Impact of different feeding strategies and plant presence on the performance of shallow horizontal subsurface-flow constructed wetlands

Science of the Total Environment 378 (2007) 253 – 262 www.elsevier.com/locate/scitotenv

Impact of different feeding strategies and plant presence on the performance of shallow horizontal subsurface-flow constructed wetlands Aracelly Caselles-Osorio a,b , Joan García a,⁎ a

Environmental Engineering Division, Department of Hydraulic, Maritime and Environmental Engineering, Technical University of Catalonia, c/ Jordi Girona 1-3, Mòdul D-1, 08034 Barcelona, Spain b Department of Biology, Atlantic University, km 7 Highway Old Colombia Port, Barranquilla, Colombia Received 5 October 2006; received in revised form 5 February 2007; accepted 25 February 2007 Available online 11 April 2007

Abstract The aim of this investigation was to evaluate the effect of continuous and intermittent feeding strategies on contaminant removal efficiency of shallow horizontal subsurface-flow constructed wetlands (SSF CWs). Also it was tested the effect of the presence of plant aboveground biomass on removal efficiency. Two experimental wetlands planted with common reed were subjected to a three-phase, 10-month experiment involving a common source of settled urban wastewater with a hydraulic loading rate of 26 mm/d during the first and second phases and 39 mm/d during the third. In the first and second phases one of the wetlands was fed continuously while the other was fed intermittently. In the third phase both systems were operated intermittently, but in one the macrophyte aboveground biomass was cut in order to study the effect of plant aboveground biomass on the removal efficiency. The intermittently fed system presented systematically more oxidised environmental conditions and higher ammonium removal efficiencies (on average 80 and 99% for the first and the second phases respectively) compared with the continuously fed system (71 and 85%). The mass amount of ammonium removed ranged from 0.58 to 0.67 g N/m2 d for the intermittently fed system and from 0.52 to 0.58 g N/m2 d for the continuously fed system. Sulphate removal was higher in the continuously fed system (on average 76 and 79% for the first and second phases respectively) compared with the intermittently fed system (51 and 58%). In the third phase the wetland that operated with aboveground biomass exhibited more oxidised environmental conditions and better removal efficiencies (on average 81% for COD and 98% for ammonium) than the wetland operated without aboveground biomass (73% for COD and 72% for ammonium). The results of this study indicate that the intermittent feeding strategy improved the removal of ammonium and the presence of aboveground biomass enhanced the removal of COD and ammonium. © 2007 Elsevier B.V. All rights reserved. Keywords: Redox potential; Oxygen; Sulphur cycle; Sulphate reduction; Nitrification; Batch; Reed beds

1. Introduction ⁎ Corresponding author. Tel.: +34 93 4016464; fax: +34 93 4017357. E-mail addresses: [email protected] (A. Caselles-Osorio), [email protected] (J. García). 0048-9697/$ - see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.scitotenv.2007.02.031

The contaminant removal efficiencies attained in horizontal subsurface-flow constructed wetlands (SSF CWs) depend on the oxidation–reduction conditions and gradients prevailing within the systems (Kadlec et al.,

254

A. Caselles-Osorio, J. García / Science of the Total Environment 378 (2007) 253–262

2000). This trend has been clearly demonstrated for contaminants such as ammonium, the removal of which requires the establishment of oxygen-enriched zones (García et al., 2005; Wiessner et al., 2005a). The redox state present in SSF CWs is influenced by a variety of factors including organic load, mode of operation (batch, continuous or intermittent), type and development of macrophytes, and water depth (Kadlec and Knight, 1996; García et al., 2004). Several studies have been devoted to evaluate the effect of the mode of operation on redox conditions and the removal efficiency of SSF CWs. In general, batch operation promotes more oxidised conditions and therefore better performance than continuous operation. Stein et al. (2003) observed an ammonium removal efficiency of 57% in batch-operated experimental systems, in comparison with 42% in continuous systems. In fact, Tanner et al. (1999) reported almost complete removal of ammonium in experimental wetlands operated in batch mode with small water level fluctuations. On the other hand, Vymazal and Masa (2003) found that changes in the water level in full-scale SSF CWs, with variations of between 8 and 15 cm, had a positive effect on the elimination of several pollutants, including COD and ammonium. Strong redox gradients at the microscale within the SSF CWs have been linked to the presence of macrophytes. The measurements of Bezbaruah and Zhang (2004) using microelectrodes in experimental wetlands showed that the redox potential at the surface of lateral roots of Scirpus validus was higher than that observed in the bulk water. The increased redox potential near the surface of the roots was related to the presence of oxygen released by the plants. Despite the evidence of oxygen release by the macrophytes in SFF CWs, what is less clear is the net contribution of this oxygen to contaminant removal. Tanner (2001) reviewed several studies in which planted and unplanted wetlands were compared; he concluded that macrophytes only marginally increase the rate of elimination of organic matter but clearly increase the rate of removal of ammonium. Recent investigations have clearly demonstrated that water depth affects the redox conditions and the removal efficiency of SSF CWs. García et al. (2005) found that pilot SSF CWs with a mean water depth of 0.27 m exhibited higher redox potential values than systems with a water depth of 0.5 m. In addition, the shallower wetlands were more efficient for removing COD and ammonium. Headley et al. (2005) observed that doubling the water depth of SSF CWs resulted in no improvement of BOD5 removal and a decline in total nitrogen removal. In a previous study conducted in our

