Long-term retention performance is a common performance indicator for low-impact development practices, such as rain barrels, rain gardens, and green roofs. This paper introduces a numerical approach for the estimation of annual retention ratios of stormwater by bioretention. The annual retention ratio is taken as the ratio of the annual accumulated volume of stormwater retained by bioretention over the total volume of runoff draining into the system. The hydrologic model Storm Water Management Model (SWMM) is used to simulate the relevant flows of a bioretention system with parametric variations of the watershed area ratio and hydraulic conductivity of the soil media. Under these two dominant performance-governing parameters, retention ratios are calculated using the 10-year (2004–2013) rainfall record in Hong Kong at 1-min intervals. This indicator can be readily applied to estimate the long-term retention performance of a bioretention using particular values of watershed area ratio and hydraulic conductivity of soil media under the climate of Hong Kong. The study also analyzes the influence of variation of annual precipitation on the estimated retention performance. Flow data monitored on a pilot-scale physical model of bioretention during a number of rainfall events are used to validate the numerical simulation.

INTRODUCTION

Bioretention, or rain gardens, is one form of the structural stormwater best management practices (BMP) for control and treatment of urban stormwater runoff. The potential hydrologic benefits include the reduction of both the peak flow of surface runoff into the storm drain and the total volume of runoff generated from the same rainfall event. A bioretention cell retains a portion of surface runoff from the vicinity within its depressed soil surface and therefore provides on-site retention/detention of stormwater (Prince George's County 2007).

Similar to a conventional detention facility, Davis (2008) suggested three metrics to quantify the hydrologic performance of a bioretention system, namely the peak flow reduction ratio, the outflow volume ratio, and the flow peak delay ratio. The hydrologic performance is affected to various degrees by a number of design parameters, such as depth of soil media, hydraulic conductivity of soil media, plant species, watershed area ratio, ponding volume, and hydraulic conductivity of in-situ soil. In addition, the local precipitation pattern also significantly influences the hydrological performance of a bioretention system. The pattern includes rainfall depth, intensity and duration, seasonal distribution, and annual fluctuations. For instance, Li et al. (2009) suggested that a good peak flow reduction can be achieved by a well-designed bioretention system for small rainfall events less than 20 mm total rainfall depth, but the exact value of the appropriate rainfall depth was observed to vary in actual field monitoring (Hunt et al. 2008; Brown & Hunt 2011).

Hong Kong has a subtropical climate with distinct wet and dry seasons affected by monsoons. In the recent 10-year period, the annual rainfall depths were very high, ranging between 1,480 and 3,218 mm (Table 1) with an average of 2,400 mm. Many urban areas are characterized by almost complete impervious surface and/or steep topography, and thus experience high risk of flash floods. To alleviate urban flooding problems, Hong Kong has developed extensive hardcore stormwater drainage systems including capacity expansion of stormwater drains, massive stormwater diversion schemes and huge retention basins (Lee et al. 2008; Yu & Lee 2009). However, bioretention or other infiltration-based stormwater BMP is not widely implemented in Hong Kong. Potential hydrologic benefits are expected from the implementation of infiltration-based BMPs, but prior evaluations are needed to estimate the extent of the benefits.

Table 1

Annual precipitation depth of Hong Kong

Year Depth (mm) Year Depth (mm) 
2004 1,743 2009 2,186 
2005 3,218 2010 2,376 
2006 2,632 2011 1,480 
2007 1,711 2012 1,929 
2008 3,071 2013 2,624 
Year Depth (mm) Year Depth (mm) 
2004 1,743 2009 2,186 
2005 3,218 2010 2,376 
2006 2,632 2011 1,480 
2007 1,711 2012 1,929 
2008 3,071 2013 2,624 

The purpose of this paper is to estimate the possible hydrologic benefit of implementing bioretention in Hong Kong under the specific rainfall characteristics. A small-sized bioretention cell was constructed for the pilot-scale test while numerical modeling was made with the Storm Water Management Model 5.0.022 with LID Control module (SWMM) (US Environmental Protection Agency (US EPA) 2009) using the recent 10-year rainfall record of Hong Kong as input. SWMM has been used in simulating the hydraulics of other low-impact development (LID) practices, such as green roofs (Burszta-Adamiak & Mrowiec 2013) and rain harvesting devices (Walsh et al. 2014). The numerical model study aims to evaluate the hydrologic performance under parametric variations of the major design parameters including the watershed ratio and the hydraulic conductivity of the subsoil. It was expected that the results of the present study can provide a statistically-based decision on the potential use of bioretention in Hong Kong.

