Rainwater is widely collected and stored in cisterns for domestic use due to extreme water scarcity and poor water quality on the Loess Plateau, China. However, little information exists on the effects of cistern materials and construction methods on water quality. This study assessed the quality of rainwater and harvested rainwater stored in different types of cisterns, using multivariate statistical analysis techniques. The harvesting system consisted of ditches underground to direct the water stream collected by catchments through pipes to the two-stage grit chambers, the rainwater runoff ultimately being fed into the cisterns. Water samples were collected monthly over 6 months from October 2008 to April 2009. The harvested rainwater generally did not meet drinking water standards due to severe bacterial contamination. To determine the factors affecting water quality, principal components and factor analysis divided the water's physicochemical composition into four principal components: ions, suspended pollutants, reducing substances, and acidity-alkalinity, which collectively accounted for 80.4% of the total variance. Cluster analysis confirmed the results. It was determined that pollutants in harvested rainwater were mainly derived from the catchments, surrounding fields, and local atmosphere. Furthermore, factor scores ranked water quality to select the optimum material and construction method. It was concluded that cement was superior to stabilized soil for rainwater storage. However, considering water quality, cost, and environmental friendliness the stabilized soil cistern using the stiff construction method was worthy of consideration.

INTRODUCTION

Throughout the world, water supply and water safety have become critical issues in many countries. These concerns are principally derived from intensive urbanization, population growth, land-use transformation, pollution, and changing climate patterns (Vialle et al. 2011; Gikas & Tsihrintzis 2012). Drinking water safety usually involves two aspects: the difficulty of getting water and the quality of water.

Compared to surface and ground water, which are often reported to be contaminated, have decreased volume, and are usually very costly to develop as new water sources, rainwater capture and storage appears to be one of the most promising alternatives for supplying fresh water (Zhu et al. 2004). Recently, a number of countries throughout the world have studied rainwater harvesting, including Bangladesh, Brazil, Greece, China, Ghana, Pakistan, Vietnam, and America (Zhu et al. 2004; Sazakli et al. 2007; Islam et al. 2010; Wilbers et al. 2013; Alves et al. 2014; Brahman et al. 2014; Thomas et al. 2014; Cobbina et al. 2015). Usually, harvested rainwater can be a good alternative water resource for potable water, cleaning water, and irrigation water. Although rainwater harvesting is an attractive solution to cope with water scarcity from the ecological perspective, the quality of the harvested rainwater should be a concern due to the potential for health risks caused by physiochemical and microbiological pollutants (Vialle et al. 2011; de Man et al. 2014).

The quality of harvested rainwater can be affected by numerous factors, including geographical location, rainwater quality, spatial conditions (proximity of industry, main roads, coastline, etc.), harvesting system (catchment area and storage types), storage time, and the management of water (Martin et al. 2010; Amin et al. 2013). Harvested rainwater usually has good physicochemical parameters, but often has a high level of microbiological contamination. As a result, the sanitary quality of harvested rainwater has received widespread attention. The pollution in harvested rainwater can be attributed to the combustion of fossil fuels in vehicles and buildings, emissions of industries, agricultural activities (runoff or leaching of fertilizer and pesticide) in rural areas and fecal deposits from animals (Evans et al. 2006).

Harvested rainwater has a long history of use on the Loess Plateau in northwest China, an area that has water scarcity and poor water quality. The average annual precipitation on the Loess Plateau ranges from 200 mm to 750 mm (Li et al. 2010), and mainly occurs from June to September. Because of the lack of a reliable supply of surface water and groundwater, rainwater harvesting has always played a prominent role in the water supply for domestic and agricultural use in rural areas on the Loess Plateau (Zhu et al. 2004). Local people have preferred to use a ready-made yard as catchment to collect rainwater and construct permanent cisterns underground to store harvested rainwater. The collected water can then be used throughout the whole year for domestic use, such as drinking, laundry and other uses. Although harvested rainwater has alleviated the water shortage in this area, it should be noted that it often does not meet drinking water standards. Thus, it is necessary to monitor the harvested rainwater in order to find appropriate ways to improve the water quality.

The quality of stored rainwater depends on the conditions of the catchment and storage tank and the antecedent dry period (Lee et al. 2011). Zhao et al. (2010) reported that water quality varied differently over time in different types of cisterns. O'Hogain et al. (2012) demonstrated an improved harvesting system. Also, advanced handling and management of water could obviously improve the quality of harvested rainwater (Moreira Neto et al. 2012).

The objectives of this research were to: (a) assess the quality of rainwater and stored rainwater with respect to drinking water standards; (b) extract the principal components of water physicochemical composition to find factors affecting water quality; and (c) compare water quality from different types of cisterns to make recommendations on best materials and construction methods for rainwater storage.

MATERIALS AND METHODS

Study region and rainwater harvesting system

A rainwater harvesting system (Figure 1) was constructed in the Water-Saving Engineering Technology Research Center in Yangling, Shaanxi, China, in July 2007. Yangling District (108°00′E–108°07′E, 34°12′N–34°20′N) is located on Guanzhong Plain. The average annual temperature in this region is 13°C. The average annual rainfall is 638 mm, concentrated mostly in June to September, and the average annual evaporation is 1,110 mm. As shown in Figure 1(a) and 1(b), in the rainwater harvesting system, rainwater is first collected by the stabilized-soil catchments and then directed via a collecting ditch (0.3 m × 0.3 m × 3.5 m) through a water pipe to the two-stage grit chambers. After a two-step precipitation, the water stream is passed through a fine-mesh filter, and then fed into a cistern with an 8 m3 capacity. People can use buckets to get water since no pumps were installed considering the cost of electricity. The valve at the end of the water pipe (Φ110, Figure 1(b)) can be closed to avoid an overflow. Excess water can flow upward from the ditch to the catchments and finally is fed into a channel.
Figure 1

Schematic diagram of yard rainwater harvesting system and configuration of water cisterns.

