You are on page 1of 9

e c o l o g i c a l e n g i n e e r i n g 3 5 ( 2 0 0 9 ) 175–183

available at www.sciencedirect.com

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

A comparative study of surface and subsurface flow


constructed wetlands for treatment of combined sewer
overflows: A greenhouse experiment

Annelies M.K. Van de Moortel a,∗ , Diederik P.L. Rousseau b ,


Filip M.G. Tack a , Niels De Pauw c
a Department of Applied Analytical and Physical Chemistry, Laboratory of Analytical Chemistry and Applied Ecochemistry, Ghent
University, Coupure Links 653, 9000 Ghent, Belgium
b Department of Environmental Resources, UNESCO-IHE Institute for Water Education, P.O. Box 3015, 2601 DA Delft, The Netherlands
c Department of Applied Ecology and Environmental Biology, Laboratory of Environmental Toxicology and Aquatic Ecology, Ghent

University, J. Plateaustraat 22, 9000 Ghent, Belgium

a r t i c l e i n f o a b s t r a c t

Article history: The use of surface flow (SFCWs) and subsurface flow constructed wetlands (SFCWs) for the
Received 7 December 2007 treatment of combined sewer overflows was assessed at pilot scale. Synthetic wastewa-
Received in revised form ter was applied in three batches with decreasing concentrations to mimic concentration
27 August 2008 profiles that are obtained in the field during overflow events. Three simulated combined
Accepted 29 August 2008 sewer overflows were applied on each wetland. Composite water samples (60 in total) were
taken for a period of 8 days to study the removal of total nitrogen (Ntot), NH4 –N, NO3 –N,
total COD (CODtot) and total phosphorus. Redox potential, which was monitored at vari-
Keywords: ous locations along the wetlands, was more negative in the SSFCWs. In general, removal
Waste water treatment occurred faster in the SSFCWs and the final concentrations were lower. The removal of Ntot
Ammonium was only 36.6 ± 3.3% in the SFCWs due to nitrification-limiting conditions. The conditions in
Nitrogen the SSFCWs, in contrast, seemed to promote Ntot removal (removal efficiency 96.7 ± 1.9%).
Phosphorus The removal of P was hampered in both wetland types by reducing conditions. P that was
Phragmites initially removed was released again from the substrates later on. First-order removal rate
Fractionation constants were derived for the removal of both CODtot (SSFCWs: 1.1 ± 0.3 m d−1 ; SFCWs:
Treatment wetlands 0.17 ± 0.06 m d−1 ) and Ntot (SSFCWs: 0.4 ± 0.1 m d−1 ; SFCWs: 1.7 ± 0.5 m d−1 ).
Stormwater © 2008 Elsevier B.V. All rights reserved.
k–C* model
First-order removal

1. Introduction nearest surface water, thereby potentially causing a sudden


pollution shock. The study of devices applicable for the tempo-
During rain events, the capacity of combined sewer systems ral storage and treatment of stormwater and combined sewer
may be too confined to allow transport of all waste- and overflows has been pointed out as one of the 10 priority areas
stormwater to a treatment plant. Combined sewer overflows in urban wet weather flow research (Heaney et al., 1999). The
(CSOs) are provided for diverting the excess water to the occurrence of uncontrolled or poorly controlled discharges


Corresponding author. Tel.: +32 9 264 59 95.
E-mail addresses: Annelies.Vandemoortel@UGent.be (A.M.K. Van de Moortel), d.rousseau@unesco-ihe.org (D.P.L. Rousseau).
0925-8574/$ – see front matter © 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.ecoleng.2008.08.015
176 e c o l o g i c a l e n g i n e e r i n g 3 5 ( 2 0 0 9 ) 175–183

