Science of the Total Environment 468–469 (2014) 228–239

Contents lists available at ScienceDirect

Science of the Total Environment

j ourna l homepage: www.e lsev ie r .com/ locate /sc i totenv

Dissolved organic carbon and trihalomethane precursor removal at a UKupland water treatment works

Rachel Gough a,⁎, Peter J. Holliman b, Naomi Willis c, Christopher Freeman a

a School of Biological Sciences, Bangor University, Deiniol Road, Bangor, Gwynedd, LL57 2UW, UKb School of Chemistry, Bangor University, Deiniol Road, Bangor, Gwynedd, LL57 2UW, UKc Dŵr Cymru Welsh Water Head Office, Pentwyn Road, Nelson, Treharris, Mid Glamorgan, CF46 6LY, UK

H I G H L I G H T S

• The role of different treatment processes was assessed over a twelve month period.• Coagulation–flocculation enhanced the formation of brominated trihalomethanes.• Reaction kinetics of trihalomethane formation varied seasonally.

⁎ Corresponding author. Tel.: +44 01248 382353.E-mail address: chpc16@bangor.ac.uk (R. Gough).

0048-9697/$ – see front matter © 2013 Elsevier B.V. All rihttp://dx.doi.org/10.1016/j.scitotenv.2013.08.048

a b s t r a c t

a r t i c l e i n f o

Article history:Received 14 May 2013Received in revised form 2 August 2013Accepted 17 August 2013Available online xxxx

Editor: Simon James Pollard

Keywords:CoagulationDissolved organic carbonSize exclusion chromatographyTrihalomethanesXAD fractionation

The removal of dissolved organic carbon (DOC) during potablewater treatment is important formaintaining aes-thetic water quality standards, minimising concentrations of micro-pollutants, controlling bacterial regrowthwithin distribution systems and, crucially, because it contains a sub-component that can act as trihalomethane(THM) precursors. In this study, the concentration and characteristics of rawwater DOC and THM formation po-tential (THMFP) entering an upland potable water treatmentworks were analysed over twelve months. Correla-tions between raw water DOC characteristics, standardised THMFP (STHMFP) and % DOC removal were alsoinvestigated. DOC and THMprecursor removal during a series of treatment stageswas examined over this period,as well as potential selectivity in the removal of DOC fractions, to assess the importance of different treatmentstages for DOC removal and THMamelioration. Though THMFP removal remained high and fairly stable through-out the study period (83–89%), the data suggest that this wasmostly the result of high DOC removal rates ratherthan the selective removal of THM precursors. Whilst this chemical agnosticism makes DOC removal more ro-bust, it may make the overall process more vulnerable to exceeding permissible THM concentrations underchanging climatic conditions. The kinetics of the reaction between DOC and chlorine appeared to vary seasonally,indicating temporal changes in the proportions of fast- and slow-reacting precursors with implications for THMconcentrations at the point of delivery to the consumer. The initial treatment stages, comprising coagulation–flocculation and dissolved air floatation (DAF) were by far the most important in terms of bulk DOC removaland the preferential removal of THM precursors, though, surprisingly, DOC quality was also modified followingchlorination and secondary rapid gravity filtration (RGF). Though net THM concentration decreased followinginitial treatment stages, a doubling in the proportion of brominated THMs (BrTHMs), which are reported to bemore carcinogenic, was also observed.

© 2013 Elsevier B.V. All rights reserved.

1. Introduction

Although there is no regulatory standard for total organic carbon(TOC), the Water Supply (Water Quality) Regulations (2010), whichapply in England and Wales, recommend that potable water TOC con-centration should be monitored, reflecting its relationship to otherwater quality parameters (DWI, 2010). The removal of dissolved organ-ic carbon (DOC), operationally defined as organicmatterwhich can pass

ghts reserved.

through a 0.45 μm filter (Kitis et al., 2001; Thurman, 1985), is often nec-essary under the general requirement that drinkingwater be acceptableto themajority of consumers in terms of its aesthetic properties (colour,odour and taste) (Davies et al., 2004; WHO, 2011). DOC can also be re-sponsible for the transport of regulated organic and inorganic micro-pollutants such as pesticides and metal contaminants (Gao et al.,1998; Rothwell et al., 2007). Biodegradable DOC in finished water canalso be responsible for bacterial regrowthwithinwater distribution sys-tems (Liu et al., 2002; Prévost et al., 1998). Perhaps most significantfrom a public health perspective, is the role of DOC, and humicsubstances in particular, as reaction precursor in the formation of

Raw water

Flocculators

DAF plant

Primay RGF

Mixing chamber

Secondary RGF

Mixing chamber

Contact tank

Supply to distribution

Ca(OH)2, Cl2(g)

Ca(OH)2, Cl2(g)

Raw

Post-DAF

Post-1RGF

Post-2RGF

Final water

Ca(OH)2, Al2(SO4)3

Mixing chamber

Key:

Treatment stage

Dosing

Sample collection

(NH3)

Fig. 1. Flowdiagram showing sequence of treatment stages, chemical dosing and samplingpoints.

229R. Gough et al. / Science of the Total Environment 468–469 (2014) 228–239

disinfection by-products (DBPs); particularly trihalomethanes (THMs)during chlorination (Adin et al., 1991; Owen et al., 1995; Rook, 1974;Symons et al., 1975).

The Water Supply (Water Quality) Regulations (2010) specify amaximum total THMs concentration of 100 μg L−1, measured at theend of the distribution system (i.e. consumers' taps) (DWI, 2010).World Health Organisation (WHO) guidelines includemaximumvaluesfor individual THMspecies, reflecting thehigher toxicity associatedwithbrominated species (60 μg L−1 for CHCl2Br and 100 μg L−1 for CHClBr2and CHBr3 compared with 300 μg L−1 for CHCl3) (WHO, 2011).Chloramination, an alternative method of disinfection, has been intro-duced at some water treatment works (WTWs) to reduce THM levelsin finished water. This involves dosing with NH3 after chlorinationwhich effectively removes free chlorine from solution by forming chlo-ramine species (NH2Cl, NHCl2 and NCl3), whilst maintaining a disinfec-tion residual within the distribution system (Guay et al., 2005).

Coagulation using Al or Fe salts, followed by flocculation and clarifi-cation by sedimentation or dissolved air floatation (DAF), is the mostwidely-used method of DOC removal (Matilainen et al., 2010). Othermethods are available, including membrane filtration, ion exchange,activated carbon filtration/adsorption, ozonation and biodegradation,but coagulation tends to offer a preferable balance between cost andDOC removal efficiency (Sharp et al., 2006c). Used in combinationwith metal coagulants, synthetic and natural polyelectrolytes may im-prove coagulation efficiency and floc characteristics. Prehydrolysedcoagulants such as PACl are reported to be effective over a widerrange of raw water conditions than conventional coagulants (Boltoand Gregory, 2007; Yan et al., 2009). Enhanced coagulation, which re-fers to the optimisation of coagulation conditions for DOC removal, isrecognised as the Best Available Technology (BAT) for controlling disin-fection by-product (DBP) levels in chlorinated drinking water (US EPA,1999).

Reported DOC removal rates by coagulation vary substantially(Matilainen et al., 2010), reflecting the range of impacting factors, in-cluding raw water and DOC characteristics, and coagulation conditionssuch as pH, coagulant type and dose, temperature and sequence ofchemical addition (Letterman and Vanderbrook, 1983; Runkana et al.,2006; Uyak and Toroz, 2007; Yan et al., 2008). The fractional characterof DOC has been identified as an important factor affecting bulk DOC re-moval during coagulation, with higher molecular weight, more hydro-phobic acids relatively amenable to removal, and the lower molecularweight, more hydrophilic DOC more recalcitrant (Chow et al., 2009b;Edwards, 1997; Gu et al., 1995; Huang and Shiu, 1996; Krasner andAmy, 1995; Sharp et al., 2006b; White et al., 1997).

DOC characteristics are also reported to influence DBP formation.THMs, which are the main DBP of concern in potable water treatment,vary both in terms of total yield and speciation as a result of DOC char-acteristics. For example, the hydrophobic acid fraction is reported to beassociated with the highest standardised THM formation potential(STHMFP) (Chow et al., 2005, 2006a; Galapate et al., 1999; Zhanget al., 2009), although exceptions have been observed (Imai et al.,2003; Lu et al., 2009) suggesting that the association is site specific. Spe-cific UV absorbance (SUVA), which provides a qualitative indication ofDOC molecular weight, hydrophobicity and % aromaticity (Edzwaldand Tobiason, 1999; Weishaar et al., 2003), is also reported to correlatepositively with STHMFP (Chow et al., 2003; Edzwald et al., 1985). An in-verse relationship is reported to exist between DOC molecular weightand STHMFP (Gang et al., 2003). Interestingly, aliphatic DOC has beenfound to produce a higher proportion of brominated THMs upon chlori-nation (Heller-Grossman et al., 1993; Teksoy et al., 2008). This suggeststhat the relationship between DOC and THMs is complex.

Understanding the role of different treatment processes in DOC andTHM precursor removal has become increasingly important since, for anumber of decades, a rising trend in DOC concentrations in surface wa-ters draining upland catchments in Northern and Western Europe andNorth America has been observed (Bouchard, 1997; Freeman et al.,

2001; Hejzlar et al., 2003; Monteith et al., 2007; Stoddard et al., 2003;Worrall et al., 2003). Increased seasonal variation and shorter-termfluctuations in DOC concentration and quality due to extreme weatherevents also present a challenge for maintaining finished water stan-dards including THM levels (Eimers et al., 2008; Elliott et al., 2005).

