TREATMENT WETLANDS - CHAPTER 14 pptx

539
14
Event-Driven Wetlands
The treatment of stormwaters of various origins is of grow-
ing concern as attempts to rectify point-source pollution
reach maturity. Runoff from urban, agricultural, and indus-
trial sources comprises a sizeable fraction of the total pollu-
tion load to receiving waters in many locations. Often, these
sources are relatively dilute compared to primary or second-
ary domestic sewage but, nevertheless, they may negatively
impact receiving waters.
Urban stormwater wetlands were rst surveyed by
Strecker et al. (1992), who documented the performance of
25 natural and constructed wetlands treating runoff. The
implementation of the technology and the knowledge base
continued to build, resulting in a compilation of data from
76 wetlands worldwide by Wong et al. (1999), and from 49
wetlands in the United States by Carleton et al. (2001). In
North America, these are all FWS systems, and that is the
predominant wetland type elsewhere as well. A typical con-
guration consists of a sedimentation basin as a forebay, fol-
lowed by some combination of marshes and deeper pools.
Design guidelines have now been promulgated by a number
of sources (for example: Schueler, 1992; Breen and Lawrence,
1998; Wong et al., 1999; LEC, 2000; Center for Watershed
Protection, 2001).
Agricultural stormwater occurs as runoff from crops and
pastures. Early work on constructed wetlands for row crop
runoff control was centered in the state of Maine (Higgins et
al., 1993) (Wengrzynek and Terrell, 1990; U.S. Department of
Agriculture, 1991). Runoff from sugarcane is being treated in

the huge FWS wetlands (16,000 ha) called stormwater treat-
ment areas (STAs) of South Florida (Goforth, 2001). However,
there has also been signicant application of constructed wet-
lands at a more modest scale for vegetable farming (Rushton
and Bahk, 2001). Nonetheless, effective control of runoff
may require consideration of the entire watershed (Crumpton,
2000). More recently, attention has been on controlling pas-
ture runoff (Tanner et al., 2003).
Industrial stormwaters have received less attention than
urban and agricultural sources. These typically will con-
tain contaminants specic to the industry in question. For
instance, rain runoff from a fertilizer plant site will poten-
tially contain high levels of nutrients, whereas rain run-
off from a petroleum facility will carry hydrocarbons. For
instance, in South Africa, a number of reed bed wetlands
were established to treat waters generated from truck wash-
ing operations at oil industry depots (Wood, 1993). An HSSF
wetland treated runoff from a 0.8-ha vehicle yard in Surprise,
Arizona, with 54–92% removal of oil and grease (Wass and
Fox, 1993).
Although rainfall runoff is the source of water for many
event-driven wetlands, there are other situations in which
a treatment wetland is subjected to episodic ows. These
include wetland systems that treat stormwater, urban or
agricultural runoff, combined sewer overows, and treat-
ment systems that are operated in batch mode. Whatever the
source, wetlands that receive pulsed ows utilize the suite
of wetland removal mechanisms in different ways than their
continuous-ow counterparts.
14.1 SOURCE CHARACTERIZATION

I
NCOMING FLOWS
The amount of water to be expected from a given water-
shed is variable and is keyed to rainfall in the contribut-
ing basin. The land uses and soils in the contributing basin
are an important modier of the runoff and inltration.
The antecedent dryness in the basin is also a contributing
factor. Detailed methods of estimating runoff are avail-
able, for instance, the SCS (U.S. Soil Conservation Service)
method (McCuen, 1982), which has been summarized by
Novotny (1995). There are also numerous computer models
that account for very detailed features of the contributing
watershed and produce both the quality and quantity of the
runoff to be treated. The focus here is the treatment wetland,
and it will be presumed that the incoming ows will be char-
acterized to the appropriate level of detail.
For purposes of rough estimation, the rational formula
may be used to estimate peak ow, Q
P
(ASCE, 2006):
QCIA
PR b

(14.1)
where
A
C
b
R
watershed area, m

runoff coefficient,


2
dimensionless
peak runoff flow, m /h
a
P
3
Q
I

 vvera
g
e rainfall intensit
y
, m/h
The values of C
R
are a function of land use and storm inten-
sity, as well as the climatological region. An example of the
range of C
R
for an arid region is shown in Table14.1. To
illustrate the regional effect, consider that agricultural land
in Arizona is rated at C
R
 0.1–0.2 for high return-frequency
events, whereas the value for agricultural land in South
Florida is C

R
y 0.5. Water is used consumptively in arid
regions, whereas the high water table and rainfall in a wet
region leads to more runoff. Modications to this simplest
formula have been presented on a site-specic basis (see, for
example, Brezonik and Stadelmann, 2002; ASCE, 2006).
© 2009 by Taylor & Francis Group, LLC
540 Treatment Wetlands
For modeling purposes, the time series of water ows
entering the treatment wetland inlet hydrograph is sometimes
needed. This time series is driven by the pattern of rainfall
associated with a particular event (represented by a hyeto-
graph), together with the collection and transfer character-
istics of the basin. In some instances, such transfer is solely
by gravity, but in some cases pumps may be used. Usually,
the inlet hydrograph contains both a rising and falling limb,
although the rising limb is normally quite steep (for an exam-
ple, see Figure 14.1).
In some instances, the events are separated by interevent
periods of no inow to the treatment wetland. These periods
are important because the wetland will act as a batch reactor
during much of these no-inow durations. A typical sequence
is (1) wetland lling with no outow, (2) ow through with
both inow and outow, (3) draining with no inow, and (4)
nally, a batch-holding mode with neither inow nor outow.
The rates and duration of these inows and outows are in
part controlled by structures. The volume of water in the wet-
land is also in part controlled by structures, but evapotrans-
piration may be an important component during interevent
periods. The durations of no-ow periods are very much a

function of the climatological region in which the system is
located. For instance, in Florida, during the rainy season, the
most frequent periods are measured in hours (4–20 hours,
Wanielista and Yousef, 1991). In contrast, the dry season
of central north island New Zealand has periods of months
with no rain, and interevent periods can be several months
(Tanner et al., 2005b).
Many stormwater wetlands are fed by pumps. These
may range in size from small sump pumps for small local-
ized urban systems to the huge pumps that send water to the
Florida stormwater treatment areas (a.k.a. STAs, or treat-
ment wetlands). Those Florida inow and outow pumps are
among the largest in the world, up to three stories tall, with
capacities up to 10,000,000 m
3
/d (Figure 14.2). A key feature
of pumped systems is intermittent feed to the treatment wet-
land, usually at xed but incremental rates, corresponding to
the number of pumps that are operated, for periods of time
dictated by conditions in the contributing watershed.
There is typically a hydraulic limitation to the event size
that may be treated in a given wetland. It is generally not
feasible to size a wetland to treat the 100-year return fre-
quency storm because of cost and footprint considerations
and because of the need for maintenance water for the (large)
wetland under normal conditions. Consequently, part of the
design decision process is the determination of the maximum
design storm to be treated. Even if the system is sized to treat
TABLE 14.1
Runoff Coefficient (C) for Use in the Rational Method in Maricopa County, Arizona

Return Period
Land Use 2–10-year 25-year 50-year 100-year
Streets and Roads
Paved roads 0.75–0.85 0.83–0.94 0.9–0.95 0.94–0.95
Gravel roadways and shoulders 0.6–0.7 0.66–0.77 0.72–0.84 0.75–0.88
I
ndustrial
Areas
Heavy 0.7–0.8 0.77–0.88 0.84–0.95 0.88–0.95
Light 0.6–0.7 0.66–0.77 0.72–0.84 0.75–0.88
B
usiness Areas
Downtown 0.75–0.85 0.83–0.94 0.9–0.95 0.94–0.95
Neighborhood 0.55–0.65 0.61–0.72 0.66–0.78 0.69–0.81
R
esidential Areas
Lawns—at 0.1–0.25 0.11–0.28 0.1–0.3 0.13–0.31
Lawns—steep 0.25–0.4 0.28–0.44 0.3–0.48 0.31–0.50
Suburban 0.3–0.4 0.33–0.44 0.36–0.48 0.38–0.5
Single family 0.45–0.55 0.5–0.61 0.54–0.66 0.56–0.69
Multi-unit 0.5–0.6 0.55–0.66 0.6–0.72 0.63–0.75
Apartments 0.6–0.7 0.66–0.77 0.72–0.84 0.75–0.88
Parks/cemeteries 0.1–0.25 0.11–0.28 0.12–0.3 0.13–0.31
Playgrounds 0.4–0.5 0.44–0.55 0.48–0.6 0.5–0.63
Agricultural areas 0.1–0.2 0.11–0.22 0.12–0.24 0.13–0.25
Bare ground 0.2–0.3 0.22–0.33 0.24–0.36 0.25–0.38
Undeveloped desert 0.3–0.4 0.33–0.44 0.36–0.48 0.38–0.5
Mountain terrain (slopes  10%)
0.6–0.8 0.66–0.88 0.72–0.95 0.75–0.95
Source: Adapted from the Drainage Design Manual for Maricopa County, Volume 1.

