Suitability of fluorescent whitening compounds (FWCs) as indicators of human faecal contamination from septic tanks in rural catchments

Suitability of fluorescent whitening compounds (FWCs) as indicators of human faecal contamination from septic tanks in rural catchments

Accepted Manuscript Suitability of fluorescent whitening compounds (FWCs) as indicators of human faecal contamination from septic tanks in rural catch...

4MB Sizes 1 Downloads 48 Views

Accepted Manuscript Suitability of fluorescent whitening compounds (FWCs) as indicators of human faecal contamination from septic tanks in rural catchments Donata Dubber, Laurence W. Gill PII:

S0043-1354(17)30829-1

DOI:

10.1016/j.watres.2017.10.005

Reference:

WR 13260

To appear in:

Water Research

Received Date: 8 June 2017 Revised Date:

23 August 2017

Accepted Date: 3 October 2017

Please cite this article as: Dubber, D., Gill, L.W., Suitability of fluorescent whitening compounds (FWCs) as indicators of human faecal contamination from septic tanks in rural catchments, Water Research (2017), doi: 10.1016/j.watres.2017.10.005. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

Fluorescent whitening compounds (FWCs) Photodecay under UV-light exposure

DSBP

DAS1 0

Exposure time

M AN U

SC

DAS1

% of initial signal

RI PT

NOM

Fluorometric detection

EP

run-off

Septic tank

AC C

ater

Percolation area

NOM

FWC detection limits in surface water

FWC

NOM

FWC

FWC

FWC

Photo detector

Surface water NOM

Distinguishes signal source

Detection limit

FWC

Water sample

TE D

Faecal bacteria/ viruses

FWC

WC

100

DAS1

DSBP

NOM

NOM concentra

ACCEPTED MANUSCRIPT

Suitability of Fluorescent Whitening Compounds (FWCs) as indicators of

2

human faecal contamination from septic tanks in rural catchments

3

Donata Dubbera* and Laurence W. Gilla

5 6

a

Department of Civil, Structural and Environmental Engineering, Trinity College Dublin, Museum Building, College Green, Dublin 2, Ireland

* Corresponding author: [email protected]

SC

4

RI PT

1

7 Abstract

9

Rural river catchments are impacted by diffuse pollution sources from agricultural

10

practices and on-site domestic wastewater treatment systems (DWWTS), mainly septic

11

tanks. Methods that can distinguish between contamination sources will significantly

12

increase water management efficiency as they will allow for the development and

13

application of targeted remediation measures. Fluorescent whitening compounds

14

(FWC), are used as optical brighteners in laundry detergents and enter the

15

environment through the discharge of domestic wastewater effluents. Due to their

16

human specific source and potential simple fluorometric measurement this represents

17

a very attractive method to be used by state monitoring agencies.

18

In this study the suitability of FWCs as chemical indicators for human faecal

19

contamination has been investigated in rural Irish catchments. It was found that no

20

quantitative measurements are possible for FWCs in natural waters when using simple

21

fluorometric methods. Hence a simple presence/absence approach needs to be

22

applied. The detectability of FWCs was quantified and found to decrease with higher

23

organic matter content of the river water which has its own fluorescence. This enabled

AC C

EP

TE D

M AN U

8

1

ACCEPTED MANUSCRIPT the establishment of equations to predict detection limits and assess the method’s

25

suitability for individual catchments based on organic matter concentrations.

26

Furthermore a modified photodecay method is suggested that increases sensitivity of

27

the technique by up to 59%. Applications at rural study sites found some removal of

28

FWCs in percolation areas of DWWTSs but they were still detectable 40 cm below the

29

infiltration depth. FWCs were also detected as distinguishable peaks in impacted

30

streams where septic tank effluents have a high contribution to the river flow.

SC

RI PT

24

31

Keywords: Fluorometry; Optical brighteners; Pollution source tracking; Photodecay;

33

Domestic wastewater; Surface Water

34

M AN U

32

1. Introduction

36

Over the years several source tracking methods (phenotypic, genotypic and chemical)

37

have been developed to identify the host origin of faecal contamination in water

38

(Scott et al. 2002). They have been reported to be a useful technique to identify and

39

locate contamination sources which is often necessary for effective remediation. This

40

is of particular interest to European countries to facilitate a targeted implementation

41

of the Programme of Measures in order to meet objectives of the Water Framework

42

Directive (2000/60/EC).

EP

AC C

43

TE D

35

44

One of the chemical methods involves the use of fluorescent whitening compounds

45

(FWC) as indicator for human faecal contamination. FWCs are used as optical

46

brighteners in laundry detergents. During the wash cycle they attach to the fabrics and

47

because they emit light in the blue range (415-445 nm) they compensate for yellowing 2

ACCEPTED MANUSCRIPT and make clothes look whiter (Hartel et al. 2007a). However, they also remain to some

49

extent in the washing liquor and enter the environment through the discharge of

50

domestic wastewater effluents (Poiger et al. 1998). It is known that two FWCs are

51

commonly used as optical brighteners in laundry detergents: Distyrylbiphenylsulfonate

52

(DSBP) and the diaminostilbene DAS 1 (Hagedorn et al. 2005b, Kramer et al. 1996,

53

Poiger et al. 1999). While washing detergents are probably the primary source, FWCs

54

are also used in toilet paper and released into the wastewater when the tissue breaks

55

down (Hagedorn et al. 2005b).

56

FWCs are regarded as good indicators of human contamination. It is possible to

57

measure them with great sensitivity using liquid chromatography (Poiger et al. 1996,

58

Stoll and Giger 1997) but the measurement of their fluorescence signal is a much

59

quicker, inexpensive and easier method which thus represents an interesting method

60

for state monitoring agencies. It does not require high technical expertise and can

61

even be performed with a handheld fluorometer in the field (Cao et al. 2009,

62

Dickerson et al. 2007, Hartel et al. 2007a).

63

Studies have already demonstrated the successful application of the fluorometric

64

method for detecting FWCs as indicator for wastewater contamination in

65

environmental waters (Dickerson et al. 2007, Hagedorn et al. 2005a, Hartel et al.

66

2007b). For example Dickerson et al. (2007) successfully used FWCs in combination

67

with a phenotypic source tracking method (antibiotic resistance analysis) to locate

68

pollution sources at two public beaches in Virgina. After repair of the identified

69

sewage leaks a repeated assessment confirmed the effectiveness of the mitigation

70

efforts made. However, in some studies problems have been reported in that the

71

background fluorescence from organic matter was shown to interfere with the

AC C

EP

TE D

M AN U

SC

RI PT

48

3

ACCEPTED MANUSCRIPT fluorometric detection of FWCs. In a first instance Hartel et al. (2007b) changed the

73

emission filter from a broad spectrum (410-600 nm) to a narrow spectrum filter at 436

74

nm. This cuts off the emission of longer wavelength which is mainly composed of the

75

fluorescence from organic matter. Even though this reduced the background

76

fluorescence by > 50%, it still prevented a successful detection of FWCs in organic rich

77

waters (Hartel et al. 2007a). To solve this problem Hartel et al. (2007a) made use of

78

the photofading effect (also called photodecay of fluorescence signals) which is

79

greater for FWCs than for natural organic matter. Hence, by exposing water samples to

80

sunlight and comparing the signal reduction it allowed them to distinguish between

81

fluorescence signals from FWCs and organic matter. Cao et al. (2009) used the same

82

general concept but developed the method further and increased its sensitivity by

83

taking advantage of the difference in shape of the photodecay curves. They introduced

84

a ratio of signal reductions after different times of UV exposure which they used

85

together with a set threshold to determine presence and absence of FWCs in a water

86

sample.

SC

M AN U

TE D

EP

87

RI PT

72

Concentrations of FWCs as well as their fate and application as indicator has been

89

investigated in lakes (Stoll and Giger 1997), urban rivers (Poiger et al. 1999), coastal

90

waters/public beaches (Dickerson et al. 2007, Hagedorn et al. 2005a, Hartel et al.

91

2007a) and storm drains (Dickerson et al. 2007, Hartel et al. 2007a) but studies about

92

the use of FWCs in order to identify diffuse pollution from failing soil based DWWTS in

93

small streams appear to have been very limited. Hayakawa et al. (2007) studied the

94

distribution and fluxes of FWAs discharged from domestic wastewater into small rivers

95

in rural catchments in Japan. They found highest concentrations in subcatchments that

AC C

88

4

ACCEPTED MANUSCRIPT were dominated by DWWTS as opposed to central wastewater treatment plants.

