You are on page 1of 5

Ecological Engineering 37 (2011) 20872091

Contents lists available at ScienceDirect

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

Short communication

Hydraulic residence time computation for constructed wetland design


Vahid Zahraeifard, Zhiqiang Deng
Department of Civil and Environmental Engineering, Louisiana State University, Baton Rouge, LA 70803, USA

a r t i c l e

i n f o

Article history:
Received 22 April 2011
Received in revised form 28 July 2011
Accepted 7 August 2011
Available online 6 September 2011
Keywords:
Constructed wetland design
Hydraulic residence time
VART model
Hydraulic loading rate

a b s t r a c t
Hydraulic residence time (HRT) is one of the key design parameters controlling the removal efciency of
contaminants and nutrients in stormwater and wastewater wetlands. The paper presents a new approach
to the estimation of HRT using the variable residence time (VART) model. The VART model is employed
to simulate the major processes (including advection, dispersion, and transient storage of contaminants/nutrients in vegetated zones) affecting HRT and thereby to produce a hydraulic residence time
distribution (HRTD) for a design wetland. The HRTD in combination with a moment-based method is
then utilized to nd a mean design HRT for the design wetland. Methods for estimation of parameters
governing the HRTD are proposed. The new approach to HRT computation is demonstrated through a case
study for the Tres Rios Demonstration (TRD) Wetlands in Arizona, USA. Modeling results show that the
design HRTs for the Hayeld wetland (H1) and the Cobble wetlands (C1 and C2) are 4.04, 4.66, and 2.65
days, respectively. The computed HRTs agree well with those reported by previous studies, conrming
the efcacy of the new approach to hydraulic design of constructed wetlands.
2011 Elsevier B.V. All rights reserved.

1. Introduction
Constructed wetlands are engineered systems that mimic natural processes to improve water quality by ltration, settling, and
bacterial decomposition in a natural-looking lined marsh. They are
capable of removing various nutrients and pollutants such as heavy
metals, BOD, and toxic compounds from contaminated waters at
relatively low cost (Yeh, 2008). Thus, constructed wetlands and
especially surface ow wetlands have been increasingly used for
treatment of urban stormwater runoff (Walker and Hurl, 2002) and
supplemental wastewater (Keefe et al., 2004). The treatment efciency of wetlands depends primarily on HRT (Danckwerts, 1953;
Walker, 1998; Conn and Fiedler, 2006). Therefore, major efforts
have been made in nding a proper method for estimation of HRT
(Levenspiel, 1972; Kadlec, 1994; Kruse et al., 2009).
The simplest method for HRT estimation is to calculate the nominal HRT that is dened as the ratio of the nominal wetland volume
to the ow rate at the outlet (Kadlec and Wallace, 2008). The
nominal HRT is based on the assumption of plug-ow reactors
(PFR). The problem with the nominal HRT is that not all wetland
volumes are involved in the ow path. Actually, many investigators
found the presence of three different hydraulic volumes or zones
in wetlands. The rst zone is the actively owing main channel;
the second one is a temporary storage zone where water and

Corresponding author. Tel.: +1 225 578 6850; fax: +1 225 578 8652.
E-mail address: zdeng@lsu.edu (Z. Deng).
0925-8574/$ see front matter 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.ecoleng.2011.08.011

