What difference might sewage treatment performance make to endocrine disruption in rivers?

Environmental Pollution 147 (2007) 194e202www.elsevier.com/locate/envpol

What difference might sewage treatment performance make to endocrinedisruption in rivers?

Andrew C. Johnson a,*, Richard J. Williams a, Pete Simpson b, Rakesh Kanda c,1

a Centre for Ecology and Hydrology, Maclean Building, Benson Lane, Crowmarsh Gifford, Wallingford, Oxfordshire, OX10 8BB, UKb Environment Agency, Waterlooville Laboratory, 4 The Meadows, Waterlooville, Hampshire, PO7 7XX, UK

c WRc-NSF Ltd., Medmenham, Marlow, Buckinghamshire, SL2 2HD, UK

Received 24 May 2006; received in revised form 1 August 2006; accepted 2 August 2006

Biological (trickling) filter sewage plants are less successful than other secondary sewage treatment plants at removing estronefrom sewage effluent.

Abstract

An assessment of the steroid estrogen removing performance of 23 different sewage treatment plants (STPs) was performed. The assessmentrelied on a model to estimate influent concentrations, and completed questionnaires on the STP treatment details from the relevant water com-panies. This information was compared with observed effluent 17b-estradiol (E2) and estrone (E1) concentrations. The 10 biological filter plants(BFP) in the study performed poorly with only 30% (SD 31) removal on average for E1. This reduced E1 removal performance of the BFPscompared to all the other STP types in the survey was statistically significant ( p < 0.001). Scenarios of all the STPs as activated sludge types,and one as all BFP types were modelled using the GREAT-ER model set up for the Aire/Calder catchment in the UK. This difference was shownto have an important effect on predicted river E1 concentrations and consequent risk classifications.� 2006 Elsevier Ltd. All rights reserved.

Keywords: Estrogens; Estrone; Activated sludge; Biological filter; Removal

1. Introduction

It is now well established that sexual disruption in UK wildfish is associated with sewage effluents, and the biologicallyactive, endocrine disrupting, substances contained withinthem (Desbrow et al., 1998; Jobling et al., 1998; Jobling andTyler, 2003). The most potent endocrine active substancescontained within these effluents are believed to be naturaland artificial steroid estrogens (of anthropogenic origin),

* Corresponding author: Tel.: þ44 (0)1491 838800; fax: þ44 (0)1491

692430.

E-mail addresses: [email protected] (A.C. Johnson), [email protected] (R.J.

Williams), [email protected] (P. Simpson).1 Present address: STL Reading, Britten Road, Reading RG2 0AU, UK.

E-mail address: [email protected] (R. Kanda).

0269-7491/$ - see front matter � 2006 Elsevier Ltd. All rights reserved.

doi:10.1016/j.envpol.2006.08.032

although other non-steroidal substances have also been impli-cated, e.g., alkylphenol and alkylphenol polyethoxylates(Blackburn and Waldock, 1995; Sole et al., 2000).

The majority of sewage in the UK is treated either by bio-logical filter, or by activated sludge. The activated sludge plant(ASP) is most commonly associated with large towns and cit-ies. Activated sludge is an intensive biological treatment inwhich bacteria are suspended in a tank and vigorously aerated,with a hydraulic retention time (HRT) of 5e20 h plus (Cooperand Downing, 1998; Johnson et al., 2000). Biological, alsoknown as percolating or trickling, filter plants (BFP) compriseof a tank with a biofilm supported on coarse media upon whichthe sewage liquor is sprayed following primary sedimentation(Boon et al., 1997). The water contact time with the biofilm isoften quite short, around 30 min. To improve the effluent qual-ity an increasing number of BFP have some form of tertiary

mailto:[email protected]



http://www.elsevier.com/locate/envpol

195A.C. Johnson et al. / Environmental Pollution 147 (2007) 194e202

treatment. These can include some sort of additional filtrationprocess, or an additional biological treatment step.

In recent years mathematical models have allowed us topredict where in catchments the greatest concentrations ofchemicals used by the human population (such as pharmaceu-ticals) are likely to occur (Anderson et al., 2004; Schowanekand Webb, 2002). However, these efforts have been hamperedby a lack of knowledge about how successful different types ofSTPs are at removing these compounds. In the absence of suchinformation, modelling has had to assume that all STPs per-form to a similar efficiency (Johnson and Williams, 2004).In the last 3 years there has been an upsurge in informationon the performance of different sewage treatment technologiesto remove steroid estrogens. Therefore, it is worthwhile re-ex-amining this area, which as others have argued (Sumpter andJohnson, 2005) could be so critical in determining lowlandwater quality. To help guide the assessment a number of test-able hypotheses were proposed:

� ASPs will remove more steroid estrogens than any otherform of sewage treatment.� BFPs will remove less steroid estrogens than any other

form of sewage treatment.� Additional tertiary biological treatment will improve ste-

roid estrogen removal for BFPs.

