The ruins of Pavlopetri lie in about three meters of water off the shore of Ponda Beach at the western end of Vatika Bay, in the southeastern corner of Laconia in southern Greece. The site sits in the narrow sea channel between the mainland and the island of Elafonisos, a short drive from the town of Neapoli. The remains cover roughly five hectares of seabed, about the size of seven football fields, in water no deeper than four meters. Most of the stones sit at depths between one and three meters.
What remains is a city plan drawn in stone. The walls are the foundations of houses once built with mud brick and timber, and most now rise only one course high, with the best-preserved reaching three courses. Streets, courtyards, rooms, and graves remain visible from the surface on a calm day. The first person to document the site was the Greek geologist Fokion Negris, who noted submerged walls here in 1904 but moved on, more interested in rocks than houses.
In 1967, the British oceanographer Nicholas Flemming, then at the National Institute of Oceanography, came to the same shore looking for evidence of sea-level change around the Peloponnese. He recognized the walls as a prehistoric city and estimated it dated to the Bronze Age, roughly 4,000 years ago.
In July and August 1968, a Cambridge University underwater exploration group spent six weeks at the site with a core team of seven people led by R.C. Jones, including a student named Anthony Harding who wrote the first draft of the report. They worked mostly with snorkels because air tanks slowed them down, measuring every wall with hand-held tape measures across a fixed grid of squares marked by small floats. Summer winds, which the Greeks call the meltemi, kept stirring the water and stopping work. Their most unusual tool was a tethered hydrogen balloon with a remotely operated camera hanging beneath it, described in the report itself as a robot camera. The images were disappointing because the seabed was so crowded with fallen stones that the walls merged with the rubble, and the team used the photos mainly to check their hand-drawn plan.
That plan, published by the British School at Athens in 1969 under the names Harding, Cadogan, and Howell, covered an area roughly 300 meters long and 100 meters wide. It showed 15 separate buildings, five streets, two stone cemeteries, and at least 37 cist graves, small box-shaped tombs lined with stone slabs. One building, labeled number seven, had walls up to 1.4 meters thick enclosing four rooms, and the report described it as the only structure at the site resembling a megaron, the long hall with a porch found in Mycenaean palaces. Then no one returned for forty years.
In 2008, Flemming came back with Nottingham archaeologist Jon Henderson and pottery specialist Chrysanthi Gallou, and found the ruins in much the same condition as the Cambridge team had drawn them, with one change: a layer of brown sponge called Chondrilla nucula had begun growing on many of the walls.
The Pavlopetri Underwater Archaeology Project began fieldwork in 2009 as a partnership between the Greek Ephorate of Underwater Antiquities and the University of Nottingham. Elias Spondilis led the Greek side, Henderson led the British side, and the permit came through the British School at Athens.
The 2009 season used two survey tools. The first was an automated total station on shore, a surveying instrument that fires a laser at a prism held over each stone and records its position with millimeter accuracy. The second was an MS 1000, a sector-scanning sonar made by Kongsberg Mesotech that rotates in place on the seabed and sweeps a sound beam around like clock hands, building a three-dimensional picture of the walls from echoes. The Nautilus Marine Group International, a North American marine engineering company, brought the MS 1000 to test its accuracy. The site was exactly the kind of test environment the company needed: shallow, full of clear walls, and already mapped stone by stone with the total station.
The sonar found a larger city than anyone expected. The Greek and British directors reported more than 9,000 square meters of new buildings to the north, closer to shore, including at least 25 connected rooms starting about 10 meters from the waterline, a 40-meter-long street, and a trapezoidal building about 34 meters long and 12 to 17 meters wide with at least three rooms, large and central enough that the team considered it important.
In the 2010 season, from May 31 to June 26, the team finished recording roughly 40,000 square meters of known buildings and more than 10,000 square meters of newly discovered ones. That summer, a private sponsor funded a team from the Australian Centre for Field Robotics at the University of Sydney to travel with a camera system designed for the task. The Australians called it the diver propulsion vehicle. It carried two cameras 7.5 centimeters apart, about the same distance as human eyes, pointed downward inside a frame of carbon fiber and balsa wood, with an LED flash half a meter in front and another behind. The device also carried a depth sensor, a GPS receiver, and a motion sensor, so every pair of images came with an accurate record of where it was taken.
The two cameras see depth the way eyes do, a method called stereoscopic imaging that measures shapes through overlapping images. Software written at the Australian centre by Ian Mahon, Oscar Pizarro, Matthew Johnson-Roberson, Ari L. Friedman, and Stefan Williams merged those images into three-dimensional models. The Nottingham team reported recording more than half the site in full three dimensions in 2010, the first time such equipment had been used on a sunken city.
