A team of archaeologists has published Itiner-e, the most detailed open map yet of the roads of the Roman Empire: 299,171 km of them, section by section. I mapped every one, asked how much of the network we really know, and built a travel-time model to answer an old question: how many days was it from Rome to anywhere?
Every Roman road began, in a sense, at the golden milestone that Augustus set up in the Forum. For centuries the best maps of the network were printed atlases. Itiner-e, released in 2025 by teams at Aarhus University and the Autonomous University of Barcelona with many collaborators, puts it all in one open dataset: 14,769 road sections, digitised from atlases, surveys, excavation reports, milestones and aerial and satellite images, with 299,171 km of road from Scotland to the Sahara and from Portugal to the Euphrates. That is 7.5 times round the Earth.
Each section has a name, a type (main or secondary road), a grade of how sure we are of its course, and, when known, the period it was built and the ancient itinerary it appears in. I used it to ask three things. How much of the network is known for certain? Where was it densest, and where does it run straightest? And, by building a graph of the whole network and walking it with stated speeds, how many days did it take to get from Rome to anywhere else?
Roads: Itiner-e static version 2024 (de Soto, Pažout, Brughmans et al. 2025, doi:10.5281/zenodo.17122148), CC BY 4.0, lines lightly simplified. Travel times are modelled, not recorded (see section 02). Click any road, or anywhere on land, for how long it took from the chosen start.
To turn a map into travel times I treated the network as a graph: the 10,315 points where sections meet are the junctions, and the 14,769 sections are the links between them. Wikipedia's article on Roman roads gives the speeds: people on foot covered 20 to 25 km a day, horse-drawn carts 40 to 50 km, and a relay of horses could carry a letter about 80 km in a day. I used the middle of each range, 22.5, 45 and 80 km a day.
Hills slow everyone down. Itiner-e gives every section a passability value that falls as the road gets steeper. It is not explained in the field description, but it matches Tobler's hiking function, a standard rule for walking speed on slopes, for every section with an average slope above about 3.4 degrees (median difference 0.001), and is 1 on gentler ground. I divided each section's time by it: 27% of the network is slowed. Where the roads stop at the sea, four short crossings join them, each counted as one day whatever the traveller: the Channel, the Strait of Messina, the Strait of Otranto and the Strait of Gibraltar. Then a shortest-path search from the Forum gives a time to every junction in the empire.
Days from Rome by road
Modelled travel time from the Forum to 20 cities. Lines show the range for the low and high ends of each speed. Click a city to see its route on the map.
On foot, London is 81 days from Rome (73 to 91 days at 25 or 20 km a day): 1,672 km of road, through Gaul to Boulogne, across the Channel and up from Dover. A cart would take 41 days and a courier 23. Constantinople, then still Byzantium, is 75 days, across the Strait of Otranto and along the Via Egnatia; by land the whole way, round the head of the Adriatic, it is 107.
Alexandria is the surprise. It is only 1,952 km from Rome as the crow flies, but there is no road across the sea, so the model takes you over to Greece, through Asia Minor, down through Syria and Judaea and across the Sinai: 3,807 km of road and 180 days on foot, 2.2 times as long as London. Carthage is stranger still. Rome to Carthage by road goes through Gaul and Spain, across the Strait of Gibraltar and back along the whole of North Africa: 4,490 km, 210 days on foot, for a city 590 km away.
By road alone, Carthage was further from Rome than Alexandria. Nobody went that way: they took a ship.
That is the main limit of a roads-only model. The Mediterranean was the empire's real highway, and the roads fed it. Stanford's ORBIS model of the Roman world (Scheidel and Meeks) does the job properly, with sea routes, rivers, seasons and many kinds of traveller, and it shows how often the sea was much quicker than the road. This map does something narrower on a far more detailed road network: it shows how the land itself was tied together, and where the roads alone could and could not take you.
How much of the network lies how far from Rome
Kilometres of road by days on foot from Rome (the time to the middle of each section). Hover a bar.
Only 5% of the network is within a month's walk of Rome; by courier, 64% is. The furthest point the roads reach is Murzuq, in the Fezzan in southern Libya, 266 days on foot along desert routes that Itiner-e marks as hypothetical. 4,063 km of road (1.4%) cannot be reached by road at all: Sardinia, Corsica, Crete and Cyprus, and a few short fragments.
Itiner-e grades every section by how sure we are of where it ran. Certain sections are well documented and digitised to within 50 m in mountains and 200 m in the plains. Conjectured sections are known roads whose exact course is less sure. Hypothetical ones are roads that probably existed but cannot be placed, or desert and floodplain tracks with no single fixed line. Only 8,181 km, 2.7% of the network, is certain. 90% is conjectured and 7.4% hypothetical.
Certain, conjectured and hypothetical road
Kilometres of main and secondary road by certainty. Hover a block.
