Ecological Modelling 291 (2014) 6–14

Contents lists available at ScienceDirect

Ecological Modelling
journal homepage: www.elsevier.com/locate/ecolmodel

Assessing the effectiveness of green infrastructures on urban flooding

reduction: A community scale study
Wen Liu, Weiping Chen ∗ , Chi Peng
State Key Laboratory for Urban and Regional Ecology, Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences, Beijing 100085 China

a r t i c l e i n f o a b s t r a c t

Article history: The risk of urban flooding is increasing as a result of rapid urbanization. Green infrastructure (GI) is an
Received 23 January 2014 emerging planning and design concept to mitigate urban flooding. A community scale simulation model
Received in revised form 14 July 2014 was developed to quantify the effectiveness of GI on reducing the volume and peak flow of urban flood-
Accepted 14 July 2014
ing. Five scenarios, namely expanding green space, converting to concave green space, constructing a
runoff retention structure, converting to porous brick pavement, and combining previous four measures
Keywords:
were considered for an urban community in Beijing. The outcomes showed that the model performed
Green infrastructure
responsively to simulate the storm runoffs at varying recurrence intervals under these scenarios. Simula-
Storm water runoff
Runoff volume
tion results showed that, the impervious surfaces have the most contribution to the storm runoffs of the
Peak flow community. The reduction capacity for single GI facility was limited, especially in bigger storm events.
Reduction effects The integrated GI configuration has effective reduction percentage, such as the total runoff reduction was
Community ranged from 100% to 85.0% and the peak flow reduced 100–92.8%. This work can guide local planners and
decision makers in their actions on green infrastructures in community scale.
© 2014 Elsevier B.V. All rights reserved.

1. Introduction Conventionally, storm water is detaining and conveying through

piped drainage systems. However, the conventional approaches
Over half of the people worldwide now live in cities and the fail to address the increases of storm runoff volume and peak
percentage is expected to reach eighty by 2050 (Bettencourt and flows caused by urban development. The runoffs overburdened the
West, 2010). It causes the shortage of natural resources and creates municipal wastewater treatment works and carried pollutants such
stresses in the already stressed out urban ecosystems (Mentens as trash, bacteria and heavy metals into the receiving waters that
et al., 2006). As the impervious urban surfaces expanding, crop degraded the quality of the urban streams. The holistic modeling
lands, pastures, wilderness, and forests are displaced by dwellings, and ecological risk assessment of urban ecosystems can be used
office structures, streets, highways, and industrial and commercial for risk evaluation and urban management to minimize the dam-
establishments. Significant land use modifications that converted age by extreme weather and changing climate from the source of
the natural landscape to urban settings may have disastrous con- urban flooding (Chen et al., 2013, 2014). In recent years, the green
sequences for urban ecosystems. Impervious surfaces alter the infrastructure (GI) has been widely used in many cities in the USA,
hydrological nature of surface runoffs, prevent the infiltration and in the UK, Canada, Germany and New Zealand to mitigate urban
of surface water into the ground, greatly increase storm runoffs flooding (Ahiablame et al., 2012). The GI employs principles such
in terms of volumes and peak flow (Goonetilleke et al., 2005; as preserving and recreating natural landscape features, implemen-
Whitford et al., 2001), and consequently causing floods in cities. The ting some on-site infrastructures that work with nature to reduce
urban flooding endangers life, private properties and public infras- the storm water runoff from sources (Graham et al., 2004). It is an
tructures, erodes stream banks and channels, and contaminates economically and environmentally viable approach for developing
streams and rivers in urban areas. Urban flooding risk may be inten- sustainable and resilient communities, adapting climate change,
sified as the earth experiences more frequent weather extremes as well as for promoting smart growth and urban sustainability
with the global climate change (Foster et al., 2011; Villarreal et al., (Benedict and McMahon, 2002, 2006; Dunn, 2010; Foster et al.,
2004). 2011; Gill et al., 2007; Mell, 2009).
The GI installations consist of systems and practices that utilize
or mimic the natural processes, allowing storm water to infiltrate,
∗ Corresponding author. Tel.: +86 10 62843981; fax: +86 10 62918177. evaporate, runoff, and/or be used on-site. For example, reser-
E-mail address: wpchen@rcees.ac.cn (W. Chen). voirs and ponds may be strategically placed on the flow paths to

http://dx.doi.org/10.1016/j.ecolmodel.2014.07.012
0304-3800/© 2014 Elsevier B.V. All rights reserved.
 W. Liu et al. / Ecological Modelling 291 (2014) 6–14 7

Fig. 1. Schematic depiction of the calculation of urban storm water runoff. P is the precipitation, E is the evaporation, C is the canopy interception, F is the soil infiltration, D
is the depression, I is the inflow, Q is the outflow, R is the surface runoff.

