How every number is made
TapBuff publishes EPA’s public drinking water data, measured against EPA’s own limits and goals, and links each finding to the certification claim that addresses it. Nothing is scored or invented, and TapBuff makes no estimates of its own: where a figure is an estimate, it is an agency’s, and the page says whose. Costs are arithmetic on makers’ published figures and on inputs each page names, which you can change.
Updated
The data
EPA SDWIS, through the ECHO bulk download, edition 2026 Q3. It lists every public water system, the areas it serves, its lead and copper 90th-percentile results, and its violations. TapBuff uses active community water systems: 49,321 of them, reporting 331 million people served. A population that is plainly wrong in EPA’s file is corrected, and the utility’s report shows EPA’s figure beside the correction and why: Tri Town Water Board in Massachusetts, listed at 856,000 against 2 service connections, sells all its water to two systems that serve 85,662 between them.
EPA UCMR 5, the fifth Unregulated Contaminant Monitoring Rule, occurrence data file of Oct. 15, 2026. It holds results for 29 PFAS and lithium from nearly every utility serving more than 3,300 people, and the ZIP codes each one serves. Its additional data elements, file of Oct. 15, 2026, give the disinfectants each utility added before water entered its pipes at each sampling point and round; a report says whether that was chlorine, chloramine or both, dated to the utility’s own sampling, not what it uses today.
EPA’s compliance monitoring file, from its Six-Year Review 4: the routine results states reported for 2012 to 2019. TapBuff uses its nitrate, arsenic, fluoride, chromium, uranium, TTHM and HAA5 results, for 42,228 community systems. Georgia, Michigan, Mississippi and New Mexico sent EPA none.
States’ own results, from 2020 on, where a state publishes them in bulk: Alaska (Alaska Department of Environmental Conservation, Drinking Water Viewer, results to Oct. 14, 2026), California (California State Water Resources Control Board, Division of Drinking Water, results to July 17, 2026), Colorado (Colorado Department of Public Health and Environment, Water Quality Control Division, results to Oct. 6, 2026), Florida (Florida Department of Environmental Protection, Drinking Water Program, results to Dec. 31, 2024), Iowa (Iowa Department of Natural Resources, Water Supply Programs, results to Oct. 6, 2026), Indiana (Indiana Department of Environmental Management, Drinking Water Viewer, results to Oct. 20, 2026), Kansas (Kansas Department of Health and Environment, Public Water Supply Section, results to Sept. 16, 2026), Louisiana (Louisiana Department of Health, Drinking Water Viewer, results to Oct. 14, 2026), Massachusetts (Massachusetts Department of Environmental Protection, Drinking Water Program, results to Oct. 5, 2026), Montana (Montana Department of Environmental Quality, Public Water Supply, results to Oct. 6, 2026), New Jersey (New Jersey Department of Environmental Protection, Drinking Water Viewer, results to Oct. 13, 2026), Ohio (Ohio Environmental Protection Agency, Division of Drinking and Ground Waters, results to Dec. 31, 2025), Pennsylvania (Pennsylvania Department of Environmental Protection, Drinking Water Reporting System, results to Oct. 8, 2026), Rhode Island (Rhode Island Department of Health, Drinking Water Viewer, results to Oct. 6, 2026), Texas (Texas Commission on Environmental Quality, Texas Drinking Water Viewer, results to Oct. 6, 2026), Utah (Utah Department of Environmental Quality, Division of Drinking Water, results to Oct. 1, 2026), Virginia (Virginia Department of Health, Drinking Water Viewer, results to Oct. 14, 2026), Washington (Washington State Department of Health, Office of Drinking Water, results to Sept. 29, 2026), Wisconsin (Wisconsin Department of Natural Resources, Drinking Water Program, results to Oct. 5, 2026), West Virginia (West Virginia Department of Health, Drinking Water Viewer, results to Oct. 15, 2026), Alabama (Alabama Department of Environmental Management, Drinking Water Branch, results to Oct. 8, 2026), Delaware (Delaware Division of Public Health, Office of Drinking Water, results to Oct. 9, 2026), Illinois (Illinois Environmental Protection Agency, Bureau of Water, results to Oct. 13, 2026), Kentucky (Kentucky Division of Water, results to Sept. 23, 2026), Maryland (Maryland Department of the Environment, Water Supply Program, results to Oct. 14, 2026), Mississippi (Mississippi State Department of Health, Bureau of Public Water Supply, results to Oct. 12, 2026), Nebraska (Nebraska Department of Water, Energy, and Environment, results to Oct. 6, 2026), New Mexico (New Mexico Environment Department, Drinking Water Bureau, results to Oct. 13, 2026), Oklahoma (Oklahoma Department of Environmental Quality, Water Quality Division, results to Oct. 2, 2026), South Carolina (South Carolina Department of Environmental Services, Bureau of Water, results to Oct. 7, 2026), Vermont (Vermont Department of Environmental Conservation, Drinking Water and Groundwater Protection Division, results to Oct. 7, 2026), Wyoming (U.S. EPA Region 8, Drinking Water Program for Wyoming, results to Oct. 5, 2026), Arkansas (Arkansas Department of Health, annual drinking water quality reports, what each utility’s report lists, to 2025), Minnesota (Minnesota Department of Health, Minnesota Public Health Data Access, yearly summaries per water system to 2025), Missouri (Missouri Department of Natural Resources, annual water quality reports, what each utility’s report lists, to 2026), New York (New York State Department of Health, Environmental Public Health Tracker, yearly summaries per water system to 2024) and South Dakota (South Dakota Department of Agriculture and Natural Resources, annual drinking water reports, what each utility’s report lists, to 2025). These are the same routine results utilities report to the state, for the same seven contaminants where the state has them, read the same way as EPA’s file. Where a state has them, a report shows them, with EPA’s 2012 to 2019 file as the earlier record. A utility’s annual report lists only what was found, and for most contaminants only the highest result and its range. A running annual average, or a range that is one sample, stands as that year’s average; otherwise the report gives the highest result alone, with no average and no count of samples.
