Solar disinfection at low costs: an experimental approach towards up-scaled continuous flow systems

Despite the efforts of the UN Millennium Development Goal (MDG) for drinking water and of the UN Sustainable Development Goal 6 (SDG 6) on clean water and sanitation by 2030, results are lagging behind expectations (Sadoff et al. 2020). Almost 10% of the global population (∼700 Million people) still lack access to an improved source of drinking water (WHO/UNICEF JMP 2020), especially in poor rural areas that lack the financial resources to build and manage any sophisticated water treatment facilities to avoid basic health risks and deaths from illnesses such as diarrhoea caused by drinking contaminated water (WHO 2015; Bivins et al. 2017). Hence, cheap and simple water treatment methods, preferably based on on-site available material, are key to quickly improving the drinking water situation in these regions of the world. Such techniques can also be very useful in humanitarian crises and post-natural disaster situations where emergency water treatment is needed.

Solar radiation is an excellent free source for developing cheap water treatment techniques. UV-C and UV-B radiation can directly damage cell DNA of pathogens (Ravanat et al. 2001). However, the atmosphere blocks UV-C and most of the UV-B radiation (Matsumi & Kawasaki 2003; Haigh 2007). Instead, UV-A radiation reaches the surface of the earth and can inactivate cell membranes and DNA of waterborne pathogenic bacteria, viruses and parasites (Wegelin et al. 1994; Miller et al. 1999; Ravanat et al. 2001; Clasen et al. 2007). The inactivation typically takes less than 6 h and the effect continues even if the exposure to sunlight is halted (Ubomba-Jaswa et al. 2009). The SODIS technique takes advantage of solar radiation by placing contaminated water in a UV-A transparent container, most often a polyethylene terephthalate (PET) bottle (1.5–2.5 litres), and exposing it to sunlight for at least 6 h. However, the exposure time varies depending on the intensity of the sunlight and the sensitivity of the pathogens (comprehensive review in McGuigan et al. (2012)). The SODIS technique has been known for more than 30 years is recommended by the World Health Organization and is used by more than 4.5 million people worldwide. Furthermore, it has been found that a significant reduction in the occurrence of diarrhoea can be achieved with SODIS, especially among children (e.g. Conroy et al. 1996; Manzolillo 2019).

SODIS depends on several factors, including temperature, type of contaminant and the characteristics of the reactor material, size and configuration. The effectiveness of the SODIS technique can be improved using thermal enhancement and reflecting materials (Ayuob & Malaeb 2019). For the former option, synergistic effects occur in the inactivation process already at 45 °C, with an increasing effect for higher temperatures (McGuigan et al. 1998; Vivar et al. 2017). A recent study has also demonstrated an increase of inactivation rate at low water temperatures around 6 °C as compared to 22 °C (Villar-Navarro et al. 2021). Solar reflectors are useful for enhancing the exposure to solar UV-A radiation and can reduce the disinfection time with as much as 30% (Martin-Dominguez et al. 2005). Other options for enhancing the effectiveness of the SODIS technique include photocatalysis using semiconductors (Malato et al. 2009) and chemical additives such as lime juice/pulp (Harding & Schwab 2012) or hydrogen peroxide (Villar-Navarro et al. 2021).

SODIS is usually implemented in PET-bottles of around 2 litres, and although novel approaches have included polypropylene bags of up to 4 litres (Gutiérrez-Alfaro et al. 2017), or bigger containers of up to 19 litres, challenges remain open to the treatment of large volumes of water with SODIS (Borde et al. 2016). Constructing continuous flow systems can expand the SODIS technique. However, only limited studies have reported the design of such systems as well as the challenges for their implementation, although satisfactory results have been presented (Gill & Price 2010; Busse 2016). Gill & Price (2010), for instance, describe the installation of a 120 m long glass pipe, supported by compound parabolic collectors (CPCs), in a rural Kenyan village where water with a fairly low contamination level of 10² colony forming units (CFUs) per ml could be reduced to 0 CFU/ml after a 20 min single pass residence time. However, the system was constructed and shipped from Portugal, thus increasing the production costs of the system, and the project faced several challenges during its execution (Mac Mahon & Gill 2018).

