A 6,500-year-old village is falling into a lake, piece by piece
At Sultana, on the shore of Lake Mostiștea in south-east Romania, Copper Age people built their houses on a spur of the terrace. When the first archaeologist came, in 1923, the water had already broken the mound in half; before that it had been at least 130 metres long. In 2019 it was under 34. What you see here is the real ground, measured by an airborne laser scan.
Scroll to see what is left, what was dug and how much the water has taken.
Marius Comper · 28 September 2026
What is left: a mound 30–40 metres across
The white line traces the top of the mound as it appears in the laser survey. Its highest point is 46.7 metres above sea level; the terrace around it lies at 41–43. The mound is a natural spur of the bank; the village layers add no more than 4 metres to it.
How big it is depends on where you decide the slope begins: between 33 and 43 metres along its long axis. The archaeologists gave 30 × 35 metres in 2001 and under 34 in 2019.
By 1923 it was already broken in half
Ioan Andrieșescu, the first to dig here, found the mound in 1923 “broken almost in half and continually collapsing on the side facing the lake”. What remained was a platform 25 metres wide and at most 71 long. Before it collapsed, he writes, the mound had been at least 130 metres long, from north-east to south-west.
The dashed line is 130 metres long. We started it at the south-west end of today's mound; where the mound really began we don't know. Its north-east end runs out over the cliff and the water.
Andrieșescu dug 290 square metres right along the collapsing edge, because that part “would have fallen within a few years”.
The bank
Every year, part of the bank collapses
The water was eating the bank before that: Andrieșescu saw the mound collapsing towards the lake as early as 1923. Today's lake is the Iezer reservoir, whose dam, between Dorobanțu, Mânăstirea and Frăsinet, went into service in 1974. The national record links the erosion to the raised water level.
“Every year, portions of the terrace on which the settlement lies collapse into the lake.”
On aerial photographs from 1972, 1986 and 2003, a team working with the archaeologists found that erosion had affected 70% of the settlement in 30 years. In one sector, between 2001 and 2005, the bank retreated by about 4 metres.
There was a lake here then, too
Drill cores, radiocarbon dates and first pollen analyses show there was a lake at the foot of the mound: some of its sediments date from the village's time and are rich in goosefoot pollen. Goosefoot still grows on the reservoir's shore.
The difference is the level: the reservoir raised the water up to the cliff on which the village stands.
The digs
Almost the whole mound was dug. The results were never published
Between 1974 and 1983, a rescue excavation examined “almost entirely” what was left. Later reports say its results “remained unknown”: all that is known is that nine small houses of one phase and two of another were found. The dig covered almost the whole surface, but only the upper layers: the oldest ones, at the base, were left untouched.
The ochre rectangles are the trenches dug from 2001 to 2021, drawn from the plan published in 2025. The long trench to the south-west, SII, dates from the 1970s and still shows as a cut in the ground.
Where they hurried to dig before the bank collapsed
Trench SI is the closest to the cliff. In 2005 it had to be reopened because the edge had come 4 metres closer; in 2006 it was extended across the whole width of the mound, as an area “in danger of collapsing into the lake”.
There they found house L5, burnt, with two rooms. The report says its size could not be established.
The houses
House L2: a burnt room 7 by 7 metres
In the south-west corner, the archaeologists found house L2 in 2003. Its surviving room measured about 7 × 7 metres. The walls were posts set in the ground, covered with a thick layer of clay. The house burnt down. Inside: more than 50 whole or restorable pots, about 31 clay weights and a gold pendant.
The ochre square is the room they found. The grey hatched volume only shows the size of a whole house, modelled on houses at other Gumelnița sites. How tall the walls were and what the roof looked like is not known from this site.
About 50–100 people
The authors of the 2025 study find room for 5–10 houses at a time on the surviving part and assume as many again on the eroded part. In all, 10–20 houses of five people each, so at most 50–100 people at once.
The village had a palisade around 4450 BC. Around 4400 BC the palisade was given up, and the village spread south-west onto new terraces. House L5 was in use around 4500 BC, house L2 around 4200 BC.
What we don't know
How much of the village is really known
Of a mound at least 130 metres long, a piece 30–40 metres across remains. In the 1970s the upper layers of that piece were almost completely excavated, and the plans of those excavations are unpublished. The archaeologists could locate precisely only eight of more than 15 reported houses.
What the lake has taken can no longer be dug. What was dug without being published may still be found in archives.
Turn the mound
Drag to turn it, pinch or use the buttons to zoom. Each layer switches on and off separately.
What is the Sultana–Malu Roșu tell?