temporary store the surface runoff (Ferguson, 1998; White, 2002) In this research, a simple rainfall-runoff model with fewer
and green spaces where runoff diverted from impervious surfaces parameters requirements is developed to describe functions of
may have a chance to infiltrate and/or evaporate (Mentens et al., green infrastructures based on the water mass balance through the
2006). GI can be used in a wide range of landscape scales in place of processes of urban hydrological cycle. By using the model, the effec-
or in addition to traditional storm runoff control elements, as well tiveness of single GI facilities and the integrated GI configuration
as maintain or restore an urban ecosystem’s hydrologic and eco- on urban flooding reductions were evaluated under different storm
logical functions. Therefore, for effective control of urban flooding, recurrence periods. A typical community in Beijing is selected for
it is imperative to develop tools that optimize the GI configurations study. This study illustrates a scientifically and ecologically respon-
according to the urban hydrologic cycle. sible approach for urban storm runoff management and landscape
The hydrological performance and benefits of GI practices have planning.
been shown in numerous studies on laboratory scales, in-situ scales
and micro-scales. For example, Alfredo et al. (2009) found that
2. Methodology
green roofs can delay and prolong the roof discharge and reduce its
peak rate by 30–78% compared to a standard roof surface. Dreelin
In terms of precipitation and runoff, the urban area is divided
et al. (2006) showed that porous pavements reduced 93% of runoff
into four types of surfaces, namely impervious surface (building
on two parking lots, and it can be used to control small storms
footprints, roads, pavements, parking lots, etc.), pervious surface
(less than 2 cm). Schneider and McCuen (2006) showed that the
(green spaces, lawns, bare soils, etc.), water body (natural and
effects of cisterns on peak discharge reduction are ineffective for
man-made reservoirs, wetlands and rivers), and green infrastruc-
large storms, but very effective for small storms. Chapman and
tures (Fig. 1). In a precipitation event, the rain would be routed
Horner (2010) reported that a street-side bioretention facility in
through different processes of the hydrological cycle in each sur-
Washington can achieve 26–52% of runoff retention in real-weather
face, which are depended on the nature of surfaces and dynamic
conditions. Qin et al. (2013) concluded that the swale, permeable
factors. The water routing through each compartment is calcu-
pavement and green roof are more effective in flood reduction
lated independently and then sum up the three surfaces’ runoffs or
during heavier and shorter storm events compared with the con-
the overflows of GIs to obtain the total storm runoff. The impacts
ventional drainage system.
of green infrastructures on storm water infiltration, retention and
Influences of GI on urban storm runoff reduction can be eval-
storage capacity are accounted for in calculating storm runoff. The
uated by hydrological models such as Storm Water Management
model subroutines are described in the following sections. Based
Model (SWMM), Urban Volume and Quality (UVQ) and Model
on water mass balance, the storm runoff volume calculation is
for Urban Stormwater Improvement Conceptualisation (MUSIC)
obtained by selecting simpler calculation equations with fewer
(Huber et al., 1988; MUSIC Development Team, 2003; Mitchell
parameters from several typical urban drainage models. The reser-
et al., 2003). However, these traditional hydrological models are
voir routing is neglected in the peak flows calculations due to the
cumbersome and unsuitable to evaluate the effectiveness of GI in
micro-scale of the community.
mitigating urban flooding because of the difficulties in defining
model parameters and the over-simplifications of reactive pro-
cesses related to GI. Although the GI performance on reducing 2.1. Runoff of impervious surface
urban flooding has been extensively investigated, few studies have
attempted to examine and compare the reduction effectiveness The calculation of impervious surface runoff is based on the
between integrated GI configuration and single GI facilities under water balance (Mitchell et al., 2003). Rainfall reaches impervious
different storm recurrence periods. areas beyond the depression storage depth is directly converted to
8 W. Liu et al. / Ecological Modelling 291 (2014) 6–14

surface runoff. The equation to calculate the amount of runoff from is usually 2.45 MJ/kg, Tmax the maximum daily temperature (◦ C),
impervious surface (Rimp , mm) is: Tmin the minimum daily temperature (◦ C) and Tav is the average
 daily temperature (◦ C). The constant 0.408 is used to convert the
P − (D − ST ) − ED P − ED > D − ST radiation to evaporation equivalents in mm (Droogers and Allen,
Rimp = (1-a) 2002).
0 P − ED ≤ D − ST

ED = min(Ep , ST ) (1-b)
2.2.3. Infiltration
where P is the precipitation (mm), D the depression of impervious The Green-Ampt infiltration equation modified by Mein and
area, ST the depression storage level (mm), ED the evaporation of Larson (Mein and Larson, 1973) is used to simulate the infiltra-
depression (mm), and Ep is the potential evaporation (mm) of the tion process, which is a function of the soil suction head, porosity,
simulated time step. hydraulic conductivity and time. The infiltration rate (f, mm/min)
is calculated by

2.2. Runoff process of pervious surface 

Ks [1 + (s − i )Sf /F] t > tp
f = (4-a)
Four processes, namely interception, evaporation, infiltration i t ≤ tp
and depression, are considered simultaneously in simulating
runoffs of pervious surface and calculated in order. Precipita-
Fp
tion entering the pervious areas, firstly intercepted by vegetation tp = (4-b)
canopy, then the net rainfall reaches the ground to permeate and be i
stored in the soil, and only water excesses the depression storage
is routed as runoff. The evaporation was occurred to recover the (s − i )Sf
Fp = (4-c)
interception and depression water storage. Based on the water bal- i/Ks − 1
ance, the pervious runoff was equal to the rainfall minus the loss
amounts of interception, evaporation, infiltration and depression where Ks is the saturated hydraulic conductivity (mm/min),  s the
water. saturated water content (%),  i the initial water content (%), Sf the
suction at wetting front (mm) and F is the cumulative depth of
infiltration (mm). The depth infiltrated at that moment (Fp ) is given
2.2.1. Interception by setting f = i in Eq. (4-a) and solving for Fp . tp is the onset of surface
The vegetation interception routine is based on the algorithm ponding (min), i is the net rainfall intensity (mm/min) which can
of UFORE-hydro that simplifies the explicit simulation of canopy be calculated as:
drainage performed in the Rutter model to reduce model parameter
requirements (Wang et al., 2008). The equation to calculate water [P − (SIc − SI0 )]
i= (4-d)
retained by vegetation canopy interception (SIc , mm) is: t

