The audit trail: where the records come from, how they are made comparable, where the picture is incomplete, what has been corrected, and what is kept private. Every displayed result should be something you can interpret and challenge from this page.
A note on authorship: the prose on these pages is drafted with the help of a large language model — the figures are not. Every number is computed by a deterministic pipeline from the utilities’ own compliance reporting, and the model never touches the data. Where the two meet, the pipeline wins.
On this page
NitrateSmart is a standing interpretation of public compliance data — the nitrate testing that community water systems already report to state and federal drinking-water programs, normalized to one unit, classified on a five-band scale, and sourced. It is not a laboratory, not a regulator, and not a substitute for testing your own tap. Private wells are outside its scope: they are federally unregulated, not routinely tested, and not reported in a national source.
Four source families, each row carrying its system ID, source location, and retrieval date: EPA system records (SDWIS and UCMR 5); per-system sample results from state drinking-water portals; results transcribed from utility Consumer Confidence Reports where no machine-readable state data exist; and annual averages for the one state (Minnesota) that publishes only aggregated results. Original source files are preserved and never edited.
The active backbone is 47,581 community water systems in the lower 48. 38,762 have a nitrate result inside the current three-year window and are classified. 65 more, whose latest result predates the window, are shown with that older result and its date — for reference, counted in no statistic. The remaining 8,754 have no usable in-window result and are shown as no current data — never estimated, filled, or counted as zero. So the classified figures describe the systems that reported, not the whole country.
Each system's band is set by its most recent qualifying result within the rolling three-year window — the current reading, not the highest it has ever hit — taking the high end where that result spans a range. The full three-year range is shown alongside, so a past spike or a low season is never hidden. The aim is to be current and comprehensive, not to pick a worst case.
Four corrections make systems comparable, in a fixed order — the categories and the reasons are public; the implementation that applies them at national scale is not.
Where a system buys its water and reports no sampling of its own, it inherits its supplier's result and is labelled inherited (5,281 systems, 13.6%); the other 33,481 (86.4%) carry direct compliance samples.
Usually the units (nitrate ion vs. nitrate as nitrogen, about 4.4× apart), the analyte (nitrate vs. nitrate-plus-nitrite), the window (most recent result plus three-year range vs. an annual average), or how a non-detect is represented. If a number looks off against a Consumer Confidence Report, the units and the analyte are the first things to check.
Every classification travels with two qualifiers. Confidence describes how much testing stands behind it — how many results and how direct — not how good or bad the water is. A range profile says whether the most recent reading sits entirely within its band (uniform), dips below the band floor (variable, often a blending signal), or has no reported low end (unknown). More than a third of systems carry a low-confidence record, and what each result rests on is shown alongside it rather than hidden.
Checking this count against SDWIS will produce a larger number. The difference is 142 systems, removed for four reasons:
The figure is counted rather than estimated, and each refresh records it in the change log below.
Coverage is uneven, and the uneven parts are named rather than coloured in. Arkansas has no nitrate data in the inventory; New Hampshire and Rhode Island have too few reporting systems to characterise; Michigan is a known collection gap (its public layer does not publish nitrate in a joinable form); private wells are out of scope everywhere. Coverage is expected to grow as more state and utility sources are gathered.
Corrections are not hidden. NitrateSmart is maintained and falsifiable, and the record of what changes after launch — to a record or to a method — is part of the audit trail.
Nitrate in community water systems, from the utilities’ own compliance reporting.
Results reaching back to July 2023 · current through August 9, 2026.
Educational information, not medical or personal safety advice.