The Repository Abyss Report: Into the Dark

The Lake at the Bottom of the Sea

August 28, 2026 · Runtime 23:33 · Watch on YouTube

At 1,624 metres in the Gulf of Mexico there is a lake, with a shoreline, a beach and a rim of drowned mussels. In 2017 a NOAA robot parked over it for twenty-five minutes with two cameras running, and the log the watch typed that afternoon ends in a question mark.

Chapters

Transcript

A surface with a shoreline, at 1,624 m 0:00

One thousand six hundred and twenty-four metres down, in the Gulf of Mexico, the lights of a robot come over a rise and stop on a surface. Not a rock. Not sediment. A surface, with a far edge and a near edge, and a line where it meets the ground. There is water above it and there is something else below it, and the boundary between them is as sharp as the edge of a swimming pool.

Twenty-five minutes, and a question mark 0:25

What the vehicle has found is a lake. An actual lake, lying at the bottom of the ocean, with a shoreline and a beach and ripples that move when you push them. This recording is twenty-five minutes of a room full of scientists looking straight at it, on two cameras, with the best deep-submergence vehicle in the American fleet parked over the top of it. And in twenty-five minutes they could not agree on what to call it. The log they typed that afternoon is still in the catalogue, and it ends in a question mark. That question mark is what this episode is about.

Green Canyon 939, off Louisiana 1:01

The dive is EX seventeen-eleven, dive ten. Twelfth of December, twenty seventeen. The site is a patch of the Louisiana continental slope called Green Canyon nine three nine, about two hundred kilometres out from the coast, which the science team on board had nicknamed the Saint Tammany Basin. The ship is the NOAA vessel Okeanos Explorer. The vehicle is Deep Discoverer, with a second camera platform, Seirios, hanging in the water above it and looking down.

They went for the gas, not the lake 1:33

They did not go there for a lake. Nobody knew there was one. The dive summary says why they went: during multibeam mapping the ship had picked up several methane bubble plumes and possible hard ground about five kilometres south of the area they had originally planned to survey. Gas coming out of the seafloor is a reliable sign that something is living on it. So they moved the dive to the gas.

Seven hours fifty-two minutes 1:58

The numbers of the dive are all published, and they are worth having in front of you, because everything that follows happens inside them. In the water at fourteen thirty-two UTC. On the bottom at fifteen thirty. Off the bottom at twenty-one thirty-five, back at the surface at twenty-two twenty-four. Seven hours and fifty-two minutes of dive, six hours and five minutes of it on the seafloor. Maximum depth, one thousand six hundred and thirty-five metres.

The column on the way down 2:29

On the way down the vehicle carries a CTD, which is the standard instrument of this science: it measures conductivity, temperature and depth, and from conductivity it computes salinity. That descent is published as a file almost nobody opens. One thousand five hundred and seventy-eight one-metre bins, from four metres down to fifteen eighty-one. It contains the ordinary shape of the Gulf: warm and salty at the top, an oxygen minimum in the middle, and cold, quiet, well-oxygenated water at the bottom.

The oxygen minimum, at 433 metres 3:03

The oxygen minimum in that file sits at four hundred and thirty-three metres, at one hundred and five point five micromoles per kilogram. At the surface it is one ninety-six. On the bottom, sixteen hundred metres further down, it is back up to one ninety-six point seven. That is not a textbook diagram. That is one instrument, on one afternoon, on the way to somewhere else. Hold on to the bottom number: one ninety-six point seven. It comes back.

Empty shells and bacterial mat 3:35

The vehicle lands on soft sediment with burrows in it and almost nothing visible alive. That is the normal deep Gulf floor. Driving toward the gas targets, the first thing that changes is the ground: scattered mussel shells, empty, lying on the mud. Then patches of pale bacterial mat, which is what sulphide-eating bacteria look like when there are enough of them to see. The shells and the mats mean the same thing. Something is leaking up through the sediment here.

"Intensified bubble streams in the distance" 4:06

In the control room they can hear the ship as well as see the seafloor. Somewhere above them the expedition coordinator is watching the acoustic sounder, and he walks in with the thing everybody wants to hear.

Into the first depression 4:34

The terrain starts to roll. Instead of flat mud there are mounds, and between the mounds there are depressions, and the chemosynthetic communities are concentrated down inside them. The vehicle drops into the first depression. And the floor of the first depression is not a floor.

