PROGRAMS AND PLANS--Dissolved Trace Element Data In Reply Refer To: September 30, 1991 Mail Stop 412 OFFICE OF WATER QUALITY TECHNICAL MEMORANDUM 91.10 Subject: PROGRAMS AND PLANS--Dissolved Trace Element Data CONTAMINATION OF DISSOLVED TRACE-ELEMENT DATA: PRESENT UNDERSTANDING, RAMIFICATIONS, AND ISSUES THAT REQUIRE RESOLUTION BACKGROUND In 1986, the U.S. Geological Survey (USGS) Office of Water Quality (OWQ) began a continuing evaluation of the methods and equipment used to produce water-quality data, in general, and for the National Stream Quality Accounting Network (NASQAN) program in particular. The goal is to identify contamination and other sources of variation in data introduced by sample collection, sample processing, and analytical procedures, and then, to take precautions or change methods to correct the problems. By 1989, these activities had begun to focus to a considerable degree on the quality of dissolved trace-element data. This focus stemmed partly from concern over possible contamination in dissolved mercury and lead results, and partly from reports appearing in the scientific literature that differences in concentrations of dissolved trace elements exist between data produced for NASQAN and by several university projects. The initial observations and comments about NASQAN data for dissolved trace elements were made by Shiller and Boyle (1987) and Flegal and Coale (1989). In 1990, the USGS conducted two studies to investigate aspects of the quality of dissolved trace-element data: (a) a Blank Sample Study (BSS) to detect potential contamination in water blanks processed through precleansed field equipment, and (b) a Mississippi River Methods Comparison Study (MRMCS), wherein dissolved trace-element data were produced using three different protocols for collecting, processing, and analyzing samples, namely, the protocols used by NASQAN, Howard Taylor's National Research Program (NRP) project, and Alan Shiller's project at the University of Southern Mississippi. Then, the June 1991 issue of Environmental Science and Technology contained an article written by Windom and others entitled "Inadequacy of NASQAN Data for Assessing Metal Trends in the Nation's Rivers." The article reports that based on recent work, concentrations of dissolved cadmium, copper, lead, and zinc in 18 East Coast rivers are considerably lower than values reported under the USGS NASQAN program for samples collected during "similar" time periods and at "similar" locations. PURPOSES The purposes of this memorandum are to: (a) present the current understanding of whether USGS data for dissolved trace elements are contaminated, (b) describe preliminary plans for examining all aspects of the issue, (c) describe changes in NASQAN analytical determinations for fiscal year (FY) 1992, and (d) suggest how USGS District offices might proceed with dissolved trace-element work in the Federal-State Cooperative Program while important issues are being resolved. The memorandum includes two appendices, two tables, and 16 figures to convey results of: (a) studies of dissolved trace- element concentrations in North American rivers, and (b) preliminary interpretations from the USGS-MRMCS and BSS. The level of detail in the presentation is commensurate with the importance of the findings with regard to the quality of USGS dissolved trace-element data and the urgency for field and laboratory studies to determine how to proceed with trace-element work in USGS programs and projects. PRESENT UNDERSTANDING During 1991, newly available data from various studies have enabled the OWQ to make initial evaluations of the quality of NASQAN dissolved trace-element data. The OWQ has: (a) as noted above, begun to evaluate dissolved trace-element data from the MRMCS and the BSS, (b) reviewed selected trace-element projects and protocols of Environment Canada, and (c) carefully reviewed trace-element data recently reported in the literature. The results of the MRMCS are presently incomplete, but it appears that Howard Taylor's NRP data are comparable to Alan Shiller's data. Further, the dissolved trace-element concentrations found by Shiller and Taylor are comparable to Windom's data, and to results generated by two separate Canadian Studies (Table 1). Thus, similar concentration ranges exist for various dissolved trace elements in the Mississippi River (Taylor in the MRMCS, Shiller in the MRMCS, and Shiller and Boyle, 1987), 18 East Coast rivers (Windom and others, 1991), the St. Lawrence River (Lum and others, 1991), and small Canadian Shield streams (Robert McCrea, Environment Canada, written commun., 1991). The dissolved concentrations reported by these studies are mostly in the 10's of parts per trillion (ppt) for cadmium and lead, in the low 100's of ppt for chromium, in the low to high 100's of ppt for zinc, and between the mid 100's to 1,800 ppt (1.8 parts per billion) for copper and nickel. The six cited studies report comparable trace-element concentrations in diverse river systems despite the use of five different sampling methods (e.g., depth- and width-integrated sampling using both conventional samplers and bag samplers, surface