constituents are exchanged with the main ow channel; the third


one is completely isolated, dead water (Kadlec, 1994; Stairs and
Moore, 1994; Buchberger and Shaw, 1995; Werner and Kadlec,
2000; Martinez and Wise, 2003a). In addition to the empirical
method, eld tracer experiments are often conducted and tracer
test data are commonly utilized in combination with the moment
method for estimation of HRT (Whitmer et al., 2000; Keefe, 2001;
Persson and Wittgren, 2003; Keefe et al., 2004). The advantage of
this approach is that a residence time distribution (RTD) can be
produced and thereby a mean HRT can be obtained for a wetland.
The major limitation of this approach, however, is that tracer
experiments can only be conducted in an existing wetland while
HRT is needed in the design stage of a constructed wetland. Therefore, extensive efforts have been made in developing a numerical
model that is capable of providing a HRTD in the design phase of a
wetland.
The transport and fate of contaminants and nutrients in constructed wetlands have classically been modeled using PFR and
continuously stirred tank reactors (CSTR) (Levenspiel, 1972). The
PFR and CSTR models were improved by Kadlec and Wallace (2008)
by using the tanks-in-series model which is characterized by a
number of CSTRs in series and a gamma function for describing
RTD. Wrman and Kronnas (2005) proposed a two-path tanksin-series model. Keefe et al. (2004) applied the one-dimensional
transport with inow and storage (known as OTIS) model to study
solute transport in constructed wetlands near Phoenix, Arizona.
The OTIS model was also calibrated to match results from tracer
experiments conducted in the Orlando Easterly Wetland (Martinez

2088

V. Zahraeifard, Z. Deng / Ecological Engineering 37 (2011) 20872091

Fig. 1. Schematic concentration breakthrough curve for calculation of HRT using


the moment-based method.

and Wise, 2003b). Wang and Jawitz (2006) presented an exponential extension model for simulating RTD in cell-network treatment
wetlands. The existing models are able to simulate a specic type of
observed RTDs. However, none of the models are capable of simulating various types of RTDs commonly observed in tracer injection
experiments (Deng and Jung, 2009).
The overall goal of this paper is to present a method for determination of HRT in the design stage of a constructed wetland. The
goal is addressed by applying the variable residence time (VART)
model that is able to simulate diverse RTDs of conservative solute
in constructed wetlands. The specic objective of this paper is to
demonstrate the design procedure through the redesign of the TRD
Wetlands in Arizona, USA.
2. Materials and methods

Statistical measures for residence time distributions and the


performance of a wetland are commonly described using various
moments (Kadlec and Wallace, 2008). Based on the analysis outlined by Levenspiel (1972), the RTD for a pulse tracer injection at
time t = 0 can be determined as:
Q (t) C(t)


0

Q (t) C(t) dt

n

Qi (t) Ci (t)

Q (t) Ci (t) t
i=1 i

(1)

in which RTD(t) is the residence time distribution [T1 ], Q(t) (Qi ) is


volumetric rate of the ow exiting the wetland at time t (ti ) [L3 /T];
C(t) (Ci ) is the concentration of the tracer exiting the wetland cell
at time t (ti ) [M/L3 ]; n is the number of samples; t = time step [T]
(Fig. 1). The denominator of Eq. (1) is the total mass of tracer recovered at the exit of wetland cell [M] (Martinez and Wise, 2003b). The
rst moment of RTD(t) gives the centroid of the area under the RTD,
which is statistically dened as the expectation or expected value
of RTD and often called actual hydraulic residence time (AHRT) in
wetland design:


AHRT =

t RTD(t) dt
=

n


ti RTDi (t) t

(2)

i=1

For a constant discharge Q, a combination of Eqs. (1) and (2) yields:


ti Ci (t)
AHRT = 
Ci (t)

2.2. VART model


The variable residence time (VART) model is a numerical tool
developed by Deng and Jung (2009) for simulation of solute transport in natural streams characterized by the transient storage and
release effect. Transition storage zones are divided into two sublayers with distinct mechanisms for solute transport. The upper
layer is an advection dominated transient storage zone involving
relatively strong hyporheic exchange of solute. The lower layer is
an effective diffusion-dominated storage zone that includes deeper
stream bed extending to banks. The two types of transient storage
zones are similar to those found in treatment wetlands (Martinez
and Wise, 2003a) in terms of transient storage effect. The VART
model consists of the following equations (Deng et al., 2010):
C
A
2 C
+ Adif 
C
+U
= Ks 2 + adv
(CS C)
A
TV
t
x
x

