Abstract

The study investigated two engineered fabrics and five cloth fabrics for low cost drinking water treatment. An optimized fabric filtration method has been developed and tested. Numerical models for predicting particulate removal efficiency have been developed for each fabric as support tools for selecting optimal process configuration. Both engineered fabrics showed better performance and achieved the most effective particulate removal for the highest number of layers used. Sequential filtration was done on eight layers for representative fabrics of each type and recorded higher contaminant removal than one filtration run. Geotextile 1 was better than geotextile 2 in particulate removal and recorded Escherichia coli removals of up to 1.4 log removal value (LRV) for eight-layer normal filtration and 3.0 LRV for four-pot sequential filtration. Brushed cotton was best among the cloth fabrics in particulate removal but performed below expectation in bacterial removal. It recorded E. coli removals of only 0.04 LRV and 0.2 LRV for eight-layer normal filtration and four-pot sequential filtration, respectively. Effluent turbidity decreased exponentially with number of fabric layers, in line with porous media filtration theory. The optimized filtration method produced very clear drinking water of relatively safe quality using geotextile 1. Appropriate disinfection is still recommended to ensure continued water safety.

INTRODUCTION

Drinking water can be treated using many different methods or a combination thereof (Parsons & Jefferson 2006; Davis 2010; MWH 2012; WHO 2017a). The use of contaminated water from unprotected sources such as streams, rivers, shallow wells, etc., without any form of treatment for drinking or in the preparation of food can lead to acute and chronic diseases with devastating public health implications (Demena et al. 2003; WHO 2017a). Outbreaks of waterborne diseases such as cholera and typhoid are a widespread problem and a major cause of death in many parts of the world (WHO 2012). Such diseases occur in all countries, but are five to six times more common in developing countries (Demena et al. 2003). The problem is especially acute in less privileged communities, where hygiene and environmental sanitation are generally poor, and is exacerbated by inadequate supplies of safe water (Demena et al. 2003; WHO 2012). It is possible for certain individuals to become immune to some waterborne bacteria by continued consumption of contaminated water, but high bacteria levels can still pose a serious threat to life in combination with other infections (Vishwanath 2010). Drinking water that has been treated effectively at point-of-use (e.g., through filtration and chlorination) and stored in clean containers can help reduce the problem (WHO 2017a).

This study optimized fabric filtration initially for particle and then for bacterial removal using two nonwoven engineered and five woven cloth fabrics for drinking water treatment. The degree to which the fabrics can remove impurities from a surface water source was assessed. Experiments were done to investigate best material and process combination for best possible removals while preventing recontamination. Bacterial removal efficiency was estimated using Escherichia coli and fecal coliforms (Ritter 2010; WHO 2017a). This is because E. coli and fecal coliforms are indicators of the level of fecal contamination in water and signal the presence of pathogens. If these are present, the water should be treated (Mihelcic et al. 2009; WHO 2012). Viruses and chemical pollutants cause far fewer problems because of drinking untreated water than bacteriological agents (McAllister 2005). Hence, the first and most important step in the fight against drinking poor quality water is the elimination of bacteria (McAllister 2005) coupled with removal of turbidity so that consumers do not choose to use lower turbidity alternatives that may not be safe (Kotlarz et al. 2009; WHO 2017b).

Fabric filtration using cloth fabrics (e.g., nylon and cotton) has been used for water treatment in many poor communities since ancient times (SWICH 2018), mainly, for removing particles (Swick & Jensen 2015). However, little work has been done to optimize fabric filtration using the principles of science and engineering (Swick & Jensen 2015). Additionally, and to the authors' knowledge, no study has been done to date on use of bidim engineered nonwoven geotextile for point-of-use (PoU) drinking water treatment. Published literature generally show studies focused on use of geotextile for storm water pollution reduction (Franks et al. 2012; Paul & Tota-Maharaj 2015) and as biofilm attachment media in wastewater treatment (Yaman et al. 2008). In drinking water treatment, nonwoven geotextile has generally been used for improving the efficiency of other methods like slow sand filtration (Graham & Mbwette 1987) and in some cases, in advanced high cost standalone systems, e.g., drinking straw (Mihelcic et al. 2009).

Much published research on fabric filtration has focused on cloth fabrics (Mihelcic et al. 2009; CAWST 2011; Swick & Jensen 2015; Shrestha & Spuhler 2018; SWICH 2018), while little research has been done to investigate and optimize bidim fabrics for low cost PoU water treatment. Cloth fabric studies, to date, have largely focused on removal of particles to improve clarity, enhance acceptability (Swick & Jensen 2015), and reduce chlorine requirements in order to reduce costs and improve taste (Kotlarz et al. 2009). In this study, nonwoven engineered bidim fabrics for PoU drinking water treatment were assessed and optimized – in addition to cloth fabrics – for both bacteria and particle removal. Cloth fabrics normally loosen significantly the more they are used, increasing their pore size and becoming less effective (Shrestha & Spuhler 2018; SWICH 2018). Therefore, engineered fabrics like bidim have relative advantages for drinking water treatment since they are stronger and can therefore be reused more often. Bidim, for instance, can easily be washed without significant fabric loosening by normal hand wash. It can also be disinfected in ordinary utility ovens at temperatures of around 100 to 200 °C, as was done in this study, and is structurally stable up to 200 °C (Kaytech Engineering 2018).

Bidim is a ‘food grade’ geotextile manufactured by Kaytech Engineering, South Africa, in accordance with ISO 9001:2008, Registration No: LS1176 (Kaytech Engineering 2018). It is a nonwoven, continuous filament, needle punched, A-grade polyester geotextile for general civil engineering applications. The A-grade geotextile has nine sub-grades ranging from A1 to A10 (Kaytech Engineering 2018), from which two were chosen for this research (A8 and A10) based on availability, mechanical, and hydraulic properties. Bidim is normally applied in hydraulic applications, e.g., for erosion control, filtration and drainage, water and waste containment, hydraulic and retaining structures, and as a turbidity curtain during bay constructions (Kaytech Engineering 2018). The woven fabrics (polycotton, cotton wool, brushed cotton, 100% polyester, and 55% polyester 45% cotton) included in this research were purchased in Stellenbosch, South Africa. These were selected based on availability and affordability to indigent groups.

