1

Phosphorusfluxestotheenvironmentfrommainswaterleakage:Seasonalityandfuturescenarios1

2

Ascott,M.J.a,Gooddy,D.C.a,Lapworth,D.J.a,Davidson,P.b,Bowes,M.J.c, Jarvie,H.P.candSurridge,3

B.W.J.d4

5

aBritishGeologicalSurvey,MacleanBuilding,Crowmarsh,Oxfordshire,OX108BB,UnitedKingdom.6

bEnvironment Agency, KingsMeadowHouse, KingsMeadow Road, Reading, Berkshire, RG1 8DQ,7

UnitedKingdom.8

c Centre for Ecology & Hydrology, Maclean Building, Crowmarsh, Oxfordshire, OX10 8BB, United9

Kingdom.10

dLancasterEnvironmentCentre,LancasterUniversity,Lancaster,LA14YQ,UnitedKingdom.11

12

CorrespondingAuthor:Ascott,M.J.*13

a British Geological Survey, Maclean Building, Crowmarsh Gifford, Oxfordshire OX10 8BB, UK.14

matta@bgs.ac.uk,+44(0)149169240815

16

17

18

19

20

2

Abstract21

Accuratequantificationof sourcesofphosphorus (P)entering theenvironment isessential for the22

managementofaquaticecosystems.Pfluxesfrommainswaterleakage(MWL-P)haverecentlybeen23

identified as a potentially significant source of P in urbanised catchments. However, both the24

temporal dynamics of this flux and the potential future significance relative to P fluxes from25

wastewatertreatmentworks(WWT-P)remainpoorlyconstrained.UsingtheRiverThamescatchment26

inEnglandasanexemplar,wepresentthefirstquantificationofboththeseasonaldynamicsofcurrent27

MWL-Pfluxesandfuturefluxscenariosto2040,relativetoWWT-PloadsandtoPloadsexportedfrom28

thecatchment.ThemagnitudeoftheMWL-Pfluxshowsastrongseasonalsignal,withpipeburstand29

leakageeventsresultinginpeakPfluxesinwinter(December,January,February)thatare>150%of30

spring(March,April,May)/autumn(September,October,November)fluxes.WeestimatethatMWL-31

P isequivalenttoupto20%ofWWT-Pduringpeak leakageevents. Winterrainfalleventscontrol32

temporal variation in bothWWT-P and riverine P fluxes which consequently masks any signal in33

riverinePfluxesassociatedwithMWL-P.TheannualaverageratioofMWL-PfluxtoWWT-Pflux is34

predictedtoincreasefrom15to38%between2015and2040,associatedwithlargeincreasesinP35

removal at wastewater treatment works by 2040 relative to modest reductions in mains water36

leakage.However,furtherresearchisrequiredtounderstandthefateofMWL-Pintheenvironment.37

FuturePresearchandmanagementprogrammesshouldmorefullyconsiderMWL-Panditsseasonal38

dynamics,alongsidethelikelyimpactsofthissourceofPonwaterquality.39

Highlights40

• Seasonality+scenariosinPfluxesfrommainswaterleakage(MWL-P)quantified41

• MWL-PcomparedtowastewaterPflux(WWT-P)andcatchmentPexport42

• WinterburstMWL-Pflux>150%ofspring/autumnflux,equaltoupto20%ofWWT-P43

• MWL-P/WWT-Pratiopredictedtoincreaseto38%by2040duetoWWTPremoval44

• MWL-PfluxandseasonalityshouldbeconsideredinfuturePresearchandmanagement45

3

Keywords46

Phosphorus,sourceapportionment,eutrophication,mainswater,leakage47

48

GraphicalAbstract49

50

51

52

53

54

55

56

57

58

1500

2000

2500

Was

tew

ater

trea

tmen

t (kg

P/d

)15

020

025

030

0M

ains

leak

age

(kg

P/d)

2005 2006 2007 2008 2009 2010 20110

200

400

600

800

1000

1200

1400

1600

1800

2015 2040

P flu

x (k

g P/

d)

Wastewater Treatment WorksMains Leakage

4

1 Introduction59

60

Eutrophication associatedwith phosphorus (P) export fromagricultural land andwithwastewater61

treatmenteffluentisawell-knowncauseofwaterqualityimpairment(Elseretal.2007,Hiltonetal.62

2006,Vollenweider1968),thatcanresultinharmfulalgalblooms,fishkillsandadverseimpactson63

animalandhumanhealth(FunariandTestai2008).Effluentfromwastewatertreatmentworkshas64

beenrecognisedasamajorsourceofPattheglobalscale(Moréeetal.2013)whichhasbeenshown65

toimpairwaterqualityinriversintheEurope(Jarvieetal.,2006),Asia(Lietal.2015),andtheUSA66

(Dubrovskyetal.2010).IntheUK,Pfluxesfromwastewatereffluenthasbeenestimatedtoaccount67

for 78% of annual riverP loads (WhiteandHammond2009). The implementationof theUrban68

WastewaterTreatmentDirective(EuropeanUnion1991)resultedinsubstantialdecreasesinPloads69

fromUKwastewater treatmentworks (WWT-P), leading to largedecreases in fluvialP fluxes. For70

example,in2015,fluvialfluxesofPinEnglandandWaleswereestimatedtohavedeclinedtolessthan71

50%of1974fluxes(Worralletal.2015).72

Recentstudieshaveshownthatincatchmentswithalargeproportionofurbanlandcover,leakageof73

P added to mains water (MWL-P) as phosphate is a potentially significant but previously largely74

overlookedsourceofPwithintheenvironment(Gooddyetal.2017).Phosphateisaddedtomains75

watertoreduceplumbosolvency(Hayes2010),withtherisksassociatedwithleadindrinkingwater76

derived from pipework being a significant public health concern (Edwards et al. 2009). Lead77

consumption in humans has been associated with cognitive development problems in children78

(Bellinger et al. 1987) as well as increased risk of heart disease and stroke (Pocock et al. 1988).79

Phosphateisaddedtomainswatertoconvertleadcarbonatetoleadphosphate.Leadphosphateis80

anorderofmagnitudelesssolublethanleadcarbonateandresultsintheformationofleadphosphate81

precipitates on internal pipe surfaces (Hayes 2010). In theUK it is estimated that >95% ofwater82

suppliesaredosedwithphosphate(CIWEM2011).IntheUSA,overhalfofsuppliesaredosed(Dodrill83

5

andEdwards1995). Dosingachieves finalP concentrations in tapwater thatareestimated tobe84

between700–1900µgP/L(UKWater IndustryResearchLtd2012),withphosphatedosinghaving85

beenahighly successfulengineering solution to reduce leadconcentrations indrinkingwater. For86

example,in2011,99.0%ofrandomtapwatersamplesinEnglandandWalesmetthedrinkingwater87

limitof10µgPb/L(CIWEM2011).88

Leakageofmainswaterisagloballysignificantchallenge,withthecostofnon-revenuewater(leakage89

andunbilledconsumption)estimated tobe$14billionperyear (WorldBank2006).National-scale90

leakage rates in England andWales are reported to be up to 25% of the water that enters the91

distributionnetworkandwatercompaniesplanfutureleakageratesovera25-yearplanninghorizon92

(2015-2040).Leakageofphosphate-dosedmainswaterasapotentiallyimportantsourceofPinthe93

environmentwas firsthypothesisedbyHolmanet al. (2008). Gooddyet al. (2015)made the first94

quantificationofthisfluxonanannualbasisforEnglandandWales,whichwassubsequentlyrefined95

byAscottetal.(2016a).96

97

However, understanding the temporal variability of MWL-P is also important, because biological98

impacts of P in aquatic ecosystems also show significant temporal variability, with autotrophic99

biomass growth occurring inmany systems primarily during spring (March, April,May) and early100

summer(June,July,August)(Bowesetal.2012a).Anumberofstudies(Bireketal.2014,Cocksand101

Oakes 2011, UKWater Industry Research Ltd 2007, 2013) have highlighted seasonal dynamics in102

leakageofwaterfrommainssupplies.Higherratesofbothburstsandlow-levelcontinuous“invisible”103

leakageoccurduringwinter(December,January,February),duetopipecontractionandexpansion.104

However,workonPtodatehasonlyquantifiedtheannualfluxofPfrommainswaterleakageandthe105

temporaldynamicsof this sourcearepoorlyconstrained. Whilst substantial reductions inWWT-P106

loadstorivershavebeenachieved(Worralletal.2015),elevatedPconcentrationsremainasignificant107

concern, with 40% of the water bodies in England not achieving “good status” under theWater108

6

Framework Directive (European Union 2000) due to high reactive P concentrations (Environment109

Agency 2015). Consequently, in England andWales, substantial further investment inwastewater110

treatmentisplannedto2020withanestimatedtotalexpenditureof£2.1billion(GlobalWaterIntel111

