Science of the Total Environment 643 (2018) 301–311

Contents lists available at ScienceDirect

Science of the Total Environment

journal homepage: www.elsevier.com/locate/scitotenv

Effects of urbanization on direct runoff characteristics in urban

functional zones
Chunlin Li a, Miao Liu a,⁎, Yuanman Hu a, Tuo Shi a,b, Xiuqi Qu a,b, M. Todd Walter c
a
CAS Key Laboratory of Forest Ecology and Management, Institute of Applied Ecology, Chinese Academy of Sciences, Shenyang 110016, China
b
University of Chinese Academy of Sciences, Beijing 100049, China
c
Department of Biological and Environmental Engineering, Cornell University, 62 Riley-Robb Hall, Ithaca, NY 14853, USA

H I G H L I G H T S G R A P H I C A L A B S T R A C T

• An improved method to evaluate direct

runoff in urban areas is proposed.
• Urbanization affected the direct runoff
increments and flooding hazard distri-
bution.
• The direct runoff has significant positive
spatial autocorrelation.
• Urbanization factors have less impact
than environmental and meteorological
factors.

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

Article history: As urbanization processes, the increasing direct runoff caused by land use change has become a major challenge
Received 5 December 2017 for urban hydrological system. In this study, the impact of urbanization on direct runoff in the Shenyang urban
Received in revised form 21 May 2018 area was investigated using a modified Soil Conservation Service Curve Number model combined with remote
Accepted 17 June 2018
sensing. Urban functional zone (UFZ) was used as the basic unit for hydrological analysis. The hydrological
Available online xxxx
changes in runoff were analyzed by calculating the runoff difference between the current condition and the
Editor: Ouyang Wei pre-urbanization condition. Moran's I was used to estimate the spatial autocorrelation of the entire area. Then
we assessed the relative influence and marginal effects of factors affecting direct runoff using boosted regression
Keywords: trees (BRT). Our results showed that direct runoff was significantly related to urbanization. Under current condi-
Direct runoff tions, direct runoff increment depth affected by urbanization in the study area was 68.02 mm. For different UFZs,
Urban functional zones high-density residential, business and industrial zones tended to have large runoff volumes and high runoff co-
Boosted regression trees efficients. Through flooding hazard analysis, we found about 6.53% of the study area fell into a significant hazard
Hydrology category. The industrial zone had largest area of significant hazard land (40.97 km2) and the business zone had
Urbanization
the largest significant hazard percentage (21.19%). Moran's I results illustrated that the high-high clusters in
Shenyang were mainly concentrated in the urban center. BRT analysis indicated that runoff had the strongest cor-
relation with rainfall (52.07%), followed by impervious ratio (27.28%), normalized difference vegetation index
(14.31%), antecedent 5-day rainfall (3.02%), and UFZs (1.70%). The industrial zone, business zone and high-
density residential zone tend to have greater influence on runoff. Our study could present method for recognizing

⁎ Corresponding author.
E-mail address: lium@iae.ac.cn (M. Liu).

https://doi.org/10.1016/j.scitotenv.2018.06.211
0048-9697/© 2018 Elsevier B.V. All rights reserved.
302 C. Li et al. / Science of the Total Environment 643 (2018) 301–311

hotspots of direct runoff in large city, and may provide potential implications for green infrastructure selection
and urban planning.
© 2018 Elsevier B.V. All rights reserved.

