the CO 2 Survey. The results indicated a systematically higher difference between the shore-based and
shipboard replicate analyses than the difference
between the Certified Value for a CRM and its
shipboard determination by coulometry. The difference averages of the order 2.2 mol kg
91
, with the
shore-based (manometric) analyses giving the
higher results. The reason for this difference, and its
inter-cruise variability, is hard to explain. It might,
for example, result partly from slow abiotic hydrolysis during sample transport and storage of a dissolved or particulate organic carbon fraction not
present in the CRMs.
It was therefore not possible to ‘close the circle’
perfectly with respect to data accuracy. A small,
unexplained, but reasonably systematic deviation
remains between the shipboard and shore-based
measurements. The source of this deviation needs
further investigation. At present it appears reasonable, however, to assume that the factors leading
to such a deviation may well have applied to
pre-WOCE shore-based versus shipboard analyses
as well.
Sabine et al. (1999) tested this assumption by
comparing the shipboard data collected during the
Indian Ocean WOCE cruises with deep-water data
collected in 1977–78 during GEOSECS. A deepwater (92000 m) offset of ;22.5<3 mol kg
91
of GEOSECS-WOCE (GEOSECS data higher) was
observed after correcting for hydrography variability using multivariate analysis. Keeling manometric analyses made on stored GEOSECS samples
from the Indian Ocean were 16.5<5 mol kg
91
lower than the GEOSECS shipboard values
(Weiss et al., 1983). For the WOCE data from the
Indian Ocean, the Keeling analyses averaged
4 mol kg
91 higher than the shipboard analyses
(P. Guenther, personal communication, Fig. 6.3.4).
The overall expected difference between the two
shipboard data sets is therefore ;20.5 mol kg
91
,
which is almost exactly what was observed.
This single example points out that it might
indeed be possible to use the WOCE CRM-calibrated shipboard data set as a reference, with the
shore-based manometric analyses acting as a transfer standard between the WOCE shipboard data
and earlier shipboard data sets. This will be of
particular value for regions where it cannot be
readily assumed that deep-water values are temporally constant (e.g. for the Atlantic Ocean).
Whether the apparent consistency demonstrated in
the Indian Ocean will hold for other ocean basins
remains to be seen.
At-sea data quality evaluation
In general the shipboard data collected during the
CO 2 Survey was of very high quality, especially
for the primary parameter, TCO 2 . The most complete assessment of data quality to date has been
for the 14-month duration, nine-leg US WOCE
Indian Ocean expedition, with assessments covering TCO 2 (Johnson et al., 1998b), total alkalinity
(Millero et al., 1998) and surface pCO 2 (Sabine
and Key, 1997) having been published. Figures
6.3.5a and b show summary plots of the analyses
of Certified Reference Materials during this expedition. In the case of TCO 2 , the shipboard data
were independently calibrated and only checked
against the Certified (manometric) value. The
mean difference between the Certified and coulometric values averaged only 90.96 mol kg
91
(n:983) with the standard deviation of the difference being :1 mol kg
91 . For total alkalinity the
approach in the Indian Ocean was slightly different, in that individual data sets were adjusted, on a
cruise-by-cruise basis in order that the mean result
of the CRM analyses matched that of the Certified
value. In this case, the overall mean difference
between measured and Certified values must average zero for each cruise. Nevertheless, the relatively low standard deviation and lack of clear
trends in the data within each cruise leg show once
again that data quality was tightly controlled. The
standard deviation of the CRM analyses for alkalinity did vary significantly between the cruise legs
because of issues related to equipment maintenance and operation. This difference in performance between the primary parameter (TCO 2 )
and the second parameter (in this case total alkalinity) reflects the significant effort that went into
perfecting TCO 2 measurement prior to the Survey
(Section 6.3.3.4). For the Indian Ocean comparisons it should be noted that identical equipment
and procedures were used throughout, albeit by
various measurement groups.
The overall consistency of data sets was also
assessed empirically through the comparison of
deep profile data at cross-over locations where two
separate WOCE cruise legs occupied stations in
close (9100 km) geographical proximity to each
other (Johnson et al., 1998b; Millero et al., 1998).
