Neal on Fire

When wildfire threatens your home, the questions get personal: What’s happening? What does it mean for me? What should I do? Shreya Saxena lived that uncertainty during the 2023 Okanagan wildfires in British Columbia. That experience led her to UBC Okanagan, where she helped build the interface for a real-time wildfire risk dashboard, working with Dr. Mathieu Bourbonnais and industry partner Rogers Communications.

In this episode of Neal on Fire, Shreya shares what it takes to turn readings from 50+ field sensors into information that firefighters, researchers and the public can actually use. She explains why she pushed to split the dashboard into real-time and long-term views, why only five environmental factors made the cut, and why soil temperature, wind speed and wind direction are her top three. She also covers what AI and satellite imagery add to wildfire detection, and why better models don’t always mean better detection.

Guest: Shreya Saxena, AI Engineer at Laudroid; former UBC Okanagan wildfire dashboard developer; satellite imagery researcher

Key topics:

• Living through the 2023 Okanagan evacuation and the information she wished she’d had
• How the experience drew her into wildfire research at UBC Okanagan
• A 50+ sensor network measuring soil temperature, soil moisture, humidity, wind speed and wind direction
• Why sensors that burned in the 2023 fire kept transmitting valuable data
• Designing for first responders: a 24-48 hour real-time page vs. a one-to-two-year analytical page
• Why wildfire triggers come from combinations of conditions, not a single reading
• Quality and density over a few expensive sensors
• Separating conditions that could start a fire from evidence that a fire has started
• Where AI helps most: camera and satellite detection, and why model hype can mislead
• Combining ground sensors and satellite imagery to see before, during and after a fire, including fireweed regrowth
• Measuring success by prescribed burns enabled, not just evacuations
• Her hopes for the next five years: earlier evacuation alerts, AI and robots on the fireline, and better protection for wildlife and trees

Key quote: “Being able to display data which is usable for first responders and people who are in a state of emergency is a completely different challenge."

In this episode of Neal on Fire, host Neal Spungen talks with Gadi Benjamini, Co-Founder and CEO of FireDome, about a new category of wildfire technology. FireDome is a stationary, fully autonomous system that launches capsules of water or fire retardant up to 1,000 feet, protecting as much as 70 acres per unit.

Gadi explains its three layers of protection: a preemptive fire break, retardant coating of structures, and detection and suppression of spot fires while they are still small. He also covers why wind-driven fires, which cause most of California’s wildfire damage, favor a ground-based launcher over grounded aircraft, what the team learned moving from testing in Israel to Yuba County, California, and why FireDome shifted from protecting forests to protecting communities.

Show Notes

Guest: Gadi Benjamini, Co-Founder and CEO, FireDome
Host: Neal Spungen

In this episode:

• How FireDome works: a self-sufficient system with its own water, power and communications
• Capacity: 3,000 capsules of about 10 gallons each, launching every 5 seconds, enough to last for weeks
• Why wind matters: over 75% of California wildfire damage comes from wind-driven fires
• Testing in Yuba County after a year in Israel, with hundreds of launches so far
• The assumption FireDome threw out, and why transparency builds trust
• Integration with external detection sources and fire department incident command
• Autonomous by design, with remote shutoff for firefighters
• Business model: “wildfire resilience as a service” (OPEX) or capital purchase
• Target customers: HOAs, resorts, wineries, utilities, and critical infrastructure
• The planned mobile unit for fire departments and prescribed burns
• The five-year vision: second-by-second data transparency across the fire service

Upcoming event: FireDome will demo and test at the Red Sky Summit on November 5.

Timestamps:
00:00 Intro
00:29 What FireDome is
02:11 Yuba County test site
04:02 Where FireDome fits in the ecosystem
07:22 System capacity
08:10 The assumption FireDome threw out
10:15 Building trust
13:20 Integration with fire departments
17:39 Coverage area
19:03 Pricing: OPEX vs. CapEx
20:19 Who buys FireDome
23:05 The mobile unit
28:11 Wildfire tech in five years

Resources:
https://www.fire-dome.com/
https://nealonfire.com/

Scott Holman, founder and CEO of Red Line Safety and a retired fire captain, paramedic, and Marine Corps veteran, joins Neal on Fire to explain I’m Safe, a platform that combines wearable sensors, environmental monitoring, a self-healing LoRa mesh network, and a cloud dashboard to give incident commanders real-time awareness of firefighter location, exposure, and conditions. Learn how it works off-grid, how firefighters control what data they share, and what pilot tests have revealed.

