Evaluation of DGT as a long-term water quality monitoring tool in natural waters ; uranium as a case study

The performance of the diffusive gradient in thin film technique (DGT) was evaluated as a tool for the long-term monitoring of water quality, using uranium as a case study. DGTs with a MetsorbTM (TiO2) sorbent were deployed consecutively at two alkaline freshwater sites, the River Enborne and the River Lambourn, UK for seven-day intervals over a five-month deployment period to obtain time weighted 10


Introduction
Currently, monitoring of water quality relies on the collection of low volume spot (grab or bottle) water samples, usually on monthly, or at most weekly, time intervals.This approach has a number of limitations, being both expensive and time consuming, the possibility for introducing contamination in sample handling or during storage 1 and the potential to miss fluctuations in contaminant concentrations.For analytes having low aqueous concentrations, such as radionuclides, often large volumes (5-20  L) of water need to be collected and pre-concentrated to ensure good instrumental limits of detection. 2To overcome some of these drawbacks, continuous in field auto-samplers 3 (active samplers) that are programmed to collect samples at set time intervals or during particular flow or meteorological conditions can be used. 4This approach is costly and can also be associated with errors in terms of sample stability for monitoring both metals and nutrients. 5,6he use of in situ pre-concentration techniques, such as passive sampling devices, can overcome many of these errors in associated with spot sampling 7 and can be beneficial in 45   investigations where concentrations of a pollutant fluctuate widely, for instance from increased surface water flow as a result of a storm event, or with large tidal fluctuations. 7,8Passive samplers have the advantage of being relatively low-cost, nonmechanical, require no power and little maintenance and can be 50 deployed in a range of field sites.
Designs of passive sampler are varied and have been developed to measure a wide range of organics and metals.Examples include the Gaiasafe, 9 Chemcatcher ® for both metals, 10 organics 11 and organometallics, 12 permeable liquid membrane 55 devices 13,14 and diffusive gradients in thin films (DGT) 15 .DGT measures the labile, dissolved fraction of analytes in situ and is the most widely used technique for measuring time weighted average (TWA) concentrations of a number of metals and inorganic substances in a variety of aquatic environments.The extract analysed by a sensitive instrumental technique e.g.inductively coupled plasma-mass spectrometry (ICP-MS).
DGTs have been used for monitoring metals in the aquatic environment in a number of single short-term deployment studies (e.g. 4 d 16 , 14 d 17 and 31 d 18 ).DGTs have also been deployed in the same location during two seasons with longer-term deployment periods (ranging from 13 to 36 d 19 ) to show interseasonal variations of pollutants in the Sava River, Croatia.DGTs were also used for one-month deployments over five consecutive months 5 in Lake Llyn, Trawsfynydd, UK.The concentration of metals in highly fluctuating, transitional environments, such as estuaries, have been monitored using DGT in short-term studies 7,20,21 .Dunn et al. 22 showed that in highly fluctuating environments concentrations of metals can change significantly over 24 h and that these variations would therefore be missed by the use of infrequent spot sampling.There is little published data for freshwater systems, however, on the effects of long-term environmental changes (for instance seasonal changes in biological activity and water chemistry and flow rate) on the operational effectiveness of DGT devices.If DGT is to be used by regulatory agencies and to be a fit for purpose monitoring tool, further long-term field testing is required in conjunction with recognised standards such as the ISO 5667 1 .In an attempt to investigate this, we used DGTs to monitor the concentrations of uranium continuously over a six-month period at two freshwater sites (River Enborne and the River Lambourn, Berkshire, UK) and compared the results against those from weekly spot water sampling.The purpose of this study was to therefore evaluate the usefulness of the DGT technique and to assess any issues (such as the measurement of the DBL, changing river chemistry and seasonal changes in biological activity) that could arise as part of its use as a regulatory environmental monitoring tool.The two river sites were chosen, as they were included in a routine environmental monitoring programme undertaken by CEH.This provided weekly data to aid the interpretation of the DGT results.
Both sites were also secure and located on private property, which ensured no interference to the devices over the deployment period.Uranium has a complex aqueous chemistry and was therefore selected to demonstrate that the DGT technique can accumulate a highly reactive analyte in a system with fluctuating water quality.
Uranium is not a priority substance in the European Union's Water Framework Directive 23 due to the high concentrations that occur naturally.Environmental monitoring of anthropogenic and naturally occurring radionuclides in natural waters is a requirement of the environmental permits issued by the various environment agencies in the UK, and by the Industrial Pollution and Radiochemical Inspectorate for all users and holders of radioactive materials, under the Environmental Permitting Regulations (England and Wales) 2010 and Radioactive Substances Act 1993. 24These permits require the nuclear industry to continually undertake risk assessments of their discharges to ensure environmental impacts are as low as is reasonably practicable. 25This includes considering the use of new monitoring technologies such as DGT.7][28][29][30][31] There are a number of candidate binding phases effective for uranium.The TiO 2 -based resin, Metsorb™ used in this study showed a high capacity for uranium. 17The mean flow and base flow indices were 1.71 m 3 s -1 and 0.97 respectively. 33The River Enborne drains impermeable tertiary sand, silt and clay deposits 34  (Table S2) and from daily measurements taken by the Centre for Ecology and Hydrology (CEH), Wallingford, UK (51.6032N, -1.1134 W) using a ground flush type rain gauge (Figure S1).

