DAVE WINNACKER stood on a hill in Northern California as flames devoured the houses below him. The Alameda County Fire Department division chief had fought wildfires before, but the 2017 North Bay Fires felt different. A sense of helplessness overcame him as he watched them burn.
"I had engines assigned to them," he says. "But you couldn't stop it."
A former Marine (and now a fire chief in the Bay Area's Moraga-Orinda Fire District), Winnacker was one of more than 10,000 firefighters who battled the infernos, which raged for three weeks. The blaze tore through a quarter-million acres, killed 44 people, and destroyed over 6,000 homes. At times, it spread at a rate of one football field every three seconds. The damage totaled $13 billion, a new U.S. record that would fall the following year.
A few weeks later, Winnacker and his friend Robert Shear, a product manager who spent years developing internet-connected devices, were canoeing on the San Francisco Bay, talking about the disaster. Winnacker felt there had to be a better way to protect people from wildfires. Out on the water, he persuaded Shear to use his Silicon Valley experience to build a solution. What he envisioned felt inevitable. But it was an inevitability that hadn't yet been invented: an "evacuation autopilot" that would use sensors to detect sudden temperature spikes and determine who should evacuate and when.
Shear reached out to Charlie Crocker, a former co-worker at the software firm Autodesk, and the two got to work developing the technology as a side project. The sensors they created functioned as planned but were expensive. Then, the 2018 Butte County Camp Fire killed 86 people, some of whom died after getting stuck in traffic jams and abandoning their vehicles.
At a demonstration of the sensor technology a few months later, a group of fire chiefs asked Crocker if he'd be interested in building something different: evacuation software. At the time, the process for ordering evacuations typically involved fire chiefs leaning on the hood of a car and using markers to draw circles on a map--a century-old approach to solving one of today's most pressing problems. They would relay their recommendation to the police, who would call homes in the area or, if things became urgent, drive around delivering orders through bullhorns.
"It's no surprise when these things don't work out well," says Winnacker. "It should be a surprise when they do."
Crocker and Shear spoke with the fire chiefs about creating map zones to be used across agencies to ensure that roads wouldn't get clogged during an evacuation.
"It was a very simple idea," says Shear, "one where your jaw kind of drops. 'You mean you're not doing this already?' "
The pair took things a step further, envisioning software that would use algorithms to decide how zones should be evacuated by projecting where a fire would move. The software would identify intersections that were likely to become chokepoints and help fire chiefs make informed decisions about where to evacuate and when. Using topographical, wind, and recent rainfall data, it would model a blaze's path one, three, and five hours in the future, helping agencies put certain zones on notice.
In early 2019, the two men worked with Jonathan Cox, deputy chief for the San Mateo Division of California's Department of Forestry and Fire Protection (a.k.a. Cal Fire), and Matthew Samson, deputy chief for the South San Francisco Fire Department, to create a pitch for the San Mateo County board of supervisors highlighting the lifesaving potential of the idea. The board decided to give the two entrepreneurs $70,000 to build a platform, and the following month, in March 2019, Crocker and Shear founded Zonehaven. What had started as a side gig looked like it had the potential to become a sustainable company. Now all they needed were customers.
THE WEST COAST wildfire season that used to run from June until October now begins in May and stretches into January. Eight of the 10 biggest fires in California history have come in the past decade. "Earlier in my career, a fire of 25,000 to 30,000 acres was a major event," says Ian Larkin, unit chief of Cal Fire's San Mateo-Santa Cruz Unit. Last year, California had its first million-acre blaze, the August Complex Fire, which burned from Northern California to Washington with such intensity that the smoke blotted out the sun, turned the sky orange, and reached as far as New York City. A record 4.2 million acres were scorched in California alone in 2020, an area the size of Connecticut. Experts estimated direct damages at up to $20 billion.
Amid all that, some Silicon Valley entrepreneurs feel they're the only thing standing between climate change and the charred landscape of tomorrow.
Zack Parisa, co-founder of SilviaTerra, based in San Francisco, is one such entrepreneur. His startup's software creates maps of the forest, outlining the average ages and types of trees in each acre. "Entire towns here wonder if they're going to be wiped from the map," says Parisa.
SilviaTerra's platform can prioritize where vegetation should be removed--an essential component of wildfire prevention. Other startups offer A.I.-enabled fire detection, virtual 3-D cartography, risk-analysis systems, electrical-malfunction defenses, robotic-ball sensors, home rebuilding apps, and more. The list is robust, and growing.
But, as Parisa and many others have learned, persuading the government to pay for such technology is a tough sell.
"The need is super apparent, and the meetings are abundant, but there isn't the action that people want or expect," Parisa says. "It's exceptionally frustrating."
WHEN BROTHERS Emrah and Hakan Gultekin arrived in the Bay Area in 2016 after emigrating from Turkey, they decided to use their artificial intelligence expertise to build Chooch AI, a computer-vision platform to monitor medical procedures and manufacturing processes. After the Camp Fire, the Gultekins turned their tool on a new scourge: wildfires. They fed Chooch's model thousands of photos and videos, and soon the A.I. was able to recognize fire with 95 percent accuracy.
Early detection is critical when it comes to wildfires. Land can burn for days before humans notice, often making fires impossible to control by the time firefighters arrive. Chooch, however, could detect the first signs of fire without the need for human involvement. Emrah Gultekin says the A.I. is even more effective when used with heat-sensing infrared cameras and satellite imagery, the latter of which has become significantly less expensive in recent years. Still, the startup has failed to secure any contracts with government agencies for wildfire detection.
"We've done everything we can to train the A.I. to be a fire expert," Gultekin says. "It's up to them to actually deploy it."
