for the comparisons. A total of 42 cross-overs
were available and the salinity differences ranged
from 0 to 7.810
93 in magnitude with a mean
difference of 2.410
93
. (Note that salinity has no
dimensions. It is measured on the Practical Salinity
Scale 1978.) Aoyama et al. went further, trying to
account for systematic differences between standard seawater batches. This methodology follows
Mantyla (1980), and attempts to identify and correct for differences between the label conductivity
and actual conductivity ratio in batches of International Association for the Physical Sciences of
the Ocean (IAPSO) Standard SeaWater (SSW).
This issue is complicated by the fact that given
batches have been observed to change their salinities with time (Culkin and Ridout, 1998; Bacon
et al., 2000). Bacon et al. report that the effect of
ageing on SSW leads to salinity changes of up to
<0.002 after 24–36 months. To be within 0.001
of the label value, SSW should be used within one
year of production. Without specific knowledge of
the systematic time-dependent changes in the SSW
batches used, this effect could contribute a standard deviation of order 0.001 to deep properties.
Having applied adjustments for the authors’ best
estimates of SSW differences, the cross-over differences in the analysis by Aoyama et al. were
reduced to a mean of 1.810
93
. So while the
cross-over differences were reduced, the change
was small, suggesting that the dominant cause for
the differences was not variations in SSW. The
cause could be real variations in salinity due to
slow changes in time over the course of the Pacific
WOCE sampling, or simply a noise residue reflecting measurement error of the many different
groups doing the sampling. We note that according to the WHP plan, salinity measurements were
required to be accurate to 210
93
. Since the
above comparisons would amplify an accuracy in
the difference by a factor of 2
1/2 , one can say that
the WHP salinity measurements in the Pacific
Basin meet the requirements set out for the onetime survey.
In the Indian Ocean, Key (personal communication; also at http://geoweb.princeton.edu/staff/Key/
key.cross/crossover.html) analyses data from 18 crossovers after fitting to a smoothed potential density
for waters deeper than 2500 dbar. The mean salinity difference (uncorrected for possible variations
in SSW) is 0.810
93 . That this difference is
smaller than the Pacific may be due to the short
time in which the Indian Ocean was sampled or
the smaller number of measuring groups.
In the Atlantic Ocean Gouretski and Jancke
(1998) have used cross-overs going back in time to
pre-WOCE measurements (SAVE in 1987) and up
to A22 (1997). They have examined the salinity
variation on potential temperature surfaces for
depths exceeding 1000 m and potential temperatures below 3°C. Mean absolute cruise–cruise differences are 2.2910
93
. These are reduced to
1.8610
93 if the SSW batch adjustments proposed by Aoyama et al. are applied. Thus in all
WHP one-time survey measurements reported at
the times of the above analyses, salinity has been
measured to the expected high accuracy required.
Individual sections might be considered for
further correction if they have exceptionally high
or low systematic offsets relative to their crossing
lines. Bacon et al. (2000) point out that the time
variability of SSW is critical. A bias between
cruises of size greater than 0.003 is probably environmental change, or could be caused by measurement error during a cruise. Differences smaller
than 0.003 could be attributed to differences
between label and actual salinity of SSW. Some of
the cross-over differences identified by Gouretski
and Jancke, for example, could reflect temporal
changes in deep water properties. However, the
Atlantic cross-over differences reported above do
not show any systematic change with time.
Consistency of the WHP one-time oxygen and nutrient
measurements
The nutrient and oxygen data have not yet been
examined as systematically as salinity, but there
have been enough initial results to warrant mention. Accuracy requirements for these parameters
(see Table 3.1.2) were expressed in per cent of
ocean basin full-scale values. For the Atlantic
Ocean Gouretski and Jancke have expressed these
(we here use WHP supported units of mol kg
91 )
as 3, 2.5, 0.5 and 0.06 for oxygen, silicate, nitrate
and phosphate, respectively. Given that observational differences are amplified by a factor of 2
1/2 ,
it is found that Atlantic data satisfy the published
requirements. In the Indian Ocean, Key examines
silica cross-overs and finds a similar result. In the
Pacific Ocean, Gordon, Mordy, Wilson and Ross
(personal communication, 1999) find that 93% of
the silica cross-overs fall within the specifications,
while 87% of the nitrate and 70% of the
SECTION 3 NEW WAYS OF OBSERVING THE OCEAN
110
were available and the salinity differences ranged
from 0 to 7.810
93 in magnitude with a mean
difference of 2.410
93
. (Note that salinity has no
dimensions. It is measured on the Practical Salinity
Scale 1978.) Aoyama et al. went further, trying to
account for systematic differences between standard seawater batches. This methodology follows
Mantyla (1980), and attempts to identify and correct for differences between the label conductivity
and actual conductivity ratio in batches of International Association for the Physical Sciences of
the Ocean (IAPSO) Standard SeaWater (SSW).
