The Repository Abyss Report: Into the Dark

Seafloor Maps: How to Tell Which Parts Were Measured

October 22, 2026 · Runtime 13:04 · Watch on YouTube

Every seafloor map paints a depth in every cell, measured or not. GEBCO publishes a second grid, the type identifier, that says where each of its 3.7 billion depths came from. We read it cell by cell, and used it to correct our own last episode.

Chapters

  1. 0:00 The map the ship opened
  2. 0:42 A correction to episode 19
  3. 1:26 1. One number per cell
  4. 2:03 The whole dive inside one square
  5. 3:08 2. A measured map shows its holes
  6. 3:44 3. The type grid
  7. 4:36 Vailulu'u on the type grid
  8. 5:17 4. How a satellite predicts depth
  9. 6:02 The seamounts that keep appearing
  10. 6:47 5. Smooth is a guess
  11. 7:49 Two charts, one seamount
  12. 8:31 6. The whole planet, counted
  13. 9:12 What 'mapped' means
  14. 9:50 The six readings, turned on us
  15. 11:02 The mistake, exactly
  16. 11:59 Six readings of one map
  17. 12:37 Next: the blue goo

Transcript

The map the ship opened 0:00

Before every dive on this channel, somebody on the ship opens a map. This is the one the Okeanos Explorer opened on the morning of the twenty fourth of February, twenty seventeen, over a volcano off American Samoa. Listen to how the science lead reads it.

A correction to episode 19 0:42

In our last episode, we read one map of the far South Pacific. We told you the ridge at fifty four south, one hundred and forty west was drawn from orbit, and that what it needed was a ship's sonar. We had checked one map. This time we read a second one, and it disagrees. On this grid, ship sonar passes about five kilometres from that point.

So this is an episode about a skill. How to look at any map of the seafloor and tell which parts of it somebody actually measured. Six readings, all on public data you can download today. At the end, we turn all six on our own mistake, and find out exactly what it was.

1. One number per cell 1:26

Reading one. Every colour on a seafloor map is one depth, painted over one square cell. The map does not know what is inside the square. It knows one number, and it paints the whole square with it.

The world's reference map is the GEBCO grid, published every year by an international project. Its cells are fifteen arc seconds on a side. Forty three thousand two hundred rows, eighty six thousand four hundred columns. Three point seven billion cells, and every one of them holds exactly one depth.

The whole dive inside one square 2:03

Here is what that size means on a real dive. Listen to the dive supervisor plan the morning.

A circle a hundred and fifty metres across. At this latitude, one cell of the global grid is about four hundred and fifty metres by four hundred and sixty. The whole morning's survey would fit inside a single square of the world map, with room to spare. Over this crater, the sharpest grid we could download has cells of fifty eight metres, because a ship sat over it with a multibeam sonar.

2. A measured map shows its holes 3:08

Reading two. Look at the edges of the ship's own map. Black specks, black gaps, a ragged border where the colour simply stops. Those are places where no sonar beam came back clean. A measured map shows you its holes.

A global map never does. Every cell has a number, whether anybody measured it or not. The soft, rounded hills you see here look exactly as confident as the sharp ridges. To find out which is which, you need a second map, one that the first map almost never shows you.

3. The type grid 3:44

Reading three. Alongside its depth grid, GEBCO publishes a type grid. Same size, same cells, but instead of a depth, each cell holds a code for where its number came from. Ten is a single beam echo sounder. Eleven is multibeam. Forty is predicted from satellite gravity. Forty one is interpolated by a computer between soundings.

The full list is written inside the file itself, and it has surprises. Code forty six is a depth estimated from an iceberg that ran aground. Code forty seven, from an Argo float that touched bottom. Code forty eight, from data loggers carried by animals. In the twenty twenty six grid, cells with code forty eight add up to about one thousandth of one percent of the ocean.

Vailulu'u on the type grid 4:36

Here is the type grid around Vailulu'u. Orange is any direct measurement. Blue is satellite prediction. The orange arrives in strokes, because that is how sonar is collected: one ship, one line, one swath at a time. In this box, fifty four percent of the cells were measured. Almost all the rest is a prediction.