Main roads make up 35% of the length, and even of these only 5.4% is certain. The certain kilometres cluster where surveys have been thorough and dry country preserves the line: Arabia, Syria, Galatia and Cappadocia, and Judaea have the most. The longest single section, Dedris-Cydamus Ghadames in the Libyan desert, runs 493 km and is hypothetical. Not one kilometre of the modelled route from Rome to London runs on a section graded certain.
None of this is a weakness of the dataset. It is the dataset being honest. A road can be known from milestones and itineraries long before anyone can say which field it crossed, and Itiner-e records that difference instead of hiding it. For a travel-time model it matters less than it sounds: an error of a few hundred metres in a road's course hardly changes a journey of weeks. Leaving out every hypothetical section changes the times to Constantinople and Alexandria by less than a day.
Few roads can be dated. 3,176 sections, 19% of the length, carry the name of the emperor or magistrate under whom they were built or a start date, mostly from milestones and inscriptions. That is a record of where inscriptions survive and have been studied, not of when the network grew, so I have not tried to animate it. Within it, Claudius has the most kilometres (5,925 km), much of it in Gaul and in Lycia, where a monument at Patara lists the roads built under him, then Trajan (4,067 km) and Hadrian (3,932 km).
Kilometres of road by the emperor or magistrate named as builder
The construction period field, grouped by name (3,191 sections have one). Hover a bar.
The ancient itineraries tell a different story. 38% of the length appears in at least one: 75,486 km on the Tabula Peutingeriana, the medieval copy of a late Roman road map, and 68,861 km in the Antonine Itinerary, a list of routes and stopping places. Among the named roads the longest in the data is the Via Egnatia across the Balkans (1,165 km), then the Via Hadriana in Egypt and the Via Claudia Augusta over the Alps.
To find the longest straight runs I joined sections into continuous lines through junctions where the road carries on within 20 degrees of straight, then looked for stretches that never stray more than 200 m from a straight line, the accuracy of a certain section in the plains. Long conjectured lines drawn with only a few points are left out, because they are straight only because they were drawn that way. 46 stretches run straight for 20 km or more.
The longest straight runs
Stretches that stay within 200 m of a straight line. Click a row to see it on the map.
The longest, 56 km, crosses the chalk plain of Picardy between Saint-Quentin and Amiens, a line a modern road still follows. Then come the Via Postumia in the Veneto (49 km) and the Via Appia across the Pontine plain towards Terracina (42 km), the longest straight on a section graded certain. These are lower bounds: a tolerance of 200 m is strict, and the famous long alignments of Roman Britain and Gaul bend just enough, at river crossings and ridges, to break into shorter runs.
Measured as kilometres of road per 1,000 km² in the provinces of about AD 117 (from the Digital Atlas of the Roman Empire, with Italy as its eleven regions), the densest network is in Italia, regio I (Latium et Campania), the country round Rome: 144 km per 1,000 km². Outside Italy, Cyprus (131), Iudaea (130) and Lycia et Pamphylia (129) lead. Italy as a whole has 70, against 58 for all the provinces together.
Road density by province
Kilometres of road per 1,000 km² of land, provinces of about AD 117. Click a row to see the province on the map.
The thinnest networks are where the province was mostly desert or held only briefly: Egypt (24), Crete and Cyrenaica (20) and Trajan's short-lived Armenia and Mesopotamia (3.8), with Mauretania Tingitana in Morocco (23) not far behind. Within Italy, Umbria (regio VI) is surprisingly thin (31), a reminder that coverage is uneven. Density also follows research. Judaea, one of the most intensively surveyed landscapes in the world, has 39% of its roads graded certain, against 2.7% for the network as a whole. 5.3% of the network, 15,732 km, lies outside the provinces altogether: in Scotland, beyond the Rhine and Danube, and along desert routes into the Sahara.
Data. Itiner-e static version 2024 (Zenodo record version 1.3), 14,769 road sections with their geodesic length, average slope and passability. Section ends that meet within 1 m are one junction; the network is clean at that tolerance. Dates come from the construction period field, else the lower date; 9999 means not given.
Travel times. Time on a section is its length divided by the speed (on foot 22.5, cart 45, courier 80 km a day, the middle of Wikipedia's ranges) and by the section's passability. Each sea crossing counts as one day; the Euripus bridge to Euboea counts as road, and the Bosporus and Hellespont ferries are already in Itiner-e. Off the network, the last stretch to any point is counted at half the road speed, up to 50 km from a road. There are no rests, seasons, rivers, sea routes or borders. These are modelled times for a steady traveller, not records of real journeys.
Straight runs and density. Straight runs are found on continuous lines joined through junctions (deflection under 20 degrees), simplified with a 200 m tolerance in a local equidistant projection; conjectured stretches need at least one point every 2 km. Density uses DARE's province outlines, which are themselves approximate, and land area in an equal-area projection.
What this cannot say. Most of the network is conjectured, and only a fifth is dated, so the map shows the roads that existed at some point between about 300 BCE and 300 CE, not the network at any one moment. Coverage is uneven: some regions have been surveyed field by field, others only from old atlases. And the travel model leaves out the sea, which for most long journeys was the quicker way.