Data & Confidence | ArgoRoute

🗺️

Horizontal Accuracy: OpenStreetMap

For razor-sharp boundary detection between land and water, we use OpenStreetMap vector data: land polygons for the coastline and 3.3 million seamark features for charted rocks, wrecks, reefs and obstructions. OSM is the absolute judge of "water or land", and a charted hazard closes its cell outright.

🌊

Vertical Depth: GEBCO 2026

For the vertical dimension we use the GEBCO 2026 grid (General Bathymetric Chart of the Oceans), the global 15 arc-second bathymetry supported by the Nippon Foundation-GEBCO Seabed 2030 Project. A pixel is about 464 m across at the equator, so a depth is an area average, not a sounding — which is why every cell also records where its depth came from.

How Much of the Seafloor Is Measured?

GEBCO publishes a Type Identifier (TID) for every cell, saying whether a ship with a sounder was ever there or the depth was predicted from satellite gravity. We read it for all of them.

28.7 %

of the ocean floor measured (Seabed 2030, June 2026)

12.6 %

measured in the 0–20 m band, where draught matters most

24.1 %

of that shallow band is depth estimated from water colour by satellite

~34 %

projected by 2030 on the current pace of 1.3 points a year

Busy Water Is Not Surveyed Water

Strait or approach	Depth from direct sounding
Dover Strait	98.9 %
Strait of Gibraltar	97.9 %
Panama approaches	40.4 %
Suez	29.2 %
Cape of Good Hope	20.4 %
Malacca Strait	5.9 %

Share of graph cells whose GEBCO source is a vessel's sounder. Malacca, one of the busiest straits on earth, is 91 % predicted. A low share means GEBCO holds no measured source for that cell — a national hydrographic office may still chart it.

The Trend

Year	Seafloor mapped
2017	6 %
2020	19 %
2024	26.1 %
2025	27.3 %
2026	28.7 %

The early jump was archive ingestion; recent gains are about 1.3 points a year and slowing. Provenance warnings are a permanent feature of the product, not a bridge to a better dataset.

ArgoRoute isobaths across the Strait of Gibraltar between Tarifa and Ceuta, from the shallow shelf to the 500 m channel — The Strait of Gibraltar, Tarifa to Ceuta: 97.9 % of these cells rest on a ship's sounder. Isobaths from 5 m to 500 m, for a 2 m draught; basemap © OpenStreetMap, Protomaps.

Every Depth Carries Its Origin

One byte per hexagon says where the number came from. The route is graded by its weakest link, never by an average — a ship runs aground at one point, not on 98 % of the route.

ArgoRoute hexagons around Split, Croatia, coloured by the origin of their depth: green for track-observed, blue for GEBCO, amber for samples that straddle the shore; cells whose depth did not measure the water are outlined with an amber dash — The water around Split coloured by origin. Green cells carry a depth a vessel's sounder actually recorded; amber ones were sampled across a shoreline and read deeper than the water beside them. Dashed outlines mark cells where nothing measured the water.

Origin	What it means for the skipper
surveyed	A higher-resolution hydrographic survey measured this water.
track observed	A vessel's sounder passed here. It proves a hull went through — not which hull, so it settles the passage, never the draught.
baseline	The GEBCO grid, sampled clear of any shoreline, but not sounded: predicted from satellite gravity or interpolated between distant soundings. Check the chart.
measured	An electronic chart sounding or lidar in GEBCO's sources. A real depth, but averaged over a 464 m pixel, so a small shoal can hide inside it.
baseline, straddling shore	A 464 m sample that caught a cliff or a beach. Reads deeper than the water beside it. Sound your own depth.
canal published	The canal authority's figure, not a survey. Check the current notice to mariners.
river channel	The port authority's dredged-fairway depth. The channel moves with the season; follow the buoyage.
bridged, inferred	The raster read land; the coastline shows water. Depth is taken from either side and no survey confirms it. Transit on sonar, or not at all.
datum zero	The survey reads a shoreline or a drying flat here. Do not rely on the stated depth.

The "Prudent Mariner" Confidence Limit

While ArgoRoute computes the mathematical optimum, every route comes with a data_confidence block: the weakest origin on the route, whether every depth measured the water the hull would float in, and one sentence a navigator can act on.

Our depths are not on the chart's datum. GEBCO refers to mean sea level; nautical charts use the lowest astronomical tide. The gap is roughly half the local tidal range — about five metres in the Bristol Channel, nothing in the Baltic — and it runs the unsafe way: our figure reads deeper than the chart, most where the tide is biggest. The API says so in every response, and we do not claim our depths are comparable to a chart's soundings.

Because GEBCO averages depth across 464 m cells, micro-reefs, isolated pinnacles and sudden depth spikes may not be represented. Tides, currents, weather, traffic separation schemes and seasonal ice are not modelled. 41,530 small bodies of water remain unreachable in the current graph, mostly behind barriers the repair does not cross.

Disclaimer: Under admiralty law, the Master of the vessel bears non-delegable responsibility for safety. ArgoRoute is a passage-planning tool, never an autopilot and not an ECDIS replacement. In pilotage waters, use the chart, the eyes and the sounder.