Waste stabilization pond (WSP) systems exist in many countries affected by climate change causing increasing demand for irrigation water. There is little long-term experience with WSP in Africa, and thus, this study provides a comprehensive investigation of a WSP in Namibia. One of its treatment trains was upgraded with mechanical and anaerobic pre-treatment, and post-treatment and compared to a second train without upgrades. The pre-treatment showed positive results on COD, TSS and partially pathogen removal. E. coli were reduced to the new EU reuse standard of 1,000 MPN/100 mL for fodder irrigation, P. aeruginosa stagnated and Enterococci levels increased. The main pathogen reduction happened during pre-treatment and in the baffled facultative pond. In contrast, the post-treatment only reduced 5% of algae-based chlorophyll-a. Among the top 20 genera, Pseudomonas was most dominant. With different upgrades, hygiene requirements especially for restricted fodder irrigation can be reached. At the same time, high total tCOD and TN effluent values did not meet different reuse standards. But as the tCOD mainly consists of algae, adding biomass and TN fertilizer to the barren soil, it raised the question if all parameters set in the standards are applicable for WSP or should be adapted to their specific purpose.
Long-term seasonal effects on the effluent quality of WSP are evaluated for the first time.
Increased hydraulic retention time and reduced hydraulic load show positive effect of on the removal of E. coli.
Selected pathogens behave differently in the WSP system.
Pre-treatment has a larger effect on algae reduction than post-treatment.
Operating conditions influence the microbial community in the WSP.
Waste stabilization ponds (WSP) are common wastewater treatment options in water scarce areas, especially in Africa, and also in other continents (Ho & Goethals 2020; Janeiro et al. 2020). Treated wastewater presents a valuable resource of water as well as nutrients; however, quality assurance is required, especially regarding pathogens. In Namibia, the effluent water quality of WSP often does not fulfill the national and international requirements for water reuse. At the same time, there is a high demand on water for irrigation purposes.
Traditional WSP systems consist of anaerobic ponds (AP), facultative ponds (FP) and maturation ponds (MP) (Shilton 2005; von Sperling 2007; Verbyla et al. 2017). If there is no receiving water body, an evaporation pond (EP) is also needed. The main purpose of AP and FP is the removal and stabilization of organic matter, while that of MP is designed to remove pathogens (Verbyla et al. 2017).
Operators of WSP systems are facing fast growing urban populations with rapidly increasing inflows to their WSP, which leads to overloading and overflowing (Verbyla et al. 2013). The classical way of meeting this challenge is to increase the number of ponds and their surface. However, the related raising of evaporation is contradictory to the aim of water reuse. Therefore, sustainable solutions are needed to achieve sufficient water quality and at the same time reduce evaporation losses. WSPs have been studied extensively in Latin America (von Sperling & De Andrada 2006; von Sperling et al. 2007; Pham et al. 2014; Verbyla et al. 2016; Dias et al. 2017), Australia, New Zealand and the USA (Powell et al. 2011; Guieysse et al. 2012), but there is only punctual research in sub-Saharan countries such as Burkina Faso, Ghana, Malawi and Tanzania (Maiga et al. 2009; Konaté et al. 2013; Kihila et al. 2014; Bansah & Suglo 2016; Zacharia et al. 2019; Ngoma et al. 2020) on WSP, particularly on reuse for the irrigation of fodder crops.
Algae and cyanobacteria are indispensable for WSP. Within the algal–bacterial mutualism, they produce oxygen and consume nutrients and CO2. They also play an important role in the removal and inactivation of pathogens (Liu et al. 2020). However, only little information is available on the microbial communities in WSPs. Eland et al. (2018) quantified cyanobacterial and eukaryotic communities in two WSPs in Brazil, and Wallace et al. (2015) evaluated algae and macrophyte species distributions in three WSPs in Canada. Due to high algae contents, the effluent from the WSP presents an opportunity to add biomass into barren soil and improve its water-holding capacity (Mara 2004). But negative effects of algal toxins need to be considered (Ho & Goethals 2020). Algae contribute considerable amounts of particulate chemical oxygen demand (pCOD) to the effluent and are therefore important regarding reuse requirements.
