A PVC-pipe device as a sanitary barrier for improving rainwater quality for drinking purposes in the Brazilian semiarid region

Historically, water availability and the relationship between the quality of water and its uses have represented a major obstacle to human development. Water security is a central theme when seeking to protect the future of cities. The planning and development of cities require better efficiency in terms of the use of natural resources, including changing society and industrial habits, aiming to build sustainable productive chains based on the scarcity of resources. Rainwater is a natural resource that is unpolluted during the hydrologic cycle. The collection and storage of rainwater is an important solution when facing water scarcity. This practice is already common in arid and remote areas, where the implementation of conventional water supply systems is unprofitable or technically unviable (Sazakli et al. 2007). Water storage in tanks, usually known as cisterns, is common in areas that suffer long periods of drought, such as arid zones in northern China (Zhu et al. 2004), southern Syria (Braemer et al. 2015), and Palestine (Al-Salaymeh et al. 2011).

Between 2011 and 2017, the Brazilian social program known as ‘Água para todos’ (Water for All, WA) assisted approximately 4.5 million people through the installation of 911,000 cisterns in the Brazilian semiarid region. The Brazilian semiarid region is characterized by long dry periods (eight to nine months per year) and a rainy period concentrated in three to four months of the year (Alves et al. 2014). The aim of the WA program is to complete 1.5 million cisterns (MI 2017), delivered to the population in the semiarid region where the conventional water supply is not viable. The program builds one 16,000-liter cistern (made of concrete or polyethylene) for every house. This volume should supply a family of five with drinking water, as well as water for cooking and brushing teeth, during the entire dry season (Fonseca et al. 2014). Rainwater is used for human consumption because it is free of pathogens and many other pollutants that are usually found in surface and groundwater. However, rainwater can be contaminated by a number of processes, including microbial contamination from animal feces, chemical contamination from the atmosphere and the dissolution of compounds from roof materials, and physical contamination from the deposition of solids (McBean et al. 2013). Additionally, water handling, the method of water retrieval from cisterns, and water storage methods may also affect the quality of water used for human consumption.

Zhu et al. (2004) recommended boiling water before drinking, while water chlorination was preferred by Dillaha & Zolan (1985). These recommendations are related to the public health risk of bacterial diarrhea (from Salmonella and Campylobacter), bacterial pneumonia (from Legionella), or botulism (from Clostridium) (Lye 2002). Additionally, poor management of water in cisterns can lead to a favorable environment for the multiplication of Aedes aegypti, which is the vector for several viruses, including dengue, zika and chikungunya (Petersen et al. 2016). Physical barriers, such as plastic screens (placed at the end of gutters), first flush devices, filters, and pumps can provide initial sanitary protection and reduce the chances of rainwater being contaminated during the collection process.

In general, there is a strong consensus that diverting the first flush of rainwater is a protective factor against the consumption of contaminated water. However, most of the first flush devices used are made of concrete, for which leakages are a common problem, often leading to the abandonment of their use (Souza et al. 2011). It is also common to manually divert the first flush (Gikas & Tsihrintzis 2012). It has even been recommended to disconnect the pipes that divert water containing suspended solids from gutters (Alves et al. 2014). The manual diversion of the first flush is extremely dependent on the attitude and availability of the user. Plastic-made devices are widely commercially available. However, scientific studies reporting the behavior and proper amount of rainwater to be diverted in the field are limited.

To improve the quality of rainwater stored in cisterns, the aim of the present study was to assess the performance of a device designed to store the first millimeter of rainfall collected in cisterns. The water is diverted automatically. The device is made of PVC for reducing leakages, and it was tested from demonstration scale to practical application in the field (rural site). Additionally, we tested a hand pump as a complementary protective device.