97

However, they used the more sensitive and complex method of detection using liquid

98

chromatography (solid phase extraction followed by HPLC analysis) as opposed to the

99

fluorometric detection method. Nevertheless, their findings highlight the chance and

100

importance of a successful application in Ireland where a high proportion (28.7%) of

101

households rely on DWWTS (CSO 2016). Most of these consist of a septic tank with a

102

percolation area. Hence, the aim of this study is to assess and modify existing

103

analytical methods for the fluorometric measurement of FWCs in the Irish

104

environment and to test their suitability to distinguish between the environmental

105

impact from agricultural activities and failing DWWTS.

M AN U

SC

RI PT

96

106 2. Material and Methods

108

Standards of the FWC compounds commonly used in laundry detergents, DSBP (4,4’-

109

bis (2-sulfostyryl) biphenyl disodium salt, CAS# 27344-41-8, Synthon Chemicals GmbH,

110

Germany)

111

yl)amino]stilbene-2,2’-disulfonate, CAS# 16090-02-1, Sarchem Laboratories Inc., USA)

112

(Hartel et al 2007, Poiger et al 1999), were used to determine fluorometric

113

characteristics and detection limits as well as for the quantification of FWCs in

114

environmental samples.

115

The Suwannee River Organic Matter from the International Humic Substances Society

116

(IHSS) was used as a natural organic matter (NOM) standard to examine and quantify

117

any interference with the fluorometric measurement of FWCs and to determine the

118

fluorescence degradation of organic matter. Humic acid sodium salt (CAS# 68131-04-4,

119

Santa Cruz Biotechnology Inc.) and surface water from pristine environment were also

DAS1

(Disodium

4,4’-bis[(4-anilino-6-morpholino-1,3,5-triazin-2-

AC C

EP

and

TE D

107

5

ACCEPTED MANUSCRIPT used as organic reference material. The environmental samples were taken from the

121

Cloghoge and Glenmacnass River in the Wicklow Mountains National Park with all

122

sampling sites located far upstream of any human settlement.

123

A total of 29 Laundry detergents including 17 powder based (12 powders, 5 tabs), 10

124

liquid based (8 liquids, 2 capsules) and 2 gel detergents from various suppliers were

125

purchased and analysed. 7 detergents (4 powder based, 2 liquid based and 1 gel) did

126

not have optical brighteners on their list of ingredients. Stock solutions were prepared

127

in distilled water at concentrations of 1 g/L or 500 µL/L, using an ultrasonic bath for

128

better dissolution where needed. Effluent samples were collected from four different

129

private septic tanks serving single houses in rural Ireland with an occupancy of 3-5

130

persons.

M AN U

SC

RI PT

120

131

2.1 Fluorescence characteristics and photodecay rates

133

A LS55 Fluorescence Spectrometer (Perkin Elmer) was used for the measurement of

134

FWCs in standard solutions, detergents, septic tank effluent and environmental water

135

samples. Fluorescence PMMA cuvettes with 10 mm optical path length were used for

136

all measurements. Emission spectra were recorded at an excitation wavelength of λex =

137

350 nm (absorption maximum for DSBP and DAS1 (Kramer et al. 1996)) with the

138

emission wavelength λem ranging from 390 to 600 nm. For a direct measurement of

139

the fluorescence signals from FWCs the emission wavelength was set at λem = 436 nm

140

(Hartel et al. 2007b) with a slit width of either 5 or 10 nm. The limit of detection (LOD)

141

of the used fluorometer was measured using distilled water as blank in 20 replicates

142

and determined according to Eq 1:

AC C

EP

TE D

132

6

ACCEPTED MANUSCRIPT 143

(Eq.1)

" =  + 3 "

144

with Sblank and σblank being the average signal strength and the standard deviation of

145

the blank measurements, respectively.

146 Calibration curves with low concentrations (0 to 0.06 µg/L and 0 to 1.3 µg/L) were

148

recorded for DSBP and DAS1 and used to determine the LOD for these compounds.

149

Photodecay curves of the fluorescence signal for FWC standards, organic matter

150

reference material (NOM, humic acid and pristine river water) and for all detergents

151

were recorded after 0, 1, 5, 10 and 15 min UV exposure. UV exposure was conducted

152

in a large dark box usinga facial tanner sun lamp with 4 Philips Cleo 15W UV tubes.

153

Cuvettes with the sample were placed into a LDPE holder (Kartell labware) that was

154

centrally positioned in front of the UV tubes at a fixed height and distance of 16 and 5

155

cm respectively. The box had a small ventilation slot and was additionally ventilated

156

throughout after each exposure to minimise a potential heat build-up.

157

If not stated differently, all fluorescence measurements and analyses were always

158

carried out in at least three replicates.

SC

M AN U

TE D

EP

159

RI PT

147

2.2 FWC concentrations in laundry detergents

161

Using 5-point calibrations (DSBP 0 to 20 µg/L, DAS1 0 to 200 µg/L) obtained from the

162

FWC standard compounds, concentrations of FWCs were determined for the tested

163

detergent solutions. Based on these results the amounts of FWCs contained in 1 g or 1

164

mL of powder and liquid based detergent were calculated respectively.

AC C

160

165

7

ACCEPTED MANUSCRIPT 2.3 NOM Interferences and Detection Limit

167

The fluorescence signal reduction after 1, 5 and 10 min of UV light exposure was

168

recorded for different concentrations of NOM (4, 8, 16, 23 and 30 mg/L) spiked with

169

different amounts of DSBP or DAS1 (final concentrations ranging from 0.25 to 8 μg/L

170

and 0.5 to 16 μg/L respectively). A detection limit analysis was carried out applying the

171

photodecay signal reduction ratio (10/5 min), recommended by Cao et al (2009): %      

%      

(Eq. 2)

SC

172

RI PT

166

The method by Cao et al (2009) recommends that a ratio smaller than the threshold of

174

1.5 indicates the presence of FWC. The results were compared to those obtained from

175

the application of a new proposed ratio of the reduction after 1 min to the reduction

176

after 10 min of UV exposure.

M AN U

173

%      

%      

177

(Eq. 3)

Samples with a ratio (1/10 min) >0.25 are considered to contain FWCs. Regression

179

analysis was carried out in order to define the detection limit for DSBP and DAS1 when

180

organic matter is present in the same solution. In order to define a detection limit for

181

which the observed ratio will be statistically significantly smaller/greater than the

182

defined threshold, the 95% prediction intervals (PIs) were determined. The

183

upper/lower PI was then used to determine the point of intersection with the

184

respective threshold.

AC C

EP

TE D

178

185 186

2.4 Application in Rural Catchments

187

Two small catchments in areas designated as having a high likelihood of inadequate

188

percolation to ground (due to the presence of clayey subsoils) and with a high density 8

ACCEPTED MANUSCRIPT of DWWTS were selected to study their impact on the water quality of small streams.

190

The first catchment, located on the east coast of Ireland to the south of Dublin is 3.3

191

km2 in size, with a total of 72 DWWTSs (22/km2); 54% of the DWWTSs are within 100

192

m of the stream. An upstream and a downstream site were selected for regular

193

monitoring. While no houses were located in the catchment area of the upstream site

194

(Cat#1 upper) many houses were situated close along the stream just before the

195

downstream sampling point (Cat#1 lower) (Figure S1). The second catchment, located

196

in the south-eastern area of Ireland, is 2.5 km2 in size with a total of 85 DWWTSs

197

(34/km2); 45% of the DWWTSs are within 200 m of the stream. There are 15 houses in

198

the area upstream of the upper sampling point (Cat#2 upper) and some houses are

199

close along the stream just before the downstream sampling point (Cat#2 lower)

200

(Figure S2).

201

Grab samples for FWC analyses were collected at upstream and downstream sites as

202

well as along the river (white circles in Fig. S1 and S2). Additionally, week long

203

sampling events using the autosampler (set at 7 hour intervals) were performed in

204

May and June 2015. To protect samples from UV light, they were stored in amber glass

205

bottles and transported to the lab for analysis at the fluorometer. The presence of

206

FWCs in water samples was determined using the photodecay method modified after

207

Cao et al. (2009) applying a 1/10 min signal reduction ratio. Samples with a ratio > 0.25

208

indicate the presence of FWCs.

