A Geographically Weighted Regression Approach to Landslide Susceptibility Modeling Thesis uri icon



  • Thesis (M.S., Geography) -- University of Idaho, 2017 | Landslide activity in Oregon causes more than $1 billion in property damage every year, and has resulted in several casualties over the past decades. The steep topography of the region, high-intensity precipitation events during the winter months, and easily weathered parent material, contribute to frequent slope failures in western Oregon. This study conducted a statistical landslide susceptibility assessment to evaluate the effects of geologic, morphologic, physical, and anthropogenic factors on landslide occurrence. Slope, erosion potential, hydrologic soil classes, volcanic and sedimentary geologic material, aspect, and curvature were identified as important predictors. A comparative analysis of traditional logistic regression (LR) and geographically weighted logistic regression (GWLR) was completed for the study area. The regression results from the LR and GWLR models were compared based on AIC, percentage of deviance explained, and prediction accuracy. The outputs demonstrated that GWLR outperformed standard LR in all models. GWLR improved prediction accuracy by 6.2% compared to traditional LR.

publication date

  • June 1, 2017