Environ. Sci. Technol. 2008, 42, 1465–1471
Perchlorate in Groundwater: A Synoptic Survey of “Pristine” Sites in the Coterminous United States DAVID R. PARKER,* ANGELIA L. SEYFFERTH, AND BRANDI KIEL REESE Soil and Water Sciences Section, Department of Environmental Sciences, University of California, Riverside, California 92521
Received August 31, 2007. Revised manuscript received November 14, 2007. Accepted December 10, 2007.
Perchlorate is widely used as an oxidant in solid rocket propellants and energetic applications, and it has frequently been detected in groundwaters at concentrations relevant to human health. The possibility of naturally occurring perchlorate has only recently received significant attention. Relying primarily on domestic, agricultural, and recreational wells, we utilized a network of volunteers to help collect 326 groundwater samples from across the coterminous United States. Care was taken to avoid known, USEPA-documented sites of perchlorate use or release, as well as perchlorate contamination due to disinfection using hypochlorite. Using IC-ESI-MS and a Cl18O4internal standard, we achieved a method detection limit (MDL) of 40 ng/L perchlorate and a minimum reporting level (MRL) of 120 ng/L. Of the 326 samples, 147 (45%) were below the MDL, while 42 (13%) were between the MDL and the MRL. Of the 137 samples that could be quantified, most (109) contained 10000 ng/L) previously reported for the west-central Texas area appear to be anomalous. Perchlorate concentrations were positively correlated with nitrate levels (P < 0.001) but not with chloride concentrations. Opportunities exist for follow-up studies of perchlorate’s origins using isotope forensics and for further elucidation of the role of atmospheric processes in the formation or transport of perchlorate.
Introduction Perchlorate (ClO4-) is an emerging environmental pollutant that has led to widespread water contamination in the United States, especially in the Southwest (1). The ammonium and potassium salts of perchlorate are used in solid rocket propellants as well as in munitions, explosives, pyrotechnics, road flares, automobile air-bag systems, and an assortment of other commercial processes and products (2, 3). Large volumes of perchlorate-containing wastes have been locally released into the environment, and the solubility of the salts and nonreactivity of ClO4- lead to a highly mobile contaminant that readily migrates to ground- and surface waters. Perchlorate can competitively inhibit normal iodide uptake by the thyroid gland (4), and it has the potential to * Corresponding author phone: (951) 827-5126; fax: (951) 827 3993; e-mail:
[email protected]. 10.1021/es7021957 CCC: $40.75
Published on Web 01/25/2008
2008 American Chemical Society
cause hypothyroidism in humans with consequent adverse effects on normal growth and development. Perchlorate exposure could be important in iodine-deficient populations, as well as in pregnant women, newborns, and young children. The United States National Research Council recently recommended a maximum daily intake for perchlorate of 0.7 µg per kg body weight per day (5), a value subsequently adopted by the USEPA. This reference dose has been translated into an interim drinking-water equivalent standard of 25 µg/L perchlorate for a 70 kg adult male whose sole perchlorate exposure is from drinking water, but controversy has arisen as to whether this limit would adequately protect the most sensitive populations (e.g., developing fetuses and infants with low dietary iodine) (6). In March of 2004, the state of California issued a public health goal of 6 µg/L based on slightly different risk assessment using the Greer et al. (4) data set (7). At least seven other states have provisional or action-level drinking water limits ranging from 2 to 18 µg/L (6, 8). Perchlorate has been detected in retail milk and breastmilk samples (9, 10), in fresh greens such as lettuce (9, 11), and other produce and beverages (9, 12), indicating that foodchain transfer of perchlorate may be relatively commonplace (6), a conclusion supported by its ubiquitous appearance in human urine samples in the United States (13). Since the 1950s, there have been especially large amounts of perchlorate released in the western United States, often in association with defense-related activities (1). In one of the first published groundwater surveys using ion chromatography coupled with suppressed conductivity detection [i.e., USEPA Method 314.0 (14) or its equivalent], some 18 out of 367 (4.9%) wells tested positively for perchlorate with a method detection limit (MDL) of 4 µg/L (15). However, only nine