wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 6 3e5 0 7 4

Avai lab le a t www.sc iencedi rec t .com

journa l homepage : www.e lsev ie r . com/ loca te /wat res

Remediation of groundwater contaminated with MTBE andbenzene: The potential of vertical-flow soil filter systems

Manfred van Afferden a,*, Khaja Z. Rahman a, Peter Mosig a, Cecilia De Biase b,Martin Thullner b, Sascha E. Oswald c, Roland A. Muller a

aCentre for Environmental Biotechnology (UBZ), UFZeHelmholtz Centre for Environmental Research, Permoserstrasse 15, 04318 Leipzig,

GermanybDepartment of Environmental Microbiology, UFZeHelmholtz Centre for Environmental Research, Permoserstrasse 15, 04318 Leipzig,

Germanyc Institute for Earth and Environmental Sciences, University of Potsdam, Potsdam, Germany

a r t i c l e i n f o

Article history:

Received 23 May 2011

Received in revised form

4 July 2011

Accepted 5 July 2011

Available online 14 July 2011

Keywords:

Benzene

Groundwater remediation

Hydraulic loading rate

MTBE

Pilot-scale constructed wetland

Vertical-flow soil filter

Willow tree

* Corresponding author. Tel.: þ49 341 235 18E-mail address: manfred.afferden@ufz.de

0043-1354/$ e see front matter ª 2011 Elsevdoi:10.1016/j.watres.2011.07.010

a b s t r a c t

Field investigations on the treatment of MTBE and benzene from contaminated ground-

water in pilot or full-scale constructed wetlands are lacking hugely. The aim of this study

was to develop a biological treatment technology that can be operated in an economic,

reliable and robust mode over a long period of time. Two pilot-scale vertical-flow soil filter

eco-technologies, a roughing filter (RF) and a polishing filter (PF) with plants (willows), were

operated independently in a single-stage configuration and coupled together in a multi-

stage (RF þ PF) configuration to investigate the MTBE and benzene removal perfor-

mances. Both filters were loaded with groundwater from a refinery site contaminated with

MTBE and benzene as the main contaminants, with a mean concentration of 2970 � 816

and 13,966 � 1998 mg L�1, respectively. Four different hydraulic loading rates (HLRs) with

a stepwise increment of 60, 120, 240 and 480 L m�2 d�1 were applied over a period of 388

days in the single-stage operation. At the highest HLR of 480 L m�2 d�1, the mean

concentrations of MTBE and benzene were found to be 550 � 133 and 65 � 123 mg L�1 in the

effluent of the RF. In the effluent of the PF system, respective mean MTBE and benzene

concentrations of 49 � 77 and 0.5 � 0.2 mg L�1 were obtained, which were well below the

relevant MTBE and benzene limit values of 200 and 1 mg L�1 for drinking water quality. But

a dynamic fluctuation in the effluent MTBE concentration showed a lack of stability in

regards to the increase in the measured values by nearly 10%, which were higher than the

limit value. Therefore, both (RF þ PF) filters were combined in a multi-stage configuration

and the combined system proved to be more stable and effective with a highly efficient

reduction of the MTBE and benzene concentrations in the effluent. Nearly 70% of MTBE and

98% of benzene were eliminated from the influent groundwater by the first vertical filter

(RF) and the remaining amount was almost completely diminished (w100% reduction) after

passing through the second filter (PF), with a mean MTBE and benzene concentration of

5 � 10 and 0.6 � 0.2 mg L�1 in the final effluent. The emission rate of volatile organic

compounds mass into the air from the systems was less than 1% of the inflowmass loading

rate. The results obtained in this study not only demonstrate the feasibility of vertical-flow

soil filter systems for treating groundwater contaminated with MTBE and benzene, but can

48; fax: þ49 341 235 1830.(M. van Afferden).

ier Ltd. All rights reserved.

wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 6 3e5 0 7 45064

also be considered a major step forward towards their application under full-scale condi-

tions for commercial purposes in the oil and gas industries.

ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction presence of sufficient oxygen and fertilizer. But very little is

Due to the widespread use of fuels, BTEX compounds

(benzene, toluene, ethylbenzene, m-, o-, and p-xylene) and

MTBE (methyl tertiary-butyl ether) are frequently detected

groundwater contaminants, with releases occurring during

their production, transportation and storage (Baehr et al.,

1999; Deeb et al., 2000; Squillace et al., 1996). MTBE has

received considerable attention in recent times as it migrates

much more quickly through the soil than most of the petro-

leum distillates due to its high water solubility (up to 51 g L�1,

USEPA, 2004). Its presence in the environment is considered as

a health and drinking water problem and classifies MTBE as

a possible human carcinogen (Johnson et al., 2000). MTBE is

relatively resistant to biological degradation under anaerobic

conditions (Moreels et al., 2006), but several studies have

shown a biodegradability under aerobic conditions (Deeb

et al., 2000; Ferreira et al., 2006; Schmidt et al., 2004).

Benzene is carcinogenic and the most water soluble BTEX

compound. It can also be degraded by many microorganisms

under aerobic conditions (Yerushalmi et al., 2002). The present

limit concentrations established by the United States Envi-

ronmental Protection Agency and the German guideline value

are 200 mg L�1 for MTBE and 1 mg L�1 for benzene in drinking

water (USEPA, 2005; DVGW, 2001).

The physico-chemical properties especially the high water

solubility and the low carbon adsorption coefficient of MTBE

make it difficult to treat these organic contaminants by using

conventional groundwater treatment technologies and

represent some unique remediation challenges. The active ex-

situ remedial methods include air stripping and removal with

granular activated carbon, vapour extraction, advanced

chemical oxidation and multiphase high-vacuum extraction

(Davis and Powers, 2000; Deeb et al., 2003; Sutherland et al.,

2004; Wilhelm et al., 2002). However, the cost associated

with the construction, maintenance and operation of these

treatments diminishes their feasibility.

Constructed wetland (CW) systems represent an effective

and inexpensive option for treating municipal wastewater

and becoming available due to their wide range of applica-

tions (Cooper, 1999; Kadlec and Wallace, 2009). They are also

accepted as an alternative method to the commonly used

engineering-based treatment technologies for the removal of

organic contaminants from surface water or groundwater

(Rubin and Ramaswami, 2001; Kassenga et al., 2004; Lorah and

Voytek, 2004). In general, the vertical-flow constructed

wetlands or soil filters are gaining popularity due to their

greater oxygen transfer capacity and smaller size as compared

to the horizontal-flow wetland systems (Cooper, 1999; Kadlec

and Wallace, 2009). The findings of Eke and Scholz (2008)

suggested that intermittently flooded vertical-flow con-

structed wetlands are able to effectively treat benzene

from hydrocarbon-contaminated wastewater streams in the

known about the technical use of vertical-flow constructed

wetlands for the removal of both MTBE and benzene from

heavily contaminated groundwater.