© 2009 by Taylor & Francis Group, LLC
Event-Driven Wetlands 541
a modest-sized storm ow, there remains the possibility that
dry conditions might persist in some years to the point that
the integrity of the wetland ecosystem is jeopardized. If such
detrimental dryout conditions are expected, then a source
of irrigation water for ecosystem maintenance should be
identied.
INCOMING CONCENTRATIONS AND LOADS
Concentrations of most parameters in stormwater are time
dependent, as are the ows. Stormwater concentrations and
loads are episodic due to periods of dryfall and deposition,
followed by the rst ush of runoff after rain, followed by
exponential decreases in runoff constituent concentrations as
storages rinse from the landscape, and nally, dry conditions
and deposition until the next storm event. The time series
of concentrations in the inow to the wetland is called the
chemograph. An example chemograph for an agricultural
runoff wetland, targeting nitrogen reduction, is shown in
Figure 14.3.
In some watersheds, the chemograph is not synchronized
with the hydrograph, but instead provides higher concentra-
tions early in the inow event. This phenomenon is termed
rst-ush behavior, referring to the surge of pollutants con-
tained in the rst water to leave the contributing basin. For
instance, Wanielista and Yousef (1991) report that, in Flor-
ida, the rst 25 mm of runoff from urban systems typically
carries 90% of the pollution. However, for the large agricultural
0.0
0.5

1.0
1.5
2.0
2.5
3.0
0 50 100 150 200
Days from May 1, 2002
Flow (L/s) or Rain (cm)
Inflow
Rainfall
FIGURE 14.1 Time series of ows entering a treatment wetland from an improved pasture in Toenepi, New Zealand. Several rain events
occurred during this winter wet season, as indicated by the repeated spikes in inow. The runoff coefcients were 0.23–0.30 and were the
result of tile drains. (Adapted from Tanner et al. (2005b) Agriculture, Ecosystems and Environment 105(1–2): 145–162.)
FIGURE 14.2 The outow pump station from STA1E (Stormwater Treatment Area 1 East) of the Everglades Protection Project. The capac-
ity of this pump station is 9,740,000 m
3
/d, and it serves to drain a 2,700-ha FWS wetland.
© 2009 by Taylor & Francis Group, LLC
542 Treatment Wetlands
watershed of South Florida, there is no such rst-ush effect.
Despite the site-specic nature of the chemograph, there is
no reliable database documenting dynamic time series of
concentrations in any given watershed. Of necessity, average
concentrations of some sort must be used, of which the ow-
weighted concentration is most useful. This may be evalu-
ated on a long-term basis or as an event mean concentration
for the inlet water:
C
QCdt
Qdt

e

¯
¯
(14.2)
where
C
C


instantaneous concentration, mg/L
event
e
mean concentration, mg/L
instantaneous fQ  llow, m /h
time, h
3
t 
and where integration is over the period of one event. The
numerator is the mass of pollutant entering or leaving during
the event. As described in Chapter 6, the performance of the
wetland may be described in terms of the loads applied to
and emanating from the system.
Tables 14.2 and 14.3 provide long-term mean concen-
trations for constituents in urban stormwater. The averages
are ow-weighted to provide realistic estimates of the total
constituent load that escapes during multiple storm events.
Instantaneous concentrations may rise considerably higher
than these averages. Pollutant concentrations and loads gen-
erally range from low levels from undeveloped and park

lands to low-density residential and commercial, to agri-
cultural, to higher-density residential and commercial, and
nally to high-density commercial, industrial, and agricul-
tural land uses. Mean concentrations per event for BOD
5
vary from below detection for undeveloped lands to 20 mg/L
for high-density urban areas. Total suspended solids concen-
trations vary from about 10 mg/L for undeveloped areas up
to 150 mg/L for high-density urban areas. Typical concentra-
tions for other stormwater pollutants are also summarized in
Tables 14.2 and 14.3.
The mass loading rates represent normalized pollut-
ant loads that are somewhat independent of local rainfall
amounts. Because pollutant loads per area per time are rela-
tively constant between similar land use areas, variable local
rainfall washes these loads off the land in a few large events
or over many smaller events. Urban pollutant loads increase
with the imperviousness of the watershed. Although 20 to
40% of the material on street surfaces is organic, it does not
biodegrade easily because it comes from leaf and wood litter,
rubber, and road-surface material (Novotny, 1992). The high
metal content of highway solids comes from vehicle emis-
sion. Novotny (1992) reported that the average total nitro-
gen load from urban lands is 5 kg/ha·yr (1 to 38.5 kg/ha·yr),
and the total phosphorus load averages 1 kg/ha·yr (0.5 to
6.25 kg/ha·yr).
Constructed wetlands are being increasingly used to
treat runoff from intensive animal operations (CH2M Hill
and Payne Engineering, 1997; Tanner et al., 2003) and from
crops (U.S. Department of Agriculture, 1991; Crumpton,

2000). Concentrations and loads from agricultural land uses
vary considerably. Flows and loads are typically highest from
areas with high animal densities or high fertilization rates.
Runoff pollutant concentrations from animal feedlots can
be extremely high unless runoff is collected and treated.
Pollutants from feedlot runoff typically include high levels
of organic and inorganic solids and associated nutrients.
Nutrient concentrations and loads from row crops and pas-
tures depend on fertilization practices and type of soil.
There exist many computer models for the amounts of
contaminants that may be expected in runoff from different
0
5
10
15
20
25
30
35
40
45
–5–4–3–2–10123456789
Time (days)
NO
x
-N (mg/L)
FIGURE 14.3 The time series of nitrate nitrogen arriving at an agricultural runoff treatment wetland in McDowell County, North Carolina.
The rain event started at time zero and drove runoff that entered the wetland by stream ow. (Adapted from Kao and Wu (2001) Water Sci-
ence and Technology 45(3): 169–174.)
© 2009 by Taylor & Francis Group, LLC

Event-Driven Wetlands 543
landscapes. Novotny (1995) summarized the characteristics
of six models for urban watersheds, and seven for agricul-
tural watersheds, and the number has grown since that time.
A large part of the design of event-driven treatment wetlands
is the determination of the ows and loads to be treated.
TABLE 14.2
Pollutant Concentrations for Source Areas for Stormwaters
Constituent
TSS
a
(mg/L)
TP
b
(mg/L)
TN
c
(mg/L)
E coli
a
(1,000 #/mL)
Cu
a
(Kg/L)
Pb
a
(Kg/L)
Zn
a
(Kg/L)

Residential roof 19 0.11 1.5 0.26 20 21 312
Commercial roof 9 0.14 2.1 1.1 7 17 256
Industrial roof 17 — — 5.8 62 43 1390
Comm./res. parking 27 0.15 1.9 1.8 51 28 139
Industrial parking 228 — — 2.7 34 85 224
Residential street 172 0.55 1.4 37 25 51 173
Commercial street 468 — — 12 73 170 450
Rural highway 51 — 22 — 22 80 80
Urban highway 142 0.32 3 — 54 400 329
Lawns 602 2.1 9.1 24 17 17 50
Landscaping 37 — — 94 94 29 263
Driveway 173 0.56 2.1 17 17 — 107
Gas station 31 — — — 88 80 290
Auto recycler 335 — — — 103 182 520
Heavy industrial 124 — — — 148 290 1,600
Source: Data from Center for Watershed Protection (2001) New York State Stormwater Management Design Manual. Report by the Center for Watershed
Protection (CWP) for the New York State Department of Environmental Conservation, Albany, New York.
a
Data from Claytor and Schueler (1996) Design of Stormwater Filtering Systems. Center for Watershed Protection: Ellicott City, Maryland.
b
Average of data from Steuer et al. (1997) Sources of contamination in an urban basin in Marquette, Michigan, and an analysis of concentration, loads, and data
quality. Water Resources Investigation Report 97–4242, U.S. Geological Survey; Bannerman et al. (1993) Water Science and Technology 28(35): 241–259;
Waschbusch (2000) Sources of phosphorus in stormwater and street dirt from two urban residential basins in Madison, Wisconsin, 1994–1995. Proceedings of the
National Conference on Tools for Urban Water Resource Management and Protection; U.S. Environmental Protection Agency: Washington, D.C, pp. 15–55.
c
Data from Steuer et al. (1997) Sources of contamination in an urban basin in Marquette, Michigan, and an analysis of concentration, loads, and data quality.
Water Resources Investigation Report 97–4242, U.S. Geological Survey.
TABLE 14.3
Typical Concentration Data for Pollutants in Urban Stormwater
Constituent Units Urban Runoff Source

TSS mg/L 54.5 Smullen and Cave (1998)
TP mg/L 0.26 Smullen and Cave (1998)
TN mg/L 2.00 Smullen and Cave (1998)
Cu
Mg/L
11.1 Smullen and Cave (1998)
Pb
Mg/L
50.7 Smullen and Cave (1998)
Zn
Mg/L
129 Smullen and Cave (1998)
Fecal coliforms 1,000 CFU/mL 1.5 Schueler (1999)
Source: Data from Center for Watershed Protection (2001) New York state stormwater management design manual. Report by
the Center for Watershed Protection (CWP) for the New York State Department of Environmental Conservation, Albany, New
York; pooled NURP/USGS Smullen and Cave (1998) Updating the U.S. nationwide urban runoff quality database. 3rd Interna-
tional Conference on Diffuse Pollution, 31 August–4 September 1998; and Schueler (1999) Watershed Protection
Techniques 3(1): 551–596.
HYDROLOGY OF PULSED AND SEASONAL SYSTEMS
Event-driven wetlands are dynamic in all respects, and the
principal underlying hydraulics exhibit variable water depths
and ows. The behavior is strongly conditioned by the nature
© 2009 by Taylor & Francis Group, LLC
544 Treatment Wetlands
of inow and outow structures that may be designed to
improve detention and treatment (Somes and Wong, 1997).
The most general situation may have several complicating
factors, but here a simple and common case is explored for
purposes of illustration. It will be presumed that the wetland
is relatively small and consequently behaves as a level-pool

system, with no gradients in stage. Stormwater wetlands
experience event ows and concentrations, followed by peri-
ods of batch operation. It is then necessary to account for the
dynamics of water storage within the wetland.
The dynamic, level-pool water mass balance for the wet-
land for unsteady state inows and meteorology is
dAh
dt
QQ APETI
()
()  
io
(14.3)
where
A
Q
Q



wetland wetted area, m
inflow, m /d
2
i
3
o
ooutflow, m /d
rainfall, m/d
evapotransp
3