Still, a diver had to push the device through the water. The part it did not cover was waiting for the machine the title calls an underwater drone. This was an autonomous underwater vehicle, or AUV, a robot submarine that follows a programmed path without a cable and without a pilot. The Australian team used a modified Iver 2, a torpedo-shaped AUV made by OceanServer Technology, and fitted it with a stereo camera system similar to the one on the diver propulsion vehicle. The Sydney centre also operated a larger AUV called Sirius.
The water was so shallow that the drone could maintain a GPS signal while its cameras stayed close enough to the stones to photograph them, something that almost never happens in deep surveys where the robot loses satellite contact once it dives. Engineers spent the week before the 2011 season tuning its depth, steering, and image-processing software. During the 2011 season, more than 600,000 images were taken covering the rest of the known site, and the paper reported that mapping it took less than 24 hours of underwater work. Divers and drones together completed coverage of the entire city in three weeks of fieldwork, a task the Cambridge divers had spent six weeks doing by hand and in far less detail.
The drone found a map. It showed the walls, rooms, and streets with millimeter accuracy, and detected faint walls in an area the team called N, so eroded they were nearly level with the seabed and hard to see with the naked eye; divers confirmed and measured them with the total station. The images also became the raw material for a documentary. The BBC and Discovery Channel had filmed the project since 2010, and in 2011 a graphics team from Prime Focus built digital buildings directly from the survey data as it came in. The film, titled “City Under the Waves: Pavlopetri,” directed by Paul Olding, aired on BBC 2 on October 9, 2011. The University of Nottingham later reported 2.03 million viewers for that broadcast. It premiered in North America on Discovery on June 2, 2012.
The maps beneath those graphics contained a slower second story about what the sea was doing to the stones, and reading it required digging first. Dating came from pottery. In 2009, the team recovered 442 ceramic fragments from the seabed surface along with an iron nail and a chip of obsidian, the black volcanic glass used to make blades, and classified them by period. About 3 percent were from the late Neolithic, the final phase of the Stone Age. Forty percent were from the Early Bronze Age, 15 percent from the Middle Bronze Age, and 25 percent from the Late Bronze Age, with only a few pieces from the Classical and Roman periods.
That small Neolithic percentage pushed the beginning of the site further back. On October 22, 2009, the University of Southampton announced that the pottery showed people had been at Pavlopetri for at least 5,000 years, about 1,200 years earlier than previously thought, and Flemming said the city could be 5,000 to 6,000 years old. By 2010, Nottingham was describing settlement about 5,500 years ago and calling Pavlopetri the oldest known sunken city in the world.
The Early Bronze Age pottery shows who the city traded with. Storage jars carry cord-impressed and fingertip patterns, and cups, sauceboats, plates, and portable hearths match styles from the Cycladic islands, western Crete, and the northeastern Aegean. The 2010 season added roof tiles, street drains, and large intact storage vessels. Excavation began on June 2, 2011, and ran through June 22 that year. The team dug a small trench next to one of the best-preserved structures, a strongly built building at the junction of three streets with a layout different from the ordinary houses around it. Long, narrow walls beside it suggested a storeroom, and the trench filled with sherds of pithoi, the tall pottery storage jars used for grain, oil, and wine across the Aegean in the Bronze Age.
A remarkable object settled into place in the gray sand. It was a pithos jar from the early Neopalatial period, about 3,700 years old, decorated with rows of stamped seals and finger marks, still sealed with a large piece of schist, a flaky layered rock, and closely resembling jars found at Kastri on the nearby island of Kythera. A second, smaller pithos jar was found nearby, and a dark circle of organic material beside them may have been the remains of a basket or barrel. These finds connected the city to Crete. The Nottingham team reported that deposits dating to between 3,700 and 3,400 years ago showed close trading contacts with the Minoan palaces of Crete, and the team proposed two explanations: either a Minoan trading colony, the first found on the Greek mainland, or a local port so influenced by Crete that its potters made Cretan shapes from local clay.
Later studies of the site’s end, such as a 2022 pottery study by Kavanagh and colleagues, indicated the site was abandoned about 2,900 years ago, with only small numbers of later pottery sherds and no newer buildings at the site. The city was inhabited for more than 2,000 years and then deserted.
The sand that had preserved those jars intact was already moving. Divers noticed that some pithoi the sea had exposed were breaking apart quickly. That pattern pointed to the real discovery of the drone survey: the walls that had disappeared. The 1968 plan and the robot maps from 2009 to 2011 cover the same city 43 years apart. Overlaying them gave the team a before-and-after picture of the site. Henderson and Flemming presented the comparison in a paper in the International Journal of Nautical Archaeology, published on July 16, 2026, by which time Henderson had moved to the University of Edinburgh.
