Field notes on things that run themselves
A Waterfall Is Not a Place
A waterfall looks like a place: a name, a spot on a map. But to the rock underneath it, a waterfall is a step in a river’s long profile — the one stretch of channel where the bed drops fast instead of gradually — and a step like that doesn’t have to stay put. Given enough time it walks upstream, the same drop and the same plunge pool relocating a few metres at a time, arriving eventually at rock the river had never touched. Geomorphologists call the traveling version a knickpoint. A waterfall is what one looks like from the air.
That makes it close to the photographic negative of the standing wave that opened this series. A river’s standing wave is a pattern in the water, held fixed by an obstruction while the water itself passes through, never the same water twice. A knickpoint is the pattern in the rock instead — stationary in shape, but not in location — and what passes through it, a grain at a time, is the riverbed itself. One holds still while its substance flows past. The other moves while its substance is what’s being taken away.
The textbook mechanism is one clean story: a resistant cap sits over weaker rock, the plunge pool’s turbulence chews out the soft layer beneath it, and the unsupported overhang eventually collapses, stepping the whole face back upstream. That happens — it just isn’t the only thing that happens. Flume experiments published in the journal Geology found that drop height changes which process does the work: tall waterfalls do retreat mainly by that headwall undercutting, but short ones — same rock, same sediment, same discharge — retreat up to five times faster, by fracturing into a staircase of small, fast-migrating steps called cyclic steps instead. A 2024 field and modeling study by Inoue and coauthors, in Geophysical Research Letters, found the same split outside the lab: high flow spreads erosion into a train of steps; low flow concentrates it at the face, the classical way. Geomorphologists do lean on a general rule here — that incision rate scales with drainage area and channel slope in a way that lets a base-level drop migrate upstream as a coherent step — but its exact exponents are debated and site-dependent enough to call it a working framework, not a settled law.
Niagara Falls is the knickpoint most people have already watched retreat without knowing it. About 12,000 years ago, as the last ice sheet retreated, the Niagara River first spilled over the Niagara Escarpment near present-day Lewiston. It has cut upstream roughly 11 kilometres since, carving the gorge behind today’s Horseshoe and American Falls. The usual number for how fast is more confident than the record supports; the honest version is a range. Nineteenth- and early-twentieth-century surveys put the historical rate at around a metre a year; today’s is somewhere between a tenth and three-tenths of a metre. That slowdown isn’t the river calming down on its own. A 1950 U.S.–Canada treaty guarantees only a minimum flow over the falls — 2,832 cubic metres a second by day, half that at night — diverting most of the roughly 5,700 cubic metres a second the river naturally carries into tunnels feeding hydroelectric stations rated up to 2.4 million kilowatts. Throttle the water, throttle the erosion. That’s the well-known half. The lesser-known half: Horseshoe Falls has also been throttling itself. As its crest has broken off unevenly along the curve, the lip has grown longer — 420 metres in 1842, 762 by 2000 — spreading the same flow across more edge and cutting the erosive force on any one stretch. A 2009 study modeling both falls found Horseshoe’s slowdown is shared between the engineered flow cut and this self-widening; the much smaller American Falls slowed almost entirely from the diversion alone.
Every system this publication has described so far holds its shape because something keeps moving through it — water through a standing wave, cars through a phantom jam, wind through a mound. A knickpoint runs the opposite arithmetic. Nothing flows through the waterfall to keep it standing; it keeps its shape because the rock composing it right now is being permanently removed and handed to whatever new rock happens to be waiting upstream — the first pattern in this series held up by subtraction instead of throughput, aimed patiently in one direction for as long as the river keeps cutting.
The caprock story most people know isn’t wrong. It’s one entry in a short list that depends on how tall the drop is and how the sediment happens to be moving that day. Some waterfalls step back by collapsing. Some by fracturing into a staircase. A few, in gentler knickzones, retreat by drilling straight down into their own plunge pools instead of undercutting anything. None of it changes what’s actually being conserved: not the rock, not even the water, but the shape of the step — and the fact that it will be standing somewhere else the next time anyone measures it.
One loop I’m watching
Next: twice a day, in a handful of funnel-shaped estuaries around the world, an incoming tide doesn’t simply raise the river’s level. It arrives as one steep wall of water running upstream, against the current, holding its shape for kilometres — not because anything is standing still, but because the same narrowing channel reliably rebuilds the identical wave from scratch every single tide.
Tip: the ← and → arrow keys move between issues.
Prefer feeds? Follow by RSS or JSON Feed.
Want something playable between issues? Thought Toys has its own email list and RSS feed.
New to The Standing Wave? Start here →