Effects of Digital Elevation Model resolution on evaluation of landslide susceptibility with a logistic regression model.

Abstract

The use of statistical methods together with the GIS technologies is currently one of the most efficient tools in the assessment of landslide susceptibility. The correlation between the physical phenomenon and its triggering factors depends on several factors, including the resolution at which the elevation data are represented in a Digital Elevation Model (DEM). The resolution becomes increasingly important as the use of DEM data is extended for spatial prediction of terrain attributes such as slope, aspect, plan and profile curvature, etc., which are considered as triggering factors of the landslides. Many methods exist in scientific literature to capture and model the correlation between landslides and triggering factors but very few studies have investigated the effects of DEM resolution on it. Among these, the logistic regression method has provided successful results in many applications. The objective of this study is to investigate the effects of DEM resolution on assessment of landslide susceptibility when a logistic regression model is used. The model will be applied several times to a study area located in the eastern Sicily using DEMs with different resolution. Results of the susceptibility analysis will be discussed and compared by means of the Area under the ROC (Receiver Operating Characteristic) curve. Moreover, the influence of DEM resolution on the logistic regression coefficients will be assessed.