The lake that did not rise 117 metres

The largest lake in Norway read under 8 metres for decades and reads over 120 today. The water did not move; the ruler did. What the zero of a gauge is, why a record only means something on one reference, and what we do when the zero moves.

Open the archive for Mjøsa at Hamar and the lake appears to do something impossible. From 1976 until November 1998 its daily level runs between 1.64 and 7.93 metres. From that month on, the same station reports between 119 and 125. Nothing happened to the lake that month. Nobody built a dam 117 metres high across Norway’s largest lake. The agency changed what zero means.

Both sets of numbers are correct. They are just not the same measurement, and anything that reads them as one series, a record low, a typical level for the date, a chart, comes out wrong.

What a gauge measures

A water-level gauge does not measure how deep the water is. It measures how high the surface stands above a fixed mark, and that mark is the gauge’s zero. On a staff gauge bolted to a bridge pier the zero is wherever the bottom of the staff happens to sit, usually chosen a little below the lowest water anyone expects, so the readings stay positive. A reading of 2.40 metres says the surface is 2.40 metres above that mark. It says nothing about the river bed, and nothing about the sea.

That is why a level only means something at its own gauge. Two stations a kilometre apart on the same river can read 0.8 and 3.1 on the same calm afternoon, and both be right.

Two ways to choose a zero

The first is a local zero: a mark that suits the site. Most river gauges started this way, and many still work this way, because for the people using them the question is how the river compares with itself.

The second is to tie the zero to the national height system, so the gauge reports the elevation of the water surface above sea level. Lake Muskoka at Bala works this way, which is why its readings sit near 226 metres rather than near 2. A lake level in that form can be laid against a map: the contour of a shoreline, the sill of a boathouse, a flood line drawn by an engineer.

Agencies move gauges from the first kind to the second as networks are surveyed and modernised. When they do, the published series can simply continue, old numbers followed by new ones, with nothing in the numbers themselves to mark the change.

When the zero moves

Mjøsa at Hamar is the clearest case we have met. Otta at Lalm, a river station in the mountains to the north-west, is another: between 0.27 and 5.15 metres until 1995, between 353 and 359 since.

Left alone, those series produce confident nonsense. The all-time low for Mjøsa becomes 1.64 metres, about 118 metres below anything the lake has done since. The typical level for a date becomes an average of two unrelated numbers. A chart of the last fifty years shows a cliff.

Shifts that large are easy to catch. The small ones are the trouble. At Folla at Dølplass the level drops by 0.78 metres in June 2005 and stays there. A river can do that by itself: a regulated reach, a bed scoured by a flood, a run of dry years. So can a surveyor with a new benchmark. The way to tell them apart is to look at more than the average. When the zero moves, low water, ordinary water and high water all shift by the same amount, on the same day. When the river changes, they do not. At Folla all three moved together.

And some jumps are not errors at all. Sira at Langhølen, in the south-west, runs at a third of a metre for months and then stands four to six metres higher for days at a time. Those are real high-water events on a regulated river, and they stay in the record.

What we did with Norway

When Nibi Metrics boarded Norway’s network this month we read fifty years of daily values for 617 stations, about 7.4 million numbers, and checked every station’s history against the level it reports today. At 74 stations part of the history sat on another reference or could not be that water at all: 365,423 values, just under five per cent of everything we read. At 17 of them it was more than half the level record. At twelve stations there was a smaller shift of the zero, from 0.5 metres to 12, that we confirmed one by one.

On those stations the level history on Nibi starts where today’s reference starts. Mjøsa’s begins in November 1998. Its record high on the page, 125.14 metres on August 13, 2023, is the highest since then, not the highest the lake has ever stood; earlier floods are in the agency’s archive under the older numbers. The flow record is untouched, because discharge was computed by the agency from the rating in force at the time and does not depend on where the zero sat.

We did not convert the old values. Mjøsa’s older numbers look as if adding about 117 and a half metres would bring them into line, and the temptation is obvious. But nothing in the published series states that offset, and a record built on a number we worked out ourselves would be ours, not the agency’s. A shorter record that is right is worth more than a longer one that might be.

Reading a level on any page

Three habits follow from all this.

Read the difference, not the level. The large number on a location page is how far the water stands from what is typical for the date at that gauge. It does not care where the zero is.

Do not compare levels between gauges unless both report height above sea level. Compare their differences from typical instead.

And look at how far back the record goes. An all-time high means the highest in the record at that station, on the reference it uses now, and that record can be shorter than the station is old.

The figures are from the daily series published by the Norwegian Water Resources and Energy Directorate (NVE), licensed under NLOD, as read by Nibi Metrics in October 2026.