Watch the corridor move

Real stage readings from the Mississippi main stem, headwaters to Gulf. Pick a window and the river replays itself. Readings VerApex captured when they happened are sealed and every one resolves to a permanent proof; readings from before this archive existed are shown as source-reported — what NWS or USGS says today about a past hour, which nobody witnessed at the time. The page never blurs the two.

Frame time · UTC—

Highest station

—

Peak stage

—

Above flood stage

—

Frame provenance

—

Reporting coverage across the window

▸ drag across the hydrograph below to zoom the window to a selection

Memphis, TN

flood 34.0 ft · action 28.0 ft

▸ click any station above to change

Two kinds of reading, and the difference matters Sealed readings were captured by VerApex at the time they describe and written to the ledger; each one resolves to a proof. This archive begins — and grows by a day every day, so the sealed span widens on its own. Source-reported readings are what NWS or USGS serves today about a past hour. They are read live from the source, are never sealed by being displayed here, and carry no proof link. They prove what the source says now — not what it said then, because nobody was watching then. Reach is a hard limit, not a choice: NWS retains about 30 days and covers the whole corridor; USGS retains 120 days but publishes gage height for only 29 of these stations. Scrub past 30 days and the corridor visibly thins. That thinning is real and is left visible on purpose.

Held readings, and why they are dimmer Gauges report on staggered 15-to-60-minute cadences, so at a five-minute frame the average moment has an actual reading from only about 19 of the plotted stations. Drawing just those would make the corridor flicker, and would read as gauges going dark rather than as frames finer than the data. Each station therefore shows its most recent reading — the way a stage hydrograph is read anyway. A carried-forward value is dimmed, never presented as fresh, and it is dropped entirely after two hours. Past that, "still 4.3 ft" would stop being the last reading and start being a claim about a gauge that has gone quiet.

Why stage is plotted against each station's own flood stage Absolute stage is not comparable across stations — every gauge sits on its own vertical datum, so a 73 ft reading upstream and a 4 ft reading downstream say nothing about each other. Normalizing each reading to a fraction of that station's flood stage is how hydrologists read a corridor, and it needs no assumption the data does not support. It also sets who can appear: a station is plotted only if NWS publishes a flood stage for it and its gauge is not on a pool-elevation datum, where the ratio would be arithmetically true and physically meaningless. That is 67 of the corridor's 102 stations. The other 35 are still captured, still sealed, and still on the live dashboard — they just have no honest denominator for this particular chart.

What people do with this

Maritime litigation

The grounding at mile 232

"Your Honor, the defendant claims the shoaling was unforeseeable. Here is the river in the seventy-two hours before the grounding."

A tow grounds on a falling river. The operator argues conditions changed without warning. The replay shows a sustained fall over three days, published hourly by NWS, available to anyone with a browser — including the defendant's dispatch desk.

The exhibit shows Stage trend at the incident location, hour by hour The nearest gauge and its distance, stated on the record Each frame's ledger proof, sealed before suit was filed

$999–2,999EXHIBIT PACKAGE

Claims & adjusting

When did the water actually arrive

"The policyholder says they evacuated at 9 p.m. The crest reached their reach at 3 a.m. Their account holds."

A flood claim turns on sequence. The adjuster needs to know when the crest arrived, not just that it did. The replay places the arrival to the hour and, just as usefully, rules out timelines that do not fit — supporting a fast pay or a defensible denial without an expert engagement.

The exhibit shows Crest arrival time at the nearest station Duration above each flood category Upstream lead time — what was knowable, and when

$299–999VERIFIED RECORD

Inland marine ops

Three days of warning on draft

"The trough hits Vicksburg Thursday. Cut tow size now, not Thursday morning."

Low water strands barges more reliably than high water does. Because a wave takes days to travel, an upstream reading is a downstream forecast. Dispatch watches the trough move and sizes tows against it — the same replay, run forward instead of backward.

The operator sees Propagation timing between reaches Low-water threshold crossings, per station Alerting the moment a reach crosses a set limit

$299–999/moAPI + ALERTS

Parametric insurance

Did the trigger actually fire

"Policy pays on 48 sustained hours above minor flood. Here is the window, sealed, with the source URL."

Parametric settlement is only as clean as its trigger source. The replay makes the trigger window visually unambiguous and mechanically checkable — the insurer, the insured, and the reinsurer all read the same sealed record and none of them has to take the others' word for it.

The settlement shows Exact hours above threshold, continuous or not On-demand attestation over the trigger window Independent verification path for every party

NegotiatedSETTLEMENT ORACLE

Public infrastructure

Reconstructing the bridge strike

"Current and stage at the structure, in the ninety minutes before impact."

A barge-mounted crane strikes a bridge. The investigation needs conditions at the structure at the moment of impact. Historically that meant a records request and a consultant. The replay produces it in minutes, with provenance attached to every value.

The investigation gets Conditions at the nearest gauge, minute-resolved where the source published it Explicit gaps where no reading exists — never interpolated A certified dataset suitable for the record

$4,999CERTIFIED DATASET

Commodity risk

Freight constraint, three days early

"Constraint reaches the export elevators Friday. Position accordingly."

Grain freight rates move with river navigability. The corridor replay turns a scatter of gauge readings into a propagating constraint with a visible arrival time — which is a tradable signal, and one that can be verified after the fact rather than argued about.

The desk sees Constraint propagation with arrival estimates Historical analogues from the archive Attested history for post-trade review

$10k–100k/yrDATASET LICENSE

Behind every frame

The replay is a view, never a source Each frame is assembled from sealed observations — it does not smooth, interpolate, or fill. Where a source published nothing, the replay shows nothing and says so. Where two sources disagreed, the conflict is carried through rather than averaged away. Every point a viewer can click resolves to a content-addressed record on the ledger and the public source URL it came from. The animation is persuasive; the record underneath it is what survives cross-examination.

VerApex · verapex.io · sources: NOAA National Water Prediction Service, USGS Water Services
Flood categories are official NWS thresholds. Crest simulated for demonstration. Production replay reads sealed records from the archive.