Skip to the story
mariuscomper.uk Română
AD 105
water at about 35 m above sea level

Trajan’s Bridge · Drobeta-Turnu Severin and Kostol

Walk across Trajan’s Bridge

By AD 105 the Romans had finished a timber bridge on twenty stone piers, from Drobeta, on today’s Romanian bank, to Pontes, on the Serbian one. The army and its supplies crossed on it. Hadrian took the timber off, and the piers crumbled over the centuries that followed. Today the stumps of those in the river lie under the reservoir of the Iron Gates II dam. We rebuilt the bridge in the real valley, with the Danube lowered to a level from before the dams.

Scroll to cross with the legion.

The gate at Drobeta

The bridge started at the camp of Drobeta. On Trajan’s Column an arched gate stands at its end. Its form comes from the Column; its size is an estimate.

One pier still stands at the Romanian end, in the grounds of the Iron Gates Region Museum. From there to the pier on the Serbian bank at Kostol is 1,074 metres in a straight line.

Over the river

Cassius Dio, a Roman senator writing a century later, marvelled above all at the river:

“...in a river so deep, in water so full of eddies, and on a bottom so muddy...”
Cassius Dio, Roman History 68.13, translated by Earnest Cary (1925)

Apollodorus of Damascus picked the point where the valley was narrowest. There the current ran even faster.

Twenty piers

“...twenty piers of squared stone one hundred and fifty feet in height above the foundations and sixty in width, and these, standing at a distance of one hundred and seventy feet from one another, are connected by arches.”
Cassius Dio, 68.13

A Roman foot was about 30 centimetres. But Dio’s numbers don’t match what the divers found. The 2003 survey, as reported in Ion Šerban’s published abstract, gives piers of 14.7 by 10.2 metres, narrower than Dio’s 60 feet. The model uses those sizes.

At Pontes

On the south bank, in Moesia, the bridge ran into the fort of Pontes, near today’s village of Kostol. Bricks found there carry the stamps of the legions IV Flavia Felix, VII Claudia, V Macedonica and XIII Gemina.

Behind the column lies Dacia. Trajan’s second war against Decebalus began the same year, 105.

What is certain and what is rebuilt

In white is what is known: the two ends of the bridge, how many piers there were, their size as reported from the underwater survey, and the fact that the bridge was timber on stone piers. Only the Column shows that last point.

In orange is what we rebuilt: the exact shape of the arches, the posts, the deck, the parapets, the gates, the ramps. Even the deck height is uncertain. Sources give between about 9 and 19 metres above the water. The model puts it at 15.

Relief on Trajan’s Column: flattened timber arches of several rings on stone piers, with an X-braced parapet above; below, the emperor with men in togas.
The bridge on Trajan’s Column, Rome, AD 113. Conrad Cichorius, plate LXXII, 1900, public domain.

Taken apart

Hadrian removes the timber

Trajan’s successor did not want an open road south. Dio writes that Hadrian

“...was afraid that it might also make it easy for the barbarians, once they had overpowered the guard at the bridge, to cross into Moesia, and so he removed the superstructure.”
Cassius Dio, 68.13

The year is unknown. The story that Hadrian pulled it down out of envy of Apollodorus does not come from Dio: he records their quarrel, but not in connection with the bridge.

The piers remain

In Dio’s day, about a century later, the piers still stood, with nothing on them:

“...merely the piers are standing, affording no means of crossing, as if they had been erected for the sole purpose of demonstrating that there is nothing which human ingenuity cannot accomplish.”
Cassius Dio, 68.13

Marsigli surveys the ruins

Over time the piers crumbled almost to the water. In 1689 Luigi Ferdinando Marsigli, an officer in the Austrian army, surveyed the site and published his plates in 1726. He put the bridge’s length at about 3,010 feet.

Watercolour from 1726: the bend of the Danube at Severin, with coloured banks.
The Danube at Severin, watercolour from Danubius Pannonico-Mysicus, 1726. Luigi Ferdinando Marsigli, public domain.

A dry summer

In the 19th century, in a summer of very low water, all twenty piers could be seen. The English Wikipedia gives the year as 1858, the Romanian one as 1856. We found no period source, so we give both years.

Around 1875 a travel guide, Murray’s, counted thirteen piers still visible and gave the depth of the water as 18 feet, about 5.5 metres.