Design of field trial
A continuous monitoring programme was used to assess the performance of DGTs over part of three riverine seasons, from summer through to autumn and winter.During these periods it was expected a wide variation in biological activity, flow regime and water chemistry would occur.DGTs were deployed between Perspex plates (15 x 7 cm, up to 8 devices per plate) (  To assess the influence of the DBL on the uptake of uranium, devices containing Metsorb™ were also deployed, with diffusive layer (polyacrylamide (PAM) gel) thicknesses (including 0.015 cm to account for the Supor membrane) of 0.015, 0.055, 0.095 and 0.135 cm, as per Warnken et al. 35 S1).Water quality analysis was undertaken at CEH laboratories (see procedures in SI and Table S1).Discharge data for each site was obtained from the CEH National River Flow Archive (Figure S1), where measurements were taken at the crump weir located 51.3791 N, -  36 The diffusive gels and filter membranes were stored in 0.01 M NaNO 3 prior to 90 deployments to ensure ionic equilibrium between the diffusive gel and the deployment environment.The PAM binding gels were prepared with 1 g Metsorb™ HMRP powder (TiO 2 with an organic binder, < 50 µm; Graver Technologies, Glasgow, USA) according to the method described by Bennett et al. 37 A disk of 95 (0.2 µm pore size) Supor polyethylene sulfone (Pall Corporation, Portsmouth, UK) that was first acid washed in 1% HNO 3 , tripled rinsed in Milli-Q water and stored in 0.01 M NaNO 3 was used as the outer membrane.DGT mouldings were obtained from DGT Research Ltd. (Lancaster, UK) and washed for 24 h in 10% 100 HNO 3 , and then rinsed three times in Milli-Q water prior to use.The devices were assembled according to Davison et al. 15 and stored at 4 o C in zip lock plastic bags, containing 1-2 mL of 0.01 M NaNO 3 in Milli-Q water (ionic strength matched to freshwater deployment site) to ensure the diffusion properties of the gels 105 were not altered, and to prevent the gels drying out.

Measurement of total uranium
This journal is © The Royal Society of Chemistry [year]   Uranium was determined in all solutions by ICP-MS using an Agilent 7500ce series instrument (Agilent Technologies Inc., Japan).Total uranium was measured under normal plasma conditions in 'no gas mode', with the sample introduction system fitted with a micromist nebuliser.The instrument blank for uranium was 6 ng L -1 while the limit of detection (calculated by the Agilent Chemstation software) for uranium was 2 ng L -1 , with a measurement relative standard deviation better than 3%.Bismuth (m/z = 209; 25 µg L -1 ) was used as an internal standard to compensate for any potential instrument drift.The certified 10 fluvial reference material SLRS-5 (National Research Council Canada, Canada) was analysed directly for uranium and found to be within 1% of the stated values.The filtered and acidified spot water samples were analysed directly with no further dilution.