laboratory, in which shallow experimental SSF CWs were fed intermittently (to avoid solid sedimentation and adsorption onto the walls of influent tanks and pipes), it was observed that the removal efficiencies for COD and ammonium were quite high, averaging 80 and 90% respectively (Caselles-Osorio and García, 2007). In this study mass surface removal rates up to 17.5 g/m2 d for COD and 1.3 g N/m2 d for ammonium were attained. While the results were encouraging, particularly for ammonium, it was not clear whether the intermittent feeding strategy had a positive effect on removal efficiency. The objectives of the present investigation were to evaluate 1) the influence of the hydraulic regime (continuous or intermittent) on urban wastewater contaminant removal efficiency, and 2) the effect of the presence of plant aboveground biomass on removal efficiency. To our knowledge the differences in continuous versus intermittent feeding had not been explored in shallow SSF CWs prior to this study. 2. Materials and methods The two experimental SSF CWs used in this study (named A and B) consisted of plastic containers (1.1 m long, 0.7 m wide and 0.38 m high) filled with gravel extracted from a pilot SSF CW system located in Les Franqueses del Vallès, Barcelona, Spain. A detailed description of the pilot system can be found elsewhere (García et al., 2004, 2005). Each container had a drainage pipe located on the flat bottom for effluent discharge (Fig. 1). The gravel layer (D60 = 3.5 mm, Cu = 1.7 and porosity of 40%) was 0.3 m deep and the water level was kept 0.05 m below the gravel surface to give a water depth of 0.25 m. One vertical, perforated tube was inserted into the gravel near the inlet zone of each wetland system to enable measurements of various physical and chemical parameters. This tube was made of metal mesh (perforated along its entire length) and was installed at the bottom of the SSF CWs. The wetlands were planted in June 2005 with developed rhizomes of common reed (Phragmites australis) and placed on the roof of the building of the Department of Hydraulic, Maritime and Environmental Engineering (Technical University of Catalonia, Barcelona, Spain). By September 2005 the plants were well established and covered the entire surface of the wetlands. Experiments started in October 2005. The two wetlands were fed with settled urban wastewater, which was obtained on a daily basis from the municipal sewer located near the Department building. Experiments were conducted during a period of 10 months and included three phases of operation, in which the hydraulic regime (intermittent or continuous feeding) and

A. Caselles-Osorio, J. García / Science of the Total Environment 378 (2007) 253–262

255

Fig. 1. Schematic representation of the study setup. Note that both wetlands received the same settled urban wastewater but with a different feeding strategy in Phases I and II. In Phase III both wetlands were fed with the same wastewater and intermittently.

the presence of macrophyte aboveground biomass were tested (Table 1). Note that each phase had a sufficient duration to guarantee a representative number of samples, that ranged between 29 and 45 depending on the phase. Intermittent feeding was carried out on a daily basis by pouring the corresponding amount of settled fresh wastewater into the inlet zone over a period of 20 min. Continuous feeding was achieved by means of a peristaltic pump that conveyed the settled fresh wastewater from a storage tank to the wetland. The effect of the hydraulic regime was studied in Phases I and II, in which a flow of 20 L/d was used to give a nominal hydraulic retention time (HRT) of 3.3 days and a hydraulic loading rate (HLR) of 26 mm/d. Note that Phases I and II only differ in the fact that the hydraulic regimes were reversed between the wetlands. Thus, both wetlands were operated with intermittent and continuous feeding depending on the phase. The scientific objective of crossing the feeding mode between the two wetlands was to confirm that the

findings were not wetland-specific and really related to the mode of operation. The effect of the presence or absence of macrophyte aboveground biomass was studied in Phase III, in which a flow of 30 L/d (intermittently supplied in both SSF CWs) was used to give an HRT of 2.1 days and an HLR of 39 mm/d. Note that in this phase it was necessary to increase the flow rate (with respect to the previous phases) to obtain representative effluent volumes during sampling, as the rate of evapotranspiration was very high at the time of these experiments. Prior to start this phase, both experimental units were operated intermittently for a period of two weeks to ensure the same initial conditions. When the experiments started both wetlands produced effluents with a very similar quality (the ammonium effluent concentration ranged between 1 and 2 mg N/L in the two systems). At this point, the aboveground biomass of the wetland B was cut to near the level of the gravel, and the short stems remaining were covered with a synthetic