209

An estimation of FWC detectability in surface water was carried out for both FWC

210

compounds. Using the average DSBP and DAS1 concentration from all measured liquid

211

and powdered detergents respectively, and with the recommended detergent dosage

212

(70 g for powders, 35 mL for liquids) the FWC load per wash was determined. Based on

AC C

EP

TE D

M AN U

SC

RI PT

189

9

ACCEPTED MANUSCRIPT a household of 3 persons (avg. occupancy rate in rural Ireland = 2.85 (CSO 2016)) with

214

a wastewater production of 304 L/d (avg. wastewater production in studied DWWTS

215

(Dubber and Gill 2014)) and assuming that a washing machine is run every 2 days with

216

50% of the FWCs removed during the wash cycle through adsorption to cloth

217

(Dickerson et al. 2007), the FWC concentration in a septic tank was estimated.

218

Conductivity, temperature and water level (CTD) sensors (OTT hydrometry, UK) were

219

installed at the upstream and downstream sites in both study catchments from which

220

continuously recorded flow data was available. This data was taken into account to

221

determine dilution in the study catchments and to estimate expected FWC

222

concentrations in the river water at the studied sampling sites.

M AN U

SC

RI PT

213

223

Finally, in order to assess the detectability in effluent percolating through the soil,

225

samples were taken from two DWWTS (3-4 PE) with percolation areas in more

226

permeable subsoils comprised of sandy silt and sandy loam. The sites had been set up

227

such that half of the percolation area at each site (2 trenches) was receiving primary

228

effluent from a septic tank while the other two trenches were receiving secondary

229

treated effluent from a small packaged treatment plant. Primary and secondary

230

effluent samples were taken and soil moisture samples were extracted from different

231

depths across the percolation areas via suction lysimeters (Soilmoisture Equipment

232

Corp., California). By using a hand pump a vacuum of approximately 0.5 bar was

233

applied to the lysimeters in order to collect percolating effluent and rainwater from a

234

horizontal depth plane within the vadose zone over a 24 hour time frame. In total 8

235

effluent samples and 35 soil moisture samples were taken in February and March 2017

AC C

EP

TE D

224

10

ACCEPTED MANUSCRIPT 236

and analysed for the presence of FWC using the photodecay method, applying the

237

1/10 min signal reduction ratio.

238 2.5 Statistical evaluation

240

The statistical data evaluation, confidence interval calculations and regression analyses

241

were carried out using the statistical software package IBM SPSS 22.0 as well as

242

Microsoft Excel Professional Plus 2013.

243

Data was tested for normality using the Shapiro-Wilk test for sample sizes n<50. For

244

normally distributed data a one-way ANOVA (with Post Hoc Dunnetts test) or

245

independent-samples T-tests were performed to identify statistically significant

246

differences in sample sets or between two samples (e.g. for photodecay rates). Where

247

normal distribution of data was not given, the Kruskal-Wallis-Test was used for sample

248

sets of n>2 using stepwise comparison approach to identify homogeneous subsets or

249

the Kolmogorov-Smirnov Test was applied to compare the means of two independent

250

samples. The comparison of measured photodecay ratios (1/10 and 5/10 min) to their

251

respective thresholds (0.25 and 1.5) was performed using a one-sample T-test.

252

For the detection limit analysis linear as well as quadratic regressions were carried out

253

and 95% prediction intervals (PI) determined. A subsequent regression of the 95% PI

254

provided the equation that was used to determine the concentration at which the

255

threshold value was reached.

AC C

EP

TE D

M AN U

SC

RI PT

239

256

11

ACCEPTED MANUSCRIPT 3. Results

258

3.1 Fluorescence characteristics and photodecay rates

259

Based on the results from the standard calibrations, the signal response from DSBP

260

was found to be about 8 times higher than for DAS1 at the same concentration. The

261

limit of detection for DSBP and DAS1 was 0.021 and 0.17 µg/L respectively.

262

Figure 1 shows the emission spectra for both FWC standards compared to natural

263

organic matter reference material (IHSS NOM standard, humic acid sodium salt and

264

pristine river water). These emission spectra revealed that FWCs not only have

265

different signal response but also their emission maxima are at different wavelengths.

266

While DSBP has its emission maximum with a distinctive sharp peak at 425 nm, the

267

emission spectrum from DAS1 is characterised by a wider peak at slightly longer

268

wavelengths between 430 and 440 nm. In comparison the emission maximum for the

269

NOM standard and for the two pristine river water samples was found to be at 452

270

nm, 454 nm and 455 nm respectively. The humic acid sodium salt has a wide and flat

271

emission peak between 489 and 503 nm with a second maximum between 534 and

272

537 nm. These maxima are also seen to form part of the NOM spectra as well as the

273

fluorescence from fulvic acid which has not been measured in this study but is

274

reported to be at around 420 nm and 450 nm for microbially and terrestrially derived

275

material (McKnight et al. 2001).

276

Figure 2 shows that the emission spectrum from a septic tank effluent sample is

277

similar to that of the liquid detergent used in the respective house. Both spectra show

278

a distinctive sharp peak at 425 nm as observed for the DSBP standard. Also shown for

AC C

EP

TE D

M AN U

SC

RI PT

257

12

ACCEPTED MANUSCRIPT comparison is a spectrum for the DAS1 standards and a powder based detergent

280

which have a broader peak between 425 and 435 nm.

281

The recorded photodecay rates of both FWC standards were significantly different

282

(p<0.01) to those of the NOM standard (see Figure 3). However, there was also a

283

significant difference (p<0.01) between the photodecay of the fluorescence signal of

284

DSBP and DAS1. The average signal reduction after 1 min of exposure to UV light was

285

1.58 ± 0.5%, 14.36 ± 2.8% and 76.64 ± 4.9% for NOM, DSBP and DAS1 respectively.

286

There were no significant differences between the photodecay observed from NOM,

287

pristine river samples and the humic acid solution. The average signal reduction after 1

288

min UV exposure for all organic material together (incl. pristine river water samples

289

and humic acid) was 1.51 ± 0.4%. Also no significant differences in photodecay were

290

observed between DSBP and most (6 out of 8) liquid detergents or between DAS1 and

291

powdered detergents (see Figure 3).

SC

M AN U

TE D

292

RI PT

279

3.2 FWC concentrations in laundry detergents

294

Linear calibration curves with slopes of 40.4 and 5.1 were obtained for DSBP and DAS1

295

respectively, with excellent linear fits for both calibrations with R2 coefficients ≥0.99

296

(Fig. S1). Due to the different signal response for DSBP and DAS1, the concentrations

297

of FWCs in detergents were determined using both calibrations. However, emission

298

spectra shape similarities and results from the photodecay analysis (signal reduction

299

after 1 min of UV exposure for DSBP around 15% and DAS1 >75%) indicate the type of

300

FWC used in the detergent.

301

All 13 powder based detergents used DAS1 with concentrations ranging from 0.22 up

302

to 4.35 mg/g (Figure S4). From 8 liquid based detergents that contained optical

AC C

EP

293

13

ACCEPTED MANUSCRIPT brighteners, 6 used DSBP in concentrations ranging from 0.35 to 2.99 mg/mL while the

304

other 2 showed indications of DAS1 being used. FWC concentrations for these were

305

recorded to be 1.6 and 2 mg/mL. Average concentration in DAS1 containing

306

detergents was 1.35 mg/g and FWC concentration in DSBP containing detergents was

307

2.35 mg/mL.

308

The 7 detergents that stated not to contain optical brighteners were either eco-

309

friendly detergents or products for coloured clothes. Here, no or only traces of DSBP

310

(<0.002 mg/mL and 0.04 mg/mL) and DAS1 (<0.02 mg/g) were found. However, for

311

one detergent, even though not stated on the package, a concentration of 0.477 mg

312

DAS1 per g detergent was found. Emission spectra recorded for the two tested gels

313

indicated the use of DSBP as optical brightener but only contained very small

314

concentrations of 0.0015 and 0.0054 mg/g.

M AN U

SC

RI PT

303

TE D

315

3.3 NOM Interferences and Detection Limit

317

Figure 4 shows results obtained from the photodecay analysis of the NOM standards

318

spiked with different concentrations of the FWC compounds (here DSBP). It can be

319

seen that for lower NOM standard concentrations the photodecay signal reduction

320

ratio (1/10 min) exceeded the threshold of 0.25 at lower DSBP concentrations than for

321

higher NOM concentrations. For instance in a 4 mg/L NOM solution a DSBP

322

concentration of 0.5 µg/L resulted in a statistically significant positive detection of this

323

FWC compound whereas a concentration of 2 µg/L DSBP was needed to be detectable

324

in the 30 mg/L NOM solution. Similar trends were observed for the NOM solutions

325

spiked with DAS1 and when applying the signal reduction ratio (10/5 min) after Cao et

AC C

EP

316

14

ACCEPTED MANUSCRIPT 326

al. (2009). This demonstrates that the detection limit of FWCs is dependent on the

327

NOM content of the sample.

