Landslide Susceptibility Assessment Using Modified Frequency Ratio Model in Kaski District, Nepal

Abstract

Landslides are the most common natural hazards in Nepal especially in the mountainous terrain. The existing topographical scenario, complex geological settings followed by the heavy rainfall in monsoon has contributed to a large number of landslide events in the Kaski district. In this study, landslide susceptibility was modeled with the consideration of twelve conditioning factors to landslides like slope, aspect, elevation, Curvature, geology, land-use, soil type, precipitation, road proximity, drainage proximity, and thrust proximity. A Google-earth-based landslide inventory map of 637 landslide locations was prepared using data from Disinventar, reports, and satellite image interpretation and was randomly subdivided into a training set (70%) with 446 Points and a test set with 191 points (30%). The relationship among the landslides and the conditioning factors were statistically evaluated through the use of Modified Frequency ratio analysis. The results from the analysis gave the highest Prediction rate (PR) of 6.77 for elevation followed by PR of 66.45 for geology and PR of 6.38 for the landcover. The analysis was then validated by calculating the Area Under a Curve (AUC) and the prediction rate was found to be 68.87%. The developed landslide susceptibility map is helpful for the locals and authorities in planning and applying different intervention measures in the Kaski District.

Comparison of Statistical and Machine-Learning Model for Analyzing Landslide Susceptibility in Sumedang Area, Indonesia

“…During the last decade, various techniques and approaches have been developed to study landslide susceptibility, including qualitative [6], statistical [7][8][9][10], numerical [1,11], and through the use of machine learning [12][13][14][15][16]. Machine learning is now becoming more widely used in landslide prevention, as it can provide optimal, accurate, efficient, and effective results with proper conditioning.…”

“…Furthermore, the landslide-susceptibility analysis in this study will use statistical and machine-learning methods; namely, frequency ratio (FR) and random forest (RF), respectively. This is based on the fact that FR had enjoyed good results in landslide statistical models [8,22,[24][25][26][27][28][29]. In addition, RF has also resulted in good performance in landslide susceptibility according to [5,12,14,21,[30][31][32][33][34].…”