model factors. A particular example is the question of how model simulations of the meridional
transport of heat are affected by the grid resolution. Evaluation of CME model cases using 1°,
1/3°, and 1/6° resolution were rendered inconclusive because possible effects due to an improved
resolution of (boundary) currents were outweighed
by inevitable changes in the efficiency of boundary
zones; e.g. Beckmann et al. (1994b) obtained a
decrease in maximum poleward heat transport
when going from 1/3° to 1/6° resolution, possibly
caused by a less effective water mass transformation in the northern Irminger Basin. In a relevant
study, Fanning and Weaver (1997a) showed in a
series of idealized mid-latitude simulations with
increasing resolution that an important mechanism
of meridional heat transport at grid sizes below
1/2° is that of baroclinic gyre transport, which
comes to rival the heat transport of the zonally
invariant meridional overturning and leads to an
increase in total meridional heat transport. The
resolution issue was revisited recently in a very
ambitious model study by Smith et al. (2000).
Using a model configuration extending into the
Nordic Seas similar to that in DYNAMO, each
grid refinement, from 2/5° to 1/5°, and finally to
1/10°, yielded a significant increase in meridional
heat transport (Fig. 2.2.12). Interestingly, even the
leap to an eddy-resolving simulation closely conforms with the behaviour of the previous eddypermitting cases, i.e. closely falls on the regression
line of Fig. 2.2.11. Hence, understanding the
causes of the resolution dependence of heat transport in these cases is again complicated because of
potential local effects such as differences in the
simulation of transport and mixing processes in
the overflow regime. It also remains to be seen
whether the linear relation between heat transport
and meridional overturning breaks down when
fully global ocean models are used and when the
imposed surface fluxes are determined by two-way
interaction with an atmospheric model.
SECTION 2 OBSERVATIONS AND MODELS
76
20°S
0
2 0 °
40°
60°N
0.5
0
0.5
1
1.5
2
Petawatts
Trenberth 98
MacDonald 96
0.1
o
0.2
o
0.4
o
Fig. 2.2.12 Net meridional heat transport (upper curves) and transport of heat by the time-varying flow (lower
curves) for North Atlantic models at 0.1° (solid), 0.2° (dashed) and 0.4° (dotted) resolution.The uncertainty range
based on an atmospheric residual calculation (Trenberth, 1998a) is indicated by the shaded region; inverse model
results based on ocean observations of MacDonald and Wunsch (1996) are indicated by multiplication signs.
From Smith et al. (2000).
transport of heat are affected by the grid resolution. Evaluation of CME model cases using 1°,
1/3°, and 1/6° resolution were rendered inconclusive because possible effects due to an improved
resolution of (boundary) currents were outweighed
by inevitable changes in the efficiency of boundary
zones; e.g. Beckmann et al. (1994b) obtained a
decrease in maximum poleward heat transport
when going from 1/3° to 1/6° resolution, possibly
caused by a less effective water mass transformation in the northern Irminger Basin. In a relevant
study, Fanning and Weaver (1997a) showed in a
series of idealized mid-latitude simulations with
increasing resolution that an important mechanism
of meridional heat transport at grid sizes below
1/2° is that of baroclinic gyre transport, which
comes to rival the heat transport of the zonally
invariant meridional overturning and leads to an
increase in total meridional heat transport. The
resolution issue was revisited recently in a very
ambitious model study by Smith et al. (2000).
Using a model configuration extending into the
Nordic Seas similar to that in DYNAMO, each
grid refinement, from 2/5° to 1/5°, and finally to
1/10°, yielded a significant increase in meridional
heat transport (Fig. 2.2.12). Interestingly, even the
leap to an eddy-resolving simulation closely conforms with the behaviour of the previous eddypermitting cases, i.e. closely falls on the regression
line of Fig. 2.2.11. Hence, understanding the
causes of the resolution dependence of heat transport in these cases is again complicated because of
potential local effects such as differences in the
simulation of transport and mixing processes in
the overflow regime. It also remains to be seen
whether the linear relation between heat transport
and meridional overturning breaks down when
fully global ocean models are used and when the
imposed surface fluxes are determined by two-way
interaction with an atmospheric model.
SECTION 2 OBSERVATIONS AND MODELS
76
20°S
0
2 0 °
40°
60°N
0.5
0
0.5
1
1.5
2
Petawatts
Trenberth 98
MacDonald 96
0.1
o
0.2
o
0.4
o
Fig. 2.2.12 Net meridional heat transport (upper curves) and transport of heat by the time-varying flow (lower
curves) for North Atlantic models at 0.1° (solid), 0.2° (dashed) and 0.4° (dotted) resolution.The uncertainty range
based on an atmospheric residual calculation (Trenberth, 1998a) is indicated by the shaded region; inverse model
results based on ocean observations of MacDonald and Wunsch (1996) are indicated by multiplication signs.
From Smith et al. (2000).