Introduction to Computational Methods in Urban History

Understanding the development of historical urban planning is essential for grasping how cities evolved over centuries. Traditional approaches relied heavily on textual records, maps, and archaeological excavations, but these methods often left gaps in spatial and temporal coverage. Recent computational advances have opened new ways to analyze past city layouts, infrastructure, and social dynamics with a precision and scale that were previously impossible. By applying algorithms, data analysis, and modeling, researchers can reconstruct ancient city plans, quantify spatial relationships, and simulate growth patterns—transforming how we interpret urban history.

Computational urban history sits at the intersection of digital humanities, geography, and data science. It allows scholars to ask questions that are fundamentally quantitative: How did street networks change after a major infrastructure project? What factors drove neighborhood segregation in 19th-century cities? How did disease spread through pre-industrial urban spaces? Answering these questions requires not only historical expertise but also the ability to handle large datasets, create iterative models, and visualize results in ways that reveal hidden patterns. The shift from purely qualitative inquiry to a mixed-methods approach has been described as a “computational turn” in historical research, and its impact on urban planning history has been especially pronounced.

Core Computational Techniques

Several computational techniques have proven particularly valuable for studying historical urban planning. Each offers a different lens through which to examine the past, and many are used in combination to build richer understandings.

Geographic Information Systems (GIS)

GIS technology allows for the mapping, analysis, and visualization of spatial data. In historical research, GIS is used to georeference old maps, digitize archaeological plans, and create layered digital atlases that track change over time. For example, historians have used GIS to overlay a 1748 map of Paris with modern street data, revealing how property boundaries, road widths, and public squares have persisted or shifted. GIS also enables spatial queries—such as measuring the distance from a medieval well to the nearest market—that would be impractical manually.

One landmark project is the Rome Lab, which uses GIS to reconstruct the urban fabric of ancient Rome from archaeological records and textual sources. Researchers have created a detailed digital model of the city as it stood in the 4th century CE, allowing them to analyze population density, land use, and even the flow of goods across neighborhoods. Similarly, the Stanford ORBIS project models the Roman road network and sea routes to estimate travel times and costs between cities, offering insights into how infrastructure influenced political and economic integration. (See Orbis: The Stanford Geospatial Network Model of the Roman World.)

Agent-Based Modeling (ABM)

Agent-based modeling simulates the actions and interactions of individual “agents” (people, households, institutions) to observe emergent patterns at the city scale. In urban history, ABM helps researchers test hypotheses about how social behaviors, economic incentives, or environmental constraints shaped development over decades or centuries. For instance, an ABM might model how medieval merchants chose locations for shops, how residents decided where to build homes relative to a cathedral, or how zoning rules (formal or informal) influenced segregation.

A notable application is the study of ancient trade networks. Researchers have built ABMs that simulate the movement of traders along the Silk Road, adjusting variables such as bandit risk, road quality, and market demand to see which route configurations were most stable. The results help explain why certain oasis cities thrived while others declined. Another study used ABM to examine the formation of street patterns in early Islamic cities, modeling how foot traffic and land ownership rules could produce the narrow, winding alleyways characteristic of those urban forms. (An example can be found in the work of Crescioli & Murgatroyd, 2020.)

Network Analysis

Network analysis treats cities as sets of nodes and edges—intersections and streets, for instance—and measures properties such as centrality, connectivity, and clustering. This technique has become essential for studying how urban form influences movement and access. Historical street networks can be extracted from old maps using digitization or automated extraction from raster images. Once in a network format, analysts can calculate metrics like betweenness centrality to identify which streets were most likely to carry through traffic, or closeness centrality to find the areas most accessible to the rest of the city.

Research on medieval European towns has shown that market squares typically held the highest centrality in the network, explaining their economic importance. Studies of 19th-century Paris under Haussmann’s redesign reveal that the new boulevards deliberately created a “legible” network that reduced the ability of residents to barricade narrow streets—a network analysis confirming historical accounts of military intent. More recent work uses network analysis to quantify the “organic” versus “planned” nature of city layouts, providing a numerical basis for typologies that were previously subjective.

Machine Learning and Deep Learning

Machine learning (ML) and deep learning offer powerful tools for processing large volumes of historical data—especially scanned maps, aerial photographs, and satellite imagery. Convolutional neural networks can be trained to recognize roads, building footprints, or green spaces in historical maps, automating the digitization that would otherwise take months. Other algorithms can detect patterns like the regular grid of a Roman colony versus the irregular fabric of a medieval town, classifying thousands of districts in hours.

A compelling use case is the project “Mapping Historical New York,” where researchers applied deep learning to extract building outlines from 19th-century fire insurance maps. The resulting dataset allowed them to study the evolution of building density and lot coverage across the city, correlating changes with immigration waves and economic booms. Similarly, natural language processing (NLP) has been used to analyze historical city council minutes, building permits, and newspaper articles, extracting references to infrastructure projects or zoning disputes. These techniques reveal not only what was built, but also the social and political context in which decisions were made.

Applications Across Historical Periods

Computational approaches have been applied to urban history across many eras, each offering unique challenges and insights.

Ancient Cities

For cities of the ancient world—Rome, Pompeii, Teotihuacan—archaeological data is often fragmentary. Computational methods help fill gaps. GIS is used to hypothesize the locations of missing structures based on known topography and building typologies. Agent-based models simulate how populations might have grown under different resource constraints. Network analysis reveals the logic of street layouts even when only foundation stones remain.

