Methanogenic activity of accumulated solids and gas emissions from planted and unplanted shallow horizontal subsurface flow constructed wetlands

Ecological Engineering 98 (2017) 297–306

Contents lists available at ScienceDirect

Ecological Engineering journal homepage: www.elsevier.com/locate/ecoleng

Methanogenic activity of accumulated solids and gas emissions from planted and unplanted shallow horizontal subsurface flow constructed wetlands T. Carballeira a,b , I. Ruiz b , M. Soto b,∗ a b

Gairesa, Outeiro 1, Lago (Valdovi˜ no), 15551 A Coru˜ na, Spain Dept. of Physical Chemistry and Chemical Engineering I, University of A Coru˜ na, Rúa da Fraga n◦ 10, 15008 A Coru˜ na, Galiza, Spain

a r t i c l e

i n f o

Article history: Received 30 January 2016 Received in revised form 9 October 2016 Accepted 31 October 2016 Available online 3 November 2016 Keywords: Subsurface flow constructed wetlands Wetland depth Plant species Methane Carbon dioxide Greenhouse gases

a b s t r a c t Anaerobic processes play an important role in horizontal subsurface flow (HSSF) constructed wetlands (CW) and methane generation contributes to overall greenhouse gas emissions. Plants and bed depth are two main factors that can influence anaerobic processes but the effect of plants’ presence and species on methane emissions on shallower HSSF (0.3 m bed depth) has not been studied. We measured CH4 and CO2 emissions from a pilot plant constituted of five HSSF units in parallel with different plant species (CW1-UN: unplanted, CW2-JE: Juncus effusus, CW3-IP: Iris pseudacorus, CW4-TL: Typha latifolia L. and CW5-PA: Phragmites australis). Shallow HSSF beds showed high methane emissions (averaging 440 mgCH4 /m2 * d) even at low loading rates (<5.0 gBOD5 /m2 * d). Differences in mean emissions between units were not significant. However, the unplanted unit showed lower methane emissions during cold periods and higher emissions during warmer periods than planted units. Temperature was the main variable determining CO2 and CH4 emission for all units, except for CO2 emissions in CW2-JE unit. Temperature models explained 63–98% of seasonal variability and predicted zero emissions of CH4 and CO2 at temperatures of 9 ± 1 ◦ C and 7.6 ± 1.3 ◦ C, respectively. Organic matter accumulation (volatile solids, VS), specific methanogenic activity (SMA) and methane potential (MP) of accumulated solids were also determined. None of these parameters showed significant differences among CW units, except the higher values of MP in CW2-JE unit. While VS content and minimum SMA clearly increased with operation time, maximum SMA and MP remained stable. Potential methane emissions can be estimated from the product of surface density of VS and SMA, the measured emissions fitting well in the estimated range from batch assays. © 2016 Elsevier B.V. All rights reserved.

1. Introduction Constructed wetlands (CWs) are engineered treatment systems for wastewater effluents in which macrophytes play several roles helping to stabilise the surface of the beds, promote microbial activity and processes, provide good conditions for physical filtration and insulate the surface against coldness (Vymazal, 2011; Button et al., 2015). Organic matter production and plant uptake of nutrients as well as root-zone oxygen and organic carbon release were identified as key factors influencing nutrient transformation and removal. Studies showed that the above-ground and below-ground parts of the macrophytes increase microorganism diversity and provide large surface areas for the development of biofilm which is

∗ Corresponding author. E-mail address: [email protected] (M. Soto). http://dx.doi.org/10.1016/j.ecoleng.2016.10.079 0925-8574/© 2016 Elsevier B.V. All rights reserved.

responsible for most of the microbial processes occurring in CWs (Button et al., 2015; Chen et al., 2014). Most studies have shown that planted horizontal subsurface flow (HSSF) CWs achieve higher treatment efficiency than unplanted systems, at least for the removal of some pollutants such as nitrogen (Tanner, 2001; Chen et al., 2014). Even recent studies also show a higher performance of planted systems in biological oxygen demand (BOD5 ) and chemical oxygen demand (COD) removal (Leto et al., 2013; Button et al., 2015; Toscano et al., 2015). Removal efficiency is usually less affected by the plant species than by the presence or absence of plants but some studies reported higher removal for selected plant species. While some studies found a higher efficiency in nutrient removal in CWs planted with Typha sp in comparison with other species (Maltais-Landry et al., 2009; Leto et al., 2013), other studies reported better results for Scirpus ˜ et al., 2007), validus (Fraser et al., 2004), Iris pseudacorus (Villasenor

298

T. Carballeira et al. / Ecological Engineering 98 (2017) 297–306

Juncus effusus (Carballeira et al., 2016) or Phragmites australis (Toscano et al., 2015). Besides plants, substrate depth in HSSF CWs appeared as a main variable in determining the system performance with the effective substrate depth usually ranging from 0.5 to 0.6 m. Shallower HSSF systems (0.2–0.3 m medium depth) force all of the wastewater through the rooting zone of the plants, which increased the treatment performance as has been shown by García et al. (2005). In contrast, Nivala et al. (2013) reported that deeper HSSF systems (0.5 m) had higher areal oxygen consumption rates (OCRs) than shallow systems (0.25 m). On the other hand, shallow beds showed the greatest effect of vegetation on areal oxygen consumption rates, presumably because the plant rhizosphere was able to occupy a greater portion of the overall bed volume and, on a volumetric basis, the shallow systems perform better than the deeper beds (Nivala et al., 2013). Some studies on shallower CW systems (0.25–0.40 m effective depth) have been carried out for secondary treatment of municipal or domestic pre-treated wastewater, for the treatment of swine, winery and tannery industry effluents and for nitrogen removal from fish farm effluents (Carballeira et al., 2016). García et al. (2005) reported that the shallow beds removed more COD, BOD5 , ammonia and phosphorus than the deeper bed at a low surface loading rate (SLR). Pedescoll et al. (2011) demonstrated that shallow beds planted with Phragmites australis are good systems that can remove COD and ammonium efficiently. The higher efficiency observed in shallower beds was related to their less reducing conditions as indicated by higher redox potential and slightly higher dissolved oxygen concentration (García et al., 2004). The difference in redox status between shallower and deeper beds can lead to differences in the biochemical processes and in particular in methane generation and emission from the wetland. If oxygen reached a higher part of the bed in shallower wetlands, anaerobic methanogenic processes would be less predominant and methane generation and emissions would probably be lower than from deeper wetlands. In fact, CWs may emit variable amounts of carbon dioxide (CO2 ), methane (CH4 ) and nitrous oxide (N2 O), which are important greenhouse gases and main contributors to the global climate change. Mander et al. (2014) point out that several studies have shown that extensive aquatic macrophyte cover significantly suppresses CH4 emission in FWS CWs and artificial riverine wetlands. However, the effect could depend on the macrophyte species, as some aerenchymatous macrophytes such as common reed (Phragmites australis) as well as willow plants can inhibit CH4 emission from wetlands while other wetland plants such as Juncus effusus and Typha latifolia L. among others were considered important emitters of methane (Mander et al., 2014). Bateganya et al. (2015) reported that planted treatments had significantly lower CH4 emissions compared to the unplanted but significantly higher CO2 emissions. On the other hand, a lower water table level caused a significant increase in CO2 and N2 O emission and a decrease in CH4 emission in comparison to a higher water table level in both natural marsh (Yang et al., 2013) and HSSF CWs (Mander et al. 2011). In addition, Mander et al. (2015) found that short-term fluctuations in the water table of HSSF treating wastewater significantly enhanced CO2 and N2 O emission. However, increasing or lowering the water table level probably causes a different effect on gas emissions than that of designing shallow CWs with a constant water table level. However, studies regarding gas emissions from shallower CW systems (0.25–0.40 m effective depth) are scarce. Maltais-Landry et al. (2009) reported CH4 emissions of 20–120 mgCH4 /m2 * d in 0.3 m deep HSSF systems treating fish farm effluent for nitrogen removal, lower values being obtained from planted and artificially aerated units. Corbella and Puigagut (2015) reported CH4 emissions ranging from 71 to 391 mgCH4 /m2 * d in 0.3 m deep HSSF (Phragmites

