PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
July 2014
639
Sensitivity of Hydrological Outputs from
SWAT to DEM Spatial Resolution
T. Goulden, R. Jamieson, C. Hopkinson, S. Sterling, A. Sinclair, and D. Hebb
Abstract
This research analyzes the sensitivity of the
SWAT
(Soil and
Water Assessment Tool) flow rate and sediment yield to
DEM
spatial resolution. A scaling analysis is performed using
1 to 50 m resolution lidar (Light Detection and Ranging)
derived
DEM
s. Thomas Brook, Canada, serves as a case study
watershed for a five year simulation period. Results indicated
monthly water yield was insensitive to
DEM
resolution, unless a
change in area was also present. Sediment yield from the 50 m
DEM
showed a 200 t reduction compared to the 1 m
DEM
. The
5 to 50 m
DEM
s also showed a reduction in channel deposition
of 45 to 90 t, respectively, compared to the 1 m
DEM
. Results
suggest parameterization using multiple resolution
DEM
s is
desirable, one commensurate with the scale used to establish
universal soil loss equations for defining hillslope scale, and
high resolution for stream network and sub-basin delineation.
Introduction
Hydrologic and hydraulic models are tools often used to
assist in planning and assessing sustainable environmental
strategies in response to ecosystem impacts such as agricul-
tural practices, land use development, or climate change. An
accurate digital landscape representation is a fundamental
requirement for hydrologic simulations because it controls
the transfer of water, sediment, nutrients, and pollutants
within the modeled environment (Moore
et al
., 1991). Digital
information on landscape shape and structure is provided to a
hydrological model through a
DEM
(Digital Elevation Model),
commonly as a regularly spaced horizontal grid of elevation
values (Collins and Moon, 1981; Moore, 1991). In applying
a
DEM
for analyzing hydrological phenomena, an important
decision must be made on the horizontal length of a
DEM
grid
cell (defined here as the spatial resolution) which corre-
sponds to the underlying spatial scale (Hutchinson and Gal-
lant, 1999). Scale driven investigations conducted under the
assumptions and constraints of a particular model structure
can guide users in selecting an appropriate
DEM
resolution.
Ideally, the choice of
DEM
resolution will simultaneously gen-
erate accurate simulated outputs, as well as a realistic model
parameterization. Such a model will be capable of generating
sound and defensible evidence for development of water
resource planning strategies.
Previous research has demonstrated that attention must be
given to the selection of
DEM
resolution because of the effects
to hydrologic outputs as well as watershed and topographic
attributes. Average terrain slope, for example, tends to in-
crease as spatial resolution increases (Evans, 1979; Chang and
Tsai, 1991; Zhang and Montgomery, 1994; Wilson
et al
., 2000;
Kienzle, 2004; Hill and Neary, 2005; Deng
et al
., 2007; Hop-
kinson
et al
., 2010). Stream lengths have shown to increase
as
DEM
resolution increases (Wang and Yin, 1998; Thieken
et
al
., 1999; Goulden
et al
., in press) while watershed area has
shown inconsistent behavior with changes in
DEM
resolution
(Cotter
et al
., 2003; Chaubey
et al
., 2005; Goulden
et al
., in
press). Peak flow rates have shown to decrease as resolution
becomes increasingly fine (Quinn et al., 1991; Zhang and
Montgomery, 1994) while the timing of peak flow tends to be
delayed (Thieken
et al
., 1999). Simulated sediment loads have
been shown to both increase (Chaplot, 2005; Cotter
et al
.,
2003; Di Luzio
et al
., 2005) and decrease (Zhao
et al
., 2010) as
DEM
resolution becomes increasingly fine. Despite the well-es-
tablished sensitivity of topographic attributes, watershed
attributes, and hydrologic outputs with
DEM
resolution, no
consistent guidelines are available for selecting a
DEM
spatial
resolution for simulated hydrological analysis.
Previous suggestions for selecting a
DEM
resolution to
represent hillslope hydrologic processes indicate that the
DEM
resolution should reflect the natural scale of the hillslope
(Quinn
et al
., 1991; Zhang and Montgomery, 1994; Beven,
1997; Hutchinson and Gallant, 2000; McMaster, 2002). Where
“hillslope hydrology” includes hydrologic processes dealing
with vegetation, overland flow, and subsurface flow (Kirkby,
1988). In cases where the
DEM
is at a resolution lower than the
natural hillslope scale, valley or hill structures can be lost and
T. Goulden is with National Ecological Observatory Network
(NEON), Inc, 1685, 38th St, Suite 100, Boulder, CO, 80301;
and formerly with Process Engineering and Applied Science,
Dalhousie University, Halifax, Canada, B3J 1Z1
neoninc.org).
R. Jamieson, and A. Sinclair Process are with Process Engi-
neering and Applied Science, Dalhousie University, 1360
Barrington Street, Halifax, Canada, B3J 1Z1.
C. Hopkinson is with Engineering and Applied Science, Dal-
housie University, 1360 Barrington Street, Halifax, Canada,
B3J 1Z1; and the Department of Geography, University of Le-
thbridge, 4401 University Drive, Lethbridge, Alberta, Canada,
T1K 3M4.
S. Sterling is with Earth Science and Environmental Science,
Dalhousie University, 1355 Oxford Street, Halifax, Canada,
B3H 4R2.
D. Hebb is with the Atlantic Food and Horticulture Research
Centre, Agriculture and Agri-Food Canada, 32 Main Street,
Kentville, Nova Scotia, Canada, B4N 1J5.
Photogrammetric Engineering & Remote Sensing
Vol. 80, No. 7, July 2014, pp. 000–000.
0099-1112/14/8007–000
© 2014 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.80.7.000