Show Notes

Episode summary
In this episode of Neal on Fire, host Neal Spungen talks with Scott Holman, founder and CEO of Red Line Safety, about a platform designed to protect firefighters with better information. Scott explains why he started the company after more than 30 years in the fire service, how the I’m Safe wearable and gateways work, how data is shared and controlled, and what the early pilots have taught the team.

About the guest
Scott Holman is the founder and CEO of Red Line Safety. He is a U.S. Marine Corps veteran and a retired fire captain and paramedic with more than 33 years of fire and EMS experience. He developed the concept after studying line-of-duty death investigations and recognizing how often better information could have changed outcomes.

Topics covered

* What motivated him to start the company after losing colleagues to line-of-duty events and work-related illness
* The burden on incident commanders, who rely on radio traffic and reports to understand where people are and how they are doing
* How I’m Safe works: a wearable with GPS, air temperature, pressure, humidity, Bluetooth, Wi-Fi, carbon monoxide, and total volatile organic compounds, with room for additional sensors
* Two wearable versions: one for all hazards and one designed for wildland use
* Integration with biometric devices such as Garmin, Polar, and Oura Ring
* Simplicity: turn it on, put it on, and the data goes to the cloud automatically
* Firefighters controlling whether and how they share personal data, including setting their own thresholds
* Dashboards for incident managers with filters, interoperability with other systems, and an open API
* How the platform enhances LCES (lookouts, communications, escape routes, and safety zones) rather than replacing it
* Gateways with thermal imaging, wind data, and air quality sensing, which can be mounted on vehicles or on tripods with solar panels
* The LoRa-based, self-healing mesh network that connects units where there is no cellular service, and possible Starlink connectivity
* A first-place win in a Department of War competition for wildland firefighter safety technology
* Early proof-of-concept work with San Jose State University and what pilot data can reveal that traditional testing cannot
* Conversations with medical directors about using the data to improve rehab and recovery protocols
* The life cycle of a wildland fire: pre-incident, active, and post-incident data
* Current status: three pilots in California and Washington, a technology readiness level of about 6 moving toward 7, and a search for the next pilot department
* A message for fire chiefs: “your people matter,” and change can be good

Key takeaways

1. Better, automatic information can reduce the burden on incident commanders and the need for constant radio check-ins.
2. Wearables, gateways, and mesh networking can work together where cell service doesn’t exist.
3. Firefighters stay in control of their own data, which supports trust and adoption.
4. The system is designed to connect with other tools, not replace them.
5. Real-world data can reveal risks that training and standard tests may miss.
6. The company is seeking departments and partners for future pilots.

Resources
https://redlinesafety.com/

Stuart Landesberg, CEO of Seneca, joins Neal on Fire to explain how a team of five autonomous aircraft, each carrying water and foam concentrate, can reach wildfires within minutes to help firefighters attack early. Learn how the Seneca strike team works, how it detects and targets fires with infrared and lidar sensing, how batteries are swapped in two minutes, and what a live fire test with CAL FIRE revealed about the future of unmanned aerial suppression.

Show Notes

Episode summary
In this episode of Neal on Fire, host Neal Spungen talks with Stuart Landesberg, CEO of Seneca, about building autonomous aerial systems that help firefighters attack fires faster, safer, and more effectively. Stuart explains the strike team design, how the aircraft find and target fires, what they can do beyond suppression, and what he learned from a live fire test with CAL FIRE.

About the guest
Stuart Landesberg is the CEO of Seneca. He previously founded and ran Grove Collaborative, taking it public in 2022 and stepping down as CEO in 2023. He lives in Marin County, California, and was motivated to start Seneca after seeing the wildfire risk in his community and hearing from fire chiefs about the importance of reaching fires early.