2. Evidence for differential estrogen removalperformance by type in the world literature

An assessment of the most recent research which include46 separate ASPs (Andersen et al., 2003; Clara et al., 2004;Johnson et al., 2005; Joss et al., 2004; Komori et al., 2004;Kreuzinger et al., 2004; Onda et al., 2003; Williams et al.,2003; Zuehlke et al., 2005) suggest removal rates for theASPs are slightly higher for E2 (91%), and E1 (78%), whilstslightly lower for ethinylestradiol (EE2, 76%) than that re-ported in a previous review (Johnson and Williams, 2004).As noted previously, E1 is a difficult steroid estrogen to pre-dict in effluents with literature data showing removal ratesranging from 45 to 99% (Johnson and Sumpter, 2001). At-tempts continue to be made to identify which feature of theASP process, or its management is associated with best steroidestrogen removal. The processes that have been considered in-clude: longer hydraulic retention time (Johnson et al., 2005),longer sludge retention time (Andersen et al., 2003; Kreu-zinger et al., 2004; Vader et al., 2000), and reduced organic(BOD) loading (Joss et al., 2004; Onda et al., 2003).

There is laboratory and field evidence that some of theslower growing microbial community, associated with nitrifi-cation and long sludge ages, are important in degrading steroidestrogens (Andersen et al., 2003; Kreuzinger et al., 2004; Shiet al., 2004; Vader et al., 2000). However, the data from mem-brane bioreactors, which have long sludge ages, whilst indicat-ing good removal of the natural steroid estrogens, appearmuch less effective with respect to EE2 removal (Claraet al., 2004; Kreuzinger et al., 2004; Joss et al., 2004). Whilst

both HRT and sludge retention time (SRT) appear to influencesteroid estrogen removal they are not infallible guides to pre-dicting ASP performance (Johnson et al., 2005).

STP which have primary treatment, without secondarytreatment, i.e. sedimentation, or sedimentation with chemicalprecipitation (to remove phosphate), had little or no removalof steroid estrogens/endocrine disrupting chemicals (Desbrowet al., 1998; Johnson et al., 2005; Svensen et al., 2003). How-ever, this group is unlikely to be important for much longer inEurope due to the requirements of the urban wastewater treat-ment directive (Council Directive, 1991).

BFPs with their short, around 30 min HRT, might be ex-pected to be less effective than ASPs at removing steroid es-trogens. There is some evidence that this is the case withcertain pharmaceuticals (Stumpf et al., 1999; Svensen et al.,2003; Ternes et al., 1999a). Unfortunately, there is still littledata reported (two STP only) regarding their performancewith respect to specific estrogen molecules. The two reportedstudies that included BFPs (Joss et al., 2004; Ternes et al.,1999a) reported E2 removals around 90% (similar to ASP),and E1 removals of 67 and 90% respectively (again withinthe expected ASP range), but a slightly lower EE2 removal(64e69%). Contrary to expectations, from this small amountof data it would be hard to argue that overall BFPs will be sig-nificantly worse performers than the average ASP at removingsteroid estrogens. To summarize, the current evidencesuggests:

� ASPs with long HRT and SRT have very good (90% plus)removal of steroid estrogens.� Physical precipitation (such as chemical P removal) and

sedimentation processes does not remove steroid estrogens� Good performance (98% removal) with membrane biore-

actors (MBRs) for natural steroid but less so with EE2.� Insufficient data to conclude BFPs would be worse than

any other biological sewage treatment in removingestrogens.

3. Materials and methods

3.1. Analytical methodology

In 2003 the UK Environment Agency conducted a survey of steroid estro-

gens in the effluents of a wide range of STPs in the UK (Environment Agency,

2006). The sampling and analysis of the effluent samples was done under con-

tract for the Environment Agency by the AstraZeneca Brixham Laboratory,

and the WRc laboratory, both based in the UK. The effluent at each site

was sampled on two separate occasions in AprileMay 2003 around midday.

Each 5 L effluent sample was filtered (0.2 mm) and the solids discarded. The

aqueous samples were spiked with a solution of deuterated (d4) E1, E2 and

EE2 to give an initial concentration of 5 ng/L for each steroid analogue.

The samples were then extracted using a C18 solid phase extraction (SPE) col-

umn. The extract from the column was derivatized with a mixture of pyridine,

N-(tert-butyldimethylsilyl)-N-methyltrifluoroacetamide and bis(trimethylsilyl)

trifluoroacetamide. The samples were cleaned up using a silica SPE column.

The samples and standards were analyzed by gas chromatographyemass spec-

trometery using selected ion monitoring (Environment Agency, 2006; Williams

et al., 2003). This method had previously been tested with six paired river water

samples which were spiked at 2 ng/L of E1, 1 ng/L E2 and 1 ng/L EE2.

196 A.C. Johnson et al. / Environmental Pollution 147 (2007) 194e202

Analysis of the 12 samples showed 7e16% relative standard deviation, detec-

tion limits of 0.4 ng/L for E1, and E2 with 0.5 ng/L for EE2, with 89e97%

recovery. Quantification was performed by the relative internal standards

method. The limit of detection was calculated using 4.65� the within batch

standard deviation of spiked samples, calculated as per NS30, a manual on an-

alytical quality control for the water industry (Cheeseman and Wilson, 1989).

In samples where the deuterated standard could not be clearly detected at the

appropriate concentration, results were recorded as non-quantifiable peak

(NQP). Only estrogens in the aqueous phase of the sewage effluent were an-

alyzed. Measuring estrogens on the effluent solid phase was not considered

critical since to date few estrogens have been detected on solids in effluent

(Andersen et al., 2003; Williams et al., 2003). This is not surprising given

the modest hydrophobicities of the molecules, and the low solid content of

sewage effluent.