2014).Similarprogrammesexist internationally, forexample in theUSA(SewageWorld2016)and112

across Europe (Water Technology 2016). In parallel,water companies continue to reduce leakage113

basedonwatersavingdrivers,whichwillinturnreduceMWL-Pfluxes.Despitetheseprogrammes,114

noestimatesofhowtherelativecontributionsofMWL-PandWWT-Ptotheenvironmentmaychange115

inthefuturehavebeenmade.IfPsourcestoaquaticecosystemsaretobemanagedeffectively,and116

the impactsofMWL-Preduced, it isessential thatthetemporaldynamicsofMWL-Pandpotential117

futurechangesinfluxesfromthissourceofParebetterconstrained.Theresearchreportedherewas118

undertakentoexaminethefollowinghypotheses:119

1. MWL-Pfluxesshowseasonaltrends,withincreasedfluxesinwinterassociatedwithincreased120

burstsandinvisibleleakage121

2. SeasonalityinMWL-PfluxescanbedistinguishedfromtemporalvariabilityinotherPsources122

suchaswastewatertreatmentworkseffluent123

3. TherelativeimportanceofMWL-Pfluxeswillincreaseinthefuture,asWWT-Pfluxesdecrease124

duetoimprovementsintertiarytreatment.125

126

Using the River Thames, a large lowland catchment in the UK, as an exemplar, we tested these127

hypotheses by deriving historic, daily MWL-P, WWT-P and riverine P fluxes. Subsequently, we128

validatedourapproachbycomparingourderivedannualfluxesofPfromMWLandWWTsourcesto129

fluxes reported in previous studies.We then undertook time series analysis usingmultiple linear130

regression modelling to develop an improved understanding of the temporal dynamics and the131

processescontrollingMWL-P,WWT-PandriverinePfluxes.Further,wederivedfuturescenariosfor132

MWL-PandWWT-P,basedon25yearplansforleakagereductionandWWT-Premovalrespectively133

7

aspublishedbytheUKenvironmentalregulatorandwatercompanies.Finally,weprovideasummary134

ofkeyresearchprioritiesinthecontextofMWL-P,andconsiderhowtomanagethissourceofPacross135

arangeofdifferenthydro-socioeconomicsettings (countrieswhichcurrentlyundertakephosphate136

dosing,countriesconsideringthisstrategy,countriesactivelyavoidingit).137

138

2 MaterialsandMethods139

2.1 Studyarea140

141

WequantifieddailyMWL-P,WWT-PandriverinePfluxesforthenon-tidalThamescatchment(Figure142

1). The Thames is a relatively large (9948 km2) lowland catchment in southern England. The143

catchmentispredominantlyunderlainbytheCretaceousChalk,withOoliticlimestonesattheheadof144

thecatchmentandareasoflowpermeabilityclaysaroundOxfordandLondon.Meanannualrainfall145

in the catchment at Oxford is 745mm (Marsh andHannaford 2008). Although the catchment is146

relativelyrural,withc.45%oflandclassifiedasarable(Fulleretal.2013),therearealsoanumberof147

majorurbanareas(London,Oxford,Reading)resultinginahighcatchmentpopulationdensity(c.960148

people/km2(Merrett2007).149

150

Watersupplyinthecatchmentisprovidedbyfourdifferentwatercompanies(ThamesWater,Affinity151

Water,SuttonandEastSurreyWaterandSouthEastWater)whichoperateacrossatotalof10largely152

self-contained water resource zones (WRZs, Environment Agency (2012)). Each individual water153

companyreportsannualleakagerates.Leakageratesarereportedtobehigh,reachingupto26%of154

water input to thedistributionnetwork (CommitteeonClimateChange2015).Thecatchmenthas155

beensubjecttopreviousresearchtoestimatetheannualcontributionofMWL-PandWWT-Ptothe156

8

RiverThames(Gooddyetal.2017).Buildingonthis initialresearch,Ascottetal.(2016a)estimated157

thatc.30%oftheMWL-PfluxinEnglandandWaleswasderivedfromtheriverbasindistrictinwhich158

thenon-tidalThamescatchmentislocated.Wastewatertreatmentoccursthroughoutthecatchment159

and137watercompanywastewatertreatmentworksserving>2000populationequivalents(p.e)are160

present.Sincethelate1990s,PfluxesfromWWTshavebeenreducedinthecatchmentthroughthe161

introductionoftertiarytreatment,primarilydrivenbytheEUUrbanWasteWaterTreatmentDirective162

(Kinniburgh and Barnett 2010, Powers et al. 2016). Consequently, P concentrations in the River163

Thamesfellbyc.90%between1992and2005(KinniburghandBarnett2010).Despitethis,thereis164

scantevidencethateutrophicationriskintheThameshasreduced(Bowesetal.2014),withanumber165

ofrivers inthecatchmentexhibitingexcessivephytoplanktonbiomassandPconcentrationsabove166

thoseatwhichthegrowthofautotrophscanbeexpected(>c.80µgSRP/l,Bowesetal.(2012b)).167

168

(b)(a)

Rivers

Water Resource Zone Boundaries

Thames Catchment

Urban Areas

WWTsCatchment Export Monitoring PointOxford Weather Station

169

Figure 1 Location of the non-tidal Thames catchment within England (a) and the water resource zones, wastewater170treatmentworks serving > 2000 p.e., catchment exportmonitoring point and theOxfordweather stationwithin the171catchment(b).ContainsOrdnanceSurveydata©Crowncopyrightanddatabaseright(2017).172

173

9

174

2.2 Pfluxtimeseriesderivationandanalysis175

2.2.1 Datasources176177

Table1summarisesthedatasourcesusedtoderivedailytimeseriesofMWL-P,WWT-Pandriverine178

Pfluxesinthisstudyfortheperiod2001-2011.MWL-Pwasestimatedbasedonpublishedleakage179

ratesandestimatesofPconcentrationsindrinkingwater.WaterqualitysamplingofWWTdischarges180

andtheRiverThamesoccursatamuchlowerfrequency(c.every10daystomonthly)incomparison181

tothedailyflowdatathatareavailable.Consequently,toderivedailyWWT-PandriverinePfluxes,182

missing concentration data were estimated using concentration-flow relationships, as detailed in183

section2.2.3.184

185

Table1DatasourcesusedtoderivedailytimeseriesofMWL-P,WWT-PandriverinePfluxesfor2001-2011.186

Flux DataTimeperiod Frequency Source

MWL-P

WRZlevelleakageratesforWRZsinnon-tidalThamescatchment

2011-2013 Annual WRMPtables

LeakageratesforThamesWater

2001-2011 Daily

CocksandOakes(2011)

PO4-Pdosingconcentration

2001-2011 Annual Comberetal.(2011)

PO4-Pdosingextent2001-2011 Annual

Hayesetal.(2008),CIWEM(2011)

RiverineP

PO4-PconcentrationforThamesatTeddington

2000-2015 Monthly

EnvironmentAgencyFlowsfortheThamesatTeddington

2000-2015 Daily

WWT-P PO4-PWWTdischargeconcentrations

2005-2011

Variable(meanfrequency=10days,medianfrequency=27days)

EnvironmentAgencyWWTdischargeflows2005-2011 Daily

187

10

188

189

2.2.2 MWL-P190

191

Annualwaterresourcezonelevelleakagerates(inML/day)for2011-2013wereextractedfromwater192

resourcesplanningtablesavailableonwatercompanywebsites(AffinityWater2014,SoutheastWater193

2014,SuttonandEastSurreyWater2014,ThamesWater2014).Dailyleakagedatafor2001to2011194

(asreportedbyCocksandOakes(2011))forthewholeThamesWatersupplyareawereusedasthe195

basis forestimatingtheMWL-Pflux.This is theonlypublicly-availablesub-annual leakagedata for196

boththecatchmentandtheUK.TheThamesWatersupplyareacovers76%ofthenon-tidalThames197

catchment that is the focus for the research reported here, but also includes London in the198

downstreamtidalThamescatchment.Toaccountforthisdiscrepancy,wecalculatedtheratioofthe199

annualleakageratesinthewaterresourcesplanningtablesforthewholeThamesWatersupplyarea200

versus the10WRZs in thenon-tidalThamescatchment.Wethenusedthis ratio toscale thedaily201

ThamesWaterleakagetimeseries(CocksandOakes2011)toprovideadailyleakagetimeseriesfor202

thenon-tidalThamescatchmenttoTeddingtonforthesameperiod.Overtheperiod2001to2011,203

theextentofPdosingofmainswaterandtheconcentrationsusedhaveincreased.Basedonpublished204

reportsofdosingconcentrations(Comberetal.2011)andthespatialextentofdosing(CIWEM2011,205

Hayesetal.2008,UKWaterIndustryResearchLtd2012),weusedtheconservativeapproachadopted206

byGooddyetal.(2017)toestimatethehistoricchangesinPdosingconcentrationsandextentswithin207

thenon-tidalThamescatchment.208

209

2.2.3 RiverinePandWWT-Pfluxderivation,fluxcomparisonandtimeseriesanalysis210

211

RiverinePfluxderivation212

11

213

The daily net export of P (2001-2016) from the catchment was derived using observed214

orthophosphate-P (PO4-P) concentration and daily flow data for the River Thames at Teddington215

(Figure1). Inorderto infillmissingconcentrationdatatoderiveadailyPfluxtimeseries,wefirst216

undertooksinglechangepointdetectionusinganasymptoticpenaltyvalueof0.05(ChenandGupta217

