This new study reveals the structure of this network for the first time, offering a data-driven perspective on infrastructure persistence that remains highly relevant today. It found that the Romans left a permanent imprint on the modern world: a correlation between the density of ancient Roman roads and major modern roads, that persisted until major historical shifts driven by the industrial revolution. This persistence was shaped by highly local conditions, including topography, river corridors, the survival of cities, and varied institutional development.

The study challenges the long-held assumption that Roman roads were uniformly straight. While Roman engineers did favor straight roads when the terrain permitted, their routes were overall less straight than modern major roads and they had to negotiate local topography. However, the Romans were not sharply constrained by the terrain; they skillfully engineered roads in less favorable landscapes using terracing, rock cutting, and retaining walls. Because of this, the absolute majority of Roman roads maintained a slope of less than 16%, ensuring they remained suitable for animal-driven carts and wagons. Interestingly, modern roads are slightly more sensitive to harsh topography, showing a stronger negative correlation with high elevation than ancient roads.

The rigorous network analysis debunks the famous proverb that "all roads lead to Rome".  Rome was spatially an appendix on the empire's terrestrial transport map. Instead, the study reveals that the late antique capital Byzantium was more central to empire-wide land mobility. Furthermore, the analysis proved that provincial capitals consistently served as key crossroads mediating regional overland communications. The consistent, high centrality of provincial capitals and main roads implies that the Romans were capable of highly sophisticated, system-level decision-making. This development of infrastructure to organize, administer, and control newly conquered territories opens fascinating new questions about exactly how the Romans planned, mapped, and executed these decisions on a continental scale for the first time in human history.

Contact: 

Tom Brughmans (Aarhus University, Aarhus, Denmark)

Email: t.b@cas.au.dk

Adam Pažout (Universitat Autònoma de Barcelona, Barcelona, Spain)

Email: Adam.Pazout@uab.cat 

Pau de Soto (Universitat Autònoma de Barcelona, Barcelona, Spain)

Email: paude.soto@uab.cat

About the research result

Studietype:
Åben dataindsamling og kortlægning inden for klassisk arkæologi og oldtidshistorie

Funding:
Danmarks Frie Forskningsfond (DFF) – Sapere Aude forskningslederstipendium til projektet MINERVA; Carlsbergfondets Young Researcher Fellowship til The Past Social Networks Project; Danish National Research Foundation (DNRF) Centre of Excellence for Urban Network Evolutions (UrbNet) samt Viator-e-projektet, som er finansieret af MCIN og af FEDER.

Link to the  publication:
https://doi.org/10.1038/s41467-026-75453-3

See more:
https://youtu.be/OTSe7MsJXbo
Kreditering: © Itiner-e, Artas Media

Link to online Atlas: https://itiner-e.org/ 

Developed by the University of Southern Denmark, UCloud AI is integrated into the national research platform UCloud and runs on the AI supercomputer BITTEN located in Sønderborg. It provides access to advanced language models through a familiar digital workspace, allowing users to work with AI alongside their existing data, storage, and applications. The service builds on the existing collaboration between SDU, Aarhus University and Aalborg University in the DeiC Interactive HPC consortium.

The platform is designed to support a wide range of users and purposes. At launch, researchers and students at Denmark’s eight universities can access the service through their existing university login, and in a little while also companies and organizations, or people who simply need a secure and accessible way to use the technology, will also be able to gain access to the service. 

“Our ambition is to make advanced AI available to everyone who can benefit from it, not only to those who have the technical expertise or resources to build their own infrastructure. With UCloud AI, we are bringing powerful models, computing resources and a secure digital environment together in one place. It is a public service built on Danish infrastructure, where users can work with AI without giving up control of their data or intellectual property, at a significantly lower cost than other providers. We believe this can open up entirely new possibilities for research, innovation and the wider society”,

Says Claudio Pica, Professor and Director of the SDU eScience Center at the University of Southern Denmark.

SDU will also collaborate with the Danish Centre for AI Innovation (DCAI), which owns and operates Gefion, Denmark’s national AI supercomputer. Researchers whose projects need to scale or have specialized computing requirements will be able to also access Gefion through UCloud AI. This collaboration points toward a future in which several supercomputing infrastructures could be accessed through a single platform, giving researchers greater flexibility and creating new opportunities for ambitious research and innovation.

Public, sovereign AI infrastructure built for sensitive work

UCloud AI is open source and operates within UCloud’s existing security framework and data processing agreements. AI processing takes place in Denmark, supporting sensitive research and organizational work while reducing dependence on external commercial providers. The platform combines AI models, data, storage, and HPC resources in one environment, with both a chat interface and an OpenAI-compatible API for developing more advanced applications.

“AI is a strategic priority across all Danish universities, and if we want to remain at the forefront of research and innovation, we also need the sovereign infrastructure to support that ambition. UCloud AI gives researchers access to advanced AI while keeping control of data and critical infrastructure in Denmark. This strengthens our ability to develop knowledge and technology on our own terms and, ultimately, to create value for society as a whole,” says Jens Ringsmose, Rector of the University of Southern Denmark and Chair of Universities Denmark.

“Secure and accessible AI infrastructure is becoming increasingly important for Danish research. From Aalborg University’s perspective, UCloud AI is an interesting addition to the national research infrastructure and a good example of how existing collaboration can give researchers access to new capabilities,” says Thomas Bak, Dean of the Technical Faculty of IT and Design at Aalborg University.

