Wildland firefighters across the West now have access to a large network of cameras monitored by artificial intelligence to alert them to wildfires when they start. The University of Oregon’s Oregon Hazards Lab, along with collaborators at other universities, has partnered with the company ALERTWest to monitor the cameras 24/7. When an algorithm detects smoke or fire, an operations center will be alerted, and dispatchers will then alert the appropriate fire manager. Doug Toomey, director of OHAZ, will join us to discuss how this system will help in the fight against wildfires.
Note: The following transcript was created by a computer and edited by a volunteer.
Dave Miller: This is Think Out Loud on OPB. I’m Dave Miller. Wildland firefighters across the West now have access to the largest public network of fire lookout cameras in the world and they don’t need to be monitored by humans. They’re connected to AI that monitors the feeds 24/7. When the algorithm detects smoke or fire, an operation center can be alerted and dispatchers can then call in the appropriate fire managers. This is a public/private partnership spearheaded by the Oregon Hazards Lab, or OHAZ, at the University of Oregon. Doug Toomey is the director of the lab and he joins us now. It’s good to have you on the show.
Doug Toomey: Yeah, thanks very much. I’m pleased to be here.
Miller: Can you give us a sense for the scale of this camera system?
Toomey: Yeah, the camera system is owned and operated by three separate universities in the West: the University of Nevada, Reno, the Alert California System at the University of California, San Diego and the University of Oregon. It’s currently in about six to seven Western states and expanding. There’s over 1,200 cameras, the vast majority of those presently in California, but expanding in Oregon and the Northwest and other states as well. And then we’ve been working with a public/private partnership with a company called ALERTWest where the public can view the cameras and where firefighters can have, through credential logins, command and control the cameras. And as you said in your intro, they have an AI detection algorithm to alert when an anomaly is detected.
Miller: Not that long ago, we heard about a short-lived fire lookout tower that was at the very top of Mount Hood – that was in [the] 1920s and ‘30s. Could you just take us back a little bit in history and give us a sense for how this basic idea of getting up high, so you have a good view and you can see where there are fires, has evolved over the years?
Toomey: Yeah, that’s a great question, of interest in my family, in particular. My grandfather came to Oregon with Frank Gilchrist. So people might know where Gilchrist, Oregon is located, to start the Gilchrist Lumber Mill. My father was on a fire lookout tower in his high school years at Odell Butte in Central Oregon. We probably had more man fire lookout towers when my father was a young man than we currently have cameras in Oregon. So we need to change that. It’s very important to have eyes on the landscape to detect and locate fires and to be able to call in resources. And we did that fairly extensively, and my father did it and others, 50, 60, 70 years ago.
Now, we can do that both with manned and unmanned lookout towers. We have a location in Oregon, on Green Mountain, that is both manned and has a camera as well. So that provides some redundancy in terms of 24/7 coverage. And I think we’re in a situation now where we can expand these cameras to augment what humans have been doing for quite some time.
Miller: AI is notoriously weird. I mean, it doesn’t, for example, know what a hand is, the way humans do. We feed it millions of pictures of stuff and we say “this is what you’re looking for, this is not what you’re looking for.” And over time, in a totally alien, nonhuman way, [it] starts to get better at what we tell it we want it to know. How good is it right now at knowing what a wildfire is and what other smoke from a campfire is, or fog or mist or clouds or just a smudgy looking house?
Toomey: Yeah, absolutely. It’s important to realize that in the system, there is a human in the loop. So it’s detecting anomalies that goes into an op center that has personnel there 24/7, and that personnel makes the final decision. So they take a look at the anomalies. They can train multiple cameras, if there’s enough cameras in that view shed area, to see what that anomaly might be. Then they determine, OK, is it a cloud, a smudge or is it a fire? It’s been very effective.
The alternative, having people 24/7 look at 1,200 cameras now, but it will probably be tens of thousands in the future throughout North America, that’s problematic. So we need to be able to use algorithms on computers to highlight an anomaly and then have a human in the loop that makes that decision. Think of it as an easy way to have your attention drawn to something and then have a person verify what that is.
