Performance evaluation of fabric aided slow sand filter in drinking water treatment

Transcription

1 Performance evaluation of fabric aided slow sand filter in drinking water treatment 703 Pulin Kumar Mondal, Rajesh Seth, and Nihar Biswas Abstract: In this study, an assessment of the performance of slow sand filter (SSF) aided with non-woven fabric (NWF) was carried out. Several laboratory-scale SSF columns were tested with simulated raw water containing varying levels of turbidity and total organic carbon (TOC). The results show that in filters with NWF, the fabric layers captured most of the incoming solids and extended the filter run time for the sand bed. The run time for the sand bed increased with the increasing of the fabric thickness from 8.9 to 44.5 mm. Turbidity, TOC, and bacterial removal efficiencies of the filters with fabric were comparable to that without fabric and representing conventional SSF. The study thus demonstrates that operation of SSF with NWF can be a feasible option for simplifying the operation of and extending the viability of the SSF process to a wider range of raw water turbidity values than that considered economical for conventional SSF. Key words: slow sand filter, non-woven fabric, filter run time, sand bed protection, filter cleaning, drinking water treatment, filter head loss. Résumé : Cette étude évalue le rendement d un filtre à sable lent (SSF) aidés d un textile non-tissé. Plusieurs colonnes de SSF à l échelle du laboratoire ont été mises à l épreuve en utilisant de l eau brute simulée présentant divers niveaux de turbidité et de carbone organique total (COT). Les résultats montrent que, dans les filtres avec textile non-tissé, les couches de textile captaient la majorité des solides entrant et prolongeaient le temps d utilisation de la portion lit de sable du filtre. Le temps d utilisation du lit de sable augmente avec l augmentation de l épaisseur du textile de 8,9 mm à 44,5 mm. La turbidité, le COT et l efficacité des filtres avec textiles à éliminer les bactéries étaient similaires à ceux des filtres sans textile représentant les SSF conventionnels. L étude démontre que le fonctionnement du SSF avec textile non-tissé peut être une option viable pour simplifier le fonctionnement et, ainsi, accroître la viabilité du procédé SSF à une plus grande gamme de valeurs de turbidité d eaux brutes que celles qui peuvent être considérées rentables pour les SSF conventionnels. Mots-clés :filtre à sable lent, textile non-tissé, temps d utilisation du filtre, protection du lit de sable, nettoyage de filtre, traitement de l eau potable, perte de charge des filtres. [Traduit par la Rédaction] Introduction Slow sand filter (SSF), being simple in technology and operation, is considered one of the more suitable low cost treatment technologies in treating drinking water, particularly for small community water supplies. Slow sand filter is the earliest technology in municipal water treatment and has been used since its introduction by James Simpson and Robert Thom in 1829 (Ellis 1985). Due to the occurrence of both physical and biological treatment processes, SSF is effective in removal of both turbidity and microorganisms. Received 11 September Revision accepted 23 March Published on the NRC Research Press Web site at jees.nrc.ca on 7 November P.K. Mondal, 1 R. Seth, 2 and N. Biswas. Department of Civil and Environmental Engineering, University of Windsor, B10 Essex Hall, 401 Sunset Avenue, Windsor, ON N9B 3P4, Canada. Written discussion of this article is welcomed and will be received by the Editor until 31 March Present address: Department of Civil Engineering, University of Toronto, Toronto, ON, Canada. 2 Corresponding author ( rseth@uwindsor.ca). Conventional SSFs typically achieve effluent turbidity of <1 nephelometric turbidity unit (NTU) (Collins et al. 1991). Researchers over the last few decades have also shown that SSF can achieve Giardia cysts and Cryptosporidium oocysts removal of 2 to 4 log (Rachwal et al. 1996; Palmateer et al. 1999). The United Nations recommends slow sand filtration as an important, reliable and cost-effective process for drinking water treatment in developing countries, especially for smaller water systems (Hendricks et al. 1991). Even though SSF is considered suitable and economical, its application is recommended only for source waters with low turbidity (<5 NTU) (Cleasby 1991). Higher turbidity in source water and excessive proliferation of algae during the summer time increase the filter cleaning frequency by clogging the filter quickly, and thus the operation cost is increased (Montiel et al. 1988; Hendricks et al. 1991). Moreover, frequent filter cleaning by scraping the top sand layer removes a large part of the active microorganisms population from the schmutzdecke and also disturbs the sand bed below the schmutzdecke. Schmutzdecke is defined as a layer of material, both deposited and synthesized, on top of the filter bed that causes head loss disproportionate to its thickness and it is characterized usually as a gelatinous mat in which microorganisms thrive and cause a major portion of the removal that occurs (AWWA 1991). The sand J. Environ. Eng. Sci. 6: (2007) doi: /s07-019