grab sampling, surface pump sampling, and manual width- integrated sampling using prepackaged contaminant-free equipment) and five sample processing techniques to remove particulate matter (e.g., conventional filtration, exhaustive filtration, cartridge filtration, continuous flow centrifugation followed by chelation techniques, and no removal of solids [for the virtually particulate-free streams studied by McCrea]). The one common feature of the six studies was the use of "ultra clean" protocols. "Ultra clean" refers to: (a) avoidance of metal samplers, (b) stringent precleansing of all containers, sampling equipment, filtration equipment, and filters, (c) use of very high quality water and acids for preparatory washing, blanks, preservation, and analyses, (d) special precaution in the collection and field handling of samples, including avoidance of all metal surfaces, use of plastic gloves and forceps, and avoiding car exhaust and atmospheric deposition (some projects conduct field processing of samples in portable laminar-flow hoods), and (e) use of a class 100 clean room, or better, for laboratory processing and analyses of samples. As one part in the overall ultra clean process, the cleansing of sample bottles and glassware varies from a few to many steps depending on the investigator. As an example, Appendix 1 describes the very detailed bottle and glassware cleansing protocol used by Lum and others (1991). In contrast to the cited trace-element concentrations for these six studies, results obtained using standard USGS protocols are much higher (Table 1, column 3). In the MRMCS, concentrations for individual District samples (collected and processed by District personnel and analyzed by the National Water-Quality Laboratory (NWQL)) exceed concentrations in comparable Taylor and Schiller samples (collected at the same time from the same cross sections) by 4-fold for copper, to greater than 100-fold for cadmium. Further, the design of the MRMCS allows the inference that the differences in concentrations result primarily from sample collection and field processing, rather than from laboratory analysis. Shiller and Boyle (1987) and Windom and others (1991) attribute such observed differences in dissolved trace-element concentrations to contamination introduced during the sampling, processing, and analysis of USGS samples. This certainly could be the cause of some, or all of the noted differences. However, the noted differences might partially result from: (a) variations in sample processing techniques (e.g., differences in particle removal procedures might cause differences in the amount of colloidal material incorporated and analyzed as "dissolved"); and (b) removal of truly dissolved trace elements by adsorption during use of ultra clean protocols processing due to sorption losses on equipment and filters. The latter could occur if the exhaustive acid cleaning opens active adsorption sites on bottles, equipment, and filters which are not thoroughly equilibrated with excess sample prior to processing aliquots collected for analysis. Appendix 2 presents scatter plots and a table comparing District and NRP data from the MRMCS. This is followed by a series of box plots of data from the BSS. Based on these results, plus those in Table 1, it appears that USGS operational program data for rivers are significantly contaminated for arsenic, boron, beryllium, cadmium, chromium, copper, lead, and zinc (see "Conclusion" section of Appendix 2). The contamination appears to result from the sample collection and sample processing steps. Based on the BSS data, and additional laboratory comparison results from the MRMCS (not presented in this memo), the NWQL does not seem to be a major source of contamination at the relatively high (i.e., compared to NRP and Shiller) reporting levels presently used in NASQAN for dissolved trace-element data. No USGS data exist to evaluate the possible contamination of dissolved trace-element data in ground-water samples. However, because USGS dissolved ground-water samples are filtered, it is possible that resultant data are contaminated for several trace elements. Until recently, nearly all water resource organizations used methods similar to USGS conventional methods for collecting, processing, and analyzing samples for dissolved trace elements. About 15 years ago, the oceanographic community began adapting knowledge of analytical research chemists to develop integrated protocols that significantly reduced contamination in trace- element results (Bruland, 1983; Windom and others, 1991). In North America, these ultra clean protocols were later applied to studies in the Great Lakes; however, they were not widely implemented in rivers. As a result, for North America, it is probable that most dissolved trace-element data collected from rivers before about 1985 overestimate ambient environmental concentrations. We know of no research or other agency data to use for evaluating the quality of the historic data base for dissolved trace elements in ground water. WHAT NEEDS TO BE DONE The OWQ has done the following: 1. Prepared this memorandum to: (a) describe the problem, (b) raise questions that need to be answered, and (c) suggest how Districts might work with cooperators on this issue. 