australis) treating municipal wastewater and found that emissions depended on the pre-treatment method. To our knowledge, there are no studies about the effect of plants presence and species on gas emissions from shallower CWs used for secondary treatment of municipal wastewater. Another point of interest is the impact of microbial communities on gas emissions (Garcia et al., 2007; Mander et al., 2014). Zhu et al. (2007) found that methane flux was directly influenced by the quantitative variation in methanogenic and methanotrophic bacteria in both wetlands. The intensity of methanogenesis can be determined throughout batch assays carried out with the accumulated solids from the wetland media (Garcia et al., 2007). The aim of this work is to study the role of plants and the usefulness of anaerobic assays on the estimation of methane and carbon dioxide emissions from shallow HSSF CWs. The impact of the seasonal development of vegetation and other environmental factors on greenhouse gas emissions is also assessed. 2. Materials and methods 2.1. Plant description The pilot plant was built in 2009 at the outdoors of the Sci˜ in A Coruna ˜ (Spain) ence Faculty of the University of A Coruna, (latitude/longitude: 43.326382, −8.410240) and was in operation from October 2009 to March 2012. The pilot plant was constituted of five HSSF CW units in parallel, including an unplanted control unit while the others were planted with a different plant species each: unplanted (CW1-UN), Juncus effusus (CW2-JE), Iris pseudacorus (CW3-IP), Typha latifolia L. (CW4-TL) and Phragmites australis (CW5-PA). Each CW unit had an overall surface of 12 m2 (3 m width × 4 m long), gravel media of 6–12 mm in size and 35 cm depth (30 cm of water depth at the outlet) and 1% slope in the flow direction. The plant location corresponded to a temperate and humid oceanic climate. Average rainfall during the monitoring periods ranged from 0.7 to 8.3 mm/d, whilst the temperature of the influent wastewater showed a variation ranging from 13 to 19 ◦ C. Other details of plant design and operation have been reported elsewhere (Carballeira et al., 2016). 2.2. Gravel sampling procedure and biological anaerobic assays At each wetland unit, four sampling points were considered, two being placed near the inlet and the other two near the outlet (Fig. 1A). To obtain the samples, a 12.7 cm diameter steel cylinder was inserted in the gravel until the bottom of the bed. Subsequently, the gravel inside the cylinder was collected by using a gardening shovel. At the same time, a sample of the liquid that remained in the cylinder was taken. A composite, representative sample was obtained from both fractions. The two samples obtained at each location (i.e. near the inlet and near the outlet) were combined in order to obtain an integrated sample. The samples were stored in plastic bottles and covered with treated effluent from the corresponding unit in order to avoid biomass aeration. Once transported to the laboratory, the gravel samples were washed by shaking and brushing them in order to withdraw the solids trapped by or attached to the gravel. The cleaned gravel particles were removed by passing the sample throughout a 0.2 mm mesh sieve. The resulting volume containing the solids was concentrated by sedimentation. This procedure has been previously tested and applied in other CW systems (Ruiz et al., 2010). Organic matter content (volatile solids, VS), specific methanogenic activity (SMA) and methane production potential (MP) were determined for the integrated samples taken from close to the inlet and outlet zones. Anaerobic assays were carried

T. Carballeira et al. / Ecological Engineering 98 (2017) 297–306

A)

Inlet

Outlet

B)

Inlet

299

Outlet

0.75 m 1m

1m

1m

1m

Fig. 1. Location of sampling points at each CW unit: (A) Sampling points for solids sampling, (B) Sampling points for gas emissions campaigns.

out in bottles of 126 mL of volume and 50 mL of liquid volume with a VS concentration of 3 g/L following the procedures described by Soto et al. (1993). Assays for maximum SMA were supplemented with 500 mgCOD/L of acetic acid while minimum SMA and MP were obtained from assays without substrate addition. In MP assays, final accumulated methane production gives the overall MP of accumulated solids, while the initial methane production rate gives the minimum SMA (SMAmin), obtained from the accumulated solids as substrate. Macro and micronutrients were added at a ratio of 1 mL/L of the stock solutions defined by Ferreiro and Soto (2003). Biological anaerobic assays were monitored following the head-space gas analysis method (Soto et al., 1993). For this, the composition of duplicate gas phase samples (0.5 mL) was determined in a gas chromatograph equipped with a thermal conductivity detector (TCD). All assays and analyses were carried out in duplicate at 20 ◦ C in a temperature controlled chamber.

2.3. Gas emission measurements The closed chamber method was used to collect emitted gases from the wetlands. Chambers with collecting surfaces of 1810 cm2 (radius R of 24 cm, total height of 66.9 cm) were used. Two plastic rigid chambers were placed in each CW unit, one at a distance of 1 m after the inlet and the other at 1 m before the outlet following the central line of the wetland (Fig. 1B). Gas samples (1 mL) were taken with a syringe through a rubber septum placed on the top of the chamber and transported directly to the laboratory. Composition of the collected samples was determined by gas chromatography (TCD detector). The measurement period was extended for 48–96 h and the detection limits were below 1.4, 4.1 and 18.2 mg/m2 * d for CH4 , N2 O and CO2 , respectively. The surface emission rate of each gas was obtained from the evolution of the percentage of that gas in the confined atmosphere. Other details of the method are available elsewhere (de la Varga et al., 2015).

2.4. Measurement campaigns and experimental conditions The amount and characteristics of accumulated solids were determined twice after 1.4 (campaign S1) and 2.4 (S2) years of operation (Table 1). In addition, Table 1 gives a summary of plant operation conditions and treatment efficiency, which have been reported in detail elsewhere (Carballeira et al., 2016). During approximately the same period, six campaigns for emissions measurement were carried out. Both I and II campaigns were carried out in winter, before harvesting (campaign I) and after aboveground plant harvesting on day 505 (II). Campaigns III–V were carried out in successive seasons in order to complete a year period, allowing us to obtain average annual emission rates. An additional campaign (VI) was carried out one year after the first one. Temperature data for the different measurement campaigns was obtained from a local meteorological service (Meteogalicia, 2014).

