An advanced assimilation method based on incremental four-dimensional variational assimilation
has been developed in a tropical Pacific configuration of the OPA Ocean General Circulation Model.
This article evaluates the impact of assimilating in situ temperature data in the OPA model and
validates model fields (currents, dynamic height anomalies) against independent data. Assimilation
of temperature data is shown to improve significantly the model representation of the tropical
system of currents and countercurrents, as well as the amplitude and variability of sea level
Recent studies have underlined the importance of ocean data assimilation for seasonal forecasts
with coupled ocean-atmosphere models [Alves et al., 1998; Ji et al., 1997; Rosati et al., 1997].
In particular, to predict seasonal anomalies associated with the El Niño-Southern Oscillation,
initial conditions of the surface and subsurface ocean state in the tropical Pacific need to
be set accurately. For this purpose, an advanced assimilation method based on incremental
four-dimensional variational assimilation (4D-Var) has been developed for the OPA Ocean General
Circulation Model (OGCM) [Madec et al., 1998] conceived by the Laboratoire d'Océanographie
Dynamique et de Climatologie (LODYC). While the general characteristics of the assimilation
system have been presented in Weaver and Vialard , this paper highlights the improvement
in ocean analyses after assimilation of in situ temperature data. The evaluation in this
configuration is a first step towards implementing the 4D-Var system (called OPAVAR) in ORCA,
the global ocean configuration of OPA, assimilating in situ temperature and salinity data as
well as TOPEX/POSEIDON and Jason altimeter data.
Basically, the OPAVAR scheme has three main characteristics. First, it is multivariate in
the sense that the model parameters (currently temperature, salinity and horizontal currents)
used to set the "analyzed" initial conditions of the control vector are all adjusted simultaneously
when assimilating data. Second, this scheme is incremental in that minimization between
observations and the model occurs month after month over one-month periods. Third, a major
hypothesis is that the model is perfect and, thus, the "analyzed" trajectory exactly verifies
the ocean model equations.
Impact of in situ temperature assimilation
In the equatorial region, the major impacts of assimilating observed in situ temperature data
are to correct model mean state biases and to improve its variability. In particular, assimilation
of temperature data considerably improves the model representation of the thermocline tightness
in the equatorial region.
This improvement can be seen in figure 1 in the central and eastern Pacific in the upper
ocean. First, below 150 meters, the model without data assimilation simulates a very weak
zonal slope of the subsurface isotherms. In contrast, the model with data assimilation
simulates a west-to-east slope characterizing a more intense upwelling at the subsurface
in the eastern Pacific. Second, above 150 meters, the model with data assimilation exhibits
a much tighter and shallower thermocline. For instance, at 110°W, the 15°C isotherm is
upwelled by about 80 meters when temperature data are assimilated. Its is likely that such
a change in the thermal structure of the equatorial upper ocean has an impact on seasonal
Impacts on simulated currents
Here, in response to the tightening of the tropical thermocline, horizontal pressure
gradients intensify. This leads to the strengthening of the tropical system of currents
and countercurrents. In particular, at the surface (not shown) the model mean surface currents
are in better agreement with Reverdin et al.  climatology. Such an improvement can be
observed in a meridional section at 140°W (figure 2). At the surface and at depth, the North
Equatorial Counter Current is much more intense than in the model without assimilation. The
tighter and shallower thermocline in the eastern Pacific is responsible for an intensification
of the Equatorial Under Current, which contributes to a reduction of the equatorial surface
currents east of 110°W and thus to the splitting in two branches of the South Equatorial
Finally, the variability of the surface currents is also improved along the Equator. For
example, at 140°W the correlation of the model zonal current to TAO surface zonal current
over the 1993-1998 period improves from 0.75 to 0.81 while the rms difference decreases
from 0.21 to 0.19 m/s. Therefore, the OPAVAR scheme significantly improves the tropical
upper ocean currents when assimilating only in situ temperature data.
Comparisons with TOPEX/POSEIDON data
To evaluate the impact of in situ temperature data assimilation on sea level anomalies,
the model dynamic height anomalies (referenced to 1,000 dbar) are compared to TOPEX/POSEIDON
(T/P) sea level anomalies over the 1993-1998 period (figure 3). Without assimilation, the
amplitude of the simulated interannual anomaly variability exhibits a reasonable pattern,
but it is still weak compared to T/P [see Vialard et al., 2001]. When assimilating temperature
data, the model recovers the 15% to 20% that were lacking in the simulation in the equatorial
band without assimilation. This result is confirmed when comparing the sea level time series:
the 4D-Var assimilation improves both the correlation and rms difference with T/P over the
entire equatorial Pacific basin (not shown).
Conclusion and perspectives
A 4D-Var assimilation scheme in a tropical Pacific configuration of the OPA model gives
very promising results. Assimilating only in situ temperature data improves the thermal
description of the upper ocean, as well as the model currents and the model dynamic height
anomalies. This result is very encouraging for the future, as these developments are a first
step towards implementing this 4D-Var scheme in ORCA, the global configuration of the OPA model,
with a view to assimilating in situ temperature and salinity measurements and altimetric data.
Alves J.O., D.L.T. Anderson, T.N. Stockdale, M.A. Balmaseda, J. Segschneider, 1998: Sensitivity of ENSO forecasts to ocean initial conditions. Proceedings of the Triangle'98 Symposium, September 1998, Kyoto, Japan.
Ji M., A. Leetma, 1997: Impact of data assimilation on ocean initialization and El Niño prediction. Mon. Wea. Rev., 125, 742-753.
Madec G., P. Delecluse, M. Imbard, C. Lévy, 1998: OPA 8.1 Ocean General Circulation Model reference manual. Note du Pôle de modélisation, Institut Pierre-Simon Laplace, N°11, 91, pp.
Reverdin G., C. Frankignoul, E. Kastenare, M.J. McPhaden, 1994: Seasonal variability in the surface currents of the equatorial Pacific. J. Geophys. Res., 99, 20323-20344
Rosati A., K. Miyakoda, R. Gudgel, 1997: The impact of ocean initial conditions on ENSO forecasting with a coupled model. Mon. Wea. Rev., 125, 754-772.
Vialard J., C. Menkes, J.P. Boulanger, P. Delecluse, E. Guilyardi, M.J. McPhaden. 2001: Oceanic mechanisms driving the SST during the 1997-1998 El Niño. J. Phys. Oceanogr. (in press).
Weaver A.T., J. Vialard, 1999: Development of an ocean incremental 4D-Var scheme for seasonal prediction. in Proceedings of the Third WMO International Symposium on Assimilation of Observations in Meteorology and Oceanograpy. 7-11. June 1999, Quebec City, Canada.