2. Initiated development of a small capability for ppt analysis within the NWQL. At present, for dissolved trace-element data, the levels of contamination contributed by the NWQL are much lower than those contributed by field activities. However, because our goal is to develop ppt capability for sampling, sample processing, and analysis, we must reduce reporting levels, and therefore laboratory contamination, to much lower levels. Further, as we systematically eliminate contamination from field activities, the present laboratory contamination levels will become more significant. We have established a target date of analyzing the first environmental samples using ppt technology in October 1992. The ppt capability is necessary to conduct studies: (a) on dissolved trace-element sampling and processing, (b) that address concentration differences resulting from factors other than contamination, and (c) of the significance of dissolved trace- element concentrations in natural systems with regard to various program objectives. After we have developed the ppt capability, resolved some issues, and determined costs, we will work with individual programs to: (a) establish specific needs for dissolved trace-element data, (b) determine whether ppt data are required, or whether ppb data will suffice, and (c) if ppt data are required, evaluate the alternative protocols for meeting the specified needs. 3. Defined experiments that need to be conducted using ppt technology to evaluate whether the ultra clean methodology causes low results by removing trace elements from environmental samples. This possibility is unlikely, but must be evaluated. These experiments can be done now by NRP projects that have ppt capability, and/or by the NWQL beginning in FY 1993. 4. Evaluated equipment for the sampling and processing of dissolved trace elements in surface waters. At present, it appears that the modified bag sampler (Leenheer and others, 1987) is the most appropriate sampling device for dissolved trace- element samples in large rivers. We may need to develop an equivalent, or use a pre-existing design (e.g., McCrea's sampler, Environment Canada) for shallow streams. Also, we must determine the most appropriate equipment and procedures for collection of dissolved trace-element data in ground water. The OWQ will need to: 1. Develop and implement guidelines for comprehensive quality assurance of all water-quality data, including trace-element data, in all media. The guidelines must cover all aspects of sample collection, sample processing, laboratory analysis, data validation, and data storage/retrieval. 2. Develop and document suitable protocols for producing dissolved trace-element data at both the ppb and ppt levels for surface and ground waters. An improved protocol for surface- water samples at the ppb level will be developed first and should be available in early 1992. 3. Evaluate methods used to remove particles from water samples prior to analysis for the "dissolved phase." At a minimum, we need to explore and compare the "standard" filtration techniques, chelation, dialysis, ultrafiltration, and super-centrifugation. Moreover, the different filtration techniques used to produce the data in Table 1 need to be compared to evaluate the possible effects of processing artifacts on the reported concentrations. 4. Provide a more complete basis for decisions about which dissolved and total recoverable trace-element data are suspect, so the USGS can begin to make informed choices regarding what to do about existing data (As previously noted, Appendix 2 provides preliminary information on this issue). 5. Evaluate how to collect, process, and analyze whole water samples for total recoverable determinations of trace elements. Such samples are not filtered, but could be contaminated by present sampling techniques. Data accuracy could be improved by using: (a) noncontaminating sampling methods, and (b) laboratory clean room capability developed for ppt determination of dissolved trace elements. 6. Evaluate methods for collecting, processing, and analyzing suspended sediment, bed sediment, and tissue samples for trace elements. Because increased emphasis on these components is likely in reconnaissance-type work, review of current procedures, and possible methods development is warranted. One problem the USGS does not face is procurement of new and expensive analytical equipment. The graphite furnace atomic absorption capability now in place in the NWQL and the ICP/MS capability presently awaiting final approval are completely adequate for ppt analyses. The challenge is to prevent contamination during the sample collection, processing, and analysis steps to enable production of accurate, reproducible data at the ppt level. WHAT THE FUTURE MAY HOLD We might have a future where: 1. For certain elements, historic data are limited to qualitative value. 