The objective of the present study was to: (i) examine the performance of two-stage sand filtration and the effect of coagulants on greywater (GW) characteristics; (ii) assess the quality of treated GW for reuse; and (iii) compare reuse options using the analytic hierarchy process (AHP). Four treatment options were examined on site. The first option was related to two-stage sand filtration (TSF) and the other three options were related to coagulation/flocculation using alum, polyaluminium chloride (PAC), and ferric chloride (FeCl3). The setup was constructed close to the GW source of a student hostel located at Sardar Vallabhbhai National Institute of Technology (SVNIT), Surat. Treated GW was safe for reuse in restricted access area irrigation (as per United States Environmental Protection Agency standards), for discharge into land for irrigation, and for industrial cooling (as per Central Pollution Control Board standards in India). All four treatment options were compared using AHP. Comparison was made in two ways: (1) comparison on the basis of the effluent quality produced and (2) comparison on the basis of removal of parameters. The selection string on the basis of the effluent quality was TSF-PAC treatment–alum treatment–FeCl3 treatment, and that on the basis of removal of parameters was TSF-PAC treatment–FeCl3 treatment–alum treatment.

## INTRODUCTION

Greywater (GW) is wastewater from the kitchen, bath, and laundry, excluding wastewater from toilets (WHO 2006; Gross et al. 2007). Light greywater (LGW) is the GW from bathroom, showers, tubs, and clothes washing machine sources. In a household, the proportion of GW flow is around 50 to 80% of the total wastewater flow (Christova-Boal et al. 1996). LGW is around 47% of GW (Ghaitidak & Yadav 2013). Hence, GW reuse can be an effective measure for saving water on the domestic level and reducing the load on wastewater treatment plants.

At present, a limited number of studies have been reported on LGW using sand filters. Ghaitidak & Yadav (2015) examined the effect of coagulants under variable pH conditions. In the present scenario, the technologies are developed just to check the efficiency of a particular system. There is a need to develop technologies by targeting the type of reuse (e.g., agriculture, gardening, flushing, washing, etc.). It is observed that a single method/technology is not capable of meeting the entire reuse standards, so there is a need to develop a flow diagram with a combination of different technologies by targeting the type of reuse. A summary of reported research on on-site GW treatment by filtration and use of coagulants is given in Table 1.

Table 1

Summary of reported research on on-site greywater treatment by filtration and use of coagulants

Reference Main features Main conclusions Remarks
Filtration
Finley et al. (2009)  Tested bath, shower, washing machine GW No significant difference in contamination levels was observed between crops irrigated with tap water, untreated GW, and treated GW Results reinforce the potential of domestic GW as an alternative irrigation source
Treatment using PST (HRT 8 h) followed by coarse filtration (HRT 24 h)  Study was not referred to any reuse standards
Tested impact of irrigating vegetable crops  Adopted very high HRT
Mandal et al. (2011)  Tested bath, wash basin GW Concluded that GW generated from a household is self-sufficient to irrigate a small garden. Study conducted for 2 months (May and June)
Treatment using filtration followed by aeration and disinfection  Tested only two samples
Katukiza et al. (2014)  Tested bath, kitchen, and laundry GW Escherichia coli log removal was >3 but could not satisfy WHO standards Tested in batch mode
Treatment using PST followed by filtration  Quantity treated was limited to 40 L/d
Coagulants
Friedler & Alﬁya (2010)  Tested bath, washbasin GW Most of the removal occurred in the coagulation-sedimentation step, while the ﬁlter acted as a polishing unit Involves large electricity consumption
Treatment using ferric chloride coagulation, sedimentation and filtration  Operated in semi-batch mode
Doses reported in terms of Fe
Kariuki et al. (2011)  Tested kitchen, laundry, hand basin greywater GW treatment system could produce effluent complying with pathogen limit in the referred standards Alum dose was not optimized
Pilot-scale study  Economic evaluation was not reported
Treatment using alum coagulation followed by disinfection  Important reuse parameters like BOD5, TSS were not monitored
Reference Main features Main conclusions Remarks
Filtration
Finley et al. (2009)  Tested bath, shower, washing machine GW No significant difference in contamination levels was observed between crops irrigated with tap water, untreated GW, and treated GW Results reinforce the potential of domestic GW as an alternative irrigation source
Treatment using PST (HRT 8 h) followed by coarse filtration (HRT 24 h)  Study was not referred to any reuse standards
Tested impact of irrigating vegetable crops  Adopted very high HRT
Mandal et al. (2011)  Tested bath, wash basin GW Concluded that GW generated from a household is self-sufficient to irrigate a small garden. Study conducted for 2 months (May and June)
Treatment using filtration followed by aeration and disinfection  Tested only two samples
Katukiza et al. (2014)  Tested bath, kitchen, and laundry GW Escherichia coli log removal was >3 but could not satisfy WHO standards Tested in batch mode
Treatment using PST followed by filtration  Quantity treated was limited to 40 L/d
Coagulants
Friedler & Alﬁya (2010)  Tested bath, washbasin GW Most of the removal occurred in the coagulation-sedimentation step, while the ﬁlter acted as a polishing unit Involves large electricity consumption
Treatment using ferric chloride coagulation, sedimentation and filtration  Operated in semi-batch mode
Doses reported in terms of Fe
Kariuki et al. (2011)  Tested kitchen, laundry, hand basin greywater GW treatment system could produce effluent complying with pathogen limit in the referred standards Alum dose was not optimized
Pilot-scale study  Economic evaluation was not reported
Treatment using alum coagulation followed by disinfection  Important reuse parameters like BOD5, TSS were not monitored

BOD5, 5-day biochemical oxygen demand; TSS, total suspended solids; PST, plain settling; HRT, hydraulic retention time.

When options are compared on the basis of a single criterion, they can be compared simply by arranging them in descending or ascending order. In the case of beneficial (e.g., profit, % removal, etc.) criteria, the options are arranged in descending order of criterion value and for non-beneficial criteria (e.g., cost, effluent concentration, etc.), options are arranged in ascending order. The option placed at the top will be the optimal option and the option placed at the bottom will be the least preferred option. However, comparison of options mostly involves multiple criteria, and selection of the optimal option becomes a complex situation. Therefore, comparison of options with multiple criteria needs a mathematical technique for scientific comparison. In the present study, treatment options were compared using the analytic hierarchy process (AHP).

AHP is a well-known multi-criteria decision-making method that has been widely applied to solve problems in many fields (Vaidya & Kumar 2006; Ishizaka & Labib 2011). However, applications of multi-criteria decision-making to GW investigations are quite limited. Chen et al. (2012) applied the preference ranking organization method for enrichment evaluation (PROMETHEE) for selecting a recycling alternative in a household laundry in Sydney. The present study gives a step-by-step solution for the use of AHP in selecting the optimal option in LGW investigations. In view of the above, the objective of the present study was to: (i) examine the performance of two-stage sand filtration and the effect of coagulants on GW characteristics; (ii) assess the quality of treated GW for reuse; and (iii) compare reuse options using AHP.

## MATERIALS AND METHODS

### Greywater, filter media, and coagulants

Greywater was collected from a student hostel (capacity 474 students) located at Sardar Vallabhbhai National Institute of Technology (SVNIT), Surat, India. The GW collection pipe was cut and GW was diverted to a collecting tank of 500-L capacity.

Naturally available river sand was used for preparation of the filters. Coarse sand (2 mm–4.75 mm), medium sand (1 mm–2 mm), and fine sand (0.125 mm–1 mm) were used. Sand was washed with potable water and sun dried. The sand was stocked in separate piles according to size, viz. coarse, medium, and fine.

Alum (Al2(SO4)3.18H2O), polyaluminium chloride (PAC), and ferric chloride (FeCl3) coagulants were used in the study. The alum contained 4.75% Al, the PAC contained 15.07% Al, and the FeCl3 contained 30.97% Fe on a weight basis. The stock was prepared regularly by dissolving 12–36 g coagulant in 1,000 mL distilled water. Normally, the entire stock was required in each coagulation run. Analytical grade coagulants were used in the entire study.

### Analytical procedures

The raw and treated GW samples were analyzed using Standard Methods (APHA 2005). Analytical grade chemicals were used in the study. Parameter pH was measured using a digital pH meter (Hanna Instruments pH 209); turbidity was measured using a digital Nephelo turbidity meter (Systronics Turbidity Meter 132); and electrical conductivity (EC25) was measured using a digital conductivity meter (Systronics Conductivity TDS Meter 308). Standardization of the instruments was checked each day before sample analysis to ascertain their proper working condition. Solids were measured by drying the samples in a hot air oven at 105 ± 2/180 ± 2 °C. Oil and grease (O&G) were analyzed by the partition-gravimetric method using petroleum ether (40 °C/60 °C). Phosphate (PO4-P) was analyzed by the stannous chloride method. Dissolved oxygen in the biochemical oxygen demand test was analyzed as per the Winkler method with azide modification. Chemical oxygen demand (COD) was analyzed by the open reflux method using potassium dichromate (K2Cr2O7) as an oxidizing agent. Metals and earth elements (Ca, Mg, Na, B, As, Al, Fe) were analyzed using an inductively coupled plasma-atomic emission spectrometer (ARCOS, M/s, Spectro, Germany).

