Bioresource Technology 98 (2007) 3044–3052

Anaerobic biodegradation tests and gas emissions from subsurfaceflow constructed wetlands

Joan Garcıa a,*, Vanessa Capel a, Anna Castro a, Isabel Ruız b, Manuel Soto b

a Environmental Engineering Division, Hydraulics, Coastal and Environmental Engineering Department, Technical University of Catalonia,

c/ Jordi Girona 1-3, 08034 Barcelona, Spainb Departamento de Quımica Fısica e Enxenaria Quımica, Universidade da Coruna, A Zapateira s/n, 15071-A Coruna, Spain

Received 21 March 2006; received in revised form 16 October 2006; accepted 17 October 2006Available online 4 December 2006

Abstract

Anaerobic tests with gravel from horizontal subsurface flow constructed wetlands (SSF) used for the treatment of urban wastewaterwere developed in order to evaluate the anaerobic biodegradability of their effluents. Two types of assays were conducted. The reactorsused for the first type were glass vials of 45 mL, that were used for only one measurement, requiring starting experiments with a numberof reactors equal to the measurements to be made. For the second type of experiments multiple measurements were done in the samereactors, by using flasks of 2.2 L. The COD of the SSF effluents used for the tests ranged from 60 to 130 mg/L. The evolution ofCO2 in the headspace of the reactors was used as indicator of anaerobic biodegradation rates. CO2 mass emission rates ranged from0.005 to 0.015 lmol/mL day. CH4 generation was not detected in the tests in relation with the refractory properties of the effluent organicmatter of the studied SSF. In situ measurement of CO2 and CH4 emissions from the gravel of the SSF ranged from 0.106 to 0.464 andfrom 0.039 to 0.107 mmol/m2 h, respectively. Several CO2 fluxes measured in the field were quite consistent with the emissions observedin the laboratory. The developed tests can help to understand the performance of SSF and improve their operation.� 2006 Elsevier Ltd. All rights reserved.

Keywords: Wastewater; Horizontal flow; Gas emissions; Methane; Carbon dioxide

1. Introduction

Horizontal subsurface flow constructed wetlands (SSF)currently are a very popular method for the sanitation ofsmall rural communities worldwide. These systems arecomplex bioreactors in which the removal of contaminantsoccurs by means of a variety of different physical, chemicaland biochemical processes (Kadlec and Knight, 1996).

The relative importance of different biochemical reac-tions involved in the degradation of organic matter in SSFhas been scarcely studied. The results obtained in severalreports suggest that aerobic respiration, denitrification, sul-phate reduction and methanogenesis are the most important

0960-8524/$ - see front matter � 2006 Elsevier Ltd. All rights reserved.

doi:10.1016/j.biortech.2006.10.016

* Corresponding author. Tel.: +34 93 401 6464; fax: +34 93 401 7357.E-mail addresses: joan.garcia@upc.edu (J. Garcıa), sotoc@udc.es (M.

Soto).

reactions, and can occur at the same time in different loca-tions of the wetlands (Burgoon et al., 1995; Aguirre et al.,2005; Garcıa et al., 2004, 2005). In the work of Garcıaet al. (2004) the rates of these reactions were indirectly esti-mated through theoretical mass balances. This approachcan be done very easily but the results are only indicativebecause the high number of assumptions needed. Despitethis inconvenience, it is quite clear that external factors suchas the organic load or intrinsic design factors like the waterdepth influence the relative rates of the biochemical reac-tions, and finally have impact on the efficiency of the system.

SSF treating municipal wastewater (non diluted primaryeffluents) can be considered mainly as anaerobic systems(Baptista, 2003). The occurrence of anaerobic reactionsis in agreement with the redox potential measurements,that usually are in average for all water depth underEH = �100 mV (Garcıa et al., 2003); values higher than

J. Garcıa et al. / Bioresource Technology 98 (2007) 3044–3052 3045

�100 mV can be found near the water surface while lowernear the bottom. Despite the importance of anaerobic bio-chemical reactions, in the literature there are no reportsdealing with tests for measuring anaerobic biodegradabilityof wastewater by gravel biomass of SSF systems. In fact,the information is presently limited to field measurementsof CH4 and other gas emissions (Brix, 1990; Tanneret al., 1997; Mander et al., 2003).

The total flux of CO2 and CH4 emitted to the atmospherein vascular plant-dominated natural wetlands (and there-fore supposedly in SSF) is mediated by three processes:gas diffusion and ebullition from sediments, and internalplant-mediated transport (Brix et al., 2001). The totalrelease by plant transport has been estimated to accountfor approximately 70% of the methanogenesis in a naturalwetland located in Czech Republic (Brix et al., 2001). Theremaining percentage not released can be accounted for dif-fusion and ebullition, and by methanotrophic oxidation(van der Nat and Middelburg, 1998). In the case of SSF,the relative release by the different processes involved isunknown. However, it could be quite different from thatobtained in natural wetlands because of the greater amountof organic matter present in wastewater.

In many SSF it has been observed that the total organicmatter concentration decreases near the inlet and after thatlittle further removal occurs (Kadlec, 2003). This is due tothe fact that a large fraction of the influent organic matter(representing in most cases more than 50% of the totalBOD5) is in the form of particles that are retained nearthe inlet. The remaining dissolved organic matter isremoved very slowly through the entire length of the SSF(Kadlec et al., 2000). The removal efficiency of this dis-solved organic matter has been reported to be in average8.9% for a SSF with a water depth of 0.5 m and 30% foranother with a depth of 0.27 m (Garcıa et al., 2005). Onequestion that arises then is whether this dissolved organicmatter is refractory to the predominant anaerobic meta-bolic pathways. Anaerobic biodegradation tests may helpto answer this question.

