How the current-state route model works

1. From a gauge reading to available depth

A river gauge does not publish depth. It publishes water level at one place, in centimetres, against a local zero. Waterway authorities also publish, for each free-flowing reach, a reference low level and the fairway depth maintained at that level. We convert one into the other:

available depth (m) = maintained depth at the reference level + (reading − reference level) ÷ 100

So a gauge sitting 20 cm under its reference level implies roughly 0.20 m less water than the maintained depth. The Rhine pairs currently published for Kaub and Köln match BfG 2026, Anlage A, Table A-1, whose GlW header is referenced to 2002; the German-Danube pairs are transcribed from KWD 2020. Maxau and several other reference values remain approximations and are marked as such in our reference data — verifying every pair against a co-vintage primary source remains open work, and future value corrections will use the model-change protocol.

2. Too little water: from depth to a reach risk

Your ship's draught is how much water it needs. We use the operator's figure where we have one and 1.60 m where we do not, then subtract a fixed 0.20 m safety allowance — nobody sails with the keel touching:

margin = available depth − draught − 0.20 m

• Below ship threshold — modeled margin below zero.
• Limited model margin — margin between zero and 0.25 m.
• Within model threshold — margin above that.

One modifier: if the gauge has fallen 8 cm or more over the past 72 hours and the margin is under 0.40 m, we move the reach one level worse. A thin margin on a falling river is not the same thing as a thin margin on a steady one.

Reaches that locks control

Not all river is the same river. Impounded stretches (the Upper Rhine, the Main, the Austrian Danube) and artificial ones (the Main–Danube Canal) hold their depth with weirs and locks, so we treat them as not depth-limited by this low-water calculation and show their class on every day chip. This does not account for every lock closure, maintenance event, local restriction, bridge constraint or operator decision.

Budapest is judgment, not a model

We have no published maintained-depth model for the Hungarian Danube, so the Budapest gauge uses calibrated thresholds instead: under 100 cm reads below model threshold and under 150 cm reads limited model margin. Those are Waterlogged judgment bands, not authority navigation limits, and every report that uses them says so on the card.

3. Too much water: the other tail

A river cruise is disrupted by high water as often as by low, and the two fail differently. On top of the low-water banding we apply the published high-water marks, on every class of reach and both kinds of gauge:

• At or above HSW (Mark II) — authority materials identify a navigation mark. We label that state at/above navigation mark; operational status still requires confirmation from the authority and operator.
• At or above Mark I — speed and wash limits apply, and this is the level class where bridge clearance starts to bind. We call that too much water — caution.

Two findings put that second rule there. Air draft binds before the legal stop marks: in December 2023 four major lines were disrupted with Kaub near 600 cm against a 640 cm navigation-stop mark, on a legally open river. And the drought-proof middle is not flood-proof: impounded reaches are exempt from low water because the locks hold the depth, but the locks close at HSW — Austria lost 27 days that way in June 2013, and the whole Austrian Danube shut for 11 days under Storm Boris in September 2024.

High-water marks are per gauge, and several of ours are still blank (the Danube's own chokepoint at Pfelling among them). A gauge with no transcribed marks carries no high-water view at all rather than an invented one, and the report says so where the marks would be.

4. From reaches to the three-state model result

• Any reach at or above its navigation mark → current threshold crossed.
• Any day crossing a free-flowing reach below its ship threshold → current threshold crossed.
• Otherwise, any reach above its first high-water caution mark → limited current margin.
• Otherwise, any day with limited model margin → limited current margin.
• Otherwise → no current threshold crossing.

The asymmetry is deliberate: low water is exempt on lock-held reaches, high water is not.

These labels describe how the latest accepted readings compare with configured ship and high-water thresholds. They do not predict cancellation, substitution, coaching, the itinerary a passenger will receive or whether any operator program applies. Day counts identify where the model threshold occurs on the mapped itinerary; they are not claim values.

Phase 2 now has a fixed passenger-level outcome taxonomy. Read what counts as sailing as booked, a modified itinerary, or a cancellation/hotel-night outcome. The percentages published against those three outcomes come from the published model, v1, and every one of them is preceded by the calibration tag: they are calibrated judgment, not a validated statistical model. The candidate model v2 runs in shadow on the same inputs and publishes nothing; it cannot replace v1 until its pre-registered hydrology and reviewed-outcome gates pass. Where required gauge coverage is incomplete, or a departure falls in the winter hiatus, the percentages are withheld and the page says which of those it is.