2.1. Moment-based method for hydraulic residence time (HRT)

RTD(t) =

and is generally attributed to the exchange of solute between the


actively owing main channel and transient storage zones. Data for
Ci values measured in existing wetlands are often truncated and
thus not available for the late-time or tail portion of wetland RTD
(Jawitz, 2004). As a result, measured RTDs are often exponentially
extrapolated to obtain meaningful moments (Sater and Levenspiel,
1966; Curl and McMillan, 1966; Martinez and Wise, 2003b; Wang
and Jawitz, 2006). The problem with the exponential extrapolation
method is that the tail portion of RTD may not always follow exponential distributions. In fact, RTDs often follow diverse distributions
such as gamma distributions (Kadlec and Wallace, 2008), powerlaw (Wrman et al., 2007) or lognormal distributions (Wrman and
Kronnas, 2005) while exponential distributions may also occur. A
new method that is able to reproduce complete and diverse RTDs
without using the extrapolation method or user-specied functions
is necessary and important. The issue can be addressed by using the
VART model.

(3)

Eq. (3) indicates that the computation of AHRT requires the solute
concentration Ci [M/L3 ] at time ti , as shown in Fig. 1. The long tailed
distribution in Fig. 1 is a typical feature of solute RTDs in wetlands

CS
1
=
(C CS ) TV =
TV
t

Adif = 4DS tS

tS =

Tmin
t

0
t Tmin

for t Tmin
for t Tmin
for t Tmin
for t Tmin

(4a)

(Tmin > 0)

(4b)

(4c)

where C and CS are solute concentrations [mg/L] in main channel


and storage zones, respectively; U is the cross-sectionally averaged ow velocity [m/s] in x direction; t is the time [s]; A is the
cross-sectional ow area of main channel [m2 ] (A = Q/U where
Q = ow discharge [m3 /s]); Aadv and Adif are the areas [m2 ] of
advection-dominated and diffusion-dominated transient storage
zones, respectively; Tmin is the minimum mean residence time [s]
for solute to travel through the advection-dominated storage zone;
tS is the time [s] since the solute release from storage zones to
the main stream; TV is the actual varying residence time of solute
[s]; KS is the longitudinal Fickian dispersion coefcient excluding
the transient storage effect [m2 /s]; DS is the effective diffusion
coefcient [m2 /s] in the transient storage zone;  is the dimensionless parameter representing the water gains ( > 1) or water
losses ( < 1). Water losses may be signicant in some wetlands
(Keefe et al., 2004). A systematic comparison between the VART
model with 4 parameters and other models indicates that the performance of VART model is comparable to other widely used 1D
solute transport models that commonly involve ve or more tting parameters and are limited to the simulation of user-specied
RTDs (Deng and Jung, 2009). Therefore, the relatively simple yet
effective VART model in combination with Eq. (3) is recommended
for determination of actual hydraulic residence time (AHRT) in the
hydraulic design of constructed wetlands.

V. Zahraeifard, Z. Deng / Ecological Engineering 37 (2011) 20872091

2089

Table 1
Redesign scenarios and VART model parameters for TRD wetlands. The redesigned wetlands have two trains and each train has ve cells. The L:W ratios are 5:1 for Cases A
and B and 2:1 for Case C. The existing (Case E) L:W ratios are 3.8 for wetland H1 and 3.9 for wetlands C1 and C2.
Wetland
Hayled 1
Case E
Case A1
Case A2
Case B1
Case B2
Case C
Cobble 1
Case E
Case A1
Case A2
Case B1
Case B2
Case C
Cobble 2
Case E
Case A1
Case A2
Case B1
Case B2
Case C

HLR (cm/day)

L (m)

W (m)

Aadv /A

KS (m2 /s)

Tmin (h)

DS /A (1/s)

AHRT (days)