3.2. Examining estrogen removal performance based onpredicted influent concentrations

The water companies involved were asked to provide detailed information

on the treatment type and practices of these works (Table 1).

The assessment of estrogen removal performance in this study was carried

out as follows:

� In the EA survey, as described above, two samples were taken from each

STP (Environment Agency, 2006). Each individual measurement was re-

corded. For each sample the flow through the STP was also recorded.

� The Johnson and Williams prediction model (Johnson and Williams,

2004) was used to calculate the predicted influent concentration of

each steroid estrogen on the basis of the recorded flow values and the hu-

man population served by the STP (Table 1).

� The measured effluent values were then considered as residues of the pre-

dicted influent value.

� These effluent values were compared to what might be expected in the

effluent in each case by using the current global removal rate, for example

65% in the case of E1 (Johnson and Williams, 2004).

It should be noted that the prediction of influent concentrations by Johnson

and Williams (2004) assumes that steroids excreted as glucuronide conjugates

in the urine become deconjugated before arriving at the STP, whilst the sulfate

conjugates do not. Recent analytical data appear to support this dichotomy

(Reddy et al., 2005). Where an LOD of for example <0.4 ng/L E2 was re-

ported in the original survey (Table 1) this may reflect in reality 0 E2, or

0.39 ng/L E2 in the effluent. For this assessment it was assumed that the

true E2 concentration was mid-way between the two, which in this example

would have been 0.2 ng/L E2.

3.3. Examining potential impact at the catchment scale ofdiffering estrogen removal efficiencies of STPs

As a form of sensitivity analysis, the potential significance of differences in

sewage treatment E1 removal performance was examined using the catchment

water quality GREAT-ER model (Schowanek et al., 2001). This combined GIS

hydrology and water quality model has been set up in the UK for the River

Aire/Calder catchment in N. Yorkshire (Schowanek and Webb, 2002). The in-

dustrialized Aire and Calder catchment in Yorkshire has within it a population

of approximately 1 900 000 people in an area of 1932 km2 (National Rivers

Authority, 1993) resulting in an estimated population density of 983 inhab/

km2. The annual mean flow from the downstream limit of the catchment at

Beal is 35.8 m3/s (Marsh and Lees, 2003), thus for each individual in the

catchment, their 24 h excretion is diluted by 1628 L. The input of E1 to every

STP in this catchment was predicted using the protocol of Johnson and Wil-

liams (2004). In the first instance every STP in the catchment was modelled

as a high performing activated sludge plant with 95% E1 removal. Thus, in

each case the model allows only 5% of its predicted influent load into the

catchment. The concentration in each reach is then predicted on the basis of

dilution and a first order decay rate. The potential effect on fish of these E1

concentrations was accomplished by converting the concentrations into low,

medium and high effect-risk classifications for vitellogenin induction (Table 2).

These risk classifications have been described and used in a previous modelling

study (Sumpter et al., 2006) and derive from fat head minnow exposure exper-

iments with E1. In the second case, the situation was modelled where every

STP in the catchment was converted into a simple BFP with 30% E1 removal.

Thus, in each case the model allows 70% of its predicted influent E1 load into

the catchment. The impact of these elevated concentrations on the risk of

vitellogenin induction was then also assessed.

It must be admitted that some caution is required in the interpretation of

the individual steroid estrogen concentrations reported in the survey and the

assumed removal rates. The removal rates are based on a prediction of the in-

fluent concentration since no actual measurements were made. The observed

effluent concentrations derive from two spot samples (not composite) taken

on separate occasions around midday. Another problem is the predicted efflu-

ent concentrations were close, or in the case of EE2, often even above the de-

tection limit (Environment Agency, 2006). Frequently for EE2 no quantifiable

measurements were available (Table 1), so this steroid estrogen could not be

used as a measure of treatment plant performance. Thus, this assessment con-

centrated on E2 and E1 due to the large number of positive determinations.

Notwithstanding these limitations, the dataset is a valuable resource since it

represents a diverse range of STP types in which detailed process information

was available. This made it worthy of examination, if only to observe whether

any gross differences in apparent performance could be ascertained.

4. Results

4.1. Quality control and confidence in the data

Before discussing the apparent implications of the data, it isworth reviewing the technique and examining how the valuescompare to other similar studies. Every sample was spikedwith deuterated analogues of the analytes. This highlightsany sample storage, or recovery problems, or interference,and where this occurred the results were discounted as NQP(Table 1). Each site was sampled on two separate occasions,and in the majority of cases these two samples gave broadlysimilar results (Table 1). The reported mean effluent concen-trations (1.8 ng/L E2, 19 ng/L E1) are typical of those foundin the literature (Baronti et al., 2000; Desbrow et al., 1998;Johnson et al., 2000, 2005; Williams et al., 2003). Only onsix occasions out of 48 was the amount of E1 reported inthe effluent greater than that predicted in the model to be pres-ent in the influent. The reported E2 concentrations were typi-cally less than half those of E1, which is consistent with previousobservations (Baronti et al., 2000; Desbrow et al., 1998; Johnsonet al., 2000; Williams et al., 2003), the relative persistence of thetwo compounds (Ternes et al., 1999b), and the differences inhuman excretion (Johnson and Williams, 2004). Each of thetwo laboratories involved in the initial survey sampled arandom mix of sewage treatment types, rather than one labo-ratory doing one type whilst the other focused on differenttypes of STP.