from CSOs and stormwater is one of the major reasons for


Table 1 – Composition of the applied synthetic waste
the long-term persistence of the poor water quality of surface water with concentration C1 (modified after Boeije et al.,
waters (Suarez and Puertas, 2005; Robson et al., 2005). 1998).
The use of constructed wetlands for treatment of domestic
Compound mg l−1
wastewater has been studied intensively (Kadlec and Knight,
1996; Scholz et al., 2002; Mungasavalli and Viraraghavan, 2006; MgCl2 51.0
Vymazal, 2007). More recently wetlands have been used for CaCl2 107.5
Urea 58.6
the retention and treatment of all kinds of urban wet weather
NH4 Cl 15.9
flows, including runoff from airports, parking lots, agricultural Peptone 30.6
runoff, combined sewer overflows and stormwater in general MgHPO·3H2 O 22.1
(Revitt et al., 2001; Debo and Reese, 2003; Fink and Mitsch, KH2 PO4 19.9
2005). Surface flow wetlands (Revitt et al., 2001; Greenway et FeSO4 ·7H2 O 4.9
al., 2006) and horizontal subsurface flow wetlands (Revitt et Starch 198.6
Milk powder 198.6
al., 2001, 2004; Shutes et al., 1999, 1997; Geary et al., 2006) have
Yeast 91.3
both been used for the treatment of urban wet weather flows.
Cr(NO3 )3 ·9H2 O 0.49
Emerging research also focuses on the use of vertical flow con- CuCl2 ·2H2 O 1.13
structed wetlands for the treatment of CSOs (Dittmer et al., MnSO4 ·H2 O 0.07
2005; Frechen et al., 2006; Welker, 2006). NiSO4 ·6H2 O 0.22
The removal of pollutants from water in constructed PbCl2 0.45
wetlands is achieved by an interactive combination of phys- ZnCl2 0.57
CdCl2 ·2H2 O 0.05
ical, chemical and biological processes as determined by
the physico-chemical conditions of the wetlands. Wong and
Geiger (1997) suggested that the k–C* model could be adapted
for the description of pollutant removal in wetlands treating filled with gravel to a height of 0.35 m; an inlet and outlet zone
stormwater. They suggested that the parameters in the k–C* was created with coarse gravel (16–32 mm) having a porosity (ε)
model would differ between wetlands used for treatment of of 0.43; the remainder of the beds was filled with finer gravel
wastewater and stormwater treatment due to the unsteady, (3–8 mm; ε = 0.38). Water levels during the experiments were
intermittent nature of the inflow of stormwater. Carleton et al. kept at a height of 0.35 m in the SSFCWs. Both types of wet-
(2001) investigated data from 35 studies on 45 wetlands used lands were planted with seedlings of common reed (Phragmites
for stormwater treatment and found that first-order removal australis) at a plant density of 25 plants m−2 (the SSFCWs in
rate constants for total phosphorus, ammonia and nitrate January 2005, the SFCWs in July 2006). The aboveground plant
were similar to values reported in the literature by Kadlec and parts of the SSFCWs were removed on a regular basis from the
Knight (1996) for wastewater treatment wetlands. Dittmer et wetland. Coverage for both types of wetland was over 90% at
al. (2005) described how a model, developed for the simula- the time of the experiments. However, the reed plants in the
tion of the treatment of municipal wastewater could be used SSFCWs formed more shoots and the shoots were stronger
to describe the removal of pollutants in a vertical flow con- and thicker. Shoot length in both wetland types was between
structed wetland used for treatment of CSOs in spite of the 1.2 and 1.5 m.
stochastic nature of the loading and loading regime of the The experiments were carried out from September to
incoming water. Based upon existing data, models for domes- November 2006 and November 2006 to February 2007. The
tic waste water can also be used for the removal of pollutants mean temperature during the tests with the SFCWs and
in CSOs without large modifications. SSFCWs was, respectively, 8 and 14 ◦ C. No artificial illumina-
The use of constructed wetlands represents a relatively tion was provided during this period. The experiments would
new approach for stormwater treatment (Dittmer et al., 2005; ideally have been conducted in the spring or summer but
Mungasavalli and Viraraghavan, 2006; Line et al., 2008). In this timing was influenced by factors beyond the control of the
study, two types of pilot-scale constructed wetlands, surface authors. All systems were batch-fed with synthetic wastewa-
(SFCWs) and subsurface flow (SSFCWs), were assessed for CSO ter (modified from Boeije et al., 1998) (Table 1). Each simulated
treatment. Besides an evaluation of the performance of both overflow event consisted of three successive batches with dif-
wetland types, emphasis was put on the difference in physico- ferent theoretical concentrations of pollutants (C1 ; C2 = 1/2C1
chemical conditions in the wetlands and their influence on and C3 = 1/4C1 ), which were successively introduced into the
the removal of nitrogen, phosphorous and chemical oxygen constructed wetlands. Dilution of the synthetic wastewater
demand (COD). was done by tapwater. The diluted synthetic wastewater was
allowed to stand for 30 min under continuous stirring. The
tapwater had a NO3 –N concentration of 3.4 ± 0.1 mg NO3 –N
2. Materials and methods l−1 . The applied concentrations were based on concentrations
measured in a full-scale CSO in Bornem, Belgium. The vol-
Four pilot-scale constructed wetlands were set up in a green- ume of one batch was 76 ± −4l, an overflow event (i.e. three
house, i.e. two replicate SFCWs (2.4 m × 0.3 m × 0.6 m) and two successive batches) corresponded with 227 ± −6l. The three
replicate SSFCWs (2.5 m × 0.625 m × 0.50 m). The SFCWs were different concentrations simulated the typically decreasing
filled with a 10 cm thick soil layer and were filled with water pollutant concentrations that enter a wetland during a sewer
to a height of 40 cm (water height of 30 cm). The SSFCWs were overflow event. In the SFCWs three tests with an overflow
e c o l o g i c a l e n g i n e e r i n g 3 5 ( 2 0 0 9 ) 175–183 177

