Longitudinal assessment of rainwater quality under tropical climatic conditions in enabling effective rainwater harvesting and reuse schemes

Journal of Cleaner Production xxx (2017) 1e12

Contents lists available at ScienceDirect

Journal of Cleaner Production journal homepage: www.elsevier.com/locate/jclepro

Longitudinal assessment of rainwater quality under tropical climatic conditions in enabling effective rainwater harvesting and reuse schemes Janet Yip Cheng Leong a, Meng Nan Chong a, b, *, Phaik Eong Poh a, b, Andreas Hermawan c, Amin Talei c a

School of Engineering, Chemical Engineering Discipline, Monash University Malaysia, Jalan Lagoon Selatan, Bandar Sunway, Selangor Darul Ehsan 47500, Malaysia Sustainable Water Alliance, Advanced Engineering Platform, Monash University Malaysia, Jalan Lagoon Selatan, Bandar Sunway, Selangor Darul Ehsan 47500, Malaysia c School of Engineering, Civil Engineering Discipline, Monash University Malaysia, Jalan Lagoon Selatan, Bandar Sunway, Selangor Darul Ehsan 47500, Malaysia b

a r t i c l e i n f o

a b s t r a c t

Article history: Received 13 June 2016 Received in revised form 26 December 2016 Accepted 26 December 2016 Available online xxx

A longitudinal harvested rainwater quality monitoring study was undertaken at 6 sites within Selangor, Malaysia over a period of 8 months. Overall, harvested rainwater is of good quality, falling within the Malaysian recreational water quality Class IIB standards with exceptions for pH (18/92), ammonia (1/92), phosphates (3/92), and total coliforms (8/92). A large number of samples tested positive for Escherichia coli (22/92), total coliforms (64/92) and Chromobacterium violaceum (7/92), showing that disinfection of harvested rainwater is mandatory prior to reuse. 2/37 harvested rainwater samples exceeded lead limits in Malaysian drinking water standards, showing that consuming rainwater without additional treatment may pose a health risk. Mixing harvested rainwater with groundwater resulted in higher phosphates and total coliforms. Rainwater collected during the wet seasons have higher concentrations of suspended solids, turbidity, and Escherichia coli than dry seasons due to the antecedent dry period. Last but not least, both principal component analysis and positive matrix factorisation were conducted on 37 samples to apportion pollutant sources in harvested rainwater. 7 principal components were identified, namely: industrial dust, steel, roadside dust, faeces, organic decay, fertilisers, and plumbing. The results from principal component analysis and positive matrix factorisation were in agreement, although the latter identified mains water top-up as an additional factor responsible for dissolved solids. Both techniques are effective at apportioning pollutant sources in harvested rainwater, and show that a rainwater harvesting system should be designed carefully to reduce contributions from steel, plumbing, organic decay, bird faeces, industrial dust and roadside dust. © 2016 Published by Elsevier Ltd.

Keywords: Alternative water Principal component analysis Positive matrix factorisation Reuse Multivariate Rainwater harvesting

1. Introduction Rainwater harvesting has long supplemented mains water supplies in households for both non-potable and potable activities (Ghisi and Mengotti de Oliveira, 2007). Rainwater has a lower concentration of pollutants than other urban sources of water, such

* Corresponding author. School of Engineering, Chemical Engineering Discipline, Monash University Malaysia, Jalan Lagoon Selatan, Bandar Sunway, Selangor Darul Ehsan 47500, Malaysia. E-mail address: [email protected] (M.N. Chong).

as greywater (Leong et al., 2016), and is thus ideal for urban reuse. In Malaysia, rainwater harvesting systems have garnered renewed attention via the 2012 Uniform Building by-Laws, which mandate the installation of rainwater harvesting systems in all new buildings with roof areas
100 m2. In spite of this new requirement, data on rainwater harvesting systems are limited. Few rainwater quality studies have been conducted in Malaysia (Yaziz et al., 1989), although numerous studies have been conducted in Australia (Huston et al., 2012), Brazil (Lara et al., 2001), Canada (Despins et al., 2009), China (Zhu et al., 2004), France (Vialle et al., 2011), Greece (Gikas and Tsihrintzis, 2012; Sazakli et al., 2007), the Netherlands (Albrechtsen, 2002), New

http://dx.doi.org/10.1016/j.jclepro.2016.12.149 0959-6526/© 2016 Published by Elsevier Ltd.

Please cite this article in press as: Leong, J.Y.C., et al., Longitudinal assessment of rainwater quality under tropical climatic conditions in enabling effective rainwater harvesting and reuse schemes, Journal of Cleaner Production (2017), http://dx.doi.org/10.1016/j.jclepro.2016.12.149

2

J.Y.C. Leong et al. / Journal of Cleaner Production xxx (2017) 1e12

Zealand (Simmons et al., 2001), Nigeria (Uba and Aghogho, 2000), Spain (Farreny et al., 2011), South Korea (Lee et al., 2012), UK (Ward et al., 2010), and the USA (Crabtree et al., 1996). Collectively, these studies show how harvested rainwater quantity is dependent on geographical climate and storage tank size (Campisano et al., 2013), whereas harvested rainwater quality is dependent on the spatial and temporal variability of rainfall (Evans et al., 2006), roofing materials (Mendez et al., 2011), and first-flush diverters (Gikas and Tsihrintzis, 2012). Successful implementation of rainwater harvesting systems therefore requires local data inventory on the physical, chemical, and microbiological characteristics of rainwater in order to minimise health risks from rainwater reuse and to design water treatment systems effectively (Chong et al., 2013). Effective design of rainwater harvesting systems additionally require source apportionment techniques to quantify the contribution of non-point sources to rainwater. Common techniques include principal component analysis (PCA), chemical mass balance (CMB) models, and positive matrix factorisation (PMF). PCA is a multivariate technique which decomposes the initial data set into principal components in order to uncover latent structures, and has been utilised in a number of rainwater quality studies (Vialle et al., 2011; Gikas and Tsihrintzis, 2012; Sazakli et al., 2007; V azquez et al., 2003). PMF is a relatively new technique (Paatero and Tapper, 1994), and holds two advantages over PCA: samples can be weighed with an uncertainty value, and solutions are constrained to be non-negative (Reff et al., 2007). PCA and PMF are preferred for source apportionment studies because they do not require pre-measured source profiles, unlike CMB. Thus, the objectives of this study are threefold: first, to monitor physical, chemical, and microbiological rainwater quality harvested from six full-scale rainwater harvesting systems in Selangor state, Malaysia; second, to assess the suitability of rainwater for nonpotable urban reuse by comparison with Malaysian Class IIB (recreational waters with body contact) and Class IV (irrigation waters) water quality standards; and third, to apportion pollutant sources using complementary multivariate chemometric techniques, such as PCA and PMF. 2. Materials and methods 2.1. Rainwater sample collection Malaysia is a tropical country with an annual rainfall of 2500 mm and temperatures ranging between 27 and 35  C. The wet monsoon season is from September to April. Six full-scale rainwater harvesting systems were randomly selected for monitoring in the state of Selangor, Malaysia. Fig. 1 shows the spatial distribution of the sites, whereas site characteristics are given in Table 1. All sites contained metal flashing on their roofs, and none have de-sludged their tanks prior to sampling. Most of the rainwater harvesting systems were installed for non-potable use, and thus Sites 1, 3, and 4 contained a mains water top-up system which leaves the tank with a minimum level of water to prevent tanks from running dry. Site 2 utilised an underground porous tank where rainwater was free to mix with groundwater at the sampling point. Sampling of harvested rainwater was carried out once every two weeks from November 2014 to June 2015 (8 months), although heavy metals were only analysed from AprileJune 2015. A sampling period of 12 months was not possible due to financial constraints. Grab samples of raw harvested rainwater were collected with 1 L Duran glass bottles from each site’s rainwater tank. Prior to sampling, all Duran glass bottles were washed with distilled water and disinfected by autoclaving at 120  C for 15 min. Samples were transported to the laboratory within 6 h, and stored at 4  C until analysis within 48 h. A total of 92 samples were collected and

