A research ship cruised through the Coral Sea, east of Australia, bearing down on Sandy Island. The digital scientific databases used by the researchers showed the island to be 15 miles long, north to south, and about three miles wide. Manhattan-sized.

But when the ship reached the place where the island should have been, the researchers saw only open ocean. The water was nearly a mile deep. Sandy Island simply wasn’t there. Or, it turned out, anywhere.

How could an island supposedly discovered in 1876, and appearing on many maps ever since, vanish? Did it sink beneath the waves like the mythical Atlantis? Or was it always a figment of some mariner’s imagination?

The bizarre and complicated story of ghostly Sandy Island is a cautionary tale about what we know and don’t know in the 21st century — and how, even with satellite technology and modern surveying instruments, the ocean can still spring a surprise.

Last October, Maria Seton, a young scientist at the University of Sydney, led a 25-day expedition to the Coral Sea aboard the Australian national research vessel RV Southern Surveyor. The researchers wanted to understand the tectonic evolution of that corner of the Pacific. They gathered magnetic and gravity data to map the sea floor and collected rock samples from the bottom at depths up to two miles.

They noticed that multiple scientific data sets, including the General Bathymetric Chart of the Oceans, showed Sandy Island clearly in a remote area west of New Caledonia. But the chart used by the ship’s master indicated only open water. Seton and her fellow sailors realized something didn’t add up.

“We had a cached version of Google Earth for the area — we had no Internet — and saw that the ‘island’ was depicted as a big black blob. This also made us very suspicious,” she said.

Seton’s “undiscovery” of the island prompted a story in the Sydney Morning Herald that went viral. This was big news in the world of cartography; experts were puzzled, and some wondered if Sandy Island had been eroded away by the waves, like some ephemeral coral atolls. Google and National Geographic quickly removed Sandy Island from all of their maps.

Seton, meanwhile, dug into the mystery and has now published an obituary of Sandy Island in EOS, the journal of the American Geophysical Union.

Her research showed that the island appeared on the 1908 edition of a British admiralty map, which indicated that Sandy Island had been discovered in 1876 in French territorial waters by the whaling ship Velocity. The location and shape of the island on the 1908 map corresponds to what can be seen in the modern, erroneous databases.

The island was repeatedly “undiscovered” over ensuing decades, but it remained a shadowy presence in the cartographic world. Some maps labeled it “ED,” for “existence doubtful.” French hydrographic maps deleted Sandy Island once and for all in 1974.

But the island kept popping up in other places. The island was clearly marked, for example, on a 1982 U.S. Defense Mapping Agency map. “Ile de Sable,” it says, giving the French name. There’s a cryptic annotation: “Reported 1876. Reported to be about 4 miles east, 1968.”

Seton’s research pointed her to the World Vector Shoreline Database (WVS), developed by the U.S. military. The database converted old, hard-copy charts to a digital format. But there were errors — perhaps decades old — lurking in the new data set.

“[I]nconsistencies in this data set exist in some of the least explored parts of our planet, a function of both human digitizing errors and errors in the original maps from which the digitizing took place,” Seton wrote in EOS.

The errors then migrated to other databases used by scientists such as the Global Self-consistent, Hierarchical, High-Resolution Shoreline Database (GSHHS), Seton found.

Christine Phillips, spokeswoman for the National Geospatial Intelligence Agency, the successor to the Defense Mapping Agency, said Sandy Island hasn’t been on military maps for many years, though she was unable to specify when the island was removed from the charts.

Modern cartography is far removed from the era when maps went blank around the edges or carried the warning Here Be Dragons. But experts in cartography say that the craft is, like any other human endeavor, vulnerable to error. The more information we assemble about the world, the more opportunity we create for making a mistake.

Moreover, appearances can be deceiving: Just because a map looks professional, and just because a digital map may have impressive bells and whistles, doesn’t mean that the underlying data has been scrubbed of errors, said David Titley, a retired rear admiral who served in the Navy for 32 years and served in the position of oceanographer and navigator.

“When we look at these computer displays, with the three-dimensional imagery and colorized, it can give you a sense that we know more than we do,” Titley said. “A lot of people in the Navy don’t always understand the difference between having a chart and having the survey data that formed that chart.”

Another key lesson: Although the planet has been studied from space and, in some places, charted all the way to the bottom of the deepest ocean trenches, “very little of it is surveyed to modern hydrographic international standards,” said Phillips, the geospatial intelligence agency spokeswoman.

It’s because of lingering uncertainties and misapprehensions that the U.S. Navy still has seven vessels that survey the oceans, Titley said.

Navigational errors can be catastrophic, as the crew of the nuclear attack submarine USS San Francisco discovered on Jan. 8, 2005. The submarine was cruising at full speed, more than 500 feet below the sea near Guam, when it slammed into an underwater mountain. Many sailors were injured, and one died later. The submarine nearly sank. Repairs cost millions of dollars.

The place where the seamount was located was marked on one map with nothing more than discolored water, Titley said. The Navy, alarmed by the accident, realized that it had a mapping problem and set out to correct the errors.

“We would find these underwater features, these underwater seamounts. They either were simply not on a chart or were misplaced by several nautical miles. Or were a significantly different depth than what had been charted,” Titley said. “It’s a really big ocean, and we certainly don’t know everything.”

But back to the central mystery of how Sandy Island came into being. Was it merely imagined? Maybe not. It’s possible that what the whaling vessel saw in 1876 was a floating raft of stone — a “pumice raft.”

Pumice is a frothy, light rock produced in volcanic eruptions. Huge mats of pumice can float on the ocean before eventually breaking to pieces.

“In volcanic terrains you can get islands that are temporarily there and then disappear. I have seen floating mats of pumice that form and drift around for several years after a marine volcanic eruption,” said Bruce Molnia, a research geologist at the U.S. Geological Survey.

In her EOS article, Seton notes that an eruption near Tonga in 2001-2002 produced a pumice raft that traveled about 2,000 miles toward Australia, and it passed within 13 miles of the location of “Sandy Island” — the island that never was.