You are on page 1of 15

Ecological Engineering 23 (2004) 189–203

Effects of wetland depth and flow rate on residence time


distribution characteristics
Jeff F. Hollanda,∗ , Jay F. Martina , Timothy Granatab , Virginie Bouchardc ,
Martin Quigleyd , Larry Browna
a Department of Food, Agricultural, and Biological Engineering, The Ohio State University,
590 Woody Hayes Dr., Columbus, OH 43210, USA
b Department of Civil and Environmental Engineering and Geodetic Science, The Ohio State University,

2070 Neil Avenue, Columbus, OH 43210, USA


c School of Natural Resources, The Ohio State University, 2021 Coffey Road, Columbus, OH 43210, USA
d Department of Horticulture and Crop Science, The Ohio State University, 2001 Fyffe Court, Columbus, OH 43210, USA

Received 6 November 2003; received in revised form 25 August 2004; accepted 1 September 2004

Abstract

The residence time distribution (RTD) representing the hydraulics of a wetland is an important tool for modeling and designing
treatment wetlands for optimal constituent removal. To correctly use RTD results, it is necessary to understand the conditions
under which this distribution remains stable. Dye tracer experiments were conducted on a stormwater treatment wetland to
investigate hydrologic factors affecting RTD characteristics. Dye was introduced into the inflow under normal flow conditions
and during simulated storm flows, providing a range of flow rates and water levels. Dye distribution in the outlet was measured
using an in situ fluorometer. Results indicate that flow rates did not have a significant effect on RTD characteristics. The RTDs
normalized for volume and flow demonstrated a greater amount of short-circuiting and a larger mixing scale when water depth
increased, demonstrating that water level can have a direct impact on the RTD of a wetland. This effect suggests that more than
one RTD may be necessary for analyzing a wetland subject to changing water levels. For the wetland in this study, increasing
the water depth elicited a decrease in hydraulic efficiency. Understanding such factors that affect hydraulic efficiency will aid in
the design and management of wetlands.
© 2004 Elsevier B.V. All rights reserved.

Keywords: Hydraulic efficiency; Rhodamine WT; Nonideal flow; Moment analysis; Hydrology; Dye tracer

1. Introduction

∗ Corresponding author. Present Address: 619 12th Avenue, 1.1. Residence time distributions
Huntington, WV 25701, USA. Tel.: +1 304 399 5108/304 523 4272;
fax: +1 614 292 9448. The residence time distribution (RTD) is a tool that
E-mail address: holland.144@osu.edu (J.F. Holland). has advanced the science and engineering of treatment

0925-8574/$ – see front matter © 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.ecoleng.2004.09.003
190 J.F. Holland et al. / Ecological Engineering 23 (2004) 189–203

wetlands. A pulse of a non-reactive tracer chemical 1996). Increasing the system volume or decreasing the
dissolved into a wetland inlet is used to measure an flow rate lengthens the retention time, thus extending
RTD. In the ideal case, a wetland experiences plug flow, the raw RTD in time. Conversely, decreasing the vol-
in which water and its dissolved constituents run uni- ume or increasing the flow shortens the retention time,
formly and without dispersion from inlet to outlet. If compressing the raw RTD. The horizontal orientation
a tracer pulse is added to an ideal wetland inlet, all of of the RTD curve is therefore a function of volume and
the tracer will exit at the same time, the residence time. flow. For this reason, it is common to use volume and
In real wetlands, flow is nonideal. Nonideal flow com- flow to normalize the time axis into dimensionless units
prises different flow path lengths, flow velocities, and (Fig. 1) (Levenspiel, 1972; AWWARF, 1996). This nor-
mechanisms of diffusion and mixing, causing a distri- malization procedure, essential for comparing RTDs of
bution of residence times, an RTD (Levenspiel, 1972). systems under different hydrologic conditions, is com-
When a tracer pulse is introduced into a nonideal wet- plicated by pulsed flow, where there may not be one
land, the outflow tracer concentration is an RTD reflect- single system volume or flow rate. Werner and Kadlec
ing the dispersive nature of the system. For simplicity, (1996) developed a dynamic normalization procedure
real wetlands are traditionally, though arguably inac- for RTD analysis of pulsed systems. While hydrologic
curately (Kadlec, 2000), modeled as ideal, plug-flow changes stretch or compress the raw RTD, the normal-
systems. RTD analysis has been developed to more ization procedure removes these effects, isolating the
accurately model treatment wetland performance, im- dispersive and mixing characteristics of the system
proving on the ideal wetland model, and leading to (Fig. 1). Dispersive and mixing characteristics are
better understanding and design of wetlands. important to resolve in a treatment wetland because
Many applications of RTD analysis have been ex- they affect the system’s treatment performance.
plored in wetland science and engineering, but the
consistency and stability of these have not been well 1.3. Hydraulic efficiency
studied. To strengthen the efficacy of RTD analysis,
its range of utility must be understood more fully. Wetland engineers use RTD analysis to quantify
The present research study explores a range of hydro- wetland design characteristics that affect treatment
logic conditions that affect RTD stability. By perform- efficiency. In order to characterize the performance of a
ing tracer studies during hydrologic manipulation of a wetland, it is useful to reduce the RTD to a single num-
small wetland, this study explores the sensitivity of the ber, the hydraulic efficiency. The hydraulic efficiency
RTD to changing flow rates and water levels. Because represents the ability of a wetland to distribute its flow
RTD analysis is used to compare the dispersive charac- uniformly throughout its volume, maximizing contact
teristics of wetlands independent of volume and flow time of pollutants in the system and optimizing the
rates, understanding the influence of variations of vol- ability to break down these pollutants. Thackston et al.
ume and flow rates is needed to improve the application (1987) introduced the concept of hydraulic efficiency
of RTD analysis. in wetlands by quantifying the relative position of the
RTD’s centroid. The RTD centroid is a measure of
1.2. Normalization of the RTD the average residence time of the RTD, representing
the true retention time of the system. Ideally, the true
In order to compare RTDs between different retention time of the system is equivalent to the theo-
wetlands or dissimilar conditions, each RTD must be retical retention time of the system, determined by the
normalized by removing the units of flow rate, system system volume and flow rate. The centroid of an RTD
volume, and tracer mass. A raw RTD can be measured falls below the theoretical retention time when short-
by adding a pulse of tracer to an aquatic system and then circuiting causes the loss of effective retention volume
measuring the resulting outflow tracer concentration as (Fig. 2). The scale of mixing in a wetland also
a function of time. Because the outlet concentration is affects its hydraulic efficiency. Complex systems,
a function of the tracer mass and wetland volume, these such as wetlands, are often described in equivalent
factors can be used to normalize the concentration number of continually stirred tank reactors, or CSTRs
axis into dimensionless units (Werner and Kadlec, (Levenspiel, 1972; Kadlec and Knight, 1996). A
J.F. Holland et al. / Ecological Engineering 23 (2004) 189–203 191