The Great Pompeii Project uses a combination of GIS, photogrammetry, and 3D modeling to reconstruct the city as it looked before the eruption of Vesuvius. Researchers have created interactive maps that allow users to explore street networks, public buildings, and water distribution systems. These tools do not just present a static picture; they also enable analysis of how the city’s layout affected daily life—for instance, by calculating walking distances between houses, baths, and forums. (More information is available at Pompeii Sites.)

Medieval and Early Modern Towns

Medieval urban planning is often described as organic, but computational analysis reveals underlying regularities. Studies of over 300 medieval towns in Central Europe have used GIS to measure plot shapes, street widths, and the placement of market squares. The results show that while no two towns are identical, they share a set of common parameters—such as the ratio of market width to street length—that suggest a shared tradition of town planning, possibly transmitted through model books or itinerant planners.

Network analysis of premodern trade routes has also expanded understanding of how towns connected. The Digital Atlas of European Towns project, for example, combines historical maps with population data and trade records to model the economic relationships between cities. These models help explain why certain towns grew into major hubs while others stagnated, linking urban form to economic function. (See the project at urbhist.org.)

Industrial Era and Modern Cities

The 19th and 20th centuries present an abundance of data: detailed maps, census records, building permits, and photographs. Computational methods have been used to study the effects of industrialization on urban form. For instance, researchers have digitized historical Sanborn fire insurance maps for hundreds of U.S. cities, creating a massive dataset of building footprints from the 1860s to the 1950s. By analyzing changes in building density and land use over time, they can track how urban sprawl, zoning, and transportation networks (including streetcars and later automobiles) reshaped cities.

Machine learning has also been applied to historical photographs and street view imagery to identify changes in architecture, street furniture, and signage. One project used deep learning to classify over 10,000 historical photographs of London, creating a timeline of how the city’s visual character evolved from the Victorian era to the present. Such analyses provide qualitative historians with quantitative evidence to support or challenge existing narratives about urban change.

Challenges and Limitations

Despite the power of computational methods, they come with significant challenges that researchers must address.

Data incompleteness and bias are foremost among them. Historical records are often fragmentary—whole blocks of a city may be missing from maps, and written records reflect the perspectives of the literate and powerful. GIS models built on incomplete data can produce misleading conclusions if gaps are filled with assumptions. Agent-based models require choices about agent behavior that may not be historically accurate. Every computational project must be transparent about its data sources and limitations, and results should be treated as suggestive rather than definitive.

Interdisciplinary collaboration is necessary but difficult. Historians and computer scientists often speak different methodological languages. A project may require careful negotiation over what counts as evidence, how to handle uncertainty, and what questions are worth asking. Successful computational urban history projects typically involve teams that include a historian, a GIS specialist, a statistician, and sometimes an archaeologist or urban planner.

Computational and infrastructural constraints also play a role. Processing high-resolution historical maps with deep learning requires significant GPU power. Storing and managing large datasets can be expensive. Smaller institutions may lack the resources to engage in large-scale projects, risking a digital divide in the discipline.

Reproducibility and methods documentation remain ongoing concerns. Many early computational studies did not share code or data, making it impossible to verify results. The field is moving toward open science practices, but the cultural shift is slow. Without transparent workflows, computational urban history risks being seen as opaque or unreliable by the broader historical community.

Future Directions

The next decade promises advances that will further integrate computational approaches into mainstream urban history. One area is the integration of multiple data types—combining maps, texts, images, and archaeological datasets into unified digital platforms. The Semantic City concept, for example, uses linked data standards to connect historical sources, allowing queries like “show me all buildings constructed between 1850 and 1870 that were more than three stories tall and located within 500 meters of a streetcar line.” Such systems would dramatically accelerate research.

Real-time simulation and interactive models are another frontier. Historians and planners could explore “what-if” scenarios: What if a city had chosen a different street grid? What if a plague had not struck? These simulations require not only better models but also more accessible user interfaces that allow non-programmers to adjust parameters and see results.

Community-sourced and citizen science projects are also growing. Platforms like Zooniverse have hosted projects where volunteers digitize historical maps, tag building functions, or transcribe handwritten records. These efforts generate large amounts of structured data at relatively low cost, while also engaging the public in historical research. As these datasets grow, they become raw material for machine learning and network analysis.

Finally, the push toward open data and reproducibility will strengthen the credibility of computational urban history. More journals are requiring data and code to be deposited alongside articles, and repositories like Figshare and Zenodo are increasingly used. This trend will enable other researchers to build on prior work, compare results across cities, and refine methods.

Conclusion

Computational approaches have fundamentally altered the study of historical urban planning. Tools like GIS, agent-based modeling, network analysis, and machine learning allow researchers to ask new kinds of questions and test theories that were formerly beyond reach. From the street network of ancient Rome to the sprawling industrial map of 19th-century Chicago, these methods reveal patterns and relationships that shape our understanding of how cities grew and functioned.

Yet the value of computational methods lies not in the technology alone, but in the questions they help answer. The most powerful work combines quantitative rigor with deep historical knowledge, using data not as a replacement for narrative but as a complement. As the field matures, historians, urbanists, and data scientists must continue to collaborate, share methods, and remain critical of their own assumptions. The past, after all, is not a dataset—but datasets can help us see it more clearly.