Ecological Indicators 58 (2015) 5563

Contents lists available at ScienceDirect

Ecological Indicators
journal homepage: www.elsevier.com/locate/ecolind

Analyses of landuse change impacts on catchment runoff using

different time indicators based on SWAT model
Bingqing Lin a , Xingwei Chen a,b, , Huaxia Yao c,d , Ying Chen a , Meibing Liu a , Lu Gao a ,
April James d
a

The Geographical Institute of Fujian Normal University, Fuzhou 350007, China

Cultivation Base of State Key Laboratory of Humid Subtropical Mountain Ecology, Fuzhou, China
c
Dorset Environmental Science Centre, Ontario Ministry of Environment and Climate Change, 1026 Bellwood Acres Road, Dorset, Ontario, Canada P0A 1E0
d
Department of Geography, Nipissing University, 100 College Drive, Box 5002, North Bay, Ontario, Canada P1B 8L7
b

a r t i c l e

i n f o

Article history:
Received 28 August 2014
Received in revised form 14 May 2015
Accepted 18 May 2015
Keywords:
Landuse change
Change in runoff
Flood indicator
Drought indicator
SWAT model
Jinjiang catchment

a b s t r a c t
Previous researches mainly focused on the runoff responses to landuse change based on annual, seasonal
or monthly time scales, there are few studies based on daily scale. We conducted a comprehensive investigation into runoff responses on the daily scale as well as annual and monthly time scales using SWAT,
and compared the impacts of time scales with different time indicators quantitatively. Jinjiang, a coastal
catchment of southeast China with a humid sub-tropical climate, was used for simulations. A calibrated
SWAT model produced satisfactory reproduction of annual, monthly and daily runoff processes over a
nine-year (20022010) period at three gauging stations. Runoff was then simulated and compared using
the same meteorological input but two different landuse scenarios (1985 and 2006, with reduced forest
and increased cropland and urbanized area). The results showed varying change in runoff among three
time scales and three catchments. The annual runoff had the smallest increase between two scenarios, monthly runoffs had medium rates (increasing in all months except OctoberNovember), and daily
runoff had the largest rates with the increase in ood peaks but decrease in drought ows, because of the
variable inuence on interception/evapotranspiration loss, percolation and antecedent soil water storage. Indicators of different time scales (annual runoff, monthly runoff, maximum 1-day and 5-day ood
runoff, minimum 1-day and 7-day runoff) proved appropriate for analysing landuse change impacts.
2015 Elsevier Ltd. All rights reserved.

1. Introduction
Water quantity and quality are critical factors affecting the ecological integrity of aquatic systems, and widespread alteration of
ow regimes requires scientic knowledge of hydrological indicators for the protection of river ecosystems (Armanini et al., 2011);
a Canadian Ecological Flow Index was developed based on benthic macroinvertebrate ow sensitivity index for Canadian rivers.
Stream runoff plays an important role in water quality and ecology
including phytoplankton, zooplankton and bacterial communities
(Godlewska et al., 2003; Sokal et al., 2010; Wu et al., 2015), and
hydrological indicators are included and used in ecological research
and management such as watershed or stream health assessments
(e.g., US EPA, 2012; Biggs et al., 2002).

Corresponding author at: The Geographical Institute of Fujian Normal

University, Shangsan Street #8, Fuzhou City 350007, Fujian Province, China.
Tel.: +86 13067215215.
E-mail address: cxwchen215@163.com (X. Chen).
http://dx.doi.org/10.1016/j.ecolind.2015.05.031
1470-160X/ 2015 Elsevier Ltd. All rights reserved.

Hydrological indicators, representing part of ecological parameters, are often altered by natural or articial processes such as
climate change and landuse change. As one of the important drivers
leading to hydrological and ecological change, landuse change may
inuence canopy interception, evapotranspiration and percolation
and eventually cause ood-drought disasters or ecological problems (Chang, 2007; Chen et al., 2009; Gebremicael et al., 2013;
Saghaan et al., 2008). The relations of hydrological components
and landuse have been researched around the world, and used to
predict the impacts of future landuse change on hydrology and
water resources (Maetens et al., 2012; Bewket and Sterk, 2005; Du
et al., 2012). Typical methodologies used in these studies included
observations from experimental catchments, time series analysis
for characteristic variables (e.g., runoff, evapotranspiration), and
simulation studies using hydrological models, as reviewed by Wei
et al. (2013).
Hydrological models including distributed physically based
models, such as SHE, TOPMODEL, HEC, VIC, IHDM, WATFLOOD
and SWAT, are capable of simulating temporal-spatial variations
in hydrological processes and assisting in understanding the

