Abstract

This paper presents the design of a constructed wetland (CW) system in an area with limited land availability, resulting in high hydraulic loads. The CW was constructed to act as a buffering/ polishing step after stabilization ponds for steel industry wastewater post-treatment. A pilot test with two different filter media (50–100mm vs 40–60mm diameter) indicated that a flow rate increase from 49.5 m3/h to 122.4 m3/h would lead to a head loss increase from 2.9cm to 8.7cm, and more than double that for the finer gravel. This was substantially higher than the calculated theoretical values, though the relation with flow rate was similar. Four full scale wetland cells (CW1, CW2, CW3 and CW4) were constructed using the coarser gravel. A design value of total head loss of 1.01m over the total system length, with a design flow of 36,000 m3/day, was expected based on pilot test results. During the first operation year (September 2017 to July 2018), the pond-CW system has received wastewater already meeting required discharge standards. The effluent from the CWs had consistently lower concentrations of all measured variables, and met the predicted values for biochemical oxygen demand (BOD5), total nitrogen (TN) and NH4+-N. Highest removal efficiencies were achieved for NH4+-N (>90%), Mn (>60%) and Fe (45%) with removal efficiencies for TN (14%), BOD5 and chemical oxygen demand (COD) (around 30%). Concentrations of phenol, CN and Cr6+ were below 10, 4 and 3μg/l, respectively, in in- and outflows. An appreciated benefit of the wetland was the ‘green element’ in the industrial landscape.

INTRODUCTION

Wastewater from the steel industry often contains organics, colour, and toxicants such as phenol, cyanides and sulphur compounds. Constructed wetlands (CWs) have been shown to be a cost-effective and environmentally friendly technology for pollutant removal from various types of wastewater, including from the steel industry (Xu et al. 2009; Yang & Hu 2005; Vymazal 2009; Huang et al. 2011). In a laboratory scale vertical flow CW with manganese ore substrate, Xu et al. (2009) reported that about 98% of iron and manganese concentrations in steel wastewater were removed, and attributed this to the presence of iron and manganese oxidation bacteria. The concentration reduction of chemical oxygen demand (COD), turbidity, N-NH4+ and total phosphorus were 55%, 90%, 67% and 93%, respectively. Similar rates were observed in a pilot-scale wetland with gravel mixed with only 4% manganese ore, providing further support for the role of biological oxidation. In another study, the mean reduction rates were 77% for COD and N-NH4+, 96% for turbidity and iron and 92% for manganese in a pilot scale subsurface gravel/manganese ore CW treating wastewater from a steel industry (Huang et al. 2011).

Although several studies of laboratory scale or pilot scale CWs treating steel industry wastewater show promising results, there are few, if any, examples of full scale application of the technology. This paper presents the initial experiences from a hectares scale constructed wetland treating wastewater from a steel industry in Vietnam.

The narrative was initiated by an emergency discharge of poorly treated steel industry wastewater and pipe cleaning solutions that resulted in a marine life disaster in central Vietnam in 2016. This prompted the Vietnamese government to request the Formosa Steel Corporation to improve their wastewater treatment with enhanced physico-chemical and biological processes, and to divert the treated flow (1,500 m3 per hour) through a buffering pond system before discharge to the ocean. In response to this, it was decided to construct a pond and wetland system on an approximately 10 ha area. Besides serving as a polishing and bio-indicator step in the treatment system, the wetland cells should also serve to prevent wash out of algae to the ocean, and create a green element in the factory landscape. It was decided to build a combined pond and wetland system, consisting of two emergency tanks and three stabilization ponds in series followed by subsurface and free water surface CWs.

For subsurface flow wetlands, the substrate composition is an essential component, and the substrate also makes up a substantial part of the total construction costs. The trade-off between achieving a high removal of particles and efficient sorption processes, and the desire to minimize the use of filtering substrate, causes a dilemma in the design phase. The available area at FHS company also posed a challenge with respect to the resulting high hydraulic load on the wetland. To address this, a pilot experiment was set up to select appropriate, locally available media, assess the head loss to be expected depending on choice of media material and to determine the design parameters for the full scale subsurface flow (SSF) wetland cells at FHS. Full scale construction of the wetland cells (CW1, CW2, CW3 and CW4) was initiated in March 2017 and they have been in operation since August 2017. This paper presents the results from the pilot experiments, the design calculations and the first treatment results from the full scale wetland system.

