Why coastlines are getting longer the more precisely you measure them

The CIA World Factbook puts the United Kingdom’s coastline at 12,429 kilometres. The World Resources Institute puts it at 19,717. The discrepancy — roughly a factor of 1.6 — is not a rounding error or a surveying dispute. Both figures are correct. They are answers to different questions wearing the same label.

This is not a country that lacks the means to measure itself. The Ordnance Survey has been mapping Britain since 1791. The UK has GPS infrastructure, satellite imagery, comprehensive hydrographic data, and more cumulative cartographic investment per square kilometre than almost anywhere on earth. It still cannot tell you how long its own coast is — not because it hasn’t tried, but because the question has no single answer. The CIA uses generalised large-scale charts. The WRI used a 1:250,000 GIS dataset derived from the US Defense Mapping Agency’s World Vector Shoreline database. Different scales, different methodologies, different numbers. Neither is wrong. Both are incomplete without a footnote.

The maddening part isn’t the gap between the two figures. It’s that the gap grows as you zoom in. Measure more precisely, count more bays and headlands and estuary edges, and the coast gets longer. There is no measurement fine enough to resolve it. This is not an engineering limitation. It is a mathematical property of the thing being measured.

Most people who’ve heard this before treat it as a curiosity — a clever parlour fact about fractals, filed somewhere between Zeno’s paradox and the Monty Hall problem. Something mildly surprising about geometry. The trouble is that coastlines are also where nations draw the boundaries of their fishing rights, their mineral claims, and their sovereign territory. Treaties reference those boundaries. Courts adjudicate disputes about them. Insurance policies are priced against them. If the number depends on how you measure it — if there is, in fact, no number — then every institution built on that number is built on a choice someone made, probably without announcing it.

That’s not a curiosity. That’s a problem.

The problem Richardson found

Lewis Fry Richardson (1881–1953) was a Quaker meteorologist who, among other things, pioneered numerical weather forecasting and spent decades trying to understand the causes of war. By the late 1940s he was investigating a specific hypothesis: that the probability of armed conflict between two countries was related to the length of their shared border. The logic was intuitive enough — more border, more friction, more war.

To test this he needed border lengths. He looked them up.

Spain, he found, recorded its border with Portugal at 987 kilometres. Portugal recorded the same border at 1,214 kilometres. A 23% difference. That’s not a rounding error. Richardson checked other pairs — Belgium and the Netherlands showed similar discrepancies. These weren’t cases where one country was lying or where the surveys were old. Both countries had competent geographers using accepted methods. They simply used different scales of measurement, and the scale turned out to matter enormously.

Richardson’s finding — now called the Richardson effect — was that measured border lengths follow a power law as the unit of measurement changes. Use a shorter ruler and you catch more detail: the small notch in the hillside, the curve of the river mouth, the way the coastline doubles back around a tidal inlet. Use a longer ruler and those features disappear, averaged out. The longer the ruler, the shorter the measured length. This was not, he insisted, a surveying error. Both Spain and Portugal were correct. They had measured the same thing at different scales, and in geography, scale is not a neutral technical choice.

The Richardson effect is most dramatic in coastlines, which are more irregular than land borders. At a ruler length of 100 kilometres, the UK coastline comes in around 2,800 kilometres. At 50 kilometres, around 3,400. Bring the ruler down to 1 kilometre and you’re past 8,000 before you’ve accounted for beach curvature, estuary fingers, and the micro-indentations of sea cliffs. Keep going and the number keeps growing.

Richardson was trying to count wars. He published his border-length findings posthumously in “The Problem of Contiguity,” in the General Systems Yearbook in 1961, eight years after his death in September 1953. His war probability work remained inconclusive — he never found the correlation he was looking for, and the findings were messy enough that he didn’t publish them as settled science. At the time, the problem looked like it might be solvable with better data. It wasn’t.

What Richardson had found was not bad data. It was a deeper property of the curves he was measuring — something that wouldn’t yield to finer instruments, something that existed in the geometry itself. He died without knowing what it was.