The comparison runs in both directions. The new surveys revealed more than two hectares of structures, mostly near the shore, that the divers in 1967, 1968, and 2008 had not reported, and their clean, unencrusted stones suggested walls that had only recently emerged from beneath the sand. The other direction is loss. A number of walls that were clear in 1968 had partially or completely eroded by 2011, and some had vanished entirely. In areas the team labeled C, D, E, and G, the sand could not be blamed for the disappearance of walls, because by 2009 those areas were exposed down to rubble on hard bottom.
The storage jars told the same story on a smaller scale. Between 2009 and 2011, divers watched pithoi emerge from the sand and then saw them break apart from currents and waves, and by later visits some had faded entirely. Conservators at the Ephorate of Underwater Antiquities in Athens assembled fresh fragments into one vessel, evidence that the breaks were recent. Repeated visits in 2017, 2020, and 2023 continued to photograph the same buildings. A rectangular building in area M, measuring 14.5 by 6.5 meters at a depth of only 1.5 meters, had lost its northeastern corner by 2023, visible in the 2010 image, and its southwestern corner had also broken apart. Building IX showed the scale of loss in deeper water. Located at 2.75 to 3 meters in the best-preserved part of the central site, it is a structure of large stones set on edge, 8 by 5 meters, and its rounded base had collapsed into its northern and southern walls, once clear in 1968, until it was barely recognizable. Black sponge covered its standing stones in 2011, thinned by 2020, and had disappeared entirely by 2023, as had stones along its western wall’s edge.
The authors listed natural causes first: sand grinding stones through wave-driven currents, limestone dissolving in seawater, and a boring sponge called Cliona viridis that burrows into rock and weakens it. They also noted that larger stones may have been displaced by anchor drag or divers, and wrote that natural and human causes cannot always be separated. Their summary was that the sea appears to be slowly excavating the city, uncovering structures and then eroding them. That process is the clear meaning of the drone survey, and it is also the point where popular versions of this story begin to drift from the actual records.
The most-viewed recent video about the city was posted on April 4, 2026, by a channel called Discovered Times, under the title “They Found An Ancient Greek City Underwater, It’s Crazy.” The video had 146,811 views when checked in late September 2026, and its description notes the use of AI-generated images and a synthesized voice. The description makes three claims. It calls Pavlopetri a complete Bronze Age city sitting intact on the seabed, older than the Great Pyramids, and says the water is so shallow that a child with a snorkel can swim over the rooftops of history. The water depth is correct, and the age claim is accurate within narrow limits. The Great Pyramid of Giza was built about 4,500 years ago, and the oldest pottery at Pavlopetri dates to about 5,500 years, though that Neolithic material made up only 3 percent of the 2009 sample, and most planned buildings belong to the Bronze Age that followed. The word “intact” is where the claim fails. No roofs survived, most walls are one course high, and the 2026 paper documents walls disappearing between surveys.
Other large videos repeat BBC phrasing, including one uploaded on May 31, 2026, to the Odyssey channel with 66,798 views, whose description matches the 2011 film, and an older Nottingham video showing the city rebuilt with computer graphics, with 248,258 views.
There are three main explanations for how a city built on dry land ended up three meters underwater, and the first is the one most videos turn to. The first theory is sudden subsidence. Pavlopetri sits on the Hellenic Arc, a curved belt of active faults across southern Greece, and the 2026 paper lists it among the forces that moved this coast. The Nottingham team leaned in this direction early after the 2010 season, when the Hellenic Centre for Marine Research surveyed the seabed between Elafonisos and Viglafia from its research vessel. The project website stated that the site appeared to have sunk in three or four separate earthquake events: the first at some point after about 3,000 years ago and the last before 1700. The idea fits the region. The same survey found that faults in Vatika Bay remain active, with terraces and sunken valleys formed by those faults. Its weakness is the ruins themselves. The 2026 paper found no evidence of a major earthquake or sudden building collapse, and that complete absence of a single disaster layer points toward something slower.
