How much of our own ocean have we seen?

Satellite altimetry gives us a low-resolution shape for the entire seafloor by measuring how seafloor mass deforms the sea surface. This is why world ocean maps exist at all. The resolution is coarse, features smaller than several kilometres are invisible.

High-resolution mapping needs a multibeam sonar, usually on a ship, sweeping a strip a few times as wide as the water is deep at around ten kilometres an hour. As of April 2026 Seabed 2030 counted 28.7 per cent of the seafloor as mapped, with almost five million square kilometres added in the preceding year.

Actually looking (with a camera, a submersible, or a remotely operated vehicle) covers a far smaller fraction than sonar mapping does. NOAA puts it at less than a thousandth of one per cent of the deep seafloor, an area about the size of Rhode Island, and more than nine tenths of the ocean is deep. A 2025 compilation of about 44,000 deep dives reached the same figure; its authors say it may be an underestimate, and that even ten times larger it would be under a hundredth of one per cent.

The difference is partly geometry. Light carries tens of metres through seawater at best, so a camera near the bottom sees a strip ten or twenty metres wide, while a ship’s sonar sweeps a swath several times as wide as the water is deep. And looking is expensive: every dive is a ship, a crew and a vehicle in the water.

Different agencies count differently: mapped at what resolution, observed by what instrument, explored to what standard. Figures from a thousandth of a per cent to a hundred all appear in reputable sources and they are not contradicting each other. They are measuring different things.

Seamounts are still being found. One reanalysis of satellite gravity data identified 19,325 that were not in the catalogue, against 24,643 that were, and a 2024 study by two of the same authors expects the SWOT satellite to reveal thousands more. Many appear rich in life, though a 2010 review found the idea that seamounts are islands of endemic species unsupported by the evidence.

Bottom trawling and climate-driven changes in ocean chemistry are already altering deep habitats that have never been surveyed, and deep-sea mining could add to them. The baseline against which those changes would be measured does not exist for most of the seafloor.

Seabed 2030 aims to compile all available seabed mapping information into a seamless map of the ocean floor by 2030, pooling data contributed by industry, government, academia and crowdsourced efforts alongside dedicated surveys. When the project began in 2017, 6 per cent of the seafloor had been mapped to an adequate resolution. At last year’s pace of about five million square kilometres, the remaining quarter of a billion would take about fifty years, so meeting the target means mapping more than ten times as fast.

Autonomous underwater vehicles work independently of the ship that launched them, which is what lets them into places a hull cannot follow: the deepest hydrothermal vents, the underside of polar ice. In deep water they also image the seafloor at resolutions a surface vessel cannot reach. For now they are slower than ships and inefficient over large areas, which several working together can offset. If the cost per square kilometre of direct observation falls far enough, the gap between mapped and seen could narrow in a way that ship-based exploration never could.