event were carried out, in the SSFCWs two tests were per- level of 5% was adapted to evaluate the significance in the
formed. In between two experiments the wetlands were fed different tests.
with tapwater to prevent drying-up. Contrary to the SSFCWs, For plotting concentration profiles as a function of time
the water in the SFCWs was drained before introducing a concentrations obtained for the different locations along the
new overflow. The water present in the SSFCWs was forced length of the constructed wetlands were pooled together. This
out of the wetlands when a new overflow was added. As such was done because of following assumption: when a new over-
the added water replaced the water that was still present flow event enters the wetland, the water that is still present
in the SSFCWs. As tapwater was used in the experiments, in the wetland will be mixed and forced out of the wetland
the effect of chlorine present in the tapwater on the biofilm as a whole. As such the mean concentration of pollutants
needs to be evaluated. The available chlorine will react with present at that time in the wetland will be the only relevant
the organic material and biofilm present in the reedbeds, but concentration.
only a small part of it will be oxidised by it due to excess of Removal efficiencies were calculated based on the amount
organic material present in the wetlands. Furthermore, if any of a certain element that was introduced to the wetlands and
negative effect was present, this would be limited to the inlet the amount that was recovered after 8 days. Concentration
zone of the wetlands. As such, we concluded that is was safe profiles describing the removal of the pollutants as a function
to use tapwater for dilution and irrigation. of time in both wetland types were fitted to the first-order k–C*
Water samples were taken from each waste water batch model (Kadlec and Knight, 1996):
before application in the constructed wetlands. In the beds
Ct − C∗
themselves, water samples were collected at different time = e−kv t
C0 − C∗
intervals (day 0, 1, 3, 6 and 8) and at various distances from
the inlet (SFCWs: 0.3; 0.95; 1.55 and 2.2 m—SSFCWs: 0.2; 0.8; where Ct the average concentration of the pollutant at time
1.4 and 2.3 m). For the SSFCWs an additional sampling event t (mg l−1 ), C* the average background concentration of the pol-
after 12 h was included to evaluate the short-term removal. lutant (mg l−1 ), C0 the initial concentration of the pollutant
The time interval between application of the first batch and (mg l−1 ), kv the volumetric removal rate constant (d−1 ), and t
collection of the samples on day 0 was approximately 4 h. In the treatment time (d).
the SFCWs water samples were taken at a depth of 15 cm below First-order rate removal constants and background concen-
the water surface. For the SSFCWs water samples were taken trations were obtained for the removal of Ntot and CODtot for
at depths of 15 and 30 cm. Water samples were taken with a both constructed wetland types. To differentiate between rate
silicone tube and syringe and the samples from both depths constants for ‘fast’ and ‘slow’ removal, influent concentrations
taken in the SSFCWs were pooled into one combined sample were either included or excluded. The kv -value obtained by
for further processing. excluding the influent concentrations was set as kv slow . kv fast
Determination of NH4 –N was performed immediately after was calculated as the difference between the kv value obtained
sampling. To preserve samples for Ntot analysis, H2 SO4 was when including the influent concentrations, and kv slow . Areal
added to a pH lower than 2. Both NH4 –N and Ntot were removal rate constants (ka ) were also calculated as ka = kv εh
determined by distillation following procedures described in with h the water height (0.35 and 0.30 m for the SSFs and
Standard methods for the examination of water and waste SFCWs respectively) and the porosity ε (0.42 and 0.95 for the
water (Eaton, 1995). Nitrate-concentrations were determined SSFCWs and SFCWs, respectively).
immediately by means of anion chromatography (Chromato-
graph 761 and Metrosep A-column, Methrohm, Switzerland).
Total phosphorus content was determined by colorimetry 3. Results
using a Jenway 6500 spectrophotometer (Jenway, Essex, UK)
after digestion with potassium persulphate as described by There were no significant differences between the two repli-
Eaton (1995). The redox potential was measured with com- cates of each wetland type and this for all considered
bined platina and gel reference electrodes (HI 3090 B/5; HI 9025, pollutants (p < 0.05). After 1 day the redox potential in the
Hanna Instruments, Ann Arbor, USA) that were permanently SSFCWs stabilised and remained stable for the further dura-
present in the constructed wetlands at a depth of 0.15 m in the tion of the experiment. The redox potential increased towards
water column and at the following distances from the inlet: the outlet zone from −200 to −70 mV whereas for the SFCWs
0.3; 0.95; 1.55 and 2.2 m for SFCWs, and 0.2; 0.8; 1.4 and 2.3 m no differences were observed along the length profile of the
for SSFCWs. Chemical oxygen demand (COD) was determined beds. In the SFCWs the redox potential showed a less rapid
using Nanocolor test kits (Macherey-Nagel, Düren, Germany). decline, but reached a similar value (−232 ± 35 mV) after day
For overflow event 1 in the SSFCWs and event 2 in the SFCWs, 3. The redox potential increased up to − 99 ± 20 mV on day 8.
water samples were filtered over 0.45 ␮m pore size filters. This Removal of Ntot occurred faster in the SSFCWs than in the
allowed partitioning total COD (CODtot, measured on unfil- SFCWs (Fig. 1) with Ntot in the SSFCWs decreasing to con-
tered samples) into soluble COD (CODsol, 0.45 ␮m filtered) centrations of 0.90 ± 0.44 mg l−1 Ntot after 8 days. The average
and particulate COD (CODpart, calculated as the difference removal efficiency for Ntot in the SSFCWs and SFCWs was
between CODtot and CODsol). 96.7 ± 1.9% and 36.6 ± 3.3% respectively (Table 2). Although
All statistical analyses were done with SPSS 15.0 (SPSS, the global N-removal in the SFCWs was low, nitrogen forms
Chicago, USA). Two-sample t-tests were used to evaluate the with time shifted from mainly organic and NO3 –N to domi-
significance of differences between the two types of con- nantly NH4 –N (Fig. 2). The NH4 –N supplied with the influent
structed wetlands and between the replicates. A confidence was not detected in the SSFCWs, indicating a fast removal
178 e c o l o g i c a l e n g i n e e r i n g 3 5 ( 2 0 0 9 ) 175–183

Fig. 2 – Fractionation of the different N-forms


(average ± 1S.D.) in the water of the SFCWs during overflow
3 as a function of time.

A substantial part of P was removed within the first 4 h


of the experiment in both types of constructed wetlands
(Fig. 1). Adsorption and precipitation are considered as the
main removal mechanisms for P. Overall removal efficien-
cies were much higher for the SSFCWs (Table 2). For both the
SFCWs and the SSFCWs a gradual release of P was observed
with time. Calculating release rates based upon the surfaces
of the beds (SSFCWs: 1.56 m2 ; SFCWs: 0.72 m2 ) and a given
retention time of 8 days showed values of 9 mg P m−2 d−1 and
14 mg P m−2 d−1 for the SSFCWs and SFCWs, respectively.
Fig. 1 – Removal patterns of Ntot, P and CODtot as a CODtot removal occurred faster (p < 0.005) in the SSFCWs
function of time for the SSFCWs (dashed line) and SFCWs and was removed to a higher extent (Fig. 1) although the influ-
(full line). Mean inlet concentrations are represented by ent concentrations were similar. The C1 batch had a CODtot
open symbols. concentration of 434 ± 50 and 450 ± 30 mg O2 l−1 , respectively
in the SFCWs and SSFCWs. After 4 h the CODtot concentra-
tion in the SSFCWs was half the concentration measured in
the SFCWs (107 ± 20 mg O2 l−1 vs. 214 ± 8 mg O2 l−1 ). The con-
centration at the end of the experiment was 95 ± 4 mg O2 l−1
occurring mostly within the first 24 h. In the SSFCWs con-
and 21 ± 15 mg O2 l−1 for the SFCWs and SSFCWs, respectively.
centrations dropped from 2.5 ± 2.2 mg l−1 NH4 –N present in
For each wetland type, a fractionation of the CODtot for one
the influent to less than 0.4 mg l−1 (data not shown) and no
overflow event (SSFCWs: event 1; SFCWs: event 3) is pre-
increase of NH4 –N was found during the further duration of
sented in Table 3. Although the composition and fractionation
the experiment. In contrast to the SSFCWs, increasing NH4 –N
were equal for both overflows on the SSFCWs and SFCWs,
concentrations were observed with time for the SFCWs:
the concentration of CODpart was higher in the SFCWs 4 h
NH4 –N concentrations increased from 2.7 ± 0.4 mg l−1 on day
after application (day 0). CODtot was removed much faster in
0 to 18.1 ± 0.2 mg l−1 on day 8. The influent concentration
the SSFCWs than in the SFCWs, as evidenced by the differ-
of nitrates was 8 ± 1.5 mg NO3 –N l−1 throughout the experi-
ence between CODtot in the influent and that on day 0. This
ments. Concentrations of NO3 –N after 1 day were lower than
could be mainly attributed to the faster removal of CODpart in
0.07 and 0.2 mg NO3 –N l−1 for the SSFCWs and SFCWs, respec-
the SSFCWs. The CODpart was for each sampling time signif-
tively.
icantly higher in the SFCWs. The overflow caused an increase
of CODpart in the SFCWs probably due to resuspension of the
sediment when the water entered the system. Moreover, the
Table 2 – Removal efficiencies (mass-based) for the CODsol was lower in the SSFCWs than in the SFCWs. The
different pollutants in the SFCWs and SSFCWs after 8 overall removal efficiency was 88 ± 4% for the SSFCWs and
days. 61 ± 7% for the SFCWs (Table 2). For the SFCWs the removal
Pollutant SFCW (%) SSFCW (%) efficiency decreased systematically over the overflow events
from 67.8 ± 2% for overflow 1 to 52 ± 0.1% for overflow 3. The
Ntot 36.6 ± 3.3 96.7 ± 1.9 COD-removal in the SSFCWs remained constant throughout
P 36.0 ± 5.0 71.1 ± 7.7
the different overflow events.
NO3 –N 99.7 ± 0.1 99.1 ± 0.5
CODtot 60.8 ± 7.1 88.1 ± 3.5
Initial contaminant removal was faster in the SSFCWs than
in the SFCWs. After 12 h the concentration of NO3 –N was
e c o l o g i c a l e n g i n e e r i n g 3 5 ( 2 0 0 9 ) 175–183 179

