- Email: [email protected]
Tracing treated wastewater in an inland catchment using anthropogenic gadolinium
Chemosphere 80 (2010) 794–799
Contents lists available at ScienceDirect
Chemosphere journal homepage: www.elsevier.com/locate/chemosphere
Tracing treated wastewater in an inland catchment using anthropogenic gadolinium Michael Glen Lawrence *, David Guimerà Bariel Advanced Water Management Centre, The University of Queensland, St. Lucia, Qld 4072, Australia
a r t i c l e
i n f o
Article history: Received 1 December 2009 Received in revised form 30 April 2010 Accepted 3 May 2010
Keywords: Australia Risk-assessment Micropollutants Drinking water Toowoomba Queensland
a b s t r a c t The discharge of treated wastewater into natural water bodies occurs worldwide; if drinking water is then extracted downstream, there is potential for micropollutants that are not fully mineralized in the wastewater treatment process to enter municipal drinking water. In Australia, drinking water treatment is typically a mixture of basic technologies such as flocculation and slow sand filtration; technologies that are not specifically designed to remove micropollutants. However, there is little awareness in Australia of the potential risk that upstream wastewater discharges may impart to the security and quality of downstream drinking water supplies. We apply a direct inductively coupled plasma mass spectrometry technique to determine the discharge of anthropogenic gadolinium from a wastewater treatment plant in Toowoomba, Queensland, Australia, that discharges into the small (147 km2) catchment of Gowrie Creek. We then continue to measure the concentrations of this wastewater tracer as Gowrie Creek flows downstream into the Condomine River, and to a community 100 km away where drinking water is extracted. Using this tracer, we demonstrate that the community has a detectable wastewater contribution within their surface drinking water supply. Ó 2010 Elsevier Ltd. All rights reserved.
1. Introduction Vast areas of eastern Australia are currently enduring, or have recently emerged from, the worst drought on record (Queensland Government, 2007). In this context, surface water supplies of more than half of all Australians were dramatically reduced; for example, in Sydney (population 4.3 million), the major water supply dam dropped to 32.5%, in Melbourne (population 3.8 million) the water supply dropped to 25% capacity, and in Brisbane (1.9 million), the water supply dropped to 16.9% capacity (New South Wales Government, 2007; Melbourne Water, 2009; SEQ Water, 2009). Severe water restrictions were applied, and community awareness of issues surrounding water security in Australia has become intense. Alternate water supplies have been investigated and desalination infrastructure is being built around the country. Yet water recycling, one of the more obvious water supply options, despite being heavily regulated, is not a widely acceptable solution for augmenting drinking water supplies. However, it is little recognized that in many inland Australian catchments, baseline water flows are often supported by the upstream discharge of wastewater. For example, previous research (Hamilton and Greenfield, 1991) based on hydraulic calculations of mean river flow and wastewater treatment plant discharge has suggested that the average wastewater contribution to the drinking water supply at Dalby, on the Condomine River, was 8.3%. However, this hypothesis has * Corresponding author. Tel.: +617 3346 3228; fax: +617 3365 4726. E-mail address: [email protected] (M.G. Lawrence). 0045-6535/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.chemosphere.2010.05.001
never been tested in practice. We present a case study where the very sensitive anthropogenic Gd tracer is used to determine the percentage of wastewater that is detectable at Dalby, approximately 100 km downstream from the wastewater discharge at Toowoomba. While this is a specific example, the applicability of this tracer is not limited to this case alone. 2. Aims of paper The primary aim of this manuscript is to quantify the extent to which wastewater contributes to downstream water supplies in regional Australia. Towards this aim we have targeted gadolinium from MRI contrast agents, which has been demonstrated to be present in wastewater from areas with advanced medical imaging (Bau and Dulski, 1996; Kümmerer and Helmers, 2000; Möller et al., 2000; Knappe et al., 2005; Rabiet et al., 2005; Verplanck et al., 2005; Bau et al., 2006; Kulaksiz and Bau, 2007; Lawrence et al., 2009; Rabiet et al., 2009). Gadolinium can be quantified directly in natural water and wastewater effluent by ICP-MS (Lawrence et al., 2006b,c, 2009); in contrast, other organic micropollutants such as pharmaceuticals and personal care products typically require >100 fold pre-concentration in natural waters prior to quantification by LC–MS/MS or GCMS. The excess Gd from MRI contrast agents is defined as GdAnth where:
½GdAnth ¼ ½GdMeasured ð1:1 ½GdNatural Þ
ð1Þ
where the factor 1.1 accounts for the small natural Gd anomaly usually observed in South East Queensland (Lawrence et al., 2006b). In
M.G. Lawrence, D.G. Bariel / Chemosphere 80 (2010) 794–799
this case the natural Gd concentration is determined from a third order polynomial fit of the non-anomalous rare earth elements (REE’s) (Möller et al., 2003). Given the current evidence (Kulaksiz and Bau, 2007) that GdAnth is refractory during estuarine mixing, and qualitatively stable in groundwater infiltration (Möller et al., 2000), there appears to be potential to treat the tracer as conservative, and thus exploit GdAnth as a tracer. As there are at least 2 MRI instruments located in Toowoomba, it was assumed that this tracer should be present. The first step towards determining whether GdAnth can be utilised as a wastewater tracer is to define the point source discharge. Previous research has suggested that there is a weekly cycle in the discharge of GdAnth in wastewater effluent. Therefore, a sampling strategy was implemented to determine the long-term (5-month) discharge of GdAnth from the Wetalla WWTP. Concurrently, three sampling campaigns were undertaken in the Gowrie Creek/Condomine River catchment to determine whether GdAnth was detectable, and if possible, estimate the amount of wastewater present. 3. Sample locations Wetalla WWTP is a biological nutrient removal plant with a design capacity of 36 ML d1 but typically operating at about half this capacity. The wastewater treatment plant services a population of 100 000 people and discharges its treated effluent into Gowrie Creek, which flows downstream where, in approximately 100 km, water is extracted for drinking water in Dalby. The flow from the WWTP is suspected to be the predominant water supply in this creek. During the sampling campaigns this was visually apparent, with little flow in Gowrie Creek above the WWTP with a substantial increase in flow below the discharge point. As MRI contrast agents are expected to be applied to only a limited number of people within Toowoomba each day, it can be expected that the influent of the wastewater treatment plant would have sporadic spikes, and inhomogeneous loads of GdAnth. However, given the large buffering capacity of the WWTP, these concentration spikes should be mixed during treatment, and the effluent should be more homogenous. Thus, it was considered that the existing effluent sampling conducted at the treatment plant, notably daily volume proportional samples of the effluent immediately prior to discharge into Gowrie Creek (an ephemeral creek in a small, 147 km2 catchment) should provide representative samples. Daily 24-h composite samples of the effluent from the Wetalla wastewater treatment plant were collected in a refrigerated (4 °C) auto-sampler triggered to collect a constant volume sub-sample of water every 100 kL. A 10 mL sub-sample of this water was then filtered with a (Millipore Millex GV 0.22 micron) syringe filter, collected in an acid washed polycarbonate culture tube, and refrigerated prior to transport to the laboratory. The samples were stored refrigerated until preparation for ICP-MS analysis as described below. Initially triplicate samples were taken, but given the consistency of the results over the first 30 d of sampling, single samples were obtained to enable a longer time-series to be collected. 4. Field sampling Similar to the wastewater effluent samples, triplicate creek and river water grab samples were collected (sampling locations are as indicated in Fig. 3), either directly in a syringe (rinsed 3 times prior to collecting the sample) or drawn into the syringe through a (rinsed) 3 m polyethylene tube connected to a sampling pole to allow the sample to be collected away from the bank. Samples were filtered as above into acid washed polycarbonate culture tubes, and kept refrigerated until return to the laboratory. The first sampling campaign was conducted to determine whether WWTP discharges