Thus, in the present paper, anaerobic tests with gravelfrom SSF are developed for the first time. Measurementof gas flux emissions from the gravel of SSF were also donein order to compare field data with the results of the labo-ratory experiments. In this study it was decided to workwith the gravel alone in order to gain experience due tothe lack of previous works. However, anaerobic tests andmeasurements of gas emissions taking into account plants,which have been ignored in the current study, must beevaluated.

2. Methods

2.1. Field gravel gas emissions

Field gas emissions were measured in two SSF (namedA1 and A2) of a pilot plant located in Les Franqueses delValles, Barcelona (Spain). This pilot plant treats the urban

wastewater of a housing scheme named Can Suquet and ismade up by an Imhoff tank for primary treatment followedby eight parallel SSF. All the SSF have approximately thesame surface area (55 m2). A detailed description of theplant can be found elsewhere (Garcıa et al., 2004). SSFA1 and A2 have a length to width ratio of 1:1 and an aver-age wetted depth of 0.5 m. A1 contains coarse graniticgravel (the diameter at which 60% of the material passedthrough the sieve was D60 = 10 mm, the uniformity coeffi-cient was Cu = D60/D10 = 1.6, 39% of initial porosity) whileA2 contains fine gravel (D60 = 3.5 mm, Cu = 1.7, 40% ofinitial porosity). The pilot plant started operation in March2001 when the SSF were planted with common reeds (Prag-

mites australis (Cav.) Trin. ex Steud.). Note that in this pilotplant all SSF are fed with the same flow of primary effluent.At the time of gas emission measurements, the SSF wereoperated with a hydraulic loading rate of 36 mm/d.

Gas samples were obtained from gas collectors made ofplastic turned round funnels (0.25 m of diameter) insertedinto the gravel, near the outlet of both SSF (at 3/4 of thelength). The funnels were installed in February 2002 mak-ing a hole in the gravel and removing all rhizomes androots of the reeds. Once the funnels were arranged, a glassvial with the base cut was attached onto the top with gumand sealed with teflon. The vials had a screw top where amicrovalve could be fitted. Note that the funnels had theirbiggest opening submerged into the water. Gas collectorsremained open in contact with the air until the start ofgas measurements. In this study two sets of measurementswere conducted, one on July 18th (M1) and another one onAugust 24th 2002 (M2). At the beginning of the experi-ments the collectors were closed. Gas samples wereobtained periodically with Hamilton� gas sampling syrin-ges from the headspace during a period of 4–7 h of incuba-tion. In each experiment 4–7 gas samples were taken. CO2

and CH4 were analysed with a gas chromatographerequipped with a flame ionisation detector. The volume ofthe headspace was estimated before each trial by measuringthe distance between the water level and the top side of thevial, and making the necessary geometrical calculations.

During the two measurement days, grab samples from theinfluent and the effluents of both SSF were collected andanalysed for temperature, pH, COD and ammonia nitrogenusing conventional methods described in APHA-AWWA-WPCF (1995). Additional data on the performance of thesetwo SFF are available from previous publications (Garcıaet al., 2004, 2005).

2.2. Anaerobic biodegradability tests

Gravel used for these experiments was obtained fromanother SSF of the eight systems located in the pilot plantof Les Franqueses del Valles (named C2 in Garcıa et al.(2004); surface area of 55 m2). This SSF has a length towidth ratio of 2:1, an average wetted depth of 0.5 m andcontains fine granitic gravel (D60 = 3.5 mm, Cu = 1.7,40% of initial porosity). Note that SSF C2 is similar to

3046 J. Garcıa et al. / Bioresource Technology 98 (2007) 3044–3052

A1 and A2 (the wetlands used for field gravel gas emis-sions) but the length to width ratio is different. The gravelcontained in SSF C2 is the same than in A2 but finer thanin A1. The results concerning field gas emissions and an-aerobic biodegradability tests can be in general comparedsince the removal efficiency of these wetlands in terms ofCOD, BOD and ammonium were very similar in a longterm evaluation (Garcıa et al., 2004, 2005). At the timeof these experiments, the SSF C2 was operated with aHLR of 36 mm/d. In all tests that were performed, twogravel samples (one from near the inlet, at 1/3 of the length,and the other near the outlet, at 3/4 of the length) wereobtained by excavating holes. These two gravel sampleswere taken in order to study the anaerobic activity of thegravel from different locations. The samples were takenfrom near the bottom of the SSF. At the laboratory, plantlitter and other residues were separated from the gravel byhand and after that it was carefully cleaned with distilledwater and arranged on a sieve for draining during onenight. On the next day, the gravel was still wet and it wasused for the experiments. Characterisation results indicatedthat the gravel contained in average about 0.8 g VSS/kggravel (5 measurements).

At the same time the gravel was collected, samples fromthe influent or the effluents (depending on the experiments)were taken and analysed for pH, COD, sulphate, ammonianitrogen and nitrate using the methods described inAPHA-AWWA-WPCF (1995). Additional data on theperformance of SFF C2 is also available from previouspublications (Garcıa et al., 2004, 2005).