Table 3 – Fractionation of CODtot (average ± 1S.D.) for the SSFCWs (overflow 1) and SFCWs (overflow 3).
CODtot CODpart CODsol

SFCW SSFCW SFCW SSFCW SFCW SSFCW

C1 434 ± 50 450 ± 30 43 ± 15 151 ± 35 361 ± 29 300 ± 5


C2 212 ± 25 227 ± 14 15 ± 18 21 ± 39 195 ± 22 205 ± 27
C3 102 ± 10 113 ± 7 7 ± 16 4±2 101 ± 14 109 ± 7
Day 0 230 ± 27 96 ± 14 50 ± 24 19 ± 8 179 ± 48 76 ± 17
Day 1 137 ± 16 50 ± 5 37 ± 10 6±4 100 ± 9 45 ± 7
Day 3 113 ± 15 20 ± 10 54 ± 15 2±4 59 ± 2 18 ± 6
Day 6 73 ± 11 16 ± 1 12 ± 11 <1 61 ± 5 <15
Day 8 n.d. 16 ± 1 n.d. <1 n.d <15

n.d.: not determined.

Table 4 – Short-term removal of Ntot, NO3 –N and P after 4, 12 and 24 h for the SSFCWs (average ± 1S.D.).
Influent 4h 12 h 24 h

Ntot (mg l−1 ) 31.9 ± 21.1 19.4 ± 5.4 (39) 12.1 ± 2.6 (62) 4.9 ± 1.3 (85)
NO3 –N (mg l−1 ) 10.2 ± 5.9 7.0 ± 0.7 (32) 0.9 ± 1.6 (91) 0.04 ± 0.03 (>99)
P (mg l−1 ) 4.0 ± 1.9 1.00 ± 0.21 (75) 1.03 ± 0.77 (74) 1.25 ± 0.85 (67)

Removal percentages (%) are shown between brackets.

Table 5 – Coefficients for the k–C*-model assessed with (+ influent) or without (− influent) influent concentrations.
R C* (mg l−1 ) kv (d−1 ) ka (m d−1 )
Ntot SFCW + Influent 0.98 18.7 ± 0.5 6.0 ± 1.6 1.7 ± 0.5
− Influent 0.98 18.3 ± 0.3 1.2 ± 0.4 0.3 ± 0.1
SSFCW + Influent 0.99 1.5 ± 0.6 2.8 ± 0.4 0.4 ± 0.1
− Influent 0.99 1.1 ± 0.1 1.9 ± 0.1 0.29 ± 0.02

CODtot SFCW + Influent 0.99 95.9 ± 10.2 0.6 ± 0.2 0.17 ± 0.06
− Influent 0.99 89.6 ± 6.9 0.4 ± 0.1 0.11 ± 0.03
SSFCW + Influent 0.99 36.7 ± 9.5 7.0 ± 1.8 1.1 ± 0.3
− Influent 0.99 24.3 ± 2.9 0.8 ± 0.1 0.12 ± 0.02

Table 6 – First-order volumetric and area-based (subscripts v and a, respectively) removal rate constants determined for
fast and slow removal in both the SFCWs and SSFCWs.
kv fast (d−1 ) kv slow (d−1 ) ka fast (m d−1 ) ka slow (m d−1 )
Ntot SFCW 4.8 ± 2.0 1.2 ± 0.4 1.4 ± 0.6 0.3 ± 0.1
SSFCW 0.9 ± 0.5 1.9 ± 0.1 0.14 ± 0.07 0.54 ± 0.03

CODtot SFCW 0.2 ± 0.3 0.4 ± 0.1 0.06 ± 0.09 0.11 ± 0.03
SSFCW 6.2 ± 1.9 0.8 ± 0.1 0.9 ± 0.3 0.12 ± 0.02

less than 1 mg l−1 (Table 4). In first instance removal of P was concentration data. The calculated constants are presented in
observed but already after 12 h the concentrations of P in Table 5. The inclusion of influent concentrations resulted in
the water started to increase. The main removal of P (75%) higher C* and k-values. Deriving kfast and kslow showed that
occurred within the first 4 h after the overflow event. This cor- fast occurring reactions were important for the removal of
responded with the sharp decline of the redox potential from Ntot in the SFCWs and the removal of CODtot in the SSFCWs
173 ± 107 mV after 4 h to −192 ± 54 mV after 24 h. (Table 6).
A considerable amount of contaminants was already
removed 4 h after application of the simulated overflow
event, suggesting high rate removal processes. During the 4. Discussion
further duration of the experiment a slower removal was
observed. First-rate removal constants for Ntot and CODtot The SSFCWs were planted in January 2005, whereas the SFCWs
were calculated with inclusion and exclusion of the influent were planted in July 2006. At the start of the experiments both
180 e c o l o g i c a l e n g i n e e r i n g 3 5 ( 2 0 0 9 ) 175–183