Topics covered

* His background as a technologist, and what led him to start Seneca
* Seneca’s mission: building technology for firefighting situations that are unsafe, inefficient, or impossible today
* The Seneca strike team: five 400-pound aircraft, each carrying 20 gallons of water and a gallon of concentrate to make foam
* Autonomous launch, navigation, and fire targeting, with onboard sensors similar in concept to those on self-driving vehicles
* Why the system is meant to support, not replace, engine crews, especially in hard-to-reach terrain or when helicopters are unavailable
* Other uses beyond suppression: hauling hose and supplies, patrol, mop-up, cold trailing, night operations, and search and rescue
* A philosophy of building a flexible tool “like a shovel” that firefighters can use in many ways
* Cost compared with traditional manned aviation
* Battery design: in-house batteries, a two-minute swap and refill, and a mobile air operations base with a generator for 24-hour operations
* Day and night operation using lidar and infrared sensing
* The iteration process, including an eighth aircraft version and heavy use of 3D printing
* Balancing speed with safety and building with input from firefighters
* A company value of asking “what would it take?”
* Where Seneca fits in the wildfire ecosystem: detection is improving, and fast initial attack is the next frontier
* How the system receives locations from cameras, sensors, satellites, and calls, and how its six cameras and two lidar sensors search for smoke and heat
* The live fire test with CAL FIRE through Fireworks, a nonprofit working with CAL FIRE on adopting new technology
* Training: a Part 107 license to fly, plus a few weeks of a few hours a day for safe operations and sustainment
* Firefighters’ healthy skepticism of technology and why it is appropriate

Key takeaways

1. Reaching a fire in minutes is a major gap that autonomous aircraft could help fill.
2. A single platform that supports many missions can be easier to justify than a one-purpose tool.
3. Quick battery swaps and mobile bases make continuous operations possible.
4. Collaboration with agencies and firefighters is central to building safe, useful technology.
5. Detection is advancing, but a fast response to what is detected is the next frontier.
6. Training is manageable, but safe sustainment of aircraft this size requires care.

Resources
https://seneca.com/

Lt. Gen. Larry James, U.S. Air Force (Ret.), a former NASA JPL leader and strategic advisor to SmartSat, joins Neal on Fire to explain what satellites can and can’t do for wildfire detection today. Learn how geosynchronous and low Earth orbit satellites compare, why coverage and latency remain hard problems, how infrared sensors developed for defense apply to ignitions, and why integration, standardization, and funding matter for firefighters in Australia and the US.

Show Notes

Episode summary
In this episode of Neal on Fire, host Neal Spungen talks with Lt. Gen. Larry James about the role of satellites, Earth observation, and AI in wildfire resilience. Larry draws on a career in military space operations, intelligence, surveillance, and reconnaissance, and NASA’s Jet Propulsion Laboratory, and shares a candid look at the strengths and shortfalls of space-based detection, as well as the need for integration and funding.

About the guest
Lt. Gen. Larry James is a retired U.S. Air Force lieutenant general with 35 years of service, including leading intelligence, surveillance, and reconnaissance for the Air Force. He held a leadership role at NASA’s Jet Propulsion Laboratory and is now a strategic advisor to SmartSat, an Australian space technology organization, and the Australasian Space Innovation Institute. He also advises startups using JPL-derived sensor and AI technology.

