An Automated Approach for Detection of Shallow Landslides from LiDAR Derived DEM Using Geomorphological Indicators in a Tropical Forest

Abstract

Landslide inventories in the tropical dense forested areas are routinely compiled by means of a terrain model interpretation (e.g. using stereo-radargrammetry; stereo-aerial photographs; stereo-optical imagery), aided with field investigations. However, construction of the landslide inventories from aerial photographs and field based studies are excessively time consuming which involves relatively high cost. Moreover, these techniques are less effective when applied to dense tropical forest where landslide scars are difficult to map from the aerial photographs. This chapter attempts an automatic procedure for detection of rotational shallow landslides from airborne based light detection and ranging (LiDAR) derived high resolution digital elevation model (DEM) in a tropical forest in Cameron Highlands, Malaysia. For the extraction of landslides from DEM, we used various geomorphic indicators such as surface roughness index, vegetation index and breaklines. The entire landslide extraction process was implemented in ArcGIS platform and custom Python scripts was used for the implementation and model construction. For modeling purpose, the Python Imaging Library (PIL) was used. The terrain zone classification was tested for various DEM resolutions of 1.5 m, 2 m, 3 m, 4 m, 5 m and 8 m. For testing purposes, the resolutions with the best results were used for further processing. To automate the classification of the terrain zones, a rule based region growing threshold was defined depending on the resolution of the DEM. Finally, a statistical description was applied to rank the extracted terrain zones according to their compliance with the landslide signature. Subsequently, the landslide probability index (LPI) was calculated by performing zonal operation using each of the geomorphic parameters. Hence, the LIDAR-derived DEM provides adequate landslide factor maps to identify the landslide occurred areas, which could be used for further landslide assessment and site-planning purposes in the tropical regions.