| Landslides represent a critical natural hazard in the Northern Tehran Basin, posing significant threats due to its complex geological setting, rugged topography, and anthropogenic activities such as road construction. This study introduces an innovative hybrid framework incorporating dynamic weighting based on Particle Swarm Optimization (PSO) and Genetic Algorithm (GA) for landslide susceptibility zonation. Unlike conventional methods (e.g., Frequency Ratio [FR], Statistical Index [SI], and Shannon Entropy [SE]), which employ static weights, our approach dynamically adjusts factor weights (e.g., distance to rivers, slope, lithology) using PSO, accounting for temporal variables such as seasonal rainfall and human activity. We compiled rainfall data and 150 landslide events (2005–2024) from local meteorological stations and geological databases. Input parameters included eight key factors (distance to rivers, distance to roads, slope, lithology, elevation, aspect, distance to faults, and land use) alongside seasonal rainfall. Results demonstrate that dynamic weighting improves prediction accuracy by 15% (AUC-ROC = 0.923 for PSO vs. 0.804 for FR), particularly during high-rainfall seasons where river proximity weight increased (vj = 8.2 vs. 7.21 in static models). The PSO-GA hybrid outperformed traditional models, with PSO (AUC-ROC = 0.923) and GA (AUC-ROC = 0.917) showing superior precision. Dynamic hazard maps accurately identified high-risk zones (e.g., near rivers with vj = 8.23 during rainy seasons). This approach offers a robust tool for landslide risk management in urbanized mountainous regions like Northern Tehran and serves as a replicable model for similar environments globally. |
