Wildfire Risk · Methodology

How we calculate your wildfire risk

A Satelife score is a weighted blend of seven public-and-private data layers. This page lays out every input, every weight, the exact formula, and — just as important — what we deliberately don't claim.

Exploded stack of the data layers Satelife combines into one wildfire score: AI risk model, fire weather, fuel types, burn probability, fire history, vegetation and terrain, and the property with its fire perimeter. — Every layer is read at one address, then combined into a single score.

Three things determine parcel-level risk

Wildfire risk at a single property is the combination of three things: how likely fire is to happen there, how much of the property is in the way, and how well the structure and its surroundings can survive it. Every Satelife report is built on those three questions.

Hazard

How likely a wildfire is to occur in the neighbourhood around the parcel — driven by official hazard maps, recent fire history, and local fire weather.

Exposure

How directly the parcel is in the path of a fire if one starts — driven by fuel load and continuity, terrain shape, and prevailing wind patterns.

Vulnerability

How well the structure and its immediate surroundings can survive a fire — driven by defensible space, building hardening, and brigade response time.

Built from public data and real imagery

Every report combines a handful of independent inputs about the property and its surroundings. Each is assessed on its own, against public datasets and recent imagery, so two reports on the same parcel land in the same place.

Public hazard data

Official fire-hazard zone

The government has already mapped where fire is more likely. We start from that map, not from scratch.

Most regions publish an official fire-hazard classification for every parcel — a public, regulator-trusted dataset that already reflects decades of fire-behavior research. It is the single strongest external signal, and it is what insurers, lenders, and local governments already rely on. We anchor our analysis to it.

USFS Wildfire Risk to Communities

Burn probability & flame length

How likely fire is to reach this spot in a given year — and how intense the flames would be if it did.

On top of the official hazard band, we read the USFS Wildfire Risk to Communities model: an annual burn-probability surface paired with a conditional flame-length layer. Together they answer two separate questions — how often fire is likely to arrive at the parcel, and how severe it would be when it does. A low-probability parcel that would burn intensely is a very different risk from a high-probability one that would only ever see low flames.

Satellite-derived land cover

Vegetation & fuel load

Fire needs fuel. We measure how much there is, how continuous it is, and how dry.

Continuous, dense, dry vegetation is the single strongest local driver of fire intensity and spread. We measure fuel density and continuity in a tight ring around the parcel from public satellite-derived land-cover datasets. Fragmented, low-load fuels read as low; continuous chaparral and dense conifer stands read as high.

Historical fire records

Fire history nearby

A parcel that has recently had fire pass nearby is statistically more likely to burn again.

We look at every recorded fire perimeter in a wide window around the parcel and score it by recency and proximity. A fire that crossed close by a few years ago weighs much more than one tens of kilometres away half a century ago. This captures real recurrence patterns that pure-fuel models miss.

Computer vision on imagery

AI imagery analysis

A computer-vision model reads your property the way a fire inspector would — but at scale.

A vision model reviews aerial and street-level imagery of the parcel and scores defensible space, roof material and condition, eaves, vents, deck construction, and ember-trap features. Confidence scores route low-certainty cases to a human reviewer. This is the layer that reflects what you specifically did or did not do to your property.

Public elevation models

Terrain & topography

Fire climbs uphill fast. Slope, aspect, and exposure all change the local risk physics.

Steep south- and west-facing slopes dry out faster and accelerate fire spread — a property at the top of a long chaparral-covered slope is materially more exposed than the same house on flat ground. We derive that geometry deterministically from public elevation models, so the same parcel scored twice always lands in the same place.

Climate & fire-weather data

Fire-weather & climate

Fuel and terrain set the stage, but weather lights the match. We fold in the long-run fire-weather pattern your parcel sits inside.

A wide grid cell around the parcel is summarised into a long-run fire-weather signal — seasonal wind rose, mean and peak wind speed, drought tendency, and a fire-weather index percentile. That signal is what separates a fuel-rich but cool coastal parcel from an identical-fuel inland one. It is a long-run climate pattern, not an operational forecast — but it materially shifts how exposed the same fuels really are.

Deeper context for higher-risk parcels

When a property warrants it, the report includes additional layers that show the wider context — real brigade response time and a physics-based view of how a plausible fire would behave on the ground. The cards below are taken straight from a real report.

Road accessibility card showing the fastest brigade route on a satellite map with surrounding stations and access roads

Response & access view

Brigade response & road access

How quickly a real fire crew can actually reach the property, and how easy it is to get out.

We look at every fire station within a reasonable radius and compute a realistic drive time over the actual road network — not straight-line distance. That same analysis surfaces single-egress and narrow-access patterns that materially worsen evacuation risk. A property with a 5-minute brigade response and a clean two-way exit is fundamentally less exposed than one with a 25-minute response and a single private lane.

Spread-simulation view

Fire-spread simulation

A physics-based view of what a plausible fire near the parcel actually looks like.

For higher-risk parcels we run a spread model over the local fuel, slope, and prevailing wind. The output is an ignition-probability surface plus a small ensemble of synthetic fires — so the report shows not only "how bad is the risk" but "what does a plausible fire here actually do, and where does it travel".

From a number to a decision

Four action tiers

Every report ends in one of four tiers, mirroring the same language official hazard maps already use. The tier is the thing you act on — the underlying number is just the explanation.

Low

No obvious wildfire-driving features at the parcel. Standard maintenance and local code compliance is normally enough.

Moderate

One or two contributing factors — some fuel load, mild slope, or aging history. Targeted mitigation is meaningful.

High

Several factors stack. Insurers and lenders are likely to ask for evidence of mitigation. A defensible-space plan is the priority.

Very High

Aligned with the highest official hazard band. Active mitigation, home hardening, and a documented evacuation plan are essential.

Regional calibration

Calibrated to each region

Fire physics is universal, but data availability and the language regulators use are not. Each region the service supports is calibrated to fit local data and local terminology, so the report reads the way a local insurer or fire authority expects.

California

Live

California has a strong, regulator-trusted public hazard classification that insurers and lenders already use. Reports are tuned to anchor on that classification and produce a tier that slots straight into existing local vocabulary.

Portugal

Pilot

Portugal does not publish a jurisdiction-wide hazard map of the same kind, so the analysis leans more on vegetation and fire history. The output is framed as an operational exposure view, matching the terminology Portuguese fire authorities and civil protection already use.

See the methodology
applied to your address

Get a free sample report for any property and see how each data layer stacks up at your parcel.

Get your free report →
Or see the full 14-step pipeline →