A novel hybrid framework for landslide susceptibility mapping using entropy-weighted fuzzy overlay and random forest: case study of Changshan County, China

This study proposes a novel hybrid approach combining entropy-weighted fuzzy overlay (EWFO) with a random forest (RF) model to address the inherent randomness and ambiguity in landslide susceptibility mapping. Nine environmental factors were used to investigate the landslide triggering mechanism, and "altitude" and "distance to the river" were identified as key constraints in the EWFO model. Landslide susceptibility maps were generated using the EWFO, RF, and EWFO & RF integrated models in Changshan County, China. The EWFO & RF model showed the highest prediction performance (AUC = 0.837), slightly outperforming the RF (AUC = 0.811) and EWFO (AUC = 0.758) models. Although the performance improvement was not significant, the fuzzification process improved the reliability of the input data and the prediction accuracy. This study highlights the effectiveness of combining EWFO with RF for accurate landslide susceptibility assessment, provides valuable insights for risk mitigation in mountainous areas, and paves the way for future geohazard research.