Walk out onto the platform of a healthy salt marsh and you are standing on an answer, not a place. The cordgrass around you tolerates only a narrow slot of the tide’s range — roughly from mid-tide up to the high-water mark. Sit lower and the roots suffocate in floodwater that never leaves; sit higher and the ground dries into territory other plants win. In a big tide the slot is generous; in a small one it can be a few tens of centimetres tall. The marsh holds thousands of hectares inside it, flat as a table, and the frame will not hold still: the sea’s rise has doubled in thirty years, to about four and a half millimetres per year.
How does a lawn hold an altitude? With the flood itself. Twice a day the tide comes in carrying mud, and the depth of the flooding sets the size of the payment: deeper water stands longer and drops more of what it carries, and the stems it drowns are the baffle that slows it and combs the sediment out. Meanwhile the grass, over most of its range, grows better the wetter it gets — more stems above the surface and more root below, where the lift is partly manufactured directly; at some sites more than a third of a marsh’s rise is made underground, by roots shouldering the soil upward. Put those together and elevation becomes self-correcting. A patch that falls behind floods deeper, gets paid more, and climbs; a patch that gets ahead is visited briefly, paid little, and stalls. That is why the platform is so eerily level. Error pays for its own correction — sinking is the act the whole system runs on.
In 2002, after nearly two decades of watching plots at North Inlet, South Carolina, James Morris and colleagues turned this into arithmetic. Plot cordgrass growth against depth in the tidal frame and you get a hump: poor when too dry, poor when truly drowned, peaking in between. As long as the marsh sits on the dry side of that peak, the loop is stable, and the marsh settles at whatever depth lets it build exactly as fast as the water rises. It does not hold an elevation at all. It holds a depth — a fixed seat in a rising frame.
So when the sea accelerates, the marsh’s answer is the strangest part of the mechanism: it moves its seat down. Model ensembles show marsh surfaces adjusting to a faster sea by riding lower in the frame, where floods are deeper and payments are larger. It keeps up by falling behind — and the distance between its seat and the crest of the growth hump is its entire budget for doing so. Which carries an ugly corollary: a marsh drawing on the last of that budget is sitting near its own maximum lushness. Peak green is not health. It is the last stop. Past the crest, the same push that used to correct now compounds — deeper flooding weakens the grass, weaker grass catches less mud — and the thermostat becomes a ratchet. In the models, a marsh pushed over the line converts to open water in thirty or forty years, and does not come back.
Where the line sits is local, because the tide sets the working room and the mud sets the wage. Marshes in nearly clear water drown at a few millimetres of rise per year; marshes in water carrying thirty to a hundred milligrams of sediment per litre can climb several centimetres a year, and a big tidal range multiplies tolerance roughly tenfold over a small one. Louisiana ran the experiment: two microtidal marshes, each facing water rising locally around ten millimetres a year over sinking ground. Old Oyster Bayou, fed about seventy milligrams per litre, held. Bayou Chitique, fed about twenty, went under. Same sea. Different wage.
The Mississippi Delta’s own sediments record the rule at scale: across 8,500 years of cores, marsh there rarely outlived rates above about three millimetres a year for more than a few centuries, and past six to nine, it drowned within decades. The sea is at four and a half and climbing, while dams and soil conservation quietly cut the wages — many marshes standing today were built on a muddier century’s pay. In southern New England, where marshes already sit past the crest, about a sixth of the vegetated marsh has gone in four decades. And it does not go the way a beach goes, shaved from the edge. It goes from the middle: bare pools open in the interior, at the lowest seats, and widen. A raised bog — No. 58’s dome — escapes water by climbing out of its reach and living on rain, on a floor of its own dead. The marsh has no such exit. The tide is its paymaster, and the one thing it cannot survive is a raise it can’t match.
≈
One loop I’m watching
Next: every glacier carries a line nobody drew — the altitude at which a year’s snowfall exactly cancels a year’s melting. Above it the glacier only gains; below it the glacier only loses; and the ice pours downhill straight through the line, so the whole glacier is a conveyor strung between the two sides of a boundary. Move that line uphill by a hundred metres and the terminus answers for a century.