The aim of this research was to better understand chemical pre-treatment of combined sewer overflows (CSOs) for subsequent ultraviolet (UV) disinfection. Approximately 200 jar tests were completed. Alum (Al2(S04)3·12H2O) resulted in a higher UV light transmission (UVT), and equivalent total suspended solids (TSS) removal, than ferric chloride (FeCl3). An alum dose of 20 mg/L increased the UVT of the raw CSO from 30 to 60% after settling. The addition of 100 mg/L of alum maximized UVT reaching approximately 85%. Flocculation did not increase UVT. However, it did improve the removal of TSS. Cationic polymers worked quickly compared with metal coagulants, but only reached a UVT of 60%. A high positive charge density on the polymer improved the removal of turbidity when compared with low charge, but did not affect UVT. If the goal is to maximise UVT, a very high alum dose may be preferred. If the goal is to minimize coagulant dose with moderate UV performance, cationic polymer at approximately 3 mg/L is recommended.

INTRODUCTION

Combined sewers carry both sewage and storm water, and are common in the older parts of most cities. Combined sewer overflows (CSOs) occur when the volume of water in the sewer exceeds the capacity, often during spring snow melt or heavy rain. In order to prevent basement flooding, sewers are designed to overflow. The release of CSOs introduces pathogens, such as Escherichia coli and viruses, to surface waters, increasing the risk of gastrointestinal illnesses (Pruss 1998; Donovan et al. 2007; Rodríguez et al. 2012). Accounting for intermittent increases in pathogens resulting from CSOs is an essential and often neglected aspect of Quantitative Microbial Risk Assessment (QMRA) used to evaluate drinking water safety (Dickenson & Sansalone 2012; Mahajan et al. 2013). In addition to adverse environmental and public health effects, the resulting closure of recreational waters, usually due to high E. coli counts, is associated with economic losses from reduced tourism (Chhetri et al. 2014). As such, a means to inactivate the pathogens in CSOs can reduce risks to public health and, in some cases, improve economic activity.

Chlorine has been used for CSO disinfection in the past. However, the formation of toxic disinfection by-products resulting from chlorination has prompted the search for less toxic alternatives, such as peracetic acid (Chhetri et al. 2014). The use of UV light for disinfection has a number of advantages including no disinfection by-products, no pH dependence, and short contact time suitable for high flow rates (Wojtenko et al. 2001). Moreover, the oocyte forming pathogen Cryptosporidium is readily inactivated by UV light, but is resistant to chlorine (Clancy et al. 1998). In a field survey in Michigan, Cryptosporidium was detected in 56% of surface water samples impacted by CSOs (Dreelin et al. 2014). Unlike chlorine, UV does not form by-products, is minimally affected by water chemistry, and has demonstrated efficacy against Cryptosporidium.

In the applications where UV disinfection is traditionally used, such as drinking water and secondary wastewater treatment, the water tends to have low total suspended solids (TSS < 15 mg/L) and low biochemical oxygen demand (BOD < 15 mg/L). In contrast, the TSS and BOD of CSOs can exceed 200 mg/L (Gasperi et al. 2012). The pollution load of CSOs tends to be highly variable, however, sometimes decreasing precipitously after a first flush (Suárez & Puertas 2005). The high concentration of particulates and organic matter in the water matrix makes the use of UV more difficult, requiring some level of pretreatment.

The UV transmission (UVT) is the fraction of light with a wavelength of 254 nm transmitted through one centimeter of water. A wavelength of 254 nm is the most effective wavelength for disinfection and is the main output of low-pressure mercury lamps (Masschelein 2002). The UVT of raw CSOs is typically about 30%, which means only about 9% of the original disinfecting power is transmitted through 2 cm of the CSO sample. In comparison, the UVT of a secondary treated effluent is typically 70%. In this case, 49% of the UV light is transmitted 2 cm, a fivefold improvement. As such, the UVT dramatically affects the size and cost of UVT equipment. The goal of this research is to improve the UVT of CSOs using a variety of chemical pretreatments.