The crater itself is code eleven, multibeam, in every cell around the dive site. That is not luck. The Okeanos maps a site before it sends the robot down, so its dives land on measured ground. Most of the ocean never gets that visit.

4. How a satellite predicts depth 5:17

Reading four is how the prediction is made. A seamount is extra mass, and its gravity pulls a little more water toward it. The sea surface above it bulges. Satellites measure the height of the sea surface, and a model turns the bumps into depths, anchored to whatever ship soundings exist nearby.

The catch is the scale. Gravity spreads out on its way up through kilometres of water. A twenty eighteen paper puts the average resolution at about eight kilometres, and the best gravity models resolve structures of about six. So a predicted cell is a four hundred and sixty metre square, painted from a guess that cannot see anything smaller than a few kilometres.

The seamounts that keep appearing 6:02

That is why seamounts keep appearing. In twenty fourteen, gravity models started showing seamounts between one and two kilometres tall that older models had missed. A twenty twenty three study found nineteen thousand three hundred and twenty five new ones, on top of a catalogue of twenty four thousand six hundred and forty three.

Ships keep correcting the satellite too. In twenty twenty one, the Okeanos mapped about forty seamounts in the North Atlantic, twenty of them with little or no data before. NOAA's report lists the surprises. A feature the satellite drew as two seamounts was one. And there was a new seamount not apparent in satellite altimetry at all.

5. Smooth is a guess 6:47

Reading five needs no code at all, only your eyes. This is real bathymetry from southeast of Samoa. Some of it is crisp, full of small ridges and scarps. Some of it is soft, like a photograph out of focus. The border between the two runs in straight bands across the image.

Here is where the bands come from. Inside the orange lines, a ship's sonar. Outside, the prediction. The seafloor did not change at the line. The source did. Once you have seen these seams, you will find them on almost every seafloor map you are shown.

We measured the difference in this box, on cells fifty nine metres wide. Where the sonar ran, the median slope is two point seven degrees. Where it did not, zero point eight. Part of that is real roughness the satellite cannot see, and part is the sonar's own noise. Either way, smooth is the signature of a guess.

Two charts, one seamount 7:49

Reading the source is not an academic habit. On the eighth of January, two thousand and five, the submarine USS San Francisco, running at flank speed at five hundred and twenty five feet, hit a seamount about three hundred and sixty miles southeast of Guam. Machinist's Mate Joseph Ashley died of his injuries the next day.

The Navy's investigation found that the seamount did not appear on the chart being used for navigation. Other charts on board, it said, clearly displayed a navigation hazard in the vicinity. Two maps of the same water, two different answers, and the crew was reading only one of them.

6. The whole planet, counted 8:31

Reading six is the whole planet. This is GEBCO's type grid for the entire world, one cell in four hundred drawn. Orange is measured. Blue is satellite prediction. Look at the open Pacific and the Indian Ocean: single ship tracks, crossing blue.

We counted every cell, weighted by its real area, from sixty degrees south to the North Pole. Thirty one percent carries a code for direct measurement. Fifty nine percent is satellite prediction, and about seven percent is interpolated. GEBCO's own official figure for twenty twenty six is twenty eight point seven percent mapped.

What 'mapped' means 9:12

The two numbers are not supposed to match. The official one follows a rule of the Seabed twenty thirty project. The ocean counts as mapped if at least one sounding falls in a grid cell whose size grows with depth: one hundred metres in shallow water, up to eight hundred metres in the deepest trenches. One sounding per square.

So mapped does not mean seen. A cell eight hundred metres wide counts as mapped with a single sounding in it. A camera on a robot sees a few metres. Everything this channel has ever shown you happened inside a handful of those squares.

The six readings, turned on us 9:50

Now, the six readings, turned on our own episode. This is the box around fifty four south, one hundred and forty west, on GEBCO's type grid. The point itself is code forty one, interpolated. But about five kilometres away the codes turn to multibeam, and within fifty kilometres, more than a third of the cells were measured.

