Quality-assurance data for routine water quality analyses by the U. S. Geological Survey laboratory in Troy, New York; July 1993 through June 1995

A laboratory for analysis of low-ionic strength water has been developed at the U.S. Geological Survey (USGS) office in Troy, N.Y., to analyze samples collected by USGS projects in the Northeast. The laboratory's quality-assurance program is based on internal and interlaboratory quality-assurance samples and quality-control procedures developed to ensure proper sample collection, processing, and analysis. The quality-assurance/quality-control data are stored in the laboratory's SAS data-management system, which provides efficient review, compilation, and plotting of quality-assurance/quality-control data. This report presents and discusses samples analyzed from July 1993 through June 1995.

Quality-control results for 18 analytical procedures were evaluated for bias and precision. Control charts show that data from seven of the analytical procedures were biased throughout the analysis period for either high-concentration or low-concentration samples but were within control limits; these procedures were: acid-neutralizing capacity, dissolved inorganic carbon, dissolved organic carbon (soil expulsions), chloride, magnesium, nitrate (colorimetric method), and pH. Three of the analytical procedures were occasionally biased but were within control limits; they were: calcium (high for high-concentration samples for May 1995), dissolved organic carbon (high for highconcentration samples from January through September 1994), and fluoride (high in samples for April and June 1994). No quality-control sample has been developed for the organic monomeric aluminum procedure.

Results from the filter-blank and analytical-blank analyses indicate that all analytical procedures in which blanks were run were within control limits, although values for a few blanks were outside the control limits. Blanks were not analyzed for acid-neutralizing capacity, dissolved inorganic carbon, fluoride, nitrate (colorimetric method), or pH.

Sampling and analysis precision are evaluated herein in terms of the coefficient of variation obtained for triplicate samples in 14 of the 18 procedures. Data-quality objectives were met by more than 90 percent of the samples analyzed in all procedures except total monomeric aluminum (85 percent of samples met objectives), total aluminum (70 percent of samples met objectives), and dissolved organic carbon (85 percent of samples met objectives). Triplicate samples were not analyzed for ammonium, fluoride, dissolved inorganic carbon, or nitrate (colorimetric method).

Results of the USGS interlaboratory Standard Reference Sample Program indicated high data quality with a median result of 3.6 of a possible 4.0. Environment Canada's LRTAP interlaboratory study results indicated that more than 85 percent of the samples met data-quality objectives in 6 of the 12 analyses; exceptions were calcium, dissolved organic carbon, chloride, pH, potassium, and sodium. Data-quality objectives were not met for calcium samples in one LRTAP study, but 94 percent of samples analyzed were within control limits for the remaining studies. Data-quality objectives were not met by 35 percent of samples analyzed for dissolved organic carbon, but 94 percent of sample values were within 20 percent of the most probable value. Data-quality objectives were not met for 30 percent of samples analyzed for chloride, but 90 percent of sample values were within 20 percent of the most probable value. Measurements of samples with a pH above 6.0 were biased high in 54 percent of the samples, although 85 percent of the samples met data-quality objectives for pH measurements below 6.0. Data-quality objectives for potassium and sodium were not met in one study (only 33 percent of the samples analyzed met the objectives), although 85 percent of the sample values were within control limits for the other studies. Measured sodium values were above the upper control limit in all studies.

Results from blind reference-sample analyses indicated that data