1. Introduction areas (Yao et al., 2015). To analyze the quality and distribution patterns
of direct runoff caused by urbanization, we used urban functional zones
Urbanization is a worldwide phenomenon, with population density (UFZs) as the spatial scale (Sanders, 1986). UFZs are the spatial patterns
continuing to grow and urban area to expand. By 2030, urban land cover of a city related to its urban functions (Tian et al., 2010). Each type of
will increase by 1.2 million km2, nearly tripling the global urban land UFZ comprises many zones with both similar structural characteristics
area circa 2000 (Seto et al., 2012). The increase of urbanization associ- and similar socioeconomic functions. Each zone is organized by a cluster
ated with population growth is one of the major changes affecting of land uses, and its function is determined by the dominant land use.
land use in big cities (Jacqueminet et al., 2013). Urban sprawl inevitably Similar urban spatial structures, human activity types and urban func-
leads to non-urban land being converted to built-up areas, resulting in a tions of same UFZ result in similar hydrological characteristics, which
significant increase in the proportion of impervious surfaces from roads, makes the UFZ a suitable scale for evaluating the hydrological impact
rooftops, parking lots and other urban surfaces (Braud et al., 2013b; of urbanization.
Mejia and Moglen, 2010b). The process of urbanization can alter In this study, a modified SCS-CN model combined with remote sens-
urban hydrological responses and negatively impact surface and down- ing was used to estimate the effect of urbanization on direct runoff var-
stream waters owing to the introduction of impervious surfaces iation in the Shenyang urban area. There were two major goals for this
(Zampella et al., 2007), removal of deep rooted vegetation and alter- paper: (1) to evaluate the hazard and spatial autocorrelation of direct
ations (Hibbs and Sharp, 2012) to the natural drainage network (Zhou runoff, and (2) to identify the main factors affecting runoff and analyze
et al., 2013). These can result in losses of infiltration, increased surface the changes of marginal effects.
runoff (Angrill et al., 2017; Fletcher et al., 2013; Weng, 2001), and the
potential to produce huge floods (Huang et al., 2008; Olang and Furst, 2. Materials and methods
2011; Quan et al., 2010; Richert et al., 2011; Yao et al., 2017). Several
studies have demonstrated that the flooding hazard in an urban area 2.1. Study area and data
can be partly attributed to the rapid replacement of natural ecosystems
by impervious urban surfaces (Hu, 2016; Kvočka et al., 2016; Shepherd, Shenyang is the largest and most important industrial city in North-
2006; Zhang et al., 2015). east China (41°11′51″–43°02′13″N, 122°25′09″–123°48′24″E). Mean
Research and practice in the last decades has shown that the impacts annual precipitation is 510–680 mm, most of which falls from June to
of urbanization on the hydrological cycle are strongly related to imper- August. The urban sprawl of Shenyang has gradually expanded from
vious surfaces (Mejia and Moglen, 2010a). However, the rainfall–runoff the central to suburban areas, and a four-ring road network was created
relationship is highly nonlinear and complex, and is dependent on nu- during the last 30 years. The entire urban area of Shenyang (called the
merous factors such as antecedent soil moisture, evaporation, infiltra- four-ring area) can be divided into four areas (zones 1–4) according to
tion and rainfall duration (Guan et al., 2016; Isik et al., 2013; Sajjad the four-ring roads (Fig. 1). The stages of urbanization are represented
et al., 2015; Zhang et al., 2012). Many studies have investigated the hy- from the first to the fourth ring areas. Our study covered the whole
drological impact of urbanization based on field data (Choi et al., 2016; four-ring area. The area of water was excluded because it did not pro-
Zhang et al., 2013). Gallo et al. (2013) assessed the effect of urban land vide direct runoff. The four-ring area is the fastest growing urbanization
cover on hydrological responses using summer runoff data from five section with most of the population and built-up areas.
catchments dominated by distinct urban land uses and found it was UFZs are the largest elements of landscape planning related directly
tightly coupled to the magnitude of rainfall (Gallo et al., 2013). Braud to functions such as dwellings, education, industry and commerce (Sun
et al. (2013a) and Braud et al. (2013b) explored several indicators to et al., 2013). QuickBird satellite images from 2015 were used to manu-
demonstrate the impact of urbanization on discharge series, and the re- ally interpret the UFZs of the Shenyang four-ring area. First, we used
sults showed a decrease of specific discharge from upstream to down- the main roads to divide the entire area into blocks, and assumed each
stream corresponding to an increase in artificial areas, except during block had the same urban function, forming one UFZ. Second, we deter-
high flows (Braud et al., 2013a). Putro et al. (2016) used historical mined the main functions of these UFZs by looking up the buildings in
data to identify the impact of climate and urbanization on selected the block on Google Maps, and eventually classified 9278 UFZs. Finally,
water quantity and quality indicators, and results indicated an upward referring to the definitions of UFZs by other researchers (Sun et al.,
trend in runoff totals in urban catchments but not in rural catchments 2013; Yao et al., 2015), we grouped the study area into 11 types of
(Putro et al., 2016). Many studies assessing the hydrological impacts UFZs. The area and description of UFZs are shown in Table 1. Field veri-
and storm water management activities of urbanization have applied fications were conducted for indeterminate UFZs. Because the WTR area
hydrological models (Choi et al., 2003; Li et al., 2016; Qin et al., 2016), cannot produce direct runoff, we did not include the WTR zones in our
such as full distributed process-based models (e.g. MIKE SHE, RHESSys, study.
TOPLATS, and WASIM), physically-based semi-distributed models (e.g.
SWAT, SWMM, and HYLUC), and conceptual lumped models (e.g.
IHACRES and NAM). The Soil Conservation Service Curve Number 2.2. Direct runoff evaluation
(SCS-CN) model, developed by the U.S. Department of Agriculture in
1954, is one of the most widely used empirical hydrological models In this study, direct runoff was simulated with the Mishra and Singh
for computing the volume of direct surface runoff (Jiao et al., 2015). It (MS) model which is a revised SCS-CN model using a separate expres-
is also an effective tool for assessing direct runoff in large urban areas sion based on antecedent 5-day rainfall to estimate antecedent moisture
which lack observed data (Ansari et al., 2016; Bartlett et al., 2016; (Mishra et al., 2002). The MS model equations are as follows:
Sahu et al., 2012; Singh et al., 2013; Tsihrintzis and Hamid, 1997). 8
Owing to clustered and unplanned development of urban areas, < ðP−Ia ÞðP−Ia þ M Þ
Q¼ ; P NIa ð1Þ
urban resources (population, impervious surface, building density, : P−Ia þ M þ S
green area, and so on) are disproportionately distributed within urban 0; P ≤Ia
 C. Li et al. / Science of the Total Environment 643 (2018) 301–311 303

Fig. 1. Regional location and 2015 urban functional zones (UFZs) of Shenyang City, China. Zone 1 is encircled by the first ring road. Zone 2 is the area between the first and second ring roads.
Zone 3 represents the area between the second and third ring roads. Zone 4 is the area between the third and fourth ring roads. The four-ring area is the entire area of zones 1–4. HRZ: high-
density residential zone; LRZ: low-density residential zone; INZ: industrial zone; BUZ: business zone; GOZ: government zone; PSZ: public service zone; AGZ: agricultural zone; DEZ: de-
velopment zone; GSZ: green space zone; PRZ: preservation zone; WTR: water.

25400 We assumed each pixel of urban land cover was composed of veg-
S¼ −254; 0 ≤CN ≤100 ð2Þ
CN etation, impervious surfaces, and bare soil according to the
vegetation-impervious surface-soil (V-I-S) model presented by
where Q is direct runoff depth (mm); P is precipitation depth (mm); S is Ridd in 1995 (Ridd, 1995). The linear spectral mixture analysis
potential maximum soil moisture retention (mm), defined by the di- (LSMA) method was used to evaluate the proportion of V-I-S in
mensionless CN parameter; Ia is the initial abstraction (mm), and Ia = each pixel from a Landsat 8 OLI image of 2015 in the Shenyang
λS. The value of λ was set to 0.2 in this study. M is antecedent moisture four-ring area. To verify the accuracy of the proportions of vegeta-
(mm), and antecedent 5-day rainfall is used to estimate M: tion, impervious surface, and soil components, we selected 50 ran-
8 dom points with area 300 m × 300 m. For each sample point, the
< SI ðP 5 −λSI Þ vegetation, impervious surface, and soil of 2015 were visually
; P ≥λSI
M ¼ P 5 þ ð1−λÞSI 5 ð3Þ interpreted on high-resolution images from Google Earth. The accu-
:
0; P 5 bλSI racy of the vegetation, impervious surface and soil maps were
assessed by root mean square error (RMSE) which is comparing the
P5 is antecedent 5-day rainfall (mm), and SI is potential maximum visual interpretation proportions from Google Earth and LSMA re-
retention in dry conditions (AMC I). SI can be taken as approximately sults. RMSEs for vegetation, impervious surface and soil were 0.18,
equal to the absolute potential maximum retention (S0): 0.20, and 0.24, respectively, which showed the estimations of vege-
tation, impervious surface, and soil from LSMA were good enough
SI ¼ S0 ¼ S þ M ð4Þ for further calculating.

Table 1
Classification and characteristics of 2015 urban functional zones (UFZs) in the Shenyang four-ring area.

