You are on page 1of 13

Uncertainty of SWAT model at different DEM

resolutions in a large mountainous watershed


Peipei Zhang, Ruimin Liu*, Yimeng Bao, Jiawei Wang, Wenwen Yu,
Zhenyao Shen
State Key Laboratory of Environment Simulation and Pollution Control, School of Environment, Beijing Normal
University, Beijing 100875, China
a r t i c l e i n f o
Article history:
Received 6 August 2013
Received in revised form
4 January 2014
Accepted 9 January 2014
Available online 21 January 2014
Keywords:
Uncertainty analysis
SWAT model
DEM resolution
Non-point Source pollution
Mountainous watershed
a b s t r a c t
The objective of this study was to enhance understanding of the sensitivity of the SWAT
model to the resolutions of Digital Elevation Models (DEMs) based on the analysis of
multiple evaluation indicators. The Xiangxi River, a large tributary of Three Gorges
Reservoir in China, was selected as the study area. A range of 17 DEM spatial resolutions,
from 30 to 1000 m, was examined, and the annual and monthly model outputs based on
each resolution were compared. The following results were obtained: (i) sediment yield
was greatly affected by DEM resolution; (ii) the prediction of dissolved oxygen load was
signicantly affected by DEM resolutions coarser than 500 m; (iii) Total Nitrogen (TN) load
was not greatly affected by the DEM resolution; (iv) Nitrate Nitrogen (NO
3
eN) and Total
Phosphorus (TP) loads were slightly affected by the DEM resolution; and (v) ow and
Ammonia Nitrogen (NH
4
eN) load were essentially unaffected by the DEM resolution. The
ow and dissolved oxygen load decreased more signicantly in the dry season than in the
wet and normal seasons. Excluding ow and dissolved oxygen, the uncertainties of the
other Hydrology/Non-point Source (H/NPS) pollution indicators were greater in the wet
season than in the dry and normal seasons. Considering the temporal distribution un-
certainties, the optimal DEM resolutions for ow was 30e200 m, for sediment and TP was
30e100 m, for dissolved oxygen and NO
3
eN was 30e300 m, for NH
4
eN was 30 to 70 m and
for TN was 30e150 m.
2014 Elsevier Ltd. All rights reserved.
1. Introduction
Water quality has become a challenging issue, threatening the
security of society and ecosystems in China. Nonpoint Source
(NPS) has been an important contribution to the pollution of
surface water (Ongley et al., 2010). Because of this wide range,
the difculty of control NPS pollution and complex un-
certainties, NPS pollution has also become a focus in the eld
of water pollution control worldwide (Wu et al., 2012). More-
over, many hydrological models have been developed to pre-
dict hydrological processes and water quality (Yuan et al.,
2007; Zhang et al., 2011; Lai et al., 2011; Shen et al., 2012b).
The ability of numerical models to accurately represent
hydrological processes and to describe nutrient loads related
to water quality greatly depends on the accuracy of the input
data. However, studies scarcely considered the uncertainties
originating from the accuracy of modeling input data (Sharma
* Corresponding author. Tel./fax: 86 10 5880 0829.
E-mail address: liurm@bnu.edu.cn (R. Liu).
Available online at www.sciencedirect.com
ScienceDirect
j ournal homepage: www. el sevi er. com/ l ocat e/ wat res
wa t e r r e s e a r c h 5 3 ( 2 0 1 4 ) 1 3 2 e1 4 4
0043-1354/$ e see front matter 2014 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.watres.2014.01.018
et al., 2011) and it need to be put more attentions (Bormann,
2008). Model input data are actually the primary sources of
uncertainty in hydrological modeling (Earls and Dixon, 2005;
Dixon and Earls, 2012; Faramarzi et al., 2013; Shen et al.,
2013b). Some studies have noted that the use of ner resolu-
tion spatial data in larger catchments does not necessarily
improve the performance of hydrological model applications
(Ndomba and Birhanu, 2008); however, generally, the preci-
sion of the modeling results increases with increasing accu-
racy of the input data (Booij, 2005; Casper et al., 2011; Katrin
et al., 2011). Much attention has been paid to uncertainty is-
sues in hydrological modeling due to their great effects on
prediction and further decision-making (Bruneau et al., 1995;
Li et al., 2010; Shen et al., 2012a).
Digital Elevation Models (DEMs) are important input data
for Hydrology (H)/NPS modeling, and DEM uncertainty has
previously attracted the attention of many researchers
(Zhang and Montgomery, 1994; Bruneau et al., 1995; Thieken
et al., 1999; Endreny et al., 2000; Gertner et al., 2002; Chaplot,
2005). DEM resolution has been proven to directly impact the
hydrologic model predictions from TOPography based hy-
drological MODEL (TOPMODEL) (Band and Moore, 1995; Lin
et al., 2010; Xu et al., 2011), the Water balance Simulation
Model (WASIM) (Bormann et al., 2009), the Soil Water
Assessment Tool (SWAT) model (Band and Moore, 1995; Lin
et al., 2013; Shen et al., 2013b), the topographic land-
atmosphere transfer scheme (TOPLATS) model (Endreny
et al., 2000; Bormann, 2006) and the Annualized Agricul-
tural Non-Point Source pollution (AnnAGNPS) model (Wang
and Lin, 2011). Cotter et al. (2003) and Chaplot (2005) found
that the results of the SWAT model were signicantly
affected by the input of the DEM data. Gericke and Venohr
(2012) proved that DEM resolution was highly relevant in
modeling sediment yield using an empirical hydrological
model. DEM accuracy also has a signicant effect on the
prediction of the water balance simulation based on the
TOPLATS model (Endreny et al., 2000), especially when the
DEMresolution coarser than 500 min the central German Dill
catchment (693 km
2
) (Bormann, 2006). The uncertainty of
DEM resolution on the topographic characteristics, hydrolo-
gy modeling based on the TOPMODEL was investigated by
previous authors (Quinn et al., 1991; Moore et al., 1993;
Wolock and Price, 1994; Zhang and Montgomery, 1994;
Bruneau et al., 1995). Additionally, Sharma et al. (2009) also
found that DEMs produced by different interpolation
methods have different contour levels, and Dixon and Earls
(2009) conrmed that resampled DEMs do not produce the
same modeling results as the original DEMs.
Research has also been undertaken to determine the
appropriate grid cell resolution for particular analyses (Albani
et al., 2004; Wechsler, 2007; Bormann et al., 2009; Kim et al.,
2012). The DEM resolution generally has different optimal
thresholds for different model usages, and changes in reso-
lutions have no impacts on the model results if the input DEM
resolution is in the optimal threshold or more accurate
(Watson et al., 1998). Moreover, high-resolution data generally
require more money, time, and computer memory. Brown
et al. (1993) showed that the sediment yield prediction of the
Areal Non-point Source Watershed Environment Response
Simulation (ANSWERS) model began to change at DEM grid
sizes greater than 120 m. Previous studies using SWAT
revealed that runoff changed slightly with decreases in DEM
resolution, whereas the sediment yield decreased greatly
(Chaplot, 2005; Song et al., 2012; Lin et al., 2013). It is therefore
essential to study the optimal resolution of spatial input data;
as an important source of uncertainty, it is the key to
improving the performance of hydrological models.
