AI system
- System
-
ALERTCalifornia AI wildfire detection (machine-learned smoke and anomaly detection on the ALERTCalifornia pan-tilt-zoom camera network)
- Model
-
Not assessed
- Release
-
Not assessed
- Application developer
-
ALERTCalifornia, University of California San Diego, with industry partner Digital Path - Role
-
Not assessed
- Configuration
-
Not assessed
- Inputs and outputs
-
Not assessed
- Autonomy
-
Not assessed
- Workflow
-
A camera network with machine-learned smoke detection watching remote forest at night, when human spotting is weakest. A verification step by trained operators that turns a model flag into a confirmed alert to the local Cal Fire station. A fire still small at detection, with a crew close enough to reach it before it grows.
- Human oversight
-
Operators checked the alert before dispatch because dust, insects, and weather can produce false detections. Fire crews responded and contained the fire.
- Relation to AI
-
The model's contribution was the detection. Its algorithm flagged a smoke column in a Cleveland National Forest camera feed at about 3 am. Human operators verified it before Cal Fire was alerted. The containment was done by Captain Yeary's crews. No source records whether a 911 call would have come later. The time gained is documented only for the program as a whole. TIME reports 77 fires identified before any 911 call in the first two months. UC San Diego's Neal Driscoll says the alerts beat 911 calls about 40 percent of the time. AFP also notes the system sometimes misreads dust, insects or weather as smoke. That is why the human verification step sits between the model and the dispatch.
- AI contribution
-
ALERTCalifornia flagged smoke in a camera feed at about 3 a.m. Operators verified the detection and notified Cal Fire, whose crews contained the fire.
Benefits
- Outcome identifiers
-
Not assessed
- Benefit
-
Cal Fire held the verified fire under a quarter acre
- Category
-
Reduced Risk
- Status
-
Not assessed
- Beneficiary groups
-
Organizations. Society.
- Beneficiaries
-
Not assessed
- Reach
-
Not assessed
- Distribution and access
-
Not assessed
- Duration
-
Not assessed
- Expected?
-
Yes
- Costs and harms
-
Not assessed
Measurement
- Studies
-
Not assessed
- Study design
-
Not assessed
- Measurement
-
Not assessed
- Setting and sample
-
Not assessed
- Measures
-
Cal Fire held the verified fire under a quarter acre
- Comparison
-
911 calls from the public as the first notice of a fire
- Results
-
Reference: not stated. Result: the AI beats 911 calls about 40 percent of the time.
- Additional value from AI
-
Not assessed
- Limitations
-
The account does not establish when a 911 call would otherwise have arrived. Figures about alerts preceding 911 calls describe the wider program rather than this fire.
- Denominators
-
Not assessed
- Materials for replication
-
Not assessed
Money
- Kind of value
-
Not assessed
- Amount
-
Not assessed
- Currency and prices
-
Not assessed
- Period and unit
-
Not assessed
- Realization
-
Not assessed
- Recipients and payers
-
Not assessed
- Gross or net
-
Not assessed
- Included costs
-
Not assessed
- Comparison
-
Not assessed
- Method
-
Not assessed
- Return and payback
-
Not assessed
People and places
- Participants
-
Developer:
ALERTCalifornia, University of California San Diego, with industry partner Digital Path. Operator: 
California Department of Forestry and Fire Protection (CAL FIRE). - Jurisdictions
-
Benefit: Mount Laguna, California,
United States. Developer: United States. Operator: United States. - Ownership and control
-
Not assessed
- Dependencies
-
Not assessed
- Cross-border connections
-
Not assessed
- Geopolitical implications
-
Not assessed
Follow-up
- Conditions for replication
-
The detection must precede any human report for the time gained to be attributable to the model. For this fire no source documents the absence of a 911 call. Human verification must be fast enough that the alert still arrives in the incipient phase. False positives from dust, insects or weather must be screened out so that crews keep trusting the alerts.
- Replication attempts
-
Not assessed
- Later findings
-
1 September 2023. After the summer trial the AI tool was rolled out to all 21 Cal Fire dispatch centers in September 2023 (day not given; first day of month used), and TIME named it one of the best inventions of 2023 on 24 October 2023.
