Methodology
Where every number here comes from
Nothing on this site is written by a model, estimated by us or inferred from something adjacent. Every figure on every page is a value from a federal data release, joined to a county and displayed. If a source does not report a value, the page shows a dash rather than a guess.
The spine is the US Census county list. Every other dataset is joined onto it by FIPS code where the source publishes one, and by state plus normalised county name where it does not.
The sources
Five datasets, and what each cannot tell you
EPA Map of Radon Zones
Current EPA releaseRadon potential zone for all 3143 counties
Built from indoor radon measurements, geology, aerial radioactivity, soil parameters and foundation types. The EPA states plainly that the map should not be used to decide whether an individual home needs testing - every home should be tested regardless of zone.
EPA Air Quality System
2021-2025Annual AQI summaries by county
Only the 1012 counties with a reporting monitor appear. The other 2131 are unmeasured, not clean. A county with fewer than 365 monitored days did not measure the whole year, and the daily AQI is set by whichever pollutant scored highest.
CDC PLACES
BRFSS 2022 and 2023, 2024 and 2025 releasesAdult asthma, COPD and current smoking prevalence
Model-based small-area estimates, not counts of diagnosed people. Each carries a confidence interval this site does not reproduce. Kentucky and Pennsylvania are absent from the current release and are filled from the prior one; every county records which release it came from.
NCI State Cancer Profiles
2018-2022 incidence, 2019-2023 mortalityLung and bronchus cancer incidence and mortality
Age-adjusted rates per 100,000 to the 2000 US standard population, pooled over five years. Counties with too few cases are suppressed and appear here as a dash. A rate describes a population, never a person.
US Census Bureau
2020 vintageCounty names, FIPS codes and county adjacency
The authority for which counties exist and which border which. Used as the spine every other source joins onto.
The join
How counties were matched
County identity is messier than it looks, and the awkward cases are documented rather than quietly dropped:
Virginia independent cities
Fairfax, Franklin, Richmond and Roanoke each name both a county and a separate independent city. The EPA file distinguishes them only by a flag on the state column, so the city rows are matched separately. Bedford, Clifton Forge and South Boston reverted to towns and no longer exist as counties; they are dropped, and their counties of the same name are not.
Alaska's redrawn census areas
The radon map predates several Alaska boundary changes. Where one zoned area became more than one county - Valdez-Cordova into Chugach and Copper River, Wrangell-Petersburg into Wrangell and Petersburg, Skagway-Hoonah-Angoon into Skagway and Hoonah-Angoon - every successor inherits the zone assigned to the ground it sits on. Wade Hampton and Shannon are matched to their renamed successors, Kusilvak and Oglala Lakota.
Spelling between agencies
The EPA air files write "Saint Louis" and "Baltimore (City)" where the Census writes "St. Louis County" and "Baltimore city". Doña Ana and LaSalle differ by a diacritic and a space. These are normalised, and the two that resist normalisation are mapped by hand.
Suppressed and missing values
NCI suppresses county rates built on too few cases, and CDC PLACES omits two states from its current release. Neither is filled with a model or a state average. A dash on this site means the source does not report it.
Derived figures
The three things we calculate
| Figure | How |
|---|---|
| National percentile | The share of counties with a lower value on that measure, over the counties where the source reports one. Counties without a value are excluded rather than treated as zero. |
| Days above moderate | The sum of the unhealthy-for-sensitive-groups, unhealthy, very unhealthy and hazardous day counts in the EPA annual summary. No reweighting. |
| State prevalence averages | County prevalence weighted by county population, over counties where both are reported. This is not the same as the state's own BRFSS estimate and is not presented as one. |
Limits
What this site is not
Reuse
Take the data
The joined county table and the five-year air quality series are published as CSV under the same terms as the federal sources they come from. Attribution to the original agencies is appreciated; attribution to us is not required.