According to Verbyla et al. (2017), the most important factor for pathogen reduction in WSP is sunlight. Temperature, turbidity, dissolved oxygen (DO), pH, sedimentation and hydraulics are also important but have varying effects on the removal of different pathogens. Most of these factors are also influenced by the amount and activity of the algae. So far, there have only been short-term investigations and literature reviews on pathogen reduction in the WSP (Nikiema et al. 2013; Ho & Goethals 2020; Janeiro et al. 2020; K'Oreje et al. 2020), but no long-term studies have been conducted on the WSP in sub-Sahara Africa.
This research presents an extensive evaluation of factors affecting the treatment performance and effluent quality of the WSP. A WSP system in Namibia was monitored over 4 years under several operational conditions, i.e. overloading of one treatment train, load reduction scenarios and additional pre- (PreT) and post-treatment (PostT) for enhancement. These scenarios were assessed regarding the removal of the COD, total suspended solids (TSS) and selected pathogens. Additionally, nutrient levels and algae content were considered as important parameters for water reuse.
The following research questions were addressed: What are the seasonal effects on the effluent water quality of WSP comparing an overloaded with an enhanced treatment train? How does a reduced hydraulic load influence the performance? What is the impact of additional treatment technologies on the performance of a WSP? How are the composition and concentrations of pathogens as well as the microbial community in general affected by these operation scenarios and improvements? Which enhancements have the highest removal potential? Can enhanced WSP fulfill Namibian and European water reuse standards?
MATERIAL AND METHODS
Configuration of the WSP system
The investigated WSP system (Figure SI 1) is located in North Namibia and consisted of two parallel treatment trains A and B, each with one primary FPs and three MPs with almost identical pond surface and volumes. (further information in Sinn & Lackner (2020)). In order to enable water reuse for irrigation treatment, train A was enhanced with a micro-sieve (MS) and in parallel with an upstream anaerobic sludge blanket (UASB) reactor as pre-treatment (PreT) to reduce TSS and organic carbon (measured as COD) and floating baffles in pond A1 to approach plug-flow conditions. As PostT, a rock filter was constructed to improve the effluent quality and reduce algae concentrations. For further information on the PreT and the PostT, the reader is referred to Sinn & Lackner (2020) and Rudolph et al. (2020), respectively. The almost identical treatment train B remained in its original setup as comparison. Operation was divided into three phases. During phase I (day 1–day 675), the PreT was constructed, and during phase II the PostT. With the start of phase III on day 1,012, the whole plant was operational (Table SI 1). For phases I + II, train B received 800 m3/d inflow, and during phase III 430 m3/d. After commissioning, train A received on average 350 m3/d in phase III. The inflow into pond B1 was raw wastewater, while in A1, it was pretreated. During the 1 year of full operation, there was a 50-day suspension of the PreT after lightning stroke on day 1,275. Train A was back to full capacity on day 1,362 (Figure 2(c)).
Sampling and analyses
During 4 years of operation, 1-L grab samples were taken regularly at the inflow and overflows of each of the eight ponds. Electrical conductivity (EC), pH, temperature and DO were analyzed directly on site with a WTW multimeter 3410 (Xylem Analytics, Germany), and samples were cooled and transported to the laboratory. The samples were analyzed after homogenizing or filtration (0.45 μm, Whatman membrane, ME 25) with Hach cuvette tests using a spectral photometer DR 2800 (Hach Lange, Germany) for COD, total nitrogen (TN), ammonium (NH4-N), nitrite (NO2-N), nitrate (NO3-N), total phosphorus (TP), phosphate (PO4-P) and potassium (K+). According to standard methods (DIN 38409-2, 1987) (DIN 1987), TSS and volatile suspended solids (VSS) were measured with glass microfiber filters (Whatman 934-AH). Furthermore, chlorophyll-a concentrations were analyzed based on DIN 38409-60, 2019 (DIN 2019).