SI0 + P − Pf − Ec SI0 + P − Pf − Ec ≤ Sc where t is the simulated time step.
SIc = (2-a)
Sc − Ec SI0 + P − Pf − Ec > Sc

where SI0 is the water storage of interception during the prior time 2.2.4. Depression storage
interval (mm), Ec the evaporation of leaf interception (mm) and Pf The empirical equation of Linsley et al. (1949) is used to simulate
is the free throughfall without contacting canopy (mm) which may the depression storage process. The depression storage (Sd0 , mm)
be calculated by: at the initial time step is calculated as:

Pf = Pe−LAI (2-b)   PC

 SI 2/3 Sd0 = Sdmax 1 − exp − (5-a)
Sdmax
c
Ec = Ep (2-c)
Sc
where Sdmax is the depression of pervious area (mm) and PC is
Sc = SL LAI (2-d) the accumulated residual rainfall (mm), which represents rainfall
minus the interception and infiltration. The calculation of depres-
where  is an extinction coefficient, LAI the leaf area index, Sc the
sion in the next time step (Sd, mm) is based on mass balance:
storage capacity of canopy (mm) and SL denotes the specific leaf
storage (mm). 
PC + Sdt − Es PC + Sdt − Es < Sdmax
Sd = (5-b)
2.2.2. Evaporation 0 PC + Sdt − Es ≥ Sdmax
The evaporation process is commonly simulated by the Penman-
Monteith equation that requires many locale-specific parameters
Es = min(Sdt , Ep ) (5-c)
often difficult to reliably define. Instead, we choose Hargreaves
and Samani (1985) that simulates the evaporation process using
only the air temperatures. The equation for calculating potential where Sdt is the depression in the previous time step (mm) and Es
evaporation (Ep , mm/d) is given as: is the evaporation of pervious depression (mm).
 RA  According to water balance, the runoff volume from the pervious
Ep = 0.0023 × 0.408
max
(Tmax − Tmin )0.5 (Tav + 17.8) (3) surface of unit area in each time step (Rper , mm) may be calculated
 as:
where RAmax is the extraterrestrial radiation of the surface related
to latitude (MJ/m2 /d),  the latent heat of vapor (MJ/kg) which Rper = P − (SIc − SI0 ) − f × t − Sd − Es (6)
 W. Liu et al. / Ecological Modelling 291 (2014) 6–14 9

2.3. Water body outflow simulation runoff volume under the concave green space configuration (Rs ,
m3 ) is expressed as:
The water bodies are devices where water may be temporarily  Ap

or permanently stored. Based on water balance, the outflow from Rs = 1− A · ˛Rimp + Qs + Qw (12)
water bodies (Qw , m3 ) is calculated by: A − Ac
⎧  A 
⎪
⎨ 0,
p
(˛Rimp + ˇRper )A + (P − Ew ) × Ac ≤ H − Vw
A − Ac
Qw =  A   A  (7)
⎪
⎩ p p
(˛Rimp + ˇRper )A + (P − Ew ) × Ac − H + Vw , (˛Rimp + ˇRper )A + (P − Ew ) × Ac > H − Vw
A − Ac A − Ac
where Ap is the rainwater of catchment area that flows into the
water body (m3 ), A the area of the community (m2 ), Ac the sur- Porous pavement is designed to temporarily store surface
face area of the water body (m2 ), ˛ the percentage of impervious runoff, allowing slow infiltration into the subsoil. The infiltration
areas (%), ˇ the percentage of pervious areas (%), Ew the evapora- functions of impervious surface may be restored by replacing it
tion of water surface (mm) estimated by potential evaporation, H with porous pavement, such as permeable bricks with cushion lay-
the water storage capacity of water body (m3 ) and Vw is the stor- ers of gravel underneath. The permeable bricks are precast concrete
age level of water body before rain (m3 ). The infiltration process is block with open voids to allow for infiltration. The runoff from the
not considered in the calculation as most urban water bodies have porous brick pavement (Qp , m3 ) is calculated by
seepage prevention measures. 
Thus, the runoff volume at each time step interval in the com- (P − t + Fc − Hp )˛Aω P − t + Fc > Hp
munity (R, m3 ) is: Qp = (13)
  0 P − t + Fc ≤ Hp
Ap
R= 1− (˛Rimp + ˇRper )A + Qw (8)
A − Ac where  is the infiltration rate of subsoil (mm/min), Fc the accumu-
lated water content volume of subsoil (mm), Hp the water storage
The total storm water runoff volume in the community is the
capacity of porous brick pavement (mm) and ω is the proportion
sum of the runoff at each time step interval.
of porous brick pavement to impervious area. The runoff volume of
2.4. Green infrastructures simulation the porous brick pavement configuration (Rp , m3 ) is expressed as:
 Ap

