Академический Документы
Профессиональный Документы
Культура Документы
Abstract: Bioretention cells are an emerging low impact development technology that address urban stormwater runoff
concerns. Field and column experiments were conducted to assess the efficacy of bioretention cells in cold conditions. Field
experiments in a prairie environment demonstrated a significant decrease (91.5%) in effluent volumes compared to influent
volumes. The majority (⬃60%) of the runoff percolated to the surrounding soils or evapotranspirated. Cold condition
performance significantly impacted high volume events and was characterized by significantly higher effluent volumes,
significantly lower runoff storage, higher effluent peak flow rates, and longer peak delays. A partially frozen surface layer caused
the changes in performance. Long-term simulation experiments on the columns indicated a significant decrease in saturated
hydraulic conductivity over the first 4 equivalent years of operation, before levelling to a constant value.
Key words: urban runoff, bioretention, LID, cold climate, hydraulic conductivity, prairie hydrology.
Résumé : Les cellules de biorétention représentent une technique émergeante de développement à faible impact qui aborde
les questions d’écoulement des eaux pluviales en zone urbaine. Des expériences sur le terrain et en colonnes ont été
réalisées pour évaluer l’efficacité des cellules de biorétention sous des conditions froides. Les expériences dans une prairie
ont démontré une diminution marquée (91,5 %) des volumes d’effluent par rapport aux volumes d’influent. La plus grande
partie (60 %) du ruissellement a percolé dans les sols environnants ou s’est dissipée par évapotranspiration. Le rendement
sous des conditions froides a grandement affecté les événements à fort volume et se caractérisait par des volumes
For personal use only.
d’effluent considérablement plus élevés, un entreposage du ruissellement considérablement plus faible, des débits de pointe
à l’effluent plus élevés et de plus longs délais pour atteindre la pointe. Une couche partiellement gelée à la surface a
modifié le rendement. Des expériences de simulation à long terme dans les colonnes ont indiqué une diminution importante
de la conductivité hydraulique saturée au cours de l’équivalent des quatre premières années de fonctionnement avant
qu’elle ne se stabilise à une valeur constante.
Mots-clés : ruissellement en milieu urbain, biorétention, LID, climat froid, conductivité hydraulique, hydrologie des
prairies.
[Traduit par la Rédaction]
1. Introduction Thorolfsson 2004; Hsieh and Davis 2005; Sharkey and Hunt
2005; Davis 2008). The primary aims of LID are to reduce the
Urban stormwater runoff is an important and abundant volume of runoff generated and to capture and treat the runoff
water resource (Hsieh and Davis 2003). Uncontrolled, urban on-site. Natural processes are used to achieve these goals, e.g.,
stormwater runoff presents a variety of complex environmen- vegetation is used to reduce and disconnect impervious areas
tal concerns including increased risk of flooding and the po- and to increase infiltration, evapotranspiration, and storage.
tential to carry a variety of contaminants into water bodies Bioretention cells, a type of LID, are small, localized infil-
(Davis 2008; Delleur 1982). Low impact development (LID) is tration and treatment basins. They are plant and soil based
a site design strategy that aims to control and treat urban facilities, consisting of a basin filled with a highly porous
stormwater runoff at or near the source. Controls and struc- growing media layer that is topped with dense vegetation and
tures are implemented on-site such that the post-development mulch (Hsieh and Davis 2003; Kim et al. 2003). Urban storm-
hydrology and water quality is improved compared to pre or water runoff is routed to the cells, where it collects on the
undeveloped site conditions (US EPA 2001; Nordberg and surface, before infiltrating into the media. As the water passes
Received 11 May 2012. Revision accepted 28 September 2012. Published at www.nrcresearchpress.com/cjce on 30 October 2012.
U.T. Khan and C. Valeo. Department of Civil Engineering, University of Calgary, 2500 University Drive N.W, Calgary, AB T2N 1N4,
Canada; Mechanical Engineering, University of Victoria, P.O. Box 3055, Station CSC, EOW Building, Room 223, Victoria, BC V8W 3P6,
Canada.
A. Chu. Department of Civil Engineering, University of Calgary, 2500 University Drive N.W, Calgary, AB T2N 1N4, Canada.
B. van Duin. Infrastructure Planning, Water Resources, The City of Calgary, Mail Code #409, Manchester Water Centre, 625 25 Avenue S.E,
Calgary, AB T2G 4K8, Canada.
Corresponding author: C. Valeo (e-mail: valeo@ucalgary.ca and valeo@uvic.ca).
Written discussion of this article is welcomed and will be received by the Editor until 31 March 2013.
Can. J. Civ. Eng. 39: 1210 –1221 (2012) doi:10.1139/l2012-110 Published by NRC Research Press
Khan et al. 1211
through the media, it undergoes various physical, chemical, drology. In these regions, design guideline revisions may be
and biological processes that remove contaminants associated required to account for altered hydraulic characteristics, such
with urban runoff (Hsieh and Davis 2005). The runoff can then as frozen soil, higher snowmelt volumes, and lower runoff
exit the cell via an under-drain beneath the cell, percolate into intensity from snowmelt. Local conditions are important to
the surrounding subsoils, or a combination of both depending take into account because of the unique freeze–thaw cycles
on the configuration. and precipitation patterns. For example, in Calgary, Alberta,
The hydrological performance indicators of bioretention westerly winds during the winter known as Chinook winds
cells include the decrease of volume and peak flow rate be- temporarily raise temperatures to above freezing. These winds
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by "University of the Philippines, Diliman-" on 11/18/16
tween the influent and effluent, the delay of the effluent, basin cause Calgary to experience frequent freeze–thaw cycles
emptying time, and changes to the long-term surface and throughout the cold months, rather than one large spring melt
near-surface hydraulic conductivity. These indicators are in- as experienced in other regions. These unique conditions can
fluenced by the design of the bioretention cell (e.g., the pres- impact bioretention cell operation and hydrologic efficacy.