4.2. Activated sludge plant performance

Unfortunately, there were only three ASP (ASP1e3) withsufficiently unambiguous treatment information that could beexamined closely. A fourth ASP is reported, although, its treat-ment practices have not been obtainable. The NQP sampleswere not plotted (Fig. 1). Each of the two samples taken

Ta

O

W

g/L)

Predicted vs

observed E1 (ng/L)

Predicted vs

observed EE2 (ng/L)

A .8 22/4, 23/11 0.6/1, 0.7/1

A 0.4 10/NQP, 16/2 0.3/NQP, 0.5/<0.5

A 0.4 9/<0.4, 9/<0.4 0.3/<0.5, 0.3/NQP

A 12/35, 10/28 0.35/NQP, 0.3/NQP

BF 15/50, 14/19 0.4/0.5, 0.3/NQP

BF 11.2/74, 7.5/100 0.3/1.3, 0.2/1.7

BF QP 2/5, 2/NQP 0.1/<0.5, 0.1/NQP

BF 11/17, 7/38 0.5/0.7, 1.4/<0.5

BF .6 13/34, 13/15 0.4/NQP, 0.4/<0.5

BF 0.4 15/25, 7/39 0.5/NQP, 0.2/NQP

BF 0.4 14/2, 12/NQP 0.4/NQP, 0.3/NQP

BF 0.4 7/18, 8/12 0.2/NQP, 0.2/NQP

BF .4 7.8/6.6, 18/41 0.2/NQP, 0.5/<0.5

BF 0.4 25/30, 27/77 0.7/NQP, 0.8/NQP

EB 18/9, 47/21 0.5/<0.5, 1.4/<0.5

EB 24/14, 18/11 0.7/0.8, 0.5/<0.5

EB 18/15, 19/14 0.5/<0.5, 0.5/NQP

EB 18/8, 12/10 0.5/1, 0.4/NQP

EB 0.4 12/NQP, 19/0.7 0.4/NQP, 0.6/<0.5

EB QP 14/0.9, 14/<0.4 0.4/NQP, 0.4/NQP

EB 0.4 15/10, 15/2.2 0.4/1.5, 0.4/<0.5

EB P 11/56, 9/71 0.3/NQP, 0.3/NQP

EB 0.4 10/4.3, 7/1.2 0.3/NQP, 0.2/NQP

W total treatment process (total HRT). SAF, submerged

ae

19

7A

.C.

Johnsonet

al./

Environm

entalP

ollution147

(2007)194e

202

ble 1

verview of STP types in the EA 2003 survey with effluent predictions followed by observations of steroid estrogens

orks Type Human PE STP flow in

survey (L/s)

Typical effluent

BOD (mg/L)

Typical effluent

NH4 (mg/L)

Bio HRT and total HRT (h) Predicted vs

observed E2 (n

SP1 ASP diffused

air, plug

flow, 12 d SRT

44500 106, 101 2 0.98 12 bio and 16 total 2.4/<0.4, 2.5/0

SP2 ASP brush

air, well

mixed, 15 d SRT

65000 348, 221 NA NA NA 1.1/NQP, 1.7/<

SP3 ASP well

mixed, 50 d SRT

108258 697, 682 3 <0.1 8 bio and 18 total 0.9/<0.4, 0.9/<

SP4 ASP 458791 2048, 2402 3.6 0.5 NA 1.3/2.6, 1.1/4

P1 BFP 5967 20.4, 23.2 6 0.85 4 bio and 9 total 1.7/7, 1.5/2

P2 BFP 83297 390, 488 11.7 1.8 NA 1.2/18, 0.8/22

P3 BFP 17769 470, 470 3 0.5 NA 0.2/<0.4, 0.2/N

P5 BFP 80624 380, 580 15 3 NA 1.2/6, 0.8/<0.4

P6 BFP 46085 236, 236 10 5 0.5 bio and 15 total 1.4/NQP, 1.4/0

P7 BFP 25082 85, 188 9 1.5 NA 1.7/<0.4, 0.8/<

P8 BFP 52375 188, 237 10 1.4 NA 1.6/0.25, 1.3/<

P9 BFP 17139 136, 111 13 3 NA 0.7/<0.4, 0.8/<

P10 BFP 5953 40, 17 NA 0.8/1.2, 1.9/<0

P11 BFP 12817 27, 25 6 1.2 NA 2.7/<0.4, 2.9/<

FP1 2-stage biological filter 58685 185, 70 12 2.5 0.5 bio and 20 total 1.9/NQP, 5/3

FP2 Biological filter

and final sand filter

10968 24, 31 6 0.5 0.5 bio and 19 total 2.6/2.9, 2/<0.4

FP3 Biological filter,

part 2-stage

31352 90, 86 5 3 NA 2/1.6, 2.1/NQP


and final Biopur SAF

18768 54, 77 4 0.6 0.5 bio and 4.25 total 2/1.3, 1.4/1.5


and final microstrainer

15674 66, 42 8 1.6 NA 1.4/NQP, 2.1/<


and final BAFF

10155 32, 31 2 0.5 4 bio and 12 total 1.5/NQP, 1.5/N

FP7 Biological filter (part

high rate biotower filter)

with final BAFF

137500 45, 63 2 0.5 NA 1.6/<0.4, 1.6/<


and final sand filter

316126 1537, 1773 5.4 2.5 NA 1.2/NQP, 1/NQ

FP9 Biological filter,

and final AS then PAC

6999 36, 55 NA NA NA 1.1/NQP, 0.7/<

here known, the hydraulic retention time associated with the secondary biological treatment step is given (bio HRT), and/or the HRT associated with the

rated filter; BAFF, biological aerated fixed filter; PAC, powdered activated charcoal.