This study comprised three main aims: (1) to investigate temporalvariations in rawwater DOC concentration and characteristics includingTHMFP over a twelve month period, (2) to investigate potential rela-tionships between % DOC removal, STHMFP and selected rawwater pa-rameters, and (3) to assess the role of different treatment processes interms of DOC and THM precursor removal and any selectivity in the re-moval of different DOC fractions.

2. Methods

2.1. Works description and sampling regime

Samplingwas undertaken at a potableWTW in anupland area of theUK where rawwater is abstracted from an adjacent 24.8 km2 reservoir.The catchment comprisesmainly coniferouswoodland plantation (30%)flanking much of the perimeter of the reservoir, grassland (38%) andpeatland habitat (32%) (Cohen, 2009). During the sampling period,raw water comprised high DOC concentration (9.0–16.2 mg L−1),high colour (52–117 Hazen), relatively low turbidity (0.21–0.62 FTU),low alkalinity (~3 mg L−1 as CaCO3) and slight acidity (pH 5.6–6.0).

During normal operation the WTW receives 38,000–40,000 m3 ofwater per day from the reservoir. A works flow diagram is shown inFig. 1. The raw water passes into a rapid mixing chamber where it isdosed with lime (0.15% w/v Ca(OH)2) for pH correction and coagulant

230 R. Gough et al. / Science of the Total Environment 468–469 (2014) 228–239

(8% w/v Al2(SO4)3(aq)). Dosed water passes under gravity to four me-chanical flocculators, arranged in parallel where pin-flocs formed dur-ing the rapid mixing aggregate into macro-flocs. During subsequentDAF, the flocs are captured by ascending micro-bubbles and buoyed tothe surface to form a sludge layer. This layer is removed by a continu-ously rotating paddle. Though sedimentation is more commonly usedfor clarification, DAF may be a more appropriate process where floccu-lation produces smaller, low density flocs or in waters rich in algae,since algal cells have a tendency to float (Teixeira and João, 2006).

The DAF stage is followed by the first set of rapid gravity filters (pri-mary RGF) which are designed to prevent floc carry-over. Followingclarification thewater is dosedwith lime and chlorine for primary disin-fection and manganese oxidation. The water then enters secondaryRGFs for manganese removal via a process of precipitation–filtration.This is followed by final pH adjustment and a second chlorine dose toachieve a free chlorine residual of 1.5 mg L−1. The contact tank priorto distribution provides a period of contact with free chlorine to ensureadequate disinfection, which is a priority during potable water treat-ment (WHO, 2011). The option of dosing with NH3 for chloraminationis also available prior to distribution if required for THM control. Chem-ical dosing is adjusted based on raw water quality which is assessedusing online monitors. Jar testing is used to verify that coagulation con-ditions (coagulant and lime dose) are optimised.

Samples were collected from sampling points along the treatmentchain as shown in Fig. 1. By comparing the post-DAF sample with theraw water sample, the combined effect of coagulation, flocculationand clarification (DAF) can be assessed. The post-1RGF sample incorpo-rates an additional filtration step and represents the culmination of allthe treatment steps designed to reduce DOC concentration, since theprimary filters prevent floc carry-over from the DAF stage. The post-2RGF sample was collected to assess whether this additional filtrationstep, which is designed to remove manganese by precipitation–filtration,affects DOC concentration and/or characteristics. The finalwater samplerepresents the total effect of all the physical and chemical treatmentsemployed at the WTW.

For consistency the final water sample was collected prior to NH3

dosing, which only took place during the final month of sampling (Au-gust 2012). Samples were collected on a monthly basis over a twelvemonth period between September 2011 and August 2012. Amore com-prehensive analysis of raw water and post-1RGF samples, includingXAD fractionation and 7 d THMFP profiles was conducted on a seasonalbasis; autumn (September 2011), winter (December 2011), spring(March 2012) and summer (June 2012). Samples were collected inamber glass bottles so as to leave noheadspace, transported immediate-ly to the laboratory and stored at 4 °C until analysis.

2.2. Analyses

Analysis of samples was focussed onmeasurement and characterisa-tion of DOC and determination of THMFP. Rawwater turbidity (FTU) andcolour (Hazen) were recorded at the WTW from online monitors. Allsamples were filtered through a 0.45 μm nylon membrane filter(Whatman) before analysis. DOC measurement was carried out using aThermalox TOC/TN analyser equipped with a non-dispersive infraredCO2 detector. UV analyses including absorbance at λ = 400 nm (usedas a proxy for colour (Mitchell and McDonald, 1992)) and at λ =254 nm were made using a Molecular Devices SpecraMax M2e multi-detection reader (spectrophotometer) with aliquots of samples pipettedinto a 96-well clear micro-plate. SUVA values were derived from the fol-lowing formula: UV Abs. 254 (cm−1) ∗ 100/DOC (mg L−1). Absorbancewas also measured at λ = 253 nm and λ = 203 nm to derive theA253:A203 ratio which is reported to correlate with the proportion ofhydroxyl-, carboxyl-, ester- and carbonyl-substituted aromatic rings(Korshin et al., 1997). These functional groups have been implicated inreactions generating DBPs (Kim and Yu, 2007). Phenolic concentrationwas measured using the Box (1983) method adapted for 300 μL micro-

plate wells. High pressure size exclusion chromatography (HPSEC) wasconducted using a Varian PL-GPC-50 DataStream unit detecting atλ = 254 nm. The HPSEC unit was interfaced to Cirrus software andequipped with a Bio Sep 2000 column. Calibration standards weresodium polystyrene sulfonate polymers with molecular weights of150,000, 77,000, 32,000, 13,000 and 4300 Da (Fluka) and cyanocobala-min (1340 Da). Themobile phase wasmilli-q water buffered with phos-phate (2 mM KH2PO4 + 2 mM K2PO4·3H2O) to pH 6.8. From themolecular weight distributions (MWDs) the followingmolecular weightindices were calculated:

Mp: peak molecular weightMn: number-average molecular weight (Eq. (1)) — the value atwhich there are equal numbers of molecules on each side

Mn ¼ ΣihiΣi

hi.

Mi

ð1Þ

Mw: weight-average molecular weight (Eq. (2)) — the value atwhich there are equal masses of molecules on each side

Mw ¼ ΣihiMi

ΣiMi: ð2Þ

Here hi is the height (from the baseline) of the SEC curve at the ithincrement and Mi is the molecular weight of the species eluting at thisincrement (obtained via calibration with standards).

2.2.1. FractionationFractionation of DOC, whichwas carried out on a seasonal basis, was

achieved by resin adsorption using a method adapted from Thurmanand Malcolm (1981) and Marhaba et al. (2003). Samples were separat-ed into five fractions: hydrophobic acid (HPOA), hydrophobic base(HPOB), hydrophilic acid (HPIA), hydrophilic base (HPIB) and hydro-philic neutral (HPIN) according to their adsorption onto macroporousresins. A column packed with Superlite™ DAX-8 resin and a secondpacked with Amberlite™ XAD-4 resin (both Supelco) were connectedusing PEEK tubing. The systemwas connected to a Cecil 1100 Series liq-uid chromatography pump to control the elution rate. The HPIN fractionwas retrieved by passing the sample through both columns at a rate of4 mL min−1 and collecting the eluent. This process also loaded theresins with the remaining fractions. The HPOA fraction was eluted bypassing 60 mL of 0.1 M NaOH through the DAX-8 column followed by40 mL of milli-q water, (both at 2 mL min−1). This was repeated forthe XAD-4 column to obtain the HPIA fraction. The HPOB fraction waseluted by passing the 60 mL of 0.1 M HCl through the DAX-8 columnfollowed by 40 mL of milli-q water, (both at 2 mL min−1). The HPIBfractionwas obtained by repeating this processwith the XAD-4 column.

2.2.2. Trihalomethane formation potential (THMFP) methodTHMFP7 d denotes the quantity of THMs formed (μg L−1) following

chlorination of a water sample for a 7 d incubation period at 25 °C.The method used was adapted from the Standing Committee ofAnalysts (1981) procedure. Samples were diluted to 1 mg L−1 DOC toderive a standardised THMFP7 d (STHMFP7 d) value which provides ameasure of DOC reactivity. THMFP7 d was calculated by multiplyingSTHMFP7 d by DOC concentration. For chlorination, 97.5 mL of dilutedsample was dosed with 2.0 mL of 0.5 M KH2PO4(aq) to buffer the solu-tion to pH 6.8. Samples were then dosed with 0.5 mL of NaOCl(aq) toprovide 5 mg of free Cl per mg of DOC. After a 7 d incubation in thedark at 25 °C, the reaction was quenched using 0.4 mL of 0.8 M

231R. Gough et al. / Science of the Total Environment 468–469 (2014) 228–239

Na2SO3(aq). Extraction of the four main THM species (CHCl3, CHBrCl2,CHBr2Cl and CHBr3) was achieved using direct immersion SPME follow-ed by quantification using a Varian 450 GC coupled with an electroncapture detector. For the seasonal samples, THM concentrations werealso measured at 1 h, 1 d and 3 d in order to profile THMFP with reac-tion time.