328 Regression analysis found a relationship between the observed photodecay ratios of

330

the spiked samples and its final FWC concentrations. For DSBP this relationship was

331

best described with a quadratic function until its inflection point. With increasing DSBP

332

concentration the photodecay ratio (1/10 min and 10/5 min) first increases/decreases

333

significantly and then the curve slowly flattens out until reaching the

334

maximum/minimum ratio of around 0.5 and 1.35 respectively which would be the

335

expected value for a pure DSBP standard solution (Figure 5). The R2 coefficients for the

336

quadratic regressions with DSBP concentrations performed individually for the

337

different NOM solutions ranged between 0.93 and 0.98 for the 1/10 min photodecay

338

ratio and from 0.71 to 0.94 for the 10/5 min ratio (Table S1). For the DAS1 spiked

339

samples a quadratic regression (R2 of 0.98 and 0.93 for the 1/10 and 10/5 min ratio

340

respectively) was obtained only for the 4 mg/L NOM solution. For all of the other NOM

341

concentrations (> 4mg/L) the coefficient of the quadratic term was not statistically

342

significant so that a linear regression was applied. This is due to the measured

343

concentrations being in the lower range where the photodecay ratio still

344

increases/decreases strongly before reaching its maximum/minimum (0.95 and 1.02,

345

respectively). The R2 coefficients for the linear regressions performed individually for

346

the different NOM solutions ranged between 0.83 and 0.95 for the 1/10 min

347

photodecay ratio and from 0.6 to 0.96 for the 10/5 min ratio (Table S2).

348

To obtain a detection limit that will ensure a positive detection of FWCs the 95%

349

prediction intervals (PI) of the regression functions from the measured photodecay

AC C

EP

TE D

M AN U

SC

RI PT

329

15

ACCEPTED MANUSCRIPT ratios were determined. The obtained functions for the lower and upper PI bounds

351

were then used to calculate the FWC’s detection limits when applying the different

352

photodecay methods and their respective thresholds (Table 1). This way, samples with

353

a FWC concentration ≥ the detection limit will be identified as positive for the

354

presence of FWCs with a confidence level of ≥97.5%. The results show that detection

355

limits for DAS1 (0.59 – 14.20 µg/L) are consistently higher than for DSBP (0.42 – 2.56

356

µg/L). Also, by using the 1/10 min photodecay ratio detection limits decrease by 34-

357

59% for DSBP and by 17-51% for DAS1 compared to the ratio proposed by Cao et al.

358

(2009).

M AN U

SC

RI PT

350

359

As already evident from the results presented in Figure 4, the detection limit increases

361

with higher NOM concentrations. However, with the exact detection limits determined

362

and plotted against the NOM concentration this relationship can be quantified (Figure

363

6) and used to make predictions for the detection limit of FWCs in water with different

364

NOM content. For DSBP a linear relationship between the detection limit and NOM

365

concentrations was found while for DAS1 a quadratic regression gave the best fit. This

366

demonstrates that the detection limit for DAS1 is much more sensitive to higher NOM

367

concentrations than DSBP.

EP

AC C

368

TE D

360

369

Figure S5 shows the regression results from plotting the average photodecay ratios

370

(1/10 min) from all DSPB or DAS1 spiked NOM samples against the FWC fluorescence

371

signal proportion (i.e. [total signal intensity – NOM unspiked signal intensity]/total

372

signal). It shows that the DSBP signal needs to account for more than 18.8% to the

373

overall fluorescence signal in order to be detected (with a 2.5% probability of error). 16

ACCEPTED MANUSCRIPT 374

Due to the higher photodecay rate of DAS1 (77% vs. 14% after 1 min UV exposure) it

375

only needs to contribute 5.5% to the overall signal to reach a photodecay ratio value

376

that is significantly greater than the threshold of 0.25.

377 3.4 Application in Rural Catchments

379

The estimation of FWC detectability in surface water was carried out for both FWC

380

compounds. The average DSBP and DAS1 concentrations from all measured liquid and

381

powdered detergents was determined as 2.35 mg/mL and 1.35 mg/g. Using a

382

detergent dosage of 35 mL and 70 g, as recommended by most manufacturer, a DSBP

383

and DAS1 load of 82.25 mg and 94.5 mg per wash was estimated, respectively.

384

Assuming that a washing machine is run about every 2 days and that there is an FWC

385

removal of 50% during the wash cycle, the daily DSBP and DAS1 load would be 20.6

386

and 23.6 mg, respectively. Based on a household of 3 persons with a wastewater

387

production of 304 L/d the estimated DSBP and DAS1 concentration in the septic tank

388

effluent would then be 67.7 μg/L and 77.7 μg/L, respectively. Taking into account

389

further dilution in the study catchments (using continuously recorded flow data), Table

390

2 shows the expected FWC concentrations in the river water at the studied sampling

391

sites. It should be noted that these estimations are based on a simplifying assumption

392

that all households upstream of the monitored sites use either DSBP or DAS1

393

containing laundry detergents. For a direct comparison the FWC detection limits,

394

predicted using the equations in Figure 6 and the organic matter background

395

fluorescence intensity (equivalent NOM concentration) observed in the catchments,

396

are also shown.

AC C

EP

TE D

M AN U

SC

RI PT

378

17

ACCEPTED MANUSCRIPT During most of the study period the estimated DSBP concentration in both catchments

398

were above the detection limit and a fluorometric detection of this FWC would

399

theoretically be possible. At the downstream sites predicted concentrations were

400

usually around and up to 3 times of the detection limit with an increased possible

401

detectability during low flow conditions in April and July 2016 when they reached

402

levels up to 5 and 10 times higher than the detection limit. For the upstream site in

403

catchment #2 the estimated DSBP levels were below the detection limits and

404

theoretically not detectable in March and April 2015. With concentrations expected to

405

be only 1.2 or at most 1.9 times the detection limit in May/June 2015 and April/July

406

2016, the detection of DSBP during this time would be considered possible but rather

407

unlikely. With no houses upstream of the upper study site in catchment #1, no FWC

408

input into the stream water and hence no detection would be expected, which was

409

indeed the result except for one sampling occasion.

410

The estimation shows that in general it will be harder to detect traces of DAS1

411

containing laundry detergents especially in natural waters with high organic matter

412

background fluorescence. While in catchment #1 where the water had a low organic

413

content (fluorescence intensity equivalent to Suwannee NOM = 2 - 3 mg/L and

414

measured TOC = 1.4 - 3.3 mg/L), the detection limit was especially in April and July

415

2016 well below the estimated DAS1 concentrations, the detection at the upstream

416

site in catchment #2 (TOC = 8.85 mg/L) was not considered to be theoretically possible

417

during any time of the sampling period. Based on these results a possible detection of

418

DAS1 at the lower site in catchment #2 would also only be expected during the

419

extreme low flow conditions that prevailed during the sampling in April and July 2016.

AC C

EP

TE D

M AN U

SC

RI PT

397

18

ACCEPTED MANUSCRIPT Based on the FWC concentration estimation and applying the relevant detection limits

421

as determined in the previous section (Table 2) the minimum proportion that septic

422

tank effluent needs to contribute to the overall stream flow for FWCs to be

423

theoretically detectable can be determined. These vary in waters with different

424

organic matter content from 0.6% up to 2.5% for the detection of DSBP and from 0.9%

425

to 16.8% for DAS1 (Table 3).

RI PT

420

SC

426

In practice, it was indeed possible to detect FWCs in the two study catchments with

428

several samples testing positive for FWCs according to the photodecay method.

429

In catchment #1 a total of 48 grab samples were taken in the period between March

430

and June 2015 as well as in April and in July 2016. Twelve of these samples had an

431

average photodecay ratio of >0.25 but only four were statistically significant (Table 4).

432

However, another three samples, for which all replicate readings were greater than

433

the threshold of 0.25, can also be considered indicative for the presence of FWC. In

434

catchment #2 a total of 59 grab samples were taken of which 6 had an average

435

photodecay ratio of >0.25. Four of these were statistically significant and one more

436

sample was considered positive for the presence of FWCs due to all replicates being

437

consistently >0.25.