from a STP (recorded here as a or b) were considered sepa-rately (Fig. 1). ASPs 1e3 removed more E2 (97% SD 2.2)and E1 (93% SD 7) than was predicted by the original model(Johnson and Williams, 2004). The information provided bythe relevant water companies reveal that ASP1 had a 12 h bi-ological HRT and a 12 d SRT, ASP2 had a 15 d SRT and ASP3had an 18 h total HRT and a 50 d SRT; thus all of these threeSTPs would fit within the group expected to perform well(Table 1). The effluent 5 day biochemical oxygen demand(BOD5) of this group was 2.5 mg/L (n ¼ 2) and the ammo-nium 0.5 mg/L (n ¼ 2). The activated sludge plant ASP4apparently removed little E2 and failed to remove E1. Thus,the performance of this small ASP group is disproportionatelyskewed with mean E2 and E1 removals of 65% (SD 50 n ¼ 6),and 62% (SD 48, n ¼ 6) respectively. Although we have noother process management information, the higher BOD5

(12 mg/L) and ammonia (1.8 mg/L) reported in the effluentof ASP4 suggests a less stringent biological performancehere than with the other ASPs (Table 1).

4.3. Simple biological filter plant performance

There were 10 STP in this class for which detailed informa-tion was available (Table 1, Fig. 2). There was a low mean E1removal of 30% (SD 31, n ¼ 18). Notwithstanding the poor E1removal performance, E2 removal was a reasonably successful

Table 2

Division of the doseeresponse curves into classes

Effect class Descriptor Vitellogenin induction

(% of maximum)

E1 concentration

(mg/L)

1 No effect 0eone-tenth of EC10 <0.0014

2 Low effect One-tenth of EC10e10 0.0014e0.014

3 Medium effect 10e50 0.014e0.058

4 High effect 50e95 0.058e5.33

5 Severe effect 95e100 >5.33

The doseeresponse curve for each chemical was divided into five sections (or

classes), ranging from no effect to severe effect. The same class boundaries

were used for all individual chemicals and mixtures

0

20

40

60

80

100

120

Predict

ed

ASP1a

ASP1b

ASP2b

ASP3a

ASP4a

ASP4b

Sewage treatment plants

Resid

ue as %

p

red

icted

in

in

flu

en

t

E2 % actual residueE1 % actual residue

Fig. 1. Predicted versus apparent removal of E2 and E1 for two samples taken

from the sewage effluent of four activated sludge plants.

70% (SD 36, n ¼ 17), albeit erratic as can be seen by the highstandard deviation. A number of BFPs removed little or no E1,but performed well with E2 removal. The raw data show highE1 effluent concentrations for these plants but only in the caseof BFP2 were these concentrations considerably greater thanthe predicted influent concentration (Table 1). Overall themean effluent BOD5 of this group was 9 mg/L (SD 4, n ¼ 9)and the ammonium 2 mg/L (SD 1.4, n ¼ 9).

4.4. Biological filter plants with additional tertiarytreatment

Nine of the STPs fell into this class, which have not beenexamined in the scientific literature so far in any detail. Thesemay be termed enhanced biological filter plants (EBFP). It isworth examining what tertiary treatment processes were in-volved (Table 1). Unlike the straight biological filter works,this group, with one exception, appear to be able to removeE1 consistently at, or better than the standard model prediction(Fig. 3). The E2 removal performance was similar, or betterthan the prediction used by Johnson and Williams (2004).The mean E2 and E1 removals were 89% (SD 6, n ¼ 10)

0

20

40

60

80

100

120

Predict

ed

BFP1a b

BFP2a b

BFP3a

BFP5a b

BFP6a b

BFP7a b

BFP8a

BFP9a b

BFP10a b

BFP11a b


Resid

ue as %

p

red

icted

in

in

flu

en

t



from the sewage effluent for 10 simple biological filter plants.

0

20

40

60

80

100

120

Predict

ed

EBFP1a b

EBFP2a b

EBFP3a b

EBFP4a b

EBFP5b

EBFP6a b

EBFP7a b

EBFP8a b

EBFP9a b


Resid

ue as %

p

red

icted

in

in

flu

en

t



from the sewage effluent for simple biological filter plants with additional ter-

tiary treatment. Note E2 for EBFP1a, EBFP3b, EBFP5a, EBFP6a and b,

EBFP8a and b, and EBFP9a were NQPs.


and 74% (SD 29, n ¼ 17) respectively. Certainly all of theSTPs whose tertiary treatment was a specialized biologicalprocess (EBFPs 1, 3, 4, 5, 7, 9) performed well (Table 1).As a subset they show an E1 removal of 84% (SD 10,n ¼ 12). It might have been assumed that the three STP(EBFPs 2, 6, and 8) whose tertiary treatment was physical,such as sand filtration, would have performed less well thanthe group with extended biological processes. Two of thosewith sand filters performed well but one failed to removeE1. The mean effluent BOD5 of this group was 5.6 mg/L(SD 3.5, n ¼ 7) and the ammonium 1.3 mg/L (SD 1, n ¼ 7)which was lower than that for the straight biological filterworks.

