Informal Urban Expansion Under Displacement: A Spatial–Relational Analysis of Building Patterns
This study introduces a novel framework combining Morphological Cells with an unsupervised Graph Attention Network to analyze nearly 50,000 buildings in Marib, Yemen, revealing that informal urban expansion under displacement follows structured spatial logics driven by tribal tenure and humanitarian constraints rather than chaotic sprawl, thereby offering a cadastral-free diagnostic tool for adaptive governance in conflict-affected cities.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Cities are often imagined as the products of master plans, drawn by architects and enforced by governments. But in many parts of the world, particularly where conflict has displaced millions, cities grow differently. They expand through a slow, self-built process where families construct shelter piece by piece, often without official permission or access to formal land records. This phenomenon, known as informal urban expansion, creates complex patterns of housing that are difficult to map or understand using traditional tools. When a city like Marib in Yemen faces a sudden influx of people fleeing war, the pressure to build shelter is immediate, and the resulting growth can appear chaotic to an outside observer. Yet, researchers have long suspected that even in these desperate conditions, human behavior follows hidden rules. The challenge for scientists has been finding a way to see those rules when the usual maps and data are missing.
To understand how these settlements form, researchers usually look at the shape and size of individual buildings. They might measure how big a house is or how many houses are packed into a single block. However, this approach treats every building as an isolated object, ignoring the crucial fact that buildings are built in relation to one another. A house is not just a shape; it is a neighbor to another house, and the way they sit together tells a story about how the community grew. In recent years, a new branch of computer science has emerged that treats cities like networks, where every building is a point connected to its neighbors by invisible lines. This method allows computers to learn not just what a building looks like, but how it fits into the larger web of the city. By combining these two ideas—measuring the physical shape of buildings and analyzing how they connect to each other—a team of researchers set out to decode the hidden logic of Marib's rapid growth.
The researchers focused on Marib, a city that transformed from a small town of roughly 55,000 people into a massive hub for over 630,000 displaced individuals in just a few years. Because the city grew so fast and outside of any official planning system, there were no government maps to guide the study. Instead, the team spent months manually tracing the outlines of nearly 50,000 buildings from high-resolution satellite images. They did not rely on automated software, which often struggles with the messy, irregular shapes of informal settlements. By carefully drawing each building by hand, they created a precise digital record of the city's physical form. They then fed this data into a computer model designed to look for patterns. The model did not just count the buildings; it learned to see how they were arranged, how close they stood to one another, and how their shapes related to the empty spaces between them.
The result was a revelation that challenged the common assumption that informal growth is simply random sprawl. The computer identified thirteen distinct types of building arrangements, which could be grouped into four main stories of how the city adapted to crisis. The first story was one of intense crowding. In specific areas designated for displaced families, the model found buildings packed tightly together in dense networks. These were not haphazard piles of shelter; they were highly organized clusters where families had coordinated to maximize space under strict humanitarian rules that limited what materials they could use. The second story was one of slow, legal expansion. On the edges of the city, where land was owned by local tribes, the model found patterns of small, widely spