Ecological Engineering 127 (2019) 468–479

Contents lists available at ScienceDirect

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

Floating treatment wetland aided nutrient removal from agricultural runoff T

using two wetland species
Jonathan T. Spanglera, David J. Sampleb, , Laurie J. Foxb, James S. Owen Jr.b, , Sarah A. Whitec
⁎ ⁎

a
Department of Biological System Engineering, Hampton Roads Agricultural Research and Extension Center, Virginia Polytechnic and State University, Virginia Beach, VA
23455, USA
b
School of Plant and Environmental Sciences, Hampton Roads Agricultural Research and Extension Center, Virginia Polytechnic and State University, Virginia Beach, VA
23455, USA
c
Department of Plant and Environmental Sciences, Clemson University, E-143 Poole Agric. Center, Clemson, SC 29634-0310, USA

ARTICLE INFO ABSTRACT

Keywords: Floating treatment wetlands (FTWs) are an emerging surface water treatment practice for reducing nutrients
Container nursery such as nitrogen (N) and phosphorus (P) in a variety of waste streams. FTWs consist of floating rafts which
Denitrification support wetland plants allowing the plant roots to be submerged below the water surface similar to hydroponic
Tailwater recovery basin (TRB) systems. N and P removal is achieved through direct plant uptake and flocculation or sedimentation facilitated
Mesocosm
by biofilms adhering to plant roots. While research has documented FTW performance for a variety of waste
Nitrogen
streams, little information is available on runoff from commercial nurseries. Nursery runoff occurs frequently,
Phosphorus
Total maximum daily load (TMDL) from storm and irrigation events, and can potentially contain elevated concentrations of total N (TN) and total P
(TP) due to crop fertilization and frequent irrigation. Nursery runoff is often captured in large ponds known as
tailwater recovery basins (TRBs) for reuse. We assessed the effectiveness of FTWs for treating nursery runoff in
TRBs over a 7-day hydraulic retention time across a 19-week growing season using two supplied nutrient
concentrations and four treatments with four replicates in a randomized complete block mesocosm experiment.
The high nutrient concentration averaged 17.1 mg·L−1 TN and 2.61 mg·L−1 TP and low nutrient concentration
averaged 5.22 mg·L−1 TN and 0.52 mg·L−1 TP. Treatments included: 1) Pontederia cordata, 2) Juncus effusus, 3) a
raft with no plants, and 4) no raft (open-water). Pontederia cordata removed 90% and 92% of TP and 84% and
89% TN, depending on initial loading; performing better than the other planted and non-planted treatments.
Juncus effusus performed better than the mat-only control treatments for TP and TN removal across nutrient
concentrations tested, but performed similarly to the open water control when nutrient load was evaluated.
Nutrient uptake by Pontederia cordata averaged 0.026 g·m−2·d−1 N and 0.0075 g·m−2·d−1 P. The study de-
monstrated that FTWs can be very effective, but nutrient removal depends greatly upon plant species selection.

1. Introduction watershed are commercial nurseries which produce a variety of plants

for resale. While commercial nurseries use several methods of produc-
Agriculture is the leading non-point source contributor to impair- tion, one method is to grow plants in pots with a media consisting of a
ments of surface waters in the USA (USEPA, 2013). Excess irrigation quick draining substrate; so frequent irrigation and application of fer-
and fertilizer use increases the transport of sediment and nutrients such tilizers are needed to maintain crop health and quality (Majsztrik et al.,
as nitrogen (N) and phosphorus (P) from farms to streams, rivers, and 2011). These container nurseries generate high volumes of runoff be-
estuaries, contributing to eutrophication (Carpenter et al., 1998; Qin, cause production areas, (“pads”), can become relatively impermeable.
2009). Eutrophication in the Chesapeake Bay estuary has led to the Majsztrik (2011) reported average commercial nursery N and P appli-
imposition of a total maximum daily load (TMDL), which requiring cation rates of approximately 680 kg·ha−1·yr−1 and 129 kg·ha−1·yr−1,
significant reduction of nutrient and sediment loads from all sources, respectively. Studies conducted in the southeastern U.S. have shown
including agriculture, within the contributing watershed (USEPA, average concentrations of total nitrogen (TN) and total phosphorus (TP)
2010). One rapidly growing agricultural practice in the Chesapeake Bay in nursery runoff to range from 8.27 to 21.7 mg·L−1 and 1.41 to

Corresponding authors.
⁎

E-mail addresses: [email protected] (J.T. Spangler), [email protected] (D.J. Sample), [email protected] (L.J. Fox), [email protected] (J.S. Owen),
[email protected] (S.A. White).

https://doi.org/10.1016/j.ecoleng.2018.12.017
Received 15 May 2018; Received in revised form 22 November 2018; Accepted 11 December 2018
0925-8574/ © 2018 Elsevier B.V. All rights reserved.
J.T. Spangler et al. Ecological Engineering 127 (2019) 468–479

Fig. 1. Floating treatment wetland cross-section diagram.

8.27 mg·L−1, respectively, during the growing season (Taylor et al., Assimilation and denitrification play a major role in N removal by
2006; White et al., 2011; White et al., 2010). To reduce runoff loads and FTWs (Borne et al., 2013; Jayaweera and Kasturiarachchi, 2004). Borne
meet TMDL reduction goals for N, P, and sediment in the Chesapeake et al., (2013) concluded that denitrification contributed more to N re-
Bay, many mid-Atlantic nurseries have begun implementing a variety of moval than plant uptake. To facilitate denitrification, a minimum of
best management practices (BMPs) (Majsztrik and Lea-Cox, 2013). 50 m2 or between 10% and 50% pond surface area coverage was re-
BMPs include efficiency improvements as well as water treatment commended to create the anoxic conditions needed (Borne et al.,
(Majsztrik et al., 2017; Majsztrik et al., 2011; Yeager et al., 2010)). One 2015b). According to Borne (2014), P removal occurs via sorption and
such BMP are tailwater recovery basins (TRBs) which are large ponds subsequent particle removal via entrapment, flocculation and sedi-
used to treat runoff through settling and to extend water supplies. TRBs mentation. Thus, sizing of FTW systems varies based on nutrient loads,
provide adequate sediment treatment, but are inefficient in removal of nutrient removal goals, pond size, and budget. Wang and Sample
dissolved pollutants (Tanner and Headley, 2011; Yeager et al., 2010), (2013) developed a model for computing N and P removal based upon
providing motivation for identifying methods of enhancing N and P raft coverage, pond volume, and rate kinetics.
reduction within them. FTWs have been assessed for treatment of urban runoff (Borne et al.,
Floating treatment wetlands (FTWs) consist of buoyant rafts upon 2013; Lynch et al., 2015; Tanner and Headley, 2011; Wang and Sample,
which emergent wetland plants are planted; with the plant roots sub- 2014; White and Cousins, 2013; Winston et al., 2013), combined
merged below the water surface (Headley and Tanner, 2006; White, sewage overflows (Van de Moortel, 2008), acid mine drainage (Smith
2013) (Fig. 1). Rafts can be homemade, some are commercially avail- and Kalin, 2000), and agricultural runoff and wastewater (Garcia
able; however, their nutrient removal efficacies are reported to be si- Chance and White, 2018; Hubbard et al., 2004; Stewart et al., 2008;
milar (Wang and Sample, 2014). FTWs are also referred to floating White and Cousins, 2013; Yang et al., 2008). Assessments of various
wetlands (Shahid et al., 2018), hydroponic root mats (Chen et al., 2016) plant taxa using replicated FTW mesocosms have been conducted;
or constructed floating wetlands (Pavlineri et al., 2017). Treatment Headley and Tanner (2007) assessed Carex virgata, Cyperus ustulatus,
occurs through direct nutrient uptake by vegetation and through in- Juncus edgariae, and Schoenoplectus tabernaemontani; Wang and Sample
creased sedimentation aided by biofilms growing on plant roots (Borne (2014), Pontederia cordata and Schoenoplectus tabernaemontani; Lynch
et al., 2015a). FTWs float, unlike constructed wetlands, allowing them et al. (2015), Juncus effusus; Chang et al. (2012b), Garcia Chance and
to adapt to water level fluctuations (Headley and Tanner, 2007; Lane White (2018), Juncus effusus and Canna flaccida; and Keizer-Vlek et al.
et al., 2016). These properties make FTWs an ideal and cost effective (2014), Iris pseudacorus and Typha angustifolia.
retrofit to TRBs to enhance treatment while potentially providing ad- Yang et al. (2008) reported Oenanthe javanica (a species native to
ditional space for growing crops (White, 2013; Winston et al., 2012). Asia) planted in FTWs resulted in TN and TP reductions of 64% and
FTWs are commonly assessed across a single growing season using 13%, respectively, when receiving agricultural runoff of 7.94 mg·L−1
replicated mesocosms. This experimental approach allows researchers TN and 1.54 mg·L−1 TP; which are within the concentrations observed
to limit extraneous variables, control influent nutrient concentrations, at a commercial nurseries. Wang et al. (2018) assessed Oenanthe java-
and identify key treatment factors; they also offer advantages of nica, but at water temperatures below 10 °C that limited shoot growth
economy in comparison with field-scale experiments. Using this ap- and uptake. TN reduction was 50%, compared with 40% without O.
proach, Chen et al. (2016) classified FTW performance by design factors javanica. This small difference was attributed primarily to microbial
such as plant species, raft dimensions, hydraulic loading rate (HLR) and activity associated with biofilms present on the plant roots.
hydraulic retention time (HRT), and/or and water application (i.e., li- The only study to date that focused on container nursery runoff
vestock runoff, wastewater, stormwater, etc.). Pavlineri et al. (2017) treatment with FTWs was White and Cousins (2013). This study as-
reviewed TN, TP, ammonia, and chemical oxygen demand (COD) sessed Juncus effusus and Canna flaccida supplied with simulated nur-
treatment through a variety of scales from microcosms to in situ ap- sery runoff using a 3-day HRT through two growing seasons. Influent
plications. An extensive review of FTWs including source waters, con- concentrations of 0.85 mg·L−1 N and 0.08 mg·L−1 P were used in the
taminants, FTW design characteristics, plants, plant-bacteria, the effects first year and 1.88 mg·L−1 N and 0.22 mg·L−1 P were used in the second
of water depth, season, temperature, aeration, and harvest on a variety year. N and P concentrations were reduced by 84% and 75%, respec-
of pollutants is provided by Shahid et al. (2018). Wang et al. (2014) tively in the first year; and 58% and 46%, respectively in the second
reviewed FTW evaluations at the mesocosm and microcosm scale. Wang year. The reduction in effectiveness the second year was attributed to
et al. (2014) and Pavlineri et al. (2017) provide performance metrics for higher influent loadings; therefore, the authors recommended that ad-
various FTW plant species for N and P removal. Available literature ditional species and nutrient concentrations be evaluated to completely
indicates that performance varies widely with specific location and assess the potential of FTWs to treat nursery runoff.
plant species. Mesocosm assessments are commonly conducted at a The objective of this research was to meet the aforementioned re-
single HRT. However, analysis of removal rates at fractions of an HRT search gap, and evaluate the efficacy of FTWs for nutrient remediation
would be useful in understanding the dynamics of nutrient removal. at container nursery operations. This study evaluated FTWs planted

469
J.T. Spangler et al. Ecological Engineering 127 (2019) 468–479

with Pontederia cordata and Juncus effusus in replicated mesocosms with 2.2. Experimental design
two nutrient concentrations representative of the range observed in
commercial nursery runoff. The species were selected because they are The experiment was set up in a randomized complete block design
native to the mid-Atlantic region of the U.S., and had potential as sal- that included four mesocosm replicates of four treatment types and two
able crops. The effectiveness of each species for reducing N and P in the nutrient concentrations (Fig. 3). The treatment combinations for the
water column was assessed on average at the HRT, and the shape of the study are listed in Table 1. For each nutrient concentration, high
removal curve was assessed using within-HRT subsamples. (17.1 mg·L−1 TN and 2.61 mg·L−1 TP) and low (5.22 mg·L−1 TN and
Additionally, an assessment of plant uptake of nutrients in roots and 0.52 mg·L−1 TP), there were four treatments that included the fol-
shoots was conducted. lowing: 1) Pontederia cordata, 2) Juncus effusus, 3) a mat and no plants,
and 4) no mat and no plants (open water). The mesocosms were
blocked by row to aid in organization of data collection and to correct
2. Methods for exposure. The northern most row was block 1 and each subsequent
southern row being blocks 2, 3 and 4.
2.1. Experiment location and equipment The water depth of each mesocosm was 47 cm when full and had a
water surface area of 0.79 m2. The 1.3 cm thick closed cell foam floating
The 19-week study was conducted in Virginia Beach, VA at the rafts used were commercially available FTW systems manufactured by
Virginia Tech Hampton Roads Agriculture Research and Extension Beemats (Beemats LLC, New Smyrna Beach, FL, USA). The mats came
Center (HRAREC; 36°53′N, 76°10′W). Weather data recorded from 5 with puzzle-piece edges (Fig. 4) for custom raft sizing and pre-cut holes
June to 19 October 2015 (136 days) included 559 mm of cumulative that can be used with aerator cups to hold the plants, also distributed by
rainfall and pooled mean, minimum and maximum temperature of Beemats. The ends of each mat were trimmed to provide a proper fit in
24.7, 7.5, and 36.3 °C across the duration of the study. Two 5867 L the mesocosms leaving a total mat surface area of 0.64 m2 (0.55 m2 if
(1550 gal) plastic storage tanks (Norwesco, St. Bonifacius, MN, USA), subtracting out the precut holes), as shown in Fig. 4. The FTW covered
also referred to as the mix tanks, were used for the batch process ma- 80.3% of the water surface area when mesocosms were planted.
keup of fertilizer solutions (one tank for high and one for low fertilizer Plants were purchased from Environmental Concerns (St. Michaels,
concentrations). Each tank drain was connected to the intake of an MD, USA) and were ≈5 cm in length (plugs). Prior to planting, the
electric water pump (Model FH40-5500, Little Giant, Oklahoma City, roots of each plant were rinsed thoroughly to remove as much of the
OK, USA). The discharge of each pump consisted of a main supply line planting substrate as possible. The rinsing process was effective for the
and a recirculation line. The recirculation line allowed for mixing of Pontederia cordata roots but was less effective for the tightly bound root
soluble fertilizer for uniform nutrient laden solution before being de- balls of the Juncus effusus. The difference in root balls after rinsing is
livered to individual mesocosms. The main supply line for each mix shown in Fig. 5a and b.
tank had four headers, one for each of the four rows of mesocosms. The Immediately after rinsing, plant roots were wrapped in coir (co-
headers were tied into the main supply line on both ends of the piping conut fiber), placed in the plastic aerator cups, and installed in the mats
layout, creating a loop to aid in equalizing the flow. Each of the four according to the design of the experiment. Each mesocosm-fitted mat
rows had eight fill lines, two fill lines per mesocosm. contained 20 planting holes. During this study, 50% of the planting
Thirty-two structural foam stock tanks (Rubbermaid Commercial holes were used. Five plants were randomly planted on each half of the
Products, Winchester, VA, USA) were used as mesocosms. Each 378.5 L mat for a total of ten plants per mesocosm (approximate density of 15
mesocosm was retrofitted with an overflow hole that allowed a max- plants/m2). A representation of this planting scheme is shown in Fig. 6.
imum water volume of 302.8 L. Every mesocosm had a plumbed main After each 7-day HRT, the mesocosms were drained and refilled with a
drain system, connected to the intake of a Honda WX15 gas powered new batch of simulated runoff. The simulated runoff was created by
water pump (Honda Power Equipment, Alpharetta, GA, USA), to fa- adding water and 66.2 g (low concentration) or 368 g (high con-
cilitate drainage of water to an appropriate conveyance system and thus centration) of water soluble fertilizer (24-8-16 Southern Agriculture
avoid flooding the experiment. The discharge of the drain pump was Insecticides Inc., Hendersonville, NC, USA) to each mix tank. After al-
connected to a series of 15.24 m (50 ft) sections of Goodyear 3.8 cm (1.5 lowing the solution to recirculate for 1 h, it was pumped to each me-
in) Spiraflex hose. The experimental equipment just after construction socosm. During the first five weeks, the high concentration treatment
is shown in Fig. 2. only received 221 g (10.6 mg·L−1 TN and 1.68 mg·L−1 TP) of fertilizer
to allow for plant acclimation.

