Utilization of rural domestic sewage tailwaters by Ipomoea aquatica in different hydroponic vegetable and constructed wetland systems

Rural sewage treatment is often more difficult than urban sewage because of widely distributed populations, fluctuations in water quality and quantity, and defective drainage networks (Yu et al. 2015). To reduce water pollution in rural areas, many researchers focus on the treatment of domestic sewage, such as improvements in sewage treatment technology (Pan & Han 2016; Rout et al. 2016; Wang et al. 2016; He et al. 2017). However, the concentration of nitrogen and phosphorus in the tailwaters of treated sewage is often still high, which causes eutrophication and other environmental problems (Lewis et al. 2011). Therefore, it is of urgent importance to develop new approaches for the effective treatment of rural sewage tailwaters.

Constructed wetlands (CWs) have low processing costs and stable operation and contribute to water purification through microbial biodegradation, plant absorption, and substrate filtration (Wu et al. 2016). CWs are often employed in the purification of sewage tailwaters after passing through a biological treatment unit (Paruch et al. 2011). Landscape plants are generally carried out in CWs, which may wither in the winter, and the transport performance of plant-mediated oxygen to the rhizosphere varies among seasons, thereby reducing the purification efficiency and overall landscape-level benefits (Steer et al. 2002). However, CWs are still a reliable treatment technology under appropriate operating parameters and the relevant parameters, namely, macrophyte species, media types, water level, hydraulic retention time (HRT), and hydraulic loading rate (HLR), have been commonly studied (Gorgoglione & Torretta 2018).

A hydroponic vegetable system (HV) is a system to cultivate vegetables in sewage, which can produce vegetables with good economic value and purify a water body, fully realizing the utilization of nitrogen and phosphorus in sewage. And the hydroponic vegetable system is not filled with substance/fillers, which avoids the filling blockage problems in constructed wetlands. Besides, compared with the constructed wetland, the contribution of plant absorption to pollutant removal in hydroponic vegetable system was significantly increased (Furumai 2008; Lu et al. 2016a, 2016b). Zheng et al. (2010) studied the pollutant removal for municipal sewage in winter via a modified free water surface system planted with edible vegetables. For further irrigation application of tailwaters after biotreatment, plants' edible security should also be monitored. Xu et al. (2016) proposed technical requirements, standards, management, and monitoring methods for reclaimed water irrigation, which provides references for our research to fully recognize their differences and holistically consider the applicability.

CWs and HVs have broad application prospects for rural tailwaters. But most previous studies in this area focus on enhancing the purification of sewage by optimizing the substrate and choosing specific plant species (Lu et al. 2015; Kim et al. 2016; Lu et al. 2016a, 2016b; Zhiming et al. 2016). Vaillant et al. (2003) treated rural domestic sewage with HV and found that the removal efficiency of chemical oxygen demand (COD) and total phosphorus (TP) could reach 82% and 38% ±9%, respectively, when the hydraulic retention time was 48 h. Song (2005) treated eutrophic surface water with a hydro-biotic filter bed. The experiment showed that when the hydraulic load was 3.0 m³ /(m²·d), total nitrogen (TN) and TP removal efficiencies reached 16.8% and 30.8%, respectively, and the removal loading reached 1.0 and 0.1 g /(m²·d). Studies have focused on the optimization of the matrix in wetlands (Zhu et al. 2011) and the selection of plant species (Zhao et al. 2011). The combination of an HV and subsurface wetland (SFCW) to purify tailwaters has been less well studied.

In this paper, four systems were constructed: HV, SFCW, and two HV/SFCW combination systems. Experiments were carried out to analyze the effects of different hydraulic loads on pollutant removal. The systems were further studied by examining the utilization of nutrients by water cabbage, Ipomoea aquatica (IA). Also, the edible safety of IA was evaluated. The combination system proposed provides an improved approach for the utilization of nitrogen and phosphorus through a combination of wastewater treatment and agricultural production.

The four systems were built in Xinbei District, Changzhou, China. The outlet water (tailwater) of the pulse biological filter was driven into four cisterns with the lifting pump. Then the sewage flowed into four test systems driven by height differences and the inlet water flow rate was controlled by valves. The treated sewage was eventually discharged into a rural canal.