Green infrastructures include wetlands, ponds, swales, rain- Rp = 1− (˛A(1 − ω)Rimp + Qp + Rper ) + Qw (14)
water tanks, vegetated filter strips, and filter strips. These A − Ac
infrastructures each retains a specified amount of rain “on-site”,
in storage devices or through groundwater recharge. In this study, 3. Model parameters and simulation scenarios
we selected three typical green infrastructures, which are more
suitable for implementing in communities of China, including stor- A community in Haidian district of Beijing is selected as a case
age pond, concave green space and permeable brick pavement to to study impacts of green infrastructures on urban flooding reduc-
evaluate their storm runoff regulatory functions. The overflow cal- tion. The community is 54,783 m2 , 30% of that is pervious green
culation of GI is that substituted the corresponding surfaces then area covered by grasses and the remaining area is impervious sur-
accordingly connected the runoff calculation through coupling with faces of streets, pavements, and structures. This community does
hydrological processes. not have receiving water body. It represents as a typical new built
The storage pond implemented near the storm water drainage community in Beijing. Table 1 summarizes the parameter values
outlet of the community, it can temporally storage the total runoff used for the model simulation and their sources.
and reduce runoff discharge. The outflow from a storage pond (Qr , The total storm runoff volume and peak flow under different
m3 ) is calculated by storm events and green infrastructure configuration are simulated.
 Four storm events representing 1, 2, 5 and 10 year of recurrence
Rc − (Hr − Vr0 ) Rc > Hr − Vr0 intervals are selected. The 24-h rainfalls for the events are 45.6,
Qr = (9) 72.0, 115.2, and 158.3 mm, respectively. The patterns of the precipi-
0 Rc ≤ Hr − Vr0
tation events are modeled according to the storm intensity formula
where  is the rainwater collection ratio (%), Rc the accumulated of Beijing and Pearson type III curve.
runoff at time step (m3 ), Hr the water storage capacity of the storage Five green infrastructure configurations are considered: (1)
pond (m3 ) and Vr0 is the water volume of the storage pond prior to increasing green space area from 30% to 40%; (2) constructing a
rain (m3 ). The runoff volume under the storage pond configuration 1500 m3 storage pond; (3) changing green area from flat to con-
(Rr , m3 ) is expressed as: caved with 5-cm depth; (4) converting 50% of impervious area into
the porous brick pavement; (5) combining the configurations of
Rr = (1 − )R + Qr (10) previous four. The reductions of runoff volumes and peak flows by
The function of concave green space is similar to swale, which implementing the GI configurations are evaluated and compared
retains storm water inside, increase infiltration and prevents the with the outcomes of no GI practice (i.e., base scenario). The simu-
runoff flowing to the adjacent road surfaces. Assume all the pervi- lation runs at a 5-min time step on the basis of storm inputs.
ous areas (green spaces) are reformed concave green spaces. The
runoff from concave green space (Qs , m3 ) is calculated by 4. Results and discussion

(it + q − fs − (1/2)hs )ˇA it + q − fs > (1/2)hs 4.1. Runoff and peak flow under different storm events
Qs = (11)
0 it + q − fs ≤ (1/2)hs
The rainfall (input) and the model predicted runoff (output)
where q is the storage level on the concave green space of the pre- hydrographs of 1-year storm recurrence interval under the base
vious time interval (mm), fs the infiltration of the concave green scenario is used to illustrate the behaviors of precipitation vs. runoff
space (mm) and hs is the depth of concave green space (mm). The (Fig. 2). The dynamics of storm runoff were coherent with storm
10 W. Liu et al. / Ecological Modelling 291 (2014) 6–14

Table 1
Parameters for the model simulation.

Parameters Values Units Source

Community’s characters
Total area 54,783 m2 Based on local investigation
Percentage of impervious areas 70 % Based on local investigation
Percentage of pervious areas 30 % Based on local investigation
Water bodies area 0 m2 Based on local investigation
Water storage capacity of water body 0 m3 Based on local investigation

Green infrastructures characters

Water storage capacity of storage pond 1500 m3 Assumed
Rainwater collection ratio 100 % Assumed
Depth of concave green space 50 mm Assumed
Infiltration rate of subsoil 0.3 mm/min Wang (2007)
Water storage capacity of porous brick pavement 32.86 mm Wang (2007)

Meteorological conditions
◦
Maximum daily temperature 31 C Beijing meteorological data
◦
Minimum daily temperature 22 C Beijing meteorological data
◦
Average daily temperature 26.5 C Beijing meteorological data

Soil properties
Saturated hydraulic conductivity 0.144 mm/min Xie et al. (1998)
Wetting front suction 69.696 mm Fu et al. (2002)
Saturated water content 40.627 % Xie et al. (1998)
Initial water content 26.279 % Xie et al. (1998)

Vegetation characters
Leaf area index 3.85 – Su and Xie (2003)
Extinction coefficient 0.3 – Wang et al. (2008)
Special leaf storage 0.2 mm Wang et al. (2008)
Runoff yield parameters
Depression of impervious area 3 mm Xu (1998)
Depression of pervious area 4 mm Lei et al. (2010)

