Hydraulic Response and Nitrogen Retention in Bioretention Mesocosms with Regulated Outlets: Part...

Hydraulic Response and NitrogenRetention in Bioretention Mesocosms

with Regulated Outlets:Part I—Hydraulic Response

William C. Lucas1,*, Margaret Greenway1

ABSTRACT: In bioretention systems used for stormwater treatment,

runoff interception improves with increased infiltration rates. However,

nitrogen retention improves with increased retention time or decreasing

infiltration rates. These contrasting responses were analyzed in 240-L

experimental mesocosms using a variety of media treatments. The

mesocosms were vegetated, except for one barren control. Dual-stage

outlets were installed to extend retention time and equalize hydraulic

responses. One unregulated treatment was free-draining. This part 1 paper

presents the media properties and hydraulic responses. The highly

aggregated media had saturated hydraulic conductivities ranging from

20.7 to 59.6 cm/h in August 2008 (austral winter), which increased to 42.8

to 110.6 cm/h in March 2009 (austral summer). The outlet regulated

mesocosms provided retention over 8 times longer than the free-draining

mesocosms, while still being able to capture large events. The outlets

provide adaptive management for bioretention design to improve both

runoff capture and nitrogen retention. Water Environ. Res., 83, 692 (2011).

KEYWORDS: bioretention, media, particle size distribution, aggrega-

tion, hydraulic conductivity, infiltration, hydraulic controls, temperature.

doi:10.2175/106143010X12780288628697

Introduction

Bioretention systems are a widely used stormwater control

measure (SCM). Recent design guidance for bioretention media

recommends the use of a loamy sand media to provide rapid

infiltration rates to improve runoff capture and minimize bypass

flow (New Jersey Department of Environmental Protection,

2009). Column studies show that, even at high infiltration rates,

both filtration and rapid sorption processes can be very effective at

retaining contaminants, such as total suspended sediments,

phosphorus, and metal cations (Hsieh and Davis, 2005). However,

biogeochemical nitrogen transformations are less effective with

a shorter retention time (Kadlec and Wallace, 2009). Rapidly

draining bioretention systems are relatively ineffective for

nitrogen retention under field stormwater loading rates (Hatt et

al., 2008).

The hydraulic performance of bioretention SCMs can vary

considerably over time. Hatt et al. (2008) observed that non-

vegetated (barren) sandy loam column experiments with initial

hydraulic conductivities ranging from 27 to 253 cm/h eventually

declined to 21 to 75 cm/h. The decline was correlated with the

extent of compaction. The effect of compaction upon reducing

infiltration rates is well-documented in the pedotransfer functions

(PDFs) developed in the soils literature (e.g., Rawls et al., 1983).

On the other hand, plant roots penetrate confining layers and

open up soil structure, increasing infiltration rates (Bartens et al.,

2009; Gilker et al., 2002). In vegetated bioretention columns,

Culbertson and Hutchinson (2004) reported that switchgrass

(Panicum virgatum) increased infiltration rates in bare soils from

0.5 to 128 cm/h. The resultant infiltration rates in deep rooted

native grass hedges are much higher than found in adjacent

shallow rooted crops planted in the same soil (Blanco-Canqui et

al., 2004; Rachman et al., 2004; Seobi et al., 2005).

As the fine particles in urban runoff pass into bioretention

systems, they are ‘‘strained’’ by the bioretention media, so most of

these particles are captured within short distances within the

media. As a result of this particle accumulation in the media,

hydraulic conductivity declines to values in the range of several

centimeters per hour (Li and Davis, 2008). The resultant increase

in ponding time allows finer particles to settle out and accumulate

on the surface of the medium as a ‘‘cake’’ deposit (Hatt et al.,

2005; Li and Davis, 2008). This deposit can reduce overall

hydraulic conductivity from between 3 and 11 cm/h (Li and Davis,

2008) to as low as 0.1 cm/h (Hatt et al., 2005). Experiments by

both Hatt et al (2008) and Li and Davis (2008) showed that

removal of this deposit with the top 2- to 3-cm clogging layer

could restore infiltration rates. Clark and Pitt (2009) also observed

a 3-cm clogging depth; however, removal of the surface 1 to 2 cm

did not restore infiltration rates to original values. These

contrasting observations may reflect differences in media

properties, sediment loading rates, and depth of clogging layer

removed.

The effects of vegetation and clogging have been documented

in a field study of 21 bioretention systems in Australia that had

been operational from 6 months to 3 years (Le Coustumer et al.,

2009). They found that infiltration rates in systems using initially

‘‘rapid’’ media (25 cm/h median initial infiltration rates) declined

over time toward a median of 10 cm/h, presumably as a result of

clogging. On the other hand, systems with initially ‘slow’ media

(1.0 cm/h) increased to 3.0 cm/h, presumably as a result of the

presence of vegetation-induced macropores. Le Coustumer et al.

(2009) hypothesized that vegetation partially mitigates clogging

1 Griffith University, Nathan, Queensland, Australia.

* 3 Lucas Lane, Malvern, PA 19355; e-mail: [email protected].

692 Water Environment Research, Volume 83, Number 8

because plant stems break up the surface ‘‘cake’’. Media

augmented with coarse fibers similar to coir peat do not seem to

clog as readily (Hatt et al., 2008), and the more rapidly infiltrating

the media, the longer it lasts before it clogs (Clark and Pitt, 2009).

These results indicate that facilities with rapid (coarser) media

maintain higher infiltration rates than facilities using slower

(finer) media.

As a result of these interactions between texture, compaction,

clogging, and vegetation, infiltration rates of bioretention systems

inevitably change over time. When infiltration rates decline,

nitrogen removal processes are likely to improve, as a result of

enhanced retention time; however, overall interception volumes

decrease, because less flow is intercepted (Davis et al., 2005;

Lucas and Greenway, 2008). On the other hand, when infiltration

rates are too fast, nitrogen removal is much less effective, as part

II of this paper demonstrates.