MATERIAL AND METHODS

Pilot experiments

Two parallel pilot CW cells were designed to study the hydraulic conductivity and assess the head loss for two filter media with different grain sizes, aiming at gaining data for design of the full scale wetland. The CW cell dimensions were: length × (width on top; width in bottom) × depth = 31.5 × (4.5; 4) × 1.0m. Each cell was filled with 1m of one of the two filter media to allow for maximum root system depth development (Grace 1989; Davis 2000; Borst et al. 2002). Cell A was filled with 10 cm of coarse sand at the bottom for protection of the HDPE liner; 60 cm of crushed limestone (50–100 mm) as the main filter media; 10 cm of peanut gravel (5–10 mm) for preventing top sand from entering the main media layer; and a top 20 cm of coarse sand for plant establishment (Figure 1). Cell B was filled identically except that the 60 cm crushed local limestone layer had a grain size of 40–60 mm.

Figure 1

Filter media layers in cell A and cell B of the pilot CW at Formosa Steel Company, Vietnam (not to scale).

Figure 1

Filter media layers in cell A and cell B of the pilot CW at Formosa Steel Company, Vietnam (not to scale).

The same hydraulic load (flow rate) was set for both cells; that is, the expected rate for the full scale wetland. A design flow of 36,000 m3/day and a first horizontal flow wetland cell with an inflow width of 74m and a 0.6m effective depth of filter media would result in a horizontal flow velocity of 33.8m/h. Based on this, the flow rate to the pilot CW cells was regulated at various rates between 50 and 120 m3/h, using valves on the discharge pipe of the feeding pumps. In addition, the CW cells outflow, Qout (m3/h), was measured as the accumulated volume of water V (m3) in the outflow well over a certain operation time T (h): Qout = V/T. The observed head losses at different flow rates were compared with theoretical head loss values, calculated using the Carmen-Kozeny formula (Equation (1); Quasim et al. 2000). 
formula
(1)
  • f = friction factor; f = 150*(1-e)/NR + 1.75.

  • NR = Reynolds number; NR = d*ν*ρw/m.

  • ν = kinematic viscosity (=0.028 m2/s).

  • ρw = density of water (=1,000 kg/m3).

  • m = absolute viscosity (=1.518*10−3 N-s/m2).

  • F = particle shape factor (usually 0.85 to 1.0).

  • e = porosity ratio, defined by experiment, measured volumetrically in the laboratory.

  • L = length of filtration module (m).

  • d = media grain diameter, m.

  • v = filtration velocity (m/s).

  • g = acceleration due to gravity (=9.81m/s2).

For comparative purposes, the equation was applied only to the 60 cm layer with limestone, ignoring the influence of the bottom layer with coarse sand, assuming a mean value for d = 0.075m for the A cell and d = 0.05m for the B cell. The measured porosity (e) was 0.45 for the coarse material (cell A) and 0.4 for the material in cell B.

Full scale wetland design

The full scale constructed wetland system consisted of 4 cells in series, with a total wet area of 4.32 ha, preceded by open ponds. The entire system of ponds and wetlands was dimensioned based on literature data; for common variables like biochemical oxygen demand (BOD5), total nitrogen (TN) and NH4+-N, the Tanks-in-Series approach was used (Equation (2); Kadlec & Wallace 2009). 
formula
(2)
where N = number of tanks in series, q = hydraulic load (m/yr), C* = background concentration of the variable concerned, and Cout and Cin are concentrations in the outflow and inflow. CW1, CW2 and CW3 were subsurface horizontal flow CWs, planted with a mixture (multi-culture) of Phragmites australis Cav., Typha angustifolia L. and Cyperus tegetiformis. The final cell, CW4, was a free water surface wetland planted with P. australis Cav. and T. angustifolia L. P. australis (Cav.), T. angustifolia L. and C. tegetiformis were chosen because they are most tolerant to the local environment conditions, and the plants could be found locally. For design purposes, a conservative N-value of 3 was used for the entire system (to represent free water surface wetlands rather than subsurface flow).

Water sampling and analytical methods

From September 2017 to July 2018, water samples were taken weekly in the influent and effluent of each wetland cell (2 samples per site [morning and afternoon] × 5 sampling sites × 40 weeks) and water quality analysed following standard methods (APHA AWWA WEF 2012). pH was measured according to method 9040C, colour was measured by the platinum–cobalt method at 455 nm (method 110.2), COD was measured by the potassium dichromate colorimetric method (method 410.1), BOD5 was measured according to method 405.1, Total suspended solids (TSS) were measured gravimetrically (method 160.2) and Cr6+ was measured by ion chromatography (method 218.7). Mn, Fe, Cd, Hg were measured by atomic absorption spectrometry (method 7000B), TN, NH4+-N and cyanide was measured by semiautomated colorimetry (method 351.2, 350 and 335.4) and phenol was measured by flame ionization detector gas chromatography (method 604). Total grease and fat was measured by hexane extractable gravimetry according to method 1664 (APHA AWWA WEF 2012).

