model with a representation of the observed ocean
mean state. This method has the obvious advantage of greatly simplifying the spin-up problem.
A second advantage is that it allows the coupled
model perturbations to grow upon a background
(mean) state that resembles reality, thus enhancing
our ability to represent correctly the growth and
pathways of slow climate variations. Clearly the
WOCE data set and the knowledge developed in
ocean state estimation provides a powerful baseline for future studies.
The assessment of past and current rates of sealevel rise has been accorded special attention in climate change assessments (Warrick and Oerlemans,
1990; Warrick et al., 1996). The need for highquality global sea-level measurements gelled nicely
with the emphasis within WOCE on a high-quality
and timely global sea-level measurement network.
It is clear from a recent review (Ocean Observations Panel for Climate, 1997) that WOCE
research and results have profoundly influenced
the strategy for monitoring and assessing the rate
of sea-level rise.
Knowledge of the influence of ocean processes
and in particular of thermal expansion on sea level
has grown through WOCE observations and
WOCE research. Bryden et al. (1996), Bindoff and
McDougall (2000) and Levitus et al. (2000) are just
three of the many studies that have exploited hydrographic and time series data to identify significant
changes in oceanic properties over long periods,
some of which are consistent with ocean thermal
expansion driven by increasing temperatures.
The extent to which these observations collectively can determine the mean state of the ocean
remains an open question (see also the discussion
in Section 7.4.1 and Talley et al., Chapter 7.1).
Wunsch (1977) was the first to provide a general
framework for the determination of the general
circulation of the oceans. He adapted methods
developed in control theory to provide an estimate
of the ocean circulation, consistent with observations and knowledge of dynamics and physics,
under the assumption that the circulation was
invariant with time. Such methods have been
applied many times in WOCE (see, for example,
Rintoul et al., Chapter 4.6; Bryden et al., Chapter
6.1). The planning for WOCE addressed the problem of temporal variability by requiring the surveys in each basin to be ‘synoptic’ (that is, the
survey times should be short compared with the
periods of the dominant variability), and by planning for selected measurements that would resolve
well the key modes of temporal variability. Time
series data do show considerable variability over
long time scales and the mean over the WOCE
period can often differ considerably from the longterm mean (see Fig. 1 of Levitus et al., 2000, for
example). There is also considerable spatial variability, suggesting that we should be cautious
about overinterpreting the representativeness of
the WOCE data for the mean state (Dickson et al.,
Chapter 7.3). Indeed, the strategy for the WOCE
synthesis phase (WOCE International Project
Office, 1997) stresses the need for interpreting the
WOCE snapshot within the context of observed
(and unobserved) variability.
The concept of a mean state has now been
largely replaced by images of an ocean with rich
temporal and spatial variability. Theory, models
and data suggest that the climate system has
natural modes of variability on long time scales so
that it is more appropriate to think of an everchanging and adjusting ocean and climate system
rather than one with a single mean state. WOCE
will provide the most comprehensive estimate to
date of the ocean state, an estimate that will also
take full account of uncertainties. Talley et al.
(Chapter 7.1) and Dickson et al. (Chapter 7.3)
address many of the issues for the WOCE synthesis related to longer-term variability, issues that
ultimately dictate the utility and representativeness
of the 1990s for past and future times. A key
theme that is emerging in the approach is that the
assumption of temporal invariance must be relaxed
to accommodate seasonal and longer-period variations in ocean state estimation (WOCE International Project Office, 1997). Such an approach
extracts greater value from the data (the temporal
variations become part of the signal rather than
treated as noise) and provides an improved framework for evaluating the representativeness of the
WOCE snapshot.
The impact of the WOCE data is clear, even
dominant, in present collections of global ocean
data. In a perverse way, it has to some extent
weakened the confidence held in historical analyses. It has become clear that older data lacked
the accuracy and precision required to delineate
subtle climate variations and that, if the variability
of the oceans in the past resembled the variability
of the oceans today (a relatively safe assumption),
SECTION 7 INSIGHTS FOR THE FUTURE
596
mean state. This method has the obvious advantage of greatly simplifying the spin-up problem.
