w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 1 0 2 e5 1 1 4

Available online at www.sciencedirect.com

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

Halophyte filter beds for treatment of saline wastewater from

aquaculture

J.M. Webb a, R. Quinta a, S. Papadimitriou a, L. Norman a, M. Rigby b, D.N. Thomas a,c,

L. Le Vay a,*
a
School of Ocean Sciences, College of Natural Sciences, Bangor University, Wales, LL59 5AB, UK
b
Llyn Aquaculture Ltd, Afonwen Farm, Chwilog, Pwllheli, Gwynedd, Wales, LL53 6TX, UK
c
Finnish Environment Institute, Marine Research Center, Helsinki, Finland

article info abstract

Article history: The expansion of aquaculture and the recent development of more intensive land-based
Received 22 December 2011 marine farms require efficient and cost-effective systems for treatment of highly
Received in revised form nutrient-rich saline wastewater. Constructed wetlands with halophytic plants offer the
11 June 2012 potential for waste-stream treatment combined with production of valuable secondary
Accepted 20 June 2012 plant crops. Pilot wetland filter beds, constructed in triplicate and planted with the salt-
Available online 28 June 2012 marsh plant Salicornia europaea, were evaluated over 88 days under commercial operating
conditions on a marine fish and shrimp farm. Nitrogen waste was primarily in the form of
Keywords: dissolved inorganic nitrogen (TDIN) and was removed by 98.2 ! 2.2% under ambient
Salicornia loadings of 109e383 mmol l"1. There was a linear relationship between TDIN uptake and
Saltmarsh loading over the range of inputs tested. At peak loadings of up to 8185 ! 590 mmol l"1
Constructed wetland (equivalent to 600 mmol N m"2 d"1), the filter beds removed between 30 and 58%
Filter bed (250 mmol N m"2 d"1) of influent TDIN. Influent dissolved inorganic phosphorus levels
Aquaculture ranged from 34 to 90 mmol l"1, with 36e89% reduction under routine operations. Dissolved
Waste water organic nitrogen (DON) loadings were lower (11e144 mmol l"1), and between 23 and 69% of
Halophyte influent DON was removed during routine operation, with no significant removal of DON
under high TDIN loading. Over the 88-day study, cumulative nitrogen removal was
1.28 mol m"2, of which 1.09 mol m"2 was retained in plant tissue, with plant uptake
ranging from 2.4 to 27.0 mmol N g"1 dry weight d"1. The results demonstrate the effec-
tiveness of N and P removal from wastewater from land-based intensive marine aqua-
culture farms by constructed wetlands planted with S. europaea.
2012 Elsevier Ltd. All rights reserved.

1. Introduction Marine and coastal aquaculture represent a significant

Recent estimates show that aquaculture provides 47% (51
million MT) of global human fish consumption (FAO, 2009). In
order to keep up with population growth and increasing per
capita fish consumption, aquaculture output is set to increase
by a further 60e100% over the next 20e30 years (FAO, 2009).
component of seafood production and current expansion of
this sector is largely the result of the intensification of marine
farming of carnivorous fish and shrimp species in coastal
cages and ponds. In the main, these are flow through or open
systems which can discharge high volumes of wastewater
containing suspended solids (excreta and food waste) and

* Corresponding author. Tel.: 44 1248 388115.