Fig. 1. The normalization procedure applied to the concentration curve of a tracer. Normalization removes effects of flow rate (Q), volume
(V), and tracer mass from the raw residence time distribution (RTD), producing the same normalized RTD from tracer studies run in similar
systems under dissimilar conditions. Certain characteristics of the tracer curve, such as the centroid and area, become unity after an ideal
normalization.

wetland can be compared to an equivalent volume of of both (Persson et al., 1999). The measurement of
many small CSTRs or a few large CSTRs. Many small hydraulic efficiency is a simple and effective method
CSTRs in series represent a system with a small mixing of characterizing wetlands, identifying factors that
scale, which has a small RTD spread and resembles affect treatment wetland performance.
plug flow (Fig. 2). As the number of equivalent CSTRs
decreases, the mixing scale and the spread of the RTD 1.4. Factors influencing hydraulic efficiency
increase (Fig. 2). A large RTD spread is considered
inefficient for a chemical reaction system (Levenspiel, Wetland shape affects the RTD and the hydraulic
1972). A metric of hydraulic efficiency may quantify efficiency. Zones of diminished mixing (ZDM), where
the short-circuiting, the mixing scale, or a combination the area of a wetland is not being optimally utilized
192 J.F. Holland et al. / Ecological Engineering 23 (2004) 189–203

Fig. 2. Conceptual diagram of the effects of mixing scale (A) and short-circuiting (B) on residence time distribution (RTD) characteristics.

for pollutant treatment (Thackston et al., 1987; Kadlec, influencing the wetland’s RTD (Koskiaho, 2003). The
1994), lower the hydraulic efficiency by reducing the inlet and outlet locations and their structure also af-
effective volume (Fig. 2). Wetland shape and aspect fect hydraulic efficiency. RTD analysis has been used
ratio are important design considerations because they to optimize inlet (Shilton and Prasad, 1996) and outlet
affect the location and abundance of ZDM (Walker, structures (Konyha et al., 1995) or both (Ta and Brignal,
1998; Persson et al., 1999). RTD analysis has been used 1998) for maximum treatment efficiency.
to design wetland shape to reduce ZDM, optimizing Although hydrologic conditions, such as flow rate
treatment efficiency (Koskiaho, 2003). and wetland depth, may affect hydraulic efficiency, few
Internal structure can affect flow dynamics and studies have quantified these effects. Natural flood-
therefore hydraulic efficiency of a wetland. Aquatic ing events in riparian wetlands have caused observ-
macrophytes, for example, enhance lateral (Nepf et al., able changes in RTD characteristics (Stern et al., 2001).
1997; Nepf, 1999) and vertical (Nepf and Koch, 1999) The water level in a treatment wetland, either caused
diffusion and flow in a wetland, potentially enhancing by flooding or controlled by wetland design and man-
the wetland’s hydraulic efficiency. If patchy vegeta- agement, should similarly affect hydraulic efficiency.
tion creates stagnant ZDM, however, the hydraulic ef- Basin morphology and vegetation, for example, would
ficiency may decrease (Thackston et al., 1987). The in- likely have different effects relative to different water
fluence of vegetation on flow patterns indicates that sea- depths. Flow rate should also affect RTD characteristics
sons or ecological succession may influence the charac- and hydraulic efficiency. Turbulent diffusion, enhanced
teristics of wetland flow. Basin morphology also affects by vegetation and other obstructions, is related to the
dispersion and flow paths through a wetland, strongly Reynolds number and thus flow velocity (Nepf et al.,
J.F. Holland et al. / Ecological Engineering 23 (2004) 189–203 193