Traditionally, the optimum coagulant dose has been determined using standard jar tests based on TSS reduction. However, the standard TSS test only counts material larger than 0.45 microns in diameter. Smaller particles, such as dissolved organics and colloids, are also absorbers of UV (Thomas 2007). For this reason, UVT is considered independently of TSS in this study.

Typical CSO control facilities are simple rectangular basins (Brombach et al. 2005). One of the original hypotheses of this work is that including flocculation may increase the UVT of the treated water. As such, the effect of increasing flocculation on UVT was examined in this work.

A full-factorial experimental design was used to study the effect of coagulant selection, dose, and flocculation time on UVT, TSS and turbidity of CSOs. The use of full-factorial design allows the detection of interactions among input variables, for example the effects of dose and flocculation time on UVT, though few were observed (Box et al. 2005). Each experiment was repeated at least four times and blocking was used to identify consistent performance trends. This is important since CSO events happen suddenly and there is often not time for additional testing as they progress.

MATERIALS AND METHODS

The following provides the details of the water sampling, as well as statistical and analytical methods used in this study.

CSO samples

Two sets of samples were used in this study: (A) actual CSO samples and (B) simulated CSO samples prepared using a 50/50 mixture of raw wastewater and tap water (Zukovs et al. 1986). In the first season (series A), it was possible to collect actual CSO samples for testing with metal coagulants. In the second season (series B) it was no longer possible to collect actual CSO samples due to construction at the CSO facility, and simulated CSOs were used to test cationic polymers. In the second season, a jar test with alum was conducted as reference for comparison purposes.

Sample collection

Actual CSO samples (A) were collected during five high rainfall events from April to October 2012. The test site was adjacent to the North Toronto CSO facility. A Manning vacuum sampler (Georgetown, TX) was triggered by high water level using a Hach Sigma 900 controller (Loveland, CO). The sampler collected 500 mL every 5 min until the 20 L sample container was full, requiring approximately 200 minutes. Samples were jar-tested the same day.

For simulated CSOs (B), samples of raw wastewater were collected from the inlet channel of Ashbridges Bay wastewater treatment plant in Toronto, ON. The four samples were collected approximately once a month from June to September 2013. Raw wastewater samples were transferred to a 20 L container and were diluted 50/50 with tap water and thoroughly mixed. All tests were completed the same day.

Jar testing

Jar tests were conducted using a standard laboratory jar tester equipped with six cells (Phipps and Bird, Richmond, VA). One liter of raw CSO was added to each jar; the coagulant was added and then rapidly mixed for 1 min at 150 rpm (velocity gradient, G = 140 s−1). Flocculation was conducted at 30 rpm (G = 20 s−1) for a duration of 0 to 5 min. Settling time was constant at 10 min.

Many jar tests use an hour for settling time (e.g. Exall & Marsalek 2013). Our relatively short settling time was selected to simulate the shorter detention time associated with CSO events, consistent with other CSO work by Li et al. (2003). After settling, a sample was decanted from each jar and the UVT, turbidity and TSS were measured immediately.

Coagulants were added in liquid form at the start of the rapid mixing phase. Metal coagulants, alum (Al2(S04)3·12H2O) and anhydrous ferric chloride (FeCl3), were dissolved in deionized water to create 1% stock solutions. As a result of previous work that showed degradation of coagulant over time (Casale 2001), stock solutions were prepared fresh 1 week before each test.

The efficacy of high molecular weight cationic polymers to improve the removal of TSS at this site has been demonstrated previously (Marsalek et al. 2005). The cationic powder (NF Inc., Riceboro, GA) was dissolved in deionized water to 0.1% 1 h prior to the test, according to the supplier's instructions. The low charge density cationic polymer (FO4190 SH) had an approximate molecular weight of 6 × 106 Da and 10% cationicity per mole. The high charge density polymer (FO4650 SH) had a similar molecular weight, but 55% cationicity per mole, according to the supplier.

