Emission of N2O, N2, CH4, and CO2 from constructed ...

Ecological Engineering 25 (2005) 528–541

Emission of N2O, N2, CH4, and CO2 from constructed wetlands for wastewater treatment and from riparian buffer zones ¨ Mander ∗ Sille Teiter, Ulo Institute of Geography, University of Tartu, 46 Vanemuise St., 51014 Tartu, Estonia Received 1 March 2005; accepted 11 July 2005

Abstract We measured nitrous oxide (N2 O), dinitrogen (N2 ), methane (CH4 ), and carbon dioxide (CO2 ) fluxes in horizontal and vertical flow constructed wetlands (CW) and in a riparian alder stand in southern Estonia using the closed chamber method in the period from October 2001 to November 2003. The replicates’ average values of N2 O, N2 , CH4 and CO2 fluxes from the riparian gray alder stand varied from −0.4 to 58 ␮g N2 O-N m−2 h−1 , 0.02–17.4 mg N2 -N m−2 h−1 , 0.1–265 ␮g CH4 -C m−2 h−1 and 55–61 mg CO2 -C m−2 h−1 , respectively. In horizontal subsurface flow (HSSF) beds of CWs, the average N2 emission varied from 0.17 to 130 and from 0.33 to 119 mg N2 -N m−2 h−1 in the vertical subsurface flow (VSSF) beds. The average N2 O-N emission from the microsites above the inflow pipes of the HSSF CWs was 6.4–31 ␮g N2 O-N m−2 h−1 , whereas the outflow microsites emitted 2.4–8 ␮g N2 O-N m−2 h−1 . In VSSF beds, the same value was 35.6–44.7 ␮g N2 O-N m−2 h−1 . The average CH4 emission from the inflow and outflow microsites in the HSSF CWs differed significantly, ranging from 640 to 9715 and from 30 to 770 ␮g CH4 -C m−2 h−1 , respectively. The average CO2 emission was somewhat higher in VSSF beds (140–291 mg CO2 C m−2 h−1 ) and at the inflow microsites of HSSF beds (61–140 mg CO2 -C m−2 h−1 ). The global warming potential (GWP) from N2 O and CH4 was comparatively high in both types of CWs (4.8 ± 9.8 and 6.8 ± 16.2 t CO2 eq ha−1 a−1 in the HSSF CW 6.5 ± 13.0 and 5.3 ± 24.7 t CO2 eq ha−1 a−1 in the hybrid CW, respectively). The GWP of the riparian alder forest from both N2 O and CH4 was relatively low (0.4 ± 1.0 and 0.1 ± 0.30 t CO2 eq ha−1 a−1 , respectively), whereas the CO2 -C flux was remarkable (3.5 ± 3.7 t ha−1 a−1 ). The global influence of CWs is not significant. Even if all global domestic wastewater were treated by wetlands, their share of the trace gas emission budget would be less than 1%. © 2005 Elsevier B.V. All rights reserved. Keywords: Carbon dioxide; Constructed wetland; Dinitrogen; Global warming potential; Methane; Nitrous oxide

1. Introduction Constructed wetlands (CW) for wastewater treatment and riparian buffer zones are important ecotech∗

Corresponding author. Tel.: +372 7 375819; fax: +372 7 375825. ¨ Mander). E-mail address: [email protected] (U.

0925-8574/$ – see front matter © 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.ecoleng.2005.07.011

nological measures for controlling water quality in agricultural catchments (Peterjohn and Correll, 1984; Kadlec and Knight, 1996; Kuusemets and Mander, 1999). Denitrification, which is generally referred to as the microbial reduction of NO3 − -N to NO2 − -N and further to gaseous forms NO, N2 O and N2 (Knowles, 1982), has been found in numerous studies to be a sig-

¨ Mander / Ecological Engineering 25 (2005) 528–541 S. Teiter, U.