Table 1 Hydraulic regimes (continuous (Cont) or intermittent (Inter)), presence/absence of macrophyte aboveground (Ab) biomass and mean surface loading rates (COD and ammonium) during the three experimental phases in the two wetlands Wetland

A B

Phase I (Oct–Dec 2005)

Phase II (Jan–Apr 2006)

Phase III (May–Jul 2006)

Hydraulic regime

Hydraulic regime

Ab biomass

Inter Cont

Surface loading rate (g/m2 d) COD

NH+4–N

7.4 7.4

0.67 0.67

Cont Inter

Surface loading rate (g/m2 d) COD

NH+4–N

8.5 8.5

0.72 0.72

Note that in Phase III both wetlands were fed with an intermittent hydraulic regime.

Yes No

Surface loading rate (g/m2 d) COD

NH+4–N

10 10

1.0 1.0

256

A. Caselles-Osorio, J. García / Science of the Total Environment 378 (2007) 253–262

rubber material (plasticine) to reduce the convection/ diffusion of air from the atmosphere to the belowground biomass. This procedure was carried out several times during Phase III to maintain a low aboveground plant biomass in wetland B and in turn reduce the amount of oxygen released by the macrophytes. During the overall study three influent and effluent samples were collected each week (mostly on Monday, Wednesday and Friday) and analysed immediately for organic matter (COD), ammonium, nitrate, nitrite and sulphates using the methods described in APHA-AWWAWPCF (2001). In the intermittently fed system, 1 L effluent samples were obtained from the entire water volume displaced when the daily influent was added. In the continuously fed system, 1 L effluent samples were taken from the water volume displaced during an entire day. The water volumes displaced in both systems were taken in small plastic tanks and completely mixed prior sampling. Water temperature, redox potential (EH) and dissolved oxygen (DO) measurements were obtained by monitoring the water within the vertical perforated tube which was inserted into the gravel at the beginning of the study. Measurements were taken at the midpoint of the water depth and, in the case of the intermittently fed wetland, before the feeding process. Water temperature was recorded with a Checktemp-1 Hanna thermometer, EH using a platinum-tipped electrode with an Ag/AgCl reference electrode (Cryson 506) and DO with a YSI 50 oxymeter. Note that these measurements were obtained from the wetland's bulk water while analyses were carried out from the effluents. Evapotranspiration (ET) values were estimated every day from the difference between influent and effluent volumes. Statistical procedures were carried out using the SYSTAT statistical software package. All of the variables were tested to ensure that they were normally distributed. One-way ANOVA procedures were used to evaluate the effect of the hydraulic regime and the absence/presence of aboveground biomass on COD, ammonium and sulphate mass removal rates. ET was taken into account in the calculation of removal rates, that in fact were mass removal loading rates. This was particularly important in Phase III, in which the evapotranspiration rates of the wetlands differed between them due to the absence of macrophyte aboveground biomass in wetland B. 3. Results and discussion 3.1. Effect of intermittent and continuous feeding The effect of intermittent and continuous feeding on the removal of COD, ammonium and sulphate was

studied in Phases I and II, in which the water temperature in the two experimental units was very similar and averaged around 14 °C (Table 2). The DO concentrations measured at the perforated tube in the two wetlands were almost in all cases below the detection limits of the oxymeter. Several DO measurements in the effluent of the two wetlands were also below the detection limits. The EH values were clearly higher in the intermittently fed system than in the continuously fed system during both phases (Table 2 and Fig. 2). Therefore, the results for EH indicate that the intermittently fed wetland operated in more oxidised conditions than the continuously fed wetland. Moreover, this trend was not wetland-specific, since it was detected for both systems. The concentration of the COD in the influent was in average 287(± 143) mg/L for Phase I and 326(± 94) mg/L for II. The COD effluent concentrations during Phase I averaged 63(± 22) and 71(± 32) mg/L respectively for the intermittently and continuously fed systems. During Phase II, the effluent concentrations for intermittently and continuously fed wetlands averaged 119(± 33) and 125(± 46) mg/L respectively, being therefore higher than in Phase I. Despite these higher effluent concentrations the mass removal rates observed in Phase II were also quite high (Table 3). The COD mass out was very similar during the two phases in both systems (Table 3 and Fig. 3). In fact, the average COD mass removal efficiencies were not statistically different between both systems according to the ANOVA method (p = 0.663). These results indicate that although the intermittently fed wetland operated in more oxidised conditions, there was no subsequent clear improvement in COD removal. The amount of organic matter removed in the two phases and the two wetlands was on average approximately 6 g COD/m2 d. The average concentration of ammonium in the influent was 26(± 11) mg N/L for Phase I and 28(± 8) mg Table 2 Average values and standard deviations (in brackets) of water temperature, redox potential (EH) and evapotranspiration (ET) for both wetlands in Phases I (n = 30) and II (n = 45) Phase Temperature (°C)

