For four centuries the world's sea charts carried islands that were never there. Deleting them turned out to be far harder than inventing them.
123
Islands struck from one chart, 1875
When Sir Frederick Evans revised Admiralty Chart 2683 — the Royal Navy’s chart of the Pacific — he deleted 123 islands in a single edition. Three of them were real.
Ten phantom islands at their charted positions · schematic, not survey-accurateEquirectangular · Daily Dose · Maps No. 027
01 · The purge
One hydrographer, one edition, 123 deletions
The 1864 Pacific chart was a landfill of rumour. Evans emptied it — and overshot.
For four hundred years the Pacific had been filling with islands reported by whalers, sealers and merchantmen: icebergs mistaken for headlands, pumice rafts, cloud banks, longitude errors, and the occasional outright invention. As the Royal Navy’s chief hydrographer, Evans took the 1864 edition of Admiralty Chart 2683 and cut 123 islands from the 1875 edition. By the end of that century roughly 200 non-existent islands were still printed on charts and atlases worldwide — a published estimate rather than a census. Evans’s problem was the one every editor has: he could not be sure which absences were real. Three of the islands he deleted did exist.
02 · Lifespans
A phantom can outlive an empire
How long each island lived between the day it was first drawn and the day it was last printed.
Hy-Brasil was placed west of Ireland on a chart of 1325 and was still being engraved into the 1860s — about 540 years on paper. Bermeja, charted by Alonso de Santa Cruz in 1539, was not formally searched for until 2009. At the other end, Crocker Land managed eight years. The pattern is not about plausibility; it is about traffic. Islands in busy water died fast. Islands nobody had reason to sail past lived for centuries.
03 · The lag
Disproving an island is the easy half
Every one of these was sailed through, sounded, or flown over long before it left the map.
James Weddell crossed the charted position of the Aurora Islands in clear weather in 1820 and saw open sea; they were still being printed into the 1870s. The German survey ship Valdivia looked for Thompson Island in 1898 and found nothing; it was still on maps in 1943. Charts are conservative by design — a hydrographer who wrongly erases a reef kills a crew, while one who leaves a ghost island in place merely embarrasses himself. So ghosts get the benefit of the doubt, sometimes for half a century.
04 · The last one
Sandy Island, 1876–2012
The final phantom died on a Thursday in the Coral Sea, with a multibeam echo sounder running.
The whaler Velocity reported Sandy Island in 1876; an Australian maritime directory listed it by 1879. France dropped it from its charts in 1974 and Australia in 1985 — but it survived in the global shoreline datasets that feed digital maps, so it reappeared on Google Earth and in scientific figures. On 22 November 2012 the R/V Southern Surveyor sailed straight over 19°13′S 159°56′E. The seabed never came shallower than about 1,300 metres. Google removed it four days later; National Geographic, three days after that.
05 · Stakes
What a phantom island can cost
Treaties, oil, and four years of a man’s life.
1783Treaty of Paris
The border between the United States and British North America was written to run through Lake Superior “northward of the Isles Royal and Phelipeaux.” Phelipeaux came from John Mitchell’s 1755 map, which shows four Lake Superior islands that do not exist. Surveyors in the 1820s could not find it.
2009UNAM seafloor survey
Mexico searched for Bermeja because the 2008–09 talks with the United States over Gulf of Mexico drilling rights turned on where the maritime boundary sat. At 22°33′N 91°22′W there was nothing on the seabed at all.
4Years chasing a mirage
The Crocker Land Expedition left in 1913 to reach land Robert Peary had reported in 1906. On 21 April 1914 the party saw it on the horizon. The Inuit guide Piugaattoq said it was a mirage. He was right; they got home in 1917.
What to notice on the map
01The phantoms cluster where sailing was hard and traffic was thin: the Southern Ocean, the empty mid-Pacific, the high Arctic. Nobody invented an island in the English Channel.
02Three of them sit inside national waters or on a border — Phelipeaux in Lake Superior, Bermeja in the Gulf of Mexico, Sandy Island inside the French Coral Sea. That is where a ghost stops being charming.
03The dates on each label are a lifespan, not a mistake: first charted, then last printed. The gap between “known to be false” and “removed” is usually decades longer than you’d guess.
04Sandy Island is the youngest corpse on this map, and the only one killed by a digital map. It died in 2012 because someone finally sailed the pixel.
Sources & notes
Sarah Laskow, Slate (Feb 2016)Sir Frederick Evans cut 123 islands from the 1875 edition of Admiralty Chart 2683 — three of which were real; roughly 200 non-existent islands were on charts and atlases by the end of the 1800s (that 200 is a published estimate, not a census).
Seton et al., Eos 94(15), AGU (2013) · “Obituary: Sandy Island (1876–2012)”R/V Southern Surveyor sailed the charted position on 22 Nov 2012; the seabed never rose above ~1,300 m, with soundings from 1,488 m to 2,353 m.
French Hydrographic Service (1974) · Australian Hydrographic Service (1985) · Google Maps (26 Nov 2012) · National Geographic Society (29 Nov 2012)The staggered removal dates for Sandy Island; it was first reported by the whaler Velocity in 1876 and listed in an Australian maritime directory by 1879.
Treaty of Paris, 1783 · John Mitchell, Map of the British and French Dominions in North America (1755)The US–British boundary was written to run “northward of the Isles Royal and Phelipeaux”; Mitchell’s map shows four non-existent Lake Superior islands (Phelipeaux, Pontchartrain, Maurepas, St. Anne). Surveyors in the 1820s found no Phelipeaux.
Universidad Nacional Autónoma de México (UNAM), 2009, for the Mexican Chamber of DeputiesA seafloor survey at 22°33′N 91°22′W found no trace of Bermeja, charted by Alonso de Santa Cruz in 1539 and relevant to 2008–09 Mexico–US talks on Gulf of Mexico drilling rights.
American Museum of Natural History · Peary–MacMillan Arctic Museum, Bowdoin CollegeCrocker Land Expedition, 1913–1917: four years in the Arctic chasing land Robert Peary reported in 1906. On 21 Apr 1914 the party “saw” it; the Inuit guide Piugaattoq called it a mirage, and was right. MacMillan made over 5,000 photographs.
Wikipedia / Angelino Dulcert (1325) · cartographic recordHy-Brasil was placed west of Ireland on a 1325 chart and is scarce on maps after 1865 — about 540 years of paper existence.
Wikipedia / Thompson Island (South Atlantic)Reported by whaling captain George Norris in 1825, ~70 km north-northeast of Bouvet Island; the German survey ship Valdivia searched in 1898 and found nothing; it was still printed on maps as late as 1943.
Wikipedia / Dougherty Island · Aurora Islands · U.S. Naval Institute Proceedings 66/4 (Apr 1940)Dougherty (reported 1841 at 59°20′S 120°20′W) was struck from International Hydrographic Bureau lists in the 1930s. James Weddell sailed the Auroras’ position in clear weather in 1820; they stayed on maps into the 1870s. Los Jardines, reported by Álvaro de Saavedra in 1529, was still printed as “E.D.” — existence doubtful — in 1940.
Note on the map: the coastlines are hand-simplified and schematic, not survey-accurate. Every island name, charted coordinate and date shown is real. The 123-cell diagram in section 01 marks three cells as the real islands Evans cut by mistake — which three is not recorded, so those cell positions are illustrative. Lifespans marked “~” end in a decade rather than a dated edition.