Abstract
Abstract. In this study, the susceptibility to landslides at Sevilla township, Valle del Cauca, located at southwest of Colombia was evaluated. The conditioning factors that involve the generation of landslides were evaluated using Geographic Information Systems (GIS) and Remote Sensing (RS) techniques. For the estimating susceptibility, an Artificial Neural Network (ANN) was implemented by applying the “Backpropagation” method to extract the synoptic weights of the conditioning variables (slopes, flow length, curvature, geology, fracture density, and land cover) on an automatic way with a data training module. The data for the analysis of the conditioning factors were carried out through a Digital Elevation Model (DEM) obtained through Radar Interferometry techniques, with Sentinel-1B satellite images for the year 2018. The results showed that Sevilla’s township has areas with high susceptibility, high slopes, and that it’s crossed by an active geological fault which implies that the earth's dynamics will condition the terrain stability.
Highlights
The susceptibility of a land is known as soil’s tendency to generate phenomena on removal, depending on its intrinsic properties, that is, it shows the predisposition of the environment and the elements that make up the landscape such as, geomorphology, geology, coverage, etc.During the years 1996 to 2002 at Sevilla Township there were 9 landslide events, which caused a total of 37 people dead, 104 injured and 141 affected population (Ingeominas, 2002)
Very high susceptibility: Present in 4.54% of the territory of Sevilla, Valle del Cauca, with a total of 26.420 ha; a hierarchy is evident in the variables of fracture density and geology, since it belongs to the pixels located in the course of the faults present in the sector
Artificial Neural Networks have a number of advantages over the different methods used to estimate susceptibility
Summary
The susceptibility of a land is known as soil’s tendency to generate phenomena on removal, depending on its intrinsic properties, that is, it shows the predisposition of the environment and the elements that make up the landscape such as, geomorphology, geology, coverage, etc. The use of GIS is a key point for mapping landslides (Guzzetti, and Paola Carrara, 1999), as it allows different calculations using large datasets This tool is commonly used because it facilitates data management and mapping susceptibilities produced by each selected variable. Kawabata & Bandibas (2009) and Pradhan & Lee (2010b, 2010a), propose a multivariate statistical technique using an ANN to calculate the landslides susceptibility produced by the set of variables considered. In this method, neurons are organized as sequential layers, each consisting of one or more neurons: the input layer, the hidden layer and the output layer. Random Forest (RF) supervised classification was performed to obtain coverage types
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