International law, meanwhile, had been using coastline measurements to divide up the ocean for decades. The question was whether it had noticed what Richardson found — and what happens when a legal system builds itself on a number that doesn’t stay still.

What Mandelbrot made of it

Fourteen years after Richardson’s death, Benoît Mandelbrot published a three-page paper in Science — “How Long Is the Coast of Britain? Statistical Self-Similarity and Fractional Dimension,” Volume 156, Number 3775, May 5, 1967, pages 636–638 — that answered Richardson’s question by showing it was unanswerable.

Richardson had described the empirical phenomenon. Mandelbrot formalised it. The length of a coastline, he showed, depends not just on the scale of measurement but on what he called the fractal dimension of the curve — a number, symbolised D, that sits between 1 (a straight line) and 2 (a filled plane). For a coastline, D is greater than 1. The west coast of Britain, Mandelbrot calculated using Richardson’s own measurements and his divider method across scales from roughly 1 to 200 kilometres, has a fractal dimension of approximately 1.25. The coastline of South Africa is close to 1.02 — nearly smooth, barely more complex than a straight line.

The significance of D > 1 is this: as you reduce the ruler length, measured length doesn’t converge. It increases without bound.

A straight line or a gentle smooth curve reaches a stable value as you measure more precisely. A fractal curve — a curve with self-similar irregularity at every scale — doesn’t. It has no true length. Not “unknown true length.” No true length.

Calculating fractal dimension

Mandelbrot's method, adapted from Richardson, is the divider method: walk a compass set to ruler length G along the coastline, count the steps N required to complete the traverse, multiply. Measured length L = N × G. Do this at several scales, plot log(G) against log(L). The slope of the line gives D − 1. For Britain's west coast, Mandelbrot found a consistent slope across multiple scales, giving D ≈ 1.25. That consistency — the same slope across 1 km to 200 km — is what self-similarity means. Each portion of the coast, zoomed in, statistically resembles the whole. South Africa's D ≈ 1.02 means a much shallower slope: length increases slowly with finer measurement. Britain's jagged Atlantic-facing coast grows fast. There is no scale at which you'd expect convergence.

Ordinary measurement error and fractal dimension are different kinds of problem. If you’re measuring a room with a tape measure and you’re getting inconsistent results, better instruments help. Finer measurement resolves the problem. With a fractal curve, finer measurement makes the problem worse — not because the instruments are bad, but because the curve itself has more structure at every scale you probe. The problem doesn’t diminish with precision. It is precision.

Richardson, meanwhile, had been correlating conflict against border length — a variable that, as Mandelbrot now showed, has no stable values. Measure every border in his dataset at the same ruler length and you get internally consistent numbers; change the ruler and they change. His correlation may have been inconclusive not because his theory was wrong, or the data bad, but because the independent variable doesn’t exist in the form he assumed. That’s speculation — Richardson’s methods had other limitations and the geography-conflict relationship is genuinely complex. But the possibility is there.

How the law solved the problem it didn’t know it had

The United Nations Convention on the Law of the Sea — UNCLOS, opened for signature in 1982, in force from 1994 — defines maritime zones from a baseline. The territorial sea extends 12 nautical miles from that baseline (Article 3). The contiguous zone reaches 24 nautical miles (Article 33). The exclusive economic zone, where a state has sovereign rights over living and mineral resources, extends 200 nautical miles (Article 55–57). All of it depends on where you draw the baseline.

UNCLOS specifies two methods. Article 5 establishes the “normal baseline” — the low-water line along the coast as marked on large-scale charts officially recognised by the coastal state. Article 7 permits “straight baselines” for coasts that are deeply indented or fringed with islands: connect the outermost points, draw a line, measure from there. Article 16 requires that the resulting charts and lists of geographical coordinates be deposited with the UN Secretary-General.

The phrase that matters in Article 5 is “large-scale charts officially recognised by the coastal state.” Scale choice is delegated to the interested party.