2.3. Tests with the effluents

Two types of tests were conducted in order to comparetheir reliability and usefulness: (1) experiments in which thereactors were used for only one measurement (and there-fore rejected afterwards) and (2) experiments in which mea-surements were done periodically in the same reactors. Thereactors used for the first type of experiments were glassvials of 45 mL. Six reactors were filled with gravel fromnear the inlet, six with gravel from near the outlet andanother six with clean gravel. Each reactor contained anamount of 45–50 g of drained wet gravel that occupiedapproximately 2/3 of the total height. Reactors with cleangravel were used as blanks to test the activity of the micro-organisms contained in effluent wastewater itself. Cleangravel was obtained drying gravel at 103 �C during onenight and after that placing the gravel in an oven at550 �C for 2 h. Once the gravel was placed into the vials,they were filled completely with effluent and closed with ascrew top with a microvalve. Next, 5 mL of water wasextracted from each reactor by means of a syringe to pro-vide a headspace for gas sampling (therefore the oxygencontent in this headspace was very low). The water levelwas slightly above the gravel surface. The water volumein each reactor was estimated through gravimetric mea-surements and density considerations, and ranged mostly

between 18 and 21 mL. All the reactors were placed inthe dark in a controlled temperature chamber at 20 �C.The reactors were gently shaken every day. In another setof experiments the reactors were incubated at 5 �C.

The evolution of CO2 content in the headspace wasselected as an indicator of anaerobic biodegradation activ-ity because in previous trials no CH4 was detected. CO2 canalso be released from aerobic or anoxic processes, but thisis not the case in these experiments because the oxygen andnitrate content of the water was negligible. Gas sampleswere taken every 3 or 4 days during a period of 20 days.Every sampling day, one reactor of each type (near inlet,near outlet and blank) was taken and two gas samples of1 mL were obtained from the headspace by means of twosyringes and analysed for gas composition. Also, waterpH was measured from each reactor. After that, the reac-tors were rejected.

For the second type of tests (periodical measurements inthe same reactor) three flasks (balloon shape) of 2.2 L wereused as reactors. Each reactor was filled with approximately1.7 kg of a different type of gravel (the same three types usedfor the tests described above). After that, the reactors werefilled with 1.5 L of effluent. The water level was a few centi-meters above the gravel level. The air remaining in the head-space was removed by injecting helium during 15 min(therefore the amount of oxygen in the headspace was verylow). The reactors were closed with the same screw top witha microvalve used in the vials and were placed in the darkchamber at 20 �C. Gas samples (1 mL) from each reactorwere taken every 3–4 days during a period of 20 days.The reactors were gently shaken every day. Syringes withgas samples for the determination of CO2 were spiked tothe gas chromatographer used in the other experiments.For calculations, empty effects on the headspace were takeninto account when necessary. In another set of experimentsthe reactors were incubated at 5 �C.

2.4. Test with the influent

This experiment was done in order to check the reliabil-ity of the tests done with effluents. It was performed at20 �C and the methodology used was the same as describedabove for the vials (one measurement per reactor). One gassample was obtained daily during a period of 5 days. CO2

and CH4 were measured through gas chromatography.Note that in all anaerobic biodegradability tests it was

not necessary to add a base in order to compensate pHdecline because in previous trials it was observed that thepH decreased only 0.1–0.3 units. The initial pH in eachassay was that of effluent or influent samples.

3. Results

3.1. Field gravel gas emissions

Table 1 shows the values of the water quality parametersmeasured in SSF A1 and A2 during the two days of gravel

Table 1Physico-chemical properties of the influent and the effluents of SSF A1 and A2 during the two gas measuring days (M1 and M2) and during a long termevaluation

Influent A1 Effluent A2 Effluent A1 Efficiency (%) A2 Efficiency (%)

M1

Temperature (�C) 22.0 20.9 22.7 – –pH 7.5 7.4 7.3 – –COD (mg/L) 210 72 60 66 71NHþ4 (mg N/L) 59 58 50 2 15

M2

Temperature (�C) 24.2 19.8 20.9 – –pH na na na – –COD (mg/L) 420 130 100 69 76NHþ4 (mg N/L) 60 50 46 17 23

Long term evaluationa

COD (mg/L) 200 ± 82 83 ± 34 76 ± 30 59 62NHþ4 (mg N/L) 45 ± 16 34 ± 14 33 ± 12 24 27BOD5 (mg/L) 140 ± 54 61 ± 28 55 ± 25 56 60

na means non-available.a These data are the averages (±SD) of 103 grab water samples taken between 2001 and 2003 (Garcıa et al., 2004, 2005).

J. Garcıa et al. / Bioresource Technology 98 (2007) 3044–3052 3047

gas emission measurements (M1 and M2). Between the twodays the applied surface organic load changed from 7.6 to15.1 g COD/m2 day because of the increase in COD influ-ent concentration. As a result in M2 the COD removalefficiency was higher than in M1 in both systems but atthe same time the effluent quality was worse in terms ofCOD concentration.

The COD removal efficiency in the two measuring dayswas quite high and slightly greater in SSF A2 than in A1.The ammonia removal efficiency was also greater in SSFA2 than in A1 but low in both systems, as it is usuallyobserved in horizontal SSF (Garcıa et al., 2005). Theoverall efficiencies observed during the two days of gasmeasurements are quite consistent with the average efficien-cies of a long term evaluation carried out in these two SSF(Table 1); even taking into account that during that evalu-ation different hydraulic loading rates were used. The efflu-ents of these systems are characterised by undetectable TSScontent and high BOD5/COD relationship, indicating thatthe remaining organic matter is readily biodegradablethrough aerobic respiration (Garcıa et al., 2004, 2005).