P
ET

 iiration, m/d
water depth, m
infiltration
h
I

 ,, m/d
time, dt 
The stage is often controlled by an outlet structure, here sup-
posed to be a rectangular weir. Thus the outow-stage rela-
tion would be given by the Francis formula (French, 1985):
Qh
oow
B
A
() (14.4)
where
hhHH
H
ow ow w
height over the weir, m ( )
wet

 lland stage, m
elevation of weir, m
out
w

o
H
Q

 fflow, m /d
calibration exponent, dimensio
3
A nnless
calibration constant, (m /d)/m
3
B
A

It is noted that the wetted area is not constant but changes with
stage according to the bathymetry of the wetland, represented
by the stage–area–volume relationship (see Chapter 2). Fur-
ther, the wetland catches rainfall over an area typically greater
than the wetted area but loses water to ET from just the wetted
area. Inltration further contributes to water losses and may
be a signicant fraction of the output from the wetland, as in
the case of the Hidden River wetland in Tampa, Florida (Carr
and Rushton, 1995).
An example of this analysis is illustrated in Figure 14.4.
For illustration, a wetland of 2,500 m
2
is subjected to a
5-cm rain event and the associated runoff from the contrib-
uting basin. The presumed conditions and wetland design
are shown in Table 14.4. The wetland does not have a level
bottom; rather, there is a presumed quadratic stage–vol-

ume relation, which implies that the wetted area increases
linearly with stage. The hypothetical sequence of events is
as follows:
The wetland starting condition is partially full,
with the water level 10 cm below the outlet weir.
A steady rain begins at time zero and persists for
one day.
Runoff from the watershed begins after t  0.5
days and persists for 1.5 days.
The wetland loses some water to ET (1.0 cm/d),
and some to inltration (1.0 cm/d).
For the rst half day (t  0–0.5 days), the wetland
lls due to direct collection of rain.
At t  0.65 days, the wetland has lled to the top of
the outow weir, and outow commences.
For the next half day (t  0.5–1.0 day), the wetland
lls because of direct rain and incoming runoff.
For the next day (t  1.0–2.0 days), the wetland lls
because of incoming runoff.
During t  2.0–3.1 days, there is no inow, and the
wetland is draining over the outow weir. At t 
3.1 days, the level has decreased to the top of the
weir, and outow stops.
After time t  3.1 days, the wetland loses water to
ET and inltration, and the level continues a slow
drop.
•
•
•
•

•
•
•
•
•
•
TABLE 14.4
Parameters for a Hypothetical Stormwater
Wetland Event Scenario
Basin
Area 
120,000 m
2
Runoff coefcient 
0.5
Wetland
WWAR 
2 %
Nominal maximum depth 
1.00 m
Wetland area full 
2,500 m
2
Wetland volume full 
1,250 m
3
Stage area: A  aH, a 
2,500 m
2
/m

Weir height 
0.90 m
Weir coefcient 
31,623 (m
3
/d)/(m
1.5
)
ET rate 
1.00 cm/d
Inltration rate 
1.00 cm/d
Rain catchment 
3,000 m
2
Rainfall
Period 
1.00 d
Rain rate 
5.00 cm/d
Basin catch 
6,000 m
3
Inow start 
0.50 d
Inow end 
2.00 d
Inow rate 
2,000 m
3

/d
Runoff 
3,000 m
3
Storm/wetland volume ratio 
2.40
© 2009 by Taylor & Francis Group, LLC
Event-Driven Wetlands 545
Because of the assumed bathymetry, the wetted area rst
grows and then shrinks through this course of events.
This simple model has been found to t event wetland
data quite well (Kadlec, 1994; Hey et al., 1994; Kadlec,
2001a). However, there are several different inlet and outlet
structures that may be used, and the level-pool model must
be modied accordingly for other situations.
FLOW AND CAPTURE
Event-driven wetlands operate with periods of ow through
and periods of water-holding or batch processing. Any water
that does not escape the wetland during a particular event will
be held until the next event, and possibly longer. Therefore,
it is subject to the water quality improvement functions of
the wetland for not only the event duration but also the
interevent period. Conversely, water that enters and leaves
the wetland during the event is subject to treatment only dur-
ing the (possibly brief) period of detention during the event.
Therefore, at the rst level of consideration, it would seem
desirable to contain the entire volume of water associated
with an event within the wetland. This requirement is often
referred back to the contributing watershed in terms of the
rain amount and runoff coefcient that generated the storm

volume. The wetland may be sized to nominally contain
the runoff volume associated with a rain event of specied
amount or return frequency.
One limit on performance is the expulsion of water that
resides in the wetland at the time of event initiation. This
water is at a background concentration if the events are widely
spaced. In this example, the time for background to be reached
was on the order of two or three days. The length of time for
the wetland water surcharge to dissipate, and thus for stage to
reach the weir elevation, was also of the order of two or three
days. If the events are more closely spaced, then the resident
antecedent water will not be at background but rather will
exhibit a concentration representing only partial treatment
of the previous event. A second limit on performance is the
0
500
1,000
1,500
2,000
2,500
3,000
0.0 1.0 2.0 3.0 4.0 5.0
Time (days)
Flow (m
3
/d) or Wetted Area (m
2
)
Inflow
Outflow

Area
(a)
FIGURE 14.4 Response of a hypothetical stormwater wetland to a one-day steady rain. Runoff into the wetland begins after half a day.
The wetland lls, and outow persists for just over three day. Inltration and ET deplete the water after the event. See Table 14.4 for
parameters.
0.7
0.8
0.9
1.0
1.1
0.0 1.0 2.0 3.0 4.0 5.0
Time (days)
Wetland Water Stage (m)
Rain
Inflow
Outflow
(b)
© 2009 by Taylor & Francis Group, LLC
546 Treatment Wetlands
ultimate removal that would be associated with a continuous
water input at the event average ow rate. This is likely to
be a relatively low percentage reduction because of the high
ow in the event.
A complicating factor in analysis is that wetlands are not
hydraulically simple and do not operate on a basis of plug-
ow displacement. Therefore, the fraction of incoming water
that remains in the system for a particular event is not deter-
mined merely from displacement. Depending on the congu-
ration of the wetland, some of the very rst water that enters
the wetland may nd its way to the exit far in advance of

nominal retention time during the event. This is partly due to
preferential ow paths, and partly due to the positioning of
water-level control structures.
If it is assumed that the wetland behaves hydraulically
as several mixed tanks in series, it is reasonable to use the
known detention time distribution to compute how much of
the incoming water is still in the wetland after a period of
ow of known magnitude. Figure 14.5 shows the results as a
function of the nominal number of wetland volumes of water
that have passed through the system. If the ow were plug
ow, then the retention is 100% until the nominal detention
time is reached, after which the new event water begins to
exit. For less ideal ow conditions, the amount of the new
event water that is held is less.
The presumption of Figure 14.5 is that all of the wetland
water is involved in ow. That can be far from true because of
wetland design factors (see Appendix B). For instance, if there
is a point inlet to the system, and a point outlet aligned directly
opposite, then the incoming pulse of water can travel directly
to the outlet, never reaching zones to the side of the most direct
ow path. That renders the corner areas essentially out of the
ow path and reduces the volumetric efciency quite mark-
edly (Agunwamba, 2006). Walker (1998) investigated this sit-
uation using computational uid mechanics (simulations) of
unvegetated FWS basins. He determined that aspect ratio
(L:W) was the major determinant of such shortcircuiting and
calculated a retention-displacement chart (Figure 14.6).
For individual, separated inow events, the concept of
retained volume provides a strong correlating factor for per-
formance. If the event is small, it is wholly contained and

held until the next displacing ow. Treatment proceeds in the
batch mode for the interevent duration. If the event is large,
then it is processed by ow through at the detention time of
the event ow. Because such ows are typically large, the
wetland will have a short detention time, which results in
decreased treatment performance. Therefore, mass reduc-
tions decrease (exponentially) as the number of nominal dis-
placements caused by the event increases.
The storage and treatment potential for individual events
may ultimately be linked to the sequence of events that
are likely to occur over a long period of record. The three
required probability distributions are for the event duration,
the event intensity, and the interevent duration (Wong and
Somes, 1995). To further complicate matters, when the wet-
land basin is “full,” water will be diverted away (bypassed).
Thus, a wetland that drains quickly, i.e., one with short deten-
tion time, will be in a position to detain more water upon
relling than a wetland that drains slowly. Wong and Geiger
(1997) examined the runoff patterns for Melbourne, Austra-
lia, and concluded that the hydrologic effectiveness (the per-
centage of runoff that is exposed to treatment) was inversely
proportional to the wetland detention time.
These factors all point to the existence of a set of perfor-
mance determinants that go beyond those for continuous-ow
treatment wetland systems. Batch interevent rate coefcients
and background concentrations are important. The inter-
nal ow patterns are also important as determinants of the
concentration and timing of the ushed antecedent water.
The dynamics of water ow and storage of the wetland are
important because they determine the volumetric constraints

on performance.
0.0
0.2
0.4
0.6
0.8
1.0
0123456
Size of Event (Wetland Volumes)
Inflow Fraction Retained
Plug Flow
N = 10
N = 3
N = 1
FIGURE 14.5 Fraction of the inow retained for a range of numbers of tanks in series for event volumes of different nominal displacements.
© 2009 by Taylor & Francis Group, LLC
Event-Driven Wetlands 547
Ultimately, the computed description of performance will
either be expressed as long-term mass removal or by dynamic
modeling that produces time series of outow and efuent
concentrations in response to the inlet time series. Most of
the stormwater wetland literature reports mass removals over
some period of record. However, there are now rst-genera-
tion dynamic models for some applications. For instance,
there is a ow and phosphorus model for application in south
Florida called the Dynamic Model for Stormwater Treatment
Areas (DMSTA) (Walker and Kadlec, 2005).
14.2 TECHNOLOGY STATUS
Data reporting falls into two general categories: event time-
series analyses, and summaries of removal efciencies. Two

concepts are needed to organize these bodies of information:
event mean concentrations and mass balancing. Because the
concentration pulses and ow pulses are often out of sync,
the event mean concentration (EMC) is used:
EMC
VC
V