2.3. Water sampling and analysis

The start of each experimental week (7-day retention time) was

designated as day 0. On day 0, grab samples were collected into 125 mL
wide-mouth Nalgene bottles (Thermo Fisher Scientific, PA, USA) from
each mix tank at a depth of 30 cm (12 in) below the water surface
before the mesocosms were filled with low or high nutrient solution to
represent the initial nutrient concentration for the 7-day retention
period. Water temperature (°C), pH, dissolved oxygen (DO, mg·L−1),
and electrical conductivity (EC, mVolt) measurements were taken in situ
for both mix tanks using a YSI Professional Plus multi-probe meter
(Yellow Springs International Inc. Ohio, USA) at a depth of 60 cm (24
in) from the water surface.
The start of each experimental week was also day 7 of the prior 7-
day retention time. On day 7, 125 mL grab samples were collected from
each of the 32 mesocosms prior to draining. Additionally, water tem-
perature, pH, DO, and EC measurements were taken in situ for each
mesocosm at a depth of 30 cm. These samples reflected the final (post
Fig. 2. Constructed system for mesocosm experiments. treatment) nutrient concentration for the retention period.

470
J.T. Spangler et al. Ecological Engineering 127 (2019) 468–479

Fig. 3. Schematic layout and experimental design.

Table 1
Treatment combinations for the floating treatment wetland study.
Treatment Mat Plants Species Concentration

1 Yes Yes Pontederia cordata Low

2 Yes Yes Pontederia cordata High
3 Yes Yes Juncus effusus Low
4 Yes Yes Juncus effusus High
5 Yes No n/a Low
6 Yes No n/a High
7 No No n/a Low
8 No No n/a High

Fig. 4. Beemats sized to fit mesocosm surface area at operational water levels.

Additionally, grab samples and in situ measurements were taken at the Fig. 5. Pontederia cordata (a) and Juncus effusus (b) roots after rinsing.
30 cm depth every other week on days 3 and 5 for each mesocosm.
analysis could be performed or were acidified to a pH of 2 in pre-
After collection, all water samples were kept on ice in a cooler until
paration for shipping. All collected water samples were analyzed at
other necessary fieldwork was complete. Subsamples were frozen until
Virginia Tech for TN and TP using automated flow injection analysis

471
J.T. Spangler et al. Ecological Engineering 127 (2019) 468–479

900
0.408 (Rn G) + u (e
T + 273 2 s
ea )
ETO =
+ (1 + 0.34u2 ) (1)

where ETO is the reference crop evapotranspiration (mm day−1), Rn is

the net radiation at the crop surface (MJ·m−2·day−1), G is the surface
heat flux (MJ·m−2·day−1), T is the air temperature at 2 m height (°C),
u2 is the wind speed at 2 m height (m·s−1), es is the saturation vapor
pressure (kPa), ea is the actual vapor pressure (kPa), es-ea is the vapor
pressure deficit (kPa), Δis the slope vapor pressure curve (kPa·°C−1),
and γ is the psychometric constant (kPa·°C−1). A water budget model
that utilized daily rainfall and potential evapotranspiration was created
to estimate daily water volume. Since the mesocosms had a maximum
fixed capacity, spills were also estimated. Weekly nutrient removal ef-
ficiency was calculated using Eq. (2):
(Ni Nf )
Removal efficiency = 100
Ni (2)

where Ni is the initial nutrient load and Nf is the nutrient load after a
Fig. 6. Representation of the planting scheme for the 2015 experiment.
specified retention time. Cumulative removal efficiency was calculated
using Eq. (3):
after persulfate digestion (persulfate digestion methods QuickChem® n n
Method 10-107-04-4B and 10-115-01-4B; Lachat Instruments, ( w=1
Ni w=1
Nf )
Removal efficiency = n 100
Loveland, CO, USA). In addition, water samples were analyzed for w=1
Ni (3)
metal content using inductively coupled plasma optical emission spec-
where the sum of each Ni and Nf was taken across all weeks.
trometry (ICP-OES 7400 Duo, Thermo Fisher Scientific, PA, USA).
Removal of TN and TP was plotted weekly, corresponding to the
The open design chosen for this study better mimicked field con-
HRT. A variety of functions available in SAS JMP® Pro 13.0.0 (SAS
ditions than if a rain shelter was used; i.e., full sunlight was available to
Institute Inc. Cary, NC, USA) were assessed using nonlinear regression
the plants; therefore, a simple water balance was employed to estimate
curve fitting. The mechanistic growth model shown in Eq. (4) was
the effects of precipitation and evapotranspiration to more accurately
chosen because of its shape, mathematical properties, and accuracy of
quantify water volume and evaporation. A water balance model was
data fit.
developed using weather data collected at the experiment site. This
process is described in more detail in the data analysis section. Nutrient removal = a (1 b e c days after load )
(4)

