A Robust Deep-Learning Model for Landslide Susceptibility Mapping: A Case Study of Kurdistan Province, Iran.

Bahareh Ghasemian,Abolfazl Jaafari,Himan Shahabi,Somayeh Renoud,Ataollah Shirzadi,Anuar Ahmad,Nadhir Al-Ansari,Marten Geertsema,Victoria R Kress

doi:10.3390/s22041573

Abstract

We mapped landslide susceptibility in Kamyaran city of Kurdistan Province, Iran, using a robust deep-learning (DP) model based on a combination of extreme learning machine (ELM), deep belief network (DBN), back propagation (BP), and genetic algorithm (GA). A total of 118 landslide locations were recorded and divided in the training and testing datasets. We selected 25 conditioning factors, and of these, we specified the most important ones by an information gain ratio (IGR) technique. We assessed the performance of the DP model using statistical measures including sensitivity, specificity, accuracy, F1-measure, and area under-the-receiver operating characteristic curve (AUC). Three benchmark algorithms, i.e., support vector machine (SVM), REPTree, and NBTree, were used to check the applicability of the proposed model. The results by IGR concluded that of the 25 conditioning factors, only 16 factors were important for our modeling procedure, and of these, distance to road, road density, lithology and land use were the four most significant factors. Results based on the testing dataset revealed that the DP model had the highest accuracy (0.926) of the compared algorithms, followed by NBTree (0.917), REPTree (0.903), and SVM (0.894). The landslide susceptibility maps prepared from the DP model with AUC = 0.870 performed the best. We consider the DP model a suitable tool for landslide susceptibility mapping.

Highlights

Landslides occur in a variety of materials and undergo various styles of movement at different rates [1]
We obtained the relative importance of the factors influencing landslide occurrence based on average merit (AM) as information gain ratio (IGR) score through the k-fold cross-validation technique (Table 2)
We illustrated the robustness of a deep-learning model against three benchmark models (SVM, Naïve Bayes tree (NBTree), and REPTree) for the prediction of landslide susceptibility in Kamyaran city, Kurdistan Province, Iran

Summary

Introduction

Landslides occur in a variety of materials and undergo various styles of movement at different rates [1]. Landslides play an important geomorphological role in the evolution of landscapes, impacting the natural (soils, ecosystems, aquatic habitat, etc.) and built (residential areas, roads, pipelines, etc.) environment [2,3]. Landslide hazards are often exacerbated by land use practices such as road building, and deforestation, and may be made worse by increases in precipitation [4]. It is important to identify areas that have a high potential for landslides and mitigate landslide damage. Landslide risk assessment methodologies can be classified into three dominant groups: qualitative, quantitative, and artificial intelligence approaches. Qualitative approaches often rely on air photo and field interpretation and expert judgment (e.g., Schwab and Geertsema [5]). Quantitative methods are based on mathematical rules and expert judgment [6]

Objectives

Methods

Results

Discussion

Conclusion