Performance of constructed wetland integrated with sand filters for treating high turbid water for drinking

In Rural and peri-urban areas of Tanzania, the use of unsafe drinking water from surface water sources like rivers, lakes, earthen dams, wells and open springs is not optional due to limited piped water supply (Mohamed et al. 2016). Therefore, to limit the outbreak of waterborne diseases it is vital to eliminate pathogens, organic matters, suspended solids and chemicals that can have serious human health effects (Gupta & Chaudhuri 1995; Groendijk & de Vries 2009; Mahmood et al. 2011). Point of use (POU) house hold water treatment technology can help individuals without access to safe and clean drinking water to treat water at their homes thus getting rid of waterborne diseases (Sobsey 2002; Sobsey et al. 2008; Jenkins et al. 2011; Kennedy et al. 2012).

In developing countries, common POU current used in homes include, biosand filters (BSF) and slow sand filter (SSF) (Sobsey et al. 2008). About 500,000 people worldwide depend on BSF for safe drinking water supply and its efficiency on faecal coliform (FC) and Escherichia coli removal have been documented (Duke et al. 2006; Stauber et al. 2006; Elliott et al. 2008). The growing interest to application of sand filters in water quality improvement is due to their low cost, convenient operation and easy maintenance while achieving high treatment efficiency (Logsdon et al. 2002; Nassar & Hajjaj 2013; Haig et al. 2014). These attributes make sand filters as cost effective technology to treat contaminated water in rural areas (Aslan & Cakici 2007; Langenbach et al. 2009).

The removal mechanisms of water contaminants by sand filters are purely physical, chemical and biological processes (Langenbach et al. 2009; Bauer et al. 2011; Ijadunola et al. 2011; Haig et al. 2014). Basically, gravity flow is the main operating force for influent flow (Young-Rojanschi & Madramootoo 2014). Physico-chemically raw water contaminants such as organic matter, suspended matters are removed by screening, sedimentation, adsorption, straining, adhesion, diffusion and flocculation (Mwiinga 2011; Bagundol et al. 2013). Furthermore, formation of biological layer by microorganisms on surface of sand bed facilitates the biological removal of water contaminants like pathogens through predation, scavenging, adsorption and bio-oxidation (Joubert & Pillay 2008; Hsieh et al. 2010; Elliott et al. 2011; Mwabi et al. 2013; Haig et al. 2014). According to Elliott et al. (2011), proteolytic enzymes produced by microbial exoproducts are responsible for pathogen reduction. Normally sand filters removal rates depend on filter depth, sand type, sand size and filtration rate (Abudi 2011). Long filter depth increase travel distance thus high pollutants removal rate (Ellis & Wood 1985).

Several authors have reported the efficiency of sand filters in removing different contaminants in water to be between 90 to 99% (pathogenic bacteria) by 85–90% (viruses), 87–96% (turbidity), 94–99% (nitrate) and >99.9% (protozoans) (Stauber et al. 2006; Aslan & Cakici 2007; Elliott et al. 2008; Jenkins et al. 2011; Mahmood et al. 2011; Kennedy et al. 2012; Mwabi et al. 2013; Young-Rojanschi & Madramootoo 2014). Nevertheless, their treatment efficiency can decrease with high turbid water (>10–50 NTU), high organic loading and high amount of microorganisms in raw water because, they tend to clog the filters and decrease the filter run time and treatment efficiency (Logsdon et al. 2002; Ray & Jain 2011). Therefore, the need for a reliable pretreatment technology before using sand filters is inevitable.