"Is that a brine pool down there?" 4:52

Watch what happens in the room. Not the discovery. The wording. Somebody says they have found what they came for. And then, immediately, somebody else asks a question with a hedge already built into it.

What the camera was looking at 5:18

Here is what they were looking at. A flat, pale, slightly luminous surface, filling the bottom of the depression. A rim of shells along its edge, the way a tideline sits on a beach. And above it, the ordinary black water the vehicle had just flown through. The pale surface is denser than the water. It sits in the hollow the way water sits in a bowl, and it has a top.

"A lake underneath the ocean" 5:43

A minute later somebody on the science watch talks the room through the picture, and the sentence they land on is the sentence this whole episode is built around. Listen to the second voice as well.

Somebody else is typing 6:16

While that is happening, somebody else is typing. Every video segment the Okeanos Explorer records gets subject terms attached to it by the watch, live, and those terms are what the national archive indexes years later. They are the finding aid. They are how anybody ever gets back to this footage. Three of them, on this dive, are the story.

16:55, 17:00, 17:20 6:39

Sixteen fifty-five: brine pool. No punctuation. Seventeen hundred: brine pool, slash, river, question mark. Seventeen twenty: brine pool end, question mark. Twenty-five minutes apart. Between the first and the last, the confidence went down, not up. The longer they looked at it, the less sure they were what they were looking at. And the ship's own written dive summary, filed afterwards, with time to think, says the same thing: a brine pool slash river. The slash is in the document.

The third kind of label 7:16

It would be easy to read that as a failure. It is the opposite. This channel has now spent two episodes on the cost of labels that sounded confident. A dive that wrote down a guess, and the guess answered the question that would have found the object. A dive that got the name exactly right, and put it in a field the search does not read. Here is the third case, and it is the only one that ages well. They did not know, and they wrote down that they did not know.

Why the ocean has lakes in it 7:47

So what is it, and why does the ocean have lakes in it. The answer starts about a hundred and sixty-five million years ago, and the scientist on the microphone explains it live, on the open feed, while the pilots are still lining the vehicle up.

The Callovian, and the Louann Salt 8:12

In the Middle Jurassic, in the Callovian, the Gulf of Mexico was a restricted basin: seawater came in, and evaporation took it out faster than circulation could replace it. What that leaves behind is salt. The deposit has a name, the Louann Salt, and it is not a layer in the sense of a stripe on a diagram.

Eight kilometres of salt 8:33

It is kilometres thick. Not metres. Kilometres. Under the northern Gulf the Louann reaches a thickness that is easier to state than to picture, and the figure the scientist gives on the open feed, off the top of her head, while a robot hovers over a lake made out of it, is this one.

Salt does not compact 9:01

Then the Gulf reopened, filled with seawater, and hundreds of metres of sediment piled on top of the salt. And salt, unlike almost everything else down there, does not compact. Buried under heavy sediment it stays about the same density while the mud above it gets denser, and eventually the arrangement is unstable: the salt starts to move. It flows, sideways and upward, and pushes domes and walls up through the sediment above it. Geologists call the whole business salt tectonics. It is why the Gulf has oil, and it is why the Gulf has lakes.

Three to eight times the salinity 9:37

Where one of those salt bodies gets close enough to the seafloor, seawater percolating down through the sediment reaches it and starts dissolving it. The result is not seawater any more. It is brine: the same water with several times the dissolved salt in it. And the number the scientist gives on the feed is the one that matters, because it is the reason the lake exists at all.

It runs downhill, like water 10:16

Salt water is heavier than fresh water. Brine that is several times saltier than seawater is heavier than seawater, by a lot. So it does not disperse upward and it does not drift away. It runs downhill along the seafloor, exactly like water on land, and it collects in whatever hollow it reaches.

Diffusion is the only mixing there is 10:48

That is the whole mechanism. A lake at the bottom of the sea is not a different kind of physics. It is the same physics, with a different fluid playing the part of the air. The part that is genuinely strange is that it stays. There is nothing solid separating the brine from the seawater on top of it. Two fluids in direct contact, for as long as the seep keeps feeding the pool, and they mix so slowly that the boundary stays sharp enough for a camera to focus on. There is almost no turbulence at sixteen hundred metres in a hollow with no current in it, so mixing is left to molecular diffusion, and molecular diffusion is slow.

