f q , the average of ͗H in ͘ is nearly 17 W m
92
,
with f s :1.0. It is not clear how best to correct the
surface flux estimates to reconcile this inconsistency, but something needs to be done. We prefer
to keep f q :0.94, so that the annual cycle of Q lat
in the North Atlantic then closely matches the
observed climatology from Kent and Taylor
(1995). There is little to be gained by adjusting the
under-sea-ice flux, because the insulation provided
by snow and ice cover makes the flux small, and
because the area is only about 10% of the ocean
surface. The simplest adjustment is to reduce uniformly the ice-free surface flux by 16.4 W m
92
.
Alternatively, the solar radiation can be suitably
reduced with f s :0.90. The three-year global
average of this corrected flux is 154 W m
92
.
The average latent, sensible and long-wave components remain at 995, 922 and 937 W m
92
,
respectively, and the corrected net flux, ͗H in ͘, is
near zero.
The global distribution of the adjusted net heat
flux (͗H in ͘ m, y 916.4 W m
92
), averaged over the
36 months, is shown in Figure 5.1.9 (see Plate
5.1.9, p. 428), which can be compared to climatological estimates shown by the WGASF (2000) and
the atlas of Beranger et al. (1999). All the features
of this three-year average are expected; equatorial
heating with peaks in the western Indian, eastern
Pacific and central Atlantic; strong heating in the
upwelling regions off the west coasts of continents;
weak heating over much of the Southern Ocean
and northeast Pacific; strong cooling of western
boundary currents (the Gulf Stream, Kuroshio and
Aghulas in particular); and pockets of strong cooling in northern high latitudes (Labrador Sea and
off Iceland). The strength of these features varies
widely between climatologies, but none in Figure
5.1.9 appear to be outliers. In particular, the
region of the eastern equatorial Pacific with more
than 100 W m
92 heating is similar to that of da
Silva et al. (1994). However, our adjustment in
this region may be compensating for a lack of
latent and sensible heat loss because of the weak
equatorial winds. There probably should be net
cooling along much of the Antarctic coast to represent the cooling of exposed surface water to the
freezing point. Further cooling is balanced by ice
formation, a form of ocean heating that should be
balanced primarily by additional ocean cooling
offshore where the Antarctic ice melts. Quantifying these sea-ice processes is beyond the scope of
this work, but clearly important, even though they
do not affect the global heat budget, only the
meridional distribution of heating.
The latitudinal variation of zonal averages (not
shown) from Figure 5.1.9 is very much like that of
the COADS (Comprehensive Ocean– Atmosphere
Data Set) climatology based on ship and buoy
observations (da Silva et al., 1994), but the amplitudes (cooling and heating maxima) tend to be
smaller. Between about 20°S and 60°N, where the
observations are most plentiful, the zonal averages
differ by at most <10 W m
92
. Such differences are
not large given the uncertainties in both (WGASF,
2000). The zonal averages are not so similar to
climatologies extracted directly from re-analysis
projects, including the NCEP (Beranger et al.,
1999). Differences with the latter are due to different radiation, the f q drying factor, different
transfer coefficients (Fig. 5.1.8), and the time
interval.
Included in the Beranger et al. (1999) atlas is
the Josey et al. (1999) COADS-based calculation
shown by Bryden and Imawaki (Chapter 6.1), who
discuss in more detail the issue of correcting for
global heat flux imbalance, and in particular the
30 W m
92 bias of this climatology. Compared with
Figure 5.1.9, it has generally greater heat flux
everywhere, but by less than the 30 W m
92 bias,
except in the Indian Ocean where the difference is
typically greater than 50 W m
92 . The atlas also
shows that differences between the zonal means
of the Josey et al. (1999) heat fluxes and other
climatologies, including Figure 5.1.9, are nearly
symmetrical about the equator. This result suggests
that poorer sampling in the southern hemisphere is
not the only source of the global imbalance.
Of more interest to WOCE, and oceanography
in general, are the meridional heat transports
implied by the mean surface heat flux. Since
this inference assumes no ocean heat storage it
becomes less robust for short time intervals such
as our three WOCE years. Also, Bryden and
Imawaki (Chapter 6.1) argue convincingly that the
error that accumulates in the integration of surface
heat flux over large areas becomes greater than the
uncertainty in direct estimates of ocean heat transport. If true, estimates can be used to constrain the
heat flux estimates, and in particular to judge
between Figure 5.1.9 (f s :1.0) and the alternative
f s :0.9. The global and Atlantic implied heat transports for both cases are shown in Figure 5.1.10.
