Environ. Sci. Technol. 2006, 40, 501-508
Comparative Evaluation of the Fate of Disinfection Byproducts at Eight Aquifer Storage and Recovery Sites P A U L P A V E L I C , * ,† PETER J. DILLON,† AND BRENTON C. NICHOLSON‡ CSIRO Land & Water, PMB 2, Glen Osmond, SA 5064, and Australian Water Quality Centre, PMB 3, Salisbury, SA 5108, Australia
Despite the growth in aquifer storage and recovery (ASR) as a technique for the provision of potable water supplies, quantitative data on the fate of disinfection byproducts that may be present in the injected water remain rare. This study evaluates the data from eight ASR sites in Australia and the United States that cover a wide range of source water compositions, hydrogeological environments, and operating conditions. Rates of attenuation and formation of trihalomethanes (THMs) in groundwater were determined using analytical techniques that took dilution effects into account. Half-lives varied by more than 2 orders of magnitude (e.g., 120 days for total THMs) and were both compound- and site- specific. Chloroform was most persistent, and more highly brominated compounds tended to be less persistent, as has generally been found. For any particular THM compound, much of the variability could be explained by contrasts in geochemical conditions within the aquifer since microbial degradation is the primary mechanism for THM attenuation. As such, bounds on the half-life were defined according to the redox state of the groundwater. In situ formation of some THMs in the aquifer after injection was directly observed at a number of sites, and was predicted to have taken place at all sites. The variance in formation estimates was large between the different methods used. Formation may be more common than previously thought because of the low frequency of groundwater sampling after injection and concomitant attenuation and mixing.
Introduction Aquifer storage and recovery (ASR) refers to the practice of injecting excess water into aquifers and later recovering that stored water from the same well during periods of demand (1). Recognition of ASR as a viable management technique is, in part, reliant on knowledge of the sustainable treatment processes that occur within the target aquifer to overcome real or perceived hazards when contaminants are introduced into the subsurface. Over the past decade or two there has been an increasing emphasis on examining the fate of disinfection byproducts (DBPs) during ASR due to concerns regarding human health effects when recovered water is used for drinking (2-6). Many of the previous studies have largely * Corresponding author phone: +61 8 8303 8742; fax: +61 8 8303 8750; e-mail:
[email protected]. † CSIRO Land & Water. ‡ Australian Water Quality Centre. 10.1021/es050768p CCC: $33.50 Published on Web 12/08/2005
2006 American Chemical Society
been qualitative in nature, with the level of analysis extending no further than to determine whether attenuation does or does not occur. Quantitative approaches to evaluating DBPs in the field have been much rarer (7, 8). The major processes that govern the fate of DBPs in groundwater include attenuation, formation, and mixing with ambient groundwater (2, 6, 8). Attenuation processes include sorption and chemical and microbial degradation. Atmospheric losses due to volatilization are generally not of relevance since the aquifers targeted for ASR are often confined, as is the case in this study. The aim of this study was to quantify the fate of DBPs (principally trihalomethanes (THMs) but also haloacetic acids (HAAs) where data permitted) at multiple ASR sites across a range of hydrogeological settings by evaluating (i) the rates of THM attenuation and whether these could be related to the physical or geochemical characteristics of the subsurface environment, and (ii) the incidences of THM formation in groundwater and whether valid predictions of THM formation within the aquifer were feasible.