analysed. On average, 16 samples were collected from each site, with the exception of Site 6. Site 6 had fewer harvested rainwater samples (N ¼ 12) due to several months where the rainwater tank was empty due to high water usage. 2.2. Analytical methods 2.2.1. Physicochemical and microbiological analysis pH, biochemical oxygen demand (BOD5), chemical oxygen demand (COD), colour, total suspended solids (TSS) and total dissolved solids (TDS) were measured according to the Standard Methods for Examination of Water and Wastewater (American Public Health Association (APHA), 2005). Turbidity was measured using a HACH Portable Turbidimeter 2100Q. Ammonia-nitrogen (NH3-N) and total phosphates (PO4-P) were measured according to HACH methods 10023 and 8190, respectively. 7 heavy metals (Co, Cu, Fe, Mn, Ni, Pb, Zn) were measured using a Perkin Elmer Optima 8000 Inductively Coupled Plasma e Optical Emission Spectroscopy (ICP-OES). Total coliforms and E. coli were enumerated in triplicates using the spread-plate method 9215C (American Public Health Association (APHA), 2005) on chromogenic Brilliance™ E. coli/ coliform Selective Agar (CM1046) and incubated at 37  C for 24 h. Purple colonies were counted as Escherichia coli (E. coli), whereas red/pink colonies were counted as other coliforms. Total coliforms were counted as the sum of both E. coli and other coliforms (purple þ pink). Glossy black colonies were isolated and sent to an external laboratory for qualitative PCR analysis, and were revealed to be Chromobacterium violaceum (C. violaceum). 2.2.2. Statistical analysis All descriptive statistical analysis was carried out in XLSTAT, a commercial software in Microsoft Excel® with significance level set at a ¼ 0.05. Harvested rainwater quality data did not follow a normal distribution according to the Shapiro-Wilk normality test, and hence the two-tailed Kruskal-Wallis test or the non-parametric equivalent of analysis of variance (ANOVA), was utilised to determine significant differences in pollutant concentrations between the six rainwater harvesting systems. If the null hypothesis in the Kruskal-Wallis test was rejected, the post hoc two-tailed DunnBonferroni test was to ascertain significant differences between paired sites. PCA was used to apportion sources of pollutants in harvested rainwater and was carried out in XLSTAT. Only components with eigenvalues
1 were retained according to Kaiser’s criterion (Kaiser, 1960), and absolute loadings > 0.4 were considered major contributors. PMF was complementary to PCA, and carried out using PMF 5.0 (US EPA, 2014). The objective of PMF was to minimise the objective function Q , given by:

Q¼

n X m X i¼1 j¼1

"

1 uij

! cij 

p X

!#2 gik f kj

(1)

k¼1

where cij and uij are the measured concentration and estimated uncertainty of species j in sample i, gik is the factor score (source contribution) of factor k to sample i, fkj is factor loading (source profile), and n, m, and p represent the number of samples, species, and sources respectively. More details can be found from in the user manual by US EPA (2014). Two input files with both the measured concentration of each species in a sample and estimated uncertainties in each sample were prepared for each chemical species. Uncertainties were estimated with method detection limits (MDL) and an assumed error of

Please cite this article in press as: Leong, J.Y.C., et al., Longitudinal assessment of rainwater quality under tropical climatic conditions in enabling effective rainwater harvesting and reuse schemes, Journal of Cleaner Production (2017), http://dx.doi.org/10.1016/j.jclepro.2016.12.149

J.Y.C. Leong et al. / Journal of Cleaner Production xxx (2017) 1e12

3

Fig. 1. Site numbers and locations in Selangor, Malaysia.

Table 1 Site characteristics of six rainwater harvesting systems situated in Selangor, Malaysia (C ¼ car washing; F ¼ first-flush diverter; G ¼ groundwater; HDPE ¼ high-density polyethylene; I ¼ irrigation; M ¼ mains water; PP ¼ polypropylene; S ¼ shower/bath; T ¼ toilet flushing; V ¼ vortex filter). Site Latitude

1 2 3 4 5 6

02 550 09.100 03 090 20.200 02 440 40.600 03 030 28.600 03 060 38.800 03 030 47.900

Longitude

Land use

Nearby Roof Roof industries material area (m2)

Tank Tank age (years Total no. material as of 2016) of tanks

Total volume of all tanks (m3)

Treatment Top-up after Endbefore sampling sampling uses

101 410 18.200 101 310 09.000 101 410 09.100 101 400 11.100 101 390 05.100 101 350 59.400

Commercial Domestic Commercial Commercial Domestic Commercial

Tertiary None Transport Tertiary None Tertiary

HDPE HDPE Concrete Metal Metal PP

9 10 412 36 6 1

None None None V, F None V

Glass Metal Metal Clay tile Metal Metal

177 586 37200 1976 487 1680

10% with the following equations (US EPA, 2014): Conc.  MDL: Uncertainty ¼ 5/6 MDL

(2)

Conc. > MDL: Uncertainty ¼ ([Error  Conc.]2 þ [0.5MDL]2)0.5 (3) Samples with concentrations beyond 4 standard deviations from the mean were assigned an uncertainty of 10 SD to minimise the impact of outliers on PMF analysis. Prior to PMF analysis, data quality was assessed using their signal-to-noise (S/N) ratio. Species with S/N ratios
2.0 was classified as “strong” in data quality. Species with S/N ratios between 0.2 and 2.0 were classified as “weak”, and their uncertainty values were tripled. Species with S/N ratios lower than 0.2 were classified as “bad” and excluded from PMF analysis. 100 base runs with an initial seed of 1 were computed for each data set to ensure that the global minimum of Q was found. The following guidelines from Huston et al. (2012) were employed to ensure that an appropriate number of factors were selected:  Qrobust z Qtrue  Observed/predicted scatter plots show good correlation between predicted and measured (observed) concentrations, with the majority of standardised residuals between 3 and þ3  G-space plots show data points lying within the source aces  Source profiles have small Discrete Difference Percentile (DDP) values

6 4 2 3 6 2

4 1 10 2 3 1

M G M M None None

T,I S,T T,I T,I T,I,C T

PMF 5.0 provides the rotational freedom parameter (Fpeak) function which controls whether more extreme values are assumed for factor loadings (Fpeak > 0) or factor scores (Fpeak < 0). In this study, altering the Fpeak value did not result in substantially better source profiles, and so base run results (Fpeak ¼ 0) are reported. The stability of the PMF solution was estimated using three error estimation methods in EPA PMF 5.0: Displacement (DISP), Bootstrap (BS), and Bootstrap-Displacement (BS-DISP). The base model run with the lowest Q value was selected for DISP. Following this, 100 BS runs with a minimum correlation value of 0.60 and block size 2 were performed. All strong species were displaced during BSDISP. 2.2.2.1. Monte Carlo simulation. Risk assessments of water quality determine the probability a water quality indicator exceeds regulatory limits. In this paper, a posterior probability distribution function was fitted to data from each water quality indicator, and Monte Carlo simulations were run for 10,000 iterations using the Microsoft Excel add-in @RISK 6.1 from Palisade Corporation. 3. Results and discussion 3.1. Physicochemical parameters Table 2 shows the univariate descriptive statistics for harvested rainwater quality for all six sampling sites and the Malaysian water

Please cite this article in press as: Leong, J.Y.C., et al., Longitudinal assessment of rainwater quality under tropical climatic conditions in enabling effective rainwater harvesting and reuse schemes, Journal of Cleaner Production (2017), http://dx.doi.org/10.1016/j.jclepro.2016.12.149

4

J.Y.C. Leong et al. / Journal of Cleaner Production xxx (2017) 1e12

Table 2 Descriptive statistics of rainwater quality in six rainwater harvesting systems alongside Malaysian water quality standards Class IIB and IV. Class IIB is for recreational waters with body contact, while Class IV is for irrigation waters. Parameter

Site no.