Previous studies have analyzed the uncertainties in
annually SWAT output due to DEM resolution, but little
attention has been given to the impact of DEM resolution on
the high-resolution temporal distribution (for example,
the monthly distribution) of H/NPS modeling predictions
(Lee et al., 2009; Song et al., 2012; Lin et al., 2013; Shen et al.,
2013b). Temporal hydrological conditions were important in
the quantication of NPS export (Chen et al., 2011; Shen et al.,
2013a). Ouyang et al. (2006) indicated that the seasonal
change in surface water quality was important for evaluating
temporal variation in river pollution. Cho et al. (2009)
assessed the seasonal and spatial uncertainties in hydro-
logic and water quality simulations of SWAT models based
on rainfall data. It is essential to study the sensitivity of
monthly modeling outputs due to the big difference of hy-
drology indicators in wet season, normal season and dry
season. So this study will conduct a high-resolution temporal
uncertainty analysis on H/NPS modeling resources fromDEM
resolution.
Therefore, the Xiangxi River Watershed, a catchment in
the Three Gorges Reservoir Area of China, was selected as the
study area, and this study attempts to examine the sensitivity
of H/NPS modeling to DEM resolutions using SWAT model.
The optimal DEM resolutions were identied for each H/NPS
indicator. Furthermore, the effects of the uncertainties of the
DEM resolutions on the temporal distribution of these in-
dicators were also examined. This study also provides insight
into the selection of the resolution of DEM inputs into the
model, with consideration of temporal uncertainty of H/NPS
modeling. The methods employed and factors considered in
this study maybe can help guiding the selective of DEM reso-
lution of other H/NPS model.
2. Methodology
2.1. Study area
The Xiangxi River is located between 110.47

and 111.13

E and
30.96

and 31.67

N in the Three Gorges Reservoir. The


drainage area covers approximately 2995 km
2
and is
controlled by the Xingshan Hydrological Gauge. The river is
94 km long, and its altitude ranges from 110 m to nearly
3088 m (Fig. 1). The Xiangxi River Watershed has a humid
subtropical monsoon climate, where the summer is hot and
wet, and the winter is cold and dry. June, July and August were
the wet season; April, May, September and October were the
normal season; January, February, March, November and
September was the dry season. 6.5% of the study area was
covered by farm land, 70.9% was forest, 5.3% was water and
4.4% was wasteland (Liu et al., 2013).
The Xiangxi River has faced serious NPS pollution and is
the focus of much research (Liu et al., 2013). Previous studies
wa t e r r e s e a r c h 5 3 ( 2 0 1 4 ) 1 3 2 e1 4 4 133
have already proven that SWAT can successfully predict the
ow, sediment yield and nutrient outputs of the Xiangxi
River Watershed (Xu et al., 2010; Wang et al., 2011; Liu et al.,
2013).
2.2. SWAT model and uncertainty analysis
SWAT, a semi-distributed physically based model, which
was originally developed by the United States Department of
Agriculture Agricultural Research Service, can predict the
effect of management decisions on water, sediment, nutrient
and pesticide yields with reasonable accuracy for large,
ungaged river basins (Arnold et al., 1998; Neitsch et al., 2005).
The ow was estimated using the modied Soil Conservation
Service (SCS) Curve Number Method (SCS, 1972), and sedi-
ment was calculated with the Modied Universal Soil Loss
Equation (Williams and Berndt, 1977). Nutrient outputs are
estimated by tracking the improvements and trans-
formations. Soluble nitrogen and phosphorus loads were
estimated with the water volume and average concentra-
tions of nitrogen and phosphorus in soil layers (Lin et al.,
2013). The calculation function of sediment transport of ni-
trogen and phosphorus loads are developed by McElroy et al.
(1976) and modied by Williams and Hann (1978). In interest
of brevity, the detailed description of SWAT was not repeated
here. More information can be found in Arnold et al. (2013)
and Neitsch et al. (2005).
In SWAT, the watershed was divided into multiple sub-
basins based on topographic features of the watershed
calculated using DEM data. For SWAT, the topographic attri-
butes of the subbasins, such as area, slope and eld slope
length, were all derived from the DEM. The length, slope,
width, and depth of the channel were also derived from the
DEM, if channels were automatically generated based on DEM
but not previously dened (Lin et al., 2013). DEM majorly
impact the ow and nutrients modeling through topographic
attributes (Chaubey et al., 2005; Jha et al., 2004). Whats more,
the process description in SWAT rely on DEM data will also
impact the modeling results, such as watershed subdivision
and HRU division in the subbasin (Jha et al., 2004).
The modeling results of 2000 and 2001 were set as the
warm-up period and 2002 to 2006 were used to analyze the
uncertainties. All model calculations are done on a daily time
scale. Due to the model parameter calibration can impact the
uncertainty come from input data (Abbaspour, 2013), so the
SWAT model will not be calibrated in this study. Whats more,
in order to separate the uncertainty resources from the other
data or the process description during modeling, the other
simulation conditions will be kept constant, these conditions
included: (1) the other input data, e.g. land use, soil, climate,
pond, reservoir and land management; (2) drainage area of
4000 ha for stream denitions to produce subbasins; (3) the
same HRUs denition thresholds of land use (5%), soil (10%)
and slope (10%); and (4) the same other default parameters by
SWAT. The SWAT outputs for ow, sediment yield, and loads
of dissolved oxygen, Nitrate Nitrogen (NO
3
eN), Ammonia Ni-
trogen (NH
4
eN), Total Nitrogen (TN) and Total Phosphorus
(TP) were used to analyze the uncertainty of the model. The
outputs obtained based on the 30 m DEM data were consid-
ered the best model outputs and were used as a baseline to
compare with the outputs of the other DEM resolutions. This
comparison will be described for three aspects of the
modeling: the watershed delineation, the annual outputs and
the average monthly outputs.
The uncertainty of the SWAT model output was quantied
using multiple indicators (Wechsler, 2007): the relative dif-
ference (RD) (Casper et al., 2011), the mean difference (MD), the
mean absolute difference (MAD), the root mean standard
Fig. 1 e Location of the Xiangxi River.
wa t e r r e s e a r c h 5 3 ( 2 0 1 4 ) 1 3 2 e1 4 4 134
difference (RMSD) and the standard deviation of the MADRD(p)
(Chaplot, 2005):
RD
P
i
P
30
P
30
(1)
MD
1
n
X
n
i1
P
i
P
30
(2)
MAD
1
n
X
n
i1
jP
i
P
30
j (3)
RMSD
(
1
n
X
n
i1
P
i
P
30

)
0:5
(4)
p
(
1
n
X
n
i1
jP
i
P
30
j MAD
2
)
0:5
(5)
where P
30
is the monthly predicted model output using the
best resolution DEM (30 m), and P
i
is the monthly predicted
model output from the coarser DEM resolutions. P
30
was used
as the standard against which to measure the coarser reso-
lutions because it had the nest spatial resolution and was
assumed to be more representative.