Miller: Do you have a sense for the hit rate right now? I mean, when the algorithm identifies an anomaly, how often is it that humans actually think it’s a serious enough situation that it was worth flagging in the first place?
Toomey: It’s been proven very productive in California over the last year or two. And that’s where it originated and tested. And then it expanded through the University of Nevada, Reno, University of Oregon – a collaboration we have with UCSD. In California, most fires historically, in the past decade or so, have been reported by 911 calls. We have situations now where fire anomalies are being detected by AI. The op center is sending out information to the fire companies and people are on site fighting that fire before the 911 call comes in. So it’s been proven to be very, very effective.
I can’t give you numbers, but that’s actually an excellent question though. How many false detections are there versus positive? But we already know that the firefighting community in California sees it as accelerating their initial attack on fires and making it much more efficient to get there.
The other aspect of these cameras, it’s not simply about detection and we’re all fascinated by this AI concept. And so we think it’s just about the detection, but once a fire starts, it’s important to know how it’s developing. It’s important to know the fire perimeter. For example, groups can make predictions of fire spread over the next hour to 12 hours if something’s happening. They can know something about the fuels on the ground, the met conditions, which direction that fire is going to go. And that helps the firefighting community respond in terms of where they should position their assets. In fact, I would say the cameras are probably much, much more useful after ignition because it provides a situational awareness for the firefighting community that keeps them safe and advances their activities.
Miller: Oh, as opposed to a person in a tower who has to flee because the winds are such, say, that there’s a chance that where they are could be engulfed in flames? The camera doesn’t have that liability.
Toomey: It doesn’t have that liability. We also don’t have those people in the towers that much anymore.
Miller: Right. I mean, yes, in that world there were thousands of people who could just hang out at the tops of mountains all the time, but that’s no longer at all feasible.
Toomey: No, that’s not feasible.
Miller: Who has access to these camera feeds? I mean, I should say that I looked at your website and I saw there’s a gigantic clickable map where I could poke around and look at, whether it was ODOT cameras or mountain lookout camera footage, all around the West. Does everybody have complete access to all of the footage from these cameras?
Toomey: They have access to what’s being seen on that camera right now and the time lapse over the last 15 minutes to 12 hours. And so as you know, there’s a website, it’s ALERTWest.live. As you noted, you can see mountain top cameras, that’s one symbol and those cameras have full pan tilt zoom capabilities. They go near infrared at night. There are also other cameras – airports have aviation cameras, ODOT has TripCheck, and those are also integrated into the system. Now, those are fixed cameras with a fixed view and you can’t move them or control them. They do provide a view that you can run an algorithm on and decide, is there an anomaly to be detected? So the public can go to that website anytime they want to and it’s fun to go there. You can see beautiful sunrises.
When the northern lights occurred a week or two ago, we had panoramas of the northern lights from cameras in Central and Eastern Oregon.
We see the cameras being used not just for fire, but Lifeflight networks and others use our cameras to know when to fly helicopters. And we know the public uses them to find out, is it a good day to go out there and go hiking? Now, that’s one way to use the cameras, is the public facing website that people can see time lapse and current images from. And if you’re in the emergency response community, you can get a credentialed login. And with the login comes full functionality of controlling the camera, of seeing anomalies that were detected, understanding what’s happened in that space over the recent history. And that’s the two ways the cameras are accessed.
Miller: Do you see any security or data concerns with having hundreds now, thousands soon, of these cameras just all over the place collecting footage? As you’d said, the public only has access to the last 15 minutes of footage. But, I imagine that somewhere there are terabytes of data that go back much longer.
Toomey: Yeah. So one of the things we do, I mean, the cameras have to be in places that are both remote but also in the wildland urban interface where people live. So if you go to the website and you go to a camera, you’ll see that at the camera itself, people’s homes and things are smudged out so that you can’t see detail. And that data of their homes is done in hardware at the camera. So nothing’s ever recorded in that terabytes of information of what someone was doing in their backyard, things like that.