The present study was conducted in the city of Caruaru in the state of Pernambuco in the northeast of Brazil (Figure 1). The Caruaru population projected for 2016 (IBGE 2010) was 351,686, distributed over an area of approximately 921,000 km². Pernambuco has 185 municipalities, with 122 of them located in the Brazilian semiarid region (Figure 1). This classification was made by the regional secretary for policies development of the Ministry of National Integration (MI 2005). The criteria used to award this classification include annual rainfall of less than 800 mm and a 60% or higher risk of drought.

The present research involved three phases: (1) identifying the amount of rainwater to be diverted; (2) defining the automatic diversion device (with no human interventions required); and (3) long-term monitoring of protective devices.

In this phase, the experiments were conducted in a rainwater catchment experimental station (RCES) (Figure 2), which was built on the university campus in the city of Caruaru. The RCES is a ground-floor building with a roof of ceramic tiles. The roof of the RCES comprises two independent surfaces with areas of 50 m² and 59 m². It is possible to pump water from a polyethylene reservoir (1,000 L) to create artificial rainfall using nozzles, which are uniformly distributed on both sets of tiles. The nozzle system comprised 28 fixed nozzles, with a 0.5 m³/h flow rate (Fabrimar^® Sempre Verde, Brazil), radial range of 4 m and maximum pressure of 20 mca for each nozzle. The collection of rainwater (artificial or not) was performed using collecting gutters and PVC pipes. Two concrete cisterns (16,000 liters each) were built to store the rainwater. The pumping station is composed of a one-stage centrifugal pump, with the following characteristics: 0.8–5.9 m³/h flow rate, manometric height of 9–30 m, and power of 1 HP (Schneider Motobombas^® BC-92, Brazil). Two hydrometers were used to measure the amount of water pumped in the precipitation simulations, with accuracy rates of 0.015 m³/h (Sappel Aquarius, Brazil). In addition, two analog gauges, with a maximal pressure of 20 mca, were used (Mecaltec, Brazil).

The RCES was installed in a rural area of the campus, close to a non-paved area used as a parking lot. Preliminary tests with AP (artificial precipitation) indicated the absence of total coliforms and Escherichia coli (E. coli) on the roof tiles of the RCES. Therefore, prior to beginning the experiment, artificial bacteriological contamination was performed by spreading 400 mL of sewage sludge evenly over the tiles (40.2 g SSV/L). The sludge was collected in an anaerobic reactor from a local sewage treatment plant, of which the microbial diversity is known (Lucena et al. 2011).

Two AP experiments were carried out in the RCES in triplicate (total of six assays). The aim of these experiments was to find the first flush volume that should be diverted to improve the quality of rainwater stored in the cisterns. The water used in the AP experiments was provided by the local sanitation company. The two precipitation intensities used were the average for the study area (23 mm/h – performed in triplicate) and another that was twice as high (46 mm/h – also performed in triplicate).

Samples of the first, second, third, and fourth millimeters of rainfall were collected. For sampling each millimeter of AP, four graded plastic buckets (named A, B, C, and D), with a maximum capacity of 59 liters each, were used to collect ten sequential samples (Table 1). Immediately after each bucket was filled, the water was mixed to standardize the water and collect samples for the analysis.

Physical-chemical analyses were conducted in accordance with Standard Methods for the Examination of Water and Wastewater (APHA 2005). The following parameters were determined: true and apparent color (2120C), turbidity (2123B), pH (4500H + B), alkalinity (2320B), total suspended solids (2540D), total dissolved solids (2540C), total coliforms (9223B), E. coli (9223B), total hardness (2340C), conductivity (2510B), and chlorides (4500 Cl-B).

Based on the results from the first phase, we designed a device to automatically divert and store the first flush of rainwater. The device design is a variation of commercially available devices (Rain harvesting^®, Waterplex_®, etc.). It is made of PVC pipes and connections, arranged to store 1 mm of rainwater. The device was named the DesviUFPE (Figure 3) and is characterized by its tightness, easy installation (like a puzzle), simple design, and low cost (approximately US$83.00 in 2017). The DesviUFPE differs from the commercially available ones, as it can be fitted to store the desired volume of diverted water according to the roof size. Additionally, the PVC pipes can be shaped to fit into the height of the house (usually the houses in rural areas are lower than houses in urban sites). It is a non-patented device.