Measurement of uranium in DGT
After exposure, the Metsorb™ binding gels were removed from the DGT and eluted (48 h) with 1 M H 2 O 2 /HNO 3 (2 mL) solution (100 mL made by combining 90 mL 1.1 M HNO 3 and 10 mL H 2 O 2 ).The eluent was then diluted 10 fold with Milli-Q water 20 prior to instrumental analysis.The concentration of uranium (µg L -1 ) measured by the ICP-MS in the eluent was multiplied by the dilution factor (×10) to give the actual uranium concentration (Ce).The absolute mass (M, ng) of the uranium in the binding gel was calculated using equation 1, where M is calculated taking into account the gel volume (Vg, cm 3 ), the eluent volume (Ve, mL), the measured concentration of uranium in the eluent (Ce, ng mL -1 ) and the elution factor (fe). 36 For this study the uptake (>90%) and elution factor (83±3%.) for U were taken from Turner et al. 17 M from equation 1 is then used to calculated the TWA concentrations (equation 2) where the concentration (C DGT , ng mL -1 ) was calculated using the mass of the analyte in the binding gel (M, ng), the thickness of the diffusive path length (diffusive gel and filter membrane) (Δg, cm), the diffusion coefficient of the analyte (D, cm 2 s -1 ) (as determined for uranium at different pH's by Hutchins et al. 30 ), deployment time (t, s) and the area of the sample exposure window (A, cm 2 ).
(2) 40   The DBL (δ) thickness was calculated using equation 3 after Warnken et al. 35 A straight line plot of 1/M versus Δg has a slope (m) of 1/(D C DGT A t) and an intercept (b) of δ/(D C DGT A t).The intercept (b) divided by the slope (m) of this plot gives the DBL thickness δ, as per equation 4. The diffusion coefficients of the uranyl ion in the diffusive gel and the water have a ratio of nearly one, 38 and so do not need to be considered for the purposes of this paper.
(3) 50 (4)   The thickness of the DBL was included in the C DGT calculations for the field trials.The DBL measurements were applied to the weekly DGT field data as follows: DBL data point 1 was applied to weeks 1-7 DGT data; DBL data point 2 was 55 applied to weeks 8-15; DBL data point 3 was applied to weeks 16-19; and DBL point 4 was applied to weeks 20-21 DGT data.An average of all four DBL readings per river was used when an average DBL was applied to the DGT data.The active sampling area (A) was 3.8 cm 2 instead of the 3.14 cm 2 used in the 60 laboratory trials, as described by Warnken et al. 35 to account for lateral spread of the analyte across the surface of the DGT device.The diffusion coefficients from Hutchins et al. 16 were used for the TWA calculations and corrected for temperature as per Zhang and Davison. 36Laboratory blanks were measured in triplicate and 65 the average concentration per disk was determined for the Metsorb™ gel disks as 0.03 ± 0.02 ng and 0.30 ± 0.10 ng for 238 U and 235 U respectively.

Measurement of uranium isotopes
235/238 U isotopic ratios were measured with an Agilent microflow 70 (100 µL min -1 ) PTFE self aspirating nebuliser, to eliminate any signal pulses caused by the peristaltic pump using a micro-mist concentric nebuliser.Isotopic ratios were determined with 3% RSD as low as 0.01 µg L -1 total uranium (0.725 x 10 -4 µg L -1 235 U).The certified reference material U005a (New Brunswick 75 Laboratories, DoE, Washington, USA) was analysed and was found to be within 99.5% of the isotopic value (0.342 x 10 -4 235/238 U).The spot water samples were measured directly without any further dilution.For the isotopic signature of uranium found with DGT, the extract was diluted 10 fold prior to analysis.

Statistical analysis
The water quality results (including the weekly spot water sample measurements) were averaged over each week (mean of the reading at the beginning and at the end of each deployment week) 85 and then subject to statistical analysis to identify any patterns between the two different techniques used to measure the uranium concentration and fluctuating water quality.All statistical analysis was performed in IBM ® SPSS ® Statistics Version 20.The non-parametric one sample Shapiro-Wilk test 90 was first used to test the data for normality (normality significance figure ≥ 0.05).If normality was established a Pearson's product-moment correlation was performed, if the data was not normally distributed then the non-parametric Spearman's ranking correlation coefficient was used (P < 0.05).

Results and discussion
Water flow rate was measured at each deployment site (Table S3) to investigate if this may affect the thickness of the DBL.Flow rates were also back calculated from the discharge data (Figure 100   S2).The DBL has been shown previously to be an important factor in the accuracy of the DGT technique in measuring TWA concentrations.Without the inclusion of the DBL in calculations, concentrations can be underestimated by up to 50%. 17 The TWA concentrations of uranium found with the DGTs were compared to the weekly averaged water quality results to determine if any statistically significant relationships existed.The mean U concentrations determined both by DGT and in the spot samples (0.26-0.38 µg L -1 ) are in line with those reported previously 17 and by CEH (Mike Bowes pers com) who reported values of 0.3 µg L -1 .The U concentrations measured in this study are in line with background U concentrations and are not particularly elevated.

