How Lake Grades Are Calculated

LakeQuality grades a lake A through F on nutrients and algae, using the condition classes EPA publishes in the National Lakes Assessment 2022 Technical Support Document (EPA 841-R-24-006, Table 6-3). The thresholds are not national averages and not our own: they are the nutrient levels EPA measured in least-disturbed reference lakes within each of nine ecoregions, so a lake is judged against the lakes it would naturally resemble rather than against a single continental yardstick. Every measurement comes from the public EPA Water Quality Portal.

What the Grade Covers, and What It Does Not

The grade is built from four measured quantities: total phosphorus, total nitrogen, chlorophyll-a and turbidity. Those describe how much nutrient a lake carries and how much algae it grows. That is the entire scope.

The grade does not measure mercury, PFAS or bacteria. It is not a swimming clearance and it is not a fish-consumption clearance. scripts/validate-grades.js puts a number on how much this matters: cross-checked against EPA ATTAINS, 56.2% of the lakes this system grades A are on a state's Clean Water Act 303(d) impaired list. There is no contradiction in that: a lake can have textbook nutrient chemistry and still be listed for mercury in fish tissue, for PFAS, or for E. coli. The grade is blind to toxics by construction, and an A is silent about them rather than reassuring.

Every lake page shows its impairment listing where one exists, alongside the grade, precisely so the two can be read together. Before swimming or keeping fish, check your state health department's current advisories — linked from each lake and state page.

Reference Condition, by Ecoregion

A phosphorus concentration that would be alarming in a Rocky Mountain lake is unremarkable in a Nebraska Sand Hills lake, which was hypereutrophic before anyone farmed near it. Grading both against one number fails the second lake for being what it has always been. EPA solves this by setting condition classes separately in each ecoregion, from the distribution of its own least-disturbed reference lakes. Quoting the Technical Support Document directly:

“The cutoff between good and fair condition class was set at the 75th percentile (Q3) of reference lakes, and the cutoff between fair and poor condition class was set at the 95th percentile (P95) of reference lakes.”

So a value at or below the Q3 cutpoint is Good, above it and at or below P95 is Fair, and anything higher is Poor. These are percentiles of reference lakes, not percentiles of our dataset — the grade is criterion-referenced and is never graded on a curve. A year in which every lake improves is a year in which every letter can improve.

The spread across ecoregions is the reason this matters. The Good/Fair cutpoint for total phosphorus runs from 16.0 µg/L in the Northern Appalachians to 486 µg/L in naturally hypereutrophic Southern Plains lakes — a thirty-fold range. The Western Mountains sit at 23.4 µg/L.

The cutpoint table

Each cell reads Good/Fair cutpoint / Fair/Poor cutpoint. Phosphorus, nitrogen and chlorophyll-a are in µg/L; turbidity is in NTU.

Source: U.S. EPA (2024), National Lakes Assessment 2022 Technical Support Document, EPA 841-R-24-006, Table 6-3, p. 88. The identical values appear as Table 6-3, p. 83 of the NLA 2017 TSD.

There are ten rows for nine ecoregions because EPA splits the Southern Plains by lake origin. The Nebraska Sand Hills natural lakes carry a reference phosphorus roughly sixteen times the man-made Southern Plains value. That split is EPA's own published evidence that naturally eutrophic lakes exist and must not be judged against a man-made scale.

How a lake gets its ecoregion

Each lake's coordinate is run point-in-polygon against EPA's Omernik Level III ecoregion boundaries, then crosswalked to the nine NLA aggregated ecoregions using EPA's own NLA 2017 site information file, which publishes both codes for every sampled lake. Across 5,721 NLA site records, 84 Level III codes map to the 9 aggregated regions with zero ambiguity.

We check the assignment against EPA's published answer rather than assuming it: running EPA's own NLA coordinates back through our pipeline, 5,600 of 5,600 points agree (100.00%), with 24 points (0.43%) falling outside every polygon — the Great Lakes and coastal-bay pattern.

One entry is inferred rather than EPA-published, and is labelled as such in the crosswalk: Level III code 49, the Northern Minnesota Wetlands, was never sampled by any NLA cycle. It is assigned to Upper Midwest on two grounds — it shares 470.5 km of border with Northern Lakes and Forests (Upper Midwest) against 329.1 km with the Lake Agassiz Plain (Temperate Plains), and its peatland lakes have Upper Midwest analogues while agricultural Temperate Plains reference conditions would be badly wrong for them.

Why Water Clarity Is Not Graded

This is the largest change from the old methodology, and it is deliberate. EPA's National Lakes Assessment assigns condition classes to phosphorus, nitrogen, chlorophyll-a and turbidity, and none to Secchi depth. We follow that, because a clarity reading on its own does not tell you whether a lake is polluted.

