Coastline paradox
The coastline paradox is the observation that a coastline has no single, well-defined length: the length you get depends on how finely you measure.[1:1] A coast is made of bays within bays and headlands upon headlands. Measure it with a long ruler and you cut straight across the smaller inlets; measure it with a shorter ruler and you follow them, so the total grows. For a smooth curve such as a circle, finer and finer measurements settle on a fixed value. For a coastline, they keep rising.[2:1]
The mathematician Hugo Steinhaus described the effect in 1954, the scientist Lewis Fry Richardson measured it, and Benoit Mandelbrot turned it into a way of describing roughness, now called the fractal dimension.[2:2][1:2] The paradox has practical consequences. Official figures for the same coast can differ many times over, depending on the measuring step and on what is counted as coast: Norway's mainland coast is 2,532 km if the line cuts across its fjords and bays, and 28,953 km if it follows them.[3:1]
How a coastline is measured
The classic method is to walk a pair of dividers, set to a fixed opening, along the coast on a map. Each step joins two points on the coastline with a straight line, and the measured length is the number of steps multiplied by the length of one step.[1:3]
Applied to a straight line, this method gives the same answer whatever the step. Applied to a circle, the answer rises a little as the step shrinks, because short steps hug the curve more closely, but it quickly settles on a limit: the circumference. Mathematicians call curves whose length settles in this way rectifiable.[2:1]
A coastline behaves differently. Each shorter step reaches into bays and around headlands that the longer step skipped, and inside those bays are smaller bays, and inside those smaller ones still. The measured length keeps increasing as the step shrinks, without settling on any value.[2:1] Measuring more carefully does not bring the answer closer to a true length; it only adds to the total.[1:1]
History
Steinhaus
In 1954 Hugo Steinhaus observed that the left bank of the Vistula, measured with ever greater precision, would give lengths 10, 100 or even 1,000 times as great as the length read off a school map. He concluded that most curves found in nature are not rectifiable, the opposite of the common belief that such curves are a mathematicians' invention.[2:2]
Richardson
Lewis Fry Richardson, a scientist known for his work on turbulence, also studied armed conflict between states.[2:3] He noted that the length of the border between Spain and Portugal, and of the border between Belgium and the Netherlands, differed by about 20% depending on which neighbour's encyclopedia reported it. Mandelbrot later attributed the gap, at least in part, to different measuring steps, and suggested that a small country such as Portugal would naturally measure its border more finely than its larger neighbour.[2:4][2:5]
Richardson measured several coastlines and land borders with steps of different lengths. When he plotted the logarithm of the measured length against the logarithm of the step, the points fell along straight lines.[4:1] He summarised this as an empirical rule: measured length equals a constant times the step length raised to the power 1 − D, where D is a second constant that differs from one coastline to another.[1:2] To Richardson, D was simply an exponent. His diagrams were found among his papers after his death and published in 1961 in a little-read yearbook.[2:6]
Mandelbrot
In a paper in Science on May 5, 1967, titled "How long is the coast of Britain? Statistical self-similarity and fractional dimension", Benoit Mandelbrot noted that Richardson's rule had attracted no attention, and proposed that the exponent D should be read as a dimension.[1:2][4:1] In that paper he was cautious: the most that could safely be said, he wrote, was that Richardson's data were compatible with coastlines being random shapes of fractional dimension, though he favoured the stronger reading.[1:4]
By his book The Fractal Geometry of Nature (1982; updated edition 1983) he went further. He proposed calling curves whose dimension exceeds 1 fractal curves, and concluded that within the range of scales that matter to geographers, coastlines can be modelled as fractal curves.[2:7] He also stated that no mathematical proof of this is possible in a natural science; his aim was to show that the length of a coastline is a genuine conceptual problem, and that D offers a workable answer.[2:8]
Fractal dimension
The fractal dimension D measures how quickly a curve's measured length grows as the measuring step shrinks. In Richardson's rule, a curve with D = 1 has a length that does not change with the step: it behaves like an ordinary smooth line. The higher D is above 1, the faster the length grows as the step shrinks, and the rougher the curve.[1:2] By the rule, halving the step multiplies the measured length by 2 raised to the power D − 1: for D = 1.02 the length grows by about 1.4%; for D = 1.25, by about 19%.