The influent to the plant comes from a local sewer receiving ˜ wastewaters from one of the faculties of the University of A Coruna and surrounding houses and was pre-treated in an up-flow anaerobic sludge bed digester. After day 650 of operation, the influent to the plant was supplemented with wine vinegar in order to increase the SLR. As indicated in Table 1, emission campaigns I–III were carried out at low SLR (2.4–3.2 gBOD5 /m2 * d), while campaigns IV–VI were carried out after increasing SLR to design conditions (4.4–5.0 gBOD5 /m2 * d). Removal efficiency of BOD5 and total nitrogen (TN) was typical of HSSF CWs and ranged from 67% to 99% and from 5% to 56%, respectively (Table 1). Other details of plant operation have been reported elsewhere (Carballeira et al., 2016). 2.5. Calculations and statistical methods The suitability of the least-squares fitting (single and multiple linear regression) was evaluated by the square of the coefficient of determination (R2 ), the adjusted R2 , the statistical F-value and probability (p). A combination of the procedures for stepwise regression and regression with the best subsets of independent variables was used to select better multivariable models (Navidi, 2006). One-way and two-way analysis of variance (ANOVA) was used to compare sets of data. An Excel programme was used for these purposes. 3. Results and discussion 3.1. Solids accumulation and methanogenic activity Results for VS accumulation and methanogenic activity indicators at campaigns S1 and S2 are given in Table 2. Fig. 2 shows cumulative methane production from anaerobic assays for solid samples from campaign S1, similar results being obtained for campaign S2. In both SMA and MP assays, a latency phase of approximately 4–8 days in methane production appeared for solids from the outlet zone while no latency time was observed for solids from the inlet zone. Overall, none of the parameters in Table 2 showed significant differences among CW units (p > 0.3, except higher values for minimum SMA and MP from inlet zone of CW2-JE unit), but they decreased in all units from near the inlet zone to near the outlet zone (p < 0.06). On the other hand, while both VS content and minimum SMA values clearly increased over time, maximum SMA and MP remained stable. In MP assays, a continuous decrease of methane production rate with time existed (Fig. 2), but an inflection point was observed at approximately 80 d of batch digestion. Thus, SMAmin was calculated from the slope of methane generation rates during the first 80 days; the results are indicated in Table 2. For at least a further 80 days, methane generation continued at reduced rates of about half of that of the first 80 days (Fig. 2). SMAmin corresponded to SMA values supported by substrate for methanogenic bacteria available from hydrolysis of VS while the actual SMA in field

300

T. Carballeira et al. / Ecological Engineering 98 (2017) 297–306

Table 1 Measurement campaigns for gas emissions (I–VI) and solids (S1, S2) and conditions of plant operation and efficiency. Campaign(operation day, initial date)a

Tmeanb (◦ C)

Tmaxb (◦ C)

HLRc (mm/d)

SLRc (g BOD5 /m2 d)

%BOD5 removalc ,d

%TNremovalc

I (495, 03/01/2011) II (509, 03/15/2011) S1 (526, 04/01/2011) III (614, 06/28/2011) IV (712, 10/04/2011) V (783, 12/14/2011) S2 (831, 01/31/2012) VI (888, 03/28/2012)

9.4 ± 0.5 11.0 ± 0.3 n/a 17.4 ± 0.9 18.7 ± 0.2 12.4 ± 1.1 n/a 16.3 ± 0.4

12.2 ± 0.8 13.0 ± 1.8 n/a 20.6 ± 1.4 22.5 ± 0.9 15.3 ± 1.0 n/a 22.8 ± 1.3

23.4 ± 1.5 23.4 ± 1.5 23.4 ± 1.5 22.0 ± 0.4 22.4 ± 0.6 22.4 ± 0.6 22.6 ± 1.5 22.6 ± 1.5

2.4 ± 0.8 2.4 ± 0.8 2.4 ± 0.8 3.2 ± 0.1 5.0 ± 0.1 5.0 ± 0.1 4.4 ± 0.3 4.4 ± 0.3

88–97 88–97 88–97 79–95 67–99 67–99 71–90 71–90

28–46 28–46 28–46 36–56 27–47 27–47 5–32 5–32

n/a: not available. a The period of the study went from 1 March 2011 (campaign I, day 495 of operation) to 28 March 2012 (campaign VI, day 888 of operation). b For emission campaigns, average daily temperature during the measurement period. c Hydraulic loading rate (HLR), surface loading rate (SLR) and removal efficiencies correspond to the nearest monitoring campaign as reported by Carballeira et al. (2016). d BOD5 was determined by using an instrument from Velp Scientifica SRL, with nitrification inhibitor (1 mL of 1 g/L allylthiourea solution). Table 2 Surface density of VS accumulated in the gravel media of HSSF units and its SMA and MP. VS (kg/m2 ) HSSF unit Campaign S1 CW1-UN CW2-JE CW3-IP CW4-TL CW5-PA Mean Campaign S2 CW1-UN CW2-JE CW3-IP CW4-TL CW5-PA Mean

SMAmax (mgCH4 /gVS* d)

SMAmin (mgCH4 /gVS* d)

MP

I

O

I

O

I

O

I

(mgCH4 /gVS) O

0.28 0.22 0.19 0.17 0.17 0.20 (0.05)

0.14 0.18 0.09 0.14 0.13 0.14 (0.03)

3.10 3.60 3.38 3.05 3.68 3.35 (0.28)

2.63 2.18 2.25 2.53 3.33 2.58 (0.45)

0.25 0.60 0.33 0.28 0.33 0.35 (0.13)

0.08 0.15 0.13 0.13 0.20 0.13 (0.05)

19.5 47.8 26.8 22.3 24.0 28.0 (11.3)

4.0 13.0 5.8 11.0 20.0 10.8 (6.3)

0.39 0.52 0.75 0.58 0.48 0.54 (0.14)

0.26 0.55 0.43 0.44 0.29 0.40 (0.12)

3.65 4.28 4.00 2.33 4.40 3.73 (0.83)

2.18 3.15 2.35 2.58 2.03 2.45 (0.45)

0.30 1.50 0.35 0.33 0.33 0.55 (0.53)

0.10 0.58 0.18 0.43 0.20 0.30 (0.20)

12.5 34.5 10.0 14.5 15.3 17.3 (9.8)

5.0 17.3 7.5 12.8 8.0 10.0 (4.8)

HSSF units: CW1-UN (unplanted), CW2-JE (Juncus effusus), CW3-IP (Iris pseudacorus), CW4-TL (Thypha latifolia L.), CW5-PA (Phragmites australis). I: inlet zone. O: outlet zone. Standard deviation is given in brackets. SMAmin corresponds to specific methane generation rates during the first 80 days of MP assays (see Fig. 2C and D for campaign S1).