ence of an impermeable liner) and by the characteristics of the Various sources assert that filter based treatment systems
corresponding catchment area and runoff. Due to these differ- such as bioretention cells function suitably in cold weather
ences, reported efficacy of bioretention cells has been highly conditions and that there is minimal effect of frozen media on
variable and inadequately reported (WER 2007a). In the lit- hydraulic function (UNHSC 2005; Davidson et al. 2008;
erature, bioretention cells tested in both field and laboratory- Roseen et al. 2006, 2009). Impacts on various performance
scale experiments, have been able to capture (or mitigate) parameters have been reported, notably an increase in effluent
100% of the influent volume for small to medium size events, volume and peak flow rate and an increase in lag time (Hunt
but ranged from 33% to 80% for larger events (Traver and and Jarrett 2004; Muthanna et al. 2008; Roseen et al. 2009).
Prokop 2003; Ermilio and Traver 2006; Davis 2008; Hatt et al. The increase in effluent volumes was attributed to a decrease
2009). Similarly, large variability was also reported for peak in evapotranspiration in cold conditions; which resulted in a
flow rate reduction and delay. Peak reduction ranged from larger fraction of volume exiting the cells as runoff (Muthanna
19% to 96% (Sharkey and Hunt 2005; Hunt et al. 2008; Lewis
et al. 2007). Higher effluent peak flow rates were attributed
et al. 2008), while delay ranged from 90 min to up to 615 min
to channelization and comparatively less homogenously dis-
(UNHSC 2005; Muthanna et al. 2008). The causes for the
tributed soil, which in turn decreased the infiltration time
For personal use only.
Fig. 1. (a) A bioretention cell in Calgary, Alberta and (b) design 2.2. Column description
details of bioretention test facility used for the field experiments. Four scaled versions of the field cell were used to test the
performance under long-term operation. The columns mea-
sured 0.5 m by 0.5 m in surface area. Two short columns
(hereupon referred to as LS1 and LS2 for “long-term short”)
were 450 mm deep (360 mm of growing media and 90 mm of
under-drain aggregate) and two tall columns (hereupon re-
ferred to as LT1 and LT2 for “long-term tall”) were 800 mm
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by "University of the Philippines, Diliman-" on 11/18/16
installed to measure the depth of the water above the notch; the ature was between –5 and ⫹5 °C and snowmelt conditions
bottom of the notch was used as the datum and was calibrated were possible. The calculated precipitation depth, d, of the
before each experiment. The pressure and ultra-sonic sensors synthetic influent varied from 23.23 to 112.47 mm. The equiv-
were connected to a Sigma 900 MAX auto-sampler and Sigma alent intensities of the experiments ranged from 25.88 to
950 flow-meter, respectively, to collect the depth over time, 134.29 mm/h and the return periods were estimated to be 25,
and calculated the flow rate of the effluent. 50, 100, and 200 years.
The soil moisture and soil temperature were measured for
each experiment. Four Delta-T SM200 soil moisture sensors 2.3.2. Column experiments: long-term simulations
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by "University of the Philippines, Diliman-" on 11/18/16
were installed in the middle of the cell at depths of 150, 300, A synthetic stormwater runoff distribution system similar to
500, and 1000 mm below the surface. A Delta-T DL6 soil the field experiments was used for the long-term simulation
moisture logger was used to collect the data. The sensors were experiments on the four columns, LT1, LT2, LS1, and LS2.
accurate for operating temperatures between 0.1 and 40 °C. Stormwater was collected from the same City of Calgary
Two HOBO soil temperature sensors were installed at depths detention pond used for the field experiments and stored at the
of 150 and 500 mm below the surface. Both the soil moisture University of Calgary’s Hydraulics Laboratory. The stormwa-
and temperature sensors were programmed to measure contin- ter was then pumped to a mixing tank where sediment was
uously at 1 min intervals. added to simulate the average Calgary TSS concentration.
To calculate the equivalent depth of precipitation applied in Submersible pumps routed the runoff to the four columns via
the form of synthetic stormwater runoff, the size and charac- vinyl tubing. The runoff drained onto the columns through an
teristics of a hypothetical catchment had to be defined. For this apparatus designed to equally distribute the runoff on the
project, the area of the bioretention cell was defined to be surface.
equal to 10% of the total catchment area, which is a typical Equation [1] was used to estimate the equivalent volume of
estimate based on guidelines that recommend between 5% and influent for the 20 year period. In this case, d equalled to
20% (Hunt and Jarrett 2004; Lanarc Consultants et al. 2005). 400 mm, the annual precipitation in Calgary, AB was equal to
The area of the cell was 32 m2, making the corresponding 0.25 m2, and I/P was the same as the field experiments, 4. This
hypothetical catchment area 320 m2. Furthermore, a typical meant the equivalent runoff each column received for a sim-
catchment will consist of pervious and impervious areas with ulated year was 0.5 m3 or 10 m3 for 20 years. The volume
For personal use only.
