Environ. Sci. Technol. 2010, 44, 3919–3925
Quantitative Evaluation of Tracers for Quantification of Wastewater Contamination of Potable Water Sources G U Y G A S S E R , †,‡ M I C H A E L R O N A , † ANNA VOLOSHENKO,† RIMMA SHELKOV,† NELI TAL,§ IRENA PANKRATOV,‡ S A R A E L H A N A N Y , ‡ A N D O V A D I A L E V * ,† Casali Institute of Applied Chemistry, Institute of Chemistry, Edmond J. Safra Campus, The Hebrew University, Jerusalem, Israel, Israeli Water Authority, Hamasger street 14, P.O. Box 20365, Tel Aviv 61203, Israel, and Shafdan Unit, Mekorot Ltd., Rishon Lezion, Israel
Received March 1, 2010. Revised manuscript received April 13, 2010. Accepted April 14, 2010.
Quantitative criteria for selection of tracers for assessment of mixing of wastewater and pristine water are proposed and evaluated for leakage from a wastewater effluent recharge system to nearby pristine water wells and the dilution of the effluents in a reclamation well by pristine water from the surrounding aquifer. Two molecular tracers were compared: carbamazepine, an organic drug whose refractory behavior was evaluated on-site, and chloride, a widely used conservative tracer. The mixing ratios and the corresponding uncertainty levels in their calculation were evaluated using actual field data. Uncertainty level analysis illuminates the effects of the analytical errors in the determination of trace micropollutants on one hand and the high level of chloride in the background on the other. Uncertainty level calculations revealed that chloride is a somewhat better tracer for the estimation of the dilution of wastewater by flow from a pristine aquifer, whereas carbamazepine is a much better tracer for the calculation of wastewater contamination of nearby drinking water wells. Surprisingly, we show that even when carbamazepine degrades to a large and unknown extent, it can still be used to estimate accurately the probability that a site is contaminated by a wastewater stream.
Introduction Once waterborne diseases became associated with microorganisms, the quest for finding indicators of wastewater contamination of drinking water sources began. Frankland and co-workers (1) identified coliforms as observable indicators of fecal contamination, and later, fecal coliforms emerged as irreplaceable indicators of fecal contamination (2). With the introduction of improved analytical techniques for assessment of trace inorganic and organic pollutants, numerous refractory contaminants associated with wastewater were proposed as sewage markers. This development led to additional consequences. The fact that advanced * Corresponding author phone: 97226585558; fax: 97226586155; e-mail:
[email protected]. † The Hebrew University. ‡ Israeli Water Authority. § Mekorot Ltd. 10.1021/es100604c
2010 American Chemical Society
Published on Web 04/29/2010
wastewater treatments leave trace micropollutants in the treated effluents became common knowledge. The public and water suppliers became increasingly aware of potential risks associated with yet unknown refractory pollutants in drinking water sources. Anthropogenic indicators are no longer sought exclusively to warn against pathogen risks but are increasingly being used to warn against unintentional mixing of drinking water sources with treated effluents, regardless of any tangible health effect associated with such mixing. The subject of this research effort is the quantification of leakage of treated effluents from a wastewater recharge site to a surrounding active aquifer. Wastewater recharge is increasingly used for safe discharge of surplus water effluents. In arid countries, it is also part of water reuse schemes, providing an additional safety barrier before irrigation or recreation water uses or as a seasonal supplement for drying rivers. Since wastewater recharge is often conducted near or within active drinking water-supplying aquifers, there is a need to quantify the ratio of wastewater effluents to water from pristine sources in wells that are located in the vicinity of wastewater infiltration facilities. Mixing ratio estimates involve two distinctly different well categories: those for drinking water supply and those pumping predominantly recharged effluents diluted by the pristine aquifer. In the first type, only a few percent of effluents may leak to the well, and for the second type, effluents are the dominant source. Peripheral water wells are often installed around the recharge system to act as a hydraulic barrier and to prevent contamination of the aquifer. Pumping from these wells should be carefully controlled to optimize leakage prevention and misuse of the pristine source, and chemical tracers provide the best means of monitoring mixing ratios in these reclamation wells. In Israel, the need for both reliable quantification and an accurate way to estimate the uncertainty levels associated with the mixing estimates based on different tracers is especially important. More than 15% of the domestic wastewater is treated with a soil aquifer treatment (SAT), which is situated within the coastal plain aquifer; the latter supplies ∼25% of Israel’s annual water consumption. Dilution Tracers. Numerous tracers for estimating the degree of mixing have been proposed. Chloride is by far