(9°N) using a multiparameter mixing analysis
(based on an approach described by Tomczak and
Large, 1989). First of all, end-member concentrations for regional water types contributing to the
water sampled along the section were specified.
Next, local biogeochemical changes to TCO 2 concentrations were estimated stoichiometrically using
deviations of measured oxygen and alkalinity from
values predicted using the (conservative) mixing
model and the end-member concentrations. These
‘biogeochemically corrected’ TCO 2 concentrations
along the section were assumed to be created by
the mixing of the end-members’ TCO 2 concentrations, each of which was assumed to be composed
of a natural and an Excess CO 2 component. The
actual Excess CO 2 component in water sampled
along the section would be expected to be lower
than the level predicted by instantaneous mixing of
the Excess CO 2 components of the end-members.
(This is because the end-members’ Excess CO 2 concentrations are increasing with time.) The contribution of an end-member’s Excess CO 2 component to
the biogeochemically corrected TCO 2 concentration along the section was therefore assigned a
functional dependence that allowed it to decrease
with depth, allowing the effects of depth-dependent
ventilation times (water-mass ages) on the Excess
CO 2 levels to be accounted for. Finally, the endmembers’ Excess CO 2 concentrations as well as the
coefficients of the depth-dependent functions were
solved iteratively with the mixing model in order to
minimize the residuals relative to the biogeochemically corrected measured values.
This approach represents perhaps the most radical departure from the long-standing methods
introduced by Brewer (1978) and Chen and
Millero (1979). It relaxes several assumptions
inherent in earlier techniques: notably no explicit
assumptions are required concerning the preformed
(surface) preindustrial concentrations, the required
stoichiometric corrections are small, and mixing of
multiple different water sources can be treated
explicitly. On the other hand, some of the procedures, assumptions and statistical weighting
involved in the approach as currently described are
difficult to assess. The Excess CO 2 distribution calculated for I1 section data using this technique
showed significant differences with the estimates
derived using the ⌬C* approach on the same data
set (Coatanoan et al., 2001). Differences were
particularly notable in the upper thermocline
(200–600 m) where the ⌬C* approach gave Excess
CO 2 levels about 12 mol kg
91 higher than the
mixing model calculations. The mixing model
approach gave slightly higher values in deep waters
(9800 m). Overall the ⌬C* approach gave water
column inventories that averaged 20% higher than
those estimated using the mixing model approach.
There is as yet no basin-scale estimate for the
entire Indian Ocean based on the mixing model
approach. Using the ⌬C* approach, Sabine et al.
(1999) estimated a total inventory for the Indian
Ocean (between 20°E and 120°E) of 20.3<3 PgC
for the year 1995.
Pacific Ocean
In contrast to the comprehensive syntheses of preWOCE Atlantic data and WOCE Indian Ocean
data discussed above, the synthesis of Pacific Data
is not as far advanced. In part this is because the
measurements were collected over several years
(1990–96) by several different groups, as opposed
to the concentrated approach employed in the
Indian Ocean during the US WOCE effort.
Brewer et al. (1997) used data collected along
P17N in the Northeastern Pacific (135°W to
160°W) to illustrate a historical and conceptual
review of preformed-CO 2 calculation techniques.
In their review the emphasis was on the calculation
of the preformed pCO 2 as a function of depth. In
particular the high sensitivity of Excess CO 2 calculations to the assumed stoichiometric value of the
C : O 2 respiratory coefficient was emphasized.
To date the only other analysis of Excess CO 2
calculations resulting from WOCE cruises was
reported by Feely et al. (1999) based on data collected along 150°W. (A time series comparison by
Ono et al. (1998) is discussed in Section 6.3.4.2.)
Feely et al. (1999) used the ⌬C
* method to calculate Excess CO 2 concentrations from 60°S to
about 55°N. They found the largest area-specific
inventories in the subtropics and the sub-Antarctic
zone north of 55°S. In the South Pacific the deepest penetration and largest inventories were at
48°S. Inventories at the equator were almost a factor of two smaller. Inventories north of 48°N were
also small. A comparison was made with two
ocean carbon cycle models. As with the Indian
Ocean and the Atlantic comparison, the measurement-based inventories were similar to the modelled inventories along most of the section, with
the exception of both high-latitude regions (955°S
SECTION 6 LARGE-SCALE OCEAN TRANSPORTS
510
(based on an approach described by Tomczak and
Large, 1989). First of all, end-member concentrations for regional water types contributing to the
water sampled along the section were specified.
