PROGRAMS AND PLANS: "Quality Assurance of WRD Water Quality Data: Past, Present, and Future" July 30, 1980 Quality of Water Branch Technical Memorandum No. 80.21 Subject: PROGRAMS AND PLANS: "Quality Assurance of WRD Water- Quality Data: Past, Present, and Future" by Linda C. Friedman. The enclosed paper, discussing primarily the evolution of laboratory quality assurance measures, was prepared originally by Linda Friedman for presentation at a Regional seminar. Because of the paramount importance of quality assurance in all of our activities, the paper is provided for your information. Please circulate the paper as widely as possible in all district and project offices. R. J. Pickering, Chief Quality of Water Branch Enclosure Key Words: water quality, quality assurance, laboratories Superseded memorandum: none WRD Distribution: A, B, FO-L, PO QUALITY ASSURANCE OF WRD WATER-QUALITY DATA: PAST, PRESENT, AND FUTURE Linda C. Friedman Although no one would question the fact that any interpretation of data depends on the validity of that data, the need for quality- assurance procedures frequently does not appear to be equally clear. However, two facts clearly indicate this need: Data usually are collected by a measurement process and errors are inherent in any measurement process. Initially, water-quality data of the U.S. Geological Survey were collected in the field and supplemented by work done in state or university laboratories; quality control depended primarily on correct preparation of standards and accurate use of techniques such as gravimetric analysis. In the period 1905-1908, comprehensive water-quality studies were begun, as W. H. Durum points out in his "Historical Profile of Quality of Water Laboratories and Activities, 1879-1973," and these studies led to the development of quality control review techniques such as looking at the cation-anion balance (Durum, 1978). In 1918, the first WRD chemical laboratory was established in Washington, D.C. Quality control at that time depended on correctly prepared standards, careful analyses, and the quality- control relationships which had been developed in the previous decade or so. Refinement of these relationships is obvious in papers such as C. S. Howard's "Determination of Total Dissolved Solids In Water Analysis," which appeared in 1933 (Howard,1933). In 1937, a second chemical laboratory was opened in the Texas District at Austin and, in the 40's, 50's, and 60's, additional district laboratories were added so that by the end of the 1960's, there were 22 WRD laboratories. Throughout most of this time, there was little change in quality-control procedures. One major addition was made in 1962, when the Standard Reference Water Sample (SRWS) program was begun as part of Marvin Skougstad's Methods Development Project. Initially, two samples of deionized water containing known concentrations of calcium, magnesium, sodium, potassium, sulfate, and chloride were prepared and sent to district laboratories. In 1965, the SRWS program was modified by replacing the deionized water solutions with natural waters (Skougstad and Fishman, 1975). Two samples were, and still are, distributed twice yearly to all participating laboratories and the quality of work of each laboratory is judged by comparing individual laboratory values with the means and standard deviations calculated from values submitted by all participating laboratories. l/ As used in this paper, the term quality assurance describes the programs and sets of procedures, including (but not limited to) quality control procedures, which are necessary to assure data reliability. The term, quality control, on the other hand, is used to describe the routine procedures used to regulate measurements and produce data of satisfactory quality. Beginning in 1972, performance of most analyses by district laboratories was phased out and these laboratories were replaced by the two central water quality laboratories in Atlanta, Georgia, and Denver, Colorado. Field analyses, however, remained very much a part of water-quality data-collection activities, and the number of state, university or other cooperator, and contractor laboratories collecting such data increased and now probably numbers in the hundreds. (An exact figure should be obtained within the next year.) The problem of assuring the quality of the data produced by all these sources is massive and is compounded by the ever-increasing amount and variety of analytical work. For example, in 1919, an estimated 480 adjusted complete analyses were made with one laboratory in operation; and by 1970, approximately 50,000 adjusted complete analyses were made by 22 laboratories (Durum, 1978). Although the term "adjusted complete" has largely lost its meaning, it is useful for comparison. Applying an estimated adjusted complete value to FY 1979 results shows that approximately 132,000 adjusted complete analyses were made by the two central water-quality laboratories. This estimate of the increase in amount of analytical work gives no indication of the increase in complexity of work. Major problems are related to assuring the quality of data produced by the huge variety of analytical methods and equipment being used today. Fifteen years ago, water samples were analyzed by the U.S. Geological Survey primarily for acidity, alkalinity, calcium, chloride, color, dissolved solids, fluoride, iron, magnesium, manganese, nitrate, pH, potassium, silica, sodium, specific conductance, sulfate, and turbidity. Analytical results were generally "when analyzed;" in other words, district personnel collected a single bottle of water and the lab analyzed all constituents from that single bottle. This procedure made things simple for both laboratory and field personnel, but gave (and gives) numerous problems in interpretation of past results; for instance, iron values were reported even if most of the iron had precipitated prior to analysis. By contrast, in