1997). Turbulent diffusion may therefore dominate the The results of this study were intended to eluci-
diffusive and mixing mechanisms at higher flow rates. date two areas of wetland science and engineering.
Different mechanisms of flow dispersion would govern First, the results will help determine under which hy-
dynamics of low flow. Convective circulation (Oldham drologic conditions an RTD remains valid for a wet-
and Sturman, 2001) and diffusion driven by wind or land. Nonideal flow models and hydraulic characteri-
molecular movement (Kadlec and Knight, 1996; Keller zations will become more robust when the limitations
and Bays, 2000) would likely dominate during long re- to their underlying RTDs are understood. This will
tention times. Because flow rate and depth affect wet- lead to better modeling and design of wetlands. Sec-
land hydraulics, studies are needed to quantify their ond, because depth and flow rate can be controlled
effects on RTD characteristics. in engineered wetlands (Kadlec and Knight, 1996;
Mitsch and Gosselink, 2000), understanding how hy-
1.5. Limitations to RTD methods draulic efficiency is affected by these parameters will
influence how treatment wetlands are designed and
Understanding depth and flow factors is important managed.
for determining conditions under which an RTD or its
derived hydraulic efficiency metrics can be applied.
Typically, a single RTD is used to characterize a wet- 2. Methods
land, a procedure that assumes that each wetland em-
bodies a single, intrinsically stable RTD. This is a 2.1. Site description
useful simplification for analyzing the efficiency of a
treatment wetland; however, it remains to be demon- This study was performed in the spring and sum-
strated under which circumstances the intrinsic RTD mer of 2003 on a 250-m2 wetland constructed on the
assumption is valid. A similar uncertainty arises when Waterman Agricultural and Natural Resource Labora-
a tracer study is performed under pulsed conditions. tory at the Ohio State University in Columbus, Ohio
Producing a normalized RTD from pulsed conditions (Fig. 3). The wetland had three adjacent inlet structures
is a practical means of assessing a wetland subject to that discharged water from storm flow, overland flow,
storm flow (Werner and Kadlec, 1996). Studies have and nursery pad irrigation runoff, and one controlled
not yet demonstrated to what extent such an RTD re- outlet structure that discharged to another wetland cell
flects the wetlands local and transient conditions versus (discharge structures from Agri Drain Corp.). To simu-
how much it represents the intrinsic state of the wet- late storm flow, water was added to the overland inflow
land. In conventional waste treatment facilities, it is structure using a vinyl hose connected to a 5-cm high-
common to run tracer tests at various flow rates and discharge irrigation riser. Flow rates were controlled
water levels to assess the system performance under by valves on the hose and riser. Although all three
varying conditions (AWWARF, 1996). A similar pro- inlets conducted flow during real storm events, sim-
tocol would be beneficial for a treatment wetland if ulated storm events were only conducted through the
studies demonstrate a flow or depth effect on wetland overland flow entrance. An adjustable weir was used
RTD characteristics. to control the depth (i.e. water level) in the wetlands.
To do this, boards with watertight seals were slid into a
1.6. Purpose of study track, raising or lowering the weir to a predetermined
level (Fig. 3).
The purpose of this study was to investigate the To determine the volume and depth of the wetland
sensitivity of the normalized RTD to different applied for different flow conditions, wetland topography and
depths and flow rates. To accomplish this goal, a se- weir positions were surveyed using a total station in
ries of dye tracer experiments was run on a small con- the summer of 2002 (Fig. 3). The survey results were
structed wetland under controlled flow rates and water processed by computer to define a relationship between
levels. Water level and flow rate were regulated with water level and wetland volume, using the assumption
adjustable-depth control structures and flow-adjustable that the water level at the outlet is uniform across the
irrigation water. wetland (Fig. 4).
194 J.F. Holland et al. / Ecological Engineering 23 (2004) 189–203

2.2. Materials tems with retention times less than 1 week (Lin et al.,
2003).
Each inlet and outlet to the wetland was monitored
by a YSI 6-series sonde (YSI, Inc.), set upstream of the 2.3. Experimental protocol
V-notch weir (Fig. 3). Capable of resolving depth to
±1 mm, the YSI devices monitored flow based on the To concurrently test the effects of high and low wa-
measured weir head (ASTM, 2002). The outlet YSI ter levels and high and low flows on the RTD, 12 ex-
device had an in situ probe specifically designed to periments were performed over a period of 13 weeks.
detect the fluorescence of the dye tracer Rhodamine The flow rates were difficult to maintain precisely dur-
WT. ing the experiments and, therefore, were quantified by
Rhodamine WT was chosen, not only because of their averages of 1.2 L s−1 for low flow and 3.2 L s−1
the fluorometer’s compatibility with existing instru- for high flow. The average depth settings were 16.6 cm
mentation, but also because Rhodamine WT is a com- for low water level and 39.8 cm for high water level
monly used dye tracer in wetland studies (Stern et (Table 1).
al., 2001). Rhodamine WT has a low natural back- Experiment sets were performed in monthly blocks
ground interference (Wilson et al., 1986) and accept- to balance the seasonal influences on each controlled
ably low adsorption and degradation rates in small sys- condition (Table 1). For the relatively small sample

Fig. 3. Diagram of the setup for the dye tracer experiments at the Waterman Agricultural and Natural Resource Laboratory wetland. The
bathymetric map of the wetland basin has contours every 10 cm, with depths labeled at 30 cm intervals. Depths are in meters relative to survey
station.
J.F. Holland et al. / Ecological Engineering 23 (2004) 189–203 195

Table 1 the study. For example, if one treatment occurred


Experimental settings for the 2003 dye tracer experiments, showing more often in the early part of the season, it would
the average flow rates applied by irrigation water and the wetland
water level settings for each tracer experiment
be impossible to determine whether observed effects
were due to treatment or season. A prescription of
Week begin Applied water First applied Second applied
level (cm) flow (L s−1 ) flow (L s−1 )
weekly changing water levels was randomized over
each monthly block. Two flow rates were assigned
Monthly block 1
19 May 16.6 2.27 0.29
to each week, and their order was chosen randomly.
26 May 39.8 2.08 1.08 Each flow pair was performed on one of each wa-
ter level per month. A protocol was developed for
Monthly block 2
16 June 39.8 1.56 3.33 the paired tracer experiments to run at the first occur-
23 June 16.6 1.51 3.36 rence of each water level each month or to postpone
Monthly block 3
the experiments until the next similar water level in
14 July 16.6 3.36 1.19 the case of inclement weather or technical constraints
2 August 39.8 1.52 5.03 (Table 1).
Preceding each dye tracer experiment, the irriga-
tion flow was set to a predetermined level (Table 1)
size of this experiment, it would have been possi- and allowed to flush the wetland for at least one re-
ble under fully randomized settings to have one set tention time to allow for thermal equalization. Rho-
of experimental prescriptions become climatologically damine WT dye was added in approximate proportion
biased by grouping at the beginning or the end of to wetland volume: approximately 4 g pure Rhodamine