Description of Study Sites Eight ASR field sites from Australia and the United States were selected for this study: Bolivar, South Australia; Jandakot, Western Australia; Charleston, SC; East Bay, CA; Las Vegas, NV; Memphis, TN; Oak Creek, WI; Talbert Gap, CA. Table 1 provides an overview of the hydrogeologic and operational characteristics at each of the ASR sites. They cover a range of subsurface environments, with aquifer composition ranging from alluvial sands and gravels to fractured rock and limestone, and groundwaters ranging from cool and aerobic to warm and anaerobic. Operating conditions are also diverse, with volumes of water injected ranging from 198 1137 997 74 24 199 199 850/yri 248 0.60 194 41 176 383 614 50 16 157 157 470/yrh ASR/4 m OBS ASR ASR ASR ASR ASRf ASRf ASR ASR ASR 55 m OBS 370 m OBS 630 m OBS 850 m PROD storage/cycle 1 recovery/cycle 9 recovery/cycle 3 storage/cycle 1 storage/cycle 2 recovery/1995-96 recove ry/1996-97 recovery/cycle 3 recovery/cycle 1 recovery/cycle 3 storage/cycle 3 NA (dual-well system) reclaimed potable potable potable Bolivar Charleston East Bay Jandakot
ambient groundwater conditions aquifer type redox stated temp (°C) DBP type tracer monitored periodc (days) site
source water
ASR phase/cycle
well ID
inj vol (ML)
rec vol storage (ML) periodb (days)
TABLE 1. Summary of Hydrogeologic and Operational Characteristics for Each of the Eight Study Sitesa 502
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 2, 2006
TABLE 2. Mean Composition of Injected Watera site
[TTHM] [TOC] [TTHM](µg/L)/ [Br-] [Cl2]b temp (µg/L) (mg/L) [TOC] (mg/L) (mg/L) (mg/L) (°C)
Bolivar 145 18.0 Charleston 77.3 3.4 East Bay 57.6 Jandakot (cycle 2) 43.1 2.0 Las Vegas (1995-96) 49.1 Memphis 2 0.5 Oak Creek (cycle 1) 12.8 6.3 Talbert Gap 13.5 1.9 a A dash means not available. injection.
b
8.1 22.6 21.5 4.0 2.1 6.8
0.5 0.01 -
1.0 2.9 1.1 0.05 1.0 1.1 1.8
20 24 15 21 18 7 25
Residual at the ASR well prior to
of a conservative tracer. Next, attenuation rates were calculated from the rate of change in the mixing-corrected DBP concentration over time. The principle assumption in both approaches is that the end members of both conservative and reactive constituents are reflected by the mean measured concentrations. The uncertainty in attenuation rates, mainly due to spatial and temporal variability in mixing fractions, was estimated by calculating the upper and lower 95% confidence limits about the mean half-life. Formation was examined by an assessment of measured concentration changes and using an existing pipeline model.
Results and Discussion THM Data. The average concentrations of total trihalomethanes (TTHMs) and associated water quality parameters at each site are presented in Table 2. Variations in TTHM yield depend on the quantity and characteristics of organic matter, chlorine dose and contact time, pH, bromide ion concentration, and temperature (9-11). Higher THMs arise from higher precursor levels, temperatures, dosage rates, and reaction times. TTHM concentrations were highest in the source water at Bolivar (145 µg/L) due to the high TOC content (18 mg/L); however, Charleston and Jandakot yielded significantly higher TTHMs per unit of TOC. Yields for Talbert Gap, Memphis, and Oak Creek were lower, while yields for East Bay and Las Vegas are unknown as TOC data were unavailable. The relative concentrations of individual THM constituents are presented in Table 3. These constituents varied substantially, from East Bay, where chloroform (CHCl3) was dominant, to Bolivar, where chlorodibromomethane (CHClBr2) and bromoform (CHBr3) were dominant. The proportions of THM constituents largely depend on the bromide ion concentration and the chlorine dose. At East Bay, where chlorinated byproducts were dominant, the bromide ion concentration was only 0.01 mg/L, while at Bolivar, where brominated byproducts were more prevalent, the bromide ion concentration was higher at 0.5 mg/L (Table 2). Bromide data were not available at the other sites. Interannual variations in the THM composition at a site appeared to be influenced by the consistency of the water source. For instance, at the Jandakot site the THM composition varied between cycles as the water was sourced from a different dam in each cycle. In contrast, more stable conditions were observed at Las Vegas since the injectant is sourced from a vast reservoir (Lake Mead) with the capacity to provide for multiple years of supply (9). Mass Balance Analysis. Figure 1 presents the mass balance results in the form of a plot of the cumulative percent recovery of TTHMs vs injectant (i.e., the tracer proxy). The Talbert Gap site was excluded since the separate injection and recovery wells made input/output analysis problematic. For five of the seven cases less TTHMs were recovered overall than were injected (i.e., the data fall below the 1:1 line). Net
TABLE 3. THM Constituents in Injected Watera site
Nb
[TTHM] (µg/L)
Bolivar Charleston East Bay Jandakot (cycle 1) Jandakot (cycle 2) Las Vegas (1995-96) Las Vegas (1996-97) Memphis Oak Creek (cycle 1) Oak Creek (cycle 3) Talbert Gap
2 1 4 2 5 3 3 1 1 1 29c
145 ( 0 77.3 57.6 ( 6 48.9 ( 25 43.1 ( 6 49.1 ( 3 48.0 ( 6 2 12.8 19.8 13.5 ( 13
a
Means and standard deviations reported.