The following indicator bacteria were determined using an IDEXX system with Quanti-Tray/2000 (IDEXX, Germany): total coliforms and Escherichia coli (Colilert-18), Enterococci (Enterolert) and Pseudomonas aeruginosa (Pseudalert). For a selected number of samples, biomass was collected in biological triplicates and centrifuged in 50 mL tubes at 8,000 g and 4 °C for 25 min. After the discharge of the supernatant, the pellets were stored at 4 °C prior to further downstream analysis. Total genomic DNA was extracted with the FastDNA™ Spin Kit for soil (MP Biomedicals, Germany) based on a modified manufacturer's protocol (Orschler et al. 2019). The DNA concentrations were determined by a Qubit 3.0 Fluorometer with a Qubit dsDNA HS kit (Thermo Fisher Scientific, Germany), and DNA was subsequently used for 16S rRNA gene amplicon sequencing. We targeted multiple hypervariable regions of 16S rRNA genes with the 16S Ion Metagenomics Kit™ (Thermo Fisher Scientific, Germany) by two separate PCRs, amplifying the V2, V4, V8 and V3, V6–7, and V9 hypervariable regions, according to the kit protocol, as described in Agrawal et al. (2020). Sequencing was performed on an Ion Torrent (ION Torrent Ion S5) using the 400-bp kit and 530 chip. Base calling and run demultiplexing were conducted by Torrent Suite version 4.4.2 (Thermo Fisher Scientific, Germany) with default parameters. DADA2 (v.1.14.1) was implemented for separating sequences for each sample; filtering the low quality and limiting the length of sequences >260 bp and filtering sequences with potential chimera. The sequences were classified based on the taxonomy in the Silva database (97% confidence threshold, version 138). Diversity and compositional analysis were performed in R (http://www.R-project.org/) using phyloseq (1.30.3) and ggplot2 (3.3.0) packages.
The performance of the enhanced train A during phase III was compared with the water quality of train B in its original setup. The effluent quality was also evaluated in comparison with the code of practice for wastewater reuse in Namibia (DWAF 2012), the newly published regulation on minimum water requirements for water reuse in the European Union (EU 2020) as well as the necessary water quality for agriculture by the Food and Agriculture Organization (FAO) (Ayers & Westcot 1985).
In each train, the MP hydraulic retention time (HRT) was calculated with the water flow divided by the related pond volume. Given the strong variations of the wastewater flow, the HRT was averaged over 21 days. This included 10 days before and after each measurement.
Seasonal effects and influence of HRT
The second summer was dry with minor rainfalls. Therefore, this period represented the influence of higher temperatures without dilution effects. All effluent values reached their low point at the end of this season but did not drop as much as the year before. During the third summer, there were small rainfalls of 207 mm in total with high water temperatures of up to 35 °C. But the lowest effluent values remained higher than the 2 years before. EC was reduced to 700 μS/cm, tCOD to 400 mg/L, TN to 40 mg/L and TP remained above 10 mg/L in the outflow of B4.
The microbial parameters showed variable effects over 4 years. For total coliforms (Figure SI 2a), the outflow concentrations of B1 remained between six and eight log values at the same level without any seasonal influence. At the outflow of B4, there was more variation but hardly any removal. The lowest effluent values were reached after heavy rainfalls at the end of the first summer season.
On the contrary, the E. coli concentrations (Figure SI 2b) showed different behavior. The outflow of B1 remained at a constant level between five and seven log values over the whole research period, while the values at the effluent of B4 showed a clear decrease after the start of phase III and the increased HRT. Enterococci (Figure SI 2c) concentrations after B1 were constantly between four and five log values, while P. aeruginosa (Figure SI 2d) concentrations were slightly increasing over 4 years from five to seven log values. The effluent concentrations at B4 remained at similar levels for P. aeruginosa even with the reduced inflow from day 1,012 onwards. For Enterococci, the effluent concentrations of B4 increased from below 1,000 MPN/100 mL to about 10,000 MPN/100 mL after the commissioning of train A.
During phase III, train B received only half the daily plant inflow and still the yearly variations were visible (Figure 1). With rainfalls similar to the first year, the dilution effect appeared but was not as distinct. The low point of EC was 600 μS/cm, while tCOD was 250 mg/L, TN at 40 mg/L and TP at 7 mg/L. With the last winter season, all effluent values increased. EC reached its high point with over 1,000 μS/cm, while TN and TP stayed within their general fluctuation. Only tCOD remained below 400 mg/L, which could be a first indication of reduced inflow loads and higher HRT.