RESULTS AND DISCUSSION

Determining hydraulic design parameters for the full scale wetland cells

The results from the pilot experiments showed significant differences in head loss between cell A and B. In cell B, with smaller grain sizes, the head loss increased from 6.6 cm to 16.3 cm when the flow rate was changed from 49.2 m3/h to 82.7 m3/h. In cell A, a flow rate increase from 49.5 m3/h to 122.4 m3/h caused an increase in head loss from 2.9 cm to 8.7 cm. The measured head losses exceeded the theoretical calculations using the Carmen–Kozeny formula but the relation with flow rate was similar (Figure 2), suggesting that the experiments gave reliable results regarding this aspect. Several assumptions in theoretical models of flows through packed beds can often not be met in real soil filtering systems, which has implications for the system design, as discussed by, for example, Schöpke (2007). One important reason that the measured head loss was higher than the theoretical in the current study was most probably that the filter media was not washed (i.e. contained finer particles), and was therefore neither homogeneous nor consolidated. This was important to consider when selecting and filling the media into the wetland cells. If a well sieved, uniform filter media cannot be used, a larger ‘safety factor’ for head loss increase must be accounted for in the design. In this case, the pilot experiments results indicated that the unit head loss in cell A was 22.8 cm/100m. Translating this to a full scale wetland with a total flow length of 450m, the head loss over the system could reach 1.01m at the maximum design flow of 36,000 m3/day. This was an important factor when determining the elevation of the four full scale wetland cells.

Figure 2

Head loss versus flow velocity values, experimental and theoretical values in pilot wetland cell A and cell B at the Formosa Steel Company, Vietnam.

Figure 2

Head loss versus flow velocity values, experimental and theoretical values in pilot wetland cell A and cell B at the Formosa Steel Company, Vietnam.

In cell B, the unit head loss was 51.8 cm/100m at a flow rate as low as 82.7 m3/h, translating to 2.33m over a 450m long wetland system. Hence, with the high hydraulic design load for the FHS CW system, this filter media (diameter 40–60 mm) was deemed unsuitable. A too high head loss would lead to the need for a significantly increased elevation of the first ponds with substantially more earth work and an increase in the pump head. An alternative solution could be to decrease the flow velocity through the wetland cross sections, but this could only be done through an increase in the media depth or wetland cell area. Besides the possibly negative effect on plant growth and the biogeochemical conditions in the cells (more anaerobic conditions), this would increase the total media volume and/or required wetland area and lead to an increase in project costs.

Treatment performance of the wetland cells

During the first year of operation, the effluent from the wastewater treatment system preceding the wetland system had a neutral pH, with little variation, and concentrations of potentially toxic compounds were, for several substances below the detection limit (Table 1). This effluent quality met the Vietnamese standards for industrial wastewater discharge to both inland waters (class A) and the sea (class B). Hence, the wetland system functioned as a polishing step during this monitoring period. As predicted, the hydraulic load was high, resulting in relatively high load of substances despite the low inflow concentrations. Concentrations were further reduced in the wetland system leading to very low levels of critical variables, such as NH4+-N, Fe and Mn, in the outflow from the system (Table 2). Concentrations of phenol, CN and Cr6+ were below the respective detection limits (10, 4 and 3 μg/l) also in the outflow from the system (values not shown).

Table 1

Quality of the wastewater entering the CWs at the Formosa Steel plant, Vietnam, and concentration limits in the Vietnamese standards for industrial wastewater discharges to different recipients (class A and B; QCVN 40:2011/BTNMT)