SSF A1

Time, hours

0 1 2 3 4 5 6 70

1

2

3

4CO2M1

CO2M2

CH4M1

CH4M2

Gas

mas

s pe

r su

rfac

e ar

ea,

mm

ol /

m2

a

Fig. 1. Changes of mass gas per surface area vs. time in the two measuring dayLes Franqueses del Valles, Barcelona.

Fig. 1 shows the changes in time of CO2 and CH4 per unitof surface area in both SSF during the two measuring days.The mass variation of both gases clearly followed a lineartrend in all cases. Note that a few outliers are not shownbecause their deviation from the general linear trend (theseare errors of gas measurement that are usually linked toproblems with syringes). Outliers were detected consideringthat their residual values were 1.5 times greater than the dis-tance between the lower (25%) and upper (75%) quartiles ofthe residuals distribution in the linear regression. Table 2shows the gas mass emission rates which in fact are theslopes of linear regressions represented in Fig. 1. Outlierswere not used for the calculation of the emission rates.

The general increase in gas emission rates from M1 toM2 is linked with the rise in organic loading rate; however,the emissions of CH4 in the second measuring day in A1were lower with no apparent reason for this fact. SSF A2had higher emission rates than A1, with the mentionedexception in the second day for CH4. This trend is in agree-ment with the greater COD and ammonia removal efficien-cies observed in SSF A2.

Time, hours

0 1 2 3 4 5 6 70

1

2

3

4SSF A2

Gas

mas

s pe

r su

rfac

e ar

ea,

mm

ol /

m2

b

s (M1 and M2) in SSF A1 (a) and A2 (b) of the pilot plant of Can Suquet,

Table 2Gravel mass emission rates (in mmol/m2 h) estimated in the twomeasuring days (M1 and M2)

SSF A1 SSF A2

M1

CO2 0.106 (0.997) 0.353 (0.983)CH4 0.039 (0.999) 0.107 (0.951)

M2

CO2 0.221 (0.930) 0.464 (0.999)CH4 0.071 (0.923) 0.040 (0.781)

In parenthesis the linear correlation coefficient.

3048 J. Garcıa et al. / Bioresource Technology 98 (2007) 3044–3052

3.2. Anaerobic biodegradability tests

3.2.1. Tests with the effluents

Table 3 shows the values of the water parameters mea-sured in the effluents of SSF C2 used for the two experi-ments conducted at different temperature. As can be seenthe physico-chemical quality of the effluents of SSF C2 isquite similar to that observed for SSF A1 and A2, whichin addition is quite consistent with the average concentra-tions calculated from the long term evaluation (Garcıaet al., 2004, 2005).

Figs. 2 and 3 show the changes in time of the CO2 massper each mL of effluent water vs. time (20 days period) inthe vials (one measurement per reactor) and in the flasks(periodical measurements) respectively. CH4 was not

Table 3Physico-chemical properties of the effluents of SSF C2 used for anaerobictests performed under different temperatures

20 �C 5 �C Long term evaluationa

pH 7.0 7.1 naTSS (mg/L) nd nd naCOD (mg/L) 86 72 72 ± 31BOD5 (mg/L) na na 55 ± 28NHþ4 (mg N/L) 38 41 31 ± 12NO�3 (mg N/L) nd nd naSO2

4 (mg N/L) 22 36 na

nd means not detected. na means non-available.a These data are the averages (±SD) of 103 grab water samples taken

between 2001 and 2003 (Garcıa et al., 2004, 2005). Influent values arethose shown in Table 1.

20 oC

Time, days0 5 10 15 20

CO

2 m

ass

per

trea

ted

wat

er,

μmol

s/m

L

0.0

0.1

0.2

0.3

0.4

0.5Near inletNear outletBlank

a

Fig. 2. Changes in time of the CO2 mass emitted per volume of effluent vs. tim5 �C (b). Each day has two values of gas mass corresponding to the two gas s

detected during these experiments. Note that in previoustrials conducted with vials and effluent water (data notshown) it was observed that after 20 days there was a clearunderestimation of CO2 concentrations. This fact couldhad been related to the increasing pressure in the headspacewhich produced a certain gas leakage and/or to an increas-ing dissolution rate of the released CO2. Moreover, in theseprevious trials it was also observed that daily measure-ments during a short period (for example 5 days) was nota good method for detecting anaerobic activity, perhapsrelated with the low organic matter content of the SSFtreated effluents. In the case of flasks we also conductedprevious trials with effluent water (data not shown) andthe underestimation of CO2 noticed in the vials did notoccur. However, experiments with flasks were finished in20 days to be consistent with the procedure used in thevials.

Table 4 shows the CO2 mass emission rates which in factare the slopes of the linear regressions represented in Figs.2 and 3. Note that for the experiment made with the vials at20 �C, the data corresponding to 20 days was not used forthe linear regressions of near inlet and outlet. These dataseemed to be underestimated because the increasing pres-sure in the headspace (Fig. 2), as it was observed in previ-ous trials for reactors incubated during more than 20 days.For near the outlet, the values obtained at 13 days neitherwere used because of their great deviation from the generallinear trend (Fig. 2). There is no apparent reason for thisclear deviation but could be a reactor-specific trend. Allthese outliers were detected considering that their residualvalues were 1.5 times greater than the distance betweenthe lower (25%) and upper (75%) quartiles of the residualsdistribution in the linear regression. In the case of theexperiment made with the flasks there were no outliersand all data were therefore used for linear regressions.

