How California Uses AI to Detect Wildfires Faster With Cameras, Satellites, and Predictive Analytics

The Core Problem: Detection Lag Costs Lives

Traditional wildfire detection relies on human observation — a hiker spots smoke, calls 911, dispatch responds. That chain takes time. In fire conditions, minutes translate directly into acres burned and lives at risk.

California’s approach flips that model. Instead of waiting for humans to report a fire, the system is designed to detect ignition automatically, often before any call is made.

AlertCalifornia: 1,260 Cameras, 915 AI-Enabled

The foundation of California’s detection network is AlertCalifornia — a system of mountaintop cameras positioned across the state. Of the 1,260 cameras currently deployed, 915 are now AI-enabled, meaning they automatically scan for smoke and push alerts directly to dispatch centers.

That’s a meaningful shift. Dispatchers aren’t waiting for calls. They’re receiving automated alerts from cameras that are watching 24/7, in locations where human observers simply aren’t present.

Cal FIRE plans to add another 200 cameras this year, expanding coverage into areas that currently have detection gaps.

FireSat: Satellites That Spot a 5×5 Meter Fire

Camera networks have limits — terrain, weather, and line-of-sight all create blind spots. That’s where FireSat comes in.

Cal FIRE joined Earth Fire Alliance’s FireSat early adopter program in 2025. The constellation now includes four satellites in orbit, with three additional spacecraft recently launched from Vandenberg Space Force Base. Once fully deployed, FireSat is expected to deliver updated imagery approximately every 20 minutes.

The detection threshold is notable: FireSat is designed to identify wildfires as small as 5 by 5 meters. That’s early-stage detection at a scale that existing satellite monitoring hasn’t been able to match.

The goal is to layer satellite data with mountaintop cameras and fixed-wing aircraft into a unified heat-detection picture across the entire state.

FIRIS: Real-Time Fire Mapping From the Air

Once a fire is confirmed, the next challenge is understanding how fast it’s moving and where it’s headed. That’s the job of the Fire Integrated Real-Time Intelligence System, or FIRIS.

FIRIS uses fixed-wing aircraft equipped with infrared sensors to deliver real-time perimeter mapping and fire spread modeling — often within minutes of ignition. Since launching statewide, the system has completed more than 2,000 missions, supporting not just wildfire response but also hazardous materials incidents and search-and-rescue operations.

For incident commanders, that kind of real-time mapping changes decision-making. Instead of estimating fire behavior from the ground, they’re working from live aerial data.

WIFTIC: Predicting Fires Before They Start

Detection is reactive. California is also investing in prediction.

The Wildfire Forecast and Threat Intelligence Integration Center — WIFTIC — combines weather data, fuel moisture conditions, lightning forecasts, and fire behavior models to identify areas at elevated risk before any ignition occurs.

That intelligence feeds directly into a pre-positioning system. When WIFTIC flags a high-risk window, Cal OES can move firefighters and equipment into position ahead of dangerous conditions. This year alone, Cal OES has approved pre-positioning requests for 20 counties, involving more than 1,400 personnel and 213 pieces of equipment.

Pre-positioning isn’t new. Doing it based on predictive AI modeling — rather than gut instinct or historical patterns alone — is a meaningful operational upgrade.

Drones and Lidar: Filling the Remaining Gaps

Beyond the headline systems, Cal FIRE is expanding its use of drones for fire mapping, prescribed burns, and tactical operations. The agency is also incorporating lidar data to assess vegetation density and fuel conditions — information that directly informs both fire behavior modeling and prescribed burn planning.

Lidar-based fuel mapping gives fire managers a more accurate picture of what’s on the ground, which makes predictive models more reliable.

What This System Actually Looks Like in Practice

Put it all together and the workflow looks something like this:

• Before ignition: WIFTIC identifies elevated risk zones based on weather and fuel data. Resources are pre-positioned.
• At ignition: AlertCalifornia cameras detect smoke automatically. FireSat satellites provide orbital confirmation. Dispatch is alerted before 911 calls arrive.
• During response: FIRIS aircraft deliver real-time perimeter mapping. Drones support tactical operations. Incident commanders work from live data, not estimates.

Each layer addresses a specific gap. The value isn’t in any single tool — it’s in how they’re integrated.

Why This Matters Beyond California

California’s system is worth paying attention to because it demonstrates what a mature, multi-layer public-sector AI deployment actually looks like in a high-stakes domain.

It’s not a single AI product. It’s a coordinated infrastructure that combines edge detection (cameras), orbital sensing (satellites), aerial intelligence (FIRIS), and predictive modeling (WIFTIC) — all feeding into operational decision-making.

Other states, countries, and emergency management agencies are already looking at this model. The underlying lesson is straightforward: AI tools work best when they’re designed to close specific gaps in a workflow, not replace the entire system.

The Practical Takeaway

If you’re tracking how AI gets applied in high-stakes, real-world environments, California’s wildfire system is one of the clearest examples available. It’s not theoretical. It’s operational, it’s layered, and it’s producing measurable outcomes — earlier detection, faster pre-positioning, and better situational awareness during active fires.

The tools themselves — computer vision, satellite remote sensing, predictive analytics, lidar — are not unique to wildfire management. The same principles apply anywhere you need to detect a problem early, model its trajectory, and position resources before conditions deteriorate.

That’s the pattern worth studying.