nificant process in nitrogen removal in riparian buffer zones (Cooper, 1990; Ambus and Lowrance, 1991; Groffman et al., 1991; Lowrance, 1992; Ambus and Christensen, 1993; Haycock and Pinay, 1993; Schipper et al., 1993; Hanson et al., 1994; Weller et al., 1994; Nelson et al., 1995; Hill, 1996; Schnabel et al., 1997; Gold et al., 1998; Hefting and de Klein, 1998; Mogge et al., 1998; Groffman et al., 2000). In the majority of these studies, nitrous oxide (N2 O) fluxes have been measured, and only a few studies devote attention to dinitrogen (molecular nitrogen; N2 ) emission (Blicher-Mathiesen et al., 1998; Watts and Seitzinger, 2000). N2 O, as one of the greenhouse gases, is increasing in the atmosphere at a rate of about 0.3% year−1 (Mosier, 1998). It has an atmospheric lifetime of about 120 years, a global warming potential of 296 relative to CO2 over a 100 year time horizon, and is responsible for about 6% of anticipated warming (IPCC, 2001). Riparian zones have the potential to be hotspots of N2 O production in the landscape (Groffman et al., 2000). Numerous studies consider emissions and sequestration of carbon dioxide (CO2 ) in wetlands (Mitsch and Gosselink, 1993; Funk et al., 1994; Hamilton et al., 1995; Lafleur et al., 1997; Joiner et al., 1999; Griffis et al., 2000; Christensen et al., 2003). Depending on meteorological and hydrological conditions, wetlands can be sources or sinks of carbon (Clark et al., 1999; Waddington and Roulet, 2000; Whiting and Chanton, 2001; Arneth et al., 2002). Likewise, riparian wetlands and wet riparian forests can be sources of CO2 (Jones and Mulholland, 1998a; Scott et al., 2004) and methane (CH4 ) (Jones and Mulholland, 1998b; Rusch and Rennenberg, 1998; Gulledge and Schimel, 2000; Rask et al., 2002), which is another greenhouse gas increasing in the atmosphere at the rate of about 0.8% year−1 (Mosier, 1998). Methane in the atmosphere has a lifetime of 8.4 years. On a 100 year time horizon, CH4 has a global warming potential of 23 relative to CO2 , and is responsible for about 20% of anticipated warming (IPCC, 2001). Both denitrification and methane formation depend on the oxygen status of the soil or sediment. In this relation, the spatial and temporal variability of fluxes of both N2 O (Robertson and Tiedje, 1984; Struwe and Kjøller, 1990; Ambus and Christensen, 1993; Pinay et al., 1993; Brooks et al., 1997; Augustin et al., 1998b; Gold et al., 1998; Jacinthe et al., 1998) and CH4 (Saarnio et al., 1997; Willison et al., 1998; Reay et al., 2001) is extremely high. In addition to oxy-

529

gen status, denitrification rates in soils are influenced by carbon availability, nitrate availability, temperature, and pH (Erich et al., 1984). Biological methane oxidation is an important environmental process preventing the release into the atmosphere of much of the CH4 produced in anoxic soils and sediments. Well-drained soil acts as a sink for atmospheric CH4 due to methane oxidation (negative emission), through either ammonia oxidizers or methanotrophs (Hanson et al., 1994). In contrast to riparian buffer zones and natural wetlands, far fewer studies have been carried out on N2 O and CH4 fluxes from CWs for wastewater treatment. Most of the data are available on the contribution of free water surface constructed wetlands to N2 O (Lund, 1999; Xue et al., 1999; Bachand and Horne, 2000; Lund et al., 2000; Spieles and Mitsch, 2000; Wild et al., 2002; Johansson et al., 2003) and CH4 (Tanner et al., 1997; Tai et al., 2002; Wild et al., 2002) emissions. Only two works (Fey et al., 1999; Tanner et al., 2002) considered the nitrous oxide fluxes from subsurface flow constructed wetlands, and only one paper considers dinitrogen emission from a CW (Mander et al., 2003). Surprisingly, we could find no published materials on CO2 emissions from constructed wetlands. The main objectives of this research were: (1) to quantify N2 O, N2 , CH4 and CO2 emission rates from two subsurface flow CWs for municipal wastewater treatment and in a grey alder stand in Estonia using the closed chamber method and (2) to compare N2 O, N2 , CH4 , and CO2 fluxes and their global warming potential (GWP) from riparian buffer zones and constructed wetlands.