One solution to both increase nitrogen retention and prevent

clogging is to provide media with high infiltration rates and then

use an outlet to regulate flowrates. However, with low effective

infiltration rates created by an outlet, overall capture would be

reduced, because peak flows during large runoff events bypass the

facility (Davis et al., 2005). To address these conflicting

objectives, we investigate the potential for using a dual-outlet

configuration, in which the treatment capability provided by

media with high saturated hydraulic conductivity is used to

rapidly filter large events, while outlets are used to provide more

retention time for small events comprising the majority of annual

runoff. We report on the hydraulic response of various media

treatments initially evaluated for phosphorus retention (Lucas and

Greenway, 2011b) and the resulting outlet-adjusted infiltration/

percolation rates.

The objective of our experiment was to manipulate the outlet

controls to improve nitrogen retention while retaining high runoff

capture efficiency. In this paper, part 1, the resultant hydraulic

properties are evaluated in a hydraulic model to estimate retention

time responses and interception volumes in a variety of runoff

events. Part II of this paper presents the resulting nitrogen

retention performance of the outlet system described in this paper.

Methods

Experimental Setup. The experiments were conducted at the

Loganholme Water Pollution Control Centre, located 25 km south

of Brisbane, Australia. The experiment comprised the bioretention

mesocosms described in Lucas and Greenway (2011b). The

mesocosms were made from 240-L ‘‘wheelie-bin’’ containers with

a mean surface area of 0.25 m2 and 99 cm deep. They were

roughened with sandy paint to minimize preferential flow down

the sides. To provide a uniform flow path through the media, the

indented bottom 14 cm was filled with gravel and overlaid by

coarse sand as a separation layer. Each media treatment was

planted with native vegetation (Lucas and Greenway, 2011b). One

barren treatment (K20nv) was used to control for vegetation

effects. All mesocosms were covered by a gravel mulch 1.0 cm

deep to reduce evapotranspiration losses and to prevent scour from

the high application rates. Figure 1a shows the experimental

setup.

Media Properties. The media matrix consisted of turf sand

with a particle size distribution (PSD) of 81.2% between 300 and

Figure 1—(a) Experimental setup, August 2008. Vegetated and barren mesocosms with compound outlets andcollection chambers. Newer systems on right side. (b) Compound outlets, showing how flows from bottom are pipedup to crossover tee (in circle) before discharge down into collection chambers. Low-flow outlet is the 4-mm tubing (insquare) shunted between vertical piping.

Lucas and Greenway

August 2011 693

420 mm, 13.2% between 420 and 1000 mm, and 3.7% over 1 mm.

There was no silt or clay and only 0.3% organic matter (Lucas and

Greenway, 2008). The following media amendments were added

to the sand in varying proportions: Krasnozem soils, neutralized

Red Mud, and alum-based water treatment residuals (WTRs)

(Lucas and Greenway, 2011b). These different media treatments

were formulated according to the ratios presented in Table 1. Coir

peat was added at 12% by volume to all media for increased water

holding capacity and infiltration performance.

Six treatments (3 replicates each) were constructed in January

2007 (experiment 1). Four treatments incorporated Krasnozem

soils, and two treatments incorporated Red Mud. Three additional

treatments using Al-WTRs were constructed 6 months later, of

which two were blended with Krasnozems (experiment 2). In

addition to the seven media components listed in Table 1, a

K10/40 treatment comprised a 20-cm lift of 40% Krasnozem

under 40 cm of 10% Krasnozem media, in anticipation that the

bottom layer with a lower permeability would control the

saturated infiltration response, while the upper layers would

initially infiltrate rapidly (Hsieh and Davis, 2005). The resulting

seven media treatments comprised K20, K10/40, K40, RM06,

RM10, WTR-K, and WTR30. The media proportions were

measured by volume and mixed with an 80-L cement mixer, to

ensure thorough mixing, and placed in 20-cm lifts. Media was

placed in the mesocosms to a depth of 70 6 5 cm to allow for

settlement. The settled media depths ranged between 60 and

67 cm, with 18 to 25 cm of freeboard.

All of the amendments contain a high proportion of clay

minerals. Krasnozem soils comprise 50 to 70% clay minerals

(Sills et al., 1974), and Red Mud comprises 40% clay, 50% silt,

and 10% sand (Snars, 2003). Coagulated from dissolved organic

compounds and clay colloids, WTRs could be classified as

entirely clay. If dispersed throughout the media, the high clay

content of these amendments would result in very low

permeability. However, Krasnozem soils are noted for their high

aggregate structure, which provides very good drainage properties

for soils with such high clay content (Isbell, 1994; Sills et al.,

1974). The dry aggregate percentage ranges from 45 to 80%, and

most are stable under wet-sieving, which is indicative of high

aggregate stability (Carter et al., 2002). The clay fraction of Red

Mud also is aggregated entirely into silt-sized particles (Snars,

2003). Likewise, WTRs form stable aggregates that have to be

ground to form sand-sized particles (Babatunde et al., 2008).

The presence and persistence of these media aggregates has a

significant effect on hydraulic responses. To examine the extent of

media aggregation over the 18-month duration of experiment 2

and 24-month duration of experiment 1, core samples were taken

from 10-cm depths from each replicate of the RM10 and K40

treatments, and from 50 cm depth of the WTR30 treatment, and

blended together for textural analysis. The PSD was evaluated

with a Beckman Coulter Multisizer (Brea, California) for particle

sizes in the range from 1 to 600 mm. Because our focus was on

hydraulic responses, aggregated clays were not dispersed

physically or chemically. Such minimally dispersed measure-

ments more accurately represent the PSD, which determines

hydraulic properties (Leys et al., 2005). Table 2 and Figure 2 give

the PSDs in the original media, as compared with that observed in

the sampled cores after the experiment.