As shown in Figure 1, four constructed systems were: a hydroponic vegetable system (HV) (Figure 1(a)), a subsurface wetland system (SFCW) (Figure 1(b)), a compound system with HV followed by SFCW (HV-SFCW) (Figure 1(c)), and a compound system with SFCW followed by HV (SFCW-HV) (Figure 1(d)). In HV-SFCW, the HV unit was built before the SFCW unit and in SFCW-HV, the HV unit and SFCW unit were constructed in the reverse order. Dimensions are shown in Table 1. The length of the four systems was the same, as was the length of HV and SFCW in the two combined systems.

The HV was filled with nonsubstrate and planted with IA. The lower layers of the SFCW was laid with gravel (Φ20–40 mm, 0.2 m), the upper layers also with gravel (Φ10–20 mm, 0.25 m), the top floors with coarse sand (Φ0.5–1.2 mm, 0.25 m), and the plants were Iris.

Tests were conducted in summer (July to October) when the water temperature was 23.5–30.1 °C. The inlet water was the effluent of domestic sewage treated by a pulse biological filter, the sewage was raised to the water tank by lift pump, and then it continuously flowed into the test device from the water tank. The inflow water quality and the HRT under different hydraulic loading in four systems are shown in Tables 2 and 3.

After filtration through 0.45 μm filters, measurements and other specific operation methods were conducted on all water samples taken from the combined systems. The test methods and instruments used for the routine physical and chemical analysis are described below.

NH₄⁺-N, NO₂⁻-N, NO₃⁻-N, and TN were measured using an Auto-Analyzer 3 instrument (Seal Analytical, Ltd, UK). TP, COD, dissolved oxygen (DO), and pH were monitored using the ascorbic acid-molybdenum blue method (EPA-600/4-79-020), DR1010 COD analyzer (HACH Co., USA), DO probe (HQ30d, HACH Co., USA), and a pH probe (pH100, YSI Co., USA), respectively. Total microcystin (TMC-LR) and extracellular microcystin (EMC-LR) were tested with HPLC (Triplus Trace GC ITQ1100, THERMO FISHER Co., USA).

The effects of HV, SFCW, HV-SFCW, and SFCW-HV on pollution removal under different hydraulic loading (0.1, 0.2, and 0.3 m³/(m²·d)) conditions were compared to determine the optimal operating load of each system and to identify which had the maximum nitrogen and phosphorus utilization. We then analyzed the effects of different combined systems on pollutant removal under the optimal operating load condition.

To observe the removal effects of different sections of each system on the pollutants, several serial sampling points were set. Sample points in HV and SFCW were set at the front end of the system, outlet, 1/4, 1/2, and 3/4 distances along the pathway. Similarly, sample points in HV-SFCW and SFCW-HV were set at the front end of the system, outlet, 1/3, and 2/3 distances along the pathway. The values of hydraulic loading in HV, SFCW, HV-SFCW, and SFCW-HV were set at 0.2 m³/(m²·d), 0.3 m³/(m²·d), and 0.3 m³/(m²·d), respectively, according to the results of the effects of hydraulic loading on the removal performance of COD, TN, and TP.

Vegetables have different absorption performances during the different stages of growth, generally, absorption ability in prophase and anaphase is less than that in the middle growth phase (Rouphael & Growth 2005); therefore, one-time full harvesting will affect the ability of the system to purify water quality. The influence of vegetable harvesting methods on tailwater purification was examined in three parallel plantings, specifically when all of the vegetables were in growth stages (26 July to 10 August). All plants in corridor 1 were harvested on 26 July. In corridor 2, half of the vegetables were harvested on 26 July, and the left harvested eight days later. In corridor 3, 1/4 of the vegetables were harvested on 25 July, and then the same proportion successively harvested on the subsequent three days. Starting on 26 July, water samples were taken daily from the effluent of HV under a hydraulic load of 0.2 m³/(m²·d).

Tailwaters with a low heavy metal content are used for hydroponic culture of cash crops. However, the plants can be enriched by heavy metal pollutants in sewage, thereby increasing the content of heavy metals in crops (Rai et al. 2019). Here, the hydroponic systems were assessed based on four segments with the same length, labeled 1, 2, 3, and 4 according to the direction of water flow. The samples were sent to the Agricultural Product Quality and Safety Supervision, Inspection, and Testing Center of the Ministry of Agriculture (Nanjing) for analysis.

At each sampling time point, 10 g sludge was taken out at 1/4, 1/2, and 3/4 points of the HV and stored in 250 mL triangular bottles. They were sealed with perforated rubber stoppers or cotton wool plugs after adding 100 mL nutrient solution containing NH₄⁺-N (0.05 mol/L) for the nitrification rate test or NO₃-N (0.03 mol/L) for the denitrification rate test. Both contents of nutrient solutions for the nitrification and denitrification tests are presented in Table 4.