Fig. 4. Volume vs. stage relationship of the experimental wetland based on survey data. This relationship is used to find the volume of the
wetland when the elevation of the water at the outlet (stage) is known.
196 J.F. Holland et al. / Ecological Engineering 23 (2004) 189–203

WT for the low water level and 10 g for high water where Mout is the amount of tracer to exit the wetland
level. Dye was added to the inlet by slowly pouring (Kadlec and Knight, 1996). The ratio of the exit mass,
15 L of a mixture of dye and wetland water into the Mout , to the added mass, M, was calculated to determine
overland flow structure (Fig. 3). At the outlet, the YSI the dye recovery. For consistency in this study, each
sensor equipped with the Rhodamine probe (Fig. 3) RTD and corresponding Mout was analyzed over the
monitored the dye concentration and weir head every interval 0 <φ <2.
5 min. Moments of the dimensionless RTD functions were
used as a basis for calculating the hydraulic efficiency
2.4. Data analysis (Kadlec and Knight, 1996). The moments of the nor-
malized RTD functions are defined as
 ∞
Although each dye tracer experiment was run un-
M0∗ = C (φ) dφ (4)
der relatively steady conditions, each RTD had to be 0
normalized in order to compare between conditions of  ∞
different flow rates and system volumes. Werner and M1∗ = φC (φ) dφ (5)
Kadlec (1996) proposed normalizing the RTD con- 0
tinuously by flow rate and by system volume, where  ∞
time on the abscissa is replaced by dimensionless flow- M2∗ = (φ − M1∗ )2 C (φ) dφ (6)
0
weighted time, φ:
 t The zeroth moment of the normalized RTD, M0∗ , is
Q(t  )  equivalent to the fraction of the mass of the dye recov-
φ= dt (1)

t0 V (t ) ered. With tracer loss corrected, M0∗ is always unity.
When M0∗ = 1, the first moment, M1∗ , is the centroid
where t is the time, t is a dummy variable of integra- of the RTD. The second moment, M2∗ , is the variance, or
tion, t0 the time of dye delivery, Q(t) the outflow rate, σ 2 , of the RTD, which accounts for the spread of the dye
and V(t) the wetland volume. Because the theoretical over time. In theory, M1∗ = 1 for a normalized RTD of a
retention time for a system with steady flow occurs at conservative tracer in a system with no dead zones, and
φ = 1, the parameter φ can be thought of as the num- M2∗ = 0 during plug flow (Werner and Kadlec, 1996).
ber of theoretical retention times. Werner and Kadlec Consequently, the deviation from the ideal case can be
(1996) defined φ under the special case where V(t) is used to quantify the hydraulic efficiency.
assumed to be a constant: Vsys . In the present study, the Different metrics of hydraulic efficiency were cal-
volume-to-stage (i.e. water level) relationship (Fig. 4) culated and compared in this study. Thackston et al.
enabled the calculation of volume as a function of time, (1987) defined hydraulic efficiency as the ratio of the
V(t), providing a more accurate description of the sys- actual and theoretical retention time:
tem volume. The dimensionless RTD function, C (φ),

is defined as λt = = M1∗ (7)
T
C(φ)V (φ)
C (φ) = (2) where λt is the hydraulic efficiency proposed by
M
Thackston et al. (1987), t̄ the system’s true retention
where C(φ) is the outflow concentration, V(φ) the sys- time, equivalent to the RTD centroid, and T the theo-
tem volume, both functions of flow-weighted time, and retical retention time determined by the surveyed vol-
M the mass of the tracer added to the system (Werner ume and measured flow. For a normalized RTD, T = 1,
and Kadlec, 1996). Although tracer degradation should and it follows that λt is the centroid of the normalized
be avoided, some loss of tracer is inevitable. In this RTD. When flow or volume varies during the tracer
study, this loss was corrected for by substituting the experiment, the steady-state parameters t̄ and T cannot
exit mass into Eq. (2): be directly calculated, but M1∗ can be computed to de-
 ∞ termine λt (Eq. (7)). Tracer spread was also used as a
Mout = Q(t)C(t) dt (3) measurement of hydraulic efficiency. While there are
0 many ways of measuring tracer spread (Persson et al.,
J.F. Holland et al. / Ecological Engineering 23 (2004) 189–203 197

1999), a normalized form of the variance was used in sided, pooled t-tests were used to compare the effects
this study. The variance of the RTD, M2∗ , describes the of water level on RTD statistics, yielding 11 degrees of
scale of mixing of the system. If the effective volume freedom for this test. All tests were used to determine
of the system is compromised by short-circuiting, i.e. with a 95% confidence level whether the differences of
λt < 1, the variance may be normalized by the square the mean RTD characteristics, λt, λp , tm , and σθ2 were
of the reduced centroid (Kadlec and Knight, 1996): nonzero.
M2∗
σθ2 = (8) 2.5. Natural events
λ2t