Total coliforms (TC) and fecal coliforms (FC) were analyzed by the multiple-tube fermentation technique (five tubes, three dilutions). A presumptive test for coliforms was performed using McConkey broth (HiMedia Laboratories Pvt. Ltd, Mumbai, India) by incubating at 35 ± 0.5 °C for 24 h. Further, TC were confirmed using Brilliant Green Bile Broth (HiMedia) by incubating at 35 ± 0.5 °C for 24 h; and FC were confirmed using EC broth (HiMedia) by incubating at 44.5 ± 0.2 °C for 24 h. Results were expressed as most probable number (MPN)/100 mL. Escherichia coli were measured by the pour plate method using EMB agar (HiMedia) by incubating at 37 °C for 24 h, and results were expressed as colony forming units (CFU) per 100 mL.

### Greywater treatment options

The setup was constructed on-site close to the GW source of a student hostel located at SVNIT, Surat. Fresh GW was collected continuously in a greywater collection tank (GWCT) of 500-L capacity after primary screening to remove the floating matter and further passed through a coarse sand filter. The study was divided in two parts. Part I deals with two-stage sand filtration and Part II deals with the effect of coagulants. The collected GW was first filtered through a coarse sand filter. Then it was passed through either two-stage sand filters or a coagulation/flocculation (COF) unit. The entire flow in the system was under gravity. A conceptual flow diagram of the experimental setup is shown in Figure 1.
Figure 1

Conceptual flow diagram of experimental setup. GWCT, raw greywater collection tank; CF, coarse sand filter; SF1, sand filter stage I; SF2, sand filter stage II; COF, coagulation/flocculation; SF, sand filter; CT1, treated water collection tank 1; CT2, treated water collection tank 2; Option-1: raw GW-CF-SF1-SF2-disinfection; Option-2: raw GW-CF-alum coagulation-SF-disinfection; Option-3: raw GW-CF-PAC coagulation-SF-disinfection; Option-4: raw GW-CF-FeCl3 coagulation-SF-disinfection.

Figure 1

Conceptual flow diagram of experimental setup. GWCT, raw greywater collection tank; CF, coarse sand filter; SF1, sand filter stage I; SF2, sand filter stage II; COF, coagulation/flocculation; SF, sand filter; CT1, treated water collection tank 1; CT2, treated water collection tank 2; Option-1: raw GW-CF-SF1-SF2-disinfection; Option-2: raw GW-CF-alum coagulation-SF-disinfection; Option-3: raw GW-CF-PAC coagulation-SF-disinfection; Option-4: raw GW-CF-FeCl3 coagulation-SF-disinfection.

### Part I: performance of two-stage sand filtration in greywater treatment

This part pertains to option 1. This option refers to the two-stage sand filtration (TSF) in this study. Two identical sand filters were operated in series in continuous mode. In the first stage, the effluent of the coarse sand filter was passed through sand filter stage I (SF1). In stage two, the effluent of the SF1 was passed through sand filter stage II (SF2). Filters were operated in the submerged mode. Filters were exposed to the atmosphere and covered by nylon mosquito net to avoid mosquitoes and flies breeding. GW treated using two-stage sand filtration was further disinfected using bleaching powder.

#### Coarse sand filter

A polyvinyl chloride (PVC) cylindrical container was used for fabricating the coarse sand filter (CF) used in the study. The CF had an internal diameter of 33 cm and depth of 56 cm. Figure 2 shows the schematic diagram of the mini coarse sand filter and coarse sand filter used in this study. The CF contained a removable mini coarse sand filter (MCF) inside. A small 3-L capacity plastic bucket half filled with sand (the same sand as in the CF) was used as the MCF. The raw GW flowing from the GWCT was first passed through the MCF and then passed through the CF. The MCF was cleaned daily using potable tap water. The hydraulic retention time (HRT) of the GW in the CF was 1 h at 10 L/h flow rate.
Figure 2

Schematic diagram of mini coarse sand filter and coarse sand filter. Layer-I, size 2,000–4,750 mm and depth 5 cm; Layer-II, size 1,000–2,000 mm and depth 32 cm.

Figure 2

Schematic diagram of mini coarse sand filter and coarse sand filter. Layer-I, size 2,000–4,750 mm and depth 5 cm; Layer-II, size 1,000–2,000 mm and depth 32 cm.

#### Sand filter stage I/II (SF1/SF2)

GW filtered from the CF was further treated using the two-stage sand filtration unit. A PVC container of 104 L capacity was used for fabricating the sand filter. The filter medium comprises coarse sand, medium sand, and fine sand. The effective size (D10) and uniformity coefficient (Cu) of the fine sand used were 0.25 mm and 2.8, respectively. The initial pore volumes in SF1 and SF2 were 26% each. A schematic diagram of the sand filters (SF1/SF2) is shown in Figure 3. Filters SF1 and SF2 were operated in series. The flow was in a downward direction and the filters were operated in submerged mode.
Figure 3

Schematic diagram of sand filter SF1/ SF2. CF, coarse sand filter; Layer-I, size 2,000–4.750 mm and depth 5 cm; Layer-II, size 1,000–2,000 mm and depth 5 cm; Layer-III, size 0.125–1,000 mm and depth 46 cm.

Figure 3

Schematic diagram of sand filter SF1/ SF2. CF, coarse sand filter; Layer-I, size 2,000–4.750 mm and depth 5 cm; Layer-II, size 1,000–2,000 mm and depth 5 cm; Layer-III, size 0.125–1,000 mm and depth 46 cm.

### Part II: effect of coagulants on greywater characteristics

This part pertains to option 2, option 3, and option 4, and refers to alum treatment, PAC treatment, and FeCl3 treatment, respectively, in the present study.

#### Coagulation/flocculation unit

The COF unit comprises a 1 L capacity plastic jar (PJ), an aquarium aerator, coagulant feeder tank (CO), and a flocculation-cum-settling tank (FST). Figure 4 shows a schematic diagram of the coagulation/flocculation unit. A plastic jar was kept in suspension in a just submerged condition in the FST. An aquarium aerator of 3 L/min capacity (consumption 3 W, 220–240 V) was used for rapid mixing of the coagulant. One outlet of the aerator was connected to the CO, and another outlet was connected to the PJ. A supply of air was required in the CO to keep the coagulant uniformly mixing and ensure a uniform coagulant concentration.
Figure 4

Schematic diagram of the coagulation/ flocculation unit. CF, coarse sand filter; FST, flocculation-cum-settling tank.

Figure 4

Schematic diagram of the coagulation/ flocculation unit. CF, coarse sand filter; FST, flocculation-cum-settling tank.

The coagulant and coarse filtered GW was first diverted into the PJ, which acts as a coagulation-cum-rapid mixing tank. Air injected by the aerator creates rapid turbulence. The turbulence thus created causes breaking down of the zeta potential, mixing of the coagulant with the GW, and gives upward/lateral movement to the GW leaving the PJ. A PVC container of 47 L capacity (diameter 33 cm, depth 56 cm) was used as the FST. The FST provides the space for floc formation, floc aggregation, and settling. The HRT of the FST was 2 h at 20 L/h flow rate. The effluent from the FST was further filtered using a sand filter.

In the present study, rapid mixing was adopted by using a pneumatic method (i.e., diffused air). The temporal mean velocity gradient (G) was calculated using Equation (1):
1
where P = total input of power in watts; μ = absolute viscosity of greywater, in N.S/m2; vol = volume of the rapid mix unit, in m3.

#### Sand filter (non-submerged)

Water passing through the coagulation/flocculation unit was further treated using a sand filter (SF). The filter configuration was the same as that of the SF1/SF2. However, this filter was operated in non-submerged mode. Filtrate from the SF was collected through a drain valve so as to operate it in non-submerged mode.

#### Disinfection of treated greywater

Greywater treated in both parts was disinfected using bleaching powder (BP). The available chlorine in the BP was 22% by weight. The chlorine dose was added to the treated GW and residual chlorine was measured after 30 min contact time.

### Statistical analysis

Statistical analysis of the results was performed using Excel 2007 and SYSTAT (Sigmaplot 12). A paired t-test (paired two sample for means) was performed on parameters monitored before and after the treatment. This test was appropriate in the present study because: (1) parameters before and after the treatment were compared, and they were of the same size; and (2) the parameters compared were a continuous variable.