Topics covered

* How advances in miniaturization, CubeSats, and lower launch costs have changed what is possible from space
* How defense infrared sensors, such as those used to detect missile launches, relate to wildfire detection, and how defense data became available in unclassified form
* NASA’s wildfire work and JPL’s Earth science research
* The “holy grail”: 24/7 coverage that can detect a very small ignition and send an instant alert
* Geosynchronous satellites: constant view, but limited sensitivity and resolution
* Low Earth orbit satellites: better resolution, but short passes and revisit gaps, and the number of satellites needed for continuous coverage
* Why a business case and a paying customer are needed to fund large constellations
* The role of aircraft, high-altitude balloons, and long-duration drones, and a systems engineering approach
* Lightning detection work with Edith Cowan University and the IGNIS program
* Similarities and differences between Australia and the US, including scale, resources, and fragmented tools and communications
* The need for national coordination, standard training, and common equipment
* A plain-language explanation of what JPL is and its history, from Explorer 1 to Mars rovers and Earth science
* The growing importance of predictive modeling, fuel loading analysis, and AI
* Unmet needs: faster detection and real-time information shared with every firefighter
* Challenges of commercializing research, including who pays and whether firefighting budgets can support new technology
* A closing note on the dedication of the researchers and firefighters he has worked with

Key takeaways

1. No single satellite type provides everything: geosynchronous gives constant coverage but less detail, while low Earth orbit gives detail but limited coverage.
2. Early detection is the key gap, and every available tool, from space to ground, should be used.
3. Integration of data across sensors and delivering it to the firefighter on the line is as important as the sensors themselves.
4. Standardization of training, equipment, and communications can improve cross-border response.
5. Commercial adoption depends on a clear return on investment and a reliable funding source.
6. Systems engineering helps identify gaps and choose the right combination of technologies.

Rod Goossen, CEO of RoGO Communications, joins Neal on Fire to explain how Drop Block, a rugged, satellite-enabled device paired with the RoGO app, gives wildland firefighters live locations, fire weather, and text messaging in areas with no cell service. Learn how the Yarnell Hill tragedy inspired the technology, why Iridium was chosen over Starlink, how it integrates with incident command platforms, and how it applies to hurricanes, military use, and insurance.

Show Notes

Episode summary
In this episode of Neal on Fire, host Neal Spungen talks with Rod Goossen, CEO of RoGO Communications, about improving first responder communications where cell service doesn’t exist. Rod explains how the Yarnell Hill tragedy led to Drop Block, how the system works, why it uses Iridium, and where it could be used beyond wildfire.

About the guest
Rod Goossen is the CEO of RoGO Communications. He has a background in avionics engineering and has worked for corporate jet manufacturers. His brother is a 30-year wildland firefighter in Colorado, and their conversations after the 2013 Yarnell Hill tragedy led to the creation of the company.

Topics covered

* The motivation: the 2013 Yarnell Hill fire, and learning that wildland firefighters still rely heavily on paper maps, compasses, and two-way radios
* How most fire incidents lack cell service, leaving crews without a clear picture of where resources and the fire are
* What Drop Block is: a small, satellite-enabled device carried by each crew boss, which connects to a phone through the RoGO app
* Three core functions: seeing the location of people and equipment, seeing live fire weather, and sending text messages point to point or to a geofenced area
* How crews coordinate today, and why locations and tasks are often based on estimates
* Safety benefits, including awareness of safety zones and changing fire behavior
* Devices for crew bosses, engines, and bulldozer crews rather than every individual
* Go-to-market status: hardware and software complete, FCC certification being finalized, and early customers including the Colorado River Fire Department and the Arizona Department of Forestry and Fire Management
* Fast deployment: eight AA batteries, about a minute to acquire a satellite, and handoff at incident check-in
* Durability: IP67-rated and designed for rough use
* Why Iridium: low cost, short burst data, and better penetration of tree canopy than higher-frequency options
* Data sent about every 15 minutes, including GPS location and weather readings such as wind speed, wind direction, humidity, and temperatures
* Cost: a satellite plan comparable to a monthly cell phone bill, and a projected 17 to 1 return on investment
* Insurance applications, including off-grid weather data for risk pricing and reduced liability
* All hazards uses: hurricanes, earthquakes, tsunamis, tornadoes, and cellular-denied areas in cities
* Supplementing existing radios and platforms such as ATAK and others rather than replacing them
* Interest from FEMA, Border Patrol, the National Guard, and special operations communities
* The biggest hurdle: education, and getting first responders to try the device

Key takeaways

1. Many wildfire incidents have no cell coverage, so crews need an alternative way to share location and weather.
2. Seeing every resource on a map reduces guesswork and supports safety.
3. Low-bandwidth satellite data is enough for location, weather, and short messages.
4. Drop Block is designed to supplement radios and incident management platforms, not replace them.
5. The same capabilities apply to hurricanes, other disasters, and defense uses.
6. Education and hands-on trials are key to adoption.

