Multitemporal UAV surveys for detection of shallow landslides and vegetation recovery

Heavy rainfall and earthquakes are major triggers of landslides. Handling the predicted increases in heavy rainfall due to climate change in East Asia requires robust knowledge regarding to the means to control landslides and those of vegetation recovery, which affects soil erosion and slope stability, at landslide sites. This study examined multitemporal high-spatial-resolution datasets from an unmanned aerial vehicle (UAV) to detect rainfall-induced landslides in 2012 and coseismic landslides triggered by the 2016 Kumamoto earthquake at Aso Volcano, Japan. The study area is characterized by grass vegetation, which is the typical landscape of the region. Vegetation recovery at the landslide sites was also monitored using these multitemporal datasets. We obtained orthorectified images and digital surface models with a spatial resolution of 0.06 m. These high-spatial-resolution datasets show that the coseismic landslides, many of which were initiated near topographic ridges, were typically located on upside hillslopes of previous rainfall-induced landslide scars. The total sediment production of the rainfall-induced and coseismic landslides was 1.2-4.8 × 10⁵ m³/km² and 2.5 × 10⁴ m³/km², respectively. Our results reveal strong vegetation recovery at both the rainfall-induced and coseismic landslide sites. The rainfall-induced landslide sites were covered with grass vegetation by 2019. However, site-specific vegetation recovery was determined primarily by topographic parameters, such as slope angle and direction, at the local scale.