b
[CHCl3] (µg/L) 30.5 ( 3 62 54.3 ( 4 4.5 ( 1 1.0 ( 1 15.0 ( 2 14.7 ( 3 0.5 5.4 9.4 8.5 ( 10
[CHCl2Br] (µg/L) 10.5 ( 3 13 3(1 12.8 ( 7 4.5 ( 1 17.0 ( 1 17.3 ( 2 0.6 4.8 6.5 2.9 ( 2
[CHClBr2] (µg/L)
[CHBr3] (µg/L)
47 ( 1 1.3 0.2 ( 0.2 24.5 ( 13 17.0 ( 3 15.0 ( 2 14.3 ( 2 0.7 2.4 3.5 1.3 ( 2
57 ( 1 1 0.1 ( 0.03 7.1 ( 4 20.6 ( 5 2.1 ( 1 1.6 ( 0.1 0.2 0.2 0.4 0.8 ( 1
Number of samples. c 1991-2000 data.
TABLE 4. DBP Attenuation Rate Data, Expressed as Half-Lives, for the Eight Sitesa site Bolivar Charleston East Bay Jandakot Las Vegasc Memphis Oak Creek Talbert Gapb
ASR phase/cycle storage/cycle 1
DBP half-lives (days)
well ID
ASR 4 m OBS recovery/cycle 9 ASR recovery/cycle 3 ASR storage/cycle 1 ASR storage/cycle 2 ASR recovery/1995-96 ASR (W029) recovery/1996-97 ASR (W029) recovery/cycle 3 ASR recovery/cycle 1 ASR recovery/cycle 3 ASR storage/cycle 3 55 m OBS NA 370 m OBS NA 630 m OBS NA 850 m PROD
TTHMs
CHCl3
CHCl2Br
CHClBr2
CHBr3
14 [12-17] 40 [36-49] 120 >130 79 [50-182] 10 [6-29] 16 [11-27] 160 [110-270] 70 100 130 56 [41-90] 18 [5-c] 32 [16-c] 340 [180-1400] 80 110 130 51 [36-92] 4 [3-6] 5 [5-6] 40 [35-47 ] Memphis All THMs were removed during the storage phase prior to pumping at Bolivar and Jandakot. Only at Memphis and Las Vegas was the recovered percentage of THMs greater than that injected. The results for Memphis were distinctly different from those of the other sites, with almost 8 times more TTHMs recovered than injected. This was even more striking given that only ∼30% of the injected water was recovered, which was the lowest of all sites due to high aquifer dispersivity and regional drift of the injectant down-gradient
of the ASR well (12). At the Las Vegas site 60% more TTHMs were recovered than injected. These phenomena illustrate that TTHM concentrations in groundwater during the storage phase can exceed those during injection, even when reductions due to mixing with ambient groundwater are taken into account. There are two known previous studies where mass balances have been applied. At the Las Vegas site, Miller et al. (5) investigated TTHMs for two ASR wells and found that they behaved conservatively. The mass of TTHM recovered was similar to that injected, with no net removal occurring beyond that through dilution at either well over the 97 day period of pumping. The results presented in this study, for a different well, also indicate near-conservative behavior, but unlike those of Miller et al. (5) also indicate formation within the aquifer. At the Lancaster site in California, Fram et al. (6) showed that, after 132% of the injected volume had been withdrawn, approximately 60% of both the injectant and TTHMs were recovered, and that, after 250% of the injected volume was recovered, the proportion of TTHM recovered had only increased to 72%. The relative removal ranking for the different sites given above is a function of the residence time of the injectant and the potential for in situ formation. Differences in operational schedules make consistent comparisons difficult to achieve from an input/output analysis and require the determination of attenuation rates which take these factors into account. Quantification of DBP Attenuation Rates. Table 4 presents calculated attenuation rate data, expressed as halflives, for total and individual THM constituents at all sites, and total HAAs at two sites. Figure 2 gives an illustration of VOL. 40, NO. 2, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
503
FIGURE 2. THM and tracer changes following the injection phase at four sites. the data by showing THMs and conservative tracer data for four sites as a function of the time of storage and/or recovery commencing from the end of the injection phase. From Table 4, the approximate order of TTHM half-lives was as follows:
Charleston < Jandakot = Bolivar = Oak Creek < Talbert Gap = Memphis < Las Vegas = East Bay Differences in the proportion of chlorinated to brominated compounds between field sites can confound such comparisons. For instance, while East Bay and Las Vegas have high half-lives, Las Vegas has a greater proportion of more easily degraded brominated compounds. Perhaps the most pronounced feature of the data was the large spread in THM attenuation rates between sites. Three orders of magnitude difference occurred between the smallest half-lives at Charleston (120 day). In many cases, the order of removal was observed to be CHBr3 > CHClBr2 > CHCl2Br > CHCl3. This is consistent with previous findings (see, e.g., refs 2 and 13). The only other known field study to report rates of attenuation