the majority of the runoff generated from the impervious applied to the columns was calculated using the dimensions of
fraction. To account for this, a ratio known as the impervious the mixing tank and the initial and final water level in the tank.
to pervious ratio (I/P) was introduced (Brown 2007). The The runoff was applied to the columns periodically: approxi-
expression I/P is the ratio of the upstream impervious area (I) mately 0.140 m3 of runoff was pumped to each column in
to the area of the pervious bioretention cell (P) that receives 90 min. After the pumping ended, the columns drained for
the runoff. For this project, an I/P of 4 was selected to 60 min. This process was repeated three times a day; staged
represent typical residential areas. Thus, if the bioretention application of runoff was necessary to prevent excessive pool-
area is 32 m2, the upstream impervious area (“I”) is 128 m2 ing of runoff on the surface of the column. The total amount
(four times the bioretention area), which represents 40% of the of sediment applied on each column was approximately 5 kg,
total catchment area (320 m2 with the other 60% being pervi- which corresponds to a TSS concentration of 444 mg/L.
ous and assumed to generate no runoff). The relationship Though the application of a large volume of runoff over a
between runoff volume and precipitation depth can be formu- short period does not exactly mimic the natural maturation
lated as process and long-term operation, it can give a conservative
estimate and highlight the anticipated changes in the hy-
[1] V ⫽ d AB (I/P ⫹ 1) drological performance.
The saturated hydraulic conductivity (Ksat) of the growing
media was estimated using a Model 2800K1 Guelph per-
where V is the volume of influent from tanks (m3); d is the meameter. The initial Ksat was measured at 0 equivalent years
equivalent precipitation depth (m); AB is the bioretention cell (prior to the application of runoff) and after 3.8, 9.2, 15.6, and
area (m2); and I/P is the impervious to pervious ratio. Equation [1] 20 equivalent years (or after the application of 1.9, 4.6, 7.8,
takes into account the runoff generated from the impervious and 10 m3 of runoff). Two Ksat values were calculated on each
area in the catchment and also the precipitation that occurs on column at each interval, with a constant head of 10 cm and
the bioretention cell itself. Using this equation, d was calcu- 15 cm, respectively.
lated for each experiment using the volume of runoff applied The PSD of the growing media was analysed throughout the
from the two storage tanks. The average intensity of the experiments to understand the mechanisms behind the poten-
artificial storm event was equal to the d divided by the duration tial decline in surface infiltration and Ksat. Media samples were
of the event. This intensity and V was used to estimate the taken at three depths each from LT1 and LT2: 0 – 0.20 m,
return period of the simulated events using intensity– dura- 0.20 – 0.4 m, and 0.4 – 0.64 m; and at two depths from LS1 and
tion–frequency (IDF) curves for Calgary (CoC 2000). LS2: 0 – 0.18 m and 0.18 – 0.36 m. The samples were taken at
A total of 24 experiments were conducted on the field the same five intervals as the Ksat measurements.
bioretention cell site, each representing a distinct precipitation
event. A summary of the experiments is provided in Table 1. 2.4. Performance measures
The experiments were conducted in the spring (Sp), summer
(S), fall (F), and winter (W). Eight experiments, indicated by 2.4.1. Change in runoff volume
an asterisk, were conducted in cold weather conditions, during The influent and effluent volumes for the field experiments
or proceeding Chinook conditions, when the mean air temper- were calculated for each experiment, with the exception of five
Precip. Return
Experiment Air Influent depth Duration Intensity period
ID Date temp. °C volume (L) (mm) (min) (mm/h) (years)
S1 7/17/08 14 8 004 50.03 116 25.88 100
S2 8/6/08 22 7 520 47.00 21 134.29 100
S3 8/15/08 20 7 736 48.35 36 80.58 100
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by "University of the Philippines, Diliman-" on 11/18/16
experiments, where no effluent drained to the outlet manhole growing media 12 h after peak soil moisture (m3); and L is the
(see Section 3.1 for details). The influent volume was calcu- other losses (m3). The variable Sg is defined as the difference
lated using the initial and final depths of water in the storage between the initial and “final” volume of water held in the
tanks and multiplying this difference by the cross-sectional growing media. The initial volume is calculated using the soil
area of each tank. The effluent volume was calculated by moisture measured immediately prior to an event. The final
multiplying the measured flow rate in the outlet manhole by volume is calculated using the soil moisture measured 12 h
the time step, and adding these values for the entire duration of after the occurrence of peak soil moisture during the event.
flow into the manhole. The change in runoff volume was The volume of water stored was calculated by multiplying the
calculated using difference in initial and final moisture in each sensor by the
representative depth and summing the volumes for each sen-
[2] DV ⫽ 1 ⫺共 Ve
Vi 兲
⫻ 100%
sor. This calculation assumes that the measurement by each
soil moisture sensor is representative of the surface area of the
cell at each depth. The term L represents the combined losses
from evapotranspiration and percolation into the soil surround-
where ⌬V is the change in runoff volume between inlet and
outlet of the bioretention cell (%); Ve is the volume of ing the bioretention cell, i.e., the volume of runoff not ac-
effluent (m3); and Vi is the volume of influent (m3). A value counted for between Ve and Sg.
of ⌬V less than 100% indicates a decrease in runoff volume
2.4.2. Change in flow rate
between the inlet and outlet, or a lower volume of effluent
The change in peak and centre-of-mass (CoM) were calcu-
than influent.
lated using the following equation:
A mass balance analysis of the runoff was conducted to
determine the fate of the artificial runoff applied to the cell.