where a is the asymptote in mass units (g or mg), b is the scale, and c is

2.4. Plant tissue sampling and analysis the growth rate in days−1. Lines were fit using average removal (n = 4)
for 2, 4, and 7 days after initial loading. This functional form in-
Tissue samples were collected from three plants from each meso- corporates an exponential growth phase, and, as time advances, an
cosm at the end of the study. Immediately following the harvest, the asymptote.
roots and shoots were separated and stored in paper bags for drying. All values are reported as mean ± the standard error of the mean
The roots and shoots from Pontederia cordata samples were separated at unless otherwise noted. SAS JMP® Pro 13.0.0 was used to perform
the crown of the plant. The roots and shoots from Juncus effusus samples statistical analyses. Normality assumptions were tested visually using
were separated at the bottom of the planting cup and the crown on the the histogram and residuals, and with the Shapiro-Wilk goodness-of-fit
plant, respectively. The modified procedure for Juncus effusus was ne- test and suggested guidelines for skew and kurtosis. Where appropriate,
cessary because the portion of roots in the planting cups was too en- analysis of variance (ANOVA) or regression (linear and nonlinear as
tangled with the planting coir for proper separation. necessary) methods were used to identify statistical significance or
Samples were dried in a forced air oven at 58 °C until all moisture identify causal relationships, respectively. Student’s t was used for
was released and consistent sample weights maintained for two con- pairwise comparisons and Tukey’s Honestly Significant Difference
secutive days. Tissue samples were then ground to ≤0.5 mm particle (HSD) was used for multiple comparisons (p ≤ 0.05). For non-normal
size using a 3379-K35 Variable Speed Digital ED-5 Wiley Mill set to 900 data, Wilcoxon/Kruskal-Wallis tests (rank sums) were used for non-
RPM (Thomas Scientific, Swedesboro, NJ, USA). Due to budget con- parametric each pair comparisons (p ≤ 0.05).
straints, all three samples of shoots from one mesocosm were ground
into one composite sample and mixed thoroughly. The same procedure
3. Results and discussion
was used for the roots. Tissue samples were analyzed for N in a nitrogen
combustion analyzer (LECO FP528 Nitrogen Combustion Analyzer,
3.1. Aqueous nitrogen and phosphorus removal
Leco Corp., St. Joseph, MI, USA). Samples were also analyzed for P on
an inductively coupled plasma-optical emission spectrometer (Spectro
The average concentration of TP and TN in the high concentration
Arcos ICP-OES, Spectro Analytical Instruments-a division of Ametek,
treatments at the start of the 7-day HRT was 2.61 ± 0.04 mg·L−1
Kleve, Germany).
(0.79 g mass load) and 17.13 ± 0.24 mg·L−1 (5.19 g mass load), re-
spectively. The average concentrations of TP and TN in the low con-
2.5. Data analysis centration treatments was 0.52 ± 0.01 mg·L−1 (0.16 g mass load) and
5.22 ± 0.03 mg·L−1 (1.58 g mass load, respectively. These concentra-
Water sample data was reported with concentration units (mg·L−1) tion levels fall within the typical N and P ranges of nursery runoff
and was converted to mass using mesocosm volume. Since mesocosm (Taylor et al., 2006; White et al., 2011; White et al., 2010).
evapotranspiration was not measured directly in the field, it was esti- Cumulative TN and TP removal for the study are shown in Table 2.
mated as potential evapotranspiration using the FAO 56 Penman- The mean removal efficiency for TP after 19 weeks ranged from 38.9%
Monteith reference crop method, shown in Eq. (1): to 92.4% for the low concentration nutrient treatments and 15.3% to

472
J.T. Spangler et al. Ecological Engineering 127 (2019) 468–479

Table 2 Table 3
Mean cumulative removal for total nitrogen (TN) and total phosphorus (TP) by Mean total nitrogen (TN) and total phosphorus (TP) removal rates (g·m−2 d−1)
nutrient concentration (n = 4 for each treatment) in the floating treatment and associated correlation values (R2) after plant establishment (weeks 5–19)
wetland experiment conducted from June 2015–October 2015 (19 weeks). (n = 4) for the floating treatment wetland study.
Treatment TP (g)1 TP (%) TN (g)1 TN (%) Treatment TP (g·m−2 d−1) TP (R2) TN (g·m−2 d−1) TN (R2)

High nutrient concentration High nutrient concentration

Control – no cover 3.67b 24.4 49.1b 49.8 Control – no cover 0.041 0.972 0.680 0.998
Control – mat cover 2.30c 15.3 24.9c 25.3 Control – mat cover 0.027 0.973 0.370 0.991
Juncus effusus 4.08b 27.1 35.0c 35.5 Juncus effusus 0.057 0.983 0.505 0.994
Pontederia cordata 13.6a 90.3 83.0a 84.3 Pontederia cordata 0.203 0.998 1.232 0.999
ANOVA F Ratio, p-value 434.95, < 0.0001 81.45, < 0.0001
Low nutrient concentration
Low nutrient concentration Control – no cover 0.016 0.994 0.219 0.992
Control – no cover 1.52c 50.5 18.4b 61.2 Control – mat cover 0.014 0.986 0.135 0.983
Control – mat cover 1.17c 38.9 10.4c 34.7 Juncus effusus 0.031 0.996 0.290 0.998
Juncus effusus 2.20b 73.4 19.9b 66.3 Pontederia cordata 0.036 0.998 0.351 0.999
Pontederia cordata 2.77a 92.4 26.7a 88.9
ANOVA F Ratio, p-value 68.18, < 0.0001 97.90, < 0.0001

1
respectively) control treatments. TP removal results suggest there is no
Means with different letter differ significantly from other means with the
advantage to using a mat-only coverage without plants in comparison
same concentration at p < 0.05.
to no coverage at all for the evaluated treatments. At the high TP
concentration, Pontederia cordata performed better than the open water
90.3% for the high concentration treatments. Pontederia cordata re-
(p < 0.0001) and mat-only (p = 0.0001) control treatments; Juncus
moved an average of 2.77 ± 0.02 g TP over the experimental period
effusus only performed better than the mat-only (p = 0.0013) control
from the low nutrient concentration treatment, which was higher
treatment.
(p = 0.0027) than the Juncus effusus (2.20 ± 0.06 g) treatments. In
The TN removal efficiency ranged from 34.7% to 88.9% and 25.3%
comparison, at the low TP concentration, Pontederia cordata and Juncus
to 84.3% for the low and high concentration treatments, respectively.
effusus performed better than the open water (p < 0.0001 and
Pontederia cordata removed an average of 26.7 ± 0.28 g TN for the low
p = 0.0006, respectively) and mat-only (p < 0.0001 and p < 0.0001,
concentration treatment and 83.0 ± 1.90 g TN for the higher

a b

c d

Fig. 7. Mean cumulative TN (high (a) and low (c) concentration) and TP (high (b) and low (d) concentration) removal for by treatment type as a function of
experimental week for the floating treatment wetland study conducted from June 2015–October 2015. Linear regression lines fitted for weeks 5 through 19 represent
removal rate after plant establishment phase.

473
J.T. Spangler et al. Ecological Engineering 127 (2019) 468–479

Fig. 8. Weekly fitted TN removal curves for high (a) and low (b) concentration Pontederia cordata treatments during the floating treatment wetland study conducted
from June 2015–October 2015.