where σθ2 is the normalized system variance, a metric of Dye tracers were also used to measure natural,
the RTD spread. If a system is modeled with continu- pulsed-flow events in the wetland. Using similar meth-
ally stirred tank reactors (CSTRs) in series, the mixing ods to the experiments with artificially simulated flow,
scale of the RTD is quantified by σθ2 = N −1 , where N is dye was added to the storm-flow wetland inlet (Fig. 3)
the theoretical number of CSTRs (Kadlec and Knight, immediately preceding a large storm event and out-
1996). A synthesized version of hydraulic efficiency, flow measurements and data analyses were made as
incorporating both effects of short-circuiting and mix- before. Because the storm-flow tracer experiments re-
ing scale, was also calculated: lied on multiple storm events to complete the RTD
    curve, their long and unpredictable time frame ex-
1 t̄ t̄ − tp tp cluded the possibility of replication. Technical con-
λp = λt 1 − = 1− = = tp∗ (9)
N T t̄ T straints also limited the quality of data. The Rho-
damine probe introduced spiked noise under high-
where λp is the hydraulic efficiency proposed by turbidity conditions. The extreme spikes (i.e. out-
Persson et al. (1999), N the equivalent number of liers) were problematic in the high water level ex-
CSTRs in series, tp the time of peak concentration, and periment and were replaced by the surrounding mean
tp∗ the normalized time of peak concentration. This con- values, degrading the signal and thus the data qual-
cept of hydraulic efficiency is simplified algebraically ity. As a result of data noise and lack of replica-
by the assumption that the RTD spread can be modeled tion, natural events were not statistically compared
by series of CSTRs. to flows created using irrigation water. Instead, the
Minimum travel time was also calculated as a char- results from the natural tracer experiments are pre-
acteristic of the RTD that identifies short-circuiting. sented as a subjective means of comparing and con-
The minimum travel time, tm , was defined as the short- trasting naturally pulsed events to artificially controlled
est normalized time of travel from the inlet to outlet, flooding.
determined by the fastest flow path through the wet-
land. Theoretically, this is the time elapsed between
the introduction of dye at the inlet and the detection of 3. Results
dye at the outlet. However, background fluctuations in
fluorescence made the exact time of the initial signal in The controlled depth varied by factor of 2.4 between
the outlet difficult to determine. Therefore, a minimum treatments and the average low and high flow rates var-
travel time was defined arbitrarily as the normalized ied by a similar factor of 2.7 (Table 1). Dye recovery for
time of the last measurement before the measured con- all experiments ranged from 75 to 95% with an average
centration reached 3% of the peak concentration. The of 84%. There were no consistent differences between
time identified with this method matched the first rise the RTD plots at different flow rates. None of the pa-
of the RTD curve. rameters for hydraulic efficiency differed significantly
Because experiments were conducted in paired between high and low flow rates (Fig. 5): λt varied
combinations of flow rates, two-sided, paired t-tests from 0.51 at low flow to 0.53 at high flow (P = 0.53),
were used to statistically compare how flow rates af- λp changed from 0.22 at low flow to 0.25 at high flow
fected the RTD characteristics. This yielded five de- (P = 0.30), tm was 0.11 at low flow and 0.12 at high
grees of freedom from the 12 tracer experiments. Two- flow (P = 0.81), and σθ2 was 0.65 for low flow and 0.55
198 J.F. Holland et al. / Ecological Engineering 23 (2004) 189–203

Fig. 5. Statistics of the residence time distribution (RTD) characteristics under differing flow rates and depths. Comparisons are made between
flow rates (averaged over two water levels) and water levels (averaged over two flow rates) (mean ± standard error) for (A) peak concentration
time, (B) minimum dye travel time, (C) RTD centroid (first moment), and (D) normalized variance of the RTD.

for high flow (P = 0.053); each of these measurements Table 2


is unitless. RTD statistics of the dye tracer experiments run at low and high water
level during natural storm events
The RTDs compared between high and low wa-
ter levels appeared distinctively different. Whereas the Statistics Low water level High water level
RTD of the low water level was unimodal, the RTD for λp 0.44 0.083
the high water level was usually bimodal, with the first λt 0.81 0.31
tm 0.41 0.026
peak typically higher and narrower than the main peak σθ2 0.23 0.41
of the low water level RTD (Fig. 6). These differences
were reflected in the RTD statistics (Fig. 5). The mean
minimum travel time, tm , was 0.15 for the low water and 0.49 for the high water level but this difference was
level and 0.077 for the high water level, a difference that not significant (P = 0.088). Each of these measurements
was statistically significant (P = 0.00012). The time to of hydraulic efficiency is unitless.
peak concentration of the RTD curve, λp , occurred at Two natural dye tracers were run during natural
the mean φ value of 0.29 during the low water level storm flow, one at high water level and one at low wa-
and 0.19 during the high water level, which also rep- ter level. The RTDs of the natural dye tracers (Fig. 7)
resents a statistically significant difference (P = 0.041). appeared different from those of the simulated storm
The mean RTD spread, or σθ2 , which was 0.47 at low flows (Fig. 6), though they showed similar, general
water level and 0.73 at high water level, changed sig- trends. For natural flow events, the RTD for the high
nificantly in response to water level (P = 0.0089). The water level had lower values of λt, λp , tm , and a larger σθ2
mean RTD centroid, λt , was 0.55 for the low water level compared to the RTD of the low water level (Table 2).
J.F. Holland et al. / Ecological Engineering 23 (2004) 189–203 199

Fig. 6. Comparison of two representative residence time distribution (RTD) curves run at different water levels.