E-mail address: l.levay@bangor.ac.uk (L. Le Vay).
0043-1354/$ e see front matter 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.watres.2012.06.034
 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 1 0 2 e5 1 1 4
dissolved metabolites in the form of organic matter and
inorganic nitrogen and phosphorus. With fish farms being
subject to best practice codes of conduct (Boyd, 2003), and
legislation imposing financial penalties on the polluter (e.g. EU
Water Framework Directive 2000/60/CE), development of
land-based intensive marine recirculating aquaculture
systems (RAS) offers significant potential to treat reduced
volumes of wastewater prior to discharge (Tal et al., 2009).
In conventional aquaculture wastewater treatment
systems, solids are removed by gravitational and/or
mechanical methods. Settlement is used to remove the denser
solids, while filtration (commonly screen filtration, expand-
able granular biofilters (EGBs), and foam fractionation) is used
for removing suspended and fine solids (Cripps and Bergheim,
2000; Piedrahita, 2003). Removal of dissolved metabolites
requires oxidation of organic matter and steps to promote
nitrification or denitrification. Ion-exchange and carbon filters
are quickly biofouled and ion-exchange filters are rendered
inactive in ion-rich seawater and these methods, although
effective, are costly both in terms of capital investment,
energy consumption and maintenance requirements.
Constructed wetlands (CW) are a well-established, cost-
effective, method for treating wastewater, such as municipal
or domestic sewage, industrial and agricultural wastewater,
landfill leachate, and stormwater runoff. A constructed
wetland is a system engineered to recreate the vegetation,
sediment, and microbial assemblages found within its natu-
rally occurring counterparts, whilst utilizing the biogeo-
chemical processes occurring therein (Vymazal, 2005). The
role of the higher plants is crucial in establishing a successful
CW since they maintain the hydraulic conductivity of the
substrate, increase microbial assemblages in the root zone
and participate in nutrient uptake. CW design usually consists
of a lined bed filled with porous media (generally rock or
gravel) and planted with emergent hydrophytes and
commonly employed designs incorporate horizontal sub-
surface flow (HF), where pre-treated waste water enters the
CW at the inlet and travels slowly down the length of the bed,
passing through the filtration media under the surface of the
bed until it reaches the outlet/discharge point. Remediation
relies on a combination of physical, chemical and biological
processes, sedimentation, precipitation, volatilization,
adsorption to soil particles, plant uptake and microbial
conversion. Several pilot scale and field studies have been
carried out and demonstrated the viability of using CWs to
treat aquaculture wastewater, mainly in freshwater and some
in brackish water systems (e.g. Lin et al., 2005), however, to
date there has been limited work using CWs with marine
systems (Lymbery et al., 2006).
Freshwater RASs are ideally suited to aquaponics, where
plants are grown in an inert medium (e.g. perlite, sand, gravel,
rockwool) supplied with wastewater from the RAS, with the
aim of reducing overall organic and inorganic nutrient load
while producing a crop (Lennard and Leonard, 2006). To apply
a similar approach to marine systems would require
a commercially-valuable halophytic plant that can thrive in
saline wastewater (Calheiros et al., 2012). One of the few
candidates is samphire (Salicornia spp.), an edible succulent
leafless genus that grows in saltmarshes and saline environ-
ments (Davy et al., 2001). Salicornia spp. have a commercial
5103
value as an oil seed crop, a seasonal vegetable and also for
potential use in the health, beauty and nutraceutical indus-
tries (Glenn et al., 1991; Rhee et al., 2009). The present study set
out to test the effectiveness of a CW planted with Salicornia
europaea agg. (L) for treating effluent water from a commer-
cially-operating marine fish and shrimp RAS.
2. Materials and methods
2.1. Filter design and operation
Triplicate pilot water treatment filter beds were installed in
a single span poly-tunnel, 5 m $ 20 m (W $ L) on an intensive
marine fish farm in Pwllheli, North Wales, UK. Each bed had
14.5 m2 surface area and 4.35 m3 volume (1 m $ 14.5 m $ 0.3 m,
W $ L $ H ). The filter beds were constructed of timber frames
on a sand base, with butyl rubber liners, and filled to 200 mm
depth with 40 mm clean, graded, single size smooth limestone
(Cefn Graianog Quarry, Chwilog, UK), to allow subsurface flow.
This was overlaid with a 100 mm layer of %6 mm, mixed M
grade quarry to a depth of 100 mm (Fig. 1). The two layers were
separated by a sheet of plastic mesh (2 mm2 pore size; www.
boddingtons-ltd.com insect mesh, ref. 47000) overlaid with
17 g m"2 geotextile frost protection fleece (http://www.
lbsgardenwarehouse.co.uk, ref. R-F2050). This combination
was designed to provide a semi-permeable barrier, preventing
sand from falling into the pore spaces between the larger
stones below, whilst allowing root growth into the lower stone
layer. At each end, a perforated 300 mm diameter PVC pipe
was fitted vertically through the gravel and sand layers,
serving as both filling and sampling points as well as water
level indicators.
Salicornia europaea agg (L) were grown from seeds taken
from 2nd generation cultivated plants. Germination took
place in a controlled environment greenhouse (photoperiod of
16 light:8 dark and temperature of 18 & C). Seeds were sown
onto the surface of P576 plug trays filled with John Innes No.1
compost and irrigated with fresh water. After two weeks,
seedlings emerged and the trays were thinned out to 1 plant
per plug. Irrigation salinity was increased to 10 psu with
TROPICMARIN! artificial sea salt and seedlings received
Phostrogen soluble plant feed (N:P:K 14:10:27 trace
elements; Bayer CropScience Ltd, Cambridge, UK). Two-
month old plants were transplanted into the filter beds at
a density of 90 m"2, so that each bed contained approximately
1250 plants.
The pilot filter beds processed waste water from
a commercially-operating intensive recirculating marine
aquaculture facility (Llyn Aquaculture Ltd) producing shrimp,
sole and turbot (Fig. 2). They were operated on a batch-
treatment, flood and drain system, as standard water
management for the facility requires once-daily discharge of
batches of waste water flushed from sediment traps, and
weekly batches of waste water used to backflush biofilter and
bead filter units. All wastewater drained to an outdoor,
uncovered settlement pond (10 $ 4 $ 0.5 m), from which it was
pumped (400 W 240 V automatic dirty water pump; www.
screwfix.com) into a covered and lined 12 m3 header tank
via a 300 L vortex separation tank containing plastic biofilter
5104 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 1 0 2 e5 1 1 4
Fig. 1 e Cross-section of pilot filter bed (a) Lengthwise and (b) width wise.
media removing suspended particulate matter. The header
tank wastewater was pumped to each of the three filter beds
via a 25 mm flexible pipe and a Kent V110 25 mm volumetric
flow meter. The filter beds were filled to just below the surface
of the upper sand layer to prevent algal fouling of the filter
surface and each was fitted with a moveable standpipe which
could be lowered to allow complete draining of the systems.
The interval between each flood and drain procedure was 24 h.
Sub-surface flow through the gravel layer was continuously
maintained by a submersible 18 W pump (Hozelock Cascade
700) located in the sump at one end of each filter bed and
discharging via a 15 mm flexible hose to the sump at the
distant end (Fig. 1b).
2.2. Simulated high nitrogen load
The pilot operation of the filter beds ran for a total of 88 days.
After the first 58 days, nitrogen loading was increased by
dosing the header tank with ammonium nitrate agricultural
grade fertiliser, to increase nitrogen concentrations to those
known to occur at peak discharge levels for this facility. Apart
from this, water management, filter bed operation and
sampling remained unchanged until the end of the test
period.
2.3. Water sampling
Water samples were taken at 0 and 24 h after filling, three
times a week from each of the 3 filter beds over the 88 days.
Water pH within each filter bed was measured using a Hanna
HI 98127 pH meter. Salinity was measured with a WTW sali-
nometer. Water temperature was measured using a Tinytag
Aquatic (TG-3100) data logger, while ambient air temperature
and humidity were measured on a Tinytag View 2 Int Temp/
RH (TV-4500) data logger (http://www.geminidataloggers.
com). Prior to filling, filter beds were flushed to remove
water remaining from the previous flooding. Immediately
after re-filling, triplicate 20 ml water samples were taken from
each filter bed using a 20 ml disposable sterile plastic syringe
fitted with a 30 cm length of acid-washed Teflon tubing.
Sample water was flushed through the syringe and tubing
twice before the sample was then passed through a Whatman
GD/X syringe filter (25 mm, nominal pore size of 0.45 mm) into
an acid washed 20 ml plastic bottle. A fourth 20 ml water
sample was taken from each bed, and was filtered similarly
into 2 $ 5 ml glass vials and fixed with phosphoric acid for
subsequent dissolved organic carbon (DOC) analyses. All
samples were frozen ("20 & C) until analyses. Following each
fill, the oxygen content, pH and the salinity of the water in
each filter bed were measured. Evaporation and transpira-
tional water loss was quantified by topping up each filter bed
to original fill levels after 24 h.
2.4. Plant growth
The plants in each filter were cropped every three weeks, with
all growth 10 cm above the sediment surface removed. The
total plant biomass removed from each filter bed at each
cropping is termed plant yield. Prior to each cropping, four
randomly selected plants were removed completely from
Fig. 2 e Schematic overview of water flow from seawater intake, through aquaculture production units to wastewater
treatment.
5106 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 1 0 2 e5 1 1 4
each filter bed. The roots were rinsed carefully to remove
sand, and plants were separated into below-ground (root) and
above-ground biomass, which were oven-dried at 50 & C and
weighed, Subsamples were then taken and were finely ground
for biochemical analysis.
2.5. Water analysis
Dissolved ammonium (NH 4 ) was determined with the fluori-
metric method of Holmes et al. (1999) using a HITACHI F2000
fluorescence spectrophotometer. The major dissolved inor-
ganic nutrients, nitrate, nitrite, dissolved inorganic phos-
phorus (DIP) and silicic acid [Si(OH)4], were determined
using standard colourimetric methodology (Grasshoff et al.,
1983) as adapted for flow injection analysis (FIA) on
a 5echannel LACHAT Instruments Quick-Chem 8000 auto-
analyzer (Hales et al., 2004). These analyses were monitored
daily using certified reference oceanic water (batch 54, Scripps
Institute of Oceanography, University of California, San Diego)
as an external standard and riverine water collected from the
River Clwyd, UK, as an internal standard. The external stan-
dard yielded: [NO" 2 ] 0.02 (!0.02) mmol kg
"1
(n 34; reported
[NO"2 ] 0.00 mmol kg "1
), [NO "
3 ] 1.27 (!0.15) mmol kg"1 (n 31;
reported [NO" "1
3 ] 1.25 mmol kg ), [DIP] 0.34 (!0.03) mmol kg
"1
"1
(n 27; reported [DIP] 0.39 mmol kg ). The internal
standard yielded, [NO" 2] 1.70 (!0.12) mmol kg"1
"
(n 12), [NO3 ] 117 (!3) mmol kg"1 (n 17),
[DIP] 2.60 (!0.18) mmol kg"1 (n 11), and
[Si(OH)4] 17 (!1) mmol kg"1 (n 13).
Dissolved organic nitrogen (DON) was determined by
subtraction of NO" "
3 ,NO2 and NH4 from the total dissolved
nitrogen (TDN) analyzed by FIA on the LACHAT autoanalyzer
using on-line peroxodisulphate oxidation coupled with
ultraviolet radiation at pH 9.0 and 100 & C (Kroon, 1993), with
a precision of 3 mmol kg"1 at 133 mmol TDN kg"1 and
1 mmol kg"1 at 9 mmol DON kg"1 based on the internal riverine
water material described above. Dissolved organic carbon
(DOC) was determined by high temperature combustion on an
MQ 1001 TOC analyzer calibrated daily with potassium
phthalate (Qian and Mopper, 1996). The performance of the
analyzer was tested daily on the certified reference material of
deep (700 m) Florida Strait seawater (Hansell Laboratory,
University of Miami, RSMAS; batch 5 FS: 47e48 mmol l"1), the
method yielded 46 (!3) mmol l"1 (n 74). Deep oceanic water,
collected on 19/11/04 at 1000 m depth in the Southern Ocean
and used throughout as an internal standard, yielded a similar
precision at a low DOC concentration (59, !3 mmol l"1, n 35).
2.6. Plant tissue analysis
Plant nitrogen content was determined using a CARLO ERBA
NA 1500 Elemental CHN Analyzer with finely ground plant
material following weighing into pre-combusted (500 & C, 3 h)
silver boats, acidification with 10% (w/v) HCl, and oven-drying
at 40 & C. Plant phosphorus was determined on ground plant
subsamples the methods of Solorzano and Sharp (1980) as
modified by Fourqueran and Zieman (1992) for particulate P
determination.
2.7. Statistical analysis
Analysis of water samples produced seven data sets of NH 4
NO" "
2 , NO3 , DIP, DOC, DON, and DOC:DON for both influent and
effluent, as well as for ambient and high nutrient loading.
Each nutrient data set was analyzed independently, to deter-
mine significant differences between influent and effluent
during the periods of ambient and high nutrient loading, and
between replicate filter beds. All statistical analyses were
conducted using MINITAB. The DON and DOC:DON data taken
during ambient nutrient loading were normally distributed
with homogeneity of variance, but the NH 4 , DIP and DOC data
taken at this time required log transformation to meet the
assumptions for normality and homogeneity of variance. All
then underwent ANOVA to compare the replicates and paired
t-test to compare influent and effluent means. Log trans-
formation had little effect on the normality and homogeneity
of variance of the pre-enrichment NO" "
2 and NO3 and DOC
data, and the influent and effluent medians were compared
using the Wilcoxon signed rank test, and differences between
the replicates were tested using the KruskaleWallis test. The
"
post-enrichment NH 4 , NO3 , DON and DOC data were nor-
mally distributed with homogeneity of variance and a one-
way ANOVA was used to compare replicates and paired
t-tests to compare influent and effluent means. The post-
enrichment DIP and NO" 2 data were log transformed, to ach-
ieve normality and homogeneity of variance, prior to ANOVA
to compare the replicates and paired t-tests to compare
influent and effluent means. Log transformation had little
effect on the normality and homogeneity of variance of the
post-enrichment DOC:DON data and the influent and effluent
medians were compared using the Wilcoxon signed rank test,
and differences between the replicates were tested using the
KruskaleWallis test.
The effluent and nutrient concentrations and DOC:DON
were (with the exception of DON during the period of high
nutrient loading) in all cases significantly lower at p % 0.001
than the measured influent levels, and in all cases there was
no significant difference at p % 0.05 between measurements
taken across the replicate filter beds. Therefore all data,
unless otherwise stated, will be quoted as the mean
concentration ! standard deviation (mmol l"1).
3. Results
Air temperature and humidity over the duration of pilot
operation are shown in Fig. 3, and influent and effluent water
temperature, salinity and pH are shown in Fig. 4. Air
temperatures within the poly-tunnel ranged from 8.2 to
55.0 & C (mean ! 1s: 23.1, !10.8 & C). Humidity ranged from 19 to
100% RH (mean ! 1s: 78 ! 25%). In all cases there was no
significant difference ( p % 0.05) between water quality
measurements taken across the replicate filter beds. Influent
water temperature across the three replicate filter beds ranged
from 16.5 to 26.1 & C (mean 20.2 ! 1.7 & C), and effluent water
temperature ranged from 17.4 to 25.7 & C (mean ! 1s:
21.5 ! 1.8). Influent salinity ranged from 10 to 29 psu
(mean ! 1s: 20 ! 5 psu), while effluent salinity ranged from 14
 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 1 0 2 e5 1 1 4 5107