Figure 1

Schematic diagram of yard rainwater harvesting system and configuration of water cisterns.

The four cisterns, abbreviated as C1, C2, C3 and C4, have similar configurations (Figure 1(c) and 1(d)), but different materials and construction methods (Fan et al. 2006; Xu & Gao 2014). The core materials of the cisterns are divided into cement and MBER (Material Becoming Earth into Rock) soil stabilizer, consisting of cementitious material (calcium silicate), alkali catalyst, surfactant, and slag in certain proportions. MBER soil stabilizer, enhancing the strength of soil through chemical reactions, makes soil into a kind of stable and durable material. A small amount of MBER can be mixed with a large amount of soil, forming stabilized soil, which costs much less than cement. It was calculated that the costs of the stabilized-soil catchment ranged from 7.1 yuan/m2 to 8.2 yuan/m2, while the concrete catchment cost 13.6 yuan per square metre (Fan et al. 2006). Using stabilized soil can not only reduce the cost of transporting sand and gravels, but also consume industrial wastes such as coal fly ash, mine refuse, and steel slag (Xu & Gao 2014). Thus, stabilized soil is a kind of economical and environmentally friendly material. The cistern built by cement with little gravels was called C1. Three construction methods of stabilized soil were used: making stabilized soil into bricks first and then using the bricks to build the cistern (C2); adding a small amount of water to the stabilized soil, classified as the stiff construction method (C3); and the plastic construction method (C4), which is the stabilized soil with water added to have a higher water content. The locations of C1, C2, C3 and C4 are shown in Figure 1(a). The stabilized soil catchments (3.5 m wide) are cut off by water collecting ditches: 4 m for C1, 5.5 m for C2, 7 m for C3, and 9 m for C4, according to the construction situation and convenience of fetching water for different uses (Figure 1(a)).

Sample collection and analysis

Rainwater samples were collected at different time intervals from one rainstorm event on October 9, 2008. Three polyethylene plastic buckets (4 L) were positioned on the roof (Figure 1(a)) and first collected rainwater for 20 minutes. Then the three water samples were mixed uniformly and loaded into a polyethylene bottle. Similarly, another four samples were obtained approximately every 40 minutes. Prior to the rainfall, the four cisterns were cleaned and disinfected to prepare for harvested rainwater storage. Then, the stored water was monitored monthly from October 2008 to April 2009. Over the monitoring period, no new rainwater was inserted in the cisterns and people did not use the stored water. From one storm event we can compare the quality of rainwater and harvested rainwater, in order to find water pollution sources and observe the effects of materials and construction methods on water quality. Fourteen parameters including pH, conductivity (EC), turbidity, chemical oxygen demand (COD), NO2, NO3, NH4+, TN, SO42−, Cl, F, Fe and total bacterial count and total coliforms were tested using the following devices and methods (Table 1). Microbiological parameters were measured according to Standard Examination Methods for Drinking Water – Microbiological Parameters (GB/T 5750.12-2006). Seven samples were tested from each cistern during the monitoring period, and were collected by pre-disinfected stainless steel buckets. Since the methods for sampling water from the four cisterns were the same, the effect of the buckets on the study results was neglected. The pH, turbidity and conductivity were measured on site. Samples collected for chemical analysis and for microbiological analysis were stored in polyethylene bottles and aseptic bottles, separately. All samples were stored in a refrigerator at 4 °C for later detection. Microbiological parameters were tested within 24 h while chemical parameters were measured within 72 h.

Table 1

Detection devices and methods used to measure water quality parameters

NumberParametersUnitsDevices and methodsNumberParametersUnitsDevices and methods
pH – HACH SENSION 1 TN mg·L−1 HACH DR2800 
Turb NTU HACH 2100P SO42− mg·L−1 HACH DR2800 
EC μS·cm−1 FE30 10 Cl mg·L−1 HACH DR2800 
COD mg·L−1 HACH DR2800 11 F mg·L−1 HACH DR2800 
NO3 mg·L−1 HACH DR2800 12 Fe mg·L−1 HACH DR2800 
NO2 mg·L−1 HACH DR2800 13 Total bacterial count CFU·mL−1 Plate count method 
NH4+ mg·L−1 HACH DR2800 14 Total coliforms MPN·(100 mL)−1 Multiple-tube fermentation technique 
NumberParametersUnitsDevices and methodsNumberParametersUnitsDevices and methods
pH – HACH SENSION 1 TN mg·L−1 HACH DR2800 
Turb NTU HACH 2100P SO42− mg·L−1 HACH DR2800 
EC μS·cm−1 FE30 10 Cl mg·L−1 HACH DR2800 
COD mg·L−1 HACH DR2800 11 F mg·L−1 HACH DR2800 
NO3 mg·L−1 HACH DR2800 12 Fe mg·L−1 HACH DR2800 
NO2 mg·L−1 HACH DR2800 13 Total bacterial count CFU·mL−1 Plate count method 
NH4+ mg·L−1 HACH DR2800 14 Total coliforms MPN·(100 mL)−1 Multiple-tube fermentation technique 

Multivariate statistical analysis

Principal components analysis (PCA) is a widely used approach to extract the main information by reducing dimensions while retaining the maximum information on the variables. Specifically, PCA is performed by three mathematical steps: (1) the standardization of measurements to avoid the effects of magnitudes and units; (2) calculation of the eigenvalues and the corresponding eigenvectors of the covariance matrix; (3) the extraction of significant components that account for a large proportion of the variance in the datasets (eigenvalues ≥ 1) (Vialle et al. 2011).