FY 1980, the capabilities of the central water- quality laboratories have been extended to include close to 800 different water quality characteristics (Figure 1). Samples analyzed today include filtered (dissolved) and unfiltered (total and total recoverable) water, suspended sediments, bottom materials, and cellular material. Preservation and pretreatment techniques are numerous, resulting in a plethora of bottles but, hopefully, also yielding more meaningful results. Unfortunately, the district laboratories added few new quality- assurance procedures as the number of phases being analyzed and the types of determinations being requested began to increase in the late 1960's and early 1970's. One of the few review checks added was to compare analyses to insure that "total" values were greater than "dissolved" values, if both had been determined on the same sample. This check did not take precision into account and often either the dissolved or total analyses would be repeated until the data "looked better'' (while in reality the values were statistically and analytically the same). The semiannual SRWS program gradually added the determination of metals and today over 40 major, nutrient, and trace constituents are included in this program, providing a measure of analytical precision. During the 1960's and 1970's, analytical methodology began changing rapidly. However, adequate records of when analytical changes were initiated were not kept. Because of this lack of record, there are doubts about the meaning of some of the chemical data collected during that period. For example, the initial procedure for determining mercury detected only inorganic mercury. An oxidant later was added which meant that some organic mercury was being detected as well as the inorganic mercury. Still later, a second oxidant was added and yielded more organic mercury. The concentrations of the two oxidants then were changed several times, each change making a greater proportion of the organic mercury susceptible to detection. Since records were not kept on which data had been obtained by which analytical procedure, an apparent increase in mercury values during the early 1970's could be due to a real increase or might be due only to methodology changes; interpretation is nearly impossible. Although a record of analytical methodology is still not part of the WATSTORE system, a record is being kept in the Central Laboratories System through the use of lab codes. Different lab codes, recorded in the "Water Quality Laboratory Services Catalog," are used for different methods so that if, in the future, it becomes important to know how a sample was analyzed and/or what the expected precision of the method used was, the data will be available on the Central Laboratories System tapes. Eventually, it is expected that such information will be stored in WATSTORE. All of the quality-control procedures developed in the past are used today in the Central Laboratories System. Heavy reliance still is placed on correct preparation of standards and careful use of analytical techniques. The analytical review checks, based largely on the relationships developed in the early l900's, are still used; however, these checks now are done by means of a computer, and result in "error messages" for the laboratory's quality-control staff. Participation in the SRWS program also continues, although the primary aim of that program now is to document multi-laboratory analytical precision. In addition, inorganic reference materials are used daily by the laboratory analysts to check their work; reference materials also are submitted as "unknowns" daily by the laboratory chief and weekly by District personnel (unknown to anyone within the Central Laboratories System). Log-in sheets accompanying all "unknown" reference materials are coded for computer recognition, and the central laboratories' quality- control sections are notified of discrepancies between analyzed and expected values within a day of when the samples were analyzed. This l-day delay in quality-control results will be eliminated when computer-analytical instrument links are completed and such information becomes immediately available. Until recently, efforts made within the USGS to document interlaboratory precision of methods had been rather limited. Within the past 2 years, data, primarily from the SRWS program, have been analyzed and precision statements for most inorganic methods have been developed and recorded in book 5, chapter Al, of the Techniques for Water-Resources Investigation (TWRI) series. Such statements of analytical precision are necessary for interpretation of data obtained by the methods. For the convenience of District personnel, the Water-Quality Laboratory Services Catalog gives an indication of method precision so that a method (lab code) with the required precision can be selected when an analysis is requested. Presently, efforts are underway to revise the computer programs so that "good" quality-assurance data (less than 1.5 standard deviations from the expected value) from unknown reference materials as well as "erroneous" data (greater than 1.5 standard deviations from the expected value) are stored in the computer files. This change will make it possible to program the computer to prepare quality-control charts that will facilitate the detection of an increase in bias or growing lack of precision before the problem becomes serious. Procedures for the quality assurance of organic analyses lag far behind those designed for inorganic analyses, largely because appropriate standard samples are not available. Although concentrated reference materials are available in ampoules for many organic compounds from the U.S. Environmental Protection Agency, dilute samples would be more useful as they could be introduced into laboratories as unknowns. Poor stability and adsorption onto sample container walls are only two of the problems expected to be