@Article{rig.2025.065667,
AUTHOR = {Nurul A. Asram, Eran S. S. Md Sadek},
TITLE = {Evaluating Shannon Entropy-Weighted Bivariate Models and Logistic Regression for Landslide Susceptibility Mapping in Jelapang, Perak, Malaysia},
JOURNAL = {Revue Internationale de Géomatique},
VOLUME = {34},
YEAR = {2025},
NUMBER = {1},
PAGES = {619--637},
URL = {http://www.techscience.com/RIG/v34n1/63340},
ISSN = {2116-7060},
ABSTRACT = {Landslides are a frequent geomorphological hazard in tropical regions, particularly where steep terrain and high precipitation coincide. This study evaluates landslide susceptibility in the Jelapang area of Perak, Malaysia, using Shannon Entropy-weighted bivariate models (i.e., Frequency Ratio, Information Value, and Weight of Evidence), in comparison with Logistic Regression. Seven conditioning factors were selected based on their geomorphological relevance and tested for multicollinearity: slope gradient, slope aspect, curvature, vegetation cover, lineament density, terrain ruggedness index, and flow accumulation. Each model generated susceptibility maps, which were validated using Receiver Operating Characteristic curves and Area Under the Curve metrics. Logistic Regression yielded the highest predictive accuracy, reflecting its strength in capturing interactions among variables. Among the bivariate models, Frequency Ratio performed best, slightly outperforming the other two methods. Zones of high susceptibility were consistently located along steep slopes, high lineament density areas, and near built environments. The study demonstrates that incorporating Shannon Entropy improves the performance of conventional bivariate methods and provides a useful framework for spatial susceptibility modeling in data-constrained environments. The comparison with Logistic Regression highlights the advantages of multivariate modeling in capturing complex spatial relationships. Limitations of the study include the use of secondary spatial data and the exclusion of dynamic parameters such as rainfall intensity. Future research should incorporate temporal datasets and investigate machine learning techniques to enhance model generalizability and predictive capability.},
DOI = {10.32604/rig.2025.065667}
}