International research on woven cloth filtration has shown promising results. For example, sari cloth filtration is used by women in India to improve water quality (SWICH 2018). If folded 3–8 times, sari cloth provides a filter of about 20 μm mesh size which increases to 100–150 μm in older cloth that becomes loosened (Colwell et al. 2003; Mihelcic et al. 2009). The initial pore size of 20 μm is small enough to remove all zooplankton, most phytoplankton, and all Vibrio cholerae (the bacteria that causes cholera) attached to the plankton as well as other particulates larger than 20 μm (Colwell et al. 2003; CDC 2015; Shrestha & Spuhler 2018; SWICH 2018). Studies done by Colwell et al. (2003) showed that cholera risk can be reduced by about 50% using sari cloth folded three times to produce an eight-layer filter. Another example is the guinea worm cloth used in Ghana for preventing PoU transmission of Guinea worm disease. It is a tightly woven monofilament cloth filter manufactured by Vestegaard, a Swiss company and has pore sizes of 100–150 micron and a 200 mm × 200 mm nylon center (CDC 2015). The cloth filters out the predatory genus Cyclops, a vector of the guinea-worm larvae which cause dracunculiasis (guinea worm disease) (Mihelcic et al. 2009; CDC 2015).

The assessment and optimization process elaborated in the methodology section, included measurement of turbidity, E. coli and fecal coliforms of raw and treated water. It was also assessed whether the filtered water can be consumed without additional treatment methods such as sand filtration, chlorination, solar disinfection, and boiling. That is, can the improved water quality meet safe levels of being consumed as is (WHO 1997, 2017a; Harvey 2007) and for which optimized process configuration and material combination. According to Shrestha & Spuhler (2018), fabric filtration has two important applications, namely, (i) as a drinking water improvement method for people with limited choices, i.e., the less privileged who cannot afford treating water another way; have a ‘better than nothing’ option, and (ii) used as first treatment stage from which water can then be disinfected or passed through additional treatment methods such as biosand filters. Although the technology might equally well apply in times of emergencies, it is primarily aimed at poor communities in developing countries due to prevalent levels of poverty and vulnerability in such settings (Demena et al. 2003; WHO 2012).

The focus of this research was therefore to investigate fabric filtration for low cost PoU drinking water treatment by testing two nonwoven engineered and five woven cloth fabrics in respect of particle and bacterial removal. The best performers of each fabric type were then optimized for both bacteria and turbidity removal. An optimized fabric filtration technique was also developed and tested. The emphasis was to attain the best process configuration to achieve best possible contaminant removal while preventing recontamination.

MATERIALS AND METHODS

Setting

This research was conducted in the Water Quality Laboratory of the Department of Civil Engineering at Stellenbosch University in Cape Town, South Africa. Raw surface water samples were obtained from Kromrivier stream, at 33°55′34.68″S and 18°51′40.56″E, next to the bridge between Ryneveld Street and Kromrivier Road, Stellenbosch, South Africa.

Study design

Laboratory experiments were conducted to optimize fabric filtration for removal of bacteria and particles (turbidity) by testing two nonwoven engineered fabrics and five woven cloth fabrics for PoU drinking water treatment. Initially, the extent to which all the fabrics can remove particles from water was assessed. More experiments were then systematically conducted on the best two of each fabric type for both bacteria and particle removal. The tests were done with a view to attaining best material combination and process configuration to give best possible contaminant removals while preventing recontamination. This was done with an assumption that fabrics producing the clearest water would be more acceptable to users as opposed to the ones producing water of marginal clarity (Kotlarz et al. 2009; CAWST 2017; WHO 2017b). In addition, fabrics removing most particles were assumed to also remove the most bacteria. Numerical models for estimating the optimal number of fabric layers were derived and tested for each fabric. These models may possibly serve as a support tool for costing and selection of optimal process configuration and material combination. Flow rates for each number of layers were measured for the two representative fabrics.

Sequential filtration was carried out in two different ways to decide on the best configuration with highest convenience and least recontamination potential. Optimization was done to produce drinking water with lowest possible bacterial counts and particulate concentrations in the easiest possible way. The first type of sequential filtration, whereby filtered water was re-filtered through eight layers of fabric multiple times (Swick & Jensen 2015) was performed up to 15 runs. In the second type – developed during this research based on the three-pot settling method (Mihelcic et al. 2009; CAWST 2011) – water was filtered through a set of four pots with eight fabric layers. Possible relationship between effluent turbidity and bacteria was assessed for the best configuration and is discussed in the results section.

Baseline study (turbidity removal only)

Initially, 300 mL water samples were filtered through a number of fabric layers ranging from one up to eight layers as proposed in various literature (e.g., Colwell et al. 2003; Mihelcic et al. 2009; Swick & Jensen 2015) where eight-layer cloth filters were used for cloth filtration. The baseline tests were done for turbidity removal only. Each filtration run was done using a new water sample and a new number of layers for each of the seven fabrics. Ten runs were conducted on each layer combination culminating into 80 runs for each fabric. Raw water and effluent turbidity were measured in duplicate for each run, thereafter percent removal values were averaged for reporting purposes. Water of low turbidity, generally <30 NTU, was used for the tests. Low turbidity is technically more difficult to remove without coagulation. For example, a study by Kotlarz et al. (2009) showed lower removals for lower turbidities and higher removals for higher turbidities. Additionally, various authors (e.g., Parsons & Jefferson 2006; Davis 2010; MWH 2012) recommend turbidities of ideally up to 10 NTU and not exceeding 20 NTU in raw water influent for direct filtration. From the initial (baseline) runs, the best performing fabrics were chosen for further experimentation.

Set up and apparatus

The general filtration set-up as shown in Figure 1 comprised: (i) a water column, (ii) the respective fabric layer combination, and (iii) a clean beaker to collect the filtered water. The fabric was gently tied over the beaker without stretching and the column was then gently secured in place. Everything was conducted in a manner to mimic as closely as possible how the filtration can be done in poor communities. The fabric materials used in the study are highlighted in Table 1.