UFZ Code Area (km2) (area Description

proportion, %)

High-density residential zone HRZ 181.46 (14.75) Typical residential communities in Shenyang urban area, including multiple family
houses and apartments with dense population.
Low-density residential zone LRZ 68.13 (5.55) Low residential buildings with lower impervious fraction and sparse population.
Industrial zone INZ 213.82 (17.42) Industrial factories, repair stations and storage land.
Business zone BUZ 39.08 (3.18) Offices for finance organizations and companies, shopping centers, restaurants and hotels.
Government zone GOZ 43.21 (3.52) Government buildings, public organization, research institutes and campuses.
Public service zone PSZ 30.74 (2.50) Public services area, such as hospitals, libraries and museums.
Agricultural zone AGZ 425.56 (34.66) Cultivated land, greenhouses and Agricultural facilities.
Development zone DEZ 45.75 (3.73) Undeveloped area and construction site.
Green space zone GSZ 103.33 (8.42) Urban parks, golf courses and scenic area with high green coverage.
Preservation zone PRZ 37.52 (3.06) Open space with natural and artificial green space, such as forest parks and natural preservation zones.
Water WTR 39.08 (3.18) Open water, including rivers, reservoirs and lakes.
304 C. Li et al. / Science of the Total Environment 643 (2018) 301–311

An improved composite CN method proposed by Fan et al. (2013) spatial scales (m2), such as the entire study area, four zones, or
was used to calculate composite CN. The composite CN value of each UFZs.
pixel can be computed as the area-weighted average of the CN values
of impervious surfaces, vegetation and soil. The composite CN calcula- ΔQ
Ir ¼  1000 ð8Þ
tion is PA

CNC ¼ SI  CNI þ SV  CN V þ SS  CNS ð5Þ where Ir is the direct runoff increment coefficient; P is the rainfall
depth (mm).
where CNC is the composite CN value; SI, SV, and SS are fractions of im-
pervious surface, vegetation and soil extracted by the LSMA, respec- 2.4. Spatial autocorrelation analyses
tively; CNI, CNV, and CNS are the initial CN values of impervious
surfaces, vegetation and soil, respectively. Moran's I is a commonly used indicator of spatial autocorrelation.
We used the MS model to evaluate the direct runoff, so the compos- Global Moran's I was used as the measure of spatial autocorrelation
ite CN was calculated under the dry antecedent moisture condition (Moran, 1950). While the global Moran's I provides an overall measure
(AMC-I). CNI under AMC-I was assigned a value of 98 according to the of spatial autocorrelation, the local Moran's I relates each observation to
lookup table of Technical Release 55 (TR-55). Vegetation and soil were its neighbors and assigns them to classes with a value indicating the de-
classified into different types according to the vegetation conditions gree of spatial autocorrelation. The local indicators of spatial association
and soil texture characteristics. The initial values of CNV and CNS in (LISA) measure the degree of spatial autocorrelation at each specific ob-
AMC-I from TR-55 are shown in Tables 2 and 3. servation using local Moran's I (Anselin, 1995). LISA represents a set of
In this study, actual daily rainfall data from May to October 2015 was localized statistical approaches that typically measure the relationship
put into the MS model to evaluate runoff. The direct runoff in the cur- between individual observations and their surrounding neighbors to
rent condition was evaluated by the current composite CN from eq. 5. uncover patterns of spatial clustering (Bone et al., 2013; Fu et al., 2014).
To express the increase of direct runoff caused by urbanization, we as- A high positive local Moran's I value implies the target value is sim-
sumed the increment of direct runoff could be characterized as that ilar to its neighborhood, and the zones are spatial clusters, which in-
achieved by replacing impervious surfaces with vegetated area in clude high–high clusters (HH, high values in a high value
good condition. Then the composite CN of pre-urbanization could be neighborhood) and low–low clusters (LL, low values in a low value
calculated by replacing the CN of impervious surface with the CN of veg- neighborhood). Meanwhile, a high negative local Moran's I value im-
etated areas in good condition. The direct runoff in pre-urbanization plies a potential spatial outlier, which is obviously different from the
conditions can be evaluated by the MS model. values of its surrounding zones. Spatial outliers include high–low out-
liers (HL, a high value in a low value neighborhood) and low–high out-
2.3. Direct runoff increment calculation liers (LH, a low value in a high value neighborhood).
For this study, we calculated the global Moran's I and local Moran's I
Three variables were defined to evaluate the potential hydrological for quantifying spatial autocorrelation of direct runoff by using ArcGIS
changes of runoff: the total amount of runoff increment (ΔQ), the direct 10.3 (ESRI Inc., Redlands, CA). The UFZs are spatial patterns of specific
runoff increment depth (ΔI) and the direct runoff increment coefficient urban functions, so we used UFZs as the appropriate neighborhood
(Ir). The volume of direct runoff increment caused by urbanization can size for the local Moran's I. The direct runoff depth of each UFZ was
be calculated using the runoff difference between the current condition used to calculate the Moran's I index and analyze the spatial
and the pre-urbanization condition. In this study, we assumed the pre- autocorrection of the entire area.
urbanization condition by replacing impervious areas with good condi-
tion vegetation. 2.5. Boosted regression trees

ΔQ ¼ Q c −Q p ð6Þ A boosted regression trees (BRT) method was chosen to analyze the
relative influence and marginal effects of factors affecting direct runoff.
where ΔQ represents the direct runoff increment volume due to urban- BRT is a self-learning method based on the classification regression tree
ization at a study unit (m3), Qc is the direct runoff volume in current algorithm, which can generate multiple regression trees by random se-
condition; Qp is the direct runoff volume in pre-urbanization conditions. lection and machine learning methods and improve model stability and
prediction accuracy (Cai et al., 2013; De'ath, 2007; Froeschke and
ΔQ Froeschke, 2011; Leathwick et al., 2006). BRT is a flexible regression
ΔI ¼  1000 ð7Þ
A modelling technique that gives important benefits for modelling eco-
logical data, which could accommodate both different types of variables
where ΔI is the direct runoff increment depth due to urbaniza-
and missing values, and is immune to the effects of extreme outliers and
tion at a study unit (mm); A is the area of a study unit at various
the inclusion of irrelevant variables (Elith et al., 2008). The relative in-
fluence of factors on runoff can be shown in a bar chart with longer
bars indicating a stronger influence on the response. The relationship
between dependent and independent variables can be presented in a
partial dependency plot, which can visually show the marginal effect
Table 2
Vegetation classification and values of the vegetation curve number CNV in dry antecedent
moisture conditions (AMC-I). NDVI represents the normalized difference vegetation
Table 3
index.
Soil texture classification and values of the soil curve number CNS in dry antecedent mois-
Vegetation type NDVI range CNV ture conditions (AMC-I).

