Multitemporal UAV surveys of geomorphological changes caused by postfire heavy rain in Kamaishi city, northeast Japan

Unmanned aerial vehicles (UAVs) are widely utilized for surveying various sediment disasters, and the high -resolution imaging they provide reveals detailed topographic characteristics. However, the availability of mul-titemporal UAV observations, which reveal geomorphological changes before and after the disaster, is still limited. In this research, we performed multitemporal UAV observations for geomorphological changes before and after extreme postfire heavy rain in the Ozaki Peninsula wildfire area in Japan. Significant geomorphological changes were observed associated with eight landslides and stream shifting, with a spatial resolution of 0.30 cm obtained by structure from motion (SfM) processing. To extract the surface changes based on multitemporal UAV observations, in addition to elevation displacement, vertex normal vectors were calculated for each mesh. We found that the elevation displacement showed comprehensive alterations in the entire area and that the changes in vertex normal vectors effectively captured the shape of these alterations. Deterioration, deposition, and scarping were shown for landslides, and the dynamics of erosion and deposition in the cross-section were for stream shifting. In addition, the detected geomorphological changes were compared with the predisaster topo-graphic characteristics. Landslides occurred with a catchment area of 500 to 2,000 m2 and a slope inclination of 30 to 40 degrees, and stream shifting occurred with a catchment area of 10,000 m2 or more and a slope inclination of less than 20 degrees. In both regions affected by landslides and stream shifting, the logarithmic stream power index (logSPI), a natural logarithmic representation of the conventional stream power index (SPI), could describe the characteristics roughly. A logSPI value of 4 to 5 represented landslides, and 5 and greater represented stream shifting, with the probability of landslides and stream shifting occurring in areas with these values exceeding 99.5% and 99.9%, respectively. Therefore, the logSPI was utilized to roughly classify the terrain and geomor-phological condition and the points among the unburnt, unlogged-burnt, and logged-burnt areas were compared. Even when the logSPI was similar, the damage was much more severe in the burnt areas. On slopes with a logSPI of approximately 4, there were signs of erosion in the logged-burnt area, and the increase in fallen trees was significant in the unlogged-burnt area, while there were no substantial changes in the unburnt area. In valleys with a logSPI of 8 or greater, sediment and woody debris were significantly deposited in the burnt areas following the postfire rain, while there were few changes in the unburnt area.