The state decides what scale its charts are drawn at. The state decides how much coastal complexity gets preserved in the baseline, and therefore how far seaward its maritime zones project.

Straight baselines under Article 7 make this more explicit. Rather than tracing the intricate low-water line, a state with a complex coast can connect outer points with straight lines and measure from there. Every notch and fjord enclosed behind those lines becomes internal waters. Every straight line pushes the outer boundary of the EEZ further from shore.

The Anglo-Norwegian Fisheries Case of 1951 — decided by the International Court of Justice before UNCLOS was drafted — established the precedent that straight baselines were legally permissible for coasts like Norway’s. The Norwegian skjærgård, the “rock rampart” of islands, islets, rocks, and skerries that lines the country’s Atlantic coast, is among the most complex coastal formations on earth. The Norwegian government has counted something in the range of 120,000 to 150,000 separate insular formations along that coast. The UK brought the case arguing that Norway’s 1935 decree, which used 48 straight baseline points to delimit its fishing waters, violated international law.

The ICJ ruled for Norway in December 1951. The court found that where the skjærgård forms a continuous coastal fringe, the outer edge of that fringe — not the inner mainland shore — constitutes the coast for baseline purposes. Norway kept its more generous delimitation. Britain lost.

What the 1951 judgment did, in practice, was ratify a methodology that produced a territorial outcome. It didn’t measure the coast. It chose a framework for measuring the coast, and that choice had beneficiaries. Norway gained; Britain’s fishing fleets did not. Courts reach decisions. They don’t make measurements. The distinction matters enormously, because the decision is dressed in the language of measurement, and most people reading the judgment wouldn’t notice.

Three cases where the convention fails

The 1825 Anglo-Russian Convention that settled the boundary between Russian Alaska and British North America had a line-drawing problem that would take seventy-eight years and a gold rush to expose.

The treaty specified that the boundary should run “parallel to the windings of the coast” at 10 marine leagues — roughly 56 kilometres — inland. The phrase was clear enough as long as no one cared about the territory. When gold was found in the Klondike in 1896 and the boundary suddenly controlled access to ports and trade routes, the question became urgent: was “the coast” the mainland shore, or the outer line of the island chain? Canada argued outer islands; the United States argued mainland.

The 1903 tribunal gave the Americans what they wanted. Canada lost access to the Alaskan panhandle and the most direct routes from the Yukon goldfields to the sea. “The windings of the coast” had no stable referent — it wound differently depending on how you looked at it. The treaty had assumed the question answered. Seventy-eight years later, the unresolved answer cost Canada accordingly.

Okinotorishima is two coral outcroppings in the Philippine Sea, approximately 1,740 kilometres south of Tokyo. At high tide, the exposed rock totals roughly 10 square metres — the footprint of a large bedroom. Japan claims them as islands. If they qualify as islands under UNCLOS, they generate a 200-nautical-mile EEZ. The area at stake is approximately 400,000 square kilometres of ocean — larger than the entire land area of Japan itself.

UNCLOS Article 121(3) says that “rocks which cannot sustain human habitation or an economic life of their own shall have no exclusive economic zone or continental shelf.” Japan’s position is that Okinotorishima is not a rock but an island. To support this position, Japan has encased the two outcroppings in concrete breakwaters and titanium nets, specifically to prevent them from being submerged by tide and erosion. The construction has cost over $600 million.

China, Taiwan, and South Korea dispute the claim. China’s position is that at high tide Okinotorishima barely exists as a surface feature, and that engineering intervention doesn’t change what it is. Japan disagrees, and Japan controls the territory.

The UNCLOS binary — island or rock — encounters something it wasn’t designed to handle: a state engineering the physical characteristics of a geographic feature to satisfy a legal classification. The fractal problem in the Okinotorishima case isn’t measurement scale. It’s that the line between “island” and “rock” is being drawn by the entity with the largest financial interest in where the line falls, using concrete.