Fig. 3 e The mean daily air temperature (a) and humidity

(b) with associated minima and maxima measured in the
experimental polytunnel (days 179-245).

to 31 psu (mean ! 1s: 22 ! 4 psu) before adjustment for

evaporation. Influent pH across the three replicate filter beds
ranged from 6.8 to 8.8 while effluent pH ranged from 6.4 to 7.4.
The mean pH of influent and effluent was 7.7 ! 0.4 and
6.9 ! 0.2, respectively.

3.1. Inorganic nutrients

3.1.1. Ammonium (NH
4)
Influent NH 4 concentrations at ambient nutrient loading over
the first 58 days of operation varied between 85.2 ! 15.9 and
344.4 ! 4.8 mmol l"1 (overall mean: 191.2 ! 81.9 mmol l"1). The
filter beds removed between 91 ! 12% and w100% of influent
NH 4 , reducing its concentration in the effluent to between
0.7 ! 0.3 and 16.5 ! 20.9 mmol l"1 (Fig. 5a).
As intended, the addition of the fertilizer increased the
NH 4 concentration in the influent to a maximum of
4349 ! 553 mmol l"1. The NH 4 concentration then declined
steadily to 1514 ! 164 mmol l"1 over the following 7 days. The
filter beds removed from 23 ! 12% to 50 ! 6% of influent NH 4,
resulting in effluent NH 4 concentration between 812 ! 121
and 3464 ! 79 mmol l"1 (Fig. 6a).
Fig. 4 e The mean salinity (a), pH (b) and temperature (c) of
waste-water entering (influent) and exiting (effluent)
triplicate pilot filter beds.
3.1.2. Nitrite (NO"
2)
The influent NO" 2 concentrations at ambient nutrient loading
over the first 58 days of operation fluctuated between 4.2 ! 0.4
and 35.0 ! 1.5 mmol l"1 (overall mean: 17.2 ! 9.1 mmol l"1). The
filter beds removed 90 ! 9% to w100% (overall mean: 97 ! 3%)
of influent NO" 2 , reducing its concentration in the effluent to
between 0.04 and 1.7 ! 2.0 mmol l"1 (overall mean:
0.4 ! 0.5 mmol l"1) (Fig. 5b).
5108 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 1 0 2 e5 1 1 4

During the period of high nutrient loading, influent NO" 2

a Ambient nutrient loading concentrations increased steadily from 27.9 ! 30.8 mmol l"1
400
on day 1 to a maximum of 220.2 ! 24.5 mmol l"1 on day 7
350 (overall mean: 135.3 ! 80.8 mmol l"1). The filter beds removed
300 between 66 ! 34% and 85 ! 4% (overall mean: 75 ! 8%) of
[NH ] (mol l )

