How Dubai Built the $12B Impossible Palm Jumeirah
In June of two thousand one, the coastline of Dubai was a shallow, unremarkable stretch of the Persian Gulf. Reclamation began that month. Within about three years the land itself existed, a trunk, a crown of fronds, and a long outer crescent, all of it sitting in open marine water where there had been nothing but seabed. The first residents didn’t move in until two thousand six, and construction on top of the island continued for years after that. But the landmass, the shape you can pick out from orbit, was formed in only about three years.
The Palm Jumeirah was promoted at its completion as the largest artificial island ever built. Whether it still holds that title depends on how you measure and who you ask, but the underlying fact isn’t really about rank. Building it meant moving something like a hundred and ten million cubic meters of sand, the figure comes from Van Oord, the lead contractor, along with about seven million tons of rock, all of it dredged, sprayed, and stacked into a deliberate shape on a seabed that had never been asked to hold anything, let alone a city.
The shape wasn’t decoration. A solid disc of that size would have trapped the water inside it and let it go stagnant, so the design was broken into a central trunk and seventeen radiating fronds, wrapped by an eleven-kilometer crescent that served as a breakwater. The island reached about five kilometers out into the Gulf and added tens of kilometers of new waterfront, eventually more than a thousand beachfront villas along the fronds. The crescent took the force of the Gulf.
What makes the Palm worth understanding isn’t that it’s large. Plenty of things are large. It’s that almost every problem involved in building it is one you’d normally solve by choosing a different place to build. Here, there was no different place. The location was the product. The engineering had to make the sea behave like ground.
The crescent went in first. Before the palm had any shape at all, the eleven-kilometer breakwater was laid down as a rock wall, because the sand island inside it couldn’t be built in open, moving water. The Gulf would have carried the fill away as fast as it was placed. So the order was protection first, then land.
The dredging and reclamation were led by the Dutch marine contractor Van Oord, with the Belgian firm Jan De Nul also credited on the work. The breakwater wasn’t a simple pile of rock. Its core was seabed sand, wrapped in a water-permeable geotextile membrane to keep that sand from washing out, then topped with a layer of roughly one-ton filter stones and, over that, two layers of armor rock up to about six tons a piece, each one set by crane rather than dumped. About seven million tons of rock went into the breakwater, which is essentially all the rock in the island, since the landmass itself is sand. Most of it was quarried inland, from sites in and around the Hajar mountain range.
The land was made by spraying. Dredgers pumped sand up from the seabed and fired it through the air in long arcs onto the target area, a technique called rainbowing, for the shape of the plume. The placement wasn’t approximate. Differential GPS positioned the sand to within about a centimeter of the design, which is how you build a recognizable palm frond out of a material that has no edges of its own.
Then the shape created a problem of its own. A continuous crescent sealed the water between the fronds, and it began to go stagnant. Openings were later cut into the crescent, each roughly a hundred meters across, to get that trapped water moving again.
By now the island existed as a shape. What it didn’t have was the ability to hold weight. Freshly sprayed sand fill is loose and saturated, with very little bearing strength, and under load it can lose that strength suddenly and behave for a moment like a liquid. So probes were driven into the fill and vibrated, often with jets of water, until the grains repacked tighter and denser. One of Nakheel’s own engineers later put it plainly: fill a bucket with tennis balls, shake it, and the level drops as they settle into a steadier arrangement. Across the whole island, vibrocompaction stiffened the fill enough that much of the Palm could be built on improved ground rather than deep piles. The loose sand had become something closer to ground.
Every piece of reclaimed land settles. Sand placed underwater doesn’t arrive as solid ground. It arrives as a loose pile of grains with water in the gaps, and over time, under its own weight and the weight of whatever is built on top, it compresses. The grains pack closer, squeezing out the water between them, and the surface drops. This isn’t a defect. It’s the ordinary behavior of fill, and every reclamation project in the world plans around it.
The engineering question is never whether the ground will settle. It’s how much it settles, and how evenly. Uneven settlement is the dangerous kind. If one part of a building’s footprint drops faster than another, the structure racks and cracks, and that’s what actually damages buildings on made ground, not the settling on its own. Vibrocompaction was aimed straight at that problem. By densifying the fill before anything was built, it forced much of the settlement to happen up front, while the ground was still empty, and it made what remained more uniform across the site. The work ran at more than two hundred thousand points across the island.
The fill made this harder than it sounds. The sand dredged from the Gulf was largely carbonate, shell and coral debris rather than quartz, and carbonate sand is softer and more crushable than ordinary beach sand, which makes its long-term settlement harder to predict. That’s part of why the compaction was tested and controlled so carefully, grid point by grid point, before construction began.