EH (mV)

Inter

Inter

I II

Cont

ET (mm/d) Cont

Inter

Cont

14 (4) 13 (5) − 89 (30) − 143 (24) 8.4 (4.3) 8.4 (4.3) 16 (4) 14 (4) − 90 (57) − 157 (48) 7.7 (4.3) 7.7 (4.3)

Temperature and EH were measured at the perforated tubes inserted into the gravel. Note that in the two phases each wetland was operated with a different hydraulic regime (intermittent (Inter), continuous (Cont)).

A. Caselles-Osorio, J. García / Science of the Total Environment 378 (2007) 253–262

257

Fig. 2. Temporal changes in redox potential (EH) in the two wetlands according to the feeding strategy: continuous (Cont) or intermittent (Inter), and the presence or absence of aboveground (Ab) plant biomass. Note that the feeding strategies were reversed in Phases I and II and both wetlands were fed intermittently in Phase III. HLR is the hydraulic loading rate.

N/L for II. Average effluent concentrations in Phase I were 0.3(± 0.2) and 5(± 3.9) mg N/L in the intermittent and continuously fed systems respectively. During Phase II, the respective effluent concentrations were greater, averaging 7.3(± 4.5) mg N/L for the intermittent system and 12(± 5.4) mg N/L for the continuously fed system. Nitrates and nitrites were not detected or had very low concentrations in the above phases. Ammonium mass out was lower in the intermittently fed wetland than in the continuously fed wetland (Table 3 and Fig. 4). The average removal efficiencies were therefore greater in the intermittent system and statistically different (p b 0.001). In both phases the amount of ammonium removed from the intermittently fed wetland was in average 0.6 g N/m2 d, while in the continuously fed wetland was 0.4 g N/m2 d. These results clearly show that, in contrast to what was observed for COD, the intermittent feeding strategy improved the removal of ammonium. The high ammonium removal rates reported in the present study are in agreement with the earlier report by Caselles-Osorio and García (2007), in

which ammonium mass removal rates in similar shallow wetlands ranged from 80 to 90%. All these high removal efficiencies are significant, given that ammonium removal efficiencies in horizontal SSF CWs are reported at lower than 50% (USEPA, 2000). Temporal changes in ammonium mass loads out showed certain interesting patterns that were not seen for COD (Fig. 4). At the beginning of Phase I both wetlands had approximately the same ammonium mass loads out; however, the loads out of the continuously fed system increased progressively until the end of this phase. When the feeding strategies were reversed in Phase II, the wetland which was changed from continuous to intermittent suddenly produced effluents with low ammonium mass load. During the second phase the ammonium mass loads out progressively increased in the wetland A (continuous), as was observed for wetland B in Phase I. Contrary to what was observed in Phase I, in Phase II the ammonium mass out of the intermittent system (wetland B) increased progressively until March (with the same pattern observed for the continuously fed

Table 3 Average values and standard deviations (in brackets) of COD, ammonium and sulphate mass loads in (Mi) and out (Mo) (all in g/m2 d), and percentage mass removal efficiencies (Rem) in the two wetlands (intermittent (Inter) and continuous (Cont)) during Phases (Ph) I and II. n = 30 in Phase I and n = 45 in Phase II NH+4–N

Ph COD Mi

Mo Inter

I II

Rem (%) Cont

Inter Cont

7.4 (3.7) 1.1 (0.6) 1.2 (0.7) 85 8.5 (2.4) 2.4 (0.8) 2.5 (0.9) 71

84 70

Mi

SO=4 Mo Inter

Rem (%) Cont

Inter Cont

0.67 (0.3) 0.005 (0.004) 0.09 (0.08) 99 0.72 (0.2) 0.14 (0.9) 0.2 (0.1) 80

85 71

Mi

Mo Inter

Rem (%) Cont

Inter Cont

4.8 (0.7) 2.0 (1.5) 1.0 (0.7) 58 5.2 (1.8) 2.5(1.8) 1.2(0.9) 51

79 76

258

A. Caselles-Osorio, J. García / Science of the Total Environment 378 (2007) 253–262

Fig. 3. Temporal changes in COD mass loads in and out in the two wetlands according to the feeding strategy: continuous (Cont) or intermittent (Inter), and the presence or absence of aboveground (Ab) plant biomass. Note that the feeding strategies were reversed in Phases I and II and both wetlands were fed intermittently in Phase III. HLR is the hydraulic loading rate.

system). This increase occurred during the winter period, in which the aboveground biomass of the macrophytes was dry (January to March). From March onwards the ammonium mass load out of the intermittent system decreased to almost zero, while in the continuous system the mass load out remained high. Both systems showed higher COD and ammonium mass loads out (and concentrations) in Phase II than in PhaseI. This may be linked to the fact that during winter (Phase II) the aboveground biomass of the macrophytes was dry.