wetland types had an actual age of 15 months (SSFCWs) and The consistent low nitrate concentrations in both wet-
5 months (SFCWs). The high plant density that was applied land types during the further duration of the experiment
to the wetlands (25 plants m−2 ) results in a faster maturation indicated limiting conditions for nitrification due to very low
process of the vegetation compared to lower plant densities dissolved oxygen levels in the water. Some authors believe
(USEPA, 2000). Another factor at which both wetlands may that the amount of oxygen escaping from the plant roots into
differ is the release of components during degradation of the anaerobic environment is very limited, and atmospheric
dead plant material. In the SFCWs, as it was recently planted, oxygen diffusion is therefore the main source of oxygen in
no decaying plant material was present. For the SSFCWs the wetlands (Armstrong and Armstrong, 1988; Brix, 1993; Kadlec
standing biomass was removed in the past and as such no and Knight, 1996). One can assume however that oxygen
litter build-up was present on top of both wetland types. supply due to atmospheric diffusion is negligible in SSFCWs
Microbial removal of pollutants is a very important pathway because the water flows between the gravel and only limited
for several contaminants. Because of the difference in age contact is possible between water and atmosphere. Then, root
between both wetland types, a difference in biofilm develop- oxygen loss would constitute the main oxygen source. How-
ment may be present. Development of microbial biomass in ever Wu et al. (2001) found in a comparative study between
the media of a SSFCW typically requires three to six months SSFCWs and SFCWs that the atmospheric oxygen-diffusion
(USEPA, 2000). Tao et al. (2006) investigated the maturation of rates of SSFCWs were 4–17 times higher than those of SFCWs.
biofilms in SFCWs used for woodwaste leachate treatment. It Wu et al. (2001) concluded their study with the suggestion that
took <1–6 weeks for maturation of the biofilm on submerged the amount of oxygen transferred through both the roots and
plant surfaces and the sedimentary microbial community, atmospheric diffusion could not fully explain the high removal
with longer maturation time needed for the attached micro- of NH4 in the SSFCWs and SFCWs under study by nitrification
bial communities than for the suspended communities. and denitrification reactions.
Although the SFCWs were much younger, it can be expected Authors have stated that constructed wetlands are char-
that in both used wetland types, a stable microbial commu- acterized by the occurrence of both aerobic and anaerobic
nity was present and that gravel, submerged plant parts, soil microsites within the wetland, allowing both aerobic and
surface and the water column were completely colonised by anaerobic reactions to occur (Vymazal, 1999). The overall good
biofilms. removal efficiency of Ntot within the SSFCWs seems to be in
N-removal in the SFCWs was very limited. Removal of accordance with the assumption that nitrification could occur
nitrogen is influenced by temperature (Gardner et al., 2000; near the roots after which diffusion of the formed nitrate to the
Wang et al., 2008). Absence of total Kjeldahl nitrogen removal anaerobic surroundings occurred, followed by denitrification.
was observed in a constructed wetland by Reuter et al. (1992). Decreasing nitrogen concentrations can go together with-
They attributed this to reduced nitrification and ammonifi- out significant increase of nitrate or nitrite as was observed
cation rates, which was associated with a poor vegetation by Akratos and Tsihrintzis (2007). Denitrification is primarily
growth. Although the reed plants in the SFCWs had been accli- determined by the rate of diffusion of the nitrates to zones
matised during more than 5 months, the lower temperatures where denitrification might take place (Flynn et al., 1999).
in the greenhouse during the test period resulted in a poor con- After 4 h no significant removal of NH4 –N was observed
dition of the reed. Below 15 ◦ C neither the micro-organisms in the SFCWs (Fig. 2). Thus, N-removal by sorption of NH4 –N
responsible for nitrogen removal nor uptake by vegetation was negligible within the first 4 h for the SFCWs, compared
functions properly (Vymazal, 1999; Akratos and Tsihrintzis, to the high contribution of this process in the SSFCWs where
2007). no NH4 could be detected 4 h after application of the over-
After 8 days, almost all the N was present in the form flow event. Adsorption onto the substrate may be of more
of NH4 –N within the SFCWs, indicating that ammonification importance in SSFCWs because of the larger surface area pro-
occurred (Fig. 2). Ammonification occurs both under aerobic vided by the gravel (Sikora et al., 1995). During a previous
and anaerobic conditions (Vymazal, 2007). Part of the Ntot experiment on the same SSFCWs as used during this study
present in the influent consisted of nitrogen associated with maximum sorption capacity of the gravel was determined at
particulate matter. This part was quickly removed by sedimen- 17 mg NH4 –N kg−1 gravel (Torres, 2005). In agreement with our
tation. The compounds that were added to act as a carbon results, Wu et al. (2001) found higher ammonium removal effi-
source, such as starch, milk powder and peptone, did not dis- ciencies in SSFCWs than in SFCWs. Depending on the pH of the
solve completely and contained also N, acting as a source of water, conversion of NH4 –N to NH3 followed by volatilisation
particulate N. The decrease of Ntot that was observed between might become important. As there is a closer contact between
the addition of the combined sewer overflow and 4 h later the water and air in the SFCWs, this reaction may, at higher
was attributed to sedimentation, sorption and nitrate-removal pH, be more important in SFCWs (Vymazal, 2007). Connolly et
reactions. The nitrates were efficiently removed after 4 h by al. (2004) described how the first step of ammonium-removal
denitrification for both the SSFCWs and SFCWs and no sig- in a vertical flow constructed wetland was adsorption of NH4 +
nificant levels of nitrates were detected later on. Such strong onto the media, followed by nitrification and denitrification
denitrification activity at redoxpotentials lower than 300 mV reactions. Sorption of NH4 –N is enhanced in more reducing
was clearly shown in a microcosm incubation experiment conditions, stimulating the removal by sorption in the SSFCWs
by Yu et al. (2007). The nitrates present in the influent were (Sikora et al., 1995).
not detected in the SSFCWs after 4 h, pointing out the quick The sorption of NH4 + onto the substrate of the SSFCWs
removal by plant uptake or turn-over to intermediate products can be followed by a gradual release. The released NH4 –N can
of the denitrification and N2 . be removed by plant uptake as plants prefer the uptake of
e c o l o g i c a l e n g i n e e r i n g 3 5 ( 2 0 0 9 ) 175–183 181