Hydraulic Dosing. An average of 56 cm of tertiary treated

effluent was applied on a weekly basis to accelerate nutrient loads

for the experiments reported in Lucas and Greenway (2011b) and

part 2 of this paper (Lucas and Greenway, 2011a). The small

amount of total suspended solids (TSS) occasionally observed in

the clarified effluent comprised readily biodegradable organic

material. As such, TSS accumulation was not anticipated to be a

factor affecting hydraulic responses in this experiment. The

hydraulic measurements used synthetic stormwater comprising

nutrients dissolved in tap water. Experimental applications were

distributed at a rate of 90 64 L/h through 25-mm tubing by use of

0.6-mm diameter sprayers, supplied by a large-diameter, dual-

manifold distribution system to diminish differential pressure

losses. Dosing rates were manipulated by a pressure regulating

valve on the supply line. The dosing used for hydraulic

measurements applied from 133 to 155 L (56 to 62 cm) to the

Table 1—Media composition of various blends. K = Krasnozem, RM = Red Mud, and WTR = water treatment residuals.Mineral proportions are by weight, while coir peat proportion is by volume.

Media Label Turf Sand Topsoil Krasnozem Red Mud WTR Coir peat

K10 93% - 7% - - 12%

K20 86% - 14% - - 12%

K40 70% - 30% - - 12%

RM06 75% 20% - 5% - 12%

RM10 71% 20% - 9% - 12%

WTR30 80% - - - 20% 12%

WTR-K 71% - 20% - 9% 12%

Table 2—Media PSDs (mm) between 1 and 600 mm, minimally dispersed; initial media and as of March 2009 (after24 months for RM10 and K40 in experiment 1 and 18 months for WTR30 in experiment 2).

Period As installed March 2009

Particle size

Clay

,2 mm

Silt 2 to

63 mm d10 d50 d90 Clay ,2 mm

Silt 2 to

63 mm d10 d50 d90

RM10 0.1% 2.4% 181 308 481 0.2% 12.6% 50 241 419

K40 0.0% 8.8% 70 280 446 0.2% 15.5% 35 281 443

WTR30 0.0% 1.6% 166 309 496 0.0% 2.4% 164 332 513

Lucas and Greenway


mesocosms over periods ranging from 90 to 110 minutes. For

comparison to bioretention facilities designed with a 25:1

impervious-source-to-treatment-area-capture ratio, the 56-cm

application is equivalent to 22 mm of runoff. Accounting for

3 mm of initial abstraction, this would be expected from a 25-mm

rainfall event at an average rainfall intensity of 17 mm/h for

90 minutes. This rate is nearly two-thirds of the 90-minute

intensity of 28 mm/h allocated for an annual recurrence interval

event for Brisbane (Queensland Ministry for Natural Resources

and Water, 2000). Assuming this quite high capture ratio, our

stormwater dosing regime represents a large event that occurs

perhaps twice per year. If the capture ratio were less, this relative

event size would be that much greater. It was our intention to test

the performance of our mesocosms under relatively intense events

where bypass flow is likely to occur.

Outlet Controls. Figure 1b shows the experimental outlet

configuration. The outlet control system regulates outflows by

using a lower outlet sized to extend retention time, with a larger

diameter outlet placed above it to convey flows ponded during

simulated runoff. The 2-mm-diameter lower outlet is set 18 cm

above the bottom of the mesocosm to provide a permanent

internal saturated zone intended to promote denitrification, while

the 10-cm-diameter upper outlet is set at or slightly below the

surface of the media. The lower outlet restricted flow to a rate of

8.1 cm/h when the ponding depth was 10 cm, which occurred at a

head of 72 6 5 cm. This flowrate provided identical percolation

responses once the surface ponding had drained.

The upper outlet controlled flows as a function of head across

the media according to Darcy’s Law. Generally close to the media

surface, the upper outlet regulated outflow when the systems were

Figure 2—Coulter multisizer PSDs of (a) RM10 (at 10 cm), (b) K40 (at 10 cm), and (c) WTR30 (at 50 cm) treatments.Initial media before experiment and media as of March 2009.

Lucas and Greenway

August 2011 695

saturated and ponding had occurred. This outlet comprised a 10-

mm riser pipe extending upward to a tee, where flows crossed

over at the level, passing into another tee before descending to the

final discharge, as shown in Figure 1b. The tees provided vacuum

breaks to prevent siphon flows. The outlet elevation controlling

flow through the media was measured at the top of the crossover

pipe, as the crossover pipe typically flowed full. By adjusting its

elevation to provide more or less head in response to media

hydraulic conductivity, this arrangement was intended to provide

similar ponding and drainage responses in replicates containing

media with varying permeabilities. This outlet configuration also

allows for adaptive management as flowrates change over time.

One treatment (WTR-Knr) was left as free discharge to compare

the effect of outlet controls on nitrogen retention. Including the

barren control, nine different treatments were examined.

Hydraulic Measurements. Hydraulic measurements were

conducted during the stormwater dosing applications in the

austral winter of August 2008 (n 5 2) and austral summer of

March 2009 (n 5 5). Measurements of ponding depth were taken

during the filling phase in August 2008 runs. The resulting

infiltration rate Q was computed as follows:

Q~Qi{(p2A2{p1A1)

tð1Þ

Where

Qi 5 inflow rate (m3/h),

A1 5 average ponding area at first interval (m2),

A2 5 average ponding area at second interval (m2),

p1 5 initial ponding depth (m),

p2 5 second ponding depth (m), and

t 5 elapsed time (hours).

Until ponding occurs, eq 1 reduces to Qi, as the infiltration rate is

limited by the application rate.

Saturated hydraulic conductivity (Ksat) was determined by

falling head measurements after the stormwater applications

ended. This was accomplished by closing off the lower outlet so

that flows were constrained to passing through the upper outlet.

