How tech, AI can help with early fire de­tec­tion

The Fresno Bee - - News - BY LEVI SUMAGAYSAY

The fast-mov­ing Camp Fire caused traf­fic grid­lock as Butte County res­i­dents tried to evac­u­ate Thurs­day, lead­ing some of them to aban­don their ve­hi­cles as the fire grew closer to the road­ways.

Had the now-deadly fire been de­tected ear­lier and of­fi­cials been able to give more no­tice, some res­i­dents might have avoided the dan­ger­ous sit­u­a­tion of such a nar­row es­cape.

How far can tech­nol­ogy go to help au­to­mat­i­cally de­tect and even pre­dict wild­fires? From coast to coast, com­puter sci­en­tists, re­searchers and oth­ers are hop­ing to make it go far­ther.

Early de­tec­tion is key as wild­fires have got­ten worse in re­cent years, ac­cord­ing to Jim Craw­ford, as­sis­tant chief of South Bay op­er­a­tions for Cal Fire’s Santa Clara unit, who has been a fire­fighter for 26 years.

“The job I signed up for 26 years ago is not the same job to­day,” he said. “In the 1990s, fire­fight­ers bat­tled a nor­mal large fire for a cou­ple of weeks. Any more than that was con­sid­ered a ‘ca­reer fire.’ Now we’re hav­ing ca­reer fires mul­ti­ple times a year.”

The Camp Fire – which as of Thurs­day night had burned more than 20,000 acres and de­stroyed thou­sands of struc­tures – is one of more than 5,500 wild­fires that have burned more than 640,000 acres in Cal­i­for­nia so far this year.

Fire agen­cies mostly find out about fires from 911 calls – which is what hap­pened with the Camp Fire, ac­cord­ing to Cal Fire Capt. Scott McLean. They also mon­i­tor cam­eras placed through­out the state, in­clud­ing on undis­closed Bay Area moun­tain­tops. Some of the cam­eras are aided by de­tec­tion al­go­rithms and satel­lites, but the sys­tems are not quite real-time and the res­o­lu­tion could be bet­ter, ex­perts say.

“Lots of peo­ple are look­ing into us­ing ar­ti­fi­cial in­tel­li­gence to fig­ure out fire lo­ca­tion, spread and be­hav­ior,” said Craig Cle­ments, as­so­ciate pro­fes­sor at the Fire Weather Re­search Lab­o­ra­tory for the Depart­ment of Me­te­o­rol­ogy and Cli­mate Sci­ence at San Jose State Univer­sity.

For ex­am­ple, there’s the Geo­sta­tion­ary Oper­a­tional En­vi­ron­men­tal Satellite (GOES) Early Fire De­tec­tion Sys­tem, a col­lab­o­ra­tion be­tween the Cen­ter for Spa­tial Tech­nolo­gies and Re­mote Sens­ing (CSTARS) at UC Davis and the U.S. For­est Ser­vice. It uses Na­tional Oceanic and At­mo­spheric Ad­min­is­tra­tion weather satel­lites and fire-de­tec­tion al­go­rithms.

CSTARS sci­en­tist Alex Koltunov, who has a Ph.D. in re­mote sens­ing, has been work­ing on and off on the sys­tem’s al­go­rithms – sets of rules pro­grammed to per­form cal­cu­la­tions or carry out cer­tain tasks – for the past sev­eral years, with the goal of op­ti­miz­ing data pro­cess­ing to help de­tect ig­ni­tions as early as pos­si­ble.

He and his col­leagues have been us­ing GOES-15, a satellite that was launched in 2010, to con­duct test­ing us­ing past im­ages and wild­fire-in­ci­dent re­ports over many states to help fine-tune the al­go­rithm, which he hopes to one day ap­ply to GOES-16 and GOES-17. Those satel­lites, launched in the past cou­ple of years, can pro­vide higher-res­o­lu­tion im­ages “every 5 min­utes or some­times every minute,” Koltunov said. Higher-res­o­lu­tion im­ages should al­low fire­fight­ers and oth­ers to de­tect wild­fires more quickly and ac­cu­rately.

But he pointed out that no sin­gle tool can be a “sil­ver bul­let” when it comes to de­tect­ing wild­fires.

“Each method has its lim­i­ta­tions,” Koltunov said. When it comes to al­go­rithm-based de­tec­tion, for in­stance, “how good is the data?”


Fire­fighter Jose Corona sprays wa­ter as flames from the Camp Fire con­sume a home in Ma­galia on Fri­day. Ex­perts say their best tool for catch­ing fires early is still 911 calls.

Newspapers in English

Newspapers from USA

© PressReader. All rights reserved.