5. How far ahead any of this means anything

The single most important thing on a report is not the colour, it is the horizon. We publish a three-rung ladder and refuse to claim above the rung a departure sits on:

• Operational, up to 14 days. The waterway authorities publish a daily level distribution at supported Rhine gauges. Waterlogged separately shows a current-state itinerary scenario: its colours mean "if the latest measured levels persisted," not forecast levels for those dates.
• Tactical, 15 to 46 days. The BfG six-week product retains weekly ENS, ESP and OBS distributions. It indicates broad range and direction at supported German gauges; it cannot supply a level for a particular cruise day. The day strip shows exposure and stops colouring dates.
• Structural, beyond 46 days. No operational river forecast exists at all, ours or anyone's. The three-state call is withdrawn and replaced with route exposure and conditioned climatology. Nearly half of river-cruise bookings are made in this window, so it is the horizon most readers are actually in.

Waterlogged does not project future stage in its route score. The current-state scenario applies the latest reading and 72-hour trend band to each route reach. Official BfG distributions are a separate evidence type, preserved at daily or weekly resolution and translated only through the ship thresholds already published on the report.

6. Historical context beyond the official forecast

Beyond the tactical window no sailing-specific forecast is available. We may show a guarded retrospective archive count: what happened in Rhine seasons grouped into the same historical Kaub comparison band. It is context about the archive, not conditioned climatology, not a passenger probability, and not a forecast for the departure.

• Counts, never sailing probabilities. The report uses forms such as "9 of 10 seasons" and states the period and definition. It does not convert that count into a chance for the passenger's cruise.
• Thin cells publish nothing. An analogue band with fewer than eight years in it produces no number at all, and the report says it cannot say.
• Never for high water. There is no honest seasonal high-water number: in our own event study the July 2021 flood sat below base odds four weeks out, and high water only becomes visible in the final week or two. We publish odds for the low tail and monitoring for the high one.
• It is Rhine archive context. If it appears beside a non-Rhine itinerary, the report states that it is not basin-specific evidence. The band edges remain a judgment heuristic rather than measured percentiles.

7. The exposure gate: is the fleet even sailing?

Before interpreting the model, we ask whether the catalog places the departure in the usual operating season. Our reference catalog contains very few departures from January to mid-March, so winter gauge events cannot be treated as passenger-outcome evidence for an ordinary sailing.

So a winter departure gets the exposure banner above the call, and a shoulder-season one (late March, or the Christmas-market weeks) is framed as the thin-fleet case it is. The gate does not change any risk level. It changes what the report is entitled to imply.

8. What we do not know

• Our route score is not a forecast. It is a current-state scenario. Where BfG publishes a distribution, that authority forecast appears in a separate panel with its full uncertainty and provenance.
• Itineraries are approximate. Day-by-day port sequences are hand-curated from marketing material and flagged approximate_v0. Your operator's actual routing may differ, and they change it in low water — which is precisely the event we are describing.
• Reference levels need verification. Several are transcribed approximations pending a line-by-line check against the source tables.
• Some high-water marks are missing. Pfelling and Hofkirchen carry none. Kienstock's are now transcribed (viadonau KWD 2020: HSW 601 cm, and the authority's own stated "HSW + 90 cm" stop at 691 cm), so 2013's Austrian closure and Storm Boris would now be visible to the engine at that gauge — Pfelling/Hofkirchen still are not.
• Some reaches are coarse. The free-flowing Wachau stretch inside the Austrian Danube is not modelled separately, so that reach reads more forgiving than it is.
• Required coverage fails closed. If a required gauge or required archive input is missing, invalid or behind its freshness target, the customer assessment is withdrawn and shown as unavailable. We do not substitute another gauge or retain the prior route result as though it were current.

9. Sources, licences and limits

Live German water levels come from PEGELONLINE (DL-DE-Zero 2.0). Voluntary provenance credit: Data source: Wasserstraßen- und Schifffahrtsverwaltung des Bundes (WSV), PEGELONLINE. Long-history Bavarian Danube files for Pfelling and Hofkirchen come from GKD under CC BY 4.0: Datenquelle: Bayerisches Landesamt für Umwelt, www.lfu.bayern.de. They are the Phase 2 Danube validation source; they do not change a published score or odds table until the evidence gate passes. Budapest and Kienstock live levels come from EuRIS Hydrometeo v3. Freshness targets in this build: PEGELONLINE 45 minutes and EuRIS 90 minutes. EuRIS identifies OVF as the Budapest originator and viadonau as the Kienstock originator. Direct OVF and DoRIS adapters remain dormant pending written permission — see Legal & disclaimers and /status.

API/Service Hydrometeo v3 incorporated from EuRIS (eurisportal.eu).

PEGELONLINE's historical files are unverified raw observations. GKD exposes both individual observations and daily means with a quality-review status. We preserve each downloaded source file before parsing, retain its station, units and time basis, and label aggregation, exclusions and model calculations as Waterlogged transformations. Both authorities disclaim accuracy, completeness, availability and resulting loss; an official source is not a warranty of a Waterlogged conclusion.

Waterlogged River Cruise Watch is an information product for travellers. It is not travel advice, it is not a navigation product, and nothing here may be used to plan the movement of a vessel. Decisions about your booking are between you, your travel adviser and your cruise line.