DC

E F

BA

0

2

4

6

8

10

12

14

16

18

DO

C (

mg

L-1

)

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

Co

lou

r (a

bs

400

nm

cm

-1)

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

Ph

eno

lics

(mg

ph

en. m

g D

OC

-1)

0

20

40

60

80

100

120

140

160

180

200

ST

HM

FP

7d (µg

TH

M m

g D

OC

-1)

HG

Fig. 2. Analyses for treatment chain samples between September 2011 and August 2012; raw (blue) including DOC concentration (A), % DOC removal (B), colour (C), A253:A203 (D), standardrepresent 5% covariance (A), the standard error (n = 3) (C,D,E and F) and the standard error d

2.2.3. Statistical analysisStatistical analysis was performed using version 20 of the SPSS Statis-

tics package (PASW). Friedman's ANOVA was conducted to investigateselectivity in the removal of DOC and THM precursors during successivetreatment stages. Post-hoc analysis employed the Wilcoxon signed-rank test incorporating a Bonferroni correction (Bonferroni, 1936).

0

10

20

30

40

50

60

70

80

90

%D

OC

rem

ova

l

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7A

253:

A20

3

0

1

2

3

4

5

6

7

SU

VA

(L m

g1

m-1

)

0

200

400

600

800

1000

1200

1400

1600

1800

2000

TH

MF

P7d

(µg

L-1

)

dark blue), post-DAF (red), post-1RGF (green), post-2RGF (purple) and final water (lightised phenolics concentration (E), SUVA (F), STHMFP7 d (G) and THMFP7 d (H). Error barserived from a detection repeatability experiment (G and H).

232 R. Gough et al. / Science of the Total Environment 468–469 (2014) 228–239

Spearman's correlation was performed to test for significant correlationsbetween raw water quality parameters, in particular between DOC re-moval rates, raw water THMFP7 d and raw water DOC characteristics.

3. Results

3.1. Temporal variations in rawwater DOC characteristics, THMFP and DOCremoval rates

Within the twelve month sampling period, maximum raw waterDOC concentration was observed in January 2012 (16.2 mg L−1) andthe minimum in August 2012 (9.0 mg L−1) (Fig. 2A). This seasonaltrend in DOC is broadly in agreement with Dŵr Cymru Welsh Water(DCWW) raw water data for preceding years, though the August mini-mum is comparatively late, and normally occurs betweenMay and June.DOC removal rates averaged 76% for the twelve month period with alow of 62% in March 2012 and a high of 83% in November 2011(Fig. 2B). Colour (λ = 400 nm), which is affected by both DOC quantityand quality, varied between 0.084 a.u. in November 2011 and 0.050 a.u.in January 2012 (Fig. 2C).

Little variation occurred in the A253:A203 ratio which showed a min-imum in January 2012 (0.48) and a maximum in June 2012 (0.57)(Fig. 2D). Raw water phenolics concentration, standardised for DOCconcentration, ranged between 0.07 mg phen. mg DOC−1 in Novemberto 0.25 mg phen. mg DOC−1 in August (Fig. 2E). Although variable, rawwater SUVA values, which provide a broad indication of hydrophobicity,molecular weight and % aromaticity, remained high during the sam-pling period (Fig. 2F). The minimum SUVA (2.8 L mg−1 m−1) in Janu-ary 2012, coincided with peak DOC concentration and the maximum(5.8 L mg−1 m−1) in August, coincided with the minimum DOC con-centration. STHMFP7 d broadly follows the same seasonal trendexhibited by SUVA. Coincidingwith the SUVAmaximum andminimum,lowest STHMFP7 dwas observed in January (63 μg THMmgDOC−1) andthe highest in August (176 μg THM mg DOC−1) (Fig. 2G).

Rawwater THMFP7 d was highest in November (1806 μg L−1), witha second peak in June (1755 μg L−1), and was lowest in February(938 μg L−1), though the whole period January 2012–April 2012 wascharacterised by relatively low THMFP7 d (Fig. 2H). Comparing thetwo measurements influencing THMFP7 d, variability in raw waterSTHMFP7 d (CV: 27%) was considerably higher than the variability inDOC concentration (CV: 16%). Percentage STHMFP7 d removal (andtherefore the preferential removal of THMprecursors) during treatmentvaried considerably, ranging from 17% in October 2011 to 69% in March2012. In contrast to its contributing factors (STHMFP7 d and DOC

HMW

MW

= 1

,000

Da

Res

po

nse

(m

V)

Retention

Fig. 3. HPSEC chromatograms for all raw water samp

concentration), % THMFP7 d removal remained fairly stable (CV: 2%),ranging between 83% in April 2012 and 89% in July 2012.

Monthly raw water HPSEC data for this sampling period indicatethat variations in DOC MWDs were minimal (Fig. 3). The raw waterMWDs include a small peak at tR = 4.2 minwhich represents highmo-lecular weight (HMW) molecules that are too large to interact with thepores of the stationary phase (Huber et al., 2011). This is followed by adominant asymmetrical peak at tR = 8.0 min with two overlappingpeaks at tR = 8.9 and 9.3 min. These data were converted to molecularweight ranges with high molecular weight (HMW) characterisedas N 1000 Da and low molecular weight (LMW) as b 1000 Da (Zouet al., 2004). According to these definitions, rawwater DOC consistentlycomprised N 94% HMWmolecules.

3.2. Relationships between raw water characteristics, % DOC removaland STHMFP

The relationship between % DOC removal rate and various DOC char-acteristics was investigated using Spearman's correlation coefficient.The only statistically significant relationship was found to be withstandardised phenolics concentration, rs = −0.778, p b 0.01, suggestingthat phenolic compounds are more recalcitrant. STHMFP7 d was found tovary as a function of DOC concentration, rs = −0.588, p b 0.05,suggesting that as rawwater DOC concentration increases, its relative re-activity with chlorine decreases. STHMFP7 d also varied positively withSUVA, rs = 0.615, p b 0.05, suggesting that more hydrophobic DOC hasa higher reactivity with chlorine. Finally, a significant positive correlationwas found between STHMFP7 d and A253:A203, rs = 0.755, p b 0.01.

3.3. Selectivity in DOC and THM precursor removal during successivetreatment stages

Friedman's ANOVAwas used to investigate changes in water qualityresulting from different treatment processes (Table 1), including selec-tivity in DOC and THM precursor removal. Unsurprisingly a substantialand statistically significant decrease in DOC concentration occurredduring the early treatment stages (coagulation–flocculation andDAF) (p b 0.05). A statistically significant reduction in SUVA from4.2 L mg−1 m−1 in the raw water to 1.8 L mg−1 m−1 following DAFtreatment (p b 0.05) also occurred as a result of these early treatments,indicating the preferential removal of hydrophobic, HMW DOC duringcoagulation–flocculation and DAF. A smaller but significant reductionin SUVA also occurred after secondary RGF. A253:A203 also fell substan-tially following the DAF stage (p b 0.05) (from 0.52 to 0.23 UV a.u.)and showed a smaller reduction following secondary RGF (p b 0.05).

LMW

Time

les between September 2011 and August 2012.

Table 1DOCquality and THMFP results for process chain samples collected between September 2011 andAugust 2012 showing statistically significant differences identifiedby Friedman'sANOVAanalysis.

Raw (a) Post-DAF (b) Post-1RGF (c) Post-2RGF (d) Final water (e)

DOC concentration (mg L−1) 11.7 ± 0.5 bcde 3.2 ± 0.2 a 2.8 ± 0.2 a 3.0 ± 0.2 a 2.8 ± 0.2 aSUVA (L mg−1 m−1) 4.2 ± 0.2 bcde 1.8 ± 0.2 ad 1.9 ± 0.2 ade 1.3 ± 0.2 abc 1.4 ± 0.2 acColour(abs. 400 nm cm−1)

0.060 ± 0.003 bcde 0.004 ± 0.001 a 0.005 ± 0.001 a 0.002 ± 0.001 a 0.004 ± 0.001 a

A253:A203 0.52 ± 0.01 bcde 0.23 ± 0.02 a 0.22 ± 0.01 ad 0.17 ± 0.01 ac 0.17 ± 0.01 aPhenolics per mg DOC (mg phen. mg DOC−1) 0.15 ± 0.02 bcde 0.25 ± 0.06 a 0.19 ± 0.06 a 0.24 ± 0.02 a 0.22 ± 0.05 a% HMW DOC 96.6 ± 0.2 bcde 94.5 ± 0.4 a 93.4 ± 0.5 a 91.8 ± 0.9 a 92.1 ± 0.8 aMp (Da) 5267 ± 25 bcde 3156 ± 120 a 3161 ± 118 a 2990 ± 23 a 3139 ± 140 aMn (Da) 3180 ± 54 bcde 2385 ± 54 a 2246 ± 55 a 2113 ± 88 a 2133 ± 82 aMw (Da) 4401 ± 27 bcde 3219 ± 44 a 3075 ± 21 a 3038 ± 40 a 3070 ± 31 aSTHMFP7 d

(μg THM mg DOC−1)121 ± 10 bcde 72 ± 7 a 72 ± 8 a 75 ± 7 a 81 ± 9 a

THMFP7 d (μg L−1) 1384 ± 87 bcde 216 ± 16 a 190 ± 14 a 216 ± 19 a 217 ± 17 a% BrTHMs 4.8 ± 1.4 bcde 11.8 ± 3.4 a 10.0 ± 2.9 ad 8.4 ± 2.4 ac 9.8 ± 2.8 a

Results given as mean ± standard error (n = 12). Letter annotations denote significantly different means (p b 0.05).