the more reduced NH4 + –N to the uptake of NO3 − (Kadlec and mean retention time. In contrast with Carleton et al. (2001),
Knight, 1996). This release could have occurred between two who found that an increase of the retention time lowered the
monitoring campaigns, when tap water was used for irrigating effect of resuspension of the P sorbed or associated with the
the beds. Nutrient-limiting conditions between two overflow sediment, our experiment showed for both the SSFCWs and
events may have enhanced desorption so that no saturation SFCWs an initial removal of P followed by a gradual release of
of the substrate occurred and adsorption sites were readily the sorbed P. The current experiment showed the importance
available when a new overflow event entered the system. of the retention time when release of P from the substrate
Evaluating the importance of plant uptake in the SFCWs is presumably to occur. Shorter retention times would have
and SSFCWs one can conclude that plant uptake will be of resulted in higher removal efficiencies. Also for the removal of
higher importance in the SSFCWs. This was mainly due to P, influence of temperature has been mentioned in the litera-
the better condition of the reed plants in the SSFCWs during ture (Wang et al., 2008), but this was described as an indirect
the test period, but might also depend on the better con- effect. Lower temperatures are associated with winter and this
tact between plant roots and water in the SSFCWs. Next to period goes together with decomposition of the litter, resulting
nutrient uptake from the water, also nutrients present in the in P-release.
soil of the SFCW are available for plant uptake. Recently, the COD removal occurred significantly faster in the SSFCWs
presence of alternative NH4 -removal pathways in constructed than in the SFCWs (p < 0.001) as a combination of enhanced
wetlands has been brought under attention (Sun and Austin, sedimentation, the presence of a higher contact surface with
2007; Vymazal, 2007). Autotrofic denitrification, better known the biofilm and probably a more balanced assembly of biofilm
as the ANAMMOX-process, could also have contributed next to in the SSFCWs as they were already in use for a longer time.
plant uptake and adsorption onto the substrate to the removal The gravel in the SSFCWs enhanced sedimentation of COD-
of nitrogen. In this process nitrite and ammonium are directly part as could be subdued from the rapid decrease of CODpart.
transformed into nitrogen gas (Mulder et al., 1995). The presence of the gravel has a dual action. It enhances
As such, we attribute the difference in removal between sedimentation by slowing down the water current, and sec-
the SSFCWs and SFCWs to a combination of characteristics ondly, the biofilm, attached to the gravel will enhance the
intrinsic to the type of system and lower temperatures dur- trapping of particulates. The gravel also prevents resuspen-
ing the testing period with the SFCWs (8 ◦ C) affecting both sion to occur, which was clearly observed in the SFCW. As
plant growth and microbial processes. Sorption of NH4 onto water from combined sewer overflows is highly loaded with
the gravel was identified as the most probable removal mech- particulate material and pollutants tend to be associated with
anism in the SSFCWs. the particulate fraction, an effective removal of this fraction
The gravel in the SSFCWs was more important in removal may strongly increase removal efficiencies of such systems.
of P than the soil substrate in the SFCWs. The gravel in the However, the capacity of removing particulate material may
SSFCWs has a higher adsorption capacity than the soil and change over time resulting in a decrease of the retention of
provides a larger contact area as water flows within the gravel suspended solids (Tanner et al., 1998). SSFCWs tend to allow
substrate rather than on top of the soil surface. Many studies a better removal of particulate material due to the filtering
demonstrated P removal for both SSFCWs and SFCWs (Green effect of the gravel, although clogging can occur, resulting in
and Upton, 1993; Kadlec and Knight, 1996; Drizo et al., 2000; lower removal performances. Kadlec and Knight (1996) stated
Braskerud, 2002). In a column-experiment simulating SFCWs that the effect of temperature on the removal of BOD for
Dunne et al. (2005) observed an initial release of P, followed by both SSFCWs and SFCWs was negligible. Also the final COD-
a decrease. This release resulted in a concentration increase sol concentration was lower in the SSFCW than in the SFCW.
of 0.4 mg l−1 after 2 days. Over the duration of this study, P in Higher concentrations of attached biofilm may have enhanced
the SFCWs increased with approximately 0.4 mg l−1 although the removal of soluble organic compounds. Stein et al. (2006)
the initial P-concentrations present in the soil were 2–4 times observed a substantial COD removal almost immediately upon
higher (1800 mg P kg−1 soil) than the soils that were used by filling during a batch-loaded column experiment simulating
Dunne et al. (2005). The increase of P in the SSFCWs was a lit- SSFCWs. Residual levels (C*) were reached by days 3–14 and
tle higher (0.6 mg l−1 ). The release of oxygen by plant roots may sometimes COD-values approached C* within 1 day.
increase the adsorption capacity of wetlands for P (Walhugala For all the investigated parameters, the SSFCWs performed
et al., 1987). The release of P coincided with decreasing redox better, resulting in a faster removal down to lower concentra-
potentials in both wetland types. Redox conditions strongly tions. This study confirms the results of Shutes (2001), which
influence the chemistry of P through indirect effects on the indicated that SSFCWs are more efficient than SFCW when
solubility of Fe and Mn minerals. In addition, the associated removing pollutants at higher application rates (Shutes, 2001).
changes in pH can affect the solubility of Ca-bound phos- Due to higher removal rates, shorter retention times can be
phate and the sorption reactions involving variably charged applied. The main drawback for both systems is the large area
surfaces (Scalenghe et al., 2002). Reduction reactions in the needed for construction and the need for a constant standing
environment result in a net alkaline reaction, which tends to volume of water to preserve the vegetation. For buffering a vol-
be counteracted by the precipitation of carbonates of Fe(II) and ume of 1000 m3 and working with a total height of 0.5 m (0.4 m
Mn(II) and the production of CO2 and organic acids originat- available for storage and a remaining water height of 0.1 m)
ing from decomposing organic material (Yu et al., 2007). Gu the required surface is 0.65 ha for the SSFCWs and 0.25 ha
and Dreschel (2008) found a correlation between the removal for the SFCWs. Also other pollutants than the ones discussed
of Ca, present in the influent water, and the removal of P. The here can be of major concern in stormwater and CSO treat-
removal of P is less influenced by the loading rate than by the ment. Metal removal and behaviour may differ between both
182 e c o l o g i c a l e n g i n e e r i n g 3 5 ( 2 0 0 9 ) 175–183