2 704 J. Environ. Eng. Sci. Vol. 6, 2007 Fig. 1. Schematic of the experimental setup. bed disturbance during scraping has been shown to have a damaging effect on the performance of SSF (Huisman and Wood 1974; Bellamy et al. 1985; Ellis 1985). Several modifications have been developed and implemented with SSF to deal with water quality issues not adequately handled by conventional SSF. Roughing filters and filter harrowing are two such modifications that allow SSF to be used for waters with higher than acceptable turbidity or algal content (Collins et al. 1991). Roughing filters are used as pretreatment for the SSF influent waters and can achieve significant reduction in turbidity, coliform bacteria, and algal content. Filter harrowing is a method of cleaning SSF where the surface of the clogged sand bed is harrowed with combtooth harrow and simultaneously the surface sumps are kept open for the drainage of supernatant water and loosened debris. However, both of these modifications increase the complexity and the cost of construction and operation of the filters (Collins et al. 1991; Tanner and Ongerth 1990). The use of synthetic NWF as filter mat above the sand bed may be another option that may allow SSF to deal with high raw water turbidity, with possibly little increase in complexity or cost of the treatment process. The feasibility of the use of NWF in SSF has been examined in a few limited studies (e.g., Graham and Mbwette 1991; Mbwette 1989). These studies have indicated that NWF can capture the incoming solids and reduce the solid loadings to the sand bed. The use of NWF was shown to increase the filter run time to a factor of 1.5 to 2 as compared to the control filter (without NWF), with no significant effect on the overall removal efficiencies of turbidity, total organic carbon (TOC), and total coliform. Klein and Berger (1994) have reported the application of NWF at an artificial groundwater recharge SSF plant in Zurich for the protection of sand bed against sunlight and algae. They reported that NWF increased the filter run time by a factor of 10 when compared to a similar facility with no NWF. Despite these studies, the effect of process parameters on the performance of the modified filter and its wider applicability to surface water of varying characteristics have not been fully evaluated. The present study was carried out to examine the effect of the use of NWF in SSF on capture of suspended solids, filter run time, and filter treatment efficiencies. Influent water quality variables considered in this study include low and high turbidity, and low and high organic content. Materials and methods Filter design and setup A laboratory scale setup was used to conduct the experiments for this study. Figure 1 shows the schematic of the setup. The setup included a mixing tank (450 L polypropylene tank), a feed tank (600 L), influent distribution manifold, filter columns, and an overflow tank (450 L). The mixing tank and feed tank were placed at the second floor of a laboratory and the filter columns were placed at the first floor of the laboratory. This arrangement provided a gravity flow of the feed to the filter columns which were controlled by valves at the distribution manifold. Figure 2 shows a picture of a portion of the laboratory setup and a close up view of the placement of NWF in the filter column. Four laboratory-scale SSF columns (made of cast acrylic tube of 150 mm internal diameter) were built based on typical SSF design considerations. All four filter columns contained 0.9 m (depth) of fine silica filter sand (silica content of >99.5%) obtained from Northern Gravel Company, Iowa, USA. Particle size analysis of the sand was done in the laboratory and the effective size and uniformity coefficient were determined to be 0.28 mm and 1.88, respectively. This filter sand was supported in the columns by a 0.1 m gravel layer and a metal strainer. Desirable characteristics of NWF for use in SSF are porosity of >80% and specific surface area (SSA) of to m 2 /m 3 (Graham and Mbwette 1991). TC Mirafi 1 S1600 from Ten Cate Nicolon Company, USA was selected and used in this study. This fabric is a non-woven geotextile composed of polypropylene fibre, which forms a stable network such that the fibres retain their relative position during the filter run. Polypropylene is inert to biological degrada-

3 Mondal et al. 705 Fig. 2. Experimental setup: (a) filter columns, (b) NWF position in the filter. tion and resistant to naturally encountered chemicals including alkalis and acids. The selected NWF has the apparent opening size of 0.15 mm and average thickness of 4.45 mm. The SSA and the porosity were calculated to be m 2 /m 3 and 87%, respectively, based on the models provided by Mbwette (1989). A brass cutting die of 150 mm inner diameter (ID) was used to cut 150 ± 0.5 mm diameter discs of the fabric for use in the laboratory SSF columns. Three filter columns (Filters 2, 3, and 4) were provided with varying thicknesses of NWF on top of the sand bed. Non-woven fabric thicknesses of 8.9, 22.3, and 44.5 mm were obtained by adding 2, 5, and 10 layers of fabric in Filters 2, 3, and 4, respectively. A rubber O-ring was provided on top of the fabric layers to prevent the water flow along the