Utilizing airborne LiDAR, reducing sediment volume estimation time to one-tenth of the conventional method

Okayama, Japan – On September 24, a research group led by Professor Takeharu Sato from the Department of Biosphere-Geosphere Science at Okayama University of Science (OUS) announced they have developed and filed a patent for a technology that can efficiently and rapidly identify areas affected by sediment movement and estimate the volume of material displaced immediately after simultaneous landslide occurrences. This technique uses Digital Surface Models (DSM) obtained from Airborne LiDAR surveys, enabling sediment volume estimation in less than one-tenth of the time required by conventional methods. It is the first method of its kind in Japan. This innovation significantly reduces labor and costs, aiding in the swift establishment of recovery systems.
 
At a press conference held at OUS, Professor Sato explained that while the conventional method involves using Airborne LiDAR, manual visual interpretation is required by engineers to account for discrepancies caused by trees and other factors when comparing pre-and post-disaster elevation values. As a result, it takes several days to weeks to accurately determine the actual sediment displacement area and calculate the volume.
However, the new method employs Artificial Intelligence (AI) to exclude tree cover and other obstacles from the post-disaster LiDAR data, and then compares it with pre-disaster data, making it possible to quickly identify areas erosion.
 
By combining this data with aerial photographs, sediment volumes can be calculated.
For example, in a disaster-affected area of 40 square kilometers (about 15.4 square miles), the conventional method requires five engineers about 12 days to estimate sediment volume. Using this new method, it can be completed in just one day. Because the sediment volume can now be estimated in less than one-tenth of the time required by conventional methods, this drastic reduction in time not only lowers labor costs but also significantly shortens the time to initiate recovery operations.
 
Professor Sato also reported case studies in Aso City, Kumamoto Prefecture, and Kure City, Hiroshima Prefecture – areas severely affected by the Kumamoto Earthquake and Western Japan Heavy Rain. In both cases, the team quickly estimated sediment distribution and volume using DSM generated from Airborne LiDAR data collected immediately after the disasters. When comparing the sediment volumes between the new method and conventional visual interpretation, the accuracy rate was found to be between 80% and 90%.
 
Professor Sato remarked, “We aim to further improve the accuracy by training the AI on a wide range of landslide examples. For larger areas, it works significantly faster, contributing to the swift recovery of disaster-affected regions. “