The mass balance of the runoff was calculated using
[3] Vi ⫽ Ve ⫹ Sg ⫹ L
[4] DQ ⫽ 1 ⫺ 共 Qe
Qi 兲
⫻ 100%
where Vi is the volume of influent and Ve is the volume of where ⌬Q is the change in either the peak or CoM flow rate
effluent, as in eq. [2]; Sg is the volume of water stored in reduction (%); Qe is the effluent peak or CoM flow rate
Fig. 2. Different shapes of the influent and effluent hydrographs from four experiments illustrate the need for a normalized measurement of
delay. The DF is calculated using ti – ts and to – ts, depicted in (a). Note that these experiments have a similar Vi and peak delay (30 or
31 min) but distinct values of DF.
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by "University of the Philippines, Diliman-" on 11/18/16
For personal use only.
(mm/h); and Qi is the influent peak or CoM flow rate (mm/h). to peak (or time to CoM) of the effluent and the time to peak
An analysis of CoM was included along with the traditional (or time to CoM) of the influent.
analysis of peak flow rates to provide an additional measure of
influent hydrographs with multiple peaks (caused by the pump t e ⫺ ti
mechanism). [5] DF ⫽
ti ⫺ ts
2.4.3. Delay factor
In addition to analysing the change in the magnitude of
the peak or CoM flow rate, the delay in the occurrences of where DF is the delay factor; te is the time of effluent peak or
these flow rates were also analysed. The delay was defined CoM; ti is the time of influent peak or CoM; and ts is the start
as the difference between the times at which Qi and Qe time of the experiment. Note that this differs from the tradi-
occur. However, analysing the delay without considering tional “lag time” since both, te and ti are measured from the
the start time of the event can be misleading. This is due to same point, ts, the time at which the experiment began. The
a number of reasons; first, for events with a similar duration functionality of DF is illustrated in Fig. 2.
and runoff volume, the shape of the influent hydrographs 3. Results and discussion
varied widely with peaks occurring at different times rela-
tive to start and total duration of the event. Second, by 3.1. Change in runoff volume
focusing on peak or CoM flow rate delay, the difference in The change in runoff volume, ⌬V for all field experi-
flow rates and volumes (indicators of the return period) are ments is shown in Fig. 3; the average ⌬V for the 24 field
not accounted for, thus making it hard to compare the delay experiments was 91.5%, i.e., a significant decrease in runoff
between events. Finally, the influence of antecedent soil volume between the influent and effluent. Additionally, ⌬V
moisture conditions in the cell affects the timing and shape was 99.9% for seven experiments (bolded rows in Table 1)
(start, duration, and peak) of the effluent hydrograph. Thus that had precipitation depths of 30.36 mm or less. Apart from
a delay factor, DF, was defined, which normalizes the time F3 (⌬V ⫽ 99.9%) and Sp3 (⌬V ⫽ 99.8%), no effluent (i.e.,
⌬V ⫽ 100%) drained to the outlet manhole for the remaining through the permeable liner, were much higher than evapo-
5 experiments listed above. In Calgary, 96% of all precipita- transpiration. This result has important implications for biore-
tion events are smaller than 32 mm (WER 2007b). This tention design as it indicates that the majority of “captured”
demonstrates that bioretention cells with similar configura- runoff is in fact lost to the surrounding soils. If this bioreten-
tions can capture the majority of precipitation and resulting tion cell was constructed with an impermeable liner, the ef-
runoff, when sized to 10% of the catchment area in regions fluent volume could theoretically be as high as ⬃70%
with similar hydrology. (L ⫹ Ve). Thus, if the objective of a specific bioretention cell
The average ⌬V was 93.5% for warm conditions and was application is to decrease runoff volume, a cell with a perme-
87.5% for cold conditions. This difference was not significant able liner would perform better than one with an impermeable
at the 95% confidence level. However, if the seven experiments liner.
with precipitation depths less than 32 mm are excluded from the Comparing the two weather conditions, Sg was signifi-
analysis (i.e., events that produce minimal to no outflow) then cantly lower in cold conditions (p ⫽ 0.043); i.e., a lower
there is a statistically significant difference (p ⫽ 0.006) between volume of runoff stored in the growing media in cold
the warm (⌬V ⫽ 91%) and cold (⌬V ⫽ 80%) weather conditions. conditions. Other losses, L were not significantly different
Thus, in cold conditions there is reduced capacity to capture or between warm and cold conditions (p ⬎ 0.05). These results
decrease the runoff volume from large, high intensity and low suggest that in cold conditions, less runoff volume is being
frequency events. held in the growing media (Sg) and the volume portion
A summary of the results from the mass balance analysis leaving via the outlet pipe is larger (⌬V).
is shown in the insert in Fig. 3. The results show each
component of the mass balance equation (eq. [3]) as a per- 3.2. Peak and centre-of-mass flow rate reduction
centage of Vi. Soil moisture data was not logged for five and delay
events, W1 through S5 (due to a soil temperature below 0.1 °C Average peak Qi was 85.6 mm/h and average peak Qe was
or sensor malfunction), so these events were excluded from 3.2 mm/h. Average CoM Qi was 34.7 mm/h and average CoM
this analysis. Due to these omissions, the percentage Ve/Vi is Qe was 0.6 mm/h. The peak and CoM change in flow rate
higher than the results shown for ⌬V above. For the remaining reduction and delay are summarized in Table 2. On average,
19 experiments, approximately 59% of the influent was ob- the peak and CoM ⌬Q for all experiments were 95.3% and
served to be “other losses” (L) and 31% was stored in the 97.5%, respectively (excluding events where no effluent
growing media (Sg). Decrease in runoff volume via Sg is minor drained to the manhole). On average, both peak and CoM ⌬Q
compared to L. On average, L was calculated to be 30 mm over were lower in cold conditions but not significantly different
the 12 h period, suggesting that losses to the surrounding soils, (p ⫽ 0.203 and p ⫽ 0.271).
