Content uploaded by Yongyong Zhang
Author content
All content in this area was uploaded by Yongyong Zhang on Sep 09, 2018
Content may be subject to copyright.
Research papers
Simulation and assessment of urbanization impacts on runoff metrics:
insights from landuse changes
Yongyong Zhang
a,
⇑
, Jun Xia
b
, Jingjie Yu
a
, Mark Randall
c
, Yichi Zhang
a
, Tongtiegang Zhao
d
, Xingyao Pan
e
,
Xiaoyan Zhai
f
, Quanxi Shao
g
a
Key Laboratory of Water Cycle and Related Land Surface Processes, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing
100101, China
b
State Key Laboratory of Water Resources & Hydropower Engineering Sciences, Wuhan University, Wuhan 430072, China
c
Department of Geosciences and Natural Resources Management, University of Copenhagen, Frederiksberg K, Denmark
d
Department of Infrastructure Engineering, The University of Melbourne, Victoria, Australia
e
Beijing Water Science and Technology Institute, Beijing 100048, China
f
China Institute of Water Resources and Hydropower Research, Beijing 100038, China
g
CSIRO Data 61, Private Bag 5, Wembley, Western Australia 6913, Australia
article info
Article history:
Received 21 January 2018
Received in revised form 11 March 2018
Accepted 12 March 2018
Available online 15 March 2018
This manuscript was handled by G. Syme,
Editor-in-Chief
Keywords:
Urbanization
Runoff metrics
Catchment hydrological model
Multivariate statistical approach
Beijing City
abstract
Urbanization-induced landuse changes alter runoff regimes in complex ways. In this study, a detailed
investigation of the urbanization impacts on runoff regimes is provided by using multiple runoff metrics
and with consideration of landuse dynamics. A catchment hydrological model is modified by coupling a
simplified flow routing module of the urban drainage system and landuse dynamics to improve long-
term urban runoff simulations. Moreover, multivariate statistical approach is adopted to mine the spatial
variations of runoff metrics so as to further identify critical impact factors of landuse changes. The Qing
River catchment as a peri-urban catchment in the Beijing metropolitan area is selected as our study
region. Results show that: (1) the dryland agriculture is decreased from 13.9% to 1.5% of the total catch-
ment area in the years 2000–2015, while the percentages of impervious surface, forest and grass are
increased from 63.5% to 72.4%, 13.5% to 16.6% and 5.1% to 6.5%, respectively. The most dramatic landuse
changes occur in the middle and downstream regions; (2) The combined landuse changes do not alter the
average flow metrics obviously at the catchment outlet, but slightly increase the high flow metrics, par-
ticularly the extreme high flows; (3) The impacts on runoff metrics in the sub-catchments are more obvi-
ous than those at the catchment outlet. For the average flow metrics, the most impacted metric is the
runoff depth in the dry season (October May) with a relative change from 10.9% to 11.6%, and the crit-
ical impact factors are the impervious surface and grass. For the high flow metrics, the extreme high flow
depth is increased most significantly with a relative change from 0.6% to 10.5%, and the critical impact
factors are the impervious surface and dryland agriculture; (4) The runoff depth metrics in the sub-
catchments are increased because of the landuse changes from dryland agriculture to impervious surface,
but are decreased because of the landuse changes from dryland agriculture or impervious surface to grass
or forest. The results of this study provide useful information for urban planning such as Sponge City
design.
Ó2018 Elsevier B.V. All rights reserved.
1. Introduction
Urbanization is the inevitable outcome along with the rapid
development of human society in the world. It has first spread
across the Western countries and then started in the developing
countries since the 1950s. In the early 1900s, the percentage of
urban population in total was just 15% in the world. By the end
of 2007, the global urbanization rate was over 50% for the first time
in human history (United Nations, 2014). It is predicted that about
64% of the developing countries and 86% of the developed coun-
tries will be urbanized by 2050 (United Nations, 2014). China, as
one of the largest developing countries, has also been experiencing
rapid urbanization since the late 1970s. The urbanization rate in
China is expected to hit 76% by 2050.
https://doi.org/10.1016/j.jhydrol.2018.03.031
0022-1694/Ó2018 Elsevier B.V. All rights reserved.
⇑
Corresponding author.
E-mail address: zhangyy003@igsnrr.ac.cn (Y. Zhang).
Journal of Hydrology 560 (2018) 247–258
Contents lists available at ScienceDirect
Journal of Hydrology
journal homepage: www.elsevier.com/locate/jhydrol
Although urbanization can improve the living conditions of res-
idents and further stimulate economic growth, it also introduces a
variety of urban environmental issues, such as waterlogging, water
shortage and pollution, and urban heat island (Xia et al., 2017). In
particular, hydrological alteration by urbanization has become a
growing concern over years. Intensive expansion of impervious
surface obstructs natural infiltration and storage of precipitation,
enhances evaporation prior to precipitation events (Braud et al.,
2013b;Ramamurthy and Bou-Zeid, 2014), and thus results in more
surface runoff yield and fast routing through land surface and drai-
nage systems (Hollis, 1975; Arnold and Gibbons, 1996; Smith et al.,
2005; Suriya and Mudgal, 2012). More surface runoff is generated
and urban waterlogging is increased due to the change of urban
micro-topography and insufficient flood control capability of urban
drainage systems. For example, waterlogging occurred in 62% of
351 investigated cities in China, and more than three severe water-
logging events happened annually in 137 cities from 2008 to 2010
according to the statistical data from the Office of State Flood Con-
trol and Drought Relief Headquarters of China. Therefore, urban
waterlogging is one of the most severe water issues in most cities
in China due to rapid urbanization. It is seriously disastrous for the
urban residents and has restricted the cities’ sustainable develop-
ment in China. The Sponge City framework has been put forward
to mitigate this urban environmental issue (MHURD, 2014). The
impact mechanism of urbanization on the hydrological cycle and
its impact magnitude are overarching scientific questions that
need to be addressed.
In the literature, the majority of studies assess the urbanization
impacts on runoff (Saghafian et al., 2008; Olang and Fürst, 2011;
Du et al., 2012; Suriya and Mudgal, 2012; Zope et al., 2016). The
general consensuses regarding impact mechanisms are that (1)
urbanization alters climate processes and thus climate variables,
such as precipitation magnitude due to enhanced convergence
(Shepherd 2005; Kaufmann et al., 2007), surface temperature and
evapotranspiration due to decreased albedo and increased thermal
storage (Ramamurthy and Bou-Zeid, 2014; Bounoua et al., 2015);
(2) urbanization induces landuse or land-cover changes by con-
verting previous land covers (e.g., dryland agriculture, grass or for-
est) to impervious lands (e.g., roads, buildings and houses) (Meija
and Moglen, 2010; Braud et al., 2013b); (3) more artificial micro-
topography and urban drainage systems disturb runoff routing
(Saghafian et al., 2008; Suriya and Mudgal, 2012). Different
approaches are adopted in these studies, including statistical anal-
ysis (e.g., correlation analysis, causality relationship analysis) for
climate variable changes (Kaufmann et al., 2007), analogical com-
parisons between urban and rural regions (Oke, 1973; Miller
et al., 2014; Zhao et al., 2014), field experiments and/or monitoring
for different underlying surfaces (Boggs and Sun, 2011; Braud et al.,
2013a), as well as hydrological simulations with consideration of
different urbanization scenarios (Lin et al., 2007; Olang and Fürst,
2011; Du et al., 2012; Suriya and Mudgal, 2012; Zope et al., 2016).