Resources
⁠https://rogocom.com

Dr. Alexandra Syphard, senior research scientist at the Conservation Biology Institute and director of science for the Global Wildfire Collective, joins Neal on Fire to explain what wildfire models do, what they can and can’t predict, and how to choose the right one. Learn how development and ignitions shape fire risk in the wildland-urban interface, why “fireproof” is a myth, how sensitivity analysis and “fit for purpose” guide modeling, and why land use planning matters for future fire outcomes.

Show Notes

Episode summary
In this episode of Neal on Fire, host Neal Spungen talks with Dr. Alexandra Syphard about wildfire modeling, a core tool for understanding wildfire risk. She explains what models are, why fire is so complex, how development and ignitions influence outcomes, and how decision makers can judge which models to trust and use.

About the guest
Dr. Alexandra Syphard is a senior research scientist at the Conservation Biology Institute and director of science for the Global Wildfire Collective. She has authored more than 120 research papers and is known for her work on wildfire risk modeling and on how ignition sources and development patterns shape fire in the wildland-urban interface.

Topics covered

* What wildfire modeling means and why models are simplifications of reality that let people ask “what if” questions
* Why wildfire is a tightly coupled system of people and ecosystems that plays out over scales from seconds to decades
* Misconceptions about development: new housing brings ignitions, changes vegetation, and changes where firefighters go
* How density matters differently depending on context, with structure loss often in low-density areas historically, and dense urban settings where structures become fuel in wind-driven fires
* Why there is “no such thing as fireproof,” and the risk of a false sense of security from home hardening and defensible space alone
* What drives model results: fire behavior models use topography, fuel, and weather, while other models are sensitive to different variables
* Sensitivity analysis and the importance of understanding data quality
* How ignitions are treated, including human-caused versus lightning ignitions, and why they are not random
* Why most ignitions are not the ones that cause the biggest impacts
* Using ignition probability maps as input to fire behavior models
* How often to update a model, and the idea of “fit for purpose”
* How remote sensing, such as FireSat, gives better data for models and empirical research
* “Garbage in, garbage out” and balancing scientific rigor with usable decision support
* Why communication and uncertainty matter, and the risk of over-trusting or under-trusting a model
* A taxonomy of wildfire models, with 10 families and many sub-categories, created for the California Department of Insurance’s effort toward a public wildfire catastrophe model
* Vetting models through peer review and knowing who built and ran them
* Why a single standard model is unlikely, given non-stationary systems and novel conditions
* Land use planning and model coupling, and examples of research being used in development decisions
* Advice for newcomers: start with your specific question, learn what others have done, and collaborate

Key takeaways

1. Models are simplified tools that help compare scenarios, not predict where a specific fire will happen.
2. Development both affects and is affected by wildfire, and the effects depend on context.
3. Mitigation reduces risk but does not make a house fireproof.
4. Ignitions are patterned, and including them in models can improve results.
5. The best model depends on the question, so a “fit for purpose” approach matters.
6. Peer review, transparency, and collaboration between disciplines help build trust in models.

Don Hallett, founder of Wildfire Innovations and inventor of RainStream, joins Neal on Fire to explain how an elevated, mobile misting tower can pre-wet up to about 18 acres above the tree canopy to help protect communities from wildfire and embers. Learn how the telescopic mast, wind ratings, patented nozzle, and water sourcing work, how fast a unit can be set up, and why it is designed to save water and free up firefighting resources.

Show Notes

Episode summary
In this episode of Neal on Fire, host Neal Spungen talks with Don Hallett, founder of Wildfire Innovations and inventor of RainStream, a mobile tower system that mists large areas from above the tree canopy. Don explains how it works, how it was developed, how it is deployed and supplied with water, and where it could fit alongside firefighters and incident commanders.