4. Discussion that volumetric fluctuations were problematic in their


procedure. Determining Vsys , the average system vol-
4.1. RTD sensitivity to hydrologic influences ume, a step necessary for normalization, is not straight-
forward under large volumetric fluctuations. This prob-
The results of this study indicate that the RTD of a lem may be avoided by defining volume as a function of
wetland is sensitive to changes in water depth, but not time, V(t), as the present study has done. Unfortunately,
significantly sensitive to changes in flow rate. If this the procedure used to determine this functional rela-
result is characteristic of wetlands in general, then it in- tionship would be impractical for many larger wetlands
dicates that a change in wetland depth by a factor of 2.4 that cannot be accurately or cost-effectively surveyed.
is enough to elicit a significant change in RTD charac- Because volume is a function of depth (Fig. 4), care
teristics. If wetland management or hydrologic factors should be taken in interpreting an RTD during large
cause depth changes of a similar magnitude, more than volumetric fluctuations, regardless of how the volume
one RTD will be needed for an appropriate hydraulic is calculated.
characterization. The results of this study also indicate Although this study demonstrates no significant ef-
that RTD characteristics may not change significantly fect of flow rate on the RTD, the limits must be con-
when flows vary by up to a factor 2.7. Flow normaliza- sidered under which this observation is made. Natural
tion of an RTD would be robust to this magnitude of flow rates may vary by magnitudes greater than those
change. The RTD normalization proposed by Werner tested in this experiment. Depth fluctuations, on the
and Kadlec (1996) should work appropriately for mod- other hand, are more likely to remain within the range
erately varying flow rates, during which the wetland’s tested in this study. It is therefore expected that flow
water level, and therefore its volume, does not change effects on the RTD may play a greater role under natu-
appreciably. Werner and Kadlec (1996) acknowledged rally pulsed conditions. In this study, not enough tracers
200 J.F. Holland et al. / Ecological Engineering 23 (2004) 189–203

Fig. 7. Residence time distribution of a dye tracer during natural, pulsed flow for low water level (A) and high water level (B).

were run under natural conditions to test this conjec- tically significant, indicated short-circuiting. This ob-
ture. The mechanisms of dispersion that may dominate servation is supported by a significant 49% increase
at low flow between pulsed events, such as convective, of the normalized travel time of the fastest flow path,
molecular, and wind-driven diffusion, are difficult to tm , during the same depth increase. The RTD spread,
isolate and measure with RTD analysis because of the σθ2 , increased by 57% as depth increased, representing
long retention times involved. Over very long retention a decrease in hydraulic efficiency. In response to in-
times, tracers tend to degrade (Lin et al., 2003). The creased depth, the hydraulic efficiency measured by λp
longer necessary time periods for low-flow tracer ex- decreased by 35%. All the parameters expressing hy-
periments under controlled conditions require a longer draulic efficiency exhibit lower efficiency for the higher
study period and introduce a higher possibility of inter- water levels. This indicates that less-efficient flow pat-
ference from inclement weather and other confounding terns were prevalent at greater water depths. Although
factors. Such further experimentation was not possible the depth effect on hydraulic efficiency affects the per-
in present study’s time frame. formance of a treatment wetland, this effect must be
weighed against other factors. The depth of a treat-
4.2. Design and management implications ment wetland is an important design consideration be-
cause of its effect on the retention time (Kadlec and
The depth effect on hydraulic efficiency observed Knight, 1996), the ecology (Batzer and Resh, 1992;
in this study may be important when wetland design Mitsch and Gosselink, 2000), and the hydrologic ef-
and management are considered. The λt parameter de- fectiveness, a measure of compliance with the mini-
creased by 12% due to the controlled depth increase. mum prescribed retention time in a stormwater treat-
This reduction in RTD centroid, although not statis- ment wetland (Wong and Somes, 1995). If the results of
J.F. Holland et al. / Ecological Engineering 23 (2004) 189–203 201

this study are representative of other wetlands, the ef- RTD analysis. The concept of short-circuiting can be
fect of depth on hydraulic efficiency should be a factor used in this study to show that 55% (λt = 0.55) of the
considered during wetland design and management. wetland volume is effectively being used at low water
The effect of flow rate on hydraulic efficiency, in level, but only 49% (λt = 0.49) is effectively used at
contrast to depth, appears to be minimal within the high water level (Fig. 5C), representing a relative 12%
range tested in this study. Each of the hydraulic ef- loss of effective treatment wetland volume at high wa-
ficiency parameters exhibited smaller changes due to ter level. The ability to quantitatively describe wetland
flow rate than due to depth, and none of these changes effectiveness makes λt an important factor to analyze
were statistically significant. This occurred in spite of treatment wetlands.
the change in flow by a factor of 2.7, which was slightly Plug-flow efficiency (Ta and Brignal, 1998), based
greater than the factor of depth change, 2.4. Of all the on the RTD spread, or σθ2 , cannot easily be used to
parameters measured, σθ2 exhibited the greatest change quantify treatment efficiency, but it is a quantitative
due to flow rate. The RTD spread, σθ2 , decreased by description of the system’s mixing scale. The recipro-
15% due to an increase in flow rate. Although this indi- cal of σθ2 is a measure of the equivalent number, N,
cates an increase in hydraulic efficiency at higher flow of CSTRs in series (Kadlec and Knight, 1996). In this
rate, this difference could be statistically due to ran- manner, the low water level of this study can be com-
dom chance. The relationship between flow rate and pared to a series of N = 2.1 CSTRs and the high water
retention time is expected to have a much greater influ- level to a series of N = 1.4 CSTRs. The possible mix-
ence on treatment efficiency (Kadlec and Knight, 1996) ing scale varies between fully mixed flow at N = 1 and
than the relationship between flow rate and hydraulic unmixed plug flow at N = ∞. Changes in mixing scale,
efficiency. however, do not directly translate into changes in treat-
ment efficiency. It can be shown for all reaction rates
4.3. Assessment of hydraulic efficiency metrics of first-order kinetics or greater that reaction comple-
tion is maximized with plug flow, σθ2 = 0. Any finite
The results of this study demonstrate that different scale of mixing (σθ2 > 0) reduces the treatment effi-
hydraulic efficiency metrics can respond uniquely to ciency unless the reaction follows zero-order kinetics,
the same treatment. In response to a depth increase, in which case plug flow and mixed flow have equiv-
a measure of short-circuiting, λt , decreased by 12%, alent reaction completions (Levenspiel, 1972; Lawler
whereas a measure of mixing scale, σθ2 , increased by and Benjamin, in press). There are other benefits to a
57%. The metric λp , a combination of mixing scale high σθ2 value, however, that may negate the theoret-
and short-circuiting, decreased under the same condi- ical loss of reaction completion. For example, mixed
tions by 35%. Although each of these observations in- flow can stabilize and equalize pulsed inputs by dilut-
dicates a decrease in hydraulic efficiency, there is not ing potentially dangerous pollutant inputs (Lawler and
one consistent measurement of hydraulic efficiency to Benjamin, in press). Therefore, in some circumstances,
report. Each metric of hydraulic efficiency used in this mixing is at most desirable, or at least neutral, even
study has unique implications on treatment effective- though some concepts of hydraulic efficiency under-
ness. Future researchers may therefore consider using value this effect. Any change in treatment performance
more than just one measurement of hydraulic efficiency due to mixing scale therefore cannot be quantified by σθ2
when assessing a wetland. alone.
In general, short-circuiting is never desirable in a Both aspects of hydraulic efficiency, the measure-
treatment wetland because treatment area is effectively ment of mixing scale and the measurement of short-
lost (Thackston et al., 1987; Kadlec, 1994). This may circuiting, are important for hydraulic characteriza-
not be of concern in wetland ecology, but it is of prac- tion. The synthesis of these two concepts proposed
tical concern regarding the performance of a treatment by Persson et al. (1999), though convenient to calcu-
wetland. The centroid of the normalized RTD, λt , rep- late, is not universally applicable. Not only is some
resents the fraction of the wetland that is not short- important information lost, but also reducing the cal-
circuited. Short-circuiting, quantified by λt , is therefore culation to identifying one single point on the RTD,
an unquestionably important factor to be resolved from the peak concentration time, may make this proce-
202 J.F. Holland et al. / Ecological Engineering 23 (2004) 189–203