15
15

228
350

60
36

25

300

24

15

275

50

0.16
0.3
0.16
0.3
0.16
0.3

0.012
0.02
0.012
0.018
0.012
0.022

69
18
69
12
69
22

1.32E07
1.20E07
2.80E07
1.70E07
2.80E07
8.40E08

4.05
4.7
4.4
2.6
3
4.6

25
25

275
340

35
28

30

300

24

25

225

44

0.16
0.3
0.16
0.3
0.16
0.3

0.031
0.023
0.031
0.025
0.031
0.024

52
45
52
35
52
48

6.10E08
1.60E07
1.60E07
1.70E07
1.90E07
1.00E07

4.17
6.3
5.9
5.1
4.7
6.1

15
15

275
350

35
28

25

250

20

15

225

44

0.58
0.3
0.58
0.3
0.58
0.3

0.013
0.011
0.013
0.012
0.013
0.024

24
32
24
17
24
33

3.85E08
1.50E08
2.90E08
2.10E08
4.10E08
1.90E08

3. Applications
In order to demonstrate the new VART model-based approach
to the AHRT, the TRD wetlands were selected for a case study. The
wetlands were selected for three reasons: (1) the wetlands perform well in terms of contaminant and nutrient removal, implying
that the hydraulic residence times and dimensions of the wetlands
were properly designed, (2) tracer injection experiments were conducted in the wetlands, providing useful data for the verication
of the VART model, and (3) other models were proposed by Keefe
et al. (2004) for the TRD wetlands, which could be used to compare
with the VART model in terms of AHRT computation.

(3)

3.1. Description of study wetlands and data gathering


The TRD Wetlands consist of two separate systems of free water
surface wetland: Cobble and Hayeld. The wetland systems are
located in Phoenix, Arizona and they receive nearly 7500 m3 /day
of secondary wastewater from a treatment plant to reduce heavy
metal concentrations before discharging into Salt River. A more
detailed description of the TRD wetland systems and tracer test
data collected from the systems can be found in Keefe et al. (2004).
The data were employed here to demonstrate the efcacy of the
new approach.
3.2. VART model-based wetland design

(5)

The general steps involved in the design of a new wetland can


be summarized as follows:
(1) Selection of hydraulic loading rate (HLR): The HLRs for BOD
and suspended solids can be determined according to US EPA
criteria (USEPA, 1988). There are no HLR standards for heavy
metals. Therefore, different HLRs should be examined, leading
to different design scenarios. Keefe et al. (2004) reported some
reference values for wetland HLRs.
(2) Determination of wetland dimensions: Based on the HLR
selected in step (1), the surface area Aw of a wetland can be
estimated using Eq. (5):
Aw =

Qi
HLR

(Aw = Length Width)

(4)

(5)

(6)

2.7
5.1
5.7
2.7
3.1
4.8

in which Qi is the net inuent discharge of the wetland [m3 /s].


Using different HLRs for the TRD wetlands, the surface area Aw
[m2 ] for the wetlands can be found. Eq. (5) also requires the
selection of L:W (length to width) ratio. Wong and Breen (2002)
and Persson and Wittgren (2003) provided some principles for
selection of L:W ratio in terms of hydraulic efciency and effective volume. Based on the USDA report (2002), water depth
in surface ow wetlands varies typically between 0.15 m and
0.46 m. Once the area Aw and water depth y are found, the crosssectional area of a wetland and consequently ow velocity in
the wetland are known.
Estimation of VART model parameters: The ratio Aadv /A in Eq.
(4) is equal to (1 ) for wetlands. The porosity for wetlands
usually varies in the range of 0.650.85 with the average of 0.75
(USEPA, 2000). The dispersion coefcient KS can be calculated
using the method proposed by Deng et al. (2001). The effective
diffusion coefcient (DS ) may vary in a wide range from positive values of 1.0 105 to 1.0 1010 m2 /s to zero and further
to negative values (Deng et al., 2010). While positive DS values
indicate the expansion of storage zones, negative values represent the contraction of storage zones (Deng et al., 2010). Typical
DS values for wetlands are positive. The parameter Tmin can be
initially estimated as the ratio L/U (where L = wetland length).
Simulation of hydraulic residence time distribution using the
VART model, resulting in the hydraulic residence time distribution shown in Fig. 1.
Calculation of AHRT using Eq. (3) and the HRTD obtained in Step
(4).
The above steps may be repeated for different design scenarios
until the best design, that meets the required removal rate or
low concentration for a contaminant or nutrient at the outlet,
is selected.

