then we must attribute far greater errors to the
estimates than we might hitherto have done. For
example, Roemmich and Cornuelle (1993) and
Gilson et al. (1998) were able to exploit the dense
XBT/XCTD sampling along the P14C track to
determine the representativeness of upper ocean
transport estimates from historical and contemporary hydrographic sections and to improve understanding of the relationship between steric height
and sea-level changes. Such results enhance the
value of data sets and provide improved estimates
of error. To the extent that WOCE has been able
to define sampling standards for measuring climate variations (see, for example, Bryden et al.,
1996; Dickson et al., Chapter 7.3), a powerful and
enduring legacy of WOCE will be that it provides
a baseline against which both future and past
changes can be measured.
7.4.6 Ocean variability and prediction:
GODAE
7.4.6.1 The changing nature of oceanography
One of the fundamental tenets of climate prediction and associated research is that the mechanisms involve interaction between two or more of
the physical earth systems components – the air,
oceans, land or ice. For seasonal time scales and
longer, the ocean is one of the important components: it provides the long-term memory, storing
and sequestering heat and water and releasing it
later, often at a different, remote location. Questions of predictability must be answered through
development of a thorough understanding of the
coupled modes of the climate system, with the
implication that modes that are ‘internal’ are of
secondary importance. In reality, there are probably few modes that are truly internal. Atmospheric
weather, for example in the form of tropical
storms, is known to interact with the underlying
ocean and provide both negative and positive feedbacks. Ocean eddies, though a product of oceanic
instability, evolve depending upon both the surface
forcing and the background state of the ocean.
WOCE treated the ocean as an independent
entity and did not mount observational or modelling projects focused on other components of the
climate system. To a large extent other WCRP programmes picked these up, though without any certainty that the specific interests of WOCE would be
included. Of course, the objectives of WOCE were
developed with a full recognition that the primary
rationale for undertaking WOCE was because of
the global ocean’s role in climate, particularly with
respect to climate change. For the atmosphere, we
are accustomed to treating weather prediction as
the primary rationale and treating its role within
the climate system as subsidiary. This is reasonable
since the dominant short-term benefit is through
weather prediction and associated applications.
Oceanography, it may be argued, has now reached
a similar point and we should develop our strategies around the multiple use of ocean observations
and the intrinsic benefits of observing and predicting the ocean for its own sake. This does not mean
we downplay the importance of climate, but that
we develop a much broader basis. We lift the study
of ocean variability and prediction to a higher
level, and place increased emphasis on the practical
benefits of routine and regular monitoring and prediction of the ocean. The investment on behalf of
climate prediction is enhanced through such a conceptual approach.
7.4.6.2 A problem in ocean estimation
We can reduce the problem to one of ocean state
estimation: How well can we estimate the state of
the ocean at any point in time given an irregularly
distributed suite of observations, with uneven
global coverage and variations in quality and type?
If we can develop the answer to this question, then
we can quantify the accuracy and information content provided to climate prediction and other applications with a dependency on ocean information.
Within the context of WOCE, these issues have
been discussed in earlier chapters of this volume.
In order to illustrate some of the nuances of ocean
estimation and, in particular, the relevance of
WOCE results, Figure 7.4.3 shows schematically
four different perspectives.
The situation in Figure 7.4.3a, a model assimilation and forecast sequence, is perhaps the most
familiar. The model is launched from a previously
determined state (the last analysis, t:t 0 9⌬t) and
integrated forward, assimilating synoptic observations to produce an estimate of the state at t:t 0 , the
initial condition (an analysis). From the current time
the model state is projected forward to produce a
prediction, in effect based on the extrapolation in
time of the information inserted prior
to t:t 0 . The data are in general not located on a
trajectory surface (line) of the model, so some
7.4 Ocean and Climate Prediction – the WOCE Legacy
597
Smith
estimates than we might hitherto have done. For
example, Roemmich and Cornuelle (1993) and
Gilson et al. (1998) were able to exploit the dense
XBT/XCTD sampling along the P14C track to
determine the representativeness of upper ocean
transport estimates from historical and contemporary hydrographic sections and to improve understanding of the relationship between steric height
and sea-level changes. Such results enhance the
value of data sets and provide improved estimates
of error. To the extent that WOCE has been able
to define sampling standards for measuring climate variations (see, for example, Bryden et al.,
1996; Dickson et al., Chapter 7.3), a powerful and
enduring legacy of WOCE will be that it provides
a baseline against which both future and past
changes can be measured.
7.4.6 Ocean variability and prediction:
GODAE
7.4.6.1 The changing nature of oceanography
One of the fundamental tenets of climate prediction and associated research is that the mechanisms involve interaction between two or more of
the physical earth systems components – the air,
oceans, land or ice. For seasonal time scales and
longer, the ocean is one of the important components: it provides the long-term memory, storing
and sequestering heat and water and releasing it
later, often at a different, remote location. Questions of predictability must be answered through
development of a thorough understanding of the
coupled modes of the climate system, with the
implication that modes that are ‘internal’ are of
secondary importance. In reality, there are probably few modes that are truly internal. Atmospheric
weather, for example in the form of tropical
storms, is known to interact with the underlying
ocean and provide both negative and positive feedbacks. Ocean eddies, though a product of oceanic
instability, evolve depending upon both the surface
forcing and the background state of the ocean.
WOCE treated the ocean as an independent
entity and did not mount observational or modelling projects focused on other components of the
climate system. To a large extent other WCRP programmes picked these up, though without any certainty that the specific interests of WOCE would be
included. Of course, the objectives of WOCE were
developed with a full recognition that the primary
rationale for undertaking WOCE was because of
the global ocean’s role in climate, particularly with
respect to climate change. For the atmosphere, we
are accustomed to treating weather prediction as
the primary rationale and treating its role within
the climate system as subsidiary. This is reasonable
since the dominant short-term benefit is through
weather prediction and associated applications.
Oceanography, it may be argued, has now reached
a similar point and we should develop our strategies around the multiple use of ocean observations
and the intrinsic benefits of observing and predicting the ocean for its own sake. This does not mean
we downplay the importance of climate, but that
we develop a much broader basis. We lift the study
of ocean variability and prediction to a higher
level, and place increased emphasis on the practical
benefits of routine and regular monitoring and prediction of the ocean. The investment on behalf of
climate prediction is enhanced through such a conceptual approach.
7.4.6.2 A problem in ocean estimation
We can reduce the problem to one of ocean state
estimation: How well can we estimate the state of
the ocean at any point in time given an irregularly
distributed suite of observations, with uneven
global coverage and variations in quality and type?
If we can develop the answer to this question, then
we can quantify the accuracy and information content provided to climate prediction and other applications with a dependency on ocean information.
Within the context of WOCE, these issues have
been discussed in earlier chapters of this volume.
In order to illustrate some of the nuances of ocean
estimation and, in particular, the relevance of
WOCE results, Figure 7.4.3 shows schematically
four different perspectives.
The situation in Figure 7.4.3a, a model assimilation and forecast sequence, is perhaps the most
familiar. The model is launched from a previously
determined state (the last analysis, t:t 0 9⌬t) and
integrated forward, assimilating synoptic observations to produce an estimate of the state at t:t 0 , the
initial condition (an analysis). From the current time
the model state is projected forward to produce a
prediction, in effect based on the extrapolation in
time of the information inserted prior
to t:t 0 . The data are in general not located on a
trajectory surface (line) of the model, so some
7.4 Ocean and Climate Prediction – the WOCE Legacy
597
Smith