dure susceptible to error, especially when the RTD comparing different water levels should be run under
is not smooth (Fig. 7). The hydraulic efficiency met- fully natural flow conditions. Unfortunately, such con-
ric λp is desirable when the RTD’s peak concentra- ditions are difficult to control and replicate. Because
tion is the most cost-effective characteristic of the controlled RTD comparisons are needed for wetland
tracer study to calculate. Similarly, the minimum travel studies, the concerns caused by using irrigation water
time, tm , which can be affected by both mixing scale were outweighed by the control advantage of the ex-
and short-circuiting, is a measurement that should not perimental design.
stand alone. Because of the different implications of Data from the natural dye tracer studies (Table 2)
each of these aspects of hydraulic efficiency, future re- appear to support the general trends of the controlled
searchers may want to consider these effects separately, study (Fig. 5) in response to depth change, but there
as this and other studies (Ta and Brignal, 1998) have are many factors that complicate this interpretation. For
done. example, the three inlets (Fig. 3) all conveyed runoff
during storm events, whereas only one inlet conducted
4.4. Limitations of this study water during the artificial storm events. When there are
multiple inlets, each inlet’s flow contribution to the final
Although the number of repetitions in this experi- RTD is weighted by its proportion of flow conducted
ment is large compared to many RTD studies of wet- (Keller and Bays, 2000). For example, the tracer added
lands, sample size for this experiment was nonethe- during low water level to the storm flow inlet may have
less small in a statistical sense. The possibility can- been initially circumvented by flow from the nursery in-
not be rejected that the differences in flow rate in let, which was flowing strongly before the storm event.
this experiment do have some effect on the RTD. The This likely increased the measured centroid and mini-
different dispersal effects discussed earlier may not mum travel time, and in doing so could have artificially
have been evident under the flow rates used in this inflated other measurements of hydraulic efficiency. It
experiment. was also clear from the noise in the high water level
The irrigation water, with an average temperature natural event that there may have been some instru-
of 14 ◦ C, was cooler than the average ambient summer mentation problems (Fig. 7) since the time series was
wetland water temperature of 20 ◦ C, and the average extremely noisy.
summer storm flow temperature of 23 ◦ C. Although
the initial flushing of the wetland minimized this tem-
perature difference between the irrigation water and 5. Conclusion
previous water in the wetland, the conditions may not
have been representative of what would occur during a This study demonstrates that changing water levels
natural storm. This effect was subjectively confirmed can significantly affect the flow characteristics through
by thermal stratification made visible by the dye tracer. a wetland. The magnitude of depth change in this study,
The same stratification was not visible when dye was a factor of 2.4, sets a baseline for identifying hydrologic
added during natural storm events with comparatively changes that are likely to affect the RTD. This study
warmer runoff. The centroids for all tracer tests using also indicates that minor fluctuations in flow rate, at
irrigation water were less than unity, indicating short- least within the factor of 2.7 tested in this study, may
circuiting was occurring under all conditions (Fig. 5C). not significantly affect the RTD. This supports analy-
Thermal stratification probably accounted for some of sis of normalized RTDs across various flow rates, such
this short-circuiting. The remaining short-circuiting is as Werner and Kadlec’s (1996) RTD characterization
likely a result of the proximity of the inlet to the out- method for pulsed-flow systems. Future researchers
let (Fig. 3). Because of the factors unique to this ex- may wish to further investigate factors affecting RTD
perimental setup, these results cannot necessarily be stability, such as the hydrologic influences investigated
extrapolated to natural flow conditions. The naturally in this study.
run tracer studies (Fig. 7) confirm this by having no- This study also indicates that water depth can af-
ticeably different RTD characteristics (Table 2) than fect hydraulic efficiency. The observed 57% increase
the controlled studies (Fig. 5). Ideally, an experiment in RTD spread and 12% loss of effective wetland
J.F. Holland et al. / Ecological Engineering 23 (2004) 189–203 203