wetland types. Furthermore, performances observed in full- biological nutrient removal. Part I. Development of a new
scale systems may differ from those in small-scale systems. synthetic sewage. Chemosphere 38, 699–709.
SSFCWs seem to be usable when CSO- and stormwater treat- Braskerud, B.C., 2002. Factors affecting phosphorus retention in
small constructed wetlands treating agricultural nonpoint
ment can be combined with tertiary treatment of wastewater.
source pollution. Ecol. Eng. 1, 41–61.
The application of this type of treatment without secondary Brix, H., 1993. Macrophyte mediated oxygen transfer in wetlands:
water inputs and to guarantee at the same time a satisfac- transport mechanisms and rates. In: Moshiri, G.A. (Ed.),
tory plant growth might prove to be quite difficult as the Constructed Wetlands for Water Quality Improvement. CRC
performance will be highly dependent on the frequency of Press LLC, Boca Raton, FL, p. 391.
overflowing. SSFCWs are sensitive to drying-up, resulting in Carleton, J.N., Grizzard, T.J., Godrej, A.N., Post, H.E., 2001. Factors
plant death, when no regular water input is present. This is affecting the performance of stormwater treatment wetlands.
Water Res. 35, 1552–1562.
important, especially during summer when high evapotran-
Connolly, R., Zhao, Y.Q., Sun, G.Z., Allen, S., 2004. Removal of
spiration rates occur. However, peak concentrations can be ammoniacal-nitrogen from an artificial landfill leachate in
removed more quickly by SSFCWs, resulting in concentrations downflow reed beds. Process Biochem. 39, 1971–
that can be discharged earlier compared to SFCWs. SFCWs can 1976.
store a larger volume of water on the same surface area than Debo, T.N., Reese, A.J., 2003. Municipal Stormwater Management.
SSFCWs and therefore a higher volume of water remaining in CRC Press LLC, Boca Raton, FL, p. 1141.
Dittmer, U., Meyer, D., Langergraber, G., 2005. Simulation of a
the bed can be used without losing the needed buffer capacity.
subsurface vertical flow constructed wetland for CSO
Therefore SFCWs seem to be more applicable as stand-alone
treatment. Water Sci. Technol. 51, 225–232.
constructed wetlands in the field without the need for sec- Drizo, A., Frost, C.A., Grace, J., Smith, K.A., 2000. Phosphate and
ondary water inputs and when the frequency of overflows is ammonium distribution in a pilot-scale constructed wetland
rather low. When surface area is scarce, when dealing with with horizontal subsurface flow using shale as a substrate.
highly frequent overflow events or rather high concentrations, Water Res. 39, 2483–2490.
preference is given to SSFCWs. Dunne, E.J., Culleton, N., O’Donovan, G., Harrington, R.,
Daly, K., 2005. Phosphorus retention and sorption by
constructed wetland soils in Southeast Ireland. Water Res. 39,
4355–4362.
5. Conclusion
Eaton, A.D., 1995. In: Eaton, A.D., Clesceri, L.S., Greenberg, A.E.
(Eds.), 1995. Standard Methods for the Examination of Water
Nitrogen, P and COD present in synthetic wastewater simulat- and Waste water, 19th ed. American Public Health
ing combined sewer overflow events were removed faster by Organisation, Washington, DC, USA, p. 1325.
SSFCWs than SFCWs. In general the end concentrations of the Fink, D.F., Mitsch, W.J., 2005. Seasonal and storm event nutrient
investigated pollutants were lower in the SSFCWs. Removal removal by a created wetland in an agricultural watershed.
of nitrogen was strongly inhibited in the SFCW. The very Ecol. Eng. 23, 313–325.
Flynn, N., Gardner, P.J., Maltby, E., 1999. The measurement and
high removal efficiencies of Ntot obtained in the SSFCWs are
analysis of denitrification rates obtained using soil
attributed to a combination of efficient ammonification, plant columns from river marginal wetlands. Soil Use Manage. 15,
uptake and effective sorption of the formed NH4 –N to the 150–156.
substrate. Removal by nitrification and denitrification upon Frechen, F.B., Schier, W., Felmeden, J., 2006. The plant-covered
desorption of the ammonium occurred probably in the time retention soil filter (RSF): the mechanical and biological
span between two testing periods. In both wetland types an combined sewer overflow (CSO) treatment plant. Eng. Life Sci.
initial removal of P was followed by a release within the first 6, 74–79.
Gardner, P.J., Flynn, N., Maltby, E., 2000. A temperature correction
day. The quick removal of P was due to sorption. More reduc-
procedure for field-based soil denitrification experiments. Soil
ing conditions resulted in a gradual release of the sorbed P. Use Manage. 16, 260–263.
Besides removal performances practical design issues such as Geary, P.M., Mendez, H., Dunstan, R.H., 2006. Design
the surface area needed to store and treat a certain amount of considerations in the performance of stormwater devices
stormwater or CSOs, dependence of other water inputs (and incorporating constructed wetlands (1833–1842). In:
the possible presence of these inputs in combination with the Proceedings of the 10th International Conference on Wetland
Systems for Water Pollution Control, September 23–29, Lisbon,
frequency of the overflow events) need to be considered when
Portugal.
choosing a wetland type for treatment of combined sewer
Green, M.B., Upton, J., 1993. Reed bed treatment for small
overflows. communities: UK experience. In: Moshiri, G.A. (Ed.),
Constructed Wetlands for Water Quality Improvement. CRC
references Press, Inc., Boca Raton, FL, pp. 517–524.
Greenway, M., Jenkins, G., Polson, C., 2006. Macrophyte zonation
in stormwater wetlands—getting it right! (1807–1819). In:
Proceedings of the 10th International Conference on Wetland
Akratos, C.S., Tsihrintzis, V.A., 2007. Effect of temperature, HRT, Systems for Water Pollution Control, September 23–29, Lisbon,
vegetation and porous media on removal efficiency of Portugal.
pilot-scale horizontal subsurface flow constructed wetlands. Gu, B., Dreschel, T., 2008. Effects of plant community and
Ecol. Eng. 29, 173–191. phosphorus loading rate on constructed wetland performance
Armstrong, J., Armstrong, W., 1988. Phragmites australis—a in Florida, USA. Wetlands 28, 81–91.
preliminary study of soil oxidizing sites and internal gas Heaney, J.P., Wright, L., Sample, D., 1999. Research needs in urban
transport pathways. New Phytol. 108, 373. wet weather flows. Water Environ. Res. 71, 2.
Boeije, G., Corstanje, R., Rottiers, A., Schowanek, D., 1998. Kadlec, R.H., Knight, R.L., 1996. Treatment Wetlands. Lewis
Adaptation of the CAS test system and synthetic sewage for Publishers, Boca Raton, FL, p. 928.
e c o l o g i c a l e n g i n e e r i n g 3 5 ( 2 0 0 9 ) 175–183 183