Factor analysis (FA) is designed to extract the most significant variables obtained from PCA. The new group of variables, known as varifactors (VFs), is extracted by rotating the axis as defined by PCA (FA can also use other methods to extract VFs). The VFs can include unobservable, hypothetical, and latent variables, whereas PCA can only be a linear combination of observable water quality variables (Kannel et al. 2007). Factors can be represented as a linear combination of variables. Then these factors are used with their variance proportions as weights to calculate factors' scores. Generally, PCA/FA is a widely used pattern to recognize pollution sources in water quality analysis.

Cluster analysis (CA) is a method to distinguish natural groups in a dataset according to the degree of similarity (using Euclidean distances as a measurement) between standardized variables. The normalized dataset was treated with the method of the average linkage between groups of similarity. In this way, each parameter is considered as a separate group at first, then different groups make up several bigger ones, and ultimately all groups merge into a big cluster. The number of groups is determined according to the needs of data interpretation (usually considering the analysis results of PCA/CA). CA can be considered a good complement to PCA/FA if clusters nearly correspond to the principal components.

RESULTS AND DISCUSSION

Quality of rainwater and harvested rainwater

Statistics of physicochemical and microbial parameters of rainwater and harvested rainwater are assessed with Chinese drinking water guidelines (GB 5749-2006) in Table 2. Figure 2(a)2(l) show box–whisker plots of pH, conductivity, turbidity, COD, NO2, NO3, NH4+, TN, SO42−, Cl, F, Fe and total bacteria count. The box contains lower and upper quartiles and the interior line is the median value. The lower and upper lines out of each box present the minimum and the maximum, respectively.
Table 2

Water quality parameters (total n = 33) compared with drinking water standards

RainwaterC1C2C3C4 
VariablesUnits(n = 5)(n = 7)(n = 7)(n = 7)(n = 7)China drinking water guidelines
pH 
 Mean – 7.48 7.81 8.01 7.96 7.96 6.5–8.5 
 SD  0.99 0.37 0.25 0.28 0.28  
Conductivity 
 Mean μS·cm−1 94.28 410 406.57 400.57 400.57 – 
 SD  45.02 21.6 20.02 13.72 13.72  
Turbidity 
 Mean NTU 4.49 1.36 4.14 1.97 1.97 
 SD  2.72 1.1 4.58 2.16 2.16  
COD 
 Mean mg L−1 2.16 2.1 4.21 5.43 5.43 
 SD  2.01 0.54 2.29 3.67 3.67  
NO2 
 Mean mg L−1 0.03 0.36 0.34 0.28 0.28 
 SD  0.03 0.21 0.16 0.15 0.15  
NO3 
 Mean mg L−1 1.54 2.81 3.57 3.26 3.26 10 
 SD  0.39 1.04 0.93 0.75 0.75  
NH4+ 
 Mean mg L−1 3.4 0.96 0.7 0.51 0.51 0.5 
 SD  1.96 0.47 0.45 0.38 0.38  
TN 
 Mean mg L−1 6.8 2.51 3.24 3.54 3.54 – 
 SD  2.22 1.01 0.78 0.69 0.69  
SO42− 
 Mean mg L−1 20.08 161.86 181.14 205.43 205.43 250 
 SD  14.92 18.27 26.57 35.04 35.04  
Cl 
 Mean mg L−1 2.83 34.85 38.45 44.22 44.22 250 
 SD  2.34 4.72 4.59 9.02 9.02  
F 
 Mean mg L−1 0.16 0.48 0.53 0.64 0.64 
 SD  0.18 0.14 0.15 0.17 0.17  
Fe 
 Mean mg L−1 0.026 0.058 0.045 0.056 0.056 0.3 
 SD  0.007 0.044 0.032 0.067 0.067  
Total bacteria 
 Mean CFU mL−1 – 1,421 1,433 1,278 1,278 100 
 SD  – 2,570 2,304 2,023 2,023  
Total coliforms 
 – MPN (100 mL)−1 
RainwaterC1C2C3C4 
VariablesUnits(n = 5)(n = 7)(n = 7)(n = 7)(n = 7)China drinking water guidelines
pH 
 Mean – 7.48 7.81 8.01 7.96 7.96 6.5–8.5 
 SD  0.99 0.37 0.25 0.28 0.28  
Conductivity 
 Mean μS·cm−1 94.28 410 406.57 400.57 400.57 – 
 SD  45.02 21.6 20.02 13.72 13.72  
Turbidity 
 Mean NTU 4.49 1.36 4.14 1.97 1.97 
 SD  2.72 1.1 4.58 2.16 2.16  
COD 
 Mean mg L−1 2.16 2.1 4.21 5.43 5.43 
 SD  2.01 0.54 2.29 3.67 3.67  
NO2 
 Mean mg L−1 0.03 0.36 0.34 0.28 0.28 
 SD  0.03 0.21 0.16 0.15 0.15  
NO3 
 Mean mg L−1 1.54 2.81 3.57 3.26 3.26 10 
 SD  0.39 1.04 0.93 0.75 0.75  
NH4+ 
 Mean mg L−1 3.4 0.96 0.7 0.51 0.51 0.5 
 SD  1.96 0.47 0.45 0.38 0.38  
TN 
 Mean mg L−1 6.8 2.51 3.24 3.54 3.54 – 
 SD  2.22 1.01 0.78 0.69 0.69  
SO42− 
 Mean mg L−1 20.08 161.86 181.14 205.43 205.43 250 
 SD  14.92 18.27 26.57 35.04 35.04  
Cl 
 Mean mg L−1 2.83 34.85 38.45 44.22 44.22 250 
 SD  2.34 4.72 4.59 9.02 9.02  
F 
 Mean mg L−1 0.16 0.48 0.53 0.64 0.64 
 SD  0.18 0.14 0.15 0.17 0.17  
Fe 
 Mean mg L−1 0.026 0.058 0.045 0.056 0.056 0.3 
 SD  0.007 0.044 0.032 0.067 0.067  
Total bacteria 
 Mean CFU mL−1 – 1,421 1,433 1,278 1,278 100 
 SD  – 2,570 2,304 2,023 2,023  
Total coliforms 
 – MPN (100 mL)−1 
Figure 2

Quality of rainwater and harvested rainwater stored in different types of cisterns over the experimental period (the dotted line indicating limits for parameters).