Variable measuredUnitMean value ± SD (min-max)Standard for industrial wastewater, class AStandard for industrial wastewater, class B
pH – 7.3 ± 0.18 (6.8–7.8) 6–9 5.5–9 
Color Pt/Co 27.8 ± 4.9 (20–39) 50 150 
COD mg/l 23.1 ± 2.6 (14–30) 75 150 
BOD5 mg/l 9.4 ± 2.3 (5 –15.8) 30 50 
TSS mg/l 17.4 ± 2.1 (14–26) 50 100 
Mn mg/l 0.43 ± 0.12 (0.08 –0.75) 0.5 
Fe mg/l 0.33 ± 0.18 (0.01 –0.86) 
TN mg/l 10.5 ± 2.8 (6.1 –19.2) 20 40 
NH4+-N mg/l 0.57 ± 0.19 (0.04 –2.1) 10 
Cr (VI) mg/l <0.003 0.05 0.1 
Cd mg/l <0.0007 0.05 0.1 
Hg mg/l <0.0003 0.005 0.01 
Phenol mg/l <0.01 0.1 0.5 
CN mg/l <0.004 0.07 0.1 
Grease and fats mg/l <0.3 10 
Variable measuredUnitMean value ± SD (min-max)Standard for industrial wastewater, class AStandard for industrial wastewater, class B
pH – 7.3 ± 0.18 (6.8–7.8) 6–9 5.5–9 
Color Pt/Co 27.8 ± 4.9 (20–39) 50 150 
COD mg/l 23.1 ± 2.6 (14–30) 75 150 
BOD5 mg/l 9.4 ± 2.3 (5 –15.8) 30 50 
TSS mg/l 17.4 ± 2.1 (14–26) 50 100 
Mn mg/l 0.43 ± 0.12 (0.08 –0.75) 0.5 
Fe mg/l 0.33 ± 0.18 (0.01 –0.86) 
TN mg/l 10.5 ± 2.8 (6.1 –19.2) 20 40 
NH4+-N mg/l 0.57 ± 0.19 (0.04 –2.1) 10 
Cr (VI) mg/l <0.003 0.05 0.1 
Cd mg/l <0.0007 0.05 0.1 
Hg mg/l <0.0003 0.005 0.01 
Phenol mg/l <0.01 0.1 0.5 
CN mg/l <0.004 0.07 0.1 
Grease and fats mg/l <0.3 10 

Arithmetic means ± SD for the period September 2017–July 2018.

Table 2

Mean (n = 80 per sampling site) water quality of the effluent from wetland cells CW1, 2, 3, 4 at Formosa Steel work, Vietnam, for the period September 2017 to July 2018. Arithmetic mean ± SD (min-max) are shown and units are as in Table 1 

Variables measuredWetland cells
CW1CW2CW3CW4
Color 20.6 ± 4.5 (18–30) 19.6 ± 2.8 (17–28) 17.9 ± 2.9 (14–25) 16.7 ± 2.6 (12–21) 
COD 20.2 ± 2.9 (15–29) 19.9 ± 2.0 (14–25) 17.4 ± 3.0 (11–23) 16.7 ± 3.3 (10–24) 
BOD5 7.7 ± 1.6 (6.0–11) 7.8 ± 1.1 (5.6–10.1) 6.7 ± 2.0 (4.9–9.1) 6.5 ± 0.9 (4.2–9.3) 
TSS 15 ± 1.5 (10–22) 14.4 ± 2.5 (8–20) 13.0 ± 2.5 (5–17) 13.4 ± 2.4 (7–18) 
Mn 0.27 ± 0.06 (0.22–0.61) 0.22 ± 0.05 (0.1–0.32) 0.17 ± 0.06 (0.03–0.41) 0.16 ± 0.08 (0.03–0.55) 
Fe 0.22 ± 0.03 (0.06–0.49) 0.25 ± 0.16 (0.04–0.56) 0.18 ± 0.09 (0.03–0.71) 0.17 ± 0.10 (0.02–0.7) 
TN 9.2 ± 1.5 (9.2–17) 9.5 ± 2.4 (8.3–16) 9.2 ± 2.1 (6.5–14.9) 8.9 ± 2.0 (6.1–15.8) 
NH4+-N 0.29 ± 0.08 (0.13–0.67) 0.15 ± 0.03 (0.09–0.42) 0.08 ± 0.02 (0.02–0.69) 0.04 ± 0.01 (0.01–0.14) 
Variables measuredWetland cells
CW1CW2CW3CW4
Color 20.6 ± 4.5 (18–30) 19.6 ± 2.8 (17–28) 17.9 ± 2.9 (14–25) 16.7 ± 2.6 (12–21) 
COD 20.2 ± 2.9 (15–29) 19.9 ± 2.0 (14–25) 17.4 ± 3.0 (11–23) 16.7 ± 3.3 (10–24) 
BOD5 7.7 ± 1.6 (6.0–11) 7.8 ± 1.1 (5.6–10.1) 6.7 ± 2.0 (4.9–9.1) 6.5 ± 0.9 (4.2–9.3) 
TSS 15 ± 1.5 (10–22) 14.4 ± 2.5 (8–20) 13.0 ± 2.5 (5–17) 13.4 ± 2.4 (7–18) 
Mn 0.27 ± 0.06 (0.22–0.61) 0.22 ± 0.05 (0.1–0.32) 0.17 ± 0.06 (0.03–0.41) 0.16 ± 0.08 (0.03–0.55) 
Fe 0.22 ± 0.03 (0.06–0.49) 0.25 ± 0.16 (0.04–0.56) 0.18 ± 0.09 (0.03–0.71) 0.17 ± 0.10 (0.02–0.7) 
TN 9.2 ± 1.5 (9.2–17) 9.5 ± 2.4 (8.3–16) 9.2 ± 2.1 (6.5–14.9) 8.9 ± 2.0 (6.1–15.8) 
NH4+-N 0.29 ± 0.08 (0.13–0.67) 0.15 ± 0.03 (0.09–0.42) 0.08 ± 0.02 (0.02–0.69) 0.04 ± 0.01 (0.01–0.14) 