The SAFIRA-project (remediation research in regionally

contaminated aquifers) is an interdisciplinary research

project focussing on innovative remediation technologies to

treat complex groundwater contamination. Within the

framework of this research project, a pilot plant was con-

structed at a refinery in Leuna, Germany, aiming at the

investigation and development of eco-technologies for the

removal of volatile organic compounds. Since the ground-

water treatment technology currently used in Leuna (pump-

and-treat system associated with an air stripping and

adsorption unit) is very expensive and requires high mainte-

nance efforts, the aim of this work was to develop an alter-

native biological treatment technology that can be operated in

an economic, reliable and robust mode over a long period of

time. Therefore, a specially designed pilot-scale subsurface

vertical-flow constructed wetland system was installed and

operated at the Leuna site for field investigations on the

removal of MTBE and benzene as the main groundwater

contaminants. In order to identify the potential factors influ-

encing the treatment efficiencies, the dynamics of MTBE and

benzene were investigated using pilot-scale single-stage and

multi-stage single-pass vertical-flow soil filter eco-

technologies with different hydraulic loading rates (HLRs) in

this study. As far as we are aware, no such biological treat-

ment system has been explored to date in pilot-scale facilities

for treating MTBE and benzene compounds from contami-

nated groundwater using the planted and unplanted vertical-

flow soil filter systems, nor has the effect of the different

hydraulic loading conditions been directly compared.

The main objectives of this study were: (i) to explore the

treatment performances of pilot-scale single-stage and multi-

stage single-pass vertical-flow soil filter systems for removing

MTBE and benzene from contaminated groundwater; (ii) to

evaluate the potential effects of the different hydraulic

loading rates (HLR) on the treatment efficiencies in both

systems; and finally (iii) to assess the feasibility of applying

a vertical-flow soil filter eco-technology to treat MTBE and

benzene contaminated groundwater under full-scale condi-

tions for commercial purposes.

2. Materials and methods

2.1. Site location and groundwater composition

The pilot-scale treatment facility was built near the Leuna

refinery in the North-East of Germany in 2007. Due to acci-

dental spills, improper handling (leaking underground storage

tanks, pipelines, etc.), and damages due to heavy bombing

Fig. 1 e Schematic diagram of the roughing filter (RF; on the

top) and the polishing filter (PF; on the bottom): (1) Inflow

feeding pipe; (2) Distribution pipe; (3) Layered filter

material; (4) Sump; (5) Plant biomass; (6) Drainage

outlet pipe.

wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 6 3e5 0 7 4 5065

during World War II, the groundwater in this area is heavily

contaminated with high concentrations of different gasoline

components (Martienssen et al., 2006). The fuel additive MTBE

and benzene are the predominant groundwater contaminants

at the site, with mean concentrations of 2970 � 816 and

13,966 � 1998 mg L�1, respectively. The mean concentrations

and standard deviations of the main organic and inorganic

compounds present in the water and the geochemical char-

acteristics of the influent groundwater observed during the

investigation period are given in Table 1.

2.2. Filter design

The two vertical-flow soil filters used in this study, the

Roughing Filter (RF) and Polishing Filter (PF), consisted of two

identical stainless steel containers (length: 2.30 m, width:

1.75 m, depth: 1.75 m), with a surface area of 4.025 m2 and

a total volume of 7.04 m3. Both filters were filled with a gran-

ular material of different grain sizes and arranged in layers of

varying configurations (Fig. 1). The filters were part of a larger

pilot plant with central maintenance facilities and operated

outdoors at the site, with their surface exposed to the local

climatic conditions.

The Roughing Filter (RF) consisted of three successive

layers of filter packing materials: a cover layer on the top

(25 cm), a main filter layer (120 cm) in themiddle and a bottom

layer (10 cm), which served as the drainage layer. The bottom

drainage layer was separated from a 20 cm deep sump by

a perforated steel plate. The cover layer was composed of

coarse expanded clay material (8e16 mm), facilitating water

distribution over the entire filter surface area and protecting

Table 1 e Influent groundwater characteristics based onsamples collected during the whole experimentaloperation period of 20 months (from September 2008 toMay 2010, except where noted).

Parameter Unit Inflow groundwater composition

Mean Standarddeviation

Number ofsamples

MTBE mg L�1 2970.18 �816.25 484

Benzene mg L�1 13,965.62 �1997.88 469

Cl� mg L�1 116.85 �9.96 44

NH4þ mg L�1 51.04 �9.34 44

SO42� mg L�1 11.09 �8.95 44

PO43� mg L�1 1.20 �0.75 44

Fe2þ mg L�1 6.73 �2.36 44

Ca2þ mg L�1 205.73 �14 44

Fetot mg L�1 6.69 �1.57 43

Ptot mg L�1 0.84 �0.18 44

Kþ mg L�1 12.36 �0.87 44

Naþ mg L�1 132.38 �8.03 44

Mg2þ mg L�1 58.02 �3.20 44

Mn2þ mg L�1 1.63 �0.23 44

O2 mg L�1 0.10 �0.07 57,075

Eh mV �432.25 �161.7 57,935

s mS cm�1 2.32 �0.40 57,946

pHa e 7.45 �0.35 54,046

Ta �C 12.20 �3.11 54,046

a Online measurement from September 2008 to April 2010.

the surface of the main layer from erosion. The 25-cm thick

cover layer was designed to reduce the emission of volatile

organic compounds. The underlying main layer consisted of

expanded clay material with a grain size in the range of fine

gravel (3e6 mm). One reason for using such a gravel material

was to prevent clogging due to a potential precipitation of iron

and carbonate within the filter bed. The advantage of the

larger pore spaces within these gravel particles reduced the

chances of filter clogging and increased the possibility of

applying higher hydraulic loads, which eventually facilitated

this filter system to serve as a potential first treatment step.

Finally, the drainage layer at the bottom consisted of crushed

gravel (8e16 mm), which prevented the washing out of fine

particles into the sump.

The Polishing Filter (PF) comprised four successive layers.

The 15-cm cover layer on the top consisted of a coarse

expanded claymaterial (8e16mm). The underlyingmain filter

layer of 120 cm was filled with zeolite material (zeosoil; grain

size 0e5 mm). The reason for using a finer material was that

the proportion of the finer particles caused a greater surface

area. Moreover, a longer hydraulic retention time is associated

with a higher degradation of organic pollutants and a homo-

geneous distribution of the contaminated groundwater within

this main filter layer. Zeolites have a larger surface area,

a special texture and inner structure, as compared to

conventional sand, and were therefore used within this filter

system. However, their smaller pore spaces are associated

with the risk of filter clogging and hence this filter systemwas

designed to serve as a potential second treatment step. To

facilitate better water discharge, the PF was constructed of

two drainage layers underlying the main layer. The upper 20-

cm drainage layer consisted of crushed gravel (8e16 mm)

wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 6 3e5 0 7 45066

followed by another 20-cm layer packed with even coarser

crushed gravel (16e32 mm) and placed at the bottom of the

filter.

The PF was planted with white willows (Salix alba) on the

top, with a density of around 5 plants m�2. Trees of almost

equal biomass (average height of 50 cm) and strength were

obtained from a local supplier and uniformly planted at the

end of August 2007. Willow trees were used due to their high

biomass productivity, their relatively high resistance to

organic contaminants, their ability to adapt to a broad range of

climatic and site specific conditions, their broad reaching root

systems, and their common use for phytoremediation

(Mleczek et al., 2010; Rentz et al., 2005). The RF was unplanted

in this investigation.