Table 4 concentrations were reduced by 84% and 58% in the first and second
Summary of mechanistic growth model fit for the FTW study TN removal by years, respectively.
day for Pontederia cordata high and low concentration treatments. The cumulative TP and TN removal as a function of experimental
Concentration AICc BIC SSE MSE RMSE R2 week for both the high and low concentration treatments is shown in
Fig. 7a–d. An evident change in slope occurred four weeks after in-
High 211.308 23.647 0.250 0.028 0.167 0.998 itiation, which became the designated time for transition from estab-
Low 109.549 −78.112 0.015 0.002 0.041 0.999
lishment phase to maximum treatment efficacy. Lynch et al. (2015)
reported a period of eight weeks after experiment initiation as the es-
tablishment phase, in contrast to the shorter period observed in this
concentration treatments, higher cumulative mass removals than all
study. The considerable difference in establishment time could be due
other treatments (p < 0.0001 and p < 0.0001, for high and low
to the much higher nutrient loads used in this study, more than 7 times
concentrations, respectively). TN removal performance for the low
greater for TN and 4 times greater for TP at the low nutrient con-
concentration solution by Juncus effusus, 19.9 ± 0.73 g, did not differ
centration. Mean removal rate and the associated R2 of maximum
from the open water control treatment (p = 0.3982), 18.4 ± 0.71 g.
treatment efficacy study phase were calculated for weeks five though
Both the Juncus effusus and open water control treatments removed
nineteen (Table 3). Pontederia cordata had higher removal rates than all
higher amounts of N (p < 0.0001 and p < 0.0001), respectively, than
other treatments at both high and low concentrations (TN and TP;
the mat-only control treatment at low N concentration, again sug-
p < 0.0001), while the mat-only treatment performed the worst.
gesting that the addition of a mat without plants adds little value for N
FTW performance studies are typically reported in g·m−2 d−1,
removal. At high concentration, the performance of Juncus effusus was
which assumes a linear nutrient removal rate over the study period. The
lower (p = 0.0178) on average than the open water control treatment.
TN removal rate within the 7-day HRT for the high and low con-
Juncus effusus, in the high concentration treatments, became infected
centration Pontederia cordata treatment for every other week from
halfway through the study by a rust disease of the Uromyces genus.
weeks three to nineteen are shown in Fig. 8a and b. Tables 4 and 5 show
Uromyces is the most reported disease for Juncus effusus in the U.S. (Farr
the summary fit for the mechanistic growth model (Eq. (4)) and asso-
and Rossman, 2017). The presence of the rust could be one contributor
ciated nonlinear regression parameters. A key difference between the
to the poor nutrient removal performance relative to the control
high and low concentration plots in Fig. 8a and b, respectively, is that at
treatments.
low concentrations (Fig. 8b), the plants had much higher growth rates
These results compare favorably with those of the aforementioned
(slope of the curve) after the initial period than those at high con-
White and Cousins (2013) study where P concentrations were reduced
centration (Fig. 8a). However, the absolute mass removal rate of TN for
by 75% and 46%, in the first and second years, respectively; N

Table 5
Nonlinear regression parameters for TN removal by Pontederia cordata high and low concentration treatments.
Experiment Week High Concentration Low Concentration

Asymptote Scale Growth Rate Asymptote Scale Growth Rate

3 3.074 1.015 0.164 2.058 1.009 0.142

5 2.866 0.991 0.255 1.453 1.007 0.211
7 5.807 1.003 0.286 1.547 1.000 0.963
9 5.416 0.999 0.554 1.607 1.000 12.307
11 5.025 0.998 0.775 1.490 1.000 16.374
13 6.667 0.999 0.596 1.514 1.000 2.281
15 5.753 0.998 0.519 1.545 1.000 2.083
17 5.554 0.999 0.500 1.471 1.000 16.746
19 3.984 1.000 0.548 1.476 1.000 28.571

474
J.T. Spangler et al. Ecological Engineering 127 (2019) 468–479

06/29/15 07/27/15 08/24/15 09/21/15

Low concentration Pontederia cordata

06/29/15 07/27/15 08/24/15 09/21/15

Low concentration Juncus effusus
Fig. 9. Plant growth over time for Pontederia cordata (top) and Juncus effusus (bottom) during the floating treatment wetland study conducted from June
2015–October 2015.

the high concentration mesocosms was higher. An additional difference greater removal of nitrate-N by increasing contact time. The study
was that as the plants matured, the asymptote was more quickly concluded that HRTs should be optimized when using FTWs for pol-
reached. Similar behavior was observed for Juncus and for high and low lution removal. Additionally, in a 1:10 scale retention pond replica,
concentrations of TP; for brevity, this information is provided in Ap- Khan et al. (2013) determined that hydraulic performance of the pond
pendix D of a companion open access article (Spangler et al., 2018). The depended heavily on the size and position of the FTW in regards to the
nonlinear uptake rate suggests that nutrient removal could be opti- pond surface and inlet.
mized by varying HRT. The reader may note that the rate TN removal in
weeks three and five was much lower rate than the remainder of the
weeks; this could be attributed to plant establishment and biomass 3.2. Plant growth
accumulation through growth. Uptake rates peaked during week 11 and
13 for both taxa. Initial nutrient load influenced the plant shoot biomass of Pontederia
In a FTW mesocosm study using synthetic greywater, Abed et al. cordata. Higher initial concentrations resulted in more Pontederia cor-
(2017) observed no significant differences in ammonium-N between the data biomass. No difference in biomass were observed by nutrient load
inflow and outflow for two-day and seven-day contact times, but cited for Juncus effusus. Root and shoot growth throughout the study for the
two low concentration treatments are shown in Fig. 9. Pontederia

475
J.T. Spangler et al. Ecological Engineering 127 (2019) 468–479

low concentration shoots (79.0 ± 5.8 g, p < 0.0001).

The opposite was found for plant root biomass, where low initial
nutrient loads increased root growth in Juncus effusus, but not
Pontederia cordata (Fig. 10), compared with high initial nutrient con-
centrations. The root dry weight of Pontederia cordata high and low
concentration treatments did not differ (p = 0.8389) but were ≈20.0 g
and 10.7 g larger than Juncus effusus treatments for high and low con-
centrations, respectively. No difference was noted for Juncus effusus
shoot dry weights between high and low concentrations (p = 0.0885);
however, the low concentration Juncus effusus root dry weight
(14.2 ± 1.9 g, p = 0.0003) was higher than plants exposed to the
greater nutrient concentration. This result could be attributed to root
morphological changes under conditions with lower P availability
(Mengel et al., 2001). Since nutrient availability is low, the roots should
grow longer in search of more nutrients.
Pontederia cordata high concentration shoots accumulated
Fig. 10. Root and shoot dry weights (mean ± standard error) by nutrient 4.87 ± 0.49 g of N, more than all other combinations of species and
treatment and species for the floating treatment wetland study conducted from nutrient treatments (p = 0.0304, Fig. 11a and b). The low concentra-
June 2015–October 2015. n = 12. tion Pontederia cordata shoots (1.84 ± 0.05 g) and the high con-
centration Juncus effusus shoots (2.05 ± 0.07 g), accumulated more N
cordata shoots from high concentration treatments had a mean dry (p = 0.0304) than the low concentration Juncus effusus shoots
weight of 260 ± 28 g, which was larger than Pontederia cordata low (1.37 ± 0.04 g). Pontederia cordata roots at the high concentrations
concentration shoots (177 ± 14 g, p = 0.0166), Juncus effusus high accumulated 0.42 ± 0.06 g, more than the other three treatment
concentration shoots (94.9 ± 8.8 g, p < 0.0001), and Juncus effusus combinations. Pontederia cordata (0.26 ± 0.02 g) and Juncus effusus
(0.18 ± 0.03 g) roots absorbed more N (p = 0.0304) than the high

a b

Fig. 11. Mean ( ± standard error) nitrogen (a) and phosphorus (b) accumulated in roots and shoots of Juncus effusus and Pontederia cordata after 19 weeks of
exposure to 2 nutrient levels (high = 17.1 mg·L−1 TN and 2.61 mg·L−1 TP; low = 5.22 mg·L−1 TN and 0.52 mg·L−1 TP) in a floating treatment wetland study.