2. The cost of doing dissolved trace-element determinations at the ppt level is so high that such analyses are inappropriate for routine measurement in operational programs. We could be looking at a future where the total number of samples for dissolved trace- element determinations per year at the ppt level (for non-NRP programs) are less than 500 compared to over 6,000 samples analyzed now using conventional methodologies. Until research is conducted and issues resolved, the number of samples appropriate for ppb analysis of dissolved trace elements is unknown. 3. Suspended sediments, bed sediments, aquatic plant tissues, and animal tissues are common matrices of study for trace elements in operational (non-research) programs for assessing anthropogenic effects on the chemistry of aquatic systems. Trace-element concentrations in sediment samples typically are in the 10's to 100's of parts per million (ppm) for copper, lead, nickel, and zinc; and in the single digit ppm for cadmium. Thus, prevention of sample contamination will not be as difficult an issue as with analysis of dissolved phase samples. 4. Dissolved trace-element data at the ppt level are produced by: a. Special teams of individuals who conduct the entire process of preparation, sample collection, sample processing, and laboratory analysis. This would require a new direction for the NWQL wherein lab personnel become members of project teams with costs paid by the project, or b. An approach wherein precleansed and disposable equipment is used for sample collection and field processing. This approach is typically used in the medical field and is being used by McCrea of Environment Canada. The observed similarity (in Table 1) of dissolved trace- element concentrations in diverse river systems leads to several hypotheses which need to be tested. Perhaps, dissolved concentrations are controlled by thermodynamic limits which, under the physiochemical conditions of fluvial systems, do not vary substantially from one system to another. Perhaps the observed low and fairly consistent concentrations of dissolved trace elements result from kinetic controls, or a combination of kinetic and thermodynamic controls. Regardless of the cause, the low dissolved trace-element concentrations recently found in diverse river systems indicate a relatively small contribution from dissolved trace elements to total concentrations and fluxes. Research is needed to determine: (a) if the concentrations of dissolved trace elements are consistently low in freshwaters, (b) if so, what the controls are, and (c) if so, how best to monitor trace elements in operational programs to assess human impacts on water quality, to determine trends, and to measure fluxes. If the concentrations of dissolved trace elements are confirmed to be in the ppt range, measurement will nevertheless continue in projects devoted to understanding processes, rates, environmental controls, and toxicology. However, because of cost, routine measurement of dissolved trace elements in operational programs such as NASQAN would probably cease, and other approaches to providing data for assessing the effects of human activities on trace elements in rivers, and trends thereof, would need to be employed. RAMIFICATIONS FOR FISCAL YEARS 1992 AND 1993 As previously noted, the following elements exhibit significant contamination for dissolved determinations: arsenic, boron, beryllium, cadmium, chromium, copper, lead, and zinc. In addition, some of the Division's recent dissolved mercury data are contaminated. Therefore, at the beginning of FY 1992, the OWQ will discontinue determining the cited list of elements plus mercury at all NASQAN and Hydrologic Benchmark stations. This list may grow as data are generated and interpreted from: (a) methods comparison studies in other climatic-geohydrologic regions, and (b) additional blank studies. For FY 1992, we will continue to determine dissolved cobalt, lithium, molybdenum, nickel, silicon, uranium, and vanadium in the NASQAN and Benchmark Programs. We will also continue to determine the major ions, plus aluminum, barium, iron, manganese, and strontium. For the future, we will need to establish specific objectives for dissolved trace- element data in the national networks and decide where, when, and how it is appropriate to collect such data. We have advised the National Water-Quality Assessment Program not to measure dissolved trace elements until the USGS has resolved the various trace- element issues and developed suitable protocols. In the FY 1992 Federal-State Cooperative Program, Districts should decide whether to determine dissolved trace elements by conventional methods for dissolved and whole water samples on a project-by-project basis, with full consideration of the environment under study, the goals of the project, and the needs of the cooperators. There may be hydrologic components with high trace-element concentrations--such as acid mine drainage and urban runoff--where present methodologies are acceptable. Although most cooperators may be unaware of the ultra clean technology, or its ramifications, it is important for Districts to advise affected cooperators of the situation, and discuss options for specific projects. In the short term, cooperators may continue to request conventional sampling, processing, and analyses