run as part of global coupled ocean–atmosphere
models, and are also coupled to biogeochemical
transport models, e.g. for carbon cycle and biological production studies. By and large, ocean climate models simulate the right magnitudes for the
large- scale transports of heat, carbon and other
sub-stances. However, they lack a proper representation of boundary currents and other important
small- and meso-scale processes, including water
mass transformation processes, which are crucial
for the thermohaline circulation, and are therefore
less well suited for direct comparison with ocean
observations.
When looking into the future, one is tempted to
draw an analogy to the situation at the beginning of
WOCE, and to predict that both lines of modelling
will ultimately converge. That process may take a
considerable time. As we will emphasize below, the
reason for this is fundamentally one of economics;
despite rapid and sustained increases in computer
power and availability, bridging the gap between
short-term, eddy-resolving simulations and global
coupled climate modelling of long duration is not
achievable in the near future on the basis of current
trends in computer performance alone.
The purpose of this chapter is to discuss some of
the issues that the modelling community is likely to
face in the next decade and beyond. While we will
not cover aspects of data assimilation into ocean
models (cf. Talley et al., Chapter 7.1), our discussion should apply to the prognostic component of
assimilation models. We begin with a discussion
of the space–time scales covered by current models,
and the prospects for improvement via increased
computing power. Next, we review several areas
wherein additional gains in model performance
might be sought, including the governing equations, sub-grid-scale parameterizations, and vertical
coordinates and spatial discretization. A considerable limitation on the fidelity of current ocean
models is the uncertainties in forcing by, and
coupling to, the atmosphere.
Lastly, as a cautionary remark, we note that
in the past important developments have often
occurred that were not anticipated at all. A striking
example is the hysteresis behaviour of the thermohaline circulation, and the associated question of instability and multiple stable equilibrium states (e.g.
Rahmstorf, 1995). The possibility of such a behaviour, which attracts so much interest today, was not
noticed in the early planning stage of WOCE.
7.2.2 Space–time scales of ocean
processes and models
The temporal and spatial scales of oceanic motions
extend over many decades. A frequency–wavenumber diagram is a convenient way to summarize
scales of processes, even though it constitutes a linear view of a non-linear system and hence is necessarily limited in its interpretation. Figure 7.2.1
indicates the temporal and spatial scales of several
relevant ocean processes. Also shown are the dispersion curves for gravity and planetary waves, in
each case barotropic, and the first four baroclinic
modes representative of mid-ocean mid-latitude
conditions. Due to limitations in the speed and
storage of computers, ocean models are restricted
to certain ranges of scales; these are shown for
several models typical of the present state-of-theart (rectangles in Fig. 7.2.1).
For global climate studies, the main interest frequently is on the ocean’s behaviour on spatial scales
from a thousand kilometres to global, and on temporal scales from a few months to many centuries
(and several thousands of years for certain palaeoclimate problems). These scales are at the heart of
the thermohaline and wind-driven ocean circulation, and are explicitly resolved by current ocean
climate models (labelled T in Fig. 7.2.1).
Motions on these large scales are however influenced by various smaller-scale processes. Boundary
currents with a scale of 50 km at mid-latitudes constitute an integral part of the circulation. Mesoscale
eddies interact with the circulation in various ways,
in particular through baroclinic/barotropic instability processes, and have similar spatial scales and
temporal scales of months. Obviously, ocean climate models are far from resolving these processes,
which therefore have to be parameterized (see
below), and even global eddy-permitting models
(labelled E in Fig. 7.2.1) are presently only marginally able to account for the dynamics at the scale of
the deformation radius.
At even smaller scales of a few kilometres, flows
through narrow straits and deep overflows across
sills can be crucial for the large-scale water mass
distribution. Convection processes that are fundamental for the thermohaline circulation have scales
of less than a kilometre. Ocean convection models (labelled C in Fig. 7.2.1) are appropriate to
describe the latter process. Finally, turbulence and
mixing (labelled L in Fig. 7.2.1) occurring at scales
SECTION 7 INSIGHTS FOR THE FUTURE
548
models, and are also coupled to biogeochemical
transport models, e.g. for carbon cycle and biological production studies. By and large, ocean climate models simulate the right magnitudes for the
large- scale transports of heat, carbon and other
sub-stances. However, they lack a proper representation of boundary currents and other important
small- and meso-scale processes, including water
mass transformation processes, which are crucial
for the thermohaline circulation, and are therefore
less well suited for direct comparison with ocean
observations.