However, the statistical analysis and analogical comparisons
focus on the relationships between urbanization and hydrological
variables, and are insufficient to quantify the urbanization impacts.
Field experiments and monitoring are restricted by costs, scale
issues and uncontrollable conditions (Zhang et al., 2016b). The
hydrological modelling approach is widely accepted for the assess-
ment of urbanization impacts, and most urban hydrological models
are able to simulate precipitation-runoff processes at city sub-
domain scale with detailed mathematical descriptions of the flow
routing in the urban drainage system. However, the runoff yield
is usually estimated based on empirical relationships between pre-
cipitation and infiltration or runoff, or Curve Number method pro-
posed by Soil Conservation Service (SCS-CN), and the underlying
surfaces are usually divided into simple impervious surface and
permeable surface, which are insufficient to characterize the
detailed discrepancies among different landuse or land covers
(Suriya and Mudgal, 2012). The evapotranspiration process is also
usually simplified during the simulation (Ramamurthy and Bou-
Zeid, 2014). Additionally, most existing catchment models (e.g.,
DWSM: Dynamic Watershed Simulation Model, HEC-HMS: Hydro-
logic Engineering Center-Hydrologic Modeling System, and
DTVGM: Distributed Time Variant Gain Model) have strong advan-
tages in runoff yield estimation with consideration of detailed lan-
duse information. However, these models do not consider flow
routing in the urban drainage system, and land surface conditions
are static as a model input, rather than dynamic (Du et al., 2012;
Isik et al., 2013; Zhang et al., 2013; Zope et al., 2016). Therefore,
the applicability of hydrological model for the urbanization impact
assessment is still a challenge.
Furthermore, hydrographic discrepancies are usually adopted to
assess the impact of different factors on runoff by changing the
impact factors and then driving the hydrological model. This
approach mainly focuses on the runoff variations at river’s cross-
sections, and is not able to directly deduce the impacts on spatial
variability of hydrological processes (Du et al., 2012; Isik et al.,
2013; Zope et al., 2016), as the abundant spatial information are
not fully leveraged, including hydrometeorological observations
and terrain datasets with high spatial resolutions, and outputs of
distributed hydrological models. Multivariate statistical approach
has competitive the advantages of being able to deeply mine mul-
tiple data sources, and quantify the entire impacts of potential fac-
tors in a comprehensive manner (Zhang et al., 2016b). The
representative methods are redundancy analysis (RDA), canonical
correspondence analysis (CCA), cluster analysis and principal com-
ponent analysis (PCA), all of which have been widely adopted to
analyze the experiment data or field observations in the ecological
sciences, medicine and finance (Van den Wollenberg, 1977;
Anderson and Willis, 2003; Lepx and Smilauer, 2003). Therefore,
multivariate statistical analysis based on the outputs of distributed
hydrological models will be a more efficient approach to compre-
hensively assess the impact of urban landuse changes on hydrolog-
ical variables.
Our study investigates the urbanization impacts on runoff met-
rics with consideration of landuse changes by the integration of a
distributed catchment model and multivariate statistical approach.
The Qing River catchment, as one of the four major catchments in
the Beijing metropolitan area, is selected as the study area. The
specific objectives are to: (1) identify the underlying surface
changes and the main transition characteristics among all the lan-
duse types; (2) develop the urban hydrological model with consid-
eration of the landuse dynamic changes and the complicated urban
flow routing processes in order to accurately capture both the
inflow and outflow hydrographs at river’s cross-sections and the
spatial distributions of runoff metrics (average, low and high flows
and runoff coefficient); (3) quantify the impact of urbanization on
runoff metrics at the average and high states by multivariate statis-
tical approach, and identify the critical impact factors of landuse
changes. This study is expected to promote the further application
of catchment hydrological model in urbanization impact assess-
ment, and provide the decision-making support for urban develop-
ment and planning, especially related to Sponge City.
2. Material and methodology
2.1. Study area
The Qing River catchment (39
°
58
0
40
°
05
0
N, 116
°
12
0
116
°
29
0
E) is situated in the northern part of Beijing metropolitan area
(Fig. 1). Its total drainage area is about 169 km
2
, which accounts
for nearly 30% of the Beijing central area. The catchment belongs
248 Y. Zhang et al. / Journal of Hydrology 560 (2018) 247–258
to the semi-humid monsoon climate region with dry winter and
hot summer. The annual average precipitation from 1975 to 2015
is around 519 mm, over 80% of which falls in the wet season (from
June to September). The annual average frequency of torrential
precipitation (daily precipitation intensity 50 mm/day) is 1.43
days/year in the wet season, while there is no torrential precipita-
tion in the dry season (from October to May). The Qing River serves
as one of the four major channels (which are Qing River, Ba River,
Tonghui River and Liangshui River) for flooding drainage in Beijing
central area. The Qing River flows through Haidian, Chaoyang and
Changping districts, and finally discharges into Wenyu River in
Shunyi district. The total river length is 23.7 km.
The Qing River catchment is a typical peri-urban catchment
with a high heterogeneity of landscape including agricultural, for-
est, grass and built-up lands. In the year 2000, the agricultural land
was only about 23 km
2
and the built-up land covered over 50% of
the total catchment area. The urban landscape has experienced
drastic changes since the late 1980s. The Zhongguancun Science
and Technology Park, often referred to as the China’s Silicon Valley,
was built with many industrial buildings along central and
crowded streets. Because of the successful application of the Olym-
pic Games in 2008, the Beijing Olympic Park was built in this
catchment with 11.59 km
2
of built-up area covering major sports
halls, conference halls, hotels and forest park. In our study, the
upstream region is primarily forest covered mountainous area in
the head stream region, and the Zhongguancun Science and Tech-
nology Park, Old Summer Palace and Shangdi Science Park are dis-
tributed in the South and North of Qing River, respectively. The
middle and downstream regions are divided by Yangfang station.
The Olympic Park is located in the middle stream region, and the
Beiyuan residence community, Qingheying Country Park and Nan-
qijia Science Park are also distributed in the South and North parts
of the downstream stream region, respectively.
2.2. Data sources
Geographical information dataset includes digital elevation
model (DEM) with a grid resolution of 30 m 30 m, and landuse
Fig. 1. Locations of study area (a), landuse maps in 2000 (b), 2005 (c), 2010 (d) and 2015 (e).
Y. Zhang et al. / Journal of Hydrology 560 (2018) 247–258 249
maps with a grid resolution of 30 m 30 m in four periods (i.e.,
2000, 2005, 2010 and 2015). All of these data are collected from
the Data Center for Resource and Environmental Sciences, Chinese
Academy of Sciences (http://www.resdc.cn/). Six major landuse
types are considered in this study according to the landuse classi-
fication of China (GBT 21010-2007) (China’s national standard,
2007), namely, impervious surface, dryland agriculture, forest,
grass, water and unused land. Observed daily precipitation data
series at four gauges are downloaded from the Beijing Water
Authority (http://www.bjwater.gov.cn/). Stormwater drainage
sub-catchments of the Qing River catchment, which are delineated
according to the topography, service areas of stormwater drainage
networks, streets and residential blocks, are obtained from the
report of Beijing Water Authority General Investigation (Pan
et al., 2015).