of DBPs is by Roberts et al. (7). Half-lives of 23 days were reported for each THM compound and are within the range calculated in this study. The injection of aerated waters into anaerobic groundwater may induce various geochemical reactions that include mineral precipitation and dissolution, cation exchange, and redox reactions, which in some cases are mediated by microorganisms. Often the subsurface environment is geochemically as well as hydraulically heterogeneous. Injected water may contain nutrients such as biodegradable organic matter that acts as a reducing agent in redox reactions. Redox gradients are dynamic, and changes can occur during the storage and recovery phases of the ASR cycle. Redox status influences the activity of indigenous microbial populations, which, in turn, influence the persistence of DBPs. THMs and HAAs can potentially be degraded by chemical hydrolysis or by other abiotic processes; however, where this occurs, half-lives are extremely long (10). Microbial processes are responsible for any substantial degradation that occurs (13, 14). Previous studies have highlighted the dependence of redox conditions within the aquifer on DBP attenuation 504
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 2, 2006
(6, 9, 13). Under aerobic conditions no degradation occurs, while the most effective removal occurs under methanogenic conditions (9). Therefore, it was not surprising that rates of attenuation were too low to measure at Las Vegas where the aquifer is aerobic, which is supported by previous studies at the site (5). Interestingly, measurable attenuation still occurred at the Oak Creek and Memphis sites where the aquifer is also aerobic. Although aerobic degradation of THMs is in theory possible, this will depend on the particular THM, the occurrence of a suitable aerobic bacterium, and the level of primary substrate available (9). A laboratory microcosm study by Landmeyer et al. (14) for Las Vegas showed that CHCl3 was not degraded by aquifer bacteria under aerobic conditions. This was also demonstrated by Fram et al. (6) for CHCl3 in addition to CHBr3. Consequently, the most likely explanation for the attenuation observed is the occurrence of microniches within the heterogeneous aquifer that are of a sufficiently reduced state to facilitate degradation. Support for this argument was found in the work of Thomas et al. (3), again for Las Vegas, where bacteria capable of reducing sulfate and iron were identified in groundwater samples in the vicinity of some ASR wells. Singer et al. (2) noted the occurrence of denitrification in the presence of dissolved oxygen, which was explained as the result of either microniches within the aquifer or nonrepresentative DO data at the Kerville ASR site, Texas. These data offer some evidence to show that attenuation rates can differ in different parts of the aquifer. At Bolivar and Oak Creek, where half-lives were also determined at observation wells over the same cycle, significant differences were observed. In both cases, attenuation was slower further out in the aquifer. The most striking difference occurred at Oak Creek, where the half-life was an order of magnitude greater during the storage phase than at the observation well situated 55 m from the ASR well for the case where only partial breakthrough of injectant had occurred. At Bolivar, where the observation well was substantially closer than at Oak Creek (4 m vs 55 m), and where >200 pore flushes had occurred in the 0-4 m zone and mixing effects were negligible over the monitored period, the half-lives at 4 m were typically between 2 and 5 times higher than those at the ASR well. During the storage phase redox conditions became progres-
FIGURE 3. Correlation between TTHM half-lives and redox state (95% probability limits given for half-lives; sites with the highest confidence in the interpreted redox state are underlined). sively more reducing until methanogenesis occurred in the vicinity of the ASR well, whereas nitrate reducing conditions were maintained at the 4 m observation well (8, 9). This is consistent with the view that the microbially active environment close to the ASR well, which becomes more reducing during storage periods, is more conducive to degradation of chloroform than that further out in the aquifer. Sorption is not considered to be a significant factor at these sites due to the low sorption coefficient for DBPs and low organic carbon content of the aquifers (9, 10). THM breakthrough data from Bolivar support this view by the absence of retardation relative to chloride ion (8). Considering the three cases where repeated cycles were analyzed, very similar half-lives were determined at two sites (at Las Vegas and Jandakot), while for Oak Creek longer halflives were determined in the latter cycle for CHCl3 (and hence TTHMs also), but the difference was less than a factor of 2. Attenuation rates may vary between cycles in response to changes in redox or thermal conditions in the aquifer. Relationship between Half-Lives and the Geochemical Environment. The key influences on DBP attenuation may be summarized as the (i) availability of suitable electron donors determined by redox conditions within the aquifer, (ii) availability of primary substrate for microbial cometabolism, and (iii) groundwater temperature. Simply put, warmer, highly reducing environments with BDOC enhance THM attenuation. An initial attempt to account for the differences in halflives focused on a single variable linear correlation analysis with respect to the temperature, TOC, dissolved oxygen, and redox potential of the groundwater during the monitored period when degradation was observed to occur (using the “Analysis ToolPak” in EXCEL). The results however were extremely poor (R2 < 0.10). Incorporating pairs of parameters (e.g., temperature and redox potential) offered little additional benefit. This suggests that none of the parameters tested could be directly related to attenuation, and therefore, either the association is highly nonlinear or other factors may be more important. An explanation was then sought by examining the geochemical environment within the zone occupied by the injected water mass during the period of monitoring in the groundwater that most precisely represents the redox conditions in the aquifer when attenuation rates were determined. This proved more successful. The most reduced redox state of the groundwater over the period of monitoring was defined in terms of one of five redox classes: aerobic, nitrate-reducing, iron-reducing, sulfate-reducing, or methanogenic. For each
case, the appropriate redox class was assigned or interpreted from published hydrochemical studies for each site (9), and where there were no such studies, through an evaluation of the available hydrochemical data. Through this undertaking, judgment was made on whether the redox class assigned to each site/well was of high or low reliability. Figure 3 demonstrates that there was a reasonable association between TTHM half-lives and the redox state. Although there was considerable variance within any particular class, the bounds differed between classes, allowing some generalizations to be drawn. Sites which were aerobic had half-lives in excess of 50 days, those that were anoxic (nitrate-reducing) had half-lives of less than 70 days (with one exception), and those that were anaerobic (sulfatereducing or methanogenic) had half-lives of less than 20 days. The only exception was the data from the Oak Creek observation well, with a redox state that was of low reliability. There are close parallels between the findings from this study and the results of the laboratory study by Bouwer and Wright (13). These authors provide strong evidence that degradation of THMs is dependent upon redox conditions, with increased degradation occurring under increasingly reduced conditions. McQuarrie and Carlson (15) also demonstrate that anoxic conditions accelerate THM attenuation in a laboratory simulation of ASR. Field Evidence and Modeling of DBP Formation. Continued formation of DBPs in aquifers immediately following injection is possible given our knowledge of DBP formation processes (9) and in light of the common practice of chlorinating the injectant at the wellhead. If the chlorine dose exceeds demand, then chlorine will not be totally consumed before injection occurs and subsequent sampling may reveal a DBP concentration increase in the aquifer. In their study on the fate of DBPs at ASR sites in the United States and United Kingdom, Singer et al. (2) observed increases in the concentrations of THMs and HAAs in groundwater during the earliest period of storage, followed by declines beyond that due to mixing. Fram et al. (6) also observed continued formation in an aerobic aquifer where little or no THM attenuation occurred. Models to describe formation have recently begun to emerge (see, e.g., refs 16-18). The model of Clark and Sivaganesan (16), developed for water distribution systems, was chosen for use here since it is based on a substantial set of water quality data (42 waters), and the key water quality parameters that are needed to derive the coefficients in the model were measured at many of the ASR study sites. The VOL. 40, NO. 2, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