Upgrade: pre-treatment and post-treatment in train A
With the commissioning of the enhanced train A tCOD (Figure 2(a)), concentrations in the effluent of pond A4 started with 300 mg/L, increased to almost 500 mg/L in the middle of the summer and fell below 200 mg/L at the end of the summer. Afterwards, it remained stable around 150 mg/L.
While the tCOD showed seasonal effects, the soluble COD (sCOD) concentration (Figure 2(b)) had only little variations of 20–40 mg/L. The effluent of pond A1 had the lowest sCOD concentrations between 60 and 80 mg/L. The final concentrations in the effluent of A4 had a higher variation between 60 and 100 mg/L. Nutrients followed a similar pattern to tCOD. TN, NH4-N and TP concentrations (Figure 2(e)–2(g)) are presented and further discussed in the Supplementary Information.
Gradients along the treatment trains
In the context of WSP, the measured tCOD not only consists of components originating from wastewater but also from biomass due to algae growth. Therefore, the sCOD also needs to be considered. At the inflow, the sCOD (Figure 3(b)) had an average concentration of 315 (±68) mg/L. At the overflow of A1, the removal was 78% with an average concentration of 71 (±16) mg/L. For B1, it was 75% with an average concentration of 80 (±25) mg/L. Afterwards, the sCOD did not further decrease in both trains and remained at 89 (±20) mg/L in A4 and 85 (±21) mg/L in B4. This trend was different for the removal of nutrients (Figure 3(c) and 3(d)), and the data are discussed further in the Supplementary Information.
For microbial indicators, different trends were measured. The inflow concentration of total coliforms (Figure SI 3a) was on average 7.3 × 107 (±4.2 × 107) MPN/100 mL. After pond A1, this was reduced by 1.1 log values to 5.7 × 106 (±1.1 × 107) MPN/100 mL, and after B1 this was reduced by 0.7 log values to 1.3 × 107 (±1.8 × 107) MPN/100 mL. Over the following ponds in train B, there were only small changes with a final effluent reduction in B4 of 0.8 log values to 1.2 × 107 (±1.5 × 107) MPN/100 mL. In train A, there was a small up and down over A2 and A3 and the final effluent concentration of A4 was higher than that of A1. Compared to the inflow concentration, a 0.9 log values reduction was observed down to 8.9 × 106 (±1.2 × 107) MPN/100 mL.
E. coli showed a different trend (Figure SI 3b), as the average inflow concentration to the plant was 1.2 × 107 (±8.1 × 106) MPN/100 mL. This value was reduced through the PreT and the baffles by 2.0 log values down to 1.2 × 105 (±1.9 × 105) MPN/100 mL after A1. In B1, the reduction was 1.1 log values down to 9.4 × 105 (±9.4 × 105) MPN/100 mL. In the following ponds B2, B3 and B4, there was a continuous removal by overall 3.8 log values down to 1.8 × 103 (±2.6 × 103) MPN/100 mL. At the outflow of ponds A2 and A3, there was a removal of 3.4 and 4.3 log values, respectively. It leveled out with pond A4, and a reduction of 4.5 log values to 3.6 × 102 (±6.8 × 102) MPN/100 mL was achieved overall.
The behavior of the Enterococci concentrations was different (Figure SI 3c). From the inflow with 1.2 × 106 (±6.6 × 105) MPN/100 mL, train A reduced their concentration by 2.7 log values after A1. In comparison, train B showed 1.6 log values in B1. The final overall log reduction in B4 was 1.8–1.9 × 104 (±3.1 × 104) MPN/100 mL. In train A, no significant further reduction compared to the inflow concentration was measured. On the contrary, it increased again between A3 and A4 with a total overall log reduction of 2.3–5.3 × 103 (±4.0 × 103) MPN/100 mL.
For P. aeruginosa (Figure SI 3d), there was a decrease in the concentration from 3.5 × 106 (±2.4 × 106) MPN/100 mL at the inflow of 1.8 log values in A1 and 0.5 log values in B1. In train B, there was a significant further reduction from B1 to B2 to B3, but then it leveled out with overall 2.6 log values reaching 7.9 × 103 (±6.8 × 103) MPN/100 mL. In contrast, further along train A, there was only little more reduction of 2.8 log values to a value of 6.0 × 103 (±5.7 × 103) MPN/100 mL in A4.