3
3
()
(14.5)
where
C  concentration in a water parcel, g/m = m
3
gg/L
water volume in the parcel, m
3
V 
In effect, the EMC is the mass average concentration over
the course of an event and may be calculated for both inlet
and outlet ows of various types for the wetland. If only the
directed inow and outow are considered, the EMC reduc-
tion is dened as
Concentration Reduction
io
i

()EMC EMC
EMC

(14.6)
The outow from the wetland corresponding to a given
inow event may total less than the inow due to inltration
and evapotranspiration. The mass load reduction for a given
pollutant is
Load Reduction
iioo
ii

()V EMC V EMC
VEMC
(14.7)
URBAN STORMWATER
Runoff from roofs, lawns, parking lots, and other urban land-
scape features often contains several classes of pollutants
(see Table 14.2). There will be small amounts of nitrogen
and phosphorus and biochemical oxygen demand (BOD), but
possibly relatively greater amounts of total suspended solids
(TSS), metals, and perhaps pesticides. An important subset
of urban runoff control is focused on highway runoff (Shutes
et al., 2001; Bulc and Sajn Slak, 2003; Pontier et al., 2003).
However, the greater number of systems are used to treat
rainwater runoff from more generalized urban areas. Here the
focus is upon gravity-driven constructed systems, as opposed
to pumped or natural systems. Such systems are often quite
small, serving small catchments (Figure 14.7).
In this section, attention is on gravity-fed, constructed
systems treating urban stormwater. Pumped systems have
been included in other summaries, such as Carleton et al.
(2001) and Wong et al. (1999), but many of these do not

reect the random and episodic nature of wetlands receiv-
ing rain-driven events and are omitted here, as are natural
wetland systems. A sampling of TSS performance is given in
Table 14.5, in which it is clear that there is a wide spectrum
of performances for all common constituents, with median
removals far from 100%.
Removal percentages are insufcient for design pur-
poses, and hence it is necessary to nd relationships between
system characteristics and performance measures. Carleton
et al. (2001) concluded that long-term pollutant removals
could be described in terms of the same kinds of rst-order,
steady-ow design equations currently employed for waste-
water treatment wetlands. They present rate coefcients that
0.0
0.2
0.4
0.6
0.8
1.0
1.2
0.0 1.0 2.0 3.0 4.0
Size of Event (Wetland Volumes)
Fraction of Inflow Retained
Plug Flow
L:W = 8
L:W = 4
L:W = 2
L:W = 1
L:W = 0.5
FIGURE 14.6 Fraction of an inow event contained in a wetland as a function of the event volume and the aspect ratio (L:W). (From Walker

(1998) Ecological Engineering 10(3): 247–262. Reprinted with permission.)
© 2009 by Taylor & Francis Group, LLC
548 Treatment Wetlands
are intended to be used in a procedure such as that described
by Wong and Geiger (1997) and Wong et al. (2006). As dis-
cussed in Chapter 6, that method has inherent drawbacks,
and calibrations for event-driven systems provide only cen-
tral tendencies. Additionally, rate coefcients derived from
single events may not be adequate for a time series of meteo-
rologically driven ows.
Data sets analyzed by Strecker et al. (1992) and extended
by Carleton et al. (2001) include about half natural wetland
systems that receive stormwater. Strecker et al. (1992) in fact
found that natural wetlands performed somewhat better than
constructed wetlands, but the use of natural wetlands is not
encouraged because of real and perceived negative impacts
(U.S. EPA, 1993e). The median areal rate coefcients reported
FIGURE 14.7 (A color version of this gure follows page 550) An urban stormwater wetland in the city of Blue Mountains, Australia.
TABLE 14.5
Suspended Solids Reduction in Constructed Urban Runoff Treatment Wetlands
Name Location Reference WWAR Area Ratio
(%)
HLR
(cm/d)
Reduction
(%)
Crookes Australia Raisin et al. (1997) 0.1 21.83 12
Mays Chapel Maryland Carleton et al. (2001) 0.6 5.55 11
Shop Creek Colorado Carleton et al. (2001) 0.6 — 25
Franklin Farms Virginia Carleton et al. (2001) 0.8 17.16 62

Lake Munson Florida Maristany and Bartel (1989) 1.1 5.19 93
Slovenia Highway Slovenia Bulc and Sajn Slak (2003) 1.1 — 74
DUST Marsh California Meiorin (1989) 1.8 — 64
Crestwood Virginia Carleton et al. (2000) 2.4 3.69 58
Greenwood Florida McCann and Olson (1994) 2.5 2.57 68
Queen Anne Maryland Carleton et al. (2001) 3.8 — 65
Clear Lake Minnesota Carleton et al. (2001) 4.9 1.71 76
Tampa Ofce Pond Florida Carleton et al. (2001) 5.1 8.16 55
West Lafayette Indiana Harbor et al. (2000) 6.3 — 75
Lake McCarrons Minnesota Carleton et al. (2001) 6.6 7.38 83
Hidden River Florida Carr and Rushton (1995) 19.5 1.04 86
Elbow Valley Calgary Amell (2004) — — 72
Hallam Valley Low Australia Wong et al. (2006) — 400 94
Hallam Valley High Australia Wong et al. (2006) — 220 94
Kaohsiung China Kao et al. (2001a) — 7.10 37
Villanova Pennsylvania Rea (2004) — 8.16 70
Median 7.10 68
Note: All are FWS except for Slovenia, which is HSSF.
© 2009 by Taylor & Francis Group, LLC
Event-Driven Wetlands 549
by Carleton et al. (2001) for FWS constructed gravity-
fed systems (N  9) are
Total Phosphorus: 8.3 m/yr
Ammonia: 5.0 m/yr
Nitrate: 6.7 m/yr
These are very low compared to the k-values reported for
continuous-ow systems in the preceding chapters. Individ-
ual events are afforded much better treatment.
Carleton et al. (2001) also suggest that wetland-to-water-
shed area ratio (WWAR) may be used as a predictor of per-

formance. Their regression equations are developed for small
data sets, including natural wetland systems, and have low
correlation coefcients (R
2
y 0.15–0.45). Within the range 0 
WWAR  0.1, the data scatter was particularly large. Regres-
sions of the data in Table 14.5 are unsatisfactory; in general,
the fractions remaining increase slightly with WWAR.
Duncan (1998; as referenced and reported by Wong et al.,
1999) suggested the use of regression equations for purposes
of sizing. Relations are based on analysis of subsets of 76
stormwater wetland systems, presumably including pumped
systems, inclusive of those in the United States and comple-
mented by systems in Australia. The proposed empirically
derived relations are
TSS Percent Remaining R
i
2


78 0 80
033 049
qC

.
(14.8)
TP Percent Remaining R
2
12 0 71
044

q
.
.(14.9)
TN Percent Remaining R
2
14 0 78
043
q
.
.

(14.10)
where
C
q
i
3
inlet concentration, g/m
hydraulic loa

 dding rate, m/yr
The range of loading rate data for this regression is approxi-
mately 2  q  200 m/yr, and the percents remaining ranged
from 10 to 100%; about 30 systems were involved in each
regression.
AGRICULTURAL STORMWATER
The acquisition of knowledge about wetland treatment of
agricultural runoff has often been spurred by regional water
quality problems as well as by the slow growth of experiences
on small-scale systems. Nutrients from agriculture have been

implicated in eutrophication of marine environments (nitro-
gen) and freshwater (phosphorus) environments. Agricul-
tural runoff is episodic by nature, and there is an extensive
literature on predicting the amounts of water and particulate
matter that leaves the elds. However, fertilization practices
dictate the availability of nutrients to be washed off the elds,
and hence the patterns of loading are different from those in
urban or industrial settings. Non-point source (NPS) pollu-
tion from agriculture may occur when nutrients are applied at
•
•
•
rates greater than crops can utilize or when timing of nutrient
applications occurs in close proximity to heavy rains (Stone
et al., 2003).
A fraction of the fertilizers applied to elds is unavoidably
lost to runoff and shallow groundwater. Tile drainage systems
collect subsurface waters and vent them to receiving streams,
where their nitrogen content, together with phosphorus, may
cause problems. Wetlands are being used at various points in
the agricultural landscape, corresponding to drainage areas
ranging from individual elds (U.S. Department of Agriculture,
1991; Tanner et al., 2003), to small-order streams (Stone
et al., 2003), to large regional landscape units of thousands
of hectares (Reilly et al., 2000). Some of these receive
pumped water at relatively constant rates, and these have
been included under the discussion of continuous ow (but
possibly variable ow) wetlands. Those that receive water as
a result of meteorological events are considered here.
Ta