In deprived communities, the access to sophisticated materials is very limited. Our hypothesis is that by using transparent plastic tubes, easily available in local stores, continuous flow SODIS systems can be constructed at low costs. Another advantage of using plastic tubes is that they are robust and resilient to damage. However, the long-term photostability and durability varies for different polymeric materials (García-Gil et al. 2020). Several further aspects have to be considered when designing full-scale continuous flow SODIS systems. However, in this preliminary study, the focus has been on bacterial removal in static batches and the UV-A transparency of the plastic tubes. Hence, we evaluate if easily available and low-cost plastic tubes can be used for the reduction of E. coli in contaminated drinking water, in order to be used in the construction of future continuous flow SODIS systems. A set-up of experiments were conducted at the Universidad de Antioquia Campus in Medellín, Colombia, a country where more than 4 million people lack access to safe drinking water (WHO/UNICEF JMP 2019, 2020).

Plastic tubes purchased on the local market in Medellín were investigated to select a suitable tube depending on the UV-A transparency of the plastic materials. We then used exposure time to solar radiation as a control variable to determine the residence time in the chosen tube, i.e. the time the contaminated water must be in the tube while being exposed to solar radiation to reach sufficient inactivation levels. More specifically, the residence time was determined using static batches of contaminated water in 20 cm long samples of the tube. Furthermore, we used aluminium foil and cardboard, which are cheap and easily accessible materials, in the experimental set-up to investigate a possible reduction in the residence time. With the estimated residence time of the tube, the length of tube needed to construct the continuous flow system can be determined. Apart from the residence time, the length of the tube will depend on the water flow rate and diameter of the tube.

Three different optically transparent plastic tubes were found and purchased on the local market in Medellín, Colombia. The three different tubes – labelled H1, H2 and H3 — and their specifications are presented in Table 1. For comparison, we also tested the UV-A transparency of PET-bottles, commonly used in the SODIS technique. Thus, two commercial PET-bottles of 1.5 litres (P1) and 2.5 litres (P2) were included in the experiment (their specifications are also included in Table 1).

We estimated the UV-A transparencies of the tubes and PET-bottles using a UV513AB digital UVA/B light meter from General Tools (General-Tools 2022) with a spectral sensitivity range of 280–400 nm and a calibration point in the UV-A band at 360 nm. The instrument does not allow a simultaneous measurement of UV-A and UV-B radiation. For comparison purposes, we instead also estimated the transparency of optical radiation, represented by photosynthetically active radiation (PAR, 400–700 nm), for all objects. The PAR radiation was measured using a LI-193 Spherical Quantum sensor from LI-COR Biosciences (LI-COR 2022), operated by a LI-1400 dataLOGGER from the same company. The experiment was conducted on 3 February 2016 at the University of Antioquia main campus in Medellín, Colombia (coordinates 6°16′8.5″N 75°34′6.6″W).

The sensor of the UVA/B light meter has a flat round surface with a diameter of approximately 10 mm. We measured UV-A intensities using cut pieces of each object with a size of 20 × 20 mm equivalent to cover the size of the sensor. The LI-193 Spherical Quantum sensor has the shape of a light bulb with a diameter of 61 mm. In order to measure the PAR transparency of the objects using the cut pieces, we covered the Quantum sensor with aluminium foil leaving an opening of 10 × 10 mm similar in size to the UV AB sensor.

The average of the 10 calculated ratios gives the transparency of the material and their standard deviation gives the 1 − σ error. The tube showing the highest UV-A transparency ratio was chosen for the rest of the experiments.

We define the residence time in the plastic tube as the time that contaminated water must be in the tube while being exposed to solar radiation to reach sufficient inactivation efficiency.