233R. Gough et al. / Science of the Total Environment 468–469 (2014) 228–239

Colour showed a substantial and statistically significant reduction fol-lowing the DAF stage (p b 0.05) in line with reduced DOC concentra-tion. Standardised phenolic concentration increased following the DAFstage (p b 0.05), supporting the view that phenolic compounds aremore recalcitrant. % HMWDOC (N1000 Da)was higher in the raw sam-ple compared with all subsequent samples (p b 0.05) and no statisticaldifference between stages thereafter. A comparison of the molecularweight averages (Mp, Mn and Mw) supports this result, each showingstatistically higher values in rawwater compared with the samples col-lected further along the treatment chain (p b 0.05), and again, no signif-icant difference between these subsequent samples. An example of theMWDs for samples collected along the treatment chain is shown inFig. 4. The first peak (tR = 4.2 min) which appears in the raw sampleis absent from all subsequent samples. A dramatic reduction in thearea of the second peak between the raw water and post-DAF stageand a further slight reduction between the two RGF stages is seen forthis month but not all, in line with reduced DOC concentration. A shifttowards lower MW molecules is also apparent in treated water com-pared with raw.

THMFP7 d and STHMFP7 d showed a dramatic and statistically signif-icant reduction in the post-DAF sample compared with the raw water(p b 0.05) and no significant change thereafter, suggesting that THMprecursors were preferentially removed during these early treatmentstages (coagulation–flocculation and DAF). Comparing THM speciation,

HMW

Res

po

nse

(m

V)

MW

= 1

,000

Da

Retenti

Fig. 4.HPSEC chromatograms for process chain samples in November 2011 (raw: blue, post-DAreferences to colour in this figure legend, the reader is referred to the web version of this artic

the % BrTHMs were found to increase significantly following the earlytreatment stages (p b 0.05) (Fig. 5). A slight but statistically significantreduction in %BrTHMs was also observed following secondary RGF(p b 0.05).

3.4. Seasonal analysis of fractional character and THMFP profiles

Fractional character and 7 d THMFP profiles for raw and post-1RGFsamples were investigated on a seasonal basis. Post-1RGF samples rep-resent the culmination of all the treatments designed to reduce DOC. Inthe rawwater samples, the HPOA fraction dominated, varying between36% in winter and 51% in spring (Fig. 6). This fraction consistently rep-resented the dominant fraction in percentage terms. The HPIA fractionvaried between 16% in summer and 28% in spring, HPIN from 35% inwinter to 18% in spring whilst the HPOB and HPIB fractions combinedconsistently represented b10% of total DOC. Compared to the rawwater samples, the contribution of the HPOA fraction in the post-1RGFsamples was less in all cases except in the summer, where the fractionalcharacter of the DOC in the post-1RGF sample remained very similar tothat of the raw water.

The 7 d STHMFP profiles (Fig. 7) show that the kinetics of the reac-tion, and specifically the % of THMs formed within the first 24 h, variedseasonally. In the raw water STHMFP1 d as a percentage of STHMFP7 d

ranged from 39% in winter to 68% in autumn. In the post-1RGF sample

LMW

on Time

F: red, post-1RGF: green, post-2RGF: pink, final water: dark blue. (For interpretation of thele.)

75

80

85

90

95

100

Raw Post-DAF Post-1RGF Post-2RGF Final water

% S

TH

MF

P

Fig. 5. Mean % contribution of THM species (CHCl3: dark grey, BrTHMs: light grey) toSTHMFP7 d for treatment chain samples. Error bars represent the standard error of themean (n = 12).

234 R. Gough et al. / Science of the Total Environment 468–469 (2014) 228–239

the STHMFP1 d percentage was also highest in autumn (56%) but waslowest in summer (38%). The 7 d THMFP profiles (Fig. 8) confirm thatTHMFP reduction during treatment remained high throughout the sam-pling period. Though temperature fluctuations and other conditionswithin the distribution system such as the role of biofilms could notbe replicated, the post-1RGF profiles provide an approximate model ofTHM concentrations at different points in the distribution system. Esti-mated post-1RGF THMFP at 36 h,which is themaximum residence timefor water in the distribution system, exceeded the regulatory limit of100 μg L−1 in autumn (127 μg L−1) and summer (131 μg L−1).

4. Discussion

4.1. Raw water quality and DOC removal rates

Comparedwith the global average for lake DOC concentrations (me-dian: 5.71 mg L−1), the reservoir water analysed in this study exhibitsrelatively high DOC concentration throughout the year (Sobek et al.,2007). Algal assemblages analysed for this water source suggest thatthe reservoir is oligotrophic; a trophic state associated with low levelsof primary production (Wetzel, 2001). Allochthonous DOC sources aretherefore likely to account for this high DOC loading. Both peatlandand coniferous forestry plantations, which cover extensive areas ofthis catchment, have been associatedwith highDOCflux; predominant-ly humic acids (Clark et al., 2008; Creed et al., 2008; Kaiser et al., 2001;Lindroos et al., 2011). The seasonal trend in DOC concentration ob-served in this study conforms with data from previous years showinga maximum in late autumn/early winter and a minimum in latespring/early summer (DCWW data), though in this case the minimum

4840 36

28

5137

49 52

2631

2733

28

3216 18

23 24 35 31

1823 25 22

0

10

20

30

40

50

60

70

80

90

100

% o

f D

OC

HPIB

HPOB

HPIN

HPIA

HPOA

Fig. 6. % contribution of HPOA, HPIA, HPIN, HPOB andHPIB fractions to the total recoveredDOC following fractionation for raw and post-1RGF samples in autumn,winter, spring andsummer.

occurred later than normal; in August. Fluvial DOC concentrationshave been found to peak in late summer/autumn as DOC which accu-mulates in the soil matrix during the summer months due to warm,aerobic conditions is flushed from the catchment (Hope et al., 1994;Neal et al., 2005). This peakmay also be influenced by litter-fall fromde-ciduous tree species in the autumn (Chow et al., 2009a). Lower streamDOC concentrations are generally observed in late winter/spring whenthe soil tends to bemorewaterlogged and cold conditions inhibitmicro-bial degradation (Halliday et al., 2012). This trend has been observed inthe inflow streams of this reservoir (data not shown). The lag betweenthis and the reservoir DOC trendmay be the result of relatively long res-idence times in the reservoir.

% DOC removal rates remained high throughout the sampling period(mean 76%) compared with previous studies where 29–70% averageDOC removal rates are reported for iron-based coagulants and 25–67%for aluminium sulphate in a range of raw water types (Matilainenet al., 2010). This is important given the rising trend in DOC concentra-tion for surface waters in many upland catchments (Bouchard, 1997;Freeman et al., 2001; Hejzlar et al., 2003; Stoddard et al., 2003;Worrall et al., 2003). Reports suggest DOC removal rates during coagu-lation depend onmany factors, including DOC and rawwater character-istics, and coagulation conditions (Letterman and Vanderbrook, 1983;Runkana et al., 2006; Sharp et al., 2006a; Uyak and Toroz, 2007; Yanet al., 2008). The high removal rates reported in this study are likelydue to the high proportion of HMW, hydrophobic humic substances inthe rawwater, which have been found to be easily removed by conven-tional coagulation–flocculation (Edwards, 1997; Randtke, 1988; Sharpet al., 2006b), allied to the higher charge densities associated with thisDOC fraction (Edzwald, 1993; Sharp et al., 2006b).

The only statistically significant relationship between % DOC re-moval and raw water characteristics was a negative correlation withstandardised phenolic content (p b 0.05). The difficulty of removingphenolic compounds by conventional coagulationmethods has been re-ported previously (Tomaszewska et al., 2004). This relates to the en-hanced aqueous solubility of phenolic compounds relative to otherDOC due to their hydrophilic hydroxyl content (Bond et al., 2009), be-cause the coagulation process is under kinetic control within a compet-itive precipitation process so the least soluble material will be removedfirst. Although phenolic content has been correlated positively withSTHMFP7 d in previous studies (Harrington et al., 1996; Imai et al.,2003) no relationship was found in this study. The positive relationshipbetween STHMFP7 d and A253:A203 (p b 0.05), which has been reportedpreviously, is likely due to the fact that the functional groups identifiedby the A253:A203 index (e.g. esters and ketones) react via haloform-likereactions during chlorination (Kim and Yu, 2007; Korshin et al., 1997).