It is a mound made of the remains of a Copper Age village of the Gumelnița culture, 300 metres north-east of the village of Sultana (according to the National Archaeological Record), Mânăstirea commune, Călărași county, on the south shore of Lake Mostiștea. Archaeologists call it a tell: a place where people rebuilt their houses many times over the remains of the old ones. Here the village layers are up to 4 metres thick and span three phases of the Gumelnița culture.
The site was found in 1923 and was the first site of the Gumelnița culture to be excavated. The Gumelnița culture as a whole, with all its villages on the Lower Danube, is on the page The hills made of houses.
The mound from above, north up, 160 metres across. White: the top of the mound in the laser survey, at the 44.5-metre contour (measured). Ochre: the excavation trenches (excavated). Dashed: the length of at least 130 metres before 1923, from a starting point we chose (inferred). Orthophoto: Stal et al., Zenodo, CC BY 4.0.
How much of the settlement has Lake Mostiștea destroyed?
Every number below comes from archaeologists who worked here. None was measured the same way from one year to the next, so each is given with its date and author.
The mound had “an elongated shape, NE–SW, of at least 130 m” (I. Andrieșescu, 1926).
The mound is “broken almost in half and continually collapsing”; the remaining platform is 25 metres wide and at most 71 long (I. Andrieșescu, 1926).
On aerial photographs, erosion affected 70% of the settlement (excavation report for 2007).
“Most of the settlement has been eroded”; a remnant of about 30 × 35 metres is left (report for 2001).
In one sector, the bank retreated by about 4 metres (report for 2005).
About 35–40 metres along the long axis (Andreescu and Moldoveanu, cited in 2022).
Under 34 metres (Covătaru and colleagues, 2022).
On the cliff below the site, repeated drone surveys show material loss (Stal and colleagues, 2022).
Romania's National Archaeological Record states: “Because of erosion caused by the raised water level of Lake Mostiștea, over the last 20 years the tell has been destroyed for the most part” (translated from Romanian). It lists the site as “gravely affected”, dated 23 December 2009.
The terrace around it is losing ground too. Comparing the 1791 map with 2018 imagery, Covătaru and colleagues found that north-west of the site the terrace edge has retreated by 200 metres. They also write that on the lake's west shore, next to Sultana village, the land has retreated by more than 200 metres since the damming of the river began in 1970. The Iezer dam, which holds the water at Sultana, went into service in 1974, according to Călărași county's risk plan.
What was found at the Sultana excavations?
Ioan Andrieșescu dug here in 1923; Barbu Ionescu of the Oltenița museum made test pits in 1958–1974; Constantin Isăcescu of the Giurgiu museum ran a rescue excavation in 1974–1983; from 2001 a team led by Radian Andreescu of the National History Museum of Romania worked with the Lower Danube Museum in Călărași. The last dig on the mound, in 2021, was by Kiel University and the Research Institute of the University of Bucharest.
The trenches on the mound. Müller and colleagues, 2025, fig. 3 (graphics: C. Covătaru), CC BY 4.0. We placed the trenches in the 3D view from this plan.
Besides houses, the excavations found a defensive ditch up to 5.9 metres deep and an earth bank about 3.4 metres high on the south-west side, a palisade that was abandoned and rebuilt, whole pots, clay weights and a gold pendant. The village cemetery lies 200 metres to the west, on the terrace.
What were the houses at Sultana like?
The plan below marks eight houses that the archaeologists could locate from the records. Names ending in “/75” come from the 1970s excavations. It is a diagram: the rectangles mark the place, not the size.
Houses of the A2 phase of the Gumelnița culture. Müller and colleagues, 2025, fig. 2, after Lazăr and colleagues, 2018, CC BY 4.0.
Two houses have measurements in the reports: the surviving room of house L2, about 7 × 7 metres, and a burnt house on the south-east edge of the settlement, found in 2007 and measured in 2010: 6 × 8 metres. For L5 and L7 the reports say their size could not be established. House L2 had walls of posts set in trenches and covered with a thick layer of clay. For L5 the report names the wall technique ceamur (in the Romanian dictionary, earth mixed with straw).
What is measured, what was dug and what is assumed
Measured today: the ground and its colours, from the laser scan (1 metre) and the drone photographs (0.5 metres) published by Stal and colleagues. The white line is the top of the mound, traced by us in these data.
Excavated here: the trenches, drawn from the 2025 plan, with an error of about 2 metres; the room of house L2, placed in trenches SX–SY where it was found; the location of house L5 in trench SI.
Inferred: the 130-metre line. Andrieșescu gives the length; the starting point is our choice.
Illustrative: the house volume. Its shape and height are taken from houses at other Gumelnița sites.
The plan of the other houses is a diagram without a scale, so their exact place on the ground cannot be fixed. For the shore of 1791, 1923 or 1972 there is no map that can be laid precisely over today's survey. And the 130 metres is a lower limit, given without a width, so the share that has been lost cannot be calculated.