The average concentration of sulphate in the influent was 183(±28) mg/L for Phase I and 200(±68) mg/L for II. During both phases, the effluent concentrations of sulphate were always greater in the intermittently fed system than in the continuously fed system. Thus, in Phases I and II the effluent concentrations for the intermittently fed wetland were 103(±56) and 127(±90) mg/L respectively, while the corresponding values for the continuously fed wetland were 60(±38) and 59(±47) mg/L respectively. The sulphate mass load out was therefore higher in the intermittently fed system than in the

Fig. 4. Temporal changes in ammonium mass loads in and out in the two wetlands according to the feeding strategy: continuous (Cont) or intermittent (Inter), and the presence or absence of aboveground (Ab) plant biomass. Note that the feeding strategies were reversed in Phases I and II and both wetlands were fed intermittently in Phase III. HLR is the hydraulic loading rate.

A. Caselles-Osorio, J. García / Science of the Total Environment 378 (2007) 253–262

continuous system (Table 3, Fig. 5). Average mass removal rates were lower in the intermittently fed system and statistically different (p b 0.001). In both phases the amount of sulphate removed from the intermittently fed wetland was in average 2.8 g/m2 d, while the average in the continuously fed wetland was 3.9 g/m2 d. The lower removal rates observed in the intermittently fed system are related to the more oxidised conditions detected. Note that the sulphate mass loads out were very similar in both wetlands in Phase II from January to March, a period in which the ammonium mass loads out were also very similar (Fig. 5). From the results of this study it is clear that the intermittent feeding strategy creates a more oxidised global environment that in turn improves the removal of ammonium. The more oxidised conditions may be the result of three non-exclusive factors: hydrodynamic behaviour, water level fluctuations and macrophytemediated effects. The feeding method affects the hydrodynamic behaviour of the system and perhaps the global redox conditions (in relation with the oxygen released by plant roots that gives place to aerobic microsites). In the intermittently fed wetland the daily wastewater flow (20 L/d) was poured in over a period of 20 min, causing greater internal turbulence and mixing than in the continuous system. The occurrence of laminar or turbulent flow in SSF CWs can be determined using the Reynolds number: Rn = Vd / υ, where V is the flow per unit of transverse area (in m/s), d is the average diameter of the substrate (in m) and υ is the water kinematic viscosity

259

(in m2/s). The flow is laminar when Rn b 1; turbulent when Rn N 10; and transitional when Rn is between 1 and 10. The Reynolds numbers were 0.017 for the continuously fed system and 1.2 for the intermittently fed system (in this case when the influent was added). Therefore, the greater turbulence in the intermittently fed wetland would allow (in the moment of wastewater addition) a larger amount of water volume to pass through aerobic (near the roots) and anaerobic microsites, while more laminar conditions in the continuous system mean that perhaps some of the water only passes through anaerobic sites, particularly at the bottom of the wetland. Note that the greater turbulence in the intermittently fed system would not provide sufficient oxygen surface reaeration to explain the observed results. The intermittent method of feeding resulted in greater fluctuations of water depth than the continuous system. In the intermittent system the addition of water resulted in a level of 0.25 m which then decreased in line with the ET rate. In the continuous system the relative decrease was lower as the wetland was fed all day. Allowing for maximum and minimum ET rates, water level fluctuations in Phases I and II ranged from 9.5 to 19 mm for the intermittent system and 1.3 to 6.9 mm for the continuous. Breen (1997) reporting on batch vs. continuous wetland systems, showed that the ET rate caused level fluctuations on the batch-loaded system. These fluctuations exposed more granular medium to the atmosphere, thus promoting more oxidised conditions. Behrends et al. (1993) reported reaeration rates four times faster in

Fig. 5. Temporal changes in sulphate mass loads in and out in the two wetlands according to the feeding strategy: continuous (Cont) or intermittent (Inter), and the presence or absence of aboveground (Ab) plant biomass. Note that the feeding strategies were reversed in Phases I and II and both wetlands were fed intermittently in Phase III. HLR is the hydraulic loading rate.

