the net moisture content of the atmospheric
column (Xie and Arkin, 1997).
Estimates based on data-assimilating atmospheric
modelling systems are difficult to assess for several
reasons. First, the model output often doesn’t
obey total mass conservation (Trenberth and
Guillemot, 1998). Second, the lack of atmospheric
profile data over the oceans is a severe limitation
for these estimates: the assimilation of scant island
station data into these systems produces ‘bull’s
eyes’ in the flux fields indicative of biases between
the models and observations over the oceans
(Trenberth and Guillemot, 1998). As the vapour
exchange with the oceans dominates the air–sea
fluxes, these flux biases are problematic for global
budgets.
Surface flux estimates of evaporation rely on
empirical relations based on either radiometric
parameters measured from satellites or marine
meteorological measurements such as wind speed,
relative humidity and sea surface temperature
measured from ships and islands. Though constantly improving, these formulae are required to
apply under a large range of conditions and can
also suffer small biases which, when accumulated
over large areas, can dominate basin totals. Shipbased measurements can also suffer biases.
Precipitation is particularly challenging to estimate over the ocean, due to its sporadic nature in
both time and space. Here, satellite estimates may
be the only means of progress, but these also rely
on empirical algorithms that require ‘tuning’. Lack
of suitable open ocean time series for calibrating
the satellite estimates is a limitation, as island stations likely suffer biases due to local small-scale
orographic effects.
It is not the intention here to catalogue or assess
the presently available flux estimates exhaustively,
but rather to highlight the need for suitable independent estimates of large areal averages that can
be used to assess the new data sets being produced, and in particular identify the small but
crucial biases that impact dramatically on ocean
circulation models.
To demonstrate the difficulties, a set of recent
flux estimates has been assembled (Table 6.2.1).
Each member of the ensemble is assumed to be as
valid as the next. As a group, these estimates and
their variance can therefore be used as a gross
quantitive measure of the present state of knowledge about the mean annual exchange of water
between the ocean and the atmosphere. Based on
the products in Table 6.2.1, ensembles of flux estimates have been formed: an eight-member ensemble of precipitation (P) estimates, a six-member
ensemble of evaporation (E) estimates, and a 13member ensemble of E–P (net freshwater flux) estimates. After mapping to a common geographic
grid, the pointwise means and standard deviation
of the ensembles were formed.
The ensemble mean E–P (Fig. 6.2.1a) shows the
major source regions of atmospheric vapour: the
subtropical oceans under the atmospheric highpressure centres. The major atmospheric water
sinks are the tropical convergence zones and the
subpolar oceans. Interestingly, while the northern
hemisphere western boundary currents undergo
strong cooling (Bryden and Imawaki, Chapter 6.1),
they are subject to only weak freshwater forcing.
Over both the Kuroshio and Gulf Stream there
occurs almost total cancellation between enhanced
evaporation (Fig. 6.2.2a) and enhanced precipitation (Fig. 6.2.3a), both of which are likely associated with atmospheric storm tracks.
Variability within the ensemble of net flux estimates (Fig. 6.2.1b) reveals a global background
uncertainty of about 250 mm yr
91
, which when
integrated over the Pacific down to 30°S adds up
to 110
9 kg s
91 of freshwater transport, which is
as large as the natural flux itself (Wijffels et al.,
1992). The largest uncertainty occurs over the
tropics and the mid-latitude storm tracks, as well
as a region in the Southeast Pacific off Chile,
which is an important water mass formation
region (McCartney, 1977).
Variability among evaporation products is
much less (Fig. 6.2.2b), suggesting the bulk of the
uncertainty in the total water flux derives from
estimates of precipitation, which is confirmed by
Figure 6.2.3b. Not surprisingly, the difficulty of
measuring precipitation over the open ocean dominates errors in the net water flux to the ocean.
6.2.3 Impacts of uncertainties on model
development
The large uncertainties in estimates of freshwater
fluxes over the oceans has retarded efforts to
model the climate system. In the realm of atmospheric modelling, the lack of a reliable benchmark
against which to compare models is a difficulty,
with differences among models often smaller than
SECTION 6 LARGE-SCALE OCEAN TRANSPORTS
476
column (Xie and Arkin, 1997).
