Method, sources and limits
Every county figure on this site is a median of public water quality samples. This page says exactly which samples, how they were processed, and what the result cannot tell you.
The source
Water Quality Portal (USGS, EPA and state agencies), at https://www.waterqualitydata.us/, retrieved 2026-09-19. The Water Quality Portal is a joint service of the USGS, the EPA and the National Water Quality Monitoring Council, and it aggregates sampling by federal, state, tribal and local agencies.
The characteristic queried is Hardness, Ca, Mg, which is total hardness expressed as calcium carbonate, reported in mg/L as CaCO3.
How a county figure is produced
- Every monitoring station in the state is retrieved, with its county FIPS code and coordinates.
- Every hardness result is retrieved and joined to its station, so each measurement lands in a county.
- Results are converted to mg/L as CaCO3. Reported units vary by agency and include forms such as
mg/l CaCO3*^; the qualifier characters are stripped before matching. Anything in a unit that cannot be interpreted with certainty is dropped rather than guessed at. - Sample fractions other than total and dissolved are dropped, because they measure something else.
- A county with fewer than three usable samples is omitted entirely rather than published thinly.
- The county figure is the median, with the 25th and 75th percentiles reported beside it. The median is used rather than the mean because hardness distributions are skewed and a handful of brine samples would drag an average badly.
Totals: 2,904,346 usable samples, 3,110 counties, 50 states with enough data for a page.
What this is not
It is not your tap water. This is the single most important limitation and it is repeated on every page. The Water Quality Portal records ambient water: wells, streams, springs and lakes. A municipal utility abstracts that water, treats it, often blends several sources and sometimes softens it centrally. A city supply can be much softer than the aquifer beneath it.
It is not evenly sampled. Monitoring effort follows research and regulatory interest, not population. A county with 2,000 samples and one with 4 both produce a median, and the sample count is published beside every figure so you can see which is which.
It is not current. Samples span many years. Hardness changes slowly, being a property of geology, so this matters less than it would for a contaminant, but a figure here is a long-run characterisation rather than a reading from this month.
Counties are not uniform. Where a county spans two geologies the median sits between them and describes neither. The middle 50% column on each state page is there to show that.
What to use instead, for your own water
- A test strip. Cheap, immediate, and measures the actual water at your tap.
- Your utility's annual water quality report, often called a Consumer Confidence Report, which every US community water system must publish.
- A laboratory test for a private well, which is worth doing for bacteria and nitrate regardless of what you conclude about hardness.
The arithmetic
Unit conversions on this site are exact ratios, not measurements: one grain per gallon is 17.118 mg/L by definition. Softener sizing uses the standard industry method of daily grain load times days between regenerations, with iron compensated at 5 grains per gallon per ppm and manganese at 4. All of it is covered by a test suite that asserts known answers, round trips every unit pair, and checks the committed aggregate for structural soundness.