The null (H0) and alternate hypothesis (H1) framed in the t-test were (Equations (2) and (3)):
2

3
where μR and μT were mean concentrations of the parameters in raw and treated GW, respectively. μT can be replaced subsequently as μ1, μ2, μ3, and μ4, which correspond to option 1, option 2, option 3, and option 4, respectively.

The level of test significance was 95%. A p-value < 0.05 indicates that H0 can be rejected, which means the mean concentration of the parameters differs significantly after treatment. A p-value >0.05 indicates failure to reject H0, which means the mean concentration of the parameters does not differ significantly after treatment.

### Analytic hierarchy process

The AHP is a method based on priority theory, and is capable of: (i) breaking down a complex problem into its component parts; (ii) arranging these components in a hierarchy model; and (iii) assigning numerical values to criteria based on their subjective judgments (Saaty 1980; Rao 2007).

In the present study, the geometric mean method of AHP was used. This method of AHP is commonly used to determine the priority weights of the attributes owing to its simplicity, easy means of finding the maximum Eigen value, and reduction in inconsistency of judgments.

To decide the importance of the attributes, an expert opinion survey was conducted. Table 2 presents the comparison scale used in AHP and corresponding linguistic variables used in the survey in the present study. A questionnaire containing the list of the criteria/attributes was circulated among environmental experts and their responses regarding the importance to be assigned to the attributes were collected in the linguistic form on the basis of the comparison scale proposed by Saaty (1990). The linguistic variables were converted to the absolute numbers. The average values of the responses of the experts were considered in forming the relative importance matrix.

Table 2

The comparison scale used in AHP and linguistic variables assigned

Intensity of importance Definition Explanation Linguistic variable
Equal importance Two activities contribute equally to one objective Extremely low
Intensity of importance (II) between 1 and 3 – Very low
Moderate importance of one over another Judgment slightly favors one activity over another Low
II between 3 and 5 – Below average
Strong importance Judgment strongly favors one activity over another Average
II between 5 and 7 – Above average
Very strong importance An activity is strongly favored and its dominance is demonstrated in practice High
II between 7 and 9 – Very high
Extreme importance The evidence favoring one activity over another is of the highest possible order of affirmation Extremely high
Intensity of importance Definition Explanation Linguistic variable
Equal importance Two activities contribute equally to one objective Extremely low
Intensity of importance (II) between 1 and 3 – Very low
Moderate importance of one over another Judgment slightly favors one activity over another Low
II between 3 and 5 – Below average
Strong importance Judgment strongly favors one activity over another Average
II between 5 and 7 – Above average
Very strong importance An activity is strongly favored and its dominance is demonstrated in practice High
II between 7 and 9 – Very high
Extreme importance The evidence favoring one activity over another is of the highest possible order of affirmation Extremely high

Note: reciprocals of the above non zero – if activity i has one of the above non-zero numbers assigned to it when compared with activity j, then j has the reciprocal value when compared with i.

## RESULTS AND DISCUSSION

### Performance of coarse filtration in greywater treatment

Filtration was carried out in continuous mode. Every week the GWCT was completely emptied by operating outlet valves kept at the bottom for draining purposes. This was necessary to avoid the accumulation of solids and contaminants in the GWCT. Characteristics of the raw and coarse filtered GW are given in Table 3.

Table 3

Characteristics of raw and coarse filtered greywater

Raw greywater

CF effluent

Parameter Min–max Avg ± SD Min–max Avg ± SD
pH 85 6.62–8.13 7.49 ± 0.30 6.4–8.34 7.52 ± 0.29
Turbidity, NTU 85 5.7–190.0 80.3 ± 30.5 3.1–98.1 39.2 ± 16.2
Temperature, °C 85 22.2–30.4 28 ± 2.9 21.6–30.7 26 ± 2.9
EC25, μS/cm 85 269–646 489 ± 83 265–637 482 ± 84
Total dissolved solids (TDS), mg/L 77 184–396 304 ± 53 182–406 299 ± 53
TSS, mg/L 77 24–215 88 ± 30 14–176 70 ± 24
Oil and grease (O&G), mg/L 59 4–80 33 ± 13 3–58 23 ± 9
Alkalinity, mg/L 79 152–260 187 ± 15 144–248 171 ± 16
BOD5, mg/L 63 30–306 102 ± 49 15–290 87 ± 43
COD, mg/L 58 68–596 243 ± 101 44–556 224 ± 95
Total coliforms (TC), MPN/100 mL 15 3.10 × 105–4.90 × 107 1.88 × 107 ± 9.86 × 106 2.10 × 105–2.10 × 107 7.90 × 106 ± 7.00 × 106
Fecal coliforms (FC), MPN/100 mL 15 1.20 × 104–8.40 × 105 2.83 × 105 ± 3.46 × E5 9.30 × 103–7.20 × 105 1.90 × 105 ± 3.20 × 105
E. coli, CFU/100 mL 15 3.00 × 102–2.10 × 103 9.71 × 102 ± 5.24 × 102 4.00 × 102–1.70 × 102 9.10 × 102 ± 4.60 × 102
Sodium adsorption ratio (SAR) 1.40–3.60 2.30 ± 0.37 – –
Boron, mg/L 0.05–0.10 0.07 ± 0.01 – –
Arsenic, mg/L < 0.01 < 0.01 – –
Ammonia nitrogen (NH3-N), mg/L 61 0.50–3.90 2.00 ± 0.93 0.30–3.36 1.69 ± 0.81
Phosphates (PO4-P), mg/L 42 0.20–1.90 0.95 ± 0.24 0.10–1.80 0.87 ± 0.23
Raw greywater

CF effluent

Parameter Min–max Avg ± SD Min–max Avg ± SD
pH 85 6.62–8.13 7.49 ± 0.30 6.4–8.34 7.52 ± 0.29
Turbidity, NTU 85 5.7–190.0 80.3 ± 30.5 3.1–98.1 39.2 ± 16.2
Temperature, °C 85 22.2–30.4 28 ± 2.9 21.6–30.7 26 ± 2.9
EC25, μS/cm 85 269–646 489 ± 83 265–637 482 ± 84
Total dissolved solids (TDS), mg/L 77 184–396 304 ± 53 182–406 299 ± 53
TSS, mg/L 77 24–215 88 ± 30 14–176 70 ± 24
Oil and grease (O&G), mg/L 59 4–80 33 ± 13 3–58 23 ± 9
Alkalinity, mg/L 79 152–260 187 ± 15 144–248 171 ± 16
BOD5, mg/L 63 30–306 102 ± 49 15–290 87 ± 43
COD, mg/L 58 68–596 243 ± 101 44–556 224 ± 95
Total coliforms (TC), MPN/100 mL 15 3.10 × 105–4.90 × 107 1.88 × 107 ± 9.86 × 106 2.10 × 105–2.10 × 107 7.90 × 106 ± 7.00 × 106
Fecal coliforms (FC), MPN/100 mL 15 1.20 × 104–8.40 × 105 2.83 × 105 ± 3.46 × E5 9.30 × 103–7.20 × 105 1.90 × 105 ± 3.20 × 105
E. coli, CFU/100 mL 15 3.00 × 102–2.10 × 103 9.71 × 102 ± 5.24 × 102 4.00 × 102–1.70 × 102 9.10 × 102 ± 4.60 × 102
Sodium adsorption ratio (SAR) 1.40–3.60 2.30 ± 0.37 – –
Boron, mg/L 0.05–0.10 0.07 ± 0.01 – –
Arsenic, mg/L < 0.01 < 0.01 – –
Ammonia nitrogen (NH3-N), mg/L 61 0.50–3.90 2.00 ± 0.93 0.30–3.36 1.69 ± 0.81
Phosphates (PO4-P), mg/L 42 0.20–1.90 0.95 ± 0.24 0.10–1.80 0.87 ± 0.23

n, number of samples; Avg, arithmetic average; SD, standard deviation; CF, coarse sand filter.

### Part I: performance of two-stage sand filtration in greywater treatment

The flow rate was adjusted to 10.2 ± 0.46 L/h. Greywater treated using two-stage sand filtration was further disinfected using bleaching powder. The average hydraulic loading rate, organic loading rate, and HRT observed in the filtration study were as given in Table 4.

Table 4

Sand filter stage I (SF1) 70.2 0.153 0.309 2.4
Sand filter stage II (SF2) 70.2 0.089 0.180 2.4

Sand filter stage I (SF1) 70.2 0.153 0.309 2.4
Sand filter stage II (SF2) 70.2 0.089 0.180 2.4

Four filtration runs were conducted for durations of 38, 53, 36, and 31 days in runs 1, 2, 3, and 4, respectively. At the end of each run the entire filter medium was removed and washed using potable water. The medium was dried and refilled in each filter before commencing the next run. Filter head loss at filter SF1 was monitored in run 3 and run 4. In run 3, the head loss was greatest (120 mm) on day 36, and in run 4 the head loss was greatest on day 31, when the filter started overflowing. The head loss observed in both the runs is shown in Figure 5. In run 3, the head loss in the filter was governed by Equation (4) with an R2 value of 0.989. In run 4, the head loss was governed by Equation (5) with an = 0.987.
4

5
where y = head loss (mm); x = run days.
Figure 5

Head loss in sand filter: (a) run 3, (b) run 4.