We prepared two experimental set-ups using 20 cm long samples (se Figure 1) cut from the tube H1 selected in the UV-transparency experiment: one set-up with five tubes placed on top of non-reflecting brown cardboard (Z set-up), and one set-up with five tubes placed on top of reflecting aluminium foil (A set-up), see Figure 2 for a schematic of the experimental set-up. Six replicates of each set-up were used and simultaneously placed on concrete in an outdoor open area, after preparation as described below. Since the selected tube had a diameter of 2.5 cm, the total volume of each 20 cm long tube sample was 100 ml.

A total of 66 tube samples were rinsed thoroughly by washing each one in 200 ml of sterilized distilled water. We tested this method prior to the experiment and concluded zero contamination in the plastic tubes. A total of 60 of the sterilized tubes were then filled with contaminated water (prepared in the laboratory), the ends of the tubes sealed with sterilized latex and finally left exposed to the sun for 6 h in the open space during the brightest part of the day (between 09.00 and 15.00 h). The experiment was conducted on 14 February 2016 at the University of Antioquia main campus in Medellín, Colombia (coordinates 6°16′8.5″N 75°34′6.6″W).

Four control samples (tubes filled with sterilized distilled water) were also included in the experiment to ensure that the sterilization of the tubes was successful and no bacterial growth occurred due to pre-contamination. We placed two of these control tubes in the sun (on cardboard) and two in the shade. The remaining two sterilized tubes were used for temperature measurement purposes (see below).

Ideally, control samples measuring the dark effect should be included, i.e. the effect on the contaminated water in the absence of light (UV and visual). However, the method used in this work for measuring the E. coli concentration saturates above 2,000 CFU (see the subsection Method for measuring E. coli concentrations below). The experiments start with E. coli concentrations of 10⁶ CFU/ml (see the subsection Preparation of the contaminated water below). With measurement volumes of 100 ml, the dark effect would have to have a reduction effect of the order of 10⁶ to be measurable. Since this is not feasible, control samples for the dark effect would be redundant and thus omitted in the experimental set-up.

Six tubes were removed from each set-up (12 in total) at five different time intervals (t = 0.5, 1, 2, 4 and 6 h, respectively), for measurements of the bacterial concentrations in terms of CFUs per 100 ml in the water samples (see description of these measurements below). We measured the contamination levels of the four control samples at the end of the experiment (after 6 h).

During the experiment, the PAR and UV-A intensities (W/m²) were monitored as the average intensity in 5-min intervals using the LI-193 Spherical Quantum sensor from LI-COR Biosciences (LI-COR 2022) and the UV513AB digital UVA/B light meter from General Tools (General-Tools 2022), respectively. Two of the sterilized tubes were filled with contaminated water and used during the experiment to monitor the water temperatures in each set-up. A thermometer was placed in each one of these tubes and the temperatures were read off every 15 min together with the air temperature in the shade during the whole experiment (6 h).

The contaminated water was prepared using sterilized distilled water and E. coli bacteria (ATCC® 25922™, ATCC 2021). E. coli bacteria were chosen since they are widely adopted as contamination indicators (e.g. Ubomba-Jaswa et al. 2009; Giannakis et al. 2014). Prior to preparing the contaminated water, the bacteria were cultivated for 24 h at a temperature of 35.5±0.5 °C in Plate Count^® agar (Merck-Millipore 2021a) using standardized Petri dishes. The E. coli bacteria were then inserted into a sample of 100 ml of sterilized distilled water. The concentration of bacteria in the water sample was estimated using a NANOCOLOR 400D photometer (Macherey-Nagel 2021). This instrument measures the absorbance of the solution, i.e. the amount of transmitted light due to the presence of microorganisms. The measured absorbance can then be translated to McFarland turbidity standards (McFarland 1907). This unit is an estimate of the concentration of microorganisms, where 0.5 McFarland corresponds to 1.5 × 10⁸ CFU/ml and an absorbance in the range 0.08–0.1 Dalynn-Biologicals (2014). E. coli bacteria were inserted into the water sample until the measured absorbance was in this range, for a bacterial concentration of approximately 1.5 × 10⁸ CFU/ml.