250 influent NO" 2 , resulting in effluent concentrations ranging

200 from 3.5 ! 0.7 to 59.8 ! 8.9 mmol l"1 (overall mean:
150 29.1 ! 23.3 mmol l"1) (Fig. 6b).
100
50
3.1.3. Nitrate (NO"
3)
0
The influent NO" 3 concentrations at ambient nutrient loading
-50 157 161 166 171 174 179 185 189 194 200 203 211
Day of the Year
over the first 58 days of operation fluctuated between 4.0 ! 0.2
b and 114 ! 5 mmol l"1 (overall mean: 38.1 ! 30.8 mmol l"1). The
40 filter beds removed 91 ! 4e100% (overall mean: 98 ! 3%) of
35 influent NO" 3 , reducing its concentration in the effluent from
30 the filter beds to between 0.0 and 2.9 ! 2.8 mmol l"1 (overall
[NO ] (mol l )

25
mean: 0.4 ! 0.7 mmol l"1) (Fig. 5c).
20
During the 7 day period of high nutrient loading, the
15
concentrations of NO" 3 in the influent reached a maximum of
10
3808 ! 111 mmol l"1 on day 218 and then declined to
5
0
723 ! 46 mmol l"1 (overall mean: 2023 ! 1370 mmol l"1) over the
-5 157 161 166 171 174 179 185 189 194 200 203 211 subsequent 7 days before returning to ambient levels shortly
Day of the Year afterwards. The filter beds removed from 33 ! 7e83 ! 2%
c (overall mean: 54 ! 21%) of influent NO" 3 , resulting in effluent
140 NO" concentrations between 135.3 ! 23.0 and
3
120 1824 ! 274 mmol l"1 (overall mean: 1032 ! 718 mmol l"1) (Fig. 6c).
100
[NO ] (mol l )

80 3.1.4. Total dissolved inorganic nitrogen (TDIN)

60 The TDIN concentration in the influent, at ambient nutrient
40 loading during the first 58 days of operation, fluctuated between
20 108.9 ! 19.5 and 383.0 ! 6.9 mmol l"1 (overall mean:
0 246.5 ! 68.3 mmol l"1). The filter beds removed 91 ! 9% to
157 161 166 171 174 179 185 189 194 200 203 211
-20 w100 ! 0.1% (overall mean: 98 ! 2%) of influent TDIN, thus
Day of the Year reducing the TDIN concentration in the effluent to between
d 0.9 ! 0.2 and 7.7 ! 5.1 mmol l"1 (overall mean: 2.8 ! 1.9 mmol l"1)
450 (Fig. 5d).
400
During the 7 day period of high nutrient loading, the TDIN
350
concentration in the influent reached a maximum of
[TDIN] (mol l )

300
250 8185 ! 590 mmol l"1 on day 218 and then steadily decreased to
200 2391 ! 248 mmol l"1 (overall mean: 4914 ! 2557 mmol l"1) over
150 the subsequent 7 days before returning to ambient levels
100 shortly afterwards (Fig. 6d). The filter beds removed
50
30 ! 34e58 ! 4% (overall mean: 40 ! 12%) of influent TDIN,
0
-50 157 161 166 171 174 179 185 189 194 200 203 211
resulting in TDIN concentrations in the effluent fluctuating
Day of the Year between 1007 ! 144 and 5291 ! 208 mmol l"1 (overall mean:
3081 ! 1812 mmol l"1). Although the percentage of TDIN
e100 removed is greatly reduced in these experimental conditions
90 when compared to ambient nutrient loading, the removal was
80
still high when considering the reduction in the TDIN
[DIP] (mol l )

70
60 concentration by 2894 ! 408 mmol l"1 in the first 24 h
50
40 period, followed by reduction between 1037 ! 30 and
30 2017 ! 36 mmol l"1 over each of the following periods of six day
20
10 each. Overall, the relationship between daily loading and
0
removal of nitrogen was linear over the range tested, and
157 161 166 171 174 179 185 189 194 200 203 211
higher rates of N removal may be achievable (Fig. 9).
Day of the Year
Influent Effluent

Fig. 5 e The mean concentration (mmol lL1) (SD) of

dissolved inorganic ammonium (a), nitrate (b), nitrite (c), exiting (effluent) triplicate Salicornia planted filter beds
TDIN (d) and DIP (e) in waste-water entering (influent) and during a period of ambient nutrient loading.
 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 1 0 2 e5 1 1 4 5109

High nutrient loading 3.1.5. Dissolved inorganic phosphate (DIP)

a 6000 Influent DIP concentrations at ambient nutrient loading over
5000 the first 58 days of operation fluctuated between 34.1 ! 0.9 and
4000 89.9 ! 0.8 mmol l"1 (overall mean: 59.1 ! 18.3 mmol l"1). The filter
[NH ] (mol l )

3000
beds removed 36 ! 6e89 ! 2% (overall mean: 67 ! 14%) of
influent DIP, resulting in an effluent DIP concentration range of
2000
10.5 ! 2.1e44.4 ! 1.3 mmol l"1 (overall mean: 19.7 ! 10.0 mmol l"1)
1000
(Fig. 5e). During the period of high nutrient loading, influent DIP
0 concentrations reached a maximum of 354.1 ! 21.7 mmol l"1 on
218 222 223 225 229 231 232 237 239 244 245
-1000 day 218. The DIP concentrations declined steadily to
Day of the Year 215.6 ! 13.2 mmol l"1 over the following 7 days (overall mean:
282.7 ! 57.1 mmol l"1) (Fig. 6e). The filter beds removed
b 300 19 ! 3e40 ! 27% (overall mean: 31 ! 9%) of influent DIP,
250 resulting in effluent DIP concentrations between 187.3 ! 6.4 and
200 227.2 ! 10.4 mmol l"1 (overall mean: 201.3 ! 2.5 mmol l"1).
[NO ] (mol l )

150

100 3.1.6. Nutrient ratios

50
The TDIN in the settled fish farm wastewater used to irrigate the
"
pilot filter beds was proportionately higher in NH 4 than NO3 or
0
"
222 223 225 229 231 232 237 239 244 245 NO2 , with mean molar ratio of 10:2:1, and had a relatively
-50
constant molar TDIN:DIP of 4:1. Over the first 58 days of oper-
Day of the Year
ation, influent TDIN and DIP concentrations fluctuated as
c 4500 a result of varying fish stock, feed loading and intervals between
4000
biofilter flushing (Fig. 5d and e). When operating under ambient
3500
3000
nutrient loading, the relative composition of TDIN in the
[NO ] (mol l )

2500 effluent from the filter beds was similar to that of the influent
" "
2000 wastewater, with a mean NH 4 :NO3 :NO2 14:2:1. However, the
1500
mean TDIN:P in the effluent declined to approximately 1:5,
1000
500
reflecting the more efficient removal of TDIN in the filters beds.
0 For the first seven days after the addition of ammonium
-500 218 222 223 225 229 231 232 237 239 244 245 nitrate fertiliser, the TDIN composition of the irrigation water
" "
Day of the Year (influent) changed, with a mean NH 4 :NO3 :NO2 24:17:1 and

d 10000 a mean TDIN:DIP 17:1.