SUVAvalues for rawwater indicate relatively highDOChydrophobic-ity and molecular weight throughout the year (mean 4.2 L mg−1 m−1)(Edzwald and Tobiason, 1999; Volk et al., 2002) suggesting high humiccontent. The high SUVA values in this reservoir are likely due to thehigh proportion of peatland and forested area in the catchment, bothof which have been found to correlate positively with surface waterSUVA (Piirsoo et al., 2012). Interestingly, peak SUVA in August(5.8 L mg−1 m−1) coincided with the DOC minimum (9.0 mg L−1),and lowest SUVA in January (2.8 L mg−1 m−1) with the DOCmaximum(16.2 mg L−1). Environmental conditions are reported to cause tempo-ral variations in SUVA; a positive relationship is reported between soilwater content and SUVA (HPOA proportion) in pore water (Chowet al., 2006b; Christ and David, 1996) and increased stream waterSUVA has also been found to correlate with rainfall events (Volk et al.,2002). However, understanding variations in SUVA where long resi-dence times are involved, such as in lakes and reservoirs, is complicateddue to the likelihood that a number of chemical and microbial transfor-mations will affect DOC quality and quantity. It has been proposed thatthe aromatic DOC fraction is relatively stable (Kalbitz et al., 2003) andtherefore consumption of microbially labile DOC by microorganismscould increase the relative proportion of aromatic carbon (Chow et al.,

0

20

40

60

80

100

120

140

160

180

200

0 50 100 150 200

ST

HM

FP

(µg

TH

M m

g D

OC

-1)

ST

HM

FP

(µg

TH

M m

g D

OC

-1)

ST

HM

FP

(µg

TH

M m

g D

OC

-1)

ST

HM

FP

(µg

TH

M m

g D

OC

-1)

time (h)

0

20

40

60

80

100

120

140

160

180

200

0 50 100 150 200

time (h)

0

20

40

60

80

100

120

140

160

180

200

0 50 100 150 200

time (h)

0

20

40

60

80

100

120

140

160

180

200

0 50 100 150 200

time (h)

A

DC

B

Fig. 7. Profile of STHMFP over a 7 d incubation period following chlorination of raw (solid line) and post-1RGF (dashed line) samples in Autumn (A), Winter (B), Spring (C) and Summer(D). Error bars represent the standard error of the mean derived from a detection repeatability experiment.

235R. Gough et al. / Science of the Total Environment 468–469 (2014) 228–239

2006b). Conversely microbial activity has been linked to an increase inthe proportion ofHPIADOCdue to acceleration of themicrobial degrada-tion of HPOA to HPIA DOC (Christ and David, 1996). Transformations ofDOC may also be driven by photo-degradation, which is reported tocause an increase in the proportion of LMW hydrophilic DOC (Waiser

0

200

400

600

800

1000

1200

1400

1600

1800

2000

0 50 100 150 200

TH

MF

P (µg

L-1

)T

HM

FP

(µg

L-1

)

time (h)

0

200

400

600

800

1000

1200

1400

1600

1800

2000

0 50 100 150 200

time (h)

A

C

Fig. 8.Profile of THMFP over a 7 d incubation period following chlorination of raw (solid line) anError bars represent the standard error of the mean derived from a detection repeatability exp

and Robarts, 2004), and precipitation which can cause partitioning ofDOC fractions. Precipitation of DOC is common in acidic waters contain-ing high concentrations of iron or aluminium; features consistent withthe reservoir in the present study (Parks and Baker, 1997; Pokrovskyand Schott, 2002). Precipitation and settling out of DOC in the reservoir

TH

MF

P (µg

L-1

)T

HM

FP

(µg

L-1

)

0

200

400

600

800

1000

1200

1400

1600

1800

2000

0 50 100 150 200

time (h)

0

200

400

600

800

1000

1200

1400

1600

1800

2000

0 50 100 150 200

time (h)

B

D

d post-1RGF (dashed line) samples inAutumn (A),Winter (B), Spring (C) and Summer (D).eriment.

236 R. Gough et al. / Science of the Total Environment 468–469 (2014) 228–239

may also explain the substantial decrease in DOC concentration in thereservoir comparedwith its input streams (data not shown). Thoughmi-crobial populations are likely to be low in this reservoir due to oligotro-phic conditions, the coincidence of the SUVA maximum with the DOCminimum in August may be the result of enhancedmicrobial processingof labile LMW HPIA DOC under higher water temperature conditions.Conversely, the suppression of microbial activity due to low water tem-perature in January may have contributed both to the higher DOC pooloverall (DOC maximum) and the relatively large pool of labile LMWHPIA DOC (indicated by the SUVA minimum).

A statistically significant correlation between raw water STHMFP7 d

and SUVA was observed (p b 0.05). A positive relationship betweenSUVA and STHMFP7 d has been reported previously with SUVA evenhaving been used as a surrogate for STHMFP (Chow et al., 2003;Edzwald et al., 1985) although it has been suggested that SUVA is nota useful indicator of DOC reactivity from distinct sources due to the re-lationship being site specific (Weishaar et al., 2003). It is perhaps unsur-prising therefore that the seasonal pattern of STHMFP7 d in the rawwater is visually similar to that of SUVA (Fig. 2F and G). During thestudy period, the reactivity of the DOC (STHMFP7 d) was more variable(CV: 27%) thanDOC concentration (CV: 16%), suggesting that variationsin rawwater THMFP over time are controlledmostly by variations in thecharacter of DOC. This also highlights the importance of the selective re-moval of THM precursors during treatment as opposed to merely bulkDOC removal. The overall reactivity of the raw water (STHMFP7 d) washighest in August 2012 (176 μg THM mg DOC−1) and the lowest inJanuary 2012 (63 μg THM mg DOC−1). The rate of halogenation in theearly stages of the reaction also varied seasonally, with the proportionof THMs formed in the first 24 h of reaction 39% in winter, comparedwith 68% in autumn. This suggests that summer and autumn presentthe highest risk for THMFP. THM precursors are known to vary in reac-tivity according to DOC functionality, with fast- and slow-reacting THMprecursors having been identified in previous studies; Gallard and vonGunten (2002) report that resorcinol-type structures and phenolic com-pounds act as fast- and slow-reacting THM precursors, respectivelywhilst Dickenson et al. (2008) identified β-diketone-acids and β-ketoacids, as fast- and slow-reacting THM precursors. Though the broad hy-drophobic/hydrophilic DOC descriptors used in this study are insuffi-cient for identifying compounds likely to influence the rate of THMformation, it is likely that temporal changes in the proportions of fast-and slow-reacting precursors are responsible for the seasonality in reac-tion kinetics observed. Interestingly, a negative correlation was identi-fied between STHMFP7 d and DOC concentration (p b 0.05), indicatingthat as DOC concentration increases, the relative proportion of THMprecursor decreases. This further emphasises the importance ofSTHMFP as well as DOC concentration in affecting raw water THMFP.The first THMFP7 d peak in November 2011 was the result of higherthan average DOC and STHMFP7 d whereas higher than averageSTHMFP7 d was solely responsible for the second peak in June 2012.The lowest raw water THMFP7 d occurred between January and April,predominantly due to low STHMFP7 d, despite the DOC concentrationmaximum in January.

The 7 d THMFP profiles (Fig. 8) provide a useful visual indication ofbulk THMFP removal during treatment. % reduction in THMFP7 d be-tween the raw and post-1RGF stages remained stable, ranging between83% and 89% over the twelve month period (Fig. 2H). These removalrates are high compared to reports in the literature. Iriarte-Velascoet al. (2007) report 31–48% bulk removal of THMFP for a low DOCsurface water under different alkalinity conditions using alum andpolyaluminium chloride coagulants. Page et al. (2002) report 55%THMFP removal on average across a range of reservoir waters andDOC concentrations using alum. Similarly Uyak and Toroz (2007) reportan average 56% precursor removal under optimum conditions averagedacross three surface water sources using ferric chloride and alum coag-ulants. The coagulation systememployed at thisWTW is thereforewell-optimised for THMFP removal. However, the 7 d THMFP post-1RGF

profiles, which are an approximate model for THM concentrationsalong the distribution system, suggest that the system may still be vul-nerable to exceeding the regulatory limit of 100 μg L−1 at the extremi-ties of the distribution system (36 h after water leaves the WTW) inautumn and summer. The majority of the reduction in THMFP is dueto bulk DOC removal (Fig. 2A),with the coagulation system less success-ful in selectively removing THM precursors (Fig. 2G). As a result ofmainly climatically-driven changes in DOC production and processing,the quality of both allochthonous and autochthonous DOC is expectedto change in the future (Ritson et al., unpublished results). Though theprecise nature of these changes is difficult to predict, it is likely thatWTWs will need to respond to increased seasonality as well asshorter-term changes in DOC quality resulting from increasingly fre-quent and severe extreme weather events (Eimers et al., 2008; Elliottet al., 2005; Jenkins et al., 2009). As a result it may be necessary toadopt a season-specific approach to coagulation optimisation focussedon the selective removal of THM precursors. In this context Tang et al.(2013) recommendmore frequent jar tests to optimise coagulation con-ditions particularly during periods of high rainfall following drought. In-troducing additional treatment processes may be necessary to enhancethe removal of recalcitrant THM precursors. For example, anion ex-change and granular activated carbon (GAC) can be successful in re-moving more recalcitrant LMW hydrophilic and neutral DOC (Bondet al., 2011). Finally, controlling THM levels may require switching to achloramination system, which has been associated with a substantialdecrease in THM formation (Kristiana et al., 2009), though this strategyis also associated with an increase in nitrogenated DBP such as N-nitrosodimethylamine (Choi and Valentine, 2002), as well as currentlyunidentifiable chlorinated by-products (Hua and Reckhow, 2007).

HPSEC analysis indicates that raw water DOC MWDs were consis-tently dominated by HMW DOC (consistently N94% contribution).Fabris et al. (2008) defined this molecular weight category as HMWhu-mics. It should be noted however that structures with few conjugatedbonds exhibit low ε to UV (Matilainen et al., 2011) and so will producelittle or no response from the detector (Leenheer and Croue, 2003). Thispoor sensitivity to these compoundswhich, some previous studies haveshown, are more likely to form THMs (Gang et al., 2003), may be thereason for the absence of a correlation between molecular weight indi-ces and THMFP in this study.