events, suggesting the model is capable of simulating the hydrolog- The model simulations track the water balance of storm events
ical processes. Under 1-, 2-, 5-, and 10-year storm return intervals, through processes including interception, infiltration, evaporation,
the maximum storm intensities were 0.9, 1.9, 2.5 and 2.8 mm/min, depression-storage, and runoffs from impervious and pervious
respectively, resulting in peak runoff flows of 47.3, 97.8, 131.1 and areas (Fig. 4). In this case, runoffs from the impervious area which
152.4 m3 /min. As the intensity increased, the peak runoff increased occupied 70% of the total surface accounted for 58.6% to 66.8% of the
from 91.9% to 97.9% of the maximum rainfall, respectively for the total precipitation. Runoffs from the pervious area (30% of the total
four storm events. surface) were considerably lesser and accounted for 3.3–12.7% of
The temporal changes of accumulative runoff under the 1-, 2-, 5-, the precipitation also depending on the storm recurrence periods.
and 10-year storm recurrence intervals on base scenario are illus- Infiltration and evaporation were important processes of the urban
trated in Fig. 3. The runoff volumes were 1546.9, 2813.3, 4858.0, and runoff hydrology. Proportional to storm intensities, infiltration and
6894.8 m3 under the 1-, 2-, 5-, and 10-year storm recurrence inter- evaporation ranged inversely from 25.4 to 16.6% and 10.0% to 2.9%,
vals respectively. The runoff volume ratio increased from 61.9% to respectively. The loss through vegetation canopy interception and
79.5% with the storm intensity increased from 1- to 10-year recur- depressions were minor, accounting for 0.5–0.2% and 2.1–0.8%,
rence periods. respectively.
The runoff generation ratios of impervious area, from 83.8%to
95.4%, were larger than those of the pervious area, from 10.9% to
400 0
42.4%. It indicated that the impervious surfaces not only have the
largest footprint, but also have the most contribution to the storm
2 runoffs of the community. Installing green infrastructures and/or
increasing green spaces would have a profound impact on mitigat-
300 4 ing urban flooding.
Stormwater runoff (m 3)

Rainfall (mm)

200
8 4.2. Green infrastructures impacts on storm runoff volume and
peak flow
10
100 4.2.1. Green space
12
If the green spaces in the community were converted to impervi-
14 ous surfaces through urban expansions, the runoff volumes would
0 be 2093.3, 3538.6, 5903.7, and 8273.7 m3 , and peak flows would
0 200 400 600 800 1000 1200 1400 1600 be 51.3, 101.7, 134.5 and 155.4 m3 /min under 1-, 2-, 5-, and 10-
Time (min)
year storm events, respectively. Compared with the outcomes of
Fig. 2. Dynamic changes of rainfall versus runoff simulation of 1-year recurrence
the base scenario, the runoff volumes increased 35.3–20.0%, and
interval storm. the peak flows have increased 8.4–2.0%, respectively. With the
 W. Liu et al. / Ecological Modelling 291 (2014) 6–14 11

10000 10000

Storm of 1-yr recurrence interval Storm of 2-yr recurrence interval

8000 8000

Accumulated volume (m 3)
Accumulated volume (m 3)

Runoff
6000 Rainfall 6000

4000 4000

2000 2000

0 0
0 200 400 600 800 1000 1200 1400 1600 0 200 400 600 800 1000 1200 1400 1600
Time (min) Time (min)
10000 10000

Storm of 5-yr recurrence interval Storm of 10-yr recurrence interval

8000 8000
Accumulated volume (m 3)

Accumulated volume (m 3)
6000 6000

4000 4000

2000 2000

0 0
0 200 400 600 800 1000 1200 1400 1600 0 200 400 600 800 1000 1200 1400 1600
Time (min) Time (min)

Fig. 3. Cumulative rainfall versus runoff, storms of 1-, 2-, 5-, and 10-year recurrence intervals.

conversion, the community would become more susceptible to 128.8 and 150.4 m3 /min, reduced 5.6–1.3% respectively under 1-,
water logging and flooding. 2-, 5-, and 10-year storm events.
If the green area of the community was expanded from the These results indicated that the reduction effectiveness of green
existing 30–40% of the surfaces, comparing to those of the base space increase was relatively small in terms of storm runoff volume
scenario, the runoff volumes would be reduced 11.8–6.7%, and and peak flow under all storm events. Thus, the urban flooding
the peak flows would be 46.0, 96.6, 129.9 and 151.4 m3 /min that could be reduced by considerably increased the green spaces of
suggested 2.8–0.7% reductions respectively under the 1-, 2-, 5-, the community. Unfortunately, given the facts that the options of
and 10-year storm events (Fig. 5A). When the green space fur- expanding green space were limited for most communities in Bei-
ther increases to 50% of the surfaces, the runoff volumes would be jing, it was not practical to resolve storm runoff and flooding by
reduced 23.6–13.3%, and the peak flows would become 44.7, 95.3, this approach alone.

Fig. 4. The relative percentage figure for mass balance of rainfall-runoff process.
12 W. Liu et al. / Ecological Modelling 291 (2014) 6–14

100 100
A B

Percentage of reduction (%)

Percentage of reduction (%)
80 80

60 60

40 40

20 20

0 0
1-yr 2-yr 5-yr 10-yr 1-yr 2-yr 5-yr 10-yr
Recurrence Interval Recurrence Interval
100 100
C D
Percentage of reduction (%)

Percentage of reduction (%)

80 80

60 60

40 40

20 20

0 0
1-yr 2-yr 5-yr 10-yr 1-yr 2-yr 5-yr 10-yr
Recurrence Interval Recurrence Interval

100
E
Percentage of reduction (%)

Total runoff
80
Peak flow

60

40

20

0
1-yr 2-yr 5-yr 10-yr
Recurrence Interval

Fig. 5. Reductions of volume and peak flow of storm runoffs achieved under green infrastructures: (A) expanding green space from 30% to 50% of the surface area, (B)
converting to concave green space with 5-cm depth, (C) providing 1500 m3 runoff storage facility, (D) converting from impervious to porous brick pavement, and (E)
combining all previous scenarios.