Next, local biogeochemical changes to TCO 2 concentrations were estimated stoichiometrically using
deviations of measured oxygen and alkalinity from
values predicted using the (conservative) mixing
model and the end-member concentrations. These
‘biogeochemically corrected’ TCO 2 concentrations
along the section were assumed to be created by
the mixing of the end-members’ TCO 2 concentrations, each of which was assumed to be composed
of a natural and an Excess CO 2 component. The
actual Excess CO 2 component in water sampled
along the section would be expected to be lower
than the level predicted by instantaneous mixing of
the Excess CO 2 components of the end-members.
(This is because the end-members’ Excess CO 2 concentrations are increasing with time.) The contribution of an end-member’s Excess CO 2 component to
the biogeochemically corrected TCO 2 concentration along the section was therefore assigned a
functional dependence that allowed it to decrease
with depth, allowing the effects of depth-dependent
ventilation times (water-mass ages) on the Excess
CO 2 levels to be accounted for. Finally, the endmembers’ Excess CO 2 concentrations as well as the
coefficients of the depth-dependent functions were
solved iteratively with the mixing model in order to
minimize the residuals relative to the biogeochemically corrected measured values.
This approach represents perhaps the most radical departure from the long-standing methods
introduced by Brewer (1978) and Chen and
Millero (1979). It relaxes several assumptions
inherent in earlier techniques: notably no explicit
assumptions are required concerning the preformed
(surface) preindustrial concentrations, the required
stoichiometric corrections are small, and mixing of
multiple different water sources can be treated
explicitly. On the other hand, some of the procedures, assumptions and statistical weighting
involved in the approach as currently described are
difficult to assess. The Excess CO 2 distribution calculated for I1 section data using this technique
showed significant differences with the estimates
derived using the ⌬C* approach on the same data
set (Coatanoan et al., 2001). Differences were
particularly notable in the upper thermocline
(200–600 m) where the ⌬C* approach gave Excess
CO 2 levels about 12 mol kg
91 higher than the
mixing model calculations. The mixing model
approach gave slightly higher values in deep waters
(9800 m). Overall the ⌬C* approach gave water
column inventories that averaged 20% higher than
those estimated using the mixing model approach.
There is as yet no basin-scale estimate for the
entire Indian Ocean based on the mixing model
approach. Using the ⌬C* approach, Sabine et al.
(1999) estimated a total inventory for the Indian
Ocean (between 20°E and 120°E) of 20.3<3 PgC
for the year 1995.
Pacific Ocean
In contrast to the comprehensive syntheses of preWOCE Atlantic data and WOCE Indian Ocean
data discussed above, the synthesis of Pacific Data
is not as far advanced. In part this is because the
measurements were collected over several years
(1990–96) by several different groups, as opposed
to the concentrated approach employed in the
Indian Ocean during the US WOCE effort.
Brewer et al. (1997) used data collected along
P17N in the Northeastern Pacific (135°W to
160°W) to illustrate a historical and conceptual
review of preformed-CO 2 calculation techniques.
In their review the emphasis was on the calculation
of the preformed pCO 2 as a function of depth. In
particular the high sensitivity of Excess CO 2 calculations to the assumed stoichiometric value of the
C : O 2 respiratory coefficient was emphasized.
To date the only other analysis of Excess CO 2
calculations resulting from WOCE cruises was
reported by Feely et al. (1999) based on data collected along 150°W. (A time series comparison by
Ono et al. (1998) is discussed in Section 6.3.4.2.)
Feely et al. (1999) used the ⌬C
* method to calculate Excess CO 2 concentrations from 60°S to
about 55°N. They found the largest area-specific
inventories in the subtropics and the sub-Antarctic
zone north of 55°S. In the South Pacific the deepest penetration and largest inventories were at
48°S. Inventories at the equator were almost a factor of two smaller. Inventories north of 48°N were
also small. A comparison was made with two
ocean carbon cycle models. As with the Indian
Ocean and the Atlantic comparison, the measurement-based inventories were similar to the modelled inventories along most of the section, with
the exception of both high-latitude regions (955°S
SECTION 6 LARGE-SCALE OCEAN TRANSPORTS
510