2. Methods 2.1. Site description A description of the Kodijärve horizontal subsurface flow (HSSF) planted sand filter (constructed in October 1996, purifies the wastewater from a hospital for about 40 population equivalents (PE); Fig. 1A) is given by Mander et al. (2001, 2003). The hybrid treatment wetland system in Kõo, Viljandi County, Estonia, consists of a two-bed vertical subsurface flow (VSSF) filter (2 m × 64 m, filled with 5–10 mm crushed limestone, planted with Phragmites australis), a HSSF filter (365 m2 , filled with 15–20 mm

530


Fig. 1. Diagrams of study sites: (A) horizontal subsurface flow planted sand filter (HSSF) in Kodijärve, (B) hybrid wetland system in Kõo and (C) riparian study area in Porijõgi. In part A: plant names in italics indicate the present dominant species in the beds, (M) automatic weather station. In part B: (1) pumping station; (2) septic tank; (3) vertical subsurface flow filter (VSSF; 2 × 64 m2 ), (a) right part, (b) left part; (4) HSSF (365 m2 ), (a) left inflow, (b) right inflow, (c) outflow; (5) 1st free-water surface wetland (FWSW; 3600 m2 ), (6) 2nd FWSW (5500 m2 ), (7) polishing pond (500 m2 ).


crushed limestone, planted with Typha latifolia and P. australis), and two free water surface wetland (FWSW) beds (3600 and 5500 m2 , planted with T. latifolia; Fig. 1B). The system was constructed in 2000 for the purification of the raw municipal wastewater generated by about 300 PE. This wetland system showed a good purification, being for BOD7, total-N and total-P 88, 65 and 72%, respectively. Regarding the very high average loading rates (96, 34 and 4.7 g m−2 for BOD7 , total N and total P, respectively), however, the purification efficiency of the VSSF and HSSF part was only 57 and 31% for BOD7 , 16 and 21% for total N, and 19 and 21% for total P, correspondingly. The Porijõgi riparian buffer zone site is a grey alder stand situated in the moraine plain of southeast Estonia (Tartu County, Sirvaku; 58◦ 13 N, 26◦ 47 E) on the right bank of a small river, the Porijõgi, which flows in a primeval valley where agricultural activities ceased in 1992. The landscape study transect in this valley crosses several plant communities: an abandoned field (last cultivated in 1992) on planosols and podzoluvisols; an abandoned cultivated grassland (last mowed in 1993) on colluvial podzoluvisol (dominated by Dactylis glomerata and Alopecurus pratensis); an 11m-wide wet grassland on gleysol (two parallel communities, one dominated by Filipendula ulmaria, another by Aegopodium podagraria); and a 20-m-wide grey alder stand (Alnus incana) on gleysol (Fig. 1C). For a more detailed description, see Kuusemets et al. (2001). The mean annual air temperature at the study sites varied from 5.0 to 5.5 ◦ C. In winter the lowest daily mean temperatures reach −20 ◦ C. The variation in long-term annual precipitation is 500–700 mm. 2.2. Sampling and laboratory analysis For the measurement of N2 O, N2 , CH4 and CO2 , two emission methods—the “closed chamber” (closed soil cover box) method (Denmead and Raupach, 1993; Hutchinson and Livingston, 1993) and the helium–oxygen (He–O) method (Butterbach-Bahl et al., 1997; Scholefield et al., 1997; Mander et al., 2003) were used. The latter was used especially for the measurement of N2 fluxes. Gas samplers (closed chambers; cover made from PVC, height 50 cm, Ø 50 cm, volume 65 l, sealed with a water-filled ring on the soil surface, painted white to avoid heating during application) were installed in five replicates in various parts

531

of the studied systems: (1) on the inlet and outlet pipes of both beds, in Kodijärve (Fig. 1A) and (2) in three different microsites (EDGE, WET and DRY; range of water table depth 45–95, 0–50 and 45–95 cm, respectively) in the Porijõgi riparian buffer zone (Fig. 1C). In the hybrid wetland system in Kõo, 8 gas samplers were installed in the vertical flow filter (4 in each bed), and 15 in the horizontal flow filter (5 on two inlet pipes and 5 on the outlet pipe; Fig. 1B). At the end of the 1 h measuring time, gas samples were taken from the enclosures of the samplers, using previously evacuated gas bottles (100 ml; see Augustin et al., 1998b). Gas sampling was carried out 15 times on the following time schedule: once a month in October and November 2001, and in March, May to December 2002, January to March, July and November 2003. Simultaneously, the soil temperature and water depth in the sampling wells was measured, and the NH4 -N and NO3 -N concentration in soil samples was analysed using the Kjeldahl method (APHA, 1989). The trace gas concentration in the collected air was determined using the gas chromatography system (electron capture detector and flame ionization detector; Loftfield et al., 1997) in the lab of the Institute of Primary Production and Microbial Ecology, Centre for Agricultural Landscape and Land Use Research (ZALF), Germany. The trace gas flux rates were calculated according to Hutchinson and Livingston (1993) from a linear change in trace gas concentration over time with reference to the internal volume of the chamber and the soil area covered. Soil temperature and groundwater tables were measured simultaneously (Augustin et al., 1998a). Intact soil cores (diameter 6.8 cm, height 6 cm) for use with the He–O method were sampled from the topsoil (0–10 cm) at gas sampler (closed chamber) sites, after gas sampling was completed, in the following order: in October and November 2001, March, June to August, and October 2002, and January to March 2003. In following text, months from November to April have been considered as “winter”, and months from May to October as “summer”. Soil samples were weighted, kept at low temperature (4 ◦ C), and transported to the ZALF laboratory. At the lab, intact soil cores were introduced into special gas-tight incubation vessels. In these vessels, N2 was removed by three subsequent slight evacuation/flushing cycles with an artificial gas mixture (21.3% O2 , 78.6% He, 337 ppm CO2 , 374 ppb N2 O, 1882 ppb CH4 and approximately 5 ppm