Even with stagnant concentrations, there was a continuous load reduction for all parameters (Figure SI 4). The inflow load was distributed with 47% into train A and 53% into train B. The tCOD (Figure SI 4a) at the inflow was 601 (±186) kg/d, and reductions of 54 and 41% at the outflows of A1 and B1 were observed, respectively. The maximum removal of 75% was reached at A4 with 70 (±54) kg/d. In comparison, B4 had almost double the effluent load of A4, with 135 (±77) kg/d and a total removal in train B of 58%.
This trend was also visible for TN (Figure SI 4b) and NH4-N (Figure SI 4c), with the highest load removal. Only looking at the rising TP concentrations (Figure 3(d)), there could be the impression that there was no removal at all. This, however, was not the case and can be shown with the daily TP loads leaving each pond (Figure SI 4d). Details are presented in the Supplementary Information.
Natural disinfection – reduction of pathogens
With regard to the corresponding log reductions of pathogens in all three operating stages (B – I + II; B – III; A – III), there was only a small difference for the reduction of total coliforms (Figure 4(a)). During the operation of B – I + II, the average log reduction was 1.1 (±0.6). After the commissioning, A – III achieved a reduction of 1.4 log values and the same deviation of ±0.6. During stage B – III, only a reduction of 1.0 (±0.5) log values occurred. However, there was an effect on the reduction of E. coli (Figure 4(b)). During B – I + II, the average log reduction was 3.1 (±1.5). This reduction increased to 3.9 (±0.6) log values during B – III. And during stage A – III, an even better value with a log reduction of 5.1 (±0.8) log values was achieved.
Correlations of these log reductions with chlorophyll-a, turbidity, temperature, precipitation, solar radiation and inflow concentrations were evaluated and seem to be influenced by a combination of several parameters. The T-test (Table SI 2) indicated that there was also no significant correlation between precipitation and log reduction. Between chlorophyll-a and log reduction, there was a good significant correlation (p < 0.01), and with all other parameters, there was a high significance (p < 0.001).
Bacterial community composition and algae
The overall microbial community composition varied between inlet samples and effluent samples from both trains (Figure SI 5). Due to PreT steps introduced in train A before sampling point A1, in comparison to train B, the samples from A1 and B1 clustered separately. Also, the samples of A3 and A4 were in different clusters. The effluent samples from both trains clustered together, except for samples from days 1,312 and 1,328. However, no significant difference between the richness and Shannon's diversity indices was observed (Table SI 3).
The phyla Proteobacteria and Actinobacteriota dominated in all samples (Figure SI 6). In the inflow, A1 and A3, Actinobacteriota accounted for 47, 49 and 40% of the relative abundance, respectively, followed by Proteobacteria (34, 32 and 30%, respectively). In A4, B1 and B4, the average relative abundance of Proteobacteria (i.e. 50, 50 and 44%, respectively) was higher than Actinobacteriota (i.e. 31, 27 and 32%, respectively).
Within the biomass sample of the influent, only an exiguous percentage of Planktothrix and unclassified species were found (Figure 6(b)). At the effluent, the relative abundance in the bacterial biomass increased to over 3% with Chlorella being the dominant genus. Only a small percentage of biomass consisted of potentially toxic Cyanobium (Figure 6(c)).
Effluent quality – seasonal effects and influence of hydraulic load
Effluent values are influenced not only by the treatment technology but also by the surrounding environment and especially the local climate (Shilton 2005; von Sperling 2007; Maiga et al. 2009). Therefore, the seasonal effects and the influence of the reduced hydraulic load were examined by comparing train B during phase I and II (B – I + II) with phase III (B – III). This information allows us to better judge the effectiveness of train A and to distinguish if the lower effluent values were caused by the enhancement measures, seasonal effects or increased HRT.
Over 4 years, regular seasonal effects were visible in train B for different parameters (Figure 1). The lowest effluent values, e.g. for tCOD or EC, were reached at the end of the summer season. This was partially due to the highest temperatures and therefore highest microbial activity, but also due to dilution effects from rainfalls, mainly toward the end of the summer season. These effects were happening independently of the enhancement measures.