r
get: Nitrogen
Nitrogen compounds are among the principal pollutants of
concern in fresh and marine waters because of their role in
eutrophication, their effect on the oxygen content of receiv-
ing waters, and their potential toxicity to aquatic invertebrate
and vertebrate species. For example, the nitrogen content
of the streams and rivers of the midwestern United States
is of particular importance at this point in history because
of hypoxia in the Gulf of Mexico, together with the asso-
ciated ecological and economic consequences (Diaz and
Solow, 1999). Both point and nonpoint sources contribute to
the nitrogen content of waters within the Mississippi River
drainage basin. About 60% of the waterborne total nitrogen
is in the form of nitrate (Goolsby and Battaglin, 2000).
The source of the nitrogen is about two thirds from agri-
culture and one third from other sources, including urban
runoff, atmospheric deposition, and point sources. The result
is a surface-water nitrate concentration, in the upper portions
of the basin, of about 4 mgN/L.
Federal and state ofcials have agreed on an action plan
that includes a component intended to promote restoration
and enhancement of natural systems for nitrogen retention
and denitrication (U.S. EPA, 2001a). The premier natural
system that has the capability to effectively remove nitrate
from surface water is the free water surface wetland, based
simply on its ability to place contaminated water in intimate
contact with the biogeochemical cycle that removes nitrogen.
More than half of the presettlement wetlands in the upper
Mississippi River basin have been lost to drainage (Dahl,

1990). It is therefore synergistic to restore wetlands that are
positioned to effectively function for nitrate reduction. Con-
siderable attention has therefore been devoted to constructed
and restored wetlands for nitrogen control in the midwest-
ern United States (Kovacic et al., 2000; Hey, 2002; Mitsch
et al., 2005; Hey et al., 2005; Kadlec, 2005b). Similarly, the
Chesapeake Bay in the eastern United States is impacted by
agricultural runoff from farms in Maryland (Whigham et
al., 1999; Jordan et al., 1999). Studies on seven constructed
© 2009 by Taylor & Francis Group, LLC
550 Treatment Wetlands
(restored) wetlands demonstrated that nutrients could be
removed from eld runoff. Wetland systems have also been
implemented for runoff control from other row crops, such as
potatoes (Wengrzynek and Terrell, 1990; U.S. Department of
Agriculture, 1991; Higgins et al., 1993).
The problem of excessive nitrogen runoff has also been
identied as detrimental to the Baltic Sea. The role of wet-
lands in southern Sweden has been thoroughly explored.
Data were acquired on eight constructed wetlands and used
to extrapolate the effects of 40 wetlands of total area 92 ha
on the runoff from a 22,400-ha catchment (Arheimer and
Wittgren, 1994; 2002). In Finland, retention performance of
constructed wetlands has been studied at four agricultural
study sites (Puustinen and Koskiaho, 2003). In Norway,
eld studies at ve sites were conducted over several years,
focusing on nutrients and sediments (Braskerud et al., 2000;
Braskerud, 2001b; 2002a; 2002b; 2003).
Rural settings often contain pastures, which form a
source of nutrients and TSS to the drainage waters. FWS

wetlands are a convenient method of intercepting some of
the nutrient loads, and have been implemented in Australia
(Raisin et al., 1997) and New Zealand (Tanner et al., 2003;
Tanner et al., 2005b). Typical on-farm systems are not large
(Figure 14.8).
Target: Phosphorus
Although phosphorus has also been studied in the various
small-scale studies discussed previously, much of the research
on agricultural phosphorus removal in treatment wetlands has
originated from the Everglades protection projects of South
Florida. The receiving Everglades ecosystem is conditioned
by very low total phosphorus, in the range of 6–12 µg/L.
Even modest amounts of eld runoff phosphorus pose a threat
to that ecosystem (Davis, 1994). Tests were performed on
dozens of mesocosm tanks, thirty 0.2-ha test cells, and four
2-ha eld scale wetlands. In parallel, a 1,500-ha, ve-cell
wetland, the Everglades Nutrient Removal Project (ENRP)
was operated as a research device. These FWS systems vari-
ously contained emergent vegetation, submerged plants, or
algae, and testing was conducted over the period 1994–2005.
A summary of results is given by Walker and Kadlec (2005).
Details of ENRP results were presented in a special issue of
Ecological Engineering (Vol. 27, Issue 4, 2006). Although
these studies were directed toward the design of pumped
event-driven systems, the trials were in general conducted
at conditions of steady ow. Plans to study event operation
were abandoned because the full-scale wetlands were placed
in operation and became the study platforms.
There are presently six Stormwater Treatment Areas
(STAs) in operation, aggregating over 16,500 ha and treat-

ing an annual average ow of 4.4 r 10
6
m
3
/d (annual average
HLR  2.7 cm/d). These FWS wetlands have been in opera-
tion for periods of two to eight years and receive episodic
deliveries of pumped inows, occasioned by the at topog-
raphy of South Florida. An example of the ows and total
phosphorus concentrations over the life of STA6 is shown
in Figure 14.9. Summaries of performance may be found in
annual reports (e.g., SFWMD, 2006).
INDUSTRIAL STORMWATER
The stormwater runoff ows and quality from “industrial”
facilities cannot be generalized because those facilities are
quite different in the chemicals that nd their way into run-
off. A broad denition of industrial is taken here, excluding
urban and agricultural runoff. This category includes inor-
ganic sources of nitrogen, such as nitrates and ammonia from
fertilizer manufacturing plants (TCI, 2005), and urea from
airport de-icing (Thoren et al., 2004).
The Simplot engineered FWS wetland (TCI, 2005) was
designed to intercept, retain, and treat previously uncontrolled,
meteorologically driven, nitrogen-contaminated groundwater
forced to the surface in wet weather conditions and otherwise
FIGURE 14.8 A constructed FWS wetland system treating tile drainage from a pasture in Toenepi, New Zealand. Wetlands are incised to
capture tile drainage by gravity.
© 2009 by Taylor & Francis Group, LLC
COLOR FiguRe 1.13  Single-home HSSF wetland in Comfort Lake, Minnesota.
COLOR FiguRe 3.33  Trees growing in the Vermontville, Michigan, constructed treatment wetland after 15 years of operation.

© 2009 by Taylor & Francis Group, LLC
COLOR FiguRe 7.10  Venting groundwater at this Wellsville, New York, site contains iron, which oxidizes upon contact with air.
COLOR FiguRe 11.6  This HSSF wetland outlet structure at Tamarack, Minnesota, has become coated with elemental sulfur.
© 2009 by Taylor & Francis Group, LLC
COLOR FiguRe 22.7  Wetland EW3 at the Des Plaines River site near Wadsworth, Illinois. The fringe zone is vegetated by cattails and 
bulrushes, whereas the interior has submerged aquatic vegetation and oating-leaved plants. Open water shows as dark areas on this false 
color infrared photo.
COLOR FiguRe 14.7  An urban stormwater wetland in the city of Blue Mountains, Australia.
© 2009 by Taylor & Francis Group, LLC
COLOR FiguRe B.8  Progress of Rhodamine WT dye tracer through a FWS wetland, cell 4 of the ENRP in Florida. The dye was intro-
duced along the upper boundary, and ow proceeds from the top to the bottom. Note the short-circuit along the left-hand side, which is 
partially redistributed by the central cross canal. (Photo courtesy of T. DeBusk.)
COLOR FiguRe B.22
  Rhodamine d
ye dispersing through an inlet deep zone, with a spikerush band providing resistance to shortcircuit-
ing. (From CH2M Hill (2003a) PSTA Research and Demonstration Project, Phases 1, 2, and 3 Summary Report, P
repared for the South 
Florida Water Management District.)
© 2009 by Taylor & Francis Group, LLC
Event-Driven Wetlands 551
contributing nutrients to the Assiniboine River. A 4-cell, 22-ha
constructed wetland was implemented in Autumn 2003, near
Brandon, Manitoba. Approximately 100,000 m
3
of nutrient-
rich water entered the wetland in 2004, over the April-through-
November operating period, corresponding to an annualized
HLR of 0.19 cm/d (0.28 cm/d instantaneous). Flows varied by
a factor of ten, depending upon rainfall. Inlet concentrations
were 400 mg/L total nitrogen (TN) and 165 mg/L ammonia

nitrogen, and were reduced by 39% for total nitrogen and
82% for ammonia. The system was zero discharge in 2004,
losing 460 mm of water. This corresponds to design net ET of
500 mm in a typical eight-month operating period.
The 18-ha FWS Kalmar Dämme wetland was built near
Kalmar Airport in southeast Sweden (Thoren et al., 2004).
Urea was used as the runway de-icing agent, and annually
contributed about 41 metric tons of total nitrogen to receiving
waters during 1998–2001. Airport runoff joined agricultural
runoff from a 48-km
2
catchment. The wetland removed 17%
of the TN load from all sources, but removal of urea was esti-
mated to be 38%, based on source allocation. Winter remov-
als were much lower than summer.
These two examples serve to illustrate the applicability
of constructed wetlands to controlling nitrogen from sources
other than municipal and animal wastewaters, or urban and
agricultural runoff. Both represent high nitrogen loadings,
which were treated to a good degree.
BATCH SYSTEMS
Not all event ows are caused by meteorology: the introduc-
tion of water to a treatment wetland may be caused solely
by a human decision. That is the case with batch-operated
systems, in which water is added episodically and drained
episodically. If this sequence occurs on a rapid cycle,
then there is no signicant period of batch holding, as
would be the case for alternative of reciprocating wetlands
0
200,000