CDC’s Environmental Public Health Tracking Network, for states with no results of their own that TapBuff can read, and for what a state’s reports don’t list: Missouri (2020 to 2023, for what its reports don’t list), Arizona (2020 to 2023), Connecticut (2020 to 2023), Maine (2020 to 2023), North Carolina (2020 to 2023), New Hampshire (2020 to 2023), Oregon (2020 to 2023) and Tennessee (2020 to 2022). Each water system’s yearly averages for arsenic, nitrate, uranium and the disinfection byproducts TTHM and HAA5, which these states supply to CDC.
Utilities’ own annual water quality reports, read by hand, for large utilities in states that publish no results of their own that TapBuff can read: City of Phoenix, Arizona (2020 to 2025), Las Vegas Valley Water District, Nevada (2021 to 2025), Suffolk County Water Authority, New York (2021 to 2024), Atlanta, Georgia (2020 to 2025), Gwinnett Co. Dept. of Water Resources, Georgia (2025), Dekalb County, Georgia (2024 to 2025), City of Tucson, Arizona (2021 to 2025), Cobb County, Georgia (2020 to 2025), City of Detroit, Michigan (2020 to 2025), D.C. Water and Sewer Authority, District of Columbia (2020 to 2025), Hnl - Windward - Pearl Harbor, Hawaii (2025), City of Mesa, Arizona (2021 to 2025), Truckee Meadows Water Authority, Nevada (2020 to 2025), North Fulton County, Georgia (2020 to 2025), North Las Vegas Utilities, Nevada (2025), City of Henderson, Nevada (2023 to 2025), Grand Rapids, Michigan (2024 to 2025), Clayton County Water Authority, Georgia (2020 to 2025), Veolia Water Idaho, Idaho (2022 to 2025), Town of Gilbert, Arizona (2020 to 2025), City of Chandler, Arizona (2023 to 2025), City of Scottsdale, Arizona (2020 to 2025), City of Glendale, Arizona (2021 to 2025), Columbus, Georgia (2020 to 2025), Cherokee County, Georgia (2023 to 2025), Waipahu - Ewa - Waianae, Hawaii (2025), Augusta - Richmond Co WS, Georgia (2020 to 2025), Kalamazoo, Michigan (2020 to 2025), City of Peoria, Arizona (2020 to 2025), Henry County Water Authority, Georgia (2020 to 2025), Savannah - Main, Georgia (2020 to 2024), Lansing Board of Water & Light, Michigan (2020 to 2025), City of Tempe, Arizona (2020 to 2025), Forsyth Co. Water & Sewer, Georgia (2020 to 2025), Meridian Water Dept, Idaho (2021 to 2025), Gainesville, Georgia (2020 to 2025), Arizona Water Co - Pinal Valley, Arizona (2023 to 2025), City of Warren, Michigan (2020 to 2025), Macon Water Authority, Georgia (2021 to 2025), Epcor - Agua Fria, Arizona (2020 to 2025), Douglasville - Douglas County WSA, Georgia (2020 to 2025), City of Sterling Heights, Michigan (2020 to 2025), Athens - Clarke Co Water System, Georgia (2020 to 2025), Columbia County, Georgia (2022 to 2025), Ann Arbor, Michigan (2020 to 2025), City of Fargo, North Dakota (2020 to 2025), Paulding County Water System, Georgia (2025), Dearborn, Michigan (2020 to 2024), City of Yuma, Arizona (2020 to 2025), Albany, Georgia (2021 to 2025), Clinton Township, Michigan (2020 to 2025), City of Nampa, Idaho (2020 to 2025), Canton Township, Michigan (2025), Dalton Utilities, Georgia (2020 to 2025), Livonia, Michigan (2023 to 2025), Houston County - Feagin Mill, Georgia (2020 to 2025), Coweta Water & Sewer Auth., Georgia (2022 to 2025), ECWA Amherst, New York (2021 to 2025), Macomb Township, Michigan (2020 to 2025), South Jordan City, Utah (2021 to 2025), Rockdale Co. Water System, Georgia (2020 to 2025), Fayette County, Georgia (2023 to 2025), Troy, Michigan (2025), Westland, Michigan (2020 to 2025), Warner Robins, Georgia (2025), City of Flint, Michigan (2020 to 2025), City of Caldwell, Idaho (2020 to 2025), Wyoming, Michigan (2024 to 2025), Wailuku, Hawaii (2025), Newton Co. Water - Sewerage Auth, Georgia (2020 to 2025), Rochester Hills, Michigan (2025), Genesee County Water System, Michigan (2020 to 2025), Waterford Township, Michigan (2020 to 2025), City of Bismarck, North Dakota (2020 to 2025), Shelby Township, Michigan (2020 to 2025), Southfield, Michigan (2021 to 2025), Bartow County, Georgia (2022 to 2025), City of Idaho Falls, Idaho (2024 to 2025), Malden Water Division (MWRA), Massachusetts (2020 to 2025), Joint Base Pearl Harbor - Hickam, Hawaii (2022 to 2025), Pontiac, Michigan (2025), Taylor, Michigan (2020 to 2025), Brookline Water and Sewer Division, Massachusetts (2020 to 2025), Marietta Board of Lights and Water, Georgia (2021 to 2025), City of St. Clair Shores, Michigan (2025), Catoosa Util. Dist. Authority, Georgia (2020 to 2025), City of Casper, Wyoming (2020 to 2025), Carson City Public Works, Nevada (2025), Cumming, Georgia (2020 to 2025), City of Pocatello, Idaho (2020 to 2025), Dearborn Heights, Michigan (2025), Grand Forks Regional WTP, North Dakota (2021 to 2025), City of Royal Oak, Michigan (2020 to 2025), University of Utah, Utah (2020 to 2025), City of Farmington Hills, Michigan (2025), City of Novi, Michigan (2025), City of Twin Falls, Idaho (2025), Southgate WSD, Colorado (2020 to 2025), City of Coeur D Alene, Idaho (2023 to 2025), Ypsilanti Community Utility Authority, Michigan (2024 to 2025), Calhoun, Georgia (2021 to 2025), Carroll County, Georgia (2020 to 2023), Winder, Georgia (2021 to 2025), Holland Board of Public Works, Michigan (2020 to 2025), Redford Township, Michigan (2022 to 2025), West Bloomfield Township, Michigan (2025), Monroe, Michigan (2021 to 2025), Green Mountain WSD, Colorado (2020 to 2025), Valdosta, Georgia (2020 to 2025), City of Minot, North Dakota (2022 to 2025), Southwest Metropolitan WSD, Colorado (2020 to 2025), Newnan Utilities, Georgia (2023 to 2025), Beverly Water Dept, Massachusetts (2020 to 2025), City of Roseville, Michigan (2025), Arlington Water Dept. (MWRA), Massachusetts (2020 to 2025), Portage, Michigan (2024 to 2025), Floyd County, Georgia (2020 to 2025), Rome, Georgia (2020 to 2025), Walton Co. Water & Sewer Auth., Georgia (2020 to 2025), Brunswick, Georgia (2020 to 2024), Salem Water Department, Massachusetts (2020 to 2025), City of Saginaw, Michigan (2020 to 2025), Bancroft Clover WSD, Colorado (2020 to 2025), Battle Creek - Verona System, Michigan (2020 to 2025), City of Midland, Michigan (2020 to 2025), Plainfield Township, Michigan (2020 to 2025), Lagrange, Georgia (2025), Bloomfield Township, Michigan (2025), Chesterfield Township, Michigan (2020 to 2025), Georgetown Township, Michigan (2025), Saginaw Charter Township, Michigan (2020 to 2025), City of Rexburg, Idaho (2024 to 2025), Smyrna, Georgia (2020 to 2024), Spalding County Water System, Georgia (2020 to 2025), Hilo, Hawaii (2025), Lincoln Park, Michigan (2020 to 2025), Schofield Barracks, Hawaii (2025), Byron - Gaines Utility Authority, Michigan (2020 to 2025), Muskegon, Michigan (2020 to 2025), Statesboro, Georgia (2025), Kentwood, Michigan (2020 to 2025), Walker County Water Authority, Georgia (2020 to 2025), Perry, Georgia (2025), City of West Fargo, North Dakota (2020 to 2025), Lihue - Kapaa, Hawaii (2024 to 2025), Lexington Water Dept. (MWRA), Massachusetts (2020 to 2025), East Point, Georgia (2022 to 2025), Jackson County Water & Sewer Authority, Georgia (2024 to 2025), Brownstown Township, Michigan (2020 to 2025), Jackson, Michigan (2020 to 2025), Orion Township, Michigan (2025), Kaysville City Water System, Utah (2020 to 2025), Thomasville, Georgia (2025), Hinesville, Georgia (2020 to 2025), City of Kuna, Idaho (2020 to 2025), Butts County/Jackson/Jenkinsburg WS, Georgia (2020 to 2025), Oconee County, Georgia (2020 to 2025), Bear Creek WSD, Colorado (2020 to 2025), Melrose Water Division (MWRA), Massachusetts (2020 to 2025), North Kona, Hawaii (2025), St. Simons Island, Georgia (2020 to 2024), ECWA Lancaster, New York (2021 to 2025), Toccoa, Georgia (2025), Pooler, Georgia (2020 to 2025), Aberdeen, South Dakota (2020 to 2025), Saugus Water Dept. (MWRA), Massachusetts (2020 to 2025), Belmont Water Dept. (MWRA), Massachusetts (2020 to 2025), City of Williston, North Dakota (2020 to 2025), Moscow Water Dept, Idaho (2021 to 2023), City of Dickinson, North Dakota (2020 to 2025), Griffin, Georgia (2024), Falls Water Company Inc, Idaho (2020 to 2025), City of Mandan, North Dakota (2020 to 2025), Kalaheo - Koloa, Hawaii (2025), ECWA Hamburg (V), New York (2021 to 2025), ECWA Boston, New York (2021 to 2025) and ECWA Newstead, New York (2021 to 2025). Each figure is the report’s own, from its table of what was detected: the highest result, and a running annual average or a range that is one value as that year’s average where every year read has one; otherwise a report shows the highest result and its year. A figure the report dates to an earlier year counts in that year, and a year whose highest the report leaves unclear is left out, not guessed. Many utilities keep only their latest report or two online, so some begin as late as 2025. Where CDC also has a utility’s yearly averages, its reports take each contaminant they list, read back to 2020, and CDC’s averages fill the rest: a report’s figures usually describe its worst site or highest result, CDC’s the whole system, so the two are never joined in one series.
States’ own PFAS results, where a state publishes them under its own PFAS rule: Pennsylvania (Pennsylvania Department of Environmental Protection, Drinking Water Reporting System, samples from Jan. 4, 2023 to Sept. 30, 2026). How reports use them is under PFAS against EPA’s limits.
EPA’s community water system service area map, version 3.0, through its census block table. About 60% of its boundaries come from states and utilities; EPA modeled the rest from building footprints, population density and connection counts. For each utility it gives the people served in every 2020 census block its area covers, and it places 43,613 of them. Some areas are drawn far bigger than the utility: a wholesaler outlined over the towns it sells to, or a small system given a whole city. Where the map puts more than twice the people a utility reports serving, its people per ZIP code, town and county are scaled down to that total, so it can’t lead a place it barely serves; whether it appears there at all still goes by the map as drawn.
The Census Bureau’s 2020 relationship files: ZIP code to census block, which turns EPA’s block table into people per utility per ZIP code, and ZIP code to county, used only when nothing else ties a utility to a ZIP code.
The site is rebuilt from fresh files when EPA publishes them. Each report names the edition it was built from.
Matching your ZIP code
A ZIP code is matched first to every utility that lists it, from UCMR 5’s ZIP code file and SDWIS’s service areas: 20,752 ZIP codes. EPA’s service area map then adds the utilities whose areas reach into each ZIP, with EPA’s estimate of the people each serves there, and results are listed most people first. For 10,961 more ZIP codes, the map is the only match.
Service areas meet at their edges, so a utility can take in a few homes of a ZIP it does not really serve. TapBuff keeps a utility in a ZIP when the utility lists the ZIP, when the ZIP holds more of its people than any other, or when at least 25 of its people are there and they make up 5% of the people served in the ZIP or a quarter of the utility’s own.
Some of the ZIP codes utilities list are mailing addresses or typing slips: a Hawaii water company listing ZIP codes across California, Texas and New York, an Oklahoma rural water district listing Plano’s, or a North Dakota city listing an Iowa ZIP code one digit off its own. Where EPA’s map doesn’t place the utility in a ZIP code it lists, TapBuff leaves the listing out if the ZIP code lies in a state not bordering the utility’s own, more than 40 miles across a shared state line from where the map places it, or more than 100 miles from there in its own state and in no county the utility says it serves. In Hawaii, a listing on another island is left out. Rural water districts often reach well past their mapped areas, so nearer listings are kept.
For the last 3,821, nothing ties a utility to the ZIP, so TapBuff shows the largest utilities serving the county that holds most of its land, and says so. All told, 44,251 of the 49,321 community water systems come up for at least one ZIP code.
