How this report works
The report combines three public data systems, each cited on every page it feeds. Lake levels and water temperature come from US Geological Survey monitoring stations, read through the modernized USGS Water Data API (api.waterdata.usgs.gov). Current air temperature comes from National Weather Service stations through api.weather.gov. Daily temperature history comes from NOAA's Regional Climate Centers, and drives the freeze model below.
Two honest coverage tiers
Very few American lakes are instrumented. Wisconsin is typical: 924 USGS lake monitoring locations exist there, but only about 26 report any continuous data in a given month, and exactly one lake in the state, Lake Monona, reports continuous water temperature. So this report refuses to blur the line: a lake is either instrumented or it is modeled, and its page says which in the first line. Across the report, 4 lakes currently carry live USGS instrumentation; the other 16,002 are covered by the freeze model with their nearest quality-controlled climate station named on the page. Lake identities come from the USGS Geographic Names Information System; a lake with no usable station within range gets no page at all. Nearby river and below-dam gages appear only as clearly labeled references, never as lake readings.
The freeze-degree-day model, stated in full
Each day with a mean temperature below 32 F accumulates freeze-degree-days equal to the shortfall: a day averaging 22 F adds 10 FDD. Seasonal accumulated FDD (AFDD, from November 1) feeds the Stefan ice-growth approximation used in US Army Corps of Engineers cold regions engineering: modeled thickness in inches equals a coefficient times the square root of AFDD in degF-days. We publish the range from a coefficient of 0.4 (snow-covered, sheltered ice, which grows slowly) to 0.7 (windswept ice with little snow), and we cap the claim there rather than publishing the theoretical clear-ice maximum.
The number this produces is a growth POTENTIAL from air temperature alone. It cannot see slush layers, springs, current, pressure ridges, or a mid-season thaw's damage, which is why it renders everywhere as "modeled range," sits beside the DNR reference table, and never gets called a measurement. When the last week included thaw days, the page says so.
What this tool will never do
It will never state or imply that ice is safe, because no ice is ever one hundred percent safe. It will never show a reading older than 24 hours as current. It will never publish a page for a lake whose data cannot support one. And it will never bury its sources: every page links the exact USGS site, NWS station, and DNR guidance it is built from, with a visible retrieval time.