A B C D Soil type Soil texture CNS

Vegetated Good condition NDVI ≥ 0.7 21 41 55 63 A Sand ≥ 50% and clay ≤ 10% 59
Fair condition 0.4 ≤ NDVI b 0.7 30 50 62 68 B Sand ≥ 50% and clay N 10% 72
Poor condition 0 ≤ NDVI b 0.4 48 62 72 76 C Sand b 50% and clay ≤ 40% 80
Non-vegetated NDVI b 0 59 72 80 85 D Sand b 50% and clay N 40% 85
 C. Li et al. / Science of the Total Environment 643 (2018) 301–311 305

Table 4
Direct runoff characteristics of 2015 of the four-ring area and different zones in Shenyang,
China.

Regions Total direct runoff Average direct runoff Direct runoff

(106 m3) depth (mm) coefficient

Zone 1 8.83 166.52 0.38

Zone 2 12.67 140.19 0.32
Zone 3 23.81 103.86 0.24
Zone 4 37.92 83.44 0.19
Four-ring area 83.23 118.94 0.27

of an influencing factor on direct runoff when other factors are held

constant.
In this study, the direct runoff depth of each pixel was selected as the
response variable. Nine variables were selected as independent vari-
ables, including meteorological factors (rainfall and antecedent 5-day Fig. 3. Direct runoff depth variations of 2015 of the urban functional zones (UFZs) in the
rainfall), natural environmental factors (impervious ratio, normalized four zones in Shenyang, China. HRZ: high-density residential zone; LRZ: low-density res-
difference vegetation index (NDVI) and digital elevation model idential zone; INZ: industrial zone; BUZ: business zone; GOZ: government zone; PSZ: pub-
lic service zone; AGZ: agricultural zone; DEZ: development zone; GSZ: green space zone;
(DEM)) and urbanization factors (UFZs, distance to urban center, build-
PRZ: preservation zone; WTR: water. Error bars were the standard deviations of direct
ing height and road density). Rainfall and antecedent 5-day rainfall runoff depth of same UFZs.
were calculated from daily rainfall data from May to October 2015.
The impervious ratio was evaluated by LSMA. NDVI was calculated
from Landsat 8 images from 2015. The distance to urban center and the gradual increase of total direct runoff, the average direct runoff
road density were calculated in ArcGIS. Building heights were extracted depth and direct runoff coefficient from zone 1 to 4 gradually decreased
from high-resolution satellite images via Barista's monoplotting func- from 166.52 to 83.44 mm, and 0.38 to 0.19, respectively. The increase of
tion (Zhang et al., 2014). All the factors were overlaid in ArcGIS, then total runoff from zone 1 to 4 mainly due to the large increase of zone
random points were created to get a 10,000-sample dataset for BRT area. The reason for the decrease of average direct runoff depth and di-
analyses performed in R using the gbm package. All the BRT models rect runoff coefficient from zone 1 to 4 is that the intensity of urbaniza-
were fitted with the following parameters: Gaussian error distribution, tion gradually decreased from the urban core area to the suburban area,
a learning rate of 0.005, a bag fraction of 0.5, and fivefold cross- and the imperious surface rate gradually decreased. Zones 1 to 4
validation. reflected the direction of urban sprawl, representing the degree of ur-
banization from high to low. The process of urban sprawl has led to
3. Results and discussion the transformation of large amounts of pervious land into impervious
surfaces, resulting in a larger runoff coefficient and depth in highly ur-
3.1. Spatial patterns of direct runoff banization area.
The direct runoff coefficients ranged from 0 to 0.59 at the UFZ scale
Direct runoff in 2015 in Shenyang was evaluated by remote sensing (Fig. 2). The variation in the direct runoff coefficients of different types
and the MS model in ArcGIS. The direct runoff characteristics of the four- of UFZs was determined by their urban functions. The BUZ had the
ring area and different zones are shown in Table 4. Total direct runoff of highest coefficient of 0.42, mainly because the BUZ represented the
the entire study area was 83.23 million m3, and zone 4 contributed the business zones in the city, including markets and office buildings,
most (37.92 million m3), followed by zones 3, 2, and 1. which do not include many green areas. The HRZ and LRZ were both res-
The average direct runoff depth of the four-ring area was idential communities, but the direct runoff coefficients were very differ-
118.94 mm, and the direct runoff coefficient was 0.27. In contrast to ent, at 0.32 and 0.18, respectively. This difference was mainly due to the
different living functions of the two residential zones. The HRZs were
the typical residential zones in the Shenyang urban area, which usually
had high building and population densities, and most areas were cov-
ered by impervious surfaces, such as buildings, roads, and parking lots.
The LRZs were usually distributed in the suburban or new developed
areas, which had a lower impervious fraction and sparser population,
such as villas and affluent residential communities. The zones
representing the green area in Shenyang urban area, GSZ, AGZ, and
PRZ, showed the three lowest direct runoff coefficients among the 10
UFZs, measuring 0.12, 0.05 and 0.02, respectively.
There were large variations of direct runoff depth among the UFZs in
the four zones. The runoff depth of most UFZs indicated a decreasing

Table 5
Direct runoff increments of 2015 of the four zones in Shenyang, China. ΔQ represents the
total amount of runoff increment; ΔI represents the direct runoff increment depth; and Ir
represents the direct runoff increment coefficient.

Region ΔQ (106 m3) ΔI (mm) Ir

Zone 1 8.75 155.86 0.36

Fig. 2. Direct runoff coefficients of 2015 of the urban functional zones (UFZs) in Shenyang,
Zone 2 12.48 121.16 0.28
China. HRZ: high-density residential zone; LRZ: low-density residential zone; INZ: indus-
Zone 3 23.18 83.33 0.19
trial zone; BUZ: business zone; GOZ: government zone; PSZ: public service zone; AGZ: ag-
Zone 4 36.37 48.48 0.11
ricultural zone; DEZ: development zone; GSZ: green space zone; PRZ: preservation zone;
Four-ring area 80.79 68.02 0.16
WTR: water.
306 C. Li et al. / Science of the Total Environment 643 (2018) 301–311

Fig. 4. Direct runoff increments of 2015 of the urban functional zones (UFZs) in Shenyang,
China. HRZ: high-density residential zone; LRZ: low-density residential zone; INZ: indus-
trial zone; BUZ: business zone; GOZ: government zone; PSZ: public service zone; AGZ: ag-
ricultural zone; DEZ: development zone; GSZ: green space zone; PRZ: preservation zone;
Fig. 6. Direct runoff increment coefficient Ir of 2015 of the urban functional zones (UFZs) in
WTR: water. ΔQ, ΔI and Ir as defined in Table 5. Error bars were the standard deviations of
the four zones in Shenyang, China. HRZ: high-density residential zone; LRZ: low-density
ΔI and Ir of same UFZs.
residential zone; INZ: industrial zone; BUZ: business zone; GOZ: government zone; PSZ:
public service zone; AGZ: agricultural zone; DEZ: development zone; GSZ: green space
trend from zone 1 to 4, while the decreased ratios were 28%–48%. The zone; PRZ: preservation zone; WTR: water. Error bars were the standard deviations of Ir
runoff of the PSZ, DEZ, GSZ in zone 4 revealed decreases of N40% com- of same UFZs.

