Photo/Illutration A computer-generated image of the Earth observation satellite Daichi-2 (Provided by JAXA)

To improve post-disaster assessment, the Japan Aerospace Exploration Agency (JAXA) is developing a system that uses satellite imagery and machine learning to assess earthquake damage on a per-building basis.

The system is being trained using detailed damage data collected from the buildings wrecked by the twin powerful earthquakes that primarily rocked Kumamoto Prefecture in 2016. 

Approximately 200,000 buildings were either completely or partially destroyed in the disaster on Japan’s southernmost main island of Kyushu.

Damage data from the sequence of quakes was meticulously collected by municipal staff who conducted on-site inspections of each building.

As a result, the dataset is both highly accurate and exceptionally valuable as a large-scale resource. With no tsunami impact involved, it also serves as a rare, undistorted record of damage caused solely by seismic activity.

JAXA has obtained the dataset from Kumamoto Prefecture and is integrating it with satellite images captured before and after the earthquakes to enhance the accuracy of its analysis.

Once the system is fully developed, it is expected to estimate damage to individual buildings within several hours of a disaster. The agency aims to put the system into operation within the next few years.

According to JAXA, its Earth observation satellites—officially known as the Advanced Land Observing Satellite (ALOS) series and nicknamed Daichi (meaning “great land” in Japanese)—are central to the project.

The Daichi-2 and Daichi-4 satellites, which orbit approximately 630 kilometers above the Earth, use sophisticated radar technology to detect surface changes as small as a few centimeters.

Integrating data from international and commercial satellites could further accelerate damage assessments.

JAXA expects the new system to significantly reduce the time required to estimate building damage caused by earthquakes. For instance, during the massive 2024 Noto Peninsula earthquake on New Year's Day, it took the national government and related agencies more than half a day to assess the overall scale of destruction.

A government report noted that the quake occurred near sunset, complicating aerial video analysis. In comparison, the active microwave radars on the Daichi-2 and Daichi-4 enable observation of the Earth's surface regardless of weather conditions or time of day, including during rain or at night.

Shiro Kawakita, program manager at JAXA’s Satellite Applications and Operations Center, expressed strong optimism about the project, stating, “If we can assess the damage quickly, it will lead to faster rescue operations.”

However, in cases such as two-story houses where only the first floor is damaged, satellite imagery captured from directly overhead may misclassify the structure as intact.

To overcome this limitation, JAXA plans to conduct further research, including the use of oblique-angle satellite images to better detect damage to lower floors.

Takashi Kimura, governor of Kumamoto Prefecture—which has entered into a cooperative agreement with JAXA—said, “I hope the lessons learned from the earthquakes can help improve disaster response efforts across the country.”

Related News