The concentration reductions achieved agreed relatively well with the design model (Equation (2)) for BOD5 and TN. There was a tendency to improved performance in the second period, when the observed concentrations of both BOD5 and TN were lower than the predicted (Table 3). No effort was made to account for the possible influence of the approximately 20% lower hydraulic load on the removal rate constant in period 2, though a positive relation between hydraulic load and k-values has been observed when applying first order rate models to monitoring data (Kadlec 2000). Regarding NH4+-N, the wetland system performed better than predicted with the model (Table 3), and this is probably a result of the initial influence of plant uptake. The k20-value used in the design model was conservatively set to the lower 0.2 percentile of the indicated range for lightly loaded ‘agronomic’ wetlands, i.e. wetlands with an annual NH4-N load <0.33 g/(m2*d) (Kadlec & Wallace 2009) similar to the load to the FHS wetland. If instead setting the k20-value to 0.55m/d (above the 0.8 percentile), the predicted outflow concentrations would have been 0.23 and 0.1 mg/l for period 1 and 2, respectively; this is still higher than the observed values. As the plants were in the establishment phase, they would have competed with nitrifying and heterotrophic bacteria for available nitrogen, resulting in the observed high NH4+-N removal efficiency (>90%; Figure 3). Those results corroborated well with previous studies on NH4+-N treatment efficiency from various industrial wastewaters: 84–94% (Vymazal 2014) and were higher than data reported by Xu et al. (2009) and Huang et al. (2011) studying wetlands treating steel wastewater.

Table 3

Hydraulic load, observed inflow and outflow concentrations and outflow concentrations modelled using the Tanks-in-Series model (P = 3, Θ = 1.08 for temp adjustment of k20)

VariableTime periodHydraulic load m/dInflow conc. mg/lLoad g/(m2*d)Remov. rate, k m/dC* mg/lPredicted outflow conc mg/lObserved outflow conc mg/l
BOD5 Sept-Dec-17 0.58 7.5 4.3 0.235a 1a 5.4 6.6 
Feb-Jul-18 0.47 11.4 5.3 0.235a 1a 7.5 6.4 
TN Sept-Dec-17 0.58 10.5 6.1 0.089b 9.3 9.4 
Feb-Jul-18 0.47 10.4 4.9 0.089b 9.0 8.5 
NH4+-N Sept-Dec-17 0.58 0.72 0.42 0.056b 0.66 0.06 
Feb-Jul-18 0.47 0.40 0.18 0.056b 0.35 0.03 
VariableTime periodHydraulic load m/dInflow conc. mg/lLoad g/(m2*d)Remov. rate, k m/dC* mg/lPredicted outflow conc mg/lObserved outflow conc mg/l
BOD5 Sept-Dec-17 0.58 7.5 4.3 0.235a 1a 5.4 6.6 
Feb-Jul-18 0.47 11.4 5.3 0.235a 1a 7.5 6.4 
TN Sept-Dec-17 0.58 10.5 6.1 0.089b 9.3 9.4 
Feb-Jul-18 0.47 10.4 4.9 0.089b 9.0 8.5 
NH4+-N Sept-Dec-17 0.58 0.72 0.42 0.056b 0.66 0.06 
Feb-Jul-18 0.47 0.40 0.18 0.056b 0.35 0.03 

aFor inflow concentrations 3–30 mg/l, 50% percentile, no temp adjustment of k (Kadlec & Wallace 2009).

bAdjusted k-value for 25 °C.

