Landslide susceptibility mapping using remote sensing data and geographic information system-based algorithms
Whether they occur due to natural triggers or human activities, landslides lead to loss of life and damages to properties which impact infrastructures, road networks and buildings. Landslide Susceptibility Map (LSM) provides the policy and decision makers with some valuable information. This study a...
Saved in:
Main Author: | |
---|---|
Format: | Thesis |
Language: | English |
Published: |
2019
|
Subjects: | |
Online Access: | http://eprints.utm.my/id/eprint/83986/1/AyubMohammadiPFAB2019.pdf http://eprints.utm.my/id/eprint/83986/ http://dms.library.utm.my:8080/vital/access/manager/Repository/vital:126525 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | Universiti Teknologi Malaysia |
Language: | English |
Summary: | Whether they occur due to natural triggers or human activities, landslides lead to loss of life and damages to properties which impact infrastructures, road networks and buildings. Landslide Susceptibility Map (LSM) provides the policy and decision makers with some valuable information. This study aims to detect landslide locations by using Sentinel-1 data, the only freely available online Radar imagery, and to map areas prone to landslide using a novel algorithm of AB-ADTree in Cameron Highlands, Pahang, Malaysia. A total of 152 landslide locations were detected by using integration of Interferometry Synthetic Aperture RADAR (InSAR) technique, Google Earth (GE) images and extensive field survey. However, 80% of the data were employed for training the machine learning algorithms and the remaining 20% for validation purposes. Seventeen triggering and conditioning factors, namely slope, aspect, elevation, distance to road, distance to river, proximity to fault, road density, river density, Normalized Difference Vegetation Index (NDVI), rainfall, land cover, lithology, soil types, curvature, profile curvature, Stream Power Index (SPI) and Topographic Wetness Index (TWI), were extracted from satellite imageries, digital elevation model (DEM), geological and soil maps. These factors were utilized to generate landslide susceptibility maps using Logistic Regression (LR) model, Logistic Model Tree (LMT), Random Forest (RF), Alternating Decision Tree (ADTree), Adaptive Boosting (AdaBoost) and a novel hybrid model from ADTree and AdaBoost models, namely AB-ADTree model. The validation was based on area under the ROC curve (AUC) and statistical measurements of Positive Predictive Value (PPV), Negative Predictive Value (NPV), sensitivity, specificity, accuracy and Root Mean Square Error (RMSE). The results showed that AUC was 90%, 92%, 88%, 59%, 96% and 94% for LR, LMT, RF, ADTree, AdaBoost and AB-ADTree algorithms, respectively. Non-parametric evaluations of the Friedman and Wilcoxon were also applied to assess the models’ performance: the findings revealed that ADTree is inferior to the other models used in this study. Using a handheld Global Positioning System (GPS), field study and validation were performed for almost 20% (30 locations) of the detected landslide locations and the results revealed that the landslide locations were correctly detected. In conclusion, this study can be applicable for hazard mitigation purposes and regional planning. |
---|