56

B. Lin et al. / Ecological Indicators 58 (2015) 5563

mechanisms of inuence behind landuse impacts. Many previous

studies have demonstrated the ability of SWAT in detecting the
impacts of landuse and climate change on hydrological components in different areas (Fan and Shibata, 2015; Nie et al., 2011;
Guo et al., 2008; Zhou et al., 2013; Gassman et al., 2007). SWAT
has been applied to investigate the inuence on water resource
by simulating annual yield for different periods which differed in
landuse (Baker and Miller, 2013; Nie et al., 2011), the response of
monthly runoff to urbanization at Charlie Creek, Florida (Dixon and
Earls, 2012), the changes in monthly and seasonal runoff over the
course of a year (increased runoff in ood season, decreased runoff
in drought season) caused by damages to natural forests in Poyang
Lake basin (Guo et al., 2008). SWAT has already been used to study
the inuence of landuse change on annual and monthly runoffs in
our study catchment, the Jinjiang catchment, a result of the rapid
economic growth, as seen elsewhere in the south-eastern coastal
region of China (Wang and Chen, 2008). However, its applications
to daily runoff changes are limited in the literature. Generally, the
studies which have focused on evaluating hydrologic change in
annual runoff do not sufciently describe changes at the daily scale.
Zhou et al. (2013) simulated change in runoff for all three time
scales (annual, monthly, daily). However, the hydrologic changes
among time scales have been rarely discussed in the literature and
what kind of time indicators is applicable for the daily runoff has
not been reported. Relative to the longer scale, the shorter-scale
runoff has less stability and more complicated relations with landuse (Wei et al., 2005). We believe that an investigation into landuse
impacts on runoff for different time scales is meaningful to illustrate
the impact mechanisms on hydrological processes. Therefore, we
chose the Jinjiang River catchment and SWAT to simulate changes
to runoff at annual, monthly and daily time scales under two landuse scenarios, and to analyze the differences in the responses with
different time indicators. Also we note and consider the role played
by precipitation in the runoff responses to landuse change.
2. Study area and data
Jinjiang catchment is located predominantly in Quanzhou City,
in south-eastern Fujian Province of China. It has an area of 5629 km2
and occupies 53.8% of Quanzhou land area, and the elevation ranges
from 0 to 1580 m. There are two major river branches within Jinjiang River, the east branch goes through Shanmei station and the
west branch goes through Anxi station, merging 2.5 km upstream
of the Shilong station. The drainage area upstream of Shilong is our
study catchment (5042 km2 , Fig. 1). The study area is characterized by a sub-tropical climate, with average annual temperature
and precipitation of 20 and 1686 mm, respectively. More than 80%
of the annual precipitation falls in the wet season (from March to
September) with both convective storms and sea-based typhoons.
The dominant landuse is forest, followed by orchard (fruit trees,
tea trees, etc.), cropland (e.g., rice and peanut land) and urbanized
area. The catchment has experienced intensive human activities
and rapid economic development over the past decades, resulting
in extensive landuse change and substantial impact on river runoff.
The spatial information and data include DEM, soil type and
landuse. The 30 m 30 m resolution DEM was obtained from the
International Scientic Data Platform of the Chinese Academy of
Sciences (http://datamiffor.csdb.cn/admin/datademMain/jsp). The
digital soil type map (1:500,000) from the Soil Fertilizer Laboratory
of Fujian Province was used to identify eleven soil types for our
catchment. Soil water characteristics (e.g., hydraulic conductivity,
available water capacity) for each soil type were obtained by using
the SPAW software developed by USDA (Saxton and Rawls, 2006).
Landuse maps are available for two separate years (1985 and
2006) and were obtained from Landsat Thematic Mapper (TM)
images by supervised classication and manual interpretation

Fig. 1. Study catchment, its river system, and gauging stations (DEM source: International Scientic Data Platform of the Chinese Academy of Sciences).

after radiometric and geometric correction. The 1985 map was

provided by the Nanjing Soil Institute of Chinese Academy of
Sciences, and the 2006 map was produced by our team. Landuse
was categorized into seven types: cropland (e.g., rice and peanut
land), forest, grassland, orchard (e.g., tea and fruit land), water,
urban, and bare land. Landuse area and their changes between
1985 and 2006 are listed in Table 1. From 1985 to 2006, the
major changes were attributed to rapid expansions of orchard
and urbanized areas, at expense of reduced cropland, forest and
grassland. Over the 21-year period, the total area of landuse change
was 44.9%, 48.9%, and 38.8% for Shilong, Anxi, and Shanmei catchments, respectively. During this period, the entire study catchment
lost cropland, grassland and forests by 49.6%, 88.3% and 9.3%,
respectively. Lost forest area was as large as 289 km2 , although
only representing 9.3% of the total watershed area. In contrast,
orchard and urban lands have signicantly increased by 1173.2%
and 227.1%, respectively. Anxi sub-catchment had undergone the
most drastic landuse change, which decreased in cropland and
forest by 54.3% and 19.1%, increased in orchard and urban by
1254.4% and 280.7%, respectively. The Shanmei sub-catchment
had smaller changes (forest lost by only 2.4%) because it belongs
to the environmental management area of the Shanmei Reservoir.
We selected 20012010 as the study period, because discharge
data before 2001 were not available. The 20012010 are used
to represent post landuse that coincides with the 2006 landuse
map, and the 9 years of data allow examination of inter-annual
variability (e.g., variation in precipitation). River discharge data
(20012010) were provided by the Water Conservation Agency
of Fujian Province for the stations at Anxi and Shilong, and by
Quanzhou City for the Shanmei station. Meteorological data (daily
precipitation, daily maximum temperature, daily minimum temperature, daily mean relative humidity, daily mean wind speed)
for 20012010 were provided by the Meteorology Agency of Fujian
Province, including two weather stations at Yongchun and Dehua,
and 32 rain gauges in the catchment (Fig. 1).
3. Methodology
3.1. Application of SWAT
SWAT (Soil and Water Assessment Tool) is a physically based
distributed hydrologic model operating on a daily time step and
is designed to predict impacts of land management practices on

B. Lin et al. / Ecological Indicators 58 (2015) 5563

57

Table 1
Characteristics of landuse changes in three catchments from 1985 to 2006.
Landuse

Cropland
Forest
Grassland
Orchard
Water
Urban
Bareland

Shilong

Anxi

Shanmei

1985 (km2 )

2006 (km2 )

19852006%
change

1985 (km2 )

2006 (km2 )

19852006%
change

1985 (km2 )

2006 (km2 )