Table 1

Properties of the nonwoven engineered and woven cloth fabrics

Fabric type Thickness (mm) Pore size (μm) Permeability (m/s × 10−3Tensile strength (kN/m) Static puncture strength (kN) Availability in South Africa Estimate cost (USD/m2
Geotextile 1 6.1 <75a 3.1a 50.0a 9.5a Available 1.76a 
Geotextile 2 5.8 <75a 2.6a 56.0a 11.7a Available 2.31a 
Polycotton <1.5 >150 NA NA NA Available 2.35 
Cotton wool <1.5 >150 NA NA NA Available 6.46 
Brushed cotton <1.5 >150 NA NA NA Available 2.35 
100% Polyester <1.5 >150 NA NA NA Available 2.35 
55% Polyester 45% Cotton <1.5 >150 NA NA NA Available 2.35 
Fabric type Thickness (mm) Pore size (μm) Permeability (m/s × 10−3Tensile strength (kN/m) Static puncture strength (kN) Availability in South Africa Estimate cost (USD/m2
Geotextile 1 6.1 <75a 3.1a 50.0a 9.5a Available 1.76a 
Geotextile 2 5.8 <75a 2.6a 56.0a 11.7a Available 2.31a 
Polycotton <1.5 >150 NA NA NA Available 2.35 
Cotton wool <1.5 >150 NA NA NA Available 6.46 
Brushed cotton <1.5 >150 NA NA NA Available 2.35 
100% Polyester <1.5 >150 NA NA NA Available 2.35 
55% Polyester 45% Cotton <1.5 >150 NA NA NA Available 2.35 

aReference (Kaytech Engineering 2018); NA, data not available.

Figure 1

General filtration set-up (top) and schematic with movable lid added for flow rate measurement (bottom).

Figure 1

General filtration set-up (top) and schematic with movable lid added for flow rate measurement (bottom).

Ordinary sequential filtration (refiltering through the same layers of fabric)

This part of the study involved filtering 750 mL of water through an eight-layer combination of geotextile 1 (bidim A8) and cotton cloth, respectively. The effluent (filtered water) was then re-filtered several times through the same eight layers of fabric as proposed by Swick & Jensen (2015). Geotextile 1 was chosen because it had the highest performance in filtration efficiency among all the fabrics and was more efficient than geotextile 2 (bidim A10). Brushed cotton fabric recorded the highest particle removal efficiency among the cloth filters and was therefore chosen for further experimentation. The bacterial and particle contents were measured for the raw water and before and after sequential filtration. The filtration set-up as depicted in Figure 1 was used.

Four-pot sequential filtration (filtering through a four-pot treatment system)

In this part of the study, filtration runs involved eight independent layers of geotextile 1 and brushed cotton tied over four clean beakers. The experimental set-up comprised four pots like the one shown in Figure 1. It essentially consisted of: (i) a moveable water column, (ii) an eight-layer fabric combination wrapped over each beaker, and (iii) four clean beakers to collect the filtered water. Effluent from the first eight-layer set was filtered through the second set, then through the third set, and finally through the last (fourth) set. Turbidity was measured for raw water and filtered water (effluent) after each step. Only geotextile 1 had bacterial counts tested after each step whereas cloth fabric had only the first and last set effluent tested for bacteria. Testing geotextile 1 for all treatment steps was done to check if bacterial removal had a similar pattern to particle removal and to assess possible relation between effluent turbidity and bacteria (CAWST 2013).

This filtration method was adapted from the three-pot treatment system (Mihelcic et al. 2009; CAWST 2011). The three-pot method is used for particle settling and requires a minimum of 24 hours waiting period (Mihelcic et al. 2009). The eight-layer four-pot sequential filtration only required a 2 hour retention time for the first pot. According to SDWF (2018), 2 hours is adequate for most particles to settle out, after which the remaining particles will require 8 days or more to settle. It was thought that four pots with eight fabric layers would remove a large proportion of contaminants from water. According to SDWF (2018), the average settling times of selected particles in water through 1 m are as follows: (i) colloids and viruses: 2 to 200 years, (ii) bacteria: 8 days, (iii) clay, algae, protozoa, and helminths: 2 hours, (iv) fine sand: 2 minutes, and (v) gravel: 1 second. This indicates that settling alone is probably not adequate. Therefore, the four-pot sequential filtration method may provide a better solution for removing more particles than the ordinary three-pot settling method.

Preparation of water of varying turbidity values for turbidity removal testing

The principle of mass conservation was applied to estimate varying water turbidities using source and filtered water, more especially, for the baseline studies, as follows: 
formula
(1)
where C = solids' concentration (mg/L); V = volume (L); C1 and C2 = solids' concentrations in the mixed samples of volume 1 and 2, respectively; C3 = concentration of the resulting mixture. 
formula
(2)
According to Walski et al. (2017), the solids' concentrations can be related to turbidity according to a generalized function: 
formula
(3)
where T = turbidity in NTU.
Substituting Equation (3) into Equation (1) and solving for turbidity (T) yields a generalized Law of Conservation of Turbidity (Walski et al. 2017): 
formula
(4)
Assuming a linear relationship between turbidity and solids' concentration (i.e., C=kT) in the mixed samples and resulting mixture (Walski et al. 2017), Equation (4) was reduced to Equation (5) and used to estimate resulting mixture turbidities for fabric testing: 
formula
(5)

Roughly only water blends with turbidity values within ±0.5 NTU of the estimated turbidity values were used in the tests. Mixtures with turbidity deviation above this were discarded. This was done to rationally keep as close as possible to the intended turbidity values.

General properties of the fabrics

Table 1 provides some typical properties (thickness, pore size, permeability, tensile strength, and static puncture strength) of each fabric material used. The values for geotextile 1 and geotextile 2 – except for thickness – were extracted from a technical data sheet provided by the manufacturer Kaytech Engineering, South Africa. The geotextile fabrics are food grade, nonwoven, continuous filament, needle punched, polyester geotextile normally used in civil engineering applications (Kaytech Engineering 2018). The cloth fabrics (polycotton, cotton wool, brushed cotton, 100% polyester, and 55% polyester 45% cotton) were generally coarser (>150 μm) and thinner (<1.5 mm) than the geotextile fabrics (see Table 1). Apparently, the cost per m2 of cloth fabrics was generally higher than that of the geotextile fabrics (Table 1), probably because geotextile fabrics are mainly sold in bulk with sufficient roll dimensions.

