OPEN-CORE · FREE API · BUILT IN BC

Wildfire intelligence
forged in fire.

INFERNIS is a machine learning engine that predicts wildfire risk across all of British Columbia. Updated daily at 1km or 5km resolution. Free to use. Open-core. Built by British Columbians, for British Columbia.

Get Your Free API Key See How It Works
84K+
Grid cells across all of BC
21
Open data sources fused
28
Engineered features per cell
0.974
AUC-ROC classification
Our Story

Born from smoke-filled summers

If you've lived through a BC summer in the last decade, you know the feeling. The sky turns orange. The air quality warnings stack up. You check the BC Wildfire dashboard obsessively, watching new fire starts appear faster than crews can respond.

The 2023 fire season burned 2.84 million hectares — more than the previous two record-setting years combined. Over 35,000 British Columbians were forced from their homes. Suppression costs alone exceeded $1 billion. And every year, the fire season starts earlier and burns longer.

INFERNIS was born from the frustration of watching this happen and asking: why isn't the best available science being turned into tools that communities and governments can actually use? Canada publishes some of the richest open environmental data in the world — ERA5 weather reanalysis, decades of fire records, daily satellite imagery, lightning detection networks. The science to turn that data into predictions has been published and validated. What was missing was an engineered system that brings it all together.

So we built one. INFERNIS fuses 21 open data sources with modern machine learning to deliver daily fire risk predictions for every corner of British Columbia — available to anyone, through a free API.

"The data exists. The science is proven. We're just connecting the dots — and giving the results away." — The INFERNIS Team, Vancouver, BC
BC FIRE SEASONS — HECTARES BURNED
2015
2017
2018
2023
2023 eclipsed all records — 2.84 million hectares burned.
How It Works

From satellites to your screen,
in plain English

Every day at 2:00 PM Pacific, INFERNIS wakes up and runs through a four-stage process. No humans need to press any buttons.

STEP 01

Gather the Data

INFERNIS pulls the latest weather observations, satellite imagery, soil moisture at four depths, vegetation indices, and lightning strike data from 21 open government and scientific sources — covering all of BC.

STEP 02

Compute Fire Weather

Using Canada's official Fire Weather Index system (the same science used by fire agencies since the 1970s), INFERNIS calculates six moisture codes and fire behavior indices — from surface litter drying to deep organic soil drought.

STEP 03

Run the ML Models

Two machine learning models analyze each cell. XGBoost evaluates 28 features per cell (AUC 0.974). A CNN captures landscape-scale spatial patterns (AUC 0.815). Together they produce calibrated fire occurrence probabilities.

STEP 04

Publish via API

Results are calibrated to each of BC's 14 biogeoclimatic zones, scored from VERY LOW to EXTREME, and published through a REST API. Anyone can query fire risk for any location in BC.

What goes in

Temperature, humidity, wind, precipitation, soil moisture at four depths, NDVI vegetation greenness, snow cover, leaf area index, terrain elevation and slope, fuel type classification, distance to nearest road, lightning strikes, and 11 fire seasons of historical records.

What comes out

A fire risk score between 0 and 1 for every grid cell in BC, a six-level danger classification (VERY LOW through EXTREME), contributing weather and fire weather data, ecological context, and up to 10-day forecast trajectories.

How accurate is it

AUC-ROC of 0.974 in 5-fold cross-validation. Walk-forward temporal backtesting (training on prior years, testing on held-out year) yields AUC 0.90–0.93 across six seasons (2019–2024). Brier score of 0.036 confirms well-calibrated probabilities.

Danger Classification

Six levels. One clear picture.

Risk scores are mapped to a familiar six-level system, calibrated independently for each of BC's 14 biogeoclimatic zones so the levels reflect real conditions in each ecosystem.

Very Low
0.00 – 0.05
Negligible risk. Wet or snow-covered conditions.
Low
0.05 – 0.15
Minor risk. Fires unlikely under current conditions.
Moderate
0.15 – 0.35
Elevated. Fires possible with an ignition source.
High
0.35 – 0.60
Significant. Fires likely to spread if ignited.
Very High
0.60 – 0.80
Severe. Aggressive fire behavior expected.
Extreme
0.80 – 1.00
Critical. Explosive fire growth potential.
Open Core Model

Free to use. Open to build on.