Observations

Minimum

Maximum

E. coli (CFU/100 mL)

1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6

16 16 16 16 12 16 16 16 16 16 12 16 16 16 16 16 12 16 16 16 16 16 12 16 16 16 16 16 12 16 16 16 16 16 12 16 16 16 16 16 12 16 16 16 16 16 12 16 16 16 16 16 12 16 16 16 16 16 12 16 16 16 16 16 12 16 7 7 7 7 2 7

0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 5.0 6.3 6.9 5.2 4.0 4.3 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0 0.0 2.0 0.0 0.0 0.8 2.8 1.4 0.4 0.5 1.0 0.0 0.0 0.3 0.0 0.0 0.3 0.1 0.1 0.0 0.1 0.1 0.1 1.0 0.0 2.0 0.0 0.0 0.0 2.2 2.0 0.1 0.0 0.0 0.0 0.00 0.00 0.00 0.00 1.00 0.00

7.67  1.00  0 2.00  0 0 5.73  1.34  1.37  4.33  3.33  6.67  8.2 7.5 10.7 7.6 5.4 6.8 1.8 3.3 3.1 3.1 3.3 1.5 6.5 10.0 51.0 5.0 2.0 8.0 35.0 166.0 94.0 43.0 17.0 39.5 92.9 57.1 18.7 10.1 1.6 4.8 0.4 1.8 1.8 0.4 2.8 1.1 0.4 2.4 0.2 0.2 0.3 1.3 5.0 46.0 15.0 13.0 5.0 5.0 139.0 94.0 443.0 199.0 491.0 63.0 6.00 6.00 6.00 6.00 6.00 6.00

Total coliforms (CFU/100 mL)

pH

BOD5 (mg/L)

COD (mg/L)

Colour (Pt-Co)

Turbidity (NTU)

NH3-N (mg/L)

PO4-P (mg/L)

TSS (mg/L)

TDS (mg/L)

Co (mg/L)

103 104 103

104 106 104 103 103 102

Mean 1.44 1.64 0 2.05 0 0 9.26 1.37 2.08 1.26 3.70 1.03 7.1 6.8 7.7 6.2 4.4 5.3 0.6 0.8 0.7 0.6 0.6 0.5 1.7 3.2 5.5 1.0 0.6 1.6 15.6 34.0 30.5 10.5 8.2 14.1 8.5 12.4 8.8 1.8 1.2 2.0 0.1 0.6 0.7 0.2 0.5 0.6 0.2 0.4 0.1 0.1 0.2 0.2 2.3 9.5 7.0 2.6 1.3 1.8 86.5 37.6 92.7 25.9 58.7 8.8 1.86 1.71 1.86 1.86 3.50 1.86

Standard deviation

 103  103  102

     

103 105 103 103 102 102

2.52  2.67  0 4.95  0 0 1.50  3.27  3.24  1.16  9.48  2.24  0.7 0.3 0.9 0.7 0.4 0.8 0.5 0.8 0.7 0.7 0.9 0.4 1.8 2.8 12.3 1.1 0.5 1.9 8.0 37.9 25.7 9.5 5.3 9.7 22.7 14.6 5.3 2.3 0.3 0.9 0.1 0.4 0.4 0.1 0.7 0.3 0.1 0.6 0.0 0.1 0.1 0.3 1.0 11.8 4.1 2.9 1.6 1.8 35.4 21.5 102.7 47.5 138.5 16.2 2.85 2.63 2.85 2.85 3.54 2.85

103 103

Class IIB

Class IV

Potable water

e

e

0

<5000

<50,000

0

6e9

5e9

6.5e9.0

<3

<12

e

<25

<100

e

<150

e

15

<50

e

5

<0.3

<2.7

1.5

0.6*

e

e

<50

<300

e

e

<4000

1000

e

e

e

102

104 105 103 103 102 102

Please cite this article in press as: Leong, J.Y.C., et al., Longitudinal assessment of rainwater quality under tropical climatic conditions in enabling effective rainwater harvesting and reuse schemes, Journal of Cleaner Production (2017), http://dx.doi.org/10.1016/j.jclepro.2016.12.149

J.Y.C. Leong et al. / Journal of Cleaner Production xxx (2017) 1e12

5

Table 2 (continued ) Parameter

Site no.

Observations

Minimum

Maximum

Mean

Standard deviation

Class IIB

Class IV

Potable water

Cu (mg/L)

1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6

7 7 7 7 2 7 7 7 7 7 2 7 7 7 7 7 2 7 7 7 7 7 2 7 7 7 7 7 2 7 7 7 7 7 2 7

0.00 0.00 0.00 0.00 46.00 0.00 6.00 29.00 7.00 0.00 6.00 8.00 4.00 1.00 0.00 2.00 4.00 2.00 0.00 0.00 0.00 0.00 1.00 0.00 0.00 0.00 0.00 0.00 4.00 1.00 0.00 14.00 34.00 54.00 106.00 57.00

7.00 16.00 14.00 12.00 53.00 9.00 30.00 229.00 247.00 11.00 15.00 49.00 29.00 71.00 11.00 14.00 15.00 14.00 5.00 5.00 5.00 5.00 6.00 5.00 7.00 7.00 9.00 8.00 11.00 110.00 15.00 129.00 92.00 139.00 149.00 320.00

2.71 8.71 8.43 6.57 49.50 4.57 19.29 138.43 71.57 5.29 10.50 19.43 17.14 20.00 4.29 8.71 9.50 7.43 1.71 1.71 1.71 1.57 3.50 1.86 2.43 3.29 2.86 2.29 7.50 18.86 5.57 57.14 49.14 95.43 127.50 146.14

2.75 5.22 4.24 4.08 4.95 2.99 9.66 85.16 82.82 4.46 6.36 13.62 9.92 24.49 3.86 4.42 7.78 3.74 2.36 2.36 2.36 2.37 3.54 2.27 3.21 3.50 3.93 3.59 4.95 40.32 5.65 37.54 19.96 30.84 30.41 90.75

20

e

1000

1000

e

300

100

200

100

50

200

20

e

e

10

5000

2000

3000

Fe (mg/L)

Mn (mg/L)

Ni (mg/L)

Pb (mg/L)

Zn (mg/L)

*Calculated using stoichiometric conversion: 1 total phosphorous (TP) ¼ 3.065 orthophosphate (PO4-P).