Figure 5

Head loss in sand filter: (a) run 3, (b) run 4.

The initial pore volume of the filter (SF1, SF2) was around 26%, which gradually reduced to 14% at the end of the run. The effective HRT varied from 2.4 to 1.7 h at a flow rate of 10 L/h. Thus, the combined HRT of both the filters varied from 4.8 to 3.4 h.

#### Characteristics of filtered greywater

The effect of filtration on GW characteristics is presented in Table 5 (see option 1). The mean raw GW pH was changed from 7.51 to 7.38, 7.19, and 7.24 in CF, SF1 and SF2 effluent, respectively. Similarly, mean alkalinity was also changed from 187 mg/L to 171, 154, 140 mg/L in CF, SF1 and SF2 effluent, respectively. Soap was a major contributor to the alkalinity. As the suspended solids (soap particles) were trapped in the reactor CF, SF1 and SF2 progressively, alkalinity was reduced. A drop in pH and alkalinity in SF1 and SF2 effluents may also be due to the release of H+ ions during the nitrification process.

Table 5

Mean raw and treated GW characteristics observed in each treatment option

Option 1 (TSF)

Option 2 (Alum treatment)

Option 3 (PAC treatment)

Option 4 (FeCl3 treatment)

Parameter Raw Treateda Raw Treatedb Raw Treatedc Raw Treatedd
pH 7.51 7.24 7.51 7.44 7.66 7.37 7.57 7.54
Turbidity, NTU 83.4 4.5 89.6 0.9 83.8 0.6 77.7 2.3
Electrical conductivity, μS/cm 488 465 454 495 497 520 500 523
Total suspended solids, mg/L 92 15 83 94 92
Oil and grease, mg/L 35 30 31 31
BOD5, mg/L 106 28 95 11 113 105 15
Boron, mg/L 0.071 0.010 0.063 0.010 0.079 0.010 0.07 0.010
Sodium adsorption ratio 2.36 2.00 1.55 1.50 3.10 2.35 2.45 2.15
Fecal coliforms, MPN/100 mL 1.97 × 105 < 2 2.50 × 105 < 2 1.40 × 105 < 2 2.60 × 105 < 2
Option 1 (TSF)

Option 2 (Alum treatment)

Option 3 (PAC treatment)

Option 4 (FeCl3 treatment)

Parameter Raw Treateda Raw Treatedb Raw Treatedc Raw Treatedd
pH 7.51 7.24 7.51 7.44 7.66 7.37 7.57 7.54
Turbidity, NTU 83.4 4.5 89.6 0.9 83.8 0.6 77.7 2.3
Electrical conductivity, μS/cm 488 465 454 495 497 520 500 523
Total suspended solids, mg/L 92 15 83 94 92
Oil and grease, mg/L 35 30 31 31
BOD5, mg/L 106 28 95 11 113 105 15
Boron, mg/L 0.071 0.010 0.063 0.010 0.079 0.010 0.07 0.010
Sodium adsorption ratio 2.36 2.00 1.55 1.50 3.10 2.35 2.45 2.15
Fecal coliforms, MPN/100 mL 1.97 × 105 < 2 2.50 × 105 < 2 1.40 × 105 < 2 2.60 × 105 < 2

aFinal effluent from option-1.

bfinal effluent from option-2.

cfinal effluent from option-3.

dfinal effluent from option-4; TSF, two-stage sand filtration.

The effect of filtration on turbidity and total suspended solids (TSS) is shown in Figures 6 and 7, respectively. Mean raw GW turbidity was reduced from 83.4 NTU to 40.3, 7.9, 4.5 NTU and mean raw GW TSS was reduced from 92 mg/L to 73, 34, 15 mg/L, in CF, SF1 and SF2 effluent, respectively. Turbidity is an indirect measure of the TSS and colloidal solids. Hence, the reduction in TSS and colloidal solids leads to the reduction of turbidity. In a similar study, using a four barrel gravel filter, Al-Hamaiedeh & Bino (2010) observed a reduction in TSS from 275 mg/L to 150 mg/L (removal 46%). Mandal et al. (2011) observed turbidity removal of 38.7 to 9.6 NTU (removal of 75%), and TSS removal of 17.6 to 12 mg/L (removal of 32%). Karabelnik et al. (2012) observed 86–90% TSS removal using a vertical filter (4–10 mm size crushed Filtralite® media) at a flow rate of 80 L/d.
Figure 6

Performance of sand filtration on turbidity. CF, coarse sand filter; SF1, sand filter stage I; SF2, sand filter stage II.

Figure 6

Performance of sand filtration on turbidity. CF, coarse sand filter; SF1, sand filter stage I; SF2, sand filter stage II.

Figure 7

Performance of sand filtration on total suspended solids. CPCB (1993; 2008) standard: 200 mg/L max; USEPA (2004) standard: 30 mg/L max; CF, coarse sand filter; SF1, sand filter stage I; SF2, sand filter stage II.

Figure 7

Performance of sand filtration on total suspended solids. CPCB (1993; 2008) standard: 200 mg/L max; USEPA (2004) standard: 30 mg/L max; CF, coarse sand filter; SF1, sand filter stage I; SF2, sand filter stage II.

In the present study, mean turbidity removal was 51.7, 90.5, and 94.6% and mean TSS removal was 20.7, 63.0, and 83.7% in CF, SF1 and SF2 effluent respectively. The better turbidity and TSS removal obtained in the present study may be due to use of fine sand filter media. The % removal of turbidity and TSS was below average in the first 8–10 days of the filter run. In the last 3–4 days of the filter run, removal was above average. Initially, at the time of commissioning, the filter pores of the media were empty; however, as the volume of GW filtered went on increasing, the filter pores gradually reduced, thereby enabling the filter media to trap the finer particles. Thus, the rate of removal increased as the filter proceeded towards maturity. The dominant mechanisms responsible for removal of turbidity and TSS may be straining and sedimentation within the granular medium.

The mean O&G concentration was reduced from 35 to 24, 11, and 8 mg/L in CF, SF1 and SF2 effluent, respectively. The corresponding removal was 31.4, 68.6, 77.1%, respectively. Since all the three reactors were operated in submerged mode, a few cm depth of standing water was always above the sand bed. The specific gravity of O&G is less than that of GW. Hence, the O&G was trapped in the standing water in the reactor.

The effect of filtration on 5-day biochemical oxygen demand (BOD5) is shown in Figure 8. The mean BOD5 concentration was reduced from 106 mg/L to 91, 53, and 28 mg/L in CF, SF1 and SF2 effluent, respectively. The corresponding mean removal was 14.2, 50.0, and 73.6% respectively. In the first 8–10 days, the removal rate was below average, whereas the removal was above average in the next days of the run. As the filter progressed, the filter pores became clogged due to trapping of solids. The biological slime layer developed after 8–10 days and became attached on the top layer of the media in the SF1 and SF2 reactors. The slime layer contains mainly facultative bacteria, fungi, algae, and protozoans (Metcalf & Eddy Inc. 2003). Removal of BOD was through the combined action of physical and biological processes. Prior to development of the slime layer, physical processes such as straining, sedimentation, and adsorption may be dominant, which mainly removed suspended BOD. However, after development of the slime layer, both physical and biological mechanisms played a significant role. Microorganisms in the slime layer degrade organic matter (dissolved) as the GW passes down through it. Over a period of time, as the thickness of the slime layer increases, the slime layers may act as a fine straining mat and improve the removal of fine particles contributing to BOD. Hence, in the last few days of the run, the BOD removal was above average due to the combined physical and biological removal mechanisms.
Figure 8

Performance of sand filtration on biochemical oxygen demand. CPCB (1993, 2008) standard: 100 mg/L max; USEPA (2004) standard: 30 mg/L max; CF, coarse sand filter; SF1, sand filter stage I; SF2, sand filter stage II.

Figure 8

Performance of sand filtration on biochemical oxygen demand. CPCB (1993, 2008) standard: 100 mg/L max; USEPA (2004) standard: 30 mg/L max; CF, coarse sand filter; SF1, sand filter stage I; SF2, sand filter stage II.

In a similar study, Al-Hamaiedeh & Bino (2010) obtained a BOD5 removal from 942 mg/L to 108 mg/L (removal of 88%). In the present study mean BOD5 removal was around 70% with a residual BOD5 of 29 mg/L. The percentage removal was slightly lower, but the final effluent quality was better than that reported by Al-Hamaiedeh & Bino (2010). The better quality of the effluent obtained in the present study may be due to the use of a fine sand medium and the contribution of bacterial slime developing on the top medium layer in the filter.