of whole water samples geared to current drinking water standards, Maximum Contaminant Levels (MCLs) for human health considerations, and aquatic health criteria set by the U.S. Environmental Protection Agency (EPA). As previously noted, an improved protocol for producing data at the ppb level for surface-water samples will be available in early 1992. This protocol will be applicable to both dissolved and whole water samples and should meet the needs of some cooperative projects. In the future, as more data are generated using ppt technology, EPA may discover that regulatory criteria and MCLs need to be revised. The improved protocol for ppb-level work will include new details for sample collection, field processing, and laboratory handling and analysis. This protocol will build upon present methods but make improvements based on work in progress by Art Horowitz, the NWQL, Howard Taylor, Alan Shiller, and Environment Canada. Simultaneously, with the writing of this protocol, research will proceed on: (a) particle removal (phase separation) procedures, and (b) possible chemical artifacts produced by the ultra clean technology. As noted, the work already initiated on development of ppt capability in the NWQL has a goal of enabling initial analysis of environmental samples in October 1992. However, experiments using the new technology will extend through FY 1993. Thus, it will probably be FY 1994 before we can reach final decisions on: (a) how USGS operational programs should approach future work on dissolved trace elements, and (b) appropriate caveats for the historic trace-element data base. For existing data, the goal is to provide a basis for decisions about which data are suspect. It may be possible to establish the maximum levels of contamination introduced during sample collection and processing. For example, certain sampling devices may contaminate samples at levels exceeding the present NWQL reporting levels, whereas other devices may not. Perhaps no sampling devices contaminate samples above the present reporting levels for certain trace elements. Such possibilities need to be systematically defined. As part of the process to sort through the dissolved trace- element issue, the OWQ will convene a panel of experts to provide advice. A number of USGS, Environment Canada, and university scientists have expressed an interest and willingness to participate. OWQ will work with the panel to address questions of: (a) what do we know now, and what conclusions can we reach; (b) what are the purposes in operational programs for collecting dissolved trace-element data, (c) are acceptable substitutes (media) available for certain purposes; (d) what additional information--from experiments and other sources--do we need for making decisions about future dissolved trace-element work in the operational program; and (e) what institutional changes are necessary to facilitate trace-element work in the USGS. The goal is to produce a steady stream of information for the Districts throughout FY's 1992 and 1993, leading to major decisions in FY 1994. If information in this memorandum prompts questions, comments, or concerns, please enter them in QWTALK so that others in the Division can share in the discussion. REFERENCES Bruland, K.W., 1983, Trace elements in sea-water, in Chemical Oceanography: New York, Academic Press, v. 8, p. 157-220. Flegal, A.R., and Coale, K., 1989, Discussion: trends in lead concentration in major U.S. rivers and their relation to historical changes in gasoline-lead consumption by R.B. Alexander and R.A. Smith: Water Resources Bulletin, v. 25, p. 1275-1277. Leenheer, J.A., Meade, R.H., Taylor, H.E., and Pereira, W.E., 1989, Sampling, fraction, and dewatering of suspended sediment from the Mississippi River for geochemical and trace-element analysis, in Mallard, G.E., and Ragone, S.E., eds., U.S. Geological Survey Toxic Substances Hydrology Program-- Proceedings of the technical meeting, Phoenix, Arizona, September 26-30, 1988: Water-Resources Investigations Report 88-4220, p. 501-511. Lum, K.R., Kaiser, K.L.E., and Jaskot, C., 1991, Distribution and fluxes of metals in the St. Lawrence River from the outflow of Lake Ontario to Quebec City: Aquatic Sciences, v. 53, no. 1, 19 p. Shiller, A.M., and Boyle, E.A., 1987, Variability of dissolved trace metals in the Mississippi River: Geochimica et Cosmochimica. Acta, v. 51, p. 3273-3277. Windom, H.L., Byrd, J.T., Smith, R.G., Jr., and Huan, F., 1991, Inadequacy of NASQAN data for assessing metal trends in the nation's rivers: Environmental Science and Technology, v. 25, no. 6, p. 1137-1142. David A. Rickert Attachments This memorandum does not supersede any previous Office of Water Quality technical memorandum. Key Words: Contamination, trace elements Distribution: A, B, S, FO, PO APPENDIX 1 (From Lum and others, 1991) Procedure for Cleaning New Polyethylene Bottles 1. Wash with detergent (1-2 percent Extran) prepared with line- distilled water (metal still) and rinse well with distilled water. Shake off excess water. 2. Add sufficient distilled-in-glass acetone, cap and shake for ca. 1 minute. Decant and shake off excess acetone. 