A storm cuts a channel through a barrier island and the sea starts filling it back in the same afternoon. Sometimes the channel wins for a decade. What decides it is not the storm, but how much water sits behind the gap and has to leave twice a day.
On 29 October 2012, Hurricane Sandy pushed a channel clean across Fire Island, off the south shore of Long Island. It opened in a low, dune-less stretch of federal wilderness at a place the maps already called Old Inlet, where a channel had stood from 1763 until the 1820s and then quietly gone. Four days later a satellite measured the new one at 72 metres across.
Now the awkward part. Sand on that coast moves constantly westward, pushed by waves arriving at an angle; longshore drift there is a daily condition, not an event. Every day for the next ten years sand was delivered to the mouth of that channel — and every day the channel was still open.
The tide is what keeps it open. Behind a barrier island sits a bay, holding a volume of water that must get in and out through whatever gaps exist. The difference between that bay at high water and at low water is the tidal prism, and it is the whole engine: twice a day it is pushed through the throat, twice a day it comes back out, and going out it scours — lifting the sand the drift delivered sideways and dumping it in a fan offshore. The gap lives on the gap between two rates.
Which inverts last issue’s foredune exactly: same coast, same sand, opposite arithmetic — the ridge is sand that arrived and stayed, the inlet is sand that arrived and did not. Nor is it No. 90’s turbidity maximum, a standing pile pinned between two opposing currents; this is a standing absence held by one current that keeps returning. And where No. 62’s sandpile finds its angle from grains alone, this shape needs a flow rate on schedule.
The geometry is startlingly lawful. In 1931, and again in 1969, Morrough O’Brien noticed that an inlet’s narrowest cross-section and the tidal prism behind it track each other across wildly unalike coasts — area rising close to proportionally with prism. Jarrett re-ran it for the Army Corps in 1976 across 108 American inlets and got an exponent between about 0.85 and 1.1, depending on the subset. It is a regularity, not a law; the scatter is real and the coefficients are argued over. But an inlet is as big as its prism can afford.
What happens when it cannot afford itself was answered in 1940, in two pages of Shore & Beach, by Francis Escoffier. Plot peak current in the throat against the size of the throat. On the left the channel is so tight it chokes its own bay, and widening it speeds the current up, because the prism it admits grows faster than the opening does. Past a peak that reverses: the bay fills freely, the prism is fixed, and each extra square metre spreads the same water thinner. Now rule one line across it — the speed below which sand stops moving, which Escoffier took as roughly a metre a second.
The line crosses the curve twice, and the crossings are not the same kind of thing. The right-hand one is stable: too wide, the current slackens and sand settles until the throat shrinks back; too narrow, it sharpens and scours until the throat grows. The left-hand one is a trapdoor. Below that width, shrinking makes the current slower, which makes it shrink faster. There is no route back up.
Fire Island’s channel spent ten years on the safe side of that curve. It widened to 190 metres by 2013, 314 by 2014, 440 by 2016 — then wandered: 368, 371, 289, 308, 272, holding that range for six years while the drift arrived without pause. On 20 October 2022 the same satellite programme measured 22 metres. Almost ten years to the day after Sandy cut it, it sealed.
Watch what left with it. All those years the channel had been building a fan of sand inside the bay — about 2.75 million square metres of it by 2020. Once the flow stopped, nothing maintained the fan either, and it began coming apart, down to 1.29 million by the end of 2023. The deposit was no more permanent than the gap: it outlived the current by about a year.
People do override this, and the bill is legible from orbit. A hurricane cut a gap in the Maryland coast at Ocean City in August 1933; jetties went in within weeks, and the town has had a navigable inlet since. But jetties do not stop drift, they intercept it. Assateague Island, downdrift, was starved: its northern end has retreated landward by several hundred metres and now sits visibly offset from its neighbour across the channel. Since 2002 the Park Service has moved sand around those jetties mechanically, settling a debt the tide used to settle for nothing.
Which is the honest shape of it. A tidal inlet is not a hole made in a coast. It is a rate being met — and it looks exactly like a hole until the morning the rate fails, when it turns out to have been made of nothing but water leaving.
≈
One loop I’m watching
Next: a shape held by the same trick on a far shorter clock. A hydraulic jump — the bright ring around the stream from a kitchen tap, or the standing wall a river throws up below a ledge — is a few centimetres of structure that an entire river passes through every second. Where this issue’s inlet is a balance of two sediment rates settled over a decade, the jump is a balance of two speeds settled continuously: fast shallow water running into slow deep water, the transition pinned exactly where the flow can no longer outrun its own ripples. Move the tap and it follows. Turn it down and it is gone. Next time.