505
FIGURE 4. Peak THM concentrations in the injectant and groundwater. parameters are chlorine dosage, pH, temperature, and TOC concentration. Formation takes the form of a growth curve that reaches a plateau once the residual oxidant (chlorine and/or bromine) in the source water is consumed. Attenuation (corrected for mixing with ambient groundwater) is characterized by exponential decay and persists until DBP concentrations are exhausted. The combined effect of these processes may give an initial peak where formation is dominant, followed by a gradual decay where attenuation becomes dominant. TTHM formation could be estimated from the available data using four methods: (a) the difference between the uncorrected concentration in the groundwater at the ASR well and the concentration in the injectant, (b) as for (a) with the groundwater concentration corrected for mixing, (c) as for (a) and (b) with groundwater also corrected for attenuation and projected back to the time of completion of formation, as determined by method d, (d) value calculated from the pipeline formation model of Clark and Sivaganesan (16). Method a offers a pragmatic method of estimating the concentrations of THMs formed in the aquifer from the net difference between the highest THM concentrations measured in the injectant prior to storage and the highest concentrations measured in the groundwater after injection had ceased. The results, presented in Figure 4, show clear evidence for increased TTHM concentrations in the aquifer at five sites: Bolivar (4 m well), Jandakot (cycle 1), Las Vegas, Memphis, and Oak Creek. The largest increase was for Memphis (>50 µg/L). Although the source water for Memphis contained sufficient residual chlorine and DOC to account for the highest level of formation observed (as will be shown), the Charleston site had significantly higher potential for formation, although none was observed largely as a result of concomitant attenuation and mixing. At the two sites where multiple cycles were considered, increases were observed during both cycles for Oak Creek and during only one cycle (1996-97) for Las Vegas. At the Talbert Gap site, the 1-5 506
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 2, 2006
year travel times to the nearest monitoring well at a distance of 370 m effectively eliminated the possibility of detecting THM formation. The extent of formation differed for each of the THM compounds. Contrary to the limited TTHM formation observed, CHCl3 formation occurred at 9 of the 12 datasets presented in Figure 4 for five of the eight sites. Often CHCl3 formation was insufficient to offset the collective decrease of other THMs, and hence, no TTHM increase was observed. Although this method provides a scan across all sites for all compounds, the findings on formation are limited due to competing processes. Figure 5 presents the measured TTHM data in recovered water as a function of time for the Jandakot site (cycle 1). The data have been corrected for mixing (maximum correction of 3 µg/L on day 42). It also shows the predicted formation curve using the Clark and Sivaganesan model (method d). Projecting the mixing-corrected TTHM data back from the time of first sampling (12 days after injection had ceased) to the time when formation was complete (∼1 day) allows a mixing- and attenuation-corrected estimate of formation by comparing the projected value with the average injected concentration (method c). Mixing effects over the projected period were negligible. This attenuation correction assumes a reliable estimate of the time for completion of formation and that the attenuation rate during the period before the first sample is the same as that measured afterward. By method a, 30 µg/L of formation is estimated to have occurred from the increase in concentration between the injectant and the groundwater. The mixing correction was negligible, but the attenuation correction (method c) was substantial, increasing the estimated initial formation from 30 to 70 µg/ L. The formation model (method d) suggested the lowest formation of 15 µg/L. The TTHM formation estimates for five sites, including the Jandakot site, are given in Table 5. The pipeline model (method d) predicted formation to have occurred at all five sites, with concentration increases ranging from 15 µg/L at
FIGURE 5. Example of three TTHM formation estimation methods for the Jandakot cycle 1 data.