For the experiments, an initial E. coli concentration of the contaminated water of 10⁶ CFU/ml was chosen. This concentration corresponds to highly contaminated water and is commonly used in similar studies (e.g. Ubomba-Jaswa et al. 2009; Giannakis et al. 2014). In this way, starting at an E. coli concentration of 1.5 × 10⁸ CFU/ml the contaminated water had to be diluted 100 times to reach the initial concentration used in the experiments.

The number of CFUs of E. coli was estimated for the tubes containing contaminated water at the specified time intervals during the experiment. This was done by first filtering all of the water in each tube (100 ml) through a sterile gridded membrane filter of cellulose nitrate (cat no. 160047GXI, Axiva 2020) where the bacteria are caught. The membranes were then placed in Chromocult^® agar (Merck-Millipore 2021b), using standard Petri dishes, for bacterial cultivation over 24 h and at a temperature of 35.5 °C. Finally, the CFUs on the membranes were counted using a stereomicroscope. This gives an estimation of the number of viable E. coli cells that were present in the samples and that could replicate in order to give rise to colonies, under the assumption that each CFU originated from one cell. Saturation is reached when there are more than 2,000 CFUs present on the membranes. Hence, for CFU numbers greater than 2,000, individual units cannot be distinguished such that the number of CFUs cannot be counted (Standard Methods for the Examination of Water and Wastewater 2012, chp. 9222B). For CFU numbers less than 2,000, the uncertainty in counting is ±0.0986 in logarithmic space (ISO 2005).

The results for the experiments of measuring the transparency of the tubes and PET-bottles are shown in Figure 3,. All three tubes and the two PET-bottles show very similar transparencies to PAR radiation of about 85%. However, the transparency to UV-A radiation differs significantly. The two PET-bottles have a UV-A transparency of 60 and 63%. H1 shows a UV-A transparency of 48%, while the corresponding numbers for H2 and H3 are 36 and 2%, respectively. Thus, H1 was chosen for the experiment of measuring the residence time. Two-sampled Student's T-tests of unequal variances were performed on the UV-A transparency ratios for all combinations of objects and all tests show p-values <10⁻⁶. The null hypothesis that the sample means are the same can thus be rejected for all combinations of objects. The low transparency of H3 is due to the presence of pigments in the plastic (see Table 1). The full experimental dataset, including the mean and standard deviations plotted in Figure 3, are presented in Supplementary Material, Appendix A.

Figure 4 shows the PAR (panel a) and UV-A (panel b) intensities measured during the experiment as a function of time (hours since the start of the experiment on the lower x-axis and time of day on the upper x-axis). The PAR and UV-A measurements show very similar behaviours, with a large scatter during the first 2 h due to varying cloud coverage, followed by 4 h of constant sunshine, a very small scatter and a constant decline until the end of the experiment. The peak in radiation occurred around 11.30, when the PAR and UV-A intensities reached values of about 850 and 95 W/m², respectively. During the experiment, the UV-A intensity never went below 29 W/m².

During the experiment – based on the selected tube H1 – six tube samples from each set-up were removed for bacterial cultivation at five specified time intervals (after 0.5, 1, 2, 4 and 6 h) and tagged A1 (0.5 h) to A5 (6 h) and Z1 to Z5 for the A and Z set-ups, respectively. The initial and estimated E. coli concentrations are presented in Tables 2 and 3 for the Z and A set-ups, respectively. Notice that the concentrations are per 100 ml, since the counts are for the full water sample of 100 ml in each tube. Hence, the initial concentration (at t=0 h) is 1.5 × 10⁸ CFU/100 ml. The mean concentrations for each time interval are given in column 8 and standard deviations in column 9 where applicable.

