| Summary: | Intense rainfall-induced shallow landslides can have severe consequences, including soil erosion and vegetation loss, making in-depth research essential for disaster risk management. However, vegetation recovery processes after shallow landslides and their influencing multivariate factors are not well known. This study aims to address this gap by investigating the vegetation recovery processes after shallow landslides and the impact of topography on this recovery. We focus on two regions in Japan: the Shobara district in Hiroshima Prefecture and the Obara district in Aichi Prefecture. The Normalized Difference Vegetation Index (<i>NDVI</i>) and Enhanced Vegetation Index (<i>EVI</i>) derived from long-term Landsat images, as well as aerial photographs and environmental datasets, are used to measure vegetation recovery. Then, statistical analysis and the Seasonal Autoregressive Integrated Moving Averages (SARIMA) model were employed to investigate the dynamic response of vegetation under different combinations of environmental conditions using <i>NDVI</i> and <i>EVI</i> time series. Historical aerial photographs and vegetation index trend analysis suggest that vegetation in the study areas will take more than ten years to return to a stable state. The results also demonstrate the influence of atmospheric and land cover conditions when monitoring vegetation response using <i>NDVI</i> and <i>EVI</i>. In Obara, concave and convergent terrain positively influenced <i>NDVI</i>, while non-steep, low-elevation, and north-facing terrain positively influenced <i>EVI</i>. In Shobara, gentle and northwest-facing slopes were positively correlated with <i>NDVI</i>, and gentle and west-facing slopes were positively correlated with <i>EVI</i>. SARIMA modeling found that <i>NDVI</i> is more suitable for modeling the middle and late stages of vegetation recovery within 10–25 years after the landslide. In comparison, <i>EVI</i> is better for modeling the early stage of vegetation recovery within 10 years after the landslide.
|
|---|