The contaminated groundwater was injected from the top

of each filter through a uniform distribution system of perfo-

rated PVC pipes, which was laid horizontally under the cover

layer. Water drained through the filter media to the bottom of

each basin, from where it was collected and discharged at the

outflow by a PVC drainage pipe.

2.3. Experimental conditions for filter operation

Contaminated groundwater was pumped by a timer-

controlled pump into an anaerobic storage plastic container

(Volume: w3 m3). Another timer-controlled pump distributed

the water as intermittent loads through distribution pipes

onto the surface of the two filter systems. This intermittent

dosing of water was chosen to provide good oxygen transfer to

the water phase (Kadlec, 2001). The pulse frequencies for the

two filters under different experimental conditions are pre-

sented in Table 2.

The experimental strategy was divided into two distinctly

different operation periods. During operational period 1 (days

0e388), both filters (RF and PF) were operated independently

as single-stage single-pass vertical-flow filter systems and

received the influent groundwater separately from the same

storage tank in parallel. Four different hydraulic loading rates

(HLRs) were applied to the systems and increased stepwise

(60, 120, 240 and 480 L m�2 d�1) over the period of 388 days

comprising four different experimental phases (phase I, II, III

Table 2 e Operation strategies and different experimental condfilter systems during the whole investigation period.

Period Stage Phase Duration(day)

Verticalfilter

Volume ofper load

1 Single I 0e86 RF 10

PF 12

II 86e235 RF 20

PF 24

III 235e297 RF 40

PF 48

IV 297e388 RF 80

PF 80

2 Multiple V 388e611 RFa 80

PFa 60

a Multi-stage combined system (RF þ PF).

and IV). The duration of each experimental phase was preset

to guarantee that a representative number of samples were

taken from each system. Detailed information on the opera-

tional strategies and loading schedules of both systems is

listed in Table 2.

During operational period 2 (days 388e611), the filters (RF

and PF) were connected to each other and operated in series as

a multi-stage single-pass vertical-flow filter system (RF þ PF).

The RF was receiving the contaminated groundwater from the

inflow storage tank and served as a first treatment step with

a hydraulic loading rate (HLR) of 960 Lm�2 d�1. The pre-treated

groundwater from the effluent of the RF was then pumped

into the second filter (PF) and passed through the second

system at a hydraulic loading rate of 480 L m�2 d�1. The

remaining 50% of the RF-effluent was sent to the nearby

technical groundwater remediation plant for further treat-

ment (stripping coupled with activated carbon adsorption)

and then re-injected into the aquifer. With the highest HLR of

960 Lm�2 d�1 in the RF system,wewere interested to see if any

hydraulic or technical problems occur, such as clogging,

overloading, etc. This experimental phase V was run over

a period of 223 days (days 388e611). Similarly to operational

period 1 (single-stage configuration), both filters (RF and PF)

were intermittently loaded with repeated pulses of ground-

water (Table 2).

The experiment started with period 1 in September 2008

and continued until the end of period 2 in May 2010. Willow

trees on the PF system showed an active growth of their

biomass, densely covering the whole filter surface area with

green and healthy shoots before the start of the experiment.

2.4. Sampling and analysis

Concentrations of dissolvedMTBE and benzene at the influent

and effluent of each system were analysed online using

a completely automated gas chromatograph (GC) equipped

with a photoionisation detector (PID) (META Water sampling

and analysis system WSS3; type: meta 3 HE II/PID, META,

Messtechnische Systeme GmbH, Dresden, Germany). An

Ultimetal column with a length of 25 m was used and the

carrier gas was synthetic air, set at 5 bar. The oven and

itions (hydraulic loading schedules) of the vertical-flow soil

water(L)

Loading pulsesper day (�)

Injectioninterval (min)

HLR (L m�2 d�1)

24 60 60

20 72 60

24 60 120

20 72 120

24 60 240

20 72 240

24 60 480

24 60 480

48 30 960

32 45 480

A

B

Fig. 2 e Influent and effluent concentrations of MTBE in the

A) RF and B) PF system during different experimental

phases (IeIV) of operational period 1.

wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 6 3e5 0 7 4 5067

injection/detection port temperatures were 60 and 80 �C,respectively. The detection limit for MTBE and benzene was

0.37 and 0.18 mg L�1, respectively. The determination of

detection and quantification limits of the calibration proce-

dure was carried out in accordance with DIN 32645 (1994).

Process control and data storage were carried out using the

installed software (metaControl) that stored all the measure-

ments and optional external signals on the hard disk of the

attached PC.

Known intermediate degradation products of MTBE, such

as tert-butyl alcohol (TBA), tert-butyl formate (TBF) and

aromatic hydrocarbons such as toluene, ethylbenzene, m-p-

Xylene, o-Xylene, 1,3,5-Trimethylbenzene, 1,2,4-

Trimethylbenzene and Naphthalene were analysed in both

influent groundwater and effluents from the filters by head-

space gas chromatography and mass spectroscopic detection

(HS-GCeMS). For headspace analysis, aqueous samples (10ml)

were stirred for 60 min at 70 �C in headspace vials (20 ml)

containing 2.5 g NaCl. Gas from the headspace (1 ml) was

injected into a GC/MS (GC: Agilent 6890, MS: Agilent 5973)

equipped with a 60 m HP1 column (Split injection 1:25, injec-

tion time 2 min). The time program was: 35 �C for 6 min, to

120 �C with 4 �C/min and to 280 �C with 20 �C/min, held at

280 �C for 5 min. The measuring time is 65 min per sample.

The detection limit for TBA was 1.56 mg L�1 and for other

substances specified above was <1 mg L�1.

2.5. Emission measurement

In principle, the contaminated groundwater comes in contact

with the atmosphere in both filter systems, and hence emis-

sions of volatile organic substances are expected in the air

during the treatment operation period. The volatile organic

compounds (VOCs) were measured in terms of total organic

carbon in a range of 0e100 mg TOC m�3 using a mobile flame

ionisation detector, FID 3-100 (JUM Engineering GmbH, Karls-

feld, Germany). The continuous flame ionisation chamberwas

heated up to 190 �C. The measurements were performed at

different heights (10, 20, 50, and 100 cm) in the air just above

the centre (middle point) of each filter surface and also at

same height immediately above the line of the inflow distri-

bution pipe (inlet point) installed below the top layer of the

filters. Moreover, the measurements were taken approx.

1e2 m downstream of each filter segment in the direction of

the out-flowingwind (at 40 cmheight; downwind) and approx.

5 m away from the filters against the wind direction (at 40 cm

height; upwind) as a background value.