19852006%
change

1222.2
3094.8
338.2
83
45.2
119.8
138.6

617.7
2805.8
39.4
1057.1
61.9
391.7
68.2

49.6
9.3
88.3
1173.2
36.9
227.1
50.8

476.7
1627
196.6
53
9.8
35.9
67

218.1
1316.3
25.7
718.1
14.8
136.7
36.2

54.3
19.1
86.9
1254.4
51.8
280.7
45.9

219.8
660.9
60.6
14.2
9.3
13
21.4

121.7
645.1
8.5
144.4
17
51.3
11.2

44.6
2.4
86
917
82.8
295.7
47.6

water, sediment and agricultural yields (Arnold et al., 1998; Liu

et al., 2003). In the application of SWAT model, the study catchment is rst divided into multiple sub-watersheds based on a DEM,
which are then further subdivided into hydrologic response units
(HRUs) based on topography, landuse and soil types. The hydrologic
cycle for each HRU is simulated based on the water balance, including precipitation, interception, surface runoff, evapotranspiration,
percolation and lateral ow from the soil prole, and return ow
from shallow aquifers (Gassman et al., 2007). Two methods exist in
SWAT for estimating surface runoff: the Soil Conservation Service
(SCS) curve number (USDA-SCS, 1972) and the Green & Ampt inltration method. A kinematic storage model (Sloan and Moore, 1984)
is used to calculate redistribution of water between soil layers.
The soil water may move laterally in the prole and contribute
to streamow, and percolate past the bottom of soil prole and
ultimately become aquifer recharge. Aquifer recharge is portioned
into shallow aquifer which ultimately becomes return ow, and
deep aquifer recharge, which is lost from watershed (Neitsch et al.,
2009; Gassman et al., 2007). Flows are simulated from the HRUs to
the sub-catchment level, and then routed through the stream system using either variable storage routing method or Muskingum
routing method. A more detailed description of the model and the
calculation formula of hydrologic cycle are provided in the SWAT
Technical Manual (Neitsch et al., 2009).
Jinjiang catchment was divided into 99 sub-catchments and 886
HRUs for the 1985 landuse, and 870 HRUs for the 2006 landuse.
The SCS curve number method was used for estimating surface
runoff, the PenmanMonteith method for potential evapotranspiration, and the kinematic storage model for lateral ows. River ow
routing was achieved by the Muskingum method.
A sensitivity analysis was conducted on model parameters
by using one-factor-at-a-time (LH-OAT) global sensitivity analysis procedure embedded in SWAT model (Van Griensven et al.,
2006), and the resulting list of sensitive parameters were calibrated against the observed runoff data from the three gauging
stations. The use of multiple gauging stations can improve the effectiveness of model calibration and increase models representation
of spatial variance (Cao et al., 2006). Years 2001, 20022007 and
20082010 were treated as the warm-up, calibration and validation
periods, respectively. Calibration and validation periods used the

2006 landuse map. During calibration and validation, calculated

annual, monthly and daily runoff were evaluated against observed
data from Anxi, Shanmei and Shilong stations using three indices:
Nash-Sutcliffe coefcient of efciency (Ens), coefcient of determination (R2 ), and Percent Bias (PBIAS, %). When all three indices are
consistent with the given criteria for a monthly time step (Ens > 0.5,
R2 > 0.7, PBIAS <25%), the model was thought applicable to the
catchment and to the impact analyses (Moriasi et al., 2007).
Given that discharge data is not available prior to 2001,
the calibrated SWAT model was used to simulate streamows
under the 1985 landuse map, which represented the pre-landuse
change scenario. Hydrological response was simulated by xing
the meteorological inputs for 20022010 and changing the landuse scenario from the 1985 to the 2006 for all time scales (annual,
monthly and daily). Changes in annual, monthly and daily runoff
between the two scenarios were calculated using a series of indicators and evaluated for quantifying landuse impacts.
3.2. Time indicators for runoff
A series of time indicators for different time scales of runoff were
selected. For annual time scale, total annual runoff was used. For
monthly time scale, total runoff in each month was used. For daily
time scales, the maximum one-day (1 d) runoff and maximum veday (5 d) runoff in a year were taken as the ood indicators, and
the minimum one-day (1 d) runoff and minimum seven-day (7 d)
runoff taken as drought indicators.
4. Results
4.1. Model calibration
Seven sensitive parameters were identied: SOL AWC,
RCHRG DP, CN2, GWQMN, ESCO, SOL K, and CANMX (Table 2). Each
sensitive parameter was manually calibrated, and the adjusted
value range or optimal values of these parameters were listed in
Table 2. Satisfactory model performance was achieved (Table 3).
The evaluation index Ens and R2 were all more than 0.9, and
PBIASs were all less than 7% for annual and monthly scales at three

Table 2
Sensitive parameters and optimal values for the Jinjiang river catchment.
No

Name

Description

Range

Adjusted value range

or optimal value

1
2
3

SOL AWC
RCHRG DP
CN2

01
01
3598

0.110.34
0.3
4494

GWQMN

5
6
7

ESCO
SOL K
CANMX

Available water capacity of the soil layer

Deep aquifer percolation fraction
Initial SCS curve number for moisture
condition
Threshold depth of water in the shallow
aquifer acquired for return ow to occur, mm
Soil evaporation compensation factor
Saturated hydraulic conductivity, mm/h
Maximum canopy storage, mm

0500

30

01
02000
0100

0.50.8
0.279.2
4

58

B. Lin et al. / Ecological Indicators 58 (2015) 5563

Fig. 2. Comparison of simulated and observed daily runoff at Shilong catchment: (a) 20022004 (part 1 of calibration period), (b) 20052007 (part 2 of calibration period),
(c) 20082010 (validation period).

catchments, both for the calibration and validation periods. As for

daily scale, the Ens and R2 were all greater than 0.75, and PBIAS
was less than 6.5% for all catchments and periods.
It is natural that the daily indices are worse than the annual or
monthly indexes, as there is more variability and greater instability
in daily runoff processes. Nevertheless, the daily runoff of 9 years
(Fig. 2, Shilong catchment example) also indicated a satisfying t
between the simulated and observed. As a result of satisfying ts in
runoffs at three time scales (Table 3), the SWAT model was thought
to be a reliable representation of hydrologic processes and can be
used to simulate responses for the catchment studied.