The working principle is based on the communicating vessels. When the rain starts, the first flush is automatically stored in the device, and cleaner rainwater is forwarded to the cistern (Figure 3).

The water stored inside the device can be discarded or used for non-human consumption purposes. The DesviUFPE was installed in a house in the semiarid rural area of Caruaru-PE. Its efficiency was assessed through the analysis of water samples collected from five points along the rainwater course (Figure 4): rainwater (point A), water that had passed over the tiles and gutters (point B), water taken from inside of the DesviUFPE (point C), water sent to the cistern (point D), and water stored inside of the cistern (point E). These samples were collected during two rainfall events that occurred during the region's rainy season. The first collection occurred during the first rainfall event after the installation of the DesviUFPE (rainfall event I). The second collection occurred one year after the first (rainfall event II).

After the water samples had been collected in sterilized plastic receptacles, the same physicochemical parameters from Phase 1 were analyzed (with the exception of total suspended solids) using the methodology described in Standard Methods for the Examination of Water and Wastewater (APHA 2005).

In this phase, we expanded the study of the DesviUFPE behavior in the field by adding the assessment of a hand pump, which is commonly used by the rural community. We collected water from eight cisterns (Table 2) installed in the rural area of Caruaru (Figure 1), 40 km distant from the urban center. The detail in Figure 3 illustrates a typical house found in the rural area of Caruaru. These typical houses have a ceramic tile roof and a wide variety of ceiling heights (in most cases). The DesviUFPE was installed successfully in every house, due to its advantage of being adjustable in many different layouts and roof areas.

Cisterns C2 and C3 (Table 2) were monitored for 24 months. Cisterns C1, C5, C6, C7, and C8 were monitored for 12 months. Four of the analyzed cisterns did not receive any protective device throughout the entire monitoring period (C3, C6, C7, and C8). Cistern C5 only had a hand pump. Cisterns C1 and C2 had two protection devices (DesviUFPE and the hand pump). Cistern C4 was monitored for a total period of 24 months, with the two protection devices (DesviUFPE and the manual pump) installed in the 13th month of monitoring (Table 2). The samples were directly collected from the cistern monthly using the hand pump (when available) or by introducing a bucket into the cistern.

The hand pump utilized can extract 1 liter of water at each pump round. This device was made using PVC pipes and connections (50 mm and 40 mm), rubber rings, and two small marbles (Figure 5).

All obtained data (same parameters as Phase 2) were analyzed using descriptive statistics tools and a level of significance of 5%. The comparison of mean values (Student's t-test) and analysis of variance (ANOVA) were also applied. Tukey's test was used to make comparisons based on the minimum significant difference.

The first millimeter was responsible for the most substantial removal of color (87.52%) and turbidity (88.40%, Figure 6) from rainwater that could reach the cistern. The diversion of the second, third, and fourth millimeters resulted in low additional contributions of 4.55%, 6.73%, and 7.86%, respectively, for color, with corresponding values of 6.39%, 7.39%, and 9.32% for turbidity. Thus, no significant differences were found between the second, third, and fourth diverted millimeters (ANOVA: p-value = 0.160078 for α = 5%).

The only parameter affected by the change in the precipitation intensity (from 23 to 46 mm/h) was turbidity (Figure 6). This influence was only relevant for the first millimeter (Student's t-test: p-value = 0.0446 for α = 5%), with 5% higher removal at the intensity of 46 mm/h when compared with the result found for the 23 mm/h intensity. The runoff velocity was higher for the higher intensity, which only affected suspended solids, measured as turbidity.