column wall and an 8 mm wide and 1 mm thick polyvinyl chloride strip was placed on top of the rubber O-ring along the column wall to keep the fabric layers in position. The fabric layers were supported by a stainless steel strainer (58% open area with 4 mm diameter perforation), which allowed a gap of 40 mm between the bottom surface of the fabric and the top surface of the sand bed for water sample collection without disturbing the filter media (Fig. 2b). One filter (Filter 1) was provided with only the sand bed (no NWF layer) to represent conventional SSF and also serve as the control filter. The depth of the supernatant water in all filters was maintained at 1.1 m by providing the influent water flow rate slightly more than the filtration rate and allowing the excess water to drain through the top drain outlet. With constant supernatant water height, the filtration rates of all the filters tended to decrease with time due to increase in head losses. The filtration rates were checked twice (or more) a day by volumetric measurement of the filter outflows, and the outlet valves were adjusted to maintain the filtration rates to within ± 10% of 0.1 m/h. Simulation of raw water Simulated surface water was used as the feed to the filters. To prepare the simulated raw water, municipal tap water available in the laboratory was aerated with compressed air and stored at laboratory temperature for more than 16 h to remove the residual chlorine. Absence of residual chlorine was checked by o-tolidine chlorine presence or absence test. Bentonite clay, obtained from Sigma Chemical Company, USA, was used to provide the desired turbidity in the raw water. According to the manufacturer, the clay had a maximum particle size of 74 mm, with 90% passing size (d 90 ) and 10% passing size (d 10 ) of 7.0 and 2.7 mm, respectively. Algae collected from a local pond was mixed with Euglena (unicellular flagellate) and five other representatives of green algae (Chlorella, Scenedesmus, Selenastrum, Ulothrix, and Volvox), which were obtained from Carolina Biological Supply, USA. The mixture of the algae was cultured in the laboratory by using Alga-Gro 1 Freshwater Medium (Carolina Biological Supply, USA). A 300 to 500 ml of the grown mix algae culture was added to 400 L of the simulated raw water to obtain chlorophyll a content of about 3to5mg/L. About 2.5 to 5.0 ml of primary settled municipal wastewater effluent was added per 1000 ml of raw water, to simulate the microbial community present in typical surface raw waters. A survival test for microorganisms, measured as total coliforms using the membrane filter procedure of Standard Methods for the Examination of Water and Wastewater (APHA, AWWA and WEF 1998), was carried out after mixing the wastewater effluent to the raw water. The low survival rate of the total coliforms in the simulated raw water was an issue during the initial part of the experiments (Phase 1 and part of Phase 2). The test revealed that even though the presence of chlorine in the stored and aerated tap water could not be detected by the o-tolidine test for chlorine presence or absence, no coliforms were detected

4 706 J. Environ. Eng. Sci. Vol. 6, 2007 Table 1. Influent water characteristics. (Phase 1: high turbidity, low TOC test). Parameter Average* ± standard deviation Temperature (8C) 20.5 ± 1.1 DO { (mg/l) 8.8 ± 0.5 ph 7.4 ± 0.3 Turbidity (NTU) 11.0 ± 0.9 Chlorophyll a (mg/l) 3.0 ± 0.8 TOC (mg/l) 0.4 ± 0.1 *Number of samples = 21 over 20 days { DO, dissolved oxygen. in the raw water 8 h after adding the wastewater effluent. However, some growth of non-coliform group of microorganisms was observed in the test plates, which confirmed the presence of microbial activity in the filters during this period. The problem of coliform survival was effectively overcome by filtering the tap water used to prepare simulated raw water through activated carbon, from day 36 of the Phase 2 experiment. Experimental plan The experiments were conducted in two different phases with varying raw water characteristics. During the Phase 1 experiment (denoted as high turbidity, low TOC test), raw water of high turbidity (9 to 12 NTU) and low TOC (<1 mg/l) content was used. Phase 2 experiment (denoted as low turbidity, low and high TOC test) was conducted with raw water of low turbidity (1 to 2 NTU). Total organic carbon content in this phase was maintained at a low level (<1 mg/l) up to 21 days of filter operation, after which it was increased to and maintained at 5 to 9 mg/l. Chlorophyll a and coliform bacteria (total coliforms and E. coli) levels were intended to be kept similar for both phases of the experiments. Because of the survival issues with the coliform bacteria, total coliforms and E. coli measurements have not been reported for Phase 1 and the first 35 days of Phase 2 experiments. A summary of the simulated influent water characteristics during the two phases are provided in Table 1 (Phase 1) and Table 2 (Phase 2). The turbidity during Phase 1 (9 to 12 NTU) was about twice the maximum recommended value of 5 NTU for SSF. The average chlorophyll a content of 3 to 5 mg/l in simulated raw water in both phases was within the typical recommended value (<5 mg/l) (Cleasby 1991). Since the