Nitrification and denitrification intensity were measured in three 1 L reactors (A, B, and C), where the carbon sources of reactors A and B were supernatant and influent, respectively, whereas no carbon source was added to reactor C as a control. Then, the bottles were placed in a water bath at 20 °C for nitrification or denitrification intensity tests. The intensity of nitrification by the sediment was calculated as the change of NO₃⁻-N concentration before and after culture. It must be noted that, in the tests of nitrification and denitrification intensity, the change of NO₃⁻-N concentration, or how much was generated or consumed, was measured by drying the sediment and expressed in per unit of mass per time. The NH₄⁺-N and NO_x⁻-N of each sample were measured three times, and nitrification and denitrification intensity tests were repeated four times.

Vegetation was sampled at three time points to evaluate growth rates and nutrient assimilation rates of the plants. All samples were washed with distilled water to eliminate residual nitrogen in the culture vessel and then incubated with the influent at 30 °C. To prevent NH₄⁺-N nitrification, 5 mg/L thiourea was added in advance.

A total of 30 mL of influent and effluent samples were collected from the system every day for dissolved TN (DTN), NO₃⁻-N, NH₄⁺-N, and dissolved TP (DTP) measurements. The removal efficiencies of NH₄⁺-N, NO₃⁻-N, DTP, and DTN in water were calculated as the change of DTN, NH₄⁺-N, NO₃⁻-N, and DTP concentration in plants while accounting for the change in the fresh weight (FW) of the plants.

To determine the plants' TN and TP contents, each plant was weighed before and after being dried in a vacuum oven at 80 °C for 24 h.

According to standard methods (Iglesias et al. 2000), a representative subsample of all biomass samples was prepared for analysis in the laboratory (washed, cut, dried, and milled). Nitrogen and phosphorus analyses were also performed according to these standard methods. The TN content was analyzed following the Kjeldahl method. In the plant TP analysis, samples were ashed at 450 °C for 2 h and concentrated HCl was added and mixed while heating. After cooling, the mixture was filtered, and phosphorus was determined by the methods described above for water samples.

The removal of TN is the sum of absorption by plants, removal by denitrification, sediment deposition, NH₄⁺-N volatilization, and other means (such as consumption by birds, insects, and/or loss of plant roots). The removal of TP can be similarly calculated as the sum of absorption by plants and sediment deposition, among other means.

Analysis of variance was conducted with IBM SPSS Statistic 25, which was used for a significance test between two or more samples. The significance level (α) was set as 0.05 and p ≤ 0.05 was used to identify significant differences.

It can be seen in Figure 2 that removal efficiency of COD significantly decreased with the increase of hydraulic load from 0.1 to 0.3 m³/(m²·d); the larger the hydraulic load, the shorter the residence time of sewage in each system, which exerts a negative influence on microorganisms for COD. SFCW achieved the highest removal efficiency of COD, with its value reaching 68.2% at a hydraulic load of 0.1 m³/(m²·d). The HV system had the lowest removal efficiency of COD under different hydraulic loads. The removal efficiency of COD was only 15.1% when the hydraulic load was 0.3 m³/(m²·d). This may have been because the effective depth of the HV system without a benthic substrate leads to fewer benthic microbes. With an increase of hydraulic load, COD in HV-SFCW and SFCW-HV decreased from 58.2% to 39.5% and from 50.3% to 34.6%, demonstrating the former performs better. When influent flowed through the SFCW first, COD and dissolved oxygen concentration (0.3–2.1 mg/L) were lower. As a result, hydroponic vegetables did not grow as well, some did not have well-developed roots or roots that had rotted, which further compromises the growth and reproduction of microorganisms in the root zone.

Removal load of COD on SFCW, HV-SFCW, and SFCW-HV increased with higher hydraulic load, as is shown in Figure 2. SFCW has the largest removal load of COD, which indicated that SFCW had the greatest degradation capacity. HV achieved the maximum removal load of 3.2 g/(m²·d) at a hydraulic load of 0.2 m³/(m²·d). The removal load increased and then decreased, with the increase of hydraulic load. This is likely because the root system of the hydroponic vegetable system was shallow, and water erosion is intensified with an increase of hydraulic load, leading to conditions not conducive to the growth and reproduction of microorganisms in the root system.