2011,KillickandEckley2014)todeterminewhethertherewasastatisticallysignificantchangeinthe218

meanandvarianceoftheconcentrationtimeseriesassociatedwithhistoricreductionsinPloadsfrom219

WWTs(KinniburghandBarnett2010).FollowingtheapproachofBowesetal.(2014),wethenused220

non-linear least-squares regression toderive separateconcentration-flow relationshipsbeforeand221

afteranychangepointsthatwereidentified:222

𝐶𝐶 = 𝐴𝐴𝑄𝑄% (1)223

WhereCisconcentration(mgP/L),Qisflow(m3/s)andAandBareempiricallyderivedconstants.The224

derivedrelationshipswereusedtoestimatePconcentrationswheredataweremissingbeforeand225

afterthechangepoint.Thecompleteconcentrationandflowtimeserieswasthenusedtoderivea226

dailyPfluxtimeseries.227

228

WastewaterTreatmentWorksPFluxEstimation229

230

WastewatertreatmentworksintheThamescatchmentfallintotwocategories:(1)monitoredsites231

whereflowsare>50m3/d;and(2)unmonitored,smallsiteswithflows<50m3/d.Outputsfromthe232

ThamesSourceApportionmentGeographicalInformationSystems(SAGIS,Comberetal.(2013))tool233

usingpopulationequivalentdataandconcentrationestimateshaveshownthatsmallunmonitored234

sitescontribute<1%ofthetotalWWT-PloadtotheRiverThames.Consequently,thesesiteshavenot235

beenconsideredfurtherintheresearchreportedhere.236

237

12

ForthemonitoredWWTsites,dailyflowmonitoringdatafor2005–2011wereusedinconjunction238

with PO4-P concentration data to derive daily P fluxes for each site. A similar approach to that239

describedabovefortheriverinePfluxestimateswasusedforinfillingofmissingconcentrationdata240

usingconcentration-flowrelationshipsoftheformofequation1.ThedatesforimplementationofP241

removalschemesatWWTsintheThamescatchmentwereprovidedbytheenvironmentalregulator242

for England. Where P removal schemes were implemented at a WWT during the study period,243

separate concentration-flow relationships were estimated for before and after implementation.244

WherenochangesintreatmentprocessesforPremovalatWWTsoccurredduringthestudyperiod,245

asingleconcentration-flowrelationshipfortheWWTwasused.ThederiveddailyPfluxesforeach246

individualWWTwerethensummedtoderiveatotal,dailycatchmentfluxofWWT-P.247

248

Fluxcomparisonandtimeseriesanalysis249

250

TovalidateourdailyestimatesofMWL-PandWWT-P,wederivedannualmean fluxes foreachof251

thesePsourcesandthencomparedthesetopreviouslypublishedestimatesofthesamefluxesforthe252

Thamescatchment(Comberetal.2013,Gooddyetal.2017,Haygarthetal.2014).Wealsocalculated253

adailytimeseriesoftheratioofMWL-PtoWWT-P.ThederivedMWL-P,WWT-PandriverinePtime254

serieswere initiallycomparedbyvisual inspection toassess the timingandmagnitudeofpeaks in255

fluxes.Wethenassessedwhetherthereweresignificantlongtermtrendsandchangesinthevariance256

inMWL-PandMWL-P/WWT-P ratiousingaMannKendall test (Pohlert 2018) anda changepoint257

analysisusinganasymptoticpenaltyvalue=0.01(KillickandEckley2014).258

259

InordertoassessthecontrolofcoldweathereventsontemporalvariabilityinMWL-P,weundertook260

correlation analyses using historic dailyminimum temperature data for the Oxford station in the261

13

Thamescatchment(Figure1).WealsoquantifiedcorrelationsbetweendailyrainfallanddailyWWT-262

P anddaily riverineP loads to assess the roleof short term rainfall events in controllingP fluxes.263

Finally,todeterminehowmuchofthevariationindailyriverinePloadcouldbeexplainedbydaily264

MWL-PanddailyWWT-P,wedevelopedsimplemultiplelinearregressionmodelsusing(1)WWT-P265

onlyand(2)WWT-P+MWL-Pasexplanatoryvariables.Wecomparedthecoefficientofdetermination266

betweenthetwomodelsandusedapartialF-test(RDevelopmentCoreTeam2016)todeterminethe267

impactofaddingtheMWL-Pvariableonmodelpredictivepower.Thereissubstantialuncertaintyin268

thefateofMWL-PintheenvironmentassociatedwiththerelativeimportanceofdifferentPretention269

(claysorption,soilandsedimentaccumulation)andtransportprocesses(groundwater,surfacewater270

andthesewernetwork)(Gooddyetal.2017).Inthiscontext,ourresearchdoesnotaimtomakea271

directcausallinkbetweentimeseriesvariabilityinWWT-P,MWL-PandriverinePfluxes.Thepurpose272

the comparison of fluxes described here is threefold: (1) to provide indications of the relative273

contributionsof thesesources to theenvironment; (2) to improveunderstandingof theprocesses274

controllingtemporalvariabilityinthesesources;and(3)toimproveunderstandingoftherelationships275

betweenthesources.276

277

278

2.3 DerivationoffuturescenariosofMWL-PandWWT-P279280

Scenarios for future reductions in both MWL-P andWWT-P were constructed. Planned leakage281

reductions foreachof the10WRZs in theThamescatchment for2015-2040wereextracted from282

watercompanywaterresourceplanningtables(AffinityWater2014,SoutheastWater2014,Sutton283

andEastSurreyWater2014,ThamesWater2014). EstimatesoffuturereductionsinWWT-Pwere284

providedbytheenvironmentalregulatorasnewdischargepermitlimitsforP(inmgP/L).InEngland285

and Wales investments in wastewater treatment are planned on a five-year Asset Management286

14

Programme(AMP)cycle.Investmentsover2015–2020(AssetManagementProgramme6(AMP6))287

arefullyplannedandcosted. Furtherintothefuture,aprovisional“longlist”of improvementsin288

WWTsthatarepotentiallyrequiredinordertomeetgoodecologicalstatusinthecatchmenthasbeen289

derivedbasedonsourceapportionmentmodelling(Comberetal.2013).Thislisthasbeensubjectto290

acostbenefitanalysiswhichhasdividedsitesintothoseatwhichenhancedPremovaliscostbeneficial291

andthoseatwhichthisisnot.Intotal,fourWWTimprovementscenariosweredeveloped,asdetailed292

below:293

1. Baseline(2015)294

2. AMP6(2020)–plannedandcostedWWTimprovementsandleakagereductionprogrammes295

for2015–2020296

3. Costbeneficialplanningperiod(2040a)–plannedleakagereductionsandcostbeneficialWWT297

improvementsfor2015–2040298

4. Full planning period (2040b) – planned leakage reductions and all potential WWT299

improvementsfor2015–2040300

ScenarioswereimplementedbyapplyingthenewdischargepermitlimitsforWWTstothebaseline301

observed concentration-flow data in 2015. Given the uncertainty in future WWT and leakage302

reductionbeyond2020,anannualaverageapproachwasusedtoderiveannualfluxesfromMWL-P303

andWWT-Pin2015,2020andthetwo2040scenarios.304

305

306

307

308

15

3 Results309310

3.1 MWL-Pfluxestimation311

312

Figure 2 shows the evolution of the MWL-P flux over the period 2001 to 2011 for the Thames313

catchment. Figure2(c)showsthetotaldaily leakage(CocksandOakes2011) for2001–2011for314

ThamesWater. Over the period 2001 – 2011, the volumeof leakagehas declineddue towater315

companies actively locating and fixing leaks in the distribution network. After accounting for this316

reduction,theleakagetimeseriesshowsarelativelystrongcorrelationatlag=0withdailyminimum317

temperature(Pearsoncorrelation,r=0.63,p=1.392x10-14).Thetimeseriesshowsaclearseasonal318

signalassociatedwithincreasedburstsandinvisibleleakageinwinter.Thewintersof2009,2010and319

2011wereparticularlycold,withmeanDecembertemperatures1.1,2.0and4.8degreesbelowlong320

termaverage(1971-2000)respectively(MetOffice2016).Thisresultedinlargeincreasesinleakage321

of up to 50% compared to the preceding autumn months (September, October, November). In322

summer(June,July,August),smallincreasesofupto6%comparedtospring(March,April,May)values323

arealsoobserved,whichhasbeenattributedtosoilexpansionandcontractionresultinginpipefailure324

(CocksandOakes2011).Figure2(d)showstheestimateddailyfluxofMWL-Pfor2001-2011forthe325

Thamescatchment.Whilstthevolumeofleakagehasreducedsubstantially,increasesinthespatial326

extentofdosingand inphosphatedosingconcentrations counteract these reductions, resulting in327

significant(MannKendalltrendtest,p<2.2x10-16)netincreasesinMWL-Pfluxesofc.300%overthe328

period 2001 to 2011. Seasonality increases through time with a significant change in variance329

identifiedat08/09/2008(singlechangepointdetectionusinganasymptoticpenaltyvalue=0.01)and330

theratioofdailyannualmaximum/minimumMWL-Pratesincreasingfromc.130%in2002–2008to331