2.3. Input data
The following data were selected as SWAT model inputs:
(1) All DEMs input SWAT including 30, 40, 50, 60, 70, 80, 90,
100,150, 200, 300, 400, 500, 600, 700, 800 and 1000 m,
where 30 m was (or was close to) the original resolution
for ASTER 30 m. DEM at lower resolutions was obtained
by resample from30 mresolution using nearest neighbor
assignment method in ArcGIS 10.0 software package.
(2) Landsat 5 TM (30 m resolution) satellite imagery was
analyzed to produce the land use data of study area in
2010 based on ENVI and eld investigation.
(3) The soil maps (30 m resolution) were got from Institute
of Geography, Chinese Academy of Sciences in 2010.
Fig. 2 e Slopes of the Xiangxi River estimated from DEM resolution of 30 me1000 m.
wa t e r r e s e a r c h 5 3 ( 2 0 1 4 ) 1 3 2 e1 4 4 135
Soil physical and chemical characteristics were
collected from Chinese soil database and local survey.
(4) 32 years (1/1/1979e12/31/2010) of daily weather gauge
data (precipitation, temperature, solar radiation, wind
speed, and relative humidity) from63 stations in or near
the watershed were got from the National Centers for
Environmental Prediction Climate Forecast System
Reanalysis.
(5) The Xiangxi River Watershed was primarily a mountain
watershed without signicant point source pollution, so
there was no input data of point source pollution. The
crop practices were set as model default values, by
which auto fertilization would be applied when soil Nor
P concentrations were below the thresholds. Land
management data was collected from Xingshan eld
survey in 2010.
All of the digital data were projected to the
WGS_1984_UTM_Zone_49N projection during the processing
with ArcGIS. All the input data was treated as the detailed
requested data format by the SWAT model (Arnold et al.,
2013).
The mean slope extracted from the 30 m DEM was 62%, but
greater slopes (maximumof 81%) were observed in the central
valley of the Xiangxi River Watershed (Fig. 2). The Xiangxi
River Watershed has smooth topography in the central
portion and higher altitudes elsewhere; approximately 80% of
the watershed was greater than 1000 m in altitude.
3. Results
3.1. Impact on topographic and watershed
characteristics
According to the topographic and watershed characteristics
calculated from the respective DEM resolutions, the
maximum altitude was underestimated by 123 m at a DEM
resolution of 700 m, and the minimum altitude was over-
estimated by 24 m (800 m resolution); furthermore, the mean
altitude decreased signicantly from 1599 m (30 m resolution)
to 1266 m(1000 mresolution) (Table 1). The maximumaltitude
decreases, and the minimum altitude increases, with
decreasing DEMresolution, and the conclusion was agree with
(Thieken et al., 1999). In general, decreasing DEM resolution
tends to slightly narrow the range of altitudes, which could be
attributed to the loss of detailed topographic characteristics.
The maximum slope changed greatly, from 85% (30 m reso-
lution) to 30% (1000 m resolution), and the minimum slope
changed from 0 (30 m resolution) to 0.18% (1000 m resolution).
The standard deviation of the altitudes and slopes decreased
withthe decreasing DEMresolutions, indicating that the lower
resolution reduced the variation in altitudes and slopes.
For the watershed delineation, the subbasins remained
unchanged when the DEM resolutions ranged from 30 to
300 m, and then, the number of subbasins decreased non-
linearly as the DEM resolution decreased (Fig. 3). As the DEM
resolutions decreased from 30 to 500 m, the number of HRUs
increased due to the narrowing of the watershed area, which
leads the graphical features were more easily to reach the
T
a
b
l
e
1
e
E
s
t
i
m
a
t
e
d
w
a
t
e
r
s
h
e
d
c
h
a
r
a
c
t
e
r
i
s
t
i
c
s
b
a
s
e
d
o
n
e
a
c
h
D
E
M
r
e
s
o
l
u
t
i
o
n
s
i
n
S
W
A
T
.
3
0
m
4
0
m
5
0
m
6
0
m
7
0
m
8
0
m
9
0
m
1
0
0
m
1
5
0
m
2
0
0
m
3
0
0
m
5
0
0
m
6
0
0
m
7
0
0
m
8
0
0
m
1
0
0
0
m
A
l
t
i
t
u
d
e
M
i
n
1
1
1
1
1
1
1
1
1
1
1
3
1
1
3
1
1
3
1
1
4
1
1
5
1
1
1
1
2
6
1
2
8
1
3
0
1
3
1
1
3
0
1
3
6
1
3
4
M
a
x
3
0
8
8
3
0
8
8
3
0
8
6
3
0
8
6
3
0
8
8
3
0
8
8
3
0
8
8
3
0
8
5
3
0
7
6
3
0
7
9
3
0
3
4
3
0
6
9
3
0
2
0
2
9
6
5
3
0
1
0
3
0
2
0
M
e
a
n
1
5
9
9
1
5
9
9
1
5
9
7
1
5
9
2
1
5
8
6
1
5
7
6
1
5
7
7
1
5
6
4
1
5
2
7
1
4
8
8
1
4
2
9
1
3
2
9
1
3
1
5
1
2
9
1
1
2
7
7
1
2
6
6
S
D
8
5
9
8
5
6
8
5
3
8
5
2
8
4
7
8
4
1
8
4
0
8
3
4
8
1
3
7
9
1
7
5
9
6
9
1
6
2
2
6
2
9
6
1
1
5
7
8
S
l
o
p
e
(
%
)
M
i
n
0
0
0
0
0
0
0
0
0
.
0
7
0
.
1
0
.
0
3
0
.
0
8
0
.
0
5
0
.
1
8
0
.
1
1
0
.
0
4
M
a
x
8
0
.
5
7
8
.
7
7
5
.
9
7
4
7
2
.
7
7
0
6
9
.
2
6
8
.
2
6
0
.
6
5
6
.
4
4
7
.
7
3
9
3
5
.
6
3
3
.
6
3
2
.
1
3
0
M
e
a
n
2
5
.
2
2
4
.
9
2
4
.
7
2
4
.
5
2
4
.
2
2
3
.
9
2
3
.
6
2
3
.
2
2
1
.
6
2
0
.
2
1
7
.
8
1
4
.
5
1
3
.
2
1
2
.
2
1
1
.
3
9
.
9
S
D
1
1
.
3
1
1
.
5
1
1
.
2
1
0
.
9
1
0
.
8
1
0
.
6
1
0
.
4
1
0
.
3
9
.
7
9
.
3
8
.
5
7
.
3
6
.
8
6
.
3
5
.
9
5
.
4
W
a
t
e
r
s
h
e
d
l
e
n
g
t
h
(
k
m
)
4
1
7
4
1
2
4
0
9
4
0
6
4
0
3
4
0
0
3
9
6
3
9
2
3
8
7
3
7
7
3
7
0
3
3
8
3
3
8
3
2
6
3
2
3
3
1
8
A
r
e
a
(
h
a
)
2
9
1
9
2
9
1
7
2
9
1
5
2
9
1
3
2
9
1
0
2
9
0
8
2
9
0
6
2
9
0
2
2
9
3
1
2
8
7
2
2
8
0
1
2
7
2
9
2
7
0
0
2
6
9
4
2
6
4
2
2
5
9
3
wa t e r r e s e a r c h 5 3 ( 2 0 1 4 ) 1 3 2 e1 4 4 136
threshold for dividing one more HRU that is a combination of
land use (5%), soil (10%) and slope (10%).