795
could be traced, and was shortly after (10 d) after a minor flood. Therefore these samples represent a potential low concentration, as the river had just been flushed. However, despite the time since the main flood event, the flood peak had not passed Loudouns Weir, thus the final sample in the first campaign was at maximum dilution. The second and third campaigns were planned to occur when the creeks had resumed a more typical flow, and in a dry period. Thus the sampling campaigns reflect a range of possible conditions and encompass maximum and minimum dilution. 5. ICP-MS analysis All wastewater and natural water samples, standards, and blanks were acidified to pH 1.5 with quartz distilled nitric acid, and spiked to 4 ppb In and Rh to act as an internal standard. Samples were then analysed using a Thermo X-series ICP-MS housed in a clean facility at the University of Queensland. Instrumental conditions were similar to those previously described (Lawrence et al., 2006a, 2009). Briefly, the instrument was tuned for maximum sensitivity for In using the high performance sample introduction system, while maintaining oxide production below 2% (measured using CeO/Ce), and Ce++/Ce below 5%. The raw instrument data was corrected for variations in internal standard (In and Re), external drift (against the National Research Council of Canada river water reference material SLRS-4), and oxide and isobaric interferences, following Eggins et al. (1997) and Aries et al. (2000). Instrument response was then calibrated against hotplate digests for the US Geological Survey (USGS) dolerite reference material W-2, using REE concentrations listed in Table 1. A similarly prepared hotplate digest of the USGS basalt reference material BIR-1 was used as a primary quality control. 6. Calculation of wastewater discharge of Gd Typically, a rare earth element (REE) pattern for an uncontaminated freshwater sample is smooth (with the exception of La, Ce, Eu and Lu). Thus an anomalously high concentration of Gd can be attributed to anthropogenic contamination. Typical examples of REE patterns are provided in Fig. 1 (from the Wetalla WWTP and sample locations A and B, as shown in Fig. 3). The [GdAnth] in each wastewater effluent sample was calculated using Eq. (1), where [GdNatural] was calculated from a linear extrapolation between the shale-normalised concentrations of Sm and Tb. The choice of method for determining [GdNatural] has very little effect (typically <1%) on the final result as [GdAnth] is typically >30 times higher than [GdNatural]. The average of the calculated concentrations for Wetalla effluent discharge over the entire sampling period was used to determine the average GdAnth ‘‘end-member” concentration. 7. Calculation of percentage wastewater present in natural waters Previous research (Morteani et al., 2006; Kulaksiz and Bau, 2007) has suggested that GdAnth is a conservative tracer during mixing over timescales of up to 3 months. This is the primary assumption made for these calculations. The second assumption is that the only source of GdAnth in the catchment is the Wetalla WWTP. As it is the major WWTP that discharges effluent into Gowrie Creek, at the only location within the catchment where MRI imaging technology is available, this initial assumption appears to be reasonable. Given we can define an average end-member discharge (as described above) after determining the [GdAnth] at a downstream location, it is mathematically simple to determine the percentage wastewater that contributes to the sample. As the
796
M.G. Lawrence, D.G. Bariel / Chemosphere 80 (2010) 794–799
Table 1 Rare earth element data (ng kg1) for the three sampling campaigns; sampling locations are as indicated in Fig. 3. Percentage wastewater is indicated in the final column where all values are calculated assuming the average discharge (58 ng kg1) from the Wetalla WWTP. The values in parentheses were calculated assuming that the end-member was the instantaneous discharge from Wetalla on the day of sampling. Sample (ng kg1)
La
Ce
Pr
Nd
Sm
Eu
Gd
Tb
Dy
Ho
Er
Tm
Yb
Lu
Gd anomaly
Anth Gd
% Wastewater
1a 1b 1c 1d 1e 1i 1m 1p 1q
19.2 11.9 16.3 16.8 124.8 43.4 24.4 52.4 6.9
32.5 18.7 34.8 37.5 338.0 88.8 49.8 99.5 0.7
5.3 3.1 5.2 5.5 34.9 13.0 8.5 16.5 0.3
22.5 14.6 26.2 28.5 154.0 59.9 44.6 72.9 2.6
5.2 3.4 7.7 8.6 37.8 14.6 14.3 17.9 3.4
1.6 0.9 2.3 2.6 11.0 4.6 4.8 5.1
5.9 67.0 54.0 63.8 145.1 21.2 34.5 18.6 0.1
0.94 0.82 1.90 2.35 6.23 2.86 4.99 2.78 0.37
5.30 4.96 12.41 15.78 33.94 17.11 41.63 15.75 0.40
1.21 1.28 3.25 3.96 6.79 3.76 10.41 3.25 0.05
3.47 3.80 10.09 12.11 17.83 10.54 29.10 8.88 0.14
0.48 0.60 1.49 1.76 2.40 1.53 3.85 1.26 0.04
2.78 3.53 9.64 11.34 13.94 9.48 21.99 7.38 0.10
0.52 0.63 1.76 2.01 2.27 1.62 3.40 1.19 0.06
1.05 14.93 5.16 5.02 3.81 1.28 1.21 1.07
62.1 42.5 49.8 103.2 2.8 3.1
107 (82) 73 86 178 5 5
2a 2b 2c 2d 2e 2f 2g 2h 2j 2m 2p 2q
11.3 5.6 19.5 19.3 75.7 34.0 19.9 21.3 10.6 12.7 84.6 1.2
16.7 10.1 43.8 49.1 176.5 66.9 44.8 44.3 17.6 23.3 178.5 0.3
3.1 1.7 6.1 6.4 20.5 9.8 6.6 7.8 4.0 3.5 26.8 0.1
14.5 9.0 31.5 32.7 89.9 42.5 32.1 41.4 23.2 17.5 127.4 1.1
4.6 3.7 10.2 11.0 22.3 11.1 10.2 13.4 8.4 5.7 31.0 1.0
1.6 1.1 3.7 4.2 6.6 3.4 3.9 5.3 3.5 2.2 8.8 0.4
4.4 70.4 58.9 59.4 57.6 11.1 53.0 63.1 26.1 7.2 31.4 0.3
0.73 0.77 2.15 2.24 3.54 1.75 2.00 2.95 2.45 0.99 4.52 0.04
4.74 4.38 15.47 16.21 19.97 10.09 13.60 20.49 26.16 6.98 26.01 0.30
1.12 1.11 3.81 4.09 4.04 2.16 3.43 5.16 9.01 2.25 5.60 0.12
3.16 3.78 11.88 11.96 10.13 6.05 10.52 14.50 28.47 9.13 14.54 0.34
0.48 0.56 1.73 1.81 1.42 0.90 1.54 2.25 4.02 1.67 2.20 0.10
2.82 3.77 10.88 10.86 8.51 5.49 9.62 13.49 22.12 10.79 13.14 0.56
0.46 0.64 1.75 1.72 1.26 0.88 1.59 2.04 3.31 1.84 2.11 0.11
0.93 16.38 4.57 4.32 2.60 1.01 4.37 3.61 2.00 1.20 1.06
65.7 44.7 44.3 33.2
113 (83) 77 76 57
39.7 43.8 11.7 0.6
68 76 20 1
3c 3e 3j 3k 3l 3m 3o 3p 3q
37.9 69.4 15.4 17.2 14.6 15.9 49.6 53.1 10.2
84.4 161.2 18.0 24.7 15.2 25.3 101.6 109.5 0.5
10.6 19.2 3.5 4.1 2.4 3.5 15.2 16.3 0.3
46.6 83.0 21.0 22.8 14.3 17.2 71.7 73.8 2.2
11.6 19.9 7.4 7.5 5.7 5.4 17.5 17.9 1.9
3.6 6.1
23.4 30.5 19.8 19.6 12.8 7.7 21.4 21.5 0.3
2.21 3.54 1.89 1.82 1.20 1.14 2.96 3.02 0.07
17.49 24.92 17.54 17.58 9.90 9.62 21.03 21.37 0.60
2.89 3.66 3.15 3.71 1.67 1.50 3.15 3.05 0.09
7.97 9.55 12.78 17.63 8.70 4.75 8.97 8.14 0.38
1.17 1.35 2.11 3.13 2.10 0.78 1.35 1.19 0.03
6.76 7.61 15.05 21.26 17.19 6.16 7.87 6.98 0.30
1.12 1.18 2.73 3.72 3.50 1.27 1.21 1.01 0.07
1.79 1.46 1.95 1.94 1.85 1.23 1.20 1.19
8.9 7.5 8.6 8.5 5.2 0.8 1.7 1.6
15 13 15 15 9 1 3 3
natural water sampling protocol entails the collection of grab samples, this study is able to determine the presence or absence of a detectable wastewater signal in the natural waters, and given the above assumption, provide an estimate of the percentage of wastewater that may be present. If the overall REE pattern is coherent (not approaching the analytical limits of uncertainty), we can confidently determine an GdAnth component when the excess Gd is >10% of the calculated natural value (e.g. Eq. (1)). Therefore, the detection of GdAnth is dependant on [GdNatural], and it is not possible to state an absolute detection limit of GdAnth, as it varies depending on the sample.
