C. Hydrographic Measurements 1. General Information
C.3. Water sample nutrient data (Joe Jennings)
Analysts, Equipment and Techniques Nutrient analyses were performed by:
P6E: Andrew A. Ross and Hernan Garcia from the College of Oceanic and Atmospheric Sciences at Oregon State University
P6C: Joe C. Jennings, Jr. from Oregon State University and Dennis Guffy of Texas A&M University
P6W: Consuelo Carbonell-Moore and Joe C. Jennings, Jr. from Oregon State University
The continuous flow analyzer used on all three legs of P6 was the Alpkem Rapid Flow Analyzer (RFA), model 300. A Keithley data acquisition system was used in parallel with analog stripchart recorders to acquire the absorbance data. The software used to process the nutrient data was developed by OSU. All of the reagent and standard materials were provided by OSU. The methods are described in Anonymous (1985) and in Gordon et. al. (in preparation, a & b).
Sampling Procedures:
Nutrient samples were drawn from all CTD/rosette casts at stations 003 through 072 leg3, 073 through 188 Leg4 and 189 through 257 Leg5. High density polyethylene (HDPE) bottles of approximately 30 ml volume were used as sample containers, and these same bottles were positioned directly in the autosampler tray. These sample tubes were routinely rinsed at least 3 times with one third to one half of their volume of sample before filling.
The nutrient samples were drawn following those for gases: Helium, tritium, dissolved oxygen and carbon dioxide. In some instances, the nutrient sampling procedure was not completed for almost 2 hours after the CTD arrived on deck.
At most stations, the RFA was started before sampling was completed to reduce the delay and minimize possible changes in nutrient concentration due to biological processes. All analyses were accomplished within a few hours of the end of the CTD/rosette casts.
Calibration and Standardization:
The volumetric flasks and pipettors used to prepare standards were gravimetrically calibrated prior to the cruise. The Eppindorf Maxipettor adjustable pipettors used to prepare mixed standards typically have a standard deviation of less than 0.002 ml on repeated deliveries of 10 ml volumes. High concentration mixed standards containing nitrate, phosphate, and silicic acid were prepared at intervals of 4 to 7 days and kept refrigerated in HDPE bottles. For almost every station, a fresh "working standard" was prepared by precise dilutions of 20 ml of the high concentration mixed standard to low nutrient seawater. This working standard has nutrient concentrations similar to those found in Deep and Bottom
waters. A separate nitrite standard solution was also added to these working standards. Corrections for the actual volumes of the flasks and pipettors were included in the preliminary data.
The WOCE Operations Manual calls for nutrient concentrations to be reported in units of micromoles per kilogram (~M kg-1). Because the salinity information required to compute density is not usually available at the time of initial computation of the nutrient concentrations, our concentrations are always originally computed as micromoles per liter. This unit conversion will be made using the corrected salinity data when it is available.
Equipment and analytical problems:
During the course of leg3, four series of standards with concentrations ranging from near zero to higher than any observed in the water column were run to check the linearity of the system response. On examining the results of these linearity checks, it became apparent that there was a significant nonlinearity present in the nitrate + nitrite channel. This nonlinearity arose from the inadvertent plumbing of the N + N channel according to the Alpkem manual and not according to the WOCE nutrient manual (Gordon et al., in preparation, a).
The data from the P6E linearity checks allowed us to apply a post cruise correction to the reported nitrate data of the form,
Ccor = K1 * Crep + K2 * Crep2,
where Ccor is the corrected concentration and Crep is the concentration reported during the cruise. K1 and K2 are constants determined by fitting the first derivatives of the concentration versus absorbance curves and from the absorbances of the standards run at each station. The derivation of this formula depends upon the observation that a quadratic equation adequately fits the concentration vs. absorbance data for all of the nonlinear cases during P6E and the first two weeks of P6C.
This correction has been applied only to the nitrate + nitrite data. It does not apply to any of the remaining analyses, including silicic acid, which have been shown to be linear to within ca. 0.1% of full-scale concentration with the methods used on P6E (cf. below and Gordon et al., in preparation, b).