Figure 3

Removal efficiencies (concentration reductions) of constructed wetland cells CW 1–4 at the Formosa Steel plant, Vietnam, during September 2017–July 2018.

Figure 3

Removal efficiencies (concentration reductions) of constructed wetland cells CW 1–4 at the Formosa Steel plant, Vietnam, during September 2017–July 2018.

In agreement with the model predictions, the reduction of TN was the lowest of all the variables (mean 14%). It is well known that the % removal is depending on the inflow concentrations as a first order rate removal model often describes the concentration reductions in CWs (Kadlec & Wallace 2009). Hence, a higher removal could have been expected in our study since the inflow contained around 10 mg/l TN, but the hydraulic load counteracted this as also confirmed by the predictions from the design model. Although nitrification-denitrification processes usually are the main nitrogen removal mechanisms in CWs (accounting for 60–96% of TN removal) (Chen et al. 2014), other authors have showed that macrophyte uptake can account for a considerable TN reduction (14–52%) in wetlands receiving low loads of N (Wu et al. 2013). Uptake in P. australis and T. orientalis accounted for about 27 and 40% of TN removal in that study. In a systematic review of CWs receiving either non-point source water or secondary or tertiary treated wastewater, Land et al. (2016) found that the average removal efficiency was 39%. However, in a study by Bulc et al. (2006), TN removal efficiency in CWs treating textile wastewater was also only 5%, suggesting that the nitrogen was bound in recalcitrant compounds. This may also be the case in the FSH CW and, in that case, the plant uptake would play little role in the efficiency, though it was likely a significant process for the NH4+-N removal. However, when comparing the predicted outflow concentrations with those observed, it appeared that the TN removal improved as the plant and microbial community developed over time. Future monitoring data will shed more light on the N removal in this CW system.

Regarding the variables BOD5 and COD, the values in the inflow were close to what sometimes is considered background values for wastewater treatment wetlands receiving raw wastewater. However, for tertiary wastewater, Kadlec & Wallace (2009) suggested a C* (background) value of 1 mg/l in horizontal subsurface flow wetlands. The results from the first months of the FSH wetland system suggest that the wetland system was performing relatively well in view of the high hydraulic load (Table 3 and Figure 3). Earlier findings on COD removal rates by CWs treating steel wastewaters ranged from 31 to 77% (Xu et al. 2009; Huang et al. 2011). Baughman et al. (2003) investigated the performance of CWs treating textile wastewater and found a COD removal efficiency ranging from 20 to 34%. The mean BOD5/COD ratio in the inflow to the FHS wetland system was 0.41, suggesting that the organic matter is not easily biodegradable. This could mean that a longer retention time than 1 day would be required for a substantial decomposition, or alternatively a finer media to allow for more filtration and sorption processes (Bulc et al. 2006). As discussed in the design section above, this was not a viable option for the FHS wetland system. These results correlated well with COD reduction rates recorded in other studies of steel wastewater and, therefore, it is plausible that the BOD5 removal efficiencies may remain at the observed level.

Based on concentration reductions, the highest removal efficiencies were achieved for Mn (>60%) and Fe (>40%) (Figure 3). The observed Mn removal rate was consistent with the average Mn removal rate by wetland treatment systems previously reported by Lesley et al. (2008) (64.5%) despite the higher inflow concentrations in that study – 1.5 mg/l versus 0.43 mg/l. On the other hand, in other studies, Xu et al. (2009) and Huang et al. (2011) found that treatment efficiencies of Mn by CWs treating wastewater from steel enterprises reached 92.5–95% when influent concentrations were in the range of 0.53–2.23 mg/l. However, in those studies, the hydraulic loads have been lower with 1.5–2 days retention time compared to <1 day in the present study. Regarding Fe removal, the efficiency was 1.5 times lower than the results recorded by Lesley et al. (2008), 48% as opposed to 70%. This could be due to the 10 times lower Fe concentrations in the influent in the FHS wetland system. Sedimentation, filtration and adsorption by surfaces could be important physical processes, which helps to decrease metal concentrations in CWs (Garcia et al. 2010). Metal remobilization can be favoured by oxidation in the rhizosphere due to oxygen released from roots of wetland plants (Garcia et al. 2010). The formation of iron plaque on the root surface of P. australis has been previously found in full scale HSSF CWs (Mantovi et al. 2003). The significance of metal uptake by plants is still a matter of debate. While several authors concluded that wetland plants play a significant role in metal uptake (Khan et al. 2009; Maine et al. 2009), other authors reported that plant uptake is negligible for total metal removal (Stottmeister et al. 2003). Some of that discrepancy can be related to load differences; plants have a limited uptake capacity and this will account for a larger proportion of the metal inflow in low loaded wetlands.

