AI technology can aid rescue efforts during natural disasters, experts say

Satellite imagery can play a key role in identifying the hardest-hit areas

Powered by automated translation

Space technology and satellites can be used to prioritise the worst-hit areas during natural disasters and help rescuers save lives on the ground, a space conference heard.

Although the AI technology used to scour thousands of images from disaster zones is in its infancy, experts say the applications have huge potential.

Speakers at the 74th International Astronautical Congress in Baku also said the latest satellite imagery and related software would be helpful for hard to reach areas or regions plagued by conflict, providing usable images in hours rather than days.

An image showing building damage, or a building that isn't damaged, can hint to rescue troops on the ground where to send help
Rene Griesbach, pre-sales regional manager for Planet Labs, a satellite development company

The earthquake that hit Turkey and Syria in the early hours of February 6, 2023 claimed the lives of more than 40,000 people and displaced 100,000 others.

The disaster has been used as a case study on how to better utilise space technology in the hunt for survivors when similar disasters happen in the future.

The United Nations Satellite Centre (UNOSAT) provides an emergency mapping service and satellite image analysis following major disaster events, complex emergencies and crisis situations.

Luca Dell’Oro, chief of the Disaster Risk Management and Climate Resilience Section at UNOSAT, said satellite imagery can play a key role in identifying the hardest-hit areas.

“In Turkey, the co-ordination response was mainly led by the National Disaster Management Authority. But in Syria, the situation was much more complex,” he said.

Mr Dell'Oro said that it was more challenging because of a lack of national capacity to co-ordinate response, and there was a huge issue with access.

“There had been 20 years of conflict and right before the earthquake, there were 15 million people in need of humanitarian assistance and 7 million internally displaced people," he added.

“We had been asking mainly to focus on Syria to estimate the level of damage, especially because of the lack of information in this area.

“Once we collected the images, we were able to assess areas that were hard to be reached.”

Earth observation technology

2023 has been a particularly deadly year for natural disasters so far, with floods in Libya and China, earthquakes in Morocco, Syria and Turkey, as well as wildfires across Europe and tornadoes in the US.

Earth observation technologies include image processing, early warning systems, remote sensing applications, artificial intelligence and communication practices to predict when natural disasters may occur.

Information sharing and co-working between governments and the private sector can play a vital role in maximising that chance of finding survivors during a disaster.

However, experts said more work is required to fine tune existing technology.

“Applying AI to imagery is a great future technology, but we are at the very beginning of that development, especially with regard to the imagery for earthquake response,” said Rene Griesbach, presales regional manager for Planet Labs, a satellite development company.

“We have been in close co-operation with Microsoft and its project, AI for Good.”

Microsoft's AI for Good Lab is a research and data visualisation hub that harnesses big data and cloud technology.

Planet Lab uses Microsoft's computing capacity to extract satellite images that can help support rescue efforts during a natural disaster or war zone.

"In this case, an image showing building damage, or a building that isn't damaged, can hint to rescue troops on the ground where to send help," said Mr Griesbach.

“We have to focus their activities too.

“This technology was actually developed for other crises in Ukraine, but then could be easily adapted for other natural disaster cases such as the earthquake in Turkey."

Updated: October 05, 2023, 3:47 AM