quality standards for both recreational waters with body contact (Class IIB), irrigation waters (Class IV), and potable water. Class IIB is more stringent than Class IV and therefore has lower maximum pollutant limits. Most samples (52/90) fell within Class IIB range of pH 6e9, with 15/92 samples within Class IV range of pH 5e9.18 samples (7 samples from Site 5, 11 samples from Site 6) were too acidic and fell outside of Class IIB and IV limits. Acidic samples may result in the leaching of various substances from collection surfaces and soils (Du et al., 2014) and thus, dilution of harvested rainwater with neutral mains water may help to increase the acidic pH in Sites 5 and 6. In contrast, the highest pH of 10.7 was found in a newly constructed rainwater harvesting system at Site 3. The high pH value is attributed to new building material residues, such as lime and gypsum, and potentially the leaching of calcium carbonate from concrete underground tank walls (Zhu et al., 2004) as the pH 10.7 sample was collected in 2014: the same year that the rainwater harvesting system began operation. A two-tailed Kruskal-Wallis test (Table 3) revealed that the median pH of harvested rainwater in Site 3 was significantly higher than Sites 4, 5, and 6. Differences between these sites may be attributed to different land use, as Site 3 is located in an area with several transportation industries, whereas Sites 4, 5, and 6 are not. Dust in the atmosphere (Gikas and Tsihrintzis, 2012), combustion of biomass which releases acidic 3 nitrates (NO 3 ), and sulphates (SO4 ) to the atmosphere (Lara et al., 2001), and the contribution of alkaline ammonium (NHþ 4 ) and calcium (Ca2þ) ions (Das et al., 2005) could all have contributed to the differences in pH between these sites. The majority of samples (53/92) fell within Class IV NH3-N

limits, and the rest within Class IIB limits. Only one harvested rainwater sample from Site 5 exceeded both standards, and the ammonia may have been sourced from bird faeces and atmospheric deposition of ammonia salts. Furthermore, median concentrations of NH3-N in Site 3 were significantly higher than in Site 4, corresponding to the higher pH in Site 3 than Site 4 as NH3-N is an alkaline buffering agent (Das et al., 2005). The NH3-N concentrations reported in this study are lower than Coombes et al. (2000), in which 68% and 24% of 200 first-flush rainwater samples exceeded NH3-N and pH requirements of the Australian Drinking Water Guidelines. If chlorine was utilised to disinfect rainwater, the chlorine would react with NH3-N to form monochloramine. Therefore, removing NH3-N from harvested rainwater would increase disinfection efficiency as free chlorine is a better disinfectant (Macauley et al., 2006). PO4-P concentrations in harvested rainwater were comparable with other literature (Sazakli et al., 2007) and likely originates from bird faeces. Site 2 has an underground high-density polyethylene (HDPE) tank with porous walls that allow groundwater to mix with harvested rainwater. As a result, Site 2 had significantly higher PO4P concentrations than Sites 3, 4, and 6 as a result of leaching of fertilisers from soil and groundwater into the tank. This implies that groundwater should not be mixed with rainwater to prevent phosphate accumulation when flushing toilets. Only one sample from each of Sites 2, 3, and 6 was unable to meet Class IIB limits for PO4-P. All harvested rainwater samples generally had low colour and turbidity within Class IIB and IV standards, showing that no aesthetic issues will be posed by rainwater reuse to flush toilets or

Please cite this article in press as: Leong, J.Y.C., et al., Longitudinal assessment of rainwater quality under tropical climatic conditions in enabling effective rainwater harvesting and reuse schemes, Journal of Cleaner Production (2017), http://dx.doi.org/10.1016/j.jclepro.2016.12.149

6

J.Y.C. Leong et al. / Journal of Cleaner Production xxx (2017) 1e12

Table 3 Kruskal-Wallis test results to ascertain if rainwater quality between sites was significantly different. The post hoc Dunn-Bonferroni test was applied to ascertain differences between paired sites. Only significant p-values are shown for the post hoc test. Parameter

E. coli

Kruskal Wallis test

Dunn-Bonferroni test

p

Compared sites

p

<0.0001

S1 S1 S2 S2 S2 S1 S1 S2 S2 S2 S2 S1 S1 S2 S2 S3 S3 S3 S4 e S2 S2 S3 S2 S2 S3 S1 S1 S2 S2 S2 S3 S3 S3 S1 S1 S1 S2 S3 S4 S2 S2 S2 S1 S2 S2 S3 S3 S3 S1 S1 S1 S2 S3 S3 S3 e S1 S2 S3 e e e S1 S1 S1

0.0031 0.0031 0.0001 0.0003 0.0001 0.0009 <0.0001 0.0017 0.0005 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 0.0015 <0.0001 <0.0001 <0.0001 0.0026 e 0.0015 0.0001 0.0018 0.0010 0.0008 0.0032 0.0006 0.0011 <0.0001 <0.0001 0.0006 <0.0001 <0.0001 0.0012 <0.0001 <0.0001 <0.0001 0.0004 <0.0001 <0.0001 0.0007 0.0010 0.0014 0.0016 <0.0001 0.0002 0.0007 <0.0001 <0.0001 <0.0001 0.0001 <0.0001 0.0022 0.0010 0.0029 <0.0001 e 0.0021 <0.0001 0.0009 e e e 0.0002 0.0018 <0.0001

Total coliforms

<0.0001

pH

<0.0001

BOD5 COD

0.4163 0.0009

Colour

0.0013

Turbidity

<0.0001

NH3-N

<0.0001

PO4-P

0.0019

TSS

<0.0001

TDS

<0.0001

Co Cu Fe

0.8732 0.0125 0.0004

Mn Ni Pb Zn

0.1115 0.8040 0.3182 <0.0001

-

S3 S6 S3 S5 S6 S5 S6 S3 S4 S5 S6 S5 S6 S5 S6 S4 S5 S6 S5

-

S4 S5 S5 S4 S5 S5 S2 S3 S4 S5 S6 S4 S5 S6 S2 S3 S6 S4 S4 S6 S3 S4 S6 S3 S5 S6 S4 S5 S6 S4 S5 S6 S6 S4 S5 S6

- S5 - S4 - S4

- S4 - S5 - S6

for landscaping. Furthermore, BOD5 was within Class IIB allowances with the exception of four samples that fell within Class IV standards. All harvested rainwater samples had COD and TSS concentrations within Class IIB allowances, and TDS concentrations within Class IV allowances. Additionally, the median of TSS and TDS in Site 3 was significantly higher than in Sites 4, 5, and 6. The low concentration of organics and solids show the excellent physicochemical quality of harvested rainwater. The differences between BOD5 and COD concentrations between sites are attributed to vehicle emissions (Zhu et al., 2004). 3.2. Heavy metal concentrations Heavy metals in harvested rainwater pose health risks when accidentally ingested during potable activities. Only 37 samples across 6 sites were analysed for heavy metal content from April to June 2015. Table 3 summarises the results of a two-tailed KruskalWallis test which was carried out to ascertain differences between sites. All heavy metals, with the exception of Pb, fell within Malaysian drinking water quality standards (Table 2). 2 samples (from Sites 5 and 6 respectively) (5.4% of 37 samples) exceeded the 10 mg/L of Pb drinking water quality limit. The number of samples exceeding the drinking water standards is lower than in another study in Australia (Huston et al., 2009). The high Pb concentrations in the harvested rainwater samples may possibly be attributed to both Pb flashing/paint on roofs, which may contribute up to 58% in rainwater tanks (Huston et al., 2012). Prolonged consumption and exposure to Pb may result in a gamut of human learning and behavioural disorders (Needleman, 2004), and thus Pb flashing should be excluded from rainwater harvesting systems. The Kruskal-Wallis test indicated no significant statistical difference between sites for Co, Mn, Ni, and Pb in harvested rainwater, implying that the roofing material and tank material did not contribute to heavy metals within the tanks. Site 5 had significantly higher median Cu and Zn concentrations than Site 1 because of acid rain dissolution of the aging metal roof. Furthermore, galvanised metal roofs have been shown to yield high concentrations of Zn in harvested rainwater (Mendez et al., 2011). Table 3 illustrates significant statistical differences in the median concentration of Zn between Site 1 and Sites 4, 5, and 6. Site 1 had the lowest Zn concentrations of all the sites because both roofing and tank materials were non-metallic: the roof was glass, whereas the tank was made out of HDPE. Additional preventative measures to reduce heavy metal concentrations in rainwater tanks include annual tank de-sludging, and annual cleaning of the roof, gutters, and debris screens (Ward et al., 2010). Sludge in a rainwater tank has been shown to contain high concentrations of heavy metals (Magyar et al., 2007), and thus regular tank de-sludging and roof cleaning help to prevent accumulation of heavy metals within the tank. 3.3. Microbiological parameters The microbiological quality of harvested rainwater was measured with faecal indicators E. coli and total coliforms. 22/92 samples (24%) tested positive for E. coli, while 64/92 samples (70%) tested positive for total coliforms. Only 8 samples exceeded both Class IIB and IV maximum limits for total coliforms, and 6 of these samples originated from Site 2. The two-tailed Kruskal Wallis test (Table 3) further revealed significant differences in total coliforms between Site 2 and Sites 3, 4, 5, and 6. The high total coliforms in Site 2 relative to other sites was attributed to groundwater seeping into the porous, underground rainwater tank, as soil contributes to bacterial contamination (Coombes et al., 2000). This implies that mixing groundwater with rainwater significantly increases the