Ponding depth to the nearest millimeter was observed at 10-

minute intervals for at least 40 minutes. Saturated hydraulic

conductivity (Ksat) then is computed as adapted from Bedient and

Huber (1988):

Ksat~AtL

Amtln

h1

h2

� �ð2Þ

Where

At 5 average ponding area (m2),

Am 5 average media cross-sectional area (m2),

L 5 media depth (m),

h1 5 initial head (m),

h2 5 final head (m), and

t 5 elapsed time (hours).

The area adjustments allow for rates to be corrected for the surface

area variations resulting from the tapered geometry of the

mesocosms. These corrections were based on rating curves

developed to determine surface area as a function of elevation.

Based on Darcy’s Law, infiltration rate Q at a single point in

time during the falling head measurements is as follows:

Q~Ksat|(pzd)

Lð3Þ

Where

p 5 ponding depth (m), and

d 5 outlet depth (m) below the media surface (m).

Head h is the sum of p and d. Equation 3 was used to determine

infiltration rates after inflow ceased and the lower outlet was

closed. Because falling head measurements were not possible

without an elevated upper outlet to control flows, no Ksat

measurements were made on the free-discharge WTR-Knr

replicates (numbers 25 to 27) until outlets were installed for the

last two measurements.

Results and Discussion

Media Aggregation. The PSDs for minimally dispersed R10,

K40, and WTR30 media are given in Table 2. Respectively, these

media would be classified as 9, 30, and 21% clay if they had been

dispersed fully. However, the clay fraction in the RM10 media

and WTR30 was only 0.1%, and no clay was observed in the K40

treatment. Instead, the majority of the amendments comprised

sand-sized aggregates with only a minor amount of fines,

dominated by silt particles. The median particle size was no

more than 310 mm in all treatments. Compared with the sand

matrix with a larger median PSD, this indicates that the majority

of media aggregates in the sand fraction were less than 310 mm.

Figure 2 presents the PSD of the original RM10, K40, and

WTR30 media, as compared with the media after 2 years

(experiment 1) and 18 months (experiment 2) after hydraulic

measurements were taken in March 2009. In this case, it can be

seen that there was a substantial shift from sand-sized aggregates

into the silt fraction in the RM10 treatment. There was a less

pronounced disaggregation trend in the Krasnozem treatment. On

the other hand, the WTR30 treatment suggests an opposite trend

of aggregation into larger particle sizes. These changes are

represented by the respective changes in clay and silt fractions

shown in Table 2. While there was only a slight increase in the

clay fraction, there was substantial increase in the silt content of

the RM10 and K40 treatments. This also is confirmed by a

corresponding shift in the d10, d50, and d90 particle sizes in the

RM10 and in the d10 size in the K40 treatment. In contrast, there

was no change in the d10 of the WTR30 treatment, while the d50

and d90 sizes increased slightly. These results indicate that most of

the amendments originally were, and still remained as, aggregates,

albeit more dispersed in the case of Krasnozem and Red Mud

amendments.

In addition to the structure provided by the aggregates, fibers in

the coir peat shrink and swell in response to moisture status, also

opening pores. It would be expected that the presence of such

pores with fissures between aggregates would promote Ksat. Note

that the effects of aggregates on soil structure are not addressed by

conventional approaches to determine hydraulic conductivity

using PDFs based on fully dispersed analyses that do not capture

the effects of aggregation and coarse organic matter (Rawls et al.,

1983).

Hydraulic Conductivity Observations. Table 3 presents the

means and coefficient of variation (CV) of saturated hydraulic

conductivities across the treatments at the end of August 2008 and

6 months later in February to March 2009. The mean Ksat values for

Lucas and Greenway


these treatments were remarkably high for media comprising such a

high proportion of clay materials. The unvegetated K20nv treatment

was the slowest media in August 2008, with a mean Ksat of 21 cm/h.

The most rapidly draining media was K10/40, with an average Ksat

of 60 cm/h. The anticipated slower drainage rates expected from an

impeding layer comprising nearly 30% clay did not occur. The

average Ksat for the other media ranged from 28 to 46 cm/h. The

K20nv and WTR treatments were the most consistent between

replicates, with the coefficient of variation at or below 21%. The

K10/40, K20, RM06, and RM10 treatments had coefficients of

variability from 33 to 59%, highlighting the variability in infiltration

rates, even with ostensibly identical plants and media.

To evaluate differences in Ksat between the K20 treatment and

the corresponding barren K20nv treatment, log-transformed

observations were analyzed with the t-test with unequal variances.

The difference between the two observations in August 2008 was

significant (p 5 0.034), while the difference between the four

observations in March 2009 was significant (p 5 0.0002). These

data suggest that plants significantly improve Ksat.

The K40 treatment had an extremely high coefficient of

variation approaching 120%. Bin 6 had a very high Ksat over

109 cm/h, bin 5 had a Ksat of 19 cm/h, and bin 4 had a Ksat of only

10.3 cm/h. We believe that the low rate of bin 4 was the result of

accidentally dousing the first lift during construction with flows of

such high pressure that the media was partly homogenized,

dispersed into the underlying stone, and then puddled. The very

high rate in replicate 6 is remarkable for the same media

comprising soil with nearly 30% clay minerals.

Further confirmation of the variability in treatment response is

evident in both the high coefficient of variation between replicates

in most treatments and the marked changes in means shown in

Table 3. Figure 3 graphically illustrates the changes in Ksat in the

replicates between the two measurement periods. Figure 3b shows

that, just 6 months later, Ksat markedly increased in all replicates

except K10/40 in bin 17. Excluding this replicate, Ksat increased

in the other two replicates by over 110%. All other treatments

increased by at least 54%, with an average increase of over 74%

and a maximum of 107%. Note that the unregulated free discharge

control treatments (bins 25 to 27) had a very high Ksat of

approximately 110 cm/h.

