Rogue waves

Why can one wave twice the height of its neighbours rise out of an otherwise ordinary sea?

On New Year's Day 1995 an instrument on the Draupner platform in the North Sea recorded a crest 18.5 metres above the still water level, in a sea state whose significant wave height was around 12 metres. Nothing about the storm was remarkable. One wave was.

The story usually told about such waves is that ordinary physics forbids them, so an exotic mechanism must be responsible — some nonlinear self-focusing that concentrates a train's energy into a single monster. It is a satisfying story. Before accepting it, though, it is worth checking the premise. Does ordinary wave physics actually forbid a wave twice the height of its neighbours?

Start with what a sea surface is. Not one wave but a superposition of many components, each with its own length, direction and phase, travelling at speeds that depend on their length. Because the phases are effectively random and independent, the surface elevation at any point is a sum of many independent contributions — which, by the usual argument for sums of many independent things, is close to a Gaussian distribution.

Now ask what that implies about wave heights. For a narrow-banded Gaussian sea, the heights follow a Rayleigh distribution, and it has a clean form: the chance that a given wave exceeds a multiple x of the significant wave height is exp(-2x²).

Put the rogue-wave definition into it. A wave is conventionally called rogue when its height exceeds twice the significant wave height. Set x = 2 and the probability is exp(-8), which is 1 in about 2,980.

Sit with that number, because it settles the question. A three-hour storm record with a mean period of ten seconds contains roughly 1,080 waves, so a fixed instrument has something like a one-in-three chance of catching a rogue wave in any such record. A ship crossing an ocean passes through vastly more waves than that, and the world's seas produce them continuously. Rogue waves are not forbidden by the ordinary statistics. They are predicted by them.

So the exotic mechanism was never needed to explain existence. What is left to explain is the discrepancy between the Rayleigh prediction and what is measured — because observed rates do run somewhat above it, and observed crests are higher than the linear theory says.

The first correction is not exotic at all. Real water waves are not sinusoids. Even in a weakly nonlinear description, each pair of components generates a bound harmonic that rides along with them, and the effect is systematic: crests become sharper and higher, troughs shallower and flatter. This second-order correction raises crest heights above the linear prediction without any energy focusing at all, and it accounts for much of the observed excess.

The second candidate is the exotic one. A narrow-banded, unidirectional wave train is unstable: sidebands grow at the expense of the carrier and the energy of many waves can concentrate into one. This is the Benjamin-Feir, or modulational, instability, and it is spectacular — in a long wave flume, and in optical fibres, where the analogous solution was observed directly. The difficulty is that its growth is strongly suppressed by two features every real storm sea possesses: broad frequency bandwidth and directional spreading. Large analyses of field data — one covering over a hundred million individual waves from fixed sensors — have found rogue-wave occurrence broadly consistent with second-order theory and no clear signature of modulational enhancement. Reanalyses of the famous events, Draupner included, have concluded much the same. This is a live and genuinely unresolved dispute, and I would not report either side as having won; but the burden has shifted, and "rogue waves are caused by modulational instability" is no longer a safe summary of the field.

There is a third mechanism, and it is the least glamorous and best documented. Waves can be focused geometrically. Running against a strong opposing current — the Agulhas off South Africa is the notorious case — wave energy is refracted and compressed, and rays converge as light does through a lens. Shoaling bathymetry does the same. Crossing seas, where two swell systems meet at an angle, appear to have been the setting for Draupner itself. These are not statistical flukes: they are places where the sea is genuinely more dangerous than the average, which is why rogue waves are not uniformly distributed over the ocean.

The correction worth making is about the reference point. The word "rogue" is defined relative to the significant wave height — itself the mean height of the highest third of the waves, and roughly twice the average wave. A rogue wave is therefore already being measured against the large end of the distribution, not the middle, and the impression that it "came from nowhere" partly reflects human sampling: an observer sees a few dozen waves and forms an expectation from those, then meets one drawn from a tail that a few dozen samples could never have revealed.

Two honest qualifications. The Rayleigh calculation above assumes a narrow-banded Gaussian sea and a stationary sea state; both are idealisations, and real records depart from them, which is one reason measured rates and theoretical rates never quite agree. And the practical question — whether a given rogue wave was a chance alignment, a bound-harmonic effect, or a focusing event — usually cannot be answered from a single point measurement at all, which is a large part of why the debate persists.

• Why a wave train's directional spread suppresses modulational instability, and how spread is measured.
• How the Agulhas Current turns a following swell into a documented shipping hazard.

Every term the collection defines is gathered in the glossary.