Figure 2

Quality of rainwater and harvested rainwater stored in different types of cisterns over the experimental period (the dotted line indicating limits for parameters).

Values of pH in rainwater and harvested rainwater ranged from 6.55 to 8.65, and from 7.36 to 8.45, respectively, both of which were generally within drinking water standards (6.5–8.5). The rainwater washing process on air pollutants is within the first 10–20 min, when rainfall water has the dominant part of pollutants (Wei et al. 2015). The initial pH value of rainwater was less than 7.0 since SO2 and NOx pollutants in the air were washed during the first 20 minutes. Then, the pH value became higher due to the low concentration of H+. Since the first-flush rainwater runoff was discarded, the initial pH value of the harvested rainwater did not change greatly during the monitoring period. In addition, the four cisterns were constructed in July 2007, and so the release of alkali materials used to build the four cisterns was decreased. Thus, the effects of cistern materials on water pH were small and pH variation in C1–C4 was less than that of the rainwater. Mean pH values in rainwater and harvested rainwater were 7.48 and 7.96, respectively, both subalkalic. There were no significant differences of pH values between rainwater and harvested rainwater (P > 0.05), indicating that pH values of harvested rainwater were mainly influenced by local atmospheric conditions instead of the rainwater collecting and storing system in this study. This result also coincided with the study by Gikas & Tsihrintzis (2012).

The mean value of conductivity (EC) in rainwater was less than 100 μS·cm−1 indicating the ion concentration of the rainwater was low. However, the conductivity range of samples collected from the cisterns was 380–440 μS·cm−1. This result suggested that some substances accumulated by catchment or released from cisterns increased the ion concentration. Differences of conductivity values between rainwater and harvested rainwater were significant (P < 0.05; also see Figure 2(b)). However, there were no obvious differences of conductivity values between cisterns (P > 0.05). Other studies have shown that the first flush flowing through a catchment can have a significantly higher level of conductivity than rainwater or stored rainwater when the first flush is discarded (Lee et al. 2010; Gikas & Tsihrintzis 2012). Turbidity values in rainwater and harvested rainwater ranged from 2.00 NTU to 8.3 NTU and 0.50 NTU to 13.9 NTU, respectively. The mean turbidity value of rainwater was 4.49 NTU, exceeding the Chinese drinking water limit (3 NTU), and 43% of harvested water samples exceeded the drinking water limitation for turbidity.

COD is an indicator of the reducing substances in water. Mean COD values were 2.16 mg L−1 in rainwater and 2.10 mg L−1 (C1), 4.22 mg L−1 (C2), 5.41 mg L−1 (C3) and 10.97 mg L−1 (C4) in harvested rainwater. Note that the catchment area increased going from the closest (C1) to the farthest (C4) cistern as shown in Figure 1(a). This suggests that with increased catchment area, the potential for more contaminated materials to move with the water stream and into the cisterns is also increased. In total, 36% of the samplings from the cisterns exceeded drinking water limits (5 mg L−1). More specifically, water samples above the COD limit accounted for 0%, 29%, 29% and 86% for C1, C2, C3 and C4, respectively. COD values for C4 were significantly higher than for rainwater, C1, C2 and C3, as shown in Figure 2(d).

Nitrogen variations in rainwater and harvested rainwater can be seen in Figure 2(e)2(g). In rainwater and harvested rainwater, nitrate (NO3) concentrations ranged from 1.1 to 2.0 mg L−1 and from 1.1 to 5.4 mg L−1, respectively; nitrite (NO2) concentrations varied from 0 to 0.060 mg L−1 and from 0.105 to 0.598 mg L−1, respectively; ammonium nitrogen (NH4+) concentrations ranged from 1.10 to 6.00 mg L−1 and from 0.10 to 1.60 mg L−1, respectively. NO3 and NO2 were the final forms of nitrogen in the harvested rainwater. Mean values of NO3 , NO2 and NH4+ in rainwater were 1.54 mg L−1, 0.026 mg L−1 and 3.4 mg L−1, respectively, and in harvested rainwater 3.3 mg L−1, 0.300 mg L−1and 0.7 mg L−1, respectively. Mean values of NO3 and NO2 in harvested rainwater were 2.1 and 11.5 times higher than in rainwater, respectively. However, the mean value of NH4+ in harvested rainwater decreased to only 20.6% of that in rainwater. NO3 and NO2 in rainwater were significantly lower than in harvested rainwater (P < 0.05); however, NH4+ was significantly higher in rainwater than in harvested rainwater (P < 0.05). This finding suggested ammonium nitrogen finally transferred to nitrate nitrogen in the storage water, which resulted in the striking increase in NO3 and NO2 (Gikas & Tsihrintzis 2012).