The paired two-sample t-test for means was used to compare the

individual replicates over the two runs in August 2008 and March

2009. The increase in Ksat was highly significant, with probability

well below 0.01 for all treatments except for the WTR30 and RM06

treatments. The increase in these treatments was still significant at a

probability ,0.05. The narrow standard deviation displayed in

Figure 3 indicates that the infiltration rates were quite consistent

over each observation period. The Ksat in the barren K20nv

treatment also increased over time. Given the absence of vegetation

in the K20nv treatment, it appears that plant-associated processes

are not responsible for this increase in infiltration rates.

As a result of decreases in viscosity, Ksat increases approxi-

mately 2% for each degree increase in temperature. Applying the

mean daily temperature to the period of observations, the mean

temperature would be 13.5uC in August 2008 (austral winter) and

26.3uC in February to March 2009 (austral summer). Using

standard temperature correction factors relative to 20uC, these

temperatures result in adjustment factors of 1.135 and 0.850,

respectively. The relative difference between these two values is

33%. Compared with observed increases from 54 to 107%,

temperature effects on viscosity would explain less than half of

the difference observed in most treatments. Measurements of the

WTR30 replicates taken in April 2010 showed an average

decrease of 9% from the March 2009 observations, which was

not significant. This suggests that the trend of increasing Ksat does

not persist. It may be that the initially low measurements had been

biased by an accumulation of TSS in the form of organic matter

flocs in the tertiary effluent at the time of measurement.

Braga et al. (2007) reported that permeability varied several-fold

as a function of temperature. As a result of both viscosity and

presumed permeability changes, the difference between maximum

and minimum Ksat varied over 300%. The overall temperature

response was fitted by the following relationship (Braga et al., 2007):

K~0:0072Tz0:0196 ð4Þ

Where

T 5 temperature (uC).

Comparing the results from eq 4 in relative terms using the

temperatures for each season in our experiment leads to an

Table 3—Mean and variation in saturated hydraulic conductivity (cm/h), as derived from falling head saturatedhydraulic conductivity measurements. Measurements from August 2008 compared with February to March 2009.Percent increase is comparison of means. WTR-Knr is the unregulated system, which was measured only in March2009. Significance of K20 means compared with K20nv indicated by asterisks: *(p,0.05) and **(p,0.001).

Treatment

August 2008 February to March 2009Treatment

percent increase

Paired sample

P-statisticMean CV Mean CV

RM06 36.2 52% 55.6 28% 54% 0.01023

RM10 26.4 27% 48.4 28% 83% 0.00030

K20 42.6* 59% 78.0** 44% 83% 0.00081

K20nv 20.7 9% 42.8 19% 107% 0.00062

K10/40 59.6 33% 85.6 33% 44% 0.00003*

K40 46.1 119% 72.9 70% 58% 0.00002

WTR30 42.4 20% 69.7 43% 64% 0.01303

WTR-K 27.7 21% 56.5 36% 104% 0.00593

WTR-Knr Not available Not available 110.6 5% Not available Not available

* Excludes replicate 17, which was the only replicate in which conductivity was reduced. Including this replicate, the P-statistic increased to

0.12008.

Lucas and Greenway

August 2011 697

increase of 78%—a value quite similar to the average increase

noted in our observations.

Outlet Regulated Observations. Regardless of the extent of

disaggregation during the experiment (Table 2 and Figure 2), the

high aggregate content in the media provided remarkably high

Ksat values in February to March 2009. The Ksat exceeded 50 cm/h

in almost 75% of the replicates and exceeded 95 cm/h in almost

30% of the replicates. Unregulated, such a high Ksat value

provides only a very short retention time. While seasonal

variations in Ksat may amplify variability in outflow responses,

the retention time afforded by even the slowest replicate is

measured in minutes, as discussed below.

Using regulated outlets to equalize flow responses, Figure 4a

shows the infiltration rates of all 24 replicates, while Figure 4c

shows the corresponding ponding depths during the rising and

falling head regimes on August 17, 2008. Figures 4b and 4d

display the corresponding responses of the K40 mesocosms. This

treatment had the greatest variation between replicates. Even with

more than a 10:1 difference between the lowest and highest Ksat,

the dual outlet configuration generated quite similar ponding and

drainage responses. The initial infiltration rate is understated,

because observations are limited by the application rate.

Infiltration rates then decreased over the first hour, as the media

became saturated. Once inflow rates exceeded outflows through

the lower outlet at saturation, ponding then increased after the first

hour, but more slowly as the upper outlets began to convey more

flow. As a result of the contribution of flows through the upper

outlet and the increasing head on both outlets, infiltration rates

increased to a maximum, until inflow ceased.

Notwithstanding this variability, infiltration rates were similar

through the majority of the observation period, as shown in

Figure 4a. Figure 4c presents the corresponding ponding eleva-

tions that resulted in the infiltration rates displayed in Figure 4a.

Because of differences between the highest and lowest Ksat values,

the ponding depths were higher for the lower Ksat replicates, while

they were lower for the highest Ksat replicate, as would be

expected by Darcy’s Law. This is more clearly displayed in

Figure 4d.

Figure 5a displays the ponding depth during the final

measurements on March 15, 2009. After several iterations of

upper outlet elevation adjustments to reduce variability in ponding

depth, the outlets provided nearly identical drainage responses. As

a result, infiltration rates displayed in Figure 5b show less

coefficient of variation than the results from 6 months earlier

shown in Figure 4a. Note that, because the draining limb does not

include the lower outlet contribution, the overall infiltration rate is

reduced by approximately 8 cm/h. Added to the upper outlet flows

observed, the resulting coefficient of variation would be reduced

even more. Because the lower outlet rate is slower than media

Ksat, this phase of the drainage response is also identical for all

replicates. As such, these outlet regulated systems can provide

similar retention times throughout the entire drainage response,

even in media with widely varying Ksat values. This is significant

in reducing effect of retention time upon nitrogen processes.