4. Results
4.1. Verication of VART model based hydraulic residence time
Fig. 2 shows that the HRTDs simulated using the VART model t
the observed ones well. The same tracer test data were employed
by Keefe et al. (2004) to calibrate the OTIS model. To achieve the
tting similar to Fig. 2, Keefe et al. (2004) added 4 reaction terms

2090

V. Zahraeifard, Z. Deng / Ecological Engineering 37 (2011) 20872091

Fig. 2. Comparison between simulated and observed HRTDs for wetlands (a) H1,
(b) C1, and (c) C2.

involving 8 additional parameters to the OTIS model. The AHRTs


(days) computed using the VART model for wetlands H1, C1, and
C2 are 4.04, 4.66 and 2.65, respectively, while the corresponding
values from the revised OTIS model were 4.05, 4.17, and 2.70.
Obviously, the VART model without any reaction term produces
the AHRTs that are very close to those obtained by Keefe et al.
(2004) using the revised OTIS model with 4 additional reaction
terms. It implies that the VART model is capable of reproducing the
residence time distributions for the TRD wetlands with the least
number of parameters.
4.2. Redesign of TRD wetlands
Tracer data are not available in the design stage. In order to
understand how uncertainties in various parameters affect wetland design results, a number of design scenarios or alternatives

Fig. 3. HRTDs for redesign scenarios for TRD wetlands (a) H1, (b) C1, and (c) C2.

are created for the TRD wetlands by assuming that they are in the
design stage. Then, the HRTDs for each scenario are simulated using
the VART model and the simulation results are shown in Fig. 3
and Table 1. The scenario using the HLRs reported by Keefe et al.
(2004) is dened as Case A (including A1 and A2). The HLRs are then
increased by 20% for wetland C1 and by 67% for wetlands H1 and
C2, forming Case B (including B1 and B2). The VART model parameters are determined using: (1) the typical values and the method
mentioned in Section 3.2, producing Cases A1 and B1, and (2) the
values calibrated using the tracer test data mentioned in Section
4.1, forming Cases A2 and B2. Case C is formed to investigate the
effect of L:W on AHRT by reducing the L:W ratio in Case A1. The
existing design, conrmed with the tracer test data and described
in Section 4.1, is dened as Case E in Table 1 and Fig. 3.
A comparison between Cases A1 and A2 or B1 and B2 indicates
that the uncertainty in VART model parameter values within their
typical ranges may cause a maximum error of 13% in AHRTs, as