At the start of P6C, the analysts ran additional standard curves to further document the extent of non-linearity, then made changes in the relative volumes of sample and buffer reagent used in the nitrate analysis to attempt to reduce this non-linearity. The first change to the nitrate pump tube configuration was made just prior to station 94. Standard curves were run to see how this change affected the linearity of the system response. The deviations from a linear response (residuals) were found to be smaller after the change in pump tubes, but still significant. A second change in pump tube sizes was made prior to station 112.
Standard curves run with this configuration had residuals which were within the
WOCE specifications for precision and accuracy, so this configuration was used for the remainder of the cruise. The pump tube configuration used on P6W was tested during and after the cruise and exhibited a linear response within WOCE specifications for precision and accuracy. No corrections to the reported nitrate data from P6W are necessary.
Phosphate phasing board failure: Starting at about station 150, the phosphate analytical channel began to experience increasing and irregular noise which resulted to decreased precision. This seemed to be electrical in origin, as routine replacement of the reagent chemicals and pump tubing did not resolve the problem. The major boards and components of the RFA used for the phosphate channel were systematically replaced wherever possible with limited and temporary success. During this period there were shipboard power failures which may have contributed to the RFA’s electrical problem. Following station 171, the phosphate noise problems became so severe as to render the data unusable and no phosphate values were reported for the final 18 stations of this leg.
Because the phosphate analysis uses a flow cell with an optical path length 2 to 3 times longer than the other analyses, it is particularly sensitive to any irregularities in the spacing of flow segmenting bubbles. Thus the air injection phasing board was the chief suspect. A replacement board was hand carried to Auckland, and it’s installation fixed the problem on the following leg.
Measurement of Precision and Bias:
Short Term Precision and Bias:
Throughout the cruise, replicate samples drawn in different sample tubes from the same Niskin bottle were analyzed to assess the precision of the RFA analyses. These replicate samples were analyzed both as adjacent samples (one after the other) and also at the beginning and end of sample runs to monitor deterioration in the samples or uncompensated instrumental drift. Except for phosphate, there was no significant difference between the precisions determined for adjacent samples and samples run at the beginning and ending of a sample run. The mean drift for phosphate was 0.014 micromol/l per hour (Leg3), 0.030 micromol/l per hour (Leg4) and -0.021 micromol/l per hour (Leg5).
That for nitrate on Leg4 was 0.12 micromol/l per hour. These drifts represent an estimate of part of the systematic error in that method.
For the other measurements there was no significant difference in the analysis of replicates as adjacent sample s and those run at approximately one hour intervals. The mean standard deviations found for the replicate analyses provide a measure of short term, intra-station precision, in micromol/l:
Leg 3
Phosphate: 0.005 Nitrate + Nitrite : 0.04 Silicic acid: 0.13 Nitrite: 0.01
Leg 4
Phosphate: 0.013 Nitrate + Nitrite : 0.05 Silicic acid: 0.20 Nitrite: 0.004
Leg 5
Phosphate: 0.016 Nitrate + Nitrite : 0.06 Silicic acid: 0.16 Nitrite: 0.022
Longer Term Precision:
On most of the sample runs during P6E, an "old" working standard from the previous station was run with the "new" working standard which had been freshly prepared. The "old" standards were kept refrigerated in plastic bottles. The average age of the "old" standards when reanalyzed was eight hours.
We calculated the difference in absorbance (peak height) between the last of the three new standards and the old standard which was run immediately after it.
This difference, with regard to sign (new - old), was tabulated and a statistical analysis was done. The results were converted to concentration units by multiplying the difference by the mean sensitivity factor for each nutrient (Table 10). It appears that the phosphate and nitrate standard concentrations increased slightly over the eight hour storage period on Leg5. This may be equipment related rather than a function of storage. The silicic acid and nitrite standards appear to have stored well. On Leg4 the phosphate standard concentrations appear to have increased slightly over the eight hour storage period. This may be equipment related rather than a function of storage; as the silicic acid, nitrate, and nitrite standards appear to have stored well.
Leg 3 differences between working standards at adjacent stations are shown in Table 12. Differences are expressed as "new" standard minus "old", and are given in concen-tration units (~M).