All samples collected from both rainwater and harvested rainwater met the drinking water standards for NO3 (10 mg L−1) and NO2 (1 mg L−1). However, all rainwater samples exceeded the drinking water limit for ammonium nitrogen (0.5 mg L−1). In the harvested rainwater, 71.4% of samples from C1, 57.1% of samples from C2 as well as C3, and 42.9% of samples from C4 exceeded the drinking water limit for ammonium. Other studies have also reported ammonia nitrogen often has a high level both in rainwater (Niu et al. 2014; Tay et al. 2014) and in harvested rainwater (Gikas & Tsihrintzis 2012; Hogain et al. 2012). Mean ammonium nitrogen concentration was 2.2 times higher than nitrate nitrogen concentration (NO3 + NO2) in rainwater, accounting for 68.5% of inorganic nitrogen. This indicates that ammonia in the local atmosphere (with low industrial activity) was a main form of nitrogen. Ammonium also had an impact on pH values (Gikas & Tsihrintzis 2012), resulting in subalkalic conditions both in rainwater and harvested rainwater. As can be seen in Figure 2(h), total nitrogen (TN) in harvested rainwater was significantly lower than in rainwater (P < 0.05) due to deposition and adsorption in cisterns.

SO42− values in rainwater and harvested rainwater ranged across 3–25 mg L−1 and 30–240 mg L−1, respectively. The mean SO42− in rainwater was 20 mg L−1. In harvested rainwater, the mean SO42− in samples collected from C1, C2, C3 and C4 was 162 mg L−1, 181 mg L−1, 205 mg L−1 and 56 mg L−1, respectively. SO42− values were shown to be significantly different between rainwater and harvested rainwater (P < 0.05; Figure 2(i)). Although all water samples (rainwater and harvested rainwater) met the drinking water standard for SO42− (250 mg L−1), SO42− values in harvested rainwater were much higher than in other studies (Vialle et al. 2011; Hogain et al. 2012). Sulfate has often been attributed to fossil fuel combustion mainly resulting from industrial, traffic, and house-heating emissions (Farreny et al. 2011). Apart from fossil fuel emissions, the materials of catchment and water cisterns contained gypsum, which accounted for a higher level of sulfate values in harvested rainwater. By comparison, sulfate values in rainwater runoff collected in the yard on the semi-arid Loess Plateau were between 2.40 and 15.62 mg L−1; from slope-land they ranged from 5.83 to 19.1 mg L−1; and from roads they were between 7.76 and 32.54 mg L−1 (Zhu et al. 2004).

Cl values in rainwater ranged between 0 and 5.88 mg L−1, and in harvested rainwater they varied from 5.61 to 58.88 mg L−1, respectively. Mean Cl values in rainwater and storage water in C1, C2, C3 and C4 were 2.83 mg L−1, 34.85 mg L−1, 38.45 mg L−1, 44.22 mg L−1 and 7.21 mg L−1, respectively. As can be seen in Figure 2(j), Cl values were significantly different between rainwater and harvested rainwater (P < 0.05). The areas around the collection yard and cisterns were farm fields. Thus, soil can be easily transferred by wind and deposited on the roof, yard and roads. With rainfall, soil particles can move with the water stream into the cisterns, increasing the Cl concentration in the storage rainwater. Fluoride in all rainwater and harvested rainwater samples was less than the drinking water limit (1 mg L−1). Mean F values in rainwater and storage water in C1, C2, C3 and C4 were 0.16 mg L−1, 0.48 mg L−1, 0.53 mg L−1, 0.64 mg L−1 and 0.54 mg L−1, respectively. Mean F values of storage water from C1, C2, C3 and C4 were 3.1, 3.3, 4.0 and 3.4 times greater than rainwater, respectively. This finding showed catchment or cistern conditions had a significant impact on the fluoride concentration in cisterns.

Mean total iron (Fe) was measured at 0.026 mg L−1 for rainwater, 0.058 mg L−1 for C1, 0.044 mg L−1 for C2, 0.056 mg L−1 for C3 and 0.145 mg L−1 for C4. All rainwater samples and harvested rainwater samples were well below the drinking water limit (0.3 mg L−1). As shown in Figure 2(l), Fe values in rainwater and harvested rainwater had no significant differences (P > 0.05), indicating the conditions of catchments and cisterns had little effect on total iron, which might be determined by local atmospheric conditions.

The microbiological parameters including total bacterial count and total coliforms were measured to assess the extent of microbiological contamination of the water. Total bacterial count was not significantly different between cisterns, which indicated that within the same catchment, the storage water had the same level of bacterial contamination. No coliform bacteria were detected in any water samples, indicating the harvested rainwater was not contaminated by feces in the yard-catchment. However, total bacterial counts in water samples from C1 ranged across 0–7,182 CFU mL−1, 54–6,502 CFU mL−1 from C2, 27–5,794 CFU mL−1 from C3, and 0–5,495 CFU mL−1 from C4. Mean total bacterial counts in C1, C2, C3 and C4 were 1,421 CFU mL−1, 1,433 CFU mL−1, 1,278 CFU mL−1 and 1,142 CFU mL−1, respectively. About 86% of all stored water samples exceeded the drinking water limit for total bacterial count (100 CFU.mL−1), which indicated the harvested rainwater did not meet the sanitary requirement.

Even though cistern water collected in the yard generally complied with drinking water limits in terms of physicochemical parameters, total bacterial load often exceeded the limits dramatically. Thus, harvested rainwater in this study was not suitable for drinking unless a process of disinfection was adopted.

Determination of factors affecting water quality

PCA/FA, and CA were applied to determine the factors and mechanisms dominating water quality. Prior to PCA, the Kaiser-Meyer-Olkin test (KMO) of the sampling sufficiency and Bartlett's test of sphericity were examined. Usually, if the KMO ≥0.5, it is appropriate to do FA. The KMO in this study was 0.689 and Bartlett's test of sphericity significance was ≤0.001 (Table 3), indicating FA was appropriate to determine factors affecting water quality in this study.

