Methodology · component
Environmental Hazard — how we measure it
One component of the ZipVitals Score. This page covers only this component — what it is built from, how those sources reach your ZIP, where they stop, and how it enters the composite. For the full recipe and the whole weight table, see the main methodology page.
Effective weight 0.1368 of the composite · higher = worse
Where this sits in the Score
What a place is exposed to from the outside — air, water, industry, radon and noise. This is the widest source set in the Score, because no single measurement captures environmental burden.
Feeds domain
Environmental Hazard
0.1368
effective weight · nominal 0.13 · higher = worse
Environmental Hazard is 1 of 9 scoring domains. Life expectancy is held out of the composite as an independent validator, so the nine renormalize to sum to 1.00 and this domain’s share lands at 0.1368.
It is built from 9 sources, listed below with what each one contributes and what it cannot tell you.
The sources feeding Environmental Hazard
Each source is labeled with what kind of number it is — observed, model-based, or administrative — because a modeled risk surface and a counted facility are not the same kind of fact. We name which is which.
EPA Toxics Release Inventory (TRI)
AdministrativeUS Environmental Protection Agency
How many facilities that report releasing listed toxic chemicals are near this ZIP, and how close the nearest one is.
The honest limit: Only facilities above a federal reporting threshold appear at all — a smaller polluter is simply not in the register, and the gaps are systematic rather than random.
EPA ECHO / FRS — contaminated sites
AdministrativeUS Environmental Protection Agency
Distance to the nearest seriously contaminated site — Superfund sites, final and proposed, and sites under corrective action for hazardous-waste releases.
The honest limit: One physical facility counts once even when it appears on several EPA programs, and a site EPA has cleaned up and delisted is not counted as a current hazard.
EPA AirToxScreen
Model-basedUS Environmental Protection Agency
Modeled added lifetime cancer risk from breathing outdoor air, summed across roughly 140 airborne toxics — the pollutants that fine-particle mass alone does not capture.
The honest limit: Not a count of cancer cases and not a prediction about any person. Concentrations are modeled block by block, but the exposure factor that turns concentration into risk is applied uniformly across a whole census tract, so within-tract differences are not resolved.
EPA RSEI — toxicity-weighted release burden
Model-basedUS Environmental Protection Agency
A modeled burden score that weights reported toxic releases by toxicity, by where the plume goes, and by the population it reaches — surfaced as a national percentile.
The honest limit: Unitless and relative: it is neither a measured concentration nor a risk estimate. We surface the per-capita score, never the raw release total, so a large industrial ZIP is not double-counted for its size.
CDC EPHT — fine particles (PM2.5)
Model-basedCDC Environmental Public Health Tracking
Days over the fine-particle standard — the pollutant most consistently tied to heart and lung outcomes.
The honest limit: Counted from a modeled national air-quality surface, not a monitor in this ZIP, because most ZIPs have none. It is a single year rather than a long-run average, so a bad wildfire or inversion year shows up in full.
Indoor radon potential
Model-basedLi et al. (Open Science Framework) with USGS geologic infill
Modeled indoor radon concentration — the naturally occurring soil gas that is a leading cause of lung cancer among people who have never smoked.
The honest limit: A modeled area value, never a test of any home; only a radon test tells you about a specific house. Where the modeled surface stops, a coarse three-bucket geologic classification stands in, and the two are never blended into one number.
BTS transportation noise
Model-basedUS Department of Transportation, Bureau of Transportation Statistics
The share of the ZIP exposed above a transportation-noise threshold, from a modeled road and aviation noise surface.
The honest limit: Framed as the share above a threshold, never as an average: most of the country sits below the modeled floor and carries no value at all, so an average would be an artifact of the map rather than a fact about the place.
EPA SDWIS — drinking-water violations
AdministrativeUS Environmental Protection Agency
Violations recorded against the public water systems that serve this ZIP — health-based limits and monitoring requirements alike.
The honest limit: Public water systems only: roughly 30% of the population drinks from private wells that no federal program monitors, and the service-area boundaries themselves are modeled. A ZIP with no violation may simply have no reporting system.
USGS private-well water quality
Model-basedUS Geological Survey
Modeled likelihood that groundwater used by private wells exceeds a reference level for arsenic, nitrate or PFAS.
