40 fire wildfire detection cameras to be install in the North Bay

The cameras can spot a fire soon after it ignites.

FireAlert camera
Technicians install an AlertWildfire camera. File photo from the University of Nevada.

Several organizations are cooperating to install a network of cameras in the North Bay area of California that can detect wildfires soon after they start. At a perfect location with a 360° view the near infrared sensors can spot the signature of heat on up to 5,000 square miles, and up to 20,000 square miles at night. If a second camera detects the same heat source or smoke, the triangulation can tell dispatchers the exact location, enabling firefighters to get to the scene quickly.

A supercomputer attached to the network can then model the fire’s spread in 30 seconds to predict where it will be burning in the next several hours.

Recently one of the $2,600 cameras was installed on a hill that overlooks the path of the deadly Tubbs fire that burned into Santa Rosa in 2017.

Below is an excerpt from the Press Democrat:

With support from PG&E, the network plans to cover Sonoma, Mendocino, Napa, Lake and Marin counties with up to 40 such cameras by the end of March.

Thirteen of the pan-tilt-zoom cameras are already operating in the North Bay, with their images available to emergency dispatchers and to the public at alertwildfire.org.

The broader goal is to establish 200 new cameras statewide this year and Gov. Gavin Newsom’s budget includes funding for 100 more, said Graham Kent, director of the seismological laboratory at the University of Nevada, Reno, that started the program.

The Sonoma County Water Agency is also supporting the camera installation project.

The AlertWildfire group, a consortium of universities, including Scripps Institution of Oceanography at UC San Diego, Sonoma State University, and Oregon State University will build and maintain the system.

“These cameras will provide us with early fire detection and a level of situational awareness that is critical as we adapt to new wildfire behavior,” Sonoma County Water Agency director and Board of Supervisors Chair James Gore said.

The fire-camera system is built to the specifications of the University of Nevada, Reno’s Seismological Lab’s earthquake monitoring communications network based in their College of Science. It features private high-speed internet connectivity capable of transmitting seismic, environmental and climate data, in addition to the live-streaming high-definition video from the fire cameras.

“These fire camera networks realize their full functionality when a cluster of cameras are deployed in one area and to supply early detection, 911 confirmation, and situational awareness as well as triangulation to locate the fires,” Neal Driscoll, a professor at UC San Diego and co-leader of AlertWildfire, said. “Sonoma County Water Agency’s vision has made the North Bay region the next fire camera cluster.”

map AlertWildfire system
The dots represent the locations of fire detecting cameras in the AlertWildfire system.

Dozens of cameras are already installed and working in Southern California, the Lake Tahoe area, and locations in Nevada.

Typos, let us know HERE, and specify which article. Please read the commenting rules before you post a comment.

Author: Bill Gabbert

After working full time in wildland fire for 33 years, he continues to learn, and strives to be a Student of Fire.

8 thoughts on “40 fire wildfire detection cameras to be install in the North Bay”

  1. Staffed Fire Lookouts are greatly undervalued, they do more than watch for fire.

    In addition to seeing where lightning strikes occur and watching to see if a fire start happens, Lookouts can also help guide firefighters into said starts. Lookouts can also a great source of local info on terrain and fire behavior.

    It is just handy to have eyes and ears with a great view, act as human repeater, watch for vandalism.

    The experience of children visiting a Fire Lookout cannot be under estimated, can last a Lifetime and influence their Career Choice. It is sad to see Manned Lookouts being phased out, big part of our History and Tradition.

    0
    0
  2. Here’s the problem. As you can see from the map, there are these cameras all over the central Sierras (one is not more than 10 air miles from my house) yet fires like the Camp Fire still happened and went undetected for long enough o get established. Why? Because they produce so many false alarms that the alarms are turned off and there is no one that is actually paid to do nothing but look at the cameras (like they have in London for their crime cameras). Until that changes, all they will be good for is confirming there is a fire after someone calls it it.

    0
    0
  3. The core usage today of ALERTWildfire is as a 911 verification tool, where dispatch can understand almost immediately if the fire has been misreported in terms of location, or whether it might need a reduced response—or perhaps response needs to be escalated significantly. In all cases, this is very valuable—sometimes extremely valuable. The embedded lightning data on the map helps enable discovery of a number of fires each year. We have run AI, but it’s still not production quality (on any platform), but we continue to make strides in this regard. The ALERTWildfire cameras are crowd-sourced (and on AWS to scale and not crash during heavy usage) so that provides another opportunity for discovery by the public when they tweet out, and has happened a few times so far, and it also allows the public to understand their situation awareness during a significant event (or weather on a non fire day). ALERTWildfire agrees and also advocates that cameras are not the only solution, but part of a comprehensive approach that starts with fuel management, where our cameras are routinely used during Rx burns to monitor the burns and distribution of smoke. We also advocate approaches similar to that adopted by SDG&E in San Diego (and partners) with pre-burn modeling before Santa Ana conditions and potential pre-deployment of assets with partners based on those results before the potential first spark. Quick discovery and/or confirmation as happened during the Lilac Fire (halted at slightly above 4000 acres) and two fires this season a couple days after the Woolsey Fire. In the latter cases, both were knocked down and assisted by location confirmation and at least in one case a Sky Crane helicopter attack—and resulted in no spread in extremely windy conditions. These are far from the only examples, but occurred during the worst historical fire conditions in San Diego County. Other incidents from Tahoe, North Bay and central Nevada tell the same tale.

    Those dispatches who use the ALERTWildfire cameras daily have fully embraced (and helped design) this new and evolving approach.

    0
    0
    1. AI-based flame detection systems are making enormous advances right now. Both of the following papers are getting very promising results:

      https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=8307064

      https://github.com/tobybreckon/fire-detection-cnn

      Both describe excellent results for very challenging scenes. I think a very promising approach moving forward would be to combine one or both of the above approaches with an infrared camera and turbulent flow detection to select candidate regions.

      0
      0
  4. i look at these cameras everyday as i do my daily routine.then look several more times aday.

    0
    0
  5. (Disclaimer: I am a researcher studying applications of computer vision to flame detection).

    This is a very challenging problem. A near-infrared camera by itself is probably not an adequate solution. The challenge comes from false positives — even a 0.01 percent false-positive rate will likely result in dozens if not hundreds of false positives per day from the system described here. Existing deployed fire detection systems have a false positive rate of 5-10 percent — which has proven acceptable for aviation and the petroleum industry.

    My own current guess is that you need a pretty high-resolution infrared camera to detect turbulent motion and an optical camera with 8-12 megapixels connected to a cleverly trained deep convolutional neural network to get reliable detection results. Plus some extremely clever programming to connect those pieces together in an intelligent way.

    0
    0
  6. OK. I understand the role of detection. 2019 will be my 50th wildland fire season. Detection is linked to suppression, which in the final analysis is simply breaking the fire triangle. If residential development ordinances and forest health/fuels treatments are not exponentially increased, the net effect of these fancy (and expensive) cameras will be to dispatch suppression resources a little quicker. When the fire environment is set for a big fire, it won’t be anything. The problem is not rapid detection, the problem is FUEL (including houses).

    0
    0

Comments are closed.