The observed daily meteorological data series from 2008 to
2015 are downloaded from different sources. The precipitation
data of three stations (i.e., Yangfang, Tianzhu and Shahe) are from
the Beijing Water Authority (http://www.bjwater.gov.cn/), and the
precipitation, minimum and maximum temperature data of Haid-
ian station are from the China Meteorological Administration
(http://www.cma.gov.cn/). As there is no runoff observation in
the whole Qing river catchment, the outlet hydrograph of Qing
River catchment is calculated based on the calibrated results for
Wenyu Catchment (Zhang et al., 2011).
2.3. Methodology
2.3.1. Transition matrix for landuse change assessment
The transition area matrix of landuse is a classical Markov chain
model, which is widely used to assess landuse dynamics
(Waggoner and Stephens, 1970; Fu et al., 2010; Hurtt et al.,
2011). This approach is advantageous to characterize the moving
directions of landuse types and their magnitudes, which clearly
reflects the changes of landuse types at a certain period.
The transition area matrix of landuse is implemented using Arc-
GIS software (version 10.0) for all the stormwater drainage sub-
catchments. The related functions of the ArcGIS software are the
dissolve function in the data management tools, and the intersect
function in the analysis tools (http://desktop.arcgis.com/en/).
2.3.2. Distributed hydrological model and its improvements
HEQM, an integrated water system model proposed by Zhang
et al. (2016a), is applied for the simulation of urban hydrological
processes. The model focuses on rainfall-runoff-biogeochemical
processes at multiple scales (i.e., field, landuse and catchment),
with extensions of vegetation growth and interaction, soil erosion
and nutrient loss, water quality migration and transformation in
the river system, as well as alteration of dam regulations. There
are eight major modules, i.e., hydrological cycle module (HCM), soil
biochemical module (SBM), crop growth module (CGM), soil ero-
sion module (SEM), overland water quality module (OQM), water
quality module of water bodies (WQM) and dam regulation mod-
ule (DRM), and a parameter analysis tool (PAT). In this study,
HCM is adopted to simulate urban hydrological processes, includ-
ing runoff yield and routing, and is briefed below.
2.3.2.1. Hydrological processes simulation. The surface runoff yield is
calculated using the mathematics-based TVGM (Time Variant Gain
Model) proposed by Xia (1991) and further extended by Wang
et al. (2009) and Ye et al. (2015). The equation is given as:
Rs
i
¼g
1;i
ðSW
u
=W
sat
Þ
g
2;i
ðPIn
i
Þð1Þ
where Rs
i
is the surface runoff yield of the i
th
landuse type (mm);
SWu and Wsat are the soil moisture of the upper layer and satura-
tion moisture, respectively (mm); Pis the precipitation amount
(mm); g
1,i
and g
2,i
are the basic coefficient of surface runoff and soil
moisture for the i
th
landuse type, respectively; In
i
is the vegetation
interception of the i
th
landuse type (mm).
The interflow and baseflow are considered to have linear rela-
tionships with the soil moistures in the upper and lower layers,
respectively. The equation is given as:
Rss
i
¼k
ss;i
SW
u
Rbs
i
¼k
bs;i
SW
l
ð2Þ
where Rss
i
and Rbs
i
are the interflow and baseflow of the i
th
landuse
type, respectively (mm); SW
u
and SW
l
are the soil moistures (mm)
of the upper and lower layers, respectively; K
ss,i
and K
bs,i
are the
yield coefficients of interflow and baseflow for the i
th
landuse type,
respectively.
The total runoff yield of a certain sub-catchment is estimated by
Rt ¼X
n
i¼1
ðRs
i
þRss
i
þRbs
i
ÞA
i
=A
c¼Rt=P
8
>
<
>
:
ð3Þ
where Rt is the total runoff yield (mm); A
i
and Aare the areas (km
2
)
of the i
th
landuse type and sub-catchment, respectively; and cis the
runoff yield coefficient.
Evapotranspiration is also one of the critical hydrological pro-
cesses, particularly for the long-term hydrological simulation.
The potential evapotranspiration is calculated using Hargreaves
method (Hargreaves and Samani, 1982) and the actual evapotran-
spiration is a function of potential evapotranspiration, leaf area
index and surface soil residues (Ritchie, 1972).
2.3.2.2. Model improvements for landuse dynamics and urban runoff
routing.
2.3.2.2.1. Landuse dynamics. The main hydrological process con-
trolled by landuse is the surface runoff yield. The areas of different
landuse types in all sub-catchments are the input to HEQM in indi-
vidual periods. The surface runoff yield for a long-term period is
calculated as:
Rs ¼
X
n
i¼1
A
t;i
Rs
i
=At
1
<t<t
2
X
n
i¼1
A
t;i
Rs
i
=At
1
<t<t
2
......
X
n
i¼1
A
t;i
Rs
i
=At
m1
<t<t
m
8
>
>
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
>
>
:
ð4Þ
where A
t,i
is the area (km
2
) of the i
th
landuse type (i= 1,2,...,n) in the
t
th
year; t
1
,t
2
,...,t
m
are the different beginning years of landuse
maps; and mis the total number of time periods. In our study, n
=7,m=4, t
1
= 2000, t
2
= 2005, t
3
= 2010 and t
4
= 2015.
2.3.2.2.2. Urban runoff routing. Urbanization significantly mod-
ifies the water pathways through roads, drainage systems
(stormwater gutters and pipes), channels and rivers (Braud et al.,
2013b). Therefore, urban runoff routing is much more complicated
than that of a typical rural catchment. It includes flow routings in
not only the land surface and rivers, but also the complex urban
drainage system (Fig. 2). The basic flow routing equation for a cer-
tain river segment is given as:
Q
r;in
¼Q
0
r;out
þQ
o
v
erl
þQ
interf
þQ
basef
þQ
p
þQ
st
if Rs >QC
p
Q
0
r;out
þQ
interf
þQ
basef
þQ
p
þQ
st
if Rs 6QC
p
(
ð5Þ
where Q
r,in
is the inflow of river segment (m
3
/s); Q’
r,out
is the out-
flow of upper river segment (m
3
/s), which is calculated using
250 Y. Zhang et al. / Journal of Hydrology 560 (2018) 247–258
Muskingum method or kinetic wave equation in HEQM; Q
overl
,Q
interf
and Q
basef
are the routings (m
3
/s) of overland flow, interflow and
baseflow, whose calculations are adopted from Neitsch et al.
(2011) in HEQM; Q
st
is the discharge from sewage treatment plants
(m
3
/s); Q
p
and QC
p
are the flow routing (m
3
/s) and drainage capacity
(mm) in the urban drainage system, which are not able to be calcu-
lated in HEQM.
It is usually not possible to calculate the flow routing of the
drainage system using the hydrodynamic equations because the
urban drainage network data is difficult to obtain in China. How-
ever, the drainage network is always designed according to the
national industry design standards and specifications of water sup-
ply and drainage design (BMEDI, 2004). The drainage capacity of
pipeline network could then be deduced according to the related
design equations. The design precipitation intensity equation is
given as:
PI ¼8:64 167 a
1
ð1þa
2
lg PRÞ=ðt
pcp
þbÞ
n
t
pcp
¼t
o
v
erl
þmt
p
¼ð1þm
1
Þt
o
v
erl
(ð6Þ
where PI and PR are the design precipitation intensity (mm/day)
and return period (year), respectively; a
1
(mm) and a
2
are the
parameters of precipitation intensity; t
pcp
and bare the design pre-
cipitation duration (min) and its adjusted parameter (min), respec-
tively; nis the precipitation decay parameter; t
overl
and t
p
are the
flow routing durations (min) in the land surface and drainage net-
work, respectively; mand m
1
are the adjusted parameters of flow
routing duration in the drainage network. In this equation, t
overl
is
determined by the overland slope and length of sub-catchment in
HEQM. a
1
,a
2
,b,nand PR are the related parameters of design pre-
cipitation intensity, all of which can be obtained according to the
design handbook (BMEDI, 2004). In Beijing City, a
1
= 11.82 mm;
a
2
= 0.811; b= 8 min, n= 0.711 and PR = 2 years.