4.2.2. Concave green space reduction, infiltration and groundwater recharge would be greatly
The ground surface of urban green spaces in China usually enhanced.
is convex and higher than the surrounding roads. As a result,
the storm water drains of the pervious green space and spills 4.2.3. Storage pond
on to the impervious road surfaces easily became runoffs. If the The storage pond may be constructed to capture the storm water
ground surface of green spaces was shaped in concave form, the thus reduce the runoff. Comparing to the base scenario, installing
storm water would be prevented from flowing to the impervious 1500 m3 of runoff storage capacity, reduced the runoff volumes by
road surfaces. We assumed the existing green spaces of the com- 97.0%, 53.3%, 30.9%, and 21.8%. Meanwhile, the peak flows were 1.6,
munity were reshaped with a depth of 5 cm. Comparing to the 20.9, 81.7 and 152.4 m3 /min, reduced 96.6%, 78.7%, 37.7%, and 0%,
base scenario, there would be no outflow from the concave green respectively, under 1-, 2-, 5-, and 10-year storm events (Fig. 5C).
space. The runoff volumes reduced 5.3–16.0% respectively under The results implied that the flooding reductions of storage pond
1-, 2-, 5-, and 10-year storm events. The peak flows were 35.9, were effective in small storms, but ineffective in large storms.
71.2, 94.2 and 108.7 m3 /min that reduced 24.2–28.6% respectively For a storm of 1-year recurrence interval, almost all of the storm
under 1-, 2-, 5-, and 10-year storm events (Fig. 5B). The concave water of the community may be retained in the storage pond.
shaped green space would temporarily store the storm water and To contain the whole runoff discharge, the storage capacity must
allow it to infiltrate over time, consequently increase groundwa- reach at least 1547.0, 2813.3, 4858.0, and 6,894.8 m3 , respectively,
ter recharge. The implementation would maximize the infiltration for the recurrence intervals of 1-, 2-, 5-, and 10-year. There is a
capacity of the green space. If the flat or raised green spaces are serious water shortage in Beijing. The stored (harvested) storm
changed to be concave to increase infiltration and retain partial water could be used in the community for green land irrigation,
impervious runoff, the ecological function in terms of storm runoff toilet flushing, and car washing. If the majority of communities
 W. Liu et al. / Ecological Modelling 291 (2014) 6–14 13

implemented the storage facilities (i.e., the decentralized, small Mentens et al. (2006) pointed out that a single green infra-
sized cisterns and tanks, and underground storage ponds) to har- structure alone will never fully eliminate the urban runoff and
vest rainwater for non-potable uses, water scarcity and flooding it needs to be combined with other runoff reduction measures.
problem in Beijing will be partly mitigated. Many communities in The outcomes of model simulation illustrate how runoff reduc-
Beijing have adopted this green infrastructure and proved to be tion measures should be integrated to optimize the effectiveness of
effective. In practice, adoption of storage facilities is depended on flooding reduction. The reduction effectiveness of GI implementing
the communities’ space and the potential uses of harvested rain- in other communities such as the older (10% pervious percentage)
water. Given that the high cost of potable water in Beijing, the and villa (50% pervious percentage) communities will have linear
storm water storage facilities were cost effective and should be changes due to the different contribution of impervious/pervious
encouraged in communities for sustainable water management. area to the total runoff.

