Wayne W. Lapham, Ground-Water Specialist

National Water-Quality Assessment Program, U.S. Geological Survey

MS 413, 12201 Sunrise Valley Drive, Reston, VA 20192

Phone: (703) 648-5805; Fax: (703) 648-6693; E-mail: wlapham.er.usgs.gov

Michael J. Moran, Hydrologist

NAWQA VOC National Synthesis, Rapid City, SD

Phone: (605) 355-4560 x244; Fax: (605) 355-4523; E-mail: mjmoran@usgs.gov

John S. Zogorski, Chief

VOC National Synthesis Project, U.S. Geological Survey, 1608 Mt. View Road, Rapid City, SD 57702

Phone: (605) 355-4560 x214; Fax: (605) 355-4523; E-mail: jszogors.cr.usgs.gov

The U.S. Geological Survey (USGS) has compiled a national retrospective data set of analyses of volatile organic compounds (VOCs) in ground water of the United States. The data are from Federal, State, and local nonpoint-source monitoring programs and were collected during 1985-95. This data set will be used to augment data collected by the USGS National Water-Quality Assessment (NAWQA) Program for assessment of the occurrence of VOCs in ground water nationwide.

Eight attributes of the retrospective data set were evaluated to determine the suitability of the data to augment NAWQA data in answering occurrence questions of varying complexity. The attributes include: the VOCs analyzed for and the associated reporting levels, water use, well-casing material, well depth, depth from land surface to the water level in the well, aquifer lithology, and distribution of sampling points. The data set generally lacks documentation of some characteristics of each well sampled, such as casing material and depth from land surface to the water level in the well. Only about 20 percent of the wells have associated documentation of aquifer lithology.

More than 90 percent of the VOC data are suitable for use in addressing simple occurrence questions relative to some current drinking-water regulations of about 5 micrograms per liter (m g/L). However, only about 15 percent of the data are suitable for addressing these simple occurrence questions at a much lower assessment level of 0.2 m g/L. Enhancing monitoring-program data bases would greatly increase the usefulness of these data in addressing complex occurrence questions, such as those that seek to explain the reasons for VOC occurrence and nonoccurrence in the Nation’s ground water. The three most important enhancements to the data set would be: an expanded VOC analyte list, recording the reporting level for each analyte for every analysis, and recording key ancillary information about each well.

In 1991, the U.S. Geological Survey (USGS) National Water-Quality Assessment (NAWQA) Program began full-scale implementation. The long-term goals of the NAWQA Program are to describe the status and trends in the quality of a large representative part of the nation’s surface-water and ground-water resources. Another goal is to provide an improved understanding of the primary natural and human factors that affect the quality of these resources. The NAWQA Program has two major operational components: 1) hydrologic investigations of large river basins and aquifer systems, referred to as Study-Unit Investigations; and 2) a National Synthesis that is organized to provide information about water-quality topics of national or regional concern. Volatile organic compounds (VOCs) are studied as part of the NAWQA Program because of the occurrence of this constituent group in many of the Nation’s water supplies (Office of Technology Assessment 1984, Tennant et al. 1992, Pankow and Cherry 1996). To facilitate an assessment of VOCs in the Nation’s ground water, a "Ground-Water VOC Retrospective Data Set," herein referred to as the retrospective data set, has been compiled.

The primary purposes of compiling the data set are: 1) to ascertain what high-quality, ambient-ground-water-monitoring data exists so that NAWQA does not duplicate these efforts; 2) to augment data collected by NAWQA; and 3) to provide data for areas that are not covered by NAWQA Study Units. The data set contains information collected in nonpoint-source ground-water-quality monitoring networks by various Federal, State and local agencies (Lapham et al. 1997). The locations of monitoring programs in the retrospective data include areas in NAWQA Study Units that have not been sampled to date (1998) and areas outside of the NAWQA Study Units. Lapham and Tadayon (1996) provided design elements for compilation of this data set and developed criteria for screening data to ensure that their inclusion would be useful for answering questions that involve the occurrence of VOCs in ground water at a national scale.