Turbidity removal efficiency prediction models

Numerical models were developed from baseline data using multiple linear regression (MLR) analysis (Juntunen et al. 2012) for predicting turbidity removal percentage by each fabric. The regression analysis was done using Analyse-it® (version 4.96.4) and Tool Pak VBA statistical software add-ins for Excel 2016. Model fitting was done using the least squares technique that minimizes the sum of squares of discrepancies between observed and predicted values (Juntunen et al. 2012). Table 2 contains a compilation of the developed models with selected model performance indicators. Seventy-two pairs of data from 10 runs on each fabric were used to develop the models from 80 pairs of observed data. The remaining data set was used for model verification. The 72 turbidity removal data points were averaged across number of layers to get layer-turbidity removal data sets which were then used in the MLR analysis to generate the models. The models were thereafter used to estimate the maximum number of layers required to achieve a possible 100% turbidity removal (Table 2).

Table 2

Turbidity removal prediction models for each fabric

Fabric Numerical model R2 RMSE CL NSE Number of layers for E = 100 
Geotextile 1 E = 13.90 + 4.22 η + 0.344 η2 0.996 1.3738 95% 0.993 11 
Geotextile 2 E = 10.74 + 1.48 η + 0.6119 η2 0.997 1.0445 95% 0.996 11 
Polycotton E = 2.93 + 2.70 η + 0.2425 η2 0.991 1.3254 95% 0.992 16 
Cotton wool E = 5.59 + 3.55 η + 0.1993 η2 0.992 1.4077 95% 0.993 15 
Brushed cotton E = 6.25 + 4.78 η + 0.0901 η2 0.990 1.6000 95% 0.989 16 
100% Polyester E = 2.04 + 4.06 η + 0.1158 η2 0.993 1.2805 95% 0.997 17 
55% Polyester 45% Cotton E = 2.68 + 5.32 η − 0.0010 η2 0.985 1.6415 95% 0.987 19 
Fabric Numerical model R2 RMSE CL NSE Number of layers for E = 100 
Geotextile 1 E = 13.90 + 4.22 η + 0.344 η2 0.996 1.3738 95% 0.993 11 
Geotextile 2 E = 10.74 + 1.48 η + 0.6119 η2 0.997 1.0445 95% 0.996 11 
Polycotton E = 2.93 + 2.70 η + 0.2425 η2 0.991 1.3254 95% 0.992 16 
Cotton wool E = 5.59 + 3.55 η + 0.1993 η2 0.992 1.4077 95% 0.993 15 
Brushed cotton E = 6.25 + 4.78 η + 0.0901 η2 0.990 1.6000 95% 0.989 16 
100% Polyester E = 2.04 + 4.06 η + 0.1158 η2 0.993 1.2805 95% 0.997 17 
55% Polyester 45% Cotton E = 2.68 + 5.32 η − 0.0010 η2 0.985 1.6415 95% 0.987 19 

E = turbidity removal efficiency in %; η = number of fabric layers; CL = confidence level.

An MLR model of the general form given in Equation (6) with N observations and P variables (Juntunen et al. 2012) assisted in the stepwise derivation of the models for each fabric. 
formula
(6)
where y = value of the response variable (turbidity removal efficiency); x = value of the predictor (explanatory) variable (number of fabric layers); β0 = a constant; β1· · ·βp = model coefficients to be estimated; ɛ = random error term (uncontrolled factors and experimental errors in the model); i indexes the N observed data.
The performance of each numerical model on turbidity removal was assessed by calculating the following (Gikas & Tsihrintzis 2012; Chen & Liu 2015): 
formula
(7)
where R2 = coefficient of determination; SSE = sum of squared errors; SST = total sum of squares. 
formula
(8)
where RMSE = root mean squared error; N= total number of observations; Pi = model predicted value; Oi = observed value. 
formula
(9)
where NSE ranges between , and the best value of NSE is 1.0; Pi = model predicted value; Oi = measured value; Omean = mean of observed values.

Sampling and filtration evaluation

At the baseline stage, only turbidity was quantified before and after fabric filtration. Thereafter, the concentrations of indicator bacteria (E. coli and fecal coliforms) and turbidity were quantified before and after filtration for geotextile 1 and brushed cotton only. Raw water was passed through the fabrics in a manner simulating as closely as possible PoU water treatment practices by users in poor communities (Figure 1). Tests for E. coli and fecal coliforms were done by Water Analytical Laboratory (WALAB) accredited to the South African National Accreditation System (SANAS) No: T0375 for microbiological analysis. The accredited fecal coliform detection method used is the biochemical method, WAL M3 while the accredited E. coli detection method used is the enzyme substrate, WAL M4. Physico-chemical tests were done in the Water Quality Laboratory at Stellenbosch University with the test apparatus being calibrated daily. All tests were performed in accordance with Standard Methods for the Examination of Water and Wastewater (APHA/AWWA/WEF 2012).

Removal effectiveness calculations

The removal percentages for turbidity were calculated using Equation (10): 
formula
(10)
Log removals for bacteria (E. coli and fecal coliforms) were calculated using Equation (11): 
formula
(11)
where LRV = log removal value; Bin = concentration of bacteria in influent; Bout = concentration of bacteria in effluent.

RESULTS AND DISCUSSION

Baseline study (turbidity removal only)

Average percentage turbidity remaining of the 10 runs on each fabric's layer combination were calculated and plotted as a function of number of layers (see Figure 2). The characteristics of the fabrics are as shown in Table 1. Geotextile 1 performed exceptionally well and consistently recorded the lowest remaining turbidity for each number of layers used. Although geotextile 2 and brushed cotton had slightly higher remaining turbidity than geotextile 1, they also recorded appreciable removals. The superior turbidity removals by geotextile 1 (Figure 2) could be attributed to its smaller pore sizes coupled with its thickness (Table 1). Geotextile 2's lower removals than geotextile 1 could be due to being slightly thinner than geotextile 1 (Table 1). Also, geotextile 2 had some observed minor loose fibers on its fabric surface, few of which were probably released into the water during filtration. The slightly higher turbidity removals by brushed cotton among the cloth fabrics could be attributed to its being a bit tighter than the others.

