Well water data & methodology

What we model

• Arsenic — probability of exceedance

  A USGS national model estimates the probability that a private well in the county area exceeds 1, 5, and 10 µg/L (the EPA limit is 10 µg/L). The thresholds are fit as three independent models, so we only show the ≥5 µg/L figure beneath the ≥10 µg/L headline when the pair is coherent.

• Nitrate — predicted concentration

  A USGS model predicts the nitrate (as N) concentration at domestic-supply depth for the county area. The EPA limit is 10 mg/L.

• Private-well population

  Modeled count of people and households on private (domestic) wells, from the USGS domestic-well population release combined with Census geography.

• Uranium & radon (New Hampshire only)

  New Hampshire publishes a state-specific groundwater model for uranium and radon; we surface it only for NH counties and label every other state as 'no state model available' rather than guess.

• Primary aquifer

  The dominant USGS principal aquifer mapped under the county — area context for where wells draw from, not a determination for any single well.

• Agricultural land use (crop pressure)

  The row-crop fraction of the county area (USDA Cropland Data Layer). Intensive row-crop land raises nitrate risk, so it is a predictor we surface — never a measurement of nitrate in a well.

• PFAS — public-system proxy

  PFAS detections in nearby PUBLIC water systems under EPA's UCMR5 program. This is a regional proxy, explicitly NOT a measurement of any private well.

• Observed samples & confidence

  Reported monitoring samples from the EPA/USGS Water Quality Portal give a local-density signal that drives the per-page data-confidence chip (sparse vs. well-corroborated).

How we build & validate it

• County boundaries are reprojected into each raster's native coordinate system before zonal averaging, so every figure is read from the model exactly where the county sits.
• We never fabricate a number. A county missing a required national metric (arsenic, nitrate, or population) is not published at all; an optional layer with no real value is shown as "not available", never as zero.
• Every served county passes a schema + honesty validator before it can reach a page, so a malformed or partial record becomes a 404 rather than a misleading "0%".
• Community-contributed lab results are only ever shown as area-level aggregates once at least ten wells have reported — never as an individual result.

Data sources

Every figure traces to a public federal dataset. The exact sources that contributed to a given county are listed on that county's page.

• USGS arsenic probability-of-exceedance modelCounty arsenic exceedance probabilities (1, 5, and 10 µg/L).
• USGS nitrate predicted concentration (domestic-supply depth)County predicted nitrate concentration.
• USGS domestic (private) well population densityEstimated people and households on private wells.
• USGS principal aquifersThe dominant principal aquifer mapped under each county.
• USGS New Hampshire uranium/radon groundwater modelNew Hampshire-only uranium and radon exceedance probabilities.
• USDA NASS Cropland Data Layer (CDL)Row-crop land-use fraction — a nitrate-risk predictor, not a measurement.
• EPA UCMR5 (Fifth Unregulated Contaminant Monitoring Rule)PFAS occurrence in nearby PUBLIC water systems — a regional proxy.
• EPA / USGS Water Quality PortalObserved monitoring samples used for local context + confidence.
• US Census TIGER/LineCounty boundaries / geography for the zonal models.