chlorophyll a represents approximately 1.5% of the dry mass of planktonic biomass (APHA, AWWA and WEF 1998), the daily average planktonic biomass concentration in the feed water was estimated to be 0.20 to 0.27 mg/l. During Phase 2, TOC content of the raw water was increased to 6.8 ± 1.5 mg/l from day 22 by adding D-glucose (CAS ) obtained from Fisher Scientific Company, Nepean, Ontario. All the filters were operated for 20 days in Phase 1 and 58 to 60 days in Phase 2, after which the required flow rate of 0.1 m/h could not be maintained in most of the filters. At the end of the Phase 1 experiment, the removed fabric layers from all of the filters were first soaked in tap water for 2 to 3 h and then rinsed by applying tap water with moderate flow from the reverse side of the fabric layer. In addition, 25 mm of the sand bed was scraped off from all the filters, Table 2. Influent water characteristics. (Phase 2: low turbidity, low and high TOC test). Parameter Average ± standard deviation Temperature (8C) 21.9 ± 1.1 (n* = 62) DO { (mg/l) 8.5 ± 0.9 (n = 62) ph 7.2 ± 0.3 (n = 62) Turbidity (NTU) 1.7 ± 0.2 (n = 62) Chlorophyll a (mg/l) 3.9 ± 1.3 (n = 25) TOC (mg/l) 0.4 ± 0.1 (n = 22; d { = 0 to 21) 6.8 ± 1.5 (n = 40; d = 22 to 60) Total coliform 5910 ± 2288 (n = 13; d = 36 to 60) (CFU /100 ml) E. coli (CFU/100 ml) 410 ± 184 (n = 13; d = 36 to 60) *n = number of samples. { DO, dissolved oxygen. { d = day. CFU, colony forming units. placed in a beaker and washed repeatedly with tap water. The fabric layers and sand were then replaced back into the respective filters and the clean bed head loss was measured at a flow rate of 0.1 m/h by using dechlorinated tap water. At the end of the Phase 2 experiment, one layer of fabric was removed from each of Filters 2 to 4 and the head loss across the remaining fabric layers was measured at a flow rate of 0.1 m/h using dechlorinated tap water. Parameter measurements and analytical methods The performance of the filters with and without NWF was monitored through the observation of filter head loss development, particle deposition, filtrate turbidity, and TOC removal. During the Phase 2 experiment, total coliform and E. coli removals were also monitored. The overall head loss and head loss across the fabric layers and the sand bed were recorded by using 3 mm ID clear acrylic piezometer tubes installed at three different locations along the filter columns (above and below fabric layers, and below the sand bed). About 300 ml of water sample was collected every day from different sampling ports for various water quality parameters analyses. All the samples were analyzed for turbidity and TOC according to the Standard Methods (APHA, AWWA and WEF 1998). Turbidity of the water samples was measured by using a Hach Ratio/XR Turbidimeter (Model 43900). Shimadzu TOC-VSH Carbon Analyzer was used for the TOC measurements, for which the method detection limit was estimated to be 0.1 mg/l. A concentration of photosynthetic pigment (chlorophyll a) was used to estimate phytoplankton biomass (algae). A standard spectrophotometric method was used for the chlorophyll a measurement. The Varian-CARY 50 Scan, UV-visible spectrophotometer, was used for the optical density measurement. After day 36 of Phase 2, the total coliforms and E. coli in raw and treated waters were enumerated simultaneously by using enzyme substrate coliform test method of the Standard Methods (APHA, AWWA and WEF 1998). This method uses the Quanti-tray/ and the Colilert 1 Test Kit from IDEXX Laboratories, Inc., USA.

5 Mondal et al. 707 Fig. 3. Variation of total head loss (NWF + sand) with time for the various filters (Phase 1: high turbidity, low TOC test). Results and discussion Head loss development The initial clean bed head losses were measured by using clean dechlorinated tap water (turbidity <0.1 NTU). The initial clean bed head losses determined at the beginning of the Phase 1 experiment were used as a reference to confirm proper cleaning of filters at the end of the filter run. At the selected filtration rate of 0.1 m/h, clean bed head losses in all of the filters were observed to be 21 to 24 mm, which were similar to the theoretical value calculated by using the empirical equation developed by Huisman and Wood (1974). The clean bed head losses across the fabric layers in Filters 2, 3, and 4 were undetectable (<1 mm). Phase 1: high turbidity, low TOC test Phase 1 experiment was conducted with high turbidity water (9 to 12 NTU). Figure 3 shows the total head loss (across fabric + sand bed) in different filters. For the first 5 days of filter operation, head losses in all the filters were less than 70 mm. After day 5, head losses started to rapidly increase in all of the filters at almost similar rates. The rate of head loss increase declined somewhat in Filter 2 after day 12. The total head loss in Filter 2 was less than that in other filters between day 12 and 20, which could not be explained. However, by day 20, head losses in all of the filters were close to or exceeded the maximum value (1000 mm) that would allow the filtration rate to be maintained at 0.1 m/h, and the filter run was terminated. Unlike the total head loss, the development of head losses across the sand beds with