Methodology & Data Sources
FireRisk is built entirely on authoritative federal and state data. Here is exactly what we use, how current it is, how precise it is, and what carries legal weight versus what is an advisory model.
Methodology v2.0 (2026-06) · last updated June 2026
See our verified data & methodology audit →How the 0–100 score is calculated
Model v2.0 (2026-06)Your FireRisk is a transparent weighted blend of four independent federal datasets, each pulled live for your exact coordinates. Every factor is a real measurement — we do not derive multiple factors from one dataset, we exclude inputs that are merely correlated proxies of another, and we do not invent numbers for things a remote model can't see. The scale is consistent across all states, so a 60 in Colorado means the same as a 60 in California.
The federal model of wildfire risk to the buildings at your location — burn probability × intensity × susceptibility (point)
Distance-decayed proximity, size & recency — negligible beyond ~15 mi (point)
Fire spreads up to 8× faster uphill (point)
County-level wildfire risk score (0–100 national percentile) — coarser, so weighted lightly
No silent defaults
If a dataset has no data for your exact point, it is dropped from the formula and the remaining weights are renormalized to sum to 100% — never filled with a guessed "moderate."
Confidence shown
Every report shows a confidence level: High (both hazard models — WRC and FEMA — returned data), Moderate (one), or Limited (neither), plus the exact source count. A search that resolves only to a city/ZIP centroid (not a full street address) is capped at Moderate and flagged, since the point may not be your specific parcel.
Continuous, not binned
Every sub-score (RPS, fire-history proximity & recency, slope, fire size) is a continuous curve anchored to published breakpoints — no "cliffs" where a hair more distance or slope jumps the score. The USFS RPS index (raw 0–~1300) uses USFS's own rating anchors (8 / 25 / 60), so two "High" homes score differently. Your raw RPS value is shown on every report for audit.
What we deliberately do not put in the score
Two of the biggest drivers of whether a home survives a wildfire — its defensible space and its structure hardening (roof, vents, deck, siding) — cannot be measured remotely from any satellite or federal raster. Earlier versions of many risk tools (and ours) showed an assumed "baseline" number for these. We removed that: presenting an assumption as a measurement is the opposite of trustworthy. Instead, those home-specific factors are scored by the free mitigation plan generator, where you answer for your actual home. Your FireRisk reflects the location's hazard; the planner reflects the home’s readiness. You need both.
Every data source, dated and cited
National "Risk to Homes" rating, all states
California legal hazard designation
Wildfire risk rating context
Recorded fire history within 25 mi
Slope & elevation
Address → coordinates & county
Regulatory vs. advisory — and resolution
Most wildfire data — including the federal USFS and FEMA layers — are advisory models, not legal designations. The one binding exception we surface is California's CAL FIRE Fire Hazard Severity Zone, which legally drives building codes, defensible-space law, and seller disclosure. We label which is which on every report.
National layers are raster datasets (30–270 m), so they describe the area around your home, not the parcel alone. Your address determines which cells are sampled, but true parcel-level risk also depends on your specific structure and defensible space — which is why the report recommends an on-site assessment for High and Very High results.
What "Risk to Structures" actually decomposes into
Our heaviest-weighted input, USFS Risk to Potential Structures (RPS), is not a vague index — it is, by definition, annual burn probability × conditional risk to a structure, where conditional risk combines modeled fire intensity (flame length) with a structure's susceptibility response function. In other words it already integrates the two questions a wildfire scientist asks — how likely is fire here and how bad would it be — into one number, computed nationally at 30 m on one consistent method. The USFS publishes its components (Burn Probability, Conditional Flame Length, Conditional RPS) as separate layers; we deliberately score the integrated RPS rather than re-combining its parts, to avoid double-counting.
Validation & scientific basis
Every input traces to peer-reviewed federal science. The risk layer (USFS Wildfire Risk to Communities — Risk to Potential Structures) is built on the Rocky Mountain Research Station's FSim large-fire simulation system (Finney et al.), LANDFIRE fuels, and the Wildfire Risk to Communities methodology (Scott, Dillon, et al.) — the same models used in federal wildfire planning and the National Cohesive Wildland Fire Management Strategy. Fire history is the official NIFC InterAgency Fire Perimeter record (2000–2024); the county layer is FEMA's National Risk Index wildfire risk score; slope is derived from a high-resolution digital elevation model (Open-Meteo / Copernicus, with USGS 3DEP as an authoritative fallback). We add no proprietary "secret sauce" on top — the score is a transparent, reproducible weighting of these public datasets, and we publish the exact formula and the raw values (including your RPS) so any expert can audit it.
FireRisk is provided for informational purposes. Risk scores and savings figures are estimates, not guarantees of insurability, premium, or grant approval. Verify any regulatory designation with your local fire authority before construction or sale.
Full formula detail
Every parameter in every sub-score, so any expert can reproduce the calculation from your raw values.