3.3%, 3.4% and 2.3% for Shilong, Anxi and Shanmei catchments. The
larger change at Anxi than at Shanmei is consistent with the more
extensive landuse change in the Anxi sub-catchment.
The obvious inter-annual variance in hydrologic change among
the 9 years is noted from Fig. 3, and is partly attributed to differences in annual precipitation. There is a clear tendency that when
precipitation was larger, change in runoff between 1985 and 2006
landuse scenarios was smaller (Fig. 4). Generally, the increase in
annual runoff might be caused by the decrease in canopy interception, evapotranspiration and percolation; however, these impacts
have a limitation, and their inuence in percentage becomes less if
the denominator runoff (or precipitation) becomes larger.

4.2. Impacts of landuse change at three temporal scales

4.2.1. Annual scale
Annual runoff increased at three catchments for all years with
landuse change (Fig. 3), an expected response when well vegetated lands were changed to orchard and urban. The increase in
annual runoff ranged from 1.3% to 8.2%, with average changes of

4.2.2. Monthly scale

Changes in monthly runoff due to landuse change varied from
month to month. The average monthly changes are shown in
Fig. 5. There were similar seasonal trends for all three catchments,
and the change at Anxi catchment was the most signicant. The
seasonal variance and changing direction have been noted. In the

B. Lin et al. / Ecological Indicators 58 (2015) 5563

59

Table 3
Model performance: calibrated and validated results for annual, monthly and daily
runoff for Shilong, Anxi and Shanmei catchments.
Calibration (20022007)

Validation (20082010)

Ens

PBIAS (%)

R2

Ens

PBIAS (%)

R2

Annual
Shilong
Anxi
Shanmei

0.978
0.941
0.972

3.81
4.00
3.76

0.99
0.97
0.99

0.970
0.967
0.970

6.36
6.66
5.01

0.99
0.99
0.99

Monthly
Shilong
Anxi
Shanmei

0.944
0.943
0.970

3.80
4.07
3.83

0.97
0.95
0.97

0.922
0.916
0.970

5.88
6.14
4.73

0.94
0.94
0.98

Daily
Shilong
Anxi
Shanmei

0.851
0.818
0.873

3.38
4.00
3.77

0.86
0.84
0.87

0.805
0.761
0.856

5.97
6.24
4.78

0.83
0.81
0.87
Fig. 5. Average change in monthly runoff between 1985 and 2006 landuse scenarios
for the three catchments. Inter-annual variation in precipitation also shown.

Fig. 3. Change in annual runoff (20022010) for three catchments between 1985
and 2006 landuse scenarios. Inter-annual variation in precipitation also shown.

wet season (March to September), monthly runoff increased by

1.38.2%, and had a negative relation with monthly precipitation
amount (seen from Fig. 5) which was similar to the annual case.
However, in the dry season (October to February), monthly runoff
decreased in the early dry season (October and November), and
then increased in the late dry season (December to February), with
increasingly larger % change. Therefore, under the same climate
conditions, landuse change generated more runoff in the wet and
late dry seasons, but less runoff in the early dry season.

4.2.3. Daily scale

Change in daily runoff between 1985 and 2006 landuse scenarios were clearly greater than that of monthly or annual time
scales. The change in maximum 1 d runoff (peak ow in a year)
ranged from 1.1% to 23.9% for the 9 years (Fig. 6a), greater than the
1.3% to 8.2% change in annual runoff, and had more spatial variance among three catchments. The average maximum 1 d runoff
increased by 6.2%, 9.5% and 3.8% at Shilong, Anxi and Shanmei
catchments, respectively. The changes were less distinct for the
maximum 5 d runoff than maximum 1 d, but still more pronounced
than the change in annual or monthly runoff (Fig. 6b). Averaged
change in 5 d runoff was 5.6%, 7.9% and 3.4% for Shilong, Anxi, and
Shanmei catchments, respectively.
In order to investigate the impacts of antecedent precipitation
on ood responses, relationships between the change in maximum 1 d runoff and the antecedent 5-day precipitation (A5 d),
and between the change in maximum 5 d runoff and antecedent
10-day precipitation were shown in Fig. 7a and b, respectively.
The gures indicated that changes in runoff due to landuse change
during oods were related to antecedent precipitation. When
the antecedent 5-day rainfall varied from 100 to 400 mm, the
change in ood peak varied from 23.9% to 1.1% (Fig. 7a). Similarly,
when the 10-day rainfall varied from 150 to 450 mm, change
in ood peak was substantially reduced. Therefore, simulated
changes in daily runoff were controlled by both landuse change
and severity of ood events.
In contrast to the increasing runoff exhibited during ood
conditions, simulated change in runoff during drought showed

Fig. 4. Relationship between change in annual runoff and annual precipitation for Shilong, Anxi and Shanmei catchments.

60

B. Lin et al. / Ecological Indicators 58 (2015) 5563

Fig. 6. Change in ood runoff with landuse change: (a) 1 d peak ow, (b) 5 d ood
ow.
Fig. 8. Change in drought runoff with landuse change: (a) minimum 1 d ow,
(b) minimum 7 d ow.

decreasing runoff with landuse change. The minimum 1 d runoff

and minimum 7 d runoff decreased in most years and locations
(Fig. 8). On average over the 9 years, the minimum 1 d runoff
decreased by 3.6%, 7.4% and 0.7%, and the minimum 7 d runoff
decreased by 2.3%, 9.4% and 0.6%, at the Shilong, Anxi and Shanmei

catchments respectively. The variations among the three catchments suggests that the largest change in drought runoff was
associated with the most landuse change (e.g., Anxi catchment),
and the inter-annual variations were related to rainfall intensity.

Fig. 7. Relationship between change in ood indices due to landuse change and antecedent rainfall: (a) 1 d peak ow, (b) 5 d ood ow.