Therefore, the decline in COD concentrations in train A from day 1,200 can also be attributed to microbial and dilution effects as it coincided with the highest temperatures and toward the end of the summer with heavy rainfalls (Figure 2(d)). Secondly, the suspension of the PreT reduced the inflow into train A from day 1,275 and resulted in a longer HRT. However, after being back to full operation on day 1,362, effluent concentrations remained low. Especially tCOD and TN were increasing slowly toward day 1,444 to only half the effluent concentrations compared to the beginning. Whether this was related to the rock filter as suggested by Rudolph et al. (2020) or it was caused only by the higher HRT cannot be finally concluded and has to be subject of future research.
The NH4-N concentration was strongly affected by rising temperatures that became visible by the reduction of the concentrations in pond A3 during summer and their stagnation at around 1 mg/L during the interruption in phase III (Figure 2). With full inflow and decreasing temperatures, the concentrations started rising again. At the same time, the behavior in A4 was different. The concentrations dropped to about 5 mg/L during summer but with the interruption and stagnation from day 1,275 onwards, they increased up to 25 mg/L and later reduced again to 1 mg/L before regular inflow started again on day 1,362. This indicates that within pond A4 and especially in the rock filter, NH4-N redissolved during stagnation periods due to anaerobic conditions. So, ideally stagnation should be avoided within the rock filter. Overall, train A with the enhancements shows a 64% better reduction of NH4-N in the effluent (Figure SI 4c), suggesting also a significant contribution of the enhancements to the increased performance.
Varying effects on pathogens were also shown by Maiga et al. (2009). In their case, increased solar radiation was more harmful to Enterococci than to E. coli. For our WSP, increased HRT would be associated with longer solar radiation and therefore better removal of Enterococci. In our system, the increased HRT had better removal of E. coli and negative effects on Enterococci. Liu et al. (2020) showed the persistence of Enterococci with low DO, while E. coli further reduced. This trend was similar in our system.
Impact of additional treatment technologies
The additional PreT and PosT technologies in train A showed positive impacts on the treatment performance compared to the original train B but still did not fulfill all reuse standards. For tCOD, a concentration of 100 mg/L is required in Namibia (DWAF 2012) and 125 mg/L in Europe (EU 2020). The best average effluent value of 317 mg/L tCOD was reached in A4 during phase III with minimum values of 127 mg/L just near the reuse standard of the EU. The increased HRT only resulted in an average outflow concentration in B4 of 415 mg/L. However, tCOD load reductions indicate an important effect of the enhancement measures in train A. The PreT and baffles already led to a 13% better result, and with the rock filter, this improved to 17%. This is considerably lower than the reported reduction of 15–25% with an MS (Lazarova & Bahri 2005; Prösl et al. 2013) and more than 60% for a UASB (Dias et al. 2017; Vassalle et al. 2020). However, those studies do not consider algae growth in the following ponds that add considerable new COD. If all pCOD is assumed to be related to algae, the sCOD would be an alternative indicator for COD reduction. For the given plant, both trains delivered a sCOD effluent quality below 90 mg/L and therefore below both reuse standards.
With regard to nutrient concentrations, the Namibian reuse standard (DWAF 2012) requires a concentration of 33 mg/L TN, which was originally aimed at effluents in areas with potential drinking water sources. In the EU there are no specific limits for nitrogen and phosphorous as these nutrients are supposed to be returned into the biological cycle (EU 2020). In order to protect crops and soil, Ayers & Westcot (1985) require severe restrictions for values above 30 mg/L TN. With average effluent values above 55 mg/L in train B, no positive effect could be attributed to the increased HRT. But clear advantages of the enhancements were visible in train A with 39 mg/L TN and minimum values of 11 mg/L. This was also visible for TN effluent loads. With the PreT and the baffles, train A reached a 10% better load reduction than train B, and with the rock filter, it increased to 16%. In total, A4 had a 34% better TN effluent concentration than B4.
For irrigation purposes, the Namibian code of practice (DWAF 2012) allows a maximum of 15 mg/L TP, which was reached by both trains with 11 mg/L. But these values were even higher than the average inflow concentration of 10 mg/L. The high evaporation of 27% (Sinn et al. 2022) has a large effect on the effluent concentration, which coincides well with Buchanan et al. (2018) who showed that there is only limited phosphorous removal in a WSP. TP assimilated by algae leaving with the final effluent might also add to the observed concentrations. Nevertheless, up to 36% of the daily TP loads were reduced in train A (Supplementary Information).
