Assessment of rainwater harvesting and maintenance practice for better drinking water quality in rural areas


 In many areas of the world, rainwater has been collected and consumed by people. Our research aims to assess the use of rainwater and the impact of operation and maintenance activities of the rainwater system on drinking water quality in rural areas where there is no access to a public drinking water system. Through the questionnaire, direct visits, interviews and sampling of water at surveyed households (HHs), it was found that 100 and 98% of surveyed HHs used rainwater for drinking and cooking, respectively. Nearly, 80% of them were aware of the necessity of frequent reservoir cleaning as well as first-flush removal. Cleaning the water reservoir had a significant impact on water quality, in particular the total dissolved solids (p-value < 0.05). The use of strainers and more frequency of cleaning the catchment roofs and gutters would make the lower turbidity in water. However, the use of strainers would reduce the dissolution of oxygen in the reservoirs. A recommendation on the frequency of maintaining the rainwater harvesting systems was proposed after assessment of the operation and maintenance behaviors at surveyed HHs.


GRAPHICAL ABSTRACT INTRODUCTION
With the increasing shortage of underground water and surface water due to human activities and climate change, rainwater harvesting (RWH) has drawn much attention as an additional water supply option for domestic and drinking purposes (Bocanegra-Martínez et al. 2014;Hanson & Vogel 2014). In developing countries, rainwater is considered as an important source for water for drinking and cooking in rural areas (Islam et al. ; Özdemir et al. ).
In Vietnam, rainwater has been widely used as a naturebased and low-cost water supply in rural areas which also gives better taste compared with underground water. The Vietnam Government has paid great attention to promoting the use of rainwater in some legal frameworks such as the In general, the rainwater has been proved to be ineligible for direct use as drinking water due to contamination of microorganisms, heavy metals and organic matter (Lee et al. , 2017;Tran et al. ). The insufficient hygiene of water quality could be due to ill maintenance of the collection system including catchment roofs, guts, standing pipes, first-flush tanks or storage tanks. It seems to be that the collection, treatment, storage and maintenance of the RWH systems at households' scale is mainly based on long-standing habits and lifestyles, in particular in rural areas. Previous studies rendered that different kinds of catchment roofs would affect the harvested rainwater to some certain extent (Lee et al. 2012;Tran et al.  Therefore, the key objective of this study was to conduct the investigation on the quality of harvested rainwater and the impact of operation and maintenance (O&M) of the RWH systems on rainwater quality. The results will support the promotion of using RWH at a large scale effectively and sustainably.

Description of studied location
The reason that Ha Nam, a northern province in Red River Delta of Vietnam, was chosen because the quality of both underground water and surface water sources has been degraded in the past 10 years. In this province, the proportion of the households (HHs) having hygienic water was only 87%, lower than that of the national average As mentioned above, the groundwater in the Red River Delta has been long claimed to be contaminated with high levels of iron and arsenic (Larsen et al. 2008). Hence, people in Ha Nam province, especially in rural areas, have had a habit of using rainwater for drinking purposes instead of groundwater for a long time. It should be noted that these areas have not yet had access to piped water.
In addition, the average precipitation in these areas was reported as high in the rainy season (i.e., from June to September) and varied unexpectedly with years ( Figure 1). Furthermore, the precipitation tends to increase recently in the rainy season probably due to the impact of climate change (General Statistic Office of Vietnam ). It should be noted that data from Ha Nam province was not available and had to be referred to data from the meteorological station in Nam Dinh, an adjacent province.
Overall, it can be seen that the issue of polluted underground water, which is the main source for water supply at present, and the increasing precipitation have led to the potential of applying RWH system at a larger scale in the rural area with proper system design and operation.