At the time intervals t=0.5 h and t=1.0 h, the number of CFUs was very high for all samples (Z1, A1, Z2 and A2) such that the membranes were saturated and lower limits are given (>2,000 CFU/100 ml). After 2 h (Z3 and A3), the number of CFUs had been reduced drastically and after 4 h (Z4 and A4), the number of E. coli CFUs was consistent with zero. Also for the control samples with sterile water (C1 and C2), the number of E. coli CFUs was consistent with zero. Hence, the sterilization of the tubes can be deemed successful.

The difference between the Z and A set-ups (i.e. the difference between using cardboard and aluminium as an underlay), can only be evaluated for the time interval t=2.0 h. At this time interval, a two-sampled Student's T-test of unequal variances was performed on the two datasets, which showed a p-value of 0.0519.

The mean concentrations presented in Tables 2 and 3 are shown in Figure 5 as a function of time. The data points represented by lower limits correspond to the label in the lower right corner. At 4 h (Z4 and A4) and 6 h (Z5 and A5), the number of CFUs had been reduced to values consistent with 0 CFU/100 ml. To include these data points, for illustrative purposes, they are presented as log C = 0. To make Figure 5 easier to read, the data points have been slightly separated in the x-direction. Error bars are the standard deviations presented in Tables 2 and 3, propagated to logarithmic space in the standard way, and added in quadrature with the count errors for measuring E. coli concentrations.

Figure 5 also includes the temperatures measured during the experiment for the Z set-up (blue dashed curve) and the A set-up (red dotted curve). The vertical axis on the right side of the chart shows the temperature scale. The different thermometers used during the experiment were calibrated and the temperatures presented in Figure 5 were adjusted accordingly. Using the spread in measurements among the three thermometers used, we estimated the temperature uncertainty to 0.61 °C.

The temperatures of the two experimental set-ups (Z and A) increase during the first 2 h of the experiment (see Figure 5). After about 2 h, the temperatures settled at around 50 and 45 °C for the Z and A set-ups, respectively. Hence, the water in the tubes placed on cardboard reached temperatures of about 5 °C higher than the water in the tubes placed on aluminium foil. The black dashed vertical line in Figure 5 indicates the time when the temperature in the tubes of the Z set-up reached 45 °C (1.75 h), i.e. the temperature above which synergistic effects between temperature and UV exposure have been found in previous studies. The ambient air temperature was only measured during the last 4 h of the experiment (11.50–15.30) due to issues with the thermometer and ranged between 28.6 and 31.6 °C with an average of 30.5 °C.

One of our aims was to study plastic tubes that are available on a local market. The first selection criteria at the time of purchase was optical transparency and all available tubes that met this criteria turned out to be made of PVC. The best option (the plastic tube labelled H1) showed a UV-transparency of about 50%. This number can be deemed successful, considering the result of a UV-A transparency of 60% for PET-bottles that are commonly used with the SODIS method. Equivalent transparency ratios for PET have been found by García-Gil et al. (2020) and Sackey et al. (2015). However, the plastic tube named H3, although also made of PVC, showed a transparency ratio very close to zero due to the inclusion of pigments in the plastic. Similarly, Xiupinget al. (2017) found a 90% shielding efficiency of UV light when adding pigment to PVC films. Considering that H1 and H3 were both made of PVC and are similar in appearance, a careful selection of the plastic material is fundamental when designing an up-scaled SODIS system based on plastic tubes.

One advantage of PVC as a material is the cheap production costs. Furthermore, PVC transmits UV-B radiation (Hu et al. 2020), which is not the case for PET (Sackey et al. 2015; García-Gil et al. 2020). However, it is important to stress that one of the main uncertainties is the chemical behaviour of the plastic in the tube and the possible release of constituents into the water when exposed to solar radiation over long periods. This has been the concern of several studies for the use of PET-bottles with SODIS (Wegelin et al. 2001; Schmid et al. 2008). García-Gil et al. (2020) developed a predictive model for material selection in SODIS systems by estimating the available radiation inside a device as a function of material and thickness. They ruled out PVC as a suitable material for SODIS systems due to poor photostability of the polymers, leading to a decline in mechanical properties and discolouring over time when exposed to natural weathering. Polypropen (PP), polycarbonate (PC), poly(methyl methacrylate) (PMMA) and PET were instead deemed suitable depending on either very low production costs or high photostability and durability. Although most likely shorter than for the other materials studied in García-Gil et al. (2020), the actual lifetime of a continuous flow SODIS system based on a PVC tube is not known. Empirical studies of such systems based on different plastic materials are, therefore, needed to understand which plastics are actually viable for implementation.

