Atlantic where they considered the direct estimates
to be reliable. While their air–sea heat exchanges
generally exhibited more heat gain by the ocean
than the divergence estimates, they argued that the
differences were within expected uncertainties in
the divergences, except for western boundary current regions where there may be strong heat losses
from the ocean that are not reflected in the climatological fluxes.
For residual estimates of ocean heat transport,
both Trenberth and Solomon (1994) and Keith
(1995) demonstrate that their residual estimates of
ocean heat transport, after they mask the continents, are in reasonable agreement with direct estimates. It is important to point out, however, that
the sum of atmospheric energy transport and such
residual estimates of ocean heat transport are still
about 1 PW less than required by the radiation
balance at the top of the atmosphere due to atmosphere–land exchanges. Whether such a deficit is
ultimately a problem with the atmospheric or
oceanic circulations remains the essential question
about the residual method.
6.1.5.2 Model ocean heat transports
Heat fluxes in ocean general circulation models
are commonly tested against direct estimates of
ocean heat transport (Bryan, 1982). In the
Atlantic, a model heat transport across 24°N that
is small compared with the direct estimate is usually a symptom of a meridional overturning in the
model that is of insufficient size or too shallow
(Böning et al., 1996). In the Pacific, small model
heat transport across 24°N is a symptom of a
problem in simulating the Kuroshio (Wilkin et al.,
1995). Comparing model heat transports with
well-determined direct estimates of ocean heat
transports has become a basic first test for the
validity of ocean and coupled atmosphere–ocean
general circulation models.
As models improve, the question arises as to
whether a model that has correctly simulated ocean
heat transport at a few latitudes can then be relied
upon to determine ocean heat transport at other
locations. It is a difficult question to answer. On
short time scales, an ocean model is nearly a global
or basin-scale inversion of the climatological data
used to initialize the model. Because ocean climatologies are horizontally sparse in the underlying
data coverage, fine features such as deep western
boundary currents may not be present in the climatology and experience suggests that inversion of a
data set without a significant circulation feature
can lead to biases in the results (Marotzke and
Wunsch, 1993). At this stage, model circulation
and heat transport must be critically compared
with the observed circulation and heat transport.
Over intermediate time scales of decades, ocean
models are slowly adjusting the initial circulation
to the air–sea fluxes forcing the model. The extent
to which the model heat transport divergences balance the air–sea fluxes becomes an important indicator of model equilibrium. Any imbalance means
that the model circulation is drifting toward a
state different from the initial or observed circulation. At this intermediate stage, three forms of
ocean heat transport need to be considered: the
heat transport associated with the model circulation; the heat transport associated with the model
air–sea exchanges; and the heat transport associated with the drift in model heat content. The drift
is a measure of the consistency between the model
circulation and the model air–sea forcing and can
be used as one measure of the uncertainty in
model heat transport. An example of such considerations is Saunders et al.’s (1999) analysis of the
Pacific Ocean heat fluxes in the global ocean
model OCCAM, where the effective surface fluxes
are determined by relaxation to observed surface
temperature and salinity. In OCCAM, the model
heat transport divergences over the Pacific are in
broad agreement with the Large et al. (1997) climatological estimates of air–sea flux and the globally averaged model drift in heat content is about
10 W m
92
. Such agreement may be taken to imply
the quality of both the model circulation and of
the air–sea forcing or flux climatology. Nevertheless, careful comparison with observed circulation
and heat transports are still required. For
OCCAM, the model heat transports are found to
be smaller than the direct estimates of ocean heat
transport across 10°N and 24°N in the Pacific,
and the reason may be smaller-than-normal wind
stresses in the tropics due to the model’s use of
winds during an El Niño year. Thus, even when a
model is near equilibrium, model heat transports
should be compared with the observed ocean circulation and heat transport to verify that the
air–sea forcing is realistic.