Design of experiments (DOE)

Table 1 shows an overview of the experimental design. Commercial software (Design Expert 8.0, Stat-Ease, Minneapolis, MN) was used for the design of experiments and data analysis. The input variables used in each series in this study were: coagulant type (two choices), coagulant concentration (three choices), and flocculation time (three choices). Each set of tests was repeated five times for actual CSOs (metal coagulants), and four times for simulated CSOs (polymers). Factors reaching the 99% confidence level based on an analysis of variance (ANOVA) were selected to be included in the final statistical model, which is shown as the surface plot in subsequent graphs.

Table 1

Overview of experimental design. In full factorial designs every possible permutation is tested, requiring 18 tests to complete one sample in a series. Series A tests used actual CSO samples; series B used a mixture of raw wastewater and tap water to simulate CSO, due to limitations on field testing

  Variables
 
Series Coagulants Coagulant concentrations (mg/L) Flocculation times (min) 
Alum (Al2(SO4)3·12H2O) or Ferric chloride (FeCl320 
60 2.5 
120 
Low charge density cationic polymer or High charge density cationic polymer 
2.5 
10 
  Variables
 
Series Coagulants Coagulant concentrations (mg/L) Flocculation times (min) 
Alum (Al2(SO4)3·12H2O) or Ferric chloride (FeCl320 
60 2.5 
120 
Low charge density cationic polymer or High charge density cationic polymer 
2.5 
10 

Analytical

The UVT was determined using a Hitachi U-3900 integrating sphere spectrophotometer (Chromabec, Waterloo, QE) at a wavelength of 254 nm. The TSS was determined according to Method 2540D in Standard Methods for the Examination of Water and Wastewater (2012). Turbidity was measured using a Hach 2100Q turbidity meter (Loveland, CO). Meter calibration at 10 nephelometric turbidity units (NTU) was verified before each set of experiments.

RESULTS AND DISCUSSION

Metal coagulants: alum and ferric chloride

CSOs are a mixture of rainwater, which has little buffering capacity, and wastewater. The low buffering capacity of rainwater can result in pH depression (pH 3.5) when adding hydrolysing coagulants, such as ferric chloride, to CSOs (e.g. Exall & Marsalek 2013). In this study, the average value does not fall below pH 6.7, within the recommended range of pH 5–7 (Figure 1). A possible reason for this is that Lake Ontario is the drinking water source and provides the matrix for wastewater. It is well-buffered with an alkalinity of 90 mg/L as CaCO3 (Chapra et al. 2012).
Figure 1

pH reduction following additions of the coagulants alum (Al2(SO4)3·12H2O) and anhydrous ferric chloride (FeCl3). Error bars represent one standard deviation (n = 54).

Figure 1

pH reduction following additions of the coagulants alum (Al2(SO4)3·12H2O) and anhydrous ferric chloride (FeCl3). Error bars represent one standard deviation (n = 54).

‘Sweep flocculation’ occurs when enough coagulant is added to exceed its solubility. The resulting Al(OH)3 or Fe(OH)3 precipitates collect other suspended material during mixing for removal by sedimentation. In the case of alum, a pH decrease from 7.4 to 6.7 is expected to decrease solubility, causing slightly more Al(OH)3 to precipitate. In the case of ferric chloride, this pH decrease causes a very slight increase in solubility (Stumm & O'Melia 1968). In either case, the high concentration of coagulant (i.e. 20–120 mg/L) suggests that sweep flocculation is likely the main removal mechanism.

Five different CSO samples were collected during high rainfall events from April to October 2012. Despite variability in water quality, the observed UVT values after chemical treatment are grouped tightly together (Figure 2). This suggests that chemical pretreatment to improve UVT is insensitive to changes in influent quality, consistent with the sweep flocculation removal mechanism. Chemically enhanced primary treatment of wastewater, a similar application with widely varying source water quality, also produced consistent effluent quality (Neupane et al. 2008).
Figure 2

UVT of treated CSO samples using alum as coagulant. Dark circles represent the actual results from different jar tests. Hatched surface is the statistical model.

