TY - JOUR
T1 - Landslide-susceptibility analysis using light detection and ranging-derived digital elevation models and logistic regression models
T2 - A case study in Mizunami City, Japan
AU - Wang, Liang Jie
AU - Sawada, Kazuhide
AU - Moriguchi, Shuji
N1 - Funding Information:
The authors gratefully acknowledge the support for this research from Young Researchers Education Center for Innovation (Grant No. 50H2300031) and Gifu Construction Research Center.
PY - 2013/1
Y1 - 2013/1
N2 - To mitigate the damage caused by landslide disasters, different mathematical models have been applied to predict landslide spatial distribution characteristics. Although some researchers have achieved excellent results around the world, few studies take the spatial resolution of the database into account. Four types of digital elevation model (DEM) ranging from 2 to 20 m derived from light detection and ranging technology to analyze landslide susceptibility in Mizunami City, Gifu Prefecture, Japan, are presented. Fifteen landslide-causative factors are considered using a logistic-regression approach to create models for landslide potential analysis. Pre-existing landslide bodies are used to evaluate the performance of the four models. The results revealed that the 20-m model had the highest classification accuracy (71.9%), whereas the 2-m model had the lowest value (68.7%). In the 2-m model, 89.4% of the landslide bodies fit in the medium to very high categories. For the 20-m model, only 83.3% of the landslide bodies were concentrated in the medium to very high classes. When the cell size decreases from 20 to 2 m, the area under the relative operative characteristic increases from 0.68 to 0.77. Therefore, higher-resolution DEMs would provide better results for landslidesusceptibility mapping.
AB - To mitigate the damage caused by landslide disasters, different mathematical models have been applied to predict landslide spatial distribution characteristics. Although some researchers have achieved excellent results around the world, few studies take the spatial resolution of the database into account. Four types of digital elevation model (DEM) ranging from 2 to 20 m derived from light detection and ranging technology to analyze landslide susceptibility in Mizunami City, Gifu Prefecture, Japan, are presented. Fifteen landslide-causative factors are considered using a logistic-regression approach to create models for landslide potential analysis. Pre-existing landslide bodies are used to evaluate the performance of the four models. The results revealed that the 20-m model had the highest classification accuracy (71.9%), whereas the 2-m model had the lowest value (68.7%). In the 2-m model, 89.4% of the landslide bodies fit in the medium to very high categories. For the 20-m model, only 83.3% of the landslide bodies were concentrated in the medium to very high classes. When the cell size decreases from 20 to 2 m, the area under the relative operative characteristic increases from 0.68 to 0.77. Therefore, higher-resolution DEMs would provide better results for landslidesusceptibility mapping.
KW - digital elevation model
KW - landslide-susceptibility analysis
KW - logistic regression model
UR - http://www.scopus.com/inward/record.url?scp=84905472411&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84905472411&partnerID=8YFLogxK
U2 - 10.1117/1.JRS.7.073561
DO - 10.1117/1.JRS.7.073561
M3 - Article
AN - SCOPUS:84905472411
VL - 7
JO - Journal of Applied Remote Sensing
JF - Journal of Applied Remote Sensing
SN - 1931-3195
IS - 1
M1 - 12236
ER -