Please cite this article in press as: Leong, J.Y.C., et al., Longitudinal assessment of rainwater quality under tropical climatic conditions in enabling effective rainwater harvesting and reuse schemes, Journal of Cleaner Production (2017), http://dx.doi.org/10.1016/j.jclepro.2016.12.149

J.Y.C. Leong et al. / Journal of Cleaner Production xxx (2017) 1e12

(a)

7

(b)

Fig. 2. (a) CM1046 agar with purple/blue Escherichia coli colonies and pink/red coliforms. (b) CM1046 agar with deep violet/black Chromobacterium violaceum. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

bacterial content of the water, and is thus not recommended. Furthermore, although E. coli was not frequently detected in harvested rainwater samples, concentrations ranged from 205 CFU/ 100 mL to 104 CFU/100 mL, showing high levels of faecal contamination likely from birds, rodents, and lizards (Gikas and Tsihrintzis, 2012). It may be possible to prevent total coliforms from entering the rainwater tank by installing a first-flush diverter. First-flush volumes contain significantly higher concentrations of total coliforms, E. coli, and enterococci (Lee et al., 2012). Indeed, in this study, Site 6 utilised a 900 L first-flush diverter, and 11/16 (69%) of harvested rainwater samples did not contain any total coliforms or E. coli. Site 6 thus contained the lowest number of coliforms due to the first-flush diverter. Furthermore, there were significant statistical differences in E. coli between Site 2 and Sites 3, 5, and 6. The results suggest that location may contribute more to E. coli concentrations than roofing material since Sites 2, 3, 5, and 6 have metal roofs. This finding is mirrored by Gikas and Tsihrintzis (2012). The results presented in this paper demonstrate that roofharvested rainwater is not suitable for potable consumption due to the high levels of faecal contamination. These results are concurrent with past studies: Sazakli et al. (2007) reports high faecal contamination in autumn, and Albrechtsen (2002) found that 11/14 samples from storage tanks contained E. coli. E. coli could therefore be specified in future water quality standards to prevent faecal contamination from birds or small mammals.

CM1046 is a chromogenic agar, which differentiates E. coli from other thermotolerant coliforms with their ability to cleave both red b-glucuronidase and blue b-galactosidase to form purple colonies (Fig. 2a). During the course of the monitoring study, glossy dark violet colonies were found in 7 separate occasions on CM1046 agar plates from Sites 1, 2 and 4 (Fig. 2b). Qualitative PCR analysis revealed these black colonies to be C. violaceum, a violet pigmented Gram negative bacterium responsible for causing septicaemia and n and Menck, 2001), rare fatalities in humans and animals (Dura further showcasing the need for rainwater disinfection. Harvested rainwater may hence be disinfected via UV irradiation (Amin and Han, 2009), chlorination (Moreira Neto et al., 2012), silver (Nawaz et al., 2012), or membrane filtration (Kim et al., 2007). Both chlorine and silver are effective chemical disinfectants, but may leave undesirable byproducts. UV irradiation and membrane filtration are hence recommended as disinfection options for rainwater harvesting systems, although UV irradiation is less energy intensive and costs less than membrane filtration. 3.4. Monte Carlo simulation Table 4 shows the distributions fitted to the data using @RISK 6.1’s built-in distribution fitting. Only finite distributions (e.g. Uniform, Pert, Triangular) with positive values were selected for fitting as concentrations cannot be negative. Based on these fitted

Table 4 Monte Carlo simulation results from @RISK 6.1 (10,000 iterations). Bolded values mark all values above 30% which present a significant concern in rainwater harvesting systems. Parameter (unit)

Class IIB

Class IV DWS Fitted distribution

E. coli (CFU/100 mL) Total coliforms (CFU/ 100 mL) pH

e e 0 <5000 <50,000 0

Triangular Triangular

e 99.2%

e 92.7%

6e9

5e9

Pert

48.4%

30.4%

BOD5 (mg/L) COD (mg/L) Colour (Pt-Co) Turbidity (NTU) NH3-N (mg/L) PO4-P (mg/L) TSS (mg/L) TDS (mg/L) Co (mg/L) Cu (mg/L) Fe (mg/L) Mn (mg/L) Ni (mg/L) Pb (mg/L) Zn (mg/L)

<3 <25 <150 <50 <0.3 0.6* <50 e e 0.02 1 0.1 0.05 e 5

<12 <100 e e <2.7 e <300 <4000 e e e 0.2 0.2 e 2

Triangular Triangular Triangular Pert Triangular Pert Triangular Triangular Triangular Triangular Triangular Triangular Triangular Triangular Triangular

1.3% 26.6% 1.2% 2.9% 80.0% 26.2% 0.0% e e 40.5% 0.0% 0.0% 0.0% e 0.0%

1.3% 0% e e 0.2% e 0.0% 0.0% e e e 0.0% 0.0% e 0.0%

6.5 e9.0 e e 15 5 1.5 e e 1000 e 1 0.3 0.1 0.02 0.01 3

Frequency of parameter exceeding Frequency of parameter exceeding Frequency of parameter Class IIB Class IV exceeding DWS

e e

e e e 0.0% 0.0% 0.0% 0.0% 83.1% 0.0%

*Calculated using stoichiometric conversion: 1 total phosphorous (TP) ¼ 3.065 orthophosphate (PO4-P).

Please cite this article in press as: Leong, J.Y.C., et al., Longitudinal assessment of rainwater quality under tropical climatic conditions in enabling effective rainwater harvesting and reuse schemes, Journal of Cleaner Production (2017), http://dx.doi.org/10.1016/j.jclepro.2016.12.149

8

J.Y.C. Leong et al. / Journal of Cleaner Production xxx (2017) 1e12

1,800

E. coli (CFU/100mL)

20

1,600

18

1,400

16

16

1,000 800

12

12

10

10

8

8

6

600

6

400

4

200

2 Dry season

Wet season

0

TSS (mg/L)

14

14

1,200

0

Turbidity (NTU)

4 2 Dry season

0

Wet season

Dry season

Wet season

Fig. 3. Seasonal variation for total coliforms, turbidity, and TSS between wet and dry seasons shown in box plots. The boxes indicate the lower and upper quartiles and the central line is the median.

distributions, a Monte Carlo simulation was carried out to assess the probability of a harvested rainwater sample exceeding Class IIB and IV limits, as well as drinking water standards (DWS). Total coliforms, pH, NH3-N, Cu, and Pb demonstrate probabilities above 30% (in bold) of exceeding Class IIB and IV values, and thus these parameters are of concern in rainwater harvesting systems. The Monte Carlo simulation results reflect the results from the longitudinal monitoring study, and thus additional treatment is warranted for rainwater reuse.

wet season may be explained by the long antecedent dry season, where contaminants accumulate on catchment surfaces (Yaziz et al., 1989). The results are in concordance with other studies: Sazakli et al. (2007) found that the autumn season contained the highest concentration of total coliforms, E. coli, and enterococci as a result of the antecedent dry season in summer, while Yaziz et al. (1989) showed that longer antecedent dry periods resulted in higher turbidity, total coliforms, and total solids. Higher doses of disinfectants, such as chlorine, would therefore be advisable in the wet season following a long antecedent dry period.