ZIP codes and utility service areas do not line up neatly. A ZIP can be served by several utilities, and some homes in it may be on private wells. Your water bill names your utility.
Looking up a town or county
A town’s or county’s utilities are the ones EPA’s service area map places inside it, ranked by the people the map puts in the census blocks within its limits: the Census Bureau’s 2020 Block Assignment Files say which town each block is in, and a block’s own code names its county. Wisconsin’s, which census.gov would not serve, comes from the same assignment in its 2020 redistricting file. A town’s ZIP codes, for narrowing it down, come from the Census Bureau’s 2020 relationship files, keeping a ZIP when at least a fifth of its land lies inside the place, or when the overlap is at least 5% of the place’s land. A utility that names the town or county in its EPA filings is listed even when the map does not put it there, after the mapped ones, and says which it is. A utility that does not name the place is left out when the map puts fewer than 25 of its people there, or fewer than 0.5% of the people the place’s largest utility serves there: a sliver along a border.
A town result is a short list to choose from, not a match. Your ZIP code, or the name on your water bill, is the precise answer.
Each town has a page listing those utilities, each with its results in a line: how many are above the legal limit, an action level, the health goal or EPA’s health benchmark, and the worst of them, counted exactly as its own report counts them. The short answer at the top says which utilities have a result above a legal limit or action level, or a health-based violation since 2016, then gives the results of the one serving the most people. It never adds a verdict of its own. A town is offered to search engines once it has ZIP codes of its own and either more than one utility or at least 1,000 people on EPA’s map; a smaller one still has its page, for the lookup, marked so search engines leave it out.
Which utilities get a report
Every active community system EPA sampled under UCMR 5, any other serving 3,300 people or more, and any smaller one with results from 2020 on, from its state or from CDC: 37,131 reports. The rest appear in ZIP results with their violation count, because EPA’s national file, which ends in 2019, is too thin to fill a report honestly. A wholesaler that sells water to other utilities and serves no one directly is left out of ZIP results.
Measuring a result
Every result is compared with two federal lines: the Maximum Contaminant Level Goal, the health goal below which EPA expects no risk, and the Maximum Contaminant Level, the enforceable limit. Lead and copper use the Lead and Copper Rule’s action levels instead of a limit. Where neither exists, as for lithium, a non-enforceable EPA benchmark is shown and labeled as one.
EPA’s limits for five PFAS (PFOA, PFOS, PFHxS, PFNA and HFPO-DA) are final but not enforced until April 26, 2029. Until then a result above one is labeled above EPA’s new limit, which takes effect in 2029, never above the legal limit, and it is counted apart from the legal limits everywhere results are summed.
UCMR 5 results are shown as the highest single sample, with the date it was taken. Compliance with a limit is judged on averages over time, not single samples, so a result above a limit is not by itself a violation, and every report says so. A UCMR 5 result counts as above a limit or benchmark only once EPA’s rounding puts it there, as EPA’s UCMR 5 Data Finder counts it: a limit written as 10 ppt is passed from 15 ppt, one written as 4.0 ppt from 4.05. A result between the two is called at the limit, not above it or below it.
Lead and copper are the 90th percentile of the homes a utility sampled, from EPA’s file. A lead figure marked “<” there is below the lab’s detection limit and is read as none detected. EPA’s file holds copper only when it exceeds the action level, so a copper result from before 2020 is shown as the last exceedance on file, not as a current finding.
Now and then a source carries a result typed in the wrong unit, or a lab’s test sample: one arsenic result of 11,000 ppb among a utility’s readings of 6 to 11. A result is left out as a likely entry error only when it is more than five times EPA’s limit, more than forty times the typical result where it was taken, has no other result at the utility within a tenth of it, and falls in a year EPA recorded no violation for that contaminant there. Each one left out is named on its report, with its date and the typical result. A violation level recorded in mg/L under a µg/L label, such as a nitrate violation written as 11 µg/L against a 10 mg/L limit, is read as the mg/L it was. A negative result is counted as not found.
Compliance results, from 2012 to 2019, are shown the way EPA judges them. Most of these limits are judged on running averages, with results below the detection limit counted as zero, so the figure is the highest yearly average, with the highest single sample beside it. Nitrate is judged sample by sample, with a confirmation sample, so its figure is the highest sample. Results are rounded to the limit’s precision before they are compared, as EPA does.
A compliance result is labeled above the legal limit only when EPA recorded a violation for it in those years: the limit is judged by the regulator, on running averages at each sampling point. A violation recorded at a level at or below EPA’s limit was judged against a stricter state standard, such as California’s 2 ppm for fluoride, so it stays in the report’s list of violations and does not set the label. A yearly average over the limit with no recorded violation is labeled above the limit’s level, no violation, and the report gives the figure and says EPA recorded none. A nitrate sample over the limit with no recorded violation is labeled above the health goal, which for nitrate is the same 10 ppm, and the report says what the next sample at that point showed. Wherever a result comes near a limit, the report says what EPA’s own violation record shows for 2012 to 2019.