B. Lin et al. / Ecological Indicators 58 (2015) 5563

For example, at Shilong station, the 10-day antecedent rainfall

before the minimum 1 d runoff in 2008 reached up to 29.6 mm, and
the lowest runoff did not decrease as in other years but increased.
Therefore, the decreasing trend in drought runoff with landuse
change exhibited by SWAT simulations may be weakened by occasional rainfall prior to the drought event.
5. Discussion
5.1. The runoff responses with different time scales
Simulated runoff changes attributable to the landuse change
(1985 versus 2006 scenarios), under meteorological inputs of
20022010 are summarized and compared in Table 4. Annual
runoff was the least sensitive to landuse change. For example,
average annual runoff for the Anxi catchment increased by 3.4%
(2.3 m3 /s), which was much smaller than the daily increase of
9.5% (104.4 m3 /s). The range of changes in annual, monthly and
daily runoff are 1.48.2% (0.94.8 m3 /s), 4.3% to 8.8% (1.4 m3 /s
to 7.1 m3 /s) and 19.5% to 23.9% (1.6 m3 /s to 197 m3 /s) at Anxi
catchment, respectively. Both the change in runoff and differences
in these changes for the three catchments increase with shorter
time scales, with the peak ows showing the largest changes with
landuse scenarios.
Runoff responses and their differences among time scales are
generally in agreement to research literatures with some exceptions, which are determined by the roles played by landuse in
hydrological processes of interception, inltration, evapotranspiration, recharge to soil or ground water. For annual time scales,
afforestation and plantation usually increase evapotranspiration
and soil inltration capacity, leading to reduced runoff (e.g., Wang
et al., 2005; Yao et al., 2009). For the landuse change applied to
our study catchment (1985 versus 2006 scenarios), an examination of SWAT generated water budget components shows that
the average annual evapotranspiration (ET, Table 5) decreased
from 532.05 mm to 520.18 mm, percolation (PERC) decreased from
433.49 mm to 384.62 mm, and runoff (water yield in streamow,
WYLD) increased from 865.57 mm to 900.42 mm. This indicates
that in our SWAT simulations deforestation and urbanization
decreased evapotranspiration, percolation loss to depth, and led to
elevated runoff (Rose and Peters, 2001; Zhang et al., 2012). According to annual water balance, changes in runoff were inuenced
mainly by the reduced evaporation and percolation losses when the
precipitation was xed. However, the impacts of landuse change
on evapotranspiration and percolation are limited (Table 4), and
the change in annual runoff becomes smaller when precipitation is
larger (Figs. 3 and 4).
Monthly runoff was determined by both the evapotranspiration
loss within a month and the antecedent water storage in the catchment. In early spring or March, soils in the study catchments were
typically dry and rainfall was small (Yu et al., 2013), which was
often seen in sub-tropical and temperate regions of China or Asia
and differs from colder regions like North America where March
could be the wettest month in a year. The forest cover was mainly
evergreen conifers and deciduous trees, therefore the losses from
canopy interception, evapotranspiration, and percolation in early
spring make up a large part of total rainfall during spring compared to the growing season. If the landuse change reduced foliar
biomass, interception and percolation loss, this reduced water loss
could mean a signicant contribution to runoff generation, resulting in the large changes in runoff displayed for March (Fig. 5).
During the growing season (AprilSeptember), simulations show
a decreasing change in runoff along the months (Fig. 5), because
there was a monthly increase in rainfall and soil water, interception
and percolation make up less of total rainfall than early spring. The
landuse change still reduced interception, evapotranspiration and

61

percolation loss, but their contribution to runoff increase becomes

relatively smaller than in early spring.
The runoff decrease in the early dry season (October and
November) could be explained by the changes in soil water storage prior to the dry season. The water stored in soil and bedrock
during the antecedent wet season could contribute to the streamow in the early dry season. However, the storage was reduced
because of landuse change from 1985 to 2006, and resulted in a
decrease of runoff in October and November. For the remaining
part (December, January and February) of the dry season, the soil
water storage did not differ much between the two landuse scenarios, but the altered vegetation condition reduced evaporation
and percolation loss and again increased the runoff. A similar pattern of monthly runoff responses was noted by Zhou et al. (2013)
in Xitiaoxi basin, which have similar catchment characteristics as
Jinjiang basin. These factors of evapotranspiration and soil storage
have contributed to the greater inuence rates in monthly runoff
than in annual runoff.
For daily runoff like peak or drought ow, canopy interception, percolation and soil water holding capacity may have a
dominant role in formulating a days runoff. The deforestation
and urbanization had reduced interception and percolation loss,
which led to a signicant increase in the maximum 1 d ood peak
or maximum 5 d runoff. However, the capacity of interception
and inltration was limited in natural vegetation, and landuse
change had less inuence on peak ow when precipitation was
larger (Fig. 7). The minimum 1 d runoff (drought) was mostly controlled by groundwater ow which was determined more by soil
water condition or rainfall recharge. The landuse change reduced
recharges to groundwater and led to a decrease in the drought
ow.
The response in peak or ood ow to landuse change has been
well documented: increased forest cover leads to reduced oods
and reduced forest cover leads to increased ood (e.g., Yao et al.,
1994). But caution should be used to infer response in drought ow
to other places. Our simulation in this study suggested that the 9.3%
reduction in forest upstream Shilong led to a decrease in drought
ow by 3.6%. A simulation study by Yao et al. (2009) for the opposite situation, reforestation also indicated a decreased drought ow
(decreasing 41% in drought ow under a 30% increase in leaf area
index). The major reasons for our results were the reduced inltration and recharge during the wet season, while in their study the
reason was the elevated evapotranspiration (or reduced soil water
storage) prior to the drought season. This may not be the case in
other catchments.
As demonstrated by our studies, the inuence on daily or
extreme runoffs is stronger than on monthly or annual runoffs,
and a better understanding of the impacts at daily scale is required
to obtain a complete assessment of landuse change impacts. The
results of Chang (2007) and Zhou et al. (2013) supported this
thought. But their studies had not yet reported the impacts on
drought ows, although they revealed the impacts of urbanization on annual, monthly runoff and peak ows. Furthermore, the
experimental catchment method as utilized by Chang (2007) had
a greater restriction in its applications than the modeling method
as we used. Our trials have covered all the time scales and both
the oods and droughts, compared the impact differences temporally and spatially, and discussed the roles played by antecedent
rainfall and landuse change severity, which could provide a deeper
insight to the impact mechanisms and broader implementations to
the landuse planning and management.
5.2. Time indicators for runoff
There have been various runoff indicators used in hydrology and
ecology for different time scales. Popular indicators for longer time