Table 2. Summary of changes in peak and CoM flow rate, delay, surface. Thus, in cold weather conditions, even though the
and DF for field experiments. initial soil moisture was similar, the maximum amount of
volume held was lower near the surface. Also, analyzing soil
⌬Q ⌬Q Peak CoM moisture readings 12 h after the peak soil moisture (i.e., the
Experiment (peak) (CoM) delay delay Peak CoM “final soil moisture”), shows that the moisture levels at the
ID (%) (%) (min) (min) DF (⫺) DF (⫺) 150 mm and 1000 mm deep sensors were significantly higher
S1 88.9 93.6 31 57 1.38 1.95 in warm conditions (p ⫽ 0.046 and p ⫽ 0.021 for the 150 mm
S2 99.0 99.3 15 28 1.75 3.33 and 1000 mm sensors, respectively). This means that either a
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by "University of the Philippines, Diliman-" on 11/18/16
S3 98.6 98.9 30 38 2.30 3.11 relatively larger volume of runoff is stored at these locations in
S4 96.5 98.8 22 41 2.29 3.41 warm conditions or that these locations are receiving less total
F1 98.6 99.3 23 43 2.35 3.69 runoff in cold conditions. The reasons for this phenomenon are
F2 97.8 99.3 30 50 1.45 3.27 discussed below. It is also important to note that there were no
F3 100.0 100.0 90 112 11.00 9.00 significant differences in moisture values for the 300 mm and
W1* — — — — — — 500 mm deep sensors.
Sp1* — — — — — —
Sp2 97.4 99.1 26 51 2.44 4.64 3.4. Cold climate hydrologic characteristics
Sp3 99.9 100.0 76 72 11.29 11.29 Cold condition experiments were characterized by higher
S5 96.8 98.0 12 41 1.32 2.58 peak and CoM Qe, higher Ve and lower Sg, as well as the soil
S6 97.2 98.9 34 48 2.89 3.67 moisture trends described above; these differences are illus-
S7* — — — — — — trated in Fig. 4 for experiments S6 and W3 as examples. The
S8* — — — — — — differences indicate that the partially frozen top layer (mea-
S9* — — — — — — sured to a depth of 150 mm by the soil temperature sensor)
S10 90.5 92.5 46 90 1.43 2.17 altered the flow path of the runoff through the bioretention cell
F4 84.8 97.1 46 52 5.18 2.86 media. The runoff is hypothesized to be short-circuiting and
W2 96.5 97.1 61 62 7.78 3.21 travelling through preferential flow paths and macro-pores to
W3 93.1 94.8 69 73 18.25 3.81 the outlet pipe, bypassing or not fully utilizing the available
For personal use only.
W4 96.6 97.6 31 66 1.66 3.00 media mass in the deeper areas of the cell. This is typical of
W5 89.9 94.6 38 70 2.03 3.12 drainage in the prairies during the winter months (Granger
S11 96.6 97.5 38 60 2.83 3.61 et al. 1984; Gray et al. 2001; van der Kamp et al. 2003; Fang
S12 92.4 96.2 71 75 19.50 3.78 et al. 2007).
Averages As the runoff is applied to the cell (Fig. 4), in cold
All 95.3 97.49 42 59.42 5.22 3.97 conditions the frozen top layer does not allow the water to
Warm 96.2 97.79 35 53.38 4.09 3.88 penetrate the growing media as it does in warm conditions.
Cold 93.5 96.84 56 72.50 7.65 4.17 The runoff does not disperse evenly and tends to travel later-
ally before finding a route down towards the perforated section
Note: *No effluent drained to the monitoring manhole for these five of the under-drain pipe. This explains the higher peak flow rate
experiments so the listed parameters could not be calculated.
(since the runoff is wetting a smaller volume of soil), longer
peak delays (due to a decrease in permeability because of the
The average delay in peak flow rate was 42 min; with a partially frozen layer), and the lower volume held in the
significantly (p ⫽ 0.039) shorter delay in warm conditions growing media (bypassing effect). It is important to note that
compared to cold conditions (35 vs. 55 min). Similarly, the since the 300 and 500 mm deep sensors do not show any
average delay in CoM flow rate was 59 min; with a signifi- difference between warm and cold conditions, the bioretention
cantly (p ⫽ 0.049) shorter lag in warm conditions compared to media itself is not altering the hydrologic response in cold
cold conditions (53 vs. 72 min). The average peak DF for all weather conditions. Rather boundary effects — specifically
experiments was 5.22; 4.10 in warm conditions and 7.65 in the surface boundary that is subject to freezing conditions —
cold conditions. The CoM DF for all experiments was 3.97; appears to affect the flow paths. This phenomenon is believed
3.88 in warm conditions and 4.17 in cold conditions. This to be unique to prairie regions experiencing frequent freeze–
supports the results from the peak and CoM delay, where there thaw cycles and its characterization improves the understand-
is a longer delay in cold weather conditions. However, using ing of bioretention operation in these conditions.
