play an important role in climate variability. This
topic is largely unexplored at present. The downstream behaviour of the overflow water is also
important. Downslope flows are currently rather
poorly represented in z-coordinate models, often
resulting in excessive mixing of dense overflow
waters, with significant downstream effects in the
subpolar gyre. Isopycnic coordinate models represent these downslope flows more naturally (see,
e.g., Roberts et al., 1996; DYNAMO Group, 1997;
Winton et al., 1998). Implementation of bottom
boundary layer models in z-coordinate models
(e.g. Beckmann and Döscher, 1997; Killworth and
Edwards, 1999) may alleviate this problem in
future.
The formation processes of AABW are complex
(Rintoul et al., Chapter 4.6), and many of the relevant processes (e.g. flow under ice shelves) are not
represented in the current generation of coupled
models. The importance of AABW in the climate
system is not as clear as for NADW, because the
heat transports involved are much less, but AABW
errors can still lead to heat transport errors of
O(0.1 PW) (see Section 2.3.5.2). Furthermore,
excessive AABW formation (associated with excess
northward transport of sea ice away from Antarctica) has been shown to result in century time scale
model drifts and excessive transport of the Antarctic
Circumpolar Current (Gent et al., 2000; see also
Rintoul et al., Chapter 4.6).
2.3.5.4 Ventilation and water mass formation
The transfer of heat, water and passive tracers
from the mixed layer to the ocean interior determines the time scale on which the mixed layer
(and hence the sea surface properties) responds to
a change in atmospheric forcing. For example, the
ventilation of the deep and intermediate water
masses provides a potential ‘heat sink’ that could
slow the rate of the climate’s response to increasing greenhouse gases. So correct modelling of water
mass formation processes is an important feature
of a climate model.
Figure 2.3.5 shows water mass transformation
rates (based on the framework of Speer and
Tziperman, 1992) for the Indian and Pacific
Oceans, in the NCAR CSM and derived from an
observational estimate of the surface buoyancy
flux (Doney et al., 1998b). The qualitative structure of the transformations is well captured in the
model, with surface fluxes tending to lighten the
light water masses and make dense water masses
denser, but the model transformation rates are
somewhat higher than those observed. This would
require higher rates of interior mixing in the
model, to balance the net destruction of the intermediate densities by the surface fluxes.
Another view of water mass formation and
ventilation is provided by the use of natural and
anthropogenic tracers such as
14 C and CFCs
(Chlorofluorocarbons – see England, 1999;
England and Maier-Reimer, 2000, for reviews). A
technical issue arises here that is peculiar to coupled models (England, 2000). The concentration
of a transient tracer in the ocean at a particular
time is a convolution of the history of the surface
flux with the ocean circulation and ventilation
processes. Errors in a simulated tracer field could
result from either errors in the fluxes or errors in
the circulation. If the primary aim of a coupled
model CFC simulation is to test the model’s ocean
circulation, use of observed winds to derive the
surface CFC flux may be more appropriate than
use of the model’s own winds (Dixon et al., 1996).
Figure 2.3.6a (see Plate 2.3.6a, p. 76) shows the
Atlantic zonal mean concentration of CFC-11 in
1982, from HadCM3 (observed winds were used
to calculate the surface flux in this case). The ventilation of deep waters in the North Atlantic is
clear, and a number of features of the North
Atlantic CFC distribution are similar to those seen
in the 1988 section reported by Doney and Bullister (1992), for example the downward spreading
of the low concentration isolines (:1 pmol l
91 )
between 25°N and 50°N, while the higher concentration isolines remain flat in this latitude band.
The overall pattern of CFC penetration (measured,
for example, by the depths of the 0.5, 1 and
2 pmol l
91 contours) agrees well with the observed
section (Fig. 2 of Doney and Bullister, 1992). In
the southern hemisphere, rather too much deep
penetration is seen when compared with the 1983
AJAX section (Warner and Weiss, 1992, Fig. 7),
probably due to excessive mixing in the model.
Simulated CFC fields are known to be sensitive
both to the surface flux formulation (England
et al., 1994; Dixon et al., 1996) and to subgrid-scale mixing parameterization (Robitaille and
Weaver, 1995; England and Hirst, 1997).