The drainage capacity (QC
p
) and flow routing (Q
p
) of pipeline
network are estimated as
QC
p
¼PI c
Q
p
¼kQC
p
A=86:4
ð7Þ
where cis the runoff coefficient, which is calculated in Eq. (3);Ais
the sub-catchment area (km
2
); and kis the coverage factor of drai-
nage network in the sub-catchment. The above estimation is based
on the assumptions that the Q
p
is discharged into nearby river in the
same sub-catchment, rather than other sub-catchments, and the
river’s backwater effect on the QC
p
is also ignored in this simplified
flow routing module.
Therefore, HEQM coupled with Eqs. (4)(7) is able to solve the
urban hydrological simulation in the catchments with no drainage
system data. All the calibrated parameters are given in Table 1,in
which two related parameters are involved in the simplified flow
routing module of the urban drainage system, i.e., the adjusted
parameter of flow routing duration in the drainage network (m
1
)
and the coverage factor of drainage network in the sub-
catchment (k).
2.3.2.3. Model setup and calibration. Sixteen sub-catchments are
delineated for the Qing River catchment (Pan et al., 2015), and
the six major landuse types mentioned above are considered for
each sub-catchment as the minimum spatial calculation units.
The daily observations of precipitation, minimum and maximum
temperatures are interpolated into all the sub-catchments by the
inverse distance weighting method. The sewage treatment plants
and sewage outfalls are also considered according to their
Fig. 2. Generalized maps of urban hydrological cycle (a), the framework of hydrological cycle module of HEQM and its improvement for the flow routing in the urban
drainage system (b).
Y. Zhang et al. / Journal of Hydrology 560 (2018) 247–258 251
geographical positions by directly adding the corresponding
sewage discharge amounts into the model. Based on field
investigation, one sewage treatment plant was built in the
No.1 sub-catchment in 2003, discharging 0.2 million m
3
/day
(i.e., 2.31 m
3
/s) of recycled water into the upper Qing River.
The hydrograph at the outlet of the Qing River catchment is
used for the calibration of the improved HEQM. The calibration
and validation periods are from 2009 to 2013, and from 2014 to
2015, respectively, and 2008 is the warming period. SCE-UA in
the PAT module of HEQM is adopted to automatically optimize
the model parameters in the calibration period, with the root mean
square error (RMSE:m
3
/s) being the optimization objective func-
tion. The parameter aggregation approach is adopted for calibra-
tion to make full use of spatial variation information without
changing the numbers of parameters (Yang et al., 2008). The g
1,i
and g
2,i
are calibrated by adjusting the relative amplitudes of the
initial parameter values, while the other parameters are calibrated
by directly replacing the initial parameter values. Moreover, the
bias, correlation coefficient and Nash-Sutcliffe efficiency are
adopted to further evaluate the simulation performance of HEQM.
The description of evaluation criteria can refer to Zhang et al.
(2016b), and the equations are given as:
RMSE ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
X
N
i¼1
ðO
i
S
i
Þ
2
=N
v
u
u
tð8Þ
bias ¼X
N
i¼1
ðO
i
S
i
Þ=X
N
i¼1
O
i
ð9Þ
r¼X
N
i¼1
ðO
i
OÞðS
i
SÞ=ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
X
N
i¼1
ðO
i
OÞ
2
ðS
i
SÞ
2
v
u
u
tð10Þ
NS ¼1X
N
i¼1
ðO
i
S
i
Þ
2
=X
N
i¼1
ðO
i
OÞ
2
ð11Þ
where bias,rand NS are the bias, correlation coefficient and Nash-
Sutcliffe efficiency, respectively; O
i
and S
i
are the i
th
observed and
simulated runoff (m
3
/s), respectively;
Oand
Sare the average
observed and simulated runoff (m
3
/s), respectively; Nis the series
length.
2.3.3. Runoff impact assessments
2.3.3.1. Runoff metrics and their variation assessment. Runoffs at
both average and high states are considered at the catchment out-
let and in sub-catchments for the urbanization impact assessment
in term of water resource utilization and flooding control. At the
outlet, the runoff metrics include annual average flow magnitude
(MFav: m
3
/s), annual average flow magnitudes in the wet season
(MFfs: m
3
/s) and dry season (MFnf: m
3
/s) for the average state,
high flows with the 1st, 5th and 10th percentiles (MF01, MF05
and MF10: m
3
/s) for the high state. At the sub-catchment scale,
the selected metrics include annual average runoff depth (Rd:
mm), annual average runoff depths in the wet season and dry sea-
son (Rdfs and Rdnf: mm) for the average state, annual average run-
off depths with the 1th, 5th and 10th percentiles (Rd01, Rd05 and
Rd10: mm) for the high state.
The discrepancies of runoff metrics among different landuse
scenarios are adopted to assess the impacts of landuse changes.
The runoff metrics of each sub-catchment are simulated by input-
ting the landuse maps of 2000, 2005, 2010 and 2015 (Scenarios 1
4) individually to drive the calibrated HEQM. Scenario 1 (landuse
2000) is set as the baseline and the runoff impact assessment is
conducted by comparing the runoff metrics of the other three Sce-
narios with those of Scenario 1.
2.3.3.2. Multivariate statistical approach for impact factor identifica-
tion. Rank analysis is a robust approach in multivariate statistics to
infer the correlation between two variable sets, and identify the
critical impact factors and their contributions. This approach is
commonly used to effectively gain insight into complex data in
ecological science, social science, psychological statistics, and so
on (Anderson and Willis, 2003; Shaw, 2003). There are two alterna-
tive methods for the rank analysis, i.e., non-symmetric Redundancy
Analysis (RDA) (Van den Wollenberg, 1977) and symmetric Canon-
ical Correlation Analysis (CCA) (Braak, 1986). The suitable method
is selected according to the gradient length calculated by the
Detrended Correspondence Analysis (DCA), which indicates that
if the length is smaller than 3.0, RDA is preferred, while if the
length is between 3.0 and 4.0, both RDA and CCA are suitable.
Otherwise, CCA is preferred (Hill and Gauch, 1980; Lepx and
Smilauer, 2003). The statistical value (p) for the impact factor
detection is set as 0.10 to determine the statistical significance. If
p0.10, the impact factor is statistically significant. The impact
factor is considered to be more important if its determination coef-
ficient (r
2
) with runoff metric changes is larger. Moreover, the cor-
responding constrained proportion is the contribution to runoff
metric changes.
3. Results
3.1. Assessment of landuse change
In the Qing River catchment, the dominating landuse type is the
impervious surface, which is mainly distributed in the upper and
middle stream regions with the area accounting for over 60% of
total catchment area and increasing steadily from 2000 to 2015
(Fig. 3). By the end of 2015, the impervious surface reaches 72.4%
of total area and increases by 8.9% compared with that of 2000.
The most dramatically decreased landuse type is dryland agricul-
ture which is located primarily around the mountainous region,
Table 1
Selected sensitive parameters, their ranges and aggregation approaches for model calibration.