When looking into the future, one is tempted to
draw an analogy to the situation at the beginning of
WOCE, and to predict that both lines of modelling
will ultimately converge. That process may take a
considerable time. As we will emphasize below, the
reason for this is fundamentally one of economics;
despite rapid and sustained increases in computer
power and availability, bridging the gap between
short-term, eddy-resolving simulations and global
coupled climate modelling of long duration is not
achievable in the near future on the basis of current
trends in computer performance alone.
The purpose of this chapter is to discuss some of
the issues that the modelling community is likely to
face in the next decade and beyond. While we will
not cover aspects of data assimilation into ocean
models (cf. Talley et al., Chapter 7.1), our discussion should apply to the prognostic component of
assimilation models. We begin with a discussion
of the space–time scales covered by current models,
and the prospects for improvement via increased
computing power. Next, we review several areas
wherein additional gains in model performance
might be sought, including the governing equations, sub-grid-scale parameterizations, and vertical
coordinates and spatial discretization. A considerable limitation on the fidelity of current ocean
models is the uncertainties in forcing by, and
coupling to, the atmosphere.
Lastly, as a cautionary remark, we note that
in the past important developments have often
occurred that were not anticipated at all. A striking
example is the hysteresis behaviour of the thermohaline circulation, and the associated question of instability and multiple stable equilibrium states (e.g.
Rahmstorf, 1995). The possibility of such a behaviour, which attracts so much interest today, was not
noticed in the early planning stage of WOCE.
7.2.2 Space–time scales of ocean
processes and models
The temporal and spatial scales of oceanic motions
extend over many decades. A frequency–wavenumber diagram is a convenient way to summarize
scales of processes, even though it constitutes a linear view of a non-linear system and hence is necessarily limited in its interpretation. Figure 7.2.1
indicates the temporal and spatial scales of several
relevant ocean processes. Also shown are the dispersion curves for gravity and planetary waves, in
each case barotropic, and the first four baroclinic
modes representative of mid-ocean mid-latitude
conditions. Due to limitations in the speed and
storage of computers, ocean models are restricted
to certain ranges of scales; these are shown for
several models typical of the present state-of-theart (rectangles in Fig. 7.2.1).
For global climate studies, the main interest frequently is on the ocean’s behaviour on spatial scales
from a thousand kilometres to global, and on temporal scales from a few months to many centuries
(and several thousands of years for certain palaeoclimate problems). These scales are at the heart of
the thermohaline and wind-driven ocean circulation, and are explicitly resolved by current ocean
climate models (labelled T in Fig. 7.2.1).
Motions on these large scales are however influenced by various smaller-scale processes. Boundary
currents with a scale of 50 km at mid-latitudes constitute an integral part of the circulation. Mesoscale
eddies interact with the circulation in various ways,
in particular through baroclinic/barotropic instability processes, and have similar spatial scales and
temporal scales of months. Obviously, ocean climate models are far from resolving these processes,
which therefore have to be parameterized (see
below), and even global eddy-permitting models
(labelled E in Fig. 7.2.1) are presently only marginally able to account for the dynamics at the scale of
the deformation radius.
At even smaller scales of a few kilometres, flows
through narrow straits and deep overflows across
sills can be crucial for the large-scale water mass
distribution. Convection processes that are fundamental for the thermohaline circulation have scales
of less than a kilometre. Ocean convection models (labelled C in Fig. 7.2.1) are appropriate to
describe the latter process. Finally, turbulence and
mixing (labelled L in Fig. 7.2.1) occurring at scales
SECTION 7 INSIGHTS FOR THE FUTURE
548