A) SMA Inlet (I)

B) SMA Outlet (I)

CW1-UN CW2-JE CW3-IP CW4-TL CW5-PA

CH4 PRODUCTION (mL)

8

CH4 PRODUCTION (mL)

10 9 8 7 6 5 4 3 2 1 0

7 6 5 4

CW1-IN CW2-JE CW3-IP CW4-TL CW5-PA

3 2 1 0

0

5

10

15 TIME (d)

20

25

0

30

12 10 8

10 15 TIME (d)

20

25

30

D) MP Outlet (I)

CW1-UN CW2-JE CW3-IP CW4-TL CW5-PA

CH4 PRODUCTION (mL)

CH4 PRODUCTION (mL)

C) MP Inlet (I) 14

5

6 4 2 0

7

CW1-UN

6

CW2-JE CW3-IP

5

CW4-TL

4

CW5-PA

3 2 1 0

0

50

100 TIME (d)

150

0

50

100 TIME (d)

150

Fig. 2. Methane production during anaerobic assays with solid samples from campaign S1: A and B: maximum SMA assays (supplemented with acetate), C and D: MP assays (no acetate added).

T. Carballeira et al. / Ecological Engineering 98 (2017) 297–306

conditions should be higher because of extra substrate provided by the influent wastewater. VS from inlet zone of CW2-JE showed a distinctly higher MP than all the other samples in both campaigns S1 (Fig. 2C, Table 2) and S2 (Table 2). MP after 160 days of degradation ranged from 4 to 27 mgCH4 /gVS (excluded CW2-JE inlet zone), and was higher for inlet zone than for outlet zone samples (p < 0.06) and particularly for inlet zone CW2-JE (35–48 mgCH4 /gVS) samples than for all others. Higher values for SMA were obtained for the inlet zone than for the outlet zone (p < 0.04) while differences between units were not significant (p > 0.3). SMAmin values were at least one order of magnitude lower than SMAmax values, indicating that hydrolysis was the limiting step in anaerobic degradation of accumulated VS. Mean VS content (mean from inlet and outlet zones) clearly increased from 0.17 kg VS/m2 during campaign S1 to 0.47 kg VS/m2 during campaign S2 (p < 0.000), which gives a solid accumulation rate of 0.12 kg VS/m2 yr from the beginning of the operation to S1 campaign and a rate of 0.36 kg VS/m2 yr from S1 to S2 campaigns. Differences in VS accumulation also existed between inlet and outlet zones, being bigger in campaign S1 (p = 0.024) than in campaign S2 (p = 0.099). As COD to VS ratio ranged from 1.3 2.0 gCOD/gVS (data not shown), MP data presented above indicates that only about 2–6% of total COD of accumulated solids (except 9% for CW2-JE inlet zone) was biodegraded in anaerobic conditions and transformed to methane during the 160 d of duration of batch assays. As the degradation rate decreased over time, this means that refractory solids will be progressively accumulated in the CW media, explaining the observed increase in VS accumulation from campaign S1 to S2. On the other hand, the product of accumulated VS by the respective MP gives the potential amount of methane that can be obtained from the accumulated solids. Considering average values from Table 2, the potential methane generation from accumulated solids was estimated to be 3553 mgCH4 /m2 in campaign S1 and 6658 mgCH4 /m2 in campaign S2 which would be evolved during a period of approximately 5 months at temperatures near to 20 ◦ C in absence of oxygen. Garcia et al. (2007) performed for the first time anaerobic tests with gravel from HSSF CWs. However, in their assays these authors did not find methane production from gravel samples unless additional substrate was added. Ruiz et al. (2010) modified this procedure and concentrated the accumulated solids by washing the gravel with tap water but only found methane production in some of the samples. Increasing solids concentration is necessary for obtaining reliable methanogenic activity values in anaerobic assays (Soto et al., 1993). The procedure applied in the present work, in which the solids were separated from the gravel by washing it with effluent water from the same CW units, solved the problems presented in the previous cited studies. This experience highlights the importance of avoiding the aeration of the accumulated solids prior to anaerobic assays. HSSF effluents usually have low oxidation-reduction potential, thus being adequate to carry out accumulated solids washing and conservation. 3.2. Gas emissions from HSSF units 3.2.1. Suitability of the applied method As stated by Mander et al. (2014), the chamber method was considered adequate to estimate gas emissions from these CWs with small areas. In addition, Waletzko and Mitsch (2014) demonstrated that the use of different accumulation chamber designs can be comparable. The same closed chamber with collecting surfaces of 1810 cm2 was used in a previous study (de la Varga et al., 2015) and its suitability and the effect of several factors (i.e. variability due to position in the transverse section of the CW, presence or absence of plants and recommended sampling period) had been assessed.

301

In this previous study, no significant differences were reported between methane emissions in chambers with or without plants inside, or between emissions from diurnal and nocturnal periods, while a measurement period of 24 h was considered sufficient. On the other hand, CO2 emission measurements should be carried out for a minimum period of 24 h, to include night-day effects and to obtain representative values. However, for both CH4 and CO2 emission measurements, extending the measurement period above 4 days markedly reduced the obtained emission values. Taking all this into account, the measurement period for the present study has been stablished as 48–96 h, in order to avoid an increase in the detection limit of N2 O. Furthermore, in all cases, plants were present in the measurement sites as the surface of the CW units was completely covered by plants. Depending on the weather, the temperature inside the chamber can rise in daytime above the ambient temperature, but an effect of this temperature increase on methane emissions has been discarded (de la Varga et al., 2015). Despite the fact that CW media presents a high spatial variability, de la Varga et al. (2015) reported that the chamber collection surface sufficed to reliably measure emissions and discarded stratification inside the chambers, hence mixing devices were not necessary. Fig. 3 shows methane emissions from each CW unit obtained for the six measurement campaigns carried out. The first three campaigns (I–III) corresponded to low SLR conditions and the other three (IV–VI) to design SLR conditions. CH4 emissions ranged from 0 to 1487 (average of 487) mgCH4 /m2 * d and CO2 emissions from 0 to 8971 (4166) mgCO2 /m2 * d. In a literature review, Mander et al. (2014) reported large ranges of variation, with average CO2 emissions of 4440 mgCO2 /m2 * d, 178 mgCH4 /m2 * d, and 3.1 mgN2 O/m2 * d in HSSF CWs. Methane emission rates from several types of CWs ranged from −208 to 36792 mgCH4 /m2 * d although more usual values were in the range of 0 to 2000 mgCH4 /m2 * d (de la Varga et al., 2015). Reported maximum CO2 emissions were approximately 40000–50000 mgCO2 /m2 * d (Mander et al., 2014; de la Varga et al., 2015) however values of up to 200,000 mgCO2 /m2 * d were found by Bateganya et al. (2015). 3.2.2. Emissions at low surface loading rate and harvesting effect Vegetation was harvested in all planted units on day 505 between gas emission campaigns I and II. Mean methane emissions in planted units were 126 (57–215) mgCH4 /m2 * d before harvesting (I) and 66 (21–119) mgCH4 /m2 * d after harvesting (II), while 37 and 17 mgCH4 /m2 * d were obtained for unplanted unit at campaigns I and II, respectively. Mean CO2 emissions also decreased from 2959 (865–6239) mgCO2 /m2 * d before harvesting to 2078 (1090–2758) mgCO2 /m2 * d after harvesting but this decrease was mainly caused by unit CW2-JE, the other units showing very similar CO2 emissions before and after harvesting. Zhu et al. (2007) reported that methane emission was significantly influenced by plant (Phragmites communis) harvesting, increasing immediately after plants were harvested and remaining high for about two weeks. In our study, an increase in methane emissions after harvesting was only observed for Typha latifolia L., the other planted units showing lower methane emissions. However, this behaviour in our study could be related to the low temperature of both campaings I and II, as Zhu et al. (2007) also reported that temperature significantly affected the effect of harvesting on methane emissions. CH4 emissions were low in both campaigns I and II, probably due to the season of the year and the low temperatures registered. Despite this, CH4 emissions decreased in all units after harvesting, except in unit CW4-TL but differences were not statistically significant. The unplanted unit also registered lower CH4 emissions in campaign II than in campaign I, indicating that other factors besides harvesting may have affected the changes of CH4 emissions. In any case, the higher CH4 emissions decrease was registered for