260

A. Caselles-Osorio, J. García / Science of the Total Environment 378 (2007) 253–262

drain and fill treatments than in static controls, due to the rapid oxygenation of the wetted gravel that was exposed to atmospheric oxygen during the drain phase. In addition, Tanner et al. (1999) indicated that fluctuating water levels improve ammonium removal because it is adsorbed onto surfaces that are later exposed to air and thus have no oxygen limitation to nitrification. Water level fluctuations in the two wetlands of the present study could have been increased (as compared to fullscale systems) due to the small size of the wetlands. However, this fact does not affect the results in terms of comparison between the intermittent and continuous systems. A macrophyte effect may also be behind the more oxidised conditions of the intermittently fed system. Bezbaruah and Zhang (2004) observed that the DO at the root surface of S. validus increased with the oxygen demand of the surrounding bulk water. Thus it is possible that, in the present study, the intermittent addition of wastewater to the system, results in a greater volume of the reactor being in contact with the wastewater, than in the continuous system. As a result this greater volume is exposed to a comparatively higher load and therefore a higher amount of oxygen is released by the macrophytes, which contributes to the more oxidised conditions.

The average concentration of influent COD and ammonium was 257(±99) mg/L and 25(±8) mg N/L. The effluent concentrations of COD in the system with aboveground plant biomass averaged 172(±79) mg/L, compared with 132(±44) mg/L for the system without plant biomass. Note that the effluent COD concentrations were rather high in both wetlands (in comparison to the results observed in previous phases) and are related with the high evapotranspiration rates observed in Phase III. These high COD concentrations are not indicative of poor performance because if considered as mass removal rates, both wetlands presented good removal efficiencies (Table 4). The effluent ammonium concentrations averaged 1.5(±0.9) mg N/L in the system with aboveground plant biomass and 13(±5) mg N/L in the system without plant biomass. The COD and ammonium mass loads out were lower in the wetland with aboveground biomass (Table 4 and Figs. 3 and 4). The average COD and ammonium mass removal efficiencies were statistically different in the two wetlands (p b 0.001). The average amount of organic matter removed was approximately 8.1 g COD/m2 d for the system with aboveground biomass and 7.3 g COD/m2 d for the system without aboveground biomass. The average amount of ammonium mass removed was 0.96 g N/m2 d in the wetland with aboveground biomass and 0.68 g N/m2 d in the other wetland. Note that ammonium mass loads out increased progressively in the wetland without aboveground biomass (Fig. 4), showing a similar pattern to that observed in Phase I for the continuously fed wetland. From these results it is clear that the more oxidised conditions in the wetland with aboveground biomass improved the removal of COD and ammonium. The average concentration of sulphate in the influent was 132(±71) mg/L. The effluent concentrations of sulphate in the system with aboveground plant biomass averaged 933(± 542) mg/L, in comparison with 97(± 73) mg/L for the system without plant biomass. The high effluent concentrations in the system with aboveground biomass caused extraordinarily high mass loads out, with many values greater than the corresponding values of sulphate mass loads in (Fig. 5 and Table 4). These

3.2. Effect of the macrophytes The effect of the macrophytes on the removal efficiency of COD, ammonium and sulphate was studied in Phase III, in which the aboveground biomass of wetland B was cut. Note that the biomass was cut mainly to reduce the convection/diffusion of air from the atmosphere to the belowground biomass. ET was lower in wetland B in relation with the lack of aboveground biomass (Table 4). EH values were clearly higher in the wetland with aboveground biomass than in the other system (Table 4 and Fig. 2). These indicate that the wetland with aboveground biomass operated in more oxidised conditions than the other wetland.

Table 4 Average values and standard deviations (in brackets) of water temperature (Temp), redox potential (EH), evapotranspiration (ET), mass loads in (Mi) and out (Mo) (all in g/m2 d), and removal efficiencies (Rem) of COD, ammonium and sulphates in the two wetlands (with and without aboveground (Ab) biomass) during Phase III Wetland

With Ab biomass Without Ab biomass n = 29.

Temp (°C)

EH (mV)

ET (mm/d)

NH+4 –N

COD Mi

Mo

Rem Mi (%)

26 (1.4) − 35 (61) 23.4 (9.5) 10 (3.8) 1.9 (1.5) 81 26 (1.4) − 164 (38) 16.8 (6.2) 10 (3.8) 2.7 (1.1) 73

SO=4 Mo

Rem Mi (%)

1.0 (0.3) 0.019 (0.016) 98 1.0 (0.3) 0.28 (0.12) 72

Mo

Rem (%)

