with the average ϳ9 Sv. Our average is only over a
three-year period and so may not be representative
of the long term.
This issue of interannual variability in the fluxes
is addressed in Figure 5.1.16 (see Plate 5.1.16,
p. 428). It shows annual flux anomalies between
1982 and 1997. Note that the radiative forcing
everywhere and the precipitation in the tropical
Pacific and Indian Oceans after 1993 are climatological. Our three-year analysis period of 1991–93
was most unrepresentative of the later WOCE
years, 1994–97, in the following respects: the positive freshwater flux anomalies in the equatorial
Indian and Pacific, the negative freshwater flux
anomaly in the equatorial Atlantic, and the greater
ocean heating throughout the Atlantic, extratropical Indian and the Pacific south of about 30°N.
The WOCE years (1991–97) differ from the earlier
years (1982–90) in several ways. There are generally negative heat flux anomalies at all latitudes of
the Indian Ocean. The equatorial Pacific heat and
freshwater flux anomalies are smaller in amplitude. There is heating over most of the South
Atlantic. The amplitude of the heat flux anomalies
associated with the North Atlantic Oscillation
(NAO) (north of 45°N in the Atlantic) is smaller.
In Figure 5.1.16 (see Plate 5.1.16, p. 428) an ‘average’ year would be characterized by a vertical zero
contour in all six panels. No such year exists, with
1992 perhaps the closest. No short period (e.g. our
analysis period, or the WOCE period) appears
to be ‘typical’ of either another short period, or
of a longer period. Prime contributors to this
observation are the irregular variability of the El
Niño-Southern Oscillation (ENSO) and the NAO.
5.1.6 Summary
In this chapter we have set out Walin’s (1982) formal theory of water mass transformation. We have
diagnosed transformation driven by surface fluxes
across density classes in various regions of the
World Ocean, over three years of the WOCE period.
Our results cover almost all the World Ocean. They
are broadly similar to, but show distinct differences
from, other diagnoses using different climatologies.
These differences arise partly from uncertainties in
the flux fields, but also from their substantial
(Marsh, 2000) temporal variability. Our results are
derived from observations over a three-year period,
which may not be representative of the long term.
The sensitivity of these transformation rates to
the surface fluxes emphasizes just how important
accurate knowledge of the surface fluxes is.
Numerical ocean models that are driven by incorrect fluxes acting on incorrect surface temperature
and density distributions will not contain the correct water masses. Models forced by relaxation
and/or climatological fluxes can easily develop
regions of ‘mismatch’ between surface fluxes and
surface properties. The strong spurious fluxes acting on fluid that is too warm and saline over an
incorrectly separated Gulf Stream are a classic
example of this problem. Coupled models, e.g. by
Doney et al. (1998b), which take into account the
finite heat capacity of the atmosphere offer more
hope in this regard.
Care must be taken in relating the formation
rates, M, derived from differentiation of the transformation rate driven by surface fluxes, F, to ‘subduction’ and the injection of water masses into the
main thermocline. In the first instance, scaling
analysis and numerical model runs (Garrett and
Tandon, 1997; Marshall et al., 1999; Nurser et al.,
1999; Tandon and Zahariev, 2000) suggest that
diffusive processes within the mixed layer and seasonal thermocline, while relatively unimportant at
higher densities, do profoundly change the transformation rates at lighter densities :ϳ25.5 m
93
.
For these lighter waters the total transformation
G calculated over the mixed layer and seasonal
thermocline can be very different from F, even
having the opposite sign (Marshall et al., 1999;
Rhines, 1993; Tandon and Garrett, 1997). Second,
the domains over which we integrate to find
G may include regions of entrainment as well as
subduction. The entrainment and subduction will
tend to cancel out, leaving a net formation rate
considerably smaller than the total subduction of
fluid from that outcrop. Over an open domain,
however, M is now the difference between the
fluid subducted down into the thermocline, and
the input of mixed layer and seasonal thermocline
waters of that density across the open boundary of
the domain. This may give something different
from the subduction or ventilation rate.
Despite these caveats, these water mass diagnostics are particularly useful in understanding the
generation of denser mode and intermediate waters,
where diffusive and recirculative processes are
not such an issue. They are powerful tools for
understanding the diabatic circulation, offering the
5.1 Ocean Surface Water Mass Transformation
335
Large and Nurser
three-year period and so may not be representative
of the long term.