Dark water has three quite different causes, and only one of them is a water-quality problem:

• Algae. The water is green and opaque because it is full of phytoplankton. This is a real problem — and it is already captured by chlorophyll-a, which measures the algae directly. Grading clarity too would count the same fact twice.
• Natural dissolved colour. Tannins leached from peat and conifer forest stain bog lakes brown. A Boundary Waters lake can be tea-coloured, nutrient-poor and pristine.
• Suspended sediment. Wind-resuspended clay or a silty inflow scatters light. A reservoir on clay soils can be permanently turbid with almost no algae in it.

Under the previous system these were indistinguishable, and 873 lakes were graded D or F on a clarity reading alone — bog lakes and clay reservoirs failed for their geology. That is the defect this rebuild exists to fix.

What we show instead: the clarity driver

Clarity is still measured, still displayed in feet on every lake page, and still explained — it is simply explained rather than scored. Each lake is classified by what is limiting its clarity:

The 10-unit test comes from EPA's Nutrient Criteria Technical Guidance Manual, p. 3-9: “In turbid lakes, it is common to see a close relationship between the total phosphorus TSI and the Secchi depth TSI, while the chlorophyll index falls 10 or 20 units below the others.” See the Secchi depth guide for how the measurement itself works.

Why Phosphorus Is Set Aside in Sediment-Driven Lakes

Removing clarity from the grade is not sufficient on its own. A clay-turbid reservoir would still fail — just through phosphorus instead of Secchi — because phosphorus binds to clay. EPA's Nutrient Criteria Technical Guidance Manual (Table 3.2, p. 3-10, after Carlson 1983/1992) states it plainly:

“Clay particles contain phosphorus, and therefore, lakes with heavy clay turbidity will have the phosphorus correlated with the clay turbidity while the algae may neither utilize all the phosphorus nor contribute significantly to the light attenuation.”

So when a lake is classified sediment-driven, its phosphorus is reported but not graded, and chlorophyll-a becomes the sole graded indicator — the algae the lake actually grows, rather than the phosphorus riding on its clay. If such a lake has no usable chlorophyll-a measurement, it is published as Not Rated rather than failed on a number we have just said is untrustworthy.

How the Letter Is Assembled

The grade is not an average of parameter grades. Averaging ordinal letters is not a valid operation, and the old implementation of it let one Secchi reading publish a letter. Instead the engine uses the cause-and-response logic state agencies use for eutrophication listing:

• Response — chlorophyll-a, the algae the lake is actually growing.
• Cause — phosphorus and nitrogen, the nutrients driving it. When both are available, the worse of the two condition classes is used.

Cause and response must agree before a lake is called degraded:

Two details are worth naming. First, the ends of the scale are pinned to EPA's national trophic bands on chlorophyll-a: an A requires the lake to be at or below the 7 µg/L mesotrophic ceiling, and an F requires it to be above the 30 µg/L hypereutrophic line. Second, a lake with loaded nutrients that is demonstrably not growing algae is downgraded one step from Good, not failed. That case is common rather than exotic — NLA-derived analysis finds only about 17% of US lake samples confidently phosphorus-limited, 14% nitrogen-limited and 69% indeterminate — and failing a lake on a nutrient reading its own algae contradict would invert the whole point of cause-and-response logic.

The two one-sided cases are not symmetric, because the ends of the scale are pinned to chlorophyll-a. A lake with chlorophyll-a but no usable nutrient measurement is graded exactly as though its nutrients were Good, so it can still reach A or F. A lake with nutrients but no chlorophyll-a cannot: the best it can do is B and the worst is D, because without the algae measurement there is nothing to place it at either end.

The Evidence Bar

A measurement has to clear three tests before it can affect a grade.

• Index period. Only June through September samples are used, matching the National Lakes Assessment's operative window; EPA's Consolidated Assessment and Listing Methodology is explicit that “only results from similar index periods [should] be compared”. There is no all-season fallback. The previous code silently reverted to whole-year data — including winter and under-ice readings — whenever a lake had no summer samples, and graded 1,071 lakes that way with no flag on the record.
• Minimum samples. A parameter needs at least 3 summer samples to enter the grade at all. Every state assessment methodology we reviewed sets a floor of 4–10 samples over 2–3 years, and EPA CALM notes that good statistical power needs roughly 30. Three sits below the lowest adopted state floor, which is why it produces a grade but never a confident one.
• Recency. The most recent sample must fall inside a 10-year window. EPA CALM Table 4-2 places data more than 5 years old — “not reflective of current conditions” — in its lowest rigour tier.

Confidence

Every rated lake carries a confidence of high or moderate. High requires at least one graded parameter with 8 or more samples across 2 or more distinct years, most recent within 5 years — at or above the state assessment floors. Everything else is moderate.

Only high-confidence grades feed rankings, “cleanest lake” lists and other comparative surfaces. A moderate grade is shown on the lake's own page, where it sits next to its sample counts and can be read for what it is, but it is not allowed to carry a comparative claim against another lake.

Not Rated Is Not an F

When the evidence does not support a letter, the lake gets no letter — never an F. This is EPA Integrated Reporting Category 3: “insufficient available data and/or information to make a use support determination”. An F is a finding about a lake. Not Rated is a statement about our evidence, and conflating the two libels lakes for being under-monitored.