Hence, the cost and lifetime of the material together with the weather conditions at the target location must be considered when designing up-scaled continuous flow SODIS systems. The possible release of constituents from different materials must also be considered and further studied, following previous studies such as Wegelin et al. (2001) and Schmid et al. (2008), as well as the recycling perspective of the materials used in the drinking water treatment technique.

The main experiment of estimating the residence time showed that less than 4 h are sufficient to reduce the E. coli concentration from 10⁶ CFU/ml to 0 CFU/ml in the selected plastic tube (H1). Considering the low concentrations found already after 2 h (see Tables 2 and 3), it can be assumed that the residence time needed to reach 0 CFU/ml is, in fact, significantly less than 4 h and, hence, lies somewhere between 2 and 4 h if a high enough exposure to sunlight is provided. Thus, future experiments must consider a finer time sampling between 2 and 4 h.

The comparison between cardboard (Z set-up) and aluminium (A set-up), with the latter as an attempt to use a cheap reflective material to enhance the UV-A exposure, indicates a higher reduction rate with the cardboard rather than with aluminium foil. One possible explanation could be the higher temperatures reached in the tubes of the Z set-up, up to 50 °C compared to the 45 °C reached by the A set-up. Several studies including McGuigan et al. (1998) and Vivar et al. (2017) have found synergistic effects between temperature and UV exposure at temperatures above 45°. Indeed, the Z set-up reached 45 °C after 1 h 45 min (see Figure 5). Although the Student's T-test did not show any significant difference between the measurements of the two set-ups after 2 h, the results indicate that cardboard is more useful than aluminium for enhancing the temperature in the tubes.

The A set-up, on the other hand, never reached 45 °C during the first 4 h of the experiment, i.e. the time needed to fully reduce the E. coli concentrations to 0 CFU/ml (see Figure 5). As mentioned above, synergistic effects between UV exposure and temperature have mainly been found above 45 °C. Vivar et al. (2017) also demonstrated that the inactivation process is slowed down and that microbial growth takes place in the temperature range 40–45 °C under low or non-continuous UV radiation. Furthermore, Giannakis et al. (2014) showed that temperatures in the range 30–40 °C may impair inactivation, but that complete disinfection can still be reached at these temperatures under sufficiently high UV-A intensities. These findings indicate a risk of impairment of the solar disinfection process in the range around optimal microbial growth temperatures. Despite varying UV-A intensities during the first 2.5 h of the experiment (see Figure 4), our results indicate that the exposure to UV-A radiation by itself was sufficient to completely reduce the E. coli to 0 CFU/ml below 45 °C and within 4 h.

In comparison, Villar-Navarro et al. (2021) found UV-A radiation by itself to be insufficient to inactivate high concentrations of bacteria in short treatment times, i.e. less than 3 h, in simulated aquaculture streams. Interestingly, the same authors found an enhancement of the inactivation with SODIS treatment at low temperatures, i.e. at 6 °C as compared to 22 °C.

The results show that the residence time in the selected plastic tube (H1) lies in the range of 2–4 h. We adopted a conservative approach by assuming a residence time of 4 h for an attempted design of an up-scaled SODIS system. Our vision of such a system is similar to the system presented in Gill & Price (2010), with the exchange of a glass pipe to a plastic tube to reduce the manufacturing costs. Figure 6 shows a simple schematic view of such a system with a plastic tube on a sloping plane, such that gravity drives the flow of contaminated water through the tube of the system.