encountered in preparing reference materials of organic compounds, especially since many organic compounds are formulated to be unstable so that they will not be long-term threats to the environment. In February 1980, a reference sample for selected herbicides and insecticides was prepared in tap water. This sample is being analyzed weekly to obtain information on sample degradation and has been distributed to approximately 30 laboratories through the SRWS program to obtain interlaboratory comparison data. Although not all data have been collected, preliminary information indicates that the sample is uniform and the pesticides relatively stable. In the meantime, the two central water-quality laboratories must rely largely on accurate preparation of organic standards and proper calibration of instruments for quality control of organic analyses. Analytical confirmation with a mass spectrometer often is used on positive results and both laboratories participate in EPA's performance evaluation program. Quality-control procedures also include collecting recovery data from analyses of "known" concentrations and analyzing some samples in duplicate. For example, routine tests are made of recoveries of standards after extraction and concentration and recoveries after interference clean-up procedures. Data on method bias now are being collected through these studies. Similarly, data on analyses of duplicate samples are being collected for pesticides in bottom materials. Because concentration differences between "duplicates" vary with compound concentrations, standard quality-control procedures which were developed for looking at duplicates in "production" processes are not appropriate. Alternatives to these procedures are being developed and, as soon as acceptable procedures are available, should be useful in monitoring the quality of data not only from the central water-quality laboratories but also from cooperator and contractor laboratories. As is true for chemical analyses, proper training of personnel and careful use of analytical techniques play a large part in the quality control of both sediment and biological analyses. In addition, eight particle-size reference materials were prepared in 1979 (Delaney and Schroder, 1979), and can be used to examine the comparability of data produced by WRD sediment laboratories and submitted periodically to the laboratories as quality assurance samples. The biology section of the Atlanta Central Laboratory maintains a reference specimen collection, monitors analyses "in- house", and participates in a program involving the interchange of samples between the section and the contractor performing phytoplankton analyses. Data from cooperator and contractor laboratories are used interchangeably with data from the WRD central water-quality laboratories. Thus, equivalent quality-assurance procedures must be applied to all laboratories providing water-quality data to the U.S. Geological Survey. Currently, a program to monitor the quality of data from non-USGS laboratories is under development; it is hoped that by the beginning of FY 1981 most of these laboratories will have become part of a USGS quality-assurance program. Meanwhile, in FY 1979, an extensive program to monitor the quality assurance of field pH and specific conductance measurements was developed. Over 10,000 reference samples have been distributed to District personnel since this program began. It is intended that the program will expand so that the quality of other field determinations, such as alkalinity, will be monitored. In addition, development of a computer program which compares field and laboratory pH and specific conductance values is being completed. It is expected that this program soon will be distributed to WRD districts and that the districts will be able to use this program in a self-monitoring process so that on a weekly or monthly basis they can compare field values with laboratory values (which, it is recognized, may differ because of actual changes in the sample). This program is intended to "fill in" the gaps between the semiannual distribution of reference materials. Another active area of quality assurance is the testing of materials (filters, bottles, and so forth) used in sample collection. For example, the 1978 contamination of lead and cadmium in samples preserved with nitric acid in ampoules indicated the need for an improved program to assure the quality of materials used for trace metal analyses. A procedure to combine standard single- and double-sampling methods (Dodge-Romig, 1959) was formulated and was used to develop a program to assure the quality of ampouled acid and polyethylene bottles. Although much work remains to be done, quality-assurance programs are undergoing rapid development. The need for the capability to compare data from samples presently being collected at site A by observer B which are analyzed by method C in laboratory D with data from samples collected ten years from now at site W by observer X and analyzed by method Y in laboratory Z clearly makes development of such programs of paramount importance. REFERENCES Delaney, B. M., and Schroder, L. J., 1979, Preparation of reference materials for particle-size analyses of silts and clays: U.S. Geological Survey Open-File Report 79-1590, 14 p. Dodge, Harold F., and Romig, Harry G., 1959, Sampling inspection tables, single and double sampling, second edition: New York, John Wiley & Sons, Inc., 224 p. Durum, W. H., 1978, Historical profile of quality of water laboratories and activities, 1879-1973: U.S. Geological Survey Open-File Report 78-432, 235 p. Howard, C. S., 1933, Determination of total dissolved solids in water analyses: Industrial and Engineering Chemistry, v. 5, no. 1, p. 4-6. Skougstad, Marvin W., and Fishman, Marvin J., 1975, Standard reference water samples: Proceedings American Water Works Association Water-Quality Technology Conference, December, 1974, p. XIX-l - p. XIX-6.