Table 3

KMO and Bartlett's test

Kaiser-Meyer-Olkin Measure of Sampling Adequacy 0.689 
Bartlett's Test of Sphericity 
 Approx. Chi-Square 263.618 
 df 66 
 Sig. 0.000 
Kaiser-Meyer-Olkin Measure of Sampling Adequacy 0.689 
Bartlett's Test of Sphericity 
 Approx. Chi-Square 263.618 
 df 66 
 Sig. 0.000 

In PCA/FA, four factors were extracted as principal factors (eigenvalues ≥1) that explained approximately 80.4% of the total variance. Specifically, the first factor (F1), the second factor (F2), the third factor (F3) and the fourth factor (F4) accounted for 30.67%, 18.02%, 16.36% and 15.32% of the total variance of the dataset, respectively (Table 4). The first factor loading contributed to NO2, SO42−, F, EC and Cl. Conductivity (EC) reflects total ionic concentration and it had positive correlations with NO2, SO42−, F, and Cl (Table 5). Specifically, correlation between EC and NO2 was 0.537, between EC and SO42− it was 0.641, between EC and F it was 0.671, and between EC and Cl it was 0.572. EC also had a close correlation with NO3 (r = 0.603, P < 0.01). Therefore, the first principal component could be interpreted as the ionic component.

Table 4

Loadings of the first four principal factors F1, F2, F3 and F4 in the total physicochemical dataset (rotated component matrix)

VariableF1F2F3F4
NO2 0.826 0.009 0.153 −0.243 
SO4−2 0.797 −0.205 −0.388 0.180 
F 0.781 −0.241 0.243 0.106 
EC 0.769 −0.154 0.236 0.456 
Cl 0.759 −0.151 −0.400 0.149 
TN −0.139 0.907 −0.006 −0.176 
Turb −0.184 0.852 0.215 −0.015 
Fe 0.050 −0.080 0.855 −0.047 
COD 0.029 0.325 0.832 0.184 
pH −0.079 −0.224 −0.012 0.850 
NO3 0.484 0.294 0.030 0.611 
NH4+ −0.537 0.467 −0.213 −0.585 
Eigenvalue 4.65 2.22 1.58 1.21 
Total variance (%) 30.67 18.02 16.36 15.32 
Cumulative (%) 30.67 48.69 65.05 80.37 
VariableF1F2F3F4
NO2 0.826 0.009 0.153 −0.243 
SO4−2 0.797 −0.205 −0.388 0.180 
F 0.781 −0.241 0.243 0.106 
EC 0.769 −0.154 0.236 0.456 
Cl 0.759 −0.151 −0.400 0.149 
TN −0.139 0.907 −0.006 −0.176 
Turb −0.184 0.852 0.215 −0.015 
Fe 0.050 −0.080 0.855 −0.047 
COD 0.029 0.325 0.832 0.184 
pH −0.079 −0.224 −0.012 0.850 
NO3 0.484 0.294 0.030 0.611 
NH4+ −0.537 0.467 −0.213 −0.585 
Eigenvalue 4.65 2.22 1.58 1.21 
Total variance (%) 30.67 18.02 16.36 15.32 
Cumulative (%) 30.67 48.69 65.05 80.37 
Table 5

Correlation coefficients between physicochemical parameters

 pHECTurbCODNO2NO3NH4+TNSO4−2ClFFe
pH 1.000            
EC 0.313 1.000           
turb −0.116 −0.212 1.000          
COD 0.035 0.265 0.401* 1.000         
NO2 −0.150 0.537** −0.059 0.062 1.000        
NO3 0.238 0.603** 0.019 0.195 0.236 1.000       
NH4+ −0.516** −0.769** 0.432* −0.140 −0.338 −0.482** 1.000      
TN −0.423* −0.756** 0.349* −0.075 −0.541** −0.515** 0.828** 1.000     
SO4−2 0.156 0.641** −0.396* −0.281 0.456** 0.321 −0.534** −0.418* 1.000    
Cl 0.160 0.572** −0.309 −0.274 0.427* 0.244 −0.463** −0.379* 0.950** 1.000   
F 0.107 0.671** −0.317 0.111 0.657** 0.432* −0.637** −0.636** 0.496** 0.438* 1.000  
Fe 0.014 0.211 0.120 0.601** 0.089 −0.058 −0.191 0.003 −0.162 −0.142 0.168 1.000 
 pHECTurbCODNO2NO3NH4+TNSO4−2ClFFe
pH 1.000            
EC 0.313 1.000           
turb −0.116 −0.212 1.000          
COD 0.035 0.265 0.401* 1.000         
NO2 −0.150 0.537** −0.059 0.062 1.000        
NO3 0.238 0.603** 0.019 0.195 0.236 1.000       
NH4+ −0.516** −0.769** 0.432* −0.140 −0.338 −0.482** 1.000      
TN −0.423* −0.756** 0.349* −0.075 −0.541** −0.515** 0.828** 1.000     
SO4−2 0.156 0.641** −0.396* −0.281 0.456** 0.321 −0.534** −0.418* 1.000    
Cl 0.160 0.572** −0.309 −0.274 0.427* 0.244 −0.463** −0.379* 0.950** 1.000   
F 0.107 0.671** −0.317 0.111 0.657** 0.432* −0.637** −0.636** 0.496** 0.438* 1.000  
Fe 0.014 0.211 0.120 0.601** 0.089 −0.058 −0.191 0.003 −0.162 −0.142 0.168 1.000 

*Correlation is significant at the 0.05 level (2-tailed).

**Correlation is significant at the 0.01 level (2-tailed).

The variables principally contributing to the second factor were TN and turbidity, representing suspended pollutants. Cisterns were close to the adjacent fields, thus soil with fertilizer could be transferred and accumulated on catchments and then moved into the cisterns with the runoff water, caused contamination and increased turbidity.

The third factor loading contributed to COD and total iron, representing reducing substances in water. COD reflected the contamination degree of reducing matter in water including organics, nitrite, ferrite and sulfide. There was also significant correlation between COD and Fe (r = 0.601, P < 0.01) (Table 5). The fourth factor was related to water acidity-alkalinity, with the most significant variables being pH and NO3.