V. Zahraeifard, Z. Deng / Ecological Engineering 37 (2011) 20872091

shown in Table 1. It means that the uncertainty in VART model


parameters would not cause signicant errors in calculated AHRTs
as long as parameter values are selected within their typical ranges.
An increase in HLR alone can cause a signicant decrease in AHRT
and increase in the concentration at outlet, as shown in Table 1 and
Fig. 3. Therefore, the HLR should be selected carefully. Anyway,
Cases B1 and B2 should not be selected for TRD wetlands as the
design HLRs are 15 cm/day for wetlands H1 and C2 and 25 cm/day
for C1 (Keefe et al., 2004). As the L:W ratio decreases from 5:1 in
Case A1 to about 2:1 in Case C, the maximum concentration at outlet also decreases. According to Persson and Wittgren (2003), such
change in the L:W ratio reduces effective volume ratio from 80% to
60%. It means that Case C has a lower treatment efciency (60%)
than other cases. Therefore, Case C is excluded from the optimum
design. Then, the optimum design can be selected among Cases A1,
A2, and C.
While all parameter values used in Cases A1, A2, and E for
wetland H1 are in their typical ranges, a comparison between
Figs. 2a and 3a clearly indicates that Case A1 produces a much lower
peak concentration while Case A2 leads to a higher peak concentration than that of Case E. Therefore, Case A2 is excluded from the
optimum design for wetland H1. While existing Case E is good, Case
A1 is even better due to the lower peak concentration. All parameter values used in Cases A1, A2, and E for wetland C1 are also in their
typical ranges. A comparison between Figs. 2b and 3b shows that
Case E produces a lower peak concentration than do Cases A1 and
A2. Therefore, the existing Case E represents the optimum design
for wetland C1. For wetland C2, the parameter Aadv /A = 0.58 used
in Cases A2 and E is too high as compared to its typical value range
of 0.150.35. As a result, the peak concentrations of Cases E and A2
are much higher than that of Case A1, as seen from Figs. 2c and 3c.
The existing wetland behaves more like a stream due to the high
Aadv /A value. Therefore, the optimum design for wetland C2 is Case
A1. It means that the wetland C2 would perform much better if the
design Case A1 was adopted.
5. Conclusions
A new approach to hydraulic design of constructed wetlands
has been proposed. The new approach is characterized by the VART
model-based computation of hydraulic residence time and a 6-step
design procedure for nding the optimum design. The efcacy of
the new approach is conrmed by both tracer test data and the
redesign of the Tres Rios Demonstration Wetlands. The redesign
results show that the existing TRD wetlands H1 and C1 were well
designed even though the performance of wetland H1 could be
further improved if the redesign Case A1 was adopted. The wetland C2 would perform much better if the redesign Case A1 was
adopted.
Acknowledgements
We would like to thank Mr. Robert F. Upham for sharing
the TRD wetland data with us. Support for this research by the
USGS/Louisiana Water Resources Research Institute and LaSPACE
is gratefully acknowledged.