METHODS AND MATERIALS

Performance indicator

An important factor affecting the hydrologic performance of bioretention is the yearly variations of annual rainfall depths in Hong Kong. This paper proposes a single-value performance indicator to quantify the benefit of stormwater runoff mitigation over a statistically lengthy period. This time period is chosen as a 1 year, which should be sufficient to cover a typical annual precipitation pattern in Hong Kong. Among the major flows in a bioretention cell, the performance indicator only takes into account the overflow and inflow as these are easy to obtain from either field monitoring or SWMM simulation. The inflow could be regarded as the runoff generated on the catchment if no bioretention were implemented and the overflow is the stormwater runoff with bioretention implemented. The difference between inflow and overflow reflects the capacity of bioretention. The proposed performance evaluation compares the yearly total accumulated volumes of the stormwater inflow into the watershed and the stormwater retained by bioretention. The ratio (RR) of the retained volume over the inflow volume is then used as the single-value performance indicator. Similar simple indicators for retention performance have been adopted in analyzing the performance of rain gardens (Jennings et al. 2013) and green roofs (Stovin et al. 2013). It should be noted that the retained stormwater may eventually return to the storm drainage system, for instance where an underdrain is installed, so that there are limitations in using RR to directly represent loading mitigation on storm drains from bioretention.

Physical model and numerical simulation

The pilot-scale physical model of a bioretention system consisted of an impervious catchment, a bioretention cell, and the related flow measurement instrumentation (Li & Lam 2013; Lam et al. 2014). The 13 × 2 m2 roof surface of a shed served as the impervious catchment (Figure 1). The runoff generated on the roof was directed into the planter-box type bioretention cell via a half-cut pipe with a chamber and a throttle pipe. The bioretention cell had a surface of size 3 × 0.3 m2. The bottom layer of the cell used gravels of depth 0.1 m where a perforated PVC underdrain was installed to simulate exfiltration into the subsoil. Above the gravel was 0.45 m thick soil media, which used a mix of 30% local public fill by weight and 70% river sand available in the construction market. The hydraulic conductivity was estimated at about 50 mm/h from the ratio of the measured flow rate in the underdrain over the cell area. A vertical standpipe was installed as the overflow outlet and the top of its opening was 95 mm above the surface of soil media, thus setting the storage depth at 95 mm. There is a 50 mm freeboard above the standpipe opening to avoid undesirable overflow of the planter box. A local plant species was transplanted into the cell and weeds were observed but left there during the monitoring period.

Figure 1

Left: pilot-scale physical model of bioretention system, consisting of a plywood-box bioretention cell and catchment area on roof of the shed. Right: outlet of standpipe and underdrain with ultrasonic flowmeter.

Figure 1

Left: pilot-scale physical model of bioretention system, consisting of a plywood-box bioretention cell and catchment area on roof of the shed. Right: outlet of standpipe and underdrain with ultrasonic flowmeter.

Flow rate measurements of inflow and overflow were made with ultrasonic flowmeters (Controlotron System 1010 UNIFLOW Universal MultiFunction Flowmeter with C3 clamp-on ultrasonic sensors) mounted on both the inflow throttle pipe and the extended horizontal pipe from the bottom of overflow standpipe. The data were recorded by a data logger (Campbell CR23X) at 5-s and 10-s intervals during two monitoring periods: 2 August–5 September 2013 and 29 March–8 April 2014, respectively.

The SWMM modeling was first calibrated and verified with monitoring data on the pilot-scale physical model. The model was then used to estimate the bioretention performance for the past 10-year rainfall of Hong Kong under a parametric variation of design parameters.

Built-in functions in SWMM including ‘LID-bioretention’, ‘sub-catchments’, and ‘rain gauges’ were used to construct the hydrologic model. A ‘rain gauge’ contained the monitored precipitation in the verification and rainfall record during 2004–2013 in the further simulation. The time step was 1 min and routing interval was 30s.

Two ‘sub-catchments’ represented the surfaces of the roof and the bioretention cell. The roof ‘catchment’ was set at 100% impervious, area 0.0026 ha, flow path 13 m, slope 5% and the runoff was directed into the bioretention ‘catchment’ which was entirely occupied by ‘LID-bioretention’ of area 0.00009 ha. The major settings of ‘LID-bioretention’ were surface storage depth 95 mm, soil media of thickness 0.4 m and hydraulic conductivity k (mm/h) = {5, 20, 40, 60, 80} for the verification simulations. In the performance simulations, the value of k was varied within a larger range of {2, 5, 10, 20, 30, 40, 60, 80 mm/h}. The surface area of the bioretention cell was also varied to arrive at 29 different values of watershed area ratio (AR) ranging from 1 to 100. Furthermore, the storage depth of the cell was changed to 150 mm as recommended by Prince George's County (2007).

