Landslide inventory mapping derived from multispectral imagery by Support Vector Machine (SVM) algorithm

Abstract

Indonesia is located right on the equator, which receives a lot of heat from the sun and rainfall. Therefore, Indonesia is prone to hydro meteorological natural disasters such as droughts, large sea waves, erosion, floods and landslides. The National Disaster Management Agency (BNPB) noted that floods are followed by landslides of the total hydro-meteorological disasters that most often occur in Indonesia. An inventory of the distribution of multi-year landslides is essential as a basis for disaster mitigation and disaster risk reduction. The research case study was carried out in an area prone to landslides around Mount Batur, Bali-Indonesia. Characteristics of areas with high rainfall and steep slopes (>45%). Detection of areas affected by landslides can be identified with multispectral remote sensing images such as Sentinel 2 Image with a spectral resolution of 13 bands and a spatial resolution ranging from 10-60 m. Data acquisition was carried out in the period 2017-2021. The Support Vector Machine (SVM) algorithm is an alternative for detecting landslide areas in this study. The result showed that the accuracy assessment of the SVM algorithm on the training and validation/testing models is more than 84%. We obtained carrying out a landslide inventory is 25.29 km2. Based on our analysis, the most extensive landslide distribution was found in Batur Village (South and Central), followed by Songan A, Sukawana, Kintamani, and Buahan Villages. This research can be used to develop the Landslide Susceptibility model so that entering the landslide inventory parameters gives good results. As well as a basis for disaster risk reduction (DRR), especially for the community, government, and tourists in this research location.

Landslide Mapping from the Cylcone Seroja in Kupang City Using Remote Sensing

“…The Landsat image was then classified using Support Vector Machines (SVM) algorithm as Supervised Classification method. This SVM algorithm is susceptible to distinguishing objects and requires simple training data [13]. For the classification classes, we took GPS of training sites from different land cover samples.…”

DCP-Net: An Efficient Image Segmentation Model for Forest Wildfires

“…These region images are then converted into feature vectors and passed to a classifier for training. Finally, classifiers such as SVM [12], Bayesian networks, random forests, BP neural networks, etc., are employed to compare these extracted features with a set of existing standard features to determine whether the image contains fire. Qiu et al [13] proposed a new algorithm to clearly and continuously define the edges of flames and fire spots.…”