5.0 (2.7) 8.8 (5.3) – 5.0 (2.7) 1.4 (1.1) 72

A. Caselles-Osorio, J. García / Science of the Total Environment 378 (2007) 253–262

very high loads were linked to high redox potentials in conjunction with high ET rates (in comparison with the wetland without aboveground biomass). Under the prevailing more oxidised conditions in the wetland with aboveground biomass, the sulphates were not removed by sulphate reduction and became concentrated due to the high ET rate. Furthermore, some of the solid phase sulphides that had been deposited as a result of sulphate reduction during the previous phases could have been subsequently oxidised, due to the significantly higher redox values in the system with aboveground biomass. Thus, sulphate concentration due to ET, lack of sulphate reduction, and sulphide oxidation and subsequent sulphate formation were the most probable causes of the high sulphate mass loads out in the system with aboveground biomass. Note that reduced sulphur formed during sulphate reduction can be stored in sediments as metal sulphides and can be oxidised by chemosynthetic bacteria using oxygen as the electron acceptor (Howarth et al., 1992). This tendency for sulphate to be conserved and concentrated has been observed by other authors in natural wetlands and is clearly correlated with a decrease in the activity of sulphate-reducing bacteria (King, 1988; Choi, 2006). In the present study, when the aboveground biomass was cut in one of the wetlands the convective transport of oxygen from the aerial parts to the roots and rhizomes was reduced, the ET rates also reduced, the water redox potential decreased and the subsequent removal of COD and ammonium diminished. Moreover, the sulphate removal rate was higher than in the system with aboveground biomass, due to the greater reducing conditions in the system without aboveground biomass. Thus, in the conditions tested in this study it is clear that the presence of the aboveground biomass of the macrophytes has a significant impact on the removal efficiency of COD and ammonium. This impact may be direct by means of the convective transport of oxygen, and/or indirect, due to increased ET rates and therefore fluctuations in water level. Water level fluctuations in the wetland with aboveground biomass could have been increased due to the small size of the wetlands. It may be possible that in larger systems the differences in efficiency between wetlands with and without biomass will be lower. Note that at the end of the study a small part of one of the experimental wetlands was dissected and it was observed that root penetration was complete. COD removal was affected by the presence or absence of macrophyte aboveground biomass but not by the continuous or intermittent mode of operation, although both treatments led to changes in the redox potential. We do not know the reasons for these patterns; nevertheless, it

261

is worth noting that the differences in redox potential between the wetlands when the macrophyte effect was studied (on average 92 mV) were greater than when the feeding strategy was examined (on average 51 and 56 in Phases I and II respectively). In the present study it was observed that the higher redox values in the intermittently fed system were correlated with enhanced ammonium removal but reduced levels of sulphate removal (and accumulation) in comparison with the continuously fed system. This inverse relationship between ammonium removal and sulphate removal has been described in other studies (García et al., 2004, 2005; Wiessner et al., 2005a,b). Wiessner et al. (2005a,b) reported that this inverse relationship coincides with high concentrations of organic matter and sulphate in the influent wastewater, and indicated that reduced sulphur compounds, such as hydrogen sulphide, are known to be potent inhibitors of plant growth and certain microbial activities, including nitrification. In fact, these authors observed an exponential decrease in ammonium removal from 75 to 35% in conjunction with an increase in sulphate removal (50% removal). The results obtained in the present study are in agreement with those of Stein and Kakizawa (2005) and other authors mentioned in the Introduction, who have observed that batch loading improves contaminant removal efficiency in comparison with continuously fed systems. However, it is important to note that batch and intermittent operation are not the same, since batch mode involves the complete draining of the wetland, which was not carried out in this study. Batch systems certainly receive the influent intermittently, but the tested wetlands were not periodically drained (changes in water level only occurred as a result of ET). The intermittent feeding mode as performed in this study constitutes a very unusual method for feeding an SSF CW. In view of our results, and particularly if ammonium needs to be removed, intermittent feeding should be considered for full-scale projects. In terms of technology the intermittent feeding only needs a tank located before the wetland in order to store the influent. The intermittent feeding might be realized by means of siphons what would not suppose an energetic cost. Future experiments should analyse the effect of discharging the wastewater in several pulses instead of the single discharge used in this study. This will allow to evaluate a more realistic feeding strategy for full scale projects. 4. Conclusions Intermittent feeding in shallow horizontal subsurface-flow constructed wetlands provided a more