4.2.4. Porous pavement

The impervious surface can be replaced with porous brick pave- 5. Conclusions
ment to decrease runoff generation. The infiltrate rate of permeable
brick far exceeds the intensities of ordinary rains allowing the storm Rapid urbanization has dramatically increased the volume and
water stored in the cushion layer and then gradually infiltrate into peak flow of storm runoffs that exposing communities to greater
soil strata below. If 50% of impervious surfaces were converted risks of flood hazards. Climate change further increased the extrem-
into porous brick pavements, the runoff volumes were reduced ities of precipitation events leading to higher frequencies and
46.2–42.0%, and the peak flows were reduced 37.9–35.7%, respec- intensities of floods. The green infrastructures softened the impacts
tively under 1-, 2-, 5-, and 10-year storm events (Fig. 5D). If all the of urban runoffs and helped to retain water in urban ecosystems.
impervious areas except for the rooftop area (accounting for 20% Increasing green spaces, reducing impervious surfaces, and build-
of the community’s footprint) were replaced by porous pavement, ing storm water retention facilities can ease the deleterious impacts
the runoff volumes and peak flows could reduce 66.5–59.6% and brought by the rapid urbanization. After the GI practices applied
54.2–51.0%, respectively, under the storm recurrence intervals. on a broad scale, it will enhance the ecological function of urban
The porous brick pavements were more effective to reduce the ecosystem, promote sustainable living and supports appropriate
storm runoffs than increasing green spaces or converting to con- urban development.
cave shape green spaces, because the runoffs from impervious area Management tools that quantify how GI configuration influ-
account for more than half of the total runoff. Converting the imper- ences storm runoff reduction were needed. We incorporate a few
vious surfaces to pervious surfaces effectively reduced the runoffs hydrological models’ algorithms, and developed a simple model
from impervious surfaces. In Beijing, impervious surfaces such as which can well address the functions of GI on the storm water
parking lots, roads, courtyards, and public squares were found reduction but with less parameters input. The simulated results
throughout the most communities. They leaded to large runoff vol- showed the model performs in simulating storm water runoffs
umes and high peak flows exasperating the flooding and even traffic were responded to varying recurrence intervals. The outcomes of
jams. Converting them to permeable surfaces would significantly scenario simulations showed the effects of single GI facility on
reduce the impervious runoff and improve current situations. flooding reduction were limited under large storms. The integrated
GI configurations were effectively reduced in all storm events.
4.2.5. Integrated approach Therefore, in order to provide effective reductions for an entire
Model simulation results illustrated that a single green infra- spectrum of storm events to control urban flooding, a more effec-
structure approach was not capable of effectively reducing both the tive strategy would be to implement integrated GI facilities. Reform
runoff volume and peak flow. The storage pond was highly effec- or planning urban landscape with appropriate green infrastruc-
tive at controlling the volume and peak flow of runoff of 1-year ture practices throughout communities is important for healthy
recurrence interval storm, however its storage capacity was inflex- and sustainable development.
ible and extremely limited. In urban communities, there was little
spare room to retrofitting a runoff storage facility with adequate Acknowledgements
capacities. Therefore, single GI facility cannot be feasibly and effec-
tively performed to control urban flooding. Options of integrated We are thankful for the financial support of the National Nat-
green infrastructures were imperative to minimize the impacts of ural Science Foundation of China (#41271501) and the National
storm runoffs. If we incorporated all of the previously examined Science & Technology Pillar Program during the Twelfth Five-year
green infrastructures, the storm water runoff of the 1- and 2-year Plan Period (#2012BAC13B04) to this research.
recurrence intervals was completely eliminated and volume and
peak flow of storm runoffs of 5- and 10-year recurrence interval
storm runoffs reduced by 94.2% and 85.6%, and 97.1% and 93.1%, References
respectively (Fig. 5E). The GI integration reduced significantly both
the volume and peak flow of storm runoffs. The reduction effective- Ahiablame, L.M., Engel, B.A., Chaubey, I., 2012. Effectiveness of low impact develop-
ment practices: literature review and suggestions for future research. Water Air
ness was more than 85%. In this manner, the community essentially Soil Pollut. 223 (7), 4253–4273, http://dx.doi.org/10.1007/s11270-012-1189-2.
does not have a need of discharging storm runoffs. The integra- Alfredo, K., Montalto, F., Goldstein, A., 2009. Observed and modeled performances of
tion of GI facilities enhanced the reduction effectiveness under prototype green roof test plots subjected to simulated low-and high-intensity
precipitations in a laboratory experiment. J. Hydrol. Eng. 15 (6), 444–457,
different storms. Therefore, implementing the integrated GI facili-
http://dx.doi.org/10.1061/(ASCE)HE. 1943-5584.0000135.
ties in communities may have considerable contribution to urban Benedict, M.A., McMahon, E.T., 2002. Green infrastructure: smart conserva-
flooding control. Besides, it increased rainwater infiltrated into soil, tion for the 21st century. Renew. Resour. J. 20 (3), 12–17, Retrieved from
http://www.sprawlwatch.org/greeninfrastructure.pdf
stored and utilized the storm water resources that saved the water
Benedict, M.A., McMahon, E.T., 2006. Green Infrastructure: Linking Landscapes and
consumption in the community. The integrated GI implemented Communities. Island Press.
throughout communities would maintain the urban ecosystem’s Bettencourt, L., West, G., 2010. A unified theory of urban living. Nature 467 (7318),
hydrologic and ecological functions, and allow water resources to 912–913, http://dx.doi.org/10.1038/467912a.
Chapman, C., Horner, R.R., 2010. Performance assessment of a street-
be maintained that provide ecological sinks to mitigate environ- drainage bioretention system. Water Environ. Res. 82 (2), 109–119,
mental change. http://dx.doi.org/10.2175/106143009X426112.
14 W. Liu et al. / Ecological Modelling 291 (2014) 6–14