Site survey
For the investigation of O&M of the RWH system at a HH scale and understand the perception of people in utilization  The remaining did not answer all questions, so they were rejected for data analysis. The number of responded HHs would render the margin of error of 6% with a confidence level of 95%. Certainly, the lower margin of error (i.e., 5%) which corresponds to 350 HHs sample size would give better representative of the population for your survey.
The distribution of samples by the HH sizes, ages of HH head, levels of education, jobs and average incomes is listed in Table 1. It can be seen from Table 1 that the sampled HHs are evenly distributed by different types of indicators. More specifically, only 3 samples (1.3%) are one-member HHs and 17 (7.3%) are large families with more than six members. Around 24% of the samples have either two or three family members, the four-person HHs share the highest percent which is about one-third of the total (33%). This result of HH size reflects quite well the data surveyed conducted via Vietnam national census, in which the average HH size was 3.8 and majority (44%) had 4-5 members in a house (United Nations ). This is also the typical size for Asian families (with three to four members per HH), while smaller HH sizes (i.e., fewer than three members per HH) are concentrated in Europe and Northern America and larger HH sizes (i.e., five or more persons per HH) are observed across much of Africa and the Middle East (United Nations ).
Looking at the age of the head of HH, only 16% of the samples is in between 20 and 40 years old, around 20% of samples is either their 40's or 50's and 38.5% of samples in between 50's and 60's is and about 25% greater than 60 years old. The percentage of the age of head of the HH fits right in the range of 20-35% for Asian countries studied by the United Nation, lower than European countries (from 35 to 45%) and higher than African countries (15-25%) (United Nations ). In term of the level of education, the majority of surveyed people only completed secondary school (51.3%), some only completed primary (12%) and a few graduated high school (19.7%). That The low income has influenced the habit of using water to some extent which will be discussed in the next sections.

Water sampling
For water quality assessment, the samples were conducted at the same time with the survey to validate the impact of system maintenance (i.e., surveyed people's O&M) on water quality. Due to limited budget, only 23 HHs (out of 234 HHs) were randomly selected to collect samples.
Water was collected from the harvested rainwater reservoirs and the groundwater's reservoirs, and then transferred to 2-L sterilized bottles for microbiological analysis (Escherichia coli and coliform) and 500-mL polyethylene bottles for on-site measurement. The main on-site parameters of concern were pH, total dissolved solids (TDS), dissolved oxygen (DO), temperature (T, C) and electrical conductivity (EC).
These parameters were recorded using a Multiparameter Por-

Valuation of the dependent variables
The main objective of this study was to understand the behavior of surveyed HHs in O&M activities of RWH systems on the water quality. For this purpose, dependent variables were indicators of rainwater quality including pH, TDS, turbidity (Tur), EC and DO ( Table 2). The accepted values were referred from the National Technical regulation (QCVN 01-1:2018/BYT) on quality of drinking water.

Valuation of the independent variables and controls
The description of the independent variables and controls is presented in Table 3. As described earlier, the main purpose of this analysis is to examine the effectiveness of O&M activities on stored rainwater quality. For instance, even though the first flush is very important, not always people remember applying it. In addition, the frequency of cleaning the rainwater reservoirs or clean the catchment roofs and guttering channels are changing sometimes. Thus, these activities would probably affect the stored water quality.

Empirical model and data analysis
The multivariate regression model capturing the relationships between water quality indicators (WQIs) and the classified independent variables and controls is expressed by the given equation, where β i is the model's constants.
The statistical analysis was also conducted to compare water quality of stored rainwater and stored groundwater by using paired T-test. Both regression and T-test were performed using SPSS Statistics 20 (IBM, USA) software.  Özdemir et al. () found that the rate of using rainwater for drinking purpose in that area was 85%, a bit lower than in the Ha Nam province which is in the Northern part. The demand of rainwater for drinking purpose would depend on many factors, one of which is the availability of piped/tap water. High percentage of using rainwater for drinking was reported from this study because the piped water was not accessible in these areas.
The general statistics for listed rainwater maintenance and utilization in the surveyed HHs is shown in Figure 3.   World Health Organization (WHO ) recommended a periodic disinfection of the tank (e.g., after long periods without rain), but not specify the exact required cleaning frequency.
For the behavior of cleaning the catchment roof and downpipes, it was found in Figure 4(b) that not many HHs had the habit of cleaning the roofs or they even did not remember of doing it (50%). It seems that they rely on the first rain of the season to wash the roofs for them. This is actually a common practice everywhere in Vietnam, where

