How this site works
Methodology
Every local page starts with a public EPA record. When the link between a ZIP code and a utility is uncertain, or the utility's record is too thin, the site says so or shows the county overview instead.
How a ZIP code is linked to a utility
ZIP codes are matched to active community water systems using three evidence types: ZIPs declared by the utility in SDWIS or UCMR 5; overlap with EPA's mapped service area; and, only as a fallback, matching reported city or county names. Each edge receives a confidence score. A primary system must reach 0.30 confidence.
A ZIP gets its own data page when its primary system has at least three UCMR 5 results, a health-based violation in the pinned 10-year window, or any lead or copper action-level exceedance on record. Other ZIP routes permanently redirect to their Census county overview. This avoids manufacturing a detailed-looking page from sparse records.
How systems are matched to states and counties
A utility's administrative mailing address can be in a different state from the people it serves. We use the state encoded in a state-issued public-water-system ID for its primary jurisdiction. For regional and tribal IDs, we use reported served states where available, with documented jurisdiction or administrative fallbacks. County links are checked against Census state and county codes and, when supplied, the county name; explicit cross-state service records are retained. An administrative fallback also requires a matching county name. Ambiguous records are excluded from county totals instead of assigned by a matching county number alone.
We corrected these geographic joins on September 15, 2026. This changed some county and state totals without adding or changing the underlying measurements. The EPA source and sampling dates remain unchanged. The distinction between administrative location and service area is documented in EPA's drinking-water data documentation.
How PFAS results are compared
UCMR 5 stores individual results in µg/L; this site displays PFAS in ppt (parts per trillion, equivalent to ng/L) and lithium, lead and copper in ppb (parts per billion, equivalent to µg/L). For PFOA, PFOS, and other PFAS with an MCL, the comparison uses the highest average from a complete sampling location: two semi-annual results for groundwater or four quarterly results for surface water, with non-detects counted as zero. A highest single sample is shown but is not called a violation.
How lead and copper are compared
Lead and copper values are the 90th percentile of selected customer-tap samples for a monitoring period. Results reported only as below a laboratory limit, negative values, and values whose magnitude conflicts with the recorded unit are excluded from action-level comparisons and identified as unusable or censored.
How violations are counted
Violation counts use one pinned 10-year window across the database, prose, and tables. Health-based violations, monitoring/reporting violations, and other public-notification or report violations are counted separately and reconciled against the total during every build.
How the site checks itself
The build independently recomputes more than 1.5 million UCMR 5 and lead/copper comparisons from the raw normalized tables, then checks the comparison attributes and visible labels in rendered HTML. A mismatch stops publication.
How local testing resources are reviewed
Local resource descriptions are compiled from the linked county, state, federal or utility source. We check which area the service covers, who provides it and what the source actually offers. A private-well program, a certified laboratory directory and a public utility's annual report serve different needs; one is not substituted for another.
The source-review date records when a link and its content were checked. It is not the date of a new test, a guarantee that a service remains available or an indication that a notice is still active. When a source explicitly provides a revision date, that is shown separately. Funding, fees, eligible properties, laboratory tests and appointment availability should be confirmed directly with the named provider.
Counties with a researched local list are distinguished from those with national directories only. We do not infer a home's supplier from its county name. Historical cleanup or court records are labeled as historical context, and are not presented as current advisories or findings about a household's water.
Martin Cabrera and Ima run Tap Water Lookup. Send corrections with the page URL and supporting source to martincabrera@neuraorkesta.com.
Data vintage
Recent station observations
County pages request the latest available observations from the USGS Water Data API. Monitoring locations are matched by their reported state and county codes. Supported measurements are shown only when observed within the previous 48 hours. We mark observations older than 6 hours as older; these display thresholds are our choice, not a USGS quality classification.
The observation time comes from the measurement record. The checked time describes retrieval by our service and does not make an old measurement new. While the section is visible, the browser checks every 10 minutes. Responses are cached briefly to limit requests to USGS. Provisional readings may be revised, and gaps, missing values and failed requests are not interpreted as zero or clean water. Station measurements do not establish drinking-water compliance or swimming safety.
EPA assessments and drinking-water records
The atlas requests published EPA ATTAINS assessments for the displayed area. Assessment cycles describe reported historical conditions, even when fetched today. EPA UCMR 5 and SDWIS records below come from dated source snapshots; they do not update at sensor frequency. Local testing and notice links are checked editorially, and the date shown beside them is a link-review date.
EPA UCMR 5 occurrence data, samples collected 2023-01-03 – 2026-06-17, final release, August 2026 (file dated Aug 28, 2026); EPA SDWIS via ECHO, 2026Q2 extract (file dated Jul 9, 2026); EPA service-area boundaries v3.0; Census 2020 ZCTAs. Database prepared 2026-09-03.