The emission of VOCs in terms of TOC in mg m�3 air was

measured at an HLR of 480 L m�2 d�1 in both filter systems

(experimental phase IV, single-stage operation). Measure-

ments were taken in different measuring cycles over the RF

and the PF system. Duration of each cycle was 60 min, which

included an inflow feeding pulsewith duration of 4e8min and

a continuous measurement of emission in the air at different

specified heights. The emission of VOCs in terms of TOC in

mg m�3 from each measuring heights and also the back-

ground values were recorded over one feeding pulse interval

in one cycle. The net emission at each particular height

(measuring points) was calculated by subtracting the back-

ground value from the measured emission value attained at

that particular height. Four cycles were carried out for the

emission estimation over the RF and only two cycles for the PF

in this experimental phase with a same HLR in both the filters.

Since wind can have a strong influence on themeasurements,

mobile walls were built around the filters to limit the move-

ment of the air above the filter beds to a wind speed range of

0.1e0.5 m s�1.

The emitted mass of VOCs in each feeding pulse was also

calculated with the assumption that the certain volume of

water feeding on the filter segment per pulse was displacing

the same volume of air which was coming out over the filter

surface. Based on this assumption as a preliminary emission

estimation study, the rate of emitted mass from each filter

surface in terms of mg TOC m�2 d�1 and percentage of emis-

sion (%) from the inflow loading mass that goes in the atmo-

sphere (air) were calculated.

3. Results

3.1. Dynamics of MTBE and benzene: single-stagesystems

The influent and effluent dynamics of MTBE and benzene in

the RF and PF system within the different experimental pha-

ses are shown in Figs. 2 and 3. During experimental phase I

(days 0e86) with an HLR of 60 L m�2 d�1, the mean MTBE

concentration in the effluent of the RF was detected to be

139 � 69 mg L�1, which was below the limit value of 200 mg L�1

for MTBE. In contrast, a relatively higher and wide range of

MTBE concentrationwith amean value of 332� 680 mg L�1 was

A

B

Fig. 3 e Influent and effluent concentrations of benzene in

the A) RF and B) PF system during different experimental

phases (IeIV) of operational period 1.

wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 6 3e5 0 7 45068

measured in the effluent of the PF system. In this phase,

a mean reduction in the MTBE concentration of 97 and 93%

was obtained, when the RF and PF single-stage system was

used, respectively. Consequently, a mean concentration of

benzene of 64 � 76 and 0.3� 0.2 mg L�1 in the effluent of the RF

and the PF, respectively, was measured. An extremely high

reduction in the benzene concentration (w100%) was

observed in the PF system.

During experimental phase II (days 86e235) with an HLR of

120 Lm�2 d�1, themean effluentMTBE concentration in the RF

increased at the beginning and then steadily slowed down to

amean value of 399� 318 mg L�1. A relatively sharp decreasing

tendency within the effluent MTBE concentration of the PF

was observed in the middle part of phase II and continued

with a very low concentration until the end of this phase (see

Fig. 2A and B). However, a mean value of 91 and 93% reduction

in the MTBE concentration was achieved in the effluent of the

RF and the PF, respectively. Similarly to the effluent dynamics

of MTBE, the effluent benzene concentration in the RF was

increased gradually and then lowered down to a mean

effluent concentration of 413 � 736 mg L�1 from a mean

influent concentration of 15,126 � 2382 mg L�1. In the PF, no

particular trend was seen in the dynamics of the effluent

benzene concentration and a relatively higher mean value of

11 � 53 mg L�1 with a great deviation was detected, as

compared to the previous phase I (see Fig. 3A and B).

In experimental phase III (corresponding to days 235e297)

with an HLR of 240 L m�2 d�1, both the systems RF and PF

started to develop differently as it was observed in the MTBE

and benzene effluent dynamics. A mean effluent value of

402 � 222 mg L�1 resulted in a mean MTBE-concentration

reduction of 84% in the RF, whereas in the PF, the effluent

MTBE concentration sharply decreased almost immediately

after changing the experimental phase and maintained a low

concentration until the end of the phase. A mean value of

43� 90 mg L�1 resulted in a remarkable reduction (w99%) of the

mean MTBE concentration in the PF system (Fig. 2A and B). In

the case of benzene, the effluent concentration varied dras-

tically in the RF even though there was a relatively constant

influent and a very high mean effluent value of

401 � 803 mg L�1 at the end of this experimental phase. In

contrast, a highly efficient reduction (w100%) in the benzene

concentration was monitored in the effluent of the PF, with

a mean value of 0.3 � 0.2 mg L�1 (Fig. 3A and B).

At a higher HLR of 480 L m�2 d�1 in the next experimental

phase IV (corresponding to days 297e388), a relatively

constant effluent MTBE concentration was observed in the RF,

with a mean value of 550 � 133 mg L�1, which contributed to

a mean MTBE-concentration reduction of 75% from the

influent. No particular trend in the reduction of the MTBE-

concentration values was detected within the effluent

dynamics of the RF and a continuous fluctuation in the MTBE

concentration with a wide range of values was observed in the

effluent of the PF. Nevertheless, the mean effluent MTBE

concentration of 49 � 77 mg L�1 in the PF was nearly 11-fold

lower than the mean MTBE concentration of

550 � 133 mg L�1 in the RF (Fig. 2A and B). Similarly to the

previous experimental phase III, the dynamics of benzene in

the effluent of the RF and the PF showed a completely opposite

trend. In the RF system, a rapid fluctuation in the benzene

concentration values showing no particular reduction trend

resulted in a mean effluent benzene concentration of

65 � 123 mg L�1, whereas a relatively constant trend in

concentration reduction was observed in the effluent of the

PF. The mean value of 0.5 � 0.2 mg L�1 in the effluent

contributed to a highly efficient (w100%) reduction in the

benzene concentration of the PF system, as compared to the

RF (see phase IV; Fig. 3A and B).

3.2. Dynamics of MTBE and benzene: multi-stagesystem

The dynamics of MTBE and benzene in both the influent and

effluent of the combined multi-stage vertical-flow soil filter

system (RF þ PF) during operational period 2, in the experi-

mental phase V (corresponding to days 388e611), are shown in

Fig. 4.

The RF system as the first filter received contaminated

groundwater at an HLR of 960 L m�2 d�1. The mean influent

MTBE-concentration value of 2760� 594 mg L�1 was reduced to

a mean effluent value of 831 � 318 mg L�1, which resulted in

a mean MTBE-concentration reduction of 69% in this treat-

ment step. This effluent of the RF system was pumped inter-

mittently onto the surface of the second filter (PF) at an HLR of

480 L m�2 d�1. The results demonstrated a remarkable (w99%)

reduction in the MTBE concentration of the effluent of the PF

with a mean value of 5 � 10 mg L�1. Although the dynamics of

the MTBE concentration in the effluent of the PF showed

a rapid fluctuation in the values during this experimental

phase, all the effluent concentration values were well below

the limit value of 200 mg L�1 for MTBE.