3.5. Seasonal variation 3.6. Principal component analysis (PCA) In order to examine the effect of seasonal variation on water quality parameters, all 92 harvested rainwater samples were categorised to either wet/monsoon (54 samples) or dry (38 samples) seasons. Fig. 3 presents the box plots comparing wet and dry seasons. Only three parameters - turbidity, TSS, and E. coli e have statistically significant higher median concentrations (p < 0.05) in wet seasons than dry seasons from a two-tailed Kruskal-Wallis test. It was interesting to note that although the wet season contained a higher mean of total coliforms (3.5  104 CFU/100 mL) compared to dry season (1.2  104 CFU/100 mL), there was no statistically significant difference between the medians for wet and dry seasons. The high concentrations of faecal indicators and solids in the

Table 5 summarises the 37 of the 92 samples analysed for PCA and PMF, and their descriptive statistics. PCA was conducted on 37 samples with 18 dependent variables, and reduced the original harvested rainwater quality dataset from 11 variables/components to 4 principal components with eigenvalues
1 (64% reduction) whilst maintaining 63% of the original data’s variability (37% loss of variance or information). The number of components are equal to the number of variables in PCA, although a component may consist of more than one variable (Vialle et al., 2011). Table 6 shows the loadings, eigenvalues, and proportion of variability represented by each principal component. PC1 has high

Table 5 Descriptive statistics of rainwater quality for both principal component analysis (PCA) and positive matrix factorisation (PMF) analysis. MDL ¼ method detection limit; S/ N ¼ signal-to-noise ratio for PMF analysis. Parameters

N

Minimum

Maximum

Mean

Median

MDL

S/N

Category

E. coli (CFU/100 mL) Total coliforms (CFU/100 mL) pH BOD5 (mg/L) COD (mg/L) Colour (Pt-Co) Turbidity (NTU) NH3-N (mg/L) PO4-P (mg/L) TSS (mg/L) TDS (mg/L) Co (mg/L) Cu (mg/L) Fe (mg/L) Mn (mg/L) Ni (mg/L) Pb (mg/L) Zn (mg/L)

37 37 37 37 37 37 37 37 37 37 37 37 37 37 37 37 37 37

0.00 0.00 4.12 0.00 0.00 2.00 0.44 0.00 0.05 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

5.67  103 1.62  105 7.99 1.82 8.00 166.00 18.70 1.80 1.32 13.00 139.00 0.01 0.05 0.25 0.07 0.01 0.11 0.32

1.53  102 1.24  104 6.30 0.52 1.90 26.59 3.71 0.50 0.21 2.75 33.72 0.00 0.01 0.05 0.01 0.00 0.01 0.07

0.00 3.33  102 6.69 0.49 1.00 20.00 1.77 0.36 0.12 2.00 25.00 0.00 0.01 0.02 0.01 0.00 0.00 0.06

1.0000 1.0000 1.0000 2.0000 3.0000 15.0000 0.0100 0.0200 0.0600 5.0000 1.0000 0.0010 0.0010 0.0010 0.0100 0.0020 0.0060 0.0006

0.0 5.4 6.7 0.0 0.5 1.6 9.0 7.7 3.3 0.3 5.9 2.1 5.4 7.8 0.9 1.2 0.5 8.3

Bad Bad Strong Bad Weak Weak Strong Strong Strong Weak Strong Bad Strong Strong Weak Bad Weak Strong

Please cite this article in press as: Leong, J.Y.C., et al., Longitudinal assessment of rainwater quality under tropical climatic conditions in enabling effective rainwater harvesting and reuse schemes, Journal of Cleaner Production (2017), http://dx.doi.org/10.1016/j.jclepro.2016.12.149

J.Y.C. Leong et al. / Journal of Cleaner Production xxx (2017) 1e12

9

Table 6 Component or loading matrix from principal component analysis (PCA), showing loadings on first four eigenvectors, PC1 e PC7. Loadings
|0.40| (bolded) are significant contributors to their respective principal components or eigenvectors. Parameter/Variable

Principal components PC1

PC2

PC3

PC4

PC5

PC6

PC7

E. coli Total coliforms pH BOD5 COD Colour Turbidity NH3-N PO4-P TSS TDS Co Cu Fe Mn Ni Pb Zn

0.26 0.21 0.68 0.41 0.38 0.75 0.69 0.03 0.17 0.65 0.45 0.21 ¡0.45 0.46 0.52 0.24 ¡0.43 ¡0.64

0.33 0.45 0.25 0.04 0.49 0.18 0.05 0.25 0.39 0.04 0.06 0.71 0.42 0.55 0.40 0.80 0.21 0.22

0.31 0.24 0.35 0.13 0.01 0.36 0.32 0.24 0.08 0.13 0.62 0.53 0.04 0.26 0.09 0.44 ¡0.43 ¡0.56

0.70 0.61 0.06 0.13 0.26 0.35 ¡0.44 0.38 0.44 0.36 0.14 0.17 0.04 0.05 0.04 0.08 0.25 0.14

0.17 0.34 0.21 0.83 0.56 0.08 0.02 0.20 0.37 0.18 0.10 0.08 0.16 0.20 0.08 0.03 0.07 0.20

0.07 0.12 0.03 0.04 0.14 0.19 0.09 0.65 0.03 0.18 0.07 0.03 0.17 0.14 ¡0.65 0.05 0.32 0.16

0.08 0.05 0.25 0.11 0.13 0.17 0.20 0.37 0.16 0.21 0.39 0.10 0.51 0.06 0.01 0.07 ¡0.47 0.05

Initial eigenvalue Proportion of variance (%) Cumulative proportion of variance (%)

4.0 22.0 22.0

2.8 15.3 37.3

2.1 11.4 48.7

1.9 10.4 59.1

1.5 8.6 67.7

1.1 6.4 74.1

1.0 5.7 79.8

loading values for pH, BOD5, colour, turbidity, TSS, TDS, Cu, Fe, Mn, Pb, and Zn, indicating an ‘industrial dust’ origin. PC2 is highly loaded on total coliforms, COD, Co, Cu, Fe, Mn, and Ni. The highest loadings for PC2 are for Co, Fe, and Ni, and hence PC2 has a ‘steel’ origin (Huston et al., 2012). PC3 was highly loaded on TDS, Co, Ni, Pb, and Zn, and is thus characterised as a ‘roadside dust’ factor. PC4

had high loadings for E. coli, total coliforms, turbidity, and PO4-P, indicating faecal contamination from small mammals and birds, which is similar to the principal component reported by Vialle et al. (2011). PC5 was highly loaded on BOD5 and COD, indicating an ‘organic decay’ origin from the leaves, lichen and mosses growing on the roofs. PC6 was characterised by NH3-N and Mn, which is

Biplot (axes PC1 and PC2: 37.32 %)

3

Total coliforms

2

COD

Ni Co

PC2 (15.30 %)

Fe Cu

1

NH3-N

E. coli Mn

Pb

Colour

Zn

TDS BOD5

0

Turbidity

TSS pH

-1 PO4-P -2 -3

-2

-1 S1

0 1 PC1 (22.02 %) S2

S3

S4

2 S5

3

4

S6

Fig. 4. Correlation biplot for PCA analysis of rainwater samples.