400,000
600,000
800,000
1,000,000
1,200,000
1,400,000
May
1999
May
2000
May
2001
May
2002
May
2003
May
2004
May
2005
May
2006
Flow (m
3
/d)
Inflow
Outflow
(a)
FIGURE 14.9 Flows (a) and TP concentrations (b) entering and leaving STA6 in Florida. The average inow rate was 161,000 m
3

/d; the
average outow was 103,000 m
3
/d. The arithmetic mean inlet TP concentration was 58 µg/L, and 25.2 µg/L at the outlet.
0
50
100
150
200
250
3
00
May
1999
May
2000
May
2001
May
2002
May
2003
May
2004
May
2005
May
2006
Total Phosphorus (µg/L)
Inlet

Outlet
(b)
© 2009 by Taylor & Francis Group, LLC
552 Treatment Wetlands
(Behrends et al., 2001). However, if the holding time is long,
then the mode would be termed sequential batch opera-
tion. The water may be simply held without ow during
the batch time, or it may be continuously recycled (Sikora
et al., 1995a). Therefore, like the interevent period of storm-
driven wetlands, batch operation is an option for treatment
wetlands. The operational advantage is that water may be
held until a specic water quality objective is met. The disad-
vantage is that some form of storage is necessary if the batch
mode is to be used to treat continuous ows.
Although it is often assumed that time in a batch system
is equivalent to transit time in a continuous-ow system, this
is a perilous assumption for wetlands. There are several rea-
sons why these are not equivalent.
Hydraulics. Continuous-ow systems display a
distribution of detention times, with different ele-
ments of water spending different lengths of time in
the owing wetland. In a batch system that is lled
and drained very quickly, every element of water
spends exactly the same time in the ecosystem.
The result is that the batch system in effect reects
a circumstance strongly akin to plug ow. Con-
sequently, only the distribution of reaction rate
coefcients attributable to mixtures will affect the
apparent rate of disappearance of a lumped pollut-
ant category (see Chapter 2).

Sorption. In continuous-ow systems, the substrate
is exposed to a relatively constant water phase con-
taminant concentration, which causes the sorption
capacity to be equilibrated and plays no further
role. In a batch system, the sorbed pollutants may
be used up by microbial and vegetative processes,
along with the water-phase materials. Therefore,
at the end of a batch, which is also the beginning
of the next, the sorption sites may be empty and
capable of immediate removal of pollutants to
sorption storage. This phenomenon was reported
by Sikora et al. (1995a).
Vegetation. The environment for vegetation is
different in the batch mode, with diffusion and
transpiration ows providing the transfer of dis-
solved substances, such as oxygen and nutrients.
In the ow mode, advective processes are present
and may dominate. Stein et al. (2003) conducted
microcosm studies on batch-loaded planted gravel
bed systems and compared them to accompany-
ing ow through systems. There were differences
attributable to plants, including more pronounced
seasonal effects and larger differences among
plant species in the batch mode. The result was
superior nitrogen and phosphorus removal in the
batch mode in all seasons and from all conditions
studied.
Microbes. Microbial populations respond to
their environment and evolve accordingly over a
period of time. In a ow through system, there are

•
•
•
•
gradients in water quality from inlet to outlet.
Over the course of time, microbial populations
adapt to different conditions near the inlet and out-
let. Therefore, the water “sees” different types of
microbes during its travel through a owing sys-
tem, which are the result of a stable growth–death
balance at each location. In contrast, the microbes
in a batch system undergo a transient in their pop-
ulation and relative abundance, i.e., they are in a
growth phase.
Despite these differences in site characteristics, there is
typically a near-exponential decline with time of pollutant
concentrations in batch wetlands, often with a residual con-
centration. This is seen at the microcosm level (e.g., Gale
et al., 1993; Van Oostrom and Russell, 1994), the mesocosm
level (e.g., Burgoon, 1993), and at eld scale (Lakhsman,
1981; Sikora et al., 1995a; Kadlec, 2001a). Declines in con-
taminant concentration are well described by the k-C* model
using the exponential decay because of the lack of detention
t
i
me distribution (DTD) effects (Figure 14.10). However, the
rate coefcients in batch systems may differ considerably
from those determined under ow through conditions. For
instance, Burgoon (1993) found a factor of two difference,
with higher rate coefcients for CBOD for batch operation.

Kadlec (2001a) also found higher values for phosphorus
k-values for batch operation.
Experiences on full-scale batch systems are limited.
Studies were conducted on three batch systems at Humboldt,
Saskatchewan, in which lagoon waters were treated in wet-
lands utilizing a ll-hold-drain strategy (Lakhsman, 1982).
In general, BOD was reduced from 90 mg/L to below 20
mg/L in four days, total Kjeldahl nitrogen (TKN) from 20
to below 3 mg/L in 20 days, and phosphorus from 12 to
below 3 mg/L in 20 days. Given an operating depth of 0.3 m,
0.0
0.2
0.4
0.6
0.8
1.0
012345678910
Time (days)
Normalized Concentration (C/C
i
)
Ammonia
Phosphorus
CBOD
FIGURE 14.10 Decline of pollutant concentrations in batch sys-
tems. Ammonia data is from FWS microcosms of Gale et al. (1993).
Phosphorus data is from an FWS eld-scale wetland of Kadlec
(2001a). CBOD data is from SSF mesocosms of Burgoon (1993). All
t lines are for the k-C* model using an exponential decay.
© 2009 by Taylor & Francis Group, LLC

Event-Driven Wetlands 553
the equivalent hydraulic loading to reduce BOD was about
7 cm/d, but reduction of nutrients needed about 1.5 cm/d.
The Oak Hammock, Manitoba, headquarters and visitor
center of Ducks Unlimited, Canada, near Winnipeg, utilizes
a lagoon batch wetland system to treat domestic wastewaters
(Pries, 1994). Waste enters unvegetated lagoon one, which
functions as a facultative pond. It then travels to a second
lagoon with partial vegetative cover of cattails, and nally
to a FWS cattail wetland and then to receiving waters (also
a marsh). About once per month, the water is drained from
the third cell and it is relled from cell two, which is in turn
is lled from cell one. The system stores water during the
winter. The discharge from the three-cell system averages
20 CFU/100 mL fecal coliforms, 0.37 mg/L total phosphorus,
1.3 mg/L TKN, and 0.2 mg/L oxidized nitrogen. The system
is very lightly loaded, with the summer discharge totaling
about 1 m/yr on less than 1 ha.
There is no large-enough database from which to fore-
cast the behavior of batch systems. Ultimately, it would be
desirable to be able to compare the long-term average per-
formance of sequential batch operation to that of continuous
ow.
COMBINED SEWER OVERFLOW (CSO)
In many communities in Europe and North America, there
is but one sewer collecting domestic wastewater and storm
ows. Wastewater treatment facilities are usually designed
to treat ows up to but not including the storm ow. During
large rain events, the excess is bypassed, which is termed a
combined sewer overow (CSO). One of the options for treat-

ing these event overows is use of treatment wetlands. Several
types of wetlands have been constructed for this purpose, and
they are all event driven.
An FWS system to treat CSO was implemented at
Houten-Oost, Netherlands, in 1984 (Lageveen and Waarden-
burg, 1985; van Oorschot, 1990). The CSOs are directed to
a series/parallel network of FWS wetlands containing bul-
rushes (Scirpus lacustris) and common reeds (Phragmites
australis; Figure 14.11). The wetlands comprise 1.5 ha and
operate at a depth of about 0.5 m. Discharge from the wet-
lands is pumped to a cascade aerator and then to a canal. A
brief monitoring study during the summer of 1987 provided
data on two overow events, both completely contained
within the wetlands. Water quality improved rapidly during
storage in the wetlands.
Severn Trent Water built many HSSF CSO systems in the
mid-1990s in the United Kingdom. All were based upon dry-
weather treatment in rotating biological contactors (RBCs),
which bypassed ows above about six times the dry weather
ow (Green and Martin, 1996). The bypassed water was
treated in reed beds. In general, these systems provide signi-
cant improvement to the CSO waters (Figure 14.12). Twelve
of these systems were surveyed in 2003–2004 and found to
be producing efuent concentrations well below the regula-
tory levels (Rousseau et al., 2005a). Varying degrees of sludge
accumulation, weed growth, surface blinding, and unequal
ow distribution were found, but these seemed to have only
minor effects on treatment performance.
A vertical ow design is in use for several stormwater
and CSO systems in the state of Hessia in Germany (Frechen

et al., 2004; Uhl and Dittmer, 2005; Dittmer et al., 2005).
These are underdrained, planted basins that ll with about
1 m of water, which is then drained through about 1 m of sand
or soil. The drainage from these retention soil lters (RSF)
is controlled at 0.09–2.7 m
3
/d per m
2
of lter area. Design
is for 30–40 m/yr total hydraulic load. The performance at
Fulda, Germany, over a 23-event, two-year period was more
than 90% removal of chemical oxygen demand (COD), BOD,
TSS, and ammonia, and 70–90% of phosphorus.
14.3 TSS IN EVENT-DRIVEN WETLANDS
Event-driven treatment wetlands utilize the same suite of
TSS processes as continuous-ow systems, but there are
important differences in the mode of operation. In contrast to
continuous-ow wetlands, there is a greater need to account
for the ow and chemical dynamics. Often, the source-water
TSS is unknown. Performance data acquisition is difcult
because ows occur during storm conditions, and close-
interval sampling is needed to dene the course of a single
event. Internal to the wetland, processes operate on a ow-
and-stop time sequence. The moderately stable gradients
established by steady ows may no longer be present. The
between-event periods are close to batch operation. Small
events may not completely displace the antecedent water in
FIGURE 14.11 The CSO treatment wetland at Houten-Oost,
Netherlands.
© 2009 by Taylor & Francis Group, LLC