ID Name Min Max Definition Aggregation
1W
fc
0.20 0.45 Soil field capacity Replacement
2W
sat
0.45 0.75 Soil saturation moisture capacity Replacement
3g
1,i
0.00 3.00 Basic runoff coefficient of i
th
landuse type Relative change
4g
2,i
0.00 3.00 Influence coefficient of soil moisture of i
th
landuse type Relative change
5K
ET
0.00 3.00 Adjustment factor of evapotranspiration Replacement
6K
r
0.00 1.00 Interflow yield coefficient Replacement
7T
g
1.00 100. Delay time for aquifer recharge (days) Replacement
8K
g
0.00 1.00 Baseflow yield coefficient Replacement
9K
sat
0.00 120. Steady state infiltration rate of soil (mm/hours) Replacement
10 m
1
0.00 1.00 Adjusted parameter of flow routing duration in the drainage network Replacement
11 k0.00 1.00 Coverage factor of drainage network in the sub-catchment Replacement
252 Y. Zhang et al. / Journal of Hydrology 560 (2018) 247–258
the Olympic Park in the middle stream region and the downstream
region. The decreasing trend is consistent with a reduction of
12.4% in coverage throughout the whole period, and the total
coverage is only 2.51 km
2
by the end of 2015. The most dramatic
decrease is from 2000 to 2005 with a reduction of 9.8% due to
the construction of the Olympic Park in the middle stream region.
Forest and grass occupy 13.5% to 18.3%, and 5.1% to 12.0% of total
catchment area, respectively, both of which are primarily located
in the mountainous region, around the Old Summer Palace and
the Olympic Park. These two landuse types slightly increase by
3.0% and 1.4% by the end of 2015, respectively, and the most dra-
matic changing region is in the northern part of the Olympic Park
from 2000 to 2005. Water, whose area accounts for 2.8% to 3.9% of
the total catchment area, decreases throughout the whole period,
particularly in the downstream regions. Unused land does not
change obviously throughout the whole period due to the tiny size
of area, i.e., only 0.1% to 0.8% of the total catchment area. Therefore,
the main transition of landuse is the changes of dryland agriculture
primarily to impervious surface, and secondarily to forest and
grass.
For the landuse changes among different sub-catchments
(Fig. 4), the impervious surface and dryland agriculture decrease
while the forest and grass increase in most sub-catchments in
the upper and middle stream regions from 2000 to 2005, including
the Zhongguancun Science and Technology Park and the Olympic
Park. The Zhongguancun Science and Technology Park is a highly
developed region with a large building areas and impervious roads.
As the local government adopts a more citywide outlook, more
trees and grass are planted. Moreover, the Olympic Park used to
be rural area with primarily dryland agriculture, which has chan-
ged to forest and grass gradually since the construction of the
Olympic Park. The landuse change rates in the Olympic Park are
most dramatic among all the sub-catchments with decreasing
rates of 24.1% and 46.2% for the impervious surface and dryland
agriculture, respectively, and increasing rates of 26.1% and 23.6%
for forest and grass, respectively. In the northern or downstream
sub-catchments, the dryland agriculture decreases obviously while
the impervious surface, forest and grass increase, among which the
change of impervious surface is the most obvious. The explanation
is that these regions were rural villages in the early 2000s, and
were gradually built up due to real estate development, auxiliary
commercial and transport facilities, and the constructions of infor-
mation industrial base and the science park (e.g., Shangdi, Nanqi-
jia). The landuse changes are consistent throughout the whole
period. From 2000 to 2010, the most dramatic changing sub-
catchments are in the downstream region with an increase rate
of 27.2% for impervious surface, and a decrease rate of 34.1% for
dryland agriculture. The grass area also increase obviously, partic-
ularly in the middle and downstream regions with the rates from
5.4% to 26.3%. From 2000 to 2015, the most dramatic changing
regions are also in the downstream region with a decrease rate
of 34.1% for dryland agriculture, and an increase rate of 29.6%
for impervious surface. The forest and grass areas increase slightly,
particularly in the middle and downstream regions with the rates
from 0.7% to 26.1% for forest, and from 5.5% to 14.2% for grass.
3.2. Performance of catchment hydrological simulation
In the calibration period, the simulated hydrograph at the
catchment outlet matches well with the observations with the bias,
rand NS being equal to 0.17, 0.76 and 0.55, respectively (Fig. 5a
and Table 2). In the validation period, although the runoff is
slightly overestimated (bias =0.28), the simulated runoff fits well
with the observed runoff with the rand NS of over 0.75 and 0.50,
respectively. Thus, the simulation performance is satisfactory.
Moreover, compared with the observed annual values of high flow
metrics (MF01, MF05 and MF10) (Fig. 5b), the simulated metrics
are also simulated very well with the bias,rand NS being equal
to 0.22, 0.95 and 0.88, respectively. Therefore, the improved HEQM
captures the runoff variations satisfactorily for both the calibration
and validation periods, particularly for the high flow events.
3.3. Impact assessments of urbanization
Compared with the baseline (landuse 2000), all the average
flow metrics (MFav, MFfs and MFnf) at the catchment outlet do
not change obviously with the range from only 2.2% to 0.3% dur-
ing the whole period (Fig. 6). However, the high flow metrics
(MF01, MF05 and MF10) slightly increase, among which the most
obvious changing metric is the extreme high flow (MF01) with
the changes ranging from 1.6% to 1.1% in Scenario 2 (landuse
2005), from 2.1% to 0.2% in Scenario 3 (landuse 2010) and from
1.7% to 8.1% in Scenario 4 (landuse 2015), respectively. Therefore,
the combined landuse changes seem not to alter the average flow
metrics at the catchment outlet, while the landuse changes from
dryland agriculture to impervious surface slightly increase the high
flow metrics.
The changes of runoff metrics at the sub-catchment scale are
more obvious than those at the catchment outlet (Fig. 7). Com-
pared with the baseline, the relative changes of Rd, Rdfs and Rdnf
in Scenario 2 are from 4.2% to 5.0%, from 2.7% to 2.7% and from
7.7% to 11.6%, respectively. The sub-catchments having decreas-
ing values are in most regions, particularly in the middle stream
region (e.g., the Olympic Park), while the sub-catchments having
increasing values are in the middle stream region. The landuse
changes from 2000 to 2005 explain 46% of spatial differences of
all the average runoff metrics, and the important impact factor is
the impervious surface (r
2
= 0.89, p< 0.10), followed by grass (r
2
= 0.58, p< 0.10), forest (r
2
= 0.51, p< 0.10) and dryland agriculture
(r
2
= 0.33, p< 0.10) (Table 3). In Scenario 3, the relative changes of
Rd, Rdfs and Rdnf are from 4.9% to 3.3%, from 2.6% to 2.3% and
from 10.9% to 7.6%, respectively. The number of sub-catchments
having decreasing values for all the three metrics is less than that
of Scenario 2, and most of them are also in the upstream regions.
The sub-catchments having increasing values are mostly in the
sub-catchments of the middle and downstream regions. The lan-
duse changes from 2000 to 2010 only explain 32% of spatial differ-
ences of all the average runoff metrics, and the impervious surface
is the only statistically significant impact factor (r
2
= 0.57, p< 0.10).
In Scenario 4, the relative changes of Rd, Rdfs and Rdnf are from
3.8% to 3.3%, from 1.6% to 1.9% and from 9.3% to 7.4%, respec-
tively. The sub-catchments having decreasing values change to the
middle stream regions, except Rdnf, while the sub-catchments
Fig. 3. Area percentages of different landuse types from 2000 to 2015.