302

T. Carballeira et al. / Ecological Engineering 98 (2017) 297–306

Fig. 3. CH4 (A) and CO2 (B) emission rates from each HSSF unit at each campaign (day/month).

the CW2-JE unit, densely populated with Juncus effusus in an active vegetative state. This unit probably suffered the highest effect of harvesting, losing its protection against the cold atmosphere. In the same way, the CW2-JE unit showed a high reduction in CO2 emissions after harvesting, while the other units were not affected. In both campaigns I and II, the unplanted unit showed lower CH4 and CO2 emissions than the planted units, which could be due to the lack of protection of this unit against the cold winter atmosphere in comparison with the planted units. Following this, campaign III was carried out in the beginning of the summer and registered a large increase of CH4 emissions in all units, and of CO2 in all except in Juncus effusus CW2-JE unit. The largest increase of emissions from campaigns I–II to campaign III was registered in the unplanted CW1-UN unit. In particular, planted units emitted 584 mgCH4 /m2 * d (range 308–903 mgCH4 /m2 * d), while the unplanted units emitted 1025 mgCH4 /m2 * d. This increase in CH4 emissions from campaigns I–II to III can be explained by the higher temperatures of the summer and the growth of methanogenic bacteria during the previous weeks as well as by the persistence of biodegradable organic matter accumulated during the winter. These results show that high methane emissions can be obtained at least during part of the year even from CWs operated at low SLR.

3.2.3. Emissions at design surface loading rate Although campaigns IV and VI were carried out in autumn and spring, the registered temperatures were really high for those periods of the year and similar to summer temperatures (Table 1). Instead, the winter campaign (V) recorded a temperature proper of that time. These conditions explained high CH4 emissions during campaigns IV and VI and low CH4 emissions during campaign V (Fig. 3). CO2 emissions behaved in a similar way to CH4 emissions, although the difference between emissions rates at high and low temperatures was less marked than for CH4 . Among units, distinct behaviour was shown by CW3-IP unit (Iris pseudacorus) which

showed virtually null CH4 and CO2 emissions during campaign VI. This could be caused by the fact that the vegetation in this unit experienced higher growth in that moment of the year, since this species began to sprout in early January, while Typha latifolia L. and Phragmites australis did not do this until March or April, and Juncus effusus showed a multi-year cycle. Note that null CO2 emissions in unit CW3-IP during campaign VI required that CO2 generation from organic matter mineralisation were compensated by plant uptake.

3.2.4. Annual mean emissions and emission factors Fig. 4 shows annual average emissions obtained from four seasonal campaigns (I, III, IV and V) corresponding to the year 2011. Considering the 5 units and the 4 campaigns, significant differences among units were not found (p > 0.12) but they existed between campaigns (p < 0.005). This applies for both CH4 and CO2 emissions. Annual CO2 emissions were higher in the planted units (mean of 4510 mgCO2 /m2 * d, range of 3300–5700 mgCO2 /m2 * d) than in the unplanted unit (2944 mgCO2 /m2 * d) but differences were not significant (p = 0.12). On the other hand, CH4 emissions in the planted units (mean of 429 mgCH4 /m2 * d, range of 329–584 mgCH4 /m2 * d) were similar to those of the unplanted unit (484 mgCH4 /m2 * d) (p = 0.72). Carballeira et al. (2016) reported for the same pilot plant that significant differences for BOD5 percentage removal between units (CW5-PA > CW2-JE = CW3-IP > CW4-TL = CW1-UN) were found at design SLR while nitrogen removal rate in planted units increased with plant biomass productivity (CW2-JE > CW4TL = CW5-PA = CW3-IP> CW1-UN). Thus, the existing differences between units in organic matter and nitrogen removal did not result in differences in methane emissions. Insignificant differences in methane emissions between plant species and between planted or unplanted units found in our study agree with the results of Emery and Fulweiler (2014) for a natural salt marsh. On the contrary, Bateganya et al. (2015) reported for HSSF CWs treating wastewater that planted treatments had significantly lower methane emissions than the unplanted units.

T. Carballeira et al. / Ecological Engineering 98 (2017) 297–306

B)

800 600 400 200 0

CO2 emission rate (mg/m2*d)

CH4 emission rate (mg/m2*d)

A) 1000

303

8000 6000 4000 2000 0

Fig. 4. Annual CH4 (A) and CO2 (B) mean emissions at each HSSF unit (error bars indicate the seasonal variation, n = 4).