To confirm the associations between the variables in the dataset, CA was conducted to classify the physicochemical parameters. The variables distributed in the four clusters generally correspond to the significant variables in the four principal factors, so dividing the parameters into four clusters is most appropriate. The dendrogram (Figure 3) shows four different clusters classified as A, B, C and D. Cluster A contained SO42−, Cl, EC , F, NO2 and NO3, coincident with the main parameters in the first factor loading in PCA/FA. The pH was classified as Cluster B, generally consistent with the dominant parameter in the fourth factor loading. Cluster C included COD and Fe, corresponding to significant variables in the third factor loading. Turbidity, TN, and NH4+ made up Cluster D, similar to the main parameters in the second factor loading. Thus, the results of (PCA/FA) can be considered reliable.
Figure 3

Dendrogram showing the four clusters formed: A – ion level; B – acidity-alkalinity; C – reducing substances; D – suspended pollutants.

Figure 3

Dendrogram showing the four clusters formed: A – ion level; B – acidity-alkalinity; C – reducing substances; D – suspended pollutants.

Effects of cistern materials and construction methods on water quality

Factor scores analysis was performed to rank the quality of the rainwater and harvested rainwater stored in different types of cisterns, according to the degree of contamination (Table 6). Higher scores correspond to heavier contamination. Thus, the rainwater had the best quality among all water samples. Harvested rainwater from the cement cistern (C1) had the best quality among all cisterns. The quality of harvested rainwater from C2 was the worst and harvested rainwater from C3 was better than other stabilized-soil cisterns. Although cement is a better material for cisterns to store rainwater than stabilized soil, the cost and environmental friendliness of stabilized soil make it a more attractive option. A cistern using stabilized soil material with the dry construction method (C3) can maintain a relatively good water quality at lower cost than cement.

Table 6

Factor scores ranked the quality of rainwater and harvested rainwater with different types of cisterns, according to the degree of contamination

Water sourceFAC1_1FAC2_1FAC3_1FAC4_1ScoreRanking
Rainwater −1.769 0.333 −0.550 −1.125 −0.93 
C1 0.374 −0.702 −0.296 −0.371 −0.15 
C2 0.647 0.660 −0.442 0.341 0.37 
C3 0.670 −0.206 −0.270 0.195 0.19 
C4 −0.428 0.010 1.400 0.639 0.25 
Water sourceFAC1_1FAC2_1FAC3_1FAC4_1ScoreRanking
Rainwater −1.769 0.333 −0.550 −1.125 −0.93 
C1 0.374 −0.702 −0.296 −0.371 −0.15 
C2 0.647 0.660 −0.442 0.341 0.37 
C3 0.670 −0.206 −0.270 0.195 0.19 
C4 −0.428 0.010 1.400 0.639 0.25 

The four factors described as PCA/FA were ranked based on their effects on water quality (Table 6). Rainwater had lower levels of ions (F1), reducing substances (F3) and pH (6.55–8.65) (F4). However, rainwater had a higher level of suspended pollutants (F2), indicating that atmospheric solid particles were the main source of pollution for rainwater. Harvested rainwater (C1, C2, C3, and C4) had higher levels of ions than rainwater due to catchment pollutants and substances released from cistern materials. C1 and C3 had a relatively lower level of suspended pollutants, indicating that less pollutants were being released from C1 and C3. The high level of suspended pollutants was the main contamination factor for harvested rainwater in C2, resulting in the worst water quality. Although C4 had lower ion concentration, the high level of reducing substances and increased pH resulted in poorer quality.

Water content is the main difference between stiff construction and plastic construction. Higher water content can increase the potential of shrinkage and swelling due to moisture changes or frost action (Katz et al. 2001). So the stabilized-soil cistern using stiff construction is more stable and durable, reducing water contamination from the cistern. Thus, harvested rainwater stored in C3 (stiff construction) had better quality than C4 (plastic construction). Shrinkage and swelling may cause cracks and even lead the surface of stabilized-soil cisterns to peel off after long-term performance. However, this effect was not obvious in the cistern using plastic construction since the water content was still limited.

CONCLUSIONS

The study assessed the quality of rainwater and harvested rainwater from four types of cisterns to determine their suitability for drinking. Additionally, four cistern construction methods were evaluated to determine the effects on water quality. The stored water generally did not meet drinking water standards, mainly due to high levels of bacterial contamination. Although coliform was not detected, the total bacterial count greatly exceeded the drinking water limit.

PCA/FA were applied to find the factors affecting water quality (physicochemical parameters). Four principal factors were identified, which collectively explained 80.37% of the total variance. Specifically, the first factor was linked to ionic concentration, accounting for 30.67% of the total variance. The second factor represented suspended pollutants, explaining 18.02% of the total variance. The third factor was related to reducing substances, accounting for 16.36% of the total variance. The fourth factor reflected acidity-alkalinity in water, explaining 15.32% of the total variance. CA also classified those parameters into four groups, confirming the results described by PCA/FA. Thus, harvested rainwater quality was mainly affected by the catchments, the surrounding fields and the local atmosphere.

Factor scores ranked water quality showing that direct rainwater was better than harvested rainwater. Cement was superior to stabilized soil for water cisterns. Stabilized soil material, with the stiff construction method (less water content) was the best construction method for the non-cement cisterns. Considering water quality, and the convenient access to adequate soil material, the stabilized soil cistern with the stiff construction method is a good option for the Loess Plateau in northwest China.