Eighty-two one-metre bins 11:28

That is the claim. And on this dive, without anybody planning it, the vehicle measured it. The CTD on Deep Discoverer does not stop logging when the vehicle reaches the seafloor. It keeps going for the entire bottom time, and the processed file for the bottom section of this dive contains eighty-two one-metre bins between fifteen seventy-nine and sixteen thirty-five metres. Those eighty-two bins include the twenty-five minutes when the vehicle was hovering directly over the pool.

34.9681 to 34.9783 11:59

Across all eighty-two bins, salinity runs from thirty-four point nine six eight one to thirty-four point nine seven eight three practical salinity units. That is a total spread of one hundredth of a unit. Temperature moves by sixteen thousandths of a degree. Oxygen stays between one ninety-six point seven and one ninety-eight point nine. At sixteen twenty-four metres exactly, which is the depth of the surface of the lake, the bins read thirty-four point nine seven, with oxygen at one ninety-seven and one ninety-eight.

An instrument that did not notice a lake 12:34

Read that again with the geometry in mind. One metre above a fluid that is several times saltier than the sea and has no oxygen in it at all, the vehicle's own conductivity sensor recorded ordinary, fully oxygenated seawater, varying by one part in three thousand. This is the strongest available evidence that the pool does not mix, and it is not an argument. It is an instrument failing to notice a lake it was flying over. Nobody wrote it up. It has been sitting in a public file since twenty twenty.

The rim of empty shells 13:08

Now look at the edge. There is a rim of mussel shells around the pool, dense, and most of them are empty. Further out, on the ordinary sediment, the same species is alive: dark, closed, wedged into the mud with the hinge buried. The ship's dive summary describes the boundary in one clause, and it is not neutral language. Live Bathymodiolus brooksi, and shells of many mussels that appeared to have drowned in the brine.

Two symbionts, one mussel 13:37

Drowned is the right word, and it is worth being precise about why. These mussels are not filter feeders in the ordinary sense. They carry bacteria inside their gills and they live on what the bacteria make. In Gulf species the arrangement is double: one symbiont oxidises methane, the other oxidises reduced sulphur compounds. So the animal needs the seep. It needs the chemistry that leaks out of the same salt system that fills the pool. It has to live close to the brine.

Fed by the lake, killed by the lake 14:09

But it also needs oxygen, and the brine has none. So the mussel bed sits in a band: close enough to the pool to be fed, far enough from it to breathe. The rim of empty shells is what that band looks like when the pool rises, or spreads, or when an animal settles on the wrong side of the line. The lake feeds them and the lake kills them, and the boundary between those two things is a few centimetres wide.

"Extensive mussel bed, all look alive" 14:36

Drive on past the pool and the story inverts. Further along the same seep field the vehicle comes over a mussel bed that is enormous and, in the watch's own words typed at the time, all look alive. Adults, juveniles, and a scatter of old shells among them. The dive summary reads that size mixture as evidence of multiple recruitment events, which is a careful way of saying this bed has been here long enough to have been colonised more than once. There are king crabs on it, feeding, and scale worms, and shrimp, and a congregation of sea cucumbers a few metres away.

Two fish possibly never filmed before 15:14

And the six hours were not only chemistry. The fish list from this one dive runs to more than a dozen species, and you can hear the room working through them in real time, spelling names out letter by letter for whoever is typing. Two of them, according to the ship's own summary, may not have been imaged underwater before at all: a cutthroat eel, Haptenchelys texis, and a deep-water skate, Rajella bathyphila. Filed in the same document, in the same tone, as everything else. That is what an exploration dive produces: a long, flat list, in which the extraordinary items are not typographically distinguished from the ordinary ones.

1977: chemosynthesis, with heat 15:56

How long have we known any of this. Less time than you would think, and the sequence is short enough to say in one breath. Nineteen seventy-seven, Galapagos Rift: the submersible Alvin finds hydrothermal vents and, around them, an entire community that does not run on sunlight. That is where chemosynthesis stops being theory. But those vents are hot, and heat was assumed to be part of the deal.