Figure 2

Average percentage turbidity remaining in the effluent from each fabric during the baseline study.

Figure 2

Average percentage turbidity remaining in the effluent from each fabric during the baseline study.

Polycotton and 100% polyester recorded roughly the same percent removals in each layer's effluent and constantly recorded the highest remaining turbidity. The pair generally had the largest pore sizes and slightly higher than 55% polyester 45% cotton; 55% polyester 45% cotton had remaining turbidity roughly equal to that of cotton wool and higher than that of brushed cotton. Cotton wool was expected to provide the best performance among the cloth fabrics but was generally loose and clearly released fibers into the treated water. It is possible that cotton wool may work more reliably if packed in a filter bag to prevent fibers from escaping into the water. Overall, the best performance in filtration efficiency was achieved using the highest number of layers of geotextile 1. This is not particularly surprising and fits well with filtration theory where more and thicker filter layers are expected to be more efficient in trapping particles than thinner and fewer layers (Parsons & Jefferson 2006; Davis 2010; MWH 2012). Likewise, filter media with smaller pore size are expected to be more efficient (Parsons & Jefferson 2006; Davis 2010; MWH 2012).

Turbidity removal prediction models

The mathematical models giving the best fit of each fabric's observed data were derived and are as listed in Table 2. Figure 3 shows each fabric's model verification plots respectively for observed and predicted turbidity removals. The estimated removals using the developed models are within sufficient accuracy of measured values, that is, the models were verified and found to generate reasonable predictions of turbidity removal efficiencies from the raw water used (Table 2 and Figure 3). These models may be helpful as a support tool for costing and selection of optimal process configuration and material combinations and may help the user to estimate the optimal number of layers for a given fabric and, correspondingly, the cost. Depending on available resources and required filter surface area, the choice of fabric and material combination can then be made.

Figure 3

Model verification plots: observed and predicted turbidity removal percent values as a function of number of layers.

Figure 3

Model verification plots: observed and predicted turbidity removal percent values as a function of number of layers.

Ordinary sequential filtration (turbidity removal performance)

Ordinary sequential filtration was done using geotextile 1 and brushed cotton and the results for turbidity removal are as shown in Figure 4. The filtration was done using 750 mL of raw water with an initial turbidity of 18 NTU through eight layers of each fabric. There was a noticeable decrease in effluent turbidity after every run until the fourth run, after which, minimal decrease was observed until the last run. Generally, turbidity decreased after each run and by the 10th run reduced to 0.8 NTU for geotextile 1 and 3.8 NTU for brushed cotton. The values are well within the recommended turbidity level (≤5 NTU) for household settings and small-scale water supplies (CAWST 2013; WHO 2017b). The exponential decrease in turbidity is not particularly surprising and is consistent with filtration theory through porous media (Davis 2010; MWH 2012; Swick & Jensen 2015).

Figure 4

Ordinary (left) and four-pot (right) sequential filtration effluent turbidity as a function of run number.

Figure 4

Ordinary (left) and four-pot (right) sequential filtration effluent turbidity as a function of run number.

Geotextile 1 performed better than brushed cotton throughout the tests. The higher particulate removals by geotextile 1 could be attributed to the smaller pore size of its filter, which is <75 μm (Kaytech Engineering 2018) vs >150 μm (Mihelcic et al. 2009) for brushed cotton and to its layer thickness of about 6 mm compared to brushed cotton with layer thickness of ≤1.5 mm. Geotextile 1 was able to meet WHO and SANS 241 standards for turbidity (≤5 NTU) after the first run while brushed cotton only met the turbidity standard after the fourth run of sequential filtration. Therefore, the eight-layer filtration for geotextile 1 may be more beneficial for small scale drinking water treatment than brushed cotton. Only geotextile 1 met the turbidity requirement (≤1 NTU) for best disinfection performance after the 10th run (WHO 2017b) to use the least possible chlorine dosage with minimal potential for disinfection by-products (CAWST 2017; WHO 2017a) and this may ably enhance taste acceptability for treated water (Kotlarz et al. 2009; WHO 2017a).

Ordinary sequential filtration: bacterial removal

The results in Table 3 show that brushed cotton could not reduce the bacteriological loads to safe levels even after 15 run cycles implying that its effluent may not provide safe drinking water, and if used, may require thorough disinfection to make the water safe. On the other hand, geotextile 1 significantly reduced bacterial loads just with one run of an eight-layer normal filtration and its effluent may require minimal disinfection. Generally, a high level of caution should be observed when using ordinary sequential filtration as some recontamination was noticed during sequential filtration. For instance, geotextile 1 recorded more bacteria in the effluent than the single run (normal) eight-layer filtration. This was noted as a disadvantage for using ordinary sequential filtration as compared to the four-pot sequential filtration method which by far gave better removals (Table 4). The four-pot sequential filtration may often perform better and need less caution and effort while ordinary sequential filtration was noted as a highly laborious and very delicate filtration process.