Figure 2

UVT of treated CSO samples using alum as coagulant. Dark circles represent the actual results from different jar tests. Hatched surface is the statistical model.

The UVT of the raw CSO samples in these tests was approximately 30%. An alum dose of 20 mg/L resulted in a UVT of approximately 60%, after settling (Figure 2). Adding alum at 20 mg/L approximately doubled UVT. In contrast, anhydrous ferric chloride produced a UVT of only 40% at 20 mg/L.

According to the statistical model, which interpolates between tested values, the maximum UVT of 85% was achieved at an alum dose of 100 mg/L. At higher doses (i.e. 120 mg/L) there was a slight decrease in UVT, possibly due to light scatter by excess precipitates.

The UVT was minimally affected by changes in flocculation time (Figure 2). Increasing the flocculation time from 0 min to 5 min had almost no effect, indicating little benefit from including flocculation to increase UVT.

Increasing flocculation time had a positive effect on TSS removal, however. This finding is indicated by the downward tilt of the response surface with increasing flocculation time (Figure 3). This emphasises the difference between UVT and TSS: a test based on TSS alone would have implied that flocculation improved performance. The ability to increase UVT and decrease TSS is likely the result of slightly different removal mechanisms.
Figure 3

Total suspended solids (TSS) of treated CSO samples using alum as coagulant. Dark circles represent the actual results from different jar tests. Hatched surface is the prediction of the statistical model. (Note that the axes have been reversed to make the response surface visible.)

Figure 3

Total suspended solids (TSS) of treated CSO samples using alum as coagulant. Dark circles represent the actual results from different jar tests. Hatched surface is the prediction of the statistical model. (Note that the axes have been reversed to make the response surface visible.)

Turbidity and TSS are used to quantify the amount of suspended material remaining in the sample after settling. For any given amount of suspended material, the UVT is higher when alum is used (Figure 4). This is likely in part due to the high UV absorbance of ferric ions in solution (Thomas 2007). However, the nature of the suspended matter may also play a role. When added to water at sufficient concentration, the insoluble alum particles, which were white, likely reflect more and absorb less UV than an equal concentration of rust-coloured ferric hydroxide precipitates, consistent with the results presented in Figure 4.
Figure 4

Changes in UV transmission (UVT) as a function of turbidity (a) and total suspended solids (b) of treated CSO samples.

Figure 4

Changes in UV transmission (UVT) as a function of turbidity (a) and total suspended solids (b) of treated CSO samples.

When using alum to increase UVT, there was no evidence of a penalty in terms of TSS reduction. No difference was observed in the ability of alum and anhydrous ferric chloride on a mass basis to reduce TSS for the conditions tested in this study (Table 2).

Table 2

Mean of all total suspended solids (TSS) measurements of treated CSO samples as a function of coagulant (n = 90)

Coagulant Mean TSS (mg/L) 
Alum (Al2(S04)3 · 12H2O) 18.9 
Ferric chloride (FeCl319.0 
ANOVA result (probability > F) 0.31 
Significant No 
Coagulant Mean TSS (mg/L) 
Alum (Al2(S04)3 · 12H2O) 18.9 
Ferric chloride (FeCl319.0 
ANOVA result (probability > F) 0.31 
Significant No 

Cationic polymers

In jar tests, the low and high charge density cationic polymers acted very quickly, producing large, dense, flocs that settled in approximately 1 min. In contrast, the metal coagulants required at least 5 min of settling.

The maximum UVT that could be achieved using cationic polymers based on four different simulated CSO samples did not exceed 65% (Figure 5). To test if this result was due to the source water, a jar test was completed using simulated CSO and alum. According to the statistical model developed using actual CSO samples, an alum dose of 80 mg/L should produce a UVT of approximately 75%. A jar test using simulated CSO at an alum dose of 80 mg/L resulted in a UVT of 78%. As discussed previously the performance of metal coagulants is relatively insensitive to changes in source water quality, in this case actual or simulated CSOs. Taken together, this suggests that although cationic polymers do act quickly, they cannot alone reach the same UVT as high doses of alum.
Figure 5

UVT of treated CSO samples using low charge density cationic polymer as coagulant. Hatched surface is the statistical model.