150%in2008–2011.332

333

16

334

Figure2:EstimatesofPdosingconcentrations(a)andextents(b)for2001–2011,dailyleakageratesreportedbyCocks335andOakes(2011)(c),andthederivedMWL-Pflux(d)336

337

338

339

3.2 WWT-Pfluxestimation340

341

Figure3reportsdatarelatedtotheWWT-PfluxfortheThamescatchment.Substantialdecreasesin342

PconcentrationinWWTeffluentsoccurredbetween2005and2007,associatedwiththeintroduction343

ofPremovaltechnology.FromMarch2008to2012,nosignificantfurtherPremovalwasimplemented344

in the catchment. Consequently, effluent P concentrations have been relatively stable and the345

frequency of P monitoring reduced. The slight increase (c. 5.4 to 5.6 mg P/L) in effluent P346

concentrationsover this periodmay reflect anoverall decrease inP removal efficiency in existing347

WWTsduetochangesinplantoperationandinfluentPloads(TchobanoglousandBurton2013).There348

700

800

900

1000

Dosi

ng (

g P/

L)μ

0.5

0.6

0.7

0.8

0.9

Dosi

ng e

xten

t (-)

2002 2004 2006 2008 2010

600

700

800

900

1000

Leak

age

(Ml/d

)

2002 2004 2006 2008 2010

200

300

400

500

600

700

800

900

MW

L-P

(kg

P/d)

(a) (b)

(c) (d)

17

isconsiderablevariationintheWWTflows(Figure3(a))particularlyduringhighflowperiodsinwinter349

associatedwithrainfalleventswhenflowscanbeuptotwiceaslargeasdryweatherflows(DWF).350

ThederivedWWT-Pflux(Figure3(c))reflectsbothlongtermchangesinPconcentrationsfollowing351

investmentinPremovaltechnologiesandshorttermvariabilityinWWTflowsassociatedwithrainfall352

events.353

354

355

Figure3TotalWWTflows(a),meandailyPconcentrationsandnumberofsamples(b)andthederivedWWT-Pfluxand356rainfalltimeseriesforOxford(c)357

358

3.3 Fluxcomparison359

360

Table2reportstheannualmeanMWL-PandWWT-Pfluxesderivedintheresearchreportedhereand361

inpreviousstudiesoftheThamescatchment.TheannualestimatesofMWL-Pinthisstudyarewithin362

therangeofvaluesreportedbyGooddyetal.(2017)whoconsideredarangeofdifferentplausible363

1000

000

2000

000

Tota

l WW

T F

low

(m

3 /d) (a)

05

1525

Num

ber o

f sam

ples

/day

5.2

5.6

6.0

6.4

Mea

n W

WT

conc

(mg

P/L

)

(b)

WWT conc.No. of samples

010

2030

40Ra

infa

ll (m

m)

2006 2008 2010 2012

1500

2500

WW

T flu

x (k

g P

/d) (c)

WWT fluxRainfall

1030

20

18

historicmainswaterdosingextents.WWT-Pfluxesbroadlycorroboratepreviousfindingsreportedby364

Haygarthetal.(2014)andComberetal.(2013).Fluxesestimatedinthecurrentstudyfor2006-2008365

aresomewhathigher(c.29%)thanestimatesmadeusingobservedsolublereactivePconcentrations366

inWWTdischargesinconjunctionwithsourceapportionmenttools(Comberetal.2013).Fluxesfrom367

thesourceapportionmentmodellingwereadjustedbytheUKenvironmentalregulator(Environment368

Agency 2016) to take into account new and enhanced P removalwhich is the likely cause of this369

discrepancy.Ourestimatesarealsolarger(c.50%)thanthosecalculatedbyHaygarthetal.(2014),370

whousedasimpleapproachbasedonpopulationequivalentsandassumedtotalPconcentrationsto371

estimateWWT-Pfluxforthecatchment.372

Table2AnnualmeanMWL-PandWWT-Pfluxesderivedinthisstudyandpreviousstudies373

374

PeriodMWL-P(ktP/yr) WWT-P(ktP/yr)

Gooddyetal.2017 Thisstudy

Haygarth et al2014

Comber et al2013 Thisstudy

2001 0.019-0.029 0.025 1.18 - -2006-2008 - - 0.52 0.672011 0.068-0.089 0.082 0.38 - 0.58

375

3.4 FluxcomparisonbetweenMWL-PandWWT-PwithRiverinePExport376

377

3.4.1 RiverinePfluxestimation378379

Figure4showstheresultsofthechangepointdetectionanalysisandthederivedconcentration-flow380

relationshipsbeforeandafter thechangepoint for theRiverThamesatTeddington.A substantial381

decrease in the mean and variance of riverine phosphate concentrations was observed in 2006,382

associatedwithreductionsinPloadsfromWWTs.Fortheperiod2000–2006,theconcentration-flow383

relationshipshowsdecreasesinconcentrationswithincreasingflow,associatedwithdilutionofWWT384

inputs. This corroborates concentration flow relationships derived by Bowes et al. (2014) in the385

19

Thamesbasin.From2006to2015,thevariabilityinconcentrationwithflowissubstantiallyreduced.386

Between0and100m3/ddecreases inconcentrationwith increasingflowareobserved,associated387

withdilutionofpoint sourceP inputs.However, from100 to500m3/d, there is limitedchange in388

concentration with increasing flow which reflects a balance of both point source dilution and389

mobilisationofdiffusePsourcesacrossthisrangeofdischarges.390

391

392

Figure4Observedphosphateconcentrations(mgP/L)intheRiverThamesatTeddingtonandmeanconcentrations393(dashedline)beforeandafterthechangepoint(a)andconcentration-flowrelationshipsforeachoftheseperiods(b).394

395

396

3.4.2 ComparisonofMWL-P,WWT-PandriverinePfluxes397398

Figure5showstheratioofMWL-PandWWT-Pfluxes(“MWL-P/WWT-P”)andtheirrelationshipwith399

Pflux intheRiverThamesatTeddington. From2005to2011there isasignificanttrend inMWL-400

P/WWT-P(Mann-Kendalltrendtest,p<2.2x10-16).TheratioofMWL-PtoWWT-Pincreasesfrom401

fromc.7-10in2005to15%in2011.Thisisduetotwofactors:(1)anincreaseintheextentofmains402

waterPdosingthroughtheperiod,asshowninFigure2;and(2)adecreaseinWWT-Ploadsdueto403

investmentinPremovaltechnology.404

2000 2005 2010 2015

0.0

0.2

0.4

0.6

0.8

1.0

1.2

Con

cent

ratio

n (m

g P/

L)

(a)

0 50 100 150 200 250 300 350

0.0

0.2

0.4

0.6

0.8

1.0

1.2

Flow (m3 /d)

Con

cent

ratio

n (m

g P

/L)

2000-2006, C = 1.036X-0.197

2006-2015, C = 0.343X-0.143

(b)

20

SeasonalityintheratioofMWL-PtoWWT-Pincreasesthroughtheperiod,duetotheincreasedextent405

ofdosing.ThereisasignificantincreaseinthevarianceofMWL-P/WWT-Pfrom22/11/2008onwards406

(singlechangepointdetectionusinganasymptoticpenaltyvalue=0.01(ChenandGupta2011,Killick407

andEckley2014)).PeaksinMWL-P/WWT-Pofupto20%occurinthewintersof2009and2011in408

comparisontovaluesaround10%in2005-2008.Thisseasonalityalsogivessomeinitialinsightinto409

theprocessescontrollingMWL-Pfluxes.Inthewinterof2009/10arapidincreaseinMWL-Poccurs410

followedbyarapidincreaseinWWT-P.ThiscausesMWL-P/WWT-Ptoriseandfallagainrapidly.Itis411

likelythesetrendsaretheresultofaparticularlycoldweatherperiodresultinginanincreaseinleaks412

andbursts.Thisislikelytobefollowedbyaperiodofactiveleakagecontroltoreduceleakagerates413

backtolevelspriortothecoldweatherperiod,ashasbeenreportedbyUKWaterIndustryResearch414

Ltd(2007). Atthesametimeaperiodofwetweatheroccurswhichresults in increasedinflowsto415

WWTs.ThisresultsinincreasesinWWToutflows(Figure3(a))butflowincreasesaresignificantlyless416

thantypicaldesignflowsforfulltreatmentofinfluentduringstorms(upto3timesDWF,with4–6417

timesDWFretained instormtanks(Sauletal.2007)). Any influent inexcessofstormtankdesign418

criteria as well as combined sewer overflows are likely to be discharged directly to the nearest419

watercourse.ThesewillallcombinetoreducetheeffectivenessoftheWWTstoremoveP.420

421

Figure5(a)and(d)showthedailyWWT-PfluxandcatchmentPexport.Thereareweakbutsignificant422

correlations betweenWWT-P anddaily rainfall (Pearson correlation, r = 0.23, p < 2.2 x 10-16) and423

catchment export anddaily rainfall (Pearson correlation, r = 0.18, p < 2.2 x 10-16). Figure6 shows424

observedandmodelledriverinePfluxesusingMLRmodelsconsidering:(1)WWT-Ponly;and(2)WWT-425