ACKNOWLEDGEMENTS

This research was supported by the Science and Technology Project of Shaanxi Province (2013KTDZ03-03-01), by the Science Foundation of China (41371276, 51309194), and by the Natural Science Foundation Project of Shaanxi Province (2016FJZDJC-S-16-6). The authors would like to thank Dr Henghui Fan and Mr Shengli Sun for the installation of the rainwater harvesting system, and also thank Dr Brian Boman, Florida, for his constructive advice in the writing of this paper.

REFERENCES

REFERENCES
Alves
F.
Köchling
T.
Luz
J.
Santos
S. M.
Gavazza
S.
2014
Water quality and microbial diversity in cisterns from semiarid areas in Brazil
.
Journal of Water and Health
12
(
3
),
513
525
.
Brahman
K. D.
Kazi
T. G.
Afridi
H. I.
Rafique
T.
Baig
J. A.
Arain
S. S.
2014
Evaluation of fresh and stored rainwater quality in fluoride and arsenic endemic area of Thar desert, Pakistan
.
Environmental Monitoring & Assessment
186
(
12
),
8611
8628
.
Cobbina
S. J.
Agoboh
Y. P.
Duwiejuah
A. B.
Bakobie
N.
2015
Evaluation of stored rainwater quality in basic schools in the tamale metropolis, Ghana
.
Water Quality Exposure & Health
7
(
4
),
1
8
.
de Man
H.
Bouwknegt
M.
van Heijnsbergen
E.
Leenen
E. J. T. M.
van Knapen
F.
de Roda Husman
A. M.
2014
Health risk assessment for splash parks that use rainwater as source water
.
Water Research
54
,
254
261
.
Fan
H.
Gao
J.
Wu
P.
Ju
W.
Sun
S.
2006
土壤固化剂集流面不同施工工艺比较 (Comparison of different construction techniques for the catchment area with soil stabilizer)
.
Transactions of the Chinese Society of Agricultural Engineering
22
(
10
),
73
77
.
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 Research
45
(
10
),
3245
3254
.
GB 5749-2006
生活饮用水卫生标准 (Standards for drinking water quality)
.
GB/T 5750.12-2006
生活饮用水标准检验方法–微生物指标 (Standard examination methods for drinking water – Microbiological parameters)
.
Gikas
G. D.
Tsihrintzis
V. A.
2012
Assessment of water quality of first-flush roof runoff and harvested rainwater
.
Journal of Hydrology
466–467
,
115
126
.
Islam
M. M.
Chou
N. F.
Kabir
M. R.
Liaw
C. H.
2010
Rainwater: a potential alternative source for scarce safe drinking and arsenic contaminated water in Bangladesh
.
Water Resources Management
24
(
14
),
3987
4008
.
Katz
L.
Rauch
A.
Liljestrand
H.
Harmon
J.
Shaw
K.
Albers
H.
2001
Mechanisms of soil stabilization with liquid ionic stabilizer
.
Transportation Research Record Journal of the Transportation Research Board
1757
(
1
),
50
57
.
Lee
J. Y.
Kim
H. J.
Han
M. Y.
2011
Quality assessment of rooftop runoff and harvested rainwater from a building catchment
.
Water Science and Technology
63
(
11
),
2725
2731
.
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
.
Resources, Conservation and Recycling
65
,
124
129
.
Niu
H.
He
Y.
Lu
X. X.
Shen
J.
Du
J.
Zhang
T.
Pu
T.
Xin
H.
Chang
L.
2014
Chemical composition of rainwater in the Yulong Snow Mountain region, Southwestern China
.
Atmospheric Research
144
,
195
206
.
O'Hogain
S.
McCarton
L.
McIntyre
N.
Pender
J.
Reid
A.
2012
Physicochemical and microbiological quality of harvested rainwater from an agricultural installation in Ireland
.
Water and Environment Journal
26
(
1
),
1
6
.
Sazakli
E.
Alexopoulos
A.
Leotsinidis
M.
2007
Rainwater harvesting, quality assessment and utilization in Kefalonia Island, Greece
.
Water Research
41
(
9
),
2039
2047
.
Tay
J. H.
Jaafar
S.
Mohd Tahir
N.
2014
Ionic composition of rainwater at selected sites of Kuantan, Pahang, Malaysia: a preliminary study
.
Bulletin of Environmental Contamination and Toxicology
92
(
3
),
329
333
.
Thomas
R. B.
Kirisits
M. J.
Lye
D. J.
Kinney
K. A.
2014
Rainwater harvesting in the United States: a survey of common system practices
.
Journal of Cleaner Production
75
(
75
),
166
173
.
Vialle
C.
Sablayrolles
C.
Lovera
M.
Jacob
S.
Huau
M. C.
Montrejaud-Vignoles
M.
2011
Monitoring of water quality from roof runoff: interpretation using multivariate analysis
.
Water Research
45
(
12
),
3765
3775
.
Wei
O.
Guo
B.
Cai
G.
Li
Q.
Sheng
H.
Liu
B.
2015
The washing effect of precipitation on particulate matter and the pollution dynamics of rainwater in downtown Beijing
.
Science of the Total Environment
505
,
306
314
.
Wilbers
G. J.
Sebesvari
Z.
Rechenburg
A.
Renaud
F. G.
2013
Effects of local and spatial conditions on the quality of harvested rainwater in the Mekong Delta, Vietnam
.
Environmental Pollution
182C
(
6
),
225
232
.
Xu
X.
Gao
J.
2014
Rainwater quality assessment of a solidified soil cistern using new construction technology
.
Journal of Southeast University
30
(
2
),
240
245
.
Zhao
W.
Gao
J.
Xu
X.
Wang
G.
2010
不同材质水窖贮存 6E8;水水质变化特征 (Characteristic changes of storing rainwater quality with different water cellars)
.
Journal of Soil and Water Conservation
24
(
1
),
20
23
.
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
.
Journal of Arid Environments
57
(
4
),
487
505
.