PROVISIONAL:all data, model outputs, and interpretations on this site are subject to change if superseding information is obtained, which isn't likely.
Floodgate ReportIndependent Investigation

The Record

This page exists for one reason: numbers in official feeds have moved — during the floods and after them — and this site watched them move, with timestamps. Everything below is stated so that no future version of any record, and no future account by any party, can quietly replace what the instruments actually served. Nothing on this page accuses anyone. It simply closes the exits.

Every revision we watched, contemporaneously documented
Hunt gauge, July 16, 2026 (live vs revised)
As served: Pulled live at ~11:00 AM during the flood: 2,540 cfs / 10.99 ft at 9:25 AM.
What it became: The revised record for the SAME timestamps now reads 33,800 cfs / 19 ft — the live feed ran more than 10× low at the peak, and the stage itself was later revised from 11 to 19 feet.
Kerrville gauge, July 16, 2026
As served: The record served through July 20 showed a peak of 49,800 cfs / 16.97 ft.
What it became: On July 21, USGS replaced the crest with an estimated block (flagged 'e', provisional): 180,000 cfs / 30.06 ft, inserted over a five-hour gap in which the gauge's own surviving readings top out at 23,300 cfs. This site updated within hours, in both directions of its own argument.
Guadalupe above Bear Creek, July 16, 2026
As served: The live feed showed a 7.0-ft crest during the event.
What it became: The verified record reads 28.92 ft — a 22-foot real-time under-read. Its discharge sensor had gone dark at 12:50 AM, before the peak.
Johnson Creek, July 16, 2026
As served: A dying sensor transmitted 95,900 cfs after its 12:55 PM malfunction; a bulk data endpoint served it from a stale cache even after USGS scrubbed it.
What it became: The true peak is 25,300 cfs at 10:10 AM. This site briefly published the bad number, caught it against the site-direct record within hours, retracted it everywhere, and then rebuilt its pipeline so no headline number can be published without a site-direct verification. The served series was also later found to contain a raw −999999 error sentinel inside the data itself.
The 2025 record — both key gauges
As served: Hunt's sensor died at 5:10 AM on July 4, 2025 (last recorded value: 106,000 cfs). Kerrville's gauge failed at its peak the same morning. Center Point failed. Above Bear Creek's flow sensor was dead.
What it became: Both published peaks — Hunt 315,000 and Kerrville 298,000 cfs — are USGS's own post-failure estimates, since approved. Hunt's estimate matches what its own rating curve yields from the surviving stage (37.52 ft → 314,290 cfs). This site labels every such number as an estimate, everywhere it appears.
The operator's ASR reports (City of Kerrville)
As served: Through September 2025, the City published its ASR operational data — storage volumes, injection activity, and the Well #4 (PZ-2) water levels that made the June 2025 injection ramp into the July 4 flood directly visible (1.05 billion gallons in storage on July 3, injecting).
What it became: Publication ceased. The last published data point is September 29, 2025 — roughly three months after the flood. The reports as published through that date are preserved in this investigation's archive; going forward, the aquifer's pressure is publicly observable only through the State of Texas's independent TWDB monitoring wells, which this site tracks live.

Why the conclusions survive all of this: no finding on this site rests on a single revisable number. The argument is anchored on complete, never-failed records (the Comfort gauge, both years; the Above Bear Creek stage record, both years), on floors that failures can only push higher, and on the convergence of independent instruments — radar, gauges, wells, surveyed high-water marks, and preserved audio. Revisions to date have moved individual numbers in both directions; they have not moved the shape of any finding.

Where every number comes from
RainfallNOAA/NWS MRMS Multi-Sensor QPE (gauge-corrected radar)mrms.nssl.noaa.gov · archived hourly grids preserved locally
Streamflow & stageUSGS National Water Information System (instantaneous values)waterdata.usgs.gov · every pull logged with its timestamp
GroundwaterTWDB / HGCD continuous monitoring wellswaterdatafortexas.org · live tracker on this site refreshes every 30 min
BasinsUSGS Watershed Boundary Dataset (HUC12)usgs.gov/national-hydrography
GeologyTWDB Hill Country Trinity Groundwater Availability Model 3.01 (2026)twdb.texas.gov · the Hammett zero-edge drawn on this site is the state's own
SoilsUSDA-NRCS SSURGO + NRCS Curve Number methodwebsoilsurvey.nrcs.usda.gov
Scanner audioBroadcastify archive, Kerr County public safetypreserved in-repo; the file's own epoch name timestamps it (10:52:38 PM CDT)
High-water marksUSGS post-flood surveyed HWMs (2025)gauge-independent peak anchors

Every source is a public government product or a preserved primary record. Nothing is proprietary. Every figure on this site can be re-derived from the sources above by anyone, free of charge — and the site states its methods on the page where each number appears, including which values are recorded, which are estimates, and which are floors.

The discipline that makes it hold
  • The verification gate.No headline value is published without a site-direct re-query of the source, outlier screening against neighboring readings, a cross-variable check (stage against flow against rain), and a unit sanity pass. This rule exists because it once caught this site's own error — see Johnson Creek above — and it has since caught error sentinels embedded in the served data itself.
  • Published nulls. When a test fails, it is published as failed: 6negative results to date, including findings against this investigation's own hypotheses — Mystic Spring thermal signature; Water-surface bulge at discharge points; Spring Branch discoloration, and more.
  • Estimates labeled as estimates.Both years' Kerrville and Hunt peaks are USGS post-failure estimates and are labeled that way in every table on this site, even where a bigger unlabeled number would make the argument look stronger.
  • Preservation. The radar grids, the gauge pulls with their timestamps, the well series, the scanner audio, and the working analyses are archived off-site. If a public feed changes tomorrow, the comparison against what it served yesterday exists.
The standing invitation

Check the work. Every exhibit is interactive, every source is public, and every method is stated where it is used. If any agency, office, or official holds records that supersede anything on this site, the provisional banner at the top of every page states exactly what happens next: the site changes to match the evidence. That is the entire policy — and it is one this site has already demonstrated on itself, publicly, within hours, in both directions.