4.2. The role of successive treatments in DOC and THM precursor removal

Friedman's ANOVA revealed that the initial treatment stages(coagulation–flocculation and DAF) were responsible for the majorityof DOC removal and the preferential removal of THM precursors (asevidenced by the fall in STHMFP7 d). These early treatment stageswere alsomost important in modifying other DOC characteristics, caus-ing substantial and statistically significant reductions in SUVA, A253:A203, colour and molecular weight and increased standardised phenoliccontent. A comparison of the raw and post-1RGF fractionation resultsalso illustrates this selective removal of DOC as the % contribution ofthe HPOA fraction is reduced after treatment in all cases except in thesummer sample which shows a slight increase (Fig. 6).

It is the preferential removal of the HMW hydrophobic, aromaticDOC indicated by these parameters andwidely reported in the literature(Edwards, 1997; Randtke, 1988; Sharp et al., 2006a) that is responsiblefor the statistically significant reduction in STHMFP7 d during thesetreatment stages. This selectivity is thought to result from the highercharge densities associated with this DOC fraction (Sharp et al.,2006a), and is important for ensuring that finished water THM levelsare sufficiently low for distribution. However, as with the raw water,the rate of halogenation in the post-1RGF samples also varied seasonallywith STHMFP1 d as a percentage of STHMFP7 d ranging between 38% insummer and 56% in autumn. This kind of variability may influence sea-sonal differences in THM levels at the point of delivery (typically 0–36 hafter chlorination).

237R. Gough et al. / Science of the Total Environment 468–469 (2014) 228–239

HPSEC has become a popular method for assessing DOC characteris-tics at different stages ofwater treatment,with a preferential removal ofHMWDOC during coagulation consistently reported (Chow et al., 2008,2009b; Croue et al., 1993; Fabris et al., 2008;Matilainen et al., 2011). TheMWDs in this study illustrate both the reduction in DOC concentrationfollowing initial treatment and the shift to lower MWs, a resultconfirmed by the statistically significant reduction in all threeMW indi-ces (Mp, Mn and Mw). However, as stated previously, there are limita-tions to this analysis due to the lack of sensitivity to LMW/aromaticcompounds.

The secondary RGFs were the only other treatment stage associatedwith a statistically significant change in DOC quality. This treatmentstage was associated with a slight reduction in SUVA and A253:A203.This did not occur during the preceding RGF stage suggesting that man-ganese removal chemistry may have contributed to the change. Chlo-rine is added prior to the secondary RGFs to oxidise Mn2+ to Mn4+,which simultaneously results in hydrolysis of the manganese to aMnO2-like solid. Thus, the action of chlorine would be to chlorinatemore reactive (aromatic and A253:A203) (Chow et al., 2003; Edzwaldet al., 1985; Kim and Yu, 2007) DOC whilst Mn precipitation would actin a similar way to coagulation, by removing some of the highly chargedaromatic DOC. Other parameters, including STHMFP7 d, were less affect-ed, although % BrTHMs decreased slightly, in line with the loss of morereactive DOC.

Interestingly, coagulation–flocculation and DAF caused a significantincrease in the proportion of BrTHMs. IC data confirm that this is not re-lated to Br− contamination during treatment, although chlorine con-tains ca. 1% Br2 as a production impurity. However, natural Br− fromsea water is also known to transfer inland via aerosol particles and pre-cipitations, resulting in the occurrence of trace levels of Br− in naturalsurfacewaters (Winchester andDuce, 1966).Where HOCl andHOBr co-exist, HOCl has been reported to react preferentially with the phenolicpolymer core of humic substances to bring about oxidative bond cleav-age. HOBr has been found to be a more powerful halogenating agent inthe subsequent electrophilic substitution steps (Ichihashi et al., 1999).DOC rich in aliphatic compounds such as ketones, indicated by lowSUVA values, are therefore associated with the formation of higher pro-portions of BrTHMs (Heller-Grossman et al., 1993; Teksoy et al., 2008).Thus, the preferential removal of the aromatic DOC fraction during co-agulation,which causes an increase in the relative proportion of aliphat-ic DOC, could be responsible for increased % BrTHMs since, under theseconditions, the competitive incorporation of HOBr is likely to be en-hanced. This may have implications for human health since brominatedTHMs are reported to be more genotoxic than their chlorinated ana-logues (Richardson et al., 2007). This issue is particularly significant inthe context of a predicted increase in sea salt deposition with risingsea surface temperatures whichmay increase Br− concentration in sur-face waters (Hurrell et al., 2004).

5. Conclusions

Water treatment is becoming increasingly important due to thegrowing global population and climate change. Chlorine-based disinfec-tion, whose benefits include low cost, and high efficacy and lifetime,continues to be widely-used. Given the widespread trend of risingDOC levels in surface waters, the predicted increase in the seasonalityof DOC loading under future climate scenarios and increasingly frequentextreme weather events, disinfection by-products arising from chlori-nation are likely to become an increasing problem. In this context, ourdata show high DOC across the sampling period (9.0–16.0 mg L−1)for this coniferous peatland catchmentwith higher DOC levels inwinterand lower DOC during the summer. However the STHMFP, which corre-latedwith SUVA as previously reported, seemed to be amore importantfactor in driving variations in THM risk (THMFP) compared with DOClevel. Importantly, variations in the reaction kinetics of THM formationidentified in this study, most likely driven by seasonal changes in the

proportions of fast- and slow-reacting THM precursors, will also affectTHM levels at the point of delivery.

Coagulation–flocculation and DAF are the first steps of water treat-ment at this site providing consistently effective DOC removal (mean76%), significantly reducing THMFP from raw to post-1RGFwater. How-ever, these processes were less successful in the targeted removal ofTHM precursors, leading to a high THM risk in summer and autumn.Mitigation strategies, such as switching disinfection to chloramination,or the introduction of more selective DOC treatments may become nec-essary in the future. Interestingly, the early treatments (coagulation,flocculation and DAF) also resulted in a doubling of the proportion ofBrTHMs, it is suggested, due to the selective removal of aromatic DOCand the consequent increase in the proportion of aliphatic DOC. Bromi-nated THMs are reported to be more carcinogenic than CHCl3, and thisissue is particularly significant given that climate change is predictedto result in increased sea salt (and hence Br) deposition withincatchments.

Unexpectedly, the combined effect of chlorination and secondaryRGFs also produced a small but statistically significant change in DOCquality, indicated by a fall in SUVA and A253:A203, as well as a slight re-duction in the proportion of BrTHMs, possibly due to manganese re-moval chemistry.

Acknowledgements

This research was part-funded by the European Social Fund (ESF)through the European Union's Convergence programme administeredby the Welsh Government. Match funding and access to water samplesand water quality data was provided by Dŵr Cymru Welsh Water(DCWW). Christopher Freeman and Peter Holliman acknowledgefunding from NERC under the first EU ERA-EnvHealth call (FP7-ENV-2007- CSA-1.2.3-01). We would like to thank David Hughes, School ofBiological Sciences, Bangor University, for assistance with the fraction-ation procedure, Nigel Brown, Treborth Botanical Gardens, for analysingalgal assemblages in reservoir samples, and Carol Weatherley and RhysLewis (DCWW) for their assistance with this research.

References

Adin A, Katzhendler J, Alkaslassy D, Rav-Acha C. Trihalomethane formation in chlorinateddrinking water: a kinetic model. Water Res 1991;25(7):797–805.

Bolto B, Gregory J. Organic polyelectrolytes in water treatment. Water Res 2007;41(11):2301–24.

Bond T, Henriet O, Goslan EH, Parsons SA, Jefferson B. Disinfection byproduct formationand fractionation behavior of natural organic matter surrogates. Environ Sci Technol2009;43(15):5982–9.

Bond T, Goslan EH, Parsons SA, Jefferson B. Treatment of disinfection by‐product precur-sors. Environ Technol 2011;32(1):1–25.

Bonferroni CE. Teoria statistica delle classi e calcolo delle probabilità. 8, Publicazioni del RInstituto Superiore di Scienze Economiche e Commerciali de Frienze; 19363–62.

Bouchard A. Recent lake acidification and recovery trends in southern Quebec, Canada.Water Air Soil Pollut 1997;94(3–4):225–45.

Box JD. Investigation of the folin-ciocalteau phenol reagent for the determination of poly-phenolic substances in natural waters. Water Res 1983;17(5):511–25.

Choi J, Valentine RL. Formation of N-nitrosodimethylamine (NDMA) from reaction withmonochloramine: a new disinfection by-product. Water Res 2002;36(4):817–24.

ChowAT, Tanji KK, Gao S. Production of dissolved organic carbon (DOC) and trihalomethane(THM) precursor from peat soils. Water Res 2003;37(18):4475–85.

Chow AT, Gao S, Dahlgren RA. Physical and chemical fractionation of dissolved organicmatter and trihalomethane precursors: a review. J Water Supply Res Technol AQUA2005;54(8):475–507.

Chow AT, Guo F, Gao S, Breuer RS. Size and XAD fractionations of trihalomethane precur-sors from soils. Chemosphere 2006a;62(10):1636–46.

Chow AT, Tanji KK, Gao S, Dahlgren RA. Temperature, water content and wet–dry cycleeffects on DOC production and carbon mineralization in agricultural peat soils. SoilBiol Biochem 2006b;38(3):477–88.

Chow CWK, Fabris R, van Leeuwen JA, Wang D, Drikas M. Assessing natural organic mat-ter treatability using high performance size exclusion chromatography. Environ SciTechnol 2008;42(17):6683–9.