Fewer and fewer

In 1906 the Danube Commission destroyed two piers that were in the way of shipping. In 1932 sixteen remained under water. In 1982 archaeologists found only twelve. The river probably took the other four.

Which piers disappeared is not known. So the model keeps all twenty as stumps.

The water

The Iron Gates II reservoir

The Iron Gates II dam, about 68 kilometres downstream, was finished in 1984. The water here rose. In 2002 the water at the Turnu Severin gauge stood on average at 41 metres above sea level. The model uses that as today’s level. We found no reading from before the dam, so we don’t know exactly how much it rose. The model’s 35 metres is an estimate.

The marks on the water show where the piers would lie beneath, if they were evenly spaced.

On the bottom

In 2003 the Serbian Institute for the Protection of Cultural Monuments ran a sonar vessel over the whole site of the bridge, funded by Serbia’s Ministry of Culture. According to reports, it found dressed stone, planks and piles. That survey is where the model’s pier sizes come from.

We found no newer published survey.

What you can see today

Only the two piers on the banks. The Romanian one stands inside a concrete enclosure, built to protect it from the reservoir.

Photograph: the ruin of a stone and brick pier, a few people high, surrounded by a concrete wall.
The pier at Drobeta, 2009. Dan Mihai Pitea / Wikimedia Commons, CC BY-SA 3.0 RO.

Look around yourself

Pick an era, then turn the scene with the mouse or a finger. The wheel, a two-finger pinch or the + and − buttons zoom in and out.

How long was the bridge?

The textbook number is 1,135 metres. We found no survey behind it. It seems to include the ramps on the banks, which are gone. What can be measured today is the distance between the two surviving bank piers: 1,074 metres on OpenStreetMap. The 2003 underwater survey gives 1,097.5 metres over the water.

MeasureValueWhere from
Between the bank piers1,074 mOpenStreetMap positions, measured by us measured
Over the water1,097.5 mThe 2003 survey, from Šerban’s abstract (2009) reported
With the ramps1,135 mThe usual number; no survey source found
Marsigli, 1689≈ 3,010 feetLetter to Montfaucon, via Britannica, 9th ed.
Spacing of the piers170 feetDio. In the model, 56.5 m from one pier’s axis to the next rebuilt
The piers14.7 × 10.2 mThe 2003 survey, from the abstract. Dio gives 60 feet wide, about 18 m reported
Deck above the water≈ 9–19 mThe sources disagree. In the model, 15 m rebuilt

It has long been called the longest arch bridge in the world for a thousand years. Constantine the Great’s bridge, also over the Danube, further downstream, may have been longer. Much less is known about it.

What is rebuilt

The place and the terrain

The bridge’s axis joins the two bank piers, placed from OpenStreetMap. On the same 1930s Romanian military map we used to rebuild Ada Kaleh, the ruin on the Romanian bank is marked “Piciorul Podului Traian”, the foot of Trajan’s Bridge. The terrain is the Copernicus GLO-30 model, the same as for Ada Kaleh, which lies about 20 kilometres upstream. The hills are today’s.

The water

Today’s level, about 41 metres, comes from the gauge at Turnu Severin: its zero is at 34.1 metres, and in 2002 the water stood on average 7.05 metres above it. The 1998 Romanian–Yugoslav convention requires the water at the town to stay below 42 metres up to a flow of 10,600 m³/s. For Roman times and for the years before the dam we found no level at all. So the model sets the river at 35 metres and says “estimated level” wherever the water is below today’s. The depth of 5.5 metres is Murray’s.

The bridge

Twenty evenly spaced piers between the two on the banks, sized as reported from the 2003 survey. The positions of those in the river are therefore interpolated. The arches, posts, deck and parapets follow the Column: flattened arches of several timber rings tied together, and an X-braced parapet. The model draws them with three rings. The number of rings, the height of the arches and the width of the deck, 14.5 metres, are our choice. The order in which the timber came off is not known. Nor is the height of the stumps.

The column of soldiers is only an illustration. It does not show any particular unit or any particular day.

Further

Sources

The first version of this page, from September 2026, had a pier-by-pier drawing of the bridge and a game about Hadrian’s decision. It used Dio’s 60-foot width. Here we moved to the sizes reported from 2003, the measured distance between the bank piers and the real terrain.