Line, D.E., Jennings, G.D., Shaffer, M.B., Calabria, J., Hunt, W.F., Sun, G., Austin, D., 2007. Completely autotrophic
2008. Evaluating the effectiveness of two stormwater nitrogen-removal over nitrite inlab-scale constructed
wetlands in North Carolina. Trans. ASABE 51, 521–528. wetlands: evidence from a mass balance study. Chemosphere
Mulder, A., van de graaf, A.A., Robertson, L.A., Kuenen, J.G., 1995. 68, 1120–1128.
Anaerobic ammonium oxidation discovered in a denitrifying Tanner, C.C., Sukias, J.P.S., Upsdell, M.P., 1998. Organic matter
fluidized bed reactor. FEMS Microbiol. Ecol. 16, 177–184. accumulation during maturation of gravel-bed constructed
Mungasavalli, D.P., Viraraghavan, T., 2006. Constructed wetlands wetlands treating farm dairy waste waters. Water Res. 32,
for stormwater management: a review. Fresenius Environ. 3046–3054.
Bull. 15, 1363–1372. Tao, W., Hall, K.J., Duff, S.J.B., 2006. Performance evaluation and
Reuter, J.E., Djohan, T., Goldman, C.R., 1992. The use of wetlands effects of hydraulic retention time and mass loading rate on
for nutrient removal from surface runoff in a cold climate treatment of woodwaste leachate in surface-flow constructed
region of California—results from a newly constructed wetlands. Ecol. Eng. 26, 252–265.
wetland at Lake Tahoe. Environ. Manage. 36, 35–53. Torres, A.M.L., 2005. Subsurface flow constructed wetlands (SSF
Revitt, D.M., Shutes, R.B.E., Jones, R.H., Forshaw, M., Winter, B., CW): pilot-scale experiments and applicability in the
2004. The performance of vegetative treatment systems for Philippines. MSc Environmental Sanitation (2004–2005), Ghent
highway runoff during wet and dry conditions. Sci. Total University, 2005.
Environ. 334–335, 264–270. USEPA, 2000. Constructed wetlands treatment of municipal
Revitt, D.M., Worall, P., Brewer, D., 2001. The integration of wastewaters. Manual of EPA/625/R-99/010, U.S. Environmental
constructed wetlands into a treatment system for airport Protection Agency, Cincinnati, OH.
runoff. Water Sci. Technol. 44, 469–476. Vymazal, J., 1999. Nitrogen removal in constructed wetlands with
Robson, M., Spence, K., Beech, L., 2005. Stream quality in a small horizontal sub-surface flow—can we determine the key
urbanised catchment. Sci. Total Environ. 357, 194–207. process? In: Vymazal, J. (Ed.), Nutrient Cycling and Retention
Scalenghe, R., Edwards, A.C., Marsan, F.A., Barberis, E., 2002. The in Natural and Constructed Wetlands. Backhuys Publishers,
effect of reducing conditions on the solubility of phosphorus Leiden, pp. 1–17.
in a diverse range of European agricultural soils. Eur. J. Soil Vymazal, J., 2007. Removal of nutrients in various types of
Sci. 53, 439–447. constructed wetlands. Sci. Total Environ. 380, 48–65.
Scholz, M., Höhn, P., Minall, R., 2002. Mature experimental Walhugala, A.G., Suzuki, T., Kurihara, Y., 1987. Removal of
constructed wetlands treating urban water receiving high nitrogen, phosphorus and COD from waste water using a sand
metal loads. Biotechnol. Prog. 18, 1257–1264. filtration system with Phragmites australis. Water Res. 21,
Shutes, R.B.E., 2001. Artificial wetlands and water quality 1217–1224.
improvement. Environ. Int. 26, 441–447. Welker, A., 2006. Vertical flow constructed wetlands for enhanced
Shutes, R.B.E., Revitt, D.M., Lagerberg, I.M., Barraud, V.C.E., 1999. CSO treatment—an alternative for elimination of organic
The design of vegetative constructed wetlands for the pollutants? In: In Proceedings of the 10th International
treatment of highway runoff. Sci. Total Environ. 235, 189–197. Conference on Wetland Systems for Water Pollution Control,
Shutes, R.B.E., Revitt, D.M., Mungur, A.S., Scholes, L.N.L., 1997. vol. 693, September 23–29, Lisbon, Portugal.
The design of wetland systems for the treatment of urban Wang, Y., Inamori, R., Kong, H., Xu, K., Inamori, Y., Kondo, T.,
runoff. Water Sci. Technol. 35, 19–25. Zhang, J., 2008. Nitrous oxide emission from polyculture
Sikora, F.J., Zhu, T., Steinberg, S.L., Coonrod, H.S., 1995. constructed wetlands: effect of plant species. Environ. Pollut.
Ammonium removal in constructed wetlands with 152, 351–360.
recirculating subsurface flow: removal rates and mechanisms. Wong, T.H.F., Geiger, W.F., 1997. Adaptation of waste water
Water Sci. Technol. 32, 193–203. surface flow wetland formulae for application in constructed
Stein, O.R., Biederman, J.A., Hook, P.B., Allen, W.C., 2006. Plant stormwater wetlands. Ecol. Eng. 9, 187–202.
species and temperature effects on the k–C* first-order model Wu, M.-Y., Franz, E.H., Chen, S., 2001. Oxygen fluxes and
for COD removal in batch-loaded SSF wetlands. Ecol. Eng. 26, ammonia removal efficiencies in constructed treatment
100–112. wetlands. Water Environ. Res. 73, 661–666.
Suarez, J., Puertas, J., 2005. Determination of COD, BOD and Yu, K., Böhme, F., Rinklelebe, J., Neue, H.-U., DeLaune, R.D., 2007.
suspended solids loads during a combined sewer overflow Major biogeochemical processes in soils—a microcosm
(CSO) event in some combined catchments in Spain. Ecol. Eng. incubation from reducing to oxidizing conditions. Soil Sci.
24, 201–219. Soc. Am. 71 (4), 1406–1417.

You might also like