on global atmospheric CO 2 levels (including
checking our forecasts of sink behaviour)?
The answers to most of these questions, including
those dealing specifically with the behaviour of the
terrestrial biosphere, will rely to a great extent on a
combination of carbon cycle models that are coupled and checked against global data sets covering
the behaviour of the oceanic and atmospheric carbon reservoirs. The complexity and variability of
carbon storage and uptake on land means that the
long-standing approach of separately determining
storage and fluxes in the ocean and atmosphere
and evaluating regional and global behaviour of
the terrestrial biosphere by difference will likely be
required well into the future. In this regard, it is
notable that the JGOFS/WOCE Global CO 2 Survey
was the first global-scale, coordinated attempt to
constrain Excess CO 2 storage in the oceans, by
measurement.
6.3.2.4 Modelling of Excess CO 2 uptake
Global inventory
As mentioned above most estimates of Excess CO 2
storage in the oceans are based on models of ocean
uptake (e.g. Sarmiento et al., 1992; Siegenthaler and
Joos, 1992; Stocker et al., 1994; Caldeira and Duffy,
2000). These studies have converged on similar
global storage estimates (about 2 PgC yr
91
) although
independent lines of evidence sometimes implied
significantly different values (e.g. Tans et al., 1990
estimated 0.5 PgC yr
91
). While the models have
generally been calibrated or validated through comparison with transient tracer uptake (e.g. bomb
14
C
and more recently the CFCs), there must always be
room for doubt concerning systematic error in indirect estimates based on models. The tracers used for
calibration are clearly imperfect analogues for
Excess CO 2 penetration, particularly in their limited
ability to assess the ventilation of water masses of
intermediate ‘age’ (Wallace, 1995; Broecker, 2001).
The possibility that a degree of unconscious tuning
or selection of models in favour of the expected
ocean carbon uptake magnitude may have occurred
cannot be completely ruled out. It is therefore
encouraging that an independent approach to estimating oceanic Excess CO 2 uptake, using measured
time series of O 2 /N 2 in the atmosphere (Keeling
et al., 1993, 1996; Bender et al., 1996; Langenfelds
et al., 1999; Battle et al., 2000), supports the consensus model based estimate of about 2 PgC yr
91
.
Regional storage and uptake
On the issue of where Excess CO 2 is taken up and
stored in the oceans (as opposed to simply how
much) there is much less consensus among models. The powerful atmospheric O 2 /N 2 time series
approach can less readily address this question due
to the homogenizing tendency of atmospheric mixing and transport and the large seasonal exchanges
of oxygen between the atmosphere and ocean.
Very significant differences in modelled regional
storage and uptake between ocean models exist.
A perhaps extreme example is to compare carbon
storage calculated with the two-dimensional zonally averaged model of Stocker et al. (1994) with
distributions predicted by the three-dimensional
Ocean Biogeochemistry General Circulation Model
(GCM) of Sarmiento et al. (1995). Figure 6.3.1
illustrates the difference in Excess CO 2 storage as
a function of latitude in the North Atlantic Ocean
for these two models. The overall inventory for the
North Atlantic north of 24°N is nearly identical
for both models, but the latitudinal distributions
of the inventory differ very greatly. Figure 6.3.1
also shows, for comparison, the distribution of
Excess CO 2 inventories in the Atlantic Ocean calculated from (pre-WOCE) ocean CO 2 data using
a so-called ‘preformed-CO 2 ’ method of calculating Excess CO 2 concentrations and inventories
(Gruber et al., 1996; see Section 6.3.4.1). The direct
estimates are clearly much more consistent with the
distributions predicted by the three-dimensional
GCM than those predicted by the two-dimensional,
zonally averaged model. This comparison shows
clearly that models can indeed converge on consistent large-scale estimates for carbon storage while
differing drastically on the details of how and where
Excess CO 2 is stored. It also shows that observation-based estimates might prove very useful in
identifying deficiencies of models.
A similar point can be made by comparing the
model-based predictions of Excess CO 2 storage in
the Southern Ocean shown by Sabine et al. (1999)
and Caldeira and Duffy (2000): the details of modelling deep ventilation in this region lead to extremely
large differences in predicted storage between different models. This was noted earlier by England
(1995), who showed the high sensitivity of Southern
Ocean CFC storage to the parameterization of mixing in models. A major, ongoing programme of
ocean carbon model intercomparisons is currently
exploring these differences (Orr et al., 2001).