1984 and 1985: the heat goes away 16:24

Nineteen eighty-four. Paull and colleagues, in Science, at the base of the Florida Escarpment: the same kind of community, the same kinds of animals, and no heat at all. Cold, sulphide-rich brine seeping out of the sediment. Nineteen eighty-five. Kennicutt and colleagues, in Nature, on this same Louisiana slope: vent-type animals in a region of hydrocarbon seepage. Two papers, one year apart, and between them the requirement for volcanic heat is gone.

1990: 121.35 practical salinity units 16:56

Nineteen ninety. MacDonald and colleagues, in Science, describe a brine pool in the northern Gulf at six hundred and fifty metres, ringed by chemosynthetic mussels, and they measure the fluid. One hundred and twenty-one point three five practical salinity units, against a seawater background of about thirty-five. That is the measured number for a Gulf brine pool, and it is from nineteen ninety. Nobody has measured this one.

A kilometre that never happened 17:25

Which brings up the question of what the vehicle was actually doing for those twenty-five minutes, and a measurement I nearly got wrong. The dive publishes a position file at one hertz: twenty-seven thousand seven hundred and forty-three fixes, latitude, longitude and depth, one every second, for the whole dive. Add up the distance between consecutive fixes across the pool window and you get one thousand and fifteen metres. A kilometre. That number is wrong.

1,015 metres, or fifteen 17:58

It is wrong because the position is acoustic. The ship works out where the vehicle is by timing sound through sixteen hundred metres of water, and every one of those fixes carries a couple of metres of scatter. Add fifteen hundred of them together and you are summing noise, not travel. Average the track over ten seconds and the kilometre becomes six hundred and twenty-four metres. Over thirty seconds, two hundred and eighteen. Over a minute, one hundred and eight. Straight line from where it started to where it ended: fifteen metres.

If it collapses when you smooth it 18:32

So the number that goes on screen is fifteen. In twenty-five minutes the vehicle finished fifteen metres from where it began. It was not surveying. It was parked, looking at the thing. And there is a rule in that, which costs nothing and would have saved me publishing a wrong kilometre: summed distance along a one-hertz acoustic track is not distance travelled, it is integrated noise. Test how the number responds to smoothing before you say it out loud. If it collapses when you smooth it, it was never a measurement.

Pool, or river 19:07

Now, finally, the question they could not answer. Pool, or river. It is not a quibble. A pool is a body of brine sitting in a closed hollow, filled from below, essentially static. A river is brine flowing downhill along the seafloor, through the hollow and out the other side, going somewhere. Same fluid, same appearance, completely different thing: one is a container, the other is a current.

Why you cannot tell from above 19:34

And from a vehicle hovering over it, you cannot tell. A flat brine surface looks flat whether or not the brine underneath is moving. There is no wind to ripple it, no bank to cut, no floating debris to carry. To decide, you would have to follow the low ground far enough to find out whether the brine ends in a rim or keeps running downhill, or you would need a current meter in the fluid itself. They had six hours of bottom time and a whole seep field to survey, and they had already spent twenty-five minutes on this one spot.

"Brine pool end?" 20:08

So at seventeen twenty they typed brine pool end, question mark, and drove on. The second question mark is not doubt about whether the pool ended. It is doubt about whether the thing that ended was a pool. They were still not sure, twenty-five minutes and two cameras later, and they said so in the only place the archive was ever going to read.

Who reads a record like that 20:31

It is worth knowing who reads a record like that, because it is not only people like me. The dive summary is explicit about why this site was on the schedule at all. It sits next to a proposed expansion of the Flower Garden Banks National Marine Sanctuary, and next to a Habitat Area of Particular Concern proposed by the Gulf of Mexico Fishery Management Council. The dive was moved outside the proposed boundaries because previous exploration and a pipeline left nowhere good to land inside them. So the baseline this dive collected — what lives here, how much of it, how it is distributed — goes into the file that management decisions get made from.

A closed item and an open one 21:13

Which is the practical reason the punctuation matters. A record that says brine pool, flatly, is a record that a later reader will treat as a mapped feature: a thing of known type, in a known place, that somebody has already characterised. A record that says brine pool slash river, question mark, is an open item. It tells the next expedition that there is something here still worth the six hours.