554 Treatment Wetlands
the wetland, but large events may replace it many times over.
The ow velocities accompanying an event may be large and
may cause resuspension.
DYNAMIC RESPONSES
The dynamics of both water and solids inuence the perfor-
mance of event-driven wetlands. If the event period is long
compared to the detention time of the wetland under the event
ow, then the rst part of the event completely ushes the
antecedent water, and the outow concentration ramps up to
th
e level dictated by the sustained event ow (Figure 14.13).
After the cessation of event ow, the wetland outows and
outlet concentrations ramp back down to nonow conditions.
The majority of the treatment occurs during the period of high
ow, and the expulsion of the highly treated antecedent water
contributes little to overall performance during the event.
In contrast, if the event period is short compared to the
detention time of the wetland under the event ow, then the
event does not completely ush the (highly treated) anteced-
ent water. Ramp-up of the exported solids load starts to occur
but stops far short of a new steady state determined by the
inow (Figure 14.14). Ramp-down of the mass export is pri-
marily due to the lower ows; TSS remained elevated in the
immediate after-event period.
Data acquisition for individual events involves close-
interval sampling, with numerous samples required to dene
an individual response. Consequently, studies akin to those
at Des Plaines are few in number. It is easier and less costly
to use ow-proportional autosampling, but the result is then

restricted to the EMC for the event.
A second aspect of event wetland dynamics is the vari-
ability from event to event during the course of time. Den-
ing such patterns is also a data-intensive effort, which has
been accomplished in only a few instances (e.g., Carr and
Rushton, 1995). Those researchers examined the water
quality from a system for 83 storm events over a period of
30 months (Figure 14.15). The inows to the Hidden River
urban stormwater wetland followed the pattern of the wet and
dry seasons for this Florida site, and it functioned with zero
discharge in the dry months. Incoming TSS concentrations
were quite variable during the year, but the outow displayed
spikes primarily during the dry season, under no-outow
conditions.
INTERSYSTEM PERFORMANCE
As for continuous-ow systems, there is a sizeable database
of stormwater wetland performance. In the United States, this
database originated with the work of Strecker et al. (1992)
and Schueler (1992), and grew through continued efforts of
Godrej et al. (1999). Table 14.5 shows that there is no clear
trend for constructed urban stormwater wetlands with the
WWAR if the ratio is greater than 1%. The median reduction
is about 70%. In contrast, agricultural runoff does not display
any trend at all with loading or WWAR (Table 14.6). That
is because eld runoff contains soil particles from the elds
that can contain a large fraction of large and heavy mineral
particles, which settle readily. Median reduction is 57%, but
that is also achieved by very small wetlands treating tilled
eld runoff.
In contrast to continuous-ow system data, event sys-

tems have much greater variability in design. Loadings are
extremely variable, as are the geometries and arrangements of
deep pools, shallow marsh zones, and interconnecting channels.
400
300
200
100
0
2:00 7:00 12:00 17:00 22:00
Sat Sun
Flow (L/s)Flow (L/s)Flow (L/s)
BOD (mg/L)
3:00 8:00 13:00
2:00 7:00 12:00 17:00 22:00
Sat Sun
3:00 8:00 13:00
0
2
4
6
8
10
Flow
Inlet
Outlet
Flow
Inlet
Outlet
Flow
Inlet

Outlet
Flow
Inlet
Outlet
1,200
1,000
600
800
400
200
0
TSS (mg/L)
0
2
4
6
8
10
2:00 7:00 12:00 17:00 22:00
Sat Sun
3:00 8:00 13:00
0
2
4
6
8
10
Flow (L/s)
2:00 7:00 12:00 17:00 22:00
Sat Sun

3:00 8:00 13:00
0
2
4
6
8
10
50
40
30
20
10
0
NH
4
-N (mg/L)
7
5
6
4
1
2
3
0
TON (mg/L)
FIGURE 14.12 Flow and concentrations in a CSO reed bed at
Lighthorne Heath, Warwickshire, United Kingdom, in a November
1993 storm event. (From Green and Martin (1996) Water Environ-
ment Research 68(6): 1054–1060. Reprinted with permission.)
© 2009 by Taylor & Francis Group, LLC

Event-Driven Wetlands 555
Therefore, the intrasystem database gives an impression
of what has been accomplished, rather than what might be
accomplished by “optimal” designs. Percent mass remov-
als are high in many cases, with 60% of the U.S. data for
stormwater wetlands of both kinds showing greater than 60%
removal (Figure 14.16).
As noted in the discussion of urban stormwater systems,
a regression has been proposed that relates TSS efuent con-
centrations to hydraulic loading and inlet TSS (Equation
14.8). That regression is entirely consistent with the intersys-
tem continuous-ow FWS wetland data (see Figure 7.18 in
Chapter 7) and may also be used to represent continuous-
ow systems.
14.4 PHOSPHORUS IN EVENT-
DRIVEN WETLANDS
Periodic runoff events from elds in agricultural produc-
tion, or from streets and parking lots, convey phosphorus to
receiving water bodies. FWS treatment wetlands have proven
to remove phosphorus from event ows as well as from steady
ows. Small pulses are very well treated, but those of longer
duration receive much less phosphorus removal.
FLOW PULSES
The dynamics of tracer and phosphorus removal and export
were studied at the Des Plaines River wetland demonstration
site. Wetland pumping events were monitored for all hydro-
logic variables, including pumping, rain, storage change, and
outow. Tracer input and response were monitored at inow,
outow, and interior stations. Phosphorus was measured
at high frequency (every six hours) for the outows, and at

lower frequency for inows and interior stations. The events
were isolated in time, with sufcient interevent spacing to
allow complete equilibration before the subsequent event. A
calibrated dynamic water mass balance was developed as
the framework for interpreting results (Kadlec, 2001a).
0
50
100
150
200
250
–4 0 4 8 1216202428
Number of Displacements
Turbidity (NTU) or Flow (m
3
/hr)
Turbidity In
Turbidity Out
Flow In
Flow Out
FIGURE 14.13 Change in efuent turbidity for a 2.5-day pumping event for Des Plaines wetland C1. The nominal detention time for this
pumping rate was 0.32 days. (From unpublished data.)
0
1,000
2,000
3,000
4,000
5,000
6,000
7,000

–20246810121416
Time (hours)
TSS Mass Flow (g/hr)
Mass In
C1 Out
C2 Out
C3 Out
FIGURE 14.14 Mass inputs and outputs for three wetlands at Des Plaines, driven by a six-hour pumping event. Nominal detention times
at the event ow rate were 7.2, 15.4, and 8.2 hours for C1, C2, and C3, respectively. The wetlands were therefore not completely ushed of
initial water. Mass removals were 73, 90, and 79%, respectively. (From unpublished data.)
© 2009 by Taylor & Francis Group, LLC
556 Treatment Wetlands
0
10
20
30
40
50
0 10 20 30 40 50 60 70 80 90
Storm Number
TSS Concentration (mg/L)
East Inflow
Outflow
(a)
FIGURE 14.15 Results of a 30-month study of the performance of the Hidden River, Florida, urban stormwater treatment wetland. TSS
and ows were monitored for 83 rain events. (Data from Carr and Rushton (1995) Integrating a native herbaceous wetland into stormwater
management. Report by the Southwest Florida Water Management District, July 1995, Boca Raton, Florida.)
0
2,000
4,000

6,000
8,000
10,000
12,000
14,000
16,000
18,000
0 6 12 18 24 30
Months from April 1991
Flows (m
3
/month)
Rain
Inflow
Outflow
(b)
TABLE 14.6
Suspended Solids Entering and Leaving FWS Agricultural Runoff Treatment Wetlands
Name Location Reference
WWAR
(%)
Mean HLR
(cm/d)
Data
(years)
Inlet
(mg/L)
Outlet
(mg/L)
Reduction

