7.1.4.4 Observing system design and
sensitivity studies
In order to achieve maximum return from longterm climate observations, a well-designed observing system is required. Ocean models and data
assimilation systems are beginning now to support the design of such an observing system and to
test its usefulness to measure climate-relevant
quantities in a quantitative and yet cost-efficient way.
A posteriori errors with and without particular
data sets provide one measure of the impact of certain observing systems. For instance, by comparing
separate ocean estimations using different data
sets, Carton et al. (1996b) found TOPEX/POSEIDON altimeter data to have a larger impact in
resolving intraseasonal variability of the tropical
Pacific Ocean than data from a mooring array or a
network of XBTs.
Forward simulations also allow the testing of
sensitivity of model solutions to varying forcings,
initial conditions, and/or model parameters, by
deducing effects of particular aspects of the circulation on other elements of the model. In comparison, the model’s adjoint can be utilized to examine
the reverse sensitivity, namely the dependence
(rather than effect) of properties on other elements
of the model. An example of such use of the
adjoint is shown in Figure 7.1.12 (see Plate
7.1.12, p. 588; from Marotzke et al., 1999). The
figure shows the sensitivity of the MIT model’s
annual mean heat transport across 29°N in the
North Atlantic with respect to changes in initial
temperature and salinity fields near the surface
and at depth, respectively. The figure clearly identifies the influence of hydrography along the
boundaries on the estimated heat flux and identifies measurements along the western and eastern
boundaries as the most important measurement
for improving heat transport estimates at this latitude. Likewise one can study the impact of local
and remote forcing fields on heat transports, or,
for example, on the strength of the meridional
overturning circulation or any other scientifically
important quantity. Such information will lead to
a better understanding of physical adjustment
processes in the ocean and allow us to define key
measurements required to understand the response
of the ocean to atmospheric changes and to
accurately measure climate fluctuations in the
atmosphere–ocean system.
7.1.5 Summary and outlook
WOCE is contributing enormously to our ability
to describe and understand the ocean, and to
creating the technical capability needed through
development and large-scale use of new observing
methods, ocean models and state estimation.
These tools and understanding are being carried
forward as a basis for study of climate variations.
Our treatment of the WOCE synthesis above was
separated into a review of the more purely databased analyses and a progress report on ocean
state estimation, combining data with models and
including new results. Our conclusions are also
separated into these two areas.
7.1.5.1 Observation-based synthesis
WOCE was originally conceived of and designed
to produce a description of the large-scale ocean
circulation and its transports of heat, mass and
fresh water, to understand the climatic state of the
ocean and to understand the relevant processes
that set that state. The practical design was for
global in-situ coverage at the space and time scales
relevant for climate, with temporal as well as complete geographic coverage, and with complementary global satellite observations. It was recognized
at the outset that description of a ‘mean’ state is a
moving target in the sense that any specific period
of observations will produce a somewhat different
answer, with the differences being complex and
regional, depending on the state of the many different, largely uncoupled, climate modes. Several
major categories of WOCE observations help to
define the climate context, particularly the upper
ocean thermal and satellite observations. However
it is also clear, and remains clear after WOCE, that
a ‘mean’ state is a relevant concept. The WOCE
data sets and modelling are providing a major step
in learning about these largest-scale processes,
which one can think of as also including the statistics of the shorter time-scale variability that
depends on the existence of this mean state. Interannual to decadal climate variations provide in
most cases only about a 10% change to the mean
state, with the exception of the strong ENSO conditions that affected the upper ocean in the tropical Pacific during WOCE.
The comprehensiveness of the WOCE data set
both in types of measurements and geographical
coverage are truly exceptional. Never before has a
global view been possible, from ongoing satellite
SECTION 7 INSIGHTS FOR THE FUTURE
542
sensitivity studies
In order to achieve maximum return from longterm climate observations, a well-designed observing system is required. Ocean models and data
assimilation systems are beginning now to support the design of such an observing system and to
test its usefulness to measure climate-relevant
quantities in a quantitative and yet cost-efficient way.
A posteriori errors with and without particular
data sets provide one measure of the impact of certain observing systems. For instance, by comparing
separate ocean estimations using different data
sets, Carton et al. (1996b) found TOPEX/POSEIDON altimeter data to have a larger impact in
resolving intraseasonal variability of the tropical
Pacific Ocean than data from a mooring array or a
network of XBTs.
Forward simulations also allow the testing of
sensitivity of model solutions to varying forcings,
initial conditions, and/or model parameters, by
deducing effects of particular aspects of the circulation on other elements of the model. In comparison, the model’s adjoint can be utilized to examine
the reverse sensitivity, namely the dependence
(rather than effect) of properties on other elements
of the model. An example of such use of the
adjoint is shown in Figure 7.1.12 (see Plate
7.1.12, p. 588; from Marotzke et al., 1999). The
figure shows the sensitivity of the MIT model’s
annual mean heat transport across 29°N in the
North Atlantic with respect to changes in initial
temperature and salinity fields near the surface
and at depth, respectively. The figure clearly identifies the influence of hydrography along the
boundaries on the estimated heat flux and identifies measurements along the western and eastern
boundaries as the most important measurement
for improving heat transport estimates at this latitude. Likewise one can study the impact of local
and remote forcing fields on heat transports, or,
for example, on the strength of the meridional
overturning circulation or any other scientifically
important quantity. Such information will lead to
a better understanding of physical adjustment
processes in the ocean and allow us to define key
measurements required to understand the response
of the ocean to atmospheric changes and to
accurately measure climate fluctuations in the
atmosphere–ocean system.
7.1.5 Summary and outlook
WOCE is contributing enormously to our ability
to describe and understand the ocean, and to
creating the technical capability needed through
development and large-scale use of new observing
methods, ocean models and state estimation.
These tools and understanding are being carried
forward as a basis for study of climate variations.
Our treatment of the WOCE synthesis above was
separated into a review of the more purely databased analyses and a progress report on ocean
state estimation, combining data with models and
including new results. Our conclusions are also
separated into these two areas.
7.1.5.1 Observation-based synthesis
WOCE was originally conceived of and designed
to produce a description of the large-scale ocean
circulation and its transports of heat, mass and
fresh water, to understand the climatic state of the
ocean and to understand the relevant processes
that set that state. The practical design was for
global in-situ coverage at the space and time scales
relevant for climate, with temporal as well as complete geographic coverage, and with complementary global satellite observations. It was recognized
at the outset that description of a ‘mean’ state is a
moving target in the sense that any specific period
of observations will produce a somewhat different
answer, with the differences being complex and
regional, depending on the state of the many different, largely uncoupled, climate modes. Several
major categories of WOCE observations help to
define the climate context, particularly the upper
ocean thermal and satellite observations. However
it is also clear, and remains clear after WOCE, that
a ‘mean’ state is a relevant concept. The WOCE
data sets and modelling are providing a major step
in learning about these largest-scale processes,
which one can think of as also including the statistics of the shorter time-scale variability that
depends on the existence of this mean state. Interannual to decadal climate variations provide in
most cases only about a 10% change to the mean
state, with the exception of the strong ENSO conditions that affected the upper ocean in the tropical Pacific during WOCE.
The comprehensiveness of the WOCE data set
both in types of measurements and geographical
coverage are truly exceptional. Never before has a
global view been possible, from ongoing satellite
SECTION 7 INSIGHTS FOR THE FUTURE
542