2091

References
Buchberger, S.G., Shaw, G.B., 1995. An approach toward rational design of constructed wetlands for wastewater treatment. Ecol. Eng. 4, 249275.
Conn, R.M., Fiedler, F.R., 2006. Increasing hydraulic residence time in constructed
stormwater treatment wetlands with designed bottom topography. Water Environ. Res. 78, 25142523.
Curl, R.L., McMillan, M.L., 1966. Accuracy in residence time measurements. AIChE J.
12, 819822.
Danckwerts, P.V., 1953. Continuous ow systems: distribution of residence times.
Chem. Eng. Sci. 2, 113.
Deng, Z.Q., Singh, V.P., Bengtsson, L., 2001. Longitudinal dispersion coefcient in
straight rivers. J. Hydraul. Eng. 127 (11), 919927.
Deng, Z.Q., Jung, H.-S., 2009. Variable residence time based model for solute transport in streams. Water Resour. Res. 45, W03415, doi:10.1029/2008WR007000.
Deng, Z.-Q., Jung, H.-S., Ghimire, B., 2010. Effect of channel size on solute
residence time distributions in rivers. Adv. Water Res. 33, 11181127,
doi:10.1016/j.advwatres.2010.06.016.
Jawitz, J.W., 2004. Moments of truncated continuous univariate distributions. Adv.
Water Res. 27, 269281, doi:10.1016/j.advwatres.2003.12.002.
Kadlec, R.H., 1994. Detention and mixing in free water wetlands. Ecol. Eng. 3,
345380.
Kadlec, R.H., Wallace, S., 2008. Treatment Wetlands, 2nd ed. CRC Press/Taylor &
Francis Group, Boca Raton, FL.
Keefe, S.H., 2001. Modeling transport and fate of organic compounds in constructed
wetlands. M.S. Thesis, Univ. of Colo., Boulder.
Keefe, S.H., Barber, L.B., Runkel, R.L., Ryan, J.N., McKnight, D.M., Wass, R.D., 2004. Conservative and reactive solute transport in constructed wetlands. Water Resour.
Res. 40, W01201, doi:10.1029/2003WR002130.
Kruse, N.A.S., Gozzard, E., Jarvis, A.P., 2009. Determination of hydraulic residence
times in several UK mine water treatment systems and their relationship to
iron removal. Mine Water Environ. 28, 115123.
Levenspiel, O., 1972. Chemical Reaction Engineering. John Wiley, Hoboken, NJ.
Martinez, C.J., Wise, W.R., 2003a. Analysis of constructed treatment wetland
hydraulic with the transient storage model OTIS. Ecol. Eng. 20, 211222.
Martinez, C.J., Wise, W.R., 2003b. Hydraulic analysis of Orlando Easterly Wetland.
ASCE J. Environ. Eng. 129, 553560.
Persson, J., Wittgren, H.B., 2003. How hydrological and hydraulic conditions affect
performance of ponds. Ecol. Eng. 21, 259269.
Sater, V.E., Levenspiel, O., 1966. Two-phase ow in packed beds. Ind. Eng. Chem.
Fundam. 5, 8692.
Stairs, D.B., Moore, J.A., 1994. Flow characteristics of constructed wetlands: tracer
studies of the hydraulic regime. In: Proc. of 4th Int. Conf. on Wetland Systems
for Water Pollution Control, Guangzhou, China.
U.S. Department of Agriculture (USDA), 2002. Part 637 Environmental Engineering,
National Engineering Handbook, 210-VI-NEH.
U.S. Environmental Protection Agency (EPA), 1988. Constructed Wetlands and
Aquatic Plant Systems for Municipal Wastewater Treatment, EPA/625/1-88/022.
U.S. Environmental Protection Agency (EPA), 2000. Constructed Wetlands Treatment of Municipal Wastewaters, EPA/625/R-99/010.
Walker, D.J., 1998. Modeling residence time in storm water ponds. Ecol. Eng. 10,
247262.
Walker, D.J., Hurl, S., 2002. The reduction of heavy metal in a storm water wetland.
Ecol. Eng. 18, 407414.
Wang, H.G., Jawitz, J.W., 2006. Hydraulic analysis of cell-network treatment wetlands. J. Hydrol. 330, 721724.
Werner, T.M., Kadlec, R.H., 2000. Wetland residence time distribution modeling.
Ecol. Eng. 15, 7790.
Whitmer, S., Baker, L., Wass, R., 2000. Loss of bromide in a wetland tracer experiment.
J. Environ. Qual. 29, 20432045.
Wong, T.H.F., Breen, P.F., 2002. Recent advances in Australian practice on the use
of constructed wetlands for stormwater treatment. In: Strecker, E.W., Huber,
W.C. (Eds.), Proc. of 9th ASCE Int. Conf. on Global Solutions for Urban Drainage.
Portland, Oregon, September 813, 2002.
Wrman, A., Kronnas, V., 2005. Effect of pond shape and vegetation heterogeneity
on ow and treatment performance of constructed wetlands. J. Hydrol. 301,
125138.
Wrman, A., Packman, A.I., Marklund, L., Harvey, J.W., Stone, S.H., 2007. Fractal
topography and subsurface water ows from uvial bedforms to the continental
shield. Geophys. Res. Lett. 34, L07402, doi:10.1029/2007GL029426.
Yeh, T.Y., 2008. Removal of metals in constructed wetlands: review. Pract.
Periodical of Haz., Toxic, and Radioactive Waste Mgmt. 12, 96101,
doi:10.1061/ASCE.1090.025X.

You might also like