Chow AT, Lee S, O'Geen AT, Orozco T, Beaudette D, Wong P, et al. Litter contributions todissolved organic matter and disinfection byproduct precursors in California oakwoodland watersheds. J Environ Qual 2009a;38(6):2334–43.

238 R. Gough et al. / Science of the Total Environment 468–469 (2014) 228–239

Chow CWK, van Leeuwen JA, Fabris R, Drikas M. Optimised coagulation using aluminiumsulfate for the removal of dissolved organic carbon. Desalination 2009b;245(1–3):120–34.

Christ MJ, David MB. Temperature and moisture effects on the production of dissolved or-ganic carbon in a spodosol. Soil Biol Biochem 1996;28(9):1191–9.

Clark JM, Lane SN, Chapman PJ, Adamson JK. Link between DOC in near surface peat andstream water in an upland catchment. Sci Total Environ 2008;404(2–3):308–15.

Cohen I. Spatial and temporal influences on the terrigenous carbon in reservoirs withinpeat-rich catchments. MPhil thesisBangor University; 2009.

Creed IF, Beall FD, Clair TA, Dillon PJ, Hesslein RH. Predicting export of dissolved organiccarbon from forested catchments in glaciated landscapes with shallow soils. GlobBiogeochem Cycles 2008;22(4).

Croue JP, Lefebvre E, Martin B, Legube B. Removal of dissolved hydrophobic and hydro-philic organic-substances during coagulation flocculation of surface waters. WaterSci Technol 1993;27(11):143–52.

Davies J-M, Roxborough M, Mazumder A. Origins and implications of drinking waterodours in lakes and reservoirs of British Columbia, Canada. Water Res 2004;38(7):1900–10.

Dickenson ERV, Summers RS, Croue J, Gallard H. Haloacetic acid and trihalomethaneformation from the chlorination and bromination of aliphatic beta-dicarbonyl acidmodel compounds. Environ Sci Technol 2008;42(9):3226–33.

DWI (Drinking Water Inspectorate). Water supply (water quality) regulations 2010:Water, England and Wales; 2010.

Edwards M. Predicting DOC, removal during enhanced coagulation. J Am Water WorksAssoc 1997;89(5):78–89.

Edzwald JK. Coagulation in drinking water treatment: particles, organics and coagulants.Control of organic material by coagulation and floc-separation processes: Proceedingsof the second conference of the IAWQ-IWSA joint specialist group on coagulation,flocculation, filtration, sedimentation and flotation, held in Geneva, Switzerland, 1–3September 1992. Pergamon Press; 1993.

Edzwald JK, Tobiason JE. Enhanced coagulation: US requirements and a broader view.Water Sci Technol 1999;40(9):63–70.

Edzwald JK, Becker WC, Wattier KL. Surrogate parameters for monitoring organic matterand THM precursors. J Am Water Works Assoc 1985;77(4):122–32. [RemovingTHMs].

Eimers MC, Buttle J, Watmough SA. Influence of seasonal changes in runoff and extremeevents on dissolved organic carbon trends in wetland-and upland-draining streams.Can J Fish Aquat Sci 2008;65(5):796–808.

Elliott AJ, Thackeray SJ, Huntingford C, Jones RG. Combining a regional climatemodel witha phytoplankton community model to predict future changes in phytoplankton inlakes. Freshwat Biol 2005;50(8):1404–11.

Fabris R, Chow CWK, Drikas M, Eikebrokk B. Comparison of NOM character in selectedAustralian and Norwegian drinking waters. Water Res 2008;42(15):4188–96.

Freeman C, Evans CD, Monteith DT, Reynolds B, Fenner N. Export of organic carbon frompeat soils. Nature 2001;412(6849):785.

Galapate RP, Baes AU, Ito K, Iwase K, Okada M. Trihalomethane formation potential pre-diction using some chemical functional groups and bulk parameters. Water Res1999;33(11):2555–60.

Gallard H, von Gunten U. Chlorination of natural organic matter: kinetics of chlorinationand of THM formation. Water Res 2002;36(1):65–74.

Gang DC, Clevenger TE, Banerji SK. Relationship of chlorine decay and THMs formation toNOM size. J Hazard Mater 2003;96(1):1–12.

Gao JP, Maguhn J, Spitzauer P, Kettrup A. Sorption of pesticides in the sediment of theTeufelsweiher pond (Southern Germany). II: competitive adsorption, desorption ofaged residues and effect of dissolved organic carbon.Water Res 1998;32(7):2089–94.

Gu B, Schmitt J, Chen Z, Liang L, McCarthy JF. Adsorption and desorption of different or-ganic matter fractions on iron oxide. Geochim Cosmochim Acta 1995;59(2):219–29.

Guay C, RodriguezM, Sérodes J. Using ozonation and chloramination to reduce the forma-tion of trihalomethanes and haloacetic acids in drinking water. Desalination2005;176(1–3):229–40.

Halliday SJ, Wade AJ, Skeffington RA, Neal C, Reynolds B, Rowland P, et al. An analysis oflong-term trends, seasonality and short-term dynamics in water quality data fromPlynlimon, Wales. Sci Total Environ 2012;434:186–200. [SI].

Harrington GW, Bruchet ARD, Singer PC. Characterization of natural organic matter andits reactivity with chlorine. ACS Symposium Series. ACS Publications; 1996.p. 138–58.

Hejzlar J, Dubrovský M, Buchtele J, Ruzicka M. The apparent and potential effects of cli-mate change on the inferred concentration of dissolved organicmatter in a temperatestream (the Malse River, South Bohemia). Sci Total Environ 2003;310(1–3):143–52.

Heller-Grossman L, Manka J, Limoni-Relis B, Rebhun M. Formation and distribution ofhaloacetic acids, THM and tox in chlorination of bromide-rich lake water. WaterRes 1993;27(8):1323–31.

Hope D, Billett MF, Cresser MS. A review of the export of carbon in river water: fluxes andprocesses. Environ Pollut 1994;84(3):301–24.

Hua G, Reckhow DA. Comparison of disinfection byproduct formation from chlorine andalternative disinfectants. Water Res 2007;41(8):1667–78.

Huang C, Shiu H. Interactions between alum and organics in coagulation. Colloids Surf APhysicochem Eng Asp 1996;113(1–2):155–63.

Huber SA, Balz A, Abert M, Pronk W. Characterisation of aquatic humic and non-humicmatter with size-exclusion chromatography – organic carbon detection – organicnitrogen detection (LC-OCD-OND). Water Res 2011;45(2):879–85.

Hurrell JW, Hoerling MP, Phillips AS, Xu T. Twentieth century North Atlantic climatechange. part I: assessing determinism. Clim Dyn 2004;23(3–4):371–89.

Ichihashi K, Teranishi K, Ichimura A. Brominated trihalomethane formation in halogena-tion of humic acid in the coexistence of hypochlorite and hypobromite ions. WaterRes 1999;33(2):477–83.

Imai A, Matsushige K, Nagai T. Trihalomethane formation potential of dissolved organicmatter in a shallow eutrophic lake. Water Res 2003;37(17):4284–94.

Iriarte-Velasco U, Álvarez-Uriarte JI, González-Velasco JR. Enhanced coagulation underchanging alkalinity-hardness conditions and its implications on trihalomethane pre-cursors removal and relationship with UV absorbance. Sep Purif Technol 2007;55(3):368–80.

Jenkins GL, Murphy JM, Sextpn DMH, Lowe JA, Jones P, Kilsby CG. Briefing report: UKclimate projections. Met Office Hadley Centre; 2009.

Kaiser K, Guggenberger G, Haumaier L, Zech W. Seasonal variations in the chemical com-position of dissolved organic matter in organic forest floor layer leachates ofold-growth Scots pine (Pinus sylvestris L.) and European beech (Fagus sylvatica L.)stands in northeastern Bavaria, Germany. Biogeochemistry 2001;55(2):103–43.

Kalbitz K, Schmerwitz J, Schwesig D, Matzner E. Biodegradation of soil-derived dissolvedorganic matter as related to its properties. Geoderma 2003;113(3–4):273–91.

KimH, YuM. Characterization of aquatic humic substances to DBPs formation in advancedtreatment processes for conventionally treated water. J HazardMater 2007;143(1–2):486–93.

Kitis M, Kilduff JE, Karanfil T. Isolation of dissolved organic matter (DOM) from surfacewaters using reverse osmosis and its impact on the reactivity of DOM to formationand speciation of disinfection by-products. Water Res 2001;35(9):2225–34.

Korshin GV, Li CW, Benjamin MM. Monitoring the properties of natural organic matterthrough UV spectroscopy: a consistent theory. Water Res 1997;31(7):1787–95.

Krasner SW, Amy G. Jar-test evaluations of enhanced coagulation. J Am Water WorksAssoc 1995;87(10):93–107.

Kristiana I, Gallard H, Joll C, Croué J. The formation of halogen-specific TOX from chlorinationand chloramination of natural organicmatter isolates.Water Res 2009;43(17):4177–86.

Leenheer JA, Croue J. Characterising dissolved aquatic organic matter. Environ Sci Technol2003;37(1):18A–26A.

Letterman RD, Vanderbrook SG. Effect of solution chemistry on coagulation with hydro-lyzed Al(III): significance of sulfate ion and pH. Water Res 1983;17(2):195–204.

Lindroos A, Derome J, Derome K, Smolander A. The effect of Scots pine, Norway spruceand silver birch on the chemical composition and stand throughfall and upper soilpercolation water in northern Finland. Boreal Environ Res 2011;16:240–50.

