Thursday, November 19, 2015

An Earthquake Response System That's Faster Than 911

A gas fire following the 2014 Napa earthquake destroyed homes at the Napa Valley Mobile Home Park. (AP Photo / Ben Margot)

by Julian Spector:

When a magnitude 6 earthquake shook Napa County in August 2014, thousands of calls came pouring in to 911, reporting everything from minor injuries to a gas fire. It was up to a small staff of dispatchers to process those calls and work with the local fire chief to figure out who needed saving and in what order.

The quick thinking of that emergency response team limited the fallout from the quake—the Bay Area’s largest since Loma Prieta in 1989—which killed one person and injured more than 200. But the technology at their disposal was largely reactive: wait for people to call in, figure out who probably needs the most help, send responders there. A fiercer quake in a more densely populated area would stretch the limits of that approach.

That’s why Ahmad Wani started One Concern, a small tech company whose mission is to overhaul the way we respond to disasters. When an earthquake hits, instead of going down the list of 911 calls or using a dispatcher’s intuition, emergency response teams can use One Concern’s algorithm to quickly identify the most likely hotspots for building damage and threatened lives and send help to those sites first. San Mateo County in Northern California has already adopted an early version of the program, and San Francisco and Alameda Counties are slated to do the same in the next month.

“We use technology to build smartphones and cool gadgets, but we have not used technology to save lives during a disaster,” Wani tells CityLab. “We should be looking into that, because we’re talking about lives here.”

Know the damage before you see it

The goal of the One Concern algorithm is to predict how much damage a building will suffer, because that’s a useful indicator of where people will be injured and need help. The algorithm analyzes how buildings behaved in past quakes and supplements that with data on soil, nearby water bodies, structure ages, recent building improvements, and more. This generates a heat map of buildings more likely to be damaged than others.

The program then compares the heat map with data on population and occupancy, based on time of day, to create a list of priority response destinations. For instance, an elementary school wouldn’t be high on that list if an earthquake struck at midnight, but if it hit at 10 a.m. on a school day it might jump to the top. The program can even send first responders with a briefing on how many people and what kinds of damage to expect.

That prediction happens in a matter of minutes and is highly accurate at the city-block level, Wani says, meaning it might miss some damage but it’s far better than an ad hoc response. In the future, response crews roaming the city might be able to use a One Concern mobile app to upload ratings of damage and photos of buildings, which would be geotagged and used to optimize the algorithm.

“The more actual data points and pictures I get, the better for the algorithm,” Wani says.

“It’s not that they didn’t want to rescue us”

Wani came to Stanford in 2013 as a graduate student to study earthquake-resistant design, but his focus changed in a way he couldn’t have predicted. Last summer he traveled home to Kashmir, India, for a nice visit with his family, but around 3 a.m. the next morning, nature intervened.

The Kashmir floods, triggered by days of monsoon rain, displaced almost a million people. “Within 15 minutes, three floors of my house were under water,” he recalls. “We were stuck in the flood for up to 7 days. I don’t know how we survived, it was bad.”

He and his family got by on some chips and wafers they’d grabbed on the way up, and apples that their neighbors tossed them from the next roof over. They saw helicopters puttering through the washed-out landscape, but for some reason they never came to the rescue.

”After the seventh day the water receded and on the eighth day morning the first helicopter stopped over our heads and it threw one loaf of bread and that was the rescue,” Wani says. “I was shocked.”

Wani flew back to Stanford a few days later, still grappling with that sense of abandonment and wondering why flood survivors were left to fend for themselves. Back in California he learned that disaster had struck this region, too, in the Napa earthquake, and that’s when he says he had a revelation of sorts. “It’s not that they didn’t want to rescue us,” he says. “The problem is, there are so many people to be rescued, how do they prioritize? They don’t know where the damage is. It’s a whole city.”

Minutes matter

Wani started to work on the prioritization problem in a computer-learning class taught by Stanford’s Andrew Ng, a leader at Coursera and Baidu, the Chinese-language search engine. What began as an academic research project grew into a mission to equip local governments with a better tool for planning their responses to earthquakes. One Concern now has seven members tackling that question.

Machine-learning holds a lot of potential for disaster response, but so far it’s been limited by the number of people with the skills to use it, Ng writes in an email to CityLab. “In disaster response, minutes matter. Intelligent software that gets first responders to the most urgent sites will save lives.”

Emergency responders themselves are excited about the possibilities, too. Don Mattei, the Office of Emergency Services supervisor at the San Mateo County Sheriff’s Office, says the program offers better situational awareness immediately after an earthquake. Previously, emergency response centers would piece together information about the damage based on 911 calls and field reports. With One Concern, if an earthquake hits San Mateo, “Now I can look at it and say, ‘Here’s where I think the damage is,’” Mattei says. “You’ll have the ability to look at things and form a strategy about how you’re going to deal with it.”

That’s useful both for the coordinators at the emergency operations center and for a patrol sergeant out in the field, Mattei adds. And if One Concern shows there’s more damage in a neighboring town, the local force can prepare to send backup more quickly.

One Concern plans to look at floods in the future, analyzing vulnerable parts of cities based on weather predictions to help plan for inundations. But that’s a long way off. For now, Wani and company are hard at work scaling up their service for San Mateo, so if an earthquake does strike, the county’s emergency response coordinators will know exactly where they need to go, even before the 911 calls come in.
You may also like:

No comments :

Post a Comment