Coupled ocean–atmosphere models have usually required flux adjustment to keep the ocean (or
atmosphere) from drifting quickly away from its
6.1 Ocean Heat Transport
467
Bryden and Imawaki
to be reliable. While their air–sea heat exchanges
generally exhibited more heat gain by the ocean
than the divergence estimates, they argued that the
differences were within expected uncertainties in
the divergences, except for western boundary current regions where there may be strong heat losses
from the ocean that are not reflected in the climatological fluxes.
For residual estimates of ocean heat transport,
both Trenberth and Solomon (1994) and Keith
(1995) demonstrate that their residual estimates of
ocean heat transport, after they mask the continents, are in reasonable agreement with direct estimates. It is important to point out, however, that
the sum of atmospheric energy transport and such
residual estimates of ocean heat transport are still
about 1 PW less than required by the radiation
balance at the top of the atmosphere due to atmosphere–land exchanges. Whether such a deficit is
ultimately a problem with the atmospheric or
oceanic circulations remains the essential question
about the residual method.
6.1.5.2 Model ocean heat transports
Heat fluxes in ocean general circulation models
are commonly tested against direct estimates of
ocean heat transport (Bryan, 1982). In the
Atlantic, a model heat transport across 24°N that
is small compared with the direct estimate is usually a symptom of a meridional overturning in the
model that is of insufficient size or too shallow
(Böning et al., 1996). In the Pacific, small model
heat transport across 24°N is a symptom of a
problem in simulating the Kuroshio (Wilkin et al.,
1995). Comparing model heat transports with
well-determined direct estimates of ocean heat
transports has become a basic first test for the
validity of ocean and coupled atmosphere–ocean
general circulation models.
As models improve, the question arises as to
whether a model that has correctly simulated ocean
heat transport at a few latitudes can then be relied
upon to determine ocean heat transport at other
locations. It is a difficult question to answer. On
short time scales, an ocean model is nearly a global
or basin-scale inversion of the climatological data
used to initialize the model. Because ocean climatologies are horizontally sparse in the underlying
data coverage, fine features such as deep western
boundary currents may not be present in the climatology and experience suggests that inversion of a
data set without a significant circulation feature
can lead to biases in the results (Marotzke and
Wunsch, 1993). At this stage, model circulation
and heat transport must be critically compared
with the observed circulation and heat transport.
Over intermediate time scales of decades, ocean
models are slowly adjusting the initial circulation
to the air–sea fluxes forcing the model. The extent
to which the model heat transport divergences balance the air–sea fluxes becomes an important indicator of model equilibrium. Any imbalance means
that the model circulation is drifting toward a
state different from the initial or observed circulation. At this intermediate stage, three forms of
ocean heat transport need to be considered: the
heat transport associated with the model circulation; the heat transport associated with the model
air–sea exchanges; and the heat transport associated with the drift in model heat content. The drift
is a measure of the consistency between the model
circulation and the model air–sea forcing and can
be used as one measure of the uncertainty in
model heat transport. An example of such considerations is Saunders et al.’s (1999) analysis of the
Pacific Ocean heat fluxes in the global ocean
model OCCAM, where the effective surface fluxes
are determined by relaxation to observed surface
temperature and salinity. In OCCAM, the model
heat transport divergences over the Pacific are in
broad agreement with the Large et al. (1997) climatological estimates of air–sea flux and the globally averaged model drift in heat content is about
10 W m
92
. Such agreement may be taken to imply
the quality of both the model circulation and of
the air–sea forcing or flux climatology. Nevertheless, careful comparison with observed circulation
and heat transports are still required. For
OCCAM, the model heat transports are found to
be smaller than the direct estimates of ocean heat
transport across 10°N and 24°N in the Pacific,
and the reason may be smaller-than-normal wind
stresses in the tropics due to the model’s use of
winds during an El Niño year. Thus, even when a
model is near equilibrium, model heat transports
should be compared with the observed ocean circulation and heat transport to verify that the
air–sea forcing is realistic.
Coupled ocean–atmosphere models have usually required flux adjustment to keep the ocean (or
atmosphere) from drifting quickly away from its
6.1 Ocean Heat Transport
467
Bryden and Imawaki