The only entry that cannot mislead you 21:40

That is the argument of this episode, and it runs against the grain of how archives usually get judged. A question mark in a metadata field looks like a defect. It is not tidy, it will never match a search cleanly, and no catalogue is proud of it. But it is the only kind of entry that cannot mislead you. It carries the observation and the confidence separately, which is the entire job. The wrong label was confident. The right name in the wrong field was confident. This one told the truth about itself.

Nine years later, the punctuation is there 22:15

Nine years later the file is still there, and so is the punctuation. Anybody who goes looking for brine pools in the NOAA catalogue will find this dive, and the record they find will tell them, in three words and a question mark, exactly how much the people who were there were prepared to claim. That is not a gap in the archive. It is the most carefully calibrated thing in it.

Two instruments, one recording 22:39

Twenty-five minutes, sixteen hundred and twenty-four metres, a shoreline, and a room full of people who could see it perfectly well and still would not name it. The instrument on the vehicle flew over the lake and recorded nothing. The instrument in the room recorded the only thing that mattered, which was uncertainty, and it recorded it accurately.

Next: methane hydrate, and something living on it 23:01

Next time: the same dive, an hour and forty minutes later. The vehicle finds a gap under an overhang of carbonate rock, and inside the gap there is something white and solid that has no business being solid at that temperature. Methane hydrate: gas locked inside a cage of water ice, held together by the pressure of sixteen hundred metres of ocean. The subject term the watch typed for that segment ends in an exclamation mark. And there is an animal living on it.

Description and sources

On 12 December 2017, dive 10 of NOAA's EX1711 went down on the Louisiana slope at Green Canyon 939, chasing methane plumes the ship had picked up on multibeam. At the bottom of the first depression was a brine pool: seawater that has dissolved Jurassic salt, become several times denser than the ocean around it, and settled into the hollow as a fluid with a top. This episode is about the twenty-five minutes the control room spent looking at it, unable to decide whether it was a pool or a river - and saying so in the subject terms that are still the archive's finding aid. It also uses three measurements computed for this video from files nobody opens: the vehicle's own CTD, a metre above the lake, reading ordinary seawater; the descent profile, with the Gulf's oxygen minimum at 433 metres; and the one-hertz position track, which says the vehicle travelled a kilometre and is wrong by a factor of sixty-seven.

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

NOAA Ocean Exploration. ROV Dive Summary, EX1711 Dive 10, 12 December 2017. Green Canyon 939 / St. Tammany Basin; max depth 1,635.0 m; "a brine pool/river with live Bathymodiolus brooksi and shells of many mussels that appeared to have drowned in the brine". https://oer.hpc.msstate.edu/okeanos/ex1711/
NOAA Ocean Exploration. EX1711 dive 10 ancillary and ROV CTD packages: RovTrack1Hz.csv (27,743 fixes) and ROVCTD_A/B_bin_avg.cnv (1,578 and 82 one-metre bins)., from the package page above.
NCEI granule metadata, EX1711 DIVE10. The subject terms BRINE POOL, BRINE POOL/RIVER? and BRINE POOL END? are published here. https://www.ncei.noaa.gov/metadata/granule/geoportal/opensearch
Paull, Hecker, Commeau et al. Science 226(4677): 965-967, 1984. https://doi.org/10.1126/science.226.4677.965
Kennicutt, Brooks, Bidigare et al. Nature 317: 351-353, 1985. https://doi.org/10.1038/317351a0
MacDonald, Reilly, Guinasso, Brooks & Bryant. Science 248(4959): 1096-1099, 1990. The 121.35 PSU measurement. https://doi.org/10.1126/science.248.4959.1096
Duperron et al. Environmental Microbiology 9(6): 1423-1438, 2007. https://doi.org/10.1111/j.1462-2920.2007.01259.x

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

About the material. Every moving shot is NOAA Ocean Exploration video from EX1711 dive 10, 12 December 2017, public domain, credited on screen. The seven passages of control-room audio are the dive's own soundtrack, unedited. Every diagram is drawn in code from the sources, and the oxygen, salinity and seafloor profiles are plotted point by point from NOAA's own published CTD and position files. There is no generated image in this video at all.

The 121.35 PSU figure is a 1990 measurement of Brine Pool NR-1 at 650 m, not of the pool shown here.

#BrinePool #GulfOfMexico #NOAA #DeepSea #OceanExploration