Liu W, Wu H, Wang Z, Ong SL, Hu JY, Ng WJ. Investigation of assimilable organic carbon(AOC) and bacterial regrowth in drinking water distribution system. Water Res2002;36(4):891–8.

Lu J, Zhang T, Ma J, Chen Z. Evaluation of disinfection by-products formation during chlo-rination and chloramination of dissolved natural organic matter fractions isolatedfrom a filtered river water. J Hazard Mater 2009;162(1):140–5.

Marhaba TF, Pu Y, Bengraine K.Modified dissolved organic matter fractionation techniquefor natural water. J Hazard Mater 2003;101(1):43–53.

Matilainen A, Vepsäläinen M, SillanpääM. Natural organic matter removal by coagulationduring drinking water treatment: a review. Adv Colloid Interface Sci 2010;159(2):189–97.

Matilainen A, Gjessing ET, Lahtinen T, Hed L, Bhatnagar A, SillanpääM. An overview of themethods used in the characterisation of natural organic matter (NOM) in relation todrinking water treatment. Chemosphere 2011;83(11):1431–42.

Mitchell G, McDonald AT. Discolouration of water by peat following induced drought andrainfall simulation. Water Res 1992;26(3):321–6.

Monteith DT, Stoddard JL, Evans CD, deWit HA, Forsius M, Hogasen T, et al. Dissolved or-ganic carbon trends resulting from changes in atmospheric deposition chemistry. Na-ture 2007;450(7169):537–40.

Neal C, Robson AJ, Neal M, Reynolds B. Dissolved organic carbon for upland acidic and acidsensitive catchments in mid-Wales. J Hydrol 2005;304(1–4):203–20.

Owen DM, Amy GL, Chowdhury ZK, Paode R, McCoy G, Viscosil K. NOM characterizationand treatability. J Am Water Works Assoc 1995;87(1):46–63.

Page DW, van Leeuwen JA, Spark KM, Drikas M, Withers N, Mulcahy DE. Effect of alumtreatment on the trihalomethane formation and bacterial regrowth potential of nat-ural and synthetic waters. Water Res 2002;36(19):4884–92.

Parks SJ, Baker LA. Sources and transport of organic carbon in an Arizona river-reservoirsystem. Water Res 1997;31(7):1751–9.

Piirsoo K, Viik M, Kõiv T, Käiro K, Laas A, Nõges T, et al. Characteristics of dissolved organicmatter in the inflows and in the outflow of Lake Võrtsjärv, Estonia. J Hydrol 2012;475:306–13.

Pokrovsky OS, Schott J. Iron colloids/organic matter associated transport of major andtrace elements in small boreal rivers and their estuaries (NW Russia). Chem Geol2002;190(1–4):141–79.

Prévost M, Rompré A, Coallier J, Servais P, Laurent P, Clément B, et al. Suspended bacterialbiomass and activity in full-scale drinking water distribution systems: impact ofwater treatment. Water Res 1998;32(5):1393–406.

Randtke SJ. Organic contaminant removal by coagulation and related process combina-tions. J Am Water Works Assoc 1988;80(5):40–56.

Richardson SD, Plewa MJ, Wagner ED, Schoeny R, Demarini DM. Occurrence,genotoxicity, and carcinogenicity of regulated and emerging disinfection by-products in drinking water: a review and roadmap for research. Mutat Res2007;636(1–3):178–242.

Ritson JP, Graham NJD, Templeton MR, Clark JM, Freeman C, Gough R. The impact of cli-mate change on the treatability of dissolved organic matter (DOM) in upland watersupplies: A UK perspective. Unpublished results.

Rook JJ. Formation of haloforms during chlorination of natural waters. J Soc Water TreatExam 1974;23:234–43.

Rothwell JJ, Evans MG, Daniels SM, Allott TEH. Baseflow and stormflow metal concentra-tions in streams draining contaminated peat moorlands in the Peak District NationalPark (UK). J Hydrol 2007;341(1–2):90–104.

Runkana V, Somasundaran P, Kapur PC. A population balance model for flocculation ofcolloidal suspensions by polymer bridging. Chem Eng Sci 2006;61(1):182–91.

239R. Gough et al. / Science of the Total Environment 468–469 (2014) 228–239

Sharp EL, Jarvis P, Parsons SA, Jefferson B. Impact of fractional character on the coagula-tion of NOM. Colloids Surf A Physicochem Eng Asp 2006a;286(1–3):104–11.

Sharp EL, Parsons SA, Jefferson B. The impact of seasonal variations in DOC arising from amoorland peat catchment on coagulation with iron and aluminium salts. EnvironPollut 2006b;140(3):436–43.

Sharp EL, Parsons SA, Jefferson B. Seasonal variations in natural organic matter and its im-pact on coagulation in water treatment. Sci Total Environ 2006c;363(1–3):183–94.

Sobek S, Tranvik LJ, Prairie YT, Kortelainen P, Cole JJ. Patterns and regulation of dissolvedorganic carbon: an analysis of 7500 widely distributed lakes. Limnol Oceanogr2007;52(3):1208–19.

Standing Committee of Analysts. Chloro and bromo trihalogenated methanes in water1980. Methods for the examination of waters and associated materials. London, UK:HMSO; 1981.

Stoddard JL, Kahl JS, Deviney FA, Dewalle DR, Driscoll CT, Herlihy AT. Response of surfacewater chemistry to the clean air act amendments of 1990. Environmental ProtectionAgency; 2003 [Report nr EPA/620/R-03/001].

Symons JM, Bellar TA, Carswell JK, Demarco J, Kropp KL, Kropp KL, et al. National organicsreconnaissance survey for halogenated organics. J Am Water Works Assoc 1975;67:634–47.

Tang R, Clark JM, Bond T, Graham N, Hughes D, Freeman C. Assessment of potential cli-mate change impacts on peatland dissolved organic carbon release and drinkingwater treatment from laboratory experiments. Environ Pollut 2013;173:270–7.

Teixeira MR, João RM. Comparing dissolved air flotation and conventional sedimentationto remove cyanobacterial cells ofMicrocystis aeruginosa Part I: the key operating con-ditions. Sep Purif Technol 2006;52(1):84–94.

Teksoy A, Alkan U, Başkaya HS. Influence of the treatment process combinations on theformation of THM species in water. Sep Purif Technol 2008;61(3):447–54.

Thurman EM, editor. Organic geochemistry of natural waters. Dordrecht, Netherlands:Kluwer Academic Publishers; 1985.

Thurman EM, Malcolm RL. Preparative isolation of aquatic humic substances. Environ SciTechnol 1981;15(4):463–6.

Tomaszewska M, Mozia S, Morawski AW. Removal of organic matter by coagulationenhanced with adsorption on PAC. Desalination 2004;161(1):79–87.

US EPA (United States Environmental Protection Agency). Enhanced coagulation andenhanced precipitative softening guidance manual; 1999.

Uyak V, Toroz I. Disinfection by-product precursors reduction by various coagulationtechniques in Istanbul water supplies. J Hazard Mater 2007;141(1):320–8.

Volk C, Wood L, Johnson B, Robinson J, Zhu HW, Kaplan L. Monitoring dissolved organiccarbon in surface and drinking waters. J Environ Monit 2002;4(1):43–7.

Waiser MJ, Robarts RD. Photodegradation of DOC in a shallow prairie wetland: evidencefrom seasonal changes in DOC optical properties and chemical characteristics.Biogeochemistry 2004;69(2):263–84.

Weishaar JL, Aiken GR, Bergamaschi BA, FramMS, Fujii R, Mopper K. Evaluation of specificultraviolet absorbance as an indicator of the chemical composition and reactivity ofdissolved organic carbon. Environ Sci Technol 2003;37(20):4702–8.

Wetzel RG. Limnology: lake and reservoir ecosystems. 3rd ed. California: Elsevier; 2001.White MC, Thompson JD, Harrington GW, Singer PC. Evaluating criteria for enhanced co-

agulation compliance. J Am Water Works Assoc 1997;89(5):64–77.WHO (World Health Organisation). Guidelines for drinking-water quality; 2011.Winchester JW, Duce RA. Coherence of iodine and bromine in the atmosphere of Hawaii,

northern Alaska, and Massachusetts. Tellus 1966;18(2–3):287–92.Worrall F, Reed M, Warburton J, Burt T. Carbon budget for a British upland peat catch-

ment. Sci Total Environ 2003;312(1–3):133–46.Yan M, Wang D, Yu J, Ni J, Edwards M, Qu J. Enhanced coagulation with

polyaluminum chlorides: role of pH/Alkalinity and speciation. Chemosphere2008;71(9):1665–73.

YanM,Wang D, Ni J, Qu J, NiW, Van Leeuwen J. Natural organic matter (NOM) removal ina typical North-China water plant by enhanced coagulation: targets and techniques.Sep Purif Technol 2009;68(3):320–7.

Zhang H, Qu J, Liu H, Wei D. Characterization of dissolved organic matter fractions and itsrelationship with the disinfection by-product formation. J Environ Sci 2009;21(1):54–61.

Zou L, Wang XC, Callahan J, Culp RA, Chen RF, Altabet MA, et al. Bacterial roles in the for-mation of high-molecular-weight dissolved organic matter in estuarine and coastalwaters: evidence from lipids and the compound-specific isotopic ratios. LimnolOceanogr 2004;49(1):297–302.