In almost seven decades of deep-sea exploration, the direct human view of the deep ocean floor is still smaller than many people would guess. A Science Advances study published in May 2025 compiled 43,681 submersible dive records dating back to 1958 and estimated that humans have visually observed less than 0.001% of the deep seafloor.
The comparison used by the researchers is deliberately plain: the area seen is roughly the size of Rhode Island. Set against the deep seafloor, which the paper defines as ocean bottom below 200 metres and which covers about 66% of Earth’s surface, that is not a large sample. It is a thin scattering of visual records across the largest habitat on the planet.
The paper, led by Katherine L. C. Bell of Ocean Discovery League with Kristen N. Johannes, Brian R. C. Kennedy and Susan E. Poulton, is titled How little we’ve seen: A visual coverage estimate of the deep seafloor. Its point is not that researchers know nothing about the ocean. It is that one particular form of knowledge, direct visual observation of the deep seabed, remains extraordinarily limited.
Visual observation is not the same as mapping
This distinction matters because ocean knowledge is not one thing. Satellites, gravity models and shipborne sonar have all helped scientists map the broad shape of the seabed: trenches, ridges, plains, slopes, seamounts and basins. Those maps are essential, but they do not show everything that matters.
A camera close to the bottom can reveal what a map cannot. It can show sponge gardens, coral communities, hydrothermal vents, nodule fields, carcass falls, trawling marks, cables, sediment texture, animal behaviour and organisms too small or too localised to appear in global bathymetry. It can also show absence: places where the seafloor is bare, disturbed or different from what researchers expected.
Bell and colleagues focused on that close visual record. They gathered records from crewed submersibles, remotely operated vehicles and other underwater platforms that had looked directly at the deep seafloor. The result was a historical tally of dives, not a claim that every dive saw the same amount or the same kind of habitat.
That is why the number is useful. It makes visible the scale of the sample. A dive that feels expansive to the people conducting it may still cover a narrow path on a global seafloor map. Decades of such dives can add up to an impressive archive and still leave nearly all of the deep seabed unseen.
A small sample, and not an evenly spread one
The study also found that visual exploration has been geographically concentrated. A large share of the compiled observations occurred near a small number of countries with long-standing access to deep-submergence technology. In the dataset, 65% of visual observations were within 200 nautical miles of the United States, Japan and New Zealand, with 97% of compiled dives conducted by five countries: the United States, Japan, New Zealand, France and Germany.
That pattern is not surprising. Deep-sea work is expensive. It requires research vessels, pressure-rated vehicles, pilots, engineers, launch and recovery systems, cameras, lights, navigation tools and teams capable of operating far from shore. Historically, only a limited number of institutions and nations have been able to sustain that capability over many years.
But concentration creates a scientific problem. If most visual records come from a small set of accessible or well-funded regions, researchers risk building global assumptions from local views. A habitat may look rare because few vehicles have passed through places where it occurs. A species may appear limited because cameras have not visited its range. A region may seem undisturbed because no baseline image exists from before a disturbance began.
The paper’s figure therefore does two things at once. It measures the smallness of the visual sample, and it warns that the sample is uneven. Less than 0.001% would be thin even if the observations were perfectly distributed. They are not.
Why this matters now
The timing of the study is important. The deep ocean is no longer a remote scientific category discussed only by oceanographers. It sits at the centre of debates over biodiversity, climate, fishing, carbon cycling, cable routes, biotechnology and possible seabed mining. Some of the places attracting commercial interest, including nodule-rich abyssal plains, are also places where ecological baselines remain sparse.
Visual records are not the only baseline that matters. Chemical measurements, biological samples, acoustic surveys, sediment cores, genetic data and physical oceanography all contribute to a clearer picture. But images and video are often the evidence that lets researchers identify habitats, compare conditions over time and communicate what is actually present before an area is disturbed.
Without that record, it becomes harder to assess damage or recovery. If a seafloor community is altered by mining, trawling, climate-driven change or industrial activity, scientists and regulators need some account of what was there before. A map can say where the seabed is. A visual record can help show what lived there.
The authors also acknowledge an important limit: not every dive record is public, and not every private or industrial survey is available to science. Oil, gas, mining and telecommunications work may have collected images that are not part of open research archives. That means the observed area may be somewhat larger than the compiled public record, but the central result is unlikely to change in practical terms. Even an order-of-magnitude correction would still leave direct visual coverage at a tiny fraction of the deep seafloor.
The hard physics of looking down
The low number is not a sign that ocean researchers have been idle. It is a sign of how difficult the place is. Below 200 metres, sunlight drops away quickly. Pressure rises by about one atmosphere every 10 metres. At several kilometres depth, vehicles must work in darkness, cold and crushing pressure while communicating through a medium that does not behave like air or space.
Radio waves do not travel well through seawater, so underwater vehicles often depend on tethers, acoustic communication or stored instructions. A remotely operated vehicle can spend hours moving carefully across a patch of seabed that feels large on a monitor and vanishingly small on a planetary scale. A crewed submersible dive can require ship time, calm enough weather, a trained team and careful planning before a single image is captured.
That is why the study’s numbers are sobering rather than embarrassing. The deep ocean is not merely underexplored because it has been neglected. It is underexplored because it is immense, technically hard to reach and expensive to observe directly.
What a better record would require
Bell and colleagues frame the gap partly as a data problem and partly as an access problem. Better shared archives would help. So would lower-cost vehicles, standardised metadata, more open industrial imagery where possible, and deeper participation by coastal and island nations whose waters and knowledge have often been underrepresented in deep-sea research.
That does not mean every square kilometre of deep seafloor must be filmed before any decision can be made. It does mean the scale of uncertainty should be visible in those decisions. A planet can be mapped in outline and still be scarcely seen in detail. The 2025 tally gives that idea a number.
Less than 0.001% is small enough to change the tone of the conversation. The deep seafloor is not a distant minor region. It is most of Earth’s solid surface beneath the ocean, and the part humans have directly observed is roughly comparable to a small U.S. state.
After nearly seventy years of submersible exploration, that is the central finding. The deep ocean has not been ignored. It has been glimpsed.