Table 6
Total nitrogen (TN) and total phosphorus (TP) mass balance after a 19 week floating treatment wetland study with 2 plant species (n = 4 for each treatment).
High Nutrient Concentration Low Nutrient Concentration

TP (g) TN (g) TP (g) TN (g)

1
Total initial load 15.0 98.6 3.0 30.0

Pontederia cordata
Total load after 7-day HRT 1.46 15.5 0.23 3.33
Load reduction 13.6 83.0 2.77 26.7
Plant uptake2 9.43 (69.5) 52.9 (63.7) 2.08 (75.1) 21.1 (78.8)
Other removal processes 4.13 30.1 0.69 5.65

Juncus effusus
Total load after 7-day HRT 10.9 63.6 0.80 10.1
Load reduction 4.08 35.0 2.20 19.9
Plant uptake2 2.94 (72.1) 21.6 (61.7) 1.54 (70.0) 15.5 (77.7)
Other removal processes 1.14 13.4 0.66 4.44

1
n = 1 for initial load data.
2
Mean uptake (% of total load reduction).

476
J.T. Spangler et al. Ecological Engineering 127 (2019) 468–479

Table 7
Mean physicochemical properties after the 7-day hydraulic retention time for the entire 19-week floating treatment wetland study (n = 76).
Treatment DO (mg·L−1)1 pH1 Temperature (°C)2 (g)1 EC (μS·cm−1)1

Low, Control – no cover 11.18a 9.05a 23.2 205c

High, Control – no cover 7.66b 7.29b 23.3 253a
Low, Control – mat cover 8.75b 7.19b 24.5 201c
High, Control – mat cover 5.74c 6.02c 24.3 268a
Low, Juncus effusus 4.24d 5.91c 24.3 194cd
High, Juncus effusus 3.35de 4.90d 24.1 264a
Low, Pontederia cordata 2.61e 4.71d 23.9 183d
High, Pontederia cordata 2.40e 4.16e 23.8 225b
ANOVA F Ratio, p-value 15.74, < 0.0001 8.64, < 0.0001 10.89, 0.1437 6.18, 0.0004

1
Means with different letter differ significantly from other means at p < 0.05.
2
Data did not meet normality assumption. Wilcoxon tests used. Reported statistical values Chi Square, p-value.

concentration Juncus effusus roots (0.11 ± 0.01 g). Chang et al. coverage area and DO is an important factor to consider when im-
(2012b) found that Juncus effusus collected 0.036 g·m−2·d−1 TN after plementing FTWs as BMPs. (Garcia Chance and White, 2018) reported
an operational period of 3 months with influent nutrient concentrations that a 50% open water area, not covered by FTWs, allowed for in-
of ≈4 mg·L−1. Chang’s results are about 5 times less than the TN uptake creased gas exchange and resulted in higher DO levels. In addition to
(0.182 g·m−2·d−1) observed in this study for Juncus effusus. Potential DO, pH was significantly lower for the FTWs with plants in comparison
reasons for additional nutrient accumulation in this study are extended to the control treatments (Table 7). Wang and Sample (2014) reported
time covered by the study, nutrient loading rate, surface area coverage significantly lower pH levels for FTWs planted with Pontederia cordata,
and planting density. but their results were still closer to neutral than those observed in this
The P sequestered in the shoots of each treatment combination study. In a pond scale study, Borne et al. (2014) observed lower pH
differed (p = 0.0027). The shoots of Pontederia cordata for the high values for FTWs in comparison to the control, but they concluded that
concentration treatment accumulated 0.90 ± 0.06 g of P per plant; the water treated with FTWs had a near neutral pH. Lower pH could
whereas the shoots of Juncus effusus accumulated only 0.27 ± 0.01 g occur because plant roots have the potential to release acidic exudates
per plant. The shoots of Pontederia cordata and Juncus effusus for the low (Coleman et al., 2001; Marschner, 1995). Diurnal fluctuations in DO
concentration treatments accumulated 0.19 ± 0.01 g and and pH can be attributed to photosynthesis and respiration processes
0.14 ± 0.01 g, respectively. Root accumulation of P was greater for (Reeder, 2011). Since physicochemical properties were consistently
Pontederia cordata for high concentration treatments (0.04 ± 0.01 g P, recorded during morning hours, respiration effects from the previous
p = 0.0304), all other treatments assimilated a similar mass of P within night could be the reason for lower pH and DO in the FTW treatments.
root tissues (0.02 ± 0.01 g P). These results contrast those reported by There was no difference between treatments for observed water tem-
Chang et al. (2012a), which was inconclusive on which species (Pon- perature on day 7.
tederia cordata or Juncus effusus) performed better for nutrient removal.
The mass balance for TP and TN, shown in Table 6, indicates that
3.4. Limitations
plant uptake of nutrients accounted for the majority of nutrient removal
processes in this FTW study. The ratio of plant uptake to total nutrient
The “open” design of the experiment allowed rain to fall directly
load reduction was similar for both plant species at a given con-
into the mesocosms. This design best mimicked nursery TRB conditions
centration. The ratio of uptake to total load reduction for TN was higher
and potentially accounted for the unexplainable error in the reported
for Pontederia cordata (p = 0.0284) and Juncus effusus (p = 0.021) for
nutrient recovery. The limited project budget did not allow for in-
low concentration treatments than high concentration treatments. If DO
stallation of a rain shelter or more frequent sample collection that may
levels are low, denitrification processes likely contribute to N removal
have ensured nutrient concentrations were not diluted nor nutrient
that is not accounted for by plant uptake. Unaccounted for P removal
removal performance overstated. In contrast, evapotranspiration effects
may be the result of binding to sediment, potentially from dead algal
could have resulted in more concentrated samples. Additionally, the
biomass. Since grab samples were taken in the middle of the water
mesocosm approach resulted in a large surface area coverage ratio
column, P representation in the samples could be low. White and
(80.3%), which could be cost prohibitive for implementation in a TRB.
Cousins (2013) reported similar mass balance results, which confirmed
The scaling of these results for field applications could overestimate
that additional removal processes occurred. For example, in a mixed
performance if close attention is not given to surface area coverage,
planting using Juncus effusus, White and Cousins (2013) reported that
species, and planting density.
32.9% of P was removed by processes other than plant uptake, which
Mid-Atlantic nurseries reduce their production during colder
compares with unaccounted for P removal of 30.0% in the low con-
months of the year when plants are dormant; requiring less or no need
centration Juncus effusus treatment.
for water treatment in the winter season. At the same time, FTW plants
and subsequent biofilms would be expected to decrease in size or even
export nutrients as plants become less active or senesce roots and
3.3. Physicochemical responses
shoots. Therefore, nutrient loading throughout the year and in different
ecophysiographic regions should be assessed when determining treat-
Planted FTWs had significantly lower DO in comparison to the
ment technology efficacy.
treatments without plants (Table 7). Tanner and Headley (2011) ob-
served similar results where treatments containing higher biomass ex-
perienced lower DO levels. Shade produced by FTWs can reduce pho- 4. Conclusions
tosynthetic activity, thus reducing DO production (Borne et al., 2015b).
If DO is lowered enough to create anoxic conditions, N removal will This mesocosm study evaluated the nutrient removal performance
occur through denitrification in addition to plant uptake. This concept of FTWs planted with Pontederia cordata and Juncus effusus under si-
supports the results shown in Table 6 where plant uptake is not solely mulated container nursery TRB conditions using representative high
responsible for TN reduction. The relationship between surface and low nutrient loading levels of TN and TP. Pontederia cordata was