532


N2 ). This was followed by the establishing of a new flow equilibrium by continuously flushing the vessel headspace with the artificial gas mixture at 10 ml/min for 12 h. For the start value, N2 and the greenhouse gas concentration in the continuous gas flow was measured. The measuring of the gas concentrations in the incubation headspace (final value) followed after closing the incubation headspace for one hour to accumulate the emission of N2 and the greenhouse gases. The final accumulation value minus the start continuous flow value served as the basis for the calculation of the emission rates. The procedures used for the determination of the actual gas emission rate are described by Mander et al. (2003). In Kodijärve, we also measured the water table once a month and took samples from 18 sampling wells and the inflows and outflows of both beds for further analyses for BOD7 , NH4 -N, NO2 -N, NO3 -N, total N, PO4 -P and total P in the lab of Tartu Environmental Research Ltd. (APHA, 1989). Water discharge was measured using tipping buckets installed in the inlet and outlet wells. Air and soil temperature, wind velocity, solar radiation and precipitation evapotranspiration were measured using a DAVIS Groweather automatic weather station installed close to the CW (Fig. 1A). 2.3. Statistical analysis The normality of variables was checked using the Kolmogorov-Smirnov, Lilliefors, and Shapiro-Wilk tests. In most cases of gas analyses the distribution differed from the normal, and hence non-parametric tests were performed. We used the Duncan Test, Wilcoxon Matched Pairs Test and the Mann–Whitney U-test to check the significance (α = 0.05 was accepted in all cases) of differences between the gas emission rates at different times and sites. The Spearman Rank Order Correlation was performed to analyse correlations between gaseous fluxes and environmental parameters.

3. Results and discussion 3.1. Temporal variation of gas emissions According to the Duncan test, a significantly higher release of all gases from CWs was observed during the warmer period (Fig. 2A–D), although the N2 O flux

showed no significant correlation with air and water temperature. However, mean topsoil temperatures (varied from 0.1 to 20.5 ◦ C) correlated significantly with the emission rates of all analysed gases (R2 values for N2 O-N, N2 -N, CH4 -C and CO2 -C are 0.32, 0.56, 0.50 and 0.54, respectively. In the HSSF wetlands, the season-dependence of CH4 emission was extremely remarkable. It resulted in significant differences in average values of CH4 fluxes from both the HSSF CW in Kodijärve and the hybrid CW in Kõo in summer (5000–21900 and 1700–14400 ␮g CH4 -C m−2 h−1 in Kodijärve and Kõo, respectively) and winter (24–300 and 16–2000 ␮g CH4 -C m−2 h−1 , respectively; Fig. 2B). The very cold winter of 2002/2003 with air temperatures from −15 to −25 ◦ C for almost 2 months apparently influenced both water purification efficiency (Noorvee et al., 2005) and gas emissions. As with purification performance, gaseous emission was significantly lower in spring and early summer than in autumn. In the riparian grey alder stand, only the CO2 emission varied in accordance to variations of water and air temperature. The average CO2 emission varied from 13.6 ± 11.3 mg CO2 -C m−2 h−1 in January to 187.8 ± 56.3 mg CO2 -C m−2 h−1 in August. The emission of N2 O from the riparian zone showed the highest values in January and March 2003 (up to 180 ␮g N2 ON m−2 h−1 from the WET microsite), remaining relatively low during the rest of the study period (from −3.3 to 24 ␮g N2 O-N m−2 h−1 ; Fig. 2A). Likewise, the results of some other investigations demonstrate that N2 O emission does not clearly depend on soil temperature, and the release of this gas from the soil in cold periods can be as high or even higher in winter as in summer (Augustin et al., 1996; Fey et al., 1999). For instance, N2 O-N fluxes through the snowpacks in winter reached 112 ␮g N2 O-N m−2 d−1 (Brooks et al., 1997), which is comparable with the lower emission values from our study sites. In the warm and dry summer of 2002, the N2 O emission from the riparian zone increased significantly with the lowering water table level (Spearman R = 0.38). This effect has been noted in several studies on natural wetlands (Martikainen et al., 1993; Dowrick et al., 1999). The average CH4 emission from the riparian alder stand varied from 0.1–29 to 1.2–265 ␮g CH4 -C m−2 h−1 in winter and summer, respectively (Fig. 2B). In our riparian study area, the emission of CH4 in the snow-covered period is signifi-

