GEBCO

From Monaco in 1903 to Seabed 2030

GEBCO, the General Bathymetric Chart of the Oceans, goes back to the 7^th International Geographical Congress in Berlin, which in 1899 named a commission on the names of undersea features. Meeting in Wiesbaden on 15 and 16 April 1903, that commission accepted the offer of Prince Albert I of Monaco, himself an oceanographer, to organise and fund a series of bathymetric charts of the world. The first edition, 24 sheets, was printed in Paris in 1905.

A second edition was printed from 1912 to 1931, and responsibility then passed to the International Hydrographic Bureau, forerunner of today's International Hydrographic Organization (IHO). The Intergovernmental Oceanographic Commission (IOC) of UNESCO was invited to co-sponsor the fifth edition, which completed world coverage at 1:10 million in 1982; GEBCO still works under the joint auspices of the IHO and the IOC.

The charts went digital with the GEBCO Digital Atlas on CD-ROM in 1994. The first global grid, the One Minute Grid, came out in 2003, the year of the centenary, followed in 2009 by a 30 arc-second grid, GEBCO_08.

In June 2017 the Nippon Foundation of Japan and GEBCO launched the Nippon Foundation-GEBCO Seabed 2030 Project, at the first UN Ocean Conference in New York, to map the entire ocean floor by 2030. At the outset only about 6% of the seabed had been mapped to an adequate resolution; with the 2026 grid the share has reached 28.7%, some 104 million km². Since 2019 the project has produced the GEBCO grid every year at 15 arc-seconds, and the 2026 grid is its eighth.

What it measures: soundings and gravity from space

The GEBCO grid is a continuous terrain model of ocean and land: one elevation in metres every 15 arc-seconds, in 43,200 rows by 86,400 columns, about 3.7 billion values, negative below sea level and positive above. Each value refers to the centre of its cell, about 460 m from north to south, and narrower from east to west away from the equator.

Behind that single surface lie two very different kinds of depth. Where a ship has surveyed, mostly with multibeam echo sounders that sweep a swath of seabed, but also with single-beam soundings, lidar, or contours digitised from charts, the depth is measured. Elsewhere, in more than two thirds of the ocean, it is predicted from space: the mass of a seamount pulls the water above it and raises the sea surface very slightly, a bump that satellite altimeters measure, and the gravity field so mapped is turned into depth, calibrated on the ships' soundings.

Those predicted depths are much coarser than the grid. SWOT, the French-American satellite launched in December 2022, sharpens marine gravity to about 8 km, against roughly 200 m for a ship's echo sounder. In unsurveyed ocean a 15 arc-second cell is therefore an interpolation of features several kilometres across, and a small seamount may be missed or its summit flattened.

The companion Type Identifier (TID) grid tells, for each cell, where its value comes from: land (0); a direct measurement, such as single-beam (10), multibeam (11), lidar (15) or a combination (17); an indirect one, such as predicted from satellite gravity (40), interpolated by a computer (41) or taken from chart contours (42); or an unknown source (70 to 72). Read with the depth, it tells a measured bank from a guessed one.

How a yearly grid is compiled

Four Regional Centers of Seabed 2030 gather the data of their oceans: the Southern Ocean at the Alfred Wegener Institute in Germany, the Pacific at Earth Sciences New Zealand, the Atlantic and Indian Oceans at the Lamont-Doherty Earth Observatory of Columbia University, and the Arctic and North Pacific at Stockholm University and the University of New Hampshire. They compile gridded data sets, largely from multibeam surveys, and pass them to the Global Center at the British Oceanographic Data Centre (BODC) in the United Kingdom.

The Global Center lays them on a base grid, version 2.8 of SRTM15+ from the Scripps Institution of Oceanography, used largely between 50S and 60N, which fuses land topography with measured and predicted seafloor; this version predicts depth from SWOT's gravity field with machine-learning methods. Outside the polar regions the regional grids are sparse, holding only the cells with data, and are merged by "remove-restore": the difference between the new data and the base is gridded and added back, so the surface follows the measurements and joins smoothly the predicted depths around them. The complete polar grids are feather-blended in, and errors reported in earlier releases are edited out.

Land comes from SRTM15+ outside the polar regions, from the BedMachine models in Greenland and Antarctica, and from other terrain models in the Arctic. The 2026 grid was released on 27 April 2026, as global files or eight 90-degree tiles in netCDF, GeoTIFF and ASCII, with a second version showing the rock beneath the ice of Greenland and Antarctica.

The grid is in the public domain: free, it may be copied, adapted and used commercially, provided the source is acknowledged and no endorsement by GEBCO, the IHO or the IOC is implied. It is supplied "as is" and is not to be used for navigation or any purpose relating to safety at sea.

At sea: the big picture, not the passage

GEBCO is good at the large shapes: the edge of the continental shelf, ridges, seamounts and trenches, the depth changes where swell steepens and currents bend. But in the open ocean most cells are predicted, not measured, and only the TID grid tells which: a depth flagged 40 may be hundreds of metres off over rough seabed, and a seamount or an isolated shoal narrower than a few kilometres may simply be missing.

Near the coast it is coarse. At about 460 m a cell can straddle a channel and the bank beside it, and a whole reef or a narrow pass may fit in one cell; where the grid is not constrained by soundings, the coastal depths are interpolated between land and the nearest data.

Its depths are assumed to refer to mean sea level, though some shallow-water data use other datums: at low tide there is less water than shown, by up to half the tidal range, unlike a chart whose depths refer to the lowest tide. GEBCO itself states that the grid is not for navigation; official charts remain the reference.

In Saily WR

Saily WR uses the GEBCO grid in its 2026 edition, at about 460 m (15 arc-seconds), over the whole globe. When you ask for a minimum depth, GEBCO gives the depth everywhere EMODnet, finer, does not cover: the open oceans and every coast outside Europe's seas and the Caribbean area.

On those coasts, given the cell size and the mean sea level datum, set the minimum depth with a generous margin and keep the coastline margin too: GEBCO keeps a route off the shelves and banks it knows, not off every rock. Where the TID grid would show a coast as interpolated rather than surveyed, the official chart is the only guide to its dangers.