DBL measurements
Several factors can affect the thickness and measurement accuracy of the DBL.These include fluctuations in water velocity, 35 the deposition of particulate matter, bio-fouling by macro-fauna and the growth of bacterial mats 39 on the active 15 sampling surface and the dissociation kinetics of organically bound metals at the solute interface of the sampler. 40,41 4b, as the calculated TWA concentration for uranium is up to 58% less than in the River Enborne (particularly when the 65 calculated DBL was higher) than measurements that account for the periodically measured DBL (Figure 3c and 4c).For the River Lambourn there was an underestimation of the TWA concentration of uranium by up to 57% when no DBL is accounted for in the calculations, with the TWA calculations 70 using the averaged DBL over the deployment time (Figure 3a) and the periodically measured DBL (Figure 3c) within ± 20%.

Effect of water flow rate on the thickness of DBL
During the first 4 months of the deployment (August to late 75 November 2011) the River Enbourn experienced below average precipitation (Table S2) in conjunction with lower flow rates (Table S3 and Figure S2) and discharge (Figure S1), and consequentially a larger DBL thickness of 0.141 ± 0.036 cm (Table1) was measured.The flow rate in September and October 80 2011 were calculated to be ≤ 2 cm s -1 (Figure S1 and S2), with the river flow where the samplers were sited likely to be even lower, as this was located outside the main channel.The sustained above average precipitation from the second week in December 2011 (Figure S2) increased the discharge and reduced the thickness of 85 the DBL to 0.086 ± 0.019 cm in December and then to 0.047 ± 0.008 cm in January 2012 (   at the River Enborne over the deployment period with a potential difference in river height of up to 1.2 m.This demonstrates the need to fully characterise the attributes of a field site prior to deployment, to ensure the devices remain submerged in a reasonably turbulent environment and are always retrievable. It is clear from the field measurements of the DBL at the River Enborne (Table 1) that the changing DBL was closely coupled to the flow rate (the 1/M v Δ g plots used for each of the DBL measurements can be seen in Figure S5a-d and the flow rate in Figure S2).As only four DBL measurements were taken over the deployment period, it was not possible to perform any statistical tests.A simple correlation could be undertaken and graphed to show that the flow is linked to the size of the DBL, as shown in Figure S4 in the supplementary information.This shows a decrease in DBL thickness with increasing flow rates over the 6 month deployment period in this study.The River Enborne shows a clearer variation in DBL thickness with flow rate than the River Lambourn most likely due to the fact that the River Enborne has a highly fluctuating flow regime.The very large DBL observed in October 2011 (Table 1), when the flow rate of the River Enborne was very low, is concurrent with that found under a laboratory setting by Warnken et al. 35 in quiescent conditions, where a large DBL of 0.15 ± 0.013 cm was observed (Table 3).Under laboratory conditions in previous studies, moderate flow rates up to 2 cm s -1 showed a reduction in the associated thickness of the DBL, with Warnken et al. 35 reporting a value of 0.044 ± 0.0014 cm, which is similar to the thickness of the DBL found in this study for the January 2012 deployments in the River Enborne.If the flow rate exceeds 2 cm s -1 (as for well stirred solutions)

