state-of-the-art high-resolution models. Global
data sets, such as hydrographic atlases, are usually
very smooth, and do not show important features
of the ocean circulation at the deformation scale.
Satellite altimetry (cf. Fu, Chapter 3.3) has been
very useful for determining synoptic and seasonal
variability, but less so for longer-term means. One
may expect that with the AIMS-phase of WOCE
this situation will greatly improve, and that
analysed data sets of higher quality will become
available for model validation.
It is obvious that a model intercomparison does
not automatically lead to model improvement.
Nevertheless, we believe that tightly organized
intercomparison studies should be one important
element of an internationally coordinated numerical experimentation programme. As differences
between different models run under identical conditions are usually of the same order as differences
within one model when run under slightly different conditions, any model intercomparison should
always be accompanied by carefully planned sensitivity experiments. Perhaps the principal result of
such studies will be a better and more quantitative
appreciation of the shortcomings of different
model concepts. The open information on model
performance associated with an intercomparison
project will be very useful not only for model
development/improvement, but also for potential
users of models in other communities, and should
ultimately lead to an overall gain in model
credibility.
7.2.6 Concluding remarks
Speaking about the future – particularly in a
rapidly evolving field such as ocean circulation
modelling – is of course a highly uncertain undertaking. Several conclusions and paths to future
progress nonetheless appear reasonably certain.
At their current horizontal (and possibly vertical) resolution, and in view of the current state-ofthe-art in sub-grid-scale parameterizations, it
appears that current coupled climate models are
seriously misrepresenting many important oceanic
processes. These misrepresentations arise from
various sources, though many are attributable to
the unique complexities of the oceanic geometry
and topography. As a consequence, important
dynamical, thermodynamical and biogeochemical
processes, such as narrow western boundary currents, localized overflows, abyssal choke points
and coastal ecosystems, are partially or wholly
unresolved in today’s climate models. While the
impact of this deficiency on climate simulations
has yet to be established, it follows that systematic
efforts to document model deficiencies and to
identify the leading-order processes responsible
for these deficiencies need to be given a high priority. Necessary components of these efforts include
a coordinated numerical experimentation programme, continued community model development and involvement, and the utilization of a
more complete observational picture of the ocean
circulation as it becomes available through
WOCE-AIMS and other future programmes.
Furthermore, prognostic models need improvement to incorporate more correctly those processes
that are at present poorly represented. Taking it as
a matter of faith that the currently employed equations of motion are ‘good enough’ in their continuous form to reproduce all important processes,
two paths to progress are indicated. On the one
hand, as computer resources increase, continued
enhancements in spatial representation are needed,
in particular through better resolution but also
through use of higher-order methods, improved
(possibly adaptive) vertical coordinates, and nested
and/or unstructured spatial grids. On the other
hand, one has to accept that for the near future
models of coarse resolution are needed for the
study of climate variability on longer time scales.
Therefore, continuing improvement of parameterizations of those processes that require sub-gridscale treatment is also necessary. In particular, this
includes such processes as narrow flows along
boundaries, through straits and over sills, transports by mesoscale eddies, topographic interactions, small-scale diapycnal mixing, and others.
Last, and perhaps most importantly, we must
pay serious attention to improvements in how
atmosphere–ocean forcing and coupling are implemented. Of all the likely sources of uncertainty in
the ocean component of climate models, errors
arising from inaccurately known air–sea fluxes are
probably still dominant. Intensive, continuing
international efforts are required to provide sustained improvements in atmosphere–ocean flux
observations, and further understanding of poorly
understood component processes.
SECTION 7 INSIGHTS FOR THE FUTURE
556
data sets, such as hydrographic atlases, are usually
very smooth, and do not show important features
of the ocean circulation at the deformation scale.
Satellite altimetry (cf. Fu, Chapter 3.3) has been
very useful for determining synoptic and seasonal
variability, but less so for longer-term means. One
may expect that with the AIMS-phase of WOCE
this situation will greatly improve, and that
analysed data sets of higher quality will become
available for model validation.
It is obvious that a model intercomparison does
not automatically lead to model improvement.
Nevertheless, we believe that tightly organized
intercomparison studies should be one important
element of an internationally coordinated numerical experimentation programme. As differences
between different models run under identical conditions are usually of the same order as differences
within one model when run under slightly different conditions, any model intercomparison should
always be accompanied by carefully planned sensitivity experiments. Perhaps the principal result of
such studies will be a better and more quantitative
appreciation of the shortcomings of different
model concepts. The open information on model
performance associated with an intercomparison
project will be very useful not only for model
development/improvement, but also for potential
users of models in other communities, and should
ultimately lead to an overall gain in model
credibility.
7.2.6 Concluding remarks
Speaking about the future – particularly in a
rapidly evolving field such as ocean circulation
modelling – is of course a highly uncertain undertaking. Several conclusions and paths to future
progress nonetheless appear reasonably certain.
At their current horizontal (and possibly vertical) resolution, and in view of the current state-ofthe-art in sub-grid-scale parameterizations, it
appears that current coupled climate models are
seriously misrepresenting many important oceanic
processes. These misrepresentations arise from
various sources, though many are attributable to
the unique complexities of the oceanic geometry
and topography. As a consequence, important
dynamical, thermodynamical and biogeochemical
processes, such as narrow western boundary currents, localized overflows, abyssal choke points
and coastal ecosystems, are partially or wholly
unresolved in today’s climate models. While the
impact of this deficiency on climate simulations
has yet to be established, it follows that systematic
efforts to document model deficiencies and to
identify the leading-order processes responsible
for these deficiencies need to be given a high priority. Necessary components of these efforts include
a coordinated numerical experimentation programme, continued community model development and involvement, and the utilization of a
more complete observational picture of the ocean
circulation as it becomes available through
WOCE-AIMS and other future programmes.
Furthermore, prognostic models need improvement to incorporate more correctly those processes
that are at present poorly represented. Taking it as
a matter of faith that the currently employed equations of motion are ‘good enough’ in their continuous form to reproduce all important processes,
two paths to progress are indicated. On the one
hand, as computer resources increase, continued
enhancements in spatial representation are needed,
in particular through better resolution but also
through use of higher-order methods, improved
(possibly adaptive) vertical coordinates, and nested
and/or unstructured spatial grids. On the other
hand, one has to accept that for the near future
models of coarse resolution are needed for the
study of climate variability on longer time scales.
Therefore, continuing improvement of parameterizations of those processes that require sub-gridscale treatment is also necessary. In particular, this
includes such processes as narrow flows along
boundaries, through straits and over sills, transports by mesoscale eddies, topographic interactions, small-scale diapycnal mixing, and others.
Last, and perhaps most importantly, we must
pay serious attention to improvements in how
atmosphere–ocean forcing and coupling are implemented. Of all the likely sources of uncertainty in
the ocean component of climate models, errors
arising from inaccurately known air–sea fluxes are
probably still dominant. Intensive, continuing
international efforts are required to provide sustained improvements in atmosphere–ocean flux
observations, and further understanding of poorly
understood component processes.
SECTION 7 INSIGHTS FOR THE FUTURE
556