Estimates based on data-assimilating atmospheric
modelling systems are difficult to assess for several
reasons. First, the model output often doesn’t
obey total mass conservation (Trenberth and
Guillemot, 1998). Second, the lack of atmospheric
profile data over the oceans is a severe limitation
for these estimates: the assimilation of scant island
station data into these systems produces ‘bull’s
eyes’ in the flux fields indicative of biases between
the models and observations over the oceans
(Trenberth and Guillemot, 1998). As the vapour
exchange with the oceans dominates the air–sea
fluxes, these flux biases are problematic for global
budgets.
Surface flux estimates of evaporation rely on
empirical relations based on either radiometric
parameters measured from satellites or marine
meteorological measurements such as wind speed,
relative humidity and sea surface temperature
measured from ships and islands. Though constantly improving, these formulae are required to
apply under a large range of conditions and can
also suffer small biases which, when accumulated
over large areas, can dominate basin totals. Shipbased measurements can also suffer biases.
Precipitation is particularly challenging to estimate over the ocean, due to its sporadic nature in
both time and space. Here, satellite estimates may
be the only means of progress, but these also rely
on empirical algorithms that require ‘tuning’. Lack
of suitable open ocean time series for calibrating
the satellite estimates is a limitation, as island stations likely suffer biases due to local small-scale
orographic effects.
It is not the intention here to catalogue or assess
the presently available flux estimates exhaustively,
but rather to highlight the need for suitable independent estimates of large areal averages that can
be used to assess the new data sets being produced, and in particular identify the small but
crucial biases that impact dramatically on ocean
circulation models.
To demonstrate the difficulties, a set of recent
flux estimates has been assembled (Table 6.2.1).
Each member of the ensemble is assumed to be as
valid as the next. As a group, these estimates and
their variance can therefore be used as a gross
quantitive measure of the present state of knowledge about the mean annual exchange of water
between the ocean and the atmosphere. Based on
the products in Table 6.2.1, ensembles of flux estimates have been formed: an eight-member ensemble of precipitation (P) estimates, a six-member
ensemble of evaporation (E) estimates, and a 13member ensemble of E–P (net freshwater flux) estimates. After mapping to a common geographic
grid, the pointwise means and standard deviation
of the ensembles were formed.
The ensemble mean E–P (Fig. 6.2.1a) shows the
major source regions of atmospheric vapour: the
subtropical oceans under the atmospheric highpressure centres. The major atmospheric water
sinks are the tropical convergence zones and the
subpolar oceans. Interestingly, while the northern
hemisphere western boundary currents undergo
strong cooling (Bryden and Imawaki, Chapter 6.1),
they are subject to only weak freshwater forcing.
Over both the Kuroshio and Gulf Stream there
occurs almost total cancellation between enhanced
evaporation (Fig. 6.2.2a) and enhanced precipitation (Fig. 6.2.3a), both of which are likely associated with atmospheric storm tracks.
Variability within the ensemble of net flux estimates (Fig. 6.2.1b) reveals a global background
uncertainty of about 250 mm yr
91
, which when
integrated over the Pacific down to 30°S adds up
to 110
9 kg s
91 of freshwater transport, which is
as large as the natural flux itself (Wijffels et al.,
1992). The largest uncertainty occurs over the
tropics and the mid-latitude storm tracks, as well
as a region in the Southeast Pacific off Chile,
which is an important water mass formation
region (McCartney, 1977).
Variability among evaporation products is
much less (Fig. 6.2.2b), suggesting the bulk of the
uncertainty in the total water flux derives from
estimates of precipitation, which is confirmed by
Figure 6.2.3b. Not surprisingly, the difficulty of
measuring precipitation over the open ocean dominates errors in the net water flux to the ocean.
6.2.3 Impacts of uncertainties on model
development
The large uncertainties in estimates of freshwater
fluxes over the oceans has retarded efforts to
model the climate system. In the realm of atmospheric modelling, the lack of a reliable benchmark
against which to compare models is a difficulty,
with differences among models often smaller than
SECTION 6 LARGE-SCALE OCEAN TRANSPORTS
476