Comparison of landslide susceptibility assessment in volcanic terrains based on a landform unit model and two multivariate logistic regression models
Por:
Paulin G.L., Bursik M., Salinas L.A.
Publicada:
1 ene 2026
Resumen:
© 2025 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license. http://creativecommons.org/licenses/by-nc-nd/4.0/Landslides occurring in deep and narrow valleys are common in Mexico because the poorly consolidated volcanic deposits are easily carved by rivers. They have the potential to impact society, disrupt infrastructure, and cause loss of life. This is the case for Ixtaccíhuatl volcano, the third highest stratovolcano in Mexico. This dormant volcano is prone to landslides due to earthquakes and the loose volcanoclastic deposits that are transported by stream floods, flash floods, and lahars during the rainy season. This study assessed and compared three landslide susceptibility models for the Xopanac watershed on the eastern flank of Iztaccíhuatl. In the watershed, unstable areas are mainly along first-order streams that carve deposits, whose stability has further decreased as a result of human activity. The susceptibility was modeled using a pre-existing landform unit model and two Multiple Logistic Regression (MLR) models, using the following variables: elevation, slope angle, aspect, curvature, drainage density, vertical erosion, land use, and with and without landform units. The resultant susceptibility maps were validated according to the Area Under the Curve (AUC) of the Receiver Operating Characteristic and by comparing the inventory map and the model in a contingency table. Both the area under the curve and the statistics from the contingency table indicate that the multiple logistic regression model using landform units has a higher predictive capacity than the multiple logistic regression model not using landform units and the pre-existing landform unit model. However, the multiple logistic regression model not using landform units is preferred over the other two models in areas with sparce information because it obtains a moderate result with fewer significant variables.
Filiaciones:
Paulin G.L.:
Instituto de Geografia, Universidad Nacional Autonoma de Mexico, Circuito Exterior, Ciudad Universitaria, Coyoacan 04510, Ciudad de Mexico, Mexico
Bursik M.:
Center for Geohazards Studies, University at Buffalo, SUNY Buffalo, Buffalo, NY 14260, USA
Salinas L.A.:
Universidad Autonoma de la Ciudad de Mexico, Colegio de Ciencias y Humanidades, Academia de la Licenciatura Proteccion Civil y Gestion de Riesgos., Ciudad de Mexico, Mexico
|