116 Geir Evensen
cal section is taken from South to North along 49°W. The layer interfaces are
shown with the thick lines before the analysis and the thin lines after the analysis.
The contour labels is the number of the layer below that particular interface. The
layer distribution shows the deepening of the mixed layer in the North which is
MICOMs way of accounting for convection. Clearly, the increase in the SST, or
model mixed layer temperature, has lead to a shalIowing of the mixed layer. Further, the vertical influence is obvious for alI of the 11 layers in the model, which
has been moved upward in the water column. That is, the pycnocline has been
lifted and the vertical density structure has been changed in a dynamicalIy balanced
manner.
A more extensive discussion of the EnKF implementation with MICOM and an
application will be reported in a manuscript in preparation.
6.7 Summary
A general overview discussion has been given on sequential data assimilation for
nonlinear ocean models. The emphasis is laid on the ensemble Kalman filter
(EnKF) which has shown to be a promising approach suitable for the nonlinear
models used in oceanography. The theoretical basis for the EnKF was discussed in
detail and the capability of the filter to handle strongly nonlinear dynamics was
demonstrated in an example with the Lorenz equations. Preliminary results were
also presented from an implementation of the EnKF with the MICOM ocean general circulation model. This is the first reported implementation of an advanced
sequential data assimilatÎon technique, where proper evolution of error statistics
are taken into account, with a realistic OGCM. It was illustrated how the ensemble
statistics contain the required information for generating a dynamicalIy balanced
analysis and thereby also project surf ace information in the vertical in a consistent
manner.
A general conclusion so far is that the EnKF should be further implemented and
explored with other dynamical models. With a relatively acceptable numericalload
which corresponds to about 100-250 model integrations, the method can be used
for operational oceanography on extant computer resources. The parallel computers are also perfectly suited for ensemble integrations where each member can be
integrated independently on a separate processor.
Acknowledgments
The work was supported by European Commission through the Environment and
Climate program under contract ENV4-CT95-0l13 (AGORA) and by the Nordic
Council of Ministers contract FSIHFjlX-96001. Further, it received support from
the Norwegian Super Computing Committee (TRU) through a grant of computing
time.
cal section is taken from South to North along 49°W. The layer interfaces are
shown with the thick lines before the analysis and the thin lines after the analysis.
The contour labels is the number of the layer below that particular interface. The
layer distribution shows the deepening of the mixed layer in the North which is
MICOMs way of accounting for convection. Clearly, the increase in the SST, or
model mixed layer temperature, has lead to a shalIowing of the mixed layer. Further, the vertical influence is obvious for alI of the 11 layers in the model, which
has been moved upward in the water column. That is, the pycnocline has been
lifted and the vertical density structure has been changed in a dynamicalIy balanced
manner.
A more extensive discussion of the EnKF implementation with MICOM and an
application will be reported in a manuscript in preparation.
6.7 Summary
A general overview discussion has been given on sequential data assimilation for
nonlinear ocean models. The emphasis is laid on the ensemble Kalman filter
(EnKF) which has shown to be a promising approach suitable for the nonlinear
models used in oceanography. The theoretical basis for the EnKF was discussed in
detail and the capability of the filter to handle strongly nonlinear dynamics was
demonstrated in an example with the Lorenz equations. Preliminary results were
also presented from an implementation of the EnKF with the MICOM ocean general circulation model. This is the first reported implementation of an advanced
sequential data assimilatÎon technique, where proper evolution of error statistics
are taken into account, with a realistic OGCM. It was illustrated how the ensemble
statistics contain the required information for generating a dynamicalIy balanced
analysis and thereby also project surf ace information in the vertical in a consistent
manner.
A general conclusion so far is that the EnKF should be further implemented and
explored with other dynamical models. With a relatively acceptable numericalload
which corresponds to about 100-250 model integrations, the method can be used
for operational oceanography on extant computer resources. The parallel computers are also perfectly suited for ensemble integrations where each member can be
integrated independently on a separate processor.
Acknowledgments
The work was supported by European Commission through the Environment and
Climate program under contract ENV4-CT95-0l13 (AGORA) and by the Nordic
Council of Ministers contract FSIHFjlX-96001. Further, it received support from
the Norwegian Super Computing Committee (TRU) through a grant of computing
time.