4.5. The potential impact of estrogen removal differencesbetween STPs on endocrine disruption at the catchmentscale

There was a pronounced difference in predicted risk of vi-tellogenin induction in the catchment by changing the type ofSTP (Figs. 4 and 5). Thus, if all the STP in the catchment werehigh performing ASPs then no effect on VTG induction is at-tributed from predicted E1 concentrations (Fig. 4). However, ifall the STPs present in the catchment were of the BFP typethen low to medium impact on fish from modelled E1 dis-charge is predicted (Fig. 5). Careful examination of Fig. 5shows that in some locations, high in the catchment, consider-able dilution does mediate the potentially poor performance ofBFPs. However, at some mid and lower sections of the

catchment, their presence might considerably increase the en-docrine disruption risk.

5. Discussion

It is worth clarifying that in this survey, removal efficiencyhas been equated only with elimination of steroid estrogensfrom the water phase. The fate of estrogens associated withsewage sludge and removed from the plant is a separate issue.The largest treatment group examined were the BFPs (10STPs,). Their E1 removal performance was highly erratic asshown by the large standard deviation, but the mean removal% indicates they were the worst performing STP group(Table 3). The biological filter works with additional tertiarytreatment (EBFPs) with one exception performed consistentlywell (9 STPs, 18 samples). The three activated sludge plantswhich had long HRT and SRT conditions were all very success-ful at removing E1 and E2, however, the apparent failure ofASP4 reminds us that this treatment does not guaranteesuccessful removal in every case. A crude relationship betweenE1 removal performance and BOD and ammonium content ofthe effluent can be seen (Table 3). Returning to the originalhypotheses:

� ASP will remove more steroid estrogens then any otherform of sewage treatment.

There is insufficient information to reliably support, or fal-sify this hypothesis. Clearly it would be wrong to assume that

Fig. 4. Predicted VTG induction risk categorization throughout the Aire/Calder catchment assuming all STP were high performance activated sludge plants (95%

E1 removal).


Fig. 5. Predicted VTG induction risk categorization throughout the Aire/Calder catchment assuming all STP were biological filter plants (30% E1 removal).

all types of ASP will be the best removers of estrogens. SomeASPs, such as ASP4, for which we do not have treatmentinformation appeared to be poor removers of steroid estrogens.

� BFPs will remove less steroid estrogens than any otherform of sewage treatment.

Some BFP seem to remove as much E1, if not more, thansome ASP. Their removal of E2 is similar to the other STP.However, the mean E1 removal percentage for ASP is greaterthan that for BFP which does suggest that it is more likely thata BFP will be a poor estrogen remover, although with thesmall data set for ASP, the means were not significantly differ-ent ( p ¼ 0.106) using the Wilcoxon rank-sum test (Helsel andHirsch, 1991). However, if the E1 removal performances ofthe simple BFP are compared against all the other STP typesin this survey, it is shown to be significantly lower( p < 0.001).

� Additional tertiary biological treatment will improve thesteroid estrogen removal performance of simple biologicalfilter plants.

The available evidence indicates that the addition of a ter-tiary biological treatment stage will improve steroid estrogenremoval compared to standard BFP (mean percentage removalwas significantly different between the two groups using theWilcoxon rank-sum test, p ¼ 0.002). The addition of a tertiarysand filtration step does not necessarily improve estrogenremoval as shown by EBFP8 (Table 1).

� Effluent BOD5 and ammonium will be a good indicator ofestrogen removal performance.

Low effluent BOD can be an indication of how successfulthe biological treatment has been in consuming all the carbo-naceous material in sewage. It may be reasonable to assumethat the more successful the overall biological treatment hasbeen, the more likely steroid estrogens will have been con-sumed also. Low ammonium is an indication of successfulnitrification, a biological process that has been associatedwith slow growing bacteria which favor long SRT in activatedsludge (Joss et al., 2004; Vader et al., 2000). Thus, both indi-cators of biological performance may also be unbiased guidesto estrogen removal in treatment. This new hypothesis was

Table 3

Summary of estrogen removal performance by different STP

Treatment type Mean E2 removal (%) Mean E1 removal (%) Mean effluent BOD5 (mg/L) Mean effluent ammonium (mg/L)

Previous ASP review

(Johnson and Williams, 2004)

82 � 11 65 � 6 NA NA

ASP 65 (SD 50, n ¼ 6) 72 (SD 48, n ¼ 8) 2.5 (n ¼ 2) 0.5 (n ¼ 2)

EBFP 89 (SD 6, n ¼ 10) 74 (SD 29, n ¼ 17) 5.6 (SD 3.3 n ¼ 8) 1.3 (SD 1, n ¼ 7)

BFP 70 (SD 36, n ¼ 17) 30 (SD 31 n ¼ 18) 9 (SD 4, n ¼ 9) 2 (SD 1.4 n ¼ 9)


tested by regression using the annual mean BOD and ammoniaconcentration data provided by the water companies in theirresponse sheets (Table 1). There was no significant trend forbetter E1 removal with lower ammonia concentrations(R2 ¼ 0.06, p ¼ 0.15) or lower BOD5 (R2 ¼ 0.1, p ¼ 0.07).