3.2. Impact of DEM resolutions on annual outputs
For ow, all of the evaluation indicators were small when the
resolution was 30e1000 m, and the greatest RD was 0.13 at
1000 mresolution, indicating that the annual owwas slightly
underestimated using the coarser DEMs (Table 2). The coarser
DEMs undoubtedly increased the uncertainties of the sedi-
ment predictions: the mean absolute difference and p
respectively approach 221 and 3428 at 700 m resolution;
furthermore, the root mean standard difference were greater
than 1000 at DEM resolutions coarser than 90 m. However, the
sediment prediction at 800 m was similar to that at 90 m, and
1000 mwas similar to 150 m; this may be due to the changes in
the delineation of the subbasins and HRUs (Arabi et al., 2006).
The dissolved oxygen load was slightly affected by the DEM
resolutions of 30e300 m, and the uncertainty greatly
increased with coarser DEM resolutions. Similar to dissolved
oxygen, the NO
3
eN load was slightly affected by the DEM
resolution; the greatest RD was only 3.7% at 500 m, although
the p-values from500 mto 1000 mwere 337e656 (Table 2). The
NO
3
eN load was unrelated to the DEM resolution from 30 to
300 m. The NH
4
eN load declined steadily from 30 to 1000 m,
especially when the DEM resolution was coarser than 300 m.
However, the RDs of NH
4
eN at 500 m and 800 m were only
14% and 10%, respectively, and the other indicators were
small enough to ignore.
The RDs of TN load for resolutions of 700 m and 1000 m
were up to 47% and 46%, respectively, which are greater RD
values than observed for the other resolutions, in a similar
pattern to that of the other indicators. However, the RDs of TN
at 500 mand 800 mwere smaller, which may be because of the
increasing numbers of HRUs and subbasins. The TP load
decreased slightly with DEM resolutions from 30 to 200 m,
whereas these RDs were greater than 130% with DEM resolu-
tions coarser than 200 m, and the maximum RD approached
1420% (800 m). Furthermore, the other indicators showed that
the uncertainty of TP was stable with changes in DEM
resolution.
3.3. Impact of DEM resolutions on monthly outputs
The uncertainty analysis related to the monthly outputs of
stream ow is illustrated in Fig. 4(a). The ow changes little
with changes in DEM resolutions from 30 to 1000 m. However,
the predicted ow decreased slightly when the resolution was
coarser than 150 m, and the RDs of ow will be bigger than 5%.
It could thus be inferred that the ow changes little with
decreasing DEM resolution.
The monthly sediment yield decreased greatly as the DEM
resolution changed from 30 m to 300 m and then uncharac-
teristically increased as the resolution changed from 500 m to
800 m. This may be due to the change in the relative distri-
bution of different land uses, leading to increased numbers of
HRUs, when the DEM resolution was coarser than 500 m. The
sediment yield at 500 m was greater than at 600 m, and at
800 m, the yield was higher than at 700 m and 1000 m
(Fig. 4(b)); thus, fewer subbasins and HRUs may have produced
greater sediment loads.
Neither dissolved oxygen load nor NO
3
eN load was sensi-
tive to DEMresolutions from30 mto 300 m, whereas when the
resolution was coarser than 300 m, both loads decreased
signicantly (Fig. 4(c) and (d)). Thus, the DEM resolution must
be 300 m or ner to model dissolved oxygen and NO
3
eN loads.
Similar to ow, NH
4
eN load changes little at DEM resolu-
tions from 30 to 150 m (Fig. 4(e)). NH
4
eN loads increased again
at 800 m, when the number of subbasins and HRUs increased
to 33 and 665, respectively, from 31 to 574 at 700 m. It could
thus be inferred that the NH
4
eN loads increased as the
number of HRUs and subbasins increased. Monthly peak
values of TN load were not sensitive to DEM resolutions of
30 me100 m, and the peak values then decreased with reso-
lutions coarser than 100 m (Fig. 4(f)). As illustrated in Fig. 4(g),
the predicted TP loads decreased slightly from 30 m to 300 m.
Then, the uncertainties increased with coarser DEM resolu-
tions, and the lowest TP load was observed at the resolution of
700 m. It is clear that the overall distribution of TP was similar
to that of TN.
3.4. Temporal variation of monthly outputs
The RDs of ow for January, February, March and August were
high, and the RDs for May, July, September and December
were in the middle; nally, the RDs for April, June, October and
November were low (Fig. 5). The variation was more signi-
cant in the dry and normal seasons than in the wet seasons.
The optimal DEM resolutions for ow estimation were
30e300 m.
The sediment results showed that the sensitivity of sedi-
ment yield does not have obvious temporal distribution rules
with coarser DEM resolutions. Dissolved oxygen load was
underestimated by SWAT from 2002 to 2006, and the absolute
RD values in the dry season were higher than in the wet and
normal seasons, similar to the pattern observed for ow.
When the DEM resolutions were coarser than 300 m, NO
3
eN
load increased in May and June and decreased in the other
months; in addition, the monthly RD was highly variable,
especially in the wet and normal seasons. NH
4
eN load was
underestimated slightly when the DEM resolutions were from
30 to 200 m (RD was within 20%), and the monthly RDs were
Fig. 3 e The number of subbasins and HRU at different DEM
resolutions.
wa t e r r e s e a r c h 5 3 ( 2 0 1 4 ) 1 3 2 e1 4 4 137
Table 2 e Comparison of the annual ow, sediment and nutrients loads in DEM resolutions 40 me1000 m with 30 m.