pared with zone 1 (Fig. 3). Different UFZs had different rates of decline.
Runoff in the PSZ had a faster decline rate than other UFZs, and it was the urbanization level increased from zone 4 to 1, ΔI and Ir increased
the second largest in zone 1 and the fourth largest in zone 4. Some correspondingly. Therefore, the urban central area (zones 1 and
UFZs were not present in the urban central area, e.g. LRZ was not pres- 2) have the most serious effect of direct runoff by urbanization.
ent in zone 1, and AGZ and PRZ were not found in zones 1 and 2. The di- The results of runoff increments were categorized by UFZs. The total
rect runoff changes in the same UFZs in different zones were mainly increased runoff volumes ΔQ of different UFZs showed very large differ-
caused by differences of building characteristics. Urban expansion is ences. The HRZ and INZ had the largest ΔQ, measuring 22.34 and 26.19
gradually extending from zone 1 to 4, so the construction ages are million m3, and the sum of these two UFZs was N60% of the total incre-
very different from the central area to the suburban area. The land prices ment of the study area (Fig. 4). Although the ΔQ of BUZ was very small
were very different, so the zones in the urban center tended to have (6.76 million m3), the ΔI and Ir of BUZ were the largest among UFZs. The
dense buildings and less green area, while the suburban area showed green zones (AGZ, GSZ, and PRZ) of Shenyang had smaller ΔI and Ir,
the opposite pattern. mainly because of the large proportion of vegetation.
The rank ordering of ΔQ among the same UFZs in different zones was
3.2. Direct runoff increment by urbanization variable (Fig. 5). The largest ΔQ in zones 1 and 2 was the HRZ, but it was
the INZ in zone 3 and 4. The ΔQ of the HRZ was very large from zone 1 to
Urbanization leads to increased impervious surfaces, resulting in 4, and the ΔQ of the LRZ and INZ increased markedly from zones 1 to 4.
runoff increments. Under the current conditions, the total amount of The Ir of most UFZs showed a gradual decrease from zones 1 to 4 (Fig. 6).
runoff increment ΔQ by urbanization in the entire study area was The BUZ had the largest Ir, and the AGZ and PRZ had the smallest Ir
80.79 million m3 (Table 5). The direct runoff increment depth ΔI was among the UFZs.
68.02 mm and the direct runoff increment coefficient Ir was 0.16. As Some research has indicated rainfall and antecedent moisture condi-
tions have a great impact on runoff (Dong et al., 2015; Yang et al., 2013).

Fig. 5. Direct runoff increment ΔQ of 2015 of the urban functional zones (UFZs) in the four
zones in Shenyang, China. HRZ: high-density residential zone; LRZ: low-density residen-
tial zone; INZ: industrial zone; BUZ: business zone; GOZ: government zone; PSZ: public
service zone; AGZ: agricultural zone; DEZ: development zone; GSZ: green space zone; Fig. 7. Direct runoff increment ΔQ of 2015 of different rainfall P and antecedent 5-day
PRZ: preservation zone; WTR: water. rainfall P5 in Shenyang, China. The size of the bubble indicates the relative amount of P5.
 C. Li et al. / Science of the Total Environment 643 (2018) 301–311 307

Fig. 10. Total area of 2015 flooding hazard in different urban functional zones (UFZs) in
Shenyang, China. HRZ: high-density residential zone; LRZ: low-density residential zone;
Fig. 8. Direct runoff increment coefficient Ir of 2015 of different rainfall P and antecedent 5- INZ: industrial zone; BUZ: business zone; GOZ: government zone; PSZ: public service
day rainfall P5 in Shenyang, China. The size of the bubble indicates the relative amount of zone; AGZ: agricultural zone; DEZ: development zone; GSZ: green space zone; PRZ: pres-
P5. ervation zone; WTR: water.

In this study, the effect of rainfall P and antecedent 5-day rainfall P5 on

the direct runoff increment ΔQ and increment coefficient Ir are shown 3.3. Flooding hazard characteristics
in Figs. 5 and 6, and the size of the bubble indicates the relative size of
P5. It was revealed that precipitation could strongly affect ΔQ and Ir. An- Daily rainfall runoff coefficients were used to evaluate the flooding
tecedent 5-day rainfall was not found to have significant effect on ΔQ hazard in the study area. The area of runoff coefficients N0.6 of each
(Fig. 7). However, under the same rainfall conditions, larger P5 could daily rainfall amount were overlaid together, and flooding hazard was
lead to higher Ir (Fig. 8). categorized according to the frequency of the runoff coefficient exceed-
ing 0.6. The cutoff values of frequency for representing different degrees
of flooding hazard were decided by five stormwater managers and pro-
fessors. Low hazard was a frequency from 1 to 5; moderate hazard was a
frequency from 6 to 10; and significant hazard was a frequency above
10. The hazard area was the potential flooding area, which should be
highly concerned by urban storm water management departments, in-
cluding monitoring the runoff depth, flooding area during rainfall, and
assessing potential flooding loss after storm. The significant hazard
area covered 77.5 km2 (6.53%) of the study area (Fig. 9). The total hazard
area in four-ring zones was 281.98 km2 (23.75%). The total areas of low,
moderate, and significant hazard increased from zone 1 to 4, while the

Fig. 11. Area ratio of 2015 flooding hazard in different urban functional zones (UFZs) in
Shenyang, China. HRZ: high-density residential zone; LRZ: low-density residential zone;
INZ: industrial zone; BUZ: business zone; GOZ: government zone; PSZ: public service
zone; AGZ: agricultural zone; DEZ: development zone; GSZ: green space zone; PRZ: pres-
Fig. 9. Flooding hazard characteristics of 2015 in different zones in Shenyang, China. Zones ervation zone; WTR: water. Error bars were the standard deviations of flooding hazard
as defined in Fig. 1. area ratios of same UFZs.
308 C. Li et al. / Science of the Total Environment 643 (2018) 301–311