Fire history sub-score (30% of total)
Base score = proximityWeight(closest fire) × recencyFactor(closest fire) × 92, then additive density & size bumps, capped at 100. Every curve is continuous — the values below are anchor points, interpolated linearly between (and clamped outside the range), so there are no jumps between bands. Recency is a multiplier (never above 1.0), so a single decades-old burn can't read "High" on its own, while a recent fire at the same distance is unaffected.
Proximity weight (anchors)
Density bonus (fires within 8 mi)
Size bonus — largest fire ≤ 12 mi (anchors)
Recency multiplier (closest fire)
Floor values
No-history floor is 6 (not 0) because long-unburned fuel loads can carry intense fire when conditions align. Incidents are capped at the nearest 14 to bound query size.
Slope & terrain sub-score (14% of total)
Slope is derived from five elevation samples: your address plus four cardinal neighbors ~100 m away, pulled in a single query from a high-resolution DEM (Open-Meteo / Copernicus, with USGS 3DEP as an authoritative fallback that returns slope directly). The maximum elevation difference between center and any neighbor is converted to degrees, mapped through a continuous curve (anchor points below, interpolated between). Values above 60° are capped (engineering bound to exclude terrain artifacts). If terrain can't be resolved for your point, slope is dropped from the formula and the remaining weights renormalized — it is never assumed flat.
8/100
0°
12/100
5°
25/100
10°
48/100
20°
68/100
30°
83/100
40°
96/100
60°
USFS WRC sub-score (44% of total)
Raw RPS value mapped through a piecewise linear curve anchored to USFS rating breakpoints. If the raw RPS is unavailable, the text rating midpoint is used as a fallback (Minimal=6, Low=26, Moderate=50, High=73, Very High=92).
RPS 0
→ 2
RPS 8
→ 30
RPS 25
→ 55
RPS 60
→ 78
RPS 150
→ 92
RPS 400
→ 98
Interpolated linearly between anchors. Values above RPS 400 are clamped to 98.
FEMA NRI sub-score (12% of total)
Uses the WFIR_RISKS field (wildfire Risk Index, 0–100 national percentile) from the FEMA NRI Counties FeatureServer, queried by your 5-digit county FIPS code and clamped to 0–100. If the county has no FEMA data — or FEMA rates it "Insufficient Data" / "No Rating" — this factor is excluded (not scored as a guessed middle value) and the remaining weights are renormalized.
Why USFS Wildfire Hazard Potential (WHP) is fetched but not weighted
WHP and WRC are both derived from LANDFIRE fuels and are strongly correlated — including WHP as a separate scored factor would double-count the fuels signal already embedded in RPS. WHP data is fetched for display context only; it carries zero weight in the 0–100 score. This is intentional, not an omission.
Home-value impact model
The assumptions behind the wildfire home-value calculator. It is an educational estimate, not an appraisal or per-address valuation — it models two transparent mechanisms from your risk tier, state, and an adjustable home value, never a fabricated AVM.
Mechanism 1 — insurance-cost capitalization
A higher wildfire insurance premium is a recurring carrying cost, so buyers capitalize it into a lower price: ΔValue ≈ annual wildfire surcharge ÷ cap rate. We use a conservative 7% capitalization rate (residential discount rates run ~5–8%; the high end understates the value hit per dollar of premium). The wildfire-attributable surcharge starts from a base homeowners premium of ~0.35% of home value, scaled by a per-tier surcharge factor (Low 0× · Moderate 0.12× · High 0.35× · Very High 0.7× · Extreme 1.0×) and a hard-market uplift in the states with the sharpest WUI premium increases (CA ×1.3, CO ×1.15). Calibrated against GAO-26-107867 (medium→high wildfire risk ≈ +8% premium) and FAIR-Plan pricing (~2–3× a standard policy).
Mechanism 2 — market risk discount
A conservative per-tier discount range applied to home value (Low 0% · Moderate 0–1% · High 2–4% · Very High 4–6% · Extreme 6–10%), capped at 10%. The ranges are anchored to the peer-reviewed California wildfire-hazard disclosure discount of ≈ 4.3% (up to ≈ 6% in Southern California — Land Economics 100(1), 2024 / RFF WP 23-26) and Redfin's high-risk-ZIP analysis (≈ 3.9%, 2020). This is the durable high-hazard effect, not the larger but transient drop right after a nearby fire, which typically recovers in 1–3 years.
Why the two are not added
The mechanisms overlap — a measured market/disclosure discount already embeds buyers pricing in higher insurance — so we do not stack them. The market discount range is the headline estimate; the capitalized insurance figure is reported separately as the mechanism explaining much of that discount. The relationship is cross-checked against Eastman, Kim & Zhou (2024), who find a ≈ 10% premium rise corresponds to a ≈ 4.6% price decline.
State median home values are adjustable defaults only. Individual homes vary widely with home hardening, views, lot, and local demand; this model is educational and is not an appraisal or financial advice. Full sources are listed on the home-value impact page.