estimating the state of the general circulation as
part of WOCE as well as those focusing on operational applications such as mesoscale forecasting
and predictions of seasonal-to-interannual climate
variability. Several of the major efforts focus on
operational assimilation and/or mesoscale forecasting. Various remote sensing and in-situ upper
ocean data sets are used, such as satellite altimetry,
sea surface temperature, upper ocean temperature
profiles from XBTs and profiling floats, and some
velocity information. Many of these data sets were
collected as part of WOCE and are continuing as
part of a global ocean observation system and/or
CLIVAR. Recent development and expansion in
assimilation activities along these lines is remarkable, and is continuing with substantial individual
and agency investments of time and funding. For
instance, the Global Ocean Data Assimilation
Experiment (GODAE) is an international initiative
centred on data assimilation and aims to conduct
regular, near-real-time global data assimilation
to demonstrate the practical utility of ocean
observations.
Only a few ocean state estimation efforts include
one of the basic WOCE data sets, hydrography,
and are focused on estimating a general circulation
rather than on prediction. An example of one such
effort in the US, in which the coauthors (DS and
IF) are engaged, is a consortium funded by NSF,
NASA and ONR – ‘Estimation of the Circulation
and Climate of the Ocean’ (ECCO), which is
building on the already existing MIT/SIO ocean
state estimation efforts and parallel ones at JPL.
The central technical goal of ECCO is a complete
global-scale ocean state estimation over the 15-year
period 1985–2000 at the highest possible resolution (given available computer resources) along
with a complete error description. But to develop
the technical understanding required to incorporate
all available data types, develop associated error
covariances, and to continue to improve models in
support of the global estimate, ECCO is beginning
with coarser-resolution global and basin-wide
models and developing near-real-time capabilities
in support of the climate community needs.
What is required for the future for successful
state estimation? The importance of accurately prescribing a priori model and data errors has been
discussed above in detail. Other issues include suitable access to data and computational resources.
Ready availability of observations is an obvious
requirement for the estimation process. Appropriate data reduction and quality control must be
applied to the observations. In light of the extent
of the WOCE observational data set, such processing, including data assemblage amenable for
model integration, is a non-trivial task.
A major limitation for ongoing state estimation
efforts in some countries is the available computational resources. In spite of advances made in
practical estimation, data assimilation is computationally intensive. Even moderate-sized general
circulation models strain extant supercomputers
when used in rigorous state estimation. The paradigm of shared resources at supercomputing facilities has so far failed to meet the requirements of
ocean state estimation in the US. An alternate
means of providing the necessary computational
resources is urgently needed.
Acknowledgements
Grant support from the National Science Foundation OCE-9712209 (LT), OCE-9730071 (DS) and
from NASA NAG5-7857 (DS) is acknowledged as
well as support from the Jet Propulsion Laboratory, California Institute of Technology, under
contract with the National Aeronautics and Space
Administration (IF).
7.1 Towards a WOCE Synthesis
545
Talley, Stammer and Fukumori
part of WOCE as well as those focusing on operational applications such as mesoscale forecasting
and predictions of seasonal-to-interannual climate
variability. Several of the major efforts focus on
operational assimilation and/or mesoscale forecasting. Various remote sensing and in-situ upper
ocean data sets are used, such as satellite altimetry,
sea surface temperature, upper ocean temperature
profiles from XBTs and profiling floats, and some
velocity information. Many of these data sets were
collected as part of WOCE and are continuing as
part of a global ocean observation system and/or
CLIVAR. Recent development and expansion in
assimilation activities along these lines is remarkable, and is continuing with substantial individual
and agency investments of time and funding. For
instance, the Global Ocean Data Assimilation
Experiment (GODAE) is an international initiative
centred on data assimilation and aims to conduct
regular, near-real-time global data assimilation
to demonstrate the practical utility of ocean
observations.
Only a few ocean state estimation efforts include
one of the basic WOCE data sets, hydrography,
and are focused on estimating a general circulation
rather than on prediction. An example of one such
effort in the US, in which the coauthors (DS and
IF) are engaged, is a consortium funded by NSF,
NASA and ONR – ‘Estimation of the Circulation
and Climate of the Ocean’ (ECCO), which is
building on the already existing MIT/SIO ocean
state estimation efforts and parallel ones at JPL.
The central technical goal of ECCO is a complete
global-scale ocean state estimation over the 15-year
period 1985–2000 at the highest possible resolution (given available computer resources) along
with a complete error description. But to develop
the technical understanding required to incorporate
all available data types, develop associated error
covariances, and to continue to improve models in
support of the global estimate, ECCO is beginning
with coarser-resolution global and basin-wide
models and developing near-real-time capabilities
in support of the climate community needs.
What is required for the future for successful
state estimation? The importance of accurately prescribing a priori model and data errors has been
discussed above in detail. Other issues include suitable access to data and computational resources.
Ready availability of observations is an obvious
requirement for the estimation process. Appropriate data reduction and quality control must be
applied to the observations. In light of the extent
of the WOCE observational data set, such processing, including data assemblage amenable for
model integration, is a non-trivial task.
A major limitation for ongoing state estimation
efforts in some countries is the available computational resources. In spite of advances made in
practical estimation, data assimilation is computationally intensive. Even moderate-sized general
circulation models strain extant supercomputers
when used in rigorous state estimation. The paradigm of shared resources at supercomputing facilities has so far failed to meet the requirements of
ocean state estimation in the US. An alternate
means of providing the necessary computational
resources is urgently needed.
Acknowledgements
Grant support from the National Science Foundation OCE-9712209 (LT), OCE-9730071 (DS) and
from NASA NAG5-7857 (DS) is acknowledged as
well as support from the Jet Propulsion Laboratory, California Institute of Technology, under
contract with the National Aeronautics and Space
Administration (IF).
7.1 Towards a WOCE Synthesis
545
Talley, Stammer and Fukumori