Hydraulic Responses. The orifice equation controls outlet

rate in the lower outlets as a function of coefficient C (typically

0.6), orifice area A (m2), and head h (m), as follows:

Q~CA|ffiffiffiffiffiffiffiffi2gh

pð5Þ

While the lower orifice should be treated as ‘‘culvert’’, inlet

controlled headloss equations are also proportional to h1/2. For

systems regulated by a single orifice at the bottom of the media, Q

is thus a function h1/2. Given that h 5 L + p 2 5 (or 72 6 5 cm), if

flows are to be constricted to approximately 8 cm/h when h 5 L

(e.g., saturated, but not ponded, conditions), increases in ponding

Figure 3—Saturated hydraulic conductivity from (a) two events in August 12–17, 2008, and (b) 5 events from February24–March 15, 2009. Means and standard deviation shown.

Lucas and Greenway


depth p represent only a relatively small increase in total head, the

effect of which is reduced even further by the half-power

relationship. Therefore, when flows through highly permeable

media are constricted enough to extend retention time, the

capacity to handle large flowrates without excessive ponding will

be restricted. As a result, substantial amounts of annual runoff

could be bypassed without any treatment in large events.

Measures that treat larger events by infiltration through the

media are far preferable to letting untreated flows bypass the

systems. Given its shallow outlet depth d, the head through the

upper outlet is much more responsive to ponding depth p, because

it responds as a function of h1 according to Darcy’s Law, and the

relative head increases rapidly. Compared with the effects of p on

a constricted lower outlet as function of h1/2, this ability to treat

high flows through the upper outlet is particularly important in

systems with limited ponding depth, such as tree planters.

Modeled Responses. Retention time is a parameter of

considerable importance for nitrogen removal processes (Kadlec

and Wallace, 2009). Plug flow is the most appropriate metric for

retention time, as it represents the average of individual retention

times for each increment of inflow, taking into account antecedent

storage provided by the 18-cm-deep internal saturated zone. The

computer program HydroCAD (HydroCAD Software Solutions

LLC, Chocorua, New Hampshire) offers explicit computations of

retention time for design storm events. HydroCAD has been used

successfully for the design of bioretention systems for urban

Figure 4—Infiltration rates (cm/h) and ponding depths (cm), August 17, 2008: (a) infiltration rates—all replicates, withstandard deviation; (b) infiltration rates—K40 replicates; (c) ponding depths—all replicates, with standard deviation;and (d) ponding depths—K40 replicates. Elapsed time from beginning of dosing.

Figure 5—(a) Ponding depth (cm) and (b) infiltration rates (cm/h) with lower outlets closed during falling head Ksat

measurements. All replicates during March 15, 2009, run. Elapsed time from beginning of falling head measurements.

Lucas and Greenway

August 2011 699

retrofits with outlets similar to that described in this paper (Lucas,

2010).

HydroCAD was used to simulate the setup used in this

experiment to reproduce observed results and compute the

resulting retention time. The system is differentiated into nodes

that represent pond, media, and stone layers, with the latter

incorporating the outlet controls. Each node has a narrow

‘‘dummy’’ column extending into the adjacent layer. This

arrangement permits flows to be modeled as if flowing

horizontally according to the head between nodes, even though

flows are actually vertical. While this simplification results in

instantaneous outflow from the media, it occurs at very low rates.

Given that outflows from the free discharge control treatment

began far sooner than the advance of wetting front would be

expected, such simplification may not be unrealistic.

Darcy’s Law then controls flow through the media as a function

of head from the surface pond to that created by outlet array. It is

represented by a coefficient based on Ksat times area divided by

media depth, applied as a rating curve. Flows between these nodes

then are routed according to standard hydraulic algorithms. Refer

to Lucas (2010) for more detailed discussion of the procedures

used to model the system. Even with a time step of 0.4 seconds,

oscillations occur in such systems as a result of the small size of

the dummy columns, so the modeled system was scaled up by a

factor of 10. Given the observations of March 15, 2009, the model

was based on media with a Ksat of 50 cm/h. Table 4 presents the

input parameters used in the model.

Figure 6a presents the response of an outlet regulated system

compared with a free discharge system in Figure 6b. Figure 6a

displays the media hydraulic grade (equivalent to pond elevation at

saturation) and the corresponding infiltration rate to the media.

Initial infiltration rates are as fast as application rates, but, once the

media is saturated, as evidenced by its hydraulic grade meeting the

surface, infiltration rates decline rapidly, because they are now

regulated by the outlet configuration. Compared with the infiltration

response displayed in Figure 5a, these ponding and infiltration

trends are similar, indicating that the model provides a reasonable

representation of actual responses. Figure 6a also displays the

resulting partition of flows through the outlets, showing how the

upper outlet starts to flow after half an hour and stops once ponding

is over. In this manner, the peak flows that would otherwise bypass

the system also are treated. Meanwhile, the lower outlet continues to

run until the media and stone are drained completely to field

capacity, or roughly 8 hours in our mesocosms.

Figure 6b displays the same responses for the free-discharge

system. Absent regulation by the outlet, infiltration rates exceed

the application rate for the entire event, so no ponding occurs.

This is indicated by the media hydraulic grade line remaining

below the media surface. Media outflows still are regulated by

Darcy’s Law, so outflows increase as the media partially fills.

Because Ksat is high enough that outflows match inflows at this

hydraulic grade, ponding would not occur, unless the application

rate was increased. The media then drains down within 45 minutes

after inflow ceases. This response corresponds to what we have

observed for the unregulated control treatment.

In the modeled representation, plug-flow retention time in the

outlet-regulated media was 108.4 minutes, with another 49.1 min-

utes required to pass through the stone. This provides a total

media/stone retention time of 157.5 minutes. Compared with the

same mesocosm treatments without outlets, the corresponding

retention times were 17.3 and 1.5 minutes, respectively, for a total

retention time of 18.8 minutes. Thus, the outlet-regulated

mesocosms provided over 8 times as long a retention time. This

considerably improved nitrogen retention performance in the

outlet-regulated mesocosms is discussed in part II of this paper

(Lucas and Greenway, 2010, 2011a).