262

A. Caselles-Osorio, J. García / Science of the Total Environment 378 (2007) 253–262

oxidised treatment environment in comparison with continuous feeding, which in turn promoted a higher level of ammonium removal (on average 80–99% compared with 71–85%) and a lower level of sulphate removal (on average 51–58% compared with 76–79%). The presence of macrophyte aboveground biomass created a more oxidised environment in comparison with wetlands without this biomass, and this in turn enhanced the removal of COD (on average 81% compared with 73%) and ammonium (on average 98% compared with 72%). Sulphate removal was greater in the system without macrophyte aboveground biomass. This research, together with previous studies conducted by the authors, shows that significant ammonium removal (N80%) can be achieved in shallow horizontal subsurface-flow constructed wetlands, and that the level of ammonium removal can be enhanced by intermittent feeding. Acknowledgements The authors wish to express their gratitude to Romina Martín, Vanesa Vivar, Carles Calventus, Enric Cuadras and Eduardo Álvarez for their invaluable help with the experimental work. Dr. Sean O'Hogain (Dublin Institute of Technology) kindly reviewed the manuscript and made multiple comments. The results of this study were possible thanks to a grant awarded by the Spanish Department of Education and Science for research projects REN2002-04113 and CTM2005-06457. The first author would like to thank the University of the Atlantic (Colombia) for the grant awarded to carry out PhD studies in Spain. References APHA-AWWA-WPCF. Standard methods for the examination of water and wastewater. 20th ed. Washington DC: American Public Health Association; 2001. Behrends LL, Coonrod HS, Bailey E, Bulls MJ. Oxygen diffusion rates in reciprocating rock biofilters: potential applications for subsurface constructed wetlands. Proceedings Subsurface Flow Constructed Wetlands Conference. August 16-17. El Paso: University of Texas; 1993. Bezbaruah AN, Zhang TC. pH, redox and oxygen microprofiles in rhizosphere of bulrush (Scirpus validus) in a constructed wetland treating municipal wastewater. Biotechnol Bioeng 2004;88:60–70.

Breen PF. The performance of vertical flow experimental wetlands under a range of operational formats and experimental conditions. Water Sci Technol 1997;35(5):167–74. Caselles-Osorio A, García J. Effect of physico-chemical pretreatment on the removal efficiency of experimental horizontal subsurfaceflow constructed wetlands. Environ Poll 2007;146(1):55–63. Choi JH. The effects of plants on the dynamics of sulphur species and zinc in wetlands sediments. MERI New Jersey Meadowlands Commission. Spring 2006 Seminar Series. Princeton University; 2006. García J, Aguirre P, Mujeriego R, Huang Y, Ortiz L, Bayona JM. Initial contaminant removal performance factors in horizontal flow reed beds used for treating urban wastewater. Water Res 2004;38(7):1669–78. García J, Aguirre P, Barragán J, Mujeriego R, Matamoros V, Bayona JM. Effect of key design parameters on the efficiency of horizontal subsurface flow constructed wetlands. Ecol Eng 2005;25:405–18. Headley TR, Herity E, Davison L. Treatment at different depths and vertical mixing within a 1-m deep horizontal subsurface-flow wetland. Ecol Eng 2005;25:567–82. Howarth RW, Stewart JWB, Ivanov MV, editors. Sulphur cycling on the continents: wetlands, terrestrial ecosystems, and associated water bodies. SCOPE. Chichester, UK: John Wiley and Sons; 1992. Kadlec RH, Knight RL. Treatment wetlands. Lewis, CRC Press. Boca Ratón, Fl. USA; 1996. 893 pp. Kadlec RH, Knight RL, Vymazal J, Brix H, Cooper P, Haberl R. Constructed wetlands for pollution control: processes, performance, design and operation. IWA scientific and technical report No.8. IWA specialist group on use of macrophytes in water pollution control. IWA Publishing; 2000. 155 pp. King GM. Patterns of sulfate reduction and the sulphur cycle in the South Caroline salt marsh. Limnol Oceanogr 1988;33(3):376–90. Stein OR, Kakizawa K. Performance differences between batch and continuous flow SSF wetlands in summer. IWA specialist group on use of macrophytes in water pollution control. Newsletter 2005;30:16–21. Stein OR, Hook PB, Biederman JA, Allen WC, Borden DJ. Does batch operation enhance oxidation in subsurface constructed wetlands? Water Sci Technol 2003;48(5):149–56. Tanner CC. Plants as ecosystem engineers in subsurface-flow treatment wetlands. Water Sci Technol 2001;44(11–12):9–17. Tanner CC, D'Eugenio J, McBride GB, Sukias JPS, Thompson K. Effect of water level fluctuation on nitrogen removal from constructed wetland mesocosms. Ecol Eng 1999;12:67–92. USEPA. Constructed wetland treatment for municipal wastewater. EPA/ 625/R-99/010. Cincinnati, OH: Office of Research and Development; 2000. 166 pp. Vymazal J, Masa M. Horizontal sub-surface flow constructed wetland with pulsing water level. Water Sci Technol 2003;48(5):143–8. Wiessner A, Kappelmeyer U, Kunschk P, Kästner M. Influence of the redox condition dynamic on the removal efficiency of a laboratoryscale constructed wetland. Water Res 2005a;39(1):248–56. Wiessner A, Kappelmeyer U, Kunschk P, Kästner M. Sulfate reduction and the removal of carbon and ammonia in a laboratory-scale constructed wetland. Water Res 2005b;39(19):4643–50.