From Richardson's data, Mandelbrot derived the following values:[1:5]
| Coastline or border | D |
|---|---|
| A border that looks straight on the map | 1.00 |
| Coast of South Africa | 1.02 |
| Coast of Australia | 1.13 |
| Land border of Portugal | 1.14 |
| Land border of Germany | 1.15 |
| West coast of Britain | 1.25 |
The coast of South Africa, one of the smoothest in the atlas, is almost an ordinary line. The west coast of Britain, chosen because it looks like one of the most irregular coasts in the world, has the highest value.[1:5]
Mathematics supplies idealised curves that show the same behaviour perfectly. In the Koch curve, each straight segment is replaced by a smaller copy of a zigzag pattern, and each segment of that copy is replaced again, without end. The finished curve is a line, yet its length is infinite, and its dimension is greater than 1.[1:6] Such a curve is self-similar: every small piece is a scaled-down copy of the whole. Mandelbrot's proposal was that real coastlines are self-similar in a statistical sense: a small stretch of coast is not an exact copy of a large one, but it is similarly irregular.[1:4]
Limits of the idea
Real coastlines are not perfect fractals. Mandelbrot himself pointed to a coast that people have reshaped, such as the coast at Chelsea. There, the measured length barely changes for steps between roughly 20 metres and 20 centimetres, and only starts to rise again when the step is small enough to register individual stones.[2:9] More generally, a coastline measured with a short enough step eventually behaves like a straight line, and further shortening adds nothing.[5:1] In theory a coast could be measured down to pebbles or grains of sand, but real coastlines do not repeat their structure at every size the way mathematical fractals do.[6:1]
The dimension of a coastline is not a single fixed number either. Different stretches of the same coast can give different values of D.[2:10] In a 1982 study of the coast of Kodiak Island, Mark Shelberg, Harold Moellering and Nina Lam found that their estimates of the dimension varied, which they read as a sign that the coast is not truly self-similar. They noted that strict self-similarity is rare in nature, and cited earlier coastline studies suggesting that the processes shaping a coast act at particular scales.[5:2][5:3]
In notes to a later reprint of his 1967 paper, Mandelbrot wrote that many non-scientists know it for the wrong reason, as a proof that a coastline gets longer the closer you look, and that the observation that natural curves lack a definite length was not original to him.[1:7]
Why coastlines are rough
Mandelbrot's own first, simple model of how coastlines form gives a dimension of 3/2, which he said was much too high to describe real coasts, but enough to show that a coastline's dimension should be expected to exceed 1.[2:11]
In 2004 Bernard Sapoval, Andrea Baldassarri and Andrea Gabrielli proposed a mechanism based on a feedback loop. Waves erode the weakest parts of a rocky coast, which makes the coast more irregular. An irregular coast in turn damps the waves, weakening their erosive power. Erosion continues until the remaining coast is strong enough to resist the weakened waves. In their computer model, the coast that results is a fractal with a dimension close to 4/3, a value tied to the mathematics of percolation, the study of how connected paths form through randomly arranged material.[7:1] The authors describe the model as minimal: it simplifies how waves are damped and assumes that eroded material is carried away. They say the value 4/3 is close to the dimension observed for many seacoasts, but they do not test the model against field measurements.[7:2]
Coastline figures in practice
Because the measured length depends on the scale, published figures for the same coastline differ, and estimates of the total length of the world's coastlines vary widely.[6:2]
In the United States, the National Oceanic and Atmospheric Administration (NOAA) gives the country's tidal shoreline, including outlying territories, as 95,471 miles. This figure was measured by hand in 1939–40 on the most detailed charts then available. It includes the shores of offshore islands, sounds, bays, rivers and creeks, followed inland until the tidal water narrows to a width of 100 feet.[8:1] A report by the Congressional Research Service gives a far shorter figure for the general coastline: 12,383 miles. It cites a measurement on other nautical charts of 29,093 miles, and a figure of 88,612 miles used for the federal coastal zone management program, and notes that coastline distances differ significantly depending on the scale of the maps or charts used.[9:1]
Norway's official figures show how much depends on what is counted:[3:1]
| What is measured | Length |
|---|---|
| Mainland, along a line cutting across fjords and bays | 2,532 km |
| Mainland, following fjords and bays | 28,953 km |
| Coastlines of islands | 71,963 km |
| Total, mainland and islands | 100,915 km |
Mandelbrot argued that the problem cannot be settled by agreeing on a standard step. A single step size might serve one agency, he wrote, but its adoption by every agency of a government is hard to imagine, and by every country all but inconceivable. In his view, the length of a coastline is not entirely objective: the person measuring is part of its definition.[2:12]
- ^a ^b ↗ length-undefinable ^a ^b ^c ^d ↗ richardson-law-ignored ^ ↗ dividers-method ^a ^b ↗ compatible-hedge ^a ^b ↗ d-values ^ ↗ koch-infinite ^ ↗ wrong-reason Mandelbrot, Benoit B. (1967-05-05). How long is the coast of Britain? Statistical self-similarity and fractional dimension. Science. https://doi.org/10.1126/science.156.3775.636 http://users.math.yale.edu/~bbm3/web_pdfs/howLongIsTheCoastOfBritain.pdf.
