Mini WorldLaplace World / Plate 50 — River meanders

Why Rivers Wind

Rivers on plains are rarely straight. In each bend the water swirls: it cuts away the outer bank and piles sand on the inner one, so the bend grows and slowly moves downstream. When a loop gets too tight, the river cuts across its neck, and the old loop is left behind as a crescent lake.

Scroll

01

Rivers on Plains Are Rarely Straight

Where the slope is gentle and the banks are sand and mud, a river goes from bend to bend. The word meander comes from a famously winding river in western Turkey, the Maeander of the ancient Greeks, today the Büyük Menderes.

The channel’s length divided by the straight distance between its ends is its sinuosity. A straight river has 1. Leopold and Wolman called rivers with a sinuosity of 1.5 or more meandering. The river on the model is about 22 m wide; once its bends have grown, its sinuosity is mostly between 2 and 3.5.

Model: a valley about 900 m long. Heights are exaggerated about 15 times

02

In a Bend, the Water Spirals

Water entering a bend is thrown outward. The surface rises a little against the outer bank, and its slope pushes the water back in. On average the two balance, but not at every depth.

Near the surface the water is fastest and is thrown hardest, so it drifts outward. Near the bed, friction slows it, the slope wins, and it flows inward. The water sinks at the outer bank and rises over the inner one, turning as it goes downstream, like a stretched spiral.

Tea leaves gather in the middle of a stirred cup for the same reason: the water at the bottom, slowed by friction, flows to the centre and takes the leaves with it. In a short paper in 1926, Einstein used the teacup to describe this circulation, and with it explained why river bends grow ever more curved.

A bend in section. The flow across is drawn enlarged; in a river it is about a tenth of the speed downstream

03

The Outer Bank Goes, the Inner Bank Grows

Under the outer bank is the deep pool. The sinking water scours the foot of the bank, and the undercut bank falls in, block by block. This is the cut bank, often a steep scarp.

The water flowing inward along the bed carries sand to the inner bank and leaves it there as a gentle slope: the point bar. As the outer bank retreats, the point bar grows, so the river stays about as wide and the whole bend moves outward.

Each step of the bar leaves a low ridge of sand. Over the years the inner bank fills with curved ridges and swales: scroll bars. In section, the bar’s sand layers dip towards the channel; each one was the bank in some year.

04

A Small Bend Grows

Suppose the channel starts nearly straight, off by less than a metre. Water flowing past these tiny bends still scours their outer banks and fills their inner ones, so each bend gets a little bigger; a bigger bend scours harder.

The scour is strongest not at the apex of a bend but a little downstream of it, because the flow takes a distance to turn. So bends do not only grow, they travel downstream as they grow.

Very short bends are smoothed away and very long ones grow slowly. The bends that grow fastest are about ten channel widths long. On the model, after about five hundred years the valley is full of them.

The model starts from a nearly straight channel

05

Cutoffs and Oxbow Lakes

As a loop grows, its two ends come closer until only a narrow neck is left. A flood spills over the neck, or the banks on either side wear through, and the river takes the short way. This is a neck cutoff.

No current runs through the abandoned loop. Its ends silt up first, and a crescent lake is left in the middle: an oxbow lake. Over the following decades to centuries mud and plants fill it in, it becomes a marsh, and in the end woodland.

A cutoff makes the river shorter and straighter at a stroke, but it soon starts to bend again. So a meandering river’s sinuosity rises and falls, staying within a range over the long run.

06

Bends Scale with the River

Leopold and Wolman compared many reaches, from streams a few feet wide to rivers a mile wide. In their 1957 data the meander wavelength was 7 to 12 times the channel width; in 1960 they measured 50 reaches and found an average of about 11 times, with bend radii mostly 2 to 3 channel widths.

So on a map without a scale bar, a creek is hard to tell from a great river. On the model the fastest-growing bends are about 11 widths long. The number depends on the friction of the bed and on how wide the river is for its depth; the model uses values usual for sand-bed rivers on plains.

The ruler along the front edge: one division is one channel width (22 m)

07

The Same River from Orbit

This is the Ucayali in Peru, one of the headwaters of the Amazon, a far bigger river than the one on the model. Both pictures are from Landsat satellites and show the same reach, one in 1985 and one in 2023, about 50 km across.

In 38 years several loops were cut off. The river carries a lot of silt and is light brown; the loops cut off are now dark oxbow lakes, still water where the silt has settled. The curved stripes either side are old channels and scroll bars.