A

B

Fig. 4 e Influent and effluent concentration along with the

limit value of A) MTBE and B) benzene in the multiphase

combined (RF D PF) system during experimental phase (V)

of operational period 2 (days 388e611).

wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 6 3e5 0 7 4 5069

In the case of benzene, the first filter (RF) received the

influent groundwater with a mean benzene concentration of

13,527 � 1638 mg L�1 and a drastic fluctuation was observed in

the effluent benzene concentration values of this filter

system. The values were spread out over a large range but

a mean effluent value of 291 � 573 mg L�1 resulted in a mean

reduction in the benzene concentration of 98% from the

influent, which did not meet the allowable limit value of

1 mg L�1. However, after passing through the second filter (PF),

a remarkably low and stable benzene concentration was

detected in the effluent of the PF system. The mean value of

Table 3 e Summary of the treatment performances in the RF a

Stage Phase Filter MTBE

Influent(mg L�1)

Effluent(mg L�1)

Remo(%)

Single I RF 3953 � 298 139 � 69 97

PF 4337 � 338 332 � 680 93

II RF 3850 � 680 399 � 318 91

PF 4207 � 456 289 � 370 93

III RF 2635 � 490 402 � 222 84

PF 3104 � 587 43 � 90 99

IV RF 2214 � 266 550 � 133 75

PF 2204 � 301 49 � 77 98

Multiple V RFa 2760 � 594 831 � 318 69

PFa 831 � 318 5 � 10 99

a Multi-stage combined system (RF þ PF), n: number of samples.

0.6 � 0.2 mg L�1 in the effluent of the second filter contributed

to an almost complete (w100%) removal of benzene in this

combined multi-stage system.

The overall treatment performances obtained in both

filters (RF and PF) during the whole operational period of this

study are summarized in Table 3.

The mean concentrations of intermediate degradation

products of MTBE (TBA, TBF) and other aromatic hydrocar-

bons such as toluene, ethylbenzene, m-p-Xylene, o-Xylene,

1,3,5-Trimethylbenzene, 1,2,4-Trimethylbenzene and Naph-

thalene in the influent groundwater and effluents of both the

RF and the PF are given in Table 4. Both TBA and TBF were

detected with a low mean concentration value in the effluent

of the RF and the PF system during the single-stage opera-

tional phase, but their concentrations were almost dimin-

ished or went below the detection limit in the final effluent

after passing the multi-stage system. Similarly, in the final

effluent of the multi-stage operational phase, the other

aromatic hydrocarbons could not be detected due to a very

low concentration value at the end (see Table 4).

3.3. Emission estimation

The emissions of the volatile organic compounds (VOCs) in

the air phase at several specified heights over the vertical-flow

soil filter systems at a given day in several measurement

cycles were registered and plotted on curves in this study. The

measurements were recorded during the single-stage opera-

tion period at the same HLR of 480 L m�2 d�1 in both filters

(experimental phase IV). An example of emission calculation

in one measurement cycle over the RF and PF is given in

Fig. 5A and B. In the RF, the inflow feeding pulse with duration

of nearly 4 min contributed to an immediate displacement of

inside trapped air to the filter surface and over a period of

approximately 17 min, the displaced air disappeared and the

emission level came back to the concentration at the back-

ground value until the end of the 60 min cycle (Fig. 5A).

Emissions of VOCs were measured in this 17 min time dura-

tion and the obtained results showed that the highest emis-

sion with a concentration of 12.27 mg TOC m�3 was recorded

nd the PF during the whole operational period of 611 days.

Benzene

val n Influent(mg L�1)

Effluent(mg L�1)

Removal(%)

n

40 15,574 � 2800 64 � 76 99 36

50 18,695 � 1578 0.3 � 0.2 100 26

92 15,126 � 2382 413 � 736 98 77

99 17,030 � 2664 11 � 53 100 61

47 13,046 � 1463 401 � 803 97 45

51 14,856 � 1115 0.3 � 0.2 100 46

83 13,649 � 1142 65 � 123 99 83

84 13,052 � 2462 0.5 � 0.2 100 46

154 13,527 � 1638 291 � 573 98 154

140 291 � 573 0.6 � 0.2 100 100

Table 4 e Concentration of intermediate degradation products of MTBE and other aromatic compounds analysed in theinfluent groundwater and the effluents of the RF and the PF system during the whole operational period of 611 days.

Substance Single-stage (phase IeIV) Multi-stage (phase V)

Influent(mg L�1)

Effluent, RF(mg L�1)

Effluent, PF(mg L�1)

n Influent(mg L�1)

Effluent, RF(mg L�1)

Effluent, PF(mg L�1)

n

TBA 53 � 12 13 � 8 4 � 3 24 41 � 6 13 � 3 2 � 1 12

TBF 4.1 � 1 2 � 1.5 2 � 1.5 24 2.7 � 1.3 1.3 � 0.5 n.d. 11

Toluene 8 � 1.4 1.1 � 0.1 n.d. 24 6.7 � 0.7 n.d. n.d. 11

Ethylbenzene 50 � 37 1.4 � 0.1 1.1 � 0.2 22 31 � 14 1.1 � 0.3 n.d. 11

m-p-Xylene 76 � 53 1.3 � 1.0 1.3 � 0.5 22 57 � 7 n.d. n.d. 11

o-Xylene 6.5 � 2.9 1.2 � 0.6 n.d. 23 6.7 � 0.9 n.d. n.d. 11

1,3,5-Trimethylbenzene 4.7 � 3.6 1.6 � 1.0 n.d. 21 2.4 � 1.5 n.d. n.d. 9

1,2,4-Trimethylbenzene 393 � 141 2.6 � 2 1.2 � 0.6 24 252 � 63 4.1 � 3.1 n.d. 12

Naphthalene 68 � 22 1.2 � 0.6 n.d. 24 36 � 9 n.d. n.d. 12

n.d.: below the limit of detection (<1 mg L�1), n: number of samples.

wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 6 3e5 0 7 45070

at the height of 10 cm over the inlet pipe (buried underneath

the top layer of the RF). The concentration of VOCs decreased

down to background level after this 17 min time period.

Therefore, the net emission at the highest concentration

measurement point was estimated as 10.55 mg TOC m�3,

which was achieved by subtracting the mean background

value of 1.72 � 0.06 mg TOC m�3 from the highest measured

emission value (H10 Inlet data set, Fig. 5A).

For better estimation in the PF system, the inflow feeding

pulse duration was adjusted to nearly 8 min and measure-

ment values were recorded over the next 8e10 min after the

dosing. Themaximum concentration of 3.18 mg TOCm�3 was

registered at the height of 10 cm over the central middle point

A

B

Fig. 5 e Concentration of VOCs emission from the surface

of the A) RF and B) PF measured overall sampling heights in

cycle 1 with the same HLR in the single-stage operation.

of the filter surface. The net emission at that highest

concentration measurement point was calculated as

0.72 mg TOC m�3, by subtracting the measured mean back-

ground value of 1.4 � 0.02 mg TOC m�3 from the highest

emission value measured in this measuring cycle over the PF

(H10 Middle data set, Fig. 5B). The regional background levels

were consistent during the whole emission measurement

experiment with values in the range of 1e2 mg TOC m�3

measured in the air.