TABLE 5. Estimated TTHM Formation from the Four Different Methods at Five Sites site
time to complete formationa (days)
time for first sample (days)
(a) uncorrected (µg/L)
(b) mixing corrected (µg/L)
(c) mixing and attenuation corrected (µg/L)
(d) formation model (µg/L)
Bolivar Charleston Jandakot (cycle 1) Memphis Oak Creek (cycle 1)
28 14 1 18 9
7 6 12 124 1
0 0 30 52 4
0 0 30 72 7
30 >200 70 360 10
79 158 15 44 36
a
Values reflect the time when 99% of the final concentration is first reached according to method d.
Jandakot to 158 µg/L at Charleston. Some uncertainty exists in applying the formation model to ASR, since the formation model, developed for pipeline systems, does not consider the organic carbon content in the aquifer, nor the bromide ion and ammonia concentrations in the injectant. TTHM formation estimated from the mixing- and attenuationcorrected method (method c) gave values ranging from 8 µg/L at Oak Creek to 360 µg/L at Memphis. The comparisons between methods c and d were often relatively poor, with results from method c higher than those from method d at three sites by up to ∼320 µg/L, and lower at two by up to ∼50 µg/L. Method c was sensitive to the effective time to complete formation when groundwater sampling had commenced before completion according to method d. Interestingly, three of the five sites (Bolivar, Charleston, and Jandakot) did not directly evidence formation (even with mixing corrections), although methods c and d suggest formation would have occurred prior to initial groundwater sampling. Formation will depend on the factors previously mentioned as well as kinetic effects. If the oxidant dose exceeds demand, then chlorine/bromine will not be totally consumed, DBP formation will be incomplete before injection occurs, and sampling may reveal an increase. Failure to measure formation within the aquifer need not necessarily imply that formation did not occur in the interval prior to sampling as removal processes can outweigh formation, as implied by Figure 5 and clearly demonstrated in Table 5. Continued formation of THMs in the aquifer would appear to be ubiquitous across the sites. The period immediately following the cessation of injection appears to be a major influence on THM concentrations in groundwater. Further data over this critical period are needed to refine the conceptual model and formation estimates.
Acknowledgments This work was made possible through the support of the American Water Works Association Research Foundation (Project No. 2618) and the Bolivar Reclaimed Water ASR
Research Project. We thank the following individuals for their contributions: Karen Barry and Joanne Vanderzalm (CSIRO Land and Water), Matt Petkewich (U.S. Geological Survey, Columbia, SC), Mike Tognolini and Hasan Abdullah (East Bay Municipal Utility District), Karen Johnston and Michael Martin (WA Water Corp.), Bill Quinn (Las Vegas Valley Water District), Fred Von Hofe (Memphis Light, Gas and Water Division), Tom Miller (CH2M HILL, St. Louis, MO), and Greg Woodside (Orange County Water District). Ray Correll (CSIRO Mathematical and Information Sciences) provided statistical advice. Peter Fox (Arizona State University) and Pieter Stuyfzand (Kiwa Water Research) provided feedback during the review of Project No. 2618. John Van Leeuwen and GuangGuo Ying of CSIRO Land and Water and the reviewers from Environmental Science & Technology provided helpful comments on the manuscript.