Where a report’s results for one of these contaminants end in 2019, or it has none, and EPA recorded a violation of the limit for it since 2020 at a level above EPA’s limit, the violation leads that result: its measured level and period, labeled above the legal limit, with the 2012 to 2019 results as the earlier record.
States publish their results in their own forms, so each is read the way EPA’s file reads that state. Colorado, Florida, Massachusetts, Ohio and Washington mark raw water by sample point or sample type; Wisconsin types its raw-water samples apart from its compliance samples, so they are left out; California marks it only in its sampling point names (“raw,” “untreated,” “influent”) and intakes; in Iowa and Utah, EPA’s file sets no raw-water samples aside, and neither does TapBuff. States that run EPA’s Drinking Water Watch lookup site are read from it sample by sample, routine and confirmation samples only; there a sampling point named as raw is raw water, judged across the whole utility. Where a Florida or Massachusetts raw sample has no source of its own, the raw-water rule below is applied across the whole utility. A Colorado result at or below its detection limit is a non-detect, as EPA treated Colorado’s results; Florida and Utah write a non-detect as zero. Special samples, outside routine monitoring, are left out. Utah’s public results hold none of its disinfection byproduct samples from the distribution system, so Utah’s TTHM and HAA5 stay with EPA’s file, and Florida published no statewide file for 2023. Uranium reported in picocuries is converted at 1.49 micrograms per picocurie, the factor EPA applied to California’s results. Where a utility reports nitrate alone, nitrate-nitrite results are set aside, as in EPA’s file.
New York publishes no samples, only each water system’s yearly summary: its samples, non-detects, highest and average. Its averages count a non-detect as zero, as EPA does, for about 93% of systems in EPA’s 2019 file; the rest count about half the reporting limit, so a few New York averages run slightly high. A New York report measures each contaminant by its highest yearly average, and nitrate by its highest yearly maximum.
Minnesota publishes yearly averages only, with no sample counts or maximums, so a Minnesota report measures nitrate by its yearly average too. Until 2023 Minnesota carried a system’s last result forward into years it took no sample, so its figures are read from 2023 on.
CDC’s Tracking Network publishes yearly averages only, so a report built on them measures nitrate by its yearly average too. A system whose samples all found nothing is shown as not found. A detected average does not count non-detects as zero: checked against Colorado’s own samples for 2021, CDC’s averages ran higher wherever some samples found nothing, so at low levels a CDC figure can read high. Each such report names CDC and the state that supplied the data; the comparisons with EPA’s limits are TapBuff’s, not CDC’s or the state’s.
Raw-water samples are left out the way EPA left them out: where they are under half of a facility’s results they are stray source samples, and where they are more, they are its compliance samples.
“Not found” means below the reporting level in every sample. It does not mean zero.
Lead and copper results are the utility’s 90th percentile across the homes it sampled in a monitoring period. They describe the system, not your home.
EPA’s lead and copper file marks every result in milligrams per liter, and some were plainly sent in micrograms. Two kinds are left out, with a note on the report giving EPA’s figure: the results Ohio sent EPA in 2006 for 2003 to 2005, every one at microgram magnitudes, and any result more than 20 times the action level (lead above 300 ppb, copper above 26 ppm). Nothing is converted, since the right unit would still be a guess.
Violations are health-based violations from Jan. 1, 2016 on, as EPA records them, with their status. Monitoring and reporting violations are counted but not listed.
Limits change. Each contaminant page states its current status with dates, including proposed changes that are not yet final.
Hardness
Hardness is calcium and magnesium dissolved in water, given in milligrams per liter as calcium carbonate, with grains per gallon at 17.1 mg/L a grain. EPA sets no limit for it, so it carries no status label, only the U.S. Geological Survey’s bands: soft up to 60 mg/L, moderately hard to 120, hard to 180, very hard above.
Few states make utilities report hardness, since it is not regulated. Where a state’s own results include it (most of California, Wisconsin, Delaware, Washington, Illinois, Utah and Massachusetts, and some utilities elsewhere), a sample’s measured total hardness is used, or else its calcium and magnesium combined as Standard Methods 2340 B does (2.497 times calcium plus 4.118 times magnesium). Raw water is set aside by the same rule as for contaminants. A total or a calcium result below the reporting level is not counted as zero, since no tap water has none, and a total lower than the same sample’s calcium, plainly sent in another unit, is left out. Where a state’s own samples show its calcium given as calcium carbonate rather than calcium (Florida’s and Delaware’s do), it is not combined. Each sampling point is taken at its median, from 2015 on where a state’s files reach that far.