the most popular tracer (3-5) because domestic activities and water evaporation increase the chloride level in sewage by some 50-150 mg/L as compared to tap levels. Under the Shafdan SAT conditions, some of the domestic water sources originate from chloride-rich surface water which further increases the difference between the chloride concentration in the wastewater effluents and that in the nearby coastal plain aquifer. However, because of the abundance of chloride in all water sources and its high background concentration, other less conservative anions such as nitrate and sulfate have also been used, as well as ratios between different (conservative) tracer anions. Vengosh et al. (6) proposed a ratio between bromide or fluoride and chloride to quantify sewage-originated and pristine water mixing. The large variability of the isotopic signature of boron in different water sources and its high concentration in laundry detergents motivated their use as effluent markers in water sources (7). Other isotope ratios were also proposed to this end (8, 9). Although organic contaminants are frequently nonconservative and prone to some degradation, many were proposed as wastewater tracers. Originally, the search focused on aggregate attributes (e.g., UV and oxygen demand parameters), but more recently, interest has shifted to VOL. 44, NO. 10, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3919
refractory organic contaminants. Almost every refractory waste-specific contaminant was put forward as a marker for sewage contamination. In particular, X-ray contrast agents (10-12), human-specific antibiotics, and pharmaceuticals (13, 14) attracted attention, since they are biorefractory by design and are mostly specific for human consumption. Health care products (13, 14), human steroidal hormones (15, 16), caffeine and its metabolites (17, 18), and more recently even artificial sweeteners (19, 20) were all considered as wastewater markers. We recently conducted a comprehensive survey of the fate of organic compounds in the wastewater treatment of the Tel Aviv area. The survey which included more than 300 micropollutants revealed that CBZ is by far the most refractory compound. CBZ is a generic drug for the control of grand mal and psychomotor epilepsy, effective in the treatment of trigeminal neuralgia, and is used in bipolar depression. CBZ was proposed by numerous researchers as a sewage marker (21-23). Because of its low degradability in wastewater plants (24-26), concentrations of up to several micrograms per liter have been detected in surface water (24, 25, 27) and groundwater (28). However, to date there is no reported systematic effort to quantify the uncertainty levels associated with the use of CBZ, or any other indicator for that matter, for source tracking. Criteria for good wastewater tracers are still intuitive or based on qualitative argumentation. The objective of this article is to compare the prediction power of the best organic tracer for a specific test-case site, CBZ, with the currently most useful marker, chloride, and thus gain insight into the selection criteria of different tracers as a function of the mixing ratio and water properties. The uncertainty analysis used here is, however, generic and can be applied for the assessment of the uncertainty involved in calculation of other water mixing situations, including crosspiping of drinking water and graywater or wastewater effluents, leakage from wastewater pipes to streams and aquifers, and source tracking in general. Dilution Calculations and Error Estimates. We consider here a simple steady-state dilution model involving the mixing of a wastewater effluent stream containing a concentration [X]ef of the tracer and a stream that comes from the surrounding aquifer with some background concentration of the tracer [X]b to give a local concentration [X]i. The mixing ratio (MR), defined here as the fraction of effluent-originated water in the sampled well, is then given by eq 1. MR ≡
[X]i - [X]b effluents ) effluents + pristine water [X]ef - [X]b
(1)
A mixing ratio of 1 implies that the site contains pure wastewater effluents, and a ratio of 0 indicates pristine water. First-order Taylor series expansion of eq 1 gives after rearrangement the relative uncertainty in the dilution estimate (the systematic errors are delineated in the Supporting Information). ∆E ≡
{(
)
1 MR estimated error ∆[X]i ) MR [X]i - [X]b 2 -([X]ef - [X]i) ∆[X]b + ([X]ef - [X]b)([X]i - [X]b)
[
(
]
2
+
)}
1 ∆[X]ef [X]ef - [X]b
2
1/2
(2)
Analysis of the relative uncertainty, ∆E, reveals that it is comprised of three sources of uncertainty, ∆[X]ef, ∆[X]i, and ∆[X]b. Each of these terms is multiplied by an amplification factor. The amplification factor of ∆[X]i increases hyperbolically as [X]i approaches [X]b. ∆[X]ef is multiplied by an amplification factor that increases hyperbolically as [X]ef 3920
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 10, 2010