of the 18 sites could be confirmed with subsequent analyses; all nine were located in California or New Mexico. A more comprehensive survey is underway based on the USEPA’s Unregulated Contaminants Monitoring Rule (UCMR) for municipal water treatment facilities, also using Method 314.0 and an MDL of 4 µg/L. As of late 2004, some 80% of the 1666 groundwater sources had complete data sets; of these, 73 (4.4%) had detectable perchlorate, but about onehalf of these were not confirmable with subsequent sampling and analysis (8). Detections were especially common in the northeastern and southwestern coastal states, but it is important to note that this likely reflects the spatial sampling density, which was driven largely by population density rather than a systematic geographical sampling scheme. With the recent, significant advances in detection by mass spectrometry, chromatographic methods are now capable of reliably quantifying sub-ppb levels of perchlorate in drinking water (16, 17). For example, Snyder et al. (18), analyzed 27 unchlorinated surface waters, groundwaters, and commercially bottled waters using LC-MS-MS; some 59% contained perchlorate above the minimum reporting level of 50 ng/L. In contrast, an earlier survey of bottled waters reported no perchlorate contamination (19), but the MDL was just 6 µg/L using an early IC-ESI-MS method. Thus, all indications are that more sensitive analytical methods will lead to much more widespread detection of perchlorate in natural waters, both surface- and groundwaters. The natural occurrence of perchlorate is only poorly understood. The nitrate-rich caliche deposits of the Atacama Desert in Chile contain substantial (g per kg) levels of perchlorate (20), and historical agricultural use of Chilean nitrate fertilizers is a possible source of groundwater contamination in some areas of the United States (6). In a ninecounty area in west-central Texas, unusually high concenVOL. 42, NO. 5, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1465
trations of perchlorate, up to 59 µg/L, have been found in groundwaters with no apparent anthropogenic source (21). In a companion study, Dasgupta et al. (22) demonstrated significant formation of perchlorate from precursor chloride under a range of laboratory conditions designed to simulate natural processes, especially those that might occur in the atmosphere. Although criticizable on realism grounds (e.g., imposed ozone concentrations of 900–1300 µL/L), this paper has provoked considerable interest in natural formation, and the possible formation due to high-energy electrical discharge during lightning storms is particularly intriguing. Additional indirect evidence for nonanthropogenic sources includes the presence of generally low levels of perchlorate in several aquifers in Texas and New Mexico where the groundwater is known to be prehistoric (based on 3H, 14C, or CFC signatures) and for which no plausible anthropogenic source could account for the estimated masses of perchlorate contained in the aquifers (23, 24). Recent advances in stable isotope techniques have begun to allow the differentiation of anthropogenic from naturally occurring perchlorate (25–27). Using both 18O/16O and 17O/ 16O ratios, Bao and Gu (25) showed that perchlorate from Atacama Desert soils had isotopic signatures distinctly different from commercially prepared samples and that these differences were consistent with atmospheric formation of the Atacama perchlorate. Subsequent work has also utilized 37 Cl/35Cl fractionation and has confirmed the unique origin of the Atacama perchlorates (26). However, two samples from the west Texas study (21) with perchlorate concentrations of 17 and 140 µg/L had isotopic signatures that were intermediate between commercial and Atacama samples, and the exact origin of perchlorate in this region remains unresolved. More data are needed to better understand the natural occurrence of perchlorate in the coterminous United States, and our goals here were twofold. First, we wished to assess the overall occurrence of perchlorate by using a much more sensitive analytical method (MDL ∼ 40 ng/L) in conjunction with a geographically driven, synoptic survey of groundwaters from across all 48 states. Second, we wished to ascertain whether there were any regional trends in the “natural” occurrence of perchlorate that would then help guide more mechanistic studies into, for example, formation in the atmosphere. Our survey specifically focused on rural, remote, and relatively pristine sites in an attempt to minimize any contributions from anthropogenic sources, including chlorination of drinking water.