Table 5 shows the grouped CO2 mass emission rates thatallow a better evaluation of the experimental results. Therate of emission of CO2 was in general quite similar nearthe inlet and the outlet. On the other hand, the rate inthe blank was lower. Emissions were lower at 5 �C thanat 20 �C; however, differences were very small. Further-more, emissions were slightly lower in the flasks (periodicalmeasurements in one reactor) than in the vials (one mea-surement per reactor).

5 oC

CO

2 m

ass

per

trea

ted

wat

er,

μmol

s/m

L

Time, days0 5 10 15 20

0.0

0.1

0.2

0.3

0.4

0.5b

e in the vials (one measurement per reactor) at temperatures of 20 (a) andamples obtained.

20 oC 5 oC

Time, days0 5 10 15 20C

O2 m

ass

per

trea

ted

wat

er,

μmol

s/m

L0.0

0.1

0.2

0.3

0.4

0.5Near inletNear outletBlank

CO

2 m

ass

per

trea

ted

wat

er,

μmol

s/m

L

Time, days0 5 10 15 20

0.0

0.1

0.2

0.3

0.4

0.5ba

Fig. 3. Changes in time of the CO2 mass emitted per volume of effluent vs. time in the flasks (periodical measurements) at temperatures of 20 (a) and 5 �C(b).

Table 4CO2 mass emission rates (lmol/mL day) obtained from the anaerobic tests of the effluents of SSF C2

Gravel origin Vials 20 �C Flasks 20 �C Vials 5 �C Flasks 5 �C

Near inlet 0.011 (0.971) 0.010 (0.760) 0.012 (0.865) 0.005 (0.963)Near outlet 0.015 (0.942) 0.008 (0.894) 0.008 (0.927) 0.006 (0.980)Blank 0.004 (0.813) 0.003 (0.859) 0.002 (0.573) na

In parenthesis the linear correlation coefficient.na means non-available.

Table 5Averages (±SD) of the CO2 mass emission rates (lmol/mL day) shown in Table 4 grouped for the different factors studied

Gravel origin Overall 20 �C 5 �C Vials Flasks

Near inlet 0.010 ± 0.003 0.011 ± 0.001 0.009 ± 0.005 0.012 ± 0.001 0.008 ± 0.004Near outlet 0.009 ± 0.004 0.012 ± 0.005 0.007 ± 0.001 0.012 ± 0.005 0.007 ± 0.001Blank 0.003 ± 0.002 0.004 ± 0.001 0.002 0.003 ± 0.001 0.003

J. Garcıa et al. / Bioresource Technology 98 (2007) 3044–3052 3049

During the experiments with vials, the pH of the waterwas measured and almost did not change from the initialvalue. At the end of the experiments, and for all tests,pH ranged between 6.8 and 7.3.

3.2.2. Test with the influent

The SSF influent used in this experiments had thefollowing characteristics: pH = 7.3, TSS = 59 mg/L,COD = 140 mg/L, NHþ4 ¼ 37 mg N/L and SO2�

4 ¼ 72mg/L. Fig. 4 shows the changes in time of the CO2 andCH4 mass produced per each mL of water vs. time (5 daysperiod) in the reactors (vials). In previous trials with influ-

Influent

Time, days0 1 2 3 4 5G

as m

ass

per

wat

er v

olum

e,μm

ol/m

L

0.00

0.05

0.10

0.15

0.20

0.25CO2 Near inlet

CO2 Near outlet

CH4 Near inlet

CH4 Near outlet

Fig. 4. Changes in time of the gas mass emitted per volume of influent vs.time in the vials (one measurement per reactor) at a temperature of 20 �C.

ent water it was observed that after approximately oneweek the concentration of the gases was underestimated(data not shown). Note that the reactor of the fifth dayfilled with gravel from the outlet was broken (Fig. 4). Inopposition to the tests done with the effluents, in the influ-ent it was observed CH4 production. CH4 emission rateswere 0.008 and 0.003 lmol/mL day for near the inlet andthe outlet respectively (linear correlation coefficient of0.866 and 0.996). CO2 emission rates were approximately2–3 times greater than those observed in the effluents:0.037 and 0.016 lmol/mL day for near the inlet and theoutlet respectively (linear correlation coefficient of 0.966and 0.919).

4. Discussion

In the present study it has been observed that CO2 andCH4 emission rates obtained from field measurements inSSF vary with a clearly linear trend during a period of 4–7 h. This linear trend was already seen in previous studieswhere gas release by diffusive processes and ebullitionwas measured (Brix, 1990; Tanner et al., 1997). In naturalwetlands it has been observed that diffusive and ebullitionrelease processes occur at nearly a constant rate duringall day long (Brix et al., 2001). The results of the presentstudy indicate that the same trend occur in SSF. The

3050 J. Garcıa et al. / Bioresource Technology 98 (2007) 3044–3052

short-term constant rate gas emissions in SSF contrastswith the long-term variations in gas emissions, such canbe the seasonal pattern of CH4 release with significantlyhigher values in summer (Tanner et al., 1997).

Differences in gas emissions rates observed in the presentstudy between SSF A1 and A2 are quite important (Table2) and do not seem to be explained by the slightly greaterefficiency of A2, that in fact only removes 5–7% moreCOD (Table 1). This trend could be related to the fact thatthe rates of the biochemical reactions involved in theremoval of organic matter in SSF change in space due tolocal site-specific properties (Kadlec, 2003). Furthermore,in this study the two tested SSF for emissions had differentsize media that could contribute to these different rates.