PandMWL-Ptogether.Themultiplelinearregressionmodelusedtodeterminewhethertheriverine426

P load can be explained by MWL-P and WWT-P resulted in a relatively modest coefficient of427

determination (adjusted R2 = 0.41). TheMLRmodel suggested thatWWT-P andMWL-P are both428

significantexplanatoryvariables(P<2x10-16). RemovingMWL-Pfromtheregressionsignificantly429

21

reduced thepredictivepowerof themodel (F=242,P<2x10-16, forpartial F-testbetweenMLR430

models considering WWT-P only and WWT-P + MWL-P) resulting in a lower coefficient of431

determination(adjustedR2=0.33).432

010

2030

40R

ainf

all (

mm

)

1000

2000

3000

WW

T flu

x (k

g P/

d)

(a)

WWT flux Rainfall

-50

510

1520

Tem

pera

ture

(°C

)

MWL-P flux Temperature

150

200

250

300

MW

L-P

Flux

(kg

P/d) (b)

0.05

0.15

0.25

MW

L-P/

STW

-P ra

tio

(c)

010

2030

40R

ainf

all (

mm

)

2005 2006 2007 2008 2009 2010 2011

020

0040

0060

00

Riv

er P

Flu

x (k

g P/

d)

(d) River flux Rainfall

433

Figure5DailyfluxesofWWT-Pandrainfall(a),MWL-Pandtemperature(b)fortheThamescatchment,theratioofMWL-434PtoWWT-P(c)andtheriverinefluxofPoutofthecatchmentandrainfall(d)435

22

436

Figure6:Observed(black)andmodelledriverinePfluxesfortheexportoftheThamescatchmentusingtheMLRmodel437consideringWWT-Ponly(blue)andWWT-PandMWL-P(red)438

439

440

3.5 FutureloadsofMWL-PandWWT-P441442

Table3reportsthederivedfutureMWL-PandWWT-Ploadsforannualaverageconditions.WWT-P443

fluxes are predicted to decrease to 2020 associated with the AMP6 programme, with further444

decreasespredictedto2040.WhenconsideringallWWTimprovementsto2040,atotaldecreasein445

WWT-Pfluxof0.41ktP/yrisestimated.SmallreductionsinMWL-Pof0.01ktP/yeararepredictedto446

occur to 2040. In total, the reduction in MWL-P loading is c. 2% of the reduction in WWT-P.447

Consequently, the ratio ofMWL-P toWWT-P is predicted to increase from c. 15% to 38% under448

averageconditions.449

450

451

2005 2006 2007 2008 2009 2010 2011

010

0020

0030

0040

0050

00

Riv

erin

e P

flux

(kg

P/d

)ObservedLinear model (WWT-P only)Linear model (WWT-P and MWL-P)

23

452

Table3FutureMWL-PandWWT-Ploadings453

454

ScenarioInformation FluxesMWL-P/WWT-P%

ScenarioName TimeSlice

WWTImprovements

LeakageImprovements

WWT-PFlux (ktP/yr)

MWL-PFlux (ktP/yr)

AnnualMean

Baseline 2015 - - 0.62 0.09 14.52

AMP6Programme 2020 AMP6

ProgrammeAMP6 LeakageProgramme 0.61 0.08 13.11

2040Plans(CostBeneficial)

2040

CBA"LongList"2040 Plan LeakageProgramme

0.57 0.08 14.04

2040 Plans (Allsites) Full"LongList" 0.21 0.08 38.10

TotalPfluximprovement(ktP/yr) 0.41 0.01

455

456

457

458

459

460

461

462

24

4 Discussion463

4.1 Seasonaldynamicsandrelative importanceofMWL-PandWWT-P to464

riverPloads465

466

EstimatesofMWL-PandWWT-PfluxesfortheThamescatchmentmadeintheresearchreportedhere467

illustratethatMWL-PisapotentiallysignificantsourceofPintheenvironmentwhichhassofarbeen468

largelyoverlookedinPsourceapportionmentandmanagementstrategies.Inparticular,thedistinct469

seasonalsignal(Figure2)inMWL-PresultsinhigherratiosofMWL-PtoWWT-Pthroughthewinter470

periods.TheseasonalityinMWL-Pthatwashypothesizedinsection1hasnotbeenreportedtodate471

andrepresentsadevelopmentinourunderstandingofthisPsource.472

473

TheprocessescontrollingseasonalityinMWL-Parecomplex.Whilstcorrelationsbetweenminimum474

air temperatures and MWL-P are reported in this paper, it should be noted that other non-475

meteorologicalprocessesalsohaveapotentiallysignificantimpactonMWL-Pvariability.Coldweather476

isthepredominantcontrolonthetimingandmagnitudeoftheinitialoutbreakofwatermainsbursts477

duringawinterleakageevent.Astemperaturesincrease,thereislikelytobesomenaturalreduction478

ininvisibleleakageassociatedwithpipeexpansion.However,ratherthananymeteorologicalfactor,479

themajorityofthereductioninleakageafteraneventistheresultofactiverepairstoburstwater480

mains(UKWaterIndustryResearchLtd2007).Thiscombinationofmeteorologicalandengineering481

processes is likely to be broadly applicable in developed countries with significant temperature482

fluctuationsandconsequentlyasimilarseasonalityinMWL-Pwouldbeexpected.Phosphatedosing483

isstartingtobedevelopedinothertemperatecountries(e.g.Ireland(IrishWater2015,Mockleretal.484

2017)).Inthesecountries,asdosingextentsandconcentrationsincrease,anincreaseintheboththe485

absolutemagnitudeandseasonalityofMWL-Pfluxes,asreportedinFigure2,wouldbeanticipated.486

25

487

RainfalleventshaveanimpactontheobservedcorrelationsbetweenWWT-P,MWL-PandriverineP488

fluxes. Winter rainfalleventscauseshort term increases inbothWWT-PandriverineP fluxesbut489

throughdifferentprocesses.Highrainfallresultsinincreasedinflowstowastewatertreatmentworks.490

Theremaybesomedilutionof influentPduringtheseeventsdueto increasedcontributionsfrom491

runoff and groundwater infiltration (De Bénédittis and Bertrand-Krajewski 2005), but these are492

unlikely to be captured in the flux estimates due to the paucity of concentrationmeasurements.493

Influentflowsbelowthepeakstormflowtreatmentdesigncriteria(typically3xDWF(Sauletal.2007))494

willbetreatedanddischargedtowatercourses. Excessflowswillbetemporarlydivertedtostorm495

tanks and flows > 6 xDWFwill be directly discharged to rivers. Winter rainfall events can cause496

increasesinriverinePfluxesthroughanumberofprocesses.Increasedagriculturalrunoff,in-stream497

mobilisation of P associated with sediments, combined sewer overflows (CSOs) and wastewater498

treatmentworksdischargescanallcontributetoincreasingPloadsduringwinterstormevents(Bowes499

et al. 2008, Jarvie et al. 2006). These processes controlling short term temporal changes in both500

riverinePfluxesandWWT-PfluxesmaskanycorrelationbetweenMWL-PandriverineP.501

502

InthecontextofMWL-P,thereissomeuncertaintyintheextentofPdosingandtheconcentrations503

used. Just aswater companiespublically releasewater resourceplanningdata,makinghistoric P504

concentrationsanddosingextentsavailablewouldbebeneficialtorefiningMWL-Pestimates.Given505

the potential significance of this flux, assessment of MWL-P using observed leakage and P506

concentrationdataatthelocalscaleusingdistrictmeteredarea(DMA)datacouldprovideimportant507

newinsights.Further,theuncertaintysurroundingtheultimatefateofMWL-Pfollowingleakageis508

significant. As discussed previously, comparisonswithWWT-P and riverine P fluxes such as those509

presentedinFigure5arebeneficialastheyprovideindicationsoftherelativecontributionofdifferent510

sourcestotheenvironment.ItislikelythatMWL-PmakessomedirectcontributiontoriverinePfluxes,511

26

buttherelativeimportanceofgroundwater,surfacewaterandthesewernetworkaspathwaysfor512

MWLP-Parelargelyunknownatthistime.Moreover,thespringandsummergrowingperiodisoften513

perceived to be themost biologically-critical period inmany receivingwaters (Jarvie et al. 2006),514

meaning thatwinterpeaks inMWL-Pwhichdodirectly contribute to riverineP loadsmayhavea515

limitedimmediateimpactonriverineecosystems.However,therearelikelytobetimelagsbetween516

the flux of P fromawatermains leak and this flux reaching a river,with a rangeof transit times517

dependingoncatchmentsize,transportpathwayandchangeinPconcentrationalongthepathway.518

Consequently, it isplausiblethatwinterpeaks in leakageofP fromwatermainsonlyreachariver519

networkinmorebiologically-criticalperiodsoftheyear.Longtransittimesarelikelyincatchments520

withasignificantlythickunsaturatedzoneandlonggroundwaterflowpaths.Inthesecatchmentsit521

isplausiblethatMWL-Pcontributesto legacyPstores ingroundwater (Holmanetal.2008)ashas522

been observed for nitrate (Ascott et al. 2017, Ascott et al. 2016b). River sediments can also523

accumulate P (Sharpley et al. 2013), with time lags of < 1 year associated with in-stream P524

remobilisationduringhighriverflowevents.Thesein-streamaccumulation-remobilisationprocesses525

will result in further lags betweenMWL-P that reaches a river and P riverine exports.Whilst not526

considered in thisstudy,agricultureremainsasignificantPsourceto theenvironment (Whiteand527