9000
8000
7000 3.2. Organic nutrients
[TDIN] (mol l )

6000
5000
4000 3.2.1. Dissolved organic carbon (DOC)
3000 Influent DOC levels, at ambient nutrient loading, over the first
2000
58 days of operation (Fig. 7a) fluctuated between 335.7 ! 96.8 and
1000
0 1360 ! 11.3 mmol l"1. The filter beds removed 3 ! 65 to 59 ! 7% of
-1000 218 222 223 225 229 231 232 237 239 244 245 inlet DOC, significantly reducing the concentration in effluent
Day of the Year to between 295.7 ! 112.2 and 823.7 ! 66.5 mmol l"1.
e Following the addition of fertilizer, influent DOC levels
400 increased (Fig. 8a) ranging between 1382 ! 215 and
350
1693 ! 61 mmol l"1 almost double that observed at ambient
300
[DIP] (mol l )

levels. During the seven days of high nutrient loading, the

250
200
filter beds ability to remove DOC deteriorated slightly, with
150 2 ! 13 to 37 ! 7% of influent DOC being removed, significantly
100 reducing effluent concentrations to between 807 ! 61 and
50 1442 ! 1467 mmol l"1.
0
218 222 223 225 229 231 232 237 239 244 245
3.2.2. Dissolved organic nitrogen (DON)
Day of the Year
Influent DON levels, at ambient nutrient loading, over the first
Influent Effluent
58 days of operation fluctuated between 45.3 ! 21.8 and
Fig. 6 e The mean concentration (SD) of dissolved 144.1 ! 6.2 mmol l"1, with concentrations in effluents fluctu-
inorganic ammonium (a), nitrate (b), nitrite (c), TDIN (d) and ating between 28.9 ! 0.6 and 66.4 ! 16.8 mmol l"1 (Fig. 7b).
DIP (e) in waste-water entering (influent) and exiting During the first 7 days of operation (days 157e164), DON
(effluent) triplicate Salicornia planted filter beds during concentrations in filter effluents were not significantly
a period of high nutrient loading. reduced compared to influent values ( p > 0.05). For the
5110 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 1 0 2 e5 1 1 4

a Ambient nutrient loading High nutrient loading

a
1600 2000
1400 1800
1600
1200
[DOC] (mol l -1)

[DOC] (mol l-1)

1400
1000
1200
800 1000
600 800
600
400
400
200
200
0 0
157 161 166 171 174 179 185 189 194 200 203 211 218 222 223 225 229 231 232 236 237 239 244
Day of the Year
b Day of the Year
b
180 900
160 800
700
140
600

[DON] (mol l -1)

[DON] (mol l-1)

120 500
100 400
80 300
200
60
100
40 0
20 218 222 223 225 229 231 232 237 239 244
0 Day of the Year
157 164 168 174 179 185 189 194 200 203 211
c
Day of the Year

Fig. 8 e The mean concentration (SD) of dissolved organic

Fig. 7 e The mean concentration (SD) of dissolved organic
carbon, DOC (a) and nitrogen, DON (b) and the DOC:DON
ratio (c) in waste-water entering (influent) and exiting
(effluent) triplicate Salicornia planted filter beds during
a period of ambient nutrient loading.
carbon, DOC (a) and nitrogen, DON (b) and the DOC:DON
ratio (c) in waste-water entering (influent) and exiting
(effluent) triplicate Salicornia planted filter beds during
a period of high nutrient loading.