Finding a way of synthesizing the large amount
of transient tracer data collected during WOCE
and before into a form easily used by modellers is
SECTION 2 OBSERVATIONS AND MODELS
90
topic is largely unexplored at present. The downstream behaviour of the overflow water is also
important. Downslope flows are currently rather
poorly represented in z-coordinate models, often
resulting in excessive mixing of dense overflow
waters, with significant downstream effects in the
subpolar gyre. Isopycnic coordinate models represent these downslope flows more naturally (see,
e.g., Roberts et al., 1996; DYNAMO Group, 1997;
Winton et al., 1998). Implementation of bottom
boundary layer models in z-coordinate models
(e.g. Beckmann and Döscher, 1997; Killworth and
Edwards, 1999) may alleviate this problem in
future.
The formation processes of AABW are complex
(Rintoul et al., Chapter 4.6), and many of the relevant processes (e.g. flow under ice shelves) are not
represented in the current generation of coupled
models. The importance of AABW in the climate
system is not as clear as for NADW, because the
heat transports involved are much less, but AABW
errors can still lead to heat transport errors of
O(0.1 PW) (see Section 2.3.5.2). Furthermore,
excessive AABW formation (associated with excess
northward transport of sea ice away from Antarctica) has been shown to result in century time scale
model drifts and excessive transport of the Antarctic
Circumpolar Current (Gent et al., 2000; see also
Rintoul et al., Chapter 4.6).
2.3.5.4 Ventilation and water mass formation
The transfer of heat, water and passive tracers
from the mixed layer to the ocean interior determines the time scale on which the mixed layer
(and hence the sea surface properties) responds to
a change in atmospheric forcing. For example, the
ventilation of the deep and intermediate water
masses provides a potential ‘heat sink’ that could
slow the rate of the climate’s response to increasing greenhouse gases. So correct modelling of water
mass formation processes is an important feature
of a climate model.
Figure 2.3.5 shows water mass transformation
rates (based on the framework of Speer and
Tziperman, 1992) for the Indian and Pacific
Oceans, in the NCAR CSM and derived from an
observational estimate of the surface buoyancy
flux (Doney et al., 1998b). The qualitative structure of the transformations is well captured in the
model, with surface fluxes tending to lighten the
light water masses and make dense water masses
denser, but the model transformation rates are
somewhat higher than those observed. This would
require higher rates of interior mixing in the
model, to balance the net destruction of the intermediate densities by the surface fluxes.
Another view of water mass formation and
ventilation is provided by the use of natural and
anthropogenic tracers such as
14 C and CFCs
(Chlorofluorocarbons – see England, 1999;
England and Maier-Reimer, 2000, for reviews). A
technical issue arises here that is peculiar to coupled models (England, 2000). The concentration
of a transient tracer in the ocean at a particular
time is a convolution of the history of the surface
flux with the ocean circulation and ventilation
processes. Errors in a simulated tracer field could
result from either errors in the fluxes or errors in
the circulation. If the primary aim of a coupled
model CFC simulation is to test the model’s ocean
circulation, use of observed winds to derive the
surface CFC flux may be more appropriate than
use of the model’s own winds (Dixon et al., 1996).
Figure 2.3.6a (see Plate 2.3.6a, p. 76) shows the
Atlantic zonal mean concentration of CFC-11 in
1982, from HadCM3 (observed winds were used
to calculate the surface flux in this case). The ventilation of deep waters in the North Atlantic is
clear, and a number of features of the North
Atlantic CFC distribution are similar to those seen
in the 1988 section reported by Doney and Bullister (1992), for example the downward spreading
of the low concentration isolines (:1 pmol l
91 )
between 25°N and 50°N, while the higher concentration isolines remain flat in this latitude band.
The overall pattern of CFC penetration (measured,
for example, by the depths of the 0.5, 1 and
2 pmol l
91 contours) agrees well with the observed
section (Fig. 2 of Doney and Bullister, 1992). In
the southern hemisphere, rather too much deep
penetration is seen when compared with the 1983
AJAX section (Warner and Weiss, 1992, Fig. 7),
probably due to excessive mixing in the model.
Simulated CFC fields are known to be sensitive
both to the surface flux formulation (England
et al., 1994; Dixon et al., 1996) and to subgrid-scale mixing parameterization (Robitaille and
Weaver, 1995; England and Hirst, 1997).
Finding a way of synthesizing the large amount
of transient tracer data collected during WOCE
and before into a form easily used by modellers is
SECTION 2 OBSERVATIONS AND MODELS
90