Hammond2009)andshouldclearlybeaccountedforinfuturePmanagementstrategies.Muchofthe528

likelytemporaldynamicsofthefateofMWL-P(higherwinterfluxeswithpotentialcatchmentlegacy529

storesandlags)areanalogoustothefateofnonpointPsourcessuchasagriculture(Sharpley2016).530

TheanticipatedimpactofmeasurestoreducePfluxesfrombothmainsleakageandagricultureshould531

betemperedwithknowledgethatcatchmentlagsandreleaseofexistingnonpointlegacyPsources532

maysignificantlydelayanywaterqualityimprovements.533

534

535

27

536

4.2 MWL-P and WWT-P Scenarios: Implications for drinking water and537

wastewatermanagement538

539

Table 3 illustrates that futureWWT-P reductions are likely to be substantially larger thanMWL-P540

reductions,andconsequentlythiswillresult inagreaterrelativecontributionofMWL-PtoP loads541

delivered into the environment in the future. Whilst policy responses to minimise MWL-P were542

previously discussed by Gooddy et al. (2017), the implications of future increases in the relative543

contribution of MWL-P have not been considered to date. Such increases are likely to occur in544

countrieswherePdosingiscurrentlybeingconsideredinfutureleadreductionstrategies(e.g.Ireland545

(IrishWater2015)andSouthKorea(Lee,pers.comm.)).Inthesecountries,thepotentialimpactsof546

MWL-P shouldbe considered inenvironmental impact assessments. Moreover, in lessdeveloped547

countries,currentwastewaterPremovalandmainswaterPdosingmaybelessextensive.Asboth548

environmentalandpublichealthstandardsimprove,itisplausiblethatinthesecountriesmainswater549

P dosing and WWT P removal may increase substantially, resulting in increases in the relative550

importanceofMWL-P.551

552

A number of European countries have adopted policies that actively avoid phosphate dosing on553

environmentalgrounds(CIWEM2011).CountriessuchastheNetherlandsundertakepHcorrection554

andcentralizedwatersofteningtoensureleadconcentrationsarebelowtherequiredstandard(Hayes555

2012).Phasbeenshowntobealimitingnutrienttobiofilmgrowthinanumberofdifferentdrinking556

watersystems(Liuetal.(2016)andreferencestherein)andconcernhasbeenraisedhistoricallythat557

thepresenceofPinwatermains(throughdosingorotherwise)wouldresult inincreasedbacterial558

counts in watermains (Miettinen et al. 1996). The current and future significance ofMWL-P in559

28

comparisontootherPsourcesidentifiedinthisstudyaddstothebodyofevidencetosupportpolicies560

that avoid phosphate dosing,where the extent of leadpiping in the distribution network is small561

enoughtosafeguardhumanhealthprotectedwithoutdosing.562

563

Acrossthesedifferenthydro-socioeconomicsettings,futuremanagementofMWL-PandWWT-Pis564

likely to be a significant challenge. Bothwater utilities and environmental regulators have often565

historicallybeendividedbetweencleanwaterandwastewater(Ofwat2017).Withinthecleanwater566

sector,theindustryhasoftenbeenfurtherdividedbetweendrinkingwaterquality(responsibleforP567

dosing) and water resources management (responsible for leakage) (Deloitte 2014). Moreover,568

regulatory drivers for changes in water management have typically addressed single issues (e.g.569

drinking water quality via the EU Drinking Water Directive (European Commision 1998),570

environmental quality via the EU Water Framework Directive (European Union 2000)), whereas571

addressingMWL-Prequirespolicyinterventionsacrossmultiplefields.WherePdosingandWWTP572

removalare in their infancy,waterandenvironmentalmanagerswillneedtoengagestakeholders573

fromacrossthesedisciplinesatanearlystageofdevelopment. Thiswillensurethatstrategiesfor574

managingPsourcestotheenvironmentconsiderbothWWT-PandMWL-P,andthatreductions in575

WWT-Parenotoffsetby increases inMWL-P. Wherethesepracticesarealreadywellestablished,576

integration of MWL-P into leakage targets and catchment P permits may be an effective policy577

intervention (Gooddy et al. 2017). Adopting these strategies will ensure that P sources to the578

environmentaremanagedeffectively,whilstsafeguardhumanhealth.579

580

4.3 ResearchprioritiesforMWL-P581

582

29

Basedontheuncertaintieshighlightedabove, thereareanumberofkeyresearchprioritieswhich583

needtobeaddressedifMWL-PanditsassociatedseasonalityaretobeeffectivelyintegratedintoP584

managementstrategies.Thefateofwaterfollowingaleakiscurrentlypoorlyunderstood.Itcanbe585

hypothesisedthatthefatewillvarydependingonthetypeofleakasfollows:(1)Visiblewinterbursts586

causingrapidtransporttorivers,groundwaterandsewers,(2)Invisiblewinterleakagewithagreater587

rechargetogroundwater,(3)Summerleakageassociatedwithchangesinsoilmoisturedeficit(SMD)588

causingpossiblelossbyevapotranspirationandtogroundwater,(4)Backgroundleakagewithaslower589

transport and greater recharge to groundwater. Background leakage is likely to have the largest590

impactinthelongterm,becausetheseleaksarerelativelydifficulttolocateandstoprelativetobursts591

(Edie2016,Lambert1994).592

593

ThefateofPwithinthesubsurfacefollowingaleakaddsfurtheruncertainty.ThepHofmainswater594

isoftenincreasedcomparedtorawuntreatedwaters(Flemetal.2015)whichwillinhibitformation595

of iron and aluminium phosphate precipitates, resulting in increased P mobility. However, after596

releaseintotheenvironmentMWL-Phasawiderangeofpotentialfates(claysorption,soil/sediment597

accumulation,flowtorivers,groundwaterorthesewernetwork).Dependingonthetimeofyearand598

thetypeofleakidentifiedabove,thefateofMWL-Pmayvarysubstantially.Visiblewinterburstsand599

invisiblewinterleakagenotcapturedbythesewernetworkmayresultinrapidPtransporttorivers600

and groundwater, potentially accelerated by high rainfall events following coldweather. Summer601

leakageassociatedwithchanges insoilmoistureandbackground leakage is likely toresult inslow602

transport and potentially sorption and accumulation of P in soils. Understanding the relative603

significanceofthesedifferentfatesandlagsinthehydrologicalsystemisessentialifMWL-Pistobe604

managedinthefuture.605

606

30

Long term changes inMWL-P seasonality are also likely to occur due to climate change. Climate607

projections(UKCP092080s,Mediumemissionsscenario)showincreasesindailymean,maximumand608

minimumtemperaturesacrosstheUK,withslightlygreaterincreasesinsummerthanwinter.Whilst609

changestoannualprecipitationtotalarepredictedtobesmall,winterandsummerprecipitationare610

predictedtoincreasebyupto33%anddecreasebyupto40%respectively(MetOffice2010,Murphy611

etal.2009).Warmerwintersmayreducewinterbursts,andwarmersummersmayincreaseleakage612

relatedtosoilmovement.ConsequentlyitisplausiblethattheseasonalcomponentofMWL-Pmay613

becomemorebimodal,withpeaksinbothwinterandsummer.Differencesinthetemporalrainfall614

distributionarealsolikelytoimpactthefateofPwithinthewidercatchment.Increasedwinterrainfall615

mayresultinmorepronouncedwinterPflushing(Johnsonetal.2009).Interactionsbetweenclimate616

change induced temperature and rainfall effects on P fluxes are likely to be complex and require617

furtherinvestigation.618

619

620

621

5 Conclusions622

623

Thisstudyhasquantifiedseasonalityandfuturescenariosof fluxesofP frommainswater leakage624

(MWL-P)totheenvironmentforthefirsttime.MWL-Pshowsastrongseasonality,withpeakfluxes625

duringburstandleakeventsinwinter>150%offluxesduringotherseasons. Duringpeakevents,626

MWL-Pisequivalenttoc.20%ofPfluxesfromwastewatertreatmentworks(WWT-P).Amoderate627

cross correlation between WWT-P and riverine P fluxes is observed as the short term temporal628

variations inbothofthesefluxesaretheresultofwinterrainfallevents. Thismasksanypotential629

31

correlationbetweenMWL-PandriverinePfluxes. Asubstantial increase intheratioofMWL-Pto630

WWT-Pispredictedto2040associatedwithimplementationofsubstantialwastewaterPremovaland631

minimalmainswaterleakagereduction.FurtherresearchisrequiredtounderstandthefateofMWL-632

Pintheenvironment,futurechangesinMWL-Ploadings,andpotentialapproachestointegratethisP633

sourceintowatermanagementstrategies.634

6 Acknowledgements635

636

TheresearchwasfundedbytheUKsNaturalEnvironmentResearchCouncil(NERC)NationalCapability637

resourcesdevolvedtotheBritishGeologicalSurvey.Thispaper ispublishedwithpermissionofthe638