Why two coastline figures are both correct

The CIA World Factbook (12,429 km) and the World Resources Institute (19,717 km) used different source datasets and resolutions. The CIA draws on generalised large-scale charts designed for navigational use, which smooth coastal complexity below a certain threshold. The WRI figure comes from the World Vector Shoreline database at 1:250,000 scale, with lengths calculated using GIS across individual line segments. More vertices in the dataset, more complexity preserved, longer measured coast. Neither is wrong. Neither comes with a footnote explaining this. The 1.6x gap isn't between two measurements of the same thing — it's between two answers to two different questions, both labelled "UK coastline." This is the coastline paradox at its most institutional.

The Pacific Islands Forum Declaration of 6 August 2021 represents the third case — and the one that most openly acknowledges what is happening.

Eighteen Pacific Island Forum member states, including Tuvalu, Kiribati, and the Marshall Islands, declared that their maritime zones would not be updated as sea-level rise physically moved their coastlines. The baselines, established and notified to the UN Secretary-General under UNCLOS, would remain fixed even as the physical features from which they were drawn subsided, shifted, or disappeared.

The declaration is explicit about the contradiction. It acknowledges that the relationship between sea-level rise and maritime zones wasn’t contemplated by UNCLOS’s drafters. It invokes legal stability, security, and predictability as reasons to maintain the zones regardless of physical change. More than a hundred states and international organisations have expressed support, according to the Forum Secretariat’s own summary of the declaration’s reception.

The economic logic is not subtle. Tuvalu’s government derives significant revenue from licensing fishing access to its EEZ. If the baseline recedes with the coastline as sea levels rise, the EEZ shrinks, and the revenue shrinks with it. The forum declaration freezes the maritime boundary at a point in time and holds it there as the physical reality beneath it dissolves.

The parallel with the ICJ’s 1951 judgment is structural: in both cases, a legal institution confronts a measurement problem and resolves it by choosing a methodology, blessing it, and moving on. The 1951 court chose straight baselines and didn’t dwell on the implications. The 2021 declaration chooses temporal fixity and says so directly. It doesn’t pretend the physical facts support the legal conclusion. It just asserts the legal conclusion will stand anyway.

The political epistemology of the ruler

Every coastline figure published without a stated methodology is structurally misleading — not an accusation of bad faith against the CIA, the WRI, or the UK Hydrographic Office, but a statement about institutional habit. The assumption that a coastline can simply be measured, that the number distinguishes good sources from bad ones, is so embedded that citing a figure without its methodology feels normal. It shouldn’t.

UNCLOS delegates scale choice to the coastal state, which means the treaty that defines the world’s maritime order builds the coastline paradox into its operating assumptions and then hands the variable to the party with the strongest interest in the outcome. Norway gained from straight baselines not by preserving coastal complexity but by replacing it — straight lines connecting the outermost points of the skjærgård push the EEZ seaward further than any traced low-water line would. Japan at Okinotorishima inverts this logic: it needs fine enough resolution to detect the outcroppings at all, so the methodology must be one that can find them and classify them as habitable land. The common pattern is not coincidence. States select the methodology that produces the territory they want, and call it technical.

The same definitional problem runs through domestic coastal governance, closer to individual landowners. FEMA’s Flood Insurance Rate Maps have historically treated the coastline as a static feature. Risk Rating 2.0 — fully implemented in April 2023 — incorporates coastal erosion as a pricing factor, but applies it as a proximity-based adjustment: erosion risk decreases linearly within 100 metres of the coast. No future shoreline is projected. The map doesn’t move.

The UK Environment Agency’s Shoreline Management Plans make the opposite choice, with unusual explicitness. They project the moving shoreline over 20-, 50-, and 100-year time horizons, and assign each policy unit one of three management options: hold the line, managed realignment, or no active intervention. A community’s designation can differ between those horizons — what the SMP holds in the near term, it may realign at 50 years and abandon at 100. The policy unit where someone’s house sits has three different coastlines depending on which temporal scale you’re reading.

