digilib@itb.ac.id +62 812 2508 8800

EVALUATION OF LANDSLIDE SUSCEPTIBILITY IN THE MOUNTAINOUS REGION OF HAKHA-FALAM AREA, CHIN STATE, MYANMAR

Penulis : THAW ZIN OO (NIM: 22015701)
Kontributor / Dosen Pembimbing :
  • Supervisor :Dr. Eng. Imam A. Sadisun
  • Co-Supervisor: Dr. Rendy Dwi Kartiko.
Jenis Koleksi : Tesis
Penerbit : Teknik Geologi
Fakultas : Fakultas Ilmu dan Teknologi Kebumian
Subjek :
Kata Kunci : Landslide susceptibility, Weight of Evidence (WoE), Area Under Curve (AUC), Seed Cell Area Index (SCAI)
Sumber :
Staf Input/Edit : Alice D   Ena Sukmana
File : 1 file
Tanggal Input : 27 Sep 2017

Hakha-Falam area of Chin State is located in the Western part of Myanmar. The purpose of this study is to evaluate and compare the results of applying the two methods, bivariate (weight of evidence or WoE) and multivariate (logistic regression combined with WoE) for estimating landslide susceptibility in Hakha - Falam area. In order to do this, first, a landslide inventory map was constructed based on landslide locations from satellite image. Second, eight influencing parameters for landslide occurrence were utilized. The slope map, the elevation map, the slope aspect map and distance from river map were derived from a digital elevation model (DEM) with resolution 30 x 30 m. The lithology map was extracted from the geological map of Myanmar of Department of Geological and Survey and Minerals Explorer (DGSE) at a scale of 1:1,500,000. The soil texture and the land cover maps were extracted from satellite image at a scale of 250 m. Rainfall map was derived from satellite image at a scale of 1 km, respectively. Actual landslide locations were used to verify and to compare the results of landslide susceptibility maps. The accuracy of the results was evaluated by AUC (Area Under Curve) analysis. The results showed that the LR-WoE method is better accuracy than WoE methods. The AUC values for the statistical index of multivariate (logistic regression with WoE) method are 0.78275 for success and 0.77750 for prediction (validation) that higher than AUC for bivariate methods (WoE) ranged 0.71074 to 0.76576 for success and 0.70179 to 0.76405 for prediction (validation). Seed Cell Area Index (SCAI) analysis of LR-WoE method has low SCAI values for high and moderate classes and higher SCAI values for very low and low classes than other methods. Combined method (LRWoE model) has more pixel accuracy and low pixel errors than WoE methods. Therefore, based on the AUC value, SCAI analysis and spatial domain comparison, combined method (LR-WoE) result gives better result than one method (WoE) results for landslide susceptibility mapping.

Artikel Terkait