area ratios decreased from zone 1 to 4. The significant hazard area of 3.4. Spatial-cluster and spatial-outlier analyses
zone 1 was 7.23 km2, which is about 12.87% of the total zone 1 area.
The total area ratio of runoff coefficient exceeding 0.6 in the urban cen- The results of the global Moran's I for the four-ring area illustrated
tral area (zones 1 and 2) were 64.76% and 46.07%, while in the suburban that significant positive spatial autocorrelation (Moran's I = 0.46, p b
area (zones 3 and 4) they were only 28.99% and 15.66%. 0.001) was observed for direct runoff. The runoff in the entire area
In terms of the different UFZs, HRZ had largest area of low haz- was clustered. For local Moran's I, most HH clusters were observed in
ard land (58.53 km2), and INZ had largest area of significant hazard zone 1 and southwest of zone 2, which represented the high-hazard
land (40.97 km2) and a relatively large low hazard area (43.07 km2) area of direct runoff, and most LH outliers were close to the HH area
(Fig. 10). In addition to INZ, the significant hazard area of HRZ and (Fig. 12). The LL clusters were mainly located in zone 4 and east of
BUZ was also very large, measuring 9.96 and 10.57 km 2 , zone 3, which represented the low hazard area of direct runoff, and
respectively. most HL outliers were close to the LL area. The LISA analysis illustrated
The area ratios of flooding hazard in different UFZs were very vari- that the HH clusters in Shenyang were mainly concentrated in zone 1,
able. The BUZ, INZ and PSZ had large significant hazard proportions, es- the central area of the city, far away from the LL clusters. The HH in
timated at 21.19%, 15.45%, and 15.34% of their zone area (Fig. 11). The the urban central area were a source of runoff, and LH outliers near
HRZ had the biggest area ratio of low hazard land (39.05%), and the pro- the HH area could be treated as runoff sinks. Drainage corridors, such
portion of significant hazard area was only 5.77%. The BUZ had the larg- as the drainage network and grass channels, can be established between
est total hazard ratio (77.48%), followed by the HRZ (55.69%), PSZ HH and LH area to transfer rainwater from HH to LH, thereby using the
(55.31%), and GOZ (51.54%). urban green space to store rainwater to mitigate the flooding hazard.

Fig. 12. Map of spatial clusters and spatial outliers of 2015 direct runoff in Shenyang, China. This map was created in ArcGIS by calculating the local Moran's I of each UFZs.
 C. Li et al. / Science of the Total Environment 643 (2018) 301–311 309

Different UFZs showed great variations of spatial autocorrelation

characteristics (Fig. 13). The AGZ had the largest LL area (280.04 km2),
followed by the GSZ (61.36 km2) and PRZ (36.06 km2). The HRZ and
BUZ had the largest HH areas, which measured 35.19 and 19.49 km2,
and the HH area of these two UFZs accounted for 77.7% of the total HH
area (Fig. 13a). The BUZ had the largest HH area ratio of the separate
UFZs. About 50% of the BUZ were HH clusters, which could produce
large amounts of direct runoff causing flooding (Fig. 13b). In the resi-
dential zones, 19% of the HRZ were HH clusters and only 8% were LL
clusters, but there were no HH clusters in the LRZ, and about 37%
were LL clusters. Most HL outliers were concentrated in the INZ, mainly
because factories and workshops were built in the suburban area near
farmland which has low direct runoff. Most LH outliers were distributed
in the GSZ and AGZ, mainly because the urban parks were located
among the high runoff areas in the urban center, and farmland was lo-
cated among new development areas in the urban fringe.

3.5. Analysis of influencing factors

Modelling using BRT was used to examine the relationships between

Fig. 13. Spatial autocorrelation characteristics of 2015 of urban functional zones (UFZs) in
direct runoff depth and environmental factors. The BRT analysis indi-
Shenyang, China. HRZ: high-density residential zone; LRZ: low-density residential zone;
INZ: industrial zone; BUZ: business zone; GOZ: government zone; PSZ: public service cated that direct runoff depth had the strongest correlation with rainfall
zone; AGZ: agricultural zone; DEZ: development zone; GSZ: green space zone; PRZ: pres- (52.07%), followed by the impervious surfaces ratio, NDVI, antecedent
ervation zone; WTR: water. 5-day rainfall, UFZs, distance to urban center, building height, DEM

Fig. 14. Relative influences and partial dependency plots of five most important variables in a boosted regression tree (BRT) predicting 2015 direct runoff in Shenyang, China. HRZ: high-
density residential zone; LRZ: low-density residential zone; INZ: industrial zone; BUZ: business zone; GOZ: government zone; PSZ: public service zone; AGZ: agricultural zone; DEZ: de-
velopment zone; GSZ: green space zone; PRZ: preservation zone; WTR: water.
310 C. Li et al. / Science of the Total Environment 643 (2018) 301–311