- ^a ^b ^c ↗ rectifiable ^a ^b ↗ steinhaus-vistula ^ ↗ richardson-background ^ ↗ encyclopedias-differ ^ ↗ portugal-finer ^ ↗ richardson-posthumous ^ ↗ fractal-curves ^ ↗ no-proof ^ ↗ chelsea-flat-zone ^ ↗ d-varies ^ ↗ d-three-halves ^ ↗ observer-intervenes Mandelbrot, Benoit B. (1983). The Fractal Geometry of Nature. W. H. Freeman, New York. ISBN 9780716711865.
- ^a ^b ↗ norway-coastline Statistics Norway (2013). Geographical survey. Statistical Yearbook of Norway 2013. https://www.ssb.no/a/english/aarbok/kart/i.html.
- ^a ^b ↗ log-log-slope Frame, Michael; Mandelbrot, Benoit; Neger, Nial. Coastlines. Fractal Geometry. Yale University. https://gauss.math.yale.edu/fractals/Panorama/Nature/Coastlines/Coastlines.html%20copy.
- ^ ↗ lower-limit ^ ↗ kodiak-variation ^ ↗ self-similarity-rare Shelberg, Mark C.; Moellering, Harold; Lam, Nina (1982). Measuring the fractal dimensions of empirical cartographic curves. Proceedings of Auto-Carto 5. https://cartogis.org/docs/proceedings/archive/auto-carto-5/pdf/measuring-the-fractal-dimensions-of-empirical-cartographic-curves.pdf.
- ^ ↗ not-true-fractals ^ ↗ estimates-vary (2017). How long are the world’s coasts? World Ocean Review 5: Coasts – A Vital Habitat Under Pressure. maribus. https://worldoceanreview.com/en/wor-5/coastal-dynamics/the-myriad-faces-of-the-coasts/how-long-are-the-worlds-coasts/.
- ^ ↗ damped-erosion ^ ↗ model-limits Sapoval, Bernard; Baldassarri, Andrea; Gabrielli, Andrea (2004-08-27). Self-stabilized fractality of seacoasts through damped erosion. Physical Review Letters. https://doi.org/10.1103/PhysRevLett.93.098501 https://arxiv.org/abs/cond-mat/0311509.
- ^ ↗ us-tidal-shoreline National Oceanic and Atmospheric Administration. Frequently asked questions: How long is the U.S. shoreline? NOAA Shoreline Website. https://web.archive.org/web/20230802161851/https://shoreline.noaa.gov/faqs.html?faq=2.
- ^ ↗ scale-matters Beaver, Janice Cheryl (2006-11-09). U.S. International Borders: Brief Facts. Congressional Research Service. https://www.everycrsreport.com/files/20061109_RS21729_9b2e66ba1f8c8da8e5e3f69827e85bc9e841dcb0.html.
Wikidata: Q1797373
Discussion
No threads yet.
Sign in to start threads and post replies.