¨ Mander / Ecological Engineering 25 (2005) 528–541 S. Teiter, U. Fig. 2. Temporal variation of emission rates of nitrous oxide (A), dinitrogen (B), methane (C) and carbon dioxide (D; average ± S.D.) from the Kodijärve HSSF CW, Kõo hybrid wetland system and the Porijõgi riparian grey alder stand, averaged over all sampling sites. For better visualization, polynomial curves are added. Hidden values in part C: (1) 21890 ± 43570; (2) 18110; (3) 27425; (4) 14020 ± 17920; (5) 1030; (6) 14410 ± 14290; (7) 17570.

533

534


cantly less than that reported by Wickland et al. (1999) for subalpine wetland sites in the Rocky Mountains (23–73% of the annual fluxes). The average dinitrogen flux from the microsites in Kodijärve was 2–3 magnitudes higher than the N2 O flux, ranging from 19.5 to 33.3 mg N2 -N m−2 h−1 . In the Kõo hybrid CW and in the Porijõgi riparian alder stand, the difference between the N2 and N2 O fluxes was 20–750 and 150–700 times, respectively. In Kõo, the variation of N2 emission was from 0.3 to 99, and in Kodijärve, from 0.6 to 17 mg N2 -N m−2 h−1 (Fig. 2A and B). In our study, CO2 emission is not connected with fluxes related to plant photosynthesis. Therefore, only data for cold periods can be considered as losses to the atmosphere. For calculating the net ecosystem CO2 exchange, a more advanced measurement technique is required. For instance, the eddy covariation technique allows the analysis of full C balance in ecosystems (Shurpali et al., 1993; Kormann et al., 2001). However, some studies on C sequestration in wetlands and

forest ecosystems (Butnor et al., 2003) allow one to estimate that about 50% of CO2 , released during soil respiration in the vegetation period cycles back to the atmosphere. It is important to take this into consideration when calculating the GWP of CWs and riparian buffer ecosystems. 3.2. Spatial variation of gas emissions The average flux of nitrous oxide from the microsites in the Kodijärve HSSF CW and Kõo hybrid CW ranged from 27 to 370 and from 72 to 500 ␮g N2 O-N m−2 h−1 , respectively (Fig. 3A). In Kodijärve, according to the Wilcoxon Matched Pairs Test, significant differences were found in average N2 O fluxes between the microsites: 325–350 ␮g N2 ON m−2 h−1 from chambers installed above the inflow pipes and 30–40 ␮g N2 O-N m−2 h−1 from chambers above the outflow pipes. In Kõo, the VSSF beds emitted more nitrous oxide than the HSSF bed (405–510 and 70–165 ␮g N2 O-N m−2 h−1 , respectively), although

Fig. 3. Emission rates of nitrous oxide (A), methane (B), carbon dioxide (C) and dinitrogen (D; average ± S.D.) from sampling sites in the Kodijärve HSSF CW, Kõo hybrid wetland system and Porijõgi riparian grey alder forest. *: Significantly differing value (p < 0.05) with at least two other microsites according to the Wilcoxon Matched Pairs Test. For the locations of sampling sites, see Fig. 1.


the differences were not significant. This kind of difference is probably due to higher initial load in the VSSF bed (see Section 2.1). In Kodijärve, differences in individual values of N2 emission from replicate soil cores varied greatly, from 170 to 130,000 ␮g N2 -N m−2 h−1 , but the variations were statistically non-significant (Fig. 3B). In contrast to the N2 emission, we found significant differences in average N2 O fluxes between the microsites: about 320–350 ␮g N2 O-N m−2 h−1 from chambers installed above the inflow pipes and