477
J.T. Spangler et al. Ecological Engineering 127 (2019) 468–479

clearly the superior plant species, removing 90.3–92.4% of TP and Coleman, J., Hench, K., Garbutt, K., Sexstone, A., Bissonnette, G., Skousen, J., 2001.
84.3–88.9% of TN at high and low nutrient loading rates, respectively. Treatment of domestic wastewater by three plant species in constructed wetlands.
Water Air Soil Pollut. 128, 283–295.
Weekly plant nutrient removal curves are generally nonlinear, in- Van de Moortel, A.M.K., 2008. Use of floating macrophyte mats for treatment of CSOs.
creasing rapidly to a limit, indicating consumption of available nu- 11th International Conference on Urban Drainage. Edinburgh, Scotland.
Farr, D.F., Rossman, A.Y., 2017. Fungal Databases, U.S. National Fungus Collections. ARS,
trients at the end of each HRT. Therefore, specific consideration should USDA.
be given to nutrient loading rates, HRTs, nutrient removal goals, and Garcia Chance, L.M., White, S.A., 2018. Aeration and plant coverage influence floating
the uptake rates when selecting plant species. In an actual nursery, treatment wetland remediation efficacy. Ecol. Eng. 122, 62–68.
Headley, T.R., Tanner, C.C., 2006. Application of Floating Wetlands for Enhanced
loading rates will be continuous and fluctuate between high and low Stormwater Treatment: A Review. Auckland Regional Council, Hamilton, New
across the season, depending upon factors such as irrigation, rainfall, Zealand.
Headley, T.R., Tanner, C.C., 2007. Floating Wetlands for Stormwater Treatment: Removal
and fertilization practices. In this study, plant uptake was not the only
of Copper, Zinc and Fine Particulates. Prepared by NIWA for Auckland Regional
means of nutrient removal. Other removal processes accounted for up Council. Auckland Regional Council.
to 30% and 38% removal for TP and TN, respectively. Although FTWs Hubbard, R.K., Gascho, G.J., Newton, G.L., 2004. Use of floating vegetation to remove
nutrients from swine lagoon wastewater. Trans. ASAE 47, 1963–1972.
with higher loading rates removed more nutrients overall, plant uptake Jayaweera, M.W., Kasturiarachchi, J.C., 2004. Removal of nitrogen and phosphorus from
at lower loading rates accounted for a larger percentage of the overall industrial wastewaters by phytoremediation using water hyacinth (Eichhornia cras-
nutrient removal. Denitrification likely played a role in reducing TN sipes (Mart.) Solms). Water Sci. Technol. 50, 217–225.
Keizer-Vlek, H.E., Verdonschot, P.F.M., Verdonschot, R.C.M., Dekkers, D., 2014. The
that was not accounted for in the mass balance. Settling could explain contribution of plant uptake to nutrient removal by floating treatment wetlands. Ecol.
the reduction of TP removal, which was not accounted for via plant Eng. 73, 684–690.
Khan, S., Melville, B.W., Shamseldin, A., 2013. Design of storm-water retention ponds
accumulation. There appears to be a 4–5 week period before the plants with floating treatment wetlands. J. Environ. Eng. 139, 1343–1349.
transition from establishment phase to maximum treatment efficacy. Lane, S., Sample, D., Andy Lazur, A., Ryan Winston, R., Streb, C., Ferrier, D., Linker, L.,
Results showed there is no advantage to using a mat-only FTW treat- Brittingham, K., 2016. Recommendations of the Expert Panel to Define Removal
Rates for Floating Treatment Wetlands in Existing Wet Ponds: Final Report.
ment without plants. Further, longer duration research is recommended Lynch, J., Fox, L.J., Owen Jr, J.S., Sample, D.J., 2015. Evaluation of commercial floating
to assess additional plant species and system scalability to broad, field- treatment wetland technologies for nutrient remediation of stormwater. Ecol. Eng.
scale application for specific climates in the eastern U.S. 75, 61–69.
Majsztrik, J.C., 2011. Modeling Nitrogen, Phosphorus and Water Dynamics in Greenhouse
and Nursery Production Systems. ProQuest Dissertations Publishing.
Declaration of interest Majsztrik, J.C., Fernandez, R.T., Fisher, P.R., Hitchcock, D.R., Lea-Cox, J., Owen Jr., J.S.,
Oki, L.R., White, S.A., 2017. Water use and treatment in container-grown specialty
crop production: a review. Water Air Soil Pollut. 228, 151.
None. This research also served as the basis for a Master’s thesis by Majsztrik, J.C., Lea-Cox, J.D., 2013. Water quality regulations in the Chesapeake Bay:
the main author from the Virginia Polytechnic Institute and State working to more precisely estimate nutrient loading rates and incentivize best
management practices in the nursery and greenhouse industry. HortScience 48,
University. 1097–1102.
Majsztrik, J.C., Ristvey, A.G., Lea-Cox, J.D., 2011. Water and nutrient management in the
production of container-grown ornamentals. Hortic. Rev. 38, 253.
Acknowledgements
Marschner, H., 1995. Mineral Nutrition of Higher Plants. Academic Press, London; San
Diego.
Funding for this work was provided in part, by the U.S. Department Mengel, K., Kirkby, E.A., Kosegarten, H., Appel, T., SpringerLink, 2001. Principles of Plant
Nutrition. Springer Netherlands, Dordrecht.
of Agriculture (USDA), National Institute of Food and Agriculture Pavlineri, N., Skoulikidis, N.T., Tsihrintzis, V.A., 2017. Constructed floating wetlands: a
(NIFA). Specialty Crop Research Initiative Project Clean WateR3 project review of research, design, operation and management aspects, and data meta-ana-
(2014-51181-22372). Funding was also provided by the Virginia lysis. Chem. Eng. J. 308, 1120–1132.
Qin, B.Q., 2009. Lake eutrophication: control countermeasures and recycling exploita-
Agricultural Experiment Station and the Hatch Program of USDA-NIFA. tion. Ecol. Eng. 35, 1569–1573.
The authors appreciate the additional field support provided by Reeder, B.C., 2011. Assessing constructed wetland functional success using diel changes
in dissolved oxygen, pH, and temperature in submerged, emergent, and open-water
Jeanette Lynch and the lab support provided by Julie Brindley, Anna habitats in the Beaver Creek Wetlands Complex, Kentucky (USA). Ecol. Eng. 37,
Birnbaum, Dil Thavarajah, and Chris Lasser. The Virginia Water 1772–1778.
Resources Research Center and the William R. Walker fellowship award Shahid, M.J., Arslan, M., Ali, S., Siddique, M., Afzal, M., 2018. Floating wetlands: a
sustainable tool for wastewater treatment. Clean Soil Air Water.
provided additional support. Smith, M.P., Kalin, M., 2000. Floating wetland vegetation covers for suspended solids
removal. Quebec 2000: Millennium Wetland Event (Selected Papers), Treatment
References Wetlands for Water Quality Removal. Quebec City, Quebec, Canada.
Spangler, J.T., Sample, D.J., Fox, L.J., Owen, J.S., White, S.A., 2018. Floating treatment
wetland aided nutrient removal from agricultural runoff using two wetland species-
Abed, S.N., Abed, S.N., Almuktar, S.A., Scholz, M., 2017. Remediation of synthetic expanded information. Data in Brief. https://doi.org/10.1016/j.dib.2018.12.037. (In
greywater in mesocosm—scale floating treatment wetlands. Ecol. Eng. 102, 303–319. Press).
Borne, K.E., 2014. Floating treatment wetland influences on the fate and removal per- Stewart, F.M., Mulholland, T., Cunningham, A.B., Kania, B.G., Osterlund, M.T., 2008.
formance of phosphorus in stormwater retention ponds. Ecol. Eng. 69, 76–82. Floating islands as an alternative to constructed wetlands for treatment of excess
Borne, K.E., Fassman-Beck, E.A., Tanner, C.C., 2014. Floating treatment wetland influ- nutrients from agricultural and municipal wastes – results of laboratory-scale tests.
ences on the fate of metals in road runoff retention ponds. Water Res. 48, 430–442. Land Contam. Reclam. 16, 25–33.
Borne, K., Fassman-Beck, E., Winston, R., Hunt, W., Tanner, C., 2015a. Implementation Tanner, C.C., Headley, T.R., 2011. Components of floating emergent macrophyte treat-
and maintenance of floating treatment wetlands for urban stormwater management. ment wetlands influencing removal of stormwater pollutants. Ecol. Eng. 37, 474–486.
J. Environ. Eng. 4015030. Taylor, M.D., White, S.A., Chandler, S.L., Klaine, S.J., Whitwell, T., 2006. Nutrient
Borne, K., Fassman-Beck, E., Winston, R., Hunt, W., Tanner, C., 2015b. Implementation management of nursery runoff water using constructed wetland systems.
and maintenance of floating treatment wetlands for urban stormwater management. HortTechnology 16, 610–614.
J. Environ. Eng. 04015030. USEPA, 2013. National Summary of State Information. Environmental Protection Agency,
Borne, K.E., Tanner, C.C., Fassman-Beck, E.A., 2013. Stormwater nitrogen removal per- Washington, D.C.: U.S.
formance of a floating treatment wetland. Water Sci. Technol. 68, 1657–1664. Wang, P., Jeelani, N., Zuo, J., Zhang, H., Zhao, D., Zhu, Z., Leng, X., An, S., 2018..
Carpenter, S.R., Caraco, N.F., Correll, D.L., Howarth, R.W., Sharpley, A.N., Smith, V.H., Nitrogen removal during the cold season by constructed floating wetlands planted
1998. Nonpoint pollution of surface waters with phosphorus and nitrogen. Issues Ecol with Oenanthe javanica. Mar. Freshwater Res. 69.
1-1. Wang, C.-Y., Sample, D.J., 2014. Assessment of the nutrient removal effectiveness of
Chang, N.B., Islam, K., Marimon, Z., Wanielista, M.P., 2012a. Assessing biological and floating treatment wetlands applied to urban retention ponds. J. Environ. Manage.
chemical signatures related to nutrient removal by floating islands in stormwater 137, 23–35.
mesocosms. Chemosphere 88, 736–743. Wang, C.-Y., Sample, D.J., Bell, C., 2014. Vegetation effects on floating treatment wetland
Chang, N.B., Islam, M.K., Wanielista, M.P., 2012b. Floating wetland mesocosm assess- nutrient removal and harvesting strategies in urban stormwater ponds. Sci. Total
ment of nutrient removal to reduce ecotoxicity in stormwater ponds. Int. J. Environ. Environ. 499, 384–393.
Sci. Technol. 9, 453–462. Wang, C.-Y., Sample, D.J., 2013. Assessing floating treatment wetlands nutrient removal
Chen, Z., Cuervo, D.P., Müller, J.A., Wiessner, A., Köser, H., Vymazal, J., Kästner, M., performance through a first order kinetics model and statistical inference. Ecol. Eng.
Kuschk, P., 2016. Hydroponic root mats for wastewater treatment—a review. 61, 292–302 Part A.
Environ. Sci. Pollut. Res. 23, 15911–15928. White, S.A., 2013. Wetland technologies for nursery and greenhouse compliance with
USEPA, 2010. Chesapeake Bay Total Maximum Daily Load for Nitrogen, Phosphorus, and nutrient regulations. HortScience 48, 1103–1108.
Sediment. USEPA Region III, Philadelphia, PA. White, S.A., Taylor, M.D., Chandler, S.L., Whitwell, T., Klaine, S.J., 2010. Remediation of