55
then it has been shown that the thickness of the DBL is not directly related to the flow rate of water. 35,42Warnken et al.   found for high flow rates, in a laboratory setting, the thickness of the DBL was 0.023 ± 0.0032 cm, which is in agreement to the DBL thicknesses (0.024 ± 0.002 cm) found by Scally et al. 43 The flow rate in this study frequently exceeded 2 cm s -1 in the River Enborne, but the lowest measured DBL was 0.046 cm, which implies other factors than flow rate may contribute to the DBL.The River Lambourn showed less variability in the thickness of the DBLs (Table 2 and Figures S6a-d) most likely as a result of the discharge remaining at a steady state over the course of the deployment period (Figure S1).The flow rate for the River Lambourn over the deployment period averages 8 cm s -1 (Table S3 and Figure S1 and S2), which is higher for most of the deployment period than in the River Enborne.The consistent and high (despite low precipitation) flow rates experienced by the River Lambourn is due to the chalk catchment and the fact the river catchment is largely ground water fed.The DBL found in October was 0.070 ± 0.022 cm, which is higher than predicted in the laboratory (Table 3) for the flow rate.Over the course of the deployment period, the thickness of the DBL increased to 0.088 ± 0.009 cm in January 2012, and decreased to 0.062 ± 0.018 which is up to two times that measured in the River Enborne and nearly four times that measured under laboratory conditions.Figure S4 does not indicate that the DBL in the River Lambourn is flow rate controlled.The flow rate therefore does not give a good indication of the thickness of the DBL in the River Lambourn, which means extraneous factors (such as biofouling) must be also taken into consideration.
DBL measurements in the field have been shown to differ significantly from those on the laboratory.Table 4 shows the thickness of DBLs found in the field, although there is a paucity of data.In a well-stirred field environment, Warnken et al. 35 found the measured the thickness (0.026 ± 0.0017 cm) of the DBL closely matched their laboratory results.Thicker DBLs in the field have been reported, by Panther et al. 44 (0.080 ± 0.013 cm for PO 4 ) and Bennett et al. 37 (0.080 ± 0.013 cm for As and Se).Hutchins et al. 16 reported a DBL thickness of 0.02 ± 0.001 cm when measuring concentrations of uranium in a freshwater system.Another consideration when comparing the thickness of DBLs found here to other field studies is the length of time the devices were deployed.DGTs are usually deployed for shorter periods (3-5 d) when examining properties of DBL.In this study, the deployment time was 7 d.A longer deployment is favourable when measuring low concentrations (ng L -1 ) of a pollutant, as this allows more of the analyte to accumulate onto the resin, however, other factors e.g.biofouling may begin to dominate the uptake process.Warnken et al. 35 suggested that when flow exceeds the 2 cm s -1 threshold, then the DBL thickness (present at 0.023 cm) could be discounted.Here a sampling area of 3.14 cm 2 can used (as opposed to 3.8 cm 2 which accounts for lateral diffusion at the DGT face) to offset the error when not accounting for the DBL, and when using a gel thickness of 0.8 mm.However, as is observed here and in other field studies (Table 4), there may be other factors influencing the thickness of the DBL than simply water flow rate.The major contributor to the thickness of the DBL is the flow rate, however, when the flow rate is decreased other influences including the effect of particulates, biological activity and dissolved organic material were found to play an increasing role but their effects are masked by the influence of high flow rate on the DBL.60

of DBL
Previous work has shown that biofouling and turbidity 35 can have an impact on the effectiveness of passive sampling devices.The River Enborne contained higher and fluctuating concentrations of suspended particulate material (SPM, mg L -1 ) than the River 70 Lambourn (Figure S7).However, when plotted using a scatter graph, no clear trend was apparent.Particulates could potentially act to increase the thickness of the DBL by acting as an additional physical barrier to diffusion across the filter membrane or by supplying a source of dissociating uranium from particulate 75 surfaces.At the diffusive interface (the surface of the filter membrane) where a concentration gradient will be present, there may be a resupply of uranium sorbed to the surface of the suspended particulates.Previous studies showed the presence of organic material in a river increases the sorption of uranium to 80 particle surfaces. 46This is supported by the fact that when the devices were retrieved, there was particulate matter collected on the active sampling surface (Figure S8).Supor ® membranes are designed to inhibit microbial growth.However, if SPM accumulated on the surface of the membranes then this will 85 provide sites for growth, with a microbial matt developing and potentially acting as a sink for the uranium. 39This could account for variability in the measurements on the thickness of the DBL depending on the depth of the microbial mat, but is an area for further work.The lower values of SPM found for the River 90 Lambourn meant that this process may not be a contributing factor to the DBL.DGTs deployed in River Lambourn accumulated algae and macro-flora over the 7 d deployment.Previous work by Turner et al. 17  0.046 ± 0.006 cm.However, rapid accumulation of macro-flora (Figures S9a-c) resulted in high variability of the DBL over the deployment period including large DBL's with associated errors (Table 2).There was little variation in flow rates at this site (Figures S1 and S2), due to its high base flow index (0.97); hence any variation occurring in the thickness of the DBL could be attributed to a biological source.Dragun et al. 19 also found limitations on the effectiveness of the DGT due to algal biofouling during long-term (13-36 d), single deployments during the spring.Ideally, DGTs should be deployed in a protective cage 10 in areas prone to the build-up of algae and macro-flora, although this would not prevent the accumulation of periphyton on the surface of the devices.This is an area for further research, as care should be taken not to reduce the water flow inside the cage.