TN removal efficiencies of the four systems decreased with increasing hydraulic load, as shown in Figure 3. SFCW had the maximum removal efficiencies ranging from 68.9–84.3%. This is because the influent is the effluent from the biological contact oxidation tank, in which the nitrogen mainly exists as NO₃⁻-N (TN was 9.5 mg/L, NO₃⁻-N 9.3 mg/L). The effective depth of SFCW was greater, which created a better hypoxic environment for denitrification, thus the removal efficiency of TN was maximized. TN removal efficiencies in HV, SFCW, and HV-SFCW were similar at a hydraulic load of 0.1 m³/(m²·d), but TN removal efficiency in SFCW-HV was far lower. This was because sewage first flowed through SFCW, which had greater depth, and water dissolved oxygen concentration was low (0.3–2.1 mg/L) when the sewage flowed through HV. This exerted a negative effect on the growth of IA and resulted in less plant absorption and microbial removal. The removal efficiency of TN in the HV system decreased from 79.9% to 45.7% with the increase in hydraulic load, and removal efficiency fluctuating substantially, which indicated hydraulic load had more influence on the removal of TN in HV.

TN removal load of SFCW, HV-SFCW, and HV-SFCW showed an increasing trend with increasing hydraulic load, and SFCW achieved the highest TN removal load, as presented in Figure 3. The removal loads of SFCW, HV- SFCW, and HV- SFCW were 2.8, 2.0, and 2.6 g/(m²·d), respectively, at a hydraulic load of 0.3 m³/(m²·d). The removal load of TN in HV increased first and then decreased. The removal load of TN reached the maximum value of 2.09 g/(m²·d) at a hydraulic load of 0.2 m³/(m²·d), and HV had the maximum TN removal capacity.

The removal efficiency of TP in HV, SFCW, and HV-SFCW decreased with the increase of hydraulic load as exhibited in Figure 4, with SFCW having the highest removal efficiency. The largest decrease in removal efficiency of TP in HV was from 78.9% to 40.8% when the hydraulic load increased from 0.1 m³/(m²·d) to 0.3 m³/(m²·d). This is because the removal of TP in HV relies on the absorption of phosphorus by plants, but the increased hydraulic load shortens residence time weakening overall plant absorption. In contrast, the predominance of phosphorus adsorption by fillers of SFCW and HV-SFCW is prominent with the increase in hydraulic load, such that the TP removal efficiency decreased less than in HV. For SFCW-HV, the sewage first flowed through SFCW, and this resulted in a lower concentration of DO in HV, exerting a negative influence on the growth of plants and bacteria in the root system, and thus the TP removal efficiency remained low. Removal efficiencies of TP in HV, SFCW, HV-SFCW, and SFCW-HV were 84.2, 81.5, 78.9, and 53.7%, respectively, at the hydraulic loading of 0.1 m³/(m²·d). TP concentrations in effluent of SFCW, HV-SFCW, and SFCW-HV reached Grade A Level 1 (GB 18918-2002), whereas TP concentration of influent was 0.84–1.34 mg/L and the hydraulic load 0.1–0.3 m³/(m²·d).

As shown in Figure 4, removal efficiency of TP is similar to TN, in SFCW, HV-SFCW, and SFCW-HV, increasing with a higher hydraulic load. SFCW achieved the best performance for the removal of TP, which was consistent with the observation of strong substrate adsorption in SFCW. The removal load of TP reached a maximum value of 0.24 g/(m²·d) at the hydraulic load of 0.3 m³/(m²·d). The lowest removal load of TP occurred in SFCW-HV at different hydraulic loads, with the removal load of TP only 0.06 g/(m²·d) at a hydraulic load of 0.1 m³/(m²·d). Removal load of TP in HV first increased from 0.07 g/(m²·d) to 0.22 g/(m²·d), and then decreased to 0.19 g/(m²·d) when the hydraulic load increased from 0.1 m³/(m²·d) to 0.3 m³/(m²·d). According to the analysis of COD, TN, and TP systems, the optimal hydraulic loads for removal were different among the four systems. Removal loads of COD, TN, and TP in HV, SFCW, and HV-SFCW were at a maximum when hydraulic loads were 0.2, 0.3, and 0.3 m³/(m²·d), and the effluent COD, TN, and TP all met Grade A Level 1. However, removal efficiencies of pollutants in SFCW-HV were relatively low, even though COD and TN concentration in effluent still reached the standard of Grade A Level 1; TP did not. Moreover, root rot of IA was observed, thus this system was least suitable for the treatment of rural sewage tailwater.