Please cite this article in press as: Leong, J.Y.C., et al., Longitudinal assessment of rainwater quality under tropical climatic conditions in enabling effective rainwater harvesting and reuse schemes, Journal of Cleaner Production (2017), http://dx.doi.org/10.1016/j.jclepro.2016.12.149

10

J.Y.C. Leong et al. / Journal of Cleaner Production xxx (2017) 1e12

sourced from fertilisers. PC7 was characterised by Cu and Pb, which indicates a ‘plumbing’ origin. Fig. 4 shows a correlation biplot. Both scores and loadings (presented as vectors) of PC1 and PC2 are presented to elucidate the differences between sites. The biplot visually confirms that Site 2 has more E. coli, total coliforms, Fe, and Mn than other sites, as several observations fall in the same direction of the loading vector. Sites 5 and 6 have low E. coli, total coliform, colour, TDS, and BOD5 concentrations. Overall, the biplot shows visually that harvested rainwater has low BOD5, COD, Ni, Co, total coliforms, TSS, and TDS concentrations, as the majority of the points are situated in the opposite direction of the aforementioned vectors. 3.7. Positive matrix factorisation (PMF) PMF was conducted on 37 samples (Table 5). Fig. 5 shows the source profiles for the harvested rainwater samples. Seven source factors were identified with a Qrobust/Qtrue ratio of 0.95. The number of source factors was decided based on the ratio of the Qrobust to Qtrue, and the relevance of resolved factors to known sources in the area. Estimated errors from DISP, BS, and BS-DISP are shown in Fig. 6. Swaps in the PMF solution indicate

rotational ambiguity, and thus the solution is not robust enough to be used for further analysis. Fig. 6 shows small error values with no swaps present, confirming that the seven-factor solution is stable and appropriate. The source profiles identified by PMF are similar to the source profiles identified by PCA. Major components for F1 include PO4-P, pH, and COD. This factor is regarded as contribution of faeces from birds and small mammals, and is similar to PC4. F2 is the key contributor to Fe and Mn, which showcases a ‘steel’ origin (Huston et al., 2012), much like PC2. F3 represents an ‘industrial dust’ origin due to the contributions of Zn, Mn, and Pb, and is similar to PC1. F4 consisted of colour, turbidity, and TSS, and is associated with ‘roadside dust’, similar to PC3. This is because a previous study has shown that a large proportion of the solids in harvested rainwater are sourced from ‘roadside dust’ (Huston et al., 2012). F5 consists mostly of Cu and Pb, and therefore has a ‘plumbing’ origin, similar to PC7. This result is concurrent with previous studies (Huston et al., 2012), showing that pipes are a potential source of heavy metals in harvested rainwater. F6 is loaded only by NH3-N, and is assumed to be the result of ‘fertilisers’, as seen in PC6. F7 consists only of TDS, which is attributed to the contribution of dissolved ions in mains water top-up.

s Fig. 5. Source (factor) profiles from PMF analysis for rainwater quality data.

Please cite this article in press as: Leong, J.Y.C., et al., Longitudinal assessment of rainwater quality under tropical climatic conditions in enabling effective rainwater harvesting and reuse schemes, Journal of Cleaner Production (2017), http://dx.doi.org/10.1016/j.jclepro.2016.12.149

J.Y.C. Leong et al. / Journal of Cleaner Production xxx (2017) 1e12

11

the variance were identified: ‘industrial dust’, ‘steel’, ‘roadside dust’, ‘faeces’, ‘organic decay’, ‘fertilisers’, and ‘plumbing’. Similarly, 7 source profiles with a Qrobust/Qtrue value of 0.95 were identified with PMF. In PMF, an additional factor accounting for TDS was found to be ‘mains water top-up’ (F7). The results of PMF agreed with PCA. Thus, materials for a rainwater harvesting system should be carefully selected to reduce contributions from ‘steel’ and ‘plumbing’, while contributions from ‘organic decay’, ‘faeces’, ‘industrial dust’ and ‘roadside dust’ may be reduced with first-flush diverters and regular tank de-sludging. Acknowledgements The authors gratefully acknowledge the financial aid from Advanced Engineering Platform of Monash University Malaysia (AEP-16-005). The authors are also indebted to the rainwater site owners for their assistance and patience in obtaining rainwater samples, as well as to Ms Hooi Leng Ser and Dr Lee Learn Han’s assistance with qualitative PCR analysis. References

Fig. 6. Error estimation results from displacement (DISP), bootstrapping (BS), and BSDISP methods in EPA PMF 5.0. Small errors show that the solution is stable and thus may be interpreted.

4. Conclusions Rainwater quality was monitored from 6 full-scale rainwater harvesting systems in Selangor, Malaysia over 8 months to yield a total of 92 samples. Harvested rainwater generally had excellent physicochemical quality, although several violations of Malaysian water quality standards for recreational waters with body contact (Class IIB) was observed: pH (18/92), NH3-N (1/92), PO4-P (3/92), and total coliforms (8/92). 22/92 samples (24%) tested positive for E. coli, 64/92 (70%) tested positive for total coliforms, and 7 instances of C. violaceum were found during monitoring. These findings indicate the importance of annual tank de-sludging, regular cleaning of roofs and gutters, and additionally indicate that disinfection is mandatory for rainwater harvesting systems in order to minimise health risks from rainwater reuse. Furthermore, 37 harvested rainwater samples were analysed for 7 heavy metals Co, Cu, Fe, Mn, Ni, Pb, and Zn. 2 samples (one each from Sites 5 and 6) exceeded Malaysian drinking water limits of 10 mg/L, showing that there may be a health risk if harvested rainwater is consumed directly without treatment for heavy metals. One of the sites (Site 2) mixed rainwater with groundwater, resulting in statistically significant (p < 0.05) higher median concentrations of PO4-P and total coliforms from other sites based on a non-parametric KruskalWallis test. Groundwater should hence not be mixed with rainwater. Furthermore, median concentrations of turbidity, TSS, and E. coli were significantly higher (p < 0.05) during the wet season as a result of the long antecedent dry period, showing the need for higher disinfectant doses during the wet season. Both PCA and PMF were utilised to extract information about the pollutant sources on 37 samples. 7 principal components which accounted for 79.8% of