The honest limit: A likelihood, never a stated concentration and never a test of any well. Private wells are federally unregulated and untested at scale, so no observed value exists at this resolution. Suppressed where almost nobody drinks from a private well, and published for the continental US only.
How these reach your ZIP
Every value is a fact about the ZIP itself. A source that is not published at ZIP level is rolled up against a published Census boundary file, weighted by population or land area as appropriate — never inherited from the county or state and rebranded as the ZIP’s. Where a legitimate ZIP value cannot be produced, the metric is shown as unavailable.
- EPA Toxics Release Inventory (TRI)
- Facility locations, with distance measured to the ZIP boundary rather than its center — so a plant just outside a large rural ZIP reads near zero.
- EPA ECHO / FRS — contaminated sites
- Facility locations, measured to the ZIP boundary.
- EPA AirToxScreen
- Census block → ZIP, population-weighted.
- EPA RSEI — toxicity-weighted release burden
- EPA's model grid → census block → ZIP, population-weighted.
- CDC EPHT — fine particles (PM2.5)
- Modeled at census-tract level, then rolled up to the ZIP.
- Indoor radon potential
- Published at ZIP level where the modeled surface reaches; the geologic infill is an area-dominant overlay on the ZIP boundary.
- BTS transportation noise
- A 30-meter modeled raster, summarized inside the ZIP boundary.
- EPA SDWIS — drinking-water violations
- Water-system service areas, intersected with the ZIP boundary.
- USGS private-well water quality
- Modeled groundwater surfaces at roughly one-kilometer resolution, weighted by the private-well population inside the ZIP.
Coverage & limitations — for this component
This is the honest part. Every limit below is a published, named limit of this component specifically, so you can judge how far to trust one of its numbers. We treat limitations as a trust signal, not fine print.
What these numbers actually are
- Air-toxics cancer risk
- An EPA model estimate of added lifetime cancer risk from breathing outdoor air, summed across roughly 140 airborne toxics, and surfaced alongside a relative national ranking. It is not a count of cancer cases and not a prediction about any person. Resolution: Concentrations are modeled block by block, but the exposure factor that converts concentration into risk is applied uniformly across a whole census tract — so within-tract differences in the final figure are not resolved.
- Days over the fine-particle (PM2.5) standard
- Counted from a modeled national air-quality surface that fuses an atmospheric model with a sparse monitor network — not a reading from a monitor in this ZIP, because most ZIPs have none. It is a single year (2021, the latest year of daily data published), not a long-run average, so a bad wildfire or inversion year shows up in full. Resolution: Modeled at census-tract level, then rolled up to the ZIP.
- Contaminated-site proximity
- Distance from the ZIP's boundary to the nearest seriously contaminated site on EPA's federal registers — Superfund sites (final and proposed) and sites under corrective action for hazardous-waste releases. One physical facility counts once even when it appears on several EPA programs, and a site EPA has cleaned up and removed from the list is not counted as a current hazard. Resolution: Facility point locations, measured to the ZIP boundary rather than its center — so a site just outside a large rural ZIP reads near zero.
- Private-well water quality (arsenic, nitrate, PFAS)
- A modeled likelihood that groundwater in this area exceeds a reference level — not a test of any well, and not a concentration. Private wells are federally unregulated and untested at scale, so no observed value exists at this resolution; a modeled likelihood is the honest alternative to publishing nothing. Resolution: A modeled national groundwater surface, rolled up to the ZIP.
Where this component’s data stops
- Drinking-water violations (SDWIS)
- EPA SDWIS covers public water systems only — the ~30% of the population on private wells is not represented. scope-limited Gated from percentile ranks pending an empirical threshold.
How it enters the Score
Every metric in this component goes through the same pipeline as the rest of the Score, then contributes to the Environmental Hazard sub-score — which contributes its 0.1368 to the composite.
- Standardize. Each metric becomes a z-score against the national distribution of ZIP codes, so a concentration, a rate and a distance become comparable.
- Cap outliers. Each z-score is capped, so one extreme ZIP cannot dominate the scale.
- Build the sub-score. The capped metrics combine into the Environmental Hazard sub-score, weighted by source and then equally within a source — so a source that publishes many metrics cannot swamp one that publishes few.
- Weight into the composite. Environmental Hazard contributes 0.1368 to the domain-weighted sum, which is rescaled to the 0–100 Score.
Go deeper
This is one component of the method. Follow any of these for the whole picture, or for this component in context.