San Francisco-based Enview knows the feeling. The A.I. startup builds 3-D terrains to determine where trees and vegetation might pose a risk to physical assets, but so far its clients are mostly in national security and energy, not wildfire prevention.
Squishy Robotics, led by UC Berkeley mechanical engineering professor Alice Agogino, manufactures a robotic ball that can be dropped into a blaze to take critical temperature and wind readings that can assist first responders. The company has earned contracts with small fire departments on the East Coast, but none in wildfire-prone areas.
The irony is not lost on the Bay Area firefighters. "Across the freeway, they're curing cancer," says South San Francisco FD's Samson, "and we just moved away from paper timecards."
And with the lack of a market comes a lack of funding. After surviving a 2014 flood in his home village in India that killed more than 500 people, Ahmad Wani launched the Menlo Park, California, disaster preparedness company One Concern in 2015. He built a mapping platform that would allow governments to run disaster simulations to reveal where infrastructure and lives were most vulnerable. Raising just $25,000 took eight months of pleading with investors.
"As soon as you go to a venture capitalist and say the words 'disaster' or 'government,' they say, 'You are a disaster. There's the door,' " Wani says. After targeting financial clients instead, he quickly secured a $23 million Series A round.
Wani's story isn't unique. Many of the startups that have found success selling wildfire tech have done so by marketing to private companies seeking protection from the wildfire problem. Jupiter Intelligence, which assesses vulnerability to climate change, recently raised $40 million in venture capital to run risk analyses for clients in sectors from agriculture to finance to real estate. Gridware, a startup that makes hardware and software designed to detect electrical malfunctions that can cause wildfires, recently won an RFP from Pacific Gas & Electric, which has paid $13.5 billion in settlements for its role in causing some of California's most destructive wildfires.
Other startups focus on consumers affected by natural disasters. Santa Rosa-based Homebound helps people who have lost homes rebuild. Tech industry veterans Jack Abraham and Nikki Pechet launched the company in 2018 after the Napa fires destroyed Abraham's house. Homebound hires contractors, navigates insurance claims, and schedules inspections, all through an online portal. Drones let homeowners monitor construction remotely, helping reduce what can be a multiyear process to an average of nine months.
Two years after launching, Homebound has 250 homes finished or in progress, many of them in the wake of wildfires in Santa Cruz and Napa and Sonoma counties. The company helped the Santa Rosa neighborhood of Coffey Park rebuild after the 2017 Tubbs Fire reduced it to a barren terrain of half-crumbled chimneys and burned-out cars.
"It is apocalyptic after one of these fires," says Pechet. "What we're seeing with wildfires, what we're seeing with climate change--these are all problems that Silicon Valley is great at tackling."
BACK IN 2019, in the months after forming Zonehaven, Crocker and Shear spent hundreds of hours collaborating with fire departments in Northern California. Those conversations led to the creation of virtually every feature they built and ultimately helped them find their market. "They haven't tried to sell us on what the solution is," says Cal Fire's Cox. "They've been partners by our side, saying, 'OK, how do we solve this?' "
In early 2020, San Mateo County signed a three-year contract with Zonehaven worth $294,000. Santa Cruz followed with a $120,000 two-year deal of its own. In all, more than a dozen counties in California came on board last year, including Sonoma, Napa, Marin, and Alameda.
Zonehaven's traction in the wildfire space remains the exception, not the rule, helped along by the fact that one of its founders happens to be friends with the local fire chief. So far, the startup's growth has been incremental. Its largest clients are county governments, none of which have paid for the service the same way: Some have freed up funds for multiyear commitments within their annual budgets, while others have received grants that paid for a single year. The cost per county typically falls between $50,000 and $100,000 a year, depending on the population.
The hope is that eventually, states like California will pay for and roll out the solution en masse. "That's really what needs to happen," says Samson. "Right now, it's hit or miss based on your county and whether it has money to try to fund something."
In August of last year, with Zonehaven six weeks away from its launch, an overnight thunderstorm sparked hundreds of fires across San Mateo and Santa Cruz counties. The next morning, Crocker's phone rang. It was a local fire chief.
"How soon can we get this thing up and running?" the chief asked.
Twenty minutes later, Zonehaven pushed its platform live. The firefighters pinpointed the locations of the fires on the map using the reports they'd received from the field. The algorithms got to work, the chiefs selected which zones to evacuate, and notices went out through Facebook, Twitter, and emergency text alerts. The messages directed residents to a live online map on the startup's website--which wasn't yet ready for the influx of visitors, causing its servers to temporarily crash. But service was quickly restored, and nearly 80,000 residents were evacuated safely. And though one person died in the fires, the response was deemed a success; similar-size blazes have led to many more fatalities in recent years.
"I think back to the Kincade Fire in 2019," says Cox. "We spent over 24 hours trying to put together an evacuation plan. On this incident, we were getting people out within a matter of minutes."
The truest test yet will come this summer, when--sadly and inevitably--wildfire season begins again. Meanwhile, Crocker, who serves as Zonehaven's CEO, and Shear, its board director, hope their now-22-person company can continue improving its fire-modeling and emergency communication capabilities in the coming months. The startup recently signed a deal with the GPS navigation app Waze to include evacuation notices. Alerts could also be pushed to Amazon Alexas or Google Homes to wake sleeping residents, while future smart homes could give permission to local authorities to turn on sprinklers or prevent HVAC systems from circulating smoky air. And the co-founders haven't ruled out a return to the early-detection sensors--if they can make the economics work. Should Zonehaven turn out to be a winner in the wildfire tech space, it could one day be one of many, given the millions of lives and billions of dollars at stake.
"We exist at the intersection of climate change, dysfunctional government, and technology," says Shear. "We all know we're not going to rely on our government to save us from climate change. The solution has to be community-led. It's up to us to solve this."