influent values. With the return to ambient loading (days

remainder of the ambient loading period (days 166e215)
effluent values were significantly lower than those observed
in the influent with filter beds removing 12 ! 44 to 68 ! 3% of
influent DON with uptake gradually increasing with time.
In the seven days following addition of fertiliser, influent
DON concentrations (Fig. 8b) were approximately six times
higher than that seen during ambient nutrient loading
ranging between 444.7 ! 133.3 and 542.8 ! 117.8 mmol l"1, with
concentrations in effluents fluctuating between 438.1 ! 107.4
and 667.0 ! 161.9 mmol l"1 (days 218e229) before returning to
ambient levels shortly afterwards (days 231e245). During the 7
days of high loading (days 218e225), DON concentrations in
filter effluents were not significantly reduced compared to
229e244) effluent values were significantly lower than those
observed in the influent with filter beds removing 23 ! 6 to
63 ! 9% of influent DON.
3.2.3. DOC:DON
The observed increases and decreases in DOC and DON are
reflected in the organic DOC:DON ratios, with a changing
pattern during the course of the experiment. Initially (days
157e166), influent DOC:DON ratio values were significantly
higher than those in the effluent, indicative of the apparent
greater removal of DON by the CW (Fig. 7c). By day 168 there
was no significant difference ( p > 0.05) between the influent
and effluent DOC:DON ratios, and between day 174e215,
DOC:DON ratios were consistently higher in the effluent than
 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 1 0 2 e5 1 1 4
Fig. 9 e Relationship between total inorganic nitrogen
(TDIN) daily load and TDIN removal from triplicate
Salicornia planted filter beds during the period of high
nutrient loading. Units are mmol N mL2 dL1, equation of
fitted curve is y [ 0.34x D 19.68, R2 [ 0.969. Mean
Salicornia europaea plant biomass during this period was
8.85 2.32 g mL2.
the influent water reflecting the significant removal ( p < 0.05)
of DON from the influent. In the 7 days following the addition
of ammonium nitrate fertiliser, there was no significant
difference ( p > 0.05) between the influent and effluent DOC:-
DON ratios (Fig. 8c), but on the return to ambient influent
levels by day 229, the previous pattern returns with DOC:DON
in the effluent significantly higher than the influent.
3.3. Nutrient removal efficiency and retention in plants
During the first 58 days of the sampling period at ambient
nutrient loading, the filter beds removed 97e100% of the
influent TDIN. During the 7-day period of nutrient enrich-
ment, the filter beds exhibited lower TDIN removal efficiency
1.2
1 0.10
0.8
[N] (mol m )
0.6
0.4
0.2
0 0.00
166 185
Day of the Year
nitrogen
Fig. 10 e The mean cumulative stored plant N and P calculated over five successive harvests from triplicate Salicornia
planted filter beds.
5111
(40e42%). It is possible to calculate nitrogen removal efficiency
with increasing N loading during the period before TDIN
was depleted in the effluent water. Fig. 9 shows the
relationship between the daily load and removal of TDIN
from the pilot CW during the period of high nutrient loading.
This shows that daily nitrogen removal increased linearly
with increased loading, yielding a TDIN removal rate of
21e239 mmol N m"2 d"1. However, the efficiency in terms of
residual TDIN concentration in the effluent declined along the
same nutrient loading gradient. Overall the time-integrated
(cumulative) apparent nitrogen removal by the filters beds
was 1.28 ! 0.05 mol N m"2. Over the same period, total N
accumulation in plant tissue (estimated from N content and
growth of plant root and shoot tissues and harvested biomass)
was 1.09 mol N m"2, indicating that up to 85% of nitrogen
removed from waste water was retained in plant tissue
(Fig. 10). During the high nutrient loading period, S. europaea
plant biomass in the filter beds did not vary significantly, with
a mean of 8.85 ! 2.32 g m"2. This suggests that under high
nutrient loading conditions, the plants were able to take up
2.4e27.0 mmol N d"1 g"1 dry weight.
At ambient nutrient loading, the DIP removal efficiency
averaged 67 ! 14%, dropping to 31 ! 9% during the period of
high nutrient loading. However, the mean daily DIP removal at
this time increased from 2.81 ! 1.24 mmol m"2 d"1 to
6.60 ! 3.05 mmol m"2 d"1. Overall, the apparent cumulative
DIP removal by the filters beds was 0.11 ! 0.01 mol m"2. Over
the same period, total P accumulation in plant tissue, esti-
mated as outlined for nitrogen above, was 0.08 mol m"2,
indicating that up to 73% of DIP removed from waste water
was retained in plant tissue (Fig. 10).
4. Discussion
The concentrations of TDIN (93e439 mmol l"1) and DIP
(34e90 mmol l"1) observed in fish farm wastewater in the
present study were within the range, or even exceeded, those
0.12
0.08
[P] mol m
0.06
0.04
0.02
208 230 246
phosphorus
5112 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 1 0 2 e5 1 1 4
measured in the wastewater from other aquaculture RAS
(Cripps and Bergheim, 2000; Porello et al., 2003; Lin et al., 2005).
Overall, under the hydraulic regime applied during the 58 day
period of ambient loading, the pilot CW acted as a highly
effective biofilter for the aquaculture effluent, removing
91e99% of influent TDIN and 41e88% of influent DIP,
comparing favourably with those reported in small-scale
experimental studies. Brown et al. (1999) achieved 98%
and 94% removal total and inorganic nitrogen in the
irrigation water and in the Lin et al. (2002a,b) experiment,
the SSF section of their CW removed 95e98% of the
0.15e21.5 mg TDIN l"1 (11e1537 mmol l"1) observed in the
influent. These levels of nutrient removal efficiency are
considerably greater than those reported by Lymbery et al.
(2006), where TDIN nutrient removal from RAS effluent
using CWs was 69%.
The addition of ammonium nitrate fertiliser to the fish
farm effluent increased TDIN loading by an order of magni-
tude above ambient levels. However, TDIN and DIP concen-
trations remained high in the effluent, comparable to those
reported in effluent from other aquaculture RAS systems
(Pagand et al., 2000; Adler et al., 2000; Lin et al., 2005). Under
the hydraulic regime applied at high nutrient loading, the CW
operated at lower removal efficiency for TDIN (30e58%) and
DIP (19e40%). However, overall assimilation of nitrogen by the
bed was significantly higher under high loading conditions
and, as indicated by the linear relationship between loading
and removal (Fig. 9), the limit of TDIN removal capacity was
not reached over the range tested in the present study. The
highest removal rate observed (263 mmol m"2 d"1) greatly
exceeds those reported in previous studies in CWs. For
example, Lin et al. (2005) observed removal rates of
17 mmol N m"2 d"1 and Konnerup et al. (2011) reported
a removal rate of 27 mmol m"2 d"1. Several studies have
investigated the efficiency of CWs under different HRT and
HLR, with a general trend of increasing efficiency with
increasing HRT (eg Lin et al., 2005; Schulz et al., 2003) sug-
gesting that during periods of high nitrogen loading, a HRT
greater than 24 h may increase N removal efficiency. Under
the hydraulic regime operated in the present study, if daily
loading remained below 50 mmol m"2 d"1, complete removal
of TDIN in waste-water was achievable in 24 h. For higher
loading levels, the effectiveness of longer retention times may
usefully be investigated.
The majority of nitrogen removal in wetlands is thought to
result from the microbial processes of nitrification and deni-
trification and to a lesser degree, sedimentation, filtration,
precipitation and volatization. Uptake by the plant is thought
to account for a minor fraction of the total reduction of
nutrients (Kadlec and Knight, 1996; Koottatep and Polprasert,
1997; Lin et al., 2002b; Haddad et al., 2006). However there is
limited research available on the uptake capacity of Salicornia
spp. and the highly effective removal of nitrogen by S. europaea
in the present study may reflect an adaptation in plants that
have evolved to compete for and sequester nitrogen in what
are periodically nitrogen-limited environments. Nitrogen
availability in salt marshes not only varies throughout the
year but also spatially and peaks in available N tend to
immediately precede a phase of vegetative growth (Jefferies,
1977). In situ additions of inorganic nitrogen to stands of