Y. Zhang et al. / Journal of Hydrology 560 (2018) 247–258 253
having increasing values are in the upper and downstream regions.
The landuse changes from 2000 to 2015 explain 65% of spatial dif-
ferences of all the average runoff metrics and the important impact
factor is the impervious surface (r
2
= 0.80, p< 0.10), followed by
dryland agriculture (r
2
= 0.45, p< 0.10) and grass (r
2
= 0.43, p<
0.10). Therefore, the most impacted metric is Rdnf and the critical
impact factor is the impervious surface, followed by grass, dryland
agriculture and forest. The average runoff metrics are gradually
Fig. 4. Relative changes of landuse types in different sub-catchments from 2000 to 2015.
Fig. 5. Simulated versus observed daily hydrographs (a) in the calibration and validation periods at the outlet of Qing river catchment, and the scatter plot (b) of simulated
versus observed high flow events (10th percentile).
Table 2
Simulation performances of hydrograph and high flow events by the improved hydrolo gical model.
Periods Bias RMSE (m
3
/s) rNS
Calibration period (2009–2013) 0.17 5.70 0.76 0.55
Validation period (2014–2015) 0.28 4.77 0.76 0.52
High flow (10th percentile) 0.22 14.20 0.95 0.88
Note: RMSE is the root mean square error; rand NS are the correlation coefficient and Nash-sutcliffe efficiency.
254 Y. Zhang et al. / Journal of Hydrology 560 (2018) 247–258
increased from 2000 to 2015 in the upper and downstream regions,
which is probably caused by the landuse changes from the dryland
agriculture and forest to the impervious surface. However, these
metrics are gradually decreased in the middle stream regions,
which is probably caused by the landuse changes from the dryland
agriculture and impervious surface to the grass and forest.
The changes of high flow metrics at the sub-catchment scale are
also little greater than those of average runoff metrics. Compared
with the baseline, the relative changes of Rd01, Rd05 and Rd10
in Scenario 2 are from 0.4% to 2.8%, from 1.9% to 1.3% and from
2.2% to 2.8%, respectively (Fig. 8). The spatial changes of Rd01 are
quite different from those of Rd05 and Rd10. Rd01 increases in
most sub-catchments, except the upper stream region and the
Olympic Park, while Rd05 and Rd10 increase in the upper and mid-
dle stream regions. The landuse changes from 2000 to 2005 explain
59% of the spatial differences of all the high flow metrics and the
important impact factor is water (r
2
= 0.56, p< 0.10), followed by
grass (r
2
= 0.45, p< 0.10), dryland agriculture (r
2
= 0.41, p< 0.10)
and forest (r
2
= 0.35, p< 0.10). The relative changes of Rd01, Rd05
and Rd10 in Scenario 3 are from 0.6% to 10.5%, from 2.2% to
1.1% and from 2.2% to 3.1%, respectively. Rd01 increases in most
sub-catchments, except the upper stream region and southern sub-
catchments in the middle stream region, while the regions with
increasing Rd05 and Rd10 are in the middle and downstream
regions. The landuse changes from 2000 to 2010 explain 60% of
spatial differences of all the high flow metrics, and the impervious
surface is the only statistically significant impact factor (r
2
= 0.55,
p< 0.10). The relative changes of Rd01, Rd05 and Rd10 in Scenario
4 are from 0.5% to 4.6%, from 4.0% to 1.1% and from 1.4% to
2.9%, respectively. Rd01 also increases in most sub-catchments,
except the upper stream region and the Olympic Park, while
Rd05 and Rd10 decrease in most sub-catchment except the Zhong-
guancun Park and Olympic Park. The landuse changes from 2000 to
2015 explain 57% of spatial differences of all the high flow metrics
and the important impact factor is also the impervious surface (r
2
= 0.59, p< 0.10), followed by dryland agriculture (r
2
= 0.37, p<
0.10). Therefore, the extreme high flow (Rd01) for most sub-
catchments from 2000 to 2015 increases consistently, while the
changes of Rd05 and Rd10 are not obvious. The critical impact fac-
tor is the impervious surface, followed by dryland agriculture,
water, grass and forest. The probable explanations are that the
decrease of high flow metrics are also caused by the landuse
changes from the dryland agriculture and impervious surface to
the grass and forest, while the increase of high flow metrics is
caused by the landuse changes from the dryland agriculture to
the impervious surface.
4. Discussion and conclusions
With the rapid urbanization, the main feature of underlying
surface changes in peri-urban regions is that the agricultural land
is shifted to impervious surface (e.g., buildings, paved roads)
(Arnold and Gibbons, 1996; Meija and Moglen, 2010; Braud
et al., 2013b). In our study catchment, the landuse transition is
consistent with the results of these studies. The total area of imper-
vious surface increases from 63.5% to 72.4%, while the dryland
agriculture decreases from 13.9% to 1.5% from 2000 to 2015. The
main causes are that the construction of the information industry
bases and science parks (e.g., Zhongguancun Science and Technol-
ogy Park and Shangdi Science Park in the upper stream region, and
Nanqijia Science Park in the downstream region), the highly inten-
sive real estate development (e.g., Beiyuan residence community in
the downstream region), and the construction of the Olympic
venues and facilities in the middle stream region. Moreover, due
to the increasing attention of green urban development, the ‘‘Green
Olympics” theme for 2008 Beijing Olympic Game and the Sponge
City, several parks are renovated and built by the Beijing Gardening
Fig. 6. Annual accumulative runoff discrepancies (a) among different scenarios in the individual years, dry seasons and wet seasons (MFav, MFnf and MFfs), and annual runoff
metric discrepancies (MF01, MF05 and MF10) (b) at the outlet.
Y. Zhang et al. / Journal of Hydrology 560 (2018) 247–258 255
and Greening Bureau (Beijing Gardening and Greening Bureau,
2007, Beijing Gardening and Greening Bureau, 2011), such as the
Old Summer Palace, the Olympic Forest Park in 2008 and the
Qingheying Country Park in 2009. Thus, the total areas of grass
and forest increase slightly whose area percentages increase from
5.1% to 6.5% and from 13.5% to 16.6%, respectively.
A comprehensive assessment of the urbanization impact on
runoff metrics is rather complicated due to the diverse types and
functions of landuse. For example, on the one hand, the expansion
from dryland agriculture to impervious surface clearly increases
the runoff yield by reducing precipitation infiltration and shorten-
ing flow routing time (Saghafian et al., 2008; Suriya and Mudgal,
2012). On the other hand, the expansion from dryland agriculture
to grass or forest enhances the water retention capacity which
likely decreases the runoff yield (Shang and Wilson, 2009). Addi-
tionally, the water area is able to generate more runoff because
of the high runoff coefficient of water surface, and also enhances
the storage capacity of water bodies. In our study, although the
impervious surface area increases consistently, the constructions
of green infrastructures (e.g., grasses, trees, parks) also replace dry-
land agriculture. Therefore, the combined landuse changes in the
whole catchment do not drastically change the average flow met-
rics at the catchment outlet, but increase the high flow metrics
marginally, such as the extreme high flow with the ranges from
Fig. 7. Spatial distribution of average flow metric discrepancies (Rd, Rdfs and Rdnf) among different scenarios at the sub-catchment scale.
