These plots monitor Rhine water level as it passes through Bonn and put it in historical context.

Few explanations:

1. Daily levels (black line) come from the German Federal Institute of Hydrology (BfG). The record starts in 1960, but it is only updated every few weeks.
2. Live levels (green line) come from the Bonn gauge itself (PEGELONLINE, WSV): one reading every 15 minutes for the last month. This fills the gap at the end of the BfG record.
3. Both sources measure from the same zero point, and on the days when both have data they agree to about 0.1 cm, so neither is adjusted.
4. The red lines show what is normal for the time of year, worked out from the years 1990 to 2021.
5. For each day of the year I did not use that one date alone. I used every value within 5 days either side, i.e., an 11-day window. That gives about 350 values per day instead of about 32. This matters for the outer percentiles: with only 32 values, "P01" and "P99" are just the single lowest and highest readings, which jump around from one day to the next.
6. Old data are corrected. On 1 November 1979 the gauge zero was moved down by exactly 1 metre. This means readings before that date are 1 metre too low. I added 100 cm to them, so the whole record can be compared directly. Without this correction the 1960s and 1970s look far drier than they really were.
7. When you select more than 3 years, the red lines and the black anomaly line are hidden, because otherwise the plots are unreadable. Click them in the legend to bring them back.

<iframe
    src="https://sites.ecmwf.int/ecbox/joaquin/rhine/index.html"
    width="100%"
    height="900"
    frameborder="0">
</iframe>

Twenty lowest daily water levels at Bonn, on the current gauge zero (1960–2026):

Only three years appear — 2018, 2022 and 2026 — and 2026 holds 11 of the 20, including the record low of 75 cm, reached twice in the last ten days of data.

Two points worth noting:

None fall before November 1979, so the datum correction doesn't affect this ranking at all. It would have mattered: on the uncorrected data the pre-1979 readings sat a metre low, and several would have crowded into this table as spurious record lows.

The 2026 events are more extreme than their raw levels suggest. The 2018 dates occur in late October when the climatological mean is ~225 cm, whereas the 2026 dates fall in late July and August when it's ~255–270 cm. So 2026 sits roughly 175–188 cm below normal against 142–155 cm in 2018 — a deeper departure, not merely a similar level at a different time of year. On the interactive page these days sit below the P01 line, which for that time of year is around 124 cm.

Twenty most negative daily anomalies at Bonn (1960–2026):

A completely different list from the lowest levels — no overlap at all. These are two clusters: the 2016/17 New Year period (14 days) and late winter 1963 (6 days). All fall in December–March, when the climatological mean is at its seasonal peak near 385 cm, so a level of ~120 cm produces a −260 cm departure. The record-low levels of 2018/2022/2026 don't appear here because they occur in late summer, when the mean is only ~225–270 cm, capping the possible departure.

Put plainly: the lowest levels are summer events, the largest anomalies are winter events. August 2026 set the absolute record at 75 cm; January 2017 was further from normal despite standing 40 cm higher.

Two things worth noting:

Six of these are pre-1979 and depend on the correction we just applied. On the uncorrected data those 1963 days would have shown anomalies near −360 cm and dominated the entire ranking — a pure artefact. They're still genuinely extreme after correction, just not record-breaking.

The 1963 P01 values look odd next to 2017's — around 198 cm for early March against ~124 cm for early January. That isn't an error: the percentiles come from the 1990–2021 climatology, which has a much higher 1st percentile in March than in January. It does mean the 1963 days sit far below their seasonal P01 while several 2017 days sit just above theirs, which is why the "below P01" column disagrees with the anomaly ranking in places.

Record low levels and largest anomalies

Calculated over the whole series, 1 January 1960 to 2026-09-06 (24,356 days), combining the BfG record with the live Bonn Pegel readings. All values are on the current gauge zero, so the years before November 1979 have been corrected upwards by 1 metre.

The 20 lowest daily water levels

The 20 most negative anomalies

What this means

The two tables answer different questions, and 2026 dominates one while being completely absent from the other.

August 2026 is a genuine record in absolute terms. Nine days in a row, from 11 to 19 August, sit below every previous reading in 66 years, with a low of 63 cm on 17-18 August. This was a sustained event, not a single spike. It also happened after the BfG record ends, so it is only visible thanks to the live gauge readings.

But it is not the largest departure from normal. Summer levels are naturally low: the normal level in mid-August is about 250 cm, so even 63 cm gives an anomaly of about -189 cm. Winter levels are normally about 385 cm, so January 2017 at 117 cm gives -266 cm. The seasonal cycle limits how negative a summer anomaly can be: the river cannot fall 260 cm below normal when normal is only 250 cm.

The largest anomalies are all winter events, in December-March, concentrated in three episodes: January 2017 (11 days), March 1963 (6 days) and late December 2016 (3 days). Six of these are from before November 1979 and depend on the 1-metre datum correction; without it they would appear far more extreme than they really were.

Comparing against the seasonal 1st percentile sharpens the picture. The 2026 days sit around 120 cm below a P01 of about 120 cm for the time of year, while the 2018 days sit near a P01 of about 84 cm. Relative to what is normal for that point in the calendar, 2026 is the more unusual event.

Table generated automatically on 07 September 2026.