A utility’s sampling points are not weighted by how much water each supplies, because no source says. San Francisco’s soft reservoir water and its harder groundwater wells are one point against nine, so a middle figure would describe nobody’s tap. A page gives a typical figure only where every point falls in one band, and otherwise the range from the softest point to the hardest, depending on the source. Nothing is estimated where nothing was measured.
California, Utah and Washington take their results at each source rather than where water enters the pipes, so a well that feeds a desalter or a softening plant is sampled before the plant. Oceanside’s wells read 740 to 1,130 mg/L; the water its desalter sends out, about 235. Where EPA’s facility list names such a plant (a desalter, reverse osmosis, nanofiltration or softening) for a utility in those states, its state results are left out, and its own report is used where it has been read.
The hardness pages group each state’s utilities by band. A state has a page once 10 of its utilities have hardness, or 3 that serve a quarter of the people its utilities serve. The hardest and softest lists rank utilities serving 100,000 people or more whose water is in one band, by their typical figure, and only where that figure is the utility’s own or a state’s figure was checked against the utility’s report and agreed. A list puts a figure where it will be quoted, and a state’s results can miss what reaches the tap: West Jordan’s were wells that supply 7% of its water, and Dayton’s were wells before its softening plants. Where a check finds that, the utility’s report takes the state’s place. A utility in more than one band has no one figure to rank by and is left off.
For large utilities in states that publish no hardness, the figure comes from the utility’s own annual water quality report, read by hand, or where that prints none, another water quality document the utility or its supplier publishes, such as a tap water analysis or monthly reports on water leaving the plant, which the page names. It is total hardness of treated water as printed, never calcium alone or untreated source water, with the document’s title, year and page. Where it gives an average it stands, with its range after it; where it gives a figure for each plant or source, the page gives the range between them. A state’s results always take the place of a report’s.
PFAS against EPA’s limits
A report shows each PFAS’s highest single UCMR 5 sample, labeled against EPA’s 2024 limits, because that is the measurement on file. One sample is not how the limits are judged: from April 26, 2029, compliance is a running annual average of each sample point’s quarterly results, from the utility’s own monitoring. UCMR 5 results are EPA’s survey, and EPA says they don’t show compliance or noncompliance.
The PFAS list compares the way EPA’s final UCMR 5 data summary (July 2026, Table 4) does. Each sampling location is averaged over its full set of samples, two for a location drawing ground water and four for surface water, with a result below the reporting level counted as zero; a location missing a sample has no average. A utility is above a limit when any one location’s average is, compared at the limit’s significant digits as EPA compares them: 4.05 ppt or more for PFOA and PFOS, 15 ppt or more for PFHxS, PFNA and HFPO-DA. The Hazard Index for PFHxS, PFNA, HFPO-DA and PFBS sums each average over EPA’s health-based water concentration (10, 10, 10 and 2,000 ppt) and counts at 1.5 or more, and only where two of the four were found at that location.
Done this way, the counts are EPA’s own: 662 of 4,277 large systems and 474 of 5,852 smaller ones above a limit, as in Table 4. A report whose single sample is above a limit says what its location average was. Utilities with a single sample above but no location average above are counted on the list’s pages, not listed. The whole table is free to download as a CSV.
Where a state publishes PFAS results under its own rule (Pennsylvania), a report shows them in place of UCMR 5’s for each compound the state tested, with UCMR 5’s as the earlier record; the PFAS list keeps to UCMR 5, so states compare alike. Pennsylvania’s rule, published January 14, 2023, sets limits of 14 ppt for PFOA and 18 ppt for PFOS, in force now and judged on a running annual average at each entry point, and has every public water system test quarterly at each entry point from 2024, or 2025 for those serving 350 people or fewer. A report gives the highest single sample, against EPA’s 2024 limits as for UCMR 5, and the highest four-quarter average at any one entry point: each quarter’s samples averaged, a result below the reporting level counted as zero, then four quarters in a row. That average is set against the state’s limit and EPA’s. A result is labeled above the state’s limit only where the state recorded a violation of it, from DEP’s violation history for each utility with a sample above the limit; where the average here is above the limit but the state recorded none, it is labeled above the limit’s level, with no violation, as for EPA’s results. The state’s own record decides compliance, and the report gives its dates and running averages. Only entry point and check samples count: raw water, in-plant and special samples are left out.
Testing labs
Every state, or EPA for the few without a program of their own, certifies the labs allowed to test drinking water, contaminant by contaminant, and publishes the list. The testing pages are those lists as the states give them: each lab’s name, city, phone number, website where the list gives one, and what it is certified to test for, as far as the list says. Contact people and email addresses are left out. A state’s page names its list, the list’s own date and the date TapBuff read it. Certifications change between lists, so a lab’s own word is the latest.