ExecutiveDirector,BritishGeologicalSurvey(NERC).639

640

References641

642

AffinityWater.LatestReports.643https://stakeholder.affinitywater.co.uk/latest-reports.aspx.Hatfield,UK:644AffinityWater,2014.645Ascott,M.J.,Gooddy,D.C.,Lapworth,D.J.andStuart,M.E.(2016a)646Estimatingtheleakagecontributionofphosphatedoseddrinkingwaterto647environmentalphosphoruspollutionatthenational-scale.ScienceofThe648TotalEnvironment572,1534-1542.649Ascott,M.J.,Gooddy,D.C.,Wang,L.,Stuart,M.E.,Lewis,M.A.,Ward,R.S.and650Binley,A.M.(2017)Globalpatternsofnitratestorageinthevadosezone.651NatureCommunications8(1),1416.652Ascott,M.J.,Wang,L.,Stuart,M.E.,Ward,R.S.andHart,A.(2016b)653Quantificationofnitratestorageinthevadose(unsaturated)zone:a654missingcomponentofterrestrialNbudgets.HydrologicalProcesses65530(12),1903-1915.656Bellinger,D.,Leviton,A.,Waternaux,C.,Needleman,H.andRabinowitz,M.657(1987)Longitudinalanalysesofprenatalandpostnatalleadexposureand658

32

earlycognitivedevelopment.NewEnglandjournalofmedicine316(17),6591037-1043.660Birek,L.,Petrovic,D.andBoylan,J.(2014)Waterleakageforecasting:the661applicationofamodifiedfuzzyevolvingalgorithm.AppliedSoftComputing66214,PartB,305-315.663Bowes,M.J.,Gozzard,E.,Johnson,A.C.,Scarlett,P.M.,Roberts,C.,Read,D.S.,664Armstrong,L.K.,Harman,S.A.andWickham,H.D.(2012a)Spatialand665temporalchangesinchlorophyll-aconcentrationsintheRiverThames666basin,UK:Arephosphorusconcentrationsbeginningtolimit667phytoplanktonbiomass?ScienceofTheTotalEnvironment426,45-55.668Bowes,M.J.,Ings,N.L.,McCall,S.J.,Warwick,A.,Barrett,C.,Wickham,H.D.,669Harman,S.A.,Armstrong,L.K.,Scarlett,P.M.,Roberts,C.,Lehmann,K.and670Singer,A.C.(2012b)NutrientandlightlimitationofperiphytonintheRiver671Thames:Implicationsforcatchmentmanagement.ScienceofTheTotal672Environment434,201-212.673Bowes,M.J.,Jarvie,H.P.,Naden,P.S.,Old,G.H.,Scarlett,P.M.,Roberts,C.,674Armstrong,L.K.,Harman,S.A.,Wickham,H.D.andCollins,A.L.(2014)675Identifyingprioritiesfornutrientmitigationusingriverconcentration–flow676relationships:TheThamesbasin,UK.JournalofHydrology517,1-12.677Bowes,M.J.,Smith,J.T.,Jarvie,H.P.andNeal,C.(2008)Modellingof678phosphorusinputstoriversfromdiffuseandpointsources.ScienceofThe679TotalEnvironment395(2),125-138.680Chen,J.andGupta,A.K.(2011)Parametricstatisticalchangepointanalysis:681withapplicationstogenetics,medicine,andfinance,SpringerScience&682BusinessMedia.683CIWEM.LeadinDrinkingWater.http://www.ciwem.org/policy-and-684international/policy-position-statements/lead-in-drinking-water.aspx,6852011.686Cocks,R.andOakes,A.(2011)ModellingtheImpactofWeatheronLeakage687andBursts.Savic,D.A.,Kapelan,Z.andButler,D.(eds),pp.351-357,688UniversityofExeter,UniversityofExeter.689Comber,S.,Cassé,F.,Brown,B.,Martin,J.,Hillis,P.andGardner,M.(2011)690Phosphatetreatmenttoreduceplumbosolvencyofdrinkingwateralso691reducesdischargesofcopperintoenvironmentalsurfacewaters.Water692andEnvironmentJournal25(2),266-270.693Comber,S.D.,Smith,R.,Daldorph,P.,Gardner,M.J.,Constantino,C.andEllor,694B.(2013)Developmentofachemicalsourceapportionmentdecision695supportframeworkforcatchmentmanagement.EnvironmentalScience&696Technology47(17),9824-9832.697CommitteeonClimateChange(2015)CCRA2:Updatedprojectionsofwater698availabilityfortheUK.699

33

DeBénédittis,J.andBertrand-Krajewski,J.-L.(2005)Infiltrationinsewer700systems:comparisonofmeasurementmethods.WaterScienceand701Technology52(3),219-227.702Deloitte(2014)WaterCountryProfiles,DeloitteToucheTohmatsuLtd,703London.704Dodrill,D.M.andEdwards,M.(1995)Corrosioncontrolonthebasisof705UtilityExperience.JournalAmericanWaterWorksAssociation87(7),74-70685.707Dubrovsky,N.M.,Burow,K.R.,Clark,G.M.,Gronberg,J.,Hamilton,P.A.,Hitt,708K.J.,Mueller,D.K.,Munn,M.D.,Nolan,B.T.andPuckett,L.J.(2010)The709qualityofourNation'swaters-NutrientsintheNation'sstreamsand710groundwater,1992-2004,USGeologicalSurvey.711Edie.LeakageExplained.712http://www.edie.net/abb/view_entry.asp?id=4268,2016.713Edwards,M.,Triantafyllidou,S.andBest,D.(2009)ElevatedBloodLeadin714YoungChildrenDuetoLead-ContaminatedDrinkingWater:Washington,715DC,2001−2004.EnvironmentalScience&Technology43(5),1618-1623.716Elser,J.J.,Bracken,M.E.S.,Cleland,E.E.,Gruner,D.S.,Harpole,W.S.,717Hillebrand,H.,Ngai,J.T.,Seabloom,E.W.,Shurin,J.B.andSmith,J.E.(2007)718Globalanalysisofnitrogenandphosphoruslimitationofprimaryproducers719infreshwater,marineandterrestrialecosystems.EcologyLetters10(12),7201135-1142.721EnvironmentAgency(2012)WaterResourcesPlanningGuideline:The722TechnicalMethodsandInstructions,EnvironmentAgency,Bristol,UK.723EnvironmentAgency(2015)WFDWaterBodies:2015statusandreasons724fornotachievinggood(RNAG)fortheupdatetotheriverbasin725managementplans-Cycle2.726EnvironmentAgency(2016)WatercompanyPloadsdischargedinto727environmentinfreshwaterRiverThamescatchment,EnvironmentAgency,728Bristol,UK.729EuropeanCommision(1998)CouncilDirective98/83/ECof3November7301998onthequalityofwaterintendedforhumanconsumption,European731Commision,Brussels.732EuropeanUnion(1991)CouncilDirectiveof21May1991concerning733urbanwastewatertreatment(91/271/EEC),EuropeanUnion,Brussels.734EuropeanUnion(2000)Directive2000/60/ECoftheEuropeanParliament735andoftheCouncilestablishingaframeworkfortheCommunityactionin736thefieldofwaterpolicy,EuropeanUnion,Brussels.737Flem,B.,Reimann,C.,Birke,M.,Banks,D.,Filzmoser,P.andFrengstad,B.738(2015)InorganicchemicalqualityofEuropeantap-water:2.Geographical739distribution.AppliedGeochemistry59,211-224.740