The core problem isn’t that the coastline paradox is unsolvable. It’s that it’s treated as solved when it isn’t — buried in technical procedures that don’t advertise the choices they embed. Courts handle scientific uncertainty through expert testimony: disclosure, questioning, cross-examination, methodology on the record. International maritime law and domestic coastal governance behave as though coastline measurement is a settled technical matter, as though the number that emerges from a particular survey is a reading of geographic reality rather than a product of a methodology selected by someone with interests.

Mandelbrot proved in 1967 that there is no geographic reality to read, at least not one with a stable length. The performance continues.

Closing

Richardson spent years trying to find whether border lengths predicted the likelihood of war. He found the border measurements were contradictory and the correlation was inconclusive. He couldn’t figure out why the numbers kept shifting. He died in September 1953 still, in some sense, counting.

Mandelbrot’s formalisation came fourteen years later. His paper explained exactly why Richardson’s numbers had no stable values — the curves he was measuring had fractal dimensions greater than 1, which means their measured length diverges without bound as measurement precision increases. Richardson had been trying to correlate something against an independent variable that doesn’t exist in the stable, measurable way he assumed.

This is speculative. Richardson’s inconclusive findings had other possible explanations — the theory was crude, the historical record of borders is itself contested, the relationship between geography and conflict is overdetermined. But the possibility remains: the man who discovered the problem may have been its first significant victim.

Coastlines are the canonical example of a thing that resists measurement by its nature — not because instruments are imprecise, but because precision makes the problem larger. The more carefully you look, the more there is to see, and the total never converges. That discovery has been in the mathematical literature since 1967. The systems that most need to engage with it are still behaving as though it hasn’t been made.

Richardson’s inconclusive numbers are the honest answer. Everyone else’s confident ones are a choice.

Gen AI Disclaimer

Some contents of this page were generated and/or edited with the help of a Generative AI.

Media

sayari bechir – Pexels

Key Sources and References

Richardson, L.F. “The Problem of Contiguity: An Appendix to Statistics of Deadly Quarrels.” General Systems: Yearbook of the Society for the Advancement of General Systems Theory, Vol. 6, 1961, pp. 139–187. Posthumous publication; Richardson died September 1953.

Mandelbrot, Benoît B. “How Long Is the Coast of Britain? Statistical Self-Similarity and Fractional Dimension.” Science, New Series, Vol. 156, No. 3775, 5 May 1967, pp. 636–638.

United Nations Convention on the Law of the Sea (UNCLOS), opened for signature 10 December 1982, in force 16 November 1994. Articles 3, 5, 7, 16, 33, 55–57, 121.

Fisheries Case (United Kingdom v. Norway), International Court of Justice, Judgment of 18 December 1951. ICJ Reports 1951, p. 116.

Alaska Boundary Dispute, Arbitral Tribunal, Award of 20 October 1903. Treaty of Saint Petersburg (Anglo-Russian Convention), 1825.

Pacific Islands Forum. “Declaration on Preserving Maritime Zones in the Face of Climate Change-Related Sea-Level Rise.” 51st Pacific Islands Forum, 6 August 2021.

Pacific Islands Forum Secretariat. International support for the Declaration on Preserving Maritime Zones in the Face of Climate Change-Related Sea-Level Rise: summary of state and organisational endorsements. forumsec.org/publications/declaration-preserving-maritime-zones-face-climate-change-related-sea-level-rise.

CIA World Factbook. United Kingdom — Geography. Coastline: 12,429 km.

World Resources Institute. Pilot Analysis of Global Ecosystems: Coastal Ecosystems. Coastline length derived from World Vector Shoreline database (US Defense Mapping Agency, 1989) at 1:250,000 scale. UK figure: 19,717 km.

UK Environment Agency / Coastal Groups. Shoreline Management Plans: policy framework applying 0–20, 20–50, and 50–100 year horizons with hold-the-line, managed realignment, and no-active-intervention designations.

FEMA. National Flood Insurance Program Risk Rating 2.0: Methodology and Data Sources. January 2022. fema.gov/flood-insurance/risk-rating.