The second theory is a slow rise in sea level. When the great ice sheets melted after the last ice age, the additional water and the changing weight on the earth’s crust kept pushing local sea levels upward for thousands of years, a process called glacial isostatic adjustment, abbreviated in the paper as GIA. For Pavlopetri, the Italian geophysicists Daniele Melini and Giorgio Spada ran six different computer models of this process. All six showed sea level rising rapidly during the first two thousand years of the city’s settlement, then slowing afterward. Henderson and Flemming used those curves to estimate a rise of about 1.5 meters between roughly 5,000 and 3,000 years ago, and another meter since then, for a total of about 2.5 meters. At the time the city was founded, the sea was rising at a rate of one meter per thousand years, then slowed to half that rate later. The ruins need more than that. The lowest walls now sit about 3 meters deep, and the city may have been built 2 meters above sea level, away from breaking waves, making the total drop about 5 meters. Rising water covers half of that, and the other half needs a different cause.
The third explanation adds the possibility of land subsidence. The paper attributes the remaining 2.5 meters to ground sinking along the faults, which geologists call tectonic subsidence, a slow lowering of the crust that happens in small jumps rather than one collapse. The evidence is in the beach rocks, sands hardened into stone along the coastline, that mark where the shore used to be. Bands of these near Pavlopetri, now submerged, indicate drops of about a quarter meter at intervals of roughly a thousand years, and the paper estimates the usual downward steps at a quarter to half a meter. After that, the sea took over the work of destruction. About 3,000 years ago, water was touching the edge of the city, and the 600-meter reef that protects it has two low sections, one near shore in front of area A and the other near Pavlopetri islet in front of area N. The walls at those two locations are exactly the most eroded.
Storms provide far more force than daily waves. Winds from the south and southeast cross 120 kilometers of open water from Crete. They blow in this direction on less than 1 percent of days, but they can push waves over two meters high toward the site. A weather station at nearby Kavo Malia recorded storms of about 72 km/h on December 1 and 2, 2017, and up to 75.6 km/h between March 17 and 21, 2018.
The island of Elafonisos was connected to the mainland by a strip of land until the Roman period, described by both Strabo and Pausanias. That land bridge sank underwater around 200 AD. After that, west winds were able to push waves and sand through the gap.
Protection came in stages. In 2011, the Greek Ministry of Culture declared Pavlopetri a protected underwater archaeological site under Law 3028 of 2002, which prohibits anchoring, diving, or fishing there without ministry permission, and the local coast guard began monitoring the waters, while mobile homes that had lined Ponda Beach were moved to a square in Viglafia. In 2014, local residents asked for help from the Cultural Heritage Rescue Alliance, a Washington-based nonprofit known as ARCH. On its first monitoring day in August 2014, local residents formed a human chain along the shore. In October 2015, the World Monuments Fund placed Pavlopetri on its 2016 World Monuments Watch list, citing threats from pollution, boat anchoring, looting, and projects such as pipelines and power plants.
In 2016, buoys were placed around the site, tied to cement blocks so light that storms dragged them over the ruins, so the scheme was abandoned. In 2018, the Greek government published the site boundaries in its official gazette, and the area appeared on international nautical charts. A 40-page master plan for management followed in 2019. In 2022, Bettina Jozef of ARCH’s Greek branch said commercial ships anchoring in Vatika Bay, some illegally cleaning their hulls, were polluting waters near the ruins, and that the plan had stalled. A snorkeling trail opened the same year, created by the Peloponnese regional government and the Ephorate of Underwater Antiquities, funded by the European Interreg MED Inherit program.
The July 2026 paper sets out what comes next. Henderson and Flemming propose an artificial reef or fabric bags filled with sand, which engineers call geotextile bags, placed to break waves without changing water flow through the site, and reuse of sand dredged near Elafonisos to rebuild the eroded beach at Ponda. They also recommend a full model of sand movement in the bay and regular photographic monitoring, something one person with one camera could do by 2026. They rejected proposals for a dive park or an underwater museum with glass walls as too intrusive. None of these protections has yet been tested against a storm like those of 2017 and 2018.
Pavlopetri is a real Bronze Age city off Ponda Beach in Laconia, first noticed by Fokion Negris in 1904, identified by Nicholas Flemming in 1967, and mapped by a Cambridge team using snorkels and hand tapes in 1968. Between 2009 and 2011, the Nottingham-Greek project, with the Australian Centre for Field Robotics, surveyed the entire city using sonar, the diver propulsion vehicle, and the Iver 2 drone, which took more than 600,000 images. The pottery indicates the first inhabitants arrived about 5,500 years ago and the last left about 2,900 years ago. The most precise figure is the subsidence itself: about 5 meters, half from rising sea level and half from land sinking, with no evidence of a single great earthquake. The drone found a map, and comparing it with the 1968 plan and later images from 2017, 2020, and 2023 shows walls fading and storage jars breaking apart after 5,000 years on the seabed. The city sank slowly, and the records show the sea now taking it apart at the same slow pace.