Literature Cited (1) Dillon, P. J. Future management of aquifer recharge. Hydrogeol. J. 2005, 13 (1), 313-316. (2) Singer, P. C.; Pyne, R. D. G.; Mallikarjun A. V. S.; Miller, C. T.; Mojonnier, C. Examining the impact of aquifer storage and recovery on DBPs. J.sAm. Water Works Assoc. 1993, 86, 85-94. (3) Thomas, J. M.; McKay, W. A..; Cole, E.; Landmeyer, J. E.; Bradley, P. M. The fate of haloacetic acids and trihalomethanes in an aquifer storage and recovery program Las Vegas, Nevada. Ground Water 2000, 38 (4), 605-614. (4) Mirecki, J. E.; Campbell, B. G.; Conlon, K. J.; Petkewich, M. D. Solute changes during aquifer storage recovery in a limestone/ clastic aquifer. Ground Water 1998, 36 (6), 394-403. (5) Miller, C. J.; Wilson, L. G.; Amy, G. L.; Brothers, K. Fate of organochlorine compounds during aquifer storage and recovery. Ground Water 1993, 31, 410-416. (6) Fram, M. S.; Bergamaschi, B. A.; Goodwin, K. D.; Fujii, R.; Clark J. F. Processes affecting the trihalomethane concentrations associated with the third injection, storage, and recovery test at Lancaster, Antelope Valley, California, March 1998 through April 1999; U.S. Geological Survey Water-Resources Investigations Report 03-4062; Sacramento, CA, 2003. VOL. 40, NO. 2, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
507
(7) Roberts, P. V.; Schreiner, J.; Hopkins, G. D. Field study of organic water quality changes during groundwater recharge in the Palo Alto Baylands. Water Res. 1982, 16, 1025-1035. (8) Pavelic, P.; Nicholson, B. C.; Dillon, P. J.; Barry, K. E. Fate of disinfection by-products in groundwater during aquifer storage and recovery with reclaimed water. J. Contam. Hydrol. 2005, 77, 351-373. (9) Dillon, P. J.; Toze, S. Water quality improvements during aquifer storage and recovery; AWWARF Project No. 2618, Report Order No. 91056F; Denver, CO, 2005. (10) Nicholson, B. C.; Dillon, P. J.; Pavelic, P. Fate of disinfection by-products during aquifer storage and recovery. In Management of Aquifer Recharge for Sustainability, Proceedings of the 4th International Symposium on Artificial Recharge (ISAR4), Adelaide, Australia, Sept 22-26, 2002; Dillon, P. J., Ed.; Swets & Zeitlinger: Lisse, The Netherlands, 2002; pp 155-160. (11) Symons, J. M.; Stevens, A. A.; Clark, R. M.; Geldreich, E. E.; Love, O. T.; DeMarco, J. Treatment techniques for controlling trihalomethanes in drinking water. J.sAm. Water Works Assoc. 1982. (12) Pavelic, P.; Dillon, P. J.; Simmons, C. T. Lumped parameter estimation of initial recovery efficiency during aquifer storage and recovery. In Management of Aquifer Recharge for Sustainability, Proceedings of the 4th International Symposium on Artificial Recharge (ISAR4), Adelaide, Australia, Sept 22-26, 2002; Dillon, P. J., Ed.; Swets & Zeitlinger: Lisse, The Netherlands, 2002; pp 285-290.
508
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 2, 2006
(13) Bouwer, E. J.; Wright, J. P. Transformations of trace halogenated aliphatics in anoxic biofilm columns. J. Contam. Hydrol. 1988, 2, 155-169. (14) Landmeyer, J. E.; Bradley, P. M.; Thomas, J. M. Biodegradation of disinfection byproducts as a potential removal process during aquifer storage recovery. J. Am. Water Resour. Assoc. 2000, 36, 861-867. (15) McQuarrie, J. P.; Carlson, K. Secondary benefits of aquifer storage and recovery: disinfection by-product control. J. Environ. Eng. 2003, 129 (5), 412-418. (16) Clark, R. M.; Sivaganesan, M. Predicting chlorine residuals and formation of TTHMs in drinking water. J. Environ. Eng. 1998, 124 (12), 1203-1210. (17) Westerhoff, P.; Debroux, J.; Amy, G. L.; Gatel, D.; Mary, V.; Cavard, J. Applying DBP models to full-scale plants. J. Am. Water Works Assoc. 2000, 92 (3), 89-102. (18) Gallard, H.; von Gunten, U. Chlorination of natural organic matter: kinetics of chlorination and of THM formation. Water Res. 2002, 36, 65-74.
Received for review April 21, 2005. Revised manuscript received October 18, 2005. Accepted October 18, 2005. ES050768P