RESULTS AND DISCUSSION

Monitored events and model verification

For the pilot-scale bioretention system, 58 rainfall events were identified during the monitoring periods using a separation dry period over 1 h. Figure 2 shows the rates of inflow and overflow for three typical events during which the total inflow was: (a) entirely retained; (b) partially retained; or (c) barely retained by bioretention. The rainfall depths of the events were 4.2 mm, 6.8 mm, and 27.6 mm, respectively, and it is obvious that the bioretention performance, in terms of both the runoff volume attenuation and the peak flow reduction, worsens with increasing precipitation depths.

Figure 2

Sample rainfall events monitored in the physical model with simulated overflow: (a) 3 August 2013, effective rainfall depth = 4.2 mm; (b) 14 August 2013, effective rainfall depth = 6.8 mm; (c) 13 August 2013, effective rainfall depth = 27.6 mm.

Figure 2

Sample rainfall events monitored in the physical model with simulated overflow: (a) 3 August 2013, effective rainfall depth = 4.2 mm; (b) 14 August 2013, effective rainfall depth = 6.8 mm; (c) 13 August 2013, effective rainfall depth = 27.6 mm.

Figure 3 summarizes the performance of the pilot-scale bioretention system for all 58 rainfall events. The volume retention ratios and peak flow reduction ratios are plotted against the event rainfall depths in Figures 3(a) and 3(b), respectively. Despites the large variations in rainfall depths, a threshold depth at ≈3 mm is evident, below which nearly 100% retention and 100% peak reduction were achieved. Precipitation above the threshold depth led to overflow and the bioretention performance deteriorated with increases in rainfall depths. The pilot-scale bioretention cell operated with a watershed area ratio of about 30. The ponding depth of 95 mm alone could retain 95/30 ≈ 3.3 mm. This value is consistent with the threshold value at ≈3 mm.

Figure 3

Bioretention performance for all 58 monitored rainfall events: (a) volume retention ratio; (b) reduction ratio of peak flow rate; (c) inflow and overflow depth, ×: three events in Figure 2; (d) comparison between modeled and monitored overflow depths.

Figure 3

Bioretention performance for all 58 monitored rainfall events: (a) volume retention ratio; (b) reduction ratio of peak flow rate; (c) inflow and overflow depth, ×: three events in Figure 2; (d) comparison between modeled and monitored overflow depths.

Figure 3(c) further suggests that this ≈3 mm retention depth averagely spread over the whole catchment was effective to all the rainfall events regardless of their precipitation depths. There was a statistically constant deficit between the 58 overflow volumes and the corresponding inflow volumes. The mean difference value was ≈2.7 mm, which is consistent with the retention depth at 3.3 mm from the physical configuration. For a better presentation, Figure 3(c) only shows the 50 rainfall events with depths of less than 30 mm but the other eight rainfall events of greater inflow depths reaching almost 100 mm also fall well onto the linear fitting line.

A calibration of the SWMM modeling is made by simulating the flow response of the physical bioretention cell for the first 45 monitored rainfall events. For summary comparison, Table 2 lists the total inflow volume, in terms of depths, accumulated over these events and the corresponding measured and computed overflow volumes. The respective retention ratios, that is, (inflow − overflow)/inflow, are also listed. The results indicate that the closest simulation is achieved by using k = 40 mm/h which is close to the measured value at 50 mm/h.

Table 2

Accumulated inflow and overflow volume (in terms of depth) over first 45 rainfall events monitored on pilot-scale bioretention cell

Measured:
 
SWMM simulated overflow with k (mm/h) value
 
Inflow (mm) Overflow (mm) 80 60 40 20 
524.2 364.0 331.7 345.4 366.7 398.4 448.0 
Retention ratio: 
 0.306 0.367 0.341 0.300 0.240 0.145 
Measured:
 
SWMM simulated overflow with k (mm/h) value
 
Inflow (mm) Overflow (mm) 80 60 40 20 
524.2 364.0 331.7 345.4 366.7 398.4 448.0 
Retention ratio: 
 0.306 0.367 0.341 0.300 0.240 0.145 

Measured and SWMM simulated overflow volumes using different hydraulic conductivity.

A second calibration takes the 58 available events into account and uses k = 40 mm/h. A more detailed comparison is made in Figure 3(d) on the measured and simulated overflow depths for each event. It is evident that the SWMM simulation can reproduce all measured data well. The coefficient of correlation is 0.991 and the Nash‒Sutcliffe coefficient for the hydrologic prediction is calculated to be 0.972 (Nash & Sutcliffe 1970).