The mean TC count of 9.10 × 106 MPN/100 mL was reduced to 8.27 × 106, 2.77 × 105, 2.29 × 104 MPN/100 mL; the FC count was reduced from 2.38 × 105 to 1.97 × 105, 2.11 × 104, 2.17 × 103 MPN/100 mL; and the E. coli count was reduced from 9.35 × 102 CFU/100 mL to 9.20 × 102, 7.00 × 102, 4.80 × 102 CFU/100 mL, in CF, SF1 and SF2 effluent, respectively. The mean log10 removal of TC was 0.04, 1.52, and 2.6; removal of FC was 0.08, 1.05, and 2.04; and E. coli removal was 0.01, 0.13, and 0.29 in CF, SF1 and SF2 effluent, respectively. The microorganisms in the bio-layer consumed the pathogens in the GW (CAWST 2008). A better removal of pathogens was observed in the present study. This may be due to the removal by microorganisms in the bio-layer, and die-off in the lower layer of the filter due to lack of food and oxygen. The dissolved oxygen level in filtration effluents (SF1 and SF2) were of the order of 0.2 to 0.4 mg/L.

Katukiza et al. (2014) obtained E. coli removal of 1.595 (log10). This is higher than that in the present study. Katukiza et al. (2014) operated filters in batch mode treating only 40 L/d of GW, whereas the present study was a continuous mode operation treating 240–250 L/d. The lesser removal of E. coli observed in the present study may be due to the continuous mode operation and treating a greater quantity of GW compared to that in the cited research.

### Part II: effect of coagulants on greywater characteristics

Greywater was filtered using a coarse sand filter and coagulated using alum, PAC, and FeCl3. Coagulated GW was filtered using a sand filter and was disinfected using bleaching powder. Floc size in the alum and PAC coagulation was up to 3 mm, and in FeCl3 was up to 2 mm. Owing to a continuous flow of GW, the flocs were escaping from the coagulation tank; therefore, the filtration was essential to trap the flocs/solids from the coagulated water. A sand filter (SF) was provided for entrapment of the flocs from the coagulated GW.

#### Rapid mixing and settling

A total of 36 coagulation runs (average 12 runs/coagulant) were conducted in the present study. From the laboratory-scale study the optimum coagulant dose was observed as 351 ± 78, 190 ± 37, and 107 ± 28 mg/L using alum, PAC, and FeCl3, respectively (Ghaitidak & Yadav 2014). They were roughly in the ratio of Alum:PAC:FeCl3 = 3:2:1. Therefore, the coagulant doses used in the on-site study were maintained in the same proportion. The doses applied were 416 ± 44, 279 ± 40, and 141 ± 12 mg/L, in alum, PAC, and FeCl3, respectively. The coagulant doses used in the present study were kept slightly higher than that in the laboratory-scale study due to unfavorable coagulation/flocculation facilities available on-site.

As per CPHEEO (1999), design criteria for rapid mix unit are: detention time = 30 to 60 s; temporal mean velocity gradient (G) > 300 s−1 (range 300 to 5,000 s−1); power requirement = 1 to 3 W per m3/h flow. In the present study power used was 1.5 W, volume of the mixing container was 1 L. Assuming μ = 0.798 × 10−3 (@ 30 °C temperature), G is 1,371 s−1. The detention time at 20 L/h flow rate is 180 s.

#### Characteristics of treated greywater

The effect of coagulants on GW characteristics is given in Table 5 (see option-2, -3, -4 for alum, PAC, and FeCl3, respectively). Mean alkalinity was reduced from 186 mg/L to 177, 63, and 59 mg/L in alum treatment; it was reduced from 189 mg/L to 174, 78, and 83 mg/L in PAC treatment; and reduced from 190 mg/L to 177, 85, and 82 mg/L in FeCl3 treatment in CF, COF, and SF effluent, respectively. The major removal of alkalinity was in the COF unit. This may be because of the necessity of alkalinity in forming hydroxide ions required for floc formation and precipitation. Around 0.45 mg/L alkalinity is required for 1.0 mg/L of alum for its complete hydrolysis into aluminium hydroxide (Metcalf & Eddy Inc. 2003). Hence, the drop in alkalinity was observed in COF effluent.

The effect of coagulants on turbidity and TSS is shown in Figures 9 and 10, respectively. Mean turbidity was reduced from 89.6 NTU to 52.7, 13.1, and 0.9 NTU in alum treatment; it was reduced from 83.8 NTU to 34.3, 3.6, and 0.6 NTU in PAC treatment; and from 77.7 NTU to 37.5, 10.0, and 2.3 NTU in FeCl3 treatment in CF, COF, and SF effluent, respectively. Mean turbidity removal was 41.2, 85.4, and 99.0% in alum; 59.0, 95.6, and 99.3% in PAC; and 51.7, 87.1, and 97.0% in FeCl3 treatment in CF, COF, and SF effluent, respectively. Mean TSS was reduced from 83 mg/L to 67, 14, and 5 mg/L in alum treatment; it was reduced from 93 mg/L to 75, 9, and 4 mg/L in PAC treatment; and from 92 mg/L to 78, 12, and 7 mg/L in FeCl3 treatment in CF, COF, and SF effluent, respectively. Mean TSS removal was 19.2, 83.1, and 93.9% in alum; 19.3, 90.3, and 95.6% in PAC; and 15.2, 86.9, and 92.3% in FeCl3 treatment in CF, COF, and SF effluent, respectively.
Figure 9

Effect of coagulant on turbidity. CF, coarse sand filter; COF, coagulation/flocculation; SF, sand filter.

Figure 9

Effect of coagulant on turbidity. CF, coarse sand filter; COF, coagulation/flocculation; SF, sand filter.

Figure 10

Effect of coagulant on total suspended solids. CPCB (1993, 2008) standard: 200 mg/L max; USEPA (2004) standard: 30 mg/L max; CF, coarse sand filter; COF, coagulation/flocculation; SF, sand filter.

Figure 10

Effect of coagulant on total suspended solids. CPCB (1993, 2008) standard: 200 mg/L max; USEPA (2004) standard: 30 mg/L max; CF, coarse sand filter; COF, coagulation/flocculation; SF, sand filter.

The effluent quality of all the three coagulation treatments was uniform, and turbidity removal was above 95% after filtration. In a similar study, Friedler & Alﬁya (2010) observed turbidity levels drop from 46 NTU to 3.9 NTU (removal of 92%), and TSS levels from 70 mg/L to 4.4 mg/L (removal of 94%) while treating LGW using FeCl3 at a dose of 22 mg-Fe/L. In the present study, similar turbidity and TSS removal was observed at a comparatively higher Fe dose (43.6 mg-Fe/L). The higher Fe dose needed in the present study may be due to more turbid raw GW, and rapid mixing by a pneumatic method as opposed to the mechanical mixing (i.e., rapid mixing at 60 rpm for 5 min, slow mixing at 30 rpm for 30 min) adopted by Friedler & Alﬁya (2010).

Figure 11 shows the effect of coagulants on BOD5. Mean BOD5 was reduced from 95 mg/L to 86, 19, and 11 mg/L in alum treatment; from 113 mg/L to 101, 15, and 7 mg/L in PAC treatment; from 105 mg/L to 93, 23, and 15 mg/L in FeCl3 treatment in CF, COF, and SF effluent, respectively. Mean BOD5 removal was 9.5, 80, and 88.4% in alum treatment; 10.6, 86.5, and 93.5% in PAC treatment; and 11.4, 78.1, and 85.7% in FeCl3 treatment in CF, COF, and SF effluent, respectively. Friedler & Alﬁya (2010) observed BOD5 removal from 103 mg/L to 44 mg/L (removal of 57%) using FeCl3. Settling time after slow mixing was 40 min, and the effective sand size in the filter column was 0.585 mm. The higher BOD5 removal observed in the present study may be due to the good coagulation efficiency achieved, higher settling time (≈2 h) after coagulation, and smaller effective sand size (0.25 mm) adopted in the sand filter as compared to those by Friedler & Alﬁya (2010).
Figure 11

Effect of coagulant on biochemical oxygen demand. CPCB (1993, 2008) standard: 100 mg/L max; USEPA (2004) standard: 30 mg/L max; CF, coarse sand filter; COF, coagulation/flocculation; SF, sand filter.

Figure 11

Effect of coagulant on biochemical oxygen demand. CPCB (1993, 2008) standard: 100 mg/L max; USEPA (2004) standard: 30 mg/L max; CF, coarse sand filter; COF, coagulation/flocculation; SF, sand filter.