438

Figure 7 shows the results from a sampling week using the autosampler at the

439

upstream site in Catchment #2 under low flow conditions in May 2015. During this

440

time three samples were statistically significantly (p<0.01) above the set threshold of

441

0.25, clearly indicating the presence of FWCs. A second sampling week a month later

442

yielded one positive sample (p<0.01) while no FWCs were detected during this time at

443

the downstream site in this catchment. During the entire monitoring period especially

AC C

EP

TE D

M AN U

427

19

ACCEPTED MANUSCRIPT 444

two sites, a midstream site in catchment #1 and the upper site in catchment #2, tested

445

repeatedly positive for the presence of FWCs.

446 From the total of 35 soil moisture samples taken within the percolation area of two

448

DWWTS sites (11 samples in February and 26 in March), 7 samples had an average

449

photodecay ratio of >0.25 (Table 5). Only four of these samples were statistically

450

significant but another two sample were considered positive for the presence of FWCs

451

due to all replicates being consistently >0.25. Both households used a DSBP based

452

washing detergents. The average photodecay ratio for primary and secondary

453

effluents ranged from 0.46 to 0.6 in February and from 0.32 to 0.44 in March. Table 5

454

also gives the installation depth for the lysimeters. It shows that at site #1 in February

455

for instance FWCs were still detectable up to a depth of 40 cm below the percolation

456

trench.

TE D

457

M AN U

SC

RI PT

447

4. Discussion

459

Emission spectra revealed that the two FWCs most commonly used in laundry

460

detergents have emission maxima at different wavelengths and different signal

461

response. Similarly, Dickerson et al. (2007) also observed a difference in the level of

462

fluorescence between optical brighteners with the signal of DSBP found to be 60 times

463

higher than for the Fluorescence Brightener 28. In this study DSBP had a signal that

464

was about 8 times stronger than that of DAS1. Even though the Fluorescence

465

Brightener 28 is similar to DAS1 their structure, which ultimately affects their

466

fluorescence characteristics, is still different. As a consequence from the different

467

signal response, it is not possible to correctly quantify an unknown mix of these

AC C

EP

458

20

ACCEPTED MANUSCRIPT compounds in surface waters when using a simple fluorometric measurement. Hence,

469

only a simple presence/absence approach can be applied for the detection of FWCs in

470

natural waters. Furthermore, the interference with organic matter, which has been

471

shown in previous studies and demonstrated in this paper, requires the use of a

472

photodecay method. This method was first described by Hartel et al. (2007a) and then

473

further developed by Cao et al. (2009) in order to distinguish between fluorescence

474

signals from organic matter and FWCs. However, the photodecay rate of the

475

fluorescence signal from the two FWCs was found to be significantly different in this

476

study (Figure 3), which has not been reported and/or considered in previous

477

publications (Cao et al. 2009, Hartel et al. 2007a). While Kramer et al. (1996) already

478

reported different photofading rates for DSBP (7%) and DAS1 (71%) in river water

479

exposed to natural sunlight for 60 min, this was not considered when the photodecay

480

methods were developed. Instead of individual standards Cao et al. (2009) used a

481

specific commercial liquid laundry detergent (Tide) as reference. According to Hartel et

482

al. (2007b) this however only contained DAS1. It was noted that using the ratio and

483

threshold levels as defined by Cao et al. (2009), the method was not sensitive enough

484

to detect small concentrations of DSBP in spiked river water. This is due to the

485

significantly lower photodecay rate that was observed for DSBP. This triggered a more

486

detailed analysis of the detection limits and a comparison with different UV exposure

487

ratios, as described in this paper. From the recorded photodecay curves (Figure 3) it

488

was evident that the highest signal reduction for the FWCs occurred within the first

489

minute of UV exposure. This data point was considered to be more distinctive than the

490

reduction after 5 min of UV exposure, especially in order to distinguish the shape of

AC C

EP

TE D

M AN U

SC

RI PT

468

21

ACCEPTED MANUSCRIPT the photodecay curve of DSBP from the one of NOM. Consequently it was selected to

492

be used for the alternative photodecay ratio to be tested.

493

The results demonstrate that the detectability of FWCs is dependent on the content of

494

organic matter and detection limit concentrations increase with higher organic matter

495

content in the sample. Even though DAS1 has the higher photodecay rates and a

496

smaller proportion of the total signal is needed for it to be detectable, its detection

497

limit is higher compared to DSBP due to its lower signal response. Furthermore the

498

detection limit of DAS1 increases more quickly with NOM background concentrations

499

than DSBP, making it again harder to detect in organic rich waters. The prediction

500

equations established in this paper now give the opportunity to estimate detection

501

limits and to assess the feasibility of using FWC as tracer in certain catchments. For

502

such a suitability assessment it would be recommended to follow a similar approach as

503

outlined in section 2.4 and include any information available. For this purpose a site

504

characterization might be needed to identify dilution, typical NOM background

505

fluorescence and detergent types in a particular region. For comparison, equivalent

506

total organic carbon (TOC) concentrations for the NOM standards are given in Table 3.

507

However, for the estimation of the FWC detection limit (Figure 6) it is recommended

508

to use the waters background fluorescence at λex = 350 nm and λem = 436 nm as a

509

measure for the equivalent NOM standard concentration. Depending on the

510

composition of the organic matter the fluorescence characteristics can change quite

511

significantly. Figure 1 demonstrates that the highest interference is expected from

512

humic- and fulvic-like organic matter. While protein-like organic matter for example

513

can contribute significantly to a higher TOC, its excitation/emission maxima at 220-

514

280/305 nm (tyrosine) and 220-280/350 nm (tryptophane) (Baker et al. 2004) mean

AC C

EP

TE D

M AN U

SC

RI PT

491

22

ACCEPTED MANUSCRIPT 515

that no additional interference of these compounds is expected for the FWC

516

measurement. So if the organic matter composition is significantly different to the

517

NOM standard the amount of TOC does not necessarily reflect the interference with

518

this fluorometric method and should not be used to predict FWC detection limits.

RI PT

519

With detection limits being up to 59% lower the new photodecay ratio (% signal

521

reduction after 1 min / % signal reduction after 10 min) performed significantly better

522

than the ratio suggested previously by Cao et al. (2009). Hence, for the analysis of

523

environmental samples taken during this study this new UV exposure ratio was used to

524

describe the photodecay rates in order to distinguish between fluorescence signals

525

originating from NOM and FWCs. Combining all results from this study the approach

526

developed by Cao et al. (2009) was modified and a new threshold of 0.25 is proposed,

527

above which the fluorometric analysis indicates the presence of FWCs (Figure 8). This

528

threshold was selected because individual measurements of all reference NOM

529

samples that were known not to contain any FWCs rarely exceeded this value and the

530

average measurements were usually well below it. But while the threshold was chosen

531

to be high enough to avoid false positives at the same time it was not set too high so

532

that samples spiked with small concentrations of FWCs would still be detected. It

533

should be noted that a minimum fluorescence signal is needed for the photodecay

534

method to be conclusive. At lower fluorescence signals the variability in the signal due

535

to instrumental factors (e.g. light source, temperature) may falsely suggest a high %

536

reduction of the signal, thereby causing incorrect conclusions to be made. From the

537

photodecay analysis of very low FWC concentrations (results not shown in this paper)

538

it appeared that about 3 times the LOD signal intensity is needed, which equates to a

AC C

EP

TE D

M AN U

SC

520

23

ACCEPTED MANUSCRIPT 539

DSBP concentration of 200 ng/L, for the method to be useful. Water samples with

540

fluorescence signals below this equivalent concentration cannot be further

541

investigated for FWC presence using the photodecay method.

542 Based on the emission spectra and photodecay curves recorded for the analysed

544

detergents it appears that DSBP is predominantly used as optical brightener in liquid

545

laundry detergents while DAS1 is used mainly in powdered detergents.

546

Environmentally friendly products and those specifically marketed for coloured

547

laundry do not contain any or only small amounts of FWCs. The different use of FWCs

548

in washing detergents together with the different detection limits (higher for DAS1)

549

makes it harder for the application in detecting pollution from single houses. For

550

instance, depending on which detergent the household is using, may mean that FWCs

551

will either not be present or not be detectable in the water matrix making it

552

impossible to detect pollution from domestic wastewater with this method. This

553

highlights an important limitation of this method, also mentioned by Cao et al. (2009),

554

and similar techniques (e.g. pharmaceuticals as indicators) for single house application

555

as opposed to urban wastewater treatment plant effluents (or combined sewer

556

overflows) which receives sewage from many households, thereby inherently

557

increasing the chances for the presence of such compounds.