62

B. Lin et al. / Ecological Indicators 58 (2015) 5563

Table 4
Summary of discharge variation (Q, m3 /s) and its percent change (R, %) attributable to landuse change (1985 versus 2006 scenarios) for all catchments and time scales.
Time indicators for runoff

1985 scenario2006 scenario

Shilong

Anxi

Shanmei
3

Q (m /s)

R (%)

Q (m /s)

R (%)

Q (m3 /s)

R (%)

Annual

Range
Average

2.78
4.8

1.75.9
3.3

0.94.8
2.3

1.48.2
3.4

0.71.2
0.9

1.34.2
2.3

Monthly

Range
Average

1.5 to 12.6
4.8

1.3 to 7.5
3.3

1.4 to 7.1
2.3

4.3 to 8.8
3.3

0.1 to 2.5
0.9

0.3 to 5.6
2.3

Maximum 5 d

Range
Average

42.987.1
63.1

2.38.7
5.6

28.271.1
45.5

3.511.9
7.9

7.816.8
10.6

1.66.3
3.4

Maximum 1 d

Range
Average

41144
122.7

1.111.8
6.2

40.4197
104.4

4.1423.9
9.5

8.740.2
20.1

1.17.6
3.8

Minimum 7 d

Range
Average

2.4 to 0.2
0.4

7.1 to 0.8
2.3

1.6 to 0.01
0.2

19.5 to 0.4
9.4

0.11 to 0.06
0.02

2.3 to 0.8
0.6

Minimum 1 d

Range
Average

1.5 to 0.4
0.3

16.5 to 6.1
3.6

0.9 to 0.01
0.2

18.3 to 0.5
7.4

0.1 to 0.03
0.03

2.1 to 0.4
0.7

Table 5
Annual water budgets of Shilong catchment under two landuse conditions of 1985 and 2006.
Year

PRECIP (mm)

SURQ (mm)

LATQ (mm)

GW Q (mm)

PERC (mm)

TLOSS (mm)

ET (mm)

WYLD (mm)

1985

2002
2003
2004
2005
2006
2007
2008
2009
2010
Mean

1739.74
1342.17
1393.89
1835.87
2205.99
1738.59
1630.88
1241.09
2049.59
1686.42

280.10
209.56
173.73
365.50
506.62
355.23
250.68
134.99
343.91
291.15

395.83
327.95
344.04
459.04
543.15
437.54
425.49
312.13
566.10
423.47

89.02
82.83
86.84
184.76
242.90
198.78
149.66
66.41
280.33
153.50

366.12
304.37
335.60
482.76
582.68
475.20
436.29
296.54
621.81
433.49

1.85
1.71
1.73
3.02
4.17
2.73
2.59
1.50
3.65
2.55

627.94
560.23
540.67
507.36
540.11
519.64
502.44
469.33
520.71
532.05

763.10
618.63
602.88
1006.28
1288.50
988.82
823.24
512.03
1186.69
865.57

2006

2002
2003
2004
2005
2006
2007
2008
2009
2010
Mean

1739.74
1342.17
1393.89
1835.87
2205.99
1738.59
1630.88
1241.09
2049.59
1686.42

356.92
267.27
234.81
445.58
600.12
433.11
341.44
182.74
442.83
367.20

386.39
313.34
332.46
439.80
520.37
423.44
414.73
298.80
545.27
408.29

76.30
66.18
69.95
150.42
200.31
167.34
130.56
53.92
235.94
127.88

332.75
261.23
296.90
422.39
508.37
426.16
399.22
263.58
551.02
384.62

2.32
2.01
2.1
3.27
4.54
3.18
3.15
1.93
4.07
2.95

596.24
560.05
524.28
512.43
544.94
494.11
463.72
473.58
512.31
520.18

817.29
644.78
635.12
1032.53
1316.26
1020.71
883.58
533.53
1219.97
900.42

Note: PRECIP is annual precipitation, SURQ is the surface runoff contribution to streamow, LATQ is the lateral ow contribution to streamow, GW Q is the groundwater
ow contribution to streamow, PERC is the percolation from the root zone to the aquifer zone, TLOSS is the water loss from reach by transmission through the streambed,
ET is the actual evapotranspiration, and WYLD is the annual water yield in streamow (WYLD = SURQ + LATQ + GWQ TLOSS). The precipitation PRECIP is basically balanced
with the sum of SURQ + LATQ + PERC + ET over a year or longer period of time, without including the minor changes in catchment water storage.

scales had included annual total runoff or annual-mean discharge,

and monthly total runoff or monthly mean discharge (Ma et al.,
2009; Baker and Miller, 2013; Li et al., 2009; Dixon and Earls, 2012),
which were used in our analyses. A ow duration curve was one
informative method of displaying the complete range of river ows
from low ows to ood events, and could be utilized for analysing
impacts on hydrology (e.g., Yao et al., 2009). There existed a variety
of low-ow measures and indices (Smakhtin, 2001), such as annual
minimum 1-day ow and minimum 7-day ow. We used these two
measures as the low-ow indicators. There could be more choices
of indicators for ood runoffs, depending on the basin size, climate
region or other factors. For example, Zhou et al. (2013) described
the entire ood process with daily runoffs from the onset of the
ood event to the end of the ood recession, and used the annual
maximum 1-day runoff to express the ood peak; Liu et al. (2012)
used the annual maximum 1-day and 7-day runoff as the ood
indicators. For our study and catchments, a ood span of ve days
was frequently seen from observed data records, and the calculation step of SWAT was one day, therefore combining these facts,
we selected the maximum 1-day and 5-day runoff to be the ood