SWMM modeling of 10-year bioretention performance for Hong Kong

In the performance simulation, the yearly total inflow volume of stormwater runoff and bioretention volume are computed with SWMM for the 10 years using different values of hydraulic conductivity and ratio of catchment areas. For each year, the simulated overflow from all minutes in that year is summed up to obtain the annual overflow volume, the difference of which to the inflow volume from annual precipitation gives the retention volume. The performance indicator for that year is calculated simply as the ratio RR between the retention volume and the inflow volume. The 10-year values can be averaged to indicate the long-term average performance of the bioretention system.

Figure 4(a) shows the 10-year averaged RR in the form of contour levels on the kAR plane. It is clear, as anticipated, that the retention ratio increases with the hydraulic conductivity and the inverse of the watershed ratio. Figure 4(a) shows that for AR between 10 and 40, the bioretention performance improves significantly from low values of RR when the hydraulic conductivity increases beyond 20 mm/h. That means that for applications with AR > 10, maintaining the hydraulic conductivity >20 mm/h is important to ensure an acceptable performance of bioretention in Hong Kong.

Figure 4

Bioretention performance with hydraulic conductivity and watershed area ratio: (a) 10-year averaged retention ratio RR; (b) coefficient of correlation between RR and annual precipitation depths over 10-year period; (c) estimated change of RR from 10-year average resulting from 1,000 mm increase in annual precipitation from 10-year average.

Figure 4

Bioretention performance with hydraulic conductivity and watershed area ratio: (a) 10-year averaged retention ratio RR; (b) coefficient of correlation between RR and annual precipitation depths over 10-year period; (c) estimated change of RR from 10-year average resulting from 1,000 mm increase in annual precipitation from 10-year average.

The effect of clogging can also be assessed in Figure 4(a). For example, a bioretention system of AR = 25 and k =40 mm/h is able to retain 50% of annual precipitation. A degradation of infiltration due to clogging from k =40 mm/h to 20 mm/h would cause RR to decrease to 40%. However, a further small degradation from k =20 to 10 mm/h, would result in the same amount of reduction in RR by 10%.

Within the 10-year period of 2004–2013, Hong Kong experiences significant fluctuations in the annual precipitation (Table 1). As such, the performance of bioretention (RR) varies each year from the 10-year average. It is worth studying how the performance is affected by the fluctuations in annual rainfall.

Figure 4(b) shows the coefficient of correlation between the yearly values of RR and the annual precipitation depths. If the condition of correlation coefficients >0.7 is chosen as the threshold for significant relevance, then there are two regions in Figure 4(b) where the bioretention performance is less likely to be affected by fluctuations in annual precipitation. The first region is at the right side of the graph with AR > 45 and the second region is the small part at the upper-left corner of Figure 4(b). Referring to Figure 4(a), the performance of bioretention is either poor (RR < 0.3) or very good (RR < 0.9) in these two regions.

For the designs of bioretention systems with performance expected to be affected by annual precipitation variations, that is, operation with coefficient of correlation >0.7 in Figure 4(b), the change in bioretention performance can be estimated from the correlation analysis. Figure 4(c) shows the predicted change in RR due to an annual precipitation being greater than the 10-year mean by 1,000 mm. The predicted changes are negative, meaning that a greater annual precipitation would result in a smaller retention ratio for those designs.

For example, a bioretention system of AR =25, and k = 40 mm/h is expected to operate with a 10-year average retention ratio at RR ≈ 0.5 (Figure 4(a)). If the annual precipitation in a particular year is close to the 10-year average value of 2,400 mm, the same performance is expected. However, in a year where the annual precipitation is greater at 3,400 mm (1,000 mm higher), the bioretention system would operate with a substantially impaired performance with RR reduced by 0.07 from 0.5, that is RR = 0.43. Similarly, in a year of lower annual precipitation at 1,400 mm, the bioretention performance will be better with RR = 0.5 + 0.07 = 0.57.

CONCLUSIONS

This paper presents a statistical approach to estimate the long-term hydrologic performance of a bioretention system, in terms of a retention ratio of the annual rainfall volume. The estimation is made with the hydrologic model SWMM with the local historic rainfall records. The contour plot of retention ratios can be used conveniently to assess the performance of a bioretention system for a wide-spectrum of the main design parameters which include the watershed area ratio and the hydraulic conductivity of the soil media; or can be used to determine the configuration of bioretention system targeted at long-term water balance.