Effects of dissolved organic matter and water quality on the thickness of the DBL
DBLs are both a physical layer where advective transport moves to diffusional transport processes, and/or an apparent layer of chemical dissociation of the analyte from a larger molecule such 20 as dissolved organic matter. 41,47Levy et al. 47 showed that in the presence of organic ligands, metals demonstrated varying degrees of kinetic limitation dependent on dissociation rates, and therefore exhibited varying apparent diffusive boundary layer (ADBL) thicknesses.The possibility of the presence of a zone of 25 chemical dissociation cannot be ignored in the case of uranium.This is due to its high affinity towards dissolved organic matter; 48 particularly when over 90% of the uranium species modelled (using Visual Minteq) were found as humic complexes (fulvic and humic acids) for the River Enborne, and ~50% of the uranium bound to humates in the River Lambourn.Figure S9 shows that the River Enborne contains up to ten times more dissolved organic carbon (DOC) than the River Lambourn, thereby affecting uranium speciation.The DOC concentration in the River Enborne increased during periods of increased precipitation due to its susceptibility to the influence of catchment run off.Warnken et al. 41 showed that the ADBL increased with metals that formed increasing strong complexes with dissolved organic matter.Uranium at low uranium:humic acid (U;HA) ratios (such as for the Rivers Enborne and Lambourn with U:HA ratios of 4.17 x 10 -5 and 1.81 x 10 -4 respectively) forms very strong humic acid complexes that have slow dissociation kinetics (k d = 4.9 x10 -5 s -1 ) compared to higher U:HA ratios (i.e.> 0.01) (k d = 10 -3 s -1 ). 49These slow dissociation kinetics may have affected the thickness of the DBL for both rivers, although this would require further studies in both the field and laboratory settings to confirm.This potential zone of dissociation may account for the presence of an extended DBL (Table 1) in the River Enborne even during periods of high flow and discharge, where the thickness of the DBL was 0.037 cm and 0.047 cm, compared to 0.023 cm in a fast moving system under laboratory conditions (Table 3).However, when plotted using a scatter graph (Figure S11), no clear trend was apparent DOC and DBL for the River Enborne, potentially because there are other stronger influencing factors such as flow rate, that make the impact of the DOC indistinguishable.Figure S11 shows a clear trend of increasing DBL thickness with increasing DOC concentrations.This may be because factors that have a greater influence on the DBL thickness such as flow rate and inorganic ligands (e.g.phosphate) are consistent over the deployment 60 period.Further work would be required to establish the relationship between the DBL and DOC when measuring uranium.
Another interesting correlation was that of phosphate and the size of the DBL.In both rivers a positive correlation was 65 observed when the DBL was plotted against the phosphate (Figure S12) this correlation being highly significant for the River Enborne (R 2 = 0.8285), which may be due to the agricultural catchment has fluctuating phosphate concentrations with run off after precipitation events, similar to that found by 70 Evan et al. 50.Further work would be required to confirm this, but the presence of phosphate and SPM may act as both source and sink of uranium on the surface of the DGT devices, thereby increasing the thickness of the DBL, acting as zone of association/dissociation.

Calculation of TWA concentrations
The TWA concentrations of uranium were calculated using varying scenarios (Figures 3 and 4), (a) the average thickness of the DBL measured over the entire deployment period; (b) not 80 accounting for a DBL; and (c) using the changing thicknesses of DBLs measured during the trial.The parameters e.g.water pH and temperature and diffusion coefficient used in these calculations are given in Table S4. Figure 3 and 4 shows that TWA concentrations generally fall between weekly spot 85 sampling data points.This was evident when there were rapid, short-lived, increases in the concentration of uranium during weeks 2 and 6 for the River Enborne, and weeks 2, 6 and 7 for the River Lambourn.During periods of relatively stability, the concentration of uranium measured in spot waters samples 90 (weeks 20-22, River Lambourn and River Enbourn; and weeks 15-16, River Enborne) corresponded well with the TWA concentrations found with the DGT.This shows the effectiveness of the DGT in measuring accurately, fluctuating concentrations, despite the difficulties of predicting the thickness of the DBL.