The mean FC count was reduced from 2.83 × 105 MPN/100 mL to 2.5 × 105, 6.5 × 104, 5.2 × 103 MPN/100 mL in alum treatment; from 2.0 × 105 MPN/100 mL to 1.4 × 105, 2.4 × 104, 9.8 × 102 MPN/100 mL in PAC treatment, from 3.0 × 105 MPN/100 mL to 2.6 × 105, 7.7 × 104, 5.5 × 103 MPN/100 mL in FeCl3 treatment in CF, COF, and SF effluent, respectively.

The TC and FC count were <2 MPN/100 mL each, and the E. coli count was <2 CFU/100 mL after disinfection. FC removal (log10) was 5.1, 4.35, and 5.11, in alum, PAC, and FeCl3 treatment, respectively. Kariuki et al. (2011) observed FC removal from 1.1 × 103 to 0 MPN/100 mL (log10 removal = 3) using alum coagulation followed by disinfection using sodium hypochlorite (dose and contact time not reported). The higher FC removal observed in the present study may be due to the chlorine dosage applied and proper contact time provided.

### Compliance of treated greywater with reuse standards

USEPA (2004) standards for restricted access area irrigation (RAI) prescribe pH = 6–9, TSS ≤ 30 mg/L, BOD5 ≤ 30 mg/L, fecal coliforms ≤ 200 MPN/100 mL, chlorine (Cl2) = 1 mg/L residual (minimum). Standards for construction (soil compaction, dust control, washing aggregate and making concrete) are the same as RAI. WHO (2006) standards for unrestricted irrigation of crops prescribe E. coli < 1,000 CFU/100 mL. CPCB (1993, 2008) standards for discharge into land for irrigation, and industrial cooling prescribe pH = 5.5–9.0, TSS < 200 mg/L, BOD5 < 100 mg/L, O&G < 10 mg/L, arsenic < 0.2 mg/L, electrical conductivity < 2,250 μS/cm, sodium adsorption ratio (SAR) < 26, boron < 2 mg/L.

In the two-stage sand filtration study, treated mean greywater TSS, BOD5, O&G, and EC25 were 15 mg/L, 28 mg/L, 8 mg/L, and 465 μS/cm, respectively. The median pH was 7.24. In alum-treated GW, the mean TSS, BOD5, O&G, and EC25 were 5 mg/L, 11 mg/L, 1 mg/L, and 495 μS/cm, respectively. In PAC-treated GW, the mean TSS, BOD5, O&G, and EC25 were 4 mg/L, 7 mg/L, 2 mg/L, and 518 μS/cm, respectively. In FeCl3-treated GW, the mean TSS, BOD5, O&G, and EC25 were 7 mg/L, 15 mg/L, 3 mg/L, and 522 μS/cm, respectively. Boron and arsenic were <0.01 mg/L. Residual chlorine was >1 mg/L in disinfected effluents from all the four treatment options. Hence, treated GW from all the four options satisfied standards for restricted access area irrigation, construction, and industrial cooling as per USEPA (2004), WHO (2006), and CPCB (1993, 2008) standards.

### Comparison of treatment options using AHP

The hierarchy structure of the AHP model used in the present study is shown in Figure 12. Treatment cost (TRC), the ability of the treatment option to work robustly (AB), and compliance of treated GW with reuse standards (CS) were the main criteria of comparison. Reuse parameters prescribed in the reuse standards referred to were considered as sub-criteria of CS. CS criteria include the sub-criteria pH, turbidity (TUR), EC25, TSS, O&G, BOD5, boron, SAR, and FC. Concentrations of arsenic were the same (i.e., <0.01 mg/L) in raw as well as treated GW, so it was not included in CS for decision-making.
Figure 12

Hierarchy structure of the AHP model used in the present study.

Figure 12

Hierarchy structure of the AHP model used in the present study.

#### Pair-wise comparison matrix and criteria weights

The pair-wise comparisons find the relative importance of the criteria, which are rated by the nine-point scale as given in Table 2. Table 6 shows the pair-wise comparison matrix (Mat A1) and weights of the main criteria. Table 7 shows the pair-wise comparison matrix and local weights for sub-criteria of CS.

Table 6

Pair-wise comparison matrix and weights of the main criteria

Mat A1

Mat A2 Mat A3 Mat A4
TRC AB CS Geometric mean Weight = A1*A2 = A3/A2
TRC 1.000 0.857 0.667 0.830 0.273 0.818 2.997
AB 1.200 1.000 0.778 0.968 0.318 0.955 3.002
CS 1.500 1.286 1.000 1.245 0.409 1.227 3.001
Mat A1

Mat A2 Mat A3 Mat A4
TRC AB CS Geometric mean Weight = A1*A2 = A3/A2
TRC 1.000 0.857 0.667 0.830 0.273 0.818 2.997
AB 1.200 1.000 0.778 0.968 0.318 0.955 3.002
CS 1.500 1.286 1.000 1.245 0.409 1.227 3.001

Mat, matrix.

Table 7

Pair-wise comparison matrix and local weights for sub-criteria of CS

pH TUR EC25 TSS O&G BOD5 Boron SAR ΔFC Weight
pH 1.000 1.400 1.750 1.167 1.167 1.000 1.400 1.400 1.000 0.135
TUR 0.714 1.000 1.250 0.833 0.833 0.714 1.000 1.000 0.714 0.096
EC25 0.571 0.800 1.000 0.667 0.667 0.571 0.800 0.800 0.571 0.077
TSS 0.857 1.200 1.500 1.000 1.000 0.857 1.200 1.200 0.857 0.115
O&G 0.857 1.200 1.500 1.000 1.000 0.857 1.200 1.200 0.857 0.115
BOD5 1.000 1.400 1.750 1.167 1.167 1.000 1.400 1.400 1.000 0.135
Boron 0.714 1.000 1.250 0.833 0.833 0.714 1.000 1.000 0.714 0.096
SAR 0.714 1.000 1.250 0.833 0.833 0.714 1.000 1.000 0.714 0.096
FC 1.000 1.400 1.750 1.167 1.167 1.000 1.400 1.400 1.000 0.135
pH TUR EC25 TSS O&G BOD5 Boron SAR ΔFC Weight
pH 1.000 1.400 1.750 1.167 1.167 1.000 1.400 1.400 1.000 0.135
TUR 0.714 1.000 1.250 0.833 0.833 0.714 1.000 1.000 0.714 0.096
EC25 0.571 0.800 1.000 0.667 0.667 0.571 0.800 0.800 0.571 0.077
TSS 0.857 1.200 1.500 1.000 1.000 0.857 1.200 1.200 0.857 0.115
O&G 0.857 1.200 1.500 1.000 1.000 0.857 1.200 1.200 0.857 0.115
BOD5 1.000 1.400 1.750 1.167 1.167 1.000 1.400 1.400 1.000 0.135
Boron 0.714 1.000 1.250 0.833 0.833 0.714 1.000 1.000 0.714 0.096
SAR 0.714 1.000 1.250 0.833 0.833 0.714 1.000 1.000 0.714 0.096
FC 1.000 1.400 1.750 1.167 1.167 1.000 1.400 1.400 1.000 0.135

λmax ≈ 9, M = 9; CI ≈ 0; RI9 = 1.45 (Saaty 1980); CR ≈ 0.

The consistency ratio, which is based on the consistency index, determines the acceptance of the weights. This is one of the essential checks in the AHP method which aims to eliminate the possibility of inconsistency in the criteria weights. The consistency of the judgment matrix was tested by calculation of the consistency index (CI) as Equation (6):
6
where λmax is the maximum Eigen value of the matrix and could be calculated from the average of matrix A4 (refer to Table 6), and M is the order of the matrix (here, λmax ≈ 3 and M = 3). In the present study, the exact values were used in comparing attributes in the relative importance matrix; therefore CI was zero.
The consistency ratio (CR) was calculated as Equation (7):
7
where RI is the random index which depends upon the size of relative importance matrix. Here, for the main criteria, CI ≈ 0 and RI3 = 0.52 (Saaty 1980). Hence, CR ≈ 0.

Saaty (1980) has suggested a CR ≤ 0.10 for concluding the consistency of the pairwise matrix and validating the weights. Here, CR ≈ 0 indicates that the matrix was consistent and the weights were valid.

#### Global priority weights

Since there were no sub-criteria in the TRC and AB their global priority weights (GPW) were as calculated in Table 6. Criterion CS has sub-criteria, and needs conversion of weights obtained in Table 7 by multiplying their criteria weight. Table 8 presents all the criteria and their global priority weights.