6. Conclusions

From the limited available data, it is clear that there arevery few types of treatment, or practice, from which infalliblepredictions on estrogen removal performance can be made.Perhaps this should not be too surprising. Effective sewagetreatment requires a successful partnership every day of theyear between bacteria, nutrient loading, mechanical equipmentand human operators. Occasionally one of these factors willchange, or fail. Consequently it must be accepted that robustconclusions on the efficacy of one sewage treatment over an-other cannot be drawn from a small dataset. For a catchmentrisk assessment of endocrine disruption it would still seem rea-sonable to model all STPs, with the exception of BFPs, on thebasis of typical ASP performance as previously suggested(Johnson and Williams, 2004). To ensure endocrine disruptionrisk is not under-predicted, however, it would appear sensibleto identify the location of simple BFPs and award them a lowerE1 removal factor. On the basis of this study, this would be30% E1 removal, whilst E2, and EE2 removal factors wouldremain as for ASP.

Acknowledgements

The authors wish to thank the Environment Agency andCEH science budget for supporting this study, and MelanieGross-Sorokin for her tireless efforts in getting the projectstarted and obtaining the treatment information. The authorsare grateful for helpful comments from Colin Neal, MikeHutchins, David Boorman and John Sumpter.

References

Andersen, H., Siegrist, H., Halling-Sørensen, B., Ternes, T.A., 2003. Fate of

estrogens in a municipal sewage treatment plant. Environ. Sci. Technol.

37, 4021e4026.

Anderson, P.D., D’Aco, V.J., Shanahan, P., Chapra, S.C., Buzby, M.E.,

Cunningham, V.L., Duplessie, B.M., Hayes, E.P., Mastrocco, F.J.,

Parke, N.J., Rader, J.C., Samuelian, J.H., Schwab, B.W., 2004. Screening

analysis of human pharmaceutical compounds in US surface waters. Envi-

ron. Sci. Technol. 38, 838e849.

Baronti, C., Curini, R., D’Assenzo, G., Di Corcia, A., Gentili, A., Samperi, R.,

2000. Monitoring natural and synthetic estrogens at activated sludge sew-

age treatment plants and in receiving river water. Environ. Sci. Technol.

24, 5059e5065.

Blackburn, M.A., Waldock, M.J., 1995. Concentrations of alkylphenols in

rivers and estuaries in England and Wales. Water Res. 29, 1623e1629.

Boon, A.G., Hemfry, J., Boon, K., Brown, M., 1997. Recent developments in

the biological filtration of sewage to produce high-quality nitrified efflu-

ents. J. Chartered Inst. Water Environ. Manage. 11, 393e412.

Clara, M., Strenn, B., Ausserleitner, M., Kreuzinger, N., 2004. Comparison of

the behaviour of selected micropollutants in a membrane bioreactor and

a conventional wastewater treatment plant. Water Sci. Technol. 50, 29e36.

Cooper, P.F., Downing, A.L., 1998. Milestones in the development of the

activated sludge process over the past eighty years. J. Chartered Inst. Water

Environ. Manage. 12, 303e313.

COUNCIL DIRECTIVE, 1991. Directive of 21 May 1991 concerning urban

waste water treatment (91/271/EEC). Off. J Eur. OJ L 135, 30.5.1991, p. 40.

Desbrow, C., Routledge, E.J., Brighty, G.C., Sumpter, J.P., Waldock, M., 1998.

Identification of estrogenic chemicals in STW effluent. 1. Chemical frac-

tionation and in vitro biological screening. Environ. Sci. Technol. 32,

1549e1558.

Cheeseman, R.V., Wilson, A.L., 1989. A Manual on Analytical Quality Con-

trol for the Water Industry, NS30. Water Research Centre.

Environment Agency, 2006. Assessment of (anti-) estrogenic and (anti-) an-

drogenic activities of final effluents from sewage treatment works. Science

Report SC020118/SR, 2006. Environment Agency, Bristol.

Helsel, D.R., Hirsch, R.M., 1991. Statistical methods in water resources. In:

Techniques of Water Resources Investigations of the United States Geolog-

ical Survey: Book 4, Hydrological Analysis and Interpretation, Chapter 5,

pp. 118e123. Internet version at <http://pubs.usgs.gov/twri/twri4a3/>,

(accessed 1.8.06.).

Jobling, S., Nolan, M., Tyler, C.R., Brighty, G., Sumpter, J.P., 1998. Widespread

sexual disruption in wild fish. Environ. Sci. Technol. 32, 2498e2506.

Jobling, S., Tyler, C.R., 2003. Endocrine disruption in wild freshwater fish.

Pure Appl. Chem. 75, 2219e2234.

Johnson, A.C., Aerni, H.-R., Gerritsen, A., Gibert, M., Giger, W., Hylland, K.,

Jurgens, M., Nakari, T., Pickering, A., Suter, M.J.-F., Svenson, A.,

Wettstein, F.E., 2005. Comparing steroid estrogen, and nonylphenol con-

tent across a range of European sewage plants with different treatment

and management practices. Water Res. 39, 47e58.

Johnson, A.C., Belfroid, A., Di Corcia, A., 2000. Estimating steroid estrogen

inputs to activated sludge treatment works and observations on their

removal from the effluent. Sci. Tot. Environ. 256, 163e173.

Johnson, A.C., Sumpter, J.P., 2001. Removal of endocrine disrupting chemicals

in activated sludge treatment works. Environ. Sci. Technol. 35, 4697e4703.