This issue of interannual variability in the fluxes
is addressed in Figure 5.1.16 (see Plate 5.1.16,
p. 428). It shows annual flux anomalies between
1982 and 1997. Note that the radiative forcing
everywhere and the precipitation in the tropical
Pacific and Indian Oceans after 1993 are climatological. Our three-year analysis period of 1991–93
was most unrepresentative of the later WOCE
years, 1994–97, in the following respects: the positive freshwater flux anomalies in the equatorial
Indian and Pacific, the negative freshwater flux
anomaly in the equatorial Atlantic, and the greater
ocean heating throughout the Atlantic, extratropical Indian and the Pacific south of about 30°N.
The WOCE years (1991–97) differ from the earlier
years (1982–90) in several ways. There are generally negative heat flux anomalies at all latitudes of
the Indian Ocean. The equatorial Pacific heat and
freshwater flux anomalies are smaller in amplitude. There is heating over most of the South
Atlantic. The amplitude of the heat flux anomalies
associated with the North Atlantic Oscillation
(NAO) (north of 45°N in the Atlantic) is smaller.
In Figure 5.1.16 (see Plate 5.1.16, p. 428) an ‘average’ year would be characterized by a vertical zero
contour in all six panels. No such year exists, with
1992 perhaps the closest. No short period (e.g. our
analysis period, or the WOCE period) appears
to be ‘typical’ of either another short period, or
of a longer period. Prime contributors to this
observation are the irregular variability of the El
Niño-Southern Oscillation (ENSO) and the NAO.
5.1.6 Summary
In this chapter we have set out Walin’s (1982) formal theory of water mass transformation. We have
diagnosed transformation driven by surface fluxes
across density classes in various regions of the
World Ocean, over three years of the WOCE period.
Our results cover almost all the World Ocean. They
are broadly similar to, but show distinct differences
from, other diagnoses using different climatologies.
These differences arise partly from uncertainties in
the flux fields, but also from their substantial
(Marsh, 2000) temporal variability. Our results are
derived from observations over a three-year period,
which may not be representative of the long term.
The sensitivity of these transformation rates to
the surface fluxes emphasizes just how important
accurate knowledge of the surface fluxes is.
Numerical ocean models that are driven by incorrect fluxes acting on incorrect surface temperature
and density distributions will not contain the correct water masses. Models forced by relaxation
and/or climatological fluxes can easily develop
regions of ‘mismatch’ between surface fluxes and
surface properties. The strong spurious fluxes acting on fluid that is too warm and saline over an
incorrectly separated Gulf Stream are a classic
example of this problem. Coupled models, e.g. by
Doney et al. (1998b), which take into account the
finite heat capacity of the atmosphere offer more
hope in this regard.
Care must be taken in relating the formation
rates, M, derived from differentiation of the transformation rate driven by surface fluxes, F, to ‘subduction’ and the injection of water masses into the
main thermocline. In the first instance, scaling
analysis and numerical model runs (Garrett and
Tandon, 1997; Marshall et al., 1999; Nurser et al.,
1999; Tandon and Zahariev, 2000) suggest that
diffusive processes within the mixed layer and seasonal thermocline, while relatively unimportant at
higher densities, do profoundly change the transformation rates at lighter densities :ϳ25.5 m
93
.
For these lighter waters the total transformation
G calculated over the mixed layer and seasonal
thermocline can be very different from F, even
having the opposite sign (Marshall et al., 1999;
Rhines, 1993; Tandon and Garrett, 1997). Second,
the domains over which we integrate to find
G may include regions of entrainment as well as
subduction. The entrainment and subduction will
tend to cancel out, leaving a net formation rate
considerably smaller than the total subduction of
fluid from that outcrop. Over an open domain,
however, M is now the difference between the
fluid subducted down into the thermocline, and
the input of mixed layer and seasonal thermocline
waters of that density across the open boundary of
the domain. This may give something different
from the subduction or ventilation rate.
Despite these caveats, these water mass diagnostics are particularly useful in understanding the
generation of denser mode and intermediate waters,
where diffusive and recirculative processes are
not such an issue. They are powerful tools for
understanding the diabatic circulation, offering the
5.1 Ocean Surface Water Mass Transformation
335
Large and Nurser