Table 3
Determination coefficients (r
2
) between landuse types and runoff metrics, and the contributions of landuse types on the spatial differences of runoff metrics at the sub-catchment
scale in different landuse scenarios.
State Average state (r
2
) High state (r
2
)
Landuse scenarios From 2000 to 2005 From 2000 to 2010 From 2000 to 2015 From 2000 to 2005 From 2000 to 2010 From 2000 to 2015
Impervious surface 0.83
*
0.57
*
0.80
*
0.02 0.55
*
0.59
*
Dryland 0.33
*
0.04 0.45
*
0.41
*
0.09 0.37
*
Forest 0.51
*
0.19 0.18 0.35
*
0.14 0.28
Grass 0.58
*
0.16 0.43
*
0.45
*
0.13 0.19
Water 0.04 0.05 0.19 0.56
*
0.08 0.3
Unused land 0.19 0.09 0.04 0.13 0.37
*
0.13
Contribution 46% 32% 65% 59% 60% 57%
Note: ‘‘
*
”indicates that the determination coefficient is statistically significant, i.e., p < 0.10.
256 Y. Zhang et al. / Journal of Hydrology 560 (2018) 247–258
2.1% to 8.1%. The other explanation is that as the catchment is
located in the northern semi-humid region, the initial rainfall loss
is usually so great that the runoff yield is quite small, particularly
in the dry season (IGCAS, 1980). Moreover, although the runoff
depth metrics tend to have different spatial variations in various
scenarios due to the combined runoff responses to the complicated
landuse changes, these metrics increase at most sub-catchments,
which are mainly in the highly developed regions, such as Zhong-
guancun Park, Shangdi Science Park and Beiyuan residence com-
munity. The most dramatic changing metrics are the runoff
depths in the dry season with the ranges from 10.9% to 11.6%,
and the extreme high flow depths with the ranges from 0.6% to
10.5%. Our findings about average and high flow assessments are
similar with those studies in the Qinghuai River Basin of China
with much greater relative changes of impervious surface (30%)
(Du et al., 2012), the Yzeron catchment, France (Braud et al.,
2013a), and the Oshiwara River Basin of India in the flooding
changes (Zope et al., 2016). Therefore, urbanization does not
change the water resource shortage dramatically, but probably
increases urban waterlogging and flooding to a certain extent.
The extreme precipitation event is still the critical trigger of urban
waterlogging and flooding, such as the event on July 21th, 2012 in
our study catchment.
Our investigation of the urbanization impacts on runoff
regimes finds that the catchment hydrological model is
improved and applied in the peri-urban catchment by coupling
a simplified flow routing module of the urban drainage system
and landuse dynamics. The modified module provides a simple
and feasible solution for the catchment as our study case where
no sufficient information about the drainage network available,
and the model performance is satisfactory by auto-optimization
technique. The assessment results of urbanization impacts on
runoff regimes are also reasonable. The findings are helpful to
clarify the urbanization impacts on hydrological processes in
the semi-humid cities and can provide scientific support for the
Sponge City planning and related decision-making. Although
the average annual runoff magnitude is not changed obviously,
more Sponge City measures (e.g., grassed swales, rain gardens,
trees) should be implemented for water conservation in the
sub-catchments. Other Sponge City measures (e.g., permeable
pavements and green roofs) should be gradually spread to
replace the traditional impervious surfaces for source control of
extreme flooding events. Moreover, the simplified flow routing
module of the urban drainage system should be extended in
order to further distinguish the probable inconsistencies of flow
routing paths between pipeline network and river network, as
well as their interactions. The capacities of urban real-time
monitoring, hydrological forecasting and emergency response
programs should be strengthened to aid in the management of
future extreme weather events.
Fig. 8. Spatial distribution of high flow metric discrepancies (Rd01, Rd05 and Rd10) among different scenarios at the sub-catchment scale.
Y. Zhang et al. / Journal of Hydrology 560 (2018) 247–258 257
Acknowledgments
This study was supported by the China Youth Innovation Promo-
tion Association, Chinese Academy of Sciences (CAS) (No.
2014041), Natural Science Foundation of China (No. 41671024),
the Program for ‘‘Bingwei” Excellent Talents in Institute of Geo-
graphic Sciences and Natural Resources Research, CAS (No.
2015RC201) and the International Fellowship Initiative, Institute
of Geographic Sciences and Natural Resources Research, CAS (No.
2017VP04). Thanks also to the Editor, Professor Geoff Syme, and
three anonymous referees for their constructive comments.
References
Anderson, M.J., Willis, T.J., 2003. Canonical analysis of principal coordinates: a
useful method of constrained ordination for ecology. Ecology 84, 511–525.
Arnold, C.L., Gibbons, C.J., 1996. Impervious surface coverage: the emergence of a
key environmental indicator. J. Am. Plan. Assoc. 62, 243–258.
Beijing General Municipal Engineering Design and Research Institute (BMEDI),
2004. Fifth Volume of the Water Supply and Drainage Design Handbook: Urban
Drainage. China Architecture & Building Press, Beijing.
Boggs, J.L., Sun, G., 2011. Urbanization alters watershed hydrology in the Piedmont
of North Carolina. Ecohydrology 4 (2), 256–264.
Bounoua, L., Zhang, P., Mostovoy, G., Thome, K., Masek, J., Imhoff, M., Shepherd, M.,
Quattrochi, D., Santanello, J., Silva, J., Wolfe, R., Toure, A.M., 2015. Impact of
urbanization on us surface climate. Environ. Res. Lett. 10 (8), 084010.
Braak, C.J.F.T., 1986. Canonical correspondence analysis: a new eigenvector
technique for multivariate direct gradient analysis. Ecology 67 (5), 1167–1179.
Braud, I., Breil, P., Thollet, F., Lagouy, M., Branger, F., Jacqueminet, C., Kermadi, S.,
Michel, K., 2013a. Evidence of the impact of urbanization on the hydrological
regime of a medium-sized periurban catchment in France. J. Hydrol. 485 (4), 5–
23.
Braud, I., Fletcher, T.D., Andrieu, H., 2013b. Hydrology of peri-urban catchments:
processes and modelling. J. Hydrol. 485 (485), 1–4.
China’s national standard Current land use condition classification (GB/T21010–
2007), 2007. General administration of quality supervision, inspection and
quarantine of China and Standardization administration of China, Beijing, China
(In Chinese).
Du, J.K., Qian, L., Rui, H.Y., Zuo, T.H., Zheng, D.P., Xu, Y.P., Xu, C.Y., 2012. Assessing the
effects of urbanization on annual runoff and flood events using an integrated
hydrological modeling system for Qinhuai river basin. China J. Hydrol. 464–465
(5), 127–139.
Fu, B., Gulinck, H., Masum, M.Z., 2010. Loess erosion in relation to land-use changes
in the ganspoel catchment, central Belgium. Land Degrad. Dev. 5 (4), 261–270.
Beijing Gardening and Greening Bureau, 2007. The 11th five year planning of Beijing
Garden greening.
Beijing Gardening and Greening Bureau, 2011. The 12th five year planning of
Beijing Garden greening.
Hargreaves, G.H., Samani, Z.A., 1982. Estimating potential evapotranspiration. J.
Irrig. Drain. Div. 108, 225–230.
Hill, M.O., Gauch, H.G., 1980. Detrended correspondence analysis: an improved
ordination technique. Vegetatio 42, 47–58.
Hollis, G.E., 1975. The effect of urbanization on floods of different recurrence
interval. Water Resour. Res. 11, 431–435.