Salicornia virginica in salt marshes not only cause a shift in
species composition in favour of S. virginica, reflecting an
ability to outcompete other salt marsh plants for available N,
but also produced significant increases in biomass (Covin and
Zedler, 1988; Boyer and Zedler, 1999).
Macrophytes play a key role in the performance of fresh-
water CW, facilitating the nitrification/denitrification process,
maintaining the hydraulic conductivity of the substrate,
increasing microbial assemblages in the root zone and
participating in nutrient uptake (Brix, 1994; Haberl et al., 1995).
Research into the role of plants in CWs tends to have focussed
on freshwater systems with wetland plants such as Phragmites
and Typha as essential components of the CW ecosystem,
releasing oxygen from their root system into the rhizosphere
and hence facilitating the establishment of the microbial
communities responsible for nitrogen removal (Faulwetter
et al., 2009). Like the rhizomes of macrophytes, the roots of
salt marsh Salicornia sp., possess aerenchyma which allow
exchange of gases between the shoot and the root and indi-
rectly aerate the surrounding soil zone potentially resulting in
increased nitrification/denitrification efficiency (Brix, 1994;
Haberl et al., 1995; Faulwetter et al., 2009). The present CW was
carefully designed to optimise plant and aerobic microbial
nutrient removal whilst promoting plant yield. The daily flood
and drain procedure, which simulated the tidal immersion
cycles experienced by saltmarsh plants, also allowed aeration
at the root zone for several hours before re-submergence, thus
reducing the build-up of anaerobic conditions usually
accountable for low nitrification (Vymazal, 2005). In addition,
large pore spaces within the stone layer reduced clogging,
whilst continuous subsurface circulation of water within each
bed improved nutrient flow across the root zone whilst
reducing the build up of anoxic zones within the CW
substrate.
In the present study, apparent DON removal depended on
the total inorganic nitrogen loading to the filter beds. Under
ambient loading, TDIN was being almost completely removed
and DON was significantly reduced and there was a slightly
higher DOC:DON ratio in effluent than influent water. By
comparison, under experimentally-elevated TDIN loadings,
when there was excess TDIN in effluent water, there was no
apparent DON removal and the DOC:DON ratio increased as
water passed through the filter beds. These results indicate that
when TDIN was available in excess, DON removal was negli-
gible suggesting that uptake of DON was by primary producers
(microbial community or plants). Bacteria can uptake both
organic and inorganic N, and the DON may then be mineralized
and made available to higher plants. Konnerup et al. (2011)
observed apparent DON removal in a CW planted with Canna
x generalis, where 90% of TDN was present as DON and TDIN
concentrations were low (30e48 mmol N l"1). The ability of
plants to uptake DON directly as free amino acids has been
widely demonstrated (see Nasholm et al. (2009) for a review),
and recently direct uptake of peptides has been measured
(Hill et al., 2011a, 2011b). Halophytic plants have been shown to
take up DON in quantities up to 20% of the DIN uptake, and DON
uptake has been shown to be independent of DIN concentra-
tion (Mozdzer et al., 2010). Direct uptake of amino acids and
peptides by Salicornia plants has also been observed (R Quinta,
unpublished data), which would suggest that, as well as
 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 1 0 2 e5 1 1 4
bacteria, the plants themselves may have directly contributed
to DOC and DON uptake within the filter beds.
Nonetheless, DOC and DON were selectively removed in
the CW as indicated by the differences in DOC:DON in the
influent and effluent. This selectivity was not uniform over
the period studied, as the difference between influent and
effluent DOC:DON shifted with time. These shifts may be
a result of alterations in the nutrient supply, as bioavailability
of DON depends on its composition (Stepanauskas et al., 2000),
together with the progressive establishment of a microbial
community and consequent population shifts: e.g. it may take
from a month to a year for the microbial community to
establish in a new CW (Lin et al., 2002a).
The CW acted as an effective biofilter for DIP in aquaculture
wastewater, removing 41e88% from influent water during the
58 day period of ambient loading, which is similar to values
reported in other aquaculture wastewater treatment systems
(Lin et al., 2002a). Adsorption to sediment is considered to be
key mechanism for P retention in CW (Kadlec and Knight,
1996). However, in the present study, over 70% of the DIP
removed from influent water was retained in plant tissue,
indicating that sediments play a less important role in this
case. The use of coarse sand with a low adsorption capacity
might suit plant growth and the associated N uptake but this
appears to have reduced the effectiveness of the substrate in P
uptake. However, phosphate uptake by S. europaea is likely to
be dependent on N supply. Webb (2005) found that in Salicornia
dolichostachya, a growth response to increasing P was only
observed with excess N supply, suggesting that it is not
limiting to growth. Over the 7 day period of nitrogen enrich-
ment, although the mean DIP removal efficiency from the
influent water dropped to 27.4%, uptake rose from
39.5 mmol l"1 to 81.5 mmol l"1, suggesting that plant P
requirements increased with increasing N supply.
Overall the pilot CW tested here performed well compared
to established alternative treatment systems. In a compara-
tive study of various hydroponic systems with comparable
influent N and P, Lennard and Leonard (2006) found gravel
beds growing Lactuca sativa removed 91% of N and 53% of P,
while Porello et al. (2003) investigated the effectiveness of
a lagoon system for phytotreatment of RAS waste and
observed only a 15% reduction in DIN from influent concen-
trations of 67e69 mmol l"1. Using a system of high rate algal
ponds for RAS effluent treatment and with DIN concentra-
tions in pond influent at 700 mmol N l"1, Pagand et al. (2000)
observed 59% removal efficiency over a 14 month period.
Present results also compare favourably with traditional bio-
filtration and mechanical filtration, which can achieve 74%
"
and 44% reduction in NH 4 and NO3 , respectively, in waste-
water from temperate marine fish farms (Borges et al., 2003).
The high removal efficiency of DIN under ambient nitrogen
loading and the significant removal of DIP indicate that con-
structed wetlands planted with Salicornia are well-suited to
the treatment of batches of highly concentrated wastewater
released during routine filter backwashing in marine RAS. The
high nutrient concentrations support high plant growth rates
and hence efficient N uptake rates. Under normal RAS fish
farm nutrient loading, the present study demonstrates that
a CW planted with the halophyte Salicornia is a feasible and
effective biological treatment alternative for marine RAS, as
5113
well as providing a high value plant crop and biomass for
composting. However, S. europaea is an annual species and
seasonal productivity and cost-effectiveness will depend on
latitude and cost of artificial lighting. Ongoing work is inves-
tigating the extension of the production season through use of
mixed communities of saltmarsh plants, including additional
biennial and perennial species that also have value as crops,
such as such Sarcocornia perennis and Aster tripolium.
5. Conclusions
( Nitrogen in aquaculture wastewater was primarily in the
form of dissolved inorganic nitrogen (TDIN) and was
removed by 98.2 ! 2.2% under ambient loadings of
109e383 mmol l"1.
( There was a linear relationship between TDIN uptake and
loading over the range of inputs tested. At peak loadings of
up to 8185 ! 590 mmol l"1 (equivalent to mmoles 600
mmol N m"2 d"1), the filter beds removed between 30 and
58% (250 mmol N m"2 d"1) of influent TDIN.
( Influent dissolved inorganic phosphate levels ranged from
34 to 90 mmol l"1, with 36e89% reduction under routine
operations.
( Dissolved organic nitrogen (DON) loadings were lower
(11e144 mmol l"1), and between 23 and 69% of influent DON
was removed during routine operation, with no significant
removal of DON under high TDIN loading.
( Over the 88-day study, cumulative nitrogen removal was
1.28 mol m"2, of which 1.09 mol m"2 was retained in plant