95
The only anomaly within the data is Week 1, which shows a much higher spot sample concentration to the TWA DGT concentration at both rivers.This may be attributed to either a high SRP concentration, very low flow and low precipitation or the use of a DBL that was determined a number of weeks after 100 this deployment.However, these are all unknowns, but again this highlights the need for the DBL to be determined regularly in a water body that has fluctuating flow and water chemistry and also the need for a toolbox approach to environmental monitoring without the reliance on one technique.Murdock et al. 5 attempted 105 to validate DGT as an in situ tool for measuring caesium.They found that over the 5-month study, both the concentrations of caesium measured by the DGT and in spot water samples were in close agreement, being within the 1 σ margin of error.As there was close agreement between the spot sample and DGT TWA 110 concentration the DBL thickness which was not measured in this study was deemed an unimportant parameter.The study was undertaken in a lake with little variation in flow and there was a constant input of caesium from the Magnox reactor sited there.They found that the longer the deployments, the more bio-fouling, susceptibility to changing flow rates, and saturation of the binding phase.Mengistu et al. 18 used DGT as a risk assessment tool, and undertook a single 31 d and a single 3 d deployment to measure seventeen metals (including uranium) in water polluted by mining tailings.They found 1-2 orders of magnitude reduction in the mass of metals accumulated in the DGT during the long-term deployments compared with the shortterm deployments.Turner et al. 17 found decreased uptake by DGT after 7 d, due to bio-fouling and saturation of the binding phase.For this reason, 7 d was chosen as the deployment period in this study.
DGT has been to measure other analytes in highly fluctuating environments, such as estuaries. 20,21,51Montero et al. 20 deployed DGTs for 10 d in 13 estuaries draining into the Bay of Biscay and found a good correlation with previously measured concentrations of cadmium, copper, nickel and zinc using spot water samples.Dunn et al. 8 used DGT to examine the effect of tidal cycles on aqueous concentrations of copper, lead, nickel and zinc, finding it to be an accurate and useful tool for short-term deployments (6 h).Neither of these studies measured the presence of a DBL as it was assumed that in a very fast flow environment this would be negligible, however it is recommended that in future studies the DBL is always measured to ensure that its influence is minimal.
In our study there was a reduction in the TWA concentration of uranium by up to 57% when no DBL thickness was taken into consideration (Figures 3b and 4b).The closest agreement between the concentrations was observed in weeks 19-21 for both deployment sites (Figures 3a and 4a) when the periodically measured DBL thicknesses over the deployment period were used.When the aqueous concentration of uranium was relatively stable, the TWA estimates (taking into account the measured DBL thickness) were 99-107% and 71-111% of those found with the spot water samples for the Rivers Enborne and Lambourn respectively.When using an averaged DBL thickness over the whole deployment period, this value rose to 124-136% for the Enborne and lowered to 70-103% for the Lambourn.Using an averaged DBL thickness has less impact on the TWA concentrations in the River Lambourn than the River Enborne, most likely due to the fluctuating flow rates at the latter site.The lower flow periods, when the DBL is greater, will increase the averaged DBL thickness and will therefore result in an overestimation of the TWA estimates (Figure 2a, weeks 17-21, 14/12/2011-8/01/2012).
To give an indication of the reliability of the DGT technique, the ratio of the TWA concentrations of uranium found with the device to the uranium concentrations found in weekly averaged spot water samples was made (Tables S5 and S6).The closer to one this ratio is the more accurate the technique can be assumed to be, although there is the possibility that the concentrations have fluctuated throughout the week.Results are in agreement with previous work undertaken at these sites, 17 approximately 86% of the dissolved uranium could be measured with accuracy.The River Enborne had an average accuracy of ~94% (38 to 205%) and the River Lambourn ~78% (27 to 138%).The failure to achieve 100% accuracy can be attributed to factors such as biofouling, variations in concentration of uranium over the 7 d deployment, and an underestimation of the thickness of the DBL as this was not measured every week. 60

Isotopic ratios of uranium
There are three naturally occurring isotopes of uranium: 238 U (99.276%), 235 U (0.718%) and 234 U (0.0056%). 49Significant quantities of uranium occur naturally in the environment, however, this element needs to be monitored due to its toxicity, 65 mobility and radiological properties. 52Isotopic composition can indicate if the uranium is of natural or anthropogenic origin as the 235:238 ratio is consistent in nature.As shown in Table 5 there is little difference between the isotopic composition of uranium measured in the spot water samples and DGT.
70 The accuracy of the DGT is within 1%, with a relative standard deviation of 2.85%, which is comparable to Turner et al., 17 where the accuracy and precision were 1 % and 10 % respectively.The 80 better precision in this study could be as a result of the longer deployment times, thereby allowing greater quantities of uranium to accumulate onto the resin.At present, slight enrichments or depletions in the 235:238 ratio would not be detectable using this technique.Further refinement would be necessary to increase the 85 accuracy.These could include using a different uranium measurement technique (such as multi-collector ICP-MS) or by removing interferences from the eluent by using an additional actinide specific resin extraction technique. 90