After each feeding pulse, it was assumed that a total

volume of 80 L water (Table 2) was flushing on the filter

surface and the same volume of 80 L entrapped air was

coming out from the filter over the surface within a short time

(8e20 min) and then disappeared. Based on this assumption

for a preliminary emission calculation, it was observed that

amass of 42.77mg TOCm�2 d�1 was emitted over the segment

of the RF system. This was calculated quantitatively by the

triangular area under the curve (actual emission measuring

zone by connecting the H10 Inlet data set, Fig. 5A) multiplied

by the HLR of 480 L m�2 d�1. After this particular emission

zonewithin the curve, the dynamics ofmeasured air emission

came back to the concentration at the background level until

the cycle ends and hence they were not taken into account in

this calculation. Comparing to the inflow TOC mass loading

rate of 7566.94 mg TOC m�2 d�1 to the filter bed, it was

observed that only 0.45% of the inflow TOC mass was emitted

over the surface and went into the surrounding atmosphere.

Similar calculation approach by using the H10 Middle data set

(Fig. 5B) estimated that the emitted VOCs mass percentage

value was even lower (0.04%) in the PF system, as compared to

the RF system under the same HLR. The summary of emission

measurement calculations in each cycle over the RF and the

PF system is given in Table 5.

4. Discussion

4.1. MTBE and benzene removal performances: single-stage systems

During the stepwise increase of the HLR (60e480 L m�2 d�1) in

the first operational period of the single-stage systems, the RF

Table 5 e Summary of the emission measurement cycles during a particular day over the surface of the RF and the PFsystem with the same HLR obtained in single-stage operation.

Stage Filter Measuringcycle

Inflow mass loading(mg TOC m�2 d�1)

Emitted mass over segment(mg TOC m�2 d�1)

Emission (%)

Single

(HLR: 480 L m�2 d�1)

RF 1 7566.94 42.77 0.45

2 50.76 0.67

3 30.23 0.40

4 21.41 0.28

PF 1 7277.39 2.98 0.04

2 1.37 0.02

wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 6 3e5 0 7 4 5071

clearly showed a decreasing tendency in the mean MTBE

removal (%), which means the lower the HLR, the better the

MTBE removal from this system (Table 3). Moreover, the

overall reductions in MTBE concentrations were not sufficient

to reach the limit value of 200 mg L�1 (USEPA, 2005; DVGW,

2001). In contrast, the PF showed a better performance of

increasing the MTBE removal (%) associated with increases in

the HLR. At the highest HLR of 480 L m�2 d�1 in the PF, the

mean effluent MTBE concentration was recorded to be

49 � 77 mg L�1, which was nearly 7 times less than the mean

effluent MTBE concentration (332 � 680 mg L�1) at the lowest

HLR of 60 Lm�2 d�1 andwell below the limit value of 200 mg L�1

for MTBE. However, although the mean MTBE concentration

of 49 � 77 mg L�1 was below the limit value, a rapid fluctuation

in the dynamics of the effluent MTBE concentration was

observed and nearly 10% of the measured values were higher

than the limit value of 200 mg L�1 (see Fig. 2B). This indicates

that theMTBE removal performance obtained in the PF system

was sufficient, but not stable.

In the case of benzene, a mean removal within a range of

97e99% was achieved in the RF, but the overall effluent

concentrations were never below the limit value of 1 mg L�1. In

contrast, the PF system exhibited a highly efficient benzene

concentration reduction and the effluent concentrations were

predominantly found to be below 1 mg L�1 (see Fig. 3A and B).

The differences in the MTBE and benzene treatment

performance observed in the RF and PF might be due to the

differences in filter designs mainly defined by the different

filtering media used as the main filter materials (expanded

clay in the RF and zeosoil in the PF; Fig. 1).

The obtained results show that a filter loaded with a fine

zeosoil (0e5 mm) filter material and plants is more efficient in

MTBE and benzene removal than a filter loaded with a coarse

expanded clay material (3e6 mm) without plants. This is

probably due to a higher reactive surface area, a better oxygen

transfer and a higher hydraulic retention time (data not

shown) within the filter loaded with fine materials. But finer

materials have the disadvantage of a possible filter clogging

and also the problem associated with water saturation at high

hydraulic loads.

Additionally the differences in effective depth and

different compactions of the filter bed, different gas exchange

rates and planteroot activity in the case of the PF that provide

oxygen to the rhizosphere (Scholz, 2006) might explain the

observed differences in treatment performance. However,

more investigations are needed beforemaking any concluding

remarks on these particular assumptions.

An effective benzene biodegradation could be expected in

the two RF and PF filter systems, since this pollutant has been

degraded in environmental systems even under hypoxic

conditions and treatment efficiencies for aerobic bioreactors

up to 100% have been described (Yerushalmi et al., 2002). In

contrast, MTBE biodegradation is reported to be by far not as

effective as benzene biodegradation. The possible reasons

might be thatMTBE ismore resistant to enzymatic attacks due

to its tertiary carbon atom and the ether bond (Davidson and

Creek, 2000). Moreover, it is reported that the biodegradation

of MTBE might be inhibited due to the presence of co-

contaminants, such as benzene, ethylbenzene, toluene and

xylene (BTEX), and the accumulation of by-products from the

biodegradation of BTEX compounds (Raynal and Pruden,

2008). An inhibition of MTBE biodegradation in the presence

of BTEX due to a potential production of by-products has also

been suggested by others (Deeb et al., 2001; Sedran et al., 2002).

These studies have focused mainly on substrate inhibition

(Park, 1999), by-product inhibition (Wilson et al., 2002) or

competitive inhibition (Sedran et al., 2002).

Therefore, the presence of BTEX compounds in the

groundwater of the refinery site was expected to inhibit MTBE

biodegradation in both the RF and PF systems. But these

inhibitory effects on MTBE biodegradation could not be

observed in both the filters during this study. Table 4 shows

the presence of BTEX and other aromatic hydrocarbons in the

influent groundwater with a mean concentration value and

still a highly efficient MTBE-concentration reduction (93e98%)

can be seen especially in the effluent of the PF during single-

stage operation (Table 3). However, very little is known

about the microbial community structure during the aerobic

MTBE degradation in the presence of BTEX.

The disappearance of MTBE metabolites, such as tert-butyl

alcohol (TBA) and tert-butyl formate (TBF) indicated the

potential of complete biodegradation within the filter

systems. As can be seen in Table 4, the mean concentrations

of TBA and TBF were remarkably decreased in the effluent of

the RF and the PF system during the single-stage operation

period and almost completely diminished or biodegraded

after passing through the second filter (PF) during the multi-

stage operation period.

4.2. MTBE and benzene removal performances: multi-stage system

The second operational period with a multi-staged combined

(RF þ PF) vertical-flow soil filter system runs very well with

wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 6 3e5 0 7 45072

a highly efficient MTBE and benzene removal. Both MTBE and

benzene effluent concentrations were always stable and well

below the allowable limit value after the second treatment

step. Moreover, the high HLR in the first filter (RF) was not

accompanied by hydraulic or technical problems, such as

clogging, overloading, etc., and it was thus very encouraging

to a further increase of HLR, which could reduce the cost of the

required land for the filter construction. Therefore, it can be

concluded that the multi-stage combined vertical-flow soil

filter system is more stable, more effective and a better option

for the removal of MTBE and benzene from contaminated

groundwater, as compared to the single-stage system.

