Methodology & sources
Every reading on Lake Level Tracker comes from USGS Water Services (current value + a 90-day trend) or USACE CWMS Data (storage, a genuine %-of-conservation-pool-full figure where the district publishes one, and standalone lakes with no USGS gauge), retrieved 2026-07-14. This page documents the exact sources, how the trend sparkline and 30-day change are computed (including a documented data-quality pass that removes a small number of contaminated sensor readings before display), why a missing datum/%-full/reference is always shown as "not reported" rather than invented, and why the site carries no basemap in this version.
Data sources
| Source | Publisher | License | Retrieved | Notes |
|---|---|---|---|---|
| USGS Water Services — Instantaneous Values (IV) & Site Service | U.S. Geological Survey | Public domain (a work of the U.S. Government, 17 U.S.C. §105) | 2026-07-14 | Reservoir elevation (parameters 62614 NGVD29 / 62615 NAVD88) and storage (00054, acre-feet), plus a 90-day series for the trend chart. |
| USACE CWMS Data | U.S. Army Corps of Engineers | Public domain (a work of the U.S. Government, 17 U.S.C. §105) | 2026-07-14 | Elevation, storage, and — for a minority of districts — a ready-made %-of-conservation-pool-full figure; also the source for lakes with no nearby USGS gauge. |
Source and vintage
Every figure on Lake Level Tracker is a real reading from USGS Water Services or USACE CWMS Data, retrieved 2026-07-14 and shown on every lake page as its "source and vintage" line. In all, this covers 1,055 lakes and reservoirs across 45 states — 535 sourced from USGS gauges and 520 standalone USACE lakes with no USGS gauge nearby (160 lakes have both sources matched and cross-checked).
Only a lake whose most recent reading is within a 60-day recency window is published — a gauge that stopped reporting does not get a stale, misleading page.
Elevation, datum, and why we never compare bare numbers across lakes
Pool elevation is always shown with its vertical datum — NGVD29 or NAVD88 — because both are in active use, sometimes within the same state, and they differ by roughly half a foot to a foot depending on location. A missing datum means we never show a lone, unlabeled elevation number, and we never compare two lakes' bare elevations without accounting for a possible datum difference.
70 lakes report storage but no elevation gauge at all — their managing agency simply does not publish a pool-elevation reading. Those pages show storage and say plainly that elevation is not reported; they never show an invented or estimated elevation.
%-full is source-specific — never computed by mixing sources
99 lakes carry a real "%-of-conservation-pool-full" figure, and it is used only where USACE CWMS Data itself publishes that exact figure for that exact lake. We never derive a %-full number by combining a USGS elevation reading with an unrelated "full pool" reference — if the source doesn't provide the ratio, the page says "not reported," not an estimate. A lake's %-full can also exceed 100% during a flood-control operation; that is a genuine reading, not an error.
The trend sparkline, the 30-day change, and a data-quality pass worth being upfront about
The inline chart on each lake page plots roughly the last 90 days of instantaneous readings, and the "recent trend" figure is the change over approximately the last 30 days of that window (rising, falling, or steady within ±0.1 ft).
While building this site we found that 12 of 1,393 lakes' 90-day series contained a small number of contaminated readings — almost always USGS's own internal "no data" placeholder value leaking into the series, plus a couple of one-off sensor glitches. These never affected the headline current-level figure (which is sourced separately and is clean for all 1,393 lakes), but they did distort the historical chart and, for 4 lakes, produced a physically-impossible 30-day change. We remove any single reading more than 500 feet from the lake's own current elevation — a threshold far beyond any real 90-day fluctuation and well short of the contaminated values — before drawing the chart or computing the trend, and we recompute the trend figure from the cleaned series when a point was actually removed. This affected 107 individual readings across those 12 lakes (0.9% of the dataset); every other lake's trend is the source's own figure, untouched.
Nearby lakes
Each lake page lists the six nearest published lakes by straight-line (great-circle) distance between site centroids — cross-state neighbors included. "Nearby" means distance, not a river-system or watershed relationship.
Honest gaps we chose not to hide
A few well-known reservoirs are missing on purpose, not by oversight. Lake Mead (Bureau of Reclamation) uses a separate API that was unreachable during data collection and is a named follow-on, not a launch blocker. The ~49 TVA reservoirs (Tennessee Valley Authority — Norris, Douglas, Watts Bar, Cherokee, and others) sit behind a bot-blocking source and are excluded at launch for the same reason. Boat-ramp status is dropped entirely: no structured, national source exists for it, so we link the managing agency's own page rather than publish a guess.
No interactive map in this version
Every lake page ships with a self-contained inline chart and zero external requests on load. An opt-in interactive map (loaded only if you click to open it) is a planned fast-follow, not something this version needed to ship a correct, fast, honest reference.
Update cadence
USGS and USACE refresh their underlying data sub-hourly to hourly; this site re-pulls periodically and moves the vintage stamp shown on every page. Between pulls, published figures are immutable.
Data: USGS Water Services (waterservices.usgs.gov) and USACE CWMS Data (cwms-data.usace.army.mil), retrieved 2026-07-14. Both public domain (a work of the U.S. Government, 17 U.S.C. §105).