Methodology

The primary dataset: NFIRS

Most of the incident-level statistics on FireVerge come from the National Fire Incident Reporting System (NFIRS), run by FEMA's U.S. Fire Administration. Each year fire departments across the country submit standardized records for the incidents they respond to — structure fires, wildland fires, vehicle fires, hazardous conditions, false alarms, and more.

FireVerge loads the five most recent reporting years (2021–2025) of the NFIRS Public Data Release. We use a rolling five-year window rather than a single year for two reasons: it smooths out year-to-year noise, and it compensates for the fact that the newest year is always incomplete on release.

From raw files to place pages

The NFIRS release is a set of large delimited text files linked by an incident key. Our pipeline:

• Parses the basic incident, address, fire-detail, casualty, and department files for each year.
• Translates every code — incident type, ignition cause, area of origin, and more — into its official NFIRS reference description. You never see a bare code like 111; you see "Building fire."
• Assigns each incident to a place by normalizing the reported city name (handling abbreviations and department-name suffixes) and resolving the county via ZIP-to-county and FIPS reference tables.
• Aggregates incidents up to the ZIP, city, county, and state level, then computes totals, causes, monthly patterns, property loss, and casualties for each.

Per-capita rates and grades

Raw counts favor big cities, so we normalize by population using U.S. Census figures to produce per-capita rates (for example, fires per 1,000 residents). This lets a small town and a large city be compared fairly.

Where FireVerge shows a composite fire-safety grade (A+ through F), it ranks a city against every other city we profile. We blend its yearly incidents per 1,000 residents (weighted 80%) and civilian fire deaths per million residents (20%), averaged over the reporting window, and grade by percentile: roughly the best 5% get an A+, and each letter below covers about the next 10%, down to F for the bottom 5%. A grade is a quick relative signal, not an official rating. Always read it alongside the underlying numbers and local context.

Wildfire, weather, and air data

Beyond NFIRS, FireVerge incorporates several federal datasets for wildfire and smoke context:

• NIFC / WFIGS — active wildfire incidents and historical fire perimeters.
• FPA-FOD (U.S. Forest Service) — decades of wildfire ignition records and causes.

Active-fire figures come from a snapshot of the NIFC feed, dated on every page that shows them, not a live connection. County wildfire risk comes from the FEMA National Risk Index. National Weather Service alerts, EPA AirNow air quality, and NASA FIRMS hotspots are not yet part of the site; where pages point to them, they link to those agencies directly.

See the Data Sources page for full details on each.

How we validate

Before publishing, we reconcile our aggregates against the U.S. Fire Administration's own published national summaries — matching national incident totals, building-fire counts, and civilian casualty figures for a reference year — and spot-check individual cities against known totals. This catches parsing errors and place-assignment mistakes.

Limitations you should know

• NFIRS is voluntary. Not every department reports, and completeness varies by department and state, so some incidents and places are undercounted.
• Missing years are flagged, not trended. When a city's count for a year is under 60% of its average for the window, a department almost certainly skipped or part-filed that year. We mark the year, show no year-over-year change for it, leave it out of grades, and keep the city out of year-over-year rankings.
• The newest year is provisional. The most recent reporting year rises for months after release as late reports arrive — expect current-year figures to be conservative.
• Place matching is imperfect. Messy free-text city fields mean a small share of incidents can be mis-assigned despite normalization.
• It's historical, not live. Incident statistics describe the past; they are not a real-time picture of current fire activity.