3. Soak in 6 M hydrochloric acid bath (Baker Instra-Analyzed) at 50 degrees C for 2 days. 4. Next soak in 2 M nitric acid bath (Baker Instra-Analyzed) at 50 degrees C for 2 days. 5. Rinse five times with Chelex-water (deionized-distilled water stored in contact with purified Chelex-100 resin, 50-100 mesh, in reagent acid bottles previously rinsed with Chelex water). Note that in preparing Chelex-water, the first batch is used to clean and condition the inner surface of the reagent bottle. This batch is discarded, and the bottle refilled with distilled-deionized water and allowed to equilibrate overnight before use. 6. Fill with 1 percent nitric acid (double sub-boiling distilled and supplied in acid-cleaned teflon bottles) prepared with Chelex water and store in zip-lock plastic bags. Procedure for Cleaning Previously Used Bottles and All Glassware 1. Rinse in deionized-distilled water. 2. Soak in 2 M nitric acid for at least 24 hours. 3. Rinse five times with Chelex-water. 4. Fill with 1 percent nitric acid prepared with Chelex water and store in zip-lock plastic bags. New sample bottles and glassware, once used, are reserved for the same type of sample. APPENDIX 2 PRELIMINARY INTERPRETATIONS FROM USGS STUDIES CONCERNING CONTAMINATION OF DISSOLVED TRACE-ELEMENT DATA The Mississippi River Methods Comparison Study Figure 1 describes and Figures 2 through 9 present sets of scatter plots for the six trace and two minor elements listed in Table 1. For cadmium (Fig. 2), chromium (Fig. 3), copper (Fig. 4), lead (Fig. 6), zinc (Fig. 7), and aluminum (Fig. 8), each set includes two plots representing: 1. Sampling effects. The differences between pairs of samples collected at nine locations wherein: o One sample in each pair was collected by District crews, and then processed (filtered) and analyzed by the NRP personnel. This is identified as the District sample. o The second sample of the pair was collected, processed, and analyzed by NRP personnel. This is identified as the NRP sample. In each case, the sampling effects include those of the sampling device, the act of collecting the sample, the churn splitter, and the act of using the churn splitter. 2. Processing effects. The differences between pairs of samples collected at 10 sites wherein: o One sample was collected by NRP personnel, processed by a District crew, and then analyzed by NRP personnel. This is identified as the District sample. o The second sample was collected, processed, and analyzed by NRP personnel. This is the second sample in item 1 above, identified as the NRP sample. Thus, for the sampling-effect plots, the only difference between the District and NRP samples is who collected the sample, whereas in the processing-effects plots, the only difference is who processed the samples. For these six figures, note that all laboratory determinations were made by the NRP. Nickel (Fig. 5) and iron (Fig. 9) were analyzed by the NWQL, rather than the NRP. Thus, for these elements (see Fig. 1), the sampling-effect plot compares nine samples collected by District crews, processed by NRP personnel, and analyzed by the NWQL versus nine samples collected and processed by NRP personnel and analyzed by the NWQL. The processing effects plot compares nine samples collected by NRP personnel, processed by District crews, and analyzed by the NWQL versus nine samples collected and processed by NRP personnel and analyzed by the NWQL. The eight sampling-effects plots exhibit positive biases (District concentrations greater than NRP concentrations) for cadmium, chromium, copper, lead, zinc, and aluminum, plus a negative bias for iron (District concentrations less than NRP concentrations). The eight processing-effects plots exhibit positive biases for cadmium, chromium, copper, lead, zinc, aluminum, and iron. Table 2 summarizes numerically the effects of sampling and processing for 25 elements and also indicates the sign and statistical significance (determined by the Sign Test) of the effects. In Table 2, it is easy to see the magnitude of sampling and processing effects relative to the median concentrations of the samples collected, processed, and analyzed by the NRP (except for Ag, Fe, and Ni, which were determined by the NWQL). The Sign Test results presented in table 1 pertain only to the signs of differences in the paired data sets; the magnitude of difference is not considered. A one-tail test was run to determine the significance of concentrations in the District samples exceeding those in the paired NRP samples. The Sign Test results are significant when concentrations of elements in the District samples are consistently higher. As an example, for zinc, the District processed samples exhibited higher concentrations than the NRP counterparts in each of 10 pairs, and the resultant p value was 0.002. For iron, the District processed samples had higher concentrations than the NRP samples in 8 of 10 sample pairs, lower concentrations in the other two, and the resultant p value was 0.11. The Wilcoxen Test which evaluates both the sign and magnitude of differences was run, but the results are not reported because numerous "less than" values occur in the paired data sets for all of the trace elements of concern. The Blank Sample Study With the results discussed to this point, consistent and sometimes major sampling and processing differences are apparent between the District and NRP protocols. Generally, District concentrations are higher than NRP concentrations. However, no cause has been illustrated. To address cause, Figures 10 through 16 show blank data for all elements listed in Table 1 except aluminum (which was not measured in the BSS). In each figure (from left to right): 1. The DIW is the concentration of the noted constituent in the deionized water used for the blanks. 