tissue, with plant uptake ranging from 2.4 to
27.0 mmol N g"1 dry weight d"1.
( The results demonstrate the effectiveness of treatment of
wastewater from land-based intensive marine aquaculture
farms by constructed wetlands planted with S. europaea.
Acknowledgements
This study was supported by a European Union FP6 CRAFT
project, Envirophyte (COOP-CT-2006-032167). RQ was sup-
ported by Fundacao para a Ciencia e a Tecnologia, Portugal,
(SFRH/BD/43234/2008)
references
Adler, P.R., Harper, J.K., Takeda, F., Wade, E.M., Summerfelt, S.T.,
2000. Economic evaluation of hydroponics and other
treatment options for phosphorus removal in aquaculture
effluent. Horticultural Science 35 (6), 993e999.
Borges, M.-T., Morais, A., Castro, P.M.L., 2003. Performance of
outdoor seawater treatment systems for recirculation in an
intensive turbot (Scophthalmus maximus) farm. Aquaculture
International 11 (6), 557e570.
Boyd, C.E., 2003. Guidelines for aquaculture effluent management
at the farm-level. Aquaculture 226 (1e4), 101e112.
Boyer, K.E., Zedler, J.B., 1999. Nitrogen addition could shift plant
community composition in a restored California salt marsh.
Restoration Ecology 7 (3), 282e287.
5114 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 1 0 2 e5 1 1 4
Brix, H., 1994. Functions of macrophytes in constructed wetlands.
Water Science and Technology 29 (4), 71e78.
Brown, J.J., Glenn, E.P., Fitzsimmons, K.M., Smith, S.E., 1999.
Halophytes for the treatment of saline aquaculture effluent.
Aquaculture 175 (3e4), 255e268.
Calheiros, C.S.C., Quiterio, P.V.B., Silva, G., Crispim, L.F.C.,
Brix, H., Moura, S.C., Castro, P.M.L., 2012. Use of constructed
wetland systems Arundo and Sarcocornia for polishing high
salinity tannery wastewater. Journal of Environmental
Management 95 (1), 66e71.
Cripps, S.J., Bergheim, A., 2000. Solids management and removal
for intensive land-based aquaculture production systems.
Aquacultural Engineering 22 (1e2), 33e56.
Covin, J.D., Zedler, J.B., 1988. Nitrogen effects on Spartina foliosa
and Salicornia virginca in the salt marsh at Tijuana Estuary,
California. Wetlands 8 (1), 51e65.
Davy, A.J., Bishop, G.F., Costa, C.S.B., 2001. Salicornia L. (Salicornia
pusilla J. Woods, S. ramosissima J. Woods, S. europaea L., S.
obscura P.W. Ball & Tutin, S. nitens P.W. Ball & Tutin, S. fragilis
P.W. Ball & Tutin and S. dolichostachya Moss). Journal of
Ecology 89 (4), 681e707.
FAO, 2009. The State of World Fisheries and Aquaculture 2008.
Food and Agriculture Organization of the United Nations,
Rome.
Faulwetter, J.L., Gagnon, V., Sundberg, C., Chazarenc, F.,
Burr, M.D., Brisson, J., Camper, A.K., Stein, O.R., 2009. Microbial
processes influencing performance of treatment wetlands:
a review. Ecological Engineering 35 (6), 987e1004.
Fourqueran, J.W., Zieman, J.C., 1992. Phosphorus limitation of
primary production in Florida Bay: evidence from C: N:P ratios
of the dominant seagrass Thalassia testudinum. Limnology and
Oceanography 37 (1), 162e171.
Glenn, E.P., OLeary, J.W., Watson, M.C., Thompson, L.T.,
Kuehl, R.O., 1991. Salicornia bigelovii Torr.: an Oilseed
halophyte for seawater irrigation. Science 251 (4997),
1065e1067.
Grasshoff, K.M., Ehrhardt, M., Kremling, K., 1983. Methods of
Seawater Analysis. Verlag Chemie.
Haberl, R., Perfler, R., Mayer, H., 1995. Constructed wetlands in
Europe. Water Science and Technology 32 (3), 305e315.
Haddad, H.R., Maine, M.A., Bonetto, C.A., 2006. Macrophyte
growth in a pilot-scale constructed wetland for industrial
wastewater treatment. Chemosphere 63 (10), 1744e1753.
Hales, B., van Greer, A., Takahashi, T., 2004. High-frequency
measurements of seawater chemistry: flow-injection analysis
of macronutrients. Limnology and Oceanography: Methods 2,
91e101.
Hill, P.W., Farrar, J., Roberts, P., Farrell, M., Grant, H.,
Newsham, K.K., Hopkins, D.W., Bardgett, R.D., Jones, D.L.,
2011a. Vascular plant success in a warming Antarctic may be
due to efficient nitrogen acquisition. Nature Climate Change 1
(1), 50e53.
Hill, P.W., Quilliam, R.S., DeLuca, T.H., Farrar, J., Farrell, M.,
Roberts, P., Newsham, K.K., Hopkins, D.W., Bardgett, R.D.,
Jones, D.L., 2011b. Acquisition and assimilation of nitrogen as
peptide-bound and D-enantiomers of amino acids by wheat.
PLoS ONE 6 (4), 1e4.
Holmes, R.M., Aminot, A., Kerouel, R., Hocker, B.A., Peterson, B.J.,
1999. A simple and precise method for measuring ammonium
in marine and fresh water. Canadian Journal of Fisheries and
Aquatic Sciences 56 (10), 1801e1808.
Jefferies, R.L., 1977. Growth responses of coastal halophytes to
inorganic nitrogen. Journal of Ecology 65 (3), 847e865.
Kadlec, R.H., Knight, R.L., 1996. Treatment Wetlands. Lewis
Publishers, Boca Raton.
Konnerup, D., Trang, N.T.D., Brix, H., 2011. Treatment of fishpond
water by recirculating horizontal and vertical flow constructed
wetlands in the tropics. Aquaculture 313 (1e4), 57e64.
Koottatep, T., Polprasert, C., 1997. Role of plant uptake on
nitrogen removal in constructed wetlands located in the
tropics. Water Science and Technology 36 (12), 1e8.
Kroon, H., 1993. Determination of nitrogen in water: comparison
of continuous flow method with on-line UV digestion with the
original Kjedahl method. Analytica Chimica Acta 276 (2),
287e293.
Lennard, W.A., Leonard, B.V., 2006. A comparison of three
different hydroponic sub-systems (gravel bed, floating and
nutrient film technique) in an aquaponic test system.
Aquaculture International 14 (6), 539e550.
Lin, Y.F., Jing, S.R., Lee, D.Y., Wang, T.W., 2002a. Nutrient removal
from aquaculture wastewater using a constructed wetlands
system. Aquaculture 209 (1e4), 169e184.
Lin, Y.F., Jing, S.R., Wang, T.W., Lee, D.Y., 2002b. Effects of
macrophytes and external carbon sources on nitrate removal
from groundwater in constructed wetlands. Environmental
Pollution 119 (3), 413e420.
Lin, Y.F., Jing, S.R., Lee, D.Y., Chang, Y.F., Chen, Y.M., Shih, K.C.,
2005. Performance of a constructed wetland treating intensive
shrimp aquaculture wastewater under high hydraulic loading
rate. Environmental Pollution 134 (3), 411e421.
Lymbery, A.J., Doupe, R.G., Bennett, T., Starcevich, M.R., 2006.
Efficacy of a subsurface-flow wetland using the estuarine
sedge Juncus kraussii to treat effluent from inland saline
aquaculture. Aquacultural Engineering 34 (1), 1e7.
Mozdzer, T.J., Zieman, J.C., McGlathery, K.J., 2010. Nitrogen
uptake by native and invasive temperate coastal
macrophytes: importance of dissolved organic nitrogen.
Estuaries and Coasts 33 (3), 784e797.
Nasholm, T., Kielland, K., Ganeteg, U., 2009. Uptake of organic
nitrogen by plants. New Phytologist 182 (1), 31e48.
Pagand, P., Blancheton, J.P., Lemoalle, J., Casellas, C., 2000. The
use of high rate algal ponds for the treatment of marine
effluent from a recirculating fish rearing system. Aquaculture
Research 31 (10), 729e736.
Piedrahita, R.H., 2003. Reducing the potential environmental
impact of tank aquaculture effluents through intensification
and recirculation. Aquaculture 226 (1e4), 35e44.
Porello, S., Lenzi, M., Persia, E., Tomassetti, P., Finoia, M.G.,
2003. Reduction of aquaculture wastewater eutrophication
by phytotreatment ponds system I. Dissolved and
particulate nitrogen and phosphorus. Aquaculture 219 (1e4),
515e529.
Qian, J., Mopper, K., 1996. High performance high temperature
combustion total organic carbon analyzer. Analytical
Chemistry 68 (18), 3090e3097.
Rhee, M.H., Park, H.-J., Cho, J.Y., 2009. Salicornia herbacea:
botanical, chemical and pharmacological review of halophyte
marsh plant. Journal of Medicinal Plants Research 3 (8),
548e555.
Schulz, C., Gelbrecht, J., Rennert, B., 2003. Treatment of rainbow
trout farm effluents in constructed wetland with emergent
plants and subsurface horizontal water flow. Aquaculture 217
(1e4), 207e221.
Solorzano, L., Sharp, J.H., 1980. Determination of total dissolved
nitrogen in natural waters. Limnology and Oceanography 25
(4), 751e754.
Stepanauskas, R., Laudon, H., Jrgensen, N.O.G., 2000. High DON
bioavailability in boreal streams during a spring flood.
Limnology and Oceanography 45 (6), 1298e1307.
Tal, Y., Schreier, H.J., Sowers, K.R., Stubblefield, J.D., Place, A.R.,
Zohar, Y., 2009. Environmentally sustainable land-based
marine aquaculture. Aquaculture 286 (1e2), 28e35.
Vymazal, J., 2005. Constructed wetlands for wastewater
treatment. Ecological Engineering 25 (5), 475e477.
Webb, J.M., 2005. Salicornia (L.): A Potential Biofilter for
Aquaculture? M.Sc. Dissertation Bangor University, Wales, UK.