Table 8

Global priority weights of criteria and sub-criteria

Criteria TRC AB CS

Sub-criteria     pH TUR EC25 TSS O&G BOD5 Boron SAR FC
GPW 0.273 0.318 0.055 0.039 0.031 0.048 0.048 0.055 0.039 0.039 0.055
Criteria TRC AB CS

Sub-criteria     pH TUR EC25 TSS O&G BOD5 Boron SAR FC
GPW 0.273 0.318 0.055 0.039 0.031 0.048 0.048 0.055 0.039 0.039 0.055

#### Comparison of reuse options on basis of effluent quality

The performance of a system can be evaluated from the quality of effluent produced by it. Reuse standards specify the limits of parameters which are mostly quantitative. In the present scenario, while comparing two options, the option producing the effluent having the lowest concentration of parameters (except pH) can be regarded as better than the other one. However, this comparison may be biased if both the options compared were not fed with the same raw wastewater. In such situations, when the raw wastewater quality is different in each option, comparison of the options based on the removal of the parameters (% removal or log removal) may be a better approach. In the present study, the treatment options were compared by both ways for the study purpose.

#### Greywater reuse options and criteria

The greywater reuse options and criteria used in decision-making are presented in Table 9. Treatment cost (TRC) indicates the cost of treatment of GW per million liters/d (MLD) flow. Treatment cost is a theoretical estimation that includes cost of construction, chemicals, energy, and manpower on an annualized basis. Treatment units were designed for a flow of 240 L/d, keeping in view the test conditions in the present study, and the TRC was worked out. Further, the TRC was converted to per MLD flow.

Table 9

Greywater reuse options and criteria used for comparison using AHP

TRC AB ΔpH TUR EC25 TSS O&G BOD5 Boron SAR FC
Option US$/MLD – – NTU μS/cm mg/L mg/L mg/L mg/L – Log10# Option 1 532 4.0 0.24 4.5 465 15 28 0.01 2.00 0.301 Option 2 9,249 6.0 0.44 0.9 495 11 0.01 1.50 0.301 Option 3 9,897 7.5 0.37 0.6 520 0.01 2.35 0.301 Option 4 3,703 7.0 0.54 2.3 523 15 0.01 2.15 0.301 TRC AB ΔpH TUR EC25 TSS O&G BOD5 Boron SAR FC Option US$/MLD – – NTU μS/cm mg/L mg/L mg/L mg/L – Log10#
Option 1 532 4.0 0.24 4.5 465 15 28 0.01 2.00 0.301
Option 2 9,249 6.0 0.44 0.9 495 11 0.01 1.50 0.301
Option 3 9,897 7.5 0.37 0.6 520 0.01 2.35 0.301
Option 4 3,703 7.0 0.54 2.3 523 15 0.01 2.15 0.301

# MPN/100 mL.

The ability of the treatment option to work robustly (AB) is a qualitative term. The qualitative judgments on AB were made and were transformed to crisp numbers between 1 and 9 using Saaty's scale (see Table 2). Average optimum doses of FeCl3 and PAC required were less than the alum dose (Ghaitidak & Yadav 2014). Less coagulant dose leads to a low cost of handling and less sludge production. Therefore, FeCl3 and PAC were rated slightly higher than alum. PAC can work better on a wide range of pH compared to FeCl3, hence the rating of PAC was kept slightly higher than that of FeCl3. Any variation in the GW quality can be easily tackled by adjusting the coagulant dose, hence the AB rating of option 2 to 4 was higher than that of option 1.

Criteria CS included pH, turbidity (TUR), EC25, TSS, O&G, BOD5, boron, SAR, and FC. Sub-criterion pH was transformed to ΔpH as Equation (8):
8
pH is a reuse standard that is preferred to be neither a minimum nor maximum. All the referred standards prescribe a range, that is, 6–9 (USEPA 2004); 5.5–9 (CPCB 1993, 2008). Basically, the pH of water varies from 0 to 14. Water at pH 7 is neutral. Therefore, Equation (8) measures the deviation of pH from 7 (which gives it a non-negative value). With this transformation, the criterion pH was renamed ΔpH.

#### Normalization of the data

When the dimensions of the criteria are different, normalization of the data is required. Normalization is the process of bringing all the criteria to a comparable platform by making them dimensionless. Table 10 presents normalized data of criteria mentioned in Table 9. Criterion AB was beneficial and all other criteria/sub-criteria were non-beneficial. Beneficial criteria were the cases of maximization and the non-beneficial criteria were the cases of minimization. Normalized values of a beneficial criterion were obtained by dividing each element of the criterion in the decision table by its largest element. In the case of non-beneficial criteria, the reciprocal of each element of the criterion in the decision table was multiplied by its smallest element.

Table 10

Normalization of the data

Option TRC AB ΔpH TUR EC25 TSS O&G BOD5 Boron SAR FC
Option 1 1.00 0.53 1.00 0.14 1.0 0.28 0.13 0.26 1.00 0.75 1.00
Option 2 0.06 0.80 0.55 0.68 0.94 0.84 1.00 0.67 1.00 1.00 1.00
Option 3 0.05 1.00 0.65 1.00 0.89 1.00 0.48 1.00 1.00 0.64 1.00
Option 4 0.14 0.93 0.44 0.27 0.89 0.60 0.33 0.49 1.00 0.70 1.00
Option TRC AB ΔpH TUR EC25 TSS O&G BOD5 Boron SAR FC
Option 1 1.00 0.53 1.00 0.14 1.0 0.28 0.13 0.26 1.00 0.75 1.00
Option 2 0.06 0.80 0.55 0.68 0.94 0.84 1.00 0.67 1.00 1.00 1.00
Option 3 0.05 1.00 0.65 1.00 0.89 1.00 0.48 1.00 1.00 0.64 1.00
Option 4 0.14 0.93 0.44 0.27 0.89 0.60 0.33 0.49 1.00 0.70 1.00

#### Selection index and ranking of options

In overall comparison (i.e., considering criteria TRC, AB, and CS), option 1 (TSF) was the optimal option with the highest SI (see Table 11). Option 3 (PAC treatment), option 2 (alum treatment), and option 4 (FeCl3 treatment) were ranked as 2, 3, and 4, respectively. Finally, the selection string was option 1–3–2–4. Considering only the criterion of CS, the optimal option was option 3 (PAC coagulation). This indicates that the PAC coagulation produced the best quality effluent. The selection string was option 3–2–4–1.

Table 11

Selection index and ranking of the options on the basis of effluent quality

Criteria (TRC + AB + CS)

Criterion (CS)

Option SI Rank SI Rank
Option 1 (TSF) 0.692 0.610
Option 2 (alum treatment, alum dose 416 ± 44 mg/L) 0.613 0.839
Option 3 (PAC treatment, PAC dose 279 ± 40 mg/L) 0.680 0.849
Option 4 (FeCl3 treatment, FeCl3 dose 141 ± 12 mg/L) 0.591 0.625
Criteria (TRC + AB + CS)

Criterion (CS)

Option SI Rank SI Rank
Option 1 (TSF) 0.692 0.610
Option 2 (alum treatment, alum dose 416 ± 44 mg/L) 0.613 0.839
Option 3 (PAC treatment, PAC dose 279 ± 40 mg/L) 0.680 0.849
Option 4 (FeCl3 treatment, FeCl3 dose 141 ± 12 mg/L) 0.591 0.625

### Comparison of reuse options on the basis of removal of parameters

#### Greywater reuse options and criteria

Treatment cost (TRC) and the ability of the treatment option to work robustly (AB) were the same as previously. The removals of parameters were calculated referring to the mean GW characteristics in Table 5. Greywater reuse options and criteria used in comparison using the AHP, on the basis of removal efficiency, are given in Table 12.

Table 12

Greywater reuse options and criteria used in comparison using AHP

TRC AB ΔpH TUR EC25 TSS O&G BOD5 Boron SAR ΔFC
Option US$/MLD – – log10 Option 1 532 4.0 0.24 94.60 4.71 83.70 77.14 73.58 85.90 15.25 4.99 Option 2 9,249 6.0 0.44 99.00 − 9.03 93.98 96.67 88.42 84.10 03.23 5.10 Option 3 9,897 7.5 0.37 99.27 − 4.97 95.53 93.26 93.49 87.30 24.19 4.85 Option 4 3,703 7.0 0.54 97.04 − 4.40 92.39 90.32 85.71 85.70 12.24 5.11 TRC AB ΔpH TUR EC25 TSS O&G BOD5 Boron SAR ΔFC Option US$/MLD – – log10
Option 1 532 4.0 0.24 94.60 4.71 83.70 77.14 73.58 85.90 15.25 4.99
Option 2 9,249 6.0 0.44 99.00 − 9.03 93.98 96.67 88.42 84.10 03.23 5.10
Option 3 9,897 7.5 0.37 99.27 − 4.97 95.53 93.26 93.49 87.30 24.19 4.85
Option 4 3,703 7.0 0.54 97.04 − 4.40 92.39 90.32 85.71 85.70 12.24 5.11

The sub-criterion pH was transformed to ΔpH using Equation (8). The percentage removal (%R) of parameters TUR, EC25, TSS, O&G, BOD5, boron, and SAR were calculated as Equation (9):
9
where CR = concentration of the parameter in raw GW; CT = concentration of the parameter in treated GW.
Removal of parameter FC was calculated using Equation (10):
10
where ΔFC = log removal; FCi = FC count in raw GW (MPN/100 mL); FCe = FC count in treated GW (MPN/100 mL).