Satellite images: NASA Scientific Visualization Studio, Landsat (NASA / USGS)

08

The Bench

The time lever runs from a few years to three thousand, or time can run by itself. Bank erodibility sets how fast the bends move. A new seed starts the river again from a slightly different straight line. With secondary flow on, the section of a bend shows the spiral.

Try thisSet bank erodibility to 2× and drag time to year 3,000: how many oxbow lakes still hold water? Set it back to 0.5× and look again.

What the model leaves out

The centreline migrates by the method of Howard and Knutson (1984): each point moves outward at a speed set by the curvature there and along a stretch of channel upstream, whose influence falls off exponentially with distance (the bend theory of Ikeda, Parker and Sawai, 1981); the length of that fall-off comes from the bed’s friction coefficient (0.01) and the depth. The width is fixed at 22 m, with a width 36 times the depth; the discharge does not change, with no floods or low water. How hard the banks are to erode varies across the valley in random patches by ±35%, with a small random jostle every year; the terraces on either side are hard to erode. A cutoff happens when two parts of the channel come within one width of each other. How fast point bars build up, how fast oxbows fill (about a thousand years) and how soon trees grow on new bars are set by the page. How fast time runs depends on bank erodibility: on the model bends move tens of centimetres to a metre or two a year; real rivers move from a few centimetres to tens of metres a year. Heights are exaggerated about 15 times; trees and buildings are to scale. In the section of a bend, the flow across is drawn as one closed swirl, and enlarged.

Easy to get wrong

Rivers do not bend to go around obstacles, and they do not need the Earth’s rotation: on an even plain a straight channel is itself unstable, and a small bend grows on its own. The Earth’s rotation does push the flow slightly to one side at mid-latitudes, but in a bend it is much weaker than the bend’s own effect. The deepest water is near the outer bank, but the scour is strongest downstream of the apex, so bends travel downstream as well as bulging outward. An oxbow lake is not a lake the water dug out; it is an old loop the river left behind. After a cutoff the river does not stay straight; it starts to bend again.

Sources

  1. Leopold & Wolman (1957): River channel patterns: braided, meandering, and straight, USGS Professional Paper 282-B
  2. Leopold & Wolman (1960): River meanders, Geological Society of America Bulletin 71, 769–793
  3. Langbein & Leopold (1966): River meanders — theory of minimum variance, USGS Professional Paper 422-H
  4. Einstein (1926): Die Ursache der Mäanderbildung der Flußläufe und des sogenannten Baerschen Gesetzes, Die Naturwissenschaften 14, 223–224
  5. Ikeda, Parker & Sawai (1981): Bend theory of river meanders. Part 1. Linear development, J. Fluid Mech. 112, 363–377
  6. Howard & Knutson (1984): Sufficient conditions for river meandering: a simulation approach, Water Resources Research 20, 1659–1667
  7. NASA Scientific Visualization Studio (2026): The Meandering Ucayali River (Landsat, 1985–2023)

Data

Satellite prints: the Ucayali River, Peru, in 1985 and 2023, from Landsat (a joint NASA and U.S. Geological Survey mission), by NASA’s Scientific Visualization Studio (visualizer Ross K. Walter, SSAI), “The Meandering Ucayali River” (May 2026); NASA and USGS imagery, in the public domain in the United States. Cropped by this site; the 10 km scale bars are the site’s, measured against the bar in NASA’s video. River migration after Alan D. Howard and Thomas R. Knutson, Sufficient conditions for river meandering: a simulation approach (Water Resources Research 20, 1984), with the bend-flow weighting of Syunsuke Ikeda, Gary Parker and Kenji Sawai, Bend theory of river meanders, part 1 (J. Fluid Mech. 112, 1981). Meander geometry from Luna B. Leopold and M. Gordon Wolman (USGS Professional Paper 282-B, 1957; GSA Bulletin 71, 1960) and Walter B. Langbein and Luna B. Leopold (USGS Professional Paper 422-H, 1966). The tea-leaf explanation: Albert Einstein, Die Naturwissenschaften 14, 223–224 (1926). The simulation, the diorama, the bend cast and the words are the site’s own.

The text, pictures and models on this page are licensed CC BY-NC 4.0: free to use and adapt for teaching and other non-commercial purposes, with credit to world.lapla.si. Third-party material listed above keeps its own licence.

World · all plates