Implementation of PCA multicollinearity method to landslide susceptibility assessment: the study case of Kabylia region

摘要

Landslides are one of the most catastrophic geo-risks observed in northern Algeria, particularly in the regions of mountain ranges (Mediterranean Kabylia), where the processes are spectacular. During the last decade, the risk of landslides has increased more and more in urban space, mainly affecting the economy and human life. The extension of urbanization aggravates natural risks and creates additional artificial risks. The present work aims to identify the zones of susceptibility to landslides for the Mediterranean Kabylia, in North Algeria, by using multivariate statistical analysis. Landslide inventory map of the study area was prepared based on 84 landslides. The seven parameters, (1) elevation, (2) lithology, (3) slope gradient, (4) NDVI (normalized difference vegetation index), (5) rainfall, (6) land use, and (7) human cause category, were prepared and were considered to be independent factors. The weights are assigned to each factor after removing the multicollinearity in the dataset using principal component analysis (PCA). The model’s landslide susceptibility maps were validated using (i) relative landslide density index, (ii) success rate, and (iii) predictive rate curves. The results of the analysis show that the moderate, high, and very high susceptibility classes to landslide density indexes represented 45.45, 71.43, and 83.33, respectively. The AUC (area under the curve) value is 0.782 indicating good model precision for identifying susceptible areas. The selection of parameters conditioning landslides is carefully made and even justified for a large number of these parameters. The PCA analysis also shows a good effect to remove multicollinearity of the parameters. Moreover, this model constitutes a first approach for assessing and planning landslides in Algeria and can be recommended for mapping the susceptibility to landslides in other regions.