Nov/1
Oct/16
Oct/1
0
Sep/16
Fig. 1. Mud of Queensland (MuQ, (Kamber et al., 2005)) normalised Rare Earth Element patterns of filtered water 200 m upstream of the Wetalla WWTP discharge, the WWTP discharge itself, and 500 m downstream of the WWTP discharge location.
Aug/31
Rare Earth Elements
20
Aug/16
La Ce Pr Nd Pm Sm Eu Gd Tb Dy Ho Er Tm Yb Lu
40
Aug/1
10 -8
60
Jul/16
Gowrie Creek 200 m upstream WWTP WWTP Discharge Gowrie Creek, 500 m downstream
80
Jul/1
10 -7
100
Jun/16
10 -6
Anthropogenic Gd
120
Jun/1
Sample/MuQ
10 -5
Anthropogenic Gd (ng/kg)
140
Fig. 2. Concentration of GdAnth measured in 24-h volume proportional (every 100 kL) samples from the Wetalla WWTP from May to October 2009. Note the clear weekly pattern in the discharge data – with a minimum observed in samples collected at 8 am on Monday and Tuesday morning. As the radiology clinic operates on weekdays only, this result is interpreted as indicated that the Saturday and Sunday minimum inputs take 48 h to pass though the treatment plant. The average discharge is 0.84 ± 0.4 g Gd.
8. Results and discussion 8.1. Discharge from WWTP The [GdAnth] was determined for each day of the sampling period using Eq. (1), and are shown in Fig. 2. Daily discharge from the
M.G. Lawrence, D.G. Bariel / Chemosphere 80 (2010) 794–799
Wetalla WWTP describes a consistent weekly pattern with an average concentration of 58 ± 23 ng kg1 throughout the time of the sampling (May–October 2009); the weekly pattern remains, even when concentrations are converted to loads (g Gd d1, data not shown). The strong weekly cycle evident in the data is consistent with the interpretation that the predominant application of MRI contrast agents occurs during the working week (Monday–Friday in the private imaging clinic), and is consistent with previous research (Knappe et al., 2005) in Berlin. In the case of Berlin (population 3.5 M), a total daily discharge of 178 g Gd was calculated, which on a population basis equates to 5 g Gd d1 per 100 000 persons. In contrast, the Wetalla WWTP, serving a population of 96 000 people has an average daily discharge of 0.84 ± 0.4 g Gd, a comparative load of 0.9 ± 0.4 g Gd d1 per 100 000 people. The recommended current dosage for both Gadovist and Magnevist, two of the common contrast agents, is 0.1 mmol kg1 body weight, which for the average 70 kg Australian adult equates to 1.1 g Gd. The discharge load, within error and assuming no degradation within the WWTP, is approximately the equivalent of one patient receiving contrast per day. In our previous research, in the more heavily populated SE-Qld region (Lawrence et al., 2009), a major WWTP was calculated to discharge 3.3 g Gd d1 per 100 000 people, however, most other WWTPs were discharging a similar load to Wetalla WWTP. Thus the lower loads in the regional centre probably reflect the lower density of MRI imaging technology (there are two MRI units known to be in operation in the catchment of this WWTP), and possibly the ongoing trend towards the application of lower concentrations of contrast agents. 8.2. River water samples within catchment Water samples were taken on three different occasions in May, July, and September, spanning the catchment from 200 m upstream of the WWTP discharge location to Loudouns Weir, which is used as a water supply for the Dalby community, approximately 100 km downstream (e.g. Fig. 3). There had been a significant rainfall event ( 75 mm 9-d prior to sampling, resulting in a tripling of WWTP discharge) in the catchment prior to the May sampling campaign – and the flood peak had not passed Loudouns Weir completely at the time of sampling; there was no other significant rainfall in the lead-up to the following sampling campaigns (a 40 mm rainfall event occurred 21 d prior to the final sampling campaign, but given the dry ground conditions, does not appear to have resulted in any run-off, and WWTP discharges remained normal in the days surrounding the event). On any one sampling campaign, up to 15 sample locations were sampled with the aim of sampling the main waterway and any tributaries that were flowing into it (not all tributaries continued to flow over the course of the sampling). As expected, and demonstrated in Fig. 3, there was no anthropogenic Gd detected above the WWTP inflow at any time. In contrast, the WWTP itself and the sample 500 m downstream from the WWTP both have large positive Gd anomalies. When the average WWTP concentration of 58 ng kg1 Gd is assumed to be the endmember, the concentrations measured at location B in May and July indicate that there is 107–113% wastewater present at this location. However, for this location, rather than the average discharge, the best value for the end-member is more likely to be the instantaneous sample collected from the WWTP just minutes prior to collecting from location B. On the first sampling campaign, the WWTP was discharging an [GdAnth] of 76 ng kg1 and on the second campaign the instantaneous WWTP [GdAnth] was 0.78 ng kg1. If these values were instead chosen to calculate the percentage wastewater present, we would estimate that wastewater contributed 81.5% and 83.2% of the total flow on these occasions. Whist there were no gauging stations above the WWTP to
797
measure the actual flow, these calculations are consistent with visual observations of the WWTP discharge, where there was clearly more water entering the creek from the WWTP. Whilst the instantaneous measure of [GdAnth] is the most logical value to choose when estimating the percentage of wastewater in a sample close to the WWTP, this is clearly not the best end-member to choose when traveling over 100 km downstream. In that case, where hydrological data would be required to estimate which water packet from the WWTP had mixed with the creek, the average value is instead chosen (as indicated by the flood peak not having completely passed Loudouns Weir 9 d after the rainfall event, the hydraulic residence time, whilst affected by volume, is therefore at least 9 d). As the average end-member concentration varies by ±40%, the estimates of the percentage wastewater are also subject to the same uncertainty. Therefore, we quote the absolute percentage wastewater detected assuming an average discharge, with the proviso that the actual concentrations could vary by half the value stated (i.e. a value of 20% could actually range from 12% to 28%). Nonetheless, for the final sampling location, Loudouns Weir near Dalby, the weir will increase the residence time of the water, and if the residence time is >1 week (the time for one concentration cycle at the WWTP), the average end-member value is the most appropriate, and our estimates of the percent wastewater present are likely to be more accurate within the stated assumptions. Over all sampling campaigns, the percentage wastewater follows the same general pattern, decreasing as the distance from the WWTP increases. As the first sampling campaign in May occurred 1 week after a minor flood event, the entire system had been flushed, and the detection of 5% wastewater at location ‘‘M” would be considered to be proof of a significant amount of wastewater being present. However, there appears to be an anomaly during this sampling campaign, with the largest detected GdAnth component in Oakey Creek, with a calculated [GdAnth] of 100 ng kg1. Clearly the assumption that the Wetalla WWTP is the only source of GdAnth is incorrect. Unfortunately, as a result of the second point source, the calculated result of 5% wastewater at location M in the first sampling campaign is unreliable, as it is possible that we have instead encountered the very high concentration end-member from the Oakey WWTP or a leaking septic system, not the expected Wetalla WWTP signal. The township of Oakey has a small WWTP with an average daily discharge of 700 kL d1. As Oakey is close (30 km) to Toowoomba, we can only assume that a patient traveled to Toowoomba for an MRI, then traveled home prior to excreting the contrast agent. This example demonstrates that MRI itself does not have to be present at a particular location, only that advanced medical imaging is available within the excretion half-life of the contrast agent. Interestingly the detection of such a high concentration is not surprising – for example, if the entire load of 1 dose (1.1 g) of contrast agent was to be introduced into the Oakey WWTP (700 kL) over 1 d, a concentration of 1570 ng kg1 would have been expected in the effluent. It is also interesting to speculate on how commonly an GdAnth spike would occur in Oakey Creek. On a population basis, Toowoomba (96 000) is 27 times larger than Oakey (3500). On average 0.9 people are treated in Toowoomba per day, therefore by extrapolation, one person in Oakey is likely to be treated every 30 d. Our sampling campaign was therefore fortunate (or unfortunate) to capture this event, but it does indicate that caution must be applied when assessing wastewater contributions with single tracers over such large catchments. Whilst we could not detect wastewater at Loudouns Weir on the first sampling campaign, this was not unexpected, as we actually ‘‘caught up” to the river peak from the rainfall event 9 d prior to sampling, with water overtopping the weir – thus the extensive
798
M.G. Lawrence, D.G. Bariel / Chemosphere 80 (2010) 794–799
Fig. 3. A–C: Maps indicating the calculated percentage of wastewater at each location for sampling campaigns in 2009, assuming that the Wetalla WWTP is the only source of GdAnth, and that the average wastewater discharge concentration is 58 ng kg1 Gd (*indicates that these samples were calculated from the instantaneous discharge of Gd; **calculated using the average Wetalla WWTP discharge, which is not applicable for these samples). Note that Oakey Creek always presents with a Gd anomaly, which is consistent with the presence of a small (700 kL d1) WWPT that discharges into this creek.