However, whilst Emery and Fulweiler (2014) reported lower CO2 emissions from the vegetated areas of salt marshes, Bateganya et al. (2015) found that planted units had significantly higher CO2 emissions than unplanted units. Our results are in an intermediate situation because CO2 emissions from planted units were somewhat higher than from the unplanted unit but differences were not significant. Considering a BOD5 to total organic carbon (TOC) conversion factor of 0.5 gTOC/gBOD5 (Mander et al., 2014), an average influent loading rate of 1.88 gTOC/m2 * d is obtained from Table 1 and thus the methane emission factor (EF, defined as CH4 -C emitted/TOCin) was 17% for planted units and 19% for the unplanted unit. These emissions factors are in the range of those reported by Mander et al. (2014) for HSSF CW systems but higher than the mean reported value of 4.5%. Lower EF were also reported by Corbella and Puigagut (2015) for HSSF systems treating urban wastewater from anaerobic digester (1.0–2.3%) and conventional settler (0.8%) pre-treatment and by Mander et al. (2015) for HSSF systems with fluctuations in the water table (0.1–0.5%), while Bateganya et al. (2015) reported EF ranging from 7.5–12.7%. However, EF were similar or somewhat higher than those reported for another hybrid CW system treating anaerobic digester effluents (5–17%) obtained by using the same closed chamber method (de la Varga et al., 2015). Corbella and Puigagut (2015) stated that lower redox conditions and slightly higher organic loading of a wetland receiving the effluent of a HUSB reactor resulted in higher methane emissions than those of a wetland fed with primary settled wastewater. Results from these studies suggest that wastewater pre-treatment in anaerobic digesters could enhance methane emissions in CWs, probably because of lower redox conditions and enrichment in methanogenic microorganisms. This could explain the relatively high CH4 emissions in the CWs of this study in spite of the low SLR applied. On the other hand, emission factors for CO2 emission (CO2 C emitted/TOCin) was 65.6% for planted units and 42.8% for the unplanted unit, very close to those reported by de la Varga et al. (2015) which ranged from 44% to 74%. Mander et al. (2014) reported that EF for CO2 ranged from 9% to 3780% (average of 895%), in most cases being higher than 100%. The high variability of CO2 EF attracts attention as do the very high values obtained in some studies. However, an EF lower than 100% and even lower than%BOD5 removal was reasonable because of the effect of CO2 uptake by plants, net organic matter accumulation in wetland sediments and output of dissolved CO2 in the effluent. N2 O was not detected in the air samples, indicating that N2 O emissions were always under the detection limit of 2.6 mgN/m2 * d. According to Mander et al. (2014), N2 O emissions from HSSF systems are usually below 10 mgN/m2 * d (mean value of 2 mgN/m2 * d) which indicates that in our study, emissions of N2 O are not higher

than those expected based on existing published research. As the TN SLR ranged from 0.9–1.9 gN/m2 * d (TNin), the EF (TN emitted/TNin) should be below 0.3% and then lower than the mean EF of 0.8% reported by Mander et al. (2014). N2 O EFs in the range of 0.06%–0.2% were also reported by Mander et al. (2015) and Bateganya et al. (2015). 3.2.5. Correlation between CH4 and CO2 emissions Emissions from the different units correlated among them in a different manner for CO2 than for CH4 . A strong correlation was found for CO2 emissions from three of the planted units (CW3-IP, CW4-TL and CW5-PA, R2 = 0.82–0.97) while it was weaker among this units and the unplanted unit (R2 = 0.50–0.84). On the contrary, no correlation was found for CO2 emissions from the CW2-JE unit (Juncus effusus) and any other unit (R2 < 0.12). This behaviour could be related to the distinctly multiannual growing pattern of Juncus effusus in comparison to that of the other species. Thus, the seasonal pattern of CO2 emissions was influenced by the vegetative state of the plant or by the absence of plants. CH4 emissions showed a seasonal pattern different of that of CO2 emissions. CH4 emissions from CW2-JE unit did not correlate with CW1-UN and CW5-PA emissions (R2 < 0.25) but correlated with CW3-IP and CW4-TL emissions (R2 > 0.7). In addition, no correlation was found for CH4 emissions from CW5-PA and those from CW3IP and CW4-TL (R2 < 0.51). Finally, a correlation for CH4 emissions from each other pair of HSSF units existed (R2 = 0.67–0.88). When comparing emissions from each campaign at a given unit, a correlation between CH4 and CO2 emissions existed for all units, giving regression coefficients of 0.94 (CW1-UN), 0.83 (CW2-JE), 0.99 (CW3-IP), 0.93 (CW4-TL) and 0.68 (CW5PA). A lower CH4 /CO2 emission ratio was obtained for the planted units (0.10 ± 0.04 gCH4 /gCO2 ) than for the unplanted unit (0.15 ± 0.10 gCH4 /gCO2 ) at a significant level of (p = 0.08). On the other hand, when comparing emissions from each unit at a given campaign, correlation between CH4 and CO2 emissions only existed for winter campaigns I (R2 = 0.78) and V (R2 = 0.86) but not for the highest temperature campaigns III and IV (R2 < 0.4), campaign VI giving an intermediate situation (R2 = 0.57). Temperature also influenced the CH4 /CO2 emission ratio which became lower at lower temperature. 3.3. Estimation of methane emissions after accumulated solids and its methanogenic activity Methane emissions must be related to the amount of accumulated solids and its specific methanogenic activity. Thus, we attempted to estimate potential methane emissions from biomass characteristics given in Section 3.1. Firstly, maximum potential methane emissions can be obtained from the product of the amount

304

T. Carballeira et al. / Ecological Engineering 98 (2017) 297–306 Table 3 Correlations for CH4 and CO2 emissions as a function of temperature.

Maximum (esmated) Minimum (esmated)

CH4 emissions (mg/m2·d)

1200

Emissions (mg/m2 * d) = a + b * T (◦ C)

Measured (mean)

HSSF unit

1000 800 600 400 200 0 CW1-UN

CW2-JE

CW3-IP

CW4-TL

CW5-PA

Fig. 5. Range (maximum and minimum) of methane emissions estimated from VS and SMA data and comparison with annual mean methane emissions (error bars indicate the seasonal variation for measured emissions, n = 4).

of solids by the maximum specific methanogenic activity, which has been measured for 0.5 gCOD/L and 20 ◦ C (SMAmax in Table 2). Actual methane emissions could only reach these maximum potential methane emissions in situations of substrate overloading and temperatures of about 20 ◦ C or higher. Secondly, methane emissions at a situation of low availability of easily biodegradable substrate can be obtained from the product of the amount of solids by the minimum specific methanogenic activity, which has been obtained in the absence of added substrate and at 20 ◦ C (SMAmin in Table 2). Thus, methane emissions calculated from SMAmin are the methane emissions self-supported by accumulated solids which according to Fig. 2C and D were supported at sustainable rates for a period of at least 150 days. Assay values at 20 ◦ C for SMAmax and SMAmin can be corrected for temperature dependence by applying the Arrhenius equation: SMAT = SMA20 · ␪ (T−20) where ␪ is the Arrhenius constant. For biological treatment processes including constructed wetlands (Rousseau et al., 2004; Ortigara et al., 2011) and anaerobic hydrolysis of primary sludge (Ferreiro and Soto, 2003), the value of ␪ was found to be approximately 1.06. Then, this value was used for temperature correction in the present work and the estimated surface emission rate (SER) was obtained throughout the following equation: SER (mgCH4 /m2 ∗ d) = VS(g/m2 ) ∗ SMA (mgCH4 /gVS∗ d) ∗ 1.06 (T−20) where SMA is the specific methanogenic activity at 20 ◦ C and T the temperature at which the SER is being estimated. The results for the estimated SER and its comparison with the mean annual emissions measured are shown in Fig. 5. As we have only two VS and SMA determinations (campaigns S1 and S2), the maximum and minimum SER were estimated for each campaign S1 and S2 and the averages were corrected for the annual mean temperature and plotted in Fig. 5. The measured emissions and its range of variation fit well in the range of estimated maximum and minimum values. The measured mean emissions equalled 64 ± 17% of the estimated maximum, ranging from 42% for CW2-JE to 87% for CW1-UN. The extreme values were obtained for the unplanted unit and for the planted unit which had the highest plant density (data about macrophyte biomass production was reported previously by Carballeira et al. (2016)). These results suggest that the high density of Juncus effusus in CW2-JE unit favoured both anaerobic microbial activity and methane emission reduction.