Output Indicator 40 50 60 70 80 90 100 150 200 300 500 600 700 800 1000
Flow (m
3
/s) RD 0 0 0 0 0 0.01 0.01 0 0.02 0.05 0.06 0.07 0.09 0.11 0.13
MD 0.1 0.1 0.2 0.2 0.3 0.4 0.5 0.4 1.2 3.2 4.6 5.3 6.5 8.1 9
MAD 0.1 0.2 0.2 0.3 0.4 0.4 0.5 0.7 1.4 3.3 4.7 5.4 6.6 8.1 9
RMSD 0.1 0.2 0.3 0.4 0.5 0.6 0.8 0.8 2 4.8 6.8 7.8 9.4 11.3 12.5
P 0.1 0.2 0.2 0.3 0.4 0.5 0.6 0.5 1.5 3.5 5 5.7 6.7 7.8 8.7
Sediment (mlt/month) RD / / / / / / / / / / / / / / /
MD 3 14 23 33 53 65 67 118 150 209 142 182 221 53 121
MAD 4 14 23 33 53 65 67 118 150 209 142 182 221 77 122
RMSD 13 31 48 65 106 126 131 227 288 404 303 380 434 180 278
P 54 217 353 517 825 1002 1039 1820 2321 3244 2195 2827 3428 1006 1881
Dissolved oxygen (t/month) RD / / / / / / / / / / / / / / /
MD 18 16 4 13 12 10 8 28 12 60 396 483 577 1410 1483
MAD 18 17 5 14 14 13 13 31 26 65 396 483 577 1410 1483
RMSD 22 21 7 18 18 17 17 40 35 94 538 636 724 1795 1884
P 4 5 70 13 17 28 41 23 294 966 6135 7482 8940 21,849 22,971
NO
3
eN (t/month) RD 0 0.001 0.003 0.004 0.004 0.005 0.005 0.018 0.011 0.004 0.037 0.007 0.005 0.005 0.051
MD 1 0 0 0 0 0 0 0 3 8 4 12 1 12 19
MAD 1 1 2 2 2 3 3 5 7 13 39 46 46 49 62
RMSD 1 1 3 3 3 5 4 7 10 17 66 73 73 78 90
P 10 12 12 11 16 25 25 39 81 164 337 483 472 403 656
NH
4
eN (t/month) RD 0.06 0.05 0.04 0.05 0.11 0.1 0.09 0.09 0.11 0.28 0.14 0.36 0.29 0.1 0.19
MD 1 1 1 1 2 2 2 2 2 5 4 6 6 4 5
MAD 1 1 1 1 2 2 2 2 2 5 4 6 6 4 5
RMSD 2 1 2 2 4 3 3 3 3 8 7 12 11 8 10
P 18 11 12 13 29 28 23 23 26 70 56 100 91 59 73
TN (t/month) RD 0 0.01 0.01 0.01 0.01 0.01 0.02 0.02 0.05 0.08 0.05 0.08 0.47 0.03 0.46
MD 5 9 10 15 20 22 27 46 73 118 106 136 289 71 253
MAD 5 9 10 16 21 23 28 48 75 121 112 141 291 91 259
RMSD 10 17 21 31 40 45 54 93 144 233 200 248 457 171 428
P 78 135 155 244 314 351 429 728 1144 1854 1692 2151 4490 1256 3965
TP (t/month) RD 0 0.01 0.01 0.02 0.02 0.03 0.03 0.05 0.08 1.3 12.4 11.6 8.3 14.2 12.5
MD 1 2 2 3 4 5 6 10 16 25 22 25 28 14 17
MAD 1 2 2 3 4 5 6 10 16 25 23 26 29 19 22
RMSD 2 4 4 7 9 10 12 21 31 50 48 56 61 45 51
P 15 27 33 54 69 75 93 158 242 386 344 401 445 255 305
w
a
t
e
r
r
e
s
e
a
r
c
h
5
3
(
2
0
1
4
)
1
3
2
e
1
4
4
1
3
8
Fig. 4 e The overall distribution and statistical analysis of ow, sediment, dissolved oxygen, NH
4
eN, NO
3
eN, TN and TP.
wa t e r r e s e a r c h 5 3 ( 2 0 1 4 ) 1 3 2 e1 4 4 139
Fig. 5 e The temporal sensitivity of mean monthly H/NPS modeling to DEM resolution (the modeling results based on
40e1000 m DEM compared with 30 m DEM).
wa t e r r e s e a r c h 5 3 ( 2 0 1 4 ) 1 3 2 e1 4 4 140
highly variable in the wet seasons (Fig. 5 (NH
4
eN)). Thus, the
uncertainties of the NO
3
eN and NH
4
eN loads would greatly
increase in the wet and normal seasons when the DEM reso-
lutions were coarser than 300 m and 70 m, respectively. The
monthly distributions of the TN and TP loads were similar.
The RDs of TN and TP would uctuate greatly when the DEM
resolution was coarser than 150 m and 100 m, respectively.
Uncharacteristically, TP tended to be overestimated from July
to December (for resolutions of 600e1000 m), and TN was
overestimated in February (for resolutions of 500e1000 m).
As illustrated in Fig. 6, the ow was more similar to the
observed data as the DEM resolutions decreased from 30 m to
1000 m for the uncalibrated SWAT. If the ow was the only
indicator of interest, the coarser resolution DEMs could be
sufcient and would save computation time during model
calibration.
4. Discussion
As the DEMresolution decreases, the estimated results tended
to overestimate the minimum altitude and underestimate the
maximum altitude, and a similar pattern was observed for
slope; this may be attributed to the loss of detailed topo-
graphic characteristics (Wechsler, 2007), Gertner et al. (2002)
and Wechsler (2007) also noted that coarser DEM resolution
increased the uncertainties of altitude and slope. Although
Chaplot (2005) concluded that the accuracy of DEMs did not
cause any signicant bias in the estimation of altitude. The
reason may be that in our study: (1) the range of DEM reso-
lutions was larger than Chaplot (2005); (2) used resampled
DEMs; (3) the study area was a mountainous catchment.
Cotter et al. (2002), Di Luzio et al. (2005) and Ghaffari (2011) also
found that the watershed area, reach length, subbasins and
ow decreased with coarser DEM resolutions.
SWAT tended to under-predict the ow peak (Fig. 4);
similar results were found by Kalin et al. (2003) with the Ki-
nematic runoff and erosion (KINEROS) model. The differences
in the ow caused by the changes in DEM resolutions were
more signicant in the dry and normal seasons than in the
wet seasons (Fig. 5). However, the RD of the ow in the wet
season was greater, and coarser DEMs obtained better results
when compared with the observed data (Fig. 6). These results
are different to previous studies, which noted that ow was
not sensitive to DEM resolution without considering the
temporal distribution (Chaplot, 2005; Lin et al., 2013). Di Luzio
et al. (2005) noted that they calculated ow by the SCS curve
number method and that the DEM resolution does not inu-
ence the average curve number value, although it did impact
the watershed area and the retardance coefcient. Considered
with all the topography attributes calculated based on DEM
resolution, only slope was more sensitive to ow simulation
and the relative importance percentage of slope to ow was
only 0.1% of total sensitive parameters (Cibin et al., 2010). So
ow is essentially unaffected by the DEM resolution.
Mamillapalli et al. (1996) found that increasing the number
of HRUs successfully compensated for decreasing the number
of subbasins and that the accuracy of the SWAT ow pre-
dictions varied depending on the number of subbasins and
HRUs used to represent the watershed. However, ow also
decreased with the slightly decreased number of subbasins
and greatly increased number of HRUs when the DEM reso-
lution was coarser than 300 m in this study, and Mamillapalli
et al. (1996) also concluded that the observed decrease in ow
may be due to the decreased watershed area and reach length.
Flow and dissolved oxygen load decreased slightly at
optimal DEM resolutions; however, when the DEM resolution
decreased, the ow and dissolved oxygen load decreased
signicantly, and the uncertainties in the dry season were
greater than in the wet and normal seasons. While for dis-
solved oxygen was very sensitive and inuenced by many
physical factors (such as temperature, altitude, salinity, and
stream structure), so it is difcult to say which resolution for
DEM data was the best. Excluding ow and dissolved oxygen
load, the uncertainties of the other H/NPS indicators were
greater inthe wet seasons than in the dry and normal seasons,
especially when the DEMs were outside of the optimal reso-
lution range.
