Variational Data Assimilation for the Global Ocean - Data Assimilation for Atmospheric, Oceanic and Hydrologic Applications - page 316

13.4.3 Altimeter Sea Surface Height

Tab le 13.1 shows that most ocean observations are remotely sensed and measure
ocean variables only at the surface. The lack of synoptic real-time data at depth
places severe limitations on the ability of the data assimilation system to resolve and
maintain an adequate representation of the ocean mesoscale. Subsurface properties
in the ocean, therefore, must be inferred from surface-only observations. The
most important observing system for this purpose is satellite altimetry, which
measures the time varying change in SSH. Changes in sea level are strongly
correlated with changes in the depth of the thermocline in the ocean, and the ocean
dynamics generating sea level change are for the most part the mesoscale eddies
and meandering ocean fronts. The SSH data are provided as anomalies relative to a
time-mean field. The time mean removes the unknown geoid, but it also removes the
mean dynamic topography (MDT), which needs to be added back in order to allow
the data to be compared with model fields. The 3DVAR determines the satellite
altimeter SSH sampling locations in two alternative ways: (1) direct assimilation
of the along-track data at the observed locations, or (2) by first performing a 2D
horizontal analysis of SSH and then generate a sampling pattern of synthetic profile
locations within contours of sea level change that exceed a prescribed noise level
threshold (see Cummings 2005 for details). Once the altimeter sampling locations
are known there are two alternative methods available in the 3DVAR to project
the SSH data to depth in the form of synthetic temperature and salinity profiles.
One method is the Modular Ocean Data Assimilation System (MODAS) database,
which models the time averaged co-variability of dynamic height vs. temperature at
depth and temperature vs. salinity at a fixed location from an analysis of historical
profile data ( Fox et al. 2002 ). The MDT used in the MODAS method is derived from
historical hydrographic data. Note that an upgrade to the MODAS synthetic profile
capability, the Improved Synthetic Ocean Profile (ISOP) system ( Helber et al. 2012 ),
is currently being evaluated. The second “direct” method adjusts the model forecast
density field to be in agreement with the difference found between the model
forecast sea surface height field and the SSH measured by the altimeter ( Cooper
and Haines 1996 ). The MDT used in the direct method is the mean SSH from the
model derived from a hindcast run. Output of the direct method is in the form of
innovations of temperature and salinity from the forecast model background field,
which are directly input into the assimilation. An advantage of the direct method is
that it relies on model dynamics for its prior information rather than statistics fixed
at the start of the assimilation. However, a disadvantage is that it cannot explicitly
correct for forecast model errors in stratification due to model drift in the absence of
any real data constraints. MODAS does not suffer from these limitations, although
MODAS may have marginal skill due to: (1) sampling limitations of the historical
profile data, (2) non-steric signals in the altimeter data, or (3) weak correlations
between steric height and temperature at depth due to other factors, such as the
influence of salinity on steric height at high latitudes. Needless to say, neither of the
methods available for assimilating altimeter SSH data is ideal. A new method under