Get The Build Report’s stories in your inbox
Join Medium for free to get updates from this writer.
The target that came out of all this was modest. By the developer’s own account, the settlement expected from the compacted fill was on the order of twenty-five millimeters, about an inch, over a hundred years: small enough, and even enough, that the structures above wouldn’t feel it. That’s a stated design expectation rather than an independently measured result, but it sets the scale the island was built to. Buildings and infrastructure were designed with tolerance for that slow movement rather than against it. Nothing in the plan assumed the island would stay perfectly still, only that it would settle slowly and by a small, predictable amount.
Which is the context the word sinking tends to erase.
Whether it’s actually sinking is where this goes next, and we’ll get to it. If watching the impossible get built is your kind of thing, that’s what this channel does, and subscribing is what keeps the next one from getting lost.
In two thousand nine, the reports arrived. The Palm Jumeirah, they said, was sinking, by an average of about five millimeters a year. The figure came from Fugro NPA, a firm that had studied the island’s surface between two thousand three and two thousand eight. Its InSAR surveying project manager, Adam Thomas, said the company stood by its measurements: they were seeing, in his words, a number of locations where the ground was moving downwards. He raised a longer worry as well, that settling ground and rising seas together could, in time, pose a flood risk.
The source wasn’t a careless one. Fugro NPA’s specialty was interferometric synthetic aperture radar, InSAR, which reads ground motion from satellite radar and, under good conditions, resolves it at the millimeter scale. Monitoring the slow settlement of engineered structures was the exact thing the firm existed to do.
Nakheel’s response was flat and total. The claims were wholly inaccurate. The Palm was intact, and if the ground were dropping even five millimeters, the developer argued, you’d see it in cracked masonry and broken glass, and there was none. Then it added a technical point. A five-millimeter drop couldn’t have been detected from orbit at all, Nakheel said, because NASA’s laser-altimeter satellites were accurate only to about fifty millimeters, ten times the figure in question.
That objection named the wrong instrument. Laser altimetry wasn’t how the study had measured anything. Fugro’s method was radar interferometry, and resolving millimeter-scale motion is exactly what it does, so the precision limit Nakheel raised belonged to a technique no one had used. The argument from intact buildings has its own limit, and it’s the distinction from a moment ago: uniform settlement doesn’t crack anything. Ground that drops evenly can drop a fair amount and leave the villas above it untouched. Intact masonry is strong evidence against uneven, damaging settlement. It isn’t proof that nothing has moved.
Two things do pull the other way. The study window, two thousand three to two thousand eight, is the early life of the fill, when the settlement of any reclaimed ground is fastest before it slows, so a few millimeters a year in those first years isn’t a rate that runs forever. And Fugro, while standing by the data, spoke of a natural process rather than an island on its way under.
Twenty years on, the Palm is fully built and shows no sign of structural failure, and no equally public survey has since pinned the figure down. What’s left, once the noise clears, is smaller than the headline and larger than the denial: reclaimed land settling slowly, about as reclaimed land does, at a scale you can read with radar but not with your eyes.
The word sinking was carrying a lot in those headlines. Most of what it carried was the sea.
There’s a question underneath the sinking argument, and it’s the one Adam Thomas actually pointed at. What threatens a low stretch of coast isn’t how much the land settles, and not how much the sea rises. It’s the two together. A structure at the shoreline feels only the position of the water relative to the land, and settlement and sea-level rise push that number the same way. Even a little subsidence adds to whatever the sea is doing on its own.
And the sea is doing something. Global sea level has been rising by a few millimeters a year, and the rate has been climbing. In a single year that’s almost nothing, set against a crescent whose crest sits a few meters above the water. Across the design life of a coastal structure, measured in decades, it stops being nothing. The engineers looking at the Palm’s future work with a projected rise on the order of forty centimeters over the next fifty years under a high-emissions scenario.
In two thousand twenty-three, a team of university engineers took the crescent’s outer armor, the rubble revetment on its seaward face, and tested it against those conditions. Under the higher-end projections, they found, the armor would no longer meet its criteria for wave stability or for water washing over the top, though the toe that anchors the structure to the seabed still would. They set out three ways to fix it: an added armor layer laid at a gentler slope; a berm across the front; or, further out, a submerged breakwater to break the waves before they reach the island. These were preliminary designs, not construction plans. The point was that the fix is known, and unremarkable.
So the honest version of the question isn’t whether the Palm Jumeirah is sinking into the sea. It’s whether the sea is rising to meet it, and how much the crescent will have to be raised and armored to hold its line over the next hundred years. That work has already been started, on paper, the way coastal engineering usually begins, one analysis at a time.