Table 3

Bacterial removal by geotextile 1 and brushed cotton (normal eight-layer and ordinary sequential filtration)

Parameter Raw water Effluent from the tested fabrics and associated filtration process
 
Potable water standards
 
Geotextile 1 normal filtration Geotextile 1 sequential filtration Brushed cotton normal filtration Brushed cotton sequential filtration WHO SANS 241 
Fecal coliforms (CFU/100 mL) 970 41 192 870 570 
E. coli (CFU/100 mL) 870 36 178 800 520 
LRV (fecal coliforms)  1.37 0.70 0.05 0.23   
LRV (E. coli 1.38 0.69 0.04 0.22   
Parameter Raw water Effluent from the tested fabrics and associated filtration process
 
Potable water standards
 
Geotextile 1 normal filtration Geotextile 1 sequential filtration Brushed cotton normal filtration Brushed cotton sequential filtration WHO SANS 241 
Fecal coliforms (CFU/100 mL) 970 41 192 870 570 
E. coli (CFU/100 mL) 870 36 178 800 520 
LRV (fecal coliforms)  1.37 0.70 0.05 0.23   
LRV (E. coli 1.38 0.69 0.04 0.22   
Table 4

Bacterial removal by the four-pot eight-layer sequential filtration method

Parameter Pot number Geotextile 1
 
Brushed cotton
 
Influent Effluent LRV Influent Effluent LRV 
Fecal coliforms (CFU/100 mL) Pot 1 1,110 360 0.49 1,110 1,100 0.004 
E. coli (CFU/100 mL) 960 310 0.49 960 860 0.048 
Fecal coliforms (CFU/100 mL) Pot 2  134 0.43  – – 
E. coli (CFU/100 mL)  122 0.41  – – 
Fecal coliforms (CFU/100 mL) Pot 3  13 1.01  – – 
E. coli (CFU/100 mL)  13 0.97  – – 
Fecal coliforms (CFU/100 mL) Pot 4  1.11  740 0.176 
E. coli (CFU/100 mL)  1.11  620 0.142 
Total (fecal coliforms) 4 Pots 1,110 3.05 1,110 740 0.18 
Total (E. coli4 Pots 960 2.98 960 620 0.19 
Parameter Pot number Geotextile 1
 
Brushed cotton
 
Influent Effluent LRV Influent Effluent LRV 
Fecal coliforms (CFU/100 mL) Pot 1 1,110 360 0.49 1,110 1,100 0.004 
E. coli (CFU/100 mL) 960 310 0.49 960 860 0.048 
Fecal coliforms (CFU/100 mL) Pot 2  134 0.43  – – 
E. coli (CFU/100 mL)  122 0.41  – – 
Fecal coliforms (CFU/100 mL) Pot 3  13 1.01  – – 
E. coli (CFU/100 mL)  13 0.97  – – 
Fecal coliforms (CFU/100 mL) Pot 4  1.11  740 0.176 
E. coli (CFU/100 mL)  1.11  620 0.142 
Total (fecal coliforms) 4 Pots 1,110 3.05 1,110 740 0.18 
Total (E. coli4 Pots 960 2.98 960 620 0.19 

‘–’ not tested.

Four-pot eight-layer sequential filtration performance (turbidity removal)

Turbidity removal trends for the four-pot eight-layer sequential filtration was much better than that of the ordinary eight-layer sequential filtration (Figure 4). The four-pot sequential filtration method achieved effluent turbidity of 0.6 and 2.6 NTU for geotextile 1 and brushed cotton, respectively, after only four cycles (Figure 4). In contrast, it took about 10 to 15 runs to reach similar turbidity levels in the ordinary eight-layer sequential filtration (Figure 4). This could be attributed to higher recontamination potential in the latter method. Therefore, the authors recommend the use of four-pot method for sequential filtration. A tap for drawing water may also be fixed on each filtration vessel so that the fabrics can be kept intact and only untied for cleaning purposes. Figure 5 gives a visual comparison of the four-pot sequentially filtered water by geotextile 1. The figure depicts significant improvement in water's clarity from the raw water through the first to the last filtration set (pot).

Figure 5

Visual comparison of four-pot sequentially treated water by geotextile 1; from raw water through first to last set.

Figure 5

Visual comparison of four-pot sequentially treated water by geotextile 1; from raw water through first to last set.

Four-pot eight-layer sequential filtration performance (bacterial removal)

There was a consistent trend of bacteria counts decreasing after each four-pot filtration set with only 1 CFU/100 mL fecal coliforms and 1 CFU/100 mL E. coli remaining in geotextile 1's fourth pot effluent (Table 4). This was a noticeable improvement in the bacteriological quality from the raw water (influent) bacterial levels which were 1,110 CFU/100 mL and 960 CFU/100 mL, respectively (Table 4). Therefore, the four-pot eight-layer process configuration for geotextile 1 was shown to provide relatively safe water (WHO 1997; Harvey 2007; CAWST 2013) even without disinfection (Table 5). This can possibly be improved and made more convenient if a tap is provided for each pot in the proposed four-pot method. WHO and SANS 241 potable water standards recommend 0 CFU/100 mL for both E. coli and fecal coliforms. In contrast to geotextile 1's removals, brushed cotton unsatisfactorily recorded 740 CFU/100 mL and 620 CFU/100 mL in its fourth-pot effluent for fecal coliforms and E. coli, respectively. This shows that fabric choice is important regardless of the process configuration used. In this case, geotextile 1 was shown to be the best performing fabric.

Table 5

Associated risk for fecal contamination in drinking water (CAWST 2013)

E. coli level (CFU/100 mL sample) Risk (WHO 1997; Harvey 2007; CAWST 2013Recommended action (Harvey 2007
0–10 Reasonable quality Water may be consumed as it is 
11–100 Polluted Treat if possible, but may be consumed as it is 
101–1,000 Dangerous Must be treated 
>1,000 Very dangerous Rejected or must be treated thoroughly 
E. coli level (CFU/100 mL sample) Risk (WHO 1997; Harvey 2007; CAWST 2013Recommended action (Harvey 2007
0–10 Reasonable quality Water may be consumed as it is 
11–100 Polluted Treat if possible, but may be consumed as it is 
101–1,000 Dangerous Must be treated 
>1,000 Very dangerous Rejected or must be treated thoroughly 

Relationship between turbidity and presence of bacteria

Figure 6 shows that there was some form of relationship between bacterial counts and turbidity in the effluent after each pot filtration. According to various authors (e.g., WHO 1997, 2017b; CAWST 2013), higher turbidity levels are most often associated with higher levels of pathogens (viruses, protozoa, bacteria, helminths, etc.). The pathogens are often attached to particles (e.g., clay and silts) in water (Ritter 2010; CAWST 2013; WHO 2017b). The presence of particles can also indicate the presence of hazardous chemicals and increased chlorine requirements (Kotlarz et al. 2009; WHO 2017b). This result supports the importance of turbidity removal by PoU methods. The WHO (2017b) recommends that in lower resource settings and small-scale water supplies turbidity should be kept below 5 NTU. The four-pot eight-layer sequential filtration method using geotextile 1 met this recommended value. It is therefore recommended for use in poor communities.