Colour removal is often an issue in treatment of steel works wastewater. There are studies showing that CWs can be effective in decolouration of some types of wastewater (mean removal of up to 90%) (Moshiri 1993). This depends strongly on the cause of the colour, and Bulc & Ojstršek (2008) found that CWs planted with P. australis removed 70 to 90% of the colour in textile wastewater, which was substantially higher than in the FHS wetland system (40%). Probably the short retention time and the relative recalcitrant organic compounds counteracted an efficient colour removal.

CWs are well known to remove TSS (Manios et al. 2003; Bulc & Ojstršek 2008), also from difficult wastewater such as textile wastewater (Bulc & Ojstršek 2008). In this study, removal efficiencies of TSS were only 23%, and this was equivalent to 13.35 mg/l in the effluent. This is lower than previous results on TSS removal in CWs treating petrochemical and refinery wastewaters (42–44%) (Vymazal 2014). Kadlec & Wallace (2009) summarized a large number of studies on gravel wetlands treating domestic wastewater, and the results suggest that a concentration around 5 mg/l TSS is achievable. It is possible that the TSS removal in the FHS wetland system will improve as the plant root system develops and covers a larger part of the filter media. This could improve the filtration effect and support the development of a viable microbial community. Filtration, followed by aerobic microbial degradation on the surface or anaerobic degradation within the media, could affect the effectiveness of organic solids removal (Manios et al. 2003).

CONCLUSIONS

Experiments in pilot scale horizontal subsurface flow wetlands receiving treated wastewater from a steel industry showed that theoretically calculated values of head loss at different hydraulic loading rates can substantially underestimate the observed values. Hence, field testing is warranted when designing large scale wetlands for high hydraulic loads. The study showed that filter media of 50–100 mm diameter had an observed unit head loss of 22.8 cm/100m at a flow rate 122 m3/h, equal to a total head loss of 1.01m over the total length of the planned full scale wetland system, with a design flow rate of 36,000 m3/day. The discrepancy from the theoretical values was probably explained by the heterogeneity of the gravel, which was neither sieved nor washed. In case media with such non-uniform size ranges must be used, a safety factor for head loss increase should be included in the design.

During the first operational year, the four full scale CW cells, with a mean retention time of 0.9 days, received a wastewater that met the discharge standards. The inflow concentrations were further reduced in the CW system, and in the second half of the observation period they were lower than predicted from the design model for BOD5 and TN, suggesting that the system improved as the plant and microbial communities developed. Highest removal efficiencies were observed for NH4+-N (>90%), Mn (>60%), and Fe (48%). The removal of TN was lower at 15%, for BOD5 31% and for COD 28%, corresponding to 8.9, 6,5 and 16.7 mg/l, respectively, in the effluent. Those relatively low removal rates are explained by the high hydraulic load and short retention time. Despite this, the removal of colour amounted to 40%, whereas TSS was removed by 23%. The results confirm that a hybrid pond and CW system can serve as a polishing step for wastewater from the steel industry, reducing the risk for discharge of potentially toxic substances to the environment. A further benefit is the ‘green element’ created in the industrial landscape (Figure 4).

Figure 4

Hybrid pond – wetland system for Formosa Steel plant wastewater polishing treatment, Vietnam (design flow 36,000 m3/day, total area 10 ha). (1) Waste stabilization pond cells; (2) subsurface flow constructed wetland cell CW1; (3) subsurface flow constructed wetland cell CW2; (4) subsurface flow constructed wetland cell CW3; (5) free water surface flow constructed wetland cell CW4.

Figure 4

Hybrid pond – wetland system for Formosa Steel plant wastewater polishing treatment, Vietnam (design flow 36,000 m3/day, total area 10 ha). (1) Waste stabilization pond cells; (2) subsurface flow constructed wetland cell CW1; (3) subsurface flow constructed wetland cell CW2; (4) subsurface flow constructed wetland cell CW3; (5) free water surface flow constructed wetland cell CW4.