Chen, S.Q., Chen, B., Fath, B.D., 2014. Urban ecosystem modeling and global change: Mein, R.G., Larson, C.L., 1973. Modeling infiltration during a steady rain. Water
potential for rational urban management and emissions mitigation. Environ. Resour. Res. 9 (2), 384–394, http://dx.doi.org/10.1029/WR009i002p00384.
Pollut. 190, 139–149, http://dx.doi.org/10.1016/j.envpol.2014.03.032. Mell, I.C., 2009. Can green infrastructure promote urban sustainability? Proceedings
Chen, S.Q., Chen, B., Fath, B.D., 2013. Ecological risk assessment on the system scale: of the ICE-Engineering Sustainability 162 (1), 23–34, http://dx.doi.org/
a review of state-of-the-art models and future perspectives. Ecol. Model. 250, 10.1680/ensu.2009.162.1.23.
25–33, http://dx.doi.org/10.1016/j.ecolmodel.2012.10.015. Mentens, J., Raes, D., Hermy, M., 2006. Green roofs as a tool for solving the rainwa-
Dreelin, E.A., Fowler, L., Ronald Carroll, C., 2006. A test of porous pavement effec- ter runoff problem in the urbanized 21st century? Landsc. Urban Plann. 77 (3),
tiveness on clay soils during natural storm events. Water Res. 40 (4), 799–805, 217–226, http://dx.doi.org/10.1016/j.landurbplan.2005.02.010.
http://dx.doi.org/10.1016/j.watres.2005.12.002. Mitchell, V., Diaper, C., Gray, S.R., Rahilly, M., 2003. UVQ: modelling the movement of
Droogers, P., Allen, R.G., 2002. Estimating reference evapotranspiration water and contaminants through the total urban water cycle. In: Paper presented
under inaccurate data conditions. Irrig. Drainage Syst. 16 (1), 33–45, at the 28th International Hydrology and Water Resources Symposium: About
http://dx.doi.org/10.1023/A:1015508322413. Water; Symposium Proceedings.
Dunn, A.D., 2010. Siting green infrastructure: legal and policy solutions to alleviate MUSIC Development Team, 2003. MUSIC User Guide Version 2. 0. Cooperative
urban poverty and promote healthy communities. BC Envtl. Aff. L. Rev. 37, 41. Research Centre for Catchment Hydrology, Melbourne, Australia.
Ferguson, B.K., 1998. Introduction to Stormwater: Concept, Purpose, Design. Wiley Qin, H.-p., Li, Z.-x., Fu, G., 2013. The effects of low impact development on
Com. urban flooding under different rainfall characteristics. J. Environ. Manage. 129,
Foster, J., Lowe, A., Winkelman, S., 2011. The Value of Green Infrastructure for Urban 577–585, http://dx.doi.org/10.1016/j.jenvman.2013.08.026.
Climate Adaptation. Center for Clean Air Policy, February. Retrieved from- Schneider, L.E., McCuen, R.H., 2006. Assessing the hydrologic perfor-
http://ccap.org/assets/The-Value-of-Green-Infrastructure-for-Urban-Climate- mance of best management practices. J. Hydrol. Eng. 11 (3), 278–281,
Adaptation CCAP-Feb-2011.pdf http://dx.doi.org/10.1061/(ASCE)1084-0699(2006)11:3(278).
Fu, S.H., Liu, B.Y., Wu, J.D., Duan, S.H., 2002. Study on runoff calculation method Su, F., Xie, Z., 2003. A model for assessing effects of climate change on
in Beijing mountainous area. Dili Kexue (Scientia Geographica Sinica) 22 (5), runoff in China. Progr. Nat. Sci. 13 (9), 701–707, http://dx.doi.org/10.1080/
604–609 (in Chinese). 10020070312331344270.
Graham, P., Maclean, L., Medina, D., Patwardhan, A., Vasarhelyi, G., 2004. The role Villarreal, E.L., Semadeni-Davies, A., Bengtsson, L., 2004. Inner city stormwater
of water balance modeling in the transition to low impact development. Water control using a combination of best management practices. Ecol. Eng. 22 (4),
Qual. Res. J. Ca 39 (4), 331–342, Retrieved from- https://www.cawq.ca/journal/ 279–298, http://dx.doi.org/10.1016/j.ecoleng.2004.06.007.
temp/article/121.pdf Wang, X.X., 2007. Assessing the runoff intercepting capacity of greenbelts of differ-
Gill, S.E., Handley, J.F., Ennos, A.R., Pauleit, S., 2007. Adapting cities for climate ent underlying surfaces in residential district. In: Thesis of China Agricultural
change: the role of the green infrastructure. Built Environ. 33 (1), 115–133, University, Beijing (in Chinese).
http://dx.doi.org/10.2148/benv.33.1.115. Wang, J., Endreny, T.A., Nowak, D.J., 2008. Mechanistic simulation of tree effects in an
Goonetilleke, A., Thomas, E., Ginn, S., Gilbert, D., 2005. Understanding the role of urban water balance model1. JAWRA J. Am. Water Resour. Assoc. 44 (1), 75–85,
land use in urban stormwater quality management. J. Environ. Manage. 74 (1), http://dx.doi.org/10.1111/j. 1752-1688.2007.00139.x.
31–42, http://dx.doi.org/10.1016/j.jenvman.2004.08.006. White, R., 2002. Building the Ecological City. CRC Press.
Hargreaves, G.H., Samani, Z.A., 1985. Reference crop evapotranspiration from ambi- Whitford, V., Ennos, A.R., Handley, J.F., 2001. ‘City form and natural
ent air temperature. Am. Soc. Agric. Eng. 1 (2), 96–99. process’—indicators for the ecological performance of urban areas and
Huber, W.C., Dickinson, R.E., Barnwell Jr., T.O., Branch, A., 1988. Storm water man- their application to Merseyside, UK. Landsc. Urban Plann. 57 (2), 91–103,
agement model, version 4. Environ. Protect. Agency 600, 3–88, Retrieved from http://dx.doi.org/10.1016/S0169-2046(01)00192-X.
http://eng.odu.edu/cee/resources/model/mbin/swmm/swmm 1.pdf Xie, Y., Huang, G., Zhao, L., 1998. Spatial variability of field soil properties. Zhongguo
Lei, X.H., Jiang, Y.Z., Wang, H. (Eds.), 2010. The Distributed Hydrological Model of Nongyedaxue Xuebao (J. China Agric. Univ.) 3 (2), 41–45 (in Chinese).
EasyDHM. China WaterPower Press, Beijing (in Chinese). Xu, X., 1998. Simulation of storm runoff process for plain urban. Shuili Xuebao (J.
Linsley, R.K., Kohler, M.A., Paulhus, J.L., 1949. Applied Hydrology. McGraw-Hill, New Hydraulic Eng.) 8 (8), 34–37 (in Chinese).
York.