About the guest
Don Hallett is the founder of Wildfire Innovations and the inventor of RainStream. He has a background as a tradesman, including welding, piping, and structural steel work in remote areas, and grew up in Northern Canada, where his family was evacuated from wildfires when he was a child.

Topics covered

* What RainStream is: an elevated sprinkler system on a trailer, about 100 feet tall, covering up to 18 acres at maximum size
* How it differs from standard sprinklers and water cannons, with a focus on elevation, area covered, and freeing up fire resources
* The origin story: witnessing a fire grow in Wyoming and recalling childhood evacuations
* Early prototypes, including a tree-mounted version, and how elevation and wind extended the spray distance
* Misting, soaking, and humidity ranges that create a microclimate, with the goal of timed and targeted use of water
* A telescopic mast with working heights of 30, 60, 80, and 100 feet, with wind ratings that increase as the mast is lowered
* A patented nozzle designed for strong wind conditions, with an adjustable trajectory
* Using wind to extend coverage and counter ember transport
* Weather stations and cameras on the unit for monitoring
* Water sources: hydrants, ponds, rivers, and lakes, a micro-screen intake filter, and pump placement
* Deployment: crews of three to five, set up in roughly 20 to 30 minutes with a goal of 10 minutes or less in a next-generation model, and mast raising in under two minutes
* Realistic coverage of about 7 to 18 acres depending on wind
* Business model: sales, rentals, training, and municipal sharing programs
* Other uses such as dust suppression, mining, construction sites, and cooling
* Coordinating with incident commanders so hydrant water isn’t depleted
* Use of environmentally friendly retardants and additives
* Plans to complement detection technology as a defensive tool “between the ground and the air”

Key takeaways

1. Elevated misting can pre-wet large areas and add humidity before embers arrive.
2. Wind, normally a challenge, can help spread mist farther.
3. Timed and targeted use of water helps conserve limited supplies.
4. Mobile deployment lets units be moved from one threatened area to another.
5. Natural water sources and filtration extend where it can be used.
6. Working with incident commanders keeps the system aligned with firefighting priorities.

Resources
https://www.wildfireinov.com/
https://nealonfire.com/

Aren Page, founder of Grid IQ and a former NOAA wildfire researcher and utility engineer, joins Neal on Fire to explain how pre-ignition grid monitoring can pinpoint faults on power lines before they become wildfires. Learn how Grid IQ’s small, quick-install hardware and mesh network locate anomalies like vegetation contact and failing conductors, how it can reduce the need for public safety power shutoffs, and why the same technology applies to underground cables, microgrids, and railways.

Show Notes

Episode summary
In this episode of Neal on Fire, host Neal Spungen talks with Aren Page, founder of Grid IQ, about moving wildfire prevention upstream by detecting problems on the power grid before they spark a fire. Aren shares the story behind the company, how the sensors work, how they fit into existing utility infrastructure, and where the technology is headed beyond wildfire.

About the guest
Aren Page is the founder of Grid IQ. He grew up in Santa Rosa and Petaluma, California, and was motivated by the impact of the 2017 Tubbs fire on his community. His background includes burn scar analysis research at NOAA, work at a major California utility in substation design, distribution engineering, and wildfire hardening, and engineering consulting across Western states. He also grew up in electrical construction.

