studies, potentially due to their impact being less severe com-
pared to large landslides. Furthermore, the spatial resolution
needs to be relevant to the scale of the morphological features
of the landslides in order to understand the spatial and tem-
poral process evident in small landslide morphology (Glenn
et al.
, 2006). To our best knowledge, small landslide suscepti-
bility mapping has not been addressed in the literature and an
evaluation is necessary to understand and propose a means of
hazard assessment for the prevention of future events.
This paper presents a novel approach for small landslide
susceptibility mapping utilizing an airborne lidar-derived
Digital Elevation Model (
DEM
). The approach employs several
geomorphologic features to analyze the local topography, spe-
cifically: the direction cosine eigenvalue ratios (
λ
1
/
λ
2
and
λ
1
/
λ
3
), resultant length of orientation vectors, aspect, roughness,
hillshade, slope, a customized Sobel operator, and soil type.
A sample set extracted from the data is used as observations
of landslide and stable terrain to calibrate the supervised
classification algorithm of Support Vector Machine (
SVM
).
The calibrated
SVM
model is subsequently used to classify
the lidar-derived
DEM
based on the extracted surface fea-
tures. Then, as a post-classification step, flat terrain is filtered
and classified as stable terrain. Consequently, a conditional
dilation/erosion filter is applied to minimize misclassified
locations by the
SVM
algorithm, in addition to suppressing
noise and generating landslide susceptible regions (clusters).
Landslide susceptible regions are then analyzed to map areas
of potential landslide activity. Finally, in order to evaluate the
performance of our proposed approach, we assess how well
the algorithms mapped landslides match the reference inven-
tory mapped landslides.
Study Area and Data
Study Area
The study area selected was along the transportation corridor
of state route (
SR
) 666 in Zanesville, Ohio, located in north-
central Muskingum County (Approx. Latitude: N39° 58' 00",
Longitude: W81° 59' 00") along the east side of the Musking-
um River. The study area begins at the intersection of
SR
-60
within the City of Zanesville just north of Interstate 70 (I-70)
and south of the Muskingum River at mile marker (MM) 0.00,
and ends at the intersection with
SR
-208 east of the Village
of Dresden at MM 14.34 (23 km). The extent of the project
coverage is 23 kilometers in length along
SR
-666 with a vary-
ing width of 75 to 180 meters and approximately 3.0 km
2
.
The area is characterized by high vegetation densities, stream
and river channeling, and some residential development. The
study area was chosen due to the availability of an airborne
lidar-derived
DEM
, a detailed landslide inventory map, and
its prolonged history of slope instabilities, especially in areas
where the river is close to the roadway. In 2004 and 2005,
Muskingum County was declared a National Disaster Area
due to extensive flooding in both tributaries and the main
river channel. Along the road seven separate sections dam-
aged by landslides were corrected as a result of these storm
events. Figure 1 presents an overview map of the study area.
Data
The lidar data was acquired in the spring of 2012 and has a
point density of 5 pts/m
2
. The vertical accuracy of the points
was assessed after the lidar was adjusted to the hard surface
control. The vertical accuracy of the points was assessed by
the root mean square error (
RMSE
), which was 9 cm for soft
surfaces and 5 cm for hard surfaces. Additionally, the vertical
accuracy was evaluated by the standard deviation, which was
6 cm and 5 cm for soft and hard surfaces, respectively. The
bare earth, filtered from the lidar data, was subsequently used
for this investigation. The lidar point cloud was bare earth fil-
tered, and then was interpolated to a spatial resolution of 50
cm using Kriging, after evaluating the nominal point spacing
to be 45 cm. The statistical results demonstrated that Kriging
provided the minimum error between the interpolated surface
(
DEM
) and the bare-earth filtered lidar point cloud. For this
reason, it was selected as the prime interpolation method.
The preprocessing of the bare earth lidar data, including con-
verting to a regular grid, was done using LAStools (Isenburg,
2013) and Arc
GIS
®
, respectively. All sequential processing was
performed in
MATLAB
. The results of the processing steps were
integrated into the project
GIS
database.
For the project area a geo-hazard inventory and evaluation
of mass movement affecting the transportation network was
completed in 2006 by the Ohio Department of Transportation
Figure 1. Study area along transportation corridor SR-666, north of Zanesville, Ohio. The figures on the right display examples of rota-
tional slides affecting the embankment, near mile marker 7.3 (top), and 1.47 (bottom) that have since been stabilized.
240
March 2015
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