The type grid does not name the ships. NOAA's public list of multibeam surveys names two that crossed this box. NOAA Ship Surveyor, December nineteen ninety three. And the research vessel Nathaniel B. Palmer, in the middle of nineteen ninety eight. Both carried multibeam sonars, in the years the Upsweep was being recorded.

Last episode we drew this profile and said its crest, five kilometres northeast of the point, was predicted from satellite measurements. In the grid we used, that was true. In GEBCO's type grid, multibeam cells start less than half a kilometre from that crest. The type grid cannot tell us how deep the measured seafloor is there. That number lives in the ships' own files.

The mistake, exactly 11:02

So here is the mistake, exactly. The number was right: zero ship sonar cells in that box of the GMRT grid. The sentence around it was wrong. Two ships had crossed that box with multibeam sonars, and one of them belonged to NOAA. Nobody went there to look for the Upsweep. But we called it a position nobody had visited, and ships had.

That is the whole skill in one rule. Before you trust a sentence about the seafloor, ask which grid it came from, and then ask that grid for its source layer. One map told us zero. The other told us two ships. Both were telling the truth about themselves.

Which brings us back to Vailulu'u. The map with black holes in it was the honest one. It showed the robot where the ship had listened, and where it had not. Here the robot is going down into the part it had.

Six readings of one map 11:59

Six readings. The colour is one number per cell. A measured map shows its holes. The type grid says who measured each cell. Satellite prediction cannot see anything smaller than a few kilometres. Smooth terrain is a guess. And roughly seven cells in ten of the ocean are something other than a measurement.

None of this makes seafloor maps less useful. It makes them readable. The ship that opened this map knew which parts it could trust, and planned a whole day around that. Now you can do the same with any map you are shown.

Next: the blue goo 12:37

Next time, back on camera. In August twenty twenty two, southwest of St. Croix, the Okeanos filmed a bumpy blue organism several times on one dive. The scientists could not say whether it was a soft coral, a sponge, or a tunicate, and they called it blue goo. They were not allowed to collect it. We will show you why, and what is still unknown.

Description and sources

Six readings of one map: the multibeam map the Okeanos Explorer opened over Vailulu'u, American Samoa, on 24 February 2017, and the GEBCO_2026 type grid. One GEBCO cell at Vailulu'u is about 450 by 460 m; the dive supervisor planned the morning inside a 150 m circle. Counting the whole type grid by area from 60 S to the pole, 30.9% of ocean cells carry a direct-measurement code, 59.1% are predicted from satellite gravity and 6.6% are interpolated (GEBCO's official 2026 figure, by its own rule: 28.7% mapped). Southeast of Samoa, cells under ship sonar have a median slope of 2.7 degrees against 0.8 for predicted cells. Correction to episode 19: the GMRT grid has no ship sonar around 54 S 140 W, but GEBCO_2026 marks multibeam about 5 km from the point, and NCEI lists two multibeam cruises through the box, NOAA Ship Surveyor (1993-94) and R/V Nathaniel B. Palmer (1998).

PRINT-READY, FROM THIS CHANNEL

Read This Dive, Vol. 01 — the method, and five dives read with it

https://therepository.gumroad.com/l/read-this-dive

PRIMARY SOURCES

ARCHIVE PLATES: all 10 are credited on screen in the video and listed in this project's manifest.

Abyss Report reads the deep ocean through the record that resolved it -- instrument, date, position and the primary sources above.

About the material. The moving frames are NOAA Ocean Exploration video from EX1702 dive 09, Vailulu'u, American Samoa, 24-25 February 2017, public domain and credited on screen, including the ship's own mapping screen recorded on the dive briefing. Two cuts play the original control room audio; each window was re-transcribed in isolation twice and used only where both passes matched. The type-grid maps and every count from them are ours, read from the public GEBCO_2026 TID grid (public domain) by scripts/read_map.py; the relief images are the GMRT grid (Ryan et al. 2009, CC BY 4.0). The submarine photographs are U.S. Navy, public domain. Diagrams are drawn in code from the sources listed. Nothing is generated.

#DeepSea #NOAA #OceanMapping #Bathymetry #OceanExploration