Conclusions

To provide rapid infiltration performance in bioretention

systems, typical recommendations for media composition suggest

that no more than 2 to 5% be in the clay fraction (New Jersey

Department of Environmental Protection, 2009). In determining

the clay fraction, typical PSD analyses use both mechanical and

chemical dispersant methods to dissolve clay aggregates. While

this method may represent the hydraulic implications of easily

dispersed clays, this method does not characterize the aggregation

found in the media that we selected for phosphorus sorption

properties (Lucas and Greenway, 2011b). The extent of aggrega-

tion noted in all three media amendments resulted in very high

Ksat rates, even with quite high clay fractions. While there was

some disaggregation as the silt fraction increased over time, very

few clay-sized particles were observed.

Comparison of the planted treatment to its corresponding barren

treatment showed that a significant increase in Ksat was associated

with the presence of plants. There was also a substantial increase

in Ksat in summer compared with winter. This temperature effect

exceeded that attributable to viscosity alone. Furthermore,

because Ksat also increased in the barren treatment, this increase

cannot be attributable to macropores associated with plant root

formation. It is possible that other processes may be involved,

such as bias by accumulation of bacterial flocs or other responses

resulting in changes in intrinsic permeability. Further research into

this phenomenon is warranted.

Within many treatments, there was a very high coefficient of

variation in Ksat of the replicates. To minimize artifacts in

retention performance introduced by such variations in flow

responses, an innovative dual-stage outlet configuration was

Table 4—Routing parameters used in the HydroCADmodel.

Pond area m2 2.53

Pond elevation m 0.76

Media area m2 2.24

Media Ksat m/h 0.50

Flowrate L/s 0.31111

Media depth m 0.62

Darcy flow coefficient 0.00050

Stone depth m 0.14

Stone area m2 2.00

Free discharge flow

Orifice elevation m 0.00

Orifice diameter cm 1

Number of orifices 10

Controlled discharge flow

Upper elevation m 0.70

Upper orifice cm 1.0


Lower elevation m 0.18

Lower orifice cm 0.18


Lucas and Greenway


developed to manage flows. This configuration provides an

opportunity for each SCM system to be engineered more precisely

for its particular distribution of flow volumes, durations, and

frequencies. With judicious selection of design parameters, the

outlets can be configured to detain all small flow events, while

still treating large events by filtration through the media.

Our experiment demonstrates that media with high infiltration

rates offer the potential to effectively intercept even quite large

events, given the high hydraulic conductivities shown in Table 3.

Combined with the extended retention time afforded by the low

flow controls, this arrangement offers a highly effective treatment

system for runoff events of all magnitudes when Ksat is high, as is

the case with the present media.

The ability to lower the upper outlet and increase the

diameter of the lower outlet provides a consistent infiltration

rate in the face of inevitable clogging as the systems

accumulate sediments. Eventually, the media will have to be

renovated by removal of the clogging layer, after which, the

outlets can be reconfigured. This capability for adaptive

management is an emerging trend in the design of stormwater

controls. This also has important implications for the design of

bioretention systems to reduce nitrogen loadings to surface and

groundwater.

Credits

We extend our thanks to the staff of the Loganholme Water

Pollution Control Centre (Logan, Queensland, Australia) for

providing the meteorological data, space, and infrastructure to run

the experiments. We also thank the following students of Griffith

University who so ably assisted in the hydraulic observations:

Alana Scott, Jennifer-Leigh Campbell, Alison Atkinson, and

Wendy Tang. Craig Strong provided key assistance in the

interpretation of the particle size distribution results.

Submitted for publication April 7, 2010; revised manuscript

submitted September 10, 2010; accepted for publication Septem-

ber 15, 2010.

Figure 6—Flow (L/s) and elevation (m) responses: (a) outlet controlled system, and (b) free discharge system.Elapsed time from beginning of dosing.

Lucas and Greenway

August 2011 701

ReferencesBabatunde, A. O.; Zhao, Y. Q.; Yang, Y.; Kearney, P. (2008) Reuse of

Dewatered Aluminium-Coagulated Water Treatment Residual to

Immobilize Phosphorus: Batch and Column Trials Using a Con-

densed Phosphate. Chem. Eng. J., 136, 108–115.

Bartens, J.; Day, S. D.; Harris, J. R.; Dove, J. E.; Wynn, T. M. (2008)

Can Urban Tree Roots Improve Infiltration through Compacted

Subsoils for Stormwater Management? J. Environ. Qual., 37, 2048–

2057.

Bedient, P. B.; Huber, W. C. (1988) Hydrology and Floodplain Analysis;

Addison Wesley: Reading, Massachusetts.

Blanco-Canqui, H.; Gantzer, C. J.; Anderson, S. H.; Alberts, E. E.;

Thompson, A. L. (2004) Grass Barrier and Vegetative Filter Strip

Effectiveness in Reducing Runoff, Sediment, Nitrogen, and Phos-

phorus Loss. Soil Sci. Soc. Am. J., 68, 1670–1678.

Braga, A.; Horst, M.; Traver, R. G. (2007) Temperature Effects on the

Infiltration Rate Through an Infiltration Basin BMP. J. Irrigat.

Drainage Eng., 133 (6), 593–601.

Carter, M. R.; Skjemstad, J. O.; MacEwan, R. J. (2002) Comparison of

Structural Stability, Carbon Fractions and Chemistry of Krasnozem

Soils from Adjacent Forest and Pasture Areas in South-Western

Victoria. Aust. J. Soil Res., 40, 283–298.

Clark, S. E.; Pitt, R. (2009) Solids Removal in Storm-Water Filters

Modeled Using a Power Equation. J. Environ. Eng., 135 (9), 896–

909.

Culbertson, T. A.; Hutchinson, S. L. (2004) Assessing Bioretention Cell

Function in a Midwest Continental Climate. Presented at the 2004

ASABE Annual International Meeting, Ottawa, Ontario, Canada, Aug.

