Developing landslide susceptibility map using Artificial Neural Network (ANN) method for mitigation of land degradation

Abstract

<p>Landslides are one of the crucial problems that have an impact on land degradation and human life. This study aimed to develop vulnerability maps using ANN to mitigate land degradation in the Bromo Tengger Semeru with the extending area of Universal Transverse Mercator (UTM) Coordinate System Top 91277639, Bottom 911569, Left 692860, and Right 706860. The method applied the Artificial Neural Network (ANN) model using RStudio machine learning. Landslides were mapped using Sentinel Image and Orthomozaic photo interpretation from data acquisition using Unmanned Aerial Vehicle (UAV). The landslide control factor data was obtained through DEMNAS (National Digital Elevation Model) with a spatial resolution of 8 meters. Data normalisation was conducted using the Mix-Max method before it was processed using RStudio. The landslide existing for ANN workflow was processed using the Bioclim model. The results showed landslide susceptibility was categorised into four classes i.e., low susceptibility (29.83%), which was spatially spread on most in the lower slopes, moderate susceptibility (3.11%), high susceptibility (2.99%), and very high susceptibility (15.94) which is scattered on the upper slope to the middle slope of the watershed. The most significant factor influencing the landslide is the topography factor, with a Relative Importance (RI) value of 0.86; the hydrological factor, with an RI of 0.833 and the surface feature, with an RI of 0.355. The results of the landslide susceptibility model are very proper for land degradation mitigation strategies. It has high accuracy through an Area Under Curve (AUC) of 0.965 and a Precision Recall Curve (PRC) of 0.976.</p>