The past month's Rim Fire blazed through more than a quarter million acres of land in California, but the tensest moment came on August 27th, when the fire came close to threatening San Francisco's water supply. If the blaze had advanced any closer than it did, it would start dumping ash into reservoirs, spurring a potential public health crisis. Stopping the fire required the forestry equivalent of a goal line stand, dropping three DC-10s worth of retardant in an attempt to push back the blaze.
"YOU HAVE VERY LITTLE ROOM FOR ERROR."
The planes were easy to assemble, but the larger problem was knowing where to drop the payload. According to Russ Johnson, disaster response chief for a mapping company called Esri, firefighters usually target fires as they cross ridges, allowing the retardant will flow downhill and cover more ground. But the threat to the water supply had forced their hand. The fire had to be stopped on a relatively flat plane, and even minor changes in slope would make an immense difference in the payload's effect on the fire. Suddenly, firefighters needed detailed meter-to-meter elevation data on a patch of land that no one had cared about just hours before. "You have very little room for error," Johnson says, "and the consequences can be really devastating."
WHERE ARE THE IMMEDIATE THREATS?
Fortunately, they found the information they needed, pulled from government survey data and routed through Esri’s servers. The payload was dropped, together with brush-clearing bulldozers and efforts from thousands of firefighters, and the fire turned back. It’s a sign of how crucial mapping technology has become in modern disaster response. FEMA still takes the lead providing relief, but a new breed of mapping companies has sprung up to make that relief smarter and more directed, manipulating geographic information systems (GIS) to show responders where their help is needed most. Simply corralling publicly available data can be the difference between containing a fire and watching it slip out of control.
Esri is one of those companies, currently working with FEMA to manage the Colorado floods. The important thing to remember, Johnson says, is that when first responders arrive in a flood-ravaged town, they know very little about what the town was like before. Where are the immediate threats? Where are the hospitals and power plants that will be needed once the waters subside? Working from household income data, responders can steer resources towards poorer neighborhoods where residents are less able to self-evacuate, or identify potential infrastructure failures before they occur.
As long as cell service stays up, Esri collects data from citizens on the ground, who can send in geolocated cell phone photos through Esri’s app infrastructure. "In the old days, you would be allocating new sets of resources on paper every 12 hours, and then you would get radio updates and try to make adjustments." Johnson says. "With GIS and mobile, you can model everything as it happens." And because of the open nature of the data, it can all be published on the web, as the company has done for both the Colorado floods and the Rim Fire.
"IF A MODEL HAS AN ERROR OF LESS THAN 50 PERCENT, IT IS REGARDED AS EXCEPTIONAL."
For wildfires, the same technology could someday be used to go on the offensive, predicting where fires are likely to break out and how existing fires will spread. After the Australian brush fires of 2009, which left 173 dead and more than 2,000 homes destroyed, researchers at CSIRO created the Fire Danger and Fire Spread Calculator to address just this problem, bringing together data on vegetation, land mass, moisture conditions, wind, temperature and countless other data sources to produce a comprehensive look at how wildfires develop. The software is mobile, so firefighters can take the information into the field and calculate on the fly how fast fire will spread in the region they're in, and whether it’s so dangerous that a team needs to be called back.
CSIRO combines that data with lab tests on fire itself, conducted in a massive wind tunnel they’ve dubbed the Pyrotron. Despite the massive data, it’s still a work in progress. "If a model has an error of less than 50 percent, it is regarded as exceptional," says Andrew Sullivan, a fire scientist at CSIRO. But with every well-monitored wildfire, the model improves and the number ticks a little bit higher.
The result is a dance between anticipating wildfires and monitoring them. Civil air patrols send in aerial photos of the blaze while engineers work on predictive models of where it will be hours or days into the future. Just predicting that San Francisco’s aquifers were in jeopardy required tracking wind speeds, burn rates, and dozens of other factors that change from minute to minute. The best data managers can do is build out better reporting tools until, as Johnson puts it, "all these people in the field become sensors." Until we know how to predict where a fire will go, responders and scientists alike will focus on knowing where it is as it burns.
By Russell Brandom on September 21, 2013 10:00 am Email @russellbrandom
"IF A MODEL HAS AN ERROR OF LESS THAN 50 PERCENT, IT IS REGARDED AS EXCEPTIONAL."
For wildfires, the same technology could someday be used to go on the offensive, predicting where fires are likely to break out and how existing fires will spread. After the Australian brush fires of 2009, which left 173 dead and more than 2,000 homes destroyed, researchers at CSIRO created the Fire Danger and Fire Spread Calculator to address just this problem, bringing together data on vegetation, land mass, moisture conditions, wind, temperature and countless other data sources to produce a comprehensive look at how wildfires develop. The software is mobile, so firefighters can take the information into the field and calculate on the fly how fast fire will spread in the region they're in, and whether it’s so dangerous that a team needs to be called back.
CSIRO combines that data with lab tests on fire itself, conducted in a massive wind tunnel they’ve dubbed the Pyrotron. Despite the massive data, it’s still a work in progress. "If a model has an error of less than 50 percent, it is regarded as exceptional," says Andrew Sullivan, a fire scientist at CSIRO. But with every well-monitored wildfire, the model improves and the number ticks a little bit higher.
The result is a dance between anticipating wildfires and monitoring them. Civil air patrols send in aerial photos of the blaze while engineers work on predictive models of where it will be hours or days into the future. Just predicting that San Francisco’s aquifers were in jeopardy required tracking wind speeds, burn rates, and dozens of other factors that change from minute to minute. The best data managers can do is build out better reporting tools until, as Johnson puts it, "all these people in the field become sensors." Until we know how to predict where a fire will go, responders and scientists alike will focus on knowing where it is as it burns.
By Russell Brandom on September 21, 2013 10:00 am Email @russellbrandom
Aucun commentaire:
Enregistrer un commentaire