CalCOFI replicate Analysis to Verify Precision Prior to Implementation of New Equipment and Field Techniques David M.Wolgast, Jennifer Sheldon, Ralf Goericke Scripps Institution of Oceanography, La Jolla CA Introduction The purpose of this work was to compare the oxygen data from standard Winkler manual titrations with those of an automated system manufactured by the Scripps Ocean Data Facility. Oxygen concentrations have been routinely measured on CalCOFI cruises since The new SIO/CalCOFI oxygen titrator still follows very closely the methods of the original Winkler titration as modified by Carpenter (1965). In place of the visible color of the iodine-starch complex as an indicator of the endpoint, it uses the strong absorption of ultra-violet (365 nm) light by the tri-iodide ion. The concept of using UV as a measure of titration progress is not new. The inclusion of temperature to calculate density in order to measure the mass of both reagent and sample is a significant innovation that will improve accuracy and allow us to report data using the oceanographic standard units micromoles/kg seawater. Easing operator fatigue and better accuracy are obviously desirable, but application to a 57-year data set warrants a comparison first with established methods. In the grand scheme of things, how do changes in equipment and techniques affect sampling, analytical and data processing procedures? Image borrowed from LTER data management presentation. In 1990 CalCOFI technicians performed an experiment to compare results from different operators on the same manual oxygen titration equipment. Results were as follows: IDI=absolute value of the difference between values in mL/L IDI=0.006 sigma=.0016 and CV=0.4% The experiments reported here were based on the same approach. Samples flasks were rinsed simultaneously and drawn in immediate succession. All analysts followed the same stringent water sampling and preservation techniques when preparing samples during this study. ResultsDiscussion Initial titration experiments performed in our shore based lab were very positive as demonstrated in Table 1. At sea results shown in Table 2 compare favorably with the previous work done by CalCOFI and also the institutional inter- comparison performed in preparation for WOCE. Our average test sample difference matched our sample accuracy at +/- 0.01mL/L, which varied from CC1990 and WOCE difference values that range between to 0.02 mL/L. This corresponds to a relative error value of 0.17% in surface and 1.7% in anoxic basin waters. The precision of the autotitrator can be seen in Table 4, where the sigma of our standards varies less than 0.1%. Blanks suffer from the presence of iodate in seawater and constitute the largest variable in the assay. Further studies are warranted to address this subject. Suffice to say, CalCOFI has always used seawater in standards and blanks because of the ubiquitous presence of naturally occurring iodate in samples. Fig. 1 show there is no systematic error correlated with concentration. Figures 2 and 3 demonstrate the same for concentration and depth of sample. The Santa Barbara Basin has the largest range of concentration values and represents the greatest range of depths regularly sampled. References Anderson, G., 1971 Marine Technicians Handbook Oxygen Analysis. SIO Ref. No Sea Grant Pub. No. 9 CalCOFI Data Report 9003 and 9004., SIO Fef. 91-4, 27 Feb Carpenter, J. H., 1965 Limnol. Oceanogr., 10, World Ocean Circulation Experiment Report 73/91, August Conclusions Although an offset of +0.01mL/L is seen in the data (Table 2 and Fig. 1) the magnitude is insignificant in the data set. Long term research data quality and continuity depend on careful experimental comparisons such as these before project design is effectively changed by implementation of new procedures and techniques. As changes are deemed necessary comparisons should be performed for other parameters measured to update metadata on precision of measurements in an effort to assure data quality and integrity. Acknowledgements and Support We would like to thank James Wilkinson for his continued excellent assistance and the analytical assistance of Karie Holterman, Fernando Ramirez and Nicholas Reale. Funding for this work came from Scripps CalCOFI and NOAA/JIMO contract. Data trend Figure 1 Figure 2 Figure 3