Geoscience Reference
In-Depth Information
14.2.5
Background Quality Control Checks
A new feature of the ROMS 4D-Var systems is the recent introduction of a
background quality control of the observations to reject those data that are subject to
gross errors (
Hollingsworth et al. 1986
;
Lorenc and Hammon 1988
). The approach
used is based on that described by
Jarvinen and Unden
(
1997
)and
Andersson and
Jarvinen
(
1999
) and used in NWP. Specifically, the elements of the innovation vector
d
are compared with their expected error (assuming observation and background
errors are uncorrelated errors) according to:
y
i
y
i
2
=
b
<˛
1
C
o
=
b
(14.9)
y
i
y
i
where
is the
i
th observation,
is the
i
th element of the vector
H.
x
b
/
,the
background evaluated at the observation locations, and
b
are the standard
deviations of the observation and background errors at the observation points. The
threshold parameter
o
and
generally depends on each observation type, and appropriate
values can be determined from historical analyses as described by
Andersson and
Jarvinen
(
1999
). Observations that do not satisfy (
14.9
) are rejected prior to the
analysis. This has the effect of eliminating from the analysis observations that are
subject to large gross errors. In addition, observations that are inconsistent with the
model due, for example, to model deficiencies, are also eliminated from the analysis.
The introduction of the background quality control in ROMS based on (
14.9
)
has been found to yield substantial improvement in the behavior and convergence
properties of the dual 4D-Var algorithm in particular.
˛
14.3
Configuration of ROMS CCS and 4D-Var
The California Current System (CCS) is an eastern boundary current characterized
by a pronounced seasonal cycle of upwelling and by energetic mesoscale circula-
tions (
Hickey 1998
;
Checkley and Barth 2009
), and provides a challenge for linear
data assimilation methods such as 4D-Var. The ROMS CCS domain and circulation
is described in detail by
Veneziani et al.
(
2009
)and
Broquet et al.
(
2009a
,b),and
spans the region
134
ı
Wto
116
ı
Wand
31
ı
Nto
48
ı
N, with 1/10th degree resolution
in the horizontal and
42
-levels in the vertical. The model domain and bathymetry
are shown in Fig.
14.1
.
The model forcing is derived from either 6 hourly or daily averaged atmo-
spheric boundary layer fields from different sources depending on the application.
In the case of the 13 year reanalysis project described in Sect.
14.4.1
and the
near real-time analysis system described in Sect.
14.5
, daily averaged atmospheric
variables at standard heights were taken from the Naval Research Laboratory's
Coupled Ocean-Atmosphere Mesoscale Prediction System (COAMPS) (
Doyle
Search WWH ::
Custom Search