Figure 6

Correlation between bacteria and turbidity in treated effluent for geotextile1 four-pot sequential filtration.

Figure 6

Correlation between bacteria and turbidity in treated effluent for geotextile1 four-pot sequential filtration.

Flow rate variation with number of layers for the representative fabrics

Filtration flow rates for geotextile 1 and brushed cotton were estimated by filtering 300 mL of potable tap water through each layer combination in a manner mimicking low cost PoU water treatment. This was done to assess the usability of the fabrics with respect to filtration time and convenience. Tap water was initially filtered through each layer combination before commencing measurements to remove any captured air, flush out any dirt, and ensure uniform initial moisture content. The time taken for 300 mL of tap water to filter through was noted and flow rate was estimated using Equation (12). The flow rate tests were done in triplicate for each layer combination to ensure accuracy using the Figure 1 set-up with a movable lid for temporarily holding water in place. Average values were then calculated for reporting purposes. The initial maximum head in a cylinder of about 43 mm diameter was 200 mm. It is worth noting, that as the water volume reduced the head also reduced. Hence, the actual flow rates may be slightly higher than measured.

However, in actual applications, flow rate is expected to reduce with time as a function of solids' accumulation on each filter surface (Franks et al. 2012). Table 6 shows estimated flow rates in L/h as a function of the number of fabric layers. Flow rate was measured in mL/s and thereafter converted to L/h. It was observed that flow rate decreased with increase in the number of layers. It should be noted that when raw (untreated) water is used, flow rate will reduce faster as solids get captured on and within the fabrics (Franks et al. 2012). However, the use of fabrics has the advantage of easy removal and washing to remove trapped dirt and once again improve flow rates.

Table 6

Estimated flow rate values for geotextile 1 and brushed cotton as a function of the number of layers

Number of layers  
Geotextile 1 Flow rate (L/h) 425.20 300.84 253.52 216.87 207.29 204.55 200.74 183.99 
Brushed cotton Flow rate (L/h) 375.00 173.91 167.44 158.36 108.76 83.53 60.61 53.31 
Number of layers  
Geotextile 1 Flow rate (L/h) 425.20 300.84 253.52 216.87 207.29 204.55 200.74 183.99 
Brushed cotton Flow rate (L/h) 375.00 173.91 167.44 158.36 108.76 83.53 60.61 53.31 
There was a noticeable significant difference in flow rate between the two fabrics particularly from two layers upwards. This could be attributed to the observed rapid clogging exhibited by the brushed cotton fabric. The rapid clogging could possibly be due to brushed cotton being woven resulting in rapid caking and consequently fast clogging. According to Mulligan et al. (2009), woven fabrics are more susceptible to rapid clogging than nonwoven fabrics. Woven fabrics generally clog quickly due to accumulation of captured particles on the first layers. Nonwoven fabrics are expected to allow for more depth filtration than woven fabrics Mulligan et al. (2009). Higher flow rates by geotextile 1 could hence be attributed to the probable depth filtration due to being nonwoven and thicker. Therefore, use of geotextile 1 would be more practical for fabric filtration due to comparatively less cleaning or replacement frequency than brushed cotton. 
formula
(12)
where Q = flow rate (L/h); V = volume of filtered water (L); t = filtration time (hours).

CONCLUSIONS AND RECOMMENDATIONS

Due to its low cost and simple operation, fabric filtration using authentic nonwoven engineered fabrics is a promising technology that can improve water security in poor communities, especially when using an optimized process such as the developed four-pot system. The representative cloth fabric (brushed cotton) performed comparatively poorly when compared to the geotextile for bacterial removal. It may therefore not be reliable due to pore sizes being too large to remove microbes. E. coli and fecal coliform levels in the brushed cotton effluent exceeded WHO drinking water guidelines and SANS 241 standards throughout the study. The nonwoven bidim fabrics performed exceptionally well and successfully removed most bacteria, especially when multiple layered and with the four-pot system.

Multiple layered fabrics recorded higher turbidity and bacterial removals compared to single layered fabrics. Sequential fabric filtration was more effective than normal (single) filtration in turbidity removals and produced clear water of acceptable turbidity (CAWST 2017; WHO 2017b). However, ordinary sequential filtration encountered some recontamination on bacterial removals. The optimized four-pot sequential filtration method with eight layers of bidim A8 (geotextile 1) produced very clear drinking water of relatively safe quality (WHO 1997; Harvey 2007; CAWST 2013). However, an appropriate disinfection step, e.g., solar disinfection or chlorination, is still recommended to ensure continued water safety. Bidim A10 (geotextile 2) also performed remarkably well and is generally easier to wash by hand than geotextile 1. It should therefore also be considered in future applications. Both the ordinary and four-pot sequential filtration methods should be subjected to field testing using bidim geotextile to assess acceptability, sustainability, and long-term performance.