Coagulation and flocculation by inorganic and organic compounds are amongst pretreatment technologies, which are used to clean high turbid water however, they are not cost effective to rural communities (Silva et al. 2013; Ramavandi 2014). Furthermore, they cannot remove other water contaminants like heavy metals, organic materials, pesticides, oil and excess nutrients. Roughing filters have also been used prior to sand filters for water turbid removal (Nkwonta & Ochieng 2009; Mwiinga 2011). However, it is generally limited to turbidity removal due to absence of macrophytes as it has been reported that vegetated system perform better in FC removal than unplanted system (Davies & Bavor 2000). Normally, vegetation helps in settling of fine particles and their associated bacteria, but also increases the physical contact area between the bacteria, plant roots and the wetland substrate (Gerba et al. 1999). Moreover, plants roots produce antibacterial which aid in pathogens removal (Vymazal 2005) and stabilization of the hydraulic conductivity thus prevent filter clogging (Cordesius & Hedstrom 2009). Therefore, constructed wetland can be considered as viable pretreatment technology that can reduce water turbidity and other pollutants found in water prior to sand filters.

Horizontal subsurface flow constructed wetland (HSSFCW) is a robust system that need low energy, easy operation and maintenance thus allowing decentralization of water treatment system (Vymazal 2009; Wu et al. 2014). The system combines physical, chemical and biological processes to remove water pollutants (Sundaravadivel & Vigneswaran 2001; Stottmeister et al. 2003). In this system physico-chemically pollutants are removed by sedimentation, filtration, adsorption, precipitation, sorption, photochemical oxidation, disinfection and volatization (Karim et al. 2004). Biologically water pollutants are removed by predation, scavenging, natural die-off, nitrification, denitrification, algae and plant uptake, aerobic and anaerobic biodegradation (Karim et al. 2004; Vacca et al. 2005; Vymazal 2005; Díaz et al. 2010).

Reports from several studies indicate that the efficiency of HSSFCW system to reduce pollutants in water ranges between 80–90% (chemical oxygen demand (COD)), 50–80% (turbidity) and >90% (FC) (Cordesius & Hedstrom 2009). On other hand it has reduced 95.9% and 90.6% for tannins and COD, respectively, form tannins extracting company (Njau & Renalda 2010). Also, in a study conducted at the University of Dar es salaam (Tanzania) demonstrated the removal efficiency of 93.2 ± 6.13%, 92.6 ± 6.05%, 71 ± 6.2%, 58.1 ± 135.56% and 40.1 ± 14.5% for FC, Escherichia coli, organic matter, nitrate and phosphate, respectively (Mairi et al. 2013). This study aimed to assess the removal efficiency of physico-chemical and biological pollutants from high turbid water by integrating constructed wetland and sand filters and justify the possible application of such technology in rural communities to increase the availability of potable water.

The experimental integrated treatment system was set and executed in the ventilated green house and water quality Laboratory at Nelson Mandela African Institution of Science and Technology (NM-AIST) in Arusha region Tanzania. The area is located at Latitudes 03^o24′S, Longitude 036^o 47′E with elevation of 1,204 m above sea level. HSSFCW was set in the green house while the integrated parts of sand filters were set in water quality laboratory. Turbid water of >1,000 NTU from Nadosaito dam in Monduli district, Tanzania was used in this experiment.

The systems used in this study were composed of HSSFCW, SSF and BSF.

Water samples from the integrated system were collected from three points, CW influent, CW effluent or SSF/BSF influents and SSF/BSF effluents. Water samples for physico-chemical analysis were collected by using 500 ml polyethylene sampling bottles. These bottles were washed and rinsed thoroughly with distilled water then re-rinsed three times with the respective water samples before sample collection. Samples for microbial analysis were collected by using 250 ml glass bottles that were first washed and rinsed with distilled water followed by sterilization by autoclaving at 121°C for 2 hours. Before sampling all bottles were re-rinsed three times with the respective water samples. All samples were stored in ice cool box at 4°C before laboratory analysis.

Physico-chemical parameters like pH, temperature, dissolved oxygen (DO), electrical conductivity (EC) and total dissolved solids (TDS) were determined onsite by Hanna Multiparameter (HI 9829), whereas turbidity was determined by Hanna Turbidometer (HI 93703). The rest of parameters were analysed at NM-AIST laboratory. Total suspended solids (TSS) were analysed by gravimetric method at 105°C for 1 hour as per standard methods for the examination of water and wastewater (APHA 2012). Nitrate were determined by cadmium reduction method using HACH DR 2800 spectrophotometer while, COD by dichromate digestion method by using Hanna COD and Multiparameter photometer (HI 83099) as described by standard methods for the examination of water and wastewater (APHA 2012).