- Availability
-
Not assessed
- Institutional decision
-
Not assessed
- Grounds
-
Not assessed
- Action and result
-
Not assessed
- Success criterion
-
Not assessed
Evidence
- Source numbers
-
AIBD-0120-source-0001–0009
- Citations
-
Numbered references below.
- Language
-
Not assessed
- Source relationships
-
Independent, Developer
- Original evidence
-
Not assessed
- Access and versions
-
Not assessed
- Supported findings
-
Not assessed
- Evidence links
-
Not assessed
- Evidence strength
-
Not assessed
- Missing evidence
-
Not assessed
Grades
- Assessed benefits
-
Organizations. Society.
- Depth
-
Organizations: 2. Society: 2.
- Reasons
-
Organizations: Cal Fire's Mount Laguna station received a verified alert at about 3 am. Its crews extinguished the fire at under a quarter acre. Cal Fire intelligence specialist Suzann Leininger says the system “is getting us time to react in a faster manner”. Society: Residents, property and forest around Mount Laguna were spared a larger fire. Captain Yeary's statement that it “could have been a devastating fire” is a judgment, not a proven counterfactual.
- Observed or intended
-
Not assessed
- Rubric
-
Not assessed
- Assessor and date
-
Not assessed
- Alternative assessments
-
Not assessed
- Fill
-
Independently reported
Entry details
- Kind
-
Not assessed
- Related entries
-
Not assessed
- Field
-
Emergency services (wildfire detection and response). State fire agency using a university camera network.
- Purpose
-
The tool was built to spot smoke in ALERTCalifornia camera feeds and notify Cal Fire emergency command centers. Crews can then reach a fire in its incipient phase. This detection was that purpose.
- Dates
-
Date basis unassessed: 2023 (precision unassessed). Updated: 21 October 2026.
- Stage
-
Not assessed
- Version
-
SHA-256: 6a95bb51aedb…
References
- Other identifiers
-
oecd_aim: 2023-09-12-f6ac.
- Category mappings
-
Not assessed
- Brief
-
Not assessed
- Revisions
-
5 revisions. Latest: 21 October 2026.
Sources
-
News report Independent author
Fonsegrives R. / AFP via Phys.org (2023). California firefighters use AI to battle wildfires. AFP wire report.
Had the AI not alerted us to it, it could have gotten much bigger
-
News report Independent author
AFP via Asharq Al-Awsat English (2023). California Firefighters Use AI to Battle Wildfires. Same AFP wire text (not independent of the Phys.org copy).
Human operators were able to verify the machines were correct and alerted Yeary
-
News report Independent author
AFP via France 24 (2023). California firefighters use AI to battle wildfires. Same AFP wire text (not independent).
We beat 911 calls about 40 percent of the time. And it's going to get better.
-
News report Independent author
Fonsegrives R. / AFP via Oman Observer (2023). Firefighters use AI to battle wildfires. Same AFP wire text (not independent).
It's getting us time to react in a faster manner.
-
News report Independent author
Dixit P. / TIME (2023). AlertCalifornia and Cal Fire AI Wildfire Detector. The Best Inventions of 2023.
In the first two months, the system had correctly identified 77 fires before any 911 calls came in.
-
News report Independent author
Jennewein C. / Times of San Diego (2023). Time Magazine: California's AI Wildfire Detection System a Top Invention of 2023. Describes a second AI-first detection near Grass Valley on 11 September 2023.
the first 911 call was not made until 6:01 a.m., and firefighters were already at the scene
-
Press release Developer
CAL FIRE and ALERTCalifornia (2023). CAL FIRE and University of California San Diego's ALERTCalifornia Program Join Forces to Enhance Wildfire Response with Artificial Intelligence Implementation Trial. Agency press release, Sacramento.
a machine-learned model has been created to accurately differentiate between smoke and other particles in the air
-
Press release Developer
Scully C. / UC San Diego Today (2023). ALERTCalifornia and CAL FIRE's Fire Detection AI Program Named One of TIME's Best Inventions of 2023. University news release.
The AI tool became available to all 21 CAL FIRE 911 Dispatch Centers in September 2023.
-
Press release Developer
If the incident is vetted and confirmed by trained watchstanders, firefighters respond quickly to extinguish the fire at the incipient phase.