When we consider the removal of total coliforms, once more, the first millimeter was responsible for the most significant removal (98.5%). The concentration of total coliforms in the rainwater that passed over the roof and gutters was reduced from 536,433 (BD, Figure 7) to 7,876 MPN/100 mL (A1, Figure 7) after diverting the first millimeter. However, after the fourth millimeter, total coliforms dropped to 510 MPN/100 mL (99.7% of removal efficiency) (A4, Figure 7). At this point, it is necessary to consider the ratio between the additional volumes of water that were discharged after the first (59 liters) and the fourth (236 liters) millimeters and the benefit generated by these diversions. When we consider regions with annual precipitation of less than 800 mm, the importance of preserving water resources is more extreme. During the years 2012 and 2013, the mean monthly precipitation was 29 mm and 50 mm, respectively, in the region monitored herein (INPE 2015). If we consider, for example, a catchment surface of 100 m² (mean catchment surface area of the cisterns monitored), the maximum volume that one family could store monthly was 2,900 liters and 5,000 liters. Thus, the diversion of the first 4 millimeters (corresponding to 400 liters in this case) of rainfall could be reckless, in terms of water resource management, due to the large amount of water being discharged. The scenario is critical when more than one discharge is required in a month, due to dry deposition between rainy events.

The best performance of diverting the first flush millimeter was found for E. coli, with a 100% removal efficiency. The association of microbes with settleable organic and inorganic solid particles may be the key for explaining the removal efficiency of pathogens by sedimentation, given that the attached microbes settle down faster than microorganisms in a free phase. Bacterial indicator organisms exhibit a consistent sedimentation correlation with settleable solids in urban runoff water samples. According to Characklis et al. (2005), centrifuging fecal coliforms, E. coli, and Enterococci with suspended solids increased the removal of these microbes by 30 to 50% in comparison to raw samples without centrifugation. This helps explain the good behavior of the device in removing E. coli (100%) and total coliforms (98.5%) in association with total solids (86.3%).

Analysis of the physicochemical parameters confirmed that turbidity, true color and apparent color were removed in the greatest quantities from rainwater that passed over the roof and gutters (difference between the results found for points B and D, Figure 4); average removal efficiencies of 71%, 66%, and 69%, respectively, were found for these parameters (Table 3).

The best results, as in the first phase of this study, were found for the microbiological parameters. In the field, the DesviUFPE retained 96.2% and 96.5% of the total coliforms (difference between points B and D, Figure 8), respectively, for rainfall events I and II. This result is particularly important given that the population drinks the rainwater stored in cisterns.

Rainfall event I occurred immediately after the installation of the DesviUFPE in the field. It is interesting to note that the total coliform value found in the cistern (point E, Figure 8) in rainfall event I was 14,592 MPN/100 mL. A year after the DesviUFPE installation, this value was reduced to 2,490 MPN/100 mL (point E, rainfall event II, Figure 8). Thus, 82.9% of total coliform removal was found in the field after one year of diverting the first millimeter of the rainwater stored in the cistern. This is the consequence of diluting the water prior to device installation with better quality rainwater, which had passed through successive diversions of the first millimeter.

Cistern C1 had both the DesviUFPE and a hand pump installed, while a hand pump was the only device in cistern C5 (Table 2). The other cisterns monitored (C6, C7, and C8) were not supplied with any protective devices. We used the parameter of turbidity to exemplify the significant differences found between the cisterns monitored (ANOVA: p-value = 6.14 × 10⁻²⁵, para α = 5%). The results found for cistern C1 were the most significantly different from the others (Table 4). The mean value for turbidity in C1, with both protection devices (pump + device), was 34.4% lower than the mean value found for C5 (only pump) and 70.7% lower than C8, which had no protection devices and exhibited the worst water quality.

The best results found in C1 are mainly related to E. coli (0.75, Figure 9), considering that the water is used for drinking purposes. The E. coli value in C1 was 63.6% lower than C5 (only pump) and 78.5% lower than C8 (no protective devices installed). This reflected the great influence of both devices, with better results found for the combination of a hand pump and the first flush device (cistern C1). Cistern C1 also showed the lowest levels for other parameters such as apparent color (mean of 3.58 uH) and total coliforms (mean of 557 CFU/100 mL).