Gravel gas mass emission rates observed in the presentstudy are somewhat lower than those previously reported(Table 6), even taking into account that our measurementswere done during summer when the emissions are supposedto be the greatest. In addition to seasonal considerations,the organic surface loading rate was higher in the presentstudy than in the other works cited in Table 6. Kadlecand Knight (1996) indicated that organic carbon load isan important factor that affects CH4 emissions. These dif-ferences put into evidence the difficulties for comparinggas emission data obtained from different SSF systems.In addition to seasonal variations and the influence ofthe load, gravel gas emissions also depend on the qualityof the organic matter (for example the proportion of partic-ulate and dissolved, or the amount of short chained fattyacids), and in relation with this, the location for measure-ment. Tanner et al. (1997) and Mander et al. (2003) haveobserved that CH4 emissions were higher near the inletthan near the outlet. CH4 emissions may also depend onthe rate of CH4 oxidation near the plant rhizosphere andthe surface of the water due to methanotrophic activity.In fact, van der Nat and Middelburg (1998) reported thatthe reeds reduced the emissions in a 16% during the grow-ing season in planted containers with tap and de-ionisedwater. The lower gravel gas emissions in the current studyin comparison to the previous studies could be related withthe rather low removal efficiency shown by SSF A1 andA2.

The amount of carbon removed that represented gravelCH4 and CO2 emissions can be estimated from COD datain Table 1 and gravel gas emission rates in Table 2. Forthese calculations it is assumed that all organic matter

Table 6Gravel gas emission rates measured in different studies

References CH4 (mmol/m2 day)

Brix (1990) 19Tanner et al. (1997) 3–30Mander et al. (2003) 0.003–175This study 0.94–2.6

It is also indicated the organic surface loading rate, which in our case has bee(WEF, 1992).

a Based on BOD7.

removed was in the form of an easily biodegradable sub-strate, such as glucose for example. The results of the cal-culations indicate that gravel CH4 and CO2 emissionsrepresent as much as 1.5% and 5% of the carbon removedrespectively. If it is considered that a half of the organicload is physically retained near the inlet in the form of inertor slowly biodegradable suspended solids, then gravel CH4

and CO2 emissions still represent a low amount of thecarbon entering the system (as much as 5% and 17%respectively). The reason for this small amounts could beexplained by the fact that for these calculations it isassumed that the rates of the biochemical reactions areuniformly spatially distributed, which is not case (Garcıaet al., 2005), even for the soluble COD fraction. Tanneret al. (1997) estimated that global CH4 emissions (includingplant mediated and diffusion and ebullition) representedbetween 2% and 8% of the carbon load.

In the anaerobic tests with the effluents, CO2 had to beconsidered as an indicator activity because CH4 was notdetected (Soto et al., 1993). On the other hand, in the testwith the influent, CH4 production was well observed. Thisindicates that the method used to test the anaerobic activityis valid and the lack of CH4 production and the lower CO2

generation rate in assays with effluents is related to theanaerobic refractory nature of their organic matter.

Although CO2 is not commonly used as indicator of theanaerobic activity, it has been monitored for the control ofthe acidogenic phase (Soto et al., 1993; Garcıa-Moraleset al., 1996). On the other hand, anoxic conditions inalmost the entire water depth of SSF allow the develop-ment of microbial anaerobic reactions, including sulphatereduction and methanogenesis (Baptista, 2003). CO2 is afinal product of all these reactions, and also it is producedin intermediate anaerobic reactions. Thus, taking intoaccount the low anaerobic biodegradation rates showedby the effluents in the current study, it seems reasonablefor SSF gravel to in addition to CH4 (if detected), use alsoCO2 as indicator of activity.

In fact, the following carbon mass balance may beestablished for anaerobic and anoxic processes: Cinitial =Cremaining + Cbiomass + CCO2

þ CCH4, meaning that the

carbon removed is recovered as biomass, carbon dioxideand/or methane. In this manner, some assays dealingwith methanogenic activity determination are based onmeasures of the total (CO2 + CH4) biogas generation(Soto et al., 1993; Ruiz et al., 1998). The drawback of this

CO2 (mmol/m2 day) Loading rate (g BOD/m2 day)

105 3.7– 2.5–4.2– 0.95a

2.5–11 5.4–10.8

n estimated from data in Table 1 and considering that COD = 1.4BOD5

J. Garcıa et al. / Bioresource Technology 98 (2007) 3044–3052 3051

procedure is derived from the fact that part of CO2 gener-ated may be retained in the liquid phase if pH increases,and therefore offering lower than actual emission rates(Sperandio and Paul, 1997). On the other hand, a pHdecrease leads to a greater release of CO2 from liquid phaseand then causes an overestimation of the generation rates.So, low pH variations are required through the assaycourse. In these conditions, CO2 evolution can be used asan indicator of anaerobic biological activity, as was previ-ously used in aerobic biodegradation assays (Buitron et al.,1993; Struus and van den Berg, 1995). In the experimentswith the vials we have observed little variation of the pHduring time, and therefore CO2 can be used as a indicatorof anaerobic activity.

In the tests with effluents the anaerobic activity wasslightly lower in the flasks than in the vials. This could berelated to the fact that small reactors like the vials (in com-parison to the flasks) reduce mass transfer resistances, inspecial considering that during the experiments it wasobserved a certain detachment of the biofilm that remainedonto the top of the gravel surface.