The purpose of this paper is to summarize eight key attributes of the retrospective data set and to discuss the importance of these attributes in answering questions of occurrence, status, and distribution of VOCs in ground water at a national scale. Suggested enhancements related to these attributes are given that would improve our understanding of the occurrence of VOCs in ground water at regional and national scales.

Table 1 presents some general characteristics of the retrospective data set. Additional information about this data set is provided in Lapham et al. (1997; table 5). NAWQA Study Units and the VOC National Synthesis, in cooperation with State and local agencies, compiled the data set from 1995 to 1996. Currently (1998), data for as many as 50 VOC compounds from 43 nonpoint-source-monitoring programs or networks have been compiled into the retrospective data set (Lapham et al. 1997; table 5). About one-half of the wells were sampled from 1985 through 1989 and the other one-half were sampled from 1990 through 1995. About 75 percent of the wells in the data set for which the use-of-water is known are either public- or domestic-supply wells. The number of VOCs analyzed in a water sample range from 24 to 50 (table 1).

There is a wide range in the complexity of questions related to VOC occurrence. For example, one of the simplest, yet an important occurrence questions is "What percentage of the sampled wells contain a regulated VOC at a concentration that exceeds its Maximum Contaminant Level (MCL)?" Answering this occurrence question is relatively simple and requires only knowledge of the number of wells sampled, the number of wells that had one or more VOCs detected above MCLs, and the laboratory’s reporting level. An example of a considerably more complex question is "Are VOCs found in the deeper parts of unconfined aquifers or are they only present in the upper, more vulnerable portions of these aquifers?" Answering this question requires knowledge of the location of sampled wells and whether the aquifer in which each well is located is confined or unconfined. Additionally, information about the depth in the aquifer at which water is being withdrawn needs to be known. This additional information and other information about well construction and aquifer characteristics are essential to answer complex occurrence questions. An important effort of the NAWQA Program is to fully populate ancillary data for each sampled well so that an attempt can be made to answer complex, nationally relevant occurrence questions.

In addition to the need for ancillary data, answering occurrence questions about VOCs in ambient ground water also requires large well networks and low reporting levels. These attributes are needed because VOCs often are reported as nondetected (i.e. below today’s analytical detection capability) and when detected, the concentrations are typically low, at sub-microgram-per-liter levels. The benefit of large well networks is the establishment of significant statistical relations between VOC occurrence and hydrogeologic and anthropogenic variables. There are, of course, practical limitations to the number of wells that can be sampled in a national assessment. Resources available within the NAWQA Program today (1998) allow the sampling of about 2,500 wells per 10-year cycle for a broad-scale occurrence assessment. As previously noted, sampling by the NAWQA Program is being augmented with results of sampling of comparable monitoring networks operated by State agencies and others. Data from a network of 5,000 to 10,000 wells or more, if possible, is sought because of varied hydrogeologic settings, aquifer types, land-use practices, and chemical-release patterns across the Nation.

The benefit of using a low analytical reporting level is an increase in the detection frequency. This is illustrated in figure 1. This figure was developed from VOC analyses of water samples from wells with shallow depths to the water table and within urban areas (urban land-use studies). The NAWQA Program (Squillace et al. 1996) collected these data between 1993-95. All analyses of VOCs represented on this graph were completed at the USGS National Water Quality Laboratory (NWQL) at a reporting level of 0.2 micrograms per liter (m g/L). Figure 1 was made by artificially censoring the data set at 12 different reporting levels ranging from 0.2 to 10 m g/L. The percentage of wells that would have been recorded as having a detection of one or more VOCs at each of these 12 reporting levels was calculated. Figure 1 indicates that the percentage of wells in which one or more VOCs were detected at various reporting levels did not increase notably as the reporting level decreased from 10 m g/L to about 2.0 m g/L. However, a notable increase in the percentage of wells in which one or more VOCs were detected occurs as the reporting level decreased from 2.0 m g/L to 0.2 m g/L. The percentage of wells in which one or more VOC were detected at a reporting level of 2.0 m g/L is about 22 percent, but increases to about 54 percent at a reporting level of 0.2 m g/L.