Albrechtsen, H.-J., 2002. Microbiological investigations of rainwater and graywater collected for toilet flushing. Water Sci. Technol. 46, 311e316. American Public Health Association (APHA), 2005. Standard Methods for the Examination of Water and Wastewater. APHA-AWWA-WEF, Washington, D.C. Amin, M., Han, M., 2009. Roof-harvested rainwater for potable purposes: application of solar collector disinfection (SOCO-DIS). Water Res. 43, 5225e5235. Campisano, A., Gnecco, I., Modica, C., Palla, A., 2013. Designing domestic rainwater harvesting systems under different climatic regimes in Italy. Water Sci. Technol. 67, 2511e2518. Chong, M.N., Sidhu, J., Aryal, R., Tang, J., Gernjak, W., Escher, B., Toze, S., 2013. Urban stormwater harvesting and reuse: a probe into the chemical, toxicology and microbiological contaminants in water quality. Environ. Monit. Assess. 185, 6645e6652. Coombes, P.J., Argue, J.R., Kuczera, G., 2000. Figtree Place: a case study in water sensitive urban development (WSUD). Urban Water 1, 335e343. Crabtree, K.D., Ruskin, R.H., Shaw, S.B., Rose, J.B., 1996. The detection of Cryptosporidium oocysts and Giardia cysts in cistern water in the US Virgin Islands. Water Res. 30, 208e216. Das, R., Das, S.N., Misra, V.N., 2005. Chemical composition of rainwater and dustfall at Bhubaneswar in the east coast of India. Atmos. Environ. 39, 5908e5916. Despins, C., Farahbakhsh, K., Leidl, C., 2009. Assessment of rainwater quality from rainwater harvesting systems in Ontario, Canada. J. Water Supply Res. Technol.dAQUA 58, 117e134. Du, Y.-J., Wei, M.-L., Reddy, K.R., Liu, Z.-P., Jin, F., 2014. Effect of acid rain pH on leaching behavior of cement stabilized lead-contaminated soil. J. Hazard. Mater. 271, 131e140. n, N., Menck, C.F., 2001. Chromobacterium violaceum: a review of pharmacoDura logical and industrial perspectives. Crit. Rev. Microbiol. 27, 201e222. Evans, C., Coombes, P., Dunstan, R., 2006. Wind, rain and bacteria: the effect of weather on the microbial composition of roof-harvested rainwater. Water Res. 40, 37e44. Farreny, R., Morales-Pinzon, T., Guisasola, A., Taya, C., Rieradevall, J., Gabarrell, X., 2011. Roof selection for rainwater harvesting: quantity and quality assessments in Spain. Water Res. 45, 3245e3254. Ghisi, E., Mengotti de Oliveira, S., 2007. Potential for potable water savings by combining the use of rainwater and greywater in houses in southern Brazil. Build. Environ. 42, 1731e1742. Gikas, G.D., Tsihrintzis, V.A., 2012. Assessment of water quality of first-flush roof runoff and harvested rainwater. J. Hydrol. 466, 115e126. Huston, R., Chan, Y., Chapman, H., Gardner, T., Shaw, G., 2012. Source apportionment of heavy metals and ionic contaminants in rainwater tanks in a subtropical urban area in Australia. Water Res. 46, 1121e1132. Huston, R., Chan, Y., Gardner, T., Shaw, G., Chapman, H., 2009. Characterisation of atmospheric deposition as a source of contaminants in urban rainwater tanks. Water Res. 43, 1630e1640. Kaiser, H.F., 1960. The application of electronic computers to factor analysis. Educ. Psychol. Meas. 20, 141e151. Kim, R.-H., Lee, S., Jeong, J., Lee, J.-H., Kim, Y.-K., 2007. Reuse of greywater and rainwater using fiber filter media and metal membrane. Desalination 202, 326e332. Lara, L.B.L.S., Artaxo, P., Martinelli, L.A., Victoria, R.L., Camargo, P.B., Krusche, A., Ayers, G.P., Ferraz, E.S.B., Ballester, M.V., 2001. Chemical composition of rainwater and anthropogenic influences in the Piracicaba river basin, southeast Brazil. Atmos. Environ. 35, 4937e4945. Lee, J.Y., Bak, G., Han, M., 2012. Quality of roof-harvested rainwatereComparison of different roofing materials. Environ. Pollut. 162, 422e429.

Please cite this article in press as: Leong, J.Y.C., et al., Longitudinal assessment of rainwater quality under tropical climatic conditions in enabling effective rainwater harvesting and reuse schemes, Journal of Cleaner Production (2017), http://dx.doi.org/10.1016/j.jclepro.2016.12.149

12

J.Y.C. Leong et al. / Journal of Cleaner Production xxx (2017) 1e12

Leong, J.Y.C., Oh, K.S., Poh, P.E., Chong, M.N., 2016. Prospects of hybrid rainwatergreywater decentralised system for water recycling and reuse: a review. J. Clean. Prod. http://dx.doi.org/10.1016/j.jclepro.2016.10.167. Macauley, J.J., Qiang, Z., Adams, C.D., Surampalli, R., Mormile, M.R., 2006. Disinfection of swine wastewater using chlorine, ultraviolet light and ozone. Water Res. 40, 2017e2026. Magyar, M., Mitchell, V., Ladson, A., Diaper, C., 2007. An investigation of rainwater tanks quality and sediment dynamics. Water Sci. Technol. 56, 21e28. Mendez, C.B., Klenzendorf, J.B., Afshar, B.R., Simmons, M.T., Barrett, M.E., Kinney, K.A., Kirisits, M.J., 2011. The effect of roofing material on the quality of harvested rainwater. Water Res. 45, 2049e2059. Moreira Neto, R.F., Calijuri, M.L., Carvalho, I. d. C., Santiago, A. d. F., 2012. Rainwater treatment in airports using slow sand filtration followed by chlorination: efficiency and costs. Resour. Conserv. Recycl. 65, 124e129. Nawaz, M., Han, M., Kim, T.-i., Manzoor, U., Amin, M., 2012. Silver disinfection of Pseudomonas aeruginosa and E. coli in rooftop harvested rainwater for potable purposes. Sci. Total Environ. 431, 20e25. Needleman, H., 2004. Lead poisoning. Annu. Rev. Med. 55, 209e222. Paatero, P., Tapper, U., 1994. Positive matrix factorization: a non-negative factor model with optimal utilization of error estimates of data values. Environmetrics 5, 111e126. Reff, A., Eberly, S.I., Bhave, P.V., 2007. Receptor modeling of ambient particulate matter data using positive matrix factorization: review of existing methods. J. Air Waste Manag. Assoc. 57, 146e154. Sazakli, E., Alexopoulos, A., Leotsinidis, M., 2007. Rainwater harvesting, quality

assessment and utilization in Kefalonia Island, Greece. Water Res. 41, 2039e2047. Simmons, G., Hope, V., Lewis, G., Whitmore, J., Gao, W., 2001. Contamination of potable roof-collected rainwater in Auckland, New Zealand. Water Res. 35, 1518e1524. Uba, B.N., Aghogho, O., 2000. Rainwater quality from different roof catchments in the Port Harcourt district, Rivers State, Nigeria. J. Water Supply Res. Technol.dAQUA 49, 281e288. US EPA, 2014. EPA Positive Matrix Factorization (PMF) 5.0 Fundamentals and User Guide. Available: https://www.epa.gov/sites/production/files/2015-02/ documents/pmf_5.0_user_guide.pdf (Accessed 26 December 2016). zquez, A., Costoya, M., Pena, R.M., García, S., Herrero, C., 2003. A rainwater quality Va monitoring network: a preliminary study of the composition of rainwater in Galicia (NW Spain). Chemosphere 51, 375e386. jaudVialle, C., Sablayrolles, C., Lovera, M., Jacob, S., Huau, M.-C., Montre Vignoles, M., 2011. Monitoring of water quality from roof runoff: interpretation using multivariate analysis. Water Res. 45, 3765e3775. Ward, S., Memon, F., Butler, D., 2010. Harvested rainwater quality: the importance of appropriate design. Water Sci. Technol. 61, 1707e1714. Yaziz, M.I., Gunting, H., Sapari, N., Ghazali, A.W., 1989. Variations in rainwater quality from roof catchments. Water Res. 23, 761e765. Zhu, K., Zhang, L., Hart, W., Liu, M., Chen, H., 2004. Quality issues in harvested rainwater in arid and semi-arid Loess Plateau of northern China. J. Arid Environ. 57, 487e505.

Please cite this article in press as: Leong, J.Y.C., et al., Longitudinal assessment of rainwater quality under tropical climatic conditions in enabling effective rainwater harvesting and reuse schemes, Journal of Cleaner Production (2017), http://dx.doi.org/10.1016/j.jclepro.2016.12.149