The rate of emission of CO2 was quite similar fromgravel samples obtained near the inlet and the outlet, sug-gesting that the anaerobic activity of the biofilm growing inthe different gravel samples was similar when the organicmatter of the effluent was used as substrate. Note that infact the capacity of the biofilm to anaerobically degradethe organic matter was indicated by the rate differencesbetween the blanks and the other samples. Anaerobic reac-tion rates depend on the S0/X0 ratio (Moreno et al., 1999),and because S0 was the same in the tests with gravel fromthe inlet and the outlet, it could be reasonably assumed thatX0 was also similar in the inlet and the outlet gravel sam-ples. Therefore, for this type of anaerobic assays testingeffluent biodegradation rate, the place in SSF from wherethe gravel is collected is not relevant (at least for wetlandsoperating with a similar organic load and time like thosetested in the present study). Nevertheless, gravel samplesfrom very near to the inlet (at less than 1/4 of fractionaldistance) should be avoided because of the retained TSS(Garcıa et al., 2005).

The results of CO2 fluxes measured in the field are quiteconsistent with the emissions estimated in anaerobic biode-gradability experiments. In the case with the highest sur-face CO2 emission rate (A2 in the second period), theestimated rate per volume of water was 0.052 lmol/mL day(calculated taking into consideration average water depthand porosity), while in lowest it was 0.012 lmol/mL day(A1 in the first period). These values were of the same orderof magnitude as those found in the effluent anaerobic bio-degradability tests. Anaerobic activity was lower at 5 �Cthan at 20 �C; however, differences were small and theyare not in agreement with the kinetic models that relatetemperature and anaerobic activity (Ruiz et al., 1998). Thissuggests that assays were carried out with a very low ratioof biodegradable substrate to biomass, so substrate was the

activity limiting factor, and masked the effect oftemperature.

The anaerobic biodegradation rates of the tested effluentis very low. In all the experiments CH4 was not producedand CO2 emissions were clearly lower than in influentassays. This indicates that organic matter contained ineffluent of this SSF is refractory to anaerobic degradation(Field et al., 1988). This conclusion is in agreement withthe low removal efficiency shown by these systems. Thetwo studied SSF for emissions have shown a rather lowremoval efficiency over a long term evaluation, withBOD5 normally above 25 mg/L (Garcıa et al., 2005). Evenwhen the hydraulic loading rate was decreased from 45 to20 mm/d, then the BOD5 reduced in average approxi-mately 20 mg/L, but still the concentration was above25 mg/L. This is due to the anaerobic refractory propertiesof the effluent organic matter. Other factors that could berelated with the low anaerobic biodegradation rates arethose inhibiting the anaerobic microbial activity, such canbe high sulphide concentration or pH.

Most of the organic matter is removed anaerobically inthe long sewer and the primary treatment of the pilot sys-tem, and this explains the low COD content in the influentin comparison to other Spanish Mediterranean sites (Gar-cıa et al., 2004, 2005). In the predominantly anaerobic con-ditions of the SSF, the organic matter content can not bereduced so much because it has been already removed pre-viously. This pattern is also consistent with the fact that theremoval of organic matter occurs near the inlet (becauseTSS retention) and little further removal occurs (Kadlecand Knight, 1996). In previous studies done with the samepilot SSF it has been observed that shallower SSF (0.3 m)are more efficient than deep ones (0.5 m) because thegreater aerobic conditions in the shallower wetlands (Gar-cıa et al., 2004, 2005). This trend also is consistent with theresults of this study because organic matter present in theeffluents can be readily removed by aerobic respiration(as shown by the high BOD5/COD relationship), and notby anaerobic pathways.

5. Conclusions

The results of this study indicate that anaerobic testswith the gravel of SSF can be successfully accomplishedwith reactors of different kind in order to study SSF efflu-ent anaerobic biodegradability. Each type of reactor usedin this investigation (vials or flasks) has advantages as wellas disadvantages. Vials (one reactor for measurement)require more time for preparation but occupy less space,and replicates can be done easily. In addition, vials showhigher gas emissions rates than flasks, which is probablyrelated to their lower volume that reduce mass transferresistances. The main disadvantage of vials is the underes-timation of gas concentration after an unknown period oftime, that seems to be related to the increase in pressure inthe headspace.

3052 J. Garcıa et al. / Bioresource Technology 98 (2007) 3044–3052

In the present study CO2 had to be considered as indica-tor of anaerobic activity because CH4 was not detected inthe tests with SSF effluents. CO2 is an intermediate andfinal product of a variety of microbial anaerobic reactionsthat take place in SSF. Thus, taking into account the lowanaerobic biodegradation rates showed by the effluents(0.005–0.015 lmol CO2/mL day), it seems reasonable forSSF gravel to in addition to CH4 (if detected), use alsoCO2 as indicator of activity. The results of CO2 fluxes mea-sured in the field were quite consistent with the emissionsestimated in anaerobic biodegradability experiments.

The anaerobic biodegradation rates of the tested efflu-ents were very low in relation with the anaerobic refractoryproperties of the organic matter. This conclusion is inagreement with the rather low removal efficiency shownby the studied SSF (for COD and ammonia usually<70% and <30% respectively).

Acknowledgements

This study was possible to the technical support of PaulaAguirre and Eduardo Alvarez. In addition, the contribu-tion of the Consorci per a la Defensa dels Rius de la Conca

del Besos and Les Franqueses del Valles Town Council wasvery helpful. Diederik Rousseau kindly made constructivecomments on the paper. We are grateful to all of them.The results of this study were obtained thanks to a grantawarded by the Spanish Department of Education and Sci-ence, Research Project REN2002-04113-C03-03. The firstauthor also thanks the Spanish Department of Educationand Science for the grant received to make a research stagein Ghent University (Belgium).