As an example, the estimation method is applied to Hong Kong conditions using 10-year rainfall record during 2004–2013. It is found that the retention ratio of a bioretention system would stay at fairly constant values despite the large fluctuations in the annual precipitation depths within the 10 years. As expected, the bioretention performance is impaired in years of extreme high precipitation depths. The retention ratio of a bioretention system is highly sensitive to the hydraulic conductivity of the soil media in the range between 5 and 20 mm/h, as well as to the watershed area ratio in the range between 10 and 40. It is recommended to maintain the hydraulic conductivity of soil media above, say, 20 mm/h for the bioretention applications in Hong Kong.

In the application and interpretation of retention ratio for bioretention performance, it should be pointed out that this approach evaluates the retention volume simply as the difference between the inflow and the overflow. This volume is the portion of stormwater passing through the bioretention system and a part of it may eventually return to the storm drainage system. Other effects such as evapotranspiration and drying–wetting cycles of the bioretention soil media are also not taken into consideration, As such, the use of the retention ratio to estimate certain hydrologic performance of a bioretention system such as mitigation of storm drainage loading should be taken in light of these limitations.

ACKNOWLEDGEMENT

The research is supported by a Strategic Research Development Fund of Department of Civil Engineering, University of Hong Kong.

REFERENCES

REFERENCES
Brown
R. A.
Hunt
W. F.
2011
Impacts of media depth on effluent water quality and hydrologic performance of undersized bioretention cells
.
Journal of Irrigation and Drainage Engineering
137
(
3
),
132
143
.
Burszta-Adamiak
E.
Mrowiec
M.
2013
Modelling of green roof's hydrologic performance using EPA's SWMM
.
Water Science and Technology
68
(
1
),
36
42
.
Davis
A. P.
2008
Field performance of bioretention: hydrology impacts
.
Journal of Hydrologic Engineering
13
(
2
),
90
95
.
Hunt
W. F.
Smith
J. T.
Jadlocki
S. J.
Hathaway
J. M.
Eubanks
P. R.
2008
Pollutant removal and peak flow mitigation by a bioretention cell in urban Charlotte, NC
.
Journal of Environmental Engineering
134
(
5
),
403
408
.
Jennings
A. A.
Adeel
A. A.
Hopkins
A.
Litofsky
A. L.
Wellstead
S. W.
2013
Rain barrel – urban garden stormwater management performance
.
Journal of Environmental Engineering
139
(
5
),
757
765
.
Lam
K. M.
Li
Z. Y.
Cheng
C. C. K.
2014
Performance of bioretention improved with higher infiltration rate
. In:
Int. Conf. Intelligent Systems, Structures and Facilities: Sustainable Green Intelligent Buildings
,
Hong Kong
,
Asian Institute of Intelligent Buildings (AIIB), 7 January 2014
, pp.
28
32
.
Lee
J. H. W.
Chan
C. H. C.
Kuang
C. P.
Clark
P.
Townsend
N.
Shiu
W. Y.
2008
Hydraulic model study of the Tai Hang Tung Storage Scheme
.
Journal of Hydraulic Research
46
(
S1
),
11
23
.
Li
Z. Y.
Lam
K. M.
2013
Bioretention with shallow fill media
. In:
Proc. 35th IAHR World Congress
,
8-13 September
,
Chengdu (Tinsgua University Press, Beijing)
, pp.
1
7
.
Li
H.
Sharkey
L. J.
Hunt
W. F.
Davis
A. P.
2009
Mitigation of impervious surface hydrology using bioretention in North Carolina and Maryland
.
Journal of Hydrologic Engineering
14
(
4
),
407
415
.
Nash
J. E.
Sutcliffe
J. V.
1970
River flow forecasting through conceptual models. Part I – a discussion of principles
.
Journal of Hydrology
10
(
3
),
282
290
.
Prince George's County
2007
Bioretention Manual
.
Department of Environmental Resources, Prince George's County
,
Largo, MD
,
USA
.
Stovin
V.
Poe
S.
Berretta
C.
2013
A modeling study of long term green roof retention performance
.
Journal of Environmental Management
131
,
206
215
.
US Environmental Protection Agency (US EPA)
2009
Storm Water Management Model (SWMM) Version 5.0 (Software)
. .
Walsh
T. C.
Pomeroy
C. A.
Burian
S. J.
2014
Hydrologic modeling analysis of a passive, residential rainwater harvesting program in an urbanized, semi-arid watershed
.
Journal of Hydrology
508
,
240
253
.
Yu
D. Y.
Lee
J. H. W.
2009
Hydraulics of tangential vortex intake for urban drainage
.
Journal of Hydraulic Engineering
135
(
3
),
164
174
.