Microbial water quality was determined by analyzing FC indicator. FC were analysed by membrane filtration method where water samples were filtered through sterilized membrane filter of 0.45 micron pore size and 47 mm diameter. Thereafter, the membrane filters were placed in sterilized petri dishes that contained prepared MFC-agar medium and incubated at 44.5°C for 24 hours (APHA 2012). This allowed the faecal bacterial indictor to grow into blue colonies that were counted as colony forming unit per 100 ml of analyte (CFU/100 ml).

All data were processed in OriginPro software version 8.6 to obtain the trend of water pollutants removal by integrated systems. Descriptive statistics were used to summarize the data using IBM SPSS version 21 and the difference in removal efficiency between the integrated CW-SSF and CW-BSF was estimated by unpaired sample statistical t-test using R-software version 3.21. All the results at P < 0.05 were considered statistically significant.

The removal of pathogenic bacteria plus other water contaminants depend on biological and chemical reactions that are also affected by environmental factors like temperature, pH and DO.

Chemical and microbial activities taking place in constructed wetland and sand filters are affected by pH. Results from Tables 1–3 show that the influents and effluents pH from the integrated systems were observed to vary depending on the treatment unit. Generally, the pH was found to be decreasing from influent to effluent of CW and increasing from the influent to effluent of SSF and BSF. This trend might be due to anaerobic decomposition of organic matter in the constructed wetland and dissolution of mineral ions in sand bed filter media during infiltration (Mahmood et al. 2011).

The rate of physico-chemical and microbial activities in water is affected by temperature variations. Temperature controls the rate of pollutants removal in CW (Kadlec & Reddy 2001; Barea et al. 2005) for example, it influences the removal of pathogens in surface and subsurface flow CW (Molleda et al. 2008). From Tables 1–3 it was observed that influents had low temperature while effluents had high temperature for CW, SSF and BSF. The trend of temperature increase might be attributed to change in atmospheric weather condition of the experimental set sites that is within the green house and NM-AIST laboratory.

DO is an important parameter that ascertains physico-chemical and biological activities taking place in water (Efe et al. 2005). In CWs and sand filters DO facilitate in degradation of organic matter by aerobic microorganisms. Results from the Tables 1–3 show the decreasing trend of influents to effluents DO concentration for the integrated systems. The observed DO decreasing trend from both integrated systems might be attributed to microbial degradation of organic matters occurring in CW and sand filters. For instance, Bhatia & Goyal (2014), reported that, presence of rhizosphere in CW macrophytes provide an attachment area of aerobic microorganisms that utilizes DO on degradation of organic matter.

EC is the measure of ability of water to conduct electric current. It reflects the presence of dissolved solids and salinity in water. Results from Tables 1–3 show that, EC and TDS were observed to increase from influent to effluent values for CW, SSF and BSF. The general increasing trend of EC and TDS from a pretreatment unit and sand filters might be attributed to dissolution of ions during degradation of water pollutants in the systems.

The removal mechanisms of water turbidity in the CW is attributed to sedimentation and filtration facilitated by macrophytes roots that reduce interspaces between gravel by forming dense filter media that is capable of removing suspended particles (Yang et al. 2011). Moreover, turbid removal in sand filters is attributed to sedimentation, microbial biodegradation of suspended organic matter and filtration through the sand column.

Statistically unpaired sample t-test showed that there was significant difference in water turbid removal between CW-BSF and CW-SSF P < 0.05 (t = 3.82, df = 6.64, p-value =0.007). The improved performance of CW-BSF over CW-SSF might be attributed to batch feeding mode in BSF which allowed enough contact time between biolayer and sand column with raw water compared to continuous flow mode in SSF (Ahammed & Davra 2011; Jenkins et al. 2011). Moreover, small sand size and long filter depth of BSF compared to SSF might be the addition factor for CW-BSF better performance (Jenkins et al. 2011).