FIGURE 1. Study site, Soreq 1 percolation lagoon, and location of wells used to quantify carbamazepine degradation (names included). Also shown is the chloride level in wells that were used for chloride background evaluation (with underlined chloride levels) and in other wells (not underscored) that were not included in the background chloride estimate. For example, all the high-chloride level wells on the eastern side of the map were not included in the chloride background calculation. In addition, the figure shows CBZ levels (in squares below the locations of the wells) in wells that were used for CBZ background evaluation. Wells that were not included in any background calculation are denoted with empty symbols. Chloride level is indicated in milligrams per liter and carbmazepine level in nanograms per liter. Chloride data were taken between 2002 and 2009 and carbamazepine levels in 2007-2009. approaches [X]b. The first amplification factor multiplying ∆[X]i is always larger than the amplification factor of the second term. Hence, the second term can become dominant only in cases where the uncertainty level in the evaluation of the background concentration is much larger than the analytical uncertainty in the examination of [X]i. This is often the case since the evaluation of the background concentration is more complicated and involves a larger degree of uncertainty. The third amplification factor, multiplying ∆[X]ef, is always smaller than the first amplification factor, and they approach each other only when most of the water at the examined location originates from the percolation lagoons. Each of the terms in eq 1 involves some uncertainty level because of instrumental uncertainties, sampling variability, and time fluctuations, which can be seasonal or may include year-to-year fluctuations. The three error values, ∆[X]ef, ∆[X]i, and ∆[X]b, have somewhat different sources. The first two terms are based on pure analytical data, involving determination of the concentration of the probe in a specific, well-defined location. These uncertainty values can be reduced by averaging the concentration at the examined location over longer periods of time. The uncertainty in the calculation of the background concentration is not as welldefined and should be estimated on the basis of the concentration of the tracer in the vicinity of the examined site.
Experimental Section Study Sites. Our study concentrated on the Soreq 1 and Yavne 2 basins of the Shafdan SAT (Figure 1 and Figure S1 of the Supporting Information). The study area, by and large, consists of a permeable calcareous sand perched aquifer supported on a marine clay aquiclude and partially interrupted by a horizontal marine clay aquitard. The water table reaches its maximum height under the percolation basin (36 m below the bottom of the
TABLE 1. Retention Times, Distances from the Percolation Lagoons, and Concentrations of Carbamazepine (CBZ) and Chloride in Different Water Wells Surrounding the Percolation Lagoons chloride
CBZ
location
retention time (months)
distance (m)
concn (mg/L)
standard deviation (mg/L)
no. of observations
concn (ng/L)
standard deviation (ng/L)
no. of observations
effluents OW 282 OW 207 RE 213 RE 219 OW 62 DAN 22
0 1.5 12 13 21 0.5 8
0 0 300 250 650 0 200
288 285 290 283 300 307 311
24 24 7 18 23
12 7 2 7 9 1 1
1660 1492 1575 1545 1474 1475 1383
326 390 177 166 234 600 260
12 7 2 8 9 2 2
lagoon) and has a gradual height decrease (of ∼20 m/km) to a radial minimum located approximately below the outer ring of the recovery wells. The average hydraulic loading in the percolation lagoons is 0.30 m/day (29). Target Water Wells. Several observation and reclamation wells in the Soreq 1 and Yavne 2 recharge sites were studied in estimating carbamazepine degradation in this particular aquifer (Table 1). In addition, several water wells in the vicinity of the Soreq 1 percolation basin of the Shafdan were selected to demonstrate the application of eq 1 and the power of carbamazepine tracers (Figure 1). Dan 23 is located some 1000 m from the percolation lagoon and primarily pumps effluents diluted to a small but unknown extent by the coastal plain aquifer. The dilution by freshwater is intentional as it helps create a hydraulic barrier against leakage of the Shafdan recharged water to the nearby aquifer. Three additional wells were chosen as test cases. Rishon 5, a water well located 1700 m from the Soreq 1 site, supplied drinking water until 2005. It ceased due to a salination process, which raised concerns regarding contamination by wastewater effluents. Yavetz, located 2000 m from the periphery of the Soreq 1 site, was also included in this study since its carbamazepine level was slightly higher than that of the nearby aquifer. Silicate, a water well situated some 2000 m from the percolation lagoon, was also examined as a borderline case, where a conclusive decision regarding contamination by effluents is currently impossible. Analysis. The analytical methods are described in the Supporting Information. Briefly, anions were analyzed by EPA method 300.1 and CBZ trace levels by EPA method 1694. Selection of the Target Dilution Tracers for This Study. CBZ is present at a fairly constant level of 1-2 ng/L in most wastewater treatment plant effluents. The level of CBZ is somewhat retarded in the subsoil (30); however, it has moderate mobility based on its reported Kd, which is in the range of 0.08-1.8 for different soils, and its log octanolwater partition coefficient was reported to be 2.67 at pH 7 (30). Our CBZ detection limit is