REFERENCES

REFERENCES
APHA, AWWA, WEF
2012
Standard Methods for the Examination of Water and Wastewater
, 22nd edn.
American Public Health Association
,
Washington
, p.
1360
.
Baughman
G. L.
,
Perkins
W. S.
,
Lasier
P. J.
&
Winger
P. V.
2003
Effect of treatment in a constructed wetland on toxicity of textile wastewater
.
AATCC Review
3
,
28
30
.
Borst
M.
,
Riscassi
A. L.
,
Estime
L.
&
Fassman
E. L.
2002
Free-water depth as a management tool for constructed wetlands
.
Journal of Aquatic Plant Management
40
,
43
45
.
Bulc
T. G.
&
Ojstršek
A.
2008
The use of constructed wetland for dye-rich textile wastewater treatment
.
Journal of Hazardous Materials
155
,
76
82
.
Bulc
T. G.
,
Ojstrsek
A.
&
Vrhovšek
D.
2006
The use of constructed wetland for textile wastewater treatment
. In:
Proc. 10th Internat. Conf. Wetland Systems for Water Pollution Control
.
MAOTDR
,
Lisbon
,
Portugal
, pp.
1667
1675
.
Davis
L.
2000
.
USDA-Natural Resources Conservation Service and the US Environmental Protection Agency-Region III
,
Harrisburg, PA
.
Garcia
J.
,
Rousseau
D. P. L.
,
Morató
J.
,
Lesage
E. L. S.
,
Matamoros
V.
&
Bayona
J. M.
2010
Contaminant removal processes in subsurface-flow constructed wetlands: a review
.
Critical Reviews in Environmental Science and Technology
40
,
561
661
.
Kadlec
R. H.
2000
The inadequacy of first-order treatment models
.
Ecological Engineering
15
,
105
119
.
Kadlec
R. H.
&
Wallace
S.
2009
Treatment Wetlands
, 2nd edn.
CRC Press, Taylor and Francis Group
,
Boca Raton, FL
.
Khan
S.
,
Ahmad
I.
,
Shah
M. T.
,
Rehman
S.
&
Khaliq
A.
2009
Use of constructed wetland for the removal of heavy metals from industrial wastewater
.
Journal of Environmental Management
90
,
3451
3457
.
Land
M.
,
Granéli
W.
,
Grimvall
A.
,
Hoffmann
C. C.
,
Mitsch
W. J.
&
Tonderski
K. S.
2016
How effective are created or restored freshwater wetlands for nitrogen and phosphorus removal? A systematic review
.
Environmental Evidence
5
,
5
9
.
Maine
M. A.
,
Suñe
N.
,
Hadad
H.
,
Sánchez
G.
&
Bonetto
C.
2009
Influence of vegetation on the removal of heavy metals and nutrients in a constructed wetland
.
Journal of Environmental Management
90
,
355
363
.
Manios
T.
,
Stentiford
E. I.
&
Millner
P.
2003
Removal of total suspended solids from wastewater in constructed horizontal flow subsurface wetlands
.
Journal of Environmental Science and Health, Part A: Toxic/Hazardous Substances and Environmental Engineering
38
,
1073
1085
.
Mantovi
P.
,
Marmiroli
M.
,
Maestri
E.
,
Tagliavini
S.
,
Piccinini
S.
&
Marmiroli
N.
2003
Application of a horizontal subsurface flow constructed wetland on treatment of dairy parlor wastewater
.
Bioresource Technology
88
,
85
94
.
Moshiri
G. A.
1993
.
Taylor & Francis
,
Boca Raton, FL
.
Quasim
S.
,
Motley
E.
&
Zhu
G.
2000
.
Prentice Hall
,
Upper Saddle River, NJ
, pp.
373
400
.
Schöpke
R.
2007
Multicomponent systems in water treatment – Data compilation and use for dimensioning treatment plants. Heft 14, Schriftenreihe Siedlungswasserwirtschaft und Umwelt. Dep Water Technology and Urban Water Engineering, Brandenburg Technical University, Cottbus, Germany (in German)
.
Stottmeister
U.
,
Wießner
A.
,
Kuschk
P.
,
Kappelmeyer
U.
,
Kästner
M.
&
Bederski
O.
2003
Effects of plants and microorganisms in constructed wetlands for wastewater treatment
.
Biotechnology Advances
22
,
93
117
.
Wu
H.
,
Zhang
J.
,
Wei
R.
,
Liang
S.
,
Li
C.
&
Xie
H.
2013
Nitrogen transformations and balance in constructed wetlands for slightly polluted river water treatment using different macrophytes
.
Environmental Science and Pollution Research
20
,
443
451
.
Xu
J. C.
,
Chen
G.
,
Huang
X. F.
,
Li
G. M.
,
Liu
J.
&
Yang
N.
2009
Iron and manganese removal by using manganese ore constructed wetlands in the reclamation of steel wastewater
.
Journal of Hazardous Materials
169
,
309
317
.