computational capabilities, at the beginning of
WOCE. Remarkable advances in regional and
global-scale modelling during WOCE are closing
the gaps between observers and modellers. Analysis of many WOCE data sets now takes place with
reference to modelling, and most importantly, the
maturation of models and the collection of global
data sets is allowing ocean state estimation to
develop rapidly.
Should global observations continue to be taken?
The answer is a resounding yes, but clearly in a different way than during WOCE. The technologies
that were developed and put into widespread use
during WOCE should be included, so that the next
set of large-scale questions may be addressed. The
community has chosen to continue working with
problems that are more clearly the variability of the
ocean and atmosphere, given the large data set that
now exists to describe the ‘mean’ state. The appropriate technologies have been selected, mostly
involving autonomous measurements from satellites, floats, drifters and large moored arrays with
telemetry. Processes important to climate that were
not explored fully in WOCE, particularly in more
remote regions such as the southern hemisphere,
should be pursued. Should there be a future
WOCE-type hydrography survey? A complete repetition of WOCE hydrography in the near future is
unlikely and unnecessary given the long time scales
that these measurements address. However, monitoring of global carbon inventories and studying
the physical and biological processes that affect
them through use of transient and other tracers are
best accomplished from large-scale research ship
coverage at present and likely will remain so for at
least a decade, thus requiring at least partial coverage through the intervening years until a new global
survey is undertaken.
7.1.5.2 State estimation
Ocean data assimilation has long been a focus
of theoreticians and modellers. The tremendous
efforts of the community have brought us well
along the way to success. There are now several
ongoing WOCE-related assimilation efforts in the
US, Europe and Japan. Maturation of this ongoing
international work will result in fully successful
WOCE syntheses. Ongoing efforts in section and
box inversion and in ocean state estimation are
still largely preliminary. Model resolution, spatial
extent, and employed physics must further be
expanded and the entire WOCE data set must
eventually be incorporated. The results so far illustrate the potential for model–data combination.
Among many possible applications, one expects
the production of much better estimates of the
magnitudes and mechanisms of property flux
divergences, and the study of oceanic biogeochemical cycles that are dependent upon the circulation.
What progress is there in ocean state estimation? At the beginning of WOCE there was little
confidence in the prospect of bringing models and
data together in a full ocean state estimation. It
has now been shown that such a synthesis is technically feasible. Moreover, the field is rapidly migrating towards operational applications of ocean state
estimation, allowing us to estimate changes in the
ocean circulation on seasonal to longer climaterelevant time scales, similar to re-analysis projects
in the atmospheric community.
What are the gaps in knowledge needed for
state estimation? The attempt to accurately prescribe a priori model and data errors is as much a
state estimation problem as it is a fundamental
issue in data analysis and numerical modelling.
Such complete statistics are likely not possible for
every data type, and estimating them requires
close collaboration between observers and modellers. For climate applications, our knowledge
about details of the connection of the changing
atmospheric forcing with changes happening in the
deep ocean over hundreds of years is only rudimentary. However, if we are to estimate the deep
ocean state in the presence of large climate fluctuations, such as ENSO events, model drift has to be
separated from actual changes in the deep ocean
due to climate changes over, say, the last 500
years. There are also many technical issues to be
addressed, ranging from questions about efficiencies of methodologies, to the representation of the
ocean’s topography and its impact on connection
between basins, e.g. in the eastern North Atlantic
where the deep water is supplied through a narrow
fracture zone. The representation of mixing in general is a complex question and there is hope that
ocean state estimation will be able to contribute
substantially to estimates of the three-dimensional
and possibly time-varying mixing tensors.
What are the strategies for the future for state
estimation? A number of groups around the world
are developing the capability to assimilate oceanographic data. These efforts include those aimed at
SECTION 7 INSIGHTS FOR THE FUTURE
544
WOCE. Remarkable advances in regional and
global-scale modelling during WOCE are closing
the gaps between observers and modellers. Analysis of many WOCE data sets now takes place with
reference to modelling, and most importantly, the
maturation of models and the collection of global
data sets is allowing ocean state estimation to
develop rapidly.
Should global observations continue to be taken?
The answer is a resounding yes, but clearly in a different way than during WOCE. The technologies
that were developed and put into widespread use
during WOCE should be included, so that the next
set of large-scale questions may be addressed. The
community has chosen to continue working with
problems that are more clearly the variability of the
ocean and atmosphere, given the large data set that
now exists to describe the ‘mean’ state. The appropriate technologies have been selected, mostly
involving autonomous measurements from satellites, floats, drifters and large moored arrays with
telemetry. Processes important to climate that were
not explored fully in WOCE, particularly in more
remote regions such as the southern hemisphere,
should be pursued. Should there be a future
WOCE-type hydrography survey? A complete repetition of WOCE hydrography in the near future is
unlikely and unnecessary given the long time scales
that these measurements address. However, monitoring of global carbon inventories and studying
the physical and biological processes that affect
them through use of transient and other tracers are
best accomplished from large-scale research ship
coverage at present and likely will remain so for at
least a decade, thus requiring at least partial coverage through the intervening years until a new global
survey is undertaken.
7.1.5.2 State estimation
Ocean data assimilation has long been a focus
of theoreticians and modellers. The tremendous
efforts of the community have brought us well
along the way to success. There are now several
ongoing WOCE-related assimilation efforts in the
US, Europe and Japan. Maturation of this ongoing
international work will result in fully successful
WOCE syntheses. Ongoing efforts in section and
box inversion and in ocean state estimation are
still largely preliminary. Model resolution, spatial
extent, and employed physics must further be
expanded and the entire WOCE data set must
eventually be incorporated. The results so far illustrate the potential for model–data combination.
Among many possible applications, one expects
the production of much better estimates of the
magnitudes and mechanisms of property flux
divergences, and the study of oceanic biogeochemical cycles that are dependent upon the circulation.
What progress is there in ocean state estimation? At the beginning of WOCE there was little
confidence in the prospect of bringing models and
data together in a full ocean state estimation. It
has now been shown that such a synthesis is technically feasible. Moreover, the field is rapidly migrating towards operational applications of ocean state
estimation, allowing us to estimate changes in the
ocean circulation on seasonal to longer climaterelevant time scales, similar to re-analysis projects
in the atmospheric community.
What are the gaps in knowledge needed for
state estimation? The attempt to accurately prescribe a priori model and data errors is as much a
state estimation problem as it is a fundamental
issue in data analysis and numerical modelling.
Such complete statistics are likely not possible for
every data type, and estimating them requires
close collaboration between observers and modellers. For climate applications, our knowledge
about details of the connection of the changing
atmospheric forcing with changes happening in the
deep ocean over hundreds of years is only rudimentary. However, if we are to estimate the deep
ocean state in the presence of large climate fluctuations, such as ENSO events, model drift has to be
separated from actual changes in the deep ocean
due to climate changes over, say, the last 500
years. There are also many technical issues to be
addressed, ranging from questions about efficiencies of methodologies, to the representation of the
ocean’s topography and its impact on connection
between basins, e.g. in the eastern North Atlantic
where the deep water is supplied through a narrow
fracture zone. The representation of mixing in general is a complex question and there is hope that
ocean state estimation will be able to contribute
substantially to estimates of the three-dimensional
and possibly time-varying mixing tensors.
What are the strategies for the future for state
estimation? A number of groups around the world
are developing the capability to assimilate oceanographic data. These efforts include those aimed at
SECTION 7 INSIGHTS FOR THE FUTURE
544