FC removal in constructed wetland is due to the development of biolayer that is colonized by microorganisms that kill other organisms by predation and scavenging (Vymazal 2005). Moreover, adsorption to filter media, filtration, sedimentation and natural die off contributed to FC reduction (Karim et al. 2004; Vymazal 2005; Díaz et al. 2010). In sand filters; predation, scavenging, resource competition, adsorption and natural die off might have contributed to FC removal (Elliott et al. 2011; Mwabi et al. 2013; Haig et al. 2014). According to unpaired sample statistical t-test, no significant difference in FC removal was observed between CW-BSF and CW-SSF P > 0.05 (t = 1.71, df = 5.56, p-value = 0.142), suggesting that sand size, long filter depth and batch mode feeding in BSF are not significant factors in FC removal, instead active biolayer and retention time was probably the main removal path.

The COD removal in integrated systems is attributed to filtration, sedimentation of suspended solids, aerobic and anaerobic degradation of organic matter by microorganisms in CW (Njau & Renalda 2010; Gersberg et al. 2015) and in sand filters (Mwiinga 2011; Bagundol et al. 2013). Statistically un-paired sample t-test showed that there was no significance different in COD removal between CW-BSF and CW-SSF (P > 0.05 t = 1.22, df = 10, p-value = 0.249). This means that mechanisms for organic matter removal in sand filters were not affected by the differences between BSF and SSF.

The TSS removal from CW is attributed to sedimentation and filtration of suspended particles by dense network of plant root. Furthermore, TSS removal by sand filters is attributed to biodegradation, particles sedimentation and entrapping in sand filter column. Unpaired sample t-test analysis indicated there was statistical significant difference in TSS removal between CW-BSF and CW-SSF P < 0.05 (t = 2.72, df = 6.59, p-value = 0.031). The better performance of CW-BSF over CW-SSF in TSS removal is attributed to long filter column, small sand size and long contact time of raw water in BSF due to pause feeding mode.

The nitrate removal in CW is due to nitrification and denitrification, sedimentation, volatization and plant uptake. However, in sand filters, nitrate removal is attributed to denitrification process and sedimentation in sand column. In this case, unpaired sample t-test showed that there was no significant difference in nitrate removal between CW-BSF and CW-SSF P > 0.05 (t = 0.56, df = 9.96, p-value = 0.591).

Treatment of high turbid water for drinking by constructed wetland integrated with sand filters proved to be feasible. It was observed that CW-BSF and CW-SSF systems were capable of producing an effluent quality that meets WHO and TBS standards for drinking water. CW-BSF needed an operational total retention time of 5 days to produce effluent of 0 NTU for turbidity, 0 CFU/100 ml for FC, 6.25 mg/L for COD, 0.5 mg/L for TSS and 4.2 mg/L for nitrate whereas, CW-SSF needed a total retention time of 7 days to produce effluent of 0.6 NTU for turbidity, 0 CFU/100 ml for FC, 6.5 mg/L for COD, 1 mg/L for TSS and 1.79 mg/L for nitrate. The allowable limits for drinking water standards according to WHO and TBS are 5–25 NTU for turbidity, 0 CFU/100 ml for FC, 10 mg/L for COD, 5 mg/L for TSS and 50 mg/L for nitrate. Therefore, this study reveal that CW-BSF shows better performance in treating high turbid water and qualify to be a potential low-cost, feasible and viable technology in providing potable water to under-served communities.

• (a) Further study should be undertaken to validate the efficiency of this technology in protozoa and virus removal from drinking water.

• (b) The study recommends the scaling up and application of this technology to real environment where treatment of turbid water is challenging.

This research work was funded by Government of Tanzania through Nelson Mandela African Institution of Science and Technology (NM-AIST). Therefore the authors wish to give their gratitude to the funder and NM-AIST for coordinating this work. Furthermore, the authors wish to extend thanks to Nadosaito community, village Chairman for their cooperation and support during this study.

No conflict of interest was declared.