CH4 CW1-UN CW2-JE CW3-IP CW3-IPa CW4-TL CW5-PA CO2 CW1-UN CW2-JE CW3-IP CW3-IPa CW4-TL CW5-PA

a

b

R2

p

T SER = 0 (◦ C)b

−1368 −343 −716 −897 −757 −707

136.7 49.0 78.1 99.3 78.7 77.6

0.679 0.627 0.489 0.920 0.764 0.736

0.044 0.061 0.12 0.010 0.023 0.029

10.0 7.0 9.2 9.0 9.6 9.1

−4794 +4028 −3013 −4722 −2993 −4862

549 49.1 507 708 479 560

0.816 0.011 0.331 0.962 0.780 0.979

0.014 0.846 0.232 0.003 0.020 0.000

8.7 n/a 5.9 6.7 6.2 8.7

n/a: not applicable. a Correlation for CW3-IP excluding campaign VI. b Temperature that makes zero emissions as derived from the respective correlation.

3.4. The effect of temperature and plants on gas emissions An analysis of linear regression was carried out for both CH4 and CO2 emissions considering several variables (atmospheric T, SLR and operation time given in Table 1, and accumulated VS as well as the product of VS and SMAmin given in Table 2) as independent factors. Atmospheric T ranged from 9.4 ◦ C to 18.7 ◦ C and appeared as the main variable determining CH4 and CO2 emissions. Correlations of CH4 and CO2 emissions with T offered positive coefficients (i.e. the emissions increased as T increased) and high values for R2 at a probability p < 0.05 for most HSSF units (Table 3). Significant correlations for CW3-IP unit are obtained only when data from the campaign VI is excluded. Apart from this case, only CO2 emissions from CW2-JE appeared to be uncorrelated with temperature (Table 3). Significant correlations were not found for any of the other variables indicated above. In addition, none of these variables improved the parameters of T correlation when they were added as a second variable in a multiple regression model. Thus, the effect of SLR and the operation time on gas emissions in the conditions of the present work must be discarded. On the other hand, the results about the lack of influence of VS and SMA on methane emissions were considered inconclusive because only two values derived from campaigns S1 and S2 were available. These results indicate that T is the main factor determining CH4 and CO2 emissions in all units, except for CO2 emissions in CW2-JE unit. This could be related to the growing pattern of Juncus effusus, which is multiannual and less dependent on temperature in comparison to that of the other species. The temperature that makes null the emissions of CH4 and CO2 can be derived from correlations in Table 3, giving mean values of 9 ± 1 ◦ C and 7.6 ± 1.3 ◦ C, respectively. On the other hand, the slope (b) of the correlation equation for CH4 emissions in Table 3 is higher for the CW1-UN unit than for the other units. This means that CH4 emissions in the CW1-UN unit suffered a more pronounced effect of T than in the planted units. At low temperatures (campaigns I, II and V), CH4 emissions were clearly lower in the CW1-UN unit than in any other unit (p = 0.10), while at high temperatures (campaigns III, IV and VI) the opposite occurred, the unplanted unit showing higher CH4 emissions than the planted units (p = 0.09). Overall, this different behaviour of CH4 emissions as a function of T was significant at a probability level p = 0.03. As a lower temperature determines lower emissions and a higher temperature higher emissions, this behaviour is clearly

T. Carballeira et al. / Ecological Engineering 98 (2017) 297–306

305

B) 9000

1000 800 600 400 200

y = 143.8 ln(x) - 351,8 R²=0.919, p=0.003

Mean CW2-CW5 (gCO2/m2*d)

Mean CW2-CW5 emissions (gCH4/m2*d)

A)

8000 7000 6000 5000 4000 3000 2000

y = 1355.6 ln(x) - 6177.4 R²=0.627, p=0.061

1000 0

0 0

200 400 600 800 1000 1200 1400

0

2000

4000

6000

8000

CW1-UN (gCO2/m2*d)

CW1-UN (gCH4/m2*d)

Fig. 6. Effect of plants on CH4 (A) and CO2 (B) emissions (mean emissions for planted units vs unplanted unit emissions; dashed line: y = x).

shown by the correlation shown in Fig. 6A. In the case of CO2 emissions (Fig. 6B) the behaviour was similar at low temperatures, with clearly lower CO2 emissions in the unplanted unit than in the planted units (p = 0.033), while at high temperatures the emissions were not different (p = 0.59). The explanation for this behaviour may be related to the role of plants in protecting the wetland media against low temperatures in winter and higher temperatures in summer. Low protection against low temperatures in the unplanted unit can cause lower microbial activities, which lead to lower methane emissions in winter but also to a higher accumulation of biodegradable organic matter. In turn, this high level of substrate in the subsequent warmer period and higher temperatures (because of the lack of protection in unplanted units) lead to higher methane emissions in summer. 4. Conclusions Higher CO2 and lower CH4 emissions were found in planted rather than in unplanted HSSF units (4510 vs. 2944 mgCO2 /m2 * d and 429 vs. 484 mgCH4 /m2 * d, respectively) but at not significant levels. Measured methane emissions were higher than average emissions reported in literature for HSSF systems, indicating that shallower HSSF beds not necessarily cause lower methane emissions. On the other hand, these results show that high methane emissions can be obtained even from CWs operated at low SLR (2.4–5.0 gBOD5 /m2 d), at least during part of the year, indicating the need for more research on methane emissions prevention. Volatile solids and specific methanogenic activity indicators did not show significant differences among CW units but VS content and minimum SMA values (0.1–1.5 mgCH4 /gVS* d) clearly increased over time while maximum SMA (2.0–4.4 mgCH4 /gVS* d) remained stable. Potential methane emissions can be estimated from the product of surface density of VS and SMA at each point given the ranges of maximum (557–936 mgCH4 /m2 * d) and minimum (37–229 mgCH4 /m2 * d) emissions. The measured mean emissions equalled 64 ± 17% of the maximum estimated from VS and SMA, ranging from 42% for CW2-JE (Juncus effusus) to 87% for CW1-UN (unplanted). These results suggest that the high density of Juncus effusus in CW2-JE unit favoured both anaerobic microbial activity and methane emission reduction. Acknowledgements This research was supported by the Spanish Department of Economy and Competitiveness (Project CTM2011-28384). Tania