Sediment yield greatly decreased when the DEM resolution
was coarser than 100 m, and similar results have been
observed in previous studies (Chaplot, 2005; Di Luzio et al.,
2005; Lin et al., 2013; Shen et al., 2013b). Additionally, sedi-
ment yields were highly correlated with ow (Darren et al.,
2013), and the ow changed greatly in the wet seasons; thus,
sediment yield was undoubtedly more sensitive to DEM ac-
curacy, especially in the wet season, and the nonlinear
decrease of sediment yield may be largely due to the nonlinear
change in the number of subbasins and HRUs (Xian et al.,
2007).
NO
3
eN load was insensitive to DEM resolutions from 30 to
300 m, whereas with coarser DEMs, the temporal uncertainty
of NO
3
eN greatly increased, especially in the wet season.
NH
4
eN load declined steadily from 30 to 1000 m, especially
when the DEM resolution was coarser than 70 m. The un-
certainties of both NO
3
eN and NH
4
eN would greatly increase
in the wet and normal seasons when the DEM resolutions
were coarser than 300 mand 70 m, respectively. Chaplot (2005)
found that the estimated NO
3
eN load decreased with coarser
DEM resolutions. The impact of the DEM resolution on the
sediment and nitrogen outputs due to sediment and nitrogen
Fig. 6 e The temporal sensitivity of mean monthly ow
modeling to DEMresolution (the modeling results based on
30e1000 m DEM compared with the observed ow).
wa t e r r e s e a r c h 5 3 ( 2 0 1 4 ) 1 3 2 e1 4 4 141
erosion were estimated from the MUSLE equation involving
the mean slope gradient and length (Williams and Berndt,
1977).
The temporal trends of TN and TP loads were similar, and
the RDs of both loads at 40e100 m were within 10% in
2002e2006. However, the monthly RD of both loads changed
signicantly when the DEM resolutions were coarser than
150 m and 100 m, respectively. When the DEM resolution was
coarser than 500 m, the temporal uncertainty of TP greatly
increased. Both NH
4
eN and TP loads were slightly sensitive in
the annual average modeling results but more sensitive in
their temporal distribution.
The optimal DEM resolutions for the different H/NPS in-
dicators of interest were somehow different from one
another. The optimal thresholds may exist because of the
non-linearity of erosion processes (Chaplot, 2005).
Because the DEMs used in this study were resampled from
30 m resolution DEMs and previous studies have conrmed
that the accuracy of resampled DEMs is slightly lower than the
original DEMs (Wechsler, 2007), the conclusions in this study
should be considered cautiously. Furthermore, the uncali-
brated SWAT prediction results in this study should be
considered in the context of the natural variability of the
environmental variables of the Xiangxi River Watershed
(Garc a-Barcina et al., 2006).
5. Conclusions
Ultimately, an understanding of DEM error and the quanti-
cation of the resulting uncertainty are essential for the
advancement of hydrologic science. The impact of DEM res-
olution on Hydrology/NPS modeling in terms of both predic-
tion accuracy and uncertainty was evaluated in this study.
The watershed area and length decreased as the DEM reso-
lution decreased. When the DEM resolution was 30e300 m,
the numbers of subbasins and HRUs changed little, but when
the resolution was coarser than 300 m, the number of sub-
basins decreased, and the number of HRUs increased, both
with great uncertainties.
In the annual average modeling output, sediment yield was
greatly affected by the DEM resolution; prediction of dissolved
oxygen load was signicantly related to DEM resolutions
coarser than 500 m; TN load was not greatly affected by the
DEM resolutions; NO
3
eN and TP were only slightly affected by
the DEM resolutions; and ow and NH
4
eN were almost un-
affected by the DEM resolutions. While with considered the
uncertainty of the monthly distribution of modeling results,
The optimal DEM resolutions for ow was 30e200 m, for
sediment and TP was 30e100 m, for dissolved oxygen and
NO
3
eN was 30e300 m, for NH
4
eN was 30e70 m and for TN
was 30e150 m. The monthly RD of ow with the optimal DEM
resolution was within 5%, and the monthly RD of sediment,
DO, NO
3
eN, NH
4
eN, TN and TP was within 10%.
To reduce the prediction uncertainty, the DEM should be
studied before modeling to determine which resolution would
perform optimally for the watershed in SWAT. The choice of
the input DEM resolution depends on the watershed response
of interest, process presentation and the specic study area.
These threshold of DEM resolutions selected in this study can
be applied to optimize the input data preparation re-
quirements for SWAT analysis of other mountainous water-
sheds, especially those with a similar landscape process scale
(Vaze and Teng, 2007), or extrapolated to other similar
distributed H/NPS models. All the H/NPS models rely on DEM
data should select the optimal DEM resolution prior to
modeling, whats more, the selection of DEM resolution
should highlighted the uncertainty on H/NPS modeling in wet,
normal and dry seasons.
Acknowledgments
This study was funded by the National Natural Science
Foundation of China (Grant No. 41001352) and the Funda-
mental Research Funds for Central Universities.
r e f e r e n c e s
Abbaspour, K., 2013. User Manual for SWAT-CUP. http://www.
eawag.ch/forschung/siam/software/swat/index.
Albani, M., Klinkenberg, B., Andison, D.W., Kimmins, J.P., 2004.
The choice of window size in approximating topographic
surfaces from Digital Elevation Models. Int. J. Geogr.
Information Sci. 18, 577e593.
Arabi, M., Govindaraju, R.S., Hantush, M.M., Engel, B.A., 2006. Role
of waterhsed subdivision on modeling the effectiveness of
best management practices with SWAT. J. Am. Water Resour.
Assoc. 42, 513e528.
Arnold, J.G., Kiniry, J.R., Srinivasan, R., Williams, J.R., Haney, E.B.,
Neitsch, S.L., 2013. SWAT2012 input/output le
documentation, Texas Water Resource Institute TR-439.
Available at: http://swat.tamu.edu/media/69296/SWAT-IO-
Documentation-2012.pdf.
Arnold, J.G., Srinivasan, R., Muttiah, R.S., Williams, J.R., 1998.
Large area hydrologic modeling and assessment e part 1:
model development. J. Am. Water Resour. Assoc. 34, 73e89.
Band, L., Moore, I., 1995. Scale: landscape attributes and
geographical information systems. Hydrol. Process. 9,
401e422.
Booij, M.J., 2005. Impact of climate change on river ooding
assessed with different spatial model resolutions. J. Hydrology
303, 176e198.
Bormann, H., 2006. Impact of spatial data resolution on simulated
catchment water balances and model performance of the
multi-scale TOPLATS model. Hydrology Earth Syst. Sci. 10,
165e179.
Bormann, H., 2008. Sensitivity of a soil-vegetation-atmosphere-
transfer scheme to input data resolution and data
classication. J. Hydrology 351, 154e169.
Bormann, H., Breuer, L., Graff, T., Huisman, J.A., Croke, B., 2009.
Assessing the impact of land use change on hydrology by
ensemble modelling (LUCHEM) IV: model sensitivity to data
aggregation and spatial (re-)distribution. Adv. Water Resour.
32, 171e192.
Brown, D.G., Bian, L., Walsh, S.J., 1993. Response of a distributed
watershed erosion model to variations in input data
aggregation levels. Comput. Geosciences 19, 499e509.
Bruneau, P., Gascuel-Odoux, C., Robin, P., Merot, P., Beven, K.,
1995. Sensitivity to space and time resolution of a hydrological
model using digital elevation data. Hydrol. Process. 9, 69e81.
