All of the quality controlled HOT bottle oxygen data collected at Station ALOHA (Figure 1) were examined so that trends and suspicious data points could be identified. Analytical precision of the oxygen titrations from the 15 years of the HOT program was explored. In addition, dissolved oxygen concentrations from three depths at Station ALOHA were evaluated.Analytical precision of the dissolved oxygen concentrations has been measured for the replicate samples collected on every HOT cruise. Replicate samples are collected from the same Niskin bottle, ideally in triplicate. BEACH protocols require that at least 20 % of the oxygen samples be collected in duplicate or triplicate and the resultant precision calculated. The precision is reported as a coefficient of variance (CV%), which is defined as the standard deviation of the replicate samples divided by the sample mean and then multiplied by 100 percent. The Carpenter modifications to the Winkler titration allow for a precision of 0.1% when applied with a diligent attention to detail, while the WOCE protocols (Culberson 1991) require a precision of 0.2% or better.

     The precision estimates of the dissolved oxygen titrations for HOT-1 through HOT-152 are plotted in Figure 2. The mean precision of the fifteen years worth of data was 0.15%. The vertical lines in the plot indicate the two main transitions in methodology: the switch from the Strickland and Parsons (1972) titration techniques to the Carpenter (1965) method between HOT-10 and –11, and the switch from a visual endpoint to a potentiometric endpoint between HOT-30 and –31.


Figure 2. Precision estimates (coefficient of variance) of the dissolved oxygen titrations for HOT-1 through HOT-152. The vertical lines indicate the switch from the Strickland and Parsons (1972) to the Carpenter (1965) method between HOT-10 and –11, and the switch from a visual endpoint to a potentiometric endpoint between HOT-30 and –31. The horizontal dashed line indicates the overall mean.

The mean precision of the oxygen titrations for HOT-1 through HOT-152 was 0.15%, which is close to the values obtained by Carpenter, and implies that, overall, good analytical techniques have been applied for the historical HOT dissolved oxygen determinations. Of the 152 cruises evaluated, 26 cruises had analytical precision values greater than 0.2% with eight of those exceeding 0.3%. The exceptionally high values for four consecutive cruises (HOT-144 through HOT-147) are indicative of the measurement problems during these cruises that are discussed in the following Section (Sect. 4.2.3 and 4.2.4).

The coefficient of variance for HOT-10 was exceptionally large (0.58%) and might partially be due to the small number of replicates measured (four). An error with any one of the four sets of measurements could significantly effect the precision calculation. There were four sets of replicates for HOT-33, which also exceeded 0.3% precision. Seven of the seventeen cruises with CV% greater than 0.2% had fewer than ten replicates.

The oxygen concentrations for the first ten HOT cruises were determined via the Strickland and Parsons version of the Winkler titration rather than the Carpenter method (see Sects. 2.2.1 and 3.1.1). The precision from three of these cruises exceeded 0.2%, but the precision of the other four cruises is comparable to results achieved with the Carpenter method. As mentioned in the Year One HOT Data Report (Chiswell et al., 1990), use of the Strickland and Parsons (1972) method adds a bias into the data related to the adiabatic change in sample temperature between the time the Niskin bottle was triggered and the time at which the samples are collected on deck, so the bottle dissolved oxygen values reported from the first ten HOT cruises may be slightly lower than the actual concentrations. The Year One HOT report states that the maximum error generated by ignoring the temperature change is around 0.06%, which is less than the average analytical precision of the titration technique.

Back to Table of Contents

Previous Next