volume during a 140% increase in depth would cause a Kadlec, R.H., Knight, R.L., 1996. Treatment Wetlands. CRC Press,
relatively small detriment to treatment performance. Boca Raton, FL.
The longer retention time created by the depth in- Keller, C.H., Bays, J.S., 2000. Tracer studies for treatment wetlands.
In: Proceedings of the INTECOL Invited Papers Symposium 13
crease, however, may result in a relatively larger ben- (Part 3) on Constructed Wetlands for Wastewater and Stormwater
efit to treatment performance. Although the depth ef- Applications, pp. 173–181.
fect on hydraulic efficiency may be small, it should Konyha, K.D., Shaw, D.T., Weiler, K.W., 1995. Hydrologic design
be weighed with other factors in wetland design and of a wetland – advantages of continuous modeling. Ecol. Eng. 4
management. For example, measures to increase hy- (2), 99–116.
Koskiaho, J., 2003. Flow velocity retardation and sediment retention
draulic efficiency, such as constructed baffles (Ta and in two constructed wetland-ponds. Ecol. Eng. 19 (5), 325–337.
Brignal, 1998), may have higher material costs in a Lawler, D.F., Benjamin, M.M., in press. Water Quality Engineering:
deeper wetland, but their greater need would offset their Physical-Chemical Treatment Processes, McGraw-Hill, New
higher cost. Future experiments should test whether York.
this depth effect on hydraulic efficiency is consistent Levenspiel, O., 1972. Chemical Reaction Engineering, 2d ed. Wiley,
New York.
under different conditions and in different wetlands. Lin, A.Y.C., Debroux, J.F., Cunningham, J.A., Reinhard, M., 2003.
Comparison of rhodamine WT and bromide in the determination
of hydraulic characteristics of constructed wetlands. Ecol. Eng.
20 (1), 75–88.
Acknowledgements Mitsch, W.J., Gosselink, J.G., 2000. Wetlands, 3rd ed. Wiley, New
York.
We would like to thank Noel Cressie for his as- Nepf, H.M., 1999. Drag, turbulence, and diffusion in flow through
emergent vegetation. Water Resour. Res. 35 (2), 479–489.
sistance and advice. We are also appreciative of the
Nepf, H.M., Koch, E.W., 1999. Vertical secondary flows in sub-
design and construction work done by Dan Gill, Tim mersed plant-like arrays. Limnol. Oceanogr. 44 (4), 1072–1080.
Salzman, and Alex Daughtery. For the technical as- Nepf, H.M., Sullivan, J.A., Zavistoski, R.A., 1997. A model for dif-
sistance of Chris Gecik, Kevin Duemmel, and Carl fusion within emergent vegetation. Limnol. Oceanogr. 42 (8),
Cooper, we are greatly indebted. Many thanks also go 1735–1745.
Oldham, C.E., Sturman, J.J., 2001. The effect of emergent vegetation
to Mark Schmittgen for his assistance on the farm, to
on convective flushing in shallow wetlands: scaling and experi-
Chris Keller for his advice on using dye tracers, and to ments. Limnol. Oceanogr. 46 (6), 1486–1493.
Mark Benjamin for supplying a draft of his publication Persson, J., Somes, N.L.G., Wong, T.H.F., 1999. Hydraulics effi-
in press. This study would not have been possible with- ciency of constructed wetlands and ponds. Water Sci. Technol.
out funding from the Ohio Agricultural and Research 40 (3), 291–300.
Shilton, A.N., Prasad, J.N., 1996. Tracer studies of a gravel bed
Development Center and generous donations from Agri
wetland. Water Sci. Technol. 34 (3–4), 421–425.
Drain Corporation. Stern, D.A., Khanbilvardi, R., Alair, J.C., Richardson, W., 2001. De-
scription of flow through a natural wetland using dye tracer tests.
Ecol. Eng. 18 (2), 173–184.
Ta, C.T., Brignal, W.J., 1998. Application of computational fluid dy-
References namics technique to storage reservoir studies. Water Sci. Technol.
37 (2), 219–226.
ASTM, 2002. Standard test method for open-channel flow measure- Thackston, E.L., Shields, F.D., Schroeder, P.R., 1987. Residence time
ment of water with thin-plate weirs. In: Annual Book of ASTM distributions of shallow basins. J. Environ. Eng. ASCE 113 (6),
Standards. ASTM, Philadelphia, PA, pp. 590–597. 1319–1332.
AWWARF, 1996. Tracer Studies in Water Treatment Facilities: A Walker, D.J., 1998. Modelling residence time in stormwater ponds.
Protocol and Case Studies. American Water Works Research Ecol. Eng. 10 (3), 247–262.
Foundation, Denver, CO. Werner, T.M., Kadlec, R.H., 1996. Application of residence time
Batzer, D.P., Resh, V.H., 1992. Wetland management strategies that distributions to stormwater treatment systems. Ecol. Eng. 7 (3),
enhance waterfowl habitats can also control mosquitoes. J. Am. 213–234.
Mosquito Control Assoc. 8 (2), 117–125. Wilson, J.F., Cobb, E.D., Kilpatrick, F.A., 1986. Fluorometric Pro-
Kadlec, R.H., 1994. Detention and mixing in free-water wetlands. cedures for Dye Tracing. U.S. Geological Survey, Denver, CO.
Ecol. Eng. 3 (4), 345–380. Wong, T.H.F., Somes, N.L.G., 1995. A stochastic approach to de-
Kadlec, R.H., 2000. The inadequacy of first-order treatment wetland signing wetlands for stormwater pollution control. Water Sci.
models. Ecol. Eng. 15 (1–2), 105–119. Technol. 32 (1), 145–151.

You might also like