Assessment of vegetation dynamic and its effects in a large-scale landslide in Central Taiwan with multitemporal Landsat images

Large-scale landslides often result in severe soil displacement and the exposure of bedrock, particularly combined with heavy rainfall. This condition significantly increases the risk of sediment-related disasters. Consequently, vegetation restoration and succession following landslide events are critical strategies for mitigating such hazards and enhancing disaster resilience. In this study, we integrated multi-temporal remote sensing imagery, land use classification, and Markov chain change simulations to evaluate the dynamic restoration of vegetation in a large-scale landslide area. Field surveys were conducted to validate the observed patterns of vegetation recovery. The results showed high accuracy in land use classifications derived from eight temporal images, with overall accuracy surpassing 80% and Kappa coefficients exceeding 0.7. The primary areas of vegetation recovery were identified as forests, followed by grasslands. Spatial change simulations indicated that full vegetation stability is expected to be reached after 2075. We emphasized the efficacy of combining remote sensing and modeling techniques for long-term monitoring of vegetation dynamics and offer critical insights for formulating sustainable strategies for disaster management.