Figure 5

UVT of treated CSO samples using low charge density cationic polymer as coagulant. Hatched surface is the statistical model.

The relationship between UVT performance and flocculation time using polymers is more complex than for metal coagulants. The UVT increased with increasing flocculation time until approximately 3 min, and then decreased slightly (Figure 5), likely the result of floc disintegration. Meanwhile, increasing polymer dose from 4 to 10 mg/L had little effect on UVT; however, in practice high polymer loading could contribute to lamp fouling.

ANOVA showed that there was no statistically significant difference between the effects of high and low charge density cationic polymers on UVT or TSS (Table 3). However, the high charge density cationic polymer was much more effective at reducing turbidity. This indicates that the main contributors to turbidity are negatively charged colloids, which respond to the higher cationic charge on the polymer. The primary UV absorbers are not likely to be negatively charged.

Table 3

Mean turbidity of all treated CSO samples as a function of cationic polymer charge density (n = 78)

Coagulant Mean UVT (%/cm) Mean TSS (mg/L) Mean turbidity (NTU) 
Low charge density cationic polymer 45.9 23.9 17.3 
High charge density cationic polymer 48.3 22.0 14.4 
ANOVA result (probability > F) 0.31 0.36 <0.001 
Significant No No Yes 
Coagulant Mean UVT (%/cm) Mean TSS (mg/L) Mean turbidity (NTU) 
Low charge density cationic polymer 45.9 23.9 17.3 
High charge density cationic polymer 48.3 22.0 14.4 
ANOVA result (probability > F) 0.31 0.36 <0.001 
Significant No No Yes 

CONCLUSIONS

This study examined chemical pre-treatment of CSOs with the goal of increasing UVT. More than 200 jar tests were completed using actual and simulated CSOs. Statistical models for the effects of alum and cationic polymers on UVT were developed. Such models may preclude the need for jar tests, which are impractical during actual CSOs events. In addition, the following was found:

  • According to the statistical model, alum at 100 mg/L produced the highest UVT, approaching 85% in this study. Lower alum doses (i.e. 20 mg/L) increased the UVT of treated CSO samples from approximately 30 to 60%, after settling.

  • After brief rapid mixing, up to 5 min of additional flocculation did not increase UVT, but did improve TSS removal when using metal coagulants, likely due to different removal mechanisms.

  • Compared with the low charge density polymer, high charge density cationic polymer improved the removal of turbidity, but did not affect UVT and TSS.

  • Compared with metal coagulants, cationic polymers worked faster and produced larger flocs. However, the maximum UVT that could be achieved using cationic polymers (60%) was lower than that achieved using alum at 80 mg/L (78%). If the goal is to maximise UVT, a high alum dose may be preferred. If the goal is to minimise coagulant dose with moderate UV performance, cationic polymer at approximately 3 mg/L is recommended.

ACKNOWLEDGEMENTS

The authors would like to thank Kristin O'Connor for her help in the laboratory and comments on the manuscript. Thanks are also extended to Stephan Lee, Sam Dith and Quintin Rochfort for help with sampling. We gratefully acknowledge the support of the City of Toronto and the financial support from Natural Sciences and Engineering Research Council of Canada (NSERC) Strategic Grant Program.