COD changes among sections are presented in Figure 5(a) and 5(b). COD concentration in HV decreased linearly, i.e. degradation of COD was similar in all sections along the pathway. The concentration of COD in the effluent of HV was higher than that in SFCW. COD in SFCW decreased significantly in the first quarter, from 69 mg/L to 50 mg/L, accounting for 48.7% of the total COD removal. In subsequent stages, COD decreased by a similar amount and effluent concentration was low. COD concentration in HV-SFCW decreased from 69 mg/L to 58 mg/L in the first 2/3 segment, accounting for 40.1% of the total removal. COD concentration in the latter segment decreased from 58 mg/L to 42 mg/L, accounting for 59.9% of the total removal in the system. The COD concentration in SFCW-HV decreased significantly in the first 1/3 stage, from 69 mg/L to 55 mg/L, accounting for 54.5% of total removal. The COD concentration in SFCW-HV in the later segment decreased 55 mg/L to 48 mg/L, whereas the COD concentration of effluent was higher.

COD in the effluent of the four systems was 48.0, 30.0, 42.0, and 48.0 mg/L which met the Grade A Level 1 of pollutant discharge standard for urban sewage treatment plants (GB 18918-2002), namely, COD ≤50 mg/L.

Nitrogen concentration in HV and SFCW decreased, with removal in the first quarter relatively high and then the decline slowed. TN removal in the first quarter of HV and SFCW accounted for 44.6% and 54.2% of total removal, and the TN concentration of the effluent reached 3.71 mg/L and 3.20 mg/L, respectively. SFCW had higher TN removal than HV. TN concentration in SFCW-HV decreased significantly in the first 1/3, from 10.4 mg/L to 7.40 mg/L; however, the TN concentration remained almost unchanged in the latter segment, which explained the low overall removal efficiency of TN in SFCW-HV. TN removal efficiency in HV-SFCW was significantly higher, and the effluent TN concentration was only 4.00 mg/L. In the first 2/3 of the HV system, the TN concentration decreased from 10.40 mg/L to 6.71 mg/L, accounting for 58.1% of the total TN removal. The latter part of the HV-SFCW still performed well for TN removal, accounting for 41.9% of the total. The TN concentration of effluent of the systems was 3.27, 3.29, 4.01, and 6.11 mg/L which met the Grade A Level 1 target (GB l8918-2002), namely, TN ≤ 15 mg/L.

Changes of TP in HV and SFCW were similar to that of TN. These two systems had a large decrease in the first quarter, with an approximately linear decrease along the pathway. Proportions of removal in the first quarter of these two systems reached 43.1% and 55.9%, respectively. SFCW performed better because SFCW allowed for high phosphorus adsorption. Trends of TP in HV and SFCW were similar to TN. TP removal efficiency of SFCW-HV was lowest, with 73.3% of TP removal in the first 1/3 and removal in the last 2/3 segment accounting for only 26.7%. TP removal efficiency of HV-SFCW was relatively high, with TP concentration in the HV section decreasing from 0.89 mg/L to 0.63 mg/L, and removal amount by the SFCW in the rear segment accounting for 45.6% of the total TP removal. The TP concentrations of effluent of all four systems were 0.30, 0.20, 0.40, and 0.49 mg/L which met the Grade A Level 1 target (GB l8918-2002), namely, TP ≤ 0.5 mg/L.

The effects of harvesting methods on COD removal are shown in Figure 6. COD concentration of influent fluctuated between 50 mg/L and 70 mg/L, and that of effluent between 40 mg/L and 50 mg/L; average removal efficiencies were ∼25%. Results show that the harvesting method had little effect on the removal of COD, likely because the main mechanism controlling COD is microbial degradation.

As shown in Figure 7, the concentration of TN in the influent water was 8.08–21.43 mg/L, and the average concentration was 15.11 mg/L. Experimental results showed that corridor 1 (with full harvest) had a poor removal effect of TN in the first five days, with an average removal efficiency of only 18.2%. These plants were in early growth stages with poor absorption of nitrogen and phosphorus. After eight days, vegetables with a high growth rate were associated with enhanced absorption of nutrients, and removal efficiency of TN reached nearly 70.0%. All vegetables in corridor 1 were harvested, which reduced the TN purification effect. Therefore, the full-harvest method for vegetables will not lead to the overall stability of water purification.

TN removal efficiencies in corridor 2 (1/2 harvest method) were between 40.1% and 70.4%, and the average removal efficiency was 54.7%. Corridor 3 (with sequential harvest) performed best. Effluent concentrations of TN always remained ∼5 mg/L and TN removal efficiency was ∼70%. Vegetables in corridor 3 went through different growth periods, maximizing water purification capability.

