KHADKA Madan Kumar, TIMALSINA Niroj, ALI Kashif, GHIMIRE Motilal, ZHANG Zhiming. 2025: Assessing landslide susceptibility: A comparative analysis of bivariate statistical methods along the proposed China-Nepal Railway Corridor. Journal of Mountain Science, 22(6): 1966-1992. DOI: 10.1007/s11629-024-9294-4
Citation: KHADKA Madan Kumar, TIMALSINA Niroj, ALI Kashif, GHIMIRE Motilal, ZHANG Zhiming. 2025: Assessing landslide susceptibility: A comparative analysis of bivariate statistical methods along the proposed China-Nepal Railway Corridor. Journal of Mountain Science, 22(6): 1966-1992. DOI: 10.1007/s11629-024-9294-4

Assessing landslide susceptibility: A comparative analysis of bivariate statistical methods along the proposed China-Nepal Railway Corridor

  • Landslides pose a significant threat in Nepal, causing substantial losses of life and property every year. This risk is heightened by the region's rugged steep mountainous terrain, heavy rainfall and tectonic activity. Additionally, human activities such as constructing roads and railways in extremely sensitive regions with geological hazards are major contributors to landslides in Nepal. The China-Nepal railway passes through high-risk zones like Rasuwa, Nuwakot, and Kathmandu, with the Saprubesi-Bidur corridor being especially vulnerable. Accurate landslide assessment is crucial for planning such large-scale projects. This study evaluates landslide susceptibility within a 10-kilometer buffer zone surrounding the proposed China-Nepal Railway part of the Belt and Road Initiative (BRI). Landslide susceptibility assessment is performed using certainty factor (CF), frequency ratio (FR), statistical index (SI) and weights of evidence (WoE) models. Altogether 599 landslides were inventoried from the image series in Google Earth and validated in the field. Of these landslides, 70% were used for model development and the remaining 30% were used for validation. Nineteen conditional factors including elevation, relative relief, slope, aspect, plan curvature, profile curvature, topographical position index (TPI), stream power index (SPI), drainage density, topographic wetness index (TWI), rainfall, normalized difference vegetation index (NDVI), land cover, distance from roads, distance from rivers, geology, distance from faults, LS factor and ruggedness index were used for the mapping of landslide susceptibility. We found that the classes of factors and landslide occurrences were consistent across the CF, FR, SI, and WoE models with descriptive statistics indicating that the CF model offered the most stable estimates. Correlation analysis reveals strong relationships among the methods, particularly between WoE and SI. Conversely, FR and SI exhibit the weakest correlation, despite some variability in their distributions. Likewise, the effectiveness of the models was evaluated using the area under the curve (AUC) which revealed that the CF, FR, SI, and WoE models achieved success rates of 83.6%, 82.0%, 83.1% and 82.9% respectively. With a spatial correlation of over 90% among landslide susceptibility maps created through selected methods, any of the selected models' results could be effectively applied to the management of landslides. Additionally, the susceptibility values from the different models (CF, FR, SI, and WoE) are especially important in the development of railway projects when located within 2-4 km from a planned station. The landslide susceptibility maps can be useful and affordable planning tools for designers and engineers working on the China-Nepal Railway and other similar construction projects.
  • loading

Catalog

    Turn off MathJax
    Article Contents

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return