Johnson, A.C., Williams, R.J., 2004. A model to estimate influent and effluent

concentrations of estradiol, estrone and ethinyloestradiol at sewage treat-

ment works. Environ. Sci. Technol. 38, 3649e3658.

Joss, A., Andersen, H., Ternes, T., Richler, P.R., Siegrist, H., 2004. Removal of

estrogens in municipal wastewater treatment under aerobic and anaerobic

conditions: Consequences for plant optimisation. Environ. Sci. Technol.

38, 3047e3055.

Komori, K., Tanaka, H., Okayasu, Y., Yasojima, M., Sato, C., 2004. Analysis

and occurrence of estrogen in wastewater in Japan. Water Sci. Technol. 50,

93e100.

Kreuzinger, N., Clara, M., Strenn, B., Kroiss, H., 2004. Relevance of sludge

retention times (SRT) as design criteria for wastewater treatment plants

for the removal of endocrine disrupters and pharmaceuticals from waste-

water. Water Sci. Technol. 50, 149e156.

Marsh, T.J., Lees, M.L., 2003. Hydrometric Register and Statistics 1996e

2000. Hydrological Data UK Series. CEH Wallingford/British Geological

Survey, Wallingford, UK.

National Rivers Authority, 1993. River Aire Catchment Management Plan. Re-

port N.N.Y.R. National Rivers Authority, Leeds, UK.

Onda, K., Nakamura, Y., Takatoh, C., Miya, A., Katsu, Y., 2003. The behavior

of estrogenic substances in the biological treatment process of sewage.

Water Sci. Technol. 47, 109e116.

Reddy, S., Iden, C.R., Brownawell, B.J., 2005. Analysis of steroid conjugates

in sewage influent and effluent by liquid chromatography-tandem mass

spectrometry. Anal. Chem. 77, 7032e7038.

Schowanek, D., Webb, S., 2002. Exposure simulation for pharmaceuticals in

European surface waters with GREAT-ER. Toxicol. Lett. 131, 39e50.

Schowanek, D., Fox, K., Holt, M., Schroeder, F.R., Koch, V., Cassani, G.,

Matthies, M., Boeje, G., Vanrolleghem, P., Young, A., Morris, G.,

Gandolfi, C., Feijtel, T.C.J., 2001. GREAT-ER: a new tool for management

and risk assessment of chemicals in river basins. Contribution to GREAT-

ER#10. Water Sci. Technol. 43, 179e185.

Shi, J., Fujisawa, S., Nakai, S., Hosomi, M., 2004. Biodegradation of natural

and synthetic estrogens by nitrifying activated sludge and ammonia-

oxidizing bacterium Nitrosomonas europea. Water Res. 38, 2322e2329.

http://pubs.usgs.gov/twri/twri4a3/


Sole, M., Lopez De Alda, M.J., Castillo, M., Porte, C., Ladegaard-

Pedersen, K., Barcelo, D., 2000. Estrogenicity determination in sewage

treatment plants and surface waters from the Catalonian area (NE Spain).

Environ. Sci. Technol. 34, 5076e5083.

Stumpf, M., Ternes, T.A., Wilken, R.-D., Rodrigues, S.V., Baumann, W., 1999.

Polar drug residues in sewage and natural waters in the state of Rio de

Janeiro. Brasil. Sci. Tot. Environ. 225, 135e141.

Sumpter, J.P., Johnson, A.C., 2005. Lessons from endocrine disruption and

their application to other issues concerning trace organics in the aquatic

environment. Environ. Sci. Technol. 39, 4321e4332.

Sumpter, J.P., Johnson, A.C., Kortenkamp, A., Scholz, M., Williams, R.J.,

2006. Modelling effects of mixtures of endocrine disrupting chemicals at

the river catchment scale. Environ. Sci. Technol. 40, 5478e5489.

Svensen, A., Allard, A.-R., Ek, M., 2003. Removal of estrogenicity in Swedish

municipal sewage treatment plants. Water Res. 37, 4433e4443.

Ternes, T.A., Stumpf, M., Mueller, J., Haberer, K., Wilken, R.-D., Servos, M.,

1999a. Behaviour and occurrence of estrogens in municipal sewage

treatment plants e I. Investigations in Germany. Canada and Brazil. Sci.

Tot. Environ. 225, 81e90.

Ternes, T.A., Kreckel, P., Mueller, J., 1999b. Behaviour and occurrence of

estrogens in municipal sewage treatment plants e II. Aerobic batch exper-

iments with activated sludge. Sci. Tot. Environ. 225, 91e99.

Vader, J.S., Van Ginkel, C.G., Sperling, F.M.G.M., de Jong, J., de Boer, W., de

Graaf, J.S., van der Most, M., Stokman, P.G.W., 2000. Degradation

of ethinyl estradiol by nitrifying activated sludge. Chemosphere 41,

1239e1243.

Williams, R.J., Johnson, A.C., Smith, J.J.L., Kanda, R., 2003. Detailed in-

vestigation of steroid estrogen profiles along river stretches arising

from sewage treatment works discharges. Environ. Sci. Technol. 37,

1744e1750.

Zuehlke, S., Duennbier, U., Heberer, T., 2005. Determination of estro-

genic steroids in surface water and wastewater by liquid chromatogra-

phy-electrospray tandem mass spectrometry. J. Separation Sci. 28,

52e58.

What difference might sewage treatment performance make to endocrine disruption in rivers?

Documents