M.G. Lawrence, D.G. Bariel / Chemosphere 80 (2010) 794–799
dilution would have dropped the wastewater percentage below the detectable level. Second sampling campaign: Results for the second campaign indicate that high percentages of wastewater are detectable over a large area of the catchment, for example, at location H, approximately 40 km down river, the wastewater contribution was still calculated at 76 ± 30%. At location J, approximately 60 km from the wastewater discharge, we estimate that 19 ± 8% wastewater is present. Such an outcome is very interesting from a policy perspective. Under the current legislation, the wastewater treatment plant would not be able to distribute the effluent water for reuse as irrigation water for cropping, yet, at these locations where 19–75% wastewater is present in Gowrie Creek, the history of the water is unimportant. The question then becomes why was the wastewater considered a risk in the first instance, and has transport down the natural system mitigated this risk? A similar situation was observed during the third sampling campaign – but a lower wastewater contribution was detected at the mid-catchment locations (typically 15%). However, on this sampling campaign, GdAnth was detected at Loudouns Weir with 3 ± 1.2% wastewater detected. This value is lower than estimated by Hamilton and Greenfield (1991) (8.3% at low flow), but is within the same order of magnitude as their estimate. At the time of sampling, Gowrie Creek was drying significantly, and Lagoon Creek, Westbrook Creek, and the Condomine River were not flowing (no samples taken). Thus it would be expected that samples collected at this time were representative of a typical dry season. While we believe that the evidence of Morteani et al. (2006) and Kulaksiz and Bau (2007) is strong, it must be pointed out that the assumption that GdAnth behaves conservatively in this type of surface water environment (i.e. shallow, slow flowing, with intense solar radiation) over the time scale of transport has not yet been thoroughly tested in the literature. If environmental degradation does occur, the estimates of the extent of wastewater contamination calculated must be viewed as the minimum levels, although our estimates could be recalculated by modeling a decay rate, and estimating the hydraulic residence time.
9. Conclusions We have demonstrated that there is: A strong weekly cycle in the discharge of Gd into the environment from a typical WWTP. An average of 58 ± 23 ng kg1 of GdAnth discharged from the Wetalla WWTP over 5 months of sampling. GdAnth was easily detected in receiving waters >50 km downstream of the WWTP source. Up to 3% wastewater was detected up to 100 km away from the point source.
Acknowledgements MGL wishes to thank the Toowoomba Regional Council, and the operators of the Wetalla WWTP for their invaluable assistance, and acknowledges the financial support (salary) of the Chair in Water
799
Recycling Agreement between Water Secure, Veolia Water and the University of Queensland. References Aries, S., Valladon, M., Polve, M., Dupre, B., 2000. A routine method for oxide and hydroxide interference corrections in ICP-MS chemical analysis of environmental and geological samples. Geostand. Newslett. – J. Geostand. Geoanal. 24, 19–31. Bau, M., Dulski, P., 1996. Anthropogenic origin of positive gadolinium anomalies in river waters. Earth Planet. Sci. Lett. 143, 245–255. Bau, M., Knappe, A., Dulski, P., 2006. Anthropogenic gadolinium as a micropollutant in river waters in Pennsylvania, and in Lake Erie, northeastern United States. Chem. Erde-Geochem. 66, 143–152. Eggins, S.M., Woodhead, J.D., Kinsley, L.P.J., Mortimer, G.E., Sylvester, P., McCulloch, M.T., Hergt, J.M., Handler, M.R., 1997. A simple method for the precise determination of >=40 trace elements in geological samples by ICPMS using enriched isotope internal standardisation. Chem. Geol. 134, 311–326. Hamilton, G., Greenfield, P.F., 1991. Potable reuse of treated wastewater. In: Australian Water and Wastewater Association 14th Federal Convention. Australian Water and Wastewater Association, Perth, pp. 497–506. Kamber, B.S., Greig, A., Collerson, K.D., 2005. A new estimate for the composition of weathered young upper continental crust from alluvial sediments, Queensland, Australia. Geochim. Cosmochim. Acta 69, 1041–1058. Knappe, A., Möller, P., Dulski, P., Pekdeger, A., 2005. Positive gadolinium anomaly in surface water and ground water of the urban area Berlin, Germany. Chem. ErdeGeochem. 65, 167–189. Kulaksiz, S., Bau, M., 2007. Contrasting behaviour of anthropogenic gadolinium and natural rare earth elements in estuaries and the gadolinium input into the North Sea. Earth Planet. Sci. Lett. 260, 361–371. Kümmerer, K., Helmers, E., 2000. Hospital effluents as a source of gadolinium in the aquatic environment. Environ. Sci. Technol. 34, 573–577. Lawrence, M.G., Greig, A., Collerson, K.D., Kamber, B.S., 2006a. Direct quantification of rare earth element concentrations in natural waters by ICP-MS. Appl. Geochem. 21, 839–848. Lawrence, M.G., Greig, A., Collerson, K.D., Kamber, B.S., 2006b. Rare earth element and yttrium variability in South East Queensland waterways. Aquat. Geochem. 12, 39–72. Lawrence, M.G., Jupiter, S.D., Kamber, B.S., 2006c. Aquatic geochemistry of the rare earth elements and yttrium in the Pioneer River catchment, Australia. Mar. Freshw. Res. 57, 725–736. Lawrence, M.G., Ort, C., Keller, J., 2009. Detection of anthropogenic gadolinium in treated wastewater in South-East Queensland. Water Res. 43, 3534–3540. Melbourne Water, 2009. Water System Storage. Melbourne.. Möller, P., Dulski, P., Bau, M., Knappe, A., Pekdeger, A., Sommer-von Jarmersted, C., 2000. Anthropogenic gadolinium as a conservative tracer in hydrology. J. Geochem. Explor. 69, 409–414. Möller, P., Morteani, G., Dulski, P., 2003. Anomalous gadolinium, cerium, and yttrium contents in the Adige and Isarco river waters and in the water of their tributaries (Provinces Trento and Bolzano/Bozen, NE Italy). Acta Hydrochim. Hydrobiol. 31, 225–239. Morteani, G., Möller, P., Fuganti, A., Paces, T., 2006. Input and fate of anthropogenic estrogens and gadolinium in surface water and sewage plants in the hydrological basin of Prague (Czech Republic). Environ. Geochem. Health 28, 257–264. New South Wales Government, S.C.A., 2007. Bulk Water Storage and Supply Report – 8 February 2007. Water Storage and Supply Reports, Sydney. Queensland Government, 2007. The South East Queensland drought to 2007. In: Water, Department of Natural Resources and Mines. Queensland Government, Brisbane. Rabiet, M., Brissaud, F., Seidel, J.L., Pistre, S., Elbaz-Poulichet, F., 2005. Deciphering the presence of wastewater in a medium-sized mediterranean catchment using a multitracer approach. Appl. Geochem. 20, 1587–1596. Rabiet, M., Brissaud, F., Seidel, J.L., Pistre, S., Elbaz-Poulichet, F., 2009. Positive gadolinium anomalies in wastewater treatment plant effluents and aquatic environment in the Herault watershed (South France). Chemosphere 75, 1057– 1064. SEQ Water, 2009. Current Dam Levels. . Verplanck, P.L., Taylor, H.E., Nordstrom, D.K., Barber, L.B., 2005. Aqueous stability of gadolinium in surface waters receiving sewage treatment plant effluent, Boulder Creek, Colorado. Environ. Sci. Technol. 39, 6923–6929.