REFERENCES

REFERENCES
Box
G. E. P.
Hunter
J. S.
Hunter
W. R.
2005
Statistics for Experimenters: Design, Innovation, and Discovery
, 2nd edn.
Wiley-Interscience
,
New York
,
USA
.
Brombach
H.
Weiss
G.
Fuchs
S.
2005
A new database on urban runoff pollution: comparison of separate and combined sewer systems
.
Water Science and Technology
51
(
2
),
119
128
.
Casale
R.
2001
Improving chemical handling procedures can help reduce associated treatment problems
.
Journal of the American Water Works Association
93
(
9
),
95
106
.
Chapra
S. C.
Dove
A.
Warren
G. J.
2012
Long-term trends of great Lakes major ion chemistry
.
Journal of Great Lakes Research
38
,
550
560
.
Chhetri
R. K.
Thornberg
D.
Berner
J.
Gramstadd
R.
Öjstedt
U.
Sharma
A. K.
Andersena
H. R.
2014
Chemical disinfection of combined sewer overflow waters using performic acid or peracetic acids
.
Science of the Total Environment
490
,
1065
1072
.
Clancy
J. L.
Hargy
T. M.
Marshall
M. M.
Dyksen
J. E.
1998
UV light inactivation of Cryptosporidium oocysts
.
Journal of the American Water Works Association
90
(
9
),
92
102
.
Donovan
E.
Urice
K.
Roberts
J. D.
Harris
M.
Finley
B.
2007
Risk of gastrointestinal disease associated with exposure to pathogens in water of the lower Passaic River
.
Applied Environmental Microbiology
74
(
4
),
994
1003
.
Dreelin
E. A.
Ives
R. L.
Molloy
S.
Rose
J.
2014
Cryptosporidium and Giardia in surface water: a case study from Michigan, USA to inform management of rural water systems
.
International Journal of Environmental Research and Public Health
11
,
10480
10503
.
Gasperi
J.
Zgheib
S.
Cladière
M.
Rocher
V.
Moilleron
R.
Chebbo
G.
2012
Priority pollutants in urban stormwater: Part 2 - Case of combined sewers
.
Water Research
46
,
6693
6703
.
Li
J.
Dhanvantari
S.
Averill
D.
Biswas
N.
2003
Windsor combined sewer overflow treatability study with chemical coagulation
.
Water Quality Research Journal of Canada
38
(
2
),
317
334
.
Marsalek
J.
He
C.
Rochfort
Q.
Exall
K.
Wood
J.
Krishnappan
B.
Seto
P.
Chessie
P.
2005
Upgrading the North Toronto combined sewer overflow (CSO) storage and treatment facility
.
Enhancing Urban Environment by Environmental: Upgrading and Restoration, Nato Science Series: IV: Earth and Environmental Sciences
43
,
111
121
.
Masschelein
W. J.
2002
Ultraviolet Light in Water and Wastewater Sanitation
.
Lewis Publishers
,
New York
,
USA
.
Neupane
D. R.
Riffat
R.
Murthy
S. N.
Peric
M. R.
Wilson
T. E.
2008
Influence of source characteristics, chemicals, and flocculation on chemically enhanced primary treatment
.
Water Environment Research
80
(
4
),
331
338
.
Rodríguez
R. A.
Gundy
P. M.
Rijal
G. K.
Gerba
C. P.
2012
The impact of combined sewage overflows on the viral contamination of receiving waters
.
Food and Environmental Virology
4
,
34
40
.
Standard Methods for the Examination of Water and Wastewater
2012
22nd edn.
American Public Health Association/American Water Works Association/Water Environment Federation
,
Washington, DC
,
USA
.
Stumm
W.
O'Melia
C. R.
1968
Stoichiometry of coagulation
.
Journal of the American Water Works Association
60
(
5
),
514
539
.
Thomas
O.
(ed.)
2007
UV-Visible Spectrophotometry of Water and Wastewater
.
Elsevier
,
New York
,
USA
.
Wojtenko
I.
Stinson
M. K.
Field
R.
2001
Challenges of combined sewer overflow disinfection by ultraviolet light irradiation
.
Critical Reviews in Environmental Science and Technology
31
(
3
),
223
241
.
Zukovs
G.
Kollar
J.
Monteith
H. D.
Ho
W. A.
Ross
R. A.
1986
Disinfection of low quality wastewaters by ultraviolet light irradiation
.
Journal of the Water Pollution Control Federation
58
(
3
),
199
206
.