indicators. All the time indicators (year, month, day) chosen for our
study are among those indicators typically used, capable of reecting hydrological characteristics of the studied catchment and the
landuse change impacts. They would be proper indicators to other
researches or catchments.
6. Conclusions
SWAT was calibrated for the Jinjiang River catchment, with reasonable reproduction of annual, monthly or daily runoff processes
for nine years. It was then applied to simulate the impacts of landuse change on runoff. When the landuse changed by increasing
urbanized, orchard area and decreasing in forest area from 1985 to
2006, changes in runoff varied among the three time scales and the
three catchments. Annual runoff exhibited the smallest changes,
monthly runoff the medium changes (increasing in all months
except OctoberNovember), and daily runoff the largest changes
(increasing in ood peaks but decreasing in drought ows), because
of the variable inuence on interception/evapotranspiration loss,
percolation to depth and antecedent soil water storage.

B. Lin et al. / Ecological Indicators 58 (2015) 5563

Landuse change intensity determined the spatial variation of the

impacts, as shown by the three catchments. These catchments had
similar runoff response patterns to the landuse change, under different time scales, whereas the response magnitude was different:
greater runoff changes with greater landuse changes.
The change in runoff response was also associated to precipitation volume: smaller annual change in runoff in wetter years
but greater change in runoff in drier years, and declining monthly
change in runoff with increasing precipitation within a wet season.
It was found that commonly used indicators like minimum
1-day runoff worked well for analyzing the landuse impacts on
drought or low ows, while indicators for the impacts on oods
or ood process would more depend on catchment size and local
ood characteristics, and would be associated with the time step
of an applied model.
Acknowledgements
The study was nancially supported by the Science and Technology Major Program of Fujian Province (Grant No. 2015Y4002),
the Key Specialized Program of Public Research Institutes of Fujian
(Grant No. 2013R04) and the National Natural Science Foundation
of China (No. 41301031). We sincerely thank two reviewers and the
editor for their kind comments.
References
Armanini, D.G., Horrigan, N., Monk, W.A., Peters, D.L., Baird, D.J., 2011. Development
of a benthic macroinvertebrate ow sensitivity index for Canadian rivers. River
Res. Appl. 27, 723737.
Arnold, J.G., Srinivasan, R., Muttiah, R.S., Williams, J.R., 1998. Large area hydrologic
modelling and assessment part I: Model development. J. Am. Water Resour.
Assoc. 34 (1), 7389.
Baker, T.J., Miller, S.N., 2013. Using the Soil and Water Assessment Tool (SWAT) to
assess landuse impact on water resources in an East African watershed. J. Hydrol.
486, 100111.
Bewket, W., Sterk, G., 2005. Dynamics in land cover and its effect on stream ow
in the Chemoga watershed, Blue Nile basin. Ethiop. Hydrol. Process. 19 (2),
445458.
Biggs, B.J.F., Kilroy, C., Mulcock, C.M., Scarsbrook, M.R., 2002. New Zealand stream
health monitoring and assessment kit. Stream Monit. Man., Version 2. NIWA
Technical Report 111. 190 pp.
Cao, W., Bowden, W.B., Davie, T., Fenemor, A., 2006. Multi-variable and multi-site
calibration and validation of SWAT in a large mountainous catchment with high
spatial variability. Hydrol. Process. 20 (5), 10571073.
Chang, H., 2007. Comparative streamow characteristics in urbanizing basins in the
Portland Metropolitan Area, Oregon, USA. Hydrol. Process. 21 (2), 211222.
Chen, Y., Xu, Y., Yin, Y., 2009. Impacts of landuse change scenarios on storm-runoff
generation in Xitiaoxi basin, China. Quat. Int. 208 (1), 121128.
Dixon, B., Earls, J., 2012. Effects of urbanization on streamow using SWAT with real
and simulated meteorological data. Appl. Geogr. 35 (1), 174190.
Du, J., Qian, L., Rui, H., Zuo, T., Zheng, D., Xu, Y., Xu, C.Y., 2012. Assessing the effects of
urbanization on annual runoff and ood events using an integrated hydrological
modeling system for Qinhuai River basin, China. J. Hydrol. 464, 127139.
Fan, M., Shibata, H., 2015. Simulation of watershed hydrology and stream water
quality under landuse and climate change scenarios in Teshio River watershed,
northern Japan. Ecol. Indic. 50, 7989.
Gassman, P.W., Reyes, M.R., Green, C.H., Arnold, J.G., 2007. The Soil and Water Assessment Tool: historical development, applications, and future research directions.
Trans. ASABE 50 (4), 12111250.
Gebremicael, T.G., Mohamed, Y.A., Betrie, G.D., Zaag, P., Teferi, E., 2013. Trend analysis of runoff and sediment uxes in the Upper Blue Nile basin: a combined
analysis of statistical tests, physically-based models and landuse maps. Hydrology 482, 5768.
G., Pociecha, A., Wilk-Wozniak,

Godlewska, M., Mazurkiewicz-Boron,

E., Jelonek, M.,
2003. Effects of ood on the functioning of the Dobczyce reservoir ecosystem.
Hydrobiologia 504 (13), 305313.
Guo, H., Hu, Q., Jiang, T., 2008. Annual and seasonal streamow responses to climate and land-cover changes in the Poyang Lake basin, China. J. Hydrol. 355 (1),
106122.