the normalized measure of delay, DF, indicates that the dif-
ferences are not significant (p ⫽ 0.188 for peak DF and 3.5. Long-term simulation experiments
p ⫽ 0.861 for CoM DF). The saturated hydraulic conductivity, Ksat, for each column
is shown in Fig. 5. The figure shows the average results for LT1
3.3. Soil moisture trends and LT2 and for LS1 and LS2. Over the testing period, Ksat for
Comparing soil moisture data between warm and cold con- LT decreased from 2.25 ⫻ 10⫺5 to 1.98 ⫻ 10⫺6 m/s, a
ditions, there was no statistical difference between initial soil significant reduction of 91.2% (p ⫽ 0.017). The Ksat for LS
moisture in the growing media at each of the four depths. This decreased from 1.91 ⫻ 10⫺5 to 1.24 ⫻ 10⫺6 m/s, a significant
indicates that the amount of initial storage capacity was un- reduction of 93.3% (p ⫽ 0.015). For both sets of columns, Ksat
changed in the two weather conditions. However, the peak soil decreased rapidly over the first 4 equivalent years of runoff
moisture was significantly lower (p ⫽ 0.020) in cold condi- application. After this, the rate of decline decreased and was
tions for only one location: the sensor 150 mm below the effectively a constant value. The Ksat initially drops due to a
Fig. 4. A comparison of effluent flow rate, soil moisture (measured at a depth of 150 mm and 1000 mm from the surface), and runoff
movement through the growing media (schematic not-to-scale), for a warm (S6) and cold (W4) climate experiment. Note that the partially
frozen layer was measured to a depth of 150 mm.
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by "University of the Philippines, Diliman-" on 11/18/16
For personal use only.
combination of compaction (due to settling of the growing the overall reduction in saturated hydraulic conductivity of the
media and runoff application) and initial clogging. The clog- columns. This analysis also demonstrates how the majority of
ging in the columns was limited due to the lack of formation TSS is captured by the columns, which is responsible for the
of a crust layer. This contributed to keeping Ksat relatively large decrease in TSS concentration seen in the water quality
constant. The depth of the columns did not have a significant component of this study (Khan et al. 2012).
impact on Ksat. These results indicate that the hydraulic con- It is important to note that the rate of decrease in con-
ductivity of the columns was unaffected by the depth of the ductivity and sediment capture in this experiment is accel-
media and remained high without any maintenance after con- erated compared to what might be experienced on a field
tinuous operation. This can potentially reduce the cost of scale. There was no maintenance conducted on the columns
bioretention cell construction and operation, due to the lower and the plants did not mature at the same rate as the runoff
volume of growing media and lower frequency of maintenance application. As a result, the effects of the root function are
required. not fully accounted for as in the field experiments. Thus,
The PSD analysis of the growing media in the columns at this procedure gives a conservative estimate in terms of
different depths concluded there was effectively no change in hydrologic performance.
the distribution below a depth of 0.20 m in each column. The
change in distribution for the 0 – 0.20 m deep layer was char- 4. Conclusions
acterized by a 50% increase in particles between 50 and
200 m and a 50% decrease in particles 0.1 and 50 m. There The hydrological performance of the field bioretention cell
were no changes to the distribution for particles smaller than was high in both warm and cold conditions. The field exper-
0.1 m and larger than 200 m. This shows that the sediment iments show an average change in runoff volume (⌬V) of
from the synthetic stormwater runoff (the majority of which is 91.5% and a reduction of peak flow rate of 95.3%, further
between 50 and 200 m) was primarily captured in the top supporting the use of bioretention cells as an important tech-
20 cm for all columns. This indicates that the capture of the nology for stormwater control. Cold condition performance
sediment is due to surface filtration; as additional depth did not was characterized by significantly lower ⌬V (only for exper-
have any impact on particle capture (depth filtration). The iments with a precipitation depth larger than 32 mm), lower
capture of these particles on the surface is also responsible for peak and CoM flow rate reduction, and longer lag times. An
Fig. 5. Saturated hydraulic conductivity trends for (a) LT1 and LT2 and (b) LS1 and LS2; the black high-low bars represent the maximum
and minimum measured values by the Guelph Permeameter at each interval; the dotted lines represent the corresponding upper and lower
bounds of measurement error.
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by "University of the Philippines, Diliman-" on 11/18/16
For personal use only.
analysis of the soil moisture during experiments concluded arship). The authors would also like to thank Dr. M. Hayashi and
that a partial frozen surface layer altered how the runoff Dr. C. Ryan of the Department of Geoscience, University of
moved through the cell in cold conditions. This contributed Calgary; A. Montgomery of Mechanical Engineering, University
to the change in performance in cold conditions. However, of Victoria; and staff at the City of Calgary and University of
the efficacy remained high, indicating that bioretention cells Calgary for their assistance and suggestions.