Casper, A.F., Dixon, B., Earls, J., Gore, J.A., 2011. Linking a spatially
explicit watershed model (SWAT) with an in-stream sh
wa t e r r e s e a r c h 5 3 ( 2 0 1 4 ) 1 3 2 e1 4 4 142
habitat model (PHABSIM): a case study of setting minimum
ows and levels in a low gradient, sub-tropical river. River Res.
Appl. 27, 269e282.
Chaplot, V., 2005. Impact of DEM mesh size and soil map scale on
SWAT runoff, sediment, and NO
3
CN loads predictions. J.
Hydrology 312, 207e222.
Chaubey, I., Cotter, A.S., Costello, T.A., Soerens, T.S., 2005. Effect
of DEM data resolution on SWAT output uncertainty. Hydrol.
Process. 19, 621e628.
Chen, D., Lu, J., Shen, Y., Gong, D., Deng, O., 2011. Spatio-temporal
variations of nitrogen in an agricultural watershed in eastern
China: catchment export, stream attenuation and discharge.
Environ. Pollut. 159, 2989e2995.
Cho, J., Bosch, D., Lowrance, R., Strickland, T., Vellidis, G., 2009.
Effect of spatial distribution of rainfall on temporal and spatial
uncertainty of SWAT output. Transactions ASABE 52,
1545e1555.
Cibin, R., Sudheer, K.P., Chaubey, I., 2010. Sensitivity and
identiability of streamow generation parameters of the
SWAT model. Hydrol. Process. 24, 1133e1148.
Cotter, A.S., Chaubey, I., Costello, T.A., Nelson, M.A., Soerens, T.,
2002. Effect of spatial data resolution on SWAT output
uncertainty. Paper No. 02-2012. In: 2002 ASAE Annual
International Meeting, Chicago, Illinois, USA. July 28e31 2002.
ASAE, 2950 Niles Rd, St Joseph, MI, USA.
Cotter, A.S., Chaubey, I., Costello, T.A., Soerens, T.S.,
Nelson, M.A., 2003. Water quality model output uncertainty as
affected by spatial resolution of input data. J. Am. Water
Resour. Assoc. 39, 977e986.
Darren, L.F., Luo, Y., Zhang, M., 2013. Watershed modelling of
hydrology and water quality in the Sacramento River
watershed, California. Hydrol. Process. 27, 236e250.
Di Luzio, M., Arnold, J.G., Srinivasan, R., 2005. Effect of GIS data
quality on small watershed streamow and sediment
simulations. Hydrol. Process. 19, 629e650.
Dixon, B., Earls, J., 2009. Resample or not?! Effects of resolution of
DEMs in watershed modeling. Hydrol. Process. 23, 1714e1724.
Dixon, B., Earls, J., 2012. Effects of urbanization on streamow
using SWAT with real and simulated meteorological data.
Appl. Geogr. 35, 174e190.
Earls, J., Dixon, B., 2005. A comparative study of the effects of
input resolution on the SWAT model. River Basin Manag. III.
83, 213e222.
Endreny, T.A., Wood, E.F., Lettenmaier, D.P., 2000. Satellite-
derived digital elevation model accuracy: hydrological
modelling requirements. Hydrol. Process. 14, 177e194.
Faramarzi, M., Abbaspour, K.C., Ashraf Vaghe, S.,
Farzaneh, M.R., Zehnder, A.J.B., Srinivasan, R., Yang, H., 2013.
Modeling impacts of climate change on freshwater availability
in Africa. J. Hydrology 480, 85e101.
Garca-Barcina, J.M., Gonz lez-Oreja, J.A., De la Sota, A., 2006.
Assessing the improvement of the Bilbao estuary water
quality in response to pollution abatement measures. Water
Res. 40, 951e960.
Gericke, A., Venohr, M., 2012. Improving the estimation of
erosion-related suspended solid yields in mountainous, non-
alpine river catchments. Environ. Model. Softw. 37, 30e40.
Gertner, G., Wang, G., Fang, S., Anderson, A.B., 2002. Effect and
uncertainty of digital elevation model spatial resolutions on
predicting the topographical factor for soil loss estimation. J.
Soil. Water Conservation 57, 164e174.
Ghaffari, G., 2011. The impact of DEM resolution on runoff and
sediment modelling results. Res. J. Environ. Sci. 5, 691e702.
Jha, M., Gassman, P.W., Secchi, S., Gu, R., Arnold, J., 2004. Effect of
watershed subdivision on swat ow, sediment, and nutrient
predictions. J. Am. Water Resour. Assoc. 40, 811e825.
Kalin, L., Govindaraju, R.S., Hantush, M.M., 2003. Effect of
geomorphologic resolution on modeling of runoff hydrograph
and sedimentograph over small watersheds. J. Hydrology 276,
89e111.
Katrin, B., Georg, H.R., Nicola, F., 2011. Simulation of streamow
and sediment with the soil and water assessment tool in a
data scarce catchment in the three Gorges Region, China. J.
Environ. Qual., 37e45. http://dx.doi.org/10.2134/jeq2011.0383.
Kim, J., Noh, J., Son, K., Kim, I., 2012. Impacts of GIS data quality
on determination of runoff and suspended sediments in the
Imha watershed in Korea. Geosciences J. 16, 181e192.
Lai, Y.C., Yang, C.P., Hsieh, C.Y., Wu, C.Y., Kao, C.M., 2011.
Evaluation of non-point source pollution and river water
quality using a multimedia two-model system. J. Hydrology
409, 583e595.
Lee, G., Tachikawa, Y., Takara, K., 2009. Interaction between
topographic and process parameters due to the spatial
resolution of DEMs in distributed rainfall-runoff modeling. J.
Hydrol. Eng. 14, 1059e1069.
Li, Z., Shao, Q., Xu, Z., Cai, X., 2010. Analysis of parameter
uncertainty in semi-distributed hydrological models using
bootstrap method: a case study of SWAT model applied to
Yingluoxia watershed in northwest China. J. Hydrology 385,
76e83.
Lin, K.R., Zhang, Q.A., Chen, X.H., 2010. An evaluation of impacts
of DEM resolution and parameter correlation on TOPMODEL
modeling uncertainty. J. Hydrology 394, 370e383.
Lin, S.P., Jing, C.W., Coles, N.A., Chaplot, V., Moore, N.J., Wu, J.P.,
2013. Evaluating DEM source and resolution uncertainties in
the soil and water assessment tool. Stoch. Environ. Res. Risk
Assess. 27, 209e221.
Liu, R.M., Zhang, P.P., Wang, X.J., Chen, Y.X., Shen, Z.Y., 2013.
Assessment of effects of best management practices on
agricultural non-point source pollution in Xiangxi River
watershed. Agric. Water Manag. 117, 9e18.
Mamillapalli, S., Srinivasan, R., Arnold, J.G., Engel, B.A., 1996.
Effect of spatial variability on basin scale modeling. In: Proc.
Third International NCGIA Conference/Workshop on
Integrating GIS and Environmental Modeling, Sante Fe, New
Mexico, January, pp. 21e25.