63

Liu, C.M., Li, D.F., Tian, Y., Hao, F.H., Yang, G.L., 2003. An application study of DEM
based distributed hydrological model on macroscale watershed. Prog. Geogr. 2
(5), 437445 (in Chinese).
Liu, L.L., Jiang, T., Xu, J.G., Zai, J.Q., Luo, Y., 2012. Response of hydrological processes
to the climate change in the Zhujiang River Basin in the 21st century. Progressus
in Quisitiones Demutatione Climatis 8 (1), 2834 (in Chinese).
Li, Z., Liu, W.Z., Zhang, X.C., Zheng, F.L., 2009. Impacts of landuse change and climate
variability on hydrology in an agricultural catchment on the Loess Plateau of
China. J. Hydrol. 377 (1), 3542.
Ma, X., Xu, J., Luo, Y., Prasad Aggarwal, S., Li, J., 2009. Response of hydrological processes to land-cover and climate changes in Kejie watershed, south-west China.
Hydrol. Process. 23 (8), 11791191.
Maetens, W., Vanmaercke, M., Poesen, J., Jankauskas, B., Jankauskien, G., Ionita,
I., 2012. Effects of landuse on annual runoff and soil loss in Europe and the
Mediterranean: a meta-analysis of plot data. Prog. Phys. Geogr. 36 (5), 597651.
Moriasi, D.N., Arnold, J.G., Van Liew, M.W., Bingner, R.L., Harmel, R.D., Veith, T.L.,
2007. Model evaluation guidelines for systematic quantication of accuracy in
watershed simulations. Trans. ASABE 50 (3), 885900.
Neitsch, S.L., Arnold, J.R., Williams, J.R., 2009. Soil and Water Assessment Tool Theoretical Documentation, version 2009. http://swat.tamu.edu/documentation/
Nie, W., Yuan, Y., Kepner, W., Nash, M.S., Jackson, M., Erickson, C., 2011. Assessing
impacts of landuse and landcover changes on hydrology for the upper San Pedro
watershed. J. Hydrol. 407 (1), 105114.
Rose, S., Peters, N.E., 2001. Effects of urbanization on streamow in the Atlanta area
(Georgia, USA): a comparative hydrological approach. Hydrol. Process. 15 (8),
14411457.
Saghaan, B., Farazjoo, H., Bozorgy, B., Yazdandoost, F., 2008. Flood intensication
due to changes in landuse. Water Resour. Manage. 22 (8), 10511067.
Saxton, K.E., Rawls, W.J., 2006. Soil water characteristic estimates by texture and
organic matter for hydrologic solutions. Soil Sci. Soc. Am. J. 70 (5), 15691578.
Sloan, P.G., Moore, I.D., 1984. Modeling subsurface stormow on steeply sloping
forested watersheds. Water Resour. Res. 20 (12), 18151822.
Smakhtin, V.U., 2001. Low ow hydrology: a review. J. Hydrol. 240 (34), 147186.
Sokal, M.A., Hall, R.I., Wolfe, B.B., 2010. The role of ooding on inter-annual and
seasonal variability of lake water chemistry, phytoplankton diatom communities and macrophyte biomass in the Slave River Delta (Northwest Territories,
Canada). Ecohydrology 3 (1), 4154.
USDA, SCS, 1972. National Engineering Handbook, Section 4: Hydrology. Washington, DC.
US EPA, 2012. Identifying and Protecting Healthy Watersheds. EPA technical document. EPA Ofce of Wetlands, Oceans, and Watersheds, Washington, DC http://
water.epa.gov/healthywatersheds
Van Griensven, A., Meixner, T., Grunwald, S., Bishop, T., Diluzio, M., Srinivasan, R.,
2006. A global sensitivity analysis tool for the parameters of multi-variable
catchment models. J. Hydrol. 324 (1), 1023.
Wang, L., Chen, X.W., 2008. Simulation of hydrological effects on vegetation restoration of degraded mountain in ecosystem with SWAT model. J. Mt. Sci. 26 (1),
7175 (in Chinese).
Wang, G.X., Zhang, Y., Liu, G.M., 2005. Effects of landuse change on runoff from 1967
to 2000 in the Maying River basin. Sci. China Ser. D: Earth Sci. 35 (7), 671681
(in Chinese).
Wei, X., Liu, W., Zhou, P., 2013. Quantifying the relative contributions of forest change
and climatic variability to hydrology in large watersheds: a critical review of
research methods. Water 5 (2), 728746.
Wei, X.H., Li, W.H., Zhou, G.Y., Liu, S.R., Sun, G., 2005. Forest and
streamowconsistence and complexity. J. Nat. Resour. 20 (5), 761770
(in Chinese).
Wu, H.P., Zeng, G.M., Liang, J., Guo, S.L., Dai, J., Lu, L.H., Wei, Z., Xu, P., Li, F., Yuan, Y.J.,
He, X.X., 2015. Effect of early dry season induced by the Three Gorges Dam on
the soil microbial biomass and bacterial community structure in the Dongting
Lake wetland. Ecol. Indic. 53, 129136.
Yao, H., Hashino, M., Yoshida, H., 1994. Analyzing effects of deforestation and
afforestation on streamow by using a physically-based conceptual model. J.
Jpn. Soc. Hydrol. Water Resour. 7 (3), 196203.
Yao, H.X., Hashino, M., Xia, J., Chen, X.H., 2009. Runoff reduction by forest growth in
Hiji River basin, Japan. Hydrol. Sci. J. 54, 556570.
Yu, L., Zhang, Y.P., Sha, L.Q., Song, Q.H., Tan, Z.H., Zhou, W.J., Wu, C.S., 2013. Temporal
variation pattern and its affecting factors of soil water content in a subtropical
evergreen broadleaved forest in Ailao Mountains, Southwest China. Chin. J. Ecol.
32 (2), 332336 (in Chinese).
Zhang, M., Wei, X., Sun, P., Liu, S., 2012. The effect of forest harvesting and climatic
variability on runoff in a large watershed: the case study in the Upper Minjiang
River of Yangtze River basin. J. Hydrol. 464, 111.
Zhou, F., Xu, Y., Chen, Y., Xu, C.Y., Gao, Y., Du, J., 2013. Hydrological response
to urbanization at different spatio-temporal scales simulated by coupling of
CLUE-S and the SWAT model in the Yangtze River Delta region. J. Hydrol. 485,
113125.