#### Normalization of the data

TRC and ΔpH were non-beneficial criteria, and all other criteria/sub-criteria were beneficial. Non-beneficial criteria were the cases of minimization, and beneficial criteria were the cases of maximization. Data were normalized as mentioned earlier.

#### Selection index and ranking of options

In overall comparison (i.e., considering the criteria of TRC, AB, and CS), option 1 obtained the highest SI (see Table 13). Option 1 (TSF) was the best management option, on the basis of removal of parameters, for treating GW compared to the other three options in the present study. Option 3 (PAC treatment), option 4 (FeCl3 treatment), and option 2 (alum treatment) were ranked 2, 3, and 4, respectively. Finally, the selection string was option 1–3–4–2. Considering only criterion CS the selection string was the same as the overall comparison (i.e., option 1–3–4–2).

Table 13

Selection index and ranking of the option on the basis of removal of parameters

Criteria (TRC + AB + CS)

Criteria (CS)

Option SI Rank SI Rank
Option 1 (TSF) 0.806 0.889
Option 2 (alum treatment, alum dose 416 ± 44 mg/L) 0.522 0.617
Option 3 (PAC treatment, PAC dose 279 ± 40 mg/L) 0.652 0.783
Option 4 (FeCl3 treatment, FeCl3 dose 141 ± 12 mg/L) 0.623 0.702
Criteria (TRC + AB + CS)

Criteria (CS)

Option SI Rank SI Rank
Option 1 (TSF) 0.806 0.889
Option 2 (alum treatment, alum dose 416 ± 44 mg/L) 0.522 0.617
Option 3 (PAC treatment, PAC dose 279 ± 40 mg/L) 0.652 0.783
Option 4 (FeCl3 treatment, FeCl3 dose 141 ± 12 mg/L) 0.623 0.702

## CONCLUSIONS

Around 250 L GW was treated per day on-site. The findings of the study can be useful for treatment of GW on a household basis. In the two-stage sand filtration study, treated mean greywater TSS, BOD5, O&G, and EC25 were 15 mg/L, 28 mg/L, 8 mg/L, and 465 μS/cm, respectively. The median pH was 7.24. In alum-treated GW, the mean TSS, BOD5, O&G, and EC25 were 5 mg/L, 11 mg/L, 1 mg/L, and 495 μS/cm, respectively. In PAC-treated GW the mean TSS, BOD5, O&G, and EC25 were 4 mg/L, 7 mg/L, 2 mg/L, and 518 μS/cm, respectively. In FeCl3-treated GW, the mean TSS, BOD5, O&G, and EC25 were 7 mg/L, 15 mg/L, 3 mg/L, and 522 μS/cm, respectively. Boron and arsenic were <0.01 mg/L. Residual chlorine was >1 mg/L in all the tested effluents. Fecal coliforms were <2 MPN/100 mL and E. coli <2 CFU/100 mL after chlorination (contact time 30 min). Hence, treated GW from all the four options satisfied standards for restricted access area irrigation, construction, and industrial cooling as per USEPA (2004), WHO (2006), and CPCB (1993, 2008) standards.

The study provides a step-by-step solution for ranking of options using the AHP. The options thus ranked can be useful for choosing appropriate technology for GW reuse. Comparison of the options on the basis of effluent quality resulted in TSF being the optimal option. PAC treatment, alum treatment, and FeCl3 treatment were ranked as 2, 3, and 4, respectively. Comparison on the basis of removal of parameters resulted in TSF being the optimal option. PAC treatment, FeCl3 treatment, and alum treatment were ranked as 2, 3, and 4, respectively.

In two-stage sand filtration, the regeneration of filters at 30–40 days' intervals may be a major limitation for implementation of this study. The operational aspect also needs close supervision of flow rates, application of coagulant dose, cleaning of screening mesh, and cleaning of the mini coarse sand filter on a daily basis. Investigations on GW from different sources such as kitchen and laundry; testing of filters with variable effective sand size and depth; and application of different multi-criteria techniques, e.g., TOPSIS, fuzzy AHP, etc., for prioritizing the options may be a further scope of research.

## ACKNOWLEDGEMENT

The authors are grateful to the Indian Institute of Technology (IIT) Bombay for analysis of metals and ground elements in their Sophisticated Analytical Instrument Facility (SAIF) laboratory. The authors are grateful to the Director, SVNIT, Surat for his support during the research.

## REFERENCES

REFERENCES
Al-Hamaiedeh
H.
Bino
M.
2010
.
Desalination
256
,
115
119
.
APHA
2005
Standard Methods for the Examination of Water and Wastewater
, 21st edn.
American Public Health Association, American Water Works Association and Water Environment Federation
,
Washington, DC
,
USA
.
CAWST
2008
Biosand Filter Manual
.
Centre for Affordable Water and Sanitation Technology
,
Calgary, Alberta
,
.
Chen
Z.
Ngo
H. H.
Guo
W. S.
Listowski
A.
O'Halloran
K.
Thompson
M.
Muthukaruppan
M.
2012
.
Sci. Total Environ.
438
,
59
65
.
Christova-Boal
D.
Eden
R. E.
McFarlane
S.
1996
.
Desalination
106
,
391
397
.
CPCB
1993
The Environment (protection) Rules, 1986, Schedule VI (Amended 1993)
.
Central Pollution Control Board
,
New Delhi
,
India
.
CPCB
2008
Guidelines for Water Quality Management, Environmental Standards-Water Quality Criteria
.
Central Pollution Control Board
,
New Delhi
,
India
.
CPHEEO
1999
Manual on Water Supply and Treatment
.
Central Public Health and Environmental Engineering Organization
,
New Delhi
,
India
.
Finley
S.
Barrington
S.
Lyew
D.
2009
.
Water Air Soil Pollut.
199
,
235
245
.
Friedler
E.
Alﬁya
Y.
2010
.
Water Sci. Technol.
62
(
10
),
2357
2363
.
Ghaitidak
D. M.
K. D.
2013
.
Environ. Sci. Pollut. Res.
20
,
2795
2809
.
Ghaitidak
D. M.
K. D.
2014
An investigation on greywater treatment options for reuse: ranking by analytic hierarchy process
.
Environ. Eng. Manage. J.
.
Ghaitidak
D. M.
K. D.
2015
.
Desalin. Water Treat.
54
(
9
),
2410
2421
.
Gross
A.
Shmueli
O.
Ronen
Z.
Raveh
E.
2007
.
Chemosphere
66
,
916
923
.
Ishizaka
A.
Labib
A.
2011
.
Expert Syst. Appl.
38
,
14336
14345
.
Karabelnik
K.
Koiv
M.
Kasak
K.
Jenssen
P. D.
Mander
U.
2012
.
Ecol. Eng.
49
,
84
92
.
Kariuki
F. W.
Kotut
K.
Nganga
V. G.
2011
.
Open Environ. Eng. J.
4
,
32
39
.
Katukiza
A. Y.
Ronteltap
M.
Niwagaba
C. B.
Kansiime
F.
Lens
P. N. L.
2014
.
J. Environ. Manage.
133
,
258
267
.
Mandal
D.
Labhasetwar
P.
Dhone
S.
Dubey
A. S.
Shinde
G.
Wate
S.
2011
.
Resour. Conserv. Recy.
55
,
356
361
.
Metcalf, Eddy, Inc.
2003
Wastewater Engineering: Treatment and Reuse
.
Tata McGraw Hill Education Private Limited
,
New Delhi
,
India
.
Rao
R. V.
2007
Decision Making in the Manufacturing Environment: Using Graph Theory and Fuzzy Multiple Attribute Decision Making Methods
.
Springer-Verlag
,
London
,
UK
.
Saaty
T. L.
1980
The Analytic Hierarchy Process: Planning, Priority Setting, Resource Allocation
.
McGraw-Hill
,
New York
,
USA
.
Saaty
T. L.
1990
.
Eur. J. Oper. Res.
48
,
9
26
.
USEPA
2004
Guidelines for Water Reuse
,
Report EPA/625/R-04/108
.
United States Environmental Protection Agency
,
Washington, DC
,
USA
.
Vaidya
O. S.
Kumar
S.
2006
.
Eur. J. Oper. Res.
169
,
1
29
.
WHO
2006
WHO Guidelines for the Safe Use of Wastewater, Excreta and Greywater, Volume 4, Excreta and Greywater Use in Agriculture
.
World Health Organization
,
Geneva
,
Switzerland
.