Each unrated lake states its own reason in plain language. The reasons are:

• No nutrient or algae measurement meets the evidence bar — fewer than 3 summer samples, or nothing inside the 10-year window.
• The only measurement available is water clarity, which is not a water-quality indicator on its own.
• Clarity and phosphorus are both driven by suspended sediment, and there is no chlorophyll-a to grade on instead.
• The lake falls outside every EPA Level III ecoregion polygon, so the NLA condition thresholds cannot be applied. These are chiefly Great Lakes and coastal-bay coordinates, and waters outside the coterminous United States.

Unrated lakes keep their pages. Name, location, depth, fish, ice-out, impairment status, boat access and neighbouring lakes are all still there; only the letter is withheld.

Trophic State

Alongside the letter, each lake carries a trophic-state classification. It now comes from chlorophyll-a alone, using EPA's national bands, rather than from a Carlson index averaged across Secchi:

The classic Carlson Trophic State Index is still computed and displayed for continuity and for the clarity-driver test, but it no longer feeds any grade. See the trophic state guide for the formulas.

Provenance: Who Measured What

Every lake page names the organizations that actually submitted its samples, with their station identifiers, analytical methods, sample counts and date range, taken from the submitting-organization field the Water Quality Portal records on every row.

When provenance is missing, the page says so. It does not fall back to a plausible-sounding agency name. An earlier version filled the gap with a template that appended “DNR” to the state name, which cited agencies that do not exist: Texas's water agency is the Texas Commission on Environmental Quality, and Oklahoma's is the Department of Environmental Quality. “Source not recorded” is a true statement; a confident wrong attribution is not.

Data Sources

Licence, publisher, and attribution for every dataset are on the data sources page. US in-situ measurements come from the EPA Water Quality Portal, which aggregates monitoring data from:

• State environmental and natural-resource agencies across 36 US states and 2 Canadian provinces, including the Minnesota Pollution Control Agency and the Wisconsin Department of Natural Resources
• U.S. Geological Survey
• EPA STORET legacy records
• County, tribal and volunteer monitoring programs

Grading thresholds for US lakes come from the EPA National Lakes Assessment; ecoregion boundaries from EPA Omernik Level III; impairment listings from EPA ATTAINS; lake morphometry from HydroLAKES and state DNR sources; invasive-species listings from the USGS Nonindigenous Aquatic Species database. Ontario and Québec keep their own authority classes in official_ratings; those never blend into A–F. Satellite-product caveats are on the data sources page.

Other Data on a Lake Page

Fisheries context comes from state survey archives — 5,243 DNR fisheries reports and resources indexed across Minnesota and Wisconsin, drawn from the Wisconsin DNR fisheries portal and its Minnesota counterpart. These add fish-community composition, stocking history and habitat notes. They are context for the grade, not inputs to it: no fisheries record moves a letter.

Limitations

• Ecoregion boundaries are discontinuities. Because thresholds change at an ecoregion line, two lakes with identical chemistry on opposite sides of one can receive different letters. We measured this rather than assuming it away: of 313 neighbouring lake pairs within 15 km of each other that straddle an ecoregion boundary, 121 (39%) would receive a different letter. Unnamed Lake in MN sits 4.7 km from a lake across an ecoregion boundary; its chlorophyll-a of 12.6 µg/L classifies Good against TPL reference condition and Fair against UMW. Smoothing this away would mean inventing reference conditions EPA never published, so we disclose it instead.
• The grade is blind to toxics and pathogens. 56.2% of A-graded lakes carry an EPA 303(d) impairment listing. Read the grade together with the impairment panel, never instead of it.
• Depth is not adjusted for. Shallow lakes have less volume to dilute incoming phosphorus, more wind-driven sediment resuspension and warmer summer water. Reference conditions vary by ecoregion but not by lake depth, so a shallow lake at natural levels for its type can still grade below a deep lake in the same region.
• Monitoring effort is uneven. Some lakes have decades of records; many have a handful of samples. Confidence carries that difference, and moderate-confidence grades are kept out of rankings — but a well-monitored lake and a barely-monitored one are not equally well understood, whatever their letters.
• The Portal aggregates many protocols. Different agencies use different laboratory methods and detection limits. We report the methods per lake and use medians rather than means to limit the influence of outliers, but cross-agency variability is real.
• The condition classes are coterminous-US only. The NLA reference framework does not extend beyond the Level III ecoregion coverage, so lakes outside it cannot be rated under this system.
• A grade is a multi-year summer summary, not today's water. It cannot tell you whether there is a bloom this week. Always check current advisories from your state health department or natural-resources agency before swimming, especially with young children, pregnant family members or pets.

Data Refresh

Water Quality Portal records refresh as state agencies upload new monitoring data, most commonly on a 6–12 month cycle. Every grade, confidence value and ranking is recomputed from scratch against the new file when each refresh ships. The current dataset was last refreshed August 2026.