Materials and Methods Reagents. All reagents were ACS certified purity or better, and solutions were prepared using type I deionized water. Perchlorate standards ranging from 0.030 to 10 µg/L were prepared by dilution of a liquid 1000 mg/L perchlorate standard solution (SPEX CertiPrep, Metuchen, NJ). Perchlorate spikes were prepared from the same standard. An internal standard (ISTD) was used consisting of Cl18O4- (99+% isotopic purity) with m/z values of 107 and 109 in the same 3:1 ratio as the natural abundance of 35Cl and 37Cl (Dionex, Sunnyvale, CA). Analytical Details. All perchlorate analyses were conducted using a Dionex (Sunnyvale, CA) ion chromatograph with electrospray-ionization-mass-spectrometry detection (IC-ESI-MS) and methods very similar to published methods (17, 28). Briefly, chromatographic separation was performed using a DX500 ion chromatograph equipped with an AS50 autosampler and a GP50 pump run in the isocratic mode; the eluent was 45 mM NaOH at a flow rate of 0.3 mL/min, and an injection loop of 100 µL was used throughout. Background conductivity was maintained below 1 µS/cm with an ASRS Ultra II (2 mm) suppressor in the external water mode and an ATC-HC trap column. Separation was achieved 1466
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 5, 2008
using an IonPac AS16 (2 × 250 mm) analytical column equipped with an IonPac AG16 (2 × 50 mm) guard column. Flow from the IC was directed into a Finnigan Surveyor MSQ Plus single quadropole MS (Thermo Electron Corporation, Waltham, MA). The MS was equipped with an AXP-MS auxiliary pump, pumping 50/50 acetonitrile/water at a flow rate of 0.3 mL/ min. Using a switching valve, all of the IC flow was diverted to waste for the first 9 min of each analysis, after which the IC flow was combined with acetonitrile/water to achieve a total flow rate of 0.6 mL/min to the MS. The retention time of perchlorate was ca. 13 min under these conditions. Flow to the MS was nebulized through an ESI source using ultrahigh purity nitrogen gas at a pressure of 550 kPa. The electrospray capillary was held at 450 °C with a needle voltage of 3.5 kV, and the entrance cone was held at a voltage of 70 V. Negative ion monitoring of m/z 99((0.5), m/z 101((0.5), and m/z 107((0.5) (corresponding to 35Cl16O4-, 37Cl16O4-, and 35Cl18O -, respectively) was utilized with a dwell time of 0.30 s 4 per ion mass. Monitoring at m/z 99 was used for all quantification, and Cl18O4- (m/z 107) was used as the ISTD at a concentration of 1 µg/L in all blanks, standards, and samples to correct for any ion suppression (29). Identification of perchlorate in the unknowns was confirmed by retention times, as well as the m/z 99 to 101 ratio of 3:1. Chromeleon Version 6.6 (Dionex, Sunnyvale, CA) was used to control the instrumentation and to quantify perchlorate. The electrical conductivity of all samples was determined using a CDM 83 conductivity meter (Radiometer America, Inc., Westlake, OH). Concentrations of major anions (NO3-, SO42-, Cl-) were determined by IC using a Dionex AS11- HC column (2 × 250 mm) in conjunction with an AG11-HC guard column (2 × 50 mm) and electrical conductivity detection (30). Quality Assurance and Control. For perchlorate, blanks and spiked blanks were made up in synthetic high-carbonate matrix (17) containing the following, in mM: HCO3-, 1.6; SO42-, 1.0; and Cl-, 2.8; all as the sodium salt. Blanks were run every 20th sample and always resulted in a nondetect. Spiked matrix blanks at either 30 or 50 ng/L perchlorate (bracketing the typical method detection limit; see below and Results) were also run every 20th sample, as well as a spiked matrix blank at 250 ng/L. Every 10th sample was run in duplicate, and values were only accepted when bracketed by a pair of duplicates that both agreed to within 10%. Analytical Limits. A widely used method for calculating the method detection limit is described in EPA Method 314.0 (14) and was utilized here. Briefly, seven replicate injections of 250 ng/L perchlorate synthetic hard water were analyzed by IC-ESI-MS, and an MDL was calculated based on the Student’s t-value at the 99% confidence interval. The MDL is given by MDL ) ts
(1)
