Former FEMA Admin.: AI Can Marry Urban Planning and Disaster Response

Former FEMA Administrator Craig Fugate is now with a company called One Concern, whose existing tools for earthquakes and floods, and a third in the works for wildfires, create highly detailed maps of possible damage.

by / January 15, 2019
TNS

Survivors are still reeling from 2018’s natural disasters as the numbers are coming in: Global losses from disaster last year were 11,000 people dead or missing and $155 billion in damages, according to Swiss Re Group, the world’s second-largest reinsurer.

Estimates were 10,000 dead or missing and $350 billion for 2017, when three record-breaking hurricanes swept through the west Atlantic in the span of a month, and scientists only expect figures to get worse as climate change progresses. A study published Monday by the National Academy of Sciences showed Antarctica’s annual ice loss has increased sixfold since 1979, melting faster in each successive decade.

Tech companies that study data and make government software have taken notice, and a handful of platforms already offer ways to help governments prepare for natural disasters — Hazus is the hazard modeling tool already used by the U.S. Federal Emergency Management Agency, Esri is adding data analytics and AI to its GIS platform, and Denver-based Geospiza maps at-risk populations.

But the company that has the attention of former FEMA administrator Craig Fugate is One Concern in Menlo Park, Calif.

Talking to Government Technology last week about the future of disaster response, Fugate, as One Concern’s chief emergency management officer, said the most pressing need is for governments to transition from being reactive to proactive. He said land use and building codes will be instrumental in mitigating future risk, but until recently, no one had both the data and computing power to predict impacts of natural disasters in enough detail to recommend precautions.

“How and where we build our communities has the single biggest outcome for success in disaster, not response. And we lacked the tools to accurately display these risks in a way that we can show how bad it could be, but also what we could do differently to change that outcome,” he said. “Forecasting 36 inches of rain and issuing a flood warning is (not enough). How many of my substations are going underwater and I’m going to lose power because of that? That’s what we can answer.”

Founded in 2015 by Stanford alumni, One Concern has two tools on the market: “Seismic” to assess the near real-time impacts of a live earthquake, and “Flood” to predict potential impacts of a flood up to five days in advance. A third product for wildfires is due for release later this year. Each can run tests based on criteria — an analysis of an earthquake’s impacts on the local health-care system, for example, will include whether utilities and access routes to local hospitals have been impacted.

“This gives us the ability to go, ‘What could happen, and how does our plan respond to that?’ I’d much rather find out what’s not working than to wait for the real event,” Fugate said. “If we wait until the event, I think we missed the opportunity to change the outcome.”

Described by spokesman Ben Colombo as an “AI-native” company, One Concern’s innovation is in its combination of AI with data harvesting, cloud computing and the physical science of earthquakes, floods and fires. He said the company collects data on streets and structures from a variety of proprietary and open sources around the world, including governments, insurance companies, satellites and even crowdsourcing. And it’s constantly being updated.

Only recently, with the advent of neural networks and cloud computing, has it been possible to analyze that volume of data in a single simulation. Fugate said the results are faster and more accurate than ever, creating a higher-resolution model of disaster outcomes than any he’s seen.

“We look at over 160 variables per structure. We’re down to blocks … In some cases we’re not even sure how it’s coming up with the answer anymore, because the neural networks are so deep,” he said. “If I know which block got heavy impacts, I also know who lives there or what kind of structures or businesses are there, and one of the tools we’ve been building for emergency managers … shows not just a red area, but how much your population over 65 have been impacted.”

Besides data and bandwidth, Fugate said, one of the obstacles to the success of similar tools in the past has been the emergency management market itself. He said the only way these products become viable is if they’re sold not as a response tool for responders but a resilience tool for governments. That means positioning One Concern’s platform as a tool for saving lives, costs and infrastructure, and even for urban planning.

“From my time at FEMA, if we understand that in the really big disasters, government is going to fail by itself — you have to work as a team — then how do we provide tools in our platform that helps a community build resilience that just doesn’t stop with government?” he said. “We’re looking at how does the private sector engage in this? How do you get this down to individuals?”

A spokesman said the company has 10 clients, a majority of which are local governments in west coast cities such as San Francisco, Los Angeles and Cupertino, Calif. They’re in the process of executing a statewide contract in the southwest and have finalized another for a major developing city in Asia.

Andrew Westrope Staff Writer

Andrew Westrope is a staff writer for Government Technology. Before that, he was a reporter and editor at community newspapers for seven years. He has a Bachelor’s degree in physiology from Michigan State University and lives in Northern California.


Platforms & Programs