and road density (Fig. 14(a)). Partial dependencies of the five most in- Ansari, T.A., Katpatal, Y.B., Vasudeo, A.D., 2016. Spatial evaluation of impacts of increase in
impervious surface area on SCS-CN and runoff in Nagpur urban watersheds, India.
fluential factors from the fitted model are presented in Fig. 14(b)–(f), Arab. J. Geosci. 9.
which demonstrate that high direct runoff was most likely to be pro- Anselin, L., 1995. Local indicators of spatial association—LISA. Geogr. Anal. 27, 93–115.
duced in areas with high rainfall and high antecedent 5-day rainfall Bartlett, M.S., Parolari, A.J., Mcdonnell, J.J., Porporato, A., 2016. Beyond the SCS-CN
method: a theoretical framework for spatially lumped rainfall-runoff response.
events, and a high impervious ratio and low NDVI. Direct runoff Water Resour. Res. 52.
responded positively to antecedent 5-day rainfall, but the marginal ef- Bone, C., Wulder, M.A., White, J.C., Robertson, C., Nelson, T.A., 2013. A GIS-based risk rating
fects of antecedent 5-day rainfall increased rapidly from 0 to 23 mm, of forest insect outbreaks using aerial overview surveys and the local Moran's I statis-
tic. Appl. Geogr. 40, 161–170.
and the marginal effects basically did not increase above 23 mm. The Braud, I., Breil, P., Thollet, F., Lagouy, M., Branger, F., Jacqueminet, C., et al., 2013a. Evidence
relative influence of UFZs was small (1.70%), but the marginal effects of the impact of urbanization on the hydrological regime of a medium-sized
of different UFZs showed very large variation. The largest relative influ- periurban catchment in France. J. Hydrol. 485, 5–23.
Braud, I., Fletcher, T.D., Andrieu, H., 2013b. Hydrology of peri-urban catchments: pro-
ence of UFZs was the INZ, followed by the BUZ and HRZ. This may be due
cesses and modelling. J. Hydrol. 485, 1–4.
to the fact that green proportion of INZ is lower than other UFZs, so the Cai, W.H., Yang, J., Liu, Z.H., Hu, Y.M., Weisberg, P.J., 2013. Post-fire tree recruitment of a
INZ tend to build parking lots and impervious areas for industrial use. boreal larch forest in Northeast China. For. Ecol. Manag. 307, 20–29.
The results of analyzing the influencing factors indicated that many Choi, J.Y., Engel, B.A., Muthukrishnan, S., Harbor, J., 2003. GIS based long term hydrologic
impact evaluation for watershed urbanization. J. Am. Water Resour. Assoc. 39,
environmental factors had a great impact on direct runoff. The direct 623–635.
runoff was clearly dependent on rainfall events, that is, the more rainfall Choi, W., Nauth, K., Choi, J., Becker, S., 2016. Urbanization and rainfall–runoff relationships
and antecedent 5-day rainfall, the more direct runoff was produced. Al- in the Milwaukee River basin. Prof. Geogr. 68, 14–25.
De'ath, G., 2007. Boosted trees for ecological modeling and prediction. Ecology 88,
though meteorological factors (rainfall and antecedent 5-day rainfall) 243–251.
cannot be changed, some factors could be adjusted to improve the Dong, L.H., Xiong, L.H., Lall, U., Wang, J.W., 2015. The effects of land use change and pre-
urban hydrological cycle, such as increasing green infrastructure in cipitation change on direct runoff in Wei River watershed, China. Water Sci. Technol.
71, 289–295.
urban areas where impervious surfaces are concentrated to control Elith, J., Leathwick, J.R., Hastie, T., 2008. A working guide to boosted regression trees.
storm runoff and reduce flooding. The INZ, BUZ and HRZ had the highest J. Anim. Ecol. 77, 802–813.
positive correlations with runoff among the different UFZs, which pro- Fan, F., Deng, Y., Hu, X., Weng, Q., 2013. Estimating composite curve number using an im-
proved SCS-CN method with remotely sensed variables in Guangzhou, China. Remote
vides a way for urban planners and rainwater managers to solve the Sens. 5, 1425–1438.
urban flooding hazard at a UFZ scale, such as adjusting the pattern of dif- Fletcher, T.D., Andrieu, H., Hamel, P., 2013. Understanding, management and modelling of
ferent functional zones, or transforming the high direct runoff potential urban hydrology and its consequences for receiving waters: a state of the art. Adv.
Water Resour. 51, 261–279.
of UFZs through green infrastructures.
Froeschke, J.T., Froeschke, B.F., 2011. Spatio-temporal predictive model based on environ-
mental factors for juvenile spotted seatrout in Texas estuaries using boosted regres-
4. Conclusions sion trees. Fish. Res. 111, 131–138.
Fu, W.J., Jiang, P.K., Zhou, G.M., Zhao, K.L., 2014. Using Moran's I and GIS to study the spa-
tial pattern of forest litter carbon density in a subtropical region of southeastern
This study investigated hydrologic effects of urbanization on direct China. Biogeosciences 11, 2401–2409.
runoff characteristics in large urban area. UFZs were delineated as Gallo, E.L., Brooks, P.D., Lohse, K.A., Mclain, J.E.T., 2013. Land cover controls on summer
discharge and runoff solution chemistry of semi-arid urban catchments. J. Hydrol.
basic study unites. Direct runoff was estimated by modified SCS-CN 485, 37–53.
model combined with the classification of urban landscapes. The char- Guan, M., Sillanpää, N., Koivusalo, H., 2016. Storm runoff response to rainfall pattern,
acteristics of direct runoff and flooding hazard were analyzed in region magnitude and urbanization in a developing urban catchment. Hydrol. Process. 30,
543–557.
and UFZ scales. The spatial autocorrelation of direct runoff was explored Hibbs, B.J., Sharp, J.M., 2012. Hydrogeological impacts of urbanization. Environ. Eng.
by LISA. In addition, we investigated the relationships between direct Geosci. 18, 3–24.
runoff depth and environmental factors using BRT. The following con- Hu, H.B., 2016. Rainstorm flash flood risk assessment using genetic programming: a case
study of risk zoning in Beijing. Nat. Hazards 83, 485–500.
clusions can be summarized as that Huang, H.J., Cheng, S.J., Wen, J.C., Lee, J.H., 2008. Effect of growing watershed impervious-
ness on hydrograph parameters and peak discharge. Hydrol. Process. 22, 2075–2085.
1. For the entire study area, the direct runoff depth under the current Isik, S., Kalin, L., Schoonover, J.E., Srivastava, P., Lockaby, B.G., 2013. Modeling effects of
changing land use/cover on daily streamflow: an artificial neural network and
conditions was 68.02 mm larger than pre-urbanization conditions.
curve number based hybrid approach. J. Hydrol. 485, 103–112.
The HRZ and INZ had the largest runoff increment. Jacqueminet, C., Kermadi, S., Michel, K., Beal, D., Gagnage, M., Branger, F., et al., 2013. Land
cover mapping using aerial and VHR satellite images for distributed hydrological
2. The area of significant hazard area was 77.5 km2 (6.53%). The HRZ modelling of periurban catchments: application to the Yzeron catchment (Lyon,
had largest area of low hazard land (58.53 km2), and the BUZ had France). J. Hydrol. 485, 68–83.
largest significant hazard proportion (21.19%). Jiao, P.J., Xu, D., Wang, S.L., Yu, Y.D., Han, S.J., 2015. Improved SCS-CN method based on
storage and depletion of antecedent daily precipitation. Water Resour. Manag. 29,
3. The spatial autocorrelation results illustrated that the HH clusters in 4753–4765.
Shenyang were mainly concentrated in zone 1, the central area of the Kvočka, D., Falconer, R.A., Bray, M., 2016. Flood hazard assessment for extreme flood
city, far away from the LL clusters. The HRZ and BUZ had the largest events. Nat. Hazards 84, 1569–1599.
Leathwick, J.R., Elith, J., Francis, M.P., Hastie, T., Taylor, P., 2006. Variation in demersal fish
HH area, and the BUZ had the largest HH area ratio. species richness in the oceans surrounding New Zealand: an analysis using boosted
4. The five most influential factors on direct runoff derived from the regression trees. MEPS 321, 267–281.
BRT analysis were rainfall, impervious ratio, NDVI, antecedent 5- Li, C., Liu, M., Hu, Y., Gong, J., Xu, Y., 2016. Modeling the quality and quantity of runoff in a
highly urbanized catchment using storm water management model. Pol. J. Environ.
day rainfall and UFZs. For different UFZs, the largest relative influence Stud. 25, 1573–1581.
of UFZs was the INZ, followed by the BUZ and HRZ. Mejia, A.I., Moglen, G.E., 2010a. Impact of the spatial distribution of imperviousness on the
hydrologic response of an urbanizing basin. Hydrol. Process. 24, 3359–3373.
Acknowledgements Mejia, A.I., Moglen, G.E., 2010b. Spatial distribution of imperviousness and the space-time
variability of rainfall, runoff generation, and routing. Water Resour. Res. 46.
Mishra, S.K., Singh, V.P., Singh, V.P., Frevert, D., 2002. SCS-CN-based Hydrologic Simulation
Funding for this project was provided by the China National R&D Package.
Program (No. 2017YFC0505705) and National Natural Science Founda- Moran, P.A.P., 1950. Notes on continuous stochastic phenomena. Biometrika 37, 17–23.
tion of China (Nos. 41501198 and 41671184). Olang, L.O., Furst, J., 2011. Effects of land cover change on flood peak discharges and runoff
volumes: model estimates for the Nyando River basin, Kenya. Hydrol. Process. 25,
80–89.
References Putro, B., Kjeldsen, T.R., Hutchins, M.G., Miller, J., 2016. An empirical investigation of cli-
mate and land-use effects on water quantity and quality in two urbanising catch-
Angrill, S., Petit-Boix, A., Morales-Pinzón, T., Josa, A., Rieradevall, J., Gabarrell, X., 2017. ments in the southern United Kingdom. Sci. Total Environ. 548, 164–172.
Urban rainwater runoff quantity and quality – a potential endogenous resource in cit- Qin, H.-p., He, K.-m., Fu, G., 2016. Modeling middle and final flush effects of urban runoff
ies? J. Environ. Manag. 189, 14–21. pollution in an urbanizing catchment. J. Hydrol. 534, 638–647.
 C. Li et al. / Science of the Total Environment 643 (2018) 301–311 311

