Accès ouvert

Landslide susceptibility mapping using machine learning techniques and comparison of their performance at Ziz upper watershed, Southeastern Morocco

Article scientifique 2022 Anglais

Résumé

Abstract Landslides pose a serious threat to humans, their infrastructure and property, especially in mountainous areas. In Morocco, these risks have attracted more attention due to their harms. The current study aims to use machine learning techniques (MLTs) including Multi-Layer Perceptron (MLP), Adaptive Boosting (AdaBoost), Random Forest (RF), Instance-Based K (IBK), Naïve Bayes (NB) and Decision Tree (J48) to model landslide susceptibility and to compare their performance. Initially, 144 landslide sites were inventoried. Then, thirteen factors related to landslides were considered. Finally, the area under the receiver operating characteristic curve (AUC-ROC) method was used to compare models’ performance. The results showed that AUC values for six MLTs range from 53.0% for RF to 96.1% for J48. Both J48 (AUC = 96.1%) and NB (AUC = 89.5%) showed the best performance compared to the others. The results of this work and the generated landslide susceptibility maps can help decision makers to avoid high susceptible areas.

Citer ce document

Manaouch, M., Sadiki, M., Aghad, M., Batchi, M., Karkouri, J. (2022). Landslide susceptibility mapping using machine learning techniques and comparison of their performance at Ziz upper watershed, Southeastern Morocco. https://doi.org/10.21203/rs.3.rs-1534262/v1

Accès au document

Voir sur le dépôt source

Ce document est hébergé sur son dépôt institutionnel d'origine.

Auteur(s)

Mohamed Manaouch Mohamed Sadiki Mohamed Aghad Mohcine Batchi Jamal Al Karkouri

Statistiques

Consultations : 1

Téléchargements : 0