For more than a century, maps of the Coral Sea showed an island at 19°13′S 159°56′E, between the Chesterfield Islands and Nereus Reef. The whaling ship Velocity reported it in 1876, after a voyage in the Pacific: heavy breakers, then sandy islets. Charts of the day listed every unconfirmed hazard as a precaution, the charts put the letters ED beside it, existence doubtful, and kept the island anyway. It was removed from French charts in 1974 and Australian charts in 1985, but it had been copied into a global digital coastline, and the copies kept it on maps into the twenty-first century. On 22 November 2012 the research ship R/V Southern Surveyor sailed to the position. The sea floor never rose above 1,300 metres. There was no island. This page is the chart. Set course, and it will correct itself.
This chart is a copy of the copies that kept the island. It shows Sandy Island where the data said it was, dashed, with the letters ED, existence doubtful, which is what the real charts wrote. The soundings around it run from 1,488 to 2,353 metres. A chart with an island at that spot would not show those numbers. The vessel is standing by; set course for the charted position.
The vessel is on the track. The sounder runs; the sea floor stays deep; nothing rises to meet the hull.
ED stands for existence doubtful, the mark hydrographic offices put beside anything reported but never confirmed. Sandy Island carried it for most of its life on the charts. The mark was the chart being careful. The error was in the copies that dropped the mark and kept the island.
The log: sailed to 19°13′S 159°56′E. No island. The sea floor never rose above 1,300 metres. The chart is corrected.
The chart believed for 130 years, and the belief was the only thing that existed. This is the record of how the chart was corrected, and how it kept the island long after the correction.
The island entered the record on the word of one ship. The whaling vessel Velocity came home from a voyage in the Pacific and its master reported two unusual things: heavy breakers, and sandy islets. An Australian maritime directory for 1879 placed the islets along the meridian 159°57′E, between latitudes 19°7′S and 19°20′S. It was standard practice to chart every possible hazard; a chart that omits one is the chart that loses a ship. So the island was charted.
A different Sandy Island had appeared on Captain Cook’s 1774 chart of New Caledonia, four degrees of longitude further east. Whether the 1876 report was meant to correct that older position is not clear from the record. What is clear is that the new position was never seen again.
Charts that could not confirm it put ED beside it, existence doubtful. The sea floor where the island should have been was measured at 1,488 to 2,353 metres below sea level, which is deep water, and no survey found a reef. The French hydrographic service removed the island in 1974, after a flying recognition campaign; the Australian Hydrographic Service removed it in 1985. The finding was passed to the other national hydrographic services.
Removed from the charts, the island stayed in the data. It had been digitised into the World Vector Shoreline Database, a global coastline set built from older maps, and errors in that set travelled into everything copied from it: bathymetry, satellite land masks, the maps people actually used. Google Maps showed it until 26 November 2012. Some global data sets reported an elevation of one metre over the position, because a land mask had been laid over the digitised coastline. The mask turned an absence into a shape.
Amateur radio operators on a DX-pedition said publicly in April 2000 that the island was not there; it appeared on some maps and not others, including the 1999 Times Atlas of the World. On 22 November 2012, scientists aboard the R/V Southern Surveyor, out studying plate tectonics, noticed the discrepancy and sailed to the position to look. The sea floor never got shallower than 1,300 metres.
“The ocean floor didn’t ever get shallower than 1300 metres below the wave-base.”
On 29 November the National Geographic Society struck Sandy Island from all its map products. “Full evidence has finally been presented,” its geographer said. “Sandy Island has now been officially stricken from all National Geographic map products.”
A possible explanation is pumice sea rafts. Underwater volcanoes throw out masses of floating pumice, and a drifting raft can look like breakers and islets from a ship. A study of a 2004 eruption near Tonga found pumice rafts from it passed within 20 kilometres of the Sandy Island position about 200 days after the eruption. The chart was doing its job when it kept a hazard it could not confirm, and said so. The copies that kept the island and dropped the mark were not.
This chart is a copy of the copies, and it believes again. Sail the course again and it will correct itself again. That is how a non-island lasted a century: every new chart was printed from the old one, until a ship went and looked.
No voyage is involved. This page is a chart: the position, the dates, the soundings and the ED mark are the record’s; the course and the vessel are this page’s own. The real voyage was the Southern Surveyor’s, on 22 November 2012.
Sources: Wikipedia, “Sandy Island, New Caledonia” (fetched 6 August 2026); Wikipedia, “Phantom island”. The chart is this page’s own drawing; the position, the soundings (1,488 to 2,353 metres), the ED mark, the 1879 directory entry, the 1974 and 1985 removals, the World Vector Shoreline Database, the April 2000 DX-pedition, the 2012 voyage and both quotations are in the sources above.