This issue is complicated by the fact that given
batches have been observed to change their salinities with time (Culkin and Ridout, 1998; Bacon
et al., 2000). Bacon et al. report that the effect of
ageing on SSW leads to salinity changes of up to
<0.002 after 24–36 months. To be within 0.001
of the label value, SSW should be used within one
year of production. Without specific knowledge of
the systematic time-dependent changes in the SSW
batches used, this effect could contribute a standard deviation of order 0.001 to deep properties.
Having applied adjustments for the authors’ best
estimates of SSW differences, the cross-over differences in the analysis by Aoyama et al. were
reduced to a mean of 1.810
93
. So while the
cross-over differences were reduced, the change
was small, suggesting that the dominant cause for
the differences was not variations in SSW. The
cause could be real variations in salinity due to
slow changes in time over the course of the Pacific
WOCE sampling, or simply a noise residue reflecting measurement error of the many different
groups doing the sampling. We note that according to the WHP plan, salinity measurements were
required to be accurate to 210
93
. Since the
above comparisons would amplify an accuracy in
the difference by a factor of 2
1/2 , one can say that
the WHP salinity measurements in the Pacific
Basin meet the requirements set out for the onetime survey.
In the Indian Ocean, Key (personal communication; also at http://geoweb.princeton.edu/staff/Key/
key.cross/crossover.html) analyses data from 18 crossovers after fitting to a smoothed potential density
for waters deeper than 2500 dbar. The mean salinity difference (uncorrected for possible variations
in SSW) is 0.810
93 . That this difference is
smaller than the Pacific may be due to the short
time in which the Indian Ocean was sampled or
the smaller number of measuring groups.
In the Atlantic Ocean Gouretski and Jancke
(1998) have used cross-overs going back in time to
pre-WOCE measurements (SAVE in 1987) and up
to A22 (1997). They have examined the salinity
variation on potential temperature surfaces for
depths exceeding 1000 m and potential temperatures below 3°C. Mean absolute cruise–cruise differences are 2.2910
93
. These are reduced to
1.8610
93 if the SSW batch adjustments proposed by Aoyama et al. are applied. Thus in all
WHP one-time survey measurements reported at
the times of the above analyses, salinity has been
measured to the expected high accuracy required.
Individual sections might be considered for
further correction if they have exceptionally high
or low systematic offsets relative to their crossing
lines. Bacon et al. (2000) point out that the time
variability of SSW is critical. A bias between
cruises of size greater than 0.003 is probably environmental change, or could be caused by measurement error during a cruise. Differences smaller
than 0.003 could be attributed to differences
between label and actual salinity of SSW. Some of
the cross-over differences identified by Gouretski
and Jancke, for example, could reflect temporal
changes in deep water properties. However, the
Atlantic cross-over differences reported above do
not show any systematic change with time.
Consistency of the WHP one-time oxygen and nutrient
measurements
The nutrient and oxygen data have not yet been
examined as systematically as salinity, but there
have been enough initial results to warrant mention. Accuracy requirements for these parameters
(see Table 3.1.2) were expressed in per cent of
ocean basin full-scale values. For the Atlantic
Ocean Gouretski and Jancke have expressed these
(we here use WHP supported units of mol kg
91 )
as 3, 2.5, 0.5 and 0.06 for oxygen, silicate, nitrate
and phosphate, respectively. Given that observational differences are amplified by a factor of 2
1/2 ,
it is found that Atlantic data satisfy the published
requirements. In the Indian Ocean, Key examines
silica cross-overs and finds a similar result. In the
Pacific Ocean, Gordon, Mordy, Wilson and Ross
(personal communication, 1999) find that 93% of
the silica cross-overs fall within the specifications,
while 87% of the nitrate and 70% of the
SECTION 3 NEW WAYS OF OBSERVING THE OCEAN
110