Contents lists available at ScienceDirect
Chemosphere journal homepage: www.elsevier.com/locate/chemosphere
Tracing treated wastewater in an inland catchment using anthropogenic gadolinium Michael Glen Lawrence *, David Guimerà Bariel Advanced Water Management Centre, The University of Queensland, St. Lucia, Qld 4072, Australia
a r t i c l e
i n f o
Article history: Received 1 December 2009 Received in revised form 30 April 2010 Accepted 3 May 2010
Keywords: Australia Risk-assessment Micropollutants Drinking water Toowoomba Queensland
a b s t r a c t The discharge of treated wastewater into natural water bodies occurs worldwide; if drinking water is then extracted downstream, there is potential for micropollutants that are not fully mineralized in the wastewater treatment process to enter municipal drinking water. In Australia, drinking water treatment is typically a mixture of basic technologies such as flocculation and slow sand filtration; technologies that are not specifically designed to remove micropollutants. However, there is little awareness in Australia of the potential risk that upstream wastewater discharges may impart to the security and quality of downstream drinking water supplies. We apply a direct inductively coupled plasma mass spectrometry technique to determine the discharge of anthropogenic gadolinium from a wastewater treatment plant in Toowoomba, Queensland, Australia, that discharges into the small (147 km2) catchment of Gowrie Creek. We then continue to measure the concentrations of this wastewater tracer as Gowrie Creek flows downstream into the Condomine River, and to a community 100 km away where drinking water is extracted. Using this tracer, we demonstrate that the community has a detectable wastewater contribution within their surface drinking water supply. Ó 2010 Elsevier Ltd. All rights reserved.
1. Introduction Vast areas of eastern Australia are currently enduring, or have recently emerged from, the worst drought on record (Queensland Government, 2007). In this context, surface water supplies of more than half of all Australians were dramatically reduced; for example, in Sydney (population 4.3 million), the major water supply dam dropped to 32.5%, in Melbourne (population 3.8 million) the water supply dropped to 25% capacity, and in Brisbane (1.9 million), the water supply dropped to 16.9% capacity (New South Wales Government, 2007; Melbourne Water, 2009; SEQ Water, 2009). Severe water restrictions were applied, and community awareness of issues surrounding water security in Australia has become intense. Alternate water supplies have been investigated and desalination infrastructure is being built around the country. Yet water recycling, one of the more obvious water supply options, despite being heavily regulated, is not a widely acceptable solution for augmenting drinking water supplies. However, it is little recognized that in many inland Australian catchments, baseline water flows are often supported by the upstream discharge of wastewater. For example, previous research (Hamilton and Greenfield, 1991) based on hydraulic calculations of mean river flow and wastewater treatment plant discharge has suggested that the average wastewater contribution to the drinking water supply at Dalby, on the Condomine River, was 8.3%. However, this hypothesis has * Corresponding author. Tel.: +617 3346 3228; fax: +617 3365 4726. E-mail address: [email protected] (M.G. Lawrence). 0045-6535/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.chemosphere.2010.05.001
never been tested in practice. We present a case study where the very sensitive anthropogenic Gd tracer is used to determine the percentage of wastewater that is detectable at Dalby, approximately 100 km downstream from the wastewater discharge at Toowoomba. While this is a specific example, the applicability of this tracer is not limited to this case alone. 2. Aims of paper The primary aim of this manuscript is to quantify the extent to which wastewater contributes to downstream water supplies in regional Australia. Towards this aim we have targeted gadolinium from MRI contrast agents, which has been demonstrated to be present in wastewater from areas with advanced medical imaging (Bau and Dulski, 1996; Kümmerer and Helmers, 2000; Möller et al., 2000; Knappe et al., 2005; Rabiet et al., 2005; Verplanck et al., 2005; Bau et al., 2006; Kulaksiz and Bau, 2007; Lawrence et al., 2009; Rabiet et al., 2009). Gadolinium can be quantified directly in natural water and wastewater effluent by ICP-MS (Lawrence et al., 2006b,c, 2009); in contrast, other organic micropollutants such as pharmaceuticals and personal care products typically require >100 fold pre-concentration in natural waters prior to quantification by LC–MS/MS or GCMS. The excess Gd from MRI contrast agents is defined as GdAnth where:
½GdAnth ¼ ½GdMeasured ð1:1 ½GdNatural Þ
ð1Þ
where the factor 1.1 accounts for the small natural Gd anomaly usually observed in South East Queensland (Lawrence et al., 2006b). In
M.G. Lawrence, D.G. Bariel / Chemosphere 80 (2010) 794–799
this case the natural Gd concentration is determined from a third order polynomial fit of the non-anomalous rare earth elements (REE’s) (Möller et al., 2003). Given the current evidence (Kulaksiz and Bau, 2007) that GdAnth is refractory during estuarine mixing, and qualitatively stable in groundwater infiltration (Möller et al., 2000), there appears to be potential to treat the tracer as conservative, and thus exploit GdAnth as a tracer. As there are at least 2 MRI instruments located in Toowoomba, it was assumed that this tracer should be present. The first step towards determining whether GdAnth can be utilised as a wastewater tracer is to define the point source discharge. Previous research has suggested that there is a weekly cycle in the discharge of GdAnth in wastewater effluent. Therefore, a sampling strategy was implemented to determine the long-term (5-month) discharge of GdAnth from the Wetalla WWTP. Concurrently, three sampling campaigns were undertaken in the Gowrie Creek/Condomine River catchment to determine whether GdAnth was detectable, and if possible, estimate the amount of wastewater present. 3. Sample locations Wetalla WWTP is a biological nutrient removal plant with a design capacity of 36 ML d1 but typically operating at about half this capacity. The wastewater treatment plant services a population of 100 000 people and discharges its treated effluent into Gowrie Creek, which flows downstream where, in approximately 100 km, water is extracted for drinking water in Dalby. The flow from the WWTP is suspected to be the predominant water supply in this creek. During the sampling campaigns this was visually apparent, with little flow in Gowrie Creek above the WWTP with a substantial increase in flow below the discharge point. As MRI contrast agents are expected to be applied to only a limited number of people within Toowoomba each day, it can be expected that the influent of the wastewater treatment plant would have sporadic spikes, and inhomogeneous loads of GdAnth. However, given the large buffering capacity of the WWTP, these concentration spikes should be mixed during treatment, and the effluent should be more homogenous. Thus, it was considered that the existing effluent sampling conducted at the treatment plant, notably daily volume proportional samples of the effluent immediately prior to discharge into Gowrie Creek (an ephemeral creek in a small, 147 km2 catchment) should provide representative samples. Daily 24-h composite samples of the effluent from the Wetalla wastewater treatment plant were collected in a refrigerated (4 °C) auto-sampler triggered to collect a constant volume sub-sample of water every 100 kL. A 10 mL sub-sample of this water was then filtered with a (Millipore Millex GV 0.22 micron) syringe filter, collected in an acid washed polycarbonate culture tube, and refrigerated prior to transport to the laboratory. The samples were stored refrigerated until preparation for ICP-MS analysis as described below. Initially triplicate samples were taken, but given the consistency of the results over the first 30 d of sampling, single samples were obtained to enable a longer time-series to be collected. 4. Field sampling Similar to the wastewater effluent samples, triplicate creek and river water grab samples were collected (sampling locations are as indicated in Fig. 3), either directly in a syringe (rinsed 3 times prior to collecting the sample) or drawn into the syringe through a (rinsed) 3 m polyethylene tube connected to a sampling pole to allow the sample to be collected away from the bank. Samples were filtered as above into acid washed polycarbonate culture tubes, and kept refrigerated until return to the laboratory. The first sampling campaign was conducted to determine whether WWTP discharges
