Evaluación de la Peligrosidad de Deslizamientos de laderas empleando Técnicas de Regresión Logística/ Hazard Assessment of earthquake-triggered landslides using logistic regression tecniques

Abstract

In this paper, a hazard assessment of earthquake-triggered landslides model is applied to a local-scale area using logistic regression techniques. The probability of landslide occurrence is considered as dependent variable, which is modelized by a sigmoid function (taking values between 0 and 1), and the independent variables are the factors related to the susceptibility of the terrain. The methodology is implemented in a Geographical Information System (GIS), where the different susceptibility factors (elevation, slope, aspect, mean annual precipitation, lithology, land uses and roughness terrain) are stored, together with an inventory of landslides in order to calibrate the model in a subsequent phase. An application is developed in a 10x6 km area at Santa Tecla (El Salvador) , modelling the scenario of 13th January 2001 with the necessary information to define the susceptibility as well as the triggering function: digital cartography 1:25.000, geological maps, rainfall database and strong-motion records of the 2001 earthquake. Results show the variables that more influence susceptibility and a hazard map which is contrasted with the landslides inventory. Finally, a high concordance between the landslides inventory and high-hazard areas is observed.