SC

M AN U

TE D

EP

AC C

558

RI PT

543

559

Based on the FWC concentration estimation and derived detection limits, the

560

minimum proportion that septic tank effluent needs to contribute to the overall

561

stream flow for FWCs to be theoretically detectable can be determined, as shown in

562

Table 3. The FWC concentration estimation in surface waters in the two study 24

ACCEPTED MANUSCRIPT catchments demonstrated that the high dilution in watercourses should significantly

564

impact detectability. Considering the FWC detection limits based on the organic

565

matter background fluorescence observed in the catchments, it should, however, be

566

possible to detect FWC at low flow conditions where septic tank effluents contribute

567

significantly to the river flow. These findings agree with previous studies in other

568

areas. Hagedorn et al. (2005a) observed that most fluorescent plumes in tidal areas,

569

e.g. from malfunctioning septic tank percolation areas, were only a few square meters

570

in size and only detectable during an ebb tide. Also Hartel et al. (2007a) recommended

571

that sampling should be carried out during baseflow conditions in order to ensure

572

optimal detection. With that they raised the concern that for systems that are only

573

failing during storm flow conditions, contamination cannot be identified by this

574

method. Dickerson et al. (2007) also experienced problems with detectability based on

575

dilution at beaches in the US but when samples were taken close to potential point

576

sources, FWCs were detected successfully.

577

It should be noted that the theoretical assessment of the likely success of the method

578

does not account for any removal en-route to the water course, which would further

579

lower river water concentrations and detectability. This is reflected in the generally

580

low detection frequency during field sampling. However, in this study FWCs were

581

successfully detected in surface water samples even at times or locations where it was

582

considered impossible or unlikely based on the theoretical assessment. For example

583

several samples taken in May 2015 at the upstream site in catchment #2 tested

584

positive for FWCs even though concentrations for DSPB and DAS1 were estimated to

585

be just above or well below the detection limit respectively. The photodecay ratios

586

from the autosampler samples were ≥0.6 for some samples (Figure 7) which would

AC C

EP

TE D

M AN U

SC

RI PT

563

25

ACCEPTED MANUSCRIPT indicate the presence for DAS1 rather than DSBP which only ever achieves a ratio of

588

around 0.5. But the estimation of FWC concentrations suggests that a detection of

589

DAS1 would not be possible. These results suggest the occurrence of clearly defined

590

discharge events where effluent with elevated concentrations, possibly due to the use

591

of the washing machine, leads to an increased detectability. Even though no obvious

592

signs of an illegal direct effluent discharge were spotted around the sampling point,

593

the clear discharge pattern (Figure 7) would lead to the assumption that the site is

594

impacted by a specific DWWTS. This further highlights the importance of continuous

595

sequenced sampling in order to capture the temporal variability and pulses of

596

increased FWC concentrations, which might be missed during spot sampling. The

597

importance of a suitable choice of sampling regime was also highlighted by Hartel et

598

al. (2007b). Since fluorometric signals are likely to be diluted with increasing distance

599

from the source they recommend targeted sampling, i.e. multiple sampling over ever-

600

decreasing distances, to identify hotspots of high fluorescence/contamination.

601

Although no impact was expected at the upstream site in catchment #1 one sample

602

taken in May 2015 tested positive for FWCs. During the time of sampling filming

603

activities took place right around the site which might have had an impact. In previous

604

studies a range of products including car care products and car fluids (oils, brake fluids

605

etc.) (Hartel et al. 2007a) as well as natural sources such as algae (Hagedorn et al.

606

2005a)

607

photodecay/fluorescence method for FWCs. Most of them (except a car wash soap)

608

tested negative, however the list of tested products/compounds is not exhaustive so

609

there may be other compounds that interfere with fluorometry that were not tested

610

yet. Although not tested in Europe yet, it is also reported that some dishwashing

AC C

EP

TE D

M AN U

SC

RI PT

587

have

been

tested

for

any

26

possible

interference

with

the

ACCEPTED MANUSCRIPT 611

detergents can contain FWCs (Hagedorn et al. 2005b). During those days of filming at

612

this study site catering trucks were parked in close distance upstream of the sampling

613

point so it is considered likely that they could have caused a contamination with

614

optical brighteners or other interfering compounds.

RI PT

615

FWCs appear to be almost non-biodegradable with no biodegradation being observed

617

in aerobic biological wastewater treatment systems or during anaerobic sludge

618

digestion (Poiger et al. 1998). The primary removal processes for FWCs during sewage

619

treatment and after release to the aquatic environment are adsorption/sedimentation

620

and photochemical degradation/photolysis (Kramer et al. 1996, Poiger et al. 1998,

621

Stoll et al. 1998). An aspect that has not been studied in much detail before is the

622

removal of FWCs en-route to the river via adsorption to soil. Samples were taken from

623

the percolation area of DWWTS to verify that FWCs are still detectable after

624

percolation through soil. The results demonstrated a general detectability during both

625

sampling events even though it was lower in March. This was due to dilution from

626

higher rainfall in the days preceding the sampling in March (33.2 mm over 7 days)

627

compared to February (9.2 mm over 7 days). Again this highlights the impact of

628

dilution and importance to sample during dry weather periods in order to improve

629

detectability. However, with an average 1/10 min photodecay ratio of 0.43 ± 0.04,

630

FWCs were still easily detectable in a depth of 40 cm underneath the percolation

631

trench during the sampling in February. Even though some removal was evident, a

632

detailed study and possibly a quantitative method using HPLC, is needed to quantify

633

this removal. However, from the observed removal, it is suspected that the flow

634

pathway in the studied catchments must be more direct into the river for some of the

AC C

EP

TE D

M AN U

SC

616

27

ACCEPTED MANUSCRIPT septic tanks in order to obtain the observed signals from FWCs in the surface water.

636

This is somehow expected due to the low permeability of the subsoils in those

637

catchments where preferential pathways and surface water run-off (Keegan et al.

638

2014) as well as illegal modifications with direct discharges to surface water have been

639

frequently observed.

RI PT

635

640

Apart from the limitations in detectability brought about by detergent use, dilution,

642

removal in soil and high NOM concentration in river water, the applicability of this

643

method is further dependent on the river’s exposure to sunlight which determines the

644

extent of photofading effects and photolysis. In a small river catchment in Japan,

645

Hayakawa et al. (2007) observed changes in DSBP and DAS1 concentrations during a

646

24 hour survey even though water flow rates during those periods did not vary

647

significantly. They concluded that this change was caused by photodegradation in

648

daytime. In a more detailed study on the photochemical degradation of FWCs, Kramer

649

et al. (1996) reported that photolysis of DSBP is three times faster than for DAS1 but in

650

general it is a slower process than photofading. Half-lives for DSBP were 1.5 hours for

651

DSBP near water surface under summer noon sunlight and 4-5 hours for DAS1.

652

However, it was acknowledged that effective half-lives in the environment will be

653

longer due to variable sunlight intensity and screening of sunlight in deeper water

654

layers. They further found that dissolved NOM partly inhibits and slows down this

655

degradation process for DSBP (Kramer et al. 1996). Yamaji et al. (2010) investigated

656

the photodegradation of DSBP and DAS1 in Lake Biwa, the largest lake in Japan, and

657

found that factors such as longer residence times and a greater surface area promotes

658

photodegradation in lacustrine environments. Hence, aspects such as shading from

AC C

EP

TE D

M AN U

SC

641

28

ACCEPTED MANUSCRIPT 659

vegetation, different solar intensities across the year, exposure times and surface

660

areas also need to be considered when assessing the results from such a FWC

661

assessment.

662 Generally all of the monitoring sites where FWCs were detected also stood out during

664

monitoring of other parameters for domestic wastewater e.g. with high E. coli and

665

ammonium concentrations (data not shown here). Due to the discussed issues with

666

low detectability for DAS1 or in waters with high organic content, the non-detection of

667

FWC cannot be taken as a definite exclusion of a possible contamination with human

668

wastewater. A positive detection however will almost certainly proof that a site is

669

impacted by a malfunctioning septic tank or indicate the presence of an illegal direct

670

discharge to the river.

M AN U

SC

RI PT

663



674 675

powder based and DSBP exclusively in liquid based detergents. •

676

The signal response and photodecay rate for the two FWCs are significantly different, which affects especially the detectability of DSBP using existing

677 678

From the two FWCs used in laundry detergents, DAS1 was mainly used in

EP

673

5. Conclusions

AC C

672

TE D

671

fluorometry photodecay methods.



A modified photodecay method has been developed for the detection of FWCs

679

in surface water samples that increases sensitivity of the technique by up to

680

59% compared to existing methods.