When it comes to controlling the cameras, where they turn and what they can see, that’s why we have credentialed logins. We keep track of everybody who touches a camera and moves it. And that is a pretty effective policy and people know what their responsibilities are. They are the emergency response community, so they’re using them for that. So if someone were to misuse that camera, they would lose their logins.
Miller: There’s also the question, separate from data security [and] maybe a more philosophical one, of what it means to have cameras and footage, especially in these places that many of us might like to think of as barely touched by humans. Does that bother you at all? And we’ve been talking about all the good public safety reasons for this, whether it’s detection or, as you said, even maybe more so, firefighting awareness and capabilities. Does it bother you at all that there are now cameras trained all over these places that we could also think of as wilderness?
Toomey: Well, I’ll flip that around in two ways. I mean, first of all, it’s been touched by humans for 10,000-plus years…
Miller: Point taken.
Toomey: … And managed in a way that we have not managed well in the past 100-years, where Indigenous Nations put fire on the landscape quite routinely in a way that was good and that we need to do in the future.
The other way I’d approach that question is that many of us [who] live throughout the West are attached to the landscape and the environment, and the vast majority of people maybe don’t get out there that often. I think there’s an advantage to having cameras in places that show the natural beauty of what we see in our region. And I think when people appreciate something, they begin to love it; when they love it, they take care of it. So I actually think the cameras have a value, an aesthetic value, that many people appreciate. And so I think that’s actually a good thing.
Miller: Let me make sure I understand that last point, because I’m not sure that it makes intuitive sense to me. But is it that if people, through a screen – in a relationship mediated through a screen – get to appreciate some patch of forest from the view of a top of a mountain, that they might be more likely to visit or at least just more likely to care about it because they’ve seen a static image on their cell phone?
Toomey: Well, I’ll address it in the following way. It is one of the things I do as a professor, I’ve taught an oceanography class. And when I first taught the class, I often talked about climate change and its impact on the oceans. And after about eight or nine weeks of that, the class felt rather down about what was happening in the context of taking care of our environment. So I approached it in a different perspective and the oceans are somewhat similar. Many people have a connection only with the beach. They don’t know the vast, deep aspects of the ocean. So I flipped the class and started talking about how wonderful it was and how many things are occurring out there that are remarkable. And I felt that if you give an opportunity to people to love something, they’ll take care of it. So after eight weeks of seeing the wonders of the ocean, then we could talk a week or two about how humans are impacting climate change. And I found that to be much more motivating for people.
Now, these camera systems, it’s not just cameras, right? We have an IP enabled, wireless network across the state where we can do full-scale environmental ecosystem monitoring. I would like to see situations where when my children go to school, they can understand rainfall in the watershed, they can see temperature and changes in river levels. And I think having a real-time network like we do throughout the Pacific Northwest allows us to both have visual data, but also other data that people can see and appreciate what’s going on.
Miller: I want to go back to an important point you made a few minutes ago, which is that humans, Indigenous people especially, on this land for thousands of years, understood that fire was a part of the landscape and they managed vast acreages with that in mind. That’s very different from the way that the U.S. Forest Service approached Western forests for more than a century, where there is famously this idea that all fires had to be put out by 10 a.m. the next day.
Do you have any concerns that this more sensitive warning system, which will enable humans earlier on to see fires when they’ve started, that it’ll make it more likely that those humans will then put out potentially beneficial, low intensity fires?
Toomey: Well, there are good days and bad days for fire, right? And so the cameras are also used in the off season, the shoulder seasons for vegetation management, controlled burn or prescribed burns, whatever you would like to call them. So it’s not just about fires on red flag days. I mean, obviously, right now, there are many prescribed burns occurring throughout the West. And as we come into the July, August, September time frame, we probably will not be doing prescribed burns on those days.