Carballeira is grateful for the fellowship received from the “Program María Barbeito PGIDIT-Incite 2010–2011 (Xunta de Galicia)”. References Bateganya, N.L., Mentler, A., Langergraber, G., Busulwa, H., Hein, T., 2015. Carbon and nitrogen gaseous fluxes from subsurface flow wetland buffer strips at mesocosm scale in East Africa. Ecol. Eng. 85, 173–184. Button, M., Nivala, J., Weber, K.P., Aubron, T., Müller, R.A., 2015. Microbial community metabolic function in subsurface flow constructed wetlands of different designs. Ecol. Eng. 80, 162–171. Carballeira, T., Ruiz, I., Soto, M., 2016. Effect of plants and surface loading rate on the treatment efficiency of shallow subsurface constructed wetlands. Ecol. Eng. 90, 203–214, http://dx.doi.org/10.1016/j.ecoleng.2016.01.038. Chen, Y., Wen, Y., Zhou, Q., Vymazal, J., 2014. Effects of plant biomass on nitrogen transformation in subsurface-batch constructed wetlands: a stable isotope and mass balance assessment. Water Res. 63, 158–167. Corbella, C., Puigagut, J., 2015. Effect of primary treatment and organic loading on methane emissions from horizontal subsurface flow constructed wetland treating urban wastewater. Ecol. Eng. 80, 79–84. de la Varga, D., Ruiz, I., Álvarez, J.A., Soto, M., 2015. Methane and carbon dioxide emissions from constructed wetlands receiving anaerobically pretreated sewage. Sci. Total Environ. 538, 824–833. Emery, H.E., Fulweiler, R.W., 2014. Spartina alterniflora and invasive Phragmites australis stands have similar greenhouse gas emissions in a New England marsh. Aquat. Bot. 116, 83–92. Ferreiro, N., Soto, M., 2003. Anaerobic hydrolysis of primary sludge: influence of sludge concentration and temperature. Water Sci. Technol. 47 (12), 239–246. Fraser, L.H., Carty, S.M., Steer, D., 2004. A test of four plant species to reduce total nitrogen and total phosphorus from soil leachate in subsurface wetland microcosms. Bioresour. Technol. 94, 185–192. García, J., Aguirre, P., Mujeriego, R., Huang, Y., Ortiz, L., Bayona, J.M., 2004. Initial contaminant removal performance factors in horizontal flow reed beds used for treating urban wastewater. Water Res. 38, 1669–1678. García, J., Aguirre, P., Barragán, J., Mujeriego, R., Matamoros, V., Bayona, J.M., 2005. Effect of key design parameters on the efficiency of horizontal subsurface flow constructed wetlands. Ecol. Eng. 25 (2005), 405–418. Garcia, J., Capel, V., Castro, A., Ruiz, I., Soto, M., 2007. Anaerobic biodegradation tests and gas emissions from subsurface flow constructed wetlands. Bioresour. Technol. 98, 3044–3052. Leto, C., Tuttolomondo, T., La Bella, S., Leone, R., Licata, M., 2013. Effects of plant species in a horizontal subsurface flow constructed wetland – phytoremediation of treated urban wastewater with Cyperus alternifolius L. and Typha latifolia L. in the West of Sicily (Italy). Ecol. Eng. 61, 282–291. Maltais-Landry, G., Maranger, R., Brisson, J., e Chazarenc, F., 2009. Nitrogen transformations and retention in planted and artificially aerated constructed wetlands. Water Res. 43, 535–545. Mander, Ü., Maddison, M., Soosaar, K., Karabelnik, K., 2011. The impact of intermit-tent hydrology and fluctuating water table on greenhouse gas emissions fromsubsurface flow constructed wetlands for wastewater treatment. Wetlands 31 (6), 1023–1032. Mander, Ü., Dotro, G., Ebie, Y., Towprayoon, S., Chiemchaisri, C., Nogueira, S.F., Jamsranjav, B., Kasak, K., Truu, J., Tournebize, J., Mitsch, W.J., 2014. Greenhouse gas emission in constructed wetlands for wastewater treatment: a review. Ecol. Eng. 66, 19–35. Mander, Ü., Maddison, M., Soosaar, K., Koger, H., Teemusk, A., Truu, J., Well, R., Sebilo, M., 2015. The impact of a pulsing water table on wastewater purification and greenhouse gas emission in a horizontal subsurface flow constructed wetland. Ecol. Eng. 80, 69–78.

306

T. Carballeira et al. / Ecological Engineering 98 (2017) 297–306

Meteogalicia, 2014. Available on https://www.meteogalicia.es (last Access: 17/02/2014). Navidi, W., 2006. Statistics for Engineers and Scientists. McGraw-Hill, México. Nivala, J., Wallace, S., Headley, T., Kassa, K., Brix, H., van Afferden, M., Müller, R., 2013. Oxygen transfer and consumption in subsurface flow treatment wetlands. Ecol. Eng. 61, 544–554. Ortigara, A.R.C., Foladori, P., Andreottola, G., 2011. Kinetics of heterotrophic biomass and storage mechanism in wetland cores measured by respirometry. Water Sci. Technol. 64 (2), 409–415. Pedescoll, A., Corzo, A., Álvarez, E., Puigagut, J., García, J., 2011. Contaminant removal efficiency depending on primary treatment and operational strategy in horizontal subsurface flow treatment wetlands. Ecol. Eng. 37, 372–380. Rousseau, D.P.L., Vanrolleghem, P.A., Pauw, N.D., 2004. Model-based design of horizontal subsurface flow constructed treatment wetlands: a review. Water Res. 38, 1484–1493. Ruiz, I., Díaz, M.A., Crujeiras, B., García, J., Soto, M., 2010. Solids hydrolysis and accumulation in a hybrid anaerobic digester-constructed wetlands system. Ecol. Eng. 36, 1007–1016. Soto, M., Méndez, R., Lema, J.M., 1993. Methanogenic and non-methanogenic activity tests: theoretical basis and experimental set up. Water Res. 27 (8), 1361–1376.

Tanner, C., 2001. Plants as ecosystem engineers in subsurface-flow treatment wetlands. Water Sci. Technol. 44 (11–12), 9–17. Toscano, A., Marzo, A., Milani, M., Cirelli, G.L., Barbagallo, S., 2015. Comparison of removal efficiencies in Mediterranean pilot constructed wetlands vegetated with different plant species. Ecol. Eng. 75, 155–160. ˜ Villasenor, J., De Lucas, A., Gómez, R., Mena, J., 2007. A comparative study of five horizontal subsurface flow constructed wetlands using different plant species for domestic wastewater treatment. Environ. Technol. 28, 1333–1343. Vymazal, J., 2011. Plants used in constructed wetlands with horizontal subsurface flow: a review. Hydrobiol 674, 133–156. Waletzko, E.J., Mitsch, W.J., 2014. Methane emissions from wetlands: an in situ side-by-side comparison of two static accumulation chamber designs. Ecol. Eng. 72, 95–102. Yang, J., Liu, J., Hu, X., Li, X., Wang, Y., Li, H., 2013. Effect of water table level on CO2 , CH4 and N2 O emissions in a freshwater marsh of Northeast China. Soil Biol. Biochem. 61, 52–60. Zhu, N.W., An, P., Krishnakumar, B., Zhao, L., Sun, L., Mizuochi, M., Inamori, Y., 2007. Effect of plant harvest on methane emission from two constructed wet-lands designed for the treatment of wastewater. J. Environ. Manage. 85 (4), 936–943.