The Effects of Different Resolution DEMs

1
The Effects of Different Resolution DEMs in
Determining Overland Flow Regimes Stacy L.
Hutchinson1, J.M. Shawn Hutchinson2, Ik-Jae Kim1,
and Philip Woodford31Department of Biological
and Agricultural Engineering and 2Department of
Geography, Kansas State University, Manhattan,
Kansas 3Integrated Training Area Management, Fort
Riley, Kansas
Abstract A gully head is a unique landscape
feature where concentrated overland flow begins
to cause significant erosion. The impacts of four
different resolution digital elevation models
(DEMs), three (3, 10, and 30 m) developed using a
differential global positioning system (GPS)
survey and the USGS 30 m DEM, were used to
identify transitional flow areas on a grassland
hillslope. A simple erosion model, nLS, based on
Mannings kinematic wave theory, was used to
determine where overland flow transitioned from
sheet flow to concentrated flow. The accumulated
erosive energy was estimated using the nLS model,
where, n is Mannings coefficient, L is the
overland flow length, and S is the slope. In
addition to the DEMs, spatial analyses for soil
(SSURGO) and land cover (Kansas GAP) were
conducted in a geographic information system
(GIS). First order streams were delineated using
each resolution DEM (contributing area 900 m2)
and overlaid with the concentrated flow data
obtained from the nLS model results. The
intersected area was buffered by 3, 6, 10, or 15
m, depending on the DEM resolution. Results
showed that average topographic and hydrologic
variables varied between the different DEM
resolutions. The 3 m DEM produced the best model
accuracy, predicting two gully head locations.
The recommended buffer radius was found to be 6
m, which is 2 times of the grid size. The efforts
to develop finer data resolution should be
supported in assessing reliable erosion potential
for watershed management.
Figure 4. RMSE of random sample points (5 ) for
different resolution DEMs using IDW and TIN
interpolation methods. Results show that 10 m
DEMs offer more reliable prediction for
hydrologic modeling than 30 m DEMs. However, the
errors in 10 m DEMs were 2.44 (IDW) or 2.55 (TIN)
times greater than in 3 m DEMs. No difference
was seen between interpolation techniques.
Figure 2. The study site, 8 ha grass-hillslope
area on Fort Riley with GPS points (n gt 15,000)
and three gully head locations (red). Two areas
on the upslope and one at the downslope were
excluded from the survey to avoid field
experiments and dense vegetation. Three gully
locations (GH-1, 2, and 3) were surveyed.


Concentrated Flow A W x L x ß ß A
Ineffective Area
Uniform Sheet Flow A W x L
L
W
Figure 5. Accumulated overland flow energy
calculated using nLS with different resolution
DEMs (left). Potential erosion areas based on 1st
order stream networks and acumulated overland
flow energy (µ1.0s of nLS) with buffering using
different resolution DEMs (right).
A gully erosion area (GH-3) on the study site in
figure 2. (Taken on Oct. 15, 2005 by IJ Kim)
Data and Methods Three different resolution (3,
10, and 30 m) digital elevation models (DEMs)
were developed using a differential GPS with
post-processing. Two interpolation techniques
(inverse distance weighted, (IDW) and
triangulated irregular network, (TIN)) were used
for converting from GPS point data to raster
format files for each resolution. The vertical
and horizontal errors were assessed using root
mean square error (RMSE) on 5 randomly selected
points and two benchmarks. The USGS National
Elevation Dataset (NED) 30m DEM was also used to
develop accumulated erosive energy layers within
ESRI ArcGIS using the nLS model. Input variables
(slope and flow length) were calculated using the
deterministic eight direction method (D8). Flow
direction was estimated without the pit removal
process. This artificial process may alter the
effect of accumulating overland flow energy for
identifying the flow transition from sheet to
concentrated flow. Flow length for each cell was
determined by multiplying the flow accumulation
values by the DEM resolution. Kansas GAP
landcover data were used to create Mannings
coefficient (n) data layers. From the
information, a continuous energy accumulation
grid was developed using the equation (1). The
statistical interval (µ1.0s, in English Unit) of
mean (µ, 131) and standard deviation (s, 22.6)
was applied to reclassify the transitional areas
(i.e., gully head locations). It was assumed that
gully heads are formed along the 1st order
streams. The selected 1st order stream segments
were buffered, using 3 m and 6 m (in radius) for
the 3 m DEMs, 10 m for the 10 m DEMs, and 15 m
for the 30 m DEMs to account for data resolution
errors in identifying gully head points.
Conclusions and Future Work Finer resolution DEMs
provided more detailed land analysis than coarser
resolution DEMs. Accurate slope estimation was
very important in this study because the lengths
of gully head from the upslope are generally less
than the lengths of the hillslope. Errors in
slope estimates significantly affected the model
performance when generating the flow direction
and flow accumulation grids. Current results
suggest that 3 m or finer DEMs should be used to
determine the geographic locations of gully heads
in the model. Further analysis using a large area
and or complete watershed is needed to
investigate the impact of study area size and
different land covers. The influence of the pit
removal process should be evaluated.
Additionally, the effect of varying contributing
areas for delineating the 1st order stream
network using the same fine resolution DEM should
be analyzed. Key references McCuen, R.H. and J.M.
Spiess. 1995. Assessment of kinematic wave time
of concentration. Journal of Hydrologic
Engineering 121 (3)256-266. Meyer, A. and J.A.
Martinez-Casasnovas. 1999. Prediction of existing
gully erosion in vineyard parcels of the NE
Spain a logistic modeling approach. Soil
Tillage Research 50 319-331. Acknowledgements Thi
s work is funded through CPSON-03-02
(Characterizing and Monitoring Non-Point Source
Runoff from Military Ranges and Identifying their
Impacts to Receiving Water Bodies).
Figure 3. A methods data flow diagram for
determining transitional erosion areas in a GIS.
(C.A. means the contributing area for delineating
1st order stream networks using the flow
accumulation grids).
Table 1. Root mean square errors (RMSE) for the
GPS survey at the two temporary benchmark points
(BM1 east and BM2 west in figure 2).
Equation (1)

• n Mannings coefficient
• L Flow length (m)
• S Slope (m/m)