On those days when we know what the met conditions are, red flag days, it’s really important that fires don’t grow out of control. So there’s the use of the cameras in the off season to monitor prescribed burns, to monitor the weather conditions on those days and to do it safely, so that we have fire on the landscape during the time when it’s safer to do that. And then in the hot, dry, windy parts of the year, use the system to try and get on fires quickly so that it doesn’t threaten people’s homes, critical infrastructure or their lives.
Miller: I understand you’ve been having regular meetings with stakeholders from a bunch of different agencies that all have a hand in fighting wildfires. What have those meetings been like?
Toomey: I’m glad you asked that question. In our lab, the Oregon Hazards Lab, we do both wildfire cameras, but also the ShakeAlert system, the earthquake early warning system, that’s along the west coast. And so one of the things we learned early on was that building communities was important. So what you mentioned is the Oregon Wildfire Detection Interoperability Committee and that committee has the Governor’s Office, the Oregon Department of Forestry, the Bureau of Land Management, the National Forest Service, structural firefighters, some of the investor owned utilities and the Oregon Hazards Lab. Interestingly enough, Oregon is the first state in the country, I think, to have such a committee that is forward looking in how it’s coordinating for wildfire detection systems, and it’s been successful. I often say, Oregon is a nice place to work because most people throughout the state want to work together and I think that committee is an excellent example of that.
We’ve been meeting for a year and a half building trust, which I think is important in all these issues, when you’re doing public safety, public service work and bringing together different stakeholders. How you build trust, the time you take to do that is more essential than the technology, right? And so now we have Idaho and Washington seeing how this committee has been working and asking, “how can we set one up?” And I think we’ll see this sort of cascading effect where, whether it’s Oregon passing the first Bottle Bill, whatever, they have an impact on other states by leading the way and how to work together.
Miller: And more broadly, what do you think this project says about the role that universities can play today?
Toomey: I think public universities have a secret power [and] that we are discovering how to use that. I mean, of course, we’re here for research and education, but we’re also here to serve the public. So one of the things that we’ve been very excited about or that we’ve learned ourselves over the past seven or eight years of building this program is that we can facilitate relationships between different organizations. Think of a three-legged stool – one of those legs is the fire response community and we refer to this as a sort of use inspired research. What does that community need? So we go out to them, the Western fire chiefs or the Forest Service, BLM, and asked them, “what would you like to use?” And they provide us feedback and criticism and that’s fantastic.
A second leg of that stool is public/private partnerships, like the one we have with ALERTWest. That system grew out of university-based research. So, state, federally-funded research gave rise to early generations of detection cameras and the development of AI, and that led to a translational opportunity that grew into ALERTWest. We’re looking at other public private partnerships. For example, with FirstNet, can we get access to their towers in a way that could be cost effective? So that’s a second leg of the stool, how the university can work with those private sector agencies.
And then the third leg is, because we’re a public nonprofit organization, we can braid together different sources of funding. So we got into Governor Brown’s budget in 2018 for ShakeAlert and wildfire cameras. It was forward looking for the state to say it’s cost effective to have an alerting system that does earthquakes, wildfires, landslides and flooding. We receive money from the U.S. Geological Survey for ShakeAlert, from the Bureau of Land Management and Forest Service for wildfire detection, from the state to integrate these things, from the private sector. And now the public universities are the sort of glue member of a team that holds together the fire community, the private sector and different sources of funding. And I think, given our public service mission and the trust that we’ve earned throughout the state, we have a vital role to play.
Miller: Doug Toomey, thanks very much for your time today. I appreciate it.
Toomey: Yeah, thanks. I appreciate your interest as well.
Miller: Doug Toomey is a professor of earth sciences at the University of Oregon, where he is the director of the Oregon Hazards Lab.
Contact “Think Out Loud®”
If you’d like to comment on any of the topics in this show or suggest a topic of your own, please get in touch with us on Facebook, send an email to thinkoutloud@opb.org, or you can leave a voicemail for us at 503-293-1983. The call-in phone number during the noon hour is 888-665-5865.