are a viable method of addressing urban stormwater runoff
concerns in Calgary and areas with similar climate and References
drainage patterns. Archer, N.A., Quinton, J.N., and Hess, T.M. 2002. Below-ground
Long-term performance experiments on the columns dem- relationships of soil texture, roots and hydraulic conductivity in
onstrated a significant decrease in Ksat over the first 4 equiv- two-phase mosaic vegetation in south-east Spain. Journal of Arid
alent years of runoff application, after which the conductivity Environments, 52(4): 535–553. doi:10.1006/jare.2002.1011.
was constant. Over the 20 year equivalent runoff application Brown, C.R. 2007. Characteristics of solids removal and clogging
period, Ksat significantly decreased in the tall and short col- processes in two types of permeable pavement. M.Sc. thesis,
umns, by 91.2% and 93.3%, respectively. Compaction and University of Calgary.
initial clogging due to the sediment in the influent was respon- CoC. 2000. Stormwater management and design manual. The City of
sible for this decrease. The majority of the captured sediment Calgary Wastewater and Drainage. Calgary, Alta.
was caught in the top 20 cm of all the columns. This data can Davidson, J.D., LeFevre, N.-J., and Oberts, G. 2008. Hydrologic
be used to assess suitable maintenance procedures to improve bioretention performance and design criteria for cold climates.
long-term bioretention cell performance. Water Environment Research Foundation, Farmington, MN.
Future research should be conducted with different I/P Davis, A.P. 2008. Field performance of bioretention: Hydrology
and catchment sizes, using the methods outlined in this impacts. Journal of Hydrologic Engineering, 13(2): 90 –95. doi:
paper. This would change the average intensities of the 10.1061/(ASCE)1084-0699(2008)13:2(90).
synthetic storm events and the equivalent duration for the Delleur, J.W. 1982. Introduction to urban hydrology and stormwater
field and column experiments, respectively, allowing for management. In Urban stormwater hydrology. Edited by D.F.
performance analysis in different regions. Additionally, ex- Kibler. American Geophysical Union, Washington, DC, USA.
periments to mimic snowmelt events and intensities need to Water Resources Monograph 7. pp. 1–26.
be conducted. Future research should focus on the fate of Ermilio, J.R., and Traver, R.G. 2006. Hydrologic and pollutant re-
the percolating runoff. moval performance of a bio-infiltration BMP. Proceedings of the
2006 world environmental and water resources congress. ASCE/
Acknowledgements EWRI. Omaha, Nb, USA. pp. 1–12.
This research was supported by NSERC, the City of Calgary, Fang, X., Minke, A., Pomeroy, J., Brown, T., Westbrook, C., Guo,
Canada Lands Company, and the CWRA (with a student schol- X., and Guangul, S. 2007. A review of Canadian Prairie hy-
drology: Principles, modelling and response to land use and drain- Muthanna, T.M., Viklander, M., Gjesdahl, N., and Thorolfsson, S.T.
age change. 2 ed., Vol. 2. Centre for Hydrology, University of 2007. Heavy metal removal in cold climate bioretention. Water,
Saskatchewan, Saskatoon, Sask. Air, and Soil Pollution, 183(1– 4): 391– 402. doi:10.1007/s11270-
Granger, R.J., Gray, D.M., and Dyck, G.E. 1984. Snowmelt infiltra- 007-9387-z.
tion to frozen prairie soils. Canadian Journal of Earth Sciences, Muthanna, T.M., Viklander, M., and Thorolfsson, S.T. 2008. Sea-
21(6): 669 – 677. doi:10.1139/e84-073. sonal climatic effects on the hydrology of a rain garden. Hydro-
Gray, D.M., Toth, B., Zhao, L., Pomeroy, J.W., and Granger, R.J. logical Processes, 22(11): 1640 –1649. doi:10.1002/hyp.6732.
2001. Estimating areal snowmelt infiltration into froze soils. Hy- Nordberg, M., and Thorolfsson, S.T. 2004. Low impact development
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by "University of the Philippines, Diliman-" on 11/18/16
drological Processes, 15(16): 3095–3111. doi:10.1002/hyp.320. and bioretention areas in cold climates. Proceedings of the 2004
Hatt, B.E., Fletcher, T.D., and Deletic, A. 2009. Hydrologic and world water and environmental resources congress: Critical tran-
pollutant removal performance of stormwater biofiltration systems sitions in water and environmental resources. ASCE, Salt Lake
at the field scale. Journal of Hydrology (Amsterdam), 365(3– 4): City, Utah. pp. 3409 –3418.
310 –321. doi:10.1016/j.jhydrol.2008.12.001. Roseen, R.M., Ballestero, T.P., Houle, J.J., Avelleneda, P., Wildey,
Hsieh, C.H., and Davis, A.P. 2003. Evaluation of bioretention for R., and Briggs, J. 2006. An examination of cold climate perfor-
treatment of urban storm water runoff. Proceedings of World mance of low impact development stormwater BMPs in a northern
Water and Environmental Resources Congress 2003. ASCE. Phil- climate. Proceedings of the 13th International Conference on cold
adelphia, Pa. pp. 141–148. regions engineering. ASCE, Orono, Maine, USA. pp. 210.
Hsieh, C.H., and Davis, A.P. 2005. Evaluation and optimization of Roseen, R.M., Ballestero, T.P., Houle, J.J., Avellaneda, P., Briggs, J.,
bioretention media for treatment of urban storm water runoff. Fowler, G., and Wildey, R. 2009. Seasonal performance variations
Journal of Environmental Engineering, 131(11): 1521–1531. doi: for storm-water management systems in cold climate conditions.