time were quite varied for the different filters, as shown in Fig. 4. The head loss began to increase from day 1 in Filter 1, day 5 in Filter 2, and day 11 in Filter 3. Sand bed in Filter 4 with 10 layers of fabric showed no measurable increase in head loss across the sand bed during the entire run of 20 days. On day 20 when the total head losses in all of the filters were in the range of 950 to 1050 mm, the losses across the sand beds were measured to be 1005, 506, 163, and 22 mm in Filters 1, 2, 3, and 4, respectively. In Filter 1, the total head loss was entirely contributed by the sand bed. From Figs. 3 and 4, it is seen that during the filter run, the increase in total head loss in Filter 2 was almost equally contributed by the fabric and the sand bed, while in Filter 3 the increase was mostly contributed by the fabric. In the case of Filter 4, however, the increase in total head loss was entirely contributed by the fabric. Phase 2: low turbidity, low and high TOC test Phase 2 of the experiment was conducted with low turbidity (<2 NTU) water. Figure 5 shows the total head loss in all of the filters. The total head loss was low (<100 mm) in all of the filters with fabric up to day 48, except between days 36 and 40 when the head loss increased to about 200 mm but then declined. In Filter 1, the head loss was observed to increase steadily from day 20 to reach a value of 477 mm on day 38. The head loss subsequently declined to 98 mm by day 48. A rapid increase in total head loss was observed in all the filters from day 48 onwards, and a value of about 1000 mm was reached after 58 to 60 days of operation when the run was terminated. As in Phase 1, although the total head losses were similar, the head losses across the sand beds only were quite varied for the different filters as shown in Fig. 6. Head loss across the sand bed started to increase from the second day in Filter 1, and reached a value of 1004 mm on day 58. In Filter 2, the head loss started to rise from day 46 to reach a value of 211 mm on day 60. A small but measurable increase in head loss across the sand bed was recorded for Filter 3 from day 49 to reach a maximum value of 50 mm. In Filter 4, no measurable increase in head loss across the sand bed was recorded during the entire filter run. Capture of suspended solids Phase 1: high turbidity, low TOC test In Filter 1, the deposited layer started to form after about 3 days of filter operation, which was reflected by an increase in the measured head loss. Approximately a 3 mm thick layer of deposited particles was formed on top of the

6 708 J. Environ. Eng. Sci. Vol. 6, 2007 Fig. 4. Variation of head loss across the sand bed alone with time for the various filters (Phase 1: high turbidity, low TOC test). Fig. 5. Variation of total head loss (NWF + sand) with time for the various filters (Phase 2: low turbidity, low and high TOC test) Filter 1 (0 layer) Filter 3 (5 layers) Filter 2 (2 layers) Filter 4 (10 layers) Head loss (mm) Stage 1 Stage 2 Stage 3 1 layer = 4.45 mm NWF Day sand surface at the end of 20 days of filter operation. In Filter 2, up to day 5, there was no measurable increase in head loss across the sand bed and no deposition was visible on the sand bed. This time period was up to 11 days in the case of Filter 3. After this period, a thin grey layer of particles was observed on top of the sand bed with a concomitant increase in the measured head loss, which indicated that a fraction of the incoming particles had started to escape through the fabric layers. In Filter 4 however, there was no measurable increase in head loss across the sand bed nor was there any visible deposition on the sand bed for the entire duration of the filter run. Only the top 25 mm of sand bed was seen to be playing any role in contributing to an increase in head loss and also removing turbidity in all of the filters. Once the top 25 mm of sand was scraped off and replaced after cleaning the sand at the end of filter run, the sand bed exhibited the same head loss as the original clean bed. From the above observations, it is clear that the total thickness of the fabric layer had a significant effect on the behaviour of particle capture. The particle capture efficiency increased with the increase in the number of the fabric layers. The sand beds remained visually clean longer in filters with increasing number of layers of fabric. From the visual observation and the head loss data it can be concluded that the fabric layers in all of the filters were effective in capturing almost all of the particles until the 5th day of operation. Filter 3 did not show any breakthrough until the 11th day of operation. No breakthrough of particles from the fabric layers was observed in Filter 4, which had the highest number of fabric layers, for the entire duration of the filter run (20 days). The total head loss after cleaning and replacing the fabric layers and sand media removed at the end of Phase 1 was measured to be in the range of 22 to 24 mm for dechlorinated tap water, which is similar to the clean bed head loss at the beginning of the experiment. Phase 2: low turbidity, low and high TOC test The head loss pattern during the Phase 2 experiment (Figs. 5 and 6) can be divided into 3 stages. During stage 1