Topics covered

* The origin story: the Tubbs fire, his career path through NOAA, a utility, and engineering consulting, and why he decided to work on the source of the problem
* How grid equipment causes ignitions: vegetation contact, small arcing from worn conductors, and aging infrastructure
* Grid IQ’s pitch: “a nervous system for the power grid,” which locates a problem the way the body locates pain
* Three phases of wildfire technology: pre-ignition, post-ignition, and deployment, with Grid IQ focused on pre-ignition
* Hardware with a software backend, using proprietary signaling to pinpoint the exact location of an anomaly
* About one unit per mile of line, a mesh network, and satellite or cellular backhaul to a grid operator interface
* Working with existing reclosers, with a small unit and about a 30 minute install for overhead and underground systems
* Prototyping with a small distribution line built in his backyard
* Avoiding contact with the conductor while keeping accuracy
* A use case walkthrough: how PSPS events work, and how real-time fault location can help avoid unnecessary shutoffs and speed up response
* Why involving linemen and engineers early made installation and maintenance easier
* How data is handled: signal injection, waveform analysis, and machine learning to separate normal behavior from faults
* Target customers using distribution-level voltages, from large utilities to small municipal systems, including reporting on what problems were prevented
* Other applications: underground systems, a rail project in Tanzania, and a microgrid on tribal land in Northern California
* A plan for three to four pilots, with interest in Puerto Rico, Arizona, New Mexico, and Catalina Island
* Thoughts on automation and real-time grid visualization over the next five years

Key takeaways

1. Many ignitions begin with small faults that existing protection systems may not detect.
2. Pinpointing the location and type of an anomaly can turn a possible disaster response into a routine repair.
3. Hardware that installs quickly on existing equipment lowers the barrier for utilities.
4. Real-time visualization may reduce the need for broad power shutoffs.
5. The same approach can improve resilience for underground cables, microgrids, and railways.

Resources
https://www.gridiq.io/
https://nealonfire.com/

Felix Vieth, co-founder of the nonprofit Pyronear, joins Neal on Fire to explain how open source AI and low-cost cameras can detect wildfires early. Learn how Pyronear’s monitoring stations spot smoke, how the team reduces false positives, why open data and community contributions matter, and how deployments in France, Catalonia, and Chile are making detection affordable for fire departments and communities.

Show Notes

Episode summary
In this episode of Neal on Fire, host Neal Spungen talks with Felix Vieth, co-founder of Pyronear, a nonprofit using open source AI and affordable camera systems to detect wildfires early. Felix explains how the project began, how the system works, how it handles false positives, and why an open, community-driven model can make detection more accessible worldwide.

About the guest
Felix Vieth is a co-founder of Pyronear, based in France. The project began when volunteers from the Data for Good community wanted to apply computer science and deep learning to help firefighters detect wildfires earlier. It is now an NGO working with fire departments, researchers, engineers, and volunteers.

Topics covered
* How Pyronear started: a volunteer group aiming to use computer vision for the common good, working with French fire departments from the beginning
* Why it is open source: code, AI models (for non-commercial use), and annotated image data are shared publicly
* How detection works: monitoring stations on high vantage points, multiple cameras, a microcomputer running an AI model, alerts sent through an API, and a platform where firefighters review images
* Reducing false positives through continuous research and development, masking recurring sources such as factory smoke, and tools that help firefighters quickly verify alerts
* Community volunteers who help annotate data, including non-technical contributors
* Why the system doesn’t currently use a human in the loop before alerts reach users
* Where it is deployed: France, Catalonia in Spain, and Chile
* Partnering with local companies, NGOs, researchers, and firefighters before expanding to a new country
* Affordability: standard IP cameras and microcomputers, with a full station installed for under about 5,000 dollars or euros
* Managing a volunteer roadmap with a small core team
* Funding through grants, support from the French government’s digital innovation directorate, company contributions, and donated expertise
* Benefits of open source, including strategic autonomy and a nonprofit mission
* Deployment figures: around 30 stations in France across seven departments, about 25 cameras in Catalonia, and about 15 stations in Chile with more planned
* Why the team is focusing on camera detection before integrating satellites or ground sensors
* Future goals: significantly reducing false positives, a convenient tool without a human in the loop, and contributions from every continent

Key takeaways
1. Open data and open source AI can lower costs and widen access to wildfire detection.
2. Convenience for firefighters is central, so alert review tools are as important as the AI.
3. Recurring false positives can be handled by masking known sources and improving the model.
4. Understanding how fire services work locally is key before expanding to a new country.
5. Pyronear currently focuses on early detection with cameras, with satellites and other data as possible future complements.
6. The project welcomes volunteers, researchers, and partners from around the world.

Resources
https://pyronear.org/en/
https://nealonfire.com/