1–4; American Society of Agricultural and Biological Engineers: St.

Joseph, Michigan.

Davis, A. P.; Shokouhian, M.; Sharma, H.; Minami, C. (2005) Water

Quality Improvement Through Bioretention Media: Nitrogen and

Phosphorus Removal. Water Environ. Res., 78, 284–293.

Gilker, R. E.; Weil, R. R.; Krizek, D. T.; Momen, B. (2002) Eastern

Gamagrass Root Penetration in Adverse Subsoil Conditions. Soil Sci.

Soc. Am. J., 66, 931–938.

Hatt, B. E.; Siriwardene, N.; Deletic, A.; Fletcher, T. D. (2005) Filter

Media for Stormwater Treatment and Recycling: The Influence of

Hydraulic Properties of Flow on Pollutant Removal. Presented at the

10th International Conference on Urban Drainage, Copenhagen,

Denmark, Aug. 21–26; Institute of Environment and Resources,

Technical University of Denmark and Section of Environmental

Engineering, Aalborg University: Copenhagen, Denmark.

Hatt, B. E.; Fletcher, T. D.; Deletic, A. (2008) Hydraulic and Pollutant

Removal Performance of Fine Media Stormwater Filtration Systems.

Environ. Sci. Technol., 42, 2535–2541.

Hsieh, C. H.; Davis, A. P. (2005) Evaluation and Optimization of

Bioretention Media for Treatment of Urban Storm Water Runoff. J.

Environ. Eng., 131, 1521–1531.

Isbell, R. F. (1994) Krasnozems-A Profile. Aust. J. Soil Res., 32, 915–929.

Kadlec, R. H.; Wallace, S. (2009) Treatment Wetlands, 2nd ed.; CRC

Press: Boca Raton, Florida.

Le Coustumer, S.; Fletcher, T. D.; Deletic, A.; Barraud, S. (2009)

Hydraulic Performance of Biofilters for Stormwater Management:

First Lessons from Both Laboratory and Field Studies. Water Sci.

Technol., 56 (10), 93–100.

Li, H.; Davis, A. P. (2008) Urban Particle Capture in Bioretention Media.

I: Laboratory and Field Studies. J. Environ. Eng., 134 (6), 419–432.

Leys, J.; McTainsh, G.; Koen, T.; Moonmey, B.; Strong, C. (2005) Testing

a Statistical Curve-Fitting Procedure for Quantifying Sediment

Populations within Multi-Modal Particle-Size Distributions. Earth

Surf. Proc. Landforms, 30, 579–590.

Lucas, W. (2010) Design of Integrated Bioinfiltration-Detention Urban

Retrofits with Design Storm and Continuous Simulation Methods. J.

Hydrol. Eng., 15 (6), 486–498.

Lucas, W. C.; Greenway, M. (2011a) Hydraulic Response and Nitrogen

Retention in Bioretention Mesocosms with Regulated Outlets: Part

II—Nitrogen Retention. Water Environ. Res., 83, 704–714.

Lucas, W. C.; Greenway, M. (2010) Nitrogen Retention in Bioretention

Mesocosms Using Outlet Controls to Improve Hydraulic Perfor-

mance. Proceedings of the World Environmental and Water

Resources Congress 2010: Challenges of Change, Providence, Rhode

Island, May 16–20; American Society of Civil Engineers: Reston,

Virginia, 3038–3047.

Lucas, W. C.; Greenway, M. (2008) Nutrient Retention in Vegetated and

Nonvegetated Bioretention Mesocosms. J. Irrigat. Drainage Eng.,

134, 613–623.

Lucas, W. C.; Greenway, M. (2011b) Phosphorus Retention by Bioretention

Mesocosms Using Media Formulated for Phosphorus Sorption:

Response to Accelerated Loads. J. Irrigat. Drainage Eng., 137, 144–154.

New Jersey Department of Environmental Protection (2009) NJDEP Best

Management Practice Manual. Chapter 9.1—Bioretention Systems.

New Jersey Department of Environmental Protection, Division of

Watershed Management: Trenton, New Jersey, http://www.state.nj.

us/dep/stormwater/tier_A/pdf/NJ_SWBMP_9.1%20print.pdf (ac-

cessed April 4, 2010).

Queensland Ministry for Natural Resources and Water (2000) Queensland

Urban Drainage Manual (QUDM), Appendix A Supplement to

QUDM; Queensland Ministry for Natural Resources and Water:

Brisbane, Australia.

Rachman, A.; Anderson, S. H.; Gantzer, C. J.; Thompson, A. L. (2004)

Influence of Stiff-Stemmed Grass Hedge Systems on Infiltration Soil

Sci. Soc. Am. J., 68, 2000–2006.

Rawls, W. L.; Brakensiek, D. L.; Miller, N. (1983) Green-Ampt

Infiltration Parameters from Soil Data. ASCE J. Hydraulics Div.,

109 (1), 62–70.

Seobi, T.; Anderson, S. H.; Udawatta, R. P.; Gantzer, C. J. (2005)

Influence of Grass and Agroforestry Buffer Strips on Soil Hydraulic

Properties for an Albaqualf. Soil Sci. Soc. Am. J., 69, 893–901.

Sills, J. D.; Aylmore, L. A. G.; Quirk, J. P. (1974) Relationship Between

Pore Size Distributions and Physical Properties of Clay Soils. Aust. J.

Soil Res., 12, 107–117.

Snars, K. E. (2003) The Effects of Bayer Process Red Mud on the P

Retention and pH of Soils. Doctoral Thesis, University of Western

Australia, Perth, Australia.

Lucas and Greenway


Hydraulic Response and Nitrogen Retention in Bioretention Mesocosms with Regulated Outlets: Part...

Documents

Transcript of Hydraulic Response and Nitrogen Retention in Bioretention Mesocosms with Regulated Outlets: Part...