Lee et al. (2010) advised that when monitoring specific storage tanks for rainwater, it is important to guarantee that the monitoring interval is long enough to assess the variations caused by the changing of the seasons throughout the year. In our case, the first 12 months of this phase were affected by a severe drought, with accumulated precipitation of 450 mm. Consequently, C2 (with a hand pump and the DesviUFPE) and C3 (without protective devices) were then monitored for an additional 12 months (997 mm of accumulated precipitation in the period of 24 months). Cistern C2 performed better than C3, with apparent color, turbidity, and dissolved solid values 50%, 67%, and 45% lower than in C3, respectively. Concerning the microbiological parameters, a significant difference was also observed (Student's t test = 3.85.10⁻⁵, p < 0.05) for total coliforms, for which content in C2 was 82% lower than in C3 (average value of 4,488 MPN/100 mL in C2, and 24,897 MPN/100 mL in C3), showing the sanitary barrier effect of the devices applied.

The last part of the experiment sought to determine the performance of the protective devices, comparing their influence in the same cistern. The water from cistern C4 was monitored for 12 months before the installation of the hand pump and the DesviUFPE (period named P1, Table 5) and 12 months after installation (period named P2, Table 5). The physicochemical parameters of apparent color, true color, turbidity, and dissolved solids were affected by the barriers installed, with removals of 43%, 35%, 70%, and 48%, respectively (Table 5).

For microbiological parameters, the corresponding removal was 18% and 60% for total coliforms and E. coli, respectively. During the rainfall harvesting process, the potential for microbiological contamination through sediments on roofs and gutters, as well as the direct depositions from birds, insects, and small mammals in the harvested water, has been widely discussed (Dillaha & Zolan 1985; Evans et al. 2006; Helmreich & Horn 2009). However, very few studies have discussed the management strategies that the cistern user should apply to avoid direct contamination of the reservoir. Factors such as the presence of animals in the vicinity of the reservoir, the use of dirty recipients to remove water, and the exposure of stored water (lack of a protective cover) can contribute to an increase in the microbiological pollutants found in stored water (Hoque et al. 2006). The low removal efficiency of total coliforms found in cistern C4 (18%) after one year of diverting the first millimeter may have been affected by this type of direct contamination in the reservoir. This efficiency tends to increase as the water storage procedures are improved, together with improved management of the reservoir on behalf of the users. The better behavior for removing E. coli (60%) in comparison with total coliforms (18%) reinforces the device relevance when harvested rainwater is used for drinking purposes.

The results of Phase 1, conducted with artificial rainwater in an experimental installation on the university campus, indicated that the greater load of pollutants is removed when the first millimeter of the harvested water is diverted. Discarding the first millimeter of rainwater reduced total coliforms and E. coli by 98% and 100%, respectively. In Phase 2, during which a PVC-pipe device named the DesviUFPE was used to accumulate 1 mm of precipitation, the results indicated a 96% removal efficiency of total coliforms for two monitored rainfall events in consecutive years. When the use of the device was extended to three houses in the rural site and a monthly monitoring program was conducted for 24 months (Phase 3), the previous behavior of working as a sanitary barrier was confirmed for the DesviUFPE. The cisterns equipped with the device and a manual pump exhibited lower concentrations of apparent color, turbidity, total dissolved solids, total coliforms, and E. coli (43%, 70%, 48%, 84%, and 72%, respectively) when compared with a cistern without any sanitary protection. The reduction found in the parameters investigated here ensures a lower health risk for drinking harvested rainwater when the 1 mm diverting device was used for the sanitary protection of the cistern.

The authors would like to thank the Brazilian funding agencies CNPq (process 556202/2009-0, granted to Savia Gavazza, and process 577035/2008-8, granted to Sylvana Melo dos Santos) and FACEPE (APQ 0966-3.07/15, granted to Savia Gavazza) for supporting this research project.