(Fig. 5.1.10b), it also loses density (Fig. 5.1.14a),
since the warming acts on warmer waters where the
coefficient of expansion ␣ is large, and the cooling
acts on cool waters where ␣ is small. A similar
effect was noted over the seasonal cycle by
Zahariev and Garrett (1997). This effect is balanced
by cabbeling in the interior (e.g. Davis, 1994).
Comparison of Figures 5.1.15a and 5.1.15b
with 5.1.14a of the Atlantic–Arctic basin indicates
that the large haline transformation in the latter
between 23 kg m
93 and 26 kg m
93 occurs south of
30°N. Between 23 and 25 kg m
93 it is largely compensated by the thermal component (Fig. 5.1.14a).
Further north, thermal transformation dominates
and in the density range of subtropical mode water
26.3–26.7 kg m
93 the negative slope gives 5 Sv as
the average rate at which this water mass formed
at the surface (Fig. 5.1.15b). A similar rate of subpolar mode water formation (27.5–28 kg m
93 )
north of 42.5°N is evident in Figure 5.1.15a. Note
the formation of very dense waters with densities
28–29 kg m
93 in the Mediterranean (included in
Figure 5.1.15b but not in Figure 5.1.14a).
The overall picture in the Pacific–Indian basin is
broadly similar, though skewed towards lighter
densities. Again there is large haline transformation at lighter densities, thermally compensated in
the range 21–24 kg m
93 . However, in contrast to
the North Atlantic, there is little formation of dense
waters in the subpolar North Pacific (Fig. 5.1.15d).
Air–sea fluxes in the subtropical North Pacific,
however, generate ϳ10 Sv of thermocline waters
(Fig. 5.1.15e), comparable to the production in the
subtropical North Atlantic (Fig. 5.1.15b). Figure
5.1.15c indicates that almost all the global transformation below about 22 kg m
93 occurs in the
tropical Pacific between 10°N and 10°S. In this
region about 20 Sv of light surface water is formed
at densities below 21.3 kg m
93
.
The plot of transformation in the Southern
Ocean, Figure 5.1.14c, shows less than 2 Sv of dense
water (27.5–28 kg m
93
) being formed, mostly thermally. It is unrealistic, given the poor fluxes and
temperature and salinity fields, to expect this data
set to show creation of the extreme water types
such as Weddell Sea Bottom water that mix
to form Antarctic Bottom water (Rintoul et al.,
Chapter 4.6). The formation of Antarctic Intermediate water of density 27–27.3 kg m
93 is only
evident as a negative slope in the haline transformation, giving about 5 Sv. Even the more limited
area of Figure 5.1.15f is too large to produce a
clear signature in the total transformation. However, there is a clear signal of the 30 Sv of Subantarctic Mode Water formation between 26.5 and
27 kg m
93
, which Speer et al. (1997) show is
mostly formed in the Indian Ocean sector. Over the
whole Southern Ocean (Fig. 5.1.14c) the creation
by surface fluxes of about 30 Sv of mode waters
with densities between 26 and 27 kg m
93 is consistent with the formation of 26 Sv of such waters in
the scheme shown in Chapter 4.6, Figure 4.6.12.
Our results show some differences from those
presented by Speer et al. (1995a) who used (i) fluxes
and (sometimes) SST from the COADS data set as
analysed by Oberhuber (1988) and Wright (1988),
together with Levitus SST/SSS, and also (ii) fluxes
from Isemer and Hasse (1987) for the North
Atlantic as a comparison. They of course did not
have fluxes for the Southern Ocean (south of
30–40°S). They found water mass transformation
peaking at ϳ30 Sv at densities of 26.5 kg m
93 in the
North Atlantic, rather than our values of ϳ15 Sv
(Fig. 5.1.15a,b). Our values are closer to those
found by Marsh (2000) for the North Atlantic, who
used the Southampton Oceanography Centre (SOC)
climatology (Josey et al., 1998, 1999), which covers
the period 1980–97. Using the unadjusted climatology, he found a peak transformation rate of ϳ14 Sv
at 27.5 kg m
93
, and ϳ9 Sv at 26.3 kg m
93
. If he corrected the SOC fluxes so as to balance the net
global heat balance (reducing short-wave by 8%
and increasing latent heat loss by 13%, following
da Silva et al. (1994)), he found instead peak rates
of ϳ15 Sv at 27.5 kg
93
, and ϳ19 Sv at 26.3 kg m
93
.
Ours and Marsh’s values are in fact more similar
to the maxima obtained in the model diagnoses
of Nurser et al. (1999) and Marshall et al. (1999)
of 15–20 Sv, which were forced by climatology
(Esbensen and Kushnir, 1981) and by 1992 NMC
fluxes respectively, plus relaxation.
These discrepancies emphasize the sensitivity of
the properties and formation rates of oceanic
water masses to the surface flux field (and indeed
to the SST and SSS fields). They may arise from
interannual variability rather than simply from
uncertainties in the fields. Using the unadjusted
SOC climatology, Marsh (2000) found very considerable interannual variability in the transformation in the North Atlantic. For instance, the peak
transformation at ϳ26.3 kg m
93 ranged from as
little as 1 Sv in 1980 to 16 Sv in 1987, compared
SECTION 5 FORMATION AND TRANSPORT OF WATER MASSES
334
since the warming acts on warmer waters where the
coefficient of expansion ␣ is large, and the cooling
acts on cool waters where ␣ is small. A similar
effect was noted over the seasonal cycle by
Zahariev and Garrett (1997). This effect is balanced
by cabbeling in the interior (e.g. Davis, 1994).