795
could be traced, and was shortly after (10 d) after a minor flood. Therefore these samples represent a potential low concentration, as the river had just been flushed. However, despite the time since the main flood event, the flood peak had not passed Loudouns Weir, thus the final sample in the first campaign was at maximum dilution. The second and third campaigns were planned to occur when the creeks had resumed a more typical flow, and in a dry period. Thus the sampling campaigns reflect a range of possible conditions and encompass maximum and minimum dilution. 5. ICP-MS analysis All wastewater and natural water samples, standards, and blanks were acidified to pH 1.5 with quartz distilled nitric acid, and spiked to 4 ppb In and Rh to act as an internal standard. Samples were then analysed using a Thermo X-series ICP-MS housed in a clean facility at the University of Queensland. Instrumental conditions were similar to those previously described (Lawrence et al., 2006a, 2009). Briefly, the instrument was tuned for maximum sensitivity for In using the high performance sample introduction system, while maintaining oxide production below 2% (measured using CeO/Ce), and Ce++/Ce below 5%. The raw instrument data was corrected for variations in internal standard (In and Re), external drift (against the National Research Council of Canada river water reference material SLRS-4), and oxide and isobaric interferences, following Eggins et al. (1997) and Aries et al. (2000). Instrument response was then calibrated against hotplate digests for the US Geological Survey (USGS) dolerite reference material W-2, using REE concentrations listed in Table 1. A similarly prepared hotplate digest of the USGS basalt reference material BIR-1 was used as a primary quality control. 6. Calculation of wastewater discharge of Gd Typically, a rare earth element (REE) pattern for an uncontaminated freshwater sample is smooth (with the exception of La, Ce, Eu and Lu). Thus an anomalously high concentration of Gd can be attributed to anthropogenic contamination. Typical examples of REE patterns are provided in Fig. 1 (from the Wetalla WWTP and sample locations A and B, as shown in Fig. 3). The [GdAnth] in each wastewater effluent sample was calculated using Eq. (1), where [GdNatural] was calculated from a linear extrapolation between the shale-normalised concentrations of Sm and Tb. The choice of method for determining [GdNatural] has very little effect (typically <1%) on the final result as [GdAnth] is typically >30 times higher than [GdNatural]. The average of the calculated concentrations for Wetalla effluent discharge over the entire sampling period was used to determine the average GdAnth ‘‘end-member” concentration. 7. Calculation of percentage wastewater present in natural waters Previous research (Morteani et al., 2006; Kulaksiz and Bau, 2007) has suggested that GdAnth is a conservative tracer during mixing over timescales of up to 3 months. This is the primary assumption made for these calculations. The second assumption is that the only source of GdAnth in the catchment is the Wetalla WWTP. As it is the major WWTP that discharges effluent into Gowrie Creek, at the only location within the catchment where MRI imaging technology is available, this initial assumption appears to be reasonable. Given we can define an average end-member discharge (as described above) after determining the [GdAnth] at a downstream location, it is mathematically simple to determine the percentage wastewater that contributes to the sample. As the
796
M.G. Lawrence, D.G. Bariel / Chemosphere 80 (2010) 794–799
Table 1 Rare earth element data (ng kg1) for the three sampling campaigns; sampling locations are as indicated in Fig. 3. Percentage wastewater is indicated in the final column where all values are calculated assuming the average discharge (58 ng kg1) from the Wetalla WWTP. The values in parentheses were calculated assuming that the end-member was the instantaneous discharge from Wetalla on the day of sampling. Sample (ng kg1)
La
Ce
Pr
Nd
Sm
Eu
Gd
Tb
Dy
Ho
Er
Tm
Yb
Lu
Gd anomaly
Anth Gd
% Wastewater
1a 1b 1c 1d 1e 1i 1m 1p 1q
19.2 11.9 16.3 16.8 124.8 43.4 24.4 52.4 6.9
32.5 18.7 34.8 37.5 338.0 88.8 49.8 99.5 0.7
5.3 3.1 5.2 5.5 34.9 13.0 8.5 16.5 0.3
22.5 14.6 26.2 28.5 154.0 59.9 44.6 72.9 2.6
5.2 3.4 7.7 8.6 37.8 14.6 14.3 17.9 3.4
1.6 0.9 2.3 2.6 11.0 4.6 4.8 5.1
5.9 67.0 54.0 63.8 145.1 21.2 34.5 18.6 0.1
0.94 0.82 1.90 2.35 6.23 2.86 4.99 2.78 0.37
5.30 4.96 12.41 15.78 33.94 17.11 41.63 15.75 0.40
1.21 1.28 3.25 3.96 6.79 3.76 10.41 3.25 0.05
3.47 3.80 10.09 12.11 17.83 10.54 29.10 8.88 0.14
0.48 0.60 1.49 1.76 2.40 1.53 3.85 1.26 0.04
2.78 3.53 9.64 11.34 13.94 9.48 21.99 7.38 0.10
0.52 0.63 1.76 2.01 2.27 1.62 3.40 1.19 0.06
1.05 14.93 5.16 5.02 3.81 1.28 1.21 1.07
62.1 42.5 49.8 103.2 2.8 3.1
107 (82) 73 86 178 5 5
2a 2b 2c 2d 2e 2f 2g 2h 2j 2m 2p 2q
11.3 5.6 19.5 19.3 75.7 34.0 19.9 21.3 10.6 12.7 84.6 1.2
16.7 10.1 43.8 49.1 176.5 66.9 44.8 44.3 17.6 23.3 178.5 0.3
3.1 1.7 6.1 6.4 20.5 9.8 6.6 7.8 4.0 3.5 26.8 0.1
14.5 9.0 31.5 32.7 89.9 42.5 32.1 41.4 23.2 17.5 127.4 1.1
4.6 3.7 10.2 11.0 22.3 11.1 10.2 13.4 8.4 5.7 31.0 1.0
1.6 1.1 3.7 4.2 6.6 3.4 3.9 5.3 3.5 2.2 8.8 0.4
4.4 70.4 58.9 59.4 57.6 11.1 53.0 63.1 26.1 7.2 31.4 0.3
0.73 0.77 2.15 2.24 3.54 1.75 2.00 2.95 2.45 0.99 4.52 0.04
4.74 4.38 15.47 16.21 19.97 10.09 13.60 20.49 26.16 6.98 26.01 0.30
1.12 1.11 3.81 4.09 4.04 2.16 3.43 5.16 9.01 2.25 5.60 0.12
3.16 3.78 11.88 11.96 10.13 6.05 10.52 14.50 28.47 9.13 14.54 0.34
0.48 0.56 1.73 1.81 1.42 0.90 1.54 2.25 4.02 1.67 2.20 0.10
2.82 3.77 10.88 10.86 8.51 5.49 9.62 13.49 22.12 10.79 13.14 0.56
0.46 0.64 1.75 1.72 1.26 0.88 1.59 2.04 3.31 1.84 2.11 0.11
0.93 16.38 4.57 4.32 2.60 1.01 4.37 3.61 2.00 1.20 1.06
65.7 44.7 44.3 33.2
113 (83) 77 76 57
39.7 43.8 11.7 0.6
68 76 20 1
3c 3e 3j 3k 3l 3m 3o 3p 3q
37.9 69.4 15.4 17.2 14.6 15.9 49.6 53.1 10.2
84.4 161.2 18.0 24.7 15.2 25.3 101.6 109.5 0.5
10.6 19.2 3.5 4.1 2.4 3.5 15.2 16.3 0.3
46.6 83.0 21.0 22.8 14.3 17.2 71.7 73.8 2.2
11.6 19.9 7.4 7.5 5.7 5.4 17.5 17.9 1.9
3.6 6.1
23.4 30.5 19.8 19.6 12.8 7.7 21.4 21.5 0.3
2.21 3.54 1.89 1.82 1.20 1.14 2.96 3.02 0.07
17.49 24.92 17.54 17.58 9.90 9.62 21.03 21.37 0.60
2.89 3.66 3.15 3.71 1.67 1.50 3.15 3.05 0.09
7.97 9.55 12.78 17.63 8.70 4.75 8.97 8.14 0.38
1.17 1.35 2.11 3.13 2.10 0.78 1.35 1.19 0.03
6.76 7.61 15.05 21.26 17.19 6.16 7.87 6.98 0.30
1.12 1.18 2.73 3.72 3.50 1.27 1.21 1.01 0.07
1.79 1.46 1.95 1.94 1.85 1.23 1.20 1.19
8.9 7.5 8.6 8.5 5.2 0.8 1.7 1.6
15 13 15 15 9 1 3 3
natural water sampling protocol entails the collection of grab samples, this study is able to determine the presence or absence of a detectable wastewater signal in the natural waters, and given the above assumption, provide an estimate of the percentage of wastewater that may be present. If the overall REE pattern is coherent (not approaching the analytical limits of uncertainty), we can confidently determine an GdAnth component when the excess Gd is >10% of the calculated natural value (e.g. Eq. (1)). Therefore, the detection of GdAnth is dependant on [GdNatural], and it is not possible to state an absolute detection limit of GdAnth, as it varies depending on the sample.