Figure 8 shows the effects of the harvesting method on TP, which showed similar trends with TN. The influent concentration of TP was 0.81–1.16 mg/L, with an average concentration of 0.99 mg/L. In the first five days, TP removal was just 27.2%. After eight days, vegetables were in a more rapid growth period, system effluent TP was <0.5 mg/L, and average removal efficiency reached 52.5%. After 7 August, the effluent concentration of TP in the system began to increase. This is because most vegetables in the corridor had been harvested and the remaining plants were in a later growth stage. The removal efficiency of TP in corridor 2 was higher than that in corridor 1, with an average removal efficiency of 59.2%. TP concentration in outlet water of corridor 3 was relatively stable, ∼0.15–0.35 mg/L and removal efficiency ∼70%. Thus, the removal capacity was highest in corridor 3. This suggests that to maximize TP removal and ensure the stability of the hydroponic system, the sequential harvesting method is optimal.

Nitrification and denitrification, plant assimilation capacity, and sediment precipitation were investigated to determine the nitrogen and phosphorus removal pathways in the HV. It has been reported that HV effectiveness in wastewater treatment relies on the growth potential and ability of macrophytes to develop sufficient root systems for microbial attachment and material transformation.

In this experiment, the effluent from the pulse biofilter served as the influent of the HV to investigate the removal efficiencies and pathways of nutrients. As shown in S.1, NH₄⁺-N removal relied mainly on plant absorption (75.8%), nitrification (18.0%), and ammonia volatilization (6.2%) in the wetland system. NO_x⁻-N removal was mostly attributed to IA absorption (66.1%), followed by denitrification (33.9%). TN removal was mostly via IA absorption (70.3%), then denitrification (26.5%), followed by other mechanisms. The assimilation of nutrients into plant biomass can substantially contribute to nutrient removal efficiency (Brix 1997). Additionally, IA absorption accounted for 86% in the removal of TP, while sediment precipitation only accounted for 5.0%. For the various nutrients, IA biomass assimilation took 75.8%, 66.1%, 70.3%, and 86.0%, respectively, of the removed NH₄⁺-N, NO_x⁻-N, TN, and TP. Thus, absorption and assimilation of nutrients by plants were the primary means by which the aquatic vegetation CW system took up unwanted nutrients from the algae solution.

Monitoring results (Table 3) showed that concentrations of Cr, Pb, Cu, As, and Cd in the plants used in these hydroponics systems did not exceed standards of National Food Safety Standard Limit of Pollutants in Food (GB 2762-2012), namely, Pb ≤0.1, Cd ≤0.05; Hg ≤0.01, As ≤0.5, and Cr ≤0.5 (GB 2762-2012) and could be safely consumed. The details are shown in Table 5.

Hydroponic vegetable and subsurface constructed wetland systems, as well as combined systems, were developed to assess their performance in purifying tailwaters of rural domestic sewage. In this paper, the effects of the hydraulic load on the performance of the four different systems were quantitatively studied. Also, the edible safety of vegetables was examined. The hydroponic vegetable system combined with the constructed wetland was demonstrated to effectively achieve the goal of safe vegetables and efficient treatment of pollutants. The main conclusions are summarized as follows:

• (1)

  The best utilization of nitrogen and phosphorus was achieved when hydraulic loading for HV, SFCW, and HV-SFCW was 0.2, 0.3, and 0.3 m³/(m²·d), respectively. The results showed that SFCW-HV was not suitable for the treatment of sewage tailwater due to the root rot of vegetables in the SFCW-HV system.

• (2)

  In HV and SFCW, the concentrations of COD, TN, and TP decreased in the first section of the systems. In the HV-SFCW system, the removal efficiency of COD, TN, and TP was 39.1%, 61.1%, 55%, respectively, and the contributions of the removal of COD, TN, and TP in the HV section account for 40.7%, 58.1%, and 53.7% of the whole system, respectively.

• (3)

  The HV-SFCW system was most suitable for the treatment of rural sewage tailwaters by considering the sewage purification effects, nitrogen and phosphorus resource utilization, and other characteristics of the systems. In the HV-SFCW system, the content of heavy metals in vegetables was lower than the national standard values, which means harvested water cabbage was safe to eat.

This study was sponsored by the National Key Research and Development Project (2016YFC0400804) and Jiangsu Social Development Project (BE2018630).

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