Comparison of Figures 5.1.15a and 5.1.15b
with 5.1.14a of the Atlantic–Arctic basin indicates
that the large haline transformation in the latter
between 23 kg m
93 and 26 kg m
93 occurs south of
30°N. Between 23 and 25 kg m
93 it is largely compensated by the thermal component (Fig. 5.1.14a).
Further north, thermal transformation dominates
and in the density range of subtropical mode water
26.3–26.7 kg m
93 the negative slope gives 5 Sv as
the average rate at which this water mass formed
at the surface (Fig. 5.1.15b). A similar rate of subpolar mode water formation (27.5–28 kg m
93 )
north of 42.5°N is evident in Figure 5.1.15a. Note
the formation of very dense waters with densities
28–29 kg m
93 in the Mediterranean (included in
Figure 5.1.15b but not in Figure 5.1.14a).
The overall picture in the Pacific–Indian basin is
broadly similar, though skewed towards lighter
densities. Again there is large haline transformation at lighter densities, thermally compensated in
the range 21–24 kg m
93 . However, in contrast to
the North Atlantic, there is little formation of dense
waters in the subpolar North Pacific (Fig. 5.1.15d).
Air–sea fluxes in the subtropical North Pacific,
however, generate ϳ10 Sv of thermocline waters
(Fig. 5.1.15e), comparable to the production in the
subtropical North Atlantic (Fig. 5.1.15b). Figure
5.1.15c indicates that almost all the global transformation below about 22 kg m
93 occurs in the
tropical Pacific between 10°N and 10°S. In this
region about 20 Sv of light surface water is formed
at densities below 21.3 kg m
93
.
The plot of transformation in the Southern
Ocean, Figure 5.1.14c, shows less than 2 Sv of dense
water (27.5–28 kg m
93
) being formed, mostly thermally. It is unrealistic, given the poor fluxes and
temperature and salinity fields, to expect this data
set to show creation of the extreme water types
such as Weddell Sea Bottom water that mix
to form Antarctic Bottom water (Rintoul et al.,
Chapter 4.6). The formation of Antarctic Intermediate water of density 27–27.3 kg m
93 is only
evident as a negative slope in the haline transformation, giving about 5 Sv. Even the more limited
area of Figure 5.1.15f is too large to produce a
clear signature in the total transformation. However, there is a clear signal of the 30 Sv of Subantarctic Mode Water formation between 26.5 and
27 kg m
93
, which Speer et al. (1997) show is
mostly formed in the Indian Ocean sector. Over the
whole Southern Ocean (Fig. 5.1.14c) the creation
by surface fluxes of about 30 Sv of mode waters
with densities between 26 and 27 kg m
93 is consistent with the formation of 26 Sv of such waters in
the scheme shown in Chapter 4.6, Figure 4.6.12.
Our results show some differences from those
presented by Speer et al. (1995a) who used (i) fluxes
and (sometimes) SST from the COADS data set as
analysed by Oberhuber (1988) and Wright (1988),
together with Levitus SST/SSS, and also (ii) fluxes
from Isemer and Hasse (1987) for the North
Atlantic as a comparison. They of course did not
have fluxes for the Southern Ocean (south of
30–40°S). They found water mass transformation
peaking at ϳ30 Sv at densities of 26.5 kg m
93 in the
North Atlantic, rather than our values of ϳ15 Sv
(Fig. 5.1.15a,b). Our values are closer to those
found by Marsh (2000) for the North Atlantic, who
used the Southampton Oceanography Centre (SOC)
climatology (Josey et al., 1998, 1999), which covers
the period 1980–97. Using the unadjusted climatology, he found a peak transformation rate of ϳ14 Sv
at 27.5 kg m
93
, and ϳ9 Sv at 26.3 kg m
93
. If he corrected the SOC fluxes so as to balance the net
global heat balance (reducing short-wave by 8%
and increasing latent heat loss by 13%, following
da Silva et al. (1994)), he found instead peak rates
of ϳ15 Sv at 27.5 kg
93
, and ϳ19 Sv at 26.3 kg m
93
.
Ours and Marsh’s values are in fact more similar
to the maxima obtained in the model diagnoses
of Nurser et al. (1999) and Marshall et al. (1999)
of 15–20 Sv, which were forced by climatology
(Esbensen and Kushnir, 1981) and by 1992 NMC
fluxes respectively, plus relaxation.
These discrepancies emphasize the sensitivity of
the properties and formation rates of oceanic
water masses to the surface flux field (and indeed
to the SST and SSS fields). They may arise from
interannual variability rather than simply from
uncertainties in the fields. Using the unadjusted
SOC climatology, Marsh (2000) found very considerable interannual variability in the transformation in the North Atlantic. For instance, the peak
transformation at ϳ26.3 kg m
93 ranged from as
little as 1 Sv in 1980 to 16 Sv in 1987, compared
SECTION 5 FORMATION AND TRANSPORT OF WATER MASSES
334