SECTION 6 LARGE-SCALE OCEAN TRANSPORTS
492
checking our forecasts of sink behaviour)?
The answers to most of these questions, including
those dealing specifically with the behaviour of the
terrestrial biosphere, will rely to a great extent on a
combination of carbon cycle models that are coupled and checked against global data sets covering
the behaviour of the oceanic and atmospheric carbon reservoirs. The complexity and variability of
carbon storage and uptake on land means that the
long-standing approach of separately determining
storage and fluxes in the ocean and atmosphere
and evaluating regional and global behaviour of
the terrestrial biosphere by difference will likely be
required well into the future. In this regard, it is
notable that the JGOFS/WOCE Global CO 2 Survey
was the first global-scale, coordinated attempt to
constrain Excess CO 2 storage in the oceans, by
measurement.
6.3.2.4 Modelling of Excess CO 2 uptake
Global inventory
As mentioned above most estimates of Excess CO 2
storage in the oceans are based on models of ocean
uptake (e.g. Sarmiento et al., 1992; Siegenthaler and
Joos, 1992; Stocker et al., 1994; Caldeira and Duffy,
2000). These studies have converged on similar
global storage estimates (about 2 PgC yr
91
) although
independent lines of evidence sometimes implied
significantly different values (e.g. Tans et al., 1990
estimated 0.5 PgC yr
91
). While the models have
generally been calibrated or validated through comparison with transient tracer uptake (e.g. bomb
14
C
and more recently the CFCs), there must always be
room for doubt concerning systematic error in indirect estimates based on models. The tracers used for
calibration are clearly imperfect analogues for
Excess CO 2 penetration, particularly in their limited
ability to assess the ventilation of water masses of
intermediate ‘age’ (Wallace, 1995; Broecker, 2001).
The possibility that a degree of unconscious tuning
or selection of models in favour of the expected
ocean carbon uptake magnitude may have occurred
cannot be completely ruled out. It is therefore
encouraging that an independent approach to estimating oceanic Excess CO 2 uptake, using measured
time series of O 2 /N 2 in the atmosphere (Keeling
et al., 1993, 1996; Bender et al., 1996; Langenfelds
et al., 1999; Battle et al., 2000), supports the consensus model based estimate of about 2 PgC yr
91
.
Regional storage and uptake
On the issue of where Excess CO 2 is taken up and
stored in the oceans (as opposed to simply how
much) there is much less consensus among models. The powerful atmospheric O 2 /N 2 time series
approach can less readily address this question due
to the homogenizing tendency of atmospheric mixing and transport and the large seasonal exchanges
of oxygen between the atmosphere and ocean.
Very significant differences in modelled regional
storage and uptake between ocean models exist.
A perhaps extreme example is to compare carbon
storage calculated with the two-dimensional zonally averaged model of Stocker et al. (1994) with
distributions predicted by the three-dimensional
Ocean Biogeochemistry General Circulation Model
(GCM) of Sarmiento et al. (1995). Figure 6.3.1
illustrates the difference in Excess CO 2 storage as
a function of latitude in the North Atlantic Ocean
for these two models. The overall inventory for the
North Atlantic north of 24°N is nearly identical
for both models, but the latitudinal distributions
of the inventory differ very greatly. Figure 6.3.1
also shows, for comparison, the distribution of
Excess CO 2 inventories in the Atlantic Ocean calculated from (pre-WOCE) ocean CO 2 data using
a so-called ‘preformed-CO 2 ’ method of calculating Excess CO 2 concentrations and inventories
(Gruber et al., 1996; see Section 6.3.4.1). The direct
estimates are clearly much more consistent with the
distributions predicted by the three-dimensional
GCM than those predicted by the two-dimensional,
zonally averaged model. This comparison shows
clearly that models can indeed converge on consistent large-scale estimates for carbon storage while
differing drastically on the details of how and where
Excess CO 2 is stored. It also shows that observation-based estimates might prove very useful in
identifying deficiencies of models.
A similar point can be made by comparing the
model-based predictions of Excess CO 2 storage in
the Southern Ocean shown by Sabine et al. (1999)
and Caldeira and Duffy (2000): the details of modelling deep ventilation in this region lead to extremely
large differences in predicted storage between different models. This was noted earlier by England
(1995), who showed the high sensitivity of Southern
Ocean CFC storage to the parameterization of mixing in models. A major, ongoing programme of
ocean carbon model intercomparisons is currently
exploring these differences (Orr et al., 2001).
SECTION 6 LARGE-SCALE OCEAN TRANSPORTS
492