Quan, R.S., Liu, M., Lu, M., Zhang, L.J., Wang, J.J., Xu, S.Y., 2010. Waterlogging risk assess- Tsihrintzis, V.A., Hamid, R., 1997. Urban stormwater quantity/quality modeling using the
ment based on land use/cover change: a case study in Pudong New Area, Shanghai. SCS method and empirical equations. J. Am. Water Resour. Assoc. 33, 163–176.
Environ. Earth Sci. 61, 1113–1121. Weng, Q.H., 2001. Modeling urban growth effects on surface runoff with the integration
Richert, E., Bianchin, S., Heilmeier, H., Merta, M., Seidler, C., 2011. A method for linking re- of remote sensing and GIS. Environ. Manag. 28, 737–748.
sults from an evaluation of land use scenarios from the viewpoint of flood prevention Yang, X., You, X.Y., Ji, M., Nima, C., 2013. Influence factors and prediction of stormwater
and nature conservation. Landsc. Urban Plan. 103, 118–128. runoff of urban green space in Tianjin, China: laboratory experiment and quantitative
Ridd, M.K., 1995. Exploring a V-I-S (vegetation-impervious-soil) model for urban ecosys- theory model. Water Sci. Technol. 67, 869–876.
tem analysis through remote sensing: comparative anatomy for citiest. Int. J. Remote Yao, L., Chen, L.D., Wei, W., Sun, R.H., 2015. Potential reduction in urban runoff by green
Sens. 16, 2165–2185. spaces in Beijing: a scenario analysis. Urban For. Urban Green. 14, 300–308.
Sahu, R.K., Mishra, S.K., Eldho, T.I., 2012. Improved storm duration and antecedent mois- Yao, L., Chen, L.D., Wei, W., 2017. Exploring the linkage between urban flood risk and spa-
ture condition coupled SCS-CN concept-based model. J. Hydrol. Eng. 17, 1173–1179. tial patterns in small urbanized catchments of Beijing, China. Int. J. Environ. Res. Pub-
Sajjad, R.U., Kim, K.J., Memon, S., Sukhbaatar, C., Paule, M.C., Lee, B.Y., et al., 2015. Charac- lic Health 14.
terization of stormwater runoff from a light rail transit area. Water Environ. Res. 87, Zampella, R.A., Procopio, N.A., Lathrop, R.G., Dow, C.L., 2007. Relationship of land-use/
813–822. land-cover patterns and surface-water quality in the Mullica River basin. J. Am.
Sanders, R.A., 1986. Urban vegetation impacts on the hydrology of Dayton, Ohio. Urban Water Resour. Assoc. 43, 594–604.
Ecol. 9, 361–376. Zhang, B.A., Xie, G.D., Zhang, C.Q., Zhang, J., 2012. The economic benefits of rainwater-
Seto, K.C., Guneralp, B., Hutyra, L.R., 2012. Global forecasts of urban expansion to 2030 and runoff reduction by urban green spaces: a case study in Beijing, China. J. Environ.
direct impacts on biodiversity and carbon pools. Proc. Natl. Acad. Sci. U. S. A. 109, Manag. 100, 65–71.
16083–16088. Zhang, X., Zhang, X., Hu, S., Liu, T., Li, G., 2013. Runoff and sediment modeling in a peri-
Shepherd, J.M., 2006. Evidence of urban-induced precipitation variability in arid climate urban artificial landscape: case study of Olympic Forest Park in Beijing. J. Hydrol.
regimes. J. Arid Environ. 67, 607–628. 485, 126–138.
Singh, P.K., Yaduvanshi, B.K., Patel, S., Ray, S., 2013. SCS-CN based quantification of poten- Zhang, P.F., Hu, Y.M., Xiong, Z.P., 2014. Extraction of three-dimensional architectural data
tial of rooftop catchments and computation of ASRC for rainwater harvesting. Water from QuickBird images. J. Indian Soc. Remote Sens. 42, 409–416.
Resour. Manag. 27, 2001–2012. Zhang, B., G-D, Xie, Li, N., Wang, S., 2015. Effect of urban green space changes on the role
Sun, R., Lu, Y., Chen, L., Yang, L., Chen, A., 2013. Assessing the stability of annual temper- of rainwater runoff reduction in Beijing, China. Landsc. Urban Plan. 140, 8–16.
atures for different urban functional zones. Build. Environ. 65, 90–98. Zhou, F., Xu, Y.P., Chen, Y., Xu, C.Y., Gao, Y.Q., Du, J.K., 2013. Hydrological response to ur-
Tian, G.J., Wu, J.G., Yang, Z.F., 2010. Spatial pattern of urban functions in the Beijing met- banization at different spatio-temporal scales simulated by coupling of CLUE-S and
ropolitan region. Habitat Int. 34, 249–255. the SWAT model in the Yangtze River Delta region. J. Hydrol. 485, 113–125.