McElroy, A.D., Chiu, S.Y., Nebgen, J.W., Aleti, A., Bennett, F.W.,
1976. Loading Functions for Assessment of Water Pollution
from Nonpoint Sources. EPA document EPA 600/2-76-151.
USEPA, Athens, GA.
Moore, I.D., Lewis, A., Gallant, J.C., 1993. Terrain attributes:
estimation methods and scale effects. In: Jakeman, A.J.,
Beck, M.B., McAleer, M. (Eds.), Modelling Change in
Environmental Systems. Wiley, New York, pp. 189e214.
Ndomba, P., Birhanu, B., 2008. Problems and prospects of SWAT
model applications in Nilotic catchments: a review. Nile Basin
Water Eng. Sci. Mag. 1, 41e52.
Neitsch, S.L., Arnold, J.G., Kiniry, J.R., Williams, J.R., 2005. Soil and
Water Assessment Tool e Theoretical Documentation,
Version 2005 (Texas, USA).
Ongley, E.D., Zhang, X.L., Yu, T., 2010. Current status of
agricultural and rural non-point source pollution assessment
in China. Environ. Pollut. 158, 1159e1168.
Ouyang, Y., Nkedi-Kizza, P., Wu, Q.T., Shinde, D., Huang, C.H.,
2006. Assessment of seasonal variations in surface water
quality. Water Res. 40, 3800e3810.
Quinn, P., Beven, K., Chevallier, P., Planchon, O., 1991. The
prediction of hillslope ow paths for distributed hydrological
modelling using digital terrain models. Hydrol. Process. 5,
59e79.
Sharma, A., Tiwari, K.N., Bhadoria, P., 2009. Measuring the
accuracy of Contour Interpolated Digital Elevation Models. J.
Indian Soc. Remote Sens. 37, 139e146.
Sharma, A., Tiwari, K.N., Bhadoria, P., 2011. Determining the
optimum cell size of digital elevation model for hydrologic
application. J. Earth Syst. Sci. 120, 573e582.
wa t e r r e s e a r c h 5 3 ( 2 0 1 4 ) 1 3 2 e1 4 4 143
Shen, Z.Y., Chen, L., Chen, T., 2013a. The inuence of parameter
distribution uncertainty on hydrological and sediment
modeling: a case study of SWAT model applied to the Daning
watershed of the Three Gorges Reservoir Region, China. Stoch.
Environ. Res. Risk Assess. 27, 235e251.
Shen, Z.Y., Chen, L., Liao, Q., Liu, R.M., Huang, Q., 2013b. A
comprehensive study of the effect of GIS data on hydrology
and non-point source pollution modeling. Agric. Water
Manag. 118, 93e102.
Shen, Z.Y., Chen, L., Liao, Q., Liu, R.M., Hong, Q., 2012a. Impact of
spatial rainfall variability on hydrology and nonpoint source
pollution modeling. J. Hydrology 472e473, 205e215.
Shen, Z.Y., Liao, Q., Hong, Q., Gong, Y., 2012b. An overview of
research on agricultural non-point source pollution modelling
in China. Sep. Purif. Technol. 84, 104e111.
Song, X., Qi, Z., Du, L.P., Kou, C.L., 2012. The inuence of DEM
resolution on hydrological simulation in the Huangshui River
Basin. Adv. Mater. Res. 518e523, 4299e4302. Stafa-Zurich,
Trans Tech Publications Ltd.
Thieken, A.H., Lucke, A., Diekkruger, B., Richter, O., 1999. Scaling
input data by GIS for hydrological modelling. Hydrol. Process.
13, 611e630.
USDA-SCS (U.S. Department of Agriculture-Soil Conservation
Service), 1972. National Engineering Handbook Section 4
Hydrology. U.S. Department of Agriculture-Soil Conservation
Service, Washington, D.C.
Vaze, J., Teng, J., December 2007. Impact of DEM Resolution on
Topographic Indices and Hydrological Modelling Results.
MODSIM 2007 International Congress on Modelling and
Simulation. Modelling and Simulation Society of Australia and
New Zealand.
Wang, X.J., Liu, R.M., Gong, Y.W., 2011. Simulation of the effect of
land use/cover change on non-point source pollution load in
Xiangxi River watershed. Chin. J. Environ. Eng. 5, 1194e1200.
Wang, X.Y., Lin, Q., 2011. Effect of DEM mesh size on AnnAGNPS
simulation and slope correction. Environ. Monit. Assess. 179,
267e277.
Watson, F., Grayson, R.B., Vertessy, R.A., McMahon, T.A., 1998.
Large-scale distribution modelling and the utility of detailed
ground data. Hydrol. Process. 12, 873e888.
Wechsler, S.P., 2007. Uncertainties associated with digital
elevation models for hydrologic applications: a review.
Hydrological Earth Systerm Sci. 11, 1481e1500.
Williams, J.R., Berndt, H.D., 1977. Sediment yield prediction based
on watershed hydrology. Transactions Am. Soc. Agric. Eng. 20,
1100e1104.
Williams, J.R., Hann, R.W., 1978. Optimal Operation of Large
Agricultural Watersheds with Water Quality Constraints.
Texas Water Resources Institute, Texas A&M Univ.. Tech.
Rept. No. 96.
Wolock, D.M., Price, C.V., 1994. Effects of digital elevation model
map scale and data resolution on a topography-based
watershed model. Water Resour. Res. 30, 3041e3052.
Wu, L., Long, T.Y., Liu, X., Guo, J.S., 2012. Impacts of climate and
land-use changes on the migration of non-point source
nitrogen and phosphorus during rainfall-runoff in the Jialing
River Watershed, China. J. Hydrology 475, 26e41.
Xian, G., Crane, M., Su, J., 2007. An analysis of urban development
and its environmental impact on the Tampa Bay watershed. J.
Environ. Manage. 85, 965e976.
Xu, H., Taylor, R.G., Kingston, D.G., Jiang, T., Thompson, J.R.,
Todd, M.C., 2010. Hydrological modeling of River Xiangxi using
SWAT2005: a comparison of model parameterizations using
station and gridded meteorological observations. Quat. Int.
226, 54e59.
Xu, J., Ren, L.L., Yuan, F., Liu, X.F., 2011. Development and
Application of a Scaling Model for Characterising Topographic
Features in Hydrological Simulation. In: INT ASSOC
Hydrological Sciences, vol. 350. IAHS Publication, Wallingford,
pp. 766e772.
Yuan, D., Lin, B., Falconer, R.A., Tao, J., 2007. Development of an
integrated model for assessing the impact of diffuse and point
source pollution on coastal waters. Environ. Model. Softw. 22,
871e879.
Zhang, J., Shen, T., Liu, M., Wan, Y., Liu, J., Li, J., 2011. Research on
non-point source pollution spatial distribution of Qingdao
based on L-THIA model. Math. Comput. Model. 54, 1151e1159.
Zhang, W.H., Montgomery, D.R., 1994. Digital elevation model grid
size, landscape representation, and hydrologic simulations.
Water Resour. Res. 30, 1019e1028.
wa t e r r e s e a r c h 5 3 ( 2 0 1 4 ) 1 3 2 e1 4 4 144

You might also like