Hurtt, G.C., Chini, L.P., Frolking, S., Betts, R.A., Feddema, J., Fischer, G., Fisk, J.P.,
Hibbard, K., Houghton, R.A., Janetos, A., Jones, C.D., Kindermann, G., Kinoshita,
T., Goldewijk, K.K., Riahi, K., Shevliakova, E., Smith, S., Stehfest, E., Thomson, A.,
Thornton, P., van Vuuren, D.P., Wang, Y.P., 2011. Harmonization of land-use
scenarios for the period 1500–2100: 600 years of global gridded annual land-
use transitions, wood harvest, and resulting secondary lands. Clim. Change 109,
117–161.
Institute of Geography, Chinese Academy of Sciences (IGCAS), 1980. Hydrological
Analysis and Experiment. Science Press, Beijing.
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 curve number based hybrid approach. J. Hydrol. 485 (485), 103–112.
Kaufmann, R.K., Seto, K.C., Schneider, A., Liu, Z., Zhou, L., Wang, W., 2007. Climate
response to rapid urban growth: evidence of a human-induced precipitation
deficit. J. Climate 20 (10), 2299–2306.
Lepx, J., Smilauer, P., 2003. Multivariate Analysis of Ecological Data using CANOCO.
Cambridge University Press, Cambridge, UK.
Lin, Y.P., Hong, N.M., Wu, P.J., Wu, C.F., Verburg, P.H., 2007. Impacts of land use
change scenarios on hydrology and land use patterns in the Wu-Tu watershed
in Northern Taiwan. Landscape Urban Plann. 80 (1–2), 111–126.
Meija, A.I., Moglen, G.E., 2010. Spatial distribution of imperviousness and the
spacetime variability of rainfall, runoff generation, and routing. Water Resour.
Res. 46, W07509. https://doi.org/10.1029/2009WR008568.
Miller, J.D., Kim, H., Kjeldsen, T.R., Packman, J., Grebby, S., Dearden, R., 2014.
Assessing the impact of urbanization on storm runoff in a peri-urban catchment
using historical change in impervious cover. J. Hydrol. 515, 59–70.
Ministry of Housing and Urban-Rural Development (MHURD), 2014. The
construction guideline of sponge city in China – low impact development of
stormwater system (trail). (In Chinese).
Neitsch, S., Arnold, J., Kiniry, J., Williams, J.R., 2011. SWAT2009 Theoretical
Documentation. Texas Water Resources Institute, Temple, Texas.
Oke, T.R., 1973. City size and the urban heat island. Atmos. Environ. 7 (8), 769–779.
Olang, L.O., Fürst, 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 (1), 80–89.
Pan, X.Y., Li, Q.J., Chen, J.G., Lu, L., Meng, F.C., 2015. Urban area watershed flood
simulation with hydraulic model: a case study of Qinghe river in Beijing. J.
Hydroelectric Eng. 34 (6), 71–80 (In Chinese).
Ramamurthy, P., Bou-Zeid, E., 2014. Contribution of impervious surfaces to urban
evaporation. Water Resour. Res. 50 (4), 2889–2902.
Ritchie, J.T., 1972. A model for predicting evaporation from a row crop with
incomplete cover. Water Resour. Res. 8, 1205–1213.
Saghafian, B., Farazjoo, H., Bozorgy, B., Yazdandoost, F., 2008. Flood intensification
due to changes in land use. Water Resour. Manage. 22, 1051–1067.
Shang, J., Wilson, J.P., 2009. Watershed urbanization and changing flood behavior
across the Los Angeles metropolitan region. Nat. Hazards 48, 41–57.
Shaw, P.J.A., 2003. Multivariate Statistics for the Environmental Sciences. Hodder
Arnold, London.
Shepherd, J.M., 2005. A review of current investigations of urban-induced rainfall
and recommendations for the future. Earth Interact. 9 (12), 1–27.
Smith, J.A., Baeck, M.L., Meierdiercks, K.L., Nelson, P.A., Miller, A.J., Holland, E.J.,
2005. Field studies of the storm event hydrologic response in an urbanizing
watershed. Water Resour. Res. 41, W10413.
Suriya, S., Mudgal, B.V., 2012. Impact of urbanization on flooding: the Thirusoolam
sub watershed – a case study. J. Hydrol. 412 (1), 210–219.
United Nations, Department of Economic and Social Affairs, Population Division,
2014. World Urbanization Prospects: The 2014 Revision, CD-ROM Edition.
Van den Wollenberg, A.L., 1977. Redundancy analysis an alternative for canonical
correlation analysis. Psychometrika 42 (2), 207–219.
Waggoner, P.E., Stephens, G.R., 1970. Transition probabilities for a forest. Nature
225, 1160–1161.
Wang, G., Xia, J., Chen, Ji, 2009. Quantification of effects of climate variations and
human activities on runoff by a monthly water balance model: a case study of
the Chaobai River Basin in Northern China. Water Resour. Res. 45 (7), 206–216.
Xia, J., 1991. Identification of a constrained nonlinear hydrological system described
by Volterra Functional Series. Water Resour. Res. 27, 2415–2420.
Xia, J., Zhang, Y.Y., Xiong, L.H., He, S., Wang, L.F., Yu, Z.B., 2017. Opportunities and
challenges of the Sponge City construction related to urban water issues in
China. Sci. China: Earth Sci. 60 (4), 652–658.
Yang, J., Reichert, P., Abbaspour, K.C., Xia, J., Yang, H., 2008. Comparing uncertainty
analysis techniques for a SWAT application to the Chaohe Basin in China. J.
Hydrol. 358 (1), 1–23.
Ye, A., Duan, Q., Chu, W., Xu, J., Mao, Y., 2015. The impact of the South-North water
transfer project (CTP)’s central route on groundwater table in the Hai River
basin, North China. Hydrol. Process. 28 (23), 5755–5768.
Zhang, Y.Y., Shao, Q.X., Ye, A.Z., Xing, H.T., Xia, J., 2016a. Integrated water system
simulation by considering hydrological and biogeochemical processes: model
development, with parameter sensitivity and autocalibration. Hydrol. Earth
Syst. Sci. 20, 529–553.
Zhang, Y., Xia, J., Chen, J., Zhang, M., 2011. Water quantity and quality optimization
modeling of dams operation based on SWAT in Wenyu river catchment. China
Environ. Monit. Assess. 173 (1–4), 409–430.
Zhang, X., Zhang, X., Hu, S., Liu, T., Li, G., 2013. Runoff and sediment modeling in a
peri-urban artificial landscape: case study of Olympic forest park in Beijing. J.
Hydrol. 485 (1), 126–138.
Zhang, Y.Y., Zhou, Y.J., Shao, Q.X., Liu, H.B., Lei, Q.L., Zhai, X.Y., Wang, X.L., 2016b.
Diffuse nutrient losses and the impact factors determining their regional
differences in four catchments from North to South China. J. Hydrol. 543, 577–
594.
Zhao, L., Lee, X., Smith, R.B., Oleson, K., 2014. Strong contributions of local
background climate to urban heat islands. Nature 511 (7508), 216–219.
Zope, P.E., Eldho, T.I., Jothiprakash, V., 2016. Impacts of land use–land cover change
and urbanization on flooding: a case study of Oshiwara River Basin in Mumbai,
India. Catena 145, 142–154.
258 Y. Zhang et al. / Journal of Hydrology 560 (2018) 247–258