where t ) 3.14 for six degrees of freedom and s is the standard deviation of the measured perchlorate concentration in the seven replicates. We adopted a minimum reporting limit (MRL) 3-fold greater than the MDL (18). Sampling. Groundwater samples were obtained from 326 sites in the coterminous United States. The majority (72%) were taken by a network of volunteers who received a sampling kit, a mailer, and a questionnaire (see Supporting Information). The samples thus obtained were dominantly from domestic (single-family) wells, with the remainder coming from field research stations and an assortment of state and national parks or other recreational facilities; a small number (11) came from active springs. Water samples were obtained from tap or spigot after 60 s of flushing, and they were placed in 125 mL HDPE sample bottles. Samples
were preserved by adding either 5 drops toluene or, more commonly, 5 drops of 10% (v/v) H2O2. Upon receipt in our laboratory, samples were stored continuously at 4 °C until analysis. The remaining samples (28%) were obtained by our laboratory personnel during various excursions into the field. These samples were dominantly from wells that served very small public water supplies such as campgrounds, parks, and roadside rest areas, as well as a few residential wells. Samples were taken as above, except that most were preserved only through continuous refrigeration. For all sites, we obtained as much well information as possible including geographical coordinates, well age, and depth. A few homeowners and site managers were able to provide drilling logs. Care was taken to avoid known, documented sites of perchlorate use or release. Here, we relied on two federal databases, one from the USEPA (31, 32) and one from the U.S. General Accountability Office (33). These sites were catalogued using their geographical coordinates, and no samples were utilized that were within a 25 km radius of any documented site. In addition, the urban-rural nature of each site was characterized based on the county population density from the 2005 U.S. census (34). In most cases, we were also able to categorize the current (or at least recent) dominant land use based on aerial photographs available on the World Wide Web (e.g., from Google Earth). Finally, because hypochlorite salts are a potential source of perchlorate contamination (6), we avoided to the extent possible any wells known to have been recently disinfected using hypochlorite or that had an in-line chlorination system. Samples collected from public recreational sites were tested for free chlorine using the DPD method (35), typically on site using a field test kit (Hach Co., Loveland, CO, P/N 2060400) but in a few cases post facto in the laboratory (detection limit ∼0.05 mg/L as Cl2). Repeated analysis of several samples showed that the chlorine levels were stable for up to six months in our refrigerated samples, so that we were able to recheck any results not reliably obtained in the field. Similarly, samples from a few homeowners who had indicated recent chlorination were also checked. All samples that tested positive for both perchlorate and chlorine were excluded, but any samples containing no measurable perchlorate (1000 ng/L were most prevalent in the Southeast, the Desert Southwest, and the Great Plains regions (see Figure 3 and below). Striking regional trends in perchlorate occurrence were not apparent (Figures 2 and 3), but when tabulated by geographical province (Table 1), some patterns emerge that are at least suggestive. Detection or quantification of perchlorate was a relatively rare occurrence in the Northeast, the Midwest, and the Pacific Coast states (Table 1). Measurable levels of perchlorate were seemingly more prevalent in the Southeast, the Great Plains, the Intermountain West, and the Desert Southwest. Thus, the regional trends in our coarse-scale survey could be used to support the notion that formation and persistence are somewhat more prevalent in semiarid regions. For example, Plummer et al. (24) described how high annual ET values coupled with very low-level deposition in rainfall could lead to perchlorate concentrations approaching 4000 ng/L. However, many of our quantifiable values came from more humid regions such as the Southeast, so that the possibility VOL. 42, NO. 5, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1467
FIGURE 1. Cumulative distribution of perchlorate concentrations in the 137 groundwater samples that could be quantified by IC-ESI-MS.