Composition of pathogens and their reduction in the WSP
The different treatment technologies also have various effects on the pathogens. Their composition and concentrations were affected in various ways. Over the whole WSP, total coliforms were less dynamic than COD and nutrients. They were hardly reduced and there was no significant difference between the effluent of the enhanced train A and the original train B. In train A, the rock filter even had a negative effect with a slight recontamination of total coliforms. Such an increase could be induced by either a reduced penetration of natural UV-radiation by the rock filter or suitable conditions for regrowth in the biofilm of the rock filter.
Other pathogens showed a different behavior. E. coli were constantly reduced in B4 to 4.0 × 104 MPN/100 mL during phase I and 1.8 × 103 MPN/100 mL during phase III, and to 3.6 × 102 MPN/100 mL in A4. Therefore, the increased HRT had a first positive effect on E. coli, resulting in a reduction of one log value and further enhancement measures improved this value by another log value. However, the observed reduction of E. coli was more related to the effects of the ponds in line A rather than the rock filter. In A4, the rock filter had no additional positive effect on E. coli concentrations. In contrary, it hindered further reduction as it happened from B3 to B4. Dias et al. (2017) also reported the best E. coli removal of 2.2 log values by ponds; however, their granular rock filter still achieved a removal of 1.0 log value. Nevertheless, the reductions that were obtained here reached the required EU effluent value of 1,000 MPN/100 mL (EU 2020) for fodder irrigation with train A and were just missed with train B.
This study additionally evaluated the removal of Enterococci and P. aeruginosa. Both pathogens behaved quite differently compared to the standard indicator E. coli. Enterococci almost continuously reduced over the different ponds during phase I down to a concentration of 5.0 × 102 MPN/100 mL in B4. During phase III, the steepest decrease occurred in the FP in both treatment trains and there was hardly any change in the MPs. Also, there was no significant difference between the two trains with B4 reaching 1.9 × 104 MPN/100 mL and A4 5.3 × 103 MPN/100 mL. Enterococci are best reduced with the PreT and the baffles in A1. It seems that the reduction occurs mainly through sedimentation rather than sunlight. Differences in the inactivation of Enterococci compared to the traditional indicator E.coli were also reported by Liu et al. (2020). P. aeruginosa followed a similar pattern with the difference that, during phase III, there was some further reduction in the first MP (up to the effluents of A2 and B2) before concentrations stagnated. Also, the final concentrations of P. aeruginosa in the effluent were almost identical between A4 and B4. A positive effect of the PreT and the baffles on the reduction of P. aeruginosa was also visible. So far, there is only limited research on the behavior of P. aeruginosa in the WSP, and thus, this study provides first novel insights into the behavior of these bacteria in the WSP. Søberg et al. (2019) reported that the trends for P. aeruginosa correlated strongly with Enterococci in stormwater bioretention systems. This could not be confirmed with this research.
The effects of PreT, baffles and rock filter vary depending on each specific parameter, as train A has better effluent values due to the PreT and baffles. For the rock filter, in A4, there is no evidence of any positive effect. In contrary for E. coli and Enterococci, it has a stagnant or negative effect. This research supports the findings of Liu et al. (2020) that E. coli should not be the only indicator organism to evaluate pathogen removal.
Microbial community composition and algae development
Not only the concentrations of specific pathogens, but also the whole microbial community and algae can be affected by the different treatment technologies. Especially algae compose a considerable part of the pCOD. Given the local circumstances, the best way to estimate the algae content in the ponds was through the concentration of chlorophyll-a. This does not cover all types of algae but mainly Cyanobacteria that could potentially be toxic (Vidal et al. 2021). While the average concentrations in train A were between 1,000 and 2,000 μg/L, train B showed higher values of up to 3,300 μg/L of chlorophyll-a. Beside the positive effect of added biomass to the soil, algae also increase the tCOD (section ‘Impact of additional treatment technologies’). Another negative effect is the risk of blockages in drip irrigation systems. Therefore, a rock filter was investigated for its removal of the algae content (Rudolph et al. 2020). However, no significant chlorophyll-a removal was visible when comparing chlorophyll-a concentrations between A3 (1,375 μg/L) and A4 (1,305 μg/L). The main effects on the algae were related to the PreT, the floating baffles and the removed sludge, which all happened upfront or in pond A1. Only long-term observations will show if the rock filter develops a positive algae removal with the increasing growth of biofilm on the rock surface.