5.1 Ocean Surface Water Mass Transformation
329
Large and Nurser
92
,
with f s :1.0. It is not clear how best to correct the
surface flux estimates to reconcile this inconsistency, but something needs to be done. We prefer
to keep f q :0.94, so that the annual cycle of Q lat
in the North Atlantic then closely matches the
observed climatology from Kent and Taylor
(1995). There is little to be gained by adjusting the
under-sea-ice flux, because the insulation provided
by snow and ice cover makes the flux small, and
because the area is only about 10% of the ocean
surface. The simplest adjustment is to reduce uniformly the ice-free surface flux by 16.4 W m
92
.
Alternatively, the solar radiation can be suitably
reduced with f s :0.90. The three-year global
average of this corrected flux is 154 W m
92
.
The average latent, sensible and long-wave components remain at 995, 922 and 937 W m
92
,
respectively, and the corrected net flux, ͗H in ͘, is
near zero.
The global distribution of the adjusted net heat
flux (͗H in ͘ m, y 916.4 W m
92
), averaged over the
36 months, is shown in Figure 5.1.9 (see Plate
5.1.9, p. 428), which can be compared to climatological estimates shown by the WGASF (2000) and
the atlas of Beranger et al. (1999). All the features
of this three-year average are expected; equatorial
heating with peaks in the western Indian, eastern
Pacific and central Atlantic; strong heating in the
upwelling regions off the west coasts of continents;
weak heating over much of the Southern Ocean
and northeast Pacific; strong cooling of western
boundary currents (the Gulf Stream, Kuroshio and
Aghulas in particular); and pockets of strong cooling in northern high latitudes (Labrador Sea and
off Iceland). The strength of these features varies
widely between climatologies, but none in Figure
5.1.9 appear to be outliers. In particular, the
region of the eastern equatorial Pacific with more
than 100 W m
92 heating is similar to that of da
Silva et al. (1994). However, our adjustment in
this region may be compensating for a lack of
latent and sensible heat loss because of the weak
equatorial winds. There probably should be net
cooling along much of the Antarctic coast to represent the cooling of exposed surface water to the
freezing point. Further cooling is balanced by ice
formation, a form of ocean heating that should be
balanced primarily by additional ocean cooling
offshore where the Antarctic ice melts. Quantifying these sea-ice processes is beyond the scope of
this work, but clearly important, even though they
do not affect the global heat budget, only the
meridional distribution of heating.
The latitudinal variation of zonal averages (not
shown) from Figure 5.1.9 is very much like that of
the COADS (Comprehensive Ocean– Atmosphere
Data Set) climatology based on ship and buoy
observations (da Silva et al., 1994), but the amplitudes (cooling and heating maxima) tend to be
smaller. Between about 20°S and 60°N, where the
observations are most plentiful, the zonal averages
differ by at most <10 W m
92
. Such differences are
not large given the uncertainties in both (WGASF,
2000). The zonal averages are not so similar to
climatologies extracted directly from re-analysis
projects, including the NCEP (Beranger et al.,
1999). Differences with the latter are due to different radiation, the f q drying factor, different
transfer coefficients (Fig. 5.1.8), and the time
interval.
Included in the Beranger et al. (1999) atlas is
the Josey et al. (1999) COADS-based calculation
shown by Bryden and Imawaki (Chapter 6.1), who
discuss in more detail the issue of correcting for
global heat flux imbalance, and in particular the
30 W m
92 bias of this climatology. Compared with
Figure 5.1.9, it has generally greater heat flux
everywhere, but by less than the 30 W m
92 bias,
except in the Indian Ocean where the difference is
typically greater than 50 W m
92 . The atlas also
shows that differences between the zonal means
of the Josey et al. (1999) heat fluxes and other
climatologies, including Figure 5.1.9, are nearly
symmetrical about the equator. This result suggests
that poorer sampling in the southern hemisphere is
not the only source of the global imbalance.
Of more interest to WOCE, and oceanography
in general, are the meridional heat transports
implied by the mean surface heat flux. Since
this inference assumes no ocean heat storage it
becomes less robust for short time intervals such
as our three WOCE years. Also, Bryden and
Imawaki (Chapter 6.1) argue convincingly that the
error that accumulates in the integration of surface
heat flux over large areas becomes greater than the
uncertainty in direct estimates of ocean heat transport. If true, estimates can be used to constrain the
heat flux estimates, and in particular to judge
between Figure 5.1.9 (f s :1.0) and the alternative
f s :0.9. The global and Atlantic implied heat transports for both cases are shown in Figure 5.1.10.
5.1 Ocean Surface Water Mass Transformation
329
Large and Nurser