(%)
McDowell Co. North Carolina Kao et al. (2001a) — — 1 280 88 69
ENRP Florida Gu et al. (2006) — 3.10 5 5 2 58
Wetland D Norway Braskerud (2001a) 0.03 340 7 — — 50
Wetland A Norway Braskerud (2001a) 0.06 140 7 — — 59
Wetland B Norway Braskerud (2001a) 0.07 120 7 — — 56
Wetland C Norway Braskerud (2001a) 0.07 160 7 — — 53
Alastaro Finland Koskiaho et al. (2003) 0.50 12.80 2 56 46 18
Flytträsk Finland Koskiaho et al. (2003) 3.00 3.80 2 29 26 12
St. John Valley Maine Higgins et al. (1993) 4.80 0.67 2 611 11 97
Hovi Finland Koskiaho et al. (2003) 5.00 1.90 2 530 170 68
B1 Maryland Jordan et al. (1999) 9.29 0.59 1.3 165 135 18
Foster Maryland Jordan et al. (1999) 11.40 0.48 1.3 150 35 77
B2 Maryland Jordan et al. (1999) 13.00 0.42 1.3 110 30 73
Braun Maryland Jordan et al. (1999) 19.17 0.30 1.3 70 55 21
Median 57
© 2009 by Taylor & Francis Group, LLC
Event-Driven Wetlands 557
Internal hydraulics were characterized by tanks in series
(TIS) models calibrated to tracer studies. The residence time
distribution was describable by ve TIS for wetland C1, for
instance. Dynamic mass balances were used, in conjunction
with a rst-order areal removal model, to describe the time
sequence of phosphorus concentrations and ows. Param-
eter estimation produced instantaneous rate coefcients.
A typical response time series and the model represen-
tation are shown in Figure 14.17. This two-day pumping
event nominally ushed the wetland more than seven times,
and the hydraulic loading rate was 87 cm/d. As a result, the
outow concentration during pumping was 213 µg/L, just

below that of the incoming water to the wetland, which was
258 µg/L. During the course of pumping, this corresponds to
a high removal rate of 14 gP/m
2
·yr. After the pumping stops,
outow quickly falls to zero, and the phosphorus concen-
trations in the wetland waters relax down to a background
level of about 54 µg/L. Removal is sustained at high levels,
with k  101 m/yr. Prior studies at this site indicated that this
high rate of removal is not sustained during continuous-ow
operation. The results of this study suggest that great caution
must be exercised in transferring steady-ow results to event-
driven wetlands. For instance, the calculation approach of
Wong and Geiger (1997), using transferred steady-ow rate
constants and C* values, may not be the best descriptor of
pulsed treatment wetlands.
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
<0 0–20 20–40 40–60 60–80 80–100
Percent Reduction
Fractional Frequency
FIGURE 14.16 Distribution of TSS reductions in stormwater wetlands (N = 49). (Data from Schueler (1992) Design of stormwater wetland
systems: guidelines for creating diverse and effective stormwater wetlands in the mid-Atlantic region. Metropolitan Washington Council
of Governments: Washington, D.C.; Whigham et al. (1999) Nutrient retention and vegetation dynamics in restored freshwater wetlands on

the Maryland coastal plain. Final Report, U.S. EPA Agreement CB-993043-04, Smithsonian Environmental Research Center: Edgewater,
Maryland; Carleton et al. (2001) Water Research 35(6): 1552–1562.)
0
50
100
150
200
250
300
350
400
450
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0
Run Days
TP Concentration (µg/L)
Pumping Time
Inlet
Outlet
Relaxation Model
FIGURE 14.17 Response of wetland C1 at Des Plaines, Illinois, to a two-day pumped-pulse addition. The two-day pumping nominally ushed
the wetland 7.2 times. The concentration dropped after pumping to 54 µg/L. The model line has parameters P = 5, k = 101 m/yr, and C* =
54 µg/L. The incoming river water averaged 258 µg/L TP, and during pumping the average outow was at 213 µg/L.
© 2009 by Taylor & Francis Group, LLC
558 Treatment Wetlands
The performance of stormwater wetlands for isolated
events is contingent on the ow rate and duration of the event,
and the capacity of the wetland. The relevant variable is the
number of nominal displacements of the antecedent wetland
water that occur, as discussed in the preceding section on cap-
ture. When these concepts are applied to the Des Plaines data

set, Figure 14.18 results. Phosphorus mass removal decreases
as the number of displacements increases. At very low dis-
placement fraction, only a small amount of (treated) anteced-
ent water is “bumped” out of the system. As the number of
displacements becomes large, the system approaches continu-
ous-ow performance at the event ow rate.
EVENT SEQUENCES
In gravity-fed systems and some pumped stormwater facilities,
events are not necessarily isolated but occur in a time series
with event-to-event overlap at times. The time series of event
treatment may be viewed on a storm-by-storm basis, in which
the mass reduction for each event is considered separately
regardless of the interevent duration. For instance, the perfor-
mance of the Hidden River, Florida, wetland was documented
for 83 consecutive rain events (Figure 14.19). Because the out-
let concentrations depend on event size, wetland storage, and
interevent spacing, there is no obvious simple correspondence
–20
0
20
40
60
80
100
120
012345678
Displacements
Percent Mass Reduction
FIGURE 14.18 The effect of varying displacement on the phosphorus removal efciency of event-driven wetlands. (From Kadlec
(2001a) In Transformation of Nutrients in Natural and Constructed Wetlands. Vymazal (Ed.), Backhuys Publishers, Leiden, The

Netherlands, pp. 365–392. Reprinted with permission.)
0
50
100
150
200
250
300
350
400
450
500
0 102030405060708090
Storm Number
TP Concentration (µg/L)
East In
Out
FIGURE 14.19 Results of a 30-month study of the phosphorus removal performance of the Hidden River, Florida, urban stormwater treat-
ment wetland. Total phosphorus concentrations and ows were monitored for 83 rain events. See Figure 14.15 for the ow time series. (Data
from Carr and Rushton (1995) Integrating a native herbaceous wetland into stormwater management. Report by the Southwest Florida
Water Management District, July 1995, Boca Raton, Florida.)
© 2009 by Taylor & Francis Group, LLC
Event-Driven Wetlands 559
between outow concentrations and inow concentrations
and ow rates. Dynamic mass balance modeling is required
to analyze this complex set of interactions.
URBAN STORMWATER
Over two decades of research directed at stormwater treat-
ment have produced much information on constructed
wetland systems to improve runoff quality. Initial surveys

tabulated data for natural and constructed systems, receiving
either pumped or gravity inows (1992). Data collections have
continued, which build upon previous efforts (Godrej et al.,
1999; Carleton et al., 2001). Emphasis here is on constructed
or engineered systems that receive gravity ows, because
most pumped systems have been discussed under the heading
of continuous-ow wetlands. In addition to project reports
and published papers, the performance of some treatment
wetland systems for stormwater improvement has been cata-
logued in a Stormwater Best Management Practices (BMP)
Database (ASCE, 2003). This database provides access to
BMP performance data in a standardized format for roughly
200 BMP studies conducted over the past 15 years, including
wetlands. More statistical details on specic systems have
been organized in a database by URS (1999).
The amount of phosphorus leaving a watershed dur-
ing a specic event is often the object of research and pre-
dictive models. For example, Brezonik and Stadelmann
(2002) found good correlation for event runoff volumes in
Minnesota, with rainfall amount, drainage area, and percent
impervious area (R
2
 0.78). However, simple correlations for
event mean concentrations and event mass loads were weak,
and varied strongly with land use.
Urban stormwater wetland studies are hampered by the
difculties involved in tracking ows and concentrations in
eld situations. In gravity systems, the rainy weather that pro-
vides inows is not conducive to easy sampling. Tracking of
just one event requires many inow and outow water quality

samples and accompanying chemical analyses. Nevertheless,
there now exist several very detailed studies, such as that of
Carr and Rushton (1995) at the Hidden River, Florida, wet-
land, which was monitored several times in each of 83 rain
events. Similarly, Godrej et al. (1999) studied 33 rain events at
the Crestwood, Virginia, wetland.
Sample results for reduction of phosphorus from con-
structed, gravity-fed systems treating urban runoff waters
are shown in Table 14.7, which reports information from the
summary of Carleton et al. (2001) as well as other sources. It
must be borne in mind that percent removals, even if based
upon load or event-mean concentrations, may reect a low
level of an incoming constituent. Despite such limitations, it
is seen that reductions average 40%. WWAR is seen to not
be a very good predictor of removals for this data set across
multiple sites and conditions, which supports an observation
of Schueler (1992).
Equation 14.9 indicates a positive correlation of phosphorus
remaining with increases in hydraulic loading rate. This trend is
TABLE 14.7
Phosphorus Entering and Leaving Constructed Urban Runoff Treatment Wetlands
Name Location Reference Area Ratio
HLR
(cm/d) % Reduction TP
Armstrong Slough Florida Carleton et al. (2001) 0.3 34.65 39.7
Mays Chapel Maryland Carleton et al. (2001) 0.6 5.55 −7
Shop Creek Colorado Carleton et al. (2001) 0.6 — 36
Franklin Farms Virginia Carleton et al. (2001) 0.8 17.16 14.9
Lake Munson Florida Maristany and Bartel (1989) 1.1 5.19 62.7
Highway Slovenia Bulc and Sajn Slak (2003) 1.1 — 75

Spring Lake Minnesota Carleton et al. (2001) 1.2 — −7
DUST Marsh California Meiorin (1989) 1.8 — 48
Franklin County Ohio Carleton et al. (2001) 2.4 10.60 16
Crestwood Virginia Carleton et al. (2001) 2.4 3.69 45.9
Greenwood Florida Carleton et al. (2001) 2.5 2.57 61.5
Queen Anne Maryland Carleton et al. (2001) 3.8 — 39.4
Clear Lake Minnesota Carleton et al. (2001) 4.9 1.71 54
Tampa Pond Florida Carleton et al. (2001) 5.1 8.16 65
Lake McCarrons Minnesota Carleton et al. (2001) 6.6 7.38 41
Hidden River Florida Carr and Rushton (1995) 19.5 1.04 70
Spring Creek North Dakota Carleton et al. (2001) — 0.22 39.6
Elbow Valley Calgary Amell (2004) — — 36
Kaohsiung China Kao et al. (2001a) — 7.10 70
Median 5.6 41
Note: All are FWS except for Slovenia, which is SSF.
© 2009 by Taylor & Francis Group, LLC