Assessing Landslide Susceptibility Using Remote Sensing and GIS Techniques in Mountainous Regions

Thesis Abstract

Mountainous regions are inherently susceptible to landslides, a hazard exacerbated by climate change, deforestation, and unplanned development, posing significant risks to human life, infrastructure, and ecological integrity. Despite the increasing frequency and severity of landslide events in such regions, there remains a critical gap in spatially explicit and predictive susceptibility assessments that incorporate modern remote sensing and Geographic Information System (GIS) technologies. This study aims to systematically evaluate landslide susceptibility in the Southern Alps using a comprehensive integration of remote sensing data and GIS-based analytical techniques. The specific objectives include (i) to identify and map landslide-prone areas through the analysis of multispectral satellite imagery; (ii) to analyze terrain and land cover factors contributing to landslides; (iii) to develop a landslide susceptibility model using logistic regression analysis; and (iv) to validate the model with field-verified landslide inventories. The research adopts an interdisciplinary, quantitative approach, with a cross-sectional design utilizing spatial datasets collected over a three-year period. The population encompasses the entire mountainous catchment area of approximately 1500 square kilometers within the region, with a sample comprising 200 satellite-derived terrain parcels selected through stratified random sampling to ensure diverse land cover and slope conditions. Data collection instruments include Landsat 8 satellite imagery, Digital Elevation Models (DEMs), geological maps, land use/land cover maps, and field-based landslide inventories recorded through direct observations and drone surveys. The validity and reliability of remotely sensed data are ensured through calibration techniques, and ground truthing is employed to verify model outputs. Analytical methods involve preprocessing of remotely sensed data, followed by Landsat spectral analysis, slope and aspect derivation from DEMs, and land cover classification using supervised maximum likelihood algorithms. Spatial variables are extracted and integrated into a GIS environment, where initial correlations are assessed through exploratory data analysis. The core statistical technique employed is binary logistic regression, aimed at identifying significant predictors of landslide susceptibility. Model performance is evaluated through Receiver Operating Characteristic (ROC) curves, with an Area Under Curve (AUC) threshold set at 0.80 for acceptable predictive accuracy. The theoretical framework is anchored on the concept of natural hazard susceptibility which integrates the Human-Environmental Interaction Theory and the Slope Stability Model. Expected findings suggest that factors such as slope steepness, land cover change, proximity to geological fault lines, and soil type significantly influence landslide occurrence. The results are anticipated to produce a high-accuracy susceptibility map delineating zones at varying levels of risk, providing a spatial basis for targeted mitigation efforts. This work advances existing models by integrating multispectral remote sensing data and GIS analytical techniques for enhanced predictive reliability in mountainous terrains. The contribution to knowledge lies in refining spatially explicit landslide susceptibility models tailored to the Southern Alps, demonstrating the efficacy of remote sensing-GIS integration in hazard assessment. The findings will inform local land use planning and disaster risk reduction strategies and offer a methodological blueprint for similar studies in comparable environments. In conclusion, the study underscores the importance of spatial technology in hazard management, recommending the adoption of integrated remote sensing and GIS approaches for proactive landslide risk mitigation. Future research should explore dynamic modeling incorporating temporal satellite data to assess landslide susceptibility changes over time, thereby enhancing early warning systems and sustainable land management practices.