478
J.T. Spangler et al. Ecological Engineering 127 (2019) 468–479

nitrogen and phosphorus from nursery runoff during the spring via free water surface a retrofit to existing stormwater detention basins. World Environmental and Water
constructed wetlands. J. Environ. Hortic. 28, 20–217. Resources Congress 2012.
White, S.A., Taylor, M.D., Albano, J.P., Whitwell, T., Klaine, S.J., 2011. Phosphorus re- Winston, R.J., Hunt, W.F., Kennedy, S.G., Merriman, L.S., Chandler, J., Brown, D., 2013.
tention in lab and field-scale subsurface-flow wetlands treating plant nursery runoff. Evaluation of floating treatment wetlands as retrofits to existing stormwater retention
Ecol. Eng. 37, 1968–1976. ponds. Ecol. Eng. 54, 254–265.
White, S.A., Cousins, M.M., 2013. Floating treatment wetland aided remediation of ni- Yang, Z.F., Zheng, S.K., Chen, J.J., Sun, M., 2008. Purification of nitrate-rich agricultural
trogen and phosphorus from simulated stormwater runoff. Ecol. Eng. 61, 207–215 runoff by a hydroponic system. Bioresour. Technol. 99, 8049–8053.
Part A. Yeager, T., Million, J., Larsen, C., Stamps, B., 2010. Florida Nursery best management
Winston, R.J., Kennedy, S.G., Hunt, W.F., 2012.. Evaluation of floating wetland islands as practices: past, present, and future. HortTechnology 20, 82.

479