For TCO 2 the absolute leg-to-leg differences
6.3 Storage and Transport of Excess CO 2 in the Oceans
501
Wallace
shipboard replicate analyses than the difference
between the Certified Value for a CRM and its
shipboard determination by coulometry. The difference averages of the order 2.2 mol kg
91
, with the
shore-based (manometric) analyses giving the
higher results. The reason for this difference, and its
inter-cruise variability, is hard to explain. It might,
for example, result partly from slow abiotic hydrolysis during sample transport and storage of a dissolved or particulate organic carbon fraction not
present in the CRMs.
It was therefore not possible to ‘close the circle’
perfectly with respect to data accuracy. A small,
unexplained, but reasonably systematic deviation
remains between the shipboard and shore-based
measurements. The source of this deviation needs
further investigation. At present it appears reasonable, however, to assume that the factors leading
to such a deviation may well have applied to
pre-WOCE shore-based versus shipboard analyses
as well.
Sabine et al. (1999) tested this assumption by
comparing the shipboard data collected during the
Indian Ocean WOCE cruises with deep-water data
collected in 1977–78 during GEOSECS. A deepwater (92000 m) offset of ;22.5<3 mol kg
91
of GEOSECS-WOCE (GEOSECS data higher) was
observed after correcting for hydrography variability using multivariate analysis. Keeling manometric analyses made on stored GEOSECS samples
from the Indian Ocean were 16.5<5 mol kg
91
lower than the GEOSECS shipboard values
(Weiss et al., 1983). For the WOCE data from the
Indian Ocean, the Keeling analyses averaged
4 mol kg
91 higher than the shipboard analyses
(P. Guenther, personal communication, Fig. 6.3.4).
The overall expected difference between the two
shipboard data sets is therefore ;20.5 mol kg
91
,
which is almost exactly what was observed.
This single example points out that it might
indeed be possible to use the WOCE CRM-calibrated shipboard data set as a reference, with the
shore-based manometric analyses acting as a transfer standard between the WOCE shipboard data
and earlier shipboard data sets. This will be of
particular value for regions where it cannot be
readily assumed that deep-water values are temporally constant (e.g. for the Atlantic Ocean).
Whether the apparent consistency demonstrated in
the Indian Ocean will hold for other ocean basins
remains to be seen.
At-sea data quality evaluation
In general the shipboard data collected during the
CO 2 Survey was of very high quality, especially
for the primary parameter, TCO 2 . The most complete assessment of data quality to date has been
for the 14-month duration, nine-leg US WOCE
Indian Ocean expedition, with assessments covering TCO 2 (Johnson et al., 1998b), total alkalinity
(Millero et al., 1998) and surface pCO 2 (Sabine
and Key, 1997) having been published. Figures
6.3.5a and b show summary plots of the analyses
of Certified Reference Materials during this expedition. In the case of TCO 2 , the shipboard data
were independently calibrated and only checked
against the Certified (manometric) value. The
mean difference between the Certified and coulometric values averaged only 90.96 mol kg
91
(n:983) with the standard deviation of the difference being :1 mol kg
91 . For total alkalinity the
approach in the Indian Ocean was slightly different, in that individual data sets were adjusted, on a
cruise-by-cruise basis in order that the mean result
of the CRM analyses matched that of the Certified
value. In this case, the overall mean difference
between measured and Certified values must average zero for each cruise. Nevertheless, the relatively low standard deviation and lack of clear
trends in the data within each cruise leg show once
again that data quality was tightly controlled. The
standard deviation of the CRM analyses for alkalinity did vary significantly between the cruise legs
because of issues related to equipment maintenance and operation. This difference in performance between the primary parameter (TCO 2 )
and the second parameter (in this case total alkalinity) reflects the significant effort that went into
perfecting TCO 2 measurement prior to the Survey
(Section 6.3.3.4). For the Indian Ocean comparisons it should be noted that identical equipment
and procedures were used throughout, albeit by
various measurement groups.
The overall consistency of data sets was also
assessed empirically through the comparison of
deep profile data at cross-over locations where two
separate WOCE cruise legs occupied stations in
close (9100 km) geographical proximity to each
other (Johnson et al., 1998b; Millero et al., 1998).
For TCO 2 the absolute leg-to-leg differences
6.3 Storage and Transport of Excess CO 2 in the Oceans
501
Wallace