FIGURE 2. Geographical distribution of 147 groundwater samples that contained perchlorate levels below the method detection limit of 40 ng/L, along with 42 sites where perchlorate was detected but could not be reliably quantified. of relatively high localized inputs during thunderstorms (see below) remains an alternative explanation. We note in passing that we found no perchlorate concentrations approaching the highest values found in more detailed studies of west Texas and eastern New Mexico (21, 23), a region that we specifically sought to avoid in our synoptic survey. Land Use and Likely Sources of Perchlorate. We could discern no patterns in the distribution of perchlorate concentrations as a function of current land use. The 28 samples containing >1000 ng/L of perchlorate help illustrate this point, as they seemed to fall within two very broad groupings with respect to possible sources. Eleven of these samples came from the southeastern United States (GA, FL, NC, and SC), and at least eight of the wells were located on or in close proximity to active agricultural land. The remaining three came from rural residential land mixed with forests, but historical agricultural activity nearby cannot be ruled 1468
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 5, 2008
out. Another seven of these “high” samples came from Great Plains (KS, NE) or Midwestern (IA, WI) sites where intensive agricultural activity is presently ongoing. For all 18 of these samples, historical use of Chilean nitrate fertilizer remains a plausible, if unproven, explanation for the elevated perchlorate levels in the sampled groundwaters. In contrast, this explanation seems far less likely for the remaining 10 samples containing >1000 ng/L that were distributed as follows: six in the Desert Southwest (AZ, NM, TX), two in the Intermountain West (CO and NV), and two in California. Of these, only the single Colorado sample was in any proximity to a row-cropping area. The remaining nine all came from remote areas of native rangeland or mountain vegetation. In these cases, an agricultural source of perchlorate is not plausible, and contributions from natural sources seem more likely.
FIGURE 3. Geographical distribution of 137 groundwater samples that contained perchlorate levels above the minimum reporting level of 120 ng/L. Bubble size indicates the relative concentration of perchlorate as indicated in the legend.
FIGURE 4. Correlations between perchlorate and (a) chloride and (b) nitrate in 326 groundwater samples from the coterminous United States. Note the logarithmic (base 10) scales. Perchlorate values below the detection limit of 40 ng/L were assigned a nominal value of 20 ng/L. Similarly, when nitrate was below detection (∼0.0003 mmol/L), it was set to 0.0002 mmol/L. Correlations between Perchlorate and Major Anions. In contrast to prior studies (21, 23, 24), we found no correlation between perchlorate and chloride concentrations in groundwater (P ) 0.663, Figure 4a). With our data set, a similarly nonsignificant correlation between perchlorate and sulfate was observed (P ) 0.423; data not shown). But, a highly significant (P < 0.001), positive correlation between
perchlorate and nitrate was observed, also in contrast to some earlier studies of more limited geographic extent (21, 23). In our view, the positive correlation between perchlorate and nitrate is not necessarily unexpected. Thunderstorm activity remains the most plausible scenario for atmospheric formation of perchlorate (22), as there is little evidence for photochemical formation under realistic, earth-surface conVOL. 42, NO. 5, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1469
TABLE 2. Perchlorate in Selected Rainfall Events, 2005–06 site
date
county
state
elevation, m
storm type
[ClO4-], ng/L
[Cl-], mmo/L
[NO3-], mmol/L
1 1 2 3 4 5 6 7 7 7 7 8 9 9
12/31/05 07/04/06 03/17/06 06/28/06 03/28/06 07/23/05 07/20/06 07/24/06 07/25/06 07/25/06 07/26/06 07/29/06 07/29/06 07/29/06
Mono Mono Riverside San Bernardino Cococino Montezuma Larimer Rio Grande Rio Grande Rio Grande Rio Grande Navajo Cococino Cococino
CA CA CA CA AZ CO CO CO CO CO CO AZ AZ AZ
2100 2100 400 1200 2200 2300 2400 2600 2600 2600 2600 1900 2200 2200
snow intense hailstorm thunderstorm thunderstorm thunderstorm intense thunderstorm intense thunderstorm afternoon shower thunderstorm evening shower thunderstorm thunderstorm intense thunderstorm intense thunderstorm