A second advantage is that it allows the coupled
model perturbations to grow upon a background
(mean) state that resembles reality, thus enhancing
our ability to represent correctly the growth and
pathways of slow climate variations. Clearly the
WOCE data set and the knowledge developed in
ocean state estimation provides a powerful baseline for future studies.
The assessment of past and current rates of sealevel rise has been accorded special attention in climate change assessments (Warrick and Oerlemans,
1990; Warrick et al., 1996). The need for highquality global sea-level measurements gelled nicely
with the emphasis within WOCE on a high-quality
and timely global sea-level measurement network.
It is clear from a recent review (Ocean Observations Panel for Climate, 1997) that WOCE
research and results have profoundly influenced
the strategy for monitoring and assessing the rate
of sea-level rise.
Knowledge of the influence of ocean processes
and in particular of thermal expansion on sea level
has grown through WOCE observations and
WOCE research. Bryden et al. (1996), Bindoff and
McDougall (2000) and Levitus et al. (2000) are just
three of the many studies that have exploited hydrographic and time series data to identify significant
changes in oceanic properties over long periods,
some of which are consistent with ocean thermal
expansion driven by increasing temperatures.
The extent to which these observations collectively can determine the mean state of the ocean
remains an open question (see also the discussion
in Section 7.4.1 and Talley et al., Chapter 7.1).
Wunsch (1977) was the first to provide a general
framework for the determination of the general
circulation of the oceans. He adapted methods
developed in control theory to provide an estimate
of the ocean circulation, consistent with observations and knowledge of dynamics and physics,
under the assumption that the circulation was
invariant with time. Such methods have been
applied many times in WOCE (see, for example,
Rintoul et al., Chapter 4.6; Bryden et al., Chapter
6.1). The planning for WOCE addressed the problem of temporal variability by requiring the surveys in each basin to be ‘synoptic’ (that is, the
survey times should be short compared with the
periods of the dominant variability), and by planning for selected measurements that would resolve
well the key modes of temporal variability. Time
series data do show considerable variability over
long time scales and the mean over the WOCE
period can often differ considerably from the longterm mean (see Fig. 1 of Levitus et al., 2000, for
example). There is also considerable spatial variability, suggesting that we should be cautious
about overinterpreting the representativeness of
the WOCE data for the mean state (Dickson et al.,
Chapter 7.3). Indeed, the strategy for the WOCE
synthesis phase (WOCE International Project
Office, 1997) stresses the need for interpreting the
WOCE snapshot within the context of observed
(and unobserved) variability.
The concept of a mean state has now been
largely replaced by images of an ocean with rich
temporal and spatial variability. Theory, models
and data suggest that the climate system has
natural modes of variability on long time scales so
that it is more appropriate to think of an everchanging and adjusting ocean and climate system
rather than one with a single mean state. WOCE
will provide the most comprehensive estimate to
date of the ocean state, an estimate that will also
take full account of uncertainties. Talley et al.
(Chapter 7.1) and Dickson et al. (Chapter 7.3)
address many of the issues for the WOCE synthesis related to longer-term variability, issues that
ultimately dictate the utility and representativeness
of the 1990s for past and future times. A key
theme that is emerging in the approach is that the
assumption of temporal invariance must be relaxed
to accommodate seasonal and longer-period variations in ocean state estimation (WOCE International Project Office, 1997). Such an approach
extracts greater value from the data (the temporal
variations become part of the signal rather than
treated as noise) and provides an improved framework for evaluating the representativeness of the
WOCE snapshot.
The impact of the WOCE data is clear, even
dominant, in present collections of global ocean
data. In a perverse way, it has to some extent
weakened the confidence held in historical analyses. It has become clear that older data lacked
the accuracy and precision required to delineate
subtle climate variations and that, if the variability
of the oceans in the past resembled the variability
of the oceans today (a relatively safe assumption),
SECTION 7 INSIGHTS FOR THE FUTURE
596