The Ocean Has a Third Dimension

A THREE-DIMENSIONAL PLANET

From orbit, the ocean looks unusually well observed. Its surface temperature, height, colour, roughness, ice and winds can all be measured globally or nearly so. The difficulty begins a few metres down. An ocean is not a surface with some water underneath.

It is a moving volume, on average about 3,700 metres deep, whose temperature, salinity, oxygen, nutrients and currents all change with depth.

This matters because the surface is often answering to something far below it. Heat is stored hundreds or thousands of metres down. Dense water sinks. Nutrients rise. Currents carry properties sideways through the interior. A satellite can observe the consequence at the top while the state that produced it occupies a column several kilometres tall.

The ocean therefore exposes an awkward property of planetary observation: global coverage in two dimensions can coexist with sparse coverage in three. A map can look complete because every pixel has a value. That does not mean every layer beneath the pixel was ever observed.

THE BOTTOM

Most of the seafloor has never been directly mapped

In April 2026 the Nippon Foundation-GEBCO Seabed 2030 Project reported that 28.7% of the world's ocean floor had been mapped to modern standards. That is a remarkable increase on a few years earlier, and it still leaves more than two-thirds of the bottom of the planet without direct modern measurement. The surfaces of the Moon and Mars, for comparison, have been mapped in considerably more detail, partly because there is no ocean in the way.

The global maps are not blank in the rest. Satellite altimetry, the radar ranging of sea level described in Volume 2, picks up tiny bumps in the sea surface produced partly by the gravity of large features below, and those bumps can be turned into a broad estimate of the seafloor's shape. NOAA describes the result as low-resolution estimation that captures the overall form but can miss smaller seamounts, volcanic craters, wrecks and other features. The map of the deep seafloor is, for the most part, an inference drawn from the surface.

Proper mapping needs multibeam sonar on a ship or vehicle. A transducer sends a fan of sound pulses down through the water and times the echoes from the bottom, which is direct enough to reveal individual landforms and expensive enough that the ocean has had no difficulty keeping large areas to itself.

Depth makes the geometry worse. From a ship, the sonar footprint widens with distance, so NOAA notes that very high resolution in the deepest water needs instruments closer to the bottom: towed sleds, remotely operated vehicles or autonomous underwater vehicles. The place furthest from the observer is also the place where detail costs the most.

ARGO'S GAPS

Between the floats, and below them

Volume 2 described Argo's robotic floats, which made the upper ocean observable for the first time. The question here is what the array leaves out. In the open ocean the floats sit roughly 300 kilometres apart, and together they produce more than 100,000 temperature and salinity profiles a year. Each float samples one vertical line every ten days, while currents, eddies and fronts evolve in the spaces between. Ocean analyses fill those spaces with dynamics, interpolation and data assimilation. The result is scientifically excellent and, between the floats, mostly reconstructed.

The array is not uniform either. Argo's own status reports note regions that are over-populated and others with gaps, and keeping the coverage going takes roughly 500 to 600 replacement floats every year. Global observation is not a finished asset. It is a machine that has to be rebuilt continuously while the ocean drifts it out of position. Under the Antarctic winter sea ice, where floats cannot surface, much of what is known still comes from elephant seals.

Below 2,000 metres the problem starts again. Core Argo stops there because that is where much of the variability that matters for weather and climate is concentrated, and because the engineering was tractable. The deep ocean did not become shallower in response. Deep Argo floats, built for the pressure, profile to 4,000 or 6,000 metres, but they are still a developing part of the array, and roughly half the ocean's volume lies below the reach of the core floats.

A second extension changes what is measured rather than how deep. Biogeochemical Argo adds sensors for oxygen, pH, nitrate, chlorophyll, suspended particles and light. Temperature and salinity describe the physical ocean. These begin to describe its metabolism and carbon chemistry. They are also harder to calibrate and maintain, and every extra property adds another measurement problem to a volume of water that was already inconveniently large.

WHAT THE OCEAN TEACHES

A complete map can contain incomplete evidence

The ocean is a useful first blind spot because nothing about it is obscure in principle. Depth can be measured with sonar, temperature with thermistors, salinity with conductivity sensors and many chemical properties with dedicated instruments. The limit is coverage in space, depth and time, not knowledge of how to measure.

That makes it different from a variable that cannot be observed directly at all. The hidden state here is measurable, just not everywhere at once, so the working system combines direct measurements with interpolation and models. This is not a compromise peculiar to oceanography. It is the normal architecture of planetary knowledge.

The mistake comes later, when a continuous product is used without its observation geometry. A three-dimensional ocean model presents a value in every cell because the mathematics requires one. Some cells are pinned by a profile taken last week. Others are pinned by distant observations, physical equations and the model's own memory. On the screen they look identical. They carry very different amounts of evidence.

The ocean is not poorly observed because we forgot it was there. It is poorly observed because three dimensions are considerably more expensive than two.