The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Download
Publications Copernicus
Download
Citation
Articles | Volume XLII-4/W9
https://doi.org/10.5194/isprs-archives-XLII-4-W9-201-2018
https://doi.org/10.5194/isprs-archives-XLII-4-W9-201-2018
30 Oct 2018
 | 30 Oct 2018

GEOSPATIAL APPROACH FOR LANDSLIDE ACTIVITY ASSESSMENT AND MAPPING BASED ON VEGETATION ANOMALIES

M. R. Mohd Salleh, N. I. Ishak, K. A. Razak, M. Z. Abd Rahman, M. A. Asmadi, Z. Ismail, and M. F. Abdul Khanan

Keywords: Tropical rain forest, Landslide activity, Vegetation anomalies

Abstract. Remote sensing has been widely used for landslide inventory mapping and monitoring. Landslide activity is one of the important parameters for landslide inventory and it can be strongly related to vegetation anomalies. Previous studies have shown that remotely sensed data can be used to obtain detailed vegetation characteristics at various scales and condition. However, only few studies of utilizing vegetation characteristics anomalies as a bio-indicator for landslide activity in tropical area. This study introduces a method that utilizes vegetation anomalies extracted using remote sensing data as a bio-indicator for landslide activity analysis and mapping. A high-density airborne LiDAR, aerial photo and satellite imagery were captured over the landslide prone area along Mesilau River in Kundasang, Sabah. Remote sensing data used in characterizing vegetation into several classes of height, density, types and structure in a tectonically active region along with vegetation indices. About 13 vegetation anomalies were derived from remotely sensed data. There were about 14 scenarios were modeled by focusing in 2 landslide depth, 3 main landslide types with 3 landslide activities by using statistical approach. All scenarios show that more than 65% of the landslides are captured within 70% of the probability model indicating high model efficiency. The predictive model rate curve also shows that more than 45% of the independent landslides can be predicted within 30% of the probability model. This study provides a better understanding of remote sensing data in extracting and characterizing vegetation anomalies induced by hillslope geomorphology processes in a tectonically active region in Malaysia.