10.1061/(ASCE)0733-9372(2005)131:11(1521). Journal of Environmental Engineering, 135(3): 128 –137. doi:
Hunt, W.F., and Jarrett, A.R. 2004. Evaluating bioretention areas 10.1061/(ASCE)0733-9372(2009)135:3(128).
from two field sites in North Carolina. Proceedings of the 2004 Sharkey, L.J., and Hunt, W.F. 2005. Hydrologic and water quality
world water and environmental resources congress: Critical tran- performance of four bioretention cells in central North Carolina.
sitions in water and environmental resources management. ASCE. Proceedings of the 2005 Watershed management conference -
Salt Lake City, Utah. pp. 797– 806. Managing watersheds for human and natural impacts: Engineering,
For personal use only.
Hunt, W.F., Smith, J.T., Jadlocki, S.J., Hathaway, J.M., and Eubanks, ecological and economic challenges. ASCE/EWRI, Williamsburg,
P.R. 2008. Pollutant removal and peak flow mitigation by a biore- Va., USA. pp. 833– 842.
tention cell in urban Charlotte, N.C. Journal of Environmental Traver, R.G., and Prokop, M. 2003. Effectiveness of a bioinfiltration
Engineering, 134(5): 403– 408. doi:10.1061/(ASCE)0733-9372 BMP. Proceedings of the World water and environmental re-
(2008)134:5(403). sources congress 2003 and related symposia. ASCE/EWRI, Phil-
Khan, U.T., Valeo, C., Chu, A., and van Duin, B. 2012. Bioretention adelphia, Pa., USA. pp. 1.
cell efficacy in cold conditions: Part 2 — water quality perfor- Traver, R.G., Davis, A.P., and Hunt, W.F. 2007. Bioretention and
mance. Canadian Journal of Civil Engineering, 39(11): This issue. bioinfiltration BMPs. Stormwater: The journal for surface water
doi:10.1139/l2012-111. quality professionals, October 2007.
Kim, H., Seagren, E.A., and Davis, A.P. 2003. Engineered biore- UNHSC. 2005. 2005 Data Report. Bioretention System. 1st ed.,
tention for removal of nitrate from stormwater runoff. Water Envi- Durham, N.H., USA.
ronment Research, 75(4): 355–367. doi:10.2175/106143003X141169. US EPA. 2001. Low impact development (LID): A literature review
PMID:12934829. (# EPA-841-B-00-005). Environmental Protection Agency, Office
Lanarc Consultants Ltd., Kerr Wood Leidal Associates, and Ngan, G. of Water, Washington, D.C., USA.
2005. Infiltration rain garden. In Stormwater source control design van der Kamp, G., Hayashi, M., and Gallén, D. 2003. Comparing the
guidelines April 2005. Greater Vancouver Regional District, Van- hydrology of grassed and cultivated catchments in the semi-arid
couver, B.C. pp. 43–52. Canadian prairies. Hydrological Processes, 17: 559 –575. doi:
Le Coustumer, S., Fletcher, T.D., Deletic, A., and Barraud, S. 2007. 10.1002/hyp.1157.
Westhoff Engineering Resources (WER). 2007a. Bioretention areas.
Hydraulic performance of biofilters for stormwater management:
In Source control practices handbook. Edited by Westhoff Engi-
first lessons from both laboratory and field studies. Water Science
neering Resources. Westhoff Engineering Resources, Calgary, Al-
and Technology, 56(10): 93–100. doi:10.2166/wst.2007.735.
berta. pp. 7.1–7.38.
PMID:18048981.
Westhoff Engineering Resources (WER). 2007b. Appendix A: Key
Le Coustumer, S., Fletcher, T., Deletic, A., and Barraud, S. 2008.
aspects of interest. In Source control practices handbook. Edited by
Influence of time and design on the hydraulic performance of
Westhoff Engineering Resources. Westhoff Engineering Re-
biofiltration systems for stormwater management. Proceedings of
sources, Calgary, Alberta. pp. A.1–A.37.
the 11th International Conference on urban drainage. ICUD, Ed-
inburgh, UK. pp. 1–10.
Lewis, J.F., Hatt, B.E., Deletic, A., and Fletcher, T.D. 2008. The List of symbols
impact of vegetation on the hydraulic conductivity of stormwater
AB bioretention cell area (m2)
biofiltration systems. Proceedings of the 11th International Con-
ference on urban drainage. ICUD, Edinburgh, UK. pp. 1–10. d equivalent precipitation depth (mm)
Li, H., and Davis, A.P. 2008. Urban particle capture in bioretention Ksat saturated hydraulic conductivity
media. I: Laboratory and field studies. Journal of Environmental L other losses (m3)
Engineering, 134(6): 409–418. doi:10.1061/(ASCE)0733-9372(2008) Qe effluent peak or CoM flow rate (mm/h)
134:6(409). Qi influent peak or CoM flow rate (mm/h)
Sg is equal to the volume of water stored in growing media 12 Ve volume of effluent (m3)
hours after peak soil moisture (m3) Vi volume of influent (m3)
te time of effluent peak or CoM ⌬Q change in either the peak or CoM flow rate reduction (%)
ti time of influent peak or CoM ⌬V change in runoff volume between inlet and outlet of the
ts start time of the experiment bioretention cell (%)
TSS total suspended solids (mg/L)
V volume of influent from tanks (m3)
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by "University of the Philippines, Diliman-" on 11/18/16
For personal use only.