34

Fuller,R.M.,Smith,G.M.,Sanderson,J.,Hill,R.andThomson,A.(2013)The741UKLandCoverMap2000:constructionofaparcel-basedvectormapfrom742satelliteimages.TheCartographicJournal.743Funari,E.andTestai,E.(2008)Humanhealthriskassessmentrelatedto744cyanotoxinsexposure.Criticalreviewsintoxicology38(2),97-125.745GlobalWaterIntel.MarketProfile:ThewatermarketinEnglandandWales746to2020.747https://www.globalwaterintel.com/client_media/uploaded/Chantal/mark748et%20profile%20sample.pdf,2014.749Gooddy,D.C.,Ascott,M.J.,Lapworth,D.J.,Ward,R.S.,Jarvie,H.P.,Bowes,M.J.,750Tipping,E.,Dils,R.andSurridge,B.W.J.(2017)Mainswaterleakage:751Implicationsforphosphorussourceapportionmentandpolicyresponsesin752catchments.ScienceofTheTotalEnvironment579,702-708.753Gooddy,D.C.,Lapworth,D.J.,Ascott,M.J.,Bennett,S.A.,Heaton,T.H.E.and754Surridge,B.W.J.(2015)IsotopicFingerprintforPhosphorusinDrinking755WaterSupplies.EnvironmentalScience&Technology49(15),9020-9028.756Hayes,C.(2010)Bestpracticeguideonthecontrolofleadindrinking757water,IWAPublishing.758Hayes,C.(2012)InternalCorrosionofWaterSupplySystems,IWA759Publishing,London.760Hayes,C.,Incledion,S.andBalch,M.(2008)ExperienceinWales(UK)ofthe761optimisationofortho-phosphatedosingforcontrollingleadindrinking762water.Journalofwaterandhealth6(2),177-185.763Haygarth,P.M.,Jarvie,H.P.,Powers,S.M.,Sharpley,A.N.,Elser,J.J.,Shen,J.,764Peterson,H.M.,Chan,N.-I.,Howden,N.J.andBurt,T.(2014)Sustainable765phosphorusmanagementandtheneedforalong-termperspective:The766legacyhypothesis.EnvironmentalScience&Technology48(15),8417-7678419.768Hilton,J.,O'Hare,M.,Bowes,M.J.andJones,J.I.(2006)Howgreenismy769river?Anewparadigmofeutrophicationinrivers.ScienceofTheTotal770Environment365(1–3),66-83.771Holman,I.,Whelan,M.,Howden,N.,Bellamy,P.,Willby,N.,Rivas-Casado,M.772andMcConvey,P.(2008)Phosphorusingroundwater—anoverlooked773contributortoeutrophication?HydrologicalProcesses22(26),5121-5127.774IrishWater(2015)DraftWaterServicesStrategicPlan:APlanforthe775FutureofWaterServices,IrishWaterDublin.776Jarvie,H.P.,Neal,C.andWithers,P.J.A.(2006)Sewage-effluentphosphorus:777Agreaterrisktorivereutrophicationthanagriculturalphosphorus?778ScienceofTheTotalEnvironment360(1),246-253.779Johnson,A.C.,Acreman,M.C.,Dunbar,M.J.,Feist,S.W.,Giacomello,A.M.,780Gozlan,R.E.,Hinsley,S.A.,Ibbotson,A.T.,Jarvie,H.P.,Jones,J.I.,Longshaw,781M.,Maberly,S.C.,Marsh,T.J.,Neal,C.,Newman,J.R.,Nunn,M.A.,Pickup,782

35

R.W.,Reynard,N.S.,Sullivan,C.A.,Sumpter,J.P.andWilliams,R.J.(2009)783TheBritishriverofthefuture:Howclimatechangeandhumanactivity784mightaffecttwocontrastingriverecosystemsinEngland.ScienceofThe785TotalEnvironment407(17),4787-4798.786Killick,R.andEckley,I.A.(2014)changepoint:AnRPackagefor787ChangepointAnalysis.201458(3),19.788Kinniburgh,J.H.andBarnett,M.(2010)Orthophosphateconcentrationsin789theRiverThames:reductionsinthepastdecade.WaterandEnvironment790Journal24(2),107-115.791Lambert,A.(1994)AccountingforLosses:TheBurstsandBackground792Concept.WaterandEnvironmentJournal8(2),205-214.793Li,H.,Liu,J.,Li,G.,Shen,J.,Bergström,L.andZhang,F.(2015)Past,present,794andfutureuseofphosphorusinChineseagricultureanditsinfluenceon795phosphoruslosses.AMBIO44(2),274-285.796Liu,S.,Gunawan,C.,Barraud,N.,Rice,S.A.,Harry,E.J.andAmal,R.(2016)797Understanding,Monitoring,andControllingBiofilmGrowthinDrinking798WaterDistributionSystems.EnvironmentalScience&Technology50(17),7998954-8976.800Marsh,T.andHannaford,J.(2008)UKhydrometricregister.Acatalogueof801riverflowgaugingstationsandobservationwellsandboreholesinthe802UnitedKingdomtogetherwithsummaryhydrometricandspatialstatistics,803CentreforEcology&Hydrology.804Merrett,S.(2007)TheThamescatchment:ariverbasinatthetippingpoint.805WaterPolicy9(4),393-404.806MetOffice(2010)UKClimateProjections:Briefingreport,MetOffice,,807Exeter,UK.808MetOffice(2016)ClimateSummaries.809Miettinen,I.T.,Vartiainen,T.andMartikainen,P.J.(1996)Contaminationof810drinkingwater.Nature381(6584),654-655.811Mockler,E.M.,Deakin,J.,Archbold,M.,Gill,L.,Daly,D.andBruen,M.(2017)812SourcesofnitrogenandphosphorusemissionstoIrishriversandcoastal813waters:Estimatesfromanutrientloadapportionmentframework.Science814ofTheTotalEnvironment601-602,326-339.815Morée,A.L.,Beusen,A.H.W.,Bouwman,A.F.andWillems,W.J.(2013)816Exploringglobalnitrogenandphosphorusflowsinurbanwastesduringthe817twentiethcentury.Globalbiogeochemicalcycles27(3),836-846.818Murphy,J.M.,Sexton,D.M.H.,Jenkins,G.J.,Booth,B.B.B.,Brown,C.C.,Clark,819R.T.,Collins,M.,Harris,G.R.,Kendon,E.J.,Betts,R.A.,Brown,S.J.,Humphrey,820K.A.,McCarthy,M.P.,McDonald,R.E.,Stephens,A.,Wallace,C.,Warren,R.,821Wilby,R.andWood,R.A.(2009)UKClimateProjectionsScienceReport:822ClimateChangeProjections,p.192,MeteorologicalOfficeHadleyCentre,823Exeter,UK.824

36

Ofwat.Watersectoroverview.http://www.ofwat.gov.uk/regulated-825companies/ofwat-industry-overview/,2017.826Pocock,S.J.,Shaper,A.G.,Ashby,D.,Delves,H.T.andClayton,B.E.(1988)827TheRelationshipbetweenBloodLead,BloodPressure,Stroke,andHeart828AttacksinMiddle-AgedBritishMen.EnvironmentalHealthPerspectives78,82923-30.830Pohlert,T.(2018)Package‘trend’.831Powers,S.M.,Bruulsema,T.W.,Burt,T.P.,Chan,N.I.,Elser,J.J.,Haygarth,832P.M.,Howden,N.J.K.,Jarvie,H.P.,Lyu,Y.,Peterson,H.M.,Sharpley,A.N.,833Shen,J.,Worrall,F.andZhang,F.(2016)Long-termaccumulationand834transportofanthropogenicphosphorusinthreeriverbasins.NatureGeosci8359(5),353-356.836RDevelopmentCoreTeam(2016)R:Alanguageandenvironmentfor837statisticalcomputing,RFoundationforStatisticalComputing,Vienna,838Austria.839Saul,A.,Shepherd,W.andBoxall,J.(2007)WorldEnvironmentalandWater840ResourcesCongress2007,AmericanSocietyofCivilEngineers,Tampa,841Florida,UnitedStates.842SewageWorld.WaterandWastewaterTreatmentPlantDirectory.843http://www.sewage.net/,2016.844Sharpley,A.(2016)Managingagriculturalphosphorustominimizewater845qualityimpacts.ScientiaAgricola73,1-8.846Sharpley,A.,Jarvie,H.P.,Buda,A.,May,L.,Spears,B.andKleinman,P.847(2013)PhosphorusLegacy:OvercomingtheEffectsofPastManagement848PracticestoMitigateFutureWaterQualityImpairment.Journalof849EnvironmentalQuality42(5),1308-1326.850SoutheastWater.WRMPLibrary.http://www.southeastwater.co.uk/about-851us/our-plans/water-resources-management-plan/wrmp-library.Snodland,852UK:SoutheastWater,2014.853SuttonandEastSurreyWater.FinalWaterResourcesManagementPlan.854https://www.waterplc.com/userfiles/file/WRMP_Final_2014_Tables.pdf.855Redhill,UK:SuttonandEastSurreyWater,2014.856Tchobanoglous,G.andBurton,F.L.(2013)WastewaterEngineering:857TreatmentandResourceRecovery,McGraw-Hill,NewYork,USA.858ThamesWater.WaterResourcesManagementPlan2015-2040.859http://www.thameswater.co.uk/about-us/5392.htm.Reading,UK:Thames860Water,2014.861UKWaterIndustryResearchLtd(2007)ManagingSeasonalVariationsin862Leakage,UKWaterIndustryResearchLtd,London.863UKWaterIndustryResearchLtd(2012)AlternativestoPhosphatefor864PlumbosolvencyControlUKWaterIndustryResearchLtd,London.865

37

UKWaterIndustryResearchLtd(2013)EffectofWeatheronLeakageand866Bursts,UKWaterIndustryResearchLtd,,London.867Vollenweider,R.A.(1968)Scientificfundamentalsoftheeutrophicationof868lakesandflowingwaters,withparticularreferencetonitrogenand869phosphorousasfactorsineutrophication.870WaterTechnology.LatestIndustryProjectsInEurope.http://www.water-871technology.net/projects/region/europe/,2016.872White,P.J.andHammond,J.P.(2009)Thesourcesofphosphorusinthe873watersofGreatBritain.JournalofEnvironmentalQuality38(1),13-26.874WorldBank(2006)TheChallengeofReducingNon-RevenueWater(NRW)875inDevelopingCountries,WorldBank,WashingtonDC.876Worrall,F.,Howden,N.J.K.,Burt,T.P.andJarvie,H.P.(2015)Thefluvialflux877ofphosphorusfromtheUK1974–2012:wherehasallthephosphorus878gone?,Vienna,Austria.879

880