Some lists say whether a lab takes samples from homeowners; where one does, the page says so, and otherwise it says nothing. A lab the list itself marks as a utility’s in-house or non-commercial lab is left out, since it tests only its own system’s water. A lab named for a water utility or its plant is listed apart, with a note to call first, and is never offered as a nearby lab.
Town pages and reports name the three labs the state certifies that are nearest, within 75 miles. A lab sits at the coordinates its list gives, or else at the center of its ZIP code; a town sits at the average of its ZIP codes’ centers, and a utility at the average of the ZIP codes it serves, from the Census 2020 gazetteer. Distances are straight lines between those points, so they are approximate, and a lab across a state line counts when that state certifies it.
States whose list can’t be read by a script, because the state asks automated readers not to or offers only a search form, are named on the national page with EPA’s table of every state’s program, which links each list. They join as their lists are read.
Standards and certifications
What each NSF/ANSI standard covers comes from NSF’s own published descriptions. A filter is listed for a claim only when a certifier’s own listing (NSF, WQA, IAPMO, CSA or UL) shows that standard and that claim. Makers’ advertised claims that could not be found on a listing never count as certification.
Whole-house systems are often sold with no certification at all. They are still listed, because readers shop for them, but each is labeled: a product page puts the maker’s own claims in a separate column, and a system with no listing says so, with the date TapBuff searched NSF, WQA and IAPMO. An uncertified system never appears as a filter for a contaminant.
Each certification on the site links to the listing it came from and carries the date it was checked.
Refrigerator filters follow the same rule. Which refrigerators a filter fits comes from the fridge maker’s own parts pages, with the date each list was read; models are listed as the maker lists them. A third-party copy is included only when a certifier lists it, and it is said to fit what the originals it replaces fit only on its own maker’s word, which the page says. Refrigerator filters are kept out of the filter lists for a contaminant, because one helps only if it fits the reader’s fridge.
Five-year cost
System and cartridge prices are each maker’s own list price, with the date checked. Where a maker publishes no price, the system is priced at the regular price of the store the maker sends buyers to, and the page names the store and the date; never at an Amazon price. Cartridges are replaced at the maker’s gallon or month rating, whichever comes first for the household entered. The set a system comes with is not counted as a purchase; where a maker sells replacements only in packs, the system is counted as coming with one filter, not a pack, and a change that takes two of a cartridge is priced at two. Assumptions are stated on the calculator and can be changed.
Whole-house systems are costed at 80 gallons per person per day, the low end of the U.S. Geological Survey’s estimate of indoor home use. TapBuff does not estimate installation, electricity or the water a backwashing filter uses; the calculator takes your plumber’s quote.
A water softener’s cost includes its salt. A softener can be set to use more or less salt per regeneration, and NSF/ANSI 44 requires every capacity a maker publishes to name its setting. TapBuff uses the maker’s own capacity at the setting its efficiency rating is stated at, or, where no listing shows a rating, the maker’s default setting; never a setting picked because it is cheapest. A household’s water brings in its hardness times the gallons it uses each day; each regeneration restores the setting’s grains with the setting’s salt. No reserve is counted, so a softener that regenerates early uses more. A softener not shown to regenerate on demand gets no salt figure, since its salt follows a timer rather than the water used. Hardness is the reader’s own, or an example 10 grains per gallon, said wherever it is used; salt is a 40-lb bag at a named store’s price on a stated date. Regeneration water is given in gallons, not priced.
A shower filter is costed at 40 gallons of showers a day, EPA WaterSense’s “nearly 40 gallons per day” for the average family, all through the one shower it is on. Where a maker states no gallon rating, the capacity on the filter’s NSF/ANSI 177 listing is used, and where a maker gives a range of months, the shorter.
A refrigerator filter is costed as every filter bought in five years, with nothing for the fridge: the filter a fridge came with is long gone by the time anyone looks one up. It is replaced on its maker’s month interval, the schedule the fridge’s own filter light keeps, because how much of a household’s water runs through the fridge varies too much to estimate. Each filter’s page gives the daily use at which its gallon rating runs out first. A filter whose maker publishes no interval gets no total.
What this cannot do
EPA’s national file of routine results ends in 2019, and Georgia, Michigan, Mississippi, New Mexico, Puerto Rico, Guam and the U.S. Virgin Islands are not in it. Mississippi’s and New Mexico’s reports draw on the states’ own records instead; the others carry no routine results. Newer results, and contaminants beyond the seven TapBuff shows from it, are in your utility’s own annual Consumer Confidence Report.
No public data describes your own plumbing. A certified laboratory test of your tap is the only way to know your water.
Corrections
If a report is wrong because EPA’s data is wrong, the fix belongs with the utility or the state, which submit it; TapBuff will show the correction once EPA publishes it. If TapBuff got something wrong, it is corrected and the page’s update date changes.
To report an error, write to info@tapbuff.com with the page’s address and what looks wrong.