681 682



Detection limits of the photodecay method increases with NOM content of the water. DAS1 is harder to detect at high organic matter content than DSBP. 29

ACCEPTED MANUSCRIPT 683



The establishment of equations to predict detection limits now allows to assess

684

the method’s suitability for individual catchments based on the organic matter

685

content. •

687 688

For a successful application in rural catchments it is recommended to sample

RI PT

686

during low flow/dry weather conditions to increase detectability. •

The fluorometry photodecay method should be especially effective in areas where an illegal direct discharge from a DWWTS is suspected. However,

690

discrete-time sampling over a few days as opposed to single spot sampling

691

would be advisable here to capture discharge events. •

693 694

M AN U

692

SC

689

FWCs experience some removal in the percolation area of DWWTSs but were still detectable in deeper soil layers en-route to receiving water bodies.



Due to limitations the suitability of application of FWCs need to be considered on a case by case basis and sampling times and location carefully considered

696

for individual catchments/application to optimise detectability.

697

TE D

695

Acknowledgements

699

This research was funded by the Irish Research Council [GOIPD/2014/192]. Thanks to

700

Manuel Ruether, Department of Chemistry, Trinity College Dublin, for making the

701

fluorometer available to us. Many thanks also to Laura Brophy and Jan Knappe,

702

Department of Civil Structural and Environmental Engineering, Trinity College Dublin,

703

for their help in getting field samples and for providing flow and rainfall data. The

704

authors would also like to acknowledge Prof. Phillip Geary from the University of

705

Newcastle, Australia, for his advice and the Irish Environmental Protection Agency for

706

additional financial support [Project 2012-W-MS-12].

AC C

EP

698

30

ACCEPTED MANUSCRIPT 707 References

709

Baker, A., Ward, D., Lieten, S.H., Periera, R., Simpson, E.C. and Slater, M. (2004)

710

Measurement of protein-like fluorescence in river and waste water using a handheld

711

spectrophotometer. Water Research 38(12), 2934-2938.

712

Cao, Y.P., Griffith, J.F. and Weisberg, S.B. (2009) Evaluation of optical brightener

713

photodecay characteristics for detection of human fecal contamination. Water

714

Research 43(8), 2273-2279.

715

CSO, (2016) Census of Population 2016 - Profile 1 Housing in Ireland, Central Statistics

716

Office, Government of Ireland, Stationery Office, last accessed: 10 May 2017,

717

http://www.cso.ie/en/census/interactivetables/.

718

Dickerson, J.W., Hagedorn, C. and Hassall, A. (2007) Detection and remediation of

719

human-origin pollution at two public beaches in Virginia using multiple source tracking

720

methods. Water Research 41(16), 3758-3770.

721

Dubber, D. and Gill, L. (2014) Application of On-Site Wastewater Treatment in Ireland

722

and Perspectives on Its Sustainability. Sustainability 6 (3), 1623-1642.

723

Hagedorn, C., Saluta, M.A., Hassall, A. and Dickerson, J.W., (2005a) Fluorometric

724

Detection of Optical Brighteners as an Indicator of Human Sources of Water Pollution.

725

Part II. Development as a Source Tracking Methodology in Open Waters, Crop and Soil

726

Environmental

727

http://www.sites.ext.vt.edu/newsletter-archive/cses/2005-11/part2.html.

AC C

EP

TE D

M AN U

SC

RI PT

708

News,

last

accessed:

31

18

May

2017,

ACCEPTED MANUSCRIPT 728

Hagedorn, C., Saluta, M.A., Hassall, A. and Dickerson, J.W., (2005b) Fluorometric

729

Detection of Optical Brighteners as Indicator of Human Sources of Water Pollution.

730

Part I. Description and Detection of Optical Brighteners, Crop and Soil Environmental

731

News,

732

archive/cses/2005-11/part1.html.

733

Hartel, P.G., Hagedorn, C., McDonald, J.L., Fisher, J.A., Saluta, M.A., Dickerson, J.W.,

734

Gentit, L.C., Smith, S.L., Mantriprayada, N.S., Ritter, K.J. and Belcher, C.N. (2007a)

735

Exposing water samples to ultraviolet light improves fluorometry for detecting human

736

fecal contamination. Water Research 41(16), 3629-3642.

737

Hartel, P.G., McDonald, J.L., Gentit, L.C., Hemmings, S.N.J., Rodgers, K., Smith, K.A.,

738

Belcher, C.N., Kuntz, R.L., Rrvera-Torres, Y., Otero, E. and Schroder, E.C. (2007b)

739

Improving fluorometry as a source tracking method to detect human fecal

740

contamination. Estuaries and Coasts 30(3), 551-561.

741

Hayakawa, K., Okumura, R., Yamamoto, H., Fujiwara, M., Yamaji, N., Takada, H.,

742

Kanematsu, M. and Shimizu, Y. (2007) Distribution and fluxes of fluorescent whitening

743

agents discharged from domestic wastewater into small rivers with seasonal changes

744

of flow rates. Limnology 8(3), 251-259.

745

Keegan, M., Kilroy, K., Nolan, D., Dubber, D., Johnston, P., Misstear, B., O’Flaherty, V.,

746

Barrett, M. and Gill, L.W. (2014) Assessment of the Impact of Traditional Septic Tank

747

Soakaway Systems on Water Quality in Ireland. Water Science and Technology 70(4),

748

634 - 641.

May 2017,

http://www.sites.ext.vt.edu/newsletter-

RI PT

accessed: 18

AC C

EP

TE D

M AN U

SC

last

32

ACCEPTED MANUSCRIPT Kramer, J.B., Canonica, S., Hoigné, J. and Kaschig, J. (1996) Degradation of Fluorescent

750

Whitening Agents in Sunlit Natural Waters. Environmental Science & Technology 30(7),

751

2227-2234.

752

McKnight, D.M., Boyer, E.W., Westerhoff, P.K., Doran, P.T., Kulbe, T. and Andersen,

753

D.T. (2001) Spectrofluorometric characterization of dissolved organic matter for

754

indication of precursor organic material and aromaticity. Limnology and Oceanography

755

46(1), 38-48.

756

Poiger, T., Field, J.A., Field, T.M. and Giger, W. (1996) Occurrence of fluorescent

757

whitening agents in sewage and river water determined by solid-phase extraction and

758

high-performance liquid chromatography. Environmental Science & Technology 30(7),

759

2220-2226.

760

Poiger, T., Field, J.A., Field, T.M., Siegrist, H. and Giger, W. (1998) Behavior of

761

fluorescent whitening agents during sewage treatment. Water Research 32(6), 1939-

762

1947.

763

Poiger, T., Kari, F.G. and Giger, W. (1999) Fate of fluorescent whitening agents in the

764

River Glatt. Environmental Science & Technology 33(4), 533-539.

765

Scott, T.M., Rose, J.B., Jenkins, T.M., Farrah, S.R. and Lukasik, J. (2002) Microbial

766

source tracking: Current methodology and future directions. Applied and

767

Environmental Microbiology 68(12), 5796-5803.

AC C

EP

TE D

M AN U

SC

RI PT

749

33

ACCEPTED MANUSCRIPT Stoll, J.M.A. and Giger, W. (1997) Determination of detergent-derived fluorescent

769

whitening agent isomers in lake sediments and surface waters by liquid

770

chromatography. Analytical Chemistry 69(13), 2594-2599.

771

Stoll, J.M.A., Ulrich, M.M. and Giger, W. (1998) Dynamic behavior of fluorescent

772

whitening agents in Greifensee: Field measurements combined with mathematical

773

modeling of sedimentation and photolysis. Environmental Science & Technology

774

32(13), 1875-1881.

775

Yamaji, N., Hayakawa, K. and Takada, H. (2010) Role of Photodegradation in the Fate

776

of Fluorescent Whitening Agents (FWAs) in Lacustrine Environments. Environmental

777

Science & Technology 44(20), 7796-7801.

M AN U

SC

RI PT

768

AC C

EP

TE D

778

34

ACCEPTED MANUSCRIPT Table 1: Detection limits for FWCs at different concentrations of natural organic matter

NOM concentration 4 mg/L 8 mg/L 16 mg/L 23 mg/L 30 mg/L

Photodecay ratio 1/10 min 0.42 0.71 0.99 1.22 1.69

Photodecay ratio 10/5 min * 1.01 1.36 1.5 2.22 2.56

Photodecay ratio 1/10 min 0.59 1.66 3.03 6.48 11.33

Photodecay ratio 10/5 min * 1.19 2.46 4.74 7.80 14.20

AC C

EP

TE D

M AN U

SC

* ratio recommended by Cao et al. (2009)

DAS1 detection limit [µg/L]

RI PT

DSBP detection limit [µg/L]

ACCEPTED MANUSCRIPT Table 2: Estimation of FWC concentrations and comparison to their detection limits in river water of the study catchments. Estimated concentrations that are higher than the expected detection limit are highlighted in bold.