The NAWQA Program has lower detection levels for the analysis of VOCs in ambient ground water than used in many other monitoring programs. For about the last 10 years, the USGS has analyzed VOC samples at the USGS NWQL in Arvada, CO using an analytical method (Rose and Schroeder 1995) that has a reporting level of 0.2 m g/L for most VOCs. This reporting level was decreased even further in 1997, and now most VOCs can be detected at concentrations of 0.05 m g/L or less (Connor et al. 1998). Such low reporting levels are possible because of the simple water matrix being analyzed in contrast to, for example, contaminated ground water at a gasoline-spill site. Rigorous field and laboratory quality-control practices, application of modern, commercially available analytical equipment, and well-trained analytical chemists have been combined to achieve these low reporting levels.

Eight attributes associated with sampling wells for VOCs are considered to be key information necessary for interpretation and explanation of VOC occurrence in ground water. Analysis of these 8 key attributes in the retrospective data set provides insight into the data set’s suitability to augment NAWQA data. These 8 attributes are:

Two attributes are related to laboratory analysis of VOCs in each ground-water sample:

1. VOCs analyzed
2. Reporting level associated with each analyte for each analysis

Five attributes of each well, which are assumed to be the minimum information that should be known about a well before it is sampled:

3. Primary use of water from the well
4. Well-casing material
5. Depth of the well
6. Depth from land surface to water level in the well
7. Lithology of the aquifer contributing water to the well
8. Distribution of sampling both nationally and in four major geographic regions of the Nation.

A large number (55) of VOCs are analyzed for by NAWQA because documenting which VOCs are not detected in ground water is considered as important for a national occurrence assessment as documenting which compounds are detected. An examination of frequency of analysis for 53 of those 55 VOCs in the retrospective data set indicates that Federal, State, and local monitoring programs seldom analyze for the number of VOCs analyzed by NAWQA. The frequency of analysis of an individual VOC generally seems to be closely associated with the regulatory history of that VOC. More than one-half of the 53 compounds were analyzed for in at least one-half of the samples, and 9 compounds were analyzed for in at least 90 percent of the samples. The three most frequently analyzed compounds were 1,1,1-trichloroethane, tetrachloromethane, and trichloromethane. Twelve of the 13 most frequently analyzed VOCs are in the use category of solvents, industrial reagents, and refrigerants. Other compounds of current interest were not frequently analyzed such as methyl tert-butyl ether (MTBE), a gasoline oxygenate, and other fuel oxygenates.

Assessment of VOC occurrence cannot be completed without knowledge of the reporting level for each analyte for every analysis. The reason for this is that the level at which the assessment of occurrence is made (i.e. the assessment level) needs to be prescribed, and all data need to be censored to this prescribed assessment level before their use in an occurrence calculation. If the reporting level for each analyte for every analysis is not known, it has to be inferred, or the analyte has to be deleted from the data set. Recording the reporting level for each analyte for every analysis is particularly important when data are compiled into one collective data set from many monitoring programs or networks. Recording reporting levels is particularly important in this case because VOC analyses are done by many different laboratories possibly using different analytical methods. For example, the VOC retrospective data set discussed in this article was compiled from 43 monitoring programs or networks in 22 States, and reporting levels for VOC analyses among those programs vary by two orders of magnitude (Table 1).

Often the reporting level associated with nondetection of an analyte can be inferred because the nondetection is reported as an unknown concentration less than the reporting level. The reporting level associated with a detection of an analyte, however, usually is not recorded. This information, however, is required when censoring data to the prescribed assessment level prior to an occurrence calculation. The unrecorded reporting level for a detection of a VOC in the retrospective data set was inferred from less-than values recorded for nondetections of the same VOC in samples collected from other wells in the same monitoring program during the same round of sampling.