In principle, the overall decrease in the MTBE and benzene

concentrations obtained from the vertical-flow soil filter

systems can be caused by microbial degradation, sorption

onto solid filter packing materials and volatilization. More-

over, in the planted filter system, it might also be caused by

plant uptake followed by transport, transformation and

phyto-volatilization. A long-term operation of the vertical-

flow soil filter systems is leading to an established adsorp-

tion/desorption balance and therefore, a removal by sorption

onto the filtering media can be assumed to be negligibly small

in this investigation. Biodegradation of MTBE and benzene is

expected to be the most dominant process for the removal of

these contaminants from the groundwater. However, the

extent of degradation cannot be estimated accurately without

a long-term and complete set of data in terms of volatilization

and plant uptake from pilot-scale vertical-flow constructed

wetlands. Eke and Scholz (2008) also concluded that the

impacts of volatilization, biodegradation and adsorption on

the benzene removal are often difficult to separate quantita-

tively from each other.

For a long-term operation of the vertical-flow soil filters,

the designed hydraulic loads need to be achieved by opti-

mising the volume of the water in each loading pulse and the

associated frequency, in order to increase the dewatering

efficiency of the filters in the period of time between the

intermittent pulses, and thus promoting oxygenation and

achieving treatments with the highest possible level of

efficiency.

4.3. Emissions

After estimating the emissions of volatile organic compounds

from the vertical-flow soil filter systems, the overall results

indicated that the emissions from the planted PF systems in

the air were much lower than those from the RF system and

were only slightly above the background value (Fig. 5).

By comparing to the inflow TOC mass loading rate, only

a negligible amount (<1%) was emitted from the surface of the

both RF and the PF systems. Therefore, with a highly efficient

mean MTBE and benzene concentration reduction in the

effluent of the RF and the PF system and almost a negligible

emission rate of VOCs mass, it can be concluded that the

biodegradation is the predominant removal pathway of both

MTBE and benzene within the vertical-flow soil filter system

treating contaminated groundwater.

Volatilization of toxic organic hydrocarbons may be

increased by technological problems, such as clogging and

subsequent flooding, and may lead to serious air pollution

(Braeckevelt et al., 2008). But both of our vertical-flow soil filter

systems were almost free of technical problems such as filter

clogging, overloading, surface flooding, etc. Experimental

investigations have shown that phyto-volatilization is

a potential emission path for MTBE and benzene along with

the direct volatilization via the soil surface of a constructed

wetland (Reiche et al., 2010). However, more improved tech-

nical equipment is necessary for measuring both the VOCs

concentration in the air and the volume of air emitted from

the surface of the filter beds. Future investigations should be

carried out with the purpose of a final evaluation of the

volatilization rate of MTBE and benzene per unit area (m2) of

the filter surface and with the aim of achieving a complete

mass balance of organic compounds and discovering the role

of the cover layer for protecting volatilization.

5. Conclusions

The following conclusions can be drawn from the current

study:

1. The Polishing Filter (PF) with a finer material and plants is

more efficient in removing MTBE and benzene from

contaminated groundwater, as compared to the Roughing

Filter (RF) with a coarse material and without plants.

2. Factors, such as filter packing material, particle size, filter

depth and loading rate, are playing an important role in

achieving a robust filter operation for the removal of

organic contaminants by vertical-flow soil filter systems.

3. The MTBE removal performance decreases with an

increasing HLR in the RF, whereas the PF system is char-

acterized by a remarkable MTBE and benzene removal

performance at an increasing HLR.

4. At a higher HLR, the MTBE removal performance of

a single-stage vertical-flow soil filter system is often

stable, but not sufficient.

5. In general, a continuous reduction in both the MTBE and

benzene concentration of the effluent indicates that the

maximum treatment capacity is yet to be reached in both

the RF and the PF systems.

6. The multi-stage combined vertical-flow soil filter system

(RF þ PF) produces the most stable and sufficient effluent

concentrations to reach the limit concentrations of MTBE

and benzene for drinking water.

7. Since a negligible amount of volatile organic compounds

is going in the air from our filter systems, therefore they

are not considered to be a potential source of air pollution

affecting the surrounding environment.

8. Since the vertical-flow constructed wetlands are accu-

mulative systems (biomass, organic matter, calcareous

material, etc.), it is of great importance to assess the

optimal operation/design load of the filters treating MTBE

and benzene and to predict the cases in which hydraulic

overloads might be problematic for the filter longevity.

9. Our systems are designed to minimize clogging and after

treating groundwater by using our technology, the water

will not need to be amended further and can be released

into an aquifer or discharged into any conventional

drainage system.

wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 6 3e5 0 7 4 5073

10. This novel groundwater remediation technology promises

to be a cost effective remediation approach for treating

groundwater contaminated with MTBE and benzene on

a full-scale.

11. To prove the long-term stability and process optimization

as well as to reach a sound economical and ecological

evaluation for this new approach, a pilot-scale or full-

scale operation over an extended period of time is needed.

Further studies are intended: i) to focus on identifying the

major microbial processes that lead to an aerobic biodegra-

dation of organic contaminants, ii) to quantify volatilization,

adsorption, absorption, mineralization and other removal

mechanisms in large-scale vertical-flow soil filter systems

treating MTBE and benzene, iii) to characterize and quantify

plant uptake, phyto-sorption and phyto-volatilization of

MTBE and benzene, iv) to explore the effects of seasonal

temperature changes on the removal of MTBE and benzene, v)

to investigate the effects of iron and calcium precipitations on

the filter performance in treating MTBE and benzene

contaminated groundwater, vi) to define the design criteria for

the remediation of contaminated groundwater using vertical-

flow soil filter eco-technologies.

Acknowledgements

This work was supported by the Helmholtz Centre for Envi-

ronmental Research e UFZ in the scope of the SAFIRA II

Research Programme (Revitalization of Contaminated Land

and Groundwater at Megasites, sub-project ‘‘Compartment

Transfer e CoTra’’) and funded by a grant from the German

Federal Ministry of Education and Research (BMBF). The

authors would like to thank Francesca Loper, Grit Weichert,

Sibylle Mothes, Heidrun Paschke, and Christina Petzold for

their assistance in the field and laboratory work.

r e f e r e n c e s

Baehr, A.L., Stackelberg, P.E., Baker, R.J., 1999. Evaluation of theatmosphere as a source of volatile organic compounds inshallow groundwater. Water Resources Research 35, 127e136.

Braeckevelt, M., Mirschel, G., Wiessner, A., Rueckert, M.,Reiche, N., Vogt, C., Schultz, A., Paschke, H., Kuschk, P.,Kaestner, M., 2008. Treatment of chlorobenzene-contaminated groundwater in a pilot-scale constructedwetland. Ecological Engineering 33, 45e53.

Cooper, P.F., 1999. A review of the design and performance ofvertical-flow and hybrid reed bed treatment systems. WaterScience and Technology 40, 1e9.

Davidson, J.M., Creek, D.N., 2000. Using the gasoline additiveMTBE in forensic environmental investigations.Environmental Forensics 1 (1), 31e36.