UCloud users are ready for the new platform

At Aarhus University, researchers have already used UCloud to investigate how language models respond to more than 3,000 dilemmas from the radio programme Sara & Monopolet, examining how closely their advice resembles that of the human panel. Another example is Lex.llm, a collaboration with Lex.dk exploring how generative AI can be combined with quality-assured Danish knowledge. UCloud AI will make it easier for projects such as these to access models and develop applications without establishing their own model infrastructure. 

The platform will also provide an important link to Danish Foundation Models (DFM), a national research initiative developing open Danish language models. By making models developed through DFM available directly in UCloud AI, the same infrastructure can support the entire journey from developing and testing new AI models to making them accessible for research and new applications.

“Fluent Danish is only part of what we need from a language model. We also need to understand how it handles Danish knowledge, social norms and cultural context. Model development and practical application need to inform each other: we need access to competitive open models that we can evaluate and integrate into applications, while experience from those applications helps us identify where the models need to improve. UCloud AI provides an important connection between these activities, allowing us to study AI systematically and build with it, while reducing dependence on commercial API providers and giving us greater control over the models we use,” says Kristoffer Nielbo, Professor and Head of the Center for Humanities Computing at Aarhus University and Research Lead for Danish Foundation Models.

Facts about UCloud AI

Operated by: Developed and operated by the SDU eScience Center. UCloud forms part of the infrastructure used in the DeiC Interactive HPC consortium involving SDU, Aarhus University and Aalborg University.

Platform: Built on UCloud, serving more than 25,000 researchers and students since 2020.

Computing infrastructure: Runs on BITTEN in Sønderborg, established by SDU in collaboration with Danfoss and Hewlett Packard Enterprise, serving as a model for integrating digital infrastructure with sustainable energy systems. Through an agreement with DCAI, projects requiring additional scale or specialized computing capabilities will also be able to access Gefion, Denmark's national AI supercomputer, through UCloud.

Open source and security: UCloud is open source under the EUPL license and ISO/IEC 27001:2022 certified.

Cost: UCloud AI is designed to provide AI resources at a substantially lower cost than comparable commercial services. Access is managed through a transparent credit-based allocation model.

Launch website: ai.escience.sdu.dk

Contact: 

Senior Consultant, Ursula Lundgreen in regard to press related questions: lundgreen@imada.sdu.dk, phone.+45 22 91 42 54 

Or contact Professor Claudio Pica, directly at: pica@cp3.sdu.dk  

As Professor of Archaeological Science, he will continue his work on reconstructing the history of human diet, with a particular focus on the Holocene, the period following the last Ice Age. His research will also contribute to methodological developments in fields including palaeoecology, zooarchaeology and biomolecular archaeology.

The aim is to further integrate science into archaeology, both in southern Scandinavia and internationally.

– I see our evolution and history as a succession of dietary experiments that allowed us to become who we are. We only spread globally once we learnt to subsist on whatever was available locally, says Marcello A. Mannino.

From hunter-gatherers to the first farmers
Marcello A. Mannino’s interest in human diet and ways of life began when he was a student and continued during his PhD research. At the time, he was interested in understanding the lifeways of past hunter-gatherers.

His initial focus was on hunter-gatherer adaptations to coastal environments. In this context, he developed methods for studying the shells of marine molluscs discarded after human consumption. These analyses make it possible to reconstruct the seasons in which shellfish were collected, the mobility of hunter-gatherers and past sea surface temperatures.

– Humans have subsisted for most of their time on Earth by hunting and gathering. By investigating how people in prehistory obtained their food and survived, we can gain a deeper understanding of who we are as a species and of our world more generally
Marcello A. Mannino

Marcello A. Mannino later developed expertise in isotope analyses of bone collagen. This method can be used to reconstruct the diets of Europe’s last hunter-gatherers and first farmers. Such investigations can contribute to a better understanding of the transition to farming, which represents a key stage in human history.

The relevance of past diets today
According to Marcello A. Mannino, knowledge of past diets is relevant to wider society for several reasons. Developments in human dietary habits paralleled changes in past cultures and societies while also helping to shape them and the environments in which humans settled.

– Today, we are often environmentally disconnected from the food that is produced for us. We also tend to forget the cultural importance of diet and that food is not merely a means of obtaining essential nutrition. Knowing what our ancestors ate can inspire our understanding of what constitutes human food and inform our decisions as we seek environmentally and humanly sustainable solutions, says Marcello A. Mannino.

Following his initial employment at Aarhus University, Marcello A. Mannino established the Moesgaard Archaeo-Science Laboratory (https://cas.au.dk/en/archaeo-science). The laboratory is an all-purpose facility for processing and sorting archaeological finds and for preparing samples for various scientific analyses. Much of the work conducted in the laboratory has aimed to investigate the dietary history of Denmark, Scandinavia and beyond.

Inaugural lecture
Marcello A. Mannino will give his inaugural lecture on 9 October 2026 at 13:00 in the auditorium at Moesgaard Museum. The title of the lecture is Through the lens of diet: The archaeology of food as a key to unlock human history.

Contact
Marcello A. Mannino, professor
Department of Archeology and Heritage
School of Culture and Society
Aarhus University
Mail: marcello.mannino@cas.au.dk