How did AI and satellites assist in disaster relief efforts during the Myanmar earthquake?

Following the earthquake in Myanmar on March 28th that killed thousands, Microsoft deployed support by combining satellite imagery and AI to identify areas in need of urgent relief.

Map showing damage assessment in Myanmar after the earthquake. Photo: Microsoft

On March 29th, Planet Labs (PBC) captured satellite images of the earthquake-affected areas as soon as the clouds cleared and sent them to Microsoft's AI for Good lab. At Microsoft headquarters in Redmond, Washington (USA), around 11 PM on March 28th, a team of experts was ready to receive the images and use AI to analyze the damage, identifying collapsed and severely damaged buildings.

Before today's technological advancements, assessing damage after disasters like earthquakes and floods relied on on-site analysis. This method was time-consuming, taking days to weeks. While it provided detailed data, it couldn't meet the demands of urgent action.

These are images taken by Planet Labs PBC satellites before and after the earthquake in Myanmar. (Photo: Planet Labs PBC)

Microsoft has shared how its AI analyzes the extent of destruction from high-resolution satellite imagery. The system uses convolutional neural networks (CNNs), a type of supercomputer specializing in visual data processing, to compare images before and after a disaster. The model then applies a scoring system to quantify the severity of the damage, classifying areas into categories such as “no damage,” “minor damage,” “partially destroyed,” and “completely destroyed.”

Microsoft emphasizes that, in the context of disaster response, the value of automated damage assessment lies in the speed of inference rather than absolute accuracy. Therefore, this time, the research team developed a custom model specifically for Mandalay, because, according to Microsoft's lead data scientist Lavista Ferres, "The Earth is too diverse, natural disasters are too varied, and satellite imagery is too different to use a single model for every situation."

According to local media, Mandalay was the hardest-hit area in the March 28 earthquake. AI analysis showed that 515 buildings there suffered 80% to 100% damage, while approximately 1,524 others sustained 20% to 80% damage. Other severely affected areas included Yangon and Minkun.

This isn't the first time Microsoft's AI for Good lab has used AI to assess damage. In 2023, the team tracked devastating floods in Libya to predict risks and aid recovery efforts; analyzed damage from a major earthquake in Turkey in March; and assessed the impact of wildfires in Maui in August, analyzing more than 2,810 buildings.

Currently, rescue efforts in Myanmar are progressing slowly due to power outages, fuel shortages, communication disruptions, and landslides cutting off many areas. The lack of modern equipment is also hindering search and rescue operations, forcing many people to dig by hand in harsh, hot weather conditions.

(According to India Today, theglobeandmail)

Scenes of devastation in Myanmar after the earthquake: In Sagaing, the epicenter of the earthquake in Myanmar, authorities are struggling to bury thousands of victims who died in the disaster on March 28.

Comment (0)