Nearly all (99 percent) of the data in the retrospective data set were analyzed at a reporting level of 5 m g/L or less, and about 70 percent of the data were analyzed at a reporting level of about 0.5 m g/L or less. However, only about 27 percent of the data were analyzed at a reporting level of 0.2 m g/L or less. Consequently, most of the data can be used for an assessment of the occurrence of one or more VOCs at an assessment level of 5 m g/L or higher. Most NAWQA VOC data collected from 1993-95 were analyzed at a reporting level of 0.2 m g/L. Therefore, at most, only about 27 percent of the retrospective data are suitable to augment NAWQA data when addressing occurrence questions at the low assessment level of 0.2 m g/L.

As previously discussed, answering complex occurrence questions requires ancillary information. The findings from other studies indicate that attributes of a well often help explain the occurrence of VOCs, nitrate, or pesticides in ground water. Five of the most important attributes of a well are: 1) the use of the water from the well; 2) the well-casing material; 3) the depth of the well (or a related attribute such as the depth from land surface to the top or bottom of the interval(s) contributing water to the well); 4) the depth from land surface to the water level in the well; and 5) the lithology of the aquifer at the interval contributing water to the well. Accordingly, the degree to which these five attributes and combinations of these attributes are populated in the retrospective data set for each well will partly determine the usefulness of the retrospective data in augmenting NAWQA data for occurrence assessment.

The percentage of wells that have the indicated attribute or combination of attributes populated in the retrospective data set is shown in table 2. Data in this table demonstrates that the usefulness of the data set decreases markedly as the information needs about each well increase. For example, about 74 percent of the wells in the retrospective data set have information on well depth. Consequently, 74 percent or less of the wells are potentially suitable for use in investigating the relation between VOC occurrence and well depth. Casing material, depth from land surface to the water level in the well, and aquifer lithology are documented in less than 25 percent of the wells, and combinations of two or more of these attributes generally are documented in only a small percentage of the wells. Only about 7 percent of the wells have information on well depth, use of water from the well, water level, and aquifer lithology. Consequently, only 7 percent or less of the wells are potentially suitable for multivariate analyses, such as principal-components analysis, of the relation between VOC occurrence and these four variables.

The percentages cited in table 2 were calculated for each of the five attributes or combinations of attributes regardless of the reporting levels for the VOCs analyzed in samples from the wells. If the assessment level selected for the occurrence assessment is low (for example, 0.2 m g/L), the percentage of wells that have attributes of the well populated and also have VOC analyses at a reporting level less than or equal to 0.2 m g/L would likely be lower.

An eighth consideration in evaluating the suitability of the data set for occurrence assessment is the geographic distribution of the data. Although the data set might have an adequate distribution to address an occurrence question at the national scale, distributions might not be adequate at smaller geographic scales. Furthermore, as the occurrence questions become more complex and the assessment level is decreased, the percentage of data in the data set that can be used to address those questions decreases. Examples of the percentages of the retrospective data suitable to address a series of occurrence questions of increasing complexity for various geographic regions at assessment levels of 5.0 and 0.2 m g/L are given in table 3. The assessment level of 5.0 m g/L is used for the first example because 5.0 m g/L is the value of the MCL for several VOCs regulated in drinking water. The assessment level of 0.2 m g/L is used for the second example because this is the assessment level currently used most often by NAWQA for analysis of VOC data. Percentages in table 3 were based on the number of wells that have at least one VOC analyzed at or below the indicated reporting level (i.e. the assessment level) and that also have the indicated supporting ancillary information populated for the well.

Question 1 (table 3), the simplest question, addressed at the national scale only requires VOC concentration data analyzed at or below the indicated reporting level. As indicated in table 3, data from nearly all (4,894 of 5,320 or 92 percent) of the wells in the data set meet these criteria at an assessment level of 5.0 m g/L. Answering this question at an assessment level of 0.2 m g/L, however, requires VOC analyses that were done at a reporting level less than or equal to 0.2 m g/L. Data from only about 15 percent of the wells meet this requirement.