Davis, S., Powers, S., 2000. Alternative sorbents for removingMTBE from gasoline-contaminated ground water. Journal ofEnvironmental Engineering 126 (4), 354e360.

Deeb, R.A., Chu, K.H., Shih, T., Linder, S., Suffet, I., Kavanaugh, M.C., Alvarez-Cohen, L., 2003. MTBE and other oxygenates:

environmental sources, analysis, occurrence and treatment.Environmental Engineering Science 20 (5), 433e444.

Deeb, R.A., Hu, H.-Y., Hanson, J.R., Scow, K.M., Alvarez-Cohen, L.,2001. Substrate interaction in BTEX and MTBE mixtures by anMTBE-degrading isolate. Environmental Science andTechnology 35 (2), 312e317.

Deeb, R.A., Scow, K.M., Alvarez-Cohen, L., 2000. Aerobic MTBEbiodegradation: an examination of past studies, currentchallenges and future research directions. Biodegradation 11,171e186.

DIN 32645, 1994. “Chemische Analytik”, Nachweis-, Erfassungs-und Bestimmungsgrenze e Ermittlung unterWiederholungsbedingungen e Begriffe, Verfahren,Auswertung. Hrsg.: Deutsches Institut fur Normung, Beuth-Verlag, Berlin, pp. 1e20 (in German).

DVGW, D.V.d.G.-u.W.e.V., 2001. Verordnung zur Novellierung derTrinkwasserverordnung vom 21. Mai 2001, Bonn, Germany (inGerman).

Eke, P.E., Scholz, M., 2008. Benzene removal with vertical-flowconstructed treatment wetlands. Journal of ChemicalTechnology Biotechnology 83, 55e63.

Ferreira, N.L., Malandain, C., Fayolle-Guichard, F., 2006. Enzymesand genes involved in the aerobic biodegradation of methyltert-butyl ether (MTBE). Applied Microbiology andBiotechnology 72 (2), 252e262.

Johnson, R., Pankow, J., Bender, D., Price, C., Zogorsky, J., 2000.MTBE to what extent will past releases contaminatecommunity water supply wells? Environmental Science andTechnology 34, 210Ae217A.

Kadlec, R.H., 2001. Thermal environments of subsurfacetreatment wetlands. Water Science and Technology 44 (11-12),251e258.

Kadlec, R.H., Wallace, S.D., 2009. Treatment Wetlands, second ed.Taylor and Francis Group, Boca Raton, USA, ISBN 978-1-56670-526-4.

Kassenga, G.R., Pardue, J.H., Moe, W.M., Bowman, K.S., 2004.Hydrogen thresholds as indicators of dehalorespiration inconstructed treatment wetlands. Environmental Science andTechnology 38 (4), 1024e1030.

Lorah, M.M., Voytek, M.A., 2004. Degradation of 1,1,2,2-tetrachloroethane and accumulation of vinyl chloride inwetland sediment microcosms and in situ porewater:biogeochemical controls and associations with microbialcommunities. Journal of Contaminant Hydrology 70, 117e145.

Martienssen, M., Fabritius, H., Kukla, S., Balcke, G.U.,Hasselwander, E., Schirmer, M., 2006. Determination ofnaturally occurring MTBE biodegradation by analysingmetabolites and biodegradation by-products. Journal ofContaminant Hydrology 87, 37e53.

Mleczek, M., Rutkowski, P., Rissmann, I., Kaczmarek, Z.,Golinski, P., Szentner, K., Strazynska, K., Stachowiak, A., 2010.Biomass productivity and phytoremediation potential of Salixalba and Salix viminalis. Biomass and Bioenergy 34, 1410e1418.

Moreels, D., Bastiaens, L., Ollevier, F., Merckx, R., Diels, L.,Springael, D., 2006. Evaluation of the intrinsic methyl tert-butyl ether (MTBE) biodegradation potential of hydrocarboncontaminated subsurface soils in batch microcosm systems.FEMS Microbiology Ecology 49, 121e128.

Park, K., 1999. Biodegradation of the Fuel Oxygenate, Methyl Tert-butyl Ether (MTBE), and Treatment of MTBE ContaminatedGroundWater in Laboratory Scale Reactors. Ph.D. dissertation,State University of New Jersey, New Jersey, USA.

Raynal, M., Pruden, A., 2008. Aerobic MTBE biodegradation in thepresence of BTEX by two consortia under batch and semi-batch conditions. Biodegradation 19, 269e282.

Reiche, N., Lorenz, W., Borsdorf, H., 2010. Development andapplication of dynamic air chambers for measurement ofvolatilization fluxes of benzene and MTBE from constructed

wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 0 6 3e5 0 7 45074

wetlands planted with common reed. Chemosphere 79,162e168.

Rentz, J.A., Alvarez, P.J.J., Schnoor, J.L., 2005. Benzo[a]pyrene co-metabolism in the presence of plant root extracts andexudates: implications for phytoremediation. EnvironmentalPollution 136, 477e484.

Rubin, E., Ramaswami, A., 2001. The potential forphytoremediation of MTBE. Water Research 35, 1348e1353.

Schmidt, T.C., Schirmer, M., Weiss, H., Haderlein, S.B., 2004.Microbial degradation of methyl tert-butyl ether and tert-butylalcohol in the subsurface. Journal of Contaminant Hydrology70 (3e4), 173e203.

Scholz, M., 2006. Wetland Systems to Control Urban Runoff.Elsevier, Amsterdam.

Sedran, M.A., Pruden, A., Wilson, G.J., Suidan, M.T., Venosa, A.D.,2002. Effect of BTEX on degradation of MTBE and TBA bymixed bacterial consortium. Journal of EnvironmentalEngineering 128 (9), 830e835.

Squillace, P.J., Zogorski, J.S., Wilber, W.G., Price, C.V., 1996.Preliminary assessment of the occurrence and possiblesources of MTBE in groundwater in the United States,

1993e1994. Environmental Science and Technology 30 (5),1721e1730.

Sutherland, J., Adams, C., Kekobad, J., 2004. Treatment of MTBE byair stripping, carbon adsorption, and advanced oxidation:technical and economic comparison for five groundwaters.Water Research 38, 193e205.

USEPA, 2004. Technologies for Treating MTBE and Other FuelOxygenates. Office of Superfund Remediation and TechnologyInnovation, Washington, DC, USA. EPA-542/R-04-009.

USEPA, 2005. List of Drinking Water Contaminants and MCLsFrom: http://water.epa.gov/drink/index.cfm.

Wilhelm, M., Adams, V., Curtis, J., 2002. Carbon adsorption andair-stripping removal of MTBE from river water. Journal ofEnvironmental Engineering 128 (9), 813e823.

Wilson, R.D., Mackay, D.M., Scow, K.M., 2002. In situ MTBEbiodegradation supported by diffusive oxygen release.Environmental Science and Technology 36, 190e199.

Yerushalmi, L., Lascourreges, J.F., Guiot, S.R., 2002. Kinetics ofbenzene biotransformation under microaerophilic andoxygen-limited conditions. Biotechnology and Bioengineering79 (3), 347e355.