References

Aguirre, P., Ojeda, E., Garcıa, J., Barragan, J., Mujeriego, R., 2005. Effectof water depth on the removal of organic matter in horizontalsubsurface flow constructed wetlands. J. Environ. Sci. Health 40,1457–1466.

APHA-AWWA-WPCF, 1995. Standard Methods for the Examination ofWater and Wastewater, nineteenth ed., American Public HealthAssociation, Washington, DC.

Baptista, J.D.C., 2003. Microbial communities in subsurface flowwetlands. In: Dias, V., Vymazal, J., (Eds.), The Use of AquaticMacrophytes for Wastewater Treatment in Constructed Wetlands,Articles of the 1st International Seminar, Lisboa, Portugal, May 8–10,Fundacao Calouste Gulbenkian, Lisboa, pp. 265–276.

Brix, H., 1990. Gas exchange through the soil-atmosphere interphase andthrough dead culms of Phragmites australis in a constructed reed bedreceiving domestic sewage. Water Res. 24, 259–266.

Brix, H., Sorrell, B.K., Lorenzen, B., 2001. Are Phragmites-dominatedwetlands a net source or net sink of greenhouse gases? Aquat. Bot. 69,313–324.

Buitron, G., Koefoed, A., Capdeville, B., 1993. Control of phenolbiodegradation by using CO2 evolution rate as an activity indicator.Environ. Tech. 44, 227–236.

Burgoon, P.S., Reddy, K.R., Debusk, T.A., 1995. Performance ofsubsurface flow wetlands with batch-load and continuous-flow condi-tions. Water Environ. Res. 67, 855–862.

Field, J., Alvarez, R.S., Lettinga, G., 1988. Anaerobic tests. In: 4thSeminar on Anaerobic Degradation in Wastewater. University ofValladolid, Spain, 53–75 (in Spanish).

Garcıa, J., Ojeda, E., Sales, E., Chico, F., Pıriz, T., Aguirre, P., Mujeriego,R., 2003. Spatial variations of temperature, redox potential, andcontaminants in horizontal flow reed beds. Ecol. Eng. 21, 129–142.

Garcıa, J., Aguirre, P., Mujeriego, R., Huang, Y., Ortiz, L., Bayona, J.M.,2004. Initial contaminant performance factors in horizontal flow reedbeds used for treating urban wastewater. Water Res. 38, 1669–1678.

Garcıa, J., Aguirre, P., Barragan, J., Mujeriego, R., Matamoros, V.,Bayona, J.M., 2005. Effect of key design parameters on the efficiency ofhorizontal subsurface flow constructed wetlands: long-term perfor-mance pilot study. Ecol. Eng. 25, 405–418.

Garcıa-Morales, J.L., Nebot, E., Cantoral, J.M., Romero, L.I., Sales, D.,1996. Methanogenic and acidogenic activity test in an anaerobicthermophilic reactor. Biotechnol. Tech. 10, 249–252.

Kadlec, R.H., 2003. Effects of pollutant speciation in treatment wetlandsdesign. Ecol. Eng. 20, 1–16.

Kadlec, R.H., Knight, R.L., 1996. Treatment Wetlands. CRC Press, BocaRaton, FL.

Kadlec, R.H., Knight, R.L., Vymazal, J., Brix, H., Cooper, P., Haberl, R.2000., Constructed Wetlands for Pollution Control: Processes, Perfor-mance, Design and Operation, IWA Specialist Group on Use ofMacrophytes in Water Pollution Control, Scientific and TechnicalReport 8, IWA Publishing, London, UK.

Mander, U., Kuusemets, V., Lohmus, K., Mauring, T., Teiter, S.,Augustin, J., 2003. Nitrous oxide, dinitrogen and methane emissionin a subsurface flow constructed wetland. Water Sci. Tech. 48 (5), 135–142.

Moreno, G., Cruz, A., Buitron, G., 1999. Influence of S0/X0 ratio andmedium composition on anaerobic biodegradability test. In: Proceed-ings of the 52th Purdue Industrial Waste Conference. Ann HarborPress, Michigan, pp. 25–133 (Chap. 14).

Ruiz, I., Soto, M., Veiga, M.C., Ligero, P., Vega, A., Blazquez, R., 1998.Performance of and biomass characterisation in a UASB reactortreating domestic waste water at ambient temperature. Water SA 24,215–222.

Soto, M., Mendez, R., Lema, J.M., 1993. Methanogenic and non-methanogenic activity test. Theoretical basis and experimental set up.Water Res. 27, 1361–1376.

Sperandio, M., Paul, E., 1997. Determination of carbon dioxide evolutionrate using on-line gas analysis during dynamic biodegradation exper-iments. Biotechnol. Bioeng. 53, 243–252.

Struus, J., van den Berg, R., 1995. Standardized biodegradability tests:extrapolation to aerobic environments. Water Res. 29, 255–262.

Tanner, C.C., Adams, D.D., Downes, M.T., 1997. Methane emissionsfrom constructed wetlands treating agricultural wastewaters. J. Envi-ron. Qual. 26, 1056–1062.

van der Nat, F.J.W.A., Middelburg, J.J., 1998. Seasonal variation inmethane oxidation by the rhizosphere of Phragmites australis andScirpus lacustris. Aquat. Bot. 61, 95–110.

WEF, 1992. Design of Municipal Wastewater Treatment Plants, vol. 1,Manual of Practice 8, ASCE Manual and Report on EngineeringPractice 76.