Artificial Intelligence to Improve Proactive and Reactive Wildfire Response

This article first appeared in Medium (here) and was published here with permission from the author

Climate change has increased the frequency and intensity of wildfires in recent years. With the increased threat of wildfires, firefighters and other first responders need improved or new tools to better understand the extent and behavior of wildfires so that they may plan their response accordingly. This year, CrowdAI was selected as the exclusive technology partner with California Air National Guard and Joint Artificial Intelligence Center (JAIC — a division of the US Department of Defense) to build a custom solution to help stop the spread of wildfires.

Before CrowdAI, analysts would manually comb over hours of video and point-and-click to draw a fire perimeter, often while working around-the-clock shifts. Depending on the size and extent of the fire, this process can take hours to complete — even for just a single section of the fire perimeter. Wildfires can sometimes spread as fast as 10 mph, so by the time an analyst can manually pinpoint the perimeter, the fire has spread.

This is where automation powered by CrowdAI can provide a force multiplier.

The AI-enabled solution involves flying unmanned aircraft equipped with Full Motion Video sensors over the wildfire regions. The AI computer vision algorithm automatically analyzes the video data, detects regions with active wildfires, and provides near real-time fire location updates to a digital map application on first responders’ mobile devices.

Speed and accuracy are both critical to response efforts. Machine learning algorithms and a more automated process can be used to provide first responders with more accurate information about the fire perimeter more often and with higher accuracy. Near-real-time updates of the fire perimeter are needed to inform the public and to coordinate response efforts. Accuracy is also important, especially as it relates to the precise location of the fire perimeter.

A detailed peer-reviewed paper on this use case can be read here.

CrowdAI created a similar solution in the aftermath of the Camp Fire in California. CrowdAI used imagery from after the fire to detect structures, comparing those structures to imagery from before the fire.

Insurers are also adopting this technology to target high-risk insured assets in fire hazard severity zones and proactively contact the homeowner to remove vegetation within close proximity of their property.

In the 1970s, the average length of the wildfire season was five months. Today, it lasts more than seven. 2017 was the costliest year on record for insurers, with total U.S. wildfire peril totaling $16bn in losses.

As of October 27, 2019, over 6,100 fires have been recorded so far this calendar year, according to CAL FIRE and the US Forest Service, totaling an estimated 198,800 acres of burned land. Fire behavioral experts and climatologists forecast that heavy rains from earlier in the year have produced an excess of vegetation that may become a tinder box later in the year as the fire season gets underway.

CrowdAI is committed to building technology that is used to make the world a better place. In keeping with that commitment, CrowdAI works on pro bono projects with nonprofits administering disaster relief.

Organizations send CrowdAI individual addresses of their assets and CrowdAI sends back relevant pre and post-disaster insights.

About Richard Purcell

Head of Commercial Sales, Founding Team for

Leave a Comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.