynamic
height differences at points adjacent to the along-track estimate of the SSH slope.
The difference between the vertically averaged model and altimeter cross-track
geostrophic velocities is used to correct the relative geostrophic shear from the
model and form the velocity profile u a .
z
/
for the assimilation according to:
u a .
z
/ D u g .
z
/ u g C c
(13.12)
where u g .
is the model relative geostrophic shear profile, u g is its vertical average,
and c is the integral cross track velocity component calculated from the altimeter
slope. Assimilation of the u
z
/
v velocity vectors formed this way via the multivariate
correlations in the 3DVAR provide balanced geopotential increments, which in
turn are decomposed into balanced temperature and salinity increments using a
linearized equation of state. The velocity profiles in this scheme are very sensitive
to the reference level of no motion. One option here is to use Argo trajectory data to
infer a time dependent geopotential field at the float parking depth (cf. Davis 2005 ).
A dynamic geopotential field would go a long way in solving a long-standing
problem of hydrography: properly referencing geostrophic shear.
;
13.7.4
Hybrid Ensemble Four Dimensional Data Assimilation
A four-dimensional (4D) ensemble-enhanced data assimilation scheme for global
HYCOM is being developed to better deal with the late receipt, temporally
distributed observations than the current 3DVAR methodology. As previously noted,
a crucial aspect of all ocean data assimilation schemes is the way in which the
background error covariances are specified. The data assimilation process is optimal
if the background error covariances are perfectly known, which is never the case.
A major challenge then is to find ways to estimate accurate and comprehensive
background error covariances. Ensemble methods provide a method for doing this,
including the ability to provide a flow-dependent estimate of the background error
covariances.
When ensemble covariances are used in a variational data assimilation frame-
work to augment the existing background-error covariance, analyses are further
improved. This method is called a hybrid ensemble variational method. In com-
parison with conventional ensemble-based data assimilation, a hybrid scheme is
attractive for the following reasons. First, the hybrid schemes build upon existing
variational systems enabling the ensemble information to be incorporated relatively
Search WWH ::




Custom Search