A Weak Constraint 4D-Var Assimilation System for the Navy Coastal Ocean Model Using the Representer Method - Data Assimilation for Atmospheric, Oceanic and Hydrologic Applications - page 370

Fig. 15.2 The model domain
with bathymetry contours and
the profile locations,
including the numbered
profiles (in red )wherethe
assimilated solution is
evaluated

15.4.2 Results with MODAS Data

Starting from an initial condition on August 02, the model was integrated and
the assimilation performed for 5 days at a time, with the analysis at the end of
the 5 days becoming the initial condition for the following 5-day assimilation, the
overall assimilation experiment interval being 30 days.
In order to assess how well the assimilation fits the observations, the analysis
is examined at 5 locations in the model domain shown on Fig. 15.2 .These
locations are selected according to their geographic position with respect to the bay:
offshore (location 1), slightly outside of the bay mouth (location 2), inside the bay
(location 3), and south and north of the bay (locations 4 and 5). Results at locations
2 and 4 are similar to those at location 5, and therefore are not shown.
Examining the solution in the top 500 m at the offshore location 1, it can be seen
that the assimilation is able to correct large and small discrepancies between the
first guess and the observations for both the temperature and salinity fields, as seen
in Fig. 15.3 . In the first 5 days temperature discrepancies range between 2 K in the
upper 50 m, and about 1 K from 100 m and below. Likewise salinity discrepancies
range from 0.15psu in the upper 200 m to 0.05psu below. These discrepancies are
gradually corrected in the analysis (bottom panels of Fig. 15.3 ) and by the end
of the first 5-day assimilation window, they have vanished. For the subsequent
5-day assimilation windows, the model temperature and salinity appear to be well
constrained below 100 m with minimal to no discrepancies between the first guess
and the data. Discrepancies are confined to the upper 100 m. They are small at the
beginning of each 5-day window and grow with time. This is to be expected since
the first guess is initialized with the previous 5-day analysis at the final time, and
because the NOGAPS forcing fields are not necessarily compatible with MODAS
data. That the discrepancies are confined to the upper ocean also suggests that the
model error is driven by erroneous surface fluxes, although the simulation of the