TABLE 12: Differences between working standards at adjacent stations for Leg3 Phosphate Nitrate Silicic Acid Nitrite Mean, (~M) wrt sign
-0.006 -0.01 -0.0014 -0.001 RMS Dev (~M)
0.016 0.065 0.18 0.007 n
59 66 66 66
Leg 4 Differences between working standards at adjacent stations are shown in Table 13. Differences are expressed as "new" standard minus "old", and are given in concen-tration units (~M).
TABLE 13: Differences between working standards at adjacent stations for Leg4 Phosphate Nitrate Silicic Acid Nitrite Mean, (~M) wrt sign
-0.015 -0.0005 -0.09 -0.002 RMS Dev (~M)
0.015 0.098 0.23 0.020 n
91 109 109 108
Leg 5 Comparisons between stored and fresh working standards are shown in Table 14. Differences are expressed as "new" standard minus "old", and are given in concen-tration units (~M).
TABLE 14: Comparisons between stored and fresh working standards for Leg5 Phosphate Nitrate Silicic Acid Nitrite Mean, (~M) wrt sign
-0.015 -0.0005 -0.09 -0.002 RMS Dev (~M)
0.015 0.098 0.23 0.020 n
91 109 109 108
Comparison with other data, long term precision and bias.
P6E/P6C:
P6E ended with station 072, and P6C commenced with station 073. We plotted the nutrients from stations 070 - 075 to check for consistency of the data from leg to leg. The phosphate values at the first station of P6C (Station 073) are somewhat higher than at the other stations, but there is no indication of a systematic shift in any of the nutrients and the agreement between the two legs is good.
P6E/P19C:
During the post-cruise QC work, additional preliminary data have become available from the WOCE P19C cruise. A comparison of several stations nearest the point where these two tracks cross indicates that the phosphate and nitrate data agree well (using final P6E nitrate data), but that there is an offset in the silicic acid data between the two cruises (see also Talley, 1993). Silicic acid concentrations reported for the P6E stations are lower than those for P19C by roughly 1 ~M in the deep water (theta < 2.0~C). This is approximately the magnitude of the correction term applied to the P19C data to account for nonlinearity in the silicic acid response. However, for the RFA procedure used on P6E, the maximum departure from a linear response in the concentration range of interest is <0.1 ~M. We conclude that nonlinearity of system response in the RFA cannot account for the discrepancy in the silicic acid data.
P6E/P6C:
P6E ended with station 072, and P6C commenced with station 073. We plotted the nutrients from stations 070 - 075 to check for consistency of the data from leg to leg. The phosphate values at the first station of P6C (Stn 073) are somewhat higher than at the other stations, but there is no indication of a systematic shift in any of the nutrients and the agreement between the two legs is good.
P6C/P6W:
P6C ended with station 188, and P6W commenced with station 189. We compared the nutrients from stations 180 through 198 to check for consistency of the data from leg to leg. There are no phosphate values for the P6C stations, but there is no indication of a systematic shift in the nutrient standards. The first three stations of P6W agree with the final stations of P6C to within less than 1% of the
deep nitrate and silicic acid concentrations. There is a greater total range of nitrate concentrations in the nitrate/theta relationship in the first 10 stations of P6W (1.2 micromol/l) than in the final stations of P6C (0.5 micromol/l), but the mean values agree to within 0.2 micromol/l. The agreement of the deep water silicic acid/theta relationship between the two legs is s 1.0 micromol/lwhere theta is less than 2xC.
Nutrient Quality Control Notes
During the cruise, a first pass quality control check on the nutrient data, primarily by comparing vertical profiles and nutrient/theta relationships. Following the cruise, all nutrient data were rechecked using log notes and the analog stripchart recordings made at sea and by examining parameter/parameter plots for outliers.
Some correctable errors were found and the appropriate corrections made. The nitrate data were corrected for nonlinearity as described above. At this time, the data quality flags were edited to conform to the definitions in the WOCE Operations Manual (WOCE Report No. 67/91). Data quality flags were assigned as follows:
Quality byte Definition 2 Acceptable measurement
3 Questionable measurement; no obvious problems found, but data somewhat out of trend.
4 Bad measurement; known analytical problems or data seriously out of trend.
5 Not reported.
9 Sample not drawn, usually due to Niskin bottle failure
At several stations, the bottle tripping order was deliberately (or accidentally) different from 1-36.