The INFERNIS engine is open-core. The base system — the models, the pipeline, the API — is available for anyone to download, deploy, and use. Commercial implementations and custom integrations are available on request.

Open Core — Free

The Engine

Download the full INFERNIS prediction engine. Deploy it yourself, study the code, contribute improvements, or build your own applications on top of it.

  • Full ML pipeline (XGBoost + FireUNet CNN)
  • Data ingestion from all 21 open sources
  • FWI computation engine
  • FastAPI REST server
  • BC grid system with PostGIS
  • Free API access (50 requests/day)
  • Community support via forum
Commercial

Custom Implementations

Need INFERNIS tailored to your organization? We build custom deployments, integrations, and applications for companies and government agencies.

  • Custom API deployments with SLA guarantees
  • Integration into your existing GIS and analytics systems
  • Custom dashboards and alerting applications
  • Enterprise-grade uptime and dedicated support
  • Additional provinces and custom grid resolutions
  • Webhook-based alerts for threshold exceedances
  • Historical analysis and seasonal outlooks
The API

One call. Full picture.

Query fire risk for any location in British Columbia. Get the risk score, danger level, fire weather indices, current conditions, and ecological context — all in a single JSON response.

GET /v1/risk/49.25/-121.77 
{
  "risk": {
    "score": 0.72,
    "level": "VERY_HIGH",
    "color": "#EF4444"
  },
  "fwi": {
    "ffmc": 91.2,  // Fine fuel moisture
    "dc": 412.0,    // Drought code
    "fwi": 34.6     // Fire weather index
  },
  "conditions": {
    "temperature_c": 32.1,
    "wind_kmh": 24.5,
    "soil_moisture": 0.12,
    "snow_cover": false
  },
  "context": {
    "bec_zone": "IDF",      // Interior Douglas-fir
    "fuel_type": "C7",      // Ponderosa Pine / Douglas-fir
    "elevation_m": 845
  }
}
Get Your Free API Key
Data Foundation

Built entirely on open data

Every data source INFERNIS uses is freely available from Canadian federal agencies, the Province of BC, and international scientific organizations. No proprietary data subscriptions required.

ERA5 Weather Reanalysis

Gridded global weather from ECMWF. Temperature, humidity, wind, precipitation, evapotranspiration, and soil moisture at four depths — hourly since 1940.

Satellite Imagery

MODIS and Sentinel-2 via Google Earth Engine. NDVI vegetation greenness, leaf area index, snow cover, active fire detection, and burn severity.

Lightning Detection

Canadian Lightning Detection Network (CLDN) flash density grids. Lightning causes approximately 60% of BC wildfire ignitions.

Historical Fire Records

Canadian National Fire Database (CNFDB) point-of-origin records and BC Fire Perimeters. Eleven fire seasons of labeled training data (2015–2025).

Fuel & Vegetation

CFFDRS fuel type maps, Biogeoclimatic Ecological Classification (BEC) zones, and real-time vegetation health indices from satellite observations.

Terrain & Infrastructure

Canadian Digital Elevation Model (CDEM) providing elevation, slope, aspect, and hillshade. BC Digital Road Atlas for distance-to-nearest-road.

Who It's For

From fire crews to your phone

Fire Services & Government

Pre-position crews based on predicted risk. Plan evacuations proactively. Allocate aerial resources before fires start, not after.

Insurance & Finance

Property-level wildfire underwriting. Portfolio exposure analysis. Parametric insurance products triggered by ML-predicted risk.

Utilities & Infrastructure

Transmission corridor risk assessment. Preventive de-energization decisions. Vegetation management prioritization along power lines.

Developers & The Public

Build fire risk into your app. Community dashboards for municipalities. Integration into weather apps, outdoor recreation tools, and mapping platforms.

Deep Dive

Read the White Paper

For a complete technical breakdown of the INFERNIS system — the hybrid ML architecture, 28-feature data fusion pipeline, per-BEC-zone calibration, walk-forward backtesting methodology, and the 10-day forecast engine — read the full white paper.

Download White Paper (PDF)
BUILT IN BC

Open intelligence for environmental risk.

INFERNIS is built in British Columbia — free for the public, available for custom implementation by companies and governments.

Get in Touch GitHub