June 2015

April 2016

July 2016

1.03

1.18

1.70

0.00

1.96 1.80 0.00 0.49 3.24 2.04 0.00

0.92

1.06

2.62

0.00

3.01 3.63 0.00

1.96

2.25

1.57

0.00

SC

0.0 1.0 1.2 2.4 0.0 1.5 2.5 2.3 0.0 0.6 4.2 2.6 0.0 1.4 3.9 4.7 0.0 2.9 1.9 6.9 0.0 5.5 2.6 14.8

Estimated FWC levels in river [μg/L] DSBP DAS1 0.00 0.00 0.70 0.80 0.79 0.91 1.63 1.88 0.00 0.00

FWC detection limit [μg/L] DSBP 0.42 0.40 1.38 1.57 0.42 0.39 1.79 1.39 0.40 0.39 1.66 1.58 0.37 0.39 1.56 1.29 0.38 0.36 1.04 1.14 0.32 0.36 0.95 1.21

DAS1 0.87 0.87 7.90 11.02 0.87 0.87 15.14 8.08 0.87 0.87 12.55 11.17 0.87 0.87 10.86 6.70 0.87 0.89 3.88 4.84 0.93 0.88 3.03 5.66

RI PT

DWWTS contribution on stream flow [%]

0.43 2.82 1.77

M AN U

May 2015

672,048 2,114,559 390,920 1,069,527 554,179 1,439,760 180,924 1,115,345 495,000 3,447,360 109,454 986,618 1,242,807 1,610,159 117,772 553,527 378,549 756,657 244,432 374,518 100,060 398,494 172,800 174,864

TE D

April 2015

Cat#1 upper Cat#1 lower Cat#2 upper Cat#2 lower Cat#1 upper Cat#1 lower Cat#2 upper Cat#2 lower Cat#1 upper Cat#1 lower Cat#2 upper Cat#2 lower Cat#1 upper Cat#1 lower Cat#2 upper Cat#2 lower Cat#1 upper Cat#1 lower Cat#2 upper Cat#2 lower Cat#1 upper Cat#1 lower Cat#2 upper Cat#2 lower

Daily DWWTS output in upstream catchment [L/d] 0 21,881 4,559 25,832 0 21,881 4,559 25,832 0 21,881 4,559 25,832 0 21,881 4,559 25,832 0 21,881 4,559 25,832 0 21,881 4,559 25,832

EP

March 2015

Site

AC C

Date

Mean Daily flow [L/d]

3.16

1.26

1.45

4.67

5.36

0.00

0.00

3.72

4.27

1.78

2.05

10.00

11.48

ACCEPTED MANUSCRIPT Table 3: Septic tank effluent contribution to stream flow that is needed to detect FWCs in waters with different organic matter content.

Equivalent TOC concentration† [mg/L] 2.4 4.2 8.0 11.1 14.25

RI PT

NOM concentration* in water matrix Very low (4 mg/L) Low (8 mg/L) Medium (16 mg/L) High (23 mg/L) Very high (30 mg/L)

DWWTS contribution to stream flow in order to detect traces from DSBP based DAS1 based detergents detergents >0.6% >0.9% >1.1% >3.3% >1.5% >5.9% >1.8% >9.6% >2.5% >16.8%

AC C

EP

TE D

M AN U

SC

* based on fluorescence intensity at ʎex=350 nm, ʎem=436 nm in comparison to Suwannee River NOM standard † measured for Suwannee River NOM standard solutions using high temperature combustion method (TOC-L, Shimadzu)

ACCEPTED MANUSCRIPT Table 4: Results from the photodecay analysis of grab samples taken within the two study catchments. Only samples for average photodecay rate > 0.25 are shown. Date

Location

Average photodecay ratio 1/10 min

p-value

Replicates >0.25 out of total analysed

Midstream Downstream Midstream Upstream Midstream Midstream Downstream Midstream Midstream Upstream Midstream Downstream

0.274 0.255 0.293 0.779 0.284 0.324 0.277 0.372 0.288 0.28 0.589 0.308

0.02* 0.789 0.765 <0.01** 0.651 0.073 0.645 0.039* 0.126 0.375 <0.01** 0.079

3/3 1/3 2/3 3/3 2/3 3/3 2/3 3/3 5/5 3/5 3/3 5/5

Upstream Midstream Upstream Side stream Upstream Upstream

0.624 0.472 0.808 0.488 0.288 0.266

<0.01** <0.01** <0.01** <0.01** 0.652 0.176

April 2016 July 2016

Catchment #2 April 2015 May 2015 June 2015 July 2016

AC C

EP

TE D

* indicates statistical significance at 0.05 level ** indicates statistical significance at 0.01 level

SC

April 2015 May 2015 June 2015

M AN U

March 2015

RI PT

Catchment #1

3/3 3/3 3/3 3/3 2/3 3/3

ACCEPTED MANUSCRIPT Table 6: Results from the FWC photodecay analysis of effluent and soil moisture samples taken from two percolation area sites receiving primary effluent (PE) and secondary treated effluent (SE). Only samples for average photodecay rate > 0.25 are shown.

Feb 2017

#1 #1 #1 #2 #2 #2 #2

March 2017

Soil moisture samples Average photodecay p-value ratio 1/10 min 0.43 0.02* 0.49 <0.01** 0.35 0.134 0.45 0.02* 0.26 0.526 0.28 0.155 0.36 0.04*

AC C

EP

TE D

M AN U

* indicates statistical significance at 0.05 level ** indicates statistical significance at 0.01 level

Replicates >0.25 out of total analysed 3/3 3/3 3/3 3/3 2/3 3/3 3/3

RI PT

Site

Sampling depth [below trench] 40 cm 20 cm 35 cm 5 cm 10 cm 10 cm 5 cm

SC

Date

Receiving effluent Average Type photodecay ratio 1/10 min SE 0.52 SE 0.52 SE 0.52 PE 0.46 PE 0.37 PE 0.37 PE 0.37

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

AC C

EP

TE D

Figure 1: Emission spectra (ʎex = 350 nm) of FWC standards and organic matter reference material (IHSS NOM standard, humic acid sodium salt and pristine river water).

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

AC C

EP

TE D

Figure 2: Emission spectra (ʎex = 350 nm) of FWC standards, detergents and septic tank effluent.

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

AC C

EP

TE D

Figure 3: Fluorescence photodecay curves for FWCs (n=7), natural organic matter (n=8), pristine river water (n=3), powder based (n=13) and liquid detergents (n=6). Error bars represent 95% confidence intervals.

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

AC C

EP

TE D

Figure 4: Photodecay signal reduction ratio (1/10 min) for NOM at various concentrations spiked with different FWC (DSBP) concentrations. A ratio > 0.25 was chosen to indicate the presence of FWCs. Error bars represent 95% confidence intervals.

ACCEPTED MANUSCRIPT

SC

RI PT

a)

EP

TE D

M AN U

b)

AC C

Figure 5: Regression analysis for photodecay signal reduction ratio a) 1/10 min and b) 10/5 min recorded for NOM 4 mg/L solutions spiked with DSBP. Dashed lines are the regressions 95% PIs and arrows demonstrate the determination of detection limits. Resulting detection limits from all NOM concentrations and for DAS1 are listed in Table 1.

ACCEPTED MANUSCRIPT

SC

RI PT

a)

EP

TE D

M AN U

b)

AC C

Figure 6: Detection limit of a) DSBP and b) DAS1 depending on the samples background NOM concentration.

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

AC C

EP

TE D

Figure 7: Signal reduction ratio 1/10 min from the photodecay analysis for a week long sampling event at the upstream site in Catchment #2 during low flow conditions in May 2015. Observations greater than the threshold of 0.25 tested positive for the presence of FWCs. Error bars represent 95% confidence intervals.

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

AC C

EP

TE D

Figure 8: Suggested modified photodecay method for the detection of FWCs in surface water samples (modified after Cao et al. (2009))

ACCEPTED MANUSCRIPT Highlights

EP

TE D

M AN U

SC

RI PT

The fluorometric FWC detection is only possible with a presence/absence approach The sensitivity of an existing photodecay method has been improved by up to 59% Detection limits for FWCs increase with NOM content of water samples An approach is presented to assess the method’s suitability on a catchment basis The FWC detectability in streams and rivers increases during low flow conditions

AC C

• • • • •