REFERENCES

REFERENCES
APHA/AWWA/WEF
2012
Standard Methods for the Examination of Water and Wastewater
,
22nd edn
.
American Public Health Association (APHA)/American Water Works Association (AWWA)/Water Environment Federation (WEF)
,
Washington, DC
,
USA
.
CDC
2015
Guinea Worm Disease Global Eradication Campaign
.
Centers for Disease Control and Prevention (CDC)
,
Atlanta, CA
,
USA
. .
Center for Affordable Water and Sanitation (CAWST)
2011
Introduction to Household Water Treatment and Safe Storage Training Participant Manual No. 01, Managing Water in the Home
.
Calgary, Alberta
,
Canada
.
Center for Affordable Water and Sanitation (CAWST)
2013
Introduction to Drinking Water Quality Testing Manual
.
Calgary, Alberta
,
Canada
.
Center for Affordable Water and Sanitation (CAWST)
2017
Introduction to Drinking Water Quality for Household Water Treatment Implementers, Technical Brief
.
Calgary, Alberta
,
Canada
.
Chen
W. B.
&
Liu
W. C.
2015
Water quality modeling in reservoirs using multivariate linear regression and two neural network models
.
Advances in Artificial Neural Systems
.
doi:10.1155/2015/521721
.
Colwell
R. R.
,
Huq
A.
,
Islam
M. S.
,
Aziz
K. M. A.
,
Yunus
M.
,
Khan
N. H.
,
Mahmud
A.
,
Sack
R. B.
,
Nair
G. B.
&
Chakraborty
J.
2003
Reduction of cholera in Bangladeshi villages by simple filtration
.
Proceedings of the National Academy of Sciences of the United States of America (PNAS)
100
,
1051
1055
.
doi:10.1073/pnas.0237386100
.
Davis
M. L.
2010
Water and Wastewater Engineering: Design Principles and Practice
.
McGraw-Hill
,
New York
,
USA
.
Demena
M.
,
Workie
A.
,
Tadesse
E.
,
Mohammed
S.
&
Gebru
T.
2003
Water Borne Disease, Haramaya University for the Ethiopian Health Center Team
.
In collaboration with the Ethiopia Public Health Training Initiative, The Carter Center, the Ethiopia Ministry of Health, and the Ethiopia Ministry of Education
.
Franks
C. A.
,
Davis
A. P.
&
Aydilek
A. H.
2012
Geosynthetic filters for water quality improvement of urban storm water runoff
.
Journal of Environmental Engineering
138
,
1018
1028
.
doi:10.1061/(ASCE)EE.1943-7870.0000565
.
Gikas
D. G.
&
Tsihrintzis
V. A.
2012
A small-size vertical flow constructed wetland for on-site treatment of household wastewater
.
Ecological Engineering
44
,
337
343
.
doi:10.1016/j.ecoleng.2012.04.016
.
Graham
M. J. D.
&
Mbwette
T. S. A.
1987
Improving the efficiency of slow sand filtration with nonwoven synthetic fabrics
. In
International Reference Center (IRC) for Water Supply and Sanitation
.
Proceedings of the Filtration Society
,
The Hague
,
The Netherlands
,
255.1 87IM
.
Harvey
P.
2007
Well Factsheet: Field Water Quality Testing in Emergencies
.
Water, Engineering and Development Centre (WEDC), Loughborough University
,
Loughborough
,
UK
. .
Juntunen
P.
,
Liukkonen
M.
,
Pelo
M.
,
Lehtola
M. J.
&
Hiltunen
Y.
2012
Modelling of water quality: an application to a water treatment process
.
Applied Computational Intelligence and Soft Computing
.
doi:10.1155/2012/846321
.
Kaytech Engineering
2018
Bidim General Civil Engineering Applications
.
https://kaytech.co.za/ (accessed 20 February 2018)
.
McAllister
S.
2005
Analysis and Comparison of Sustainable Water Filters
.
EPD 397 technical report
. .
Mihelcic
J.
,
Fry
L.
,
Myre
E.
,
Phillips
L.
&
Barkdoll
B.
2009
Field Guide to Environmental Engineering for Development Workers
.
American Society of Civil Engineers
,
Reston, VA
,
USA
.
Mulligan
C.
,
Davarpanah
N.
,
Fukue
M.
&
Inoue
T.
2009
Filtration of contaminated suspended solids for the treatment of surface water
.
Chemosphere
74
(
6
),
779
786
.
doi:10.1016/j.chemosphere.2008.10
.
MWH
2012
MWH's Water Treatment: Principles and Design
,
3rd edn
.
MWH Global, Inc. John Wiley & Sons
,
Hoboken, NJ
,
USA
.
Parsons
S. A.
&
Jefferson
B.
2006
Introduction to Potable Water Treatment Processes
.
Blackwell
,
Oxford
,
UK
.
Ritter
J. A.
2010
Water Quality: Principles and Practices of Water Supply Operations
.
American Water Works Association
,
Denver, CO
,
USA
.
SDWF
2018
Conventional Water Treatment: Coagulation and Filtration
.
Safe Drinking Water Foundation (SDWF)
,
Saskatoon
,
Canada
. .
Shrestha
R.
&
Spuhler
D.
2018
Water Purification: Straining and Filtration
.
Sustainable Sanitation and Water Management (SSWM)
. .
SWICH
2018
Low-cost Water Treatment Technologies for Developing Countries
.
Safe Water International Clearinghouse (SWICH)
. .
Swick
R.
&
Jensen
J.
2015
Effectiveness of Muslin Fabric Filtration for Drinking Water Treatment in Developing Countries
. In:
Proceedings of the 2nd New England Graduate Student Water Symposium
,
September 11–13, 2015
,
University of Massachusetts
,
Amherst, MA
,
USA
.
Vishwanath
A.
2010
How Simple Cotton Cloth can be Used to Filter Water and Save Life
.
http://www.pantoto.org/waterfilter/ (accessed 22 February 2018)
.
Walski
T.
,
Minnich
K.
,
Sherman
C.
,
Strause
L.
&
Whitman
B.
2017
Can there be a Law of conservation of turbidity
.
Procedia Engineering
186
,
372
379
.
doi:10.1016/j.proeng.2017.03.233
.
WHO
1997
Guidelines for Drinking-Water Quality
,
2nd edn
,
Vol. 3
.
Surveillance and control of community supplies
.
World Health Organization
,
Geneva
,
Switzerland
. .
WHO
2012
A Toolkit for Monitoring and Evaluating Household Water Treatment and Safe Storage Programmes
.
World Health Organization
,
Geneva
,
Switzerland
.
WHO
2017a
Guidelines for Drinking-Water Quality: 4th edn Incorporating the First Addendum
.
World Health Organization (WHO)
,
Geneva
,
Switzerland
. .
WHO
2017b
Water Quality and Health-Review of Turbidity: Information for Regulators and Water Suppliers
.
World Health Organization
,
Geneva
,
Switzerland
. .
Yaman
C.
,
Martin
J. P.
&
Korkut
E. N.
2008
Geotextile Biofilters for Wastewater Treatment Study Report
.
Drexel University, Civil, Architectural and Environmental Engineering, Philadelphia, PA, USA and Fatih University, Environmental Engineering
,
Istanbul
,
Turkey
.