Development of a Remote Sensing-Based GIS Platform for Landslide Prediction

Thesis Abstract

Landslides constitute a significant natural hazard in many mountainous and hilly regions worldwide, causing substantial socio-economic losses and threatening human life and property. Despite advances in remote sensing technology and geographic information systems (GIS), there remains a critical gap in integrating these tools into an operational predictive framework capable of providing timely early warnings for susceptible areas. This study aims to develop a comprehensive remote sensing-based GIS platform tailored for landslide prediction, with objectives including the identification of key terrain and spectral indicators of landslide susceptibility, the integration of multi-temporal satellite data to analyze land surface dynamics, and the implementation of predictive modeling techniques within an accessible GIS environment. Employing a mixed-methods research design, the study combines quantitative spatial analysis with qualitative validation of model outputs. The primary population comprises 150 landslide-prone sites within the Kintambe Hills region, selected through stratified random sampling based on prior landslide inventories and expert assessments. Data collection involved high-resolution satellite imagery from Landsat 8 and Sentinel-2 platforms, digital elevation models (DEMs), soil type maps, geology layers, and rainfall datasets spanning a ten-year period. Instruments included GIS software tools (ArcGIS and QGIS), Remote SensingImage Processing software (ENVI), and field validation surveys conducted at 50 randomly selected sites to verify spectral signatures and terrain features related to landslide occurrence. Data preprocessing involved rectification, radiometric correction, and spectral enhancement to extract relevant indicators such as normalized difference vegetation index (NDVI), slope stability parameters, and land surface temperature anomalies. Spatial data were integrated using GIS layers, and statistical analysis was performed through a combination of regression analysis and machine learning techniques, specifically Random Forest classification, to model landslide susceptibility and predict future risk zones. The Analytical Hierarchy Process (AHP) theory guided the weight assignment for different variables, while the Theory of Landslide Mechanics provided a conceptual basis for understanding terrain failure processes. Expected findings include identifying critical thresholds for spectral and terrain indicators associated with landslide occurrence, generating a susceptibility map with high predictive accuracy (expected area under the curve (AUC) above 0.85), and demonstrating the efficacy of integrating multi-temporal remote sensing data within a GIS framework for real-time risk assessment. The models developed are anticipated to outperform traditional susceptibility assessments by providing dynamic, data-driven insights aligned with land surface changes. This research contributes novel methodologies by operationalizing a remote sensing-driven GIS platform that integrates spectral, topographical, and climatic data for landslide prediction, filling a gap in existing hazard modeling literature. It offers a scalable approach applicable to other vulnerable regions, enabling authorities to implement proactive landslide risk management strategies based on spatial-temporal dynamics. The platform's user-friendly interface supports decision-making processes, emergency preparedness, and land-use planning. The study concludes that remote sensing technology, combined with advanced GIS modeling, significantly enhances landslide prediction capabilities and supports sustainable land management. Recommendations include the continuous updating of remote sensing datasets to refine the model, capacity-building for local authorities in GIS application, and further research to incorporate additional environmental variables such as land use change and human activity. Future studies should explore the integration of real-time sensor networks and climate change scenarios to improve the robustness and responsiveness of landslide early warning systems.