Nov/1
Oct/16
Oct/1
0
Sep/16
Fig. 1. Mud of Queensland (MuQ, (Kamber et al., 2005)) normalised Rare Earth Element patterns of filtered water 200 m upstream of the Wetalla WWTP discharge, the WWTP discharge itself, and 500 m downstream of the WWTP discharge location.
Aug/31
Rare Earth Elements
20
Aug/16
La Ce Pr Nd Pm Sm Eu Gd Tb Dy Ho Er Tm Yb Lu
40
Aug/1
10 -8
60
Jul/16
Gowrie Creek 200 m upstream WWTP WWTP Discharge Gowrie Creek, 500 m downstream
80
Jul/1
10 -7
100
Jun/16
10 -6
Anthropogenic Gd
120
Jun/1
Sample/MuQ
10 -5
Anthropogenic Gd (ng/kg)
140
Fig. 2. Concentration of GdAnth measured in 24-h volume proportional (every 100 kL) samples from the Wetalla WWTP from May to October 2009. Note the clear weekly pattern in the discharge data – with a minimum observed in samples collected at 8 am on Monday and Tuesday morning. As the radiology clinic operates on weekdays only, this result is interpreted as indicated that the Saturday and Sunday minimum inputs take 48 h to pass though the treatment plant. The average discharge is 0.84 ± 0.4 g Gd.
8. Results and discussion 8.1. Discharge from WWTP The [GdAnth] was determined for each day of the sampling period using Eq. (1), and are shown in Fig. 2. Daily discharge from the
M.G. Lawrence, D.G. Bariel / Chemosphere 80 (2010) 794–799
Wetalla WWTP describes a consistent weekly pattern with an average concentration of 58 ± 23 ng kg1 throughout the time of the sampling (May–October 2009); the weekly pattern remains, even when concentrations are converted to loads (g Gd d1, data not shown). The strong weekly cycle evident in the data is consistent with the interpretation that the predominant application of MRI contrast agents occurs during the working week (Monday–Friday in the private imaging clinic), and is consistent with previous research (Knappe et al., 2005) in Berlin. In the case of Berlin (population 3.5 M), a total daily discharge of 178 g Gd was calculated, which on a population basis equates to 5 g Gd d1 per 100 000 persons. In contrast, the Wetalla WWTP, serving a population of 96 000 people has an average daily discharge of 0.84 ± 0.4 g Gd, a comparative load of 0.9 ± 0.4 g Gd d1 per 100 000 people. The recommended current dosage for both Gadovist and Magnevist, two of the common contrast agents, is 0.1 mmol kg1 body weight, which for the average 70 kg Australian adult equates to 1.1 g Gd. The discharge load, within error and assuming no degradation within the WWTP, is approximately the equivalent of one patient receiving contrast per day. In our previous research, in the more heavily populated SE-Qld region (Lawrence et al., 2009), a major WWTP was calculated to discharge 3.3 g Gd d1 per 100 000 people, however, most other WWTPs were discharging a similar load to Wetalla WWTP. Thus the lower loads in the regional centre probably reflect the lower density of MRI imaging technology (there are two MRI units known to be in operation in the catchment of this WWTP), and possibly the ongoing trend towards the application of lower concentrations of contrast agents. 8.2. River water samples within catchment Water samples were taken on three different occasions in May, July, and September, spanning the catchment from 200 m upstream of the WWTP discharge location to Loudouns Weir, which is used as a water supply for the Dalby community, approximately 100 km downstream (e.g. Fig. 3). There had been a significant rainfall event ( 75 mm 9-d prior to sampling, resulting in a tripling of WWTP discharge) in the catchment prior to the May sampling campaign – and the flood peak had not passed Loudouns Weir completely at the time of sampling; there was no other significant rainfall in the lead-up to the following sampling campaigns (a 40 mm rainfall event occurred 21 d prior to the final sampling campaign, but given the dry ground conditions, does not appear to have resulted in any run-off, and WWTP discharges remained normal in the days surrounding the event). On any one sampling campaign, up to 15 sample locations were sampled with the aim of sampling the main waterway and any tributaries that were flowing into it (not all tributaries continued to flow over the course of the sampling). As expected, and demonstrated in Fig. 3, there was no anthropogenic Gd detected above the WWTP inflow at any time. In contrast, the WWTP itself and the sample 500 m downstream from the WWTP both have large positive Gd anomalies. When the average WWTP concentration of 58 ng kg1 Gd is assumed to be the endmember, the concentrations measured at location B in May and July indicate that there is 107–113% wastewater present at this location. However, for this location, rather than the average discharge, the best value for the end-member is more likely to be the instantaneous sample collected from the WWTP just minutes prior to collecting from location B. On the first sampling campaign, the WWTP was discharging an [GdAnth] of 76 ng kg1 and on the second campaign the instantaneous WWTP [GdAnth] was 0.78 ng kg1. If these values were instead chosen to calculate the percentage wastewater present, we would estimate that wastewater contributed 81.5% and 83.2% of the total flow on these occasions. Whist there were no gauging stations above the WWTP to
797
measure the actual flow, these calculations are consistent with visual observations of the WWTP discharge, where there was clearly more water entering the creek from the WWTP. Whilst the instantaneous measure of [GdAnth] is the most logical value to choose when estimating the percentage of wastewater in a sample close to the WWTP, this is clearly not the best end-member to choose when traveling over 100 km downstream. In that case, where hydrological data would be required to estimate which water packet from the WWTP had mixed with the creek, the average value is instead chosen (as indicated by the flood peak not having completely passed Loudouns Weir 9 d after the rainfall event, the hydraulic residence time, whilst affected by volume, is therefore at least 9 d). As the average end-member concentration varies by ±40%, the estimates of the percentage wastewater are also subject to the same uncertainty. Therefore, we quote the absolute percentage wastewater detected assuming an average discharge, with the proviso that the actual concentrations could vary by half the value stated (i.e. a value of 20% could actually range from 12% to 28%). Nonetheless, for the final sampling location, Loudouns Weir near Dalby, the weir will increase the residence time of the water, and if the residence time is >1 week (the time for one concentration cycle at the WWTP), the average end-member value is the most appropriate, and our estimates of the percent wastewater present are likely to be more accurate within the stated assumptions. Over all sampling campaigns, the percentage wastewater follows the same general pattern, decreasing as the distance from the WWTP increases. As the first sampling campaign in May occurred 1 week after a minor flood event, the entire system had been flushed, and the detection of 5% wastewater at location ‘‘M” would be considered to be proof of a significant amount of wastewater being present. However, there appears to be an anomaly during this sampling campaign, with the largest detected GdAnth component in Oakey Creek, with a calculated [GdAnth] of 100 ng kg1. Clearly the assumption that the Wetalla WWTP is the only source of GdAnth is incorrect. Unfortunately, as a result of the second point source, the calculated result of 5% wastewater at location M in the first sampling campaign is unreliable, as it is possible that we have instead encountered the very high concentration end-member from the Oakey WWTP or a leaking septic system, not the expected Wetalla WWTP signal. The township of Oakey