A requirement for inclusion of an analysis in the data set is the location of the sampled well by latitude and longitude (Lapham et al., 1997). This information enables analysis of the VOC data by geographic region. The wells suitable for answering question 1 at an assessment level of 5.0 m g/L are distributed fairly evenly across the four geographic regions. Also, the number of wells in each of the four geographic regions is fairly large: 1,543 wells in both the northeast and southwest regions, 1,010 wells in the southeast region, and 798 wells in the northwest region. Therefore, the retrospective data set is probably adequate to address simple occurrence questions relative to drinking-water regulations at both the national and regional scales. Wells suitable for answering question 1 at an assessment level of 0.2 m g/L also are distributed fairly evenly across the four geographic regions. However, the percentages of wells in the data set suitable to address question 1 at this low assessment level in each of the four geographic regions are small (5 percent or less).

The most complex questions in table 3 (questions 4a - 4d) require the information needed to answer question 1 plus knowledge of water use and whether or not the source of the water to the well is from a consolidated or unconsolidated aquifer. Only a small percentage of the wells in the data set have this information fully documented. For example, only 7.1 percent (or about 378) of the wells have information to determine both the aquifer type and that the water use is for either domestic or public supply at an assessment level of 5.0 m g/L. Additionally, only a few percent or fewer of the wells have this information documented and have water samples analyzed for VOCs at an assessment level of 0.2 m g/L for each of the four geographic regions.

The utility of the retrospective data for occurrence assessment depends on the question being asked. The utility of the data generally decreases as the complexity of the question increases, where complexity is measured by the information required to answer the question posed. For example, information that would be required to answer a relatively simple question about the occurrence of one or more VOCs in the Nation’s ground water would include the concentration(s) of the VOC(s) of concern and the laboratory reporting levels associated with each VOC in each water sample. More complex questions also might require location information and various combinations of ancillary information, such as the depth of each well sampled, the depth from land surface to the water level in each well, the use of the water from each well, well-casing material, and the type of aquifer contributing water to each well.

Analysis of the characteristics of the retrospective data described in this article indicates that the three greatest limitations of the retrospective data for addressing occurrence questions are inadequate: 1) numbers of VOC(s) of interest frequently analyzed for; 2) information on the reporting level for each VOC; and 3) ancillary information about each sampled well. Therefore, the three most important enhancements to VOC data collected in Federal, State, and local nonpoint-source monitoring programs for use in a national assessment of VOC occurrence in drinking water would be an expanded VOC analyte list, recording the reporting level for each analyte for every analysis, and recording key ancillary information about each well.

In addition to the limitations just cited, another possible limiting feature of the retrospective data for use in augmenting NAWQA data is the relatively high laboratory reporting levels used by many laboratories for analyses of VOCs in water samples, as compared to the reporting level commonly used by NAWQA. For occurrence assessment at the assessment level of 0.2 m g/L commonly used by NAWQA, only a small percentage of the retrospective data is suitable to augment NAWQA data, even at the national scale. However, further analysis would be useful in evaluating the relative benefit of lower reporting levels in explaining the occurrence of VOCs in ground water.

Enhancing nonpoint-source ground-water monitoring programs for VOCs would enable the data to be used collectively to answer a range of simple to complex questions about the occurrence of VOCs in ground water at national and regional scales. Enhancements, such as measuring a larger number of VOCs at a reporting level that is less than often used today and recording characteristics about each well and aquifer sampled, would:

1. Result in a more complete assessment of the occurrence of VOCs in ground water nationwide by documenting the absence as well as the presence of VOCs at a low reporting level.
2. Explain possible reasons for occurrence and nonoccurrence.
3. Likely provide an early warning about an unanticipated regional or national occurrence of a VOC.
4. Provide data in support of the Safe Drinking Water Act on the occurrence of currently unregulated VOCs that might be regulated in the future.
5. Provide VOC data that might be needed in the future to address unanticipated questions of regional or national concern as they arise.

NAWQA personnel, in cooperation with State and local agencies, compiled most of the retrospective data summarized in this paper during 1995. The cooperation of these agencies is greatly appreciated.