2. The shelf blank is an aliquot of the original DIW and represents analyte contamination/loss from storage. 3. The trip blank was DIW carried to the field; the container was opened, part of the water was used, and an aliquot returned to the laboratory. This represents analyte contamination/loss from storage and the atmosphere. 4. The sampler blank was obtained by processing DIW through the sampler. This represents analyte contamination/loss from storage, the atmosphere, and the sampler. 5. The churn blank was obtained by processing DIW through the sampler and the churn. This represents analyte contamination/ loss from storage, the atmosphere, sampler, and churn. 6. The filter blank was obtained by processing DIW through the sampler, churn, and filter/filter apparatus, then collecting the filtrate after the first 500 ml were discarded. This represents analyte contamination/loss from storage, the atmosphere, sampler, churn, and filtration. The DIW and shelf blanks were prepared, and the trip, sampler, churn, and filter blanks were collected in 1990. All samples were then frozen without acidification. In 1991, all samples were thawed, acidified and held for one week, and then analyzed. In Figures 10-16, note that the NWQL reporting limits for the BSS were identical to the NASQAN reporting limits for nickel, zinc, and iron, but lower than the NASQAN reporting limits for cadmium, copper, chromium, and lead. More conclusive interpretations would have resulted from the BSS if DIW had been put through just one individual step in the field handling sequence, rather than integrating multiple steps. New experiments designed in this manner are planned. At the reporting levels used in the BSS, Figures 10 through 16 show large amounts of contamination arising from the sampling step for copper, lead, and zinc; and from the filtration step for cadmium, lead, and zinc. Except for iron, the churn step did not appear to be a source of contamination. The observed levels of contamination in the BSS results are substantial at present NASQAN reporting levels for copper, lead, zinc, and iron, but not for cadmium, chromium, and nickel. Conclusions We have three lines of evidence to reach conclusions: (a) the similarity of dissolved trace-element concentrations reported in Table 1 for diverse river systems wherein the investigators used five different sample collection methods and five different particle separation methods, (b) the comparisons of District to NRP data from the MRMCS as presented in Figures 2 through 9 and Table 2, and (c) the observed contamination in the BSS data shown in Figures 10 through 16. We also have information from many discussions with hydrologists and chemists in the USGS, Environment Canada, and several universities. Based on these multiple lines of evidence, the OWQ has provisionally categorized the USGS data for dissolved elements (trace, minor, major) as follows: 1. Noncontaminated or minimally contaminated--barium, calcium, cobalt, lithium, magnesium, molybdenum, nickel, sodium, silicon, strontium, uranium, and vanadium. 2. Significantly contaminated--arsenic, boron, beryllium, cadmium, chromium, copper, lead, and zinc. From other studies and lines of evidence, the mercury data base is known to contain contaminated results. 3. Significantly different from NRP data, but the differences may result largely from filtration artifacts, rather than contamination--aluminum, iron, and manganese. 4. As yet undetermined--selenium and silver. Additional work is needed to confirm and further explore the provisional listings. First, blank studies are needed of a different design that cover more elements. Second, methods intercomparison studies in surface waters are needed in different climatic-geohydrologic regions to test for introduced trace elements from contamination by aerosols during sample handling. Third, studies are needed to test for contamination by conventional methods for the sampling and processing of ground- water samples. The case of aluminum, iron, and manganese needs special work. In Table 2, the median sampling and processing differences are large for these elements relative to their median concentrations. However, a soon to be published article by Art Horowitz demonstrates that differences in the concentrations of aluminum and iron of the magnitudes summarized in Table 1 can arise from differences in how replicate samples are filtered. Additional work is needed on the effects of filtration on the levels of: (a) aluminum, iron, and manganese, and (b) the trace elements.