The proper set-up of a continuous flow SODIS system will be dependent on the required need of treated water. According to WHO, 50 litres of water per person and day (l/p/d) are required to meet the needs of consumption and basic hygiene (personal and food), while 100 l/p/d meet all needs (Howard & Bartram © WHO 2003). The flow rate can then be calculated assuming a system runtime of 6 h per day and a daily need of 50–200 litres of water for one household. The selected tube in this work (with a radius of 1.25 cm and the conservative residence time of 4 h) results in a system length of approximately 68–271 m.

Hence, the selected tube would result in long system lengths. For future studies, it would be desirable to reduce the system length significantly by increasing the tube diameter. Such studies should be designed to understand the relationship between the tube diameter and the residence time due to light absorbing factors, while also taking the transparency of the selected tube into consideration.

Furthermore, the design assumes that the system will stop working at night and then resume in the morning. Previous studies have shown that the total dose of UV radiation is the most important factor, and that a continuous, uninterrupted exposure gives the best results (Ubomba-Jaswa et al. 2009). Further tests are needed to evaluate the level of inactivation if the system is halted during the night.

Another uncertainty involves the presence of clouds, as they will reduce the UV-A intensity significantly. This can be seen during the first 2 h of our experiment, when the UV-A intensity dropped about 50% in the presence of clouds. The effect of clouds has to be taken into consideration when designing the system and, most importantly, during the education of how to use the SODIS technique. For this purpose, Moreno-SanSegundo et al. (2021) designed a solar calculator based on the position of the sun, atmospheric extinction, cloud-cover, and elevation to estimate the potential for using SODIS in different regions on the earth. Similar to the problem with clouds is the treatment of water with high turbidity. The efficiency of inactivating pathogens decreases with increased turbidity and very high solar radiation intensities are needed to reach complete inactivation at high turbidites (e.g. Kehoe et al. 2001).

The cost of the studied tube H1 is 1.21 USD/m (see Table 1). Considering only the material of the tube, the price of constructing the system with the above calculated lengths would exceed 80 USD due to the rather long system lengths. More interesting is the cost per unit volume of treated water, but this would require knowing the lifetime of the tube when used in the system. An increase in tube diameter would reduce the system length significantly, while the cost of the tube per meter will increase slightly (see Table 1). Further studies are needed to estimate lifetimes and then cost per unit volume of treated water for different diameters of easily available tubes purchased on local markets. Further studies should also include the residence time for water with varying turbidity, covering realisitic values found in areas of target for SODIS.

In this paper, we experimentally studied the feasibility of using cheap, easily available plastic tubes in SOlar DISinfection (SODIS) systems. We used plastic PVC tubes that are easily available in local hardware stores and obtained a UV-A transparency ratio of ∼50% with the best option, compared to ∼60% for PET-bottles. In the chosen tube, we were able to eliminate concentrations of E. coli of 10⁶ CFU/ml in static batches with residence times of less than 4 h of exposure to solar radiation.

The results are very promising in the overall idea of using cheap, easily available plastic tubes to effectively treat contaminated water in a continuous flow SODIS system. Nonetheless, we stress the need to consider several other aspects prior to the full-scale use of continuous flow SODIS systems: the UV dose that determines the inactivation level, the lethal dose to deliver in an uninterrupted manner to reach complete inactivation, the plastic constituents that might be released into the water, and the longevity of the materials used.

Therefore, we conclude that cheap, easily available plastic tubes can be used for treating contaminated water with the SODIS technique. However, more research is needed to evaluate the use of these materials compared to other material alternatives as well as techniques to ensure a safe and sustainable future for coming generations.

The work published in this article is part of the master thesis Up-scaled continuous flow SODIS systems for deprived communities: with the aim of maintaining low costs by Jonas Johansson, presented in 2016 at the University of Gothenburg. Jonas Johansson would like to thank the Swedish International Development Cooperation Agency, SIDA (Styrelsen för internationellt utvecklingssamarbete), for financial support of the fieldwork in Colombia. The authors would also like to thank the anonymous reviewers for useful comments that helped improve the paper.

All relevant data are included in the paper or its Supplementary Information.