has a small WWTP with an average daily discharge of 700 kL d1. As Oakey is close (30 km) to Toowoomba, we can only assume that a patient traveled to Toowoomba for an MRI, then traveled home prior to excreting the contrast agent. This example demonstrates that MRI itself does not have to be present at a particular location, only that advanced medical imaging is available within the excretion half-life of the contrast agent. Interestingly the detection of such a high concentration is not surprising – for example, if the entire load of 1 dose (1.1 g) of contrast agent was to be introduced into the Oakey WWTP (700 kL) over 1 d, a concentration of 1570 ng kg1 would have been expected in the effluent. It is also interesting to speculate on how commonly an GdAnth spike would occur in Oakey Creek. On a population basis, Toowoomba (96 000) is 27 times larger than Oakey (3500). On average 0.9 people are treated in Toowoomba per day, therefore by extrapolation, one person in Oakey is likely to be treated every 30 d. Our sampling campaign was therefore fortunate (or unfortunate) to capture this event, but it does indicate that caution must be applied when assessing wastewater contributions with single tracers over such large catchments. Whilst we could not detect wastewater at Loudouns Weir on the first sampling campaign, this was not unexpected, as we actually ‘‘caught up” to the river peak from the rainfall event 9 d prior to sampling, with water overtopping the weir – thus the extensive
798
M.G. Lawrence, D.G. Bariel / Chemosphere 80 (2010) 794–799
Fig. 3. A–C: Maps indicating the calculated percentage of wastewater at each location for sampling campaigns in 2009, assuming that the Wetalla WWTP is the only source of GdAnth, and that the average wastewater discharge concentration is 58 ng kg1 Gd (*indicates that these samples were calculated from the instantaneous discharge of Gd; **calculated using the average Wetalla WWTP discharge, which is not applicable for these samples). Note that Oakey Creek always presents with a Gd anomaly, which is consistent with the presence of a small (700 kL d1) WWPT that discharges into this creek.
M.G. Lawrence, D.G. Bariel / Chemosphere 80 (2010) 794–799
dilution would have dropped the wastewater percentage below the detectable level. Second sampling campaign: Results for the second campaign indicate that high percentages of wastewater are detectable over a large area of the catchment, for example, at location H, approximately 40 km down river, the wastewater contribution was still calculated at 76 ± 30%. At location J, approximately 60 km from the wastewater discharge, we estimate that 19 ± 8% wastewater is present. Such an outcome is very interesting from a policy perspective. Under the current legislation, the wastewater treatment plant would not be able to distribute the effluent water for reuse as irrigation water for cropping, yet, at these locations where 19–75% wastewater is present in Gowrie Creek, the history of the water is unimportant. The question then becomes why was the wastewater considered a risk in the first instance, and has transport down the natural system mitigated this risk? A similar situation was observed during the third sampling campaign – but a lower wastewater contribution was detected at the mid-catchment locations (typically 15%). However, on this sampling campaign, GdAnth was detected at Loudouns Weir with 3 ± 1.2% wastewater detected. This value is lower than estimated by Hamilton and Greenfield (1991) (8.3% at low flow), but is within the same order of magnitude as their estimate. At the time of sampling, Gowrie Creek was drying significantly, and Lagoon Creek, Westbrook Creek, and the Condomine River were not flowing (no samples taken). Thus it would be expected that samples collected at this time were representative of a typical dry season. While we believe that the evidence of Morteani et al. (2006) and Kulaksiz and Bau (2007) is strong, it must be pointed out that the assumption that GdAnth behaves conservatively in this type of surface water environment (i.e. shallow, slow flowing, with intense solar radiation) over the time scale of transport has not yet been thoroughly tested in the literature. If environmental degradation does occur, the estimates of the extent of wastewater contamination calculated must be viewed as the minimum levels, although our estimates could be recalculated by modeling a decay rate, and estimating the hydraulic residence time.
9. Conclusions We have demonstrated that there is: A strong weekly cycle in the discharge of Gd into the environment from a typical WWTP. An average of 58 ± 23 ng kg1 of GdAnth discharged from the Wetalla WWTP over 5 months of sampling. GdAnth was easily detected in receiving waters >50 km downstream of the WWTP source. Up to 3% wastewater was detected up to 100 km away from the point source.
Acknowledgements MGL wishes to thank the Toowoomba Regional Council, and the operators of the Wetalla WWTP for their invaluable assistance, and acknowledges the financial support (salary) of the Chair in Water
799
Recycling Agreement between Water Secure, Veolia Water and the University of Queensland. References Aries, S., Valladon, M., Polve, M., Dupre, B., 2000. A routine method for oxide and hydroxide interference corrections in ICP-MS chemical analysis of environmental and geological samples. Geostand. Newslett. – J. Geostand. Geoanal. 24, 19–31. Bau, M., Dulski, P., 1996. Anthropogenic origin of positive gadolinium anomalies in river waters. Earth Planet. Sci. Lett. 143, 245–255. Bau, M., Knappe, A., Dulski, P., 2006. Anthropogenic gadolinium as a micropollutant in river waters in Pennsylvania, and in Lake Erie, northeastern United States. Chem. Erde-Geochem. 66, 143–152. Eggins, S.M., Woodhead, J.D., Kinsley, L.P.J., Mortimer, G.E., Sylvester, P., McCulloch, M.T., Hergt, J.M., Handler, M.R., 1997. A simple method for the precise determination of >=40 trace elements in geological samples by ICPMS using enriched isotope internal standardisation. Chem. Geol. 134, 311–326. Hamilton, G., Greenfield, P.F., 1991. Potable reuse of treated wastewater. In: Australian Water and Wastewater Association 14th Federal Convention. Australian Water and Wastewater Association, Perth, pp. 497–506. Kamber, B.S., Greig, A., Collerson, K.D., 2005. A new estimate for the composition of weathered young upper continental crust from alluvial sediments, Queensland, Australia. Geochim. Cosmochim. Acta 69, 1041–1058. Knappe, A., Möller, P., Dulski, P., Pekdeger, A., 2005. Positive gadolinium anomaly in surface water and ground water of the urban area Berlin, Germany. Chem. ErdeGeochem. 65, 167–189. Kulaksiz, S., Bau, M., 2007. Contrasting behaviour of anthropogenic gadolinium and natural rare earth elements in estuaries and the gadolinium input into the North Sea. Earth Planet. Sci. Lett. 260, 361–371. Kümmerer, K., Helmers, E., 2000. Hospital effluents as a source of gadolinium in the aquatic environment. Environ. Sci. Technol. 34, 573–577. Lawrence, M.G., Greig, A., Collerson, K.D., Kamber, B.S., 2006a. Direct quantification of rare earth element concentrations in natural waters by ICP-MS. Appl. Geochem. 21, 839–848. Lawrence, M.G., Greig, A., Collerson, K.D., Kamber, B.S., 2006b. Rare earth element and yttrium variability in South East Queensland waterways. Aquat. Geochem. 12, 39–72. Lawrence, M.G., Jupiter, S.D., Kamber, B.S., 2006c. Aquatic geochemistry of the rare earth elements and yttrium in the Pioneer River catchment, Australia. Mar. Freshw. Res. 57, 725–736. Lawrence, M.G., Ort, C., Keller, J., 2009. Detection of anthropogenic gadolinium in treated wastewater in South-East Queensland. Water Res. 43, 3534–3540. Melbourne Water, 2009. Water System Storage. Melbourne.