Public well tests, added up by county

All results come from the Water Quality Portal, which publishes data from the US Geological Survey (USGS), the EPA and state agencies. We keep the raw files so every figure can be traced back.

• 1,544,453 sites recorded as wells, from 244 organizations.
• 2,985,559 test results after removing duplicates and unusable rows.
• Where a sample has calcium and magnesium but no hardness result, hardness is calculated (2.497 × calcium + 4.118 × magnesium) and marked as calculated.

The levels we compare against

EPA secondary standards, which are guidelines for taste, odor, staining and scale, not health limits: iron 0.3 mg/L, manganese 0.05 mg/L, sulfate 250 mg/L, dissolved solids 500 mg/L, pH 6.5 to 8.5. Hardness has no federal limit; we use the usual bands (soft to very hard).

Results below what the lab could measure

A "not detected" result is kept at the lab's detection limit, and typical values use half of it. If more than half of a county's wells were below detection, we say so instead of showing a number. Older records that wrote exactly 0 (55,528 results) are read as below a small nominal limit and marked.

Counties

A site's county is the one it reports, if the 2023 Census still lists it; otherwise we place it by its coordinates. In a check of 1,000 sites the two methods disagreed for 5. Connecticut figures are by its nine planning regions, which replaced its counties in 2022.

The public data doesn't say what a well is used for: a landfill monitoring well and a family's well look the same. We sort them in three steps, most reliable first.

1. Published labels. Where an agency labels each well's use (USGS's 2025 California classification), we use it: 12,070 sites left out.
2. A reviewed list of programs. 101 organizations and projects, each with a written reason: Superfund, mines, landfills, permit monitoring and city supply. 13,219 sites left out.
3. Site names. Names like MW-, PIEZ, MONITOR or LANDFILL mark monitoring wells; MINE, TAILING or SLUDGE mark mine studies. 7,676 sites left out.

Wells we can't classify stay in, because most private wells sampled by states carry no label. Of 198,252 sites with iron or manganese results, 8.5% are left out. Nothing is deleted; each county shows how many wells were left out.

Sites that clearly aren't drinking-water wells (oil-field brine, mine drainage, landfill leachate) are also kept out of every county figure: 60,298 results at 2,119 sites. The cut-offs are judgement calls.

Does keeping unlabelled wells change the answer? In California, where household wells are labelled, our county figures matched the household-only figures in 95% of cases.

Penn State's lab and the University of Wisconsin–Stevens Point test household wells and publish county results. Neither is in our source data, so they're an independent check. We compared 12 counties in each state.

• Typical levels agreed in 75% of comparisons (79% Pennsylvania, 72% Wisconsin). Our target was 80%, so we did not meet it.
• Hardness and sulfate agreed closely. Iron and manganese ran higher in our data: 17–20 points more wells over the level for iron, 13–16 for manganese.
• The gap sits in counties whose wells come mostly from older survey programs. Where a county had at least 10 wells tested since 1990, we agreed within about 5 points.
• Penn State's samples include softened water, which lowers its figures in hard-water counties. We note this rather than adjust for it.

So every county carries a label for iron and manganese

"Figures for this county come mostly from older survey wells: fewer than 10 of its wells have a sample since 2000. In Pennsylvania and Wisconsin, where we could compare, such wells showed iron and manganese over the limit more often than homeowner samples did. Test before deciding anything."

1,931 of 3,235 US counties have at least 10 kept wells tested for iron and 10 for manganese. They are home to 74.3% of the people the USGS estimates were on private wells in 2010. The bar was set before we looked at the data.

79.1% of wells tested for iron were last tested before 2000. When the same well was tested before and after 2000, the result held steady in 87% of cases for iron and manganese, 94% for hardness and 97% for sulfate. What changes between eras is mostly which wells were tested, not the water. So we keep older tests and label the counties that depend on them. Every county shows how many wells were tested since 2000 and since 2015, and the year of the last test.

• The samples aren't random. Every well was tested because a program chose it. County figures describe those wells.
• One agency dominates. In 93% of covered counties, most iron results come from USGS.
• Most tests are older than 2000.
• Unlabelled wells are kept. If one is really a monitoring or supply well, it counts as a household well.
• Some state networks test city-supply wells (Ohio, Oklahoma, Iowa, South Carolina). We can't separate them yet.
• The brine and mine cut-offs are judgement calls, and some rest on a single result.

Zip codes. We map a zip to its county with the HUD USPS ZIP Code Crosswalk (2026 second quarter, refreshed quarterly). If a zip crosses county lines, we say so. This product uses the HUD User Data API but is not endorsed or certified by HUD User.

Updates. Every quarter we re-run the same steps on newly published data. Each refresh gets a new dataset version and a note of what changed. New programs with more than 200 results are reviewed before they're published, and any county whose figures move more than twofold is checked by hand.

Who's behind this. Well Known is published by LinkABit AI Labs LLC. An independent methodology reviewer will be named here, with credentials and review date, once the review is signed. Nothing here or in the Well Known tool is for sale, and neither carries advertising or affiliate links.