Evaluation of harvested rainwater quality
In this part, the water quality of the stored rainwater shall be evaluated itself, and also compared with that of the filteredand-stored groundwater as both kinds of water were used in 23 surveyed HHs. It is worth noting that nearly 80% of these HHs filtered the groundwater before storing, in which nearly 70% employed sand filters. On the contrary, only 70% of them filtered the rainwater simply with a strainer before storing because they thought it was readily clean.
Statistical analysis of two kinds of water in term of pH, TDS, turbidity, EC, DO, E. coli and coliforms is depicted in Figure 5. It is a graphical representation of key values (i.e., minimum, 25th percentile, median, 75th percentile and the maximum).
In comparison of the two waters, it can be seen from  Table 4, where these p-values were less than 0.05 and t-values were negative for TDS and EC. The coliforms were distributed evenly from 1,000 to 2,000 CFU/100 mL in groundwater, while in rainwater, the distribution is abnormal and shifts to high values (e.g., 4,000-6,000 CFU/100 mL). Surprisingly, there is no statistically difference of turbidity in stored rainwater and groundwater (p-value > 0.05). E. coli is not presented in Figure 5 as it was the same averagely 60 CFU/100 mL for both waters. The possible explanation of higher pH in stored rainwater can be due to the fact that rainwater is mostly stored in concrete tank for a long period of time.
The tanks are often plastered inside with Portland cement to enhance waterproofing. This kind of cement is rich in CaO, leading to the high possibility of leaching CaO into the water after a long storage time, and eventually higher pH in the water. Due to the exposure to the air, it is logical that DO in the stored rainwater can be higher and especially, the coliforms can be significantly elevated. As for the TDS and EC parameters, which both implies the dissolved and charged ions in the water, there are more dissolved ions in groundwater as the dissolution of metal ores under the ground occurs more significantly. Rainwater has only some ions which it catches on the way down to the roof and through harvesting system.
Compared with the rainwater quality in previous studies, it was found that pH in this study was higher than those reported previously (Gikas et al. 2012;Lee et al. 2012) with pH of 6-7.5. The difference was due to the tank material.
They used stainless steel or plastic tanks as oppose to concrete and brick tank plastered with cement inside applied in this study. The increase of pH within the storage tank was revealed in a review of Sánchez et al. () in which pH could be up to 9.5-10.2 in brick tanks, and to 8.7-9.8 in ferro-cement rainwater tanks. In Lee et al. (2012) study, they also found much smaller number of coliforms and E. coli in the rainwater tank (i.e., less than 20 CFU/100 mL for each) than those in our study. It seems possibly that microbial contamination in rural areas occurs more significantly than in urban (city) areas due to the nearby farming and husbandry. The findings of turbidity, TDS and EC in this study were, however, in line with those in previous studies (Gikas et al. 2012;Igbinosa & Aighewi ).
Overall, the stored rainwater quality all meets the national standard of Vietnam for drinking water (QCVN 01-1/2018/BYT) with pH in the range of 6.5-8.0, TDS < 500 mg/L, turbidity < 2 FTU, except for the coliforms (should be 0 CFU/100 mL). That is the reason why people boil (98%) rainwater or filter it with reverse osmosis prior to drinking (see Figure 3). According to research by Amin finding is logical in a way that the use of strainer and more frequency of cleaning the catchment roofs and gutters would make the lower turbidity in water. Moreover, the use of strainers also hinders the dissolution of oxygen in the reservoirs (negative relation). In addition, TDS was found to be negatively and significantly impacted by the users' education level as a control (5% significant level). This indicates that people who got higher knowledge seem to neglect more often the labor work. More study on the impact of the controls on the water quality should be conducted in the future.
One more interesting was found that the intercept coefficients of pH, EC and DO were quite significant at 0.01, 5 and 5%, respectively (see Table 5). That meant the bases of pH, EC and DO parameters were 8.5, 114.1 and 4.2, respectively.
Another speaking, those were intrinsic values of pH, EC and DO regardless of any changes in the conditions of O&M for the harvested rainwater in this survey area.

CONCLUSIONS
A complex survey and water sampling in the rural province revealed that the HHs used groundwater for domestic  purposes while employed rainwater for drinking and cooking.
The majority (95%) of HHs did not add disinfectant in the reservoirs but boil water or filter with reverse osmosis unit before drinking, so the rate of having water-borne illness is low (about 8%). The surveyed HHs had good behavior of frequent cleaning the reservoirs (97%) but did not pay much attention to cleaning the catchment roofs and guttering pipes. Statistical analysis showed that only 'cleaning the reservoirs' activity had a significant relationship with dissolved solids in the reservoirs. The study also revealed that use of strainer and more frequency of cleaning the catchment roofs and gutters would make the lower turbidity in water. However, the use of strainers would reduce the dissolution of oxygen in the reservoirs.
The stored rainwater has better quality in terms of TDS and DO than those of groundwater, but both rainwater and groundwater are contaminated with microorganisms. They need to be boiled or disinfected before drinking.