7 Mondal et al. 709 Table 3. Average turbidity (NTU) of filtrates from various filters (Phase 1: high turbidity, low TOC test). Day Filter Filter 1 (0 layer*) Filter 2 (2 layers*) Filter 3 (5 layers*) Filter 4 ( layers*) *1 layer = 4.45 mm of NWF. (day 0 to 36), only Filter 1 showed a thin layer of deposited materials on the sand surface and significant amount of head loss development. No visible deposition was observed either on the fabric or sand surfaces for the other filters. This indicates that almost all of the incoming particles were captured within the fabric layers during this stage. The rapid increase in total head loss in all of the filters during stage 2 (day 36 to 44) may be due to the higher TOC content (6.8 ± 1.5 mg/l from day 22 to 60) and increased microbial activity in the raw water (from day 36; as discussed in experimental plan section). A visible white layer (possibly biogrowth) was observed on the sand (Filter 1) and the fabric layers of Filters 2 to 4. The subsequent reduction in head loss in all the filters (between day 40 and 44) could not be explained but might be due to destruction of the developed schmutzdecke by predatory action of certain microorganisms. During stage 3 (day 44 to 60), the continued accumulation of particles and biogrowth was responsible for the rapid increase in head loss observed in all of the filters. Of the total head loss ranging between 913 and 1004 mm in all the filters at the end of day 60, most of it (702, 927, and 977 mm for Filters 2, 3, and 4, respectively) was across the fabric layers in Filters 2 to 4. In contrast, the maximum increase in head loss across the sand bed was about 190 mm for Filter 2, 30 mm for Filter 3, and negligible for Filter 4. Figure 6 indicates that almost all of the incoming particles were captured in the fabric layers for 45 days in Filter 2, 48 days in Filter 3, and >58 days in Filter 4, as evidenced by no measurable increase in the head loss across the sand bed. After this time in each filter, a fraction of the suspended particles escaped the fabric layers and was captured by the sand bed, causing the head loss across the sand bed to increase. At the end of Phase 2, one layer of fabric was removed from the Filters 2 to 4 and the head loss was measured across the remaining fabric layers (1 layer in Filter 2, 4 layers in Filter 3, and 9 layers in Filter 4) by using clean dechlorinated tap water. The head loss in the remaining fabric layers was observed to be about 1 mm. This suggests that most of the incoming particles captured in the fabric layers were concentrated within the top most fabric layer. Filter run time and sand bed protection time Phase 1: high turbidity, low TOC test In this study, the laboratory setup allowed a maximum total head loss of about 1000 mm beyond which the selected Table 4. Average turbidity (NTU) of filtrates from various filters (Phase 2: low turbidity, low and high TOC test). Day Filter Filter 1 (0 layer*) Filter 2 (2 layers*) Filter 3 (5 layers*) Filter 4 (10 layers*) *1 layer = 4.45 mm of NWF. filtration rate of 0.1 m/h could not be maintained. This value was thus used to define filter run time. In Phase 1, all the filters had similar filter run times of about 20 days. In Filter 1, particles started to deposit on the sand from the first day of filter operation with a measurable increase in head loss across the sand bed. In Filter 2, the head loss across the sand was in the range of 20 to 25 mm and similar to the clean bed head loss of 22 mm up until day 5. This suggests that the presence of two layers of fabric provided almost complete protection of the sand bed against particle deposition for 5 days. This time period, i.e., the duration of the filter run for which the head loss across the sand bed is similar to the clean bed head loss, is being termed as sand bed protection time (SBPT). For Filter 2 therefore, the SBPT was 5 days. From day 6, particle deposition in the sand bed was evident from significant and continued increase in the head loss and a visible layer of particles deposited on the bed. Using the same logic, the SBPT was 11 days for Filter 3. For Filter 4, the SBPT was undetermined (>20 days). The SBPT is thus observed to increase with an increase in the thickness of fabric (varied by the number of layers) provided from 8.9 to 44.5 mm. Phase 2: low turbidity, low and high TOC test During Phase 2, the filter run times were again similar for all of the filters and varied between 58 and 60 days. These results are in contrast with the results of Mbwette (1989). Studying SSF with NWF (thickness of 7.2 to 21.6 mm) at filtration rate of 0.15 m/h with low turbidity surface water (1.1 to 4.4 NTU), Mbwette (1989) reported an increase in the filter run time by a factor of 1.5 to 2 when compared to the filter without fabric. This difference may be due to the difference in the raw water characteristics or due to the fact that a head loss of 350 mm was used to define the end of the filter run by Mbwette (1989) as compared to about 1000 mm used in the present study. Considering a head loss of 350 mm, Fig. 5 shows that this value was reached in Filter 1 (without fabric) on the 37th day, whereas Filters 2, 3, and 4 (with NWF) reached the value simultaneously on the 55th day. This corresponds to an increase in filter run time by a factor of about 1.5 due to NWF, which corroborates the findings of Mbwette (1989). However, unlike Mbwette (1989), the increase in the present study was independent of fabric thickness. The presence of fabric layers also provided protection to the sand bed against particle deposition (measured as SBPT), and this time is seen to increase with an increase in the number of the fabric layers (Fig. 6). In the figure, the SBPT is observed to be 36 days for Filter 2 and >60 days

