Document Type : Origional Article
Authors
1
Assistant Professor of Geomorphology, Department of Geography, Faculty of Literature and Humanities, University of Guilan, Rasht, Iran
2
Ph.D Student, Department of Physical Geography, Faculty of Social Sciences, University of Mohaghegh Ardabili, Ardabil, Iran.
3
Professor, Department of Physical Geography, Faculty of Social Sciences, University of Mohaghegh Ardabili, Ardabil, Iran.
Abstract
Background and Objective: Landslides are among the most significant natural hazards in the mountainous and humid regions of northern Iran, causing substantial damage to natural resources, infrastructure, and human settlements each year. The Navroud watershed in Gilan Province is considered highly susceptible to landslide occurrence due to its complex topography, relatively high precipitation, and intensive human activities. The primary objective of this study is to assess and map landslide susceptibility and to determine the relative importance of the influencing factors using the Random Forest model in combination with the capabilities of Google Earth Engine (GEE) and Geographic Information Systems (GIS) within the Navroud watershed.
Methodology: This research adopts an applied–analytical approach, in which eleven landslide-conditioning factors including slope, aspect, elevation, geology, land use, vegetation cover, slope curvature, distance to roads, distance to drainage networks, distance to faults, and annual precipitation were selected as independent variables. The required thematic layers were generated using remote sensing data, the ALOS PALSAR digital elevation model with a spatial resolution of 12.5 m, geological maps, and meteorological data. After fuzzification and normalization, the datasets were integrated into the modeling process. A landslide inventory map comprising 150 landslide locations was prepared based on satellite imagery interpretation and field observations, and the data were divided into training (70%) and validation (30%) subsets. The Random Forest model was implemented in the Google Earth Engine environment, and the final landslide susceptibility map was produced in five classes: very low, low, moderate, high, and very high.
Results and findings: The results indicate that approximately 32% of the watershed area falls within the high and very high susceptibility classes, predominantly concentrated in the elevated and steep slopes of the northwestern ad southwestern parts of the basin. Furthermore, variable importance analysis revealed that slope, elevation, precipitation, land use, and distance to roads are the most influential factors controlling landslide occurrence. Model performance evaluation using the ROC curve showed that the Area Under the Curve (AUC) value during the validation stage was 0.897, demonstrating the high accuracy and strong predictive capability of the Random Forest model in identifying landslide-prone areas.
Keywords
Subjects