Another important aspect for the reuse of algae containing water is their toxicity. The results of this study show that only a very small percentage of the Cyanobacteria were potential toxin producers. Cyanobium PCC-6307 was not dominant in samples from A4 but the second most dominant genus in samples from B4. Interestingly, Candidatus aquiluna, a photoheterotroph (Kang et al. 2012), seemed to be replaced from the second most abundant in the inflow by either Cyanobium PCC-6307 or C39 genus of the Rhodocyclaceae in the samples of A4 and B4. C39 has been mainly detected in freshwater bodies (Carney et al. 2015; Cannon et al. 2017). A high amount of algae was observed in samples from A4 and B4, which might explain the high abundance of genus C39. A previous study emphasized a plausible association between other algae and genus C39 (Cannon et al. 2017). However, whether toxins are released depends the environment and stress (Vidal et al. 2021). A more detailed evaluation of potential toxins was not possible and should be further examined in future research.
During the course of operation of both trains, a sequential change in the microbial community occurred; however, no change in the dominant phyla (i.e. Proteobacteria and Actinobacteriota) was observed (Figure SI 5, Figure SI 6). Both of these phyla include microorganisms (such as Pseudomonas and Actinobacteria) known for enhanced phosphorous removal (Lee et al. 2002), which might explain the reduction of the TP load in both trains. There also seemed to be a positive effect of PreT on the reduction of Mycobacterium (Figure 5), which is a re-emerging pathogen (Taylor et al. 2001) but has no impact on Pseudomonas. This underlines the persistent nature of Pseudomonas in the pond system.
This research detailed different effects and operating conditions on the behavior and effluent quality of a WSP system in Namibia over a 4-year period. The main findings are:
Regular seasonal effects on different parameters are visible in the overloaded and enhanced train over 4 years of operation.
Increasing HRT and reduced hydraulic load have a positive effect on the removal of E. coli and the required 40-day HRT is achieved with both trains in operation.
Mechanical and anaerobic biological PreT such as MS and UASB have a positive effect on the removal of COD, TSS and to some extent pathogens.
Within the first year of operation, the rock filter did not show any additional removal of algae compared with the original train. Long-term operation and monitoring will show if more biofilms are growing on the rocks and if they will improve removal.
Best measures to reduce the algae concentrations are emptying of the sludge in the first pond, installing PreT and baffles in the FP.
The studied pathogens exhibit different behavior: E. coli were reduced, P. aeruginosa stagnated and Enterococci levels increased. The main pathogen reduction happened during PreT and in the first pond (FP enhanced with baffles) and not in the maturation ponds.
Future detailed research is needed into the pathogen removal capacity of WSP as E. coli do not seem to be the best indicator for broader pathogen reduction.
With the enhancement, the WSP reaches the new EU water reuse standard for E. coli.
High tCOD and TN effluent values currently do not meet Namibian and European reuse standards. But a large portion of the tCOD consists of algae which add, as long as they are not toxic, valuable biomass to the barren soil. Also, TN is a valuable fertilizer, and depending on the selected crop, this can be an important asset, especially if additional water from other sources with low nutrients is used.
This work was supported by the German Federal Ministry of Education and Research within the joint research project EPoNa (grant no. 02WAV1401A-F) under the funding measure WavE. We would like to thank Peter Cornel for his initiative, commitment and invaluable advice. We would also like to thank Lydia Luvinga, Haikela Nahambelelwe, Mikael Hidulika and all the students for their regular sampling and water quality analyses over 4 years of the project and the Outapi Town Council (OTC) for their support and access to their facilities.
DATA AVAILABILITY STATEMENT
The raw fastq sequences are available in the NCBI Sequence Read Archive under BioProject number PRJNA796294: https://www.ncbi.nlm.nih.gov/bioproject/?term=PRJNA796294.
CONFLICT OF INTEREST
The authors declare there is no conflict.