Conclusions
The data presented here shows DGT can be used as a tool in long-term environmental monitoring programmes, even though seasonal variations in water flow and chemistry can have an impact on results.Water bodies with highly fluctuating flows 95 require extensive DBL measurements.The thickness of the DBL is also affected by factors such as amount of SPM and degree of biofouling.Ideally, the DBL needs to be measured for each deployment.For rivers with a high degree of biological activity, samplers should be mounted in a cage, and this particularly is 100 advisable for longer-term deployments (> 4 d).In addition, recording other physical parameters such as water temperature and pH are essential in order to obtain a reliable value for the

Field locations 70 Fig. 1
Fig. 1 Location of field sites in the UK.Site 1 (S1) is located on the River Lambourn and site 2 (S2) on the River Enborne.Both rivers are tributaries of the River Kennet within the River Thames catchment.Two freshwater field sites were used: Site 1 (51.4469N, -1.3838 75 and has a slow flowing deep channel with a pH ~7.8.The mean flow and base flow indices were 1.32 m 3 s -1 and 0.53.Mean monthly 85 meteorological data was obtained from the Met Office Benson meteorological monitoring station (51.62 N, -1.097 W) (http://www.metoffice.gov.uk/public/weather/climate/benson)Newbury Rea Newbury R This journal is © The Royal Society of Chemistry [year] Journal Name, [year], [vol], 00-00 | 3 5 m from the river bank, out of the main flow channel to allow for access.Three DGT devices containing Metsorb™ resin gel were removed and replaced every week over a 21 week period from 24/08/2011 to 18/01/2012.Procedural blanks (in triplicate) were exposed to the field environment

Fig. 2
Fig. 2 Photograph of DGTs held in place by a Perspex plate.The plate held up to eight devices.If more samplers were deployed then two Perspex plates were fixed back to back.The plate was deployed in the rivers a vertical position.
We calculated the TWA concentrations of uranium using various 105 DBL values to highlight the importance of including this variable.This journal is © The Royal Society of Chemistry [year] Journal Name, [year], [vol], 00-00 | 5

8 |
at this site, showed with daily removal of vegetation and for 95 shorter deployment times (5 days) the thickness of the DBL was Journal Name, [year], [vol], 00-00 This journal is © The Royal Society of Chemistry [year]

and preparation of DGT Chemicals
75Materials were of analytical grade or better and supplied by Fisher Scientific Ltd. (Loughborough, UK), unless otherwise specified.Milli-Q (ultra-pure) water (> 18.2 MΩ cm, Millipore, Watford, UK) was used as the laboratory water.All uranium ICP-

Table 1
Tables1 and 2show the thickness of the DBL (calculated from FiguresS5 and S6) measured in the River Enborne and River Lambourn respectively.Thickness of the diffusive boundary layer (DBL) measured at the River Enborne site. 20

Table 2
Thickness of the diffusive boundary layer (DBL) measured at the River Lambourn site.
Tables1 and 2show that the DBL thickness represents a large 30 component of the overall diffusive layer thickness.The ratio of these values in the River Lambourn throughout the deployment fluctuated between 0.65 to 0.99, and decrease in the River Enborne from 1.48 to 0.39.Figures3 and 4show how the TWA concentrations of uranium calculated over the deployment period

Table 3
Examples of the thickness of the diffusive boundary layer (DBL) found in laboratory experiments in relation to flow or stirring rates.

Table 4
Examples of the thickness of DBL calculated in other field studies.
diffusion coefficient over the trial period.These factors aside, DGT can provide valuable information on labile and bioavailable concentrations of wide range pollutants over long periods and give information that is complementary to that obtained with spot water sampling.The inclusion of this passive sampler in the 'tool box' of techniques for potential use in regulatory water monitoring programmes is justified.funding the project; Susan Atkins (University of Portsmouth) for laboratory support, Dr Gareth Old (CEH Lambourn Observatory Manager) and the Centre for Ecology and Hydrology, Wallingford, UK for use of their freshwater field site (River Lambourn) and provision of water quality data for this site; and 15 Wasings Estate Ltd. for access to the River Enborne.We thank Graver Technologies Ltd. (www.gravertech.com)for the provision of the Metsorb TM resin.We also thank the two anonymous reviewers' for their helpful comments.