8 710 J. Environ. Eng. Sci. Vol. 6, 2007 Fig. 6. Variation of head loss across the sand bed alone with time for the various filters (Phase 2: low turbidity, low and high TOC test) Filter 1 (0 layer) Filter 2 (2 layers) Head loss (mm) Filter 3 (5 layers) Filter 4 (10 layers) Stage 1 Stage 2 Stage 3 1 layer = 4.45 mm NWF Day for Filters 3 and 4. Looking at just the sand bed and a terminal head loss of 350 mm, the head loss was reached on the 37th day in Filter 1 (no fabric) and was less than the value (350 mm) in all of the filters with fabric until the end of the filter run (day 60). Hence the presence of NWF increased the run time for the sand bed by a factor of >1.6. Comparing the same terminal head loss value (350 mm) during the Phase 1 experiment (Fig. 4), it is seen that the presence of two layers (8.9 mm) of NWF in Filter 2 increased the run time for the sand bed by a factor of about 2, whereas this factor was >2 for increasing number of fabric layers (Filters 3 and 4). This effect was not examined in the study by Mbwette (1989). Treatment efficiency of filters Phase 1: high turbidity, low TOC test Turbidity removal by the four filters was comparable. The overall average turbidity removal efficiencies were >98% in all four filters, and the final filtrate turbidity in all of the filters was <0.2 NTU (Table 3). The overall average turbidity removal efficiencies by the fabric alone were 74%, 96%, and 98% for the Filters 2, 3, and 4, respectively. This indicates that thicker fabric in Filter 3 (5 layers of fabric) and Filter 4 (10 layers of fabric) were more efficient in removing turbidity, while the fabric (2 layers) in Filter 2 allowed more particles to escape. The increase in particle capture with a thicker fabric layer can be explained by using the depth and adsorptive filtration mechanisms. In thicker fabric layer, the tortuous paths for water flow are longer due to the winding nature of fibre and the probability of particle and fibre collisions and attachments by adsorption is increased. Overall average TOC removal efficiencies by the four filters were also comparable and ranged between 41 and 47%. Total organic carbon removal efficiencies in SSF in the range of 25% have been reported for natural surface waters with low TOC content (Collins et al. 1991; Cleasby 1991). Higher TOC removal observed in this study could be due to the differences in composition and characteristics of the raw water. Phase 2: low turbidity, low and high TOC test The raw water turbidity for this run was low (<2 NTU) and the turbidity of final filtrates of all of the filters were always less than 1 NTU. The treatment efficiencies for the four filters were similar. The overall average final filtrate turbidity values for the filters ranged between 0.5 and 0.65 NTU (Table 4) with corresponding removal efficiencies of 60 to 71%. Average TOC removal efficiencies were found to be 79 to 81%. These values are about twice the TOC removal values in the Phase 1 (41% to 47%), which might be due to the increased bioactivity and higher amount of easily biodegradable TOC (due to the addition of D-glucose) in the raw water during the Phase 2. With simulated raw water of high TOC content (4.5 to 7.5 mg/l) of easily biodegradable (readily assimilable) organic matter Barrett and Silverstein (1988) reported TOC removal efficiencies of about 80%, which are similar to those observed during Phase 2 of the current study. The bacteriological quality of the treated water was measured as total coliform and E. coli removal efficiencies between day 37 and 58 of the Phase 2 experiment, and the results are presented in Table 5. Again, the performances of the four filters were comparable with >2-log reductions for both total coliforms and E. coli. For the filters with fabric, 55 to 80% of this reduction for total coliforms and 70 to 100% for E. coli was attributable to the fabric layers. Conclusions The present study investigated the use of non-woven fabric (NWF) as an aid to slow sand filter (SSF) and showed the following:. Turbidity, readily-assimilable TOC, total coliform and E. coli removal efficiencies of the filters with NWF were comparable to the one without NWF and those reported for conventional SSF in literature. Filtrate turbidity values were <1 NTU, and >2-log reductions in both total coliforms and E. coli were obtained.. Non-woven fabric protected the sand bed against particle deposition. The extent of protection provided to the sand

9 Mondal et al. 711 Table 5. Total coliform and E. coli removal efficiencies (%) for the various filters from day 37 to day 58 (Phase 2: low turbidity, low and high TOC test). Total coliform Removal Efficiencies (%) Filter 1 Filter 2 Filter 3 Filter 4 Feed (CFU/ 100 ml) Total Total Fabric (2 layers { ) Total Fabric (5 layers { ) Total Fabric (10 layers { ) Average* Lowest Highest E. coli Average Lowest Highest *Number of observations = 11. { 1 layer = 4.45 mm of NWF. bed increased with increasing fabric thickness from 8.9 mm (2 layers) to 44.5 mm (10 layers), and reduced with increasing raw water turbiditiy.. Within the fabric layers, most of the particles were captured by the top-most layer, which contributed to most of the increase in head loss. Thus by allowing periodic removal of one layer at a time, use of NWF can increase the filter run time of conventional SSF. The present study demonstrates that the operation of SSF with NWF can be an option for simplifying the operation of and extending the viability of the process to a wider range of raw water turbidity values than that considered economical for conventional SSF. Further work is needed to establish this range and investigate the usefulness of the modification for other situations such as short-term episodes of high turbidity and algal blooms. The results show that most of the incoming particles are captured in the top-most fabric layer, and thus periodic removal of one fabric layer at a time could extend the filter run time before cleaning of the sand bed is required. Thus, the SSF can be provided with a number of fabric layers (e.g., 10). When a pre-set head loss is reached, one layer of fabric could be removed and the operation could be continued. When a chosen minimum number of fabric layers is reached (e.g., 5), the removed and cleaned fabric layers could be replaced back in the filter below the remaining fabric layers and the filter operation could be continued. Thus with the aid of NWF, the run time for the sand bed would be greatly extended and cleaning would only be required when a pre-set head loss across the sand bed is reached. The removal, cleaning and replacing of NWF layers is expected to be faster, easier, and less disruptive to the sand bed in the operation of the small-scale filtration systems. Acknowledgements The authors wish to thank the Department of Civil and Environmental Engineering and the Faculty of Graduate Studies and Research, University of Windsor for their support. In addition, discovery grants received from the Natural Sciences and Engineering Research Council (NSERC) of Canada are gratefully acknowledged. References APHA, AWWA, and WEF Standard methods for the examination of water and wastewater. 20th ed. American Public Health Association, Washington, D.C. AWWA Manual of design for slow sand filtration. D. Hendricks (editor), J.M. Barrett, J. Bryck, M.R. Collins, B.A. Janonis, and G.S. Logsdon. AWWA Research Foundation and American Water Works Association, Colorado, USA. Barrett, J.M., and Silverstein, J The effects of high carbon and high coliform feed waters on the performance of slow sand filters under tropical conditions, in Slow sand filtration: recent developments in water treatment technology. Edited by N.J.D. Graham. Ellis Horwood Ltd., Chichester, Bellamy, W.D., Hendricks, D.W., and Logsdon, G.S Slow sand filtration: influences of selected process variables. J. Am. Water Works Assoc. 77: Cleasby, J.L Source water quality and pre-treatment options for slow sand filters. Ch. 3. In Slow sand filtration. Edited by G.S. Logsdon. American Society of Civil Engineers, New York. Collins, M.R., Eighmy, T.T., and Malley, J.P., Jr Evaluating modifications to slow sand filters. J. Am. Water Works Assoc. 83: Ellis, K.V Slow sand filtration. Crit. Rev. Env. Contr. 15: Graham, N.J.D., and Mbwette, T.S.A Protected slow sand filtration: specification of non-woven synthetic fabric layers. Water Supply, 9: Hendricks, D.W., Barrett, J.M., Bryck, J., Collins, M.R., Janonis, B.A., and Logsdon, G.S Manual of design for slow sand filtration. American Water Works Association Research Foundation and American Water Works Association, Colorado. Huisman, L., and Wood, W.E Slow sand filtration. World Health Organization, Geneva. Klein, H.P., and Berger, C Slow sand filters covered by geotextiles. Water Supply, 12: Mbwette, T.A.S The performance of fabric protected slow sand filters. Ph.D. Thesis. Department of Civil Engineering, Imperial College of Science, Technology and Medicine, London. Montiel, A., Welte, B., and Barbier, J.M Improvement of slow sand filtration application to the Ivry rehabilitation project. Ch. 1. In Slow sand filtration: recent developments in water treatment technology. Edited by N.J.D. Graham. Ellis Horwood Ltd., Chichester.

10 712 J. Environ. Eng. Sci. Vol. 6, 2007 Palmateer, G., Manz, D., Jurkovic, A., McInnis, R., Unger, S., Kwan, K.K., and Dutka, B.J Toxicant and parasite challenge of Manz intermittent slow sand filter. Environ. Toxicol. 14: doi: /(sici) (199905)14:2<217::aid- TOX2>3.0.CO;2-L. Rachwal, A.J., Bauer, M.J., Chipps, M.J., Colbourne, J.S., and Foster, D.M Comparison between slow sand and high rate biofiltration. Part I. In Advances in slow sand and alternative biological filtration. Edited by N.J.D. Graham and M.R. Collins. John Wiley & Sons, Chichester. Tanner, S.A., and Ongerth, J.E Evaluation of slow sand filters in northern Idaho. J. Am. Water Works Assoc. 82: