Following a natural disaster every second matters

Extreme weather events are increasing at an alarming pace. According to research from the United Nations, natural disasters such as flooding and hurricanes nearly doubled during the past two decades when compared to the previous ones. These events can wreak havoc on communities, and in their aftermath, relief agencies must be deployed as quickly as possible to prevent further devastation.

“This is the time where seconds matter—every moment matters,” says Sarah Russell, who leads Project Bellwether at X. “Lives depend on how fast the first responders arrive.”

The National Guard is in charge of coordinating all disaster response carried out by the military in the United States. They play a pivotal, stabilizing role in domestic disaster operations through search and rescue, logistics, and increasingly, damage assessment. Damage assessment is a necessary first step in order to understand the impact of the event, and must be constantly updated as the disaster progresses.

However, the National Guard is saddled with a cumbersome system that can often delay the relief process by hours or even days. Damage assessment is performed manually, with humans meticulously poring over thousands of images captured by planes, searching through databases to find corresponding photographs of precise locations, and noting any changes in infrastructure. The National Guard must wait for every image to be properly analyzed before determining how to coordinate and deploy appropriate resources.

“Right now, our analysts have to spend time sorting through images to find the ones that cover the areas most affected by natural disasters,” said Col. Brian McGarry, who oversees the National Guard’s operations, plans, and training division. “They then have to correlate those images to surrounding infrastructure, label all the relevant features, and only then can highlight the significant damage and send it forward to first responder teams.”

A ‘prediction engine’ for the Earth and everything on it

The Defense Innovation Unit (DIU), a Department of Defense organization tasked with integrating commercial early-stage technology into Defense Department operations, sought a partner to develop tools that might address the inefficiencies in the National Guard’s process. The Bellwether team thought their technology would be a good fit to help tackle this challenge.

“Our moonshot is to systematize that information so that disaster response organizations and other entities can use it to make better decisions and plan for the future.”

Bellwether’s engineers are creating what Russell describes as “a prediction engine for the Earth and everything on it,” building tools that incorporate geospatial data and machine learning to better understand the physical world. Bellwether’s key insight began with the notion that it’s extremely difficult to parse and analyze Earth-related data, even though the development of satellites and sensors has led to the availability of more environmental information than ever before.

“There is so much information from so many sources about the Earth out there,” Russell explains. “Our moonshot is to systematize that information so that disaster response organizations and other entities can use it to make better decisions and plan for the future.”

Image analysis with the help of AI

In partnership with the DIU, the Bellwether team spent nine months building a prototype that uses AI and machine learning to analyze aerial imagery of disaster scenes in mere seconds. After a major weather event such as a hurricane in Texas or a flood in Kentucky, the Bellwether system quickly processes photos of the disaster area and identifies damage to critical infrastructure, using Google’s geospatial assets as a reference for what the area looked like prior to the disaster.

“Using AI and ML to do the routine tasks of georectification, identification, and labeling will greatly speed up how quickly we can get important information to the folks that need it most.”

Once every image of a disaster scene is analyzed, the Bellwether tool produces a labeled map of affected areas so the National Guard can quickly determine how to deploy its resources most effectively. The DIU was so encouraged by Bellwether’s results that they decided to partner with the team on future disaster response efforts.

“Using AI and ML to do the routine tasks of georectification, identification, and labeling will greatly speed up how quickly we can get important information to the folks that need it most,” says McGarry. “It’s all about saving lives in our communities.”

Russell is optimistic about her team’s progress. She noted that their ongoing relationship with the DIU is fruitful due to the agency’s hands-on involvement, with technical expertise in the domain and their ability to bring frontline workers into the process to offer ongoing feedback as the Bellwether team continues to develop its technology.

“The DIU supports our mission to build the most effective tools possible as natural disasters continue to occur with worsening severity,” she says. “In five years, nobody should have to wait to understand the extent of extreme weather damage and the community’s most urgent needs. It should be seen as reasonable and expected that we immediately know the state of the most important infrastructure across a landscape, and what to do next.”