Well water contamination by zip code: arsenic, nitrate, radon and more

The headline figure for each contaminant is the share of tested wells over the standard: of the wells USGS sampled in your county, how many had a most recent result above the limit. It is the closest thing public data offers to “how likely is my well to have this problem?” — but it describes wells USGS chose to sample, not a random draw of homes.

• 10% or more over puts the contaminant at the top of the “tests worth ordering” list for that county.
• Fewer than 5 tested wells and we show a count (“1 of 3”) instead of a percentage, and hatch the county on the map. No data is not good news: it means nobody measured.
• Median and 90th percentile are concentrations across the same wells. A figure like “<1” means the middle well was below the lab’s reporting limit.
• Confidence is High with 30+ tested wells including samples since 2000, Low under 5 wells, and Moderate between.

The standards we compare against

Private wells are not covered by these rules — they apply to public water systems — but they are the yardstick labs and health departments use for well results too. The water test interpreter judges your own lab report against the same numbers.

Across every well in the data, one well counted once. Treat these as a description of what USGS sampled, not a national rate for private wells: studies went where problems were suspected as often as not.

“Counties” is how many of the 3,142 US counties have at least one tested well for that contaminant. Nitrate and fluoride have been measured almost everywhere; uranium and radon in fewer counties and with far fewer wells in each (617 counties have 5+ tested wells for radon, against 2,468 for nitrate), so a blank or hatched county on those maps says more about where USGS worked than about the water.

The ten counties with the highest share of tested wells over each standard, among counties with at least 20 tested wells, and the states with the highest share among states with at least 50. Check the wells and years columns before quoting a county: some figures rest on a single study.

Share of tested wells over the standard used on the map, by state. A dash means fewer than 50 tested wells in the state, too few for a state figure.

Small figure under each share: wells tested. Bold: 10% or more over.

Lead. USGS samples are usually taken at the wellhead or from a tap run until the water is fresh from the aquifer, so they say little about lead from your own pipes, solder and fixtures, which is where most lead in drinking water comes from. EPA’s lead figure is an action level for public water systems — a trigger for corrosion control, set at 15 µg/L and lowered to 10 µg/L from November 1, 2027 — not a health-based limit. If your home has older plumbing, test a first-draw sample at the kitchen tap whatever the county shows.

Radon. There is no final federal limit for radon in drinking water. In 1999 EPA proposed a 300 pCi/L limit, or 4,000 pCi/L for systems in states running programs to cut radon in indoor air, and the rule was never finalised. EPA estimated that water at 4,000 pCi/L adds about 0.4 pCi/L to the air in a home, so radon in water matters mainly as a contributor to indoor air radon. We show both proposed figures and label them as proposed everywhere.

• Order the tests the county data points at. A standard lab panel covers bacteria, nitrate and pH; arsenic, manganese, uranium, fluoride and lead are usually add-ons, and radon needs its own sealed vial. See what a well water test costs and how to take the sample.
• Test every year regardless. The CDC recommends testing a private well at least once a year for total coliform bacteria, nitrates, total dissolved solids and pH — none of which a county figure can stand in for.
• When the results come back, enter them in the water test interpreter. For arsenic, nitrate, fluoride, uranium and lead, the usual fix is reverse osmosis at the drinking tap; manganese is handled by an iron and manganese filter.
• Hard water too? The water hardness map uses the same county data source for hardness.

Source. Groundwater results collected by the U.S. Geological Survey and published through the Water Quality Portal (USGS, EPA and the National Water Quality Monitoring Council), for sites typed as wells in all 50 states and DC, downloaded September 26, 2026. The portal’s USGS feed runs to 2023; newer USGS samples are published through a separate service not included here. The county comes from each well’s station record, and zip codes are matched to counties with the Census Bureau’s 2020 ZCTA-to-county file, exactly as on the hardness map.

What was measured. Only these USGS parameter codes, so bed-sediment, suspended-only and uncertainty records under the same names never enter the data. When one sample has results under more than one code, the first listed wins (filtered before unfiltered; nitrate as N before nitrate plus nitrite before nitrate reported as NO₃).

Nitrate is reported as nitrogen (mg/L as N), the basis of EPA’s 10 mg/L limit. We prefer USGS’s nitrate-as-N result, then nitrate plus nitrite as N (nitrite is normally a tiny fraction in well water, and EPA’s limit for the two combined is also 10 mg/L as N), and convert a result reported as NO₃ by ×0.2259 only when the unit or code says NO₃.

Units and filters. Every unit the portal returned for these codes was converted explicitly; results in any other unit, or with none, were dropped. We also dropped quality-control samples (blanks, replicates, spikes), results marked rejected or flagged for contamination, non-detects with no reporting limit, and wells whose names identify them as landfill or contamination-monitoring wells or piezometers.

“Kept” counts one result per sample, after 426,971 duplicate results for the same sample under a second parameter code were set aside. Lead also leaves out 62,986 results from before 2000 (explained below).

Non-detects. A result reported as below the lab’s reporting limit (“<1 µg/L”) counts as not exceeding when that limit is at or below the standard. When the limit is above the standard — as in some older results reported as “<50” or “<100” — the result cannot say either way, so it is left out for that standard; a well with only such results is left out of the share and the card says how many. Medians and percentiles never invent a value for a non-detect: we compute them with non-detects at zero and at their limit, and when the two differ we show the upper figure as “<” or “≤”.

One well, one vote — the most recent result. Each well counts once, represented by its most recent result that can be judged against the standard (if a well has two that day, the higher). We use the latest rather than each well’s median because it is the closest to what the well delivers now (nitrate in particular changes over time), because a median of results that are partly “below the limit” is not well defined, and because most wells were sampled only once anyway. Medians, percentiles and shares all use that same result per well.

Lead before 2000. USGS lead results from before 2000 are left out. Of 62,986 older results, 3% were over 15 µg/L, against <1% of 33,214 since — a drop too abrupt to be a change in groundwater, and many 1970s results are coarse round values such as 100 µg/L. Older trace-metal sampling and lab methods are the likelier explanation, so including them would overstate lead. Arsenic, manganese and fluoride results show no such break by decade.

Thresholds. A county needs 5 tested wells for a percentage, 20 for the rankings, and a state 50 for the state table.

Standards, checked against EPA on September 26, 2026: National Primary Drinking Water Regulations (arsenic, nitrate, uranium, fluoride MCLs; lead action level and its November 1, 2027 revision), secondary standards (manganese, fluoride), manganese health advisory, and the 1999 proposed radon rule.

Limits. Sampled wells are not a random sample of homes: USGS picked wells for studies — some to map an aquifer, some to follow a known problem, some shallow monitoring wells under farms or suburbs — so a county’s share can run high or low against its private wells. Many results are decades old. A county is not your well: neighbouring wells can differ many times over with depth and rock. Lead is mostly a plumbing question and radon has only a proposed standard. This data points to which tests are worth ordering; it is not a substitute for testing your own well.

The full county table — every county and contaminant with wells, samples, wells judged and over each standard, median, 90th percentile and years — is free to download and reuse under CC BY 4.0: credit Well Water Fixes with a link to this page. The underlying USGS data is in the public domain.

Download the CSV (1.6 MB)