Chapter 6.2. Climatologies are available from
these many different sources. The surface flux Special Analysis Center has its own data products for
the North Atlantic, tropical Pacific and tropical
Indian Oceans, including monthly means and
anomalies for many years as well as overall means.
Other primary sources are the NCEP and ECMWF
re-analysis projects, climatologies produced by the
Southampton Oceanography Centre (Josey et al.,
1999), and a large number of products accessible
through the online LDEO Data System.
Discriminating between these many different
products, which are relatively easy to access
through the WOCE Special Analysis Center website, has been a large challenge for WOCE. A major
and much-desired recent activity of the air–sea flux
community has been the comparison of the many
available flux products and parameterizations. An
international working group designated by SCOR
(Scientific Committee on Ocean Research) and the
WCRP has recently completed this comparison.
This Working Group on Air–Sea Fluxes (WGASF)
considered the requirements for surface flux data
sets, determined what data sets are currently available, and analysed their strengths and weaknesses.
It is not possible to summarize their results briefly
here; the reader is referred to their report
(WGASF, 2000).
7.1.3 Model evaluation and development
WOCE synthesis involves the development and
testing of models for a range of purposes including:
(1) revealing and studying the physics of processes
relevant to the general circulation; (2) simulation
of the ocean circulation with good representation
of eddy statistics important for defining the semipermanent features of the circulation; (3) good
simulation of the ocean circulation at resolutions
that are useful for coupled climate modelling; and
(4) use in ocean state estimation as a way of providing the most consistent representation of the
observed fields. The first three topics are covered
well in other chapters in this book.
Models used for ocean state estimation must be
evaluated with respect to their realism in simulating observed processes. The models must be reasonably consistent with ocean data prior to using
them for state estimation. Such comparisons are a
usual step in using the models to understand and
simulate processes that have not been observed as
a result of undersampling. By incorporating data
in the models through state estimation, another
stage of model evaluation is reached, ultimately
leading also to model improvements where needed
(e.g. see Fu and Smith, 1996; Stammer et al.,
1996). The two examples discussed here that indicate that models, at least in some of their state
variables, show an intriguing agreement with what
has actually been observed in the ocean.
A comparison of sea surface height time series
simulated by the POCM 4C model (R. Tokmakian,
personal communication; see Chapter 2.2) with
data from five tide gauge stations on the west coast
of the Americas is given in Fig. 2.2.3 of Böning
and Semtner (Chapter 2.2). The model noticeably
simulates most of the observed climate events, at
least qualitatively – the correlation coefficients
vary between 0.57 and 0.88. A similarly encouraging result can be found from a comparison of
velocity fields from the MIT model (Marshall
et al., 1997a,b) with TOGA-TAO velocity data at
0°N, 140°W (Fig. 7.1.4; T. Lee, personal communication, 1999). The comparison indicates reasonable skill in simulating the time mean and
variability of the upper equatorial Pacific circulation, an important prerequisite for using the model
to study seasonal to interannual variability. The
mean flow field shows some biases, however, in
the core of the undercurrent.
Uncertainties in prescribed mixing coefficients
and surface forcing fields are believed to cause
these observed model–data differences. Model
development efforts aim at reducing such differences in order to provide a best possible model for
process-oriented studies and data assimilation. See
Böning and Semtner (Chapter 2.2) and Willebrand
and Haidvogel (Chapter 7.2) for detailed discussions of ongoing and anticipated model improvement activities.
It should be noted that ocean state estimation
often includes estimation of surface flux fields that
bring the model into consistency with ocean observations. Resulting estimates need to be consistent
with meteorological centre-provided fields and their
uncertainties. Inconsistencies in turn will shed light
on errors in the atmospheric and ocean models.
7.1.4 Ocean state estimation
While data-based studies make use primarily of
simple models, such as geostrophy and mass
7.1 Towards a WOCE Synthesis
535
Talley, Stammer and Fukumori
these many different sources. The surface flux Special Analysis Center has its own data products for
the North Atlantic, tropical Pacific and tropical
Indian Oceans, including monthly means and
anomalies for many years as well as overall means.
Other primary sources are the NCEP and ECMWF
re-analysis projects, climatologies produced by the
Southampton Oceanography Centre (Josey et al.,
1999), and a large number of products accessible
through the online LDEO Data System.
Discriminating between these many different
products, which are relatively easy to access
through the WOCE Special Analysis Center website, has been a large challenge for WOCE. A major
and much-desired recent activity of the air–sea flux
community has been the comparison of the many
available flux products and parameterizations. An
international working group designated by SCOR
(Scientific Committee on Ocean Research) and the
WCRP has recently completed this comparison.
This Working Group on Air–Sea Fluxes (WGASF)
considered the requirements for surface flux data
sets, determined what data sets are currently available, and analysed their strengths and weaknesses.
It is not possible to summarize their results briefly
here; the reader is referred to their report
(WGASF, 2000).
7.1.3 Model evaluation and development
WOCE synthesis involves the development and
testing of models for a range of purposes including:
(1) revealing and studying the physics of processes
relevant to the general circulation; (2) simulation
of the ocean circulation with good representation
of eddy statistics important for defining the semipermanent features of the circulation; (3) good
simulation of the ocean circulation at resolutions
that are useful for coupled climate modelling; and
(4) use in ocean state estimation as a way of providing the most consistent representation of the
observed fields. The first three topics are covered
well in other chapters in this book.
Models used for ocean state estimation must be
evaluated with respect to their realism in simulating observed processes. The models must be reasonably consistent with ocean data prior to using
them for state estimation. Such comparisons are a
usual step in using the models to understand and
simulate processes that have not been observed as
a result of undersampling. By incorporating data
in the models through state estimation, another
stage of model evaluation is reached, ultimately
leading also to model improvements where needed
(e.g. see Fu and Smith, 1996; Stammer et al.,
1996). The two examples discussed here that indicate that models, at least in some of their state
variables, show an intriguing agreement with what
has actually been observed in the ocean.
A comparison of sea surface height time series
simulated by the POCM 4C model (R. Tokmakian,
personal communication; see Chapter 2.2) with
data from five tide gauge stations on the west coast
of the Americas is given in Fig. 2.2.3 of Böning
and Semtner (Chapter 2.2). The model noticeably
simulates most of the observed climate events, at
least qualitatively – the correlation coefficients
vary between 0.57 and 0.88. A similarly encouraging result can be found from a comparison of
velocity fields from the MIT model (Marshall
et al., 1997a,b) with TOGA-TAO velocity data at
0°N, 140°W (Fig. 7.1.4; T. Lee, personal communication, 1999). The comparison indicates reasonable skill in simulating the time mean and
variability of the upper equatorial Pacific circulation, an important prerequisite for using the model
to study seasonal to interannual variability. The
mean flow field shows some biases, however, in
the core of the undercurrent.
Uncertainties in prescribed mixing coefficients
and surface forcing fields are believed to cause
these observed model–data differences. Model
development efforts aim at reducing such differences in order to provide a best possible model for
process-oriented studies and data assimilation. See
Böning and Semtner (Chapter 2.2) and Willebrand
and Haidvogel (Chapter 7.2) for detailed discussions of ongoing and anticipated model improvement activities.
It should be noted that ocean state estimation
often includes estimation of surface flux fields that
bring the model into consistency with ocean observations. Resulting estimates need to be consistent
with meteorological centre-provided fields and their
uncertainties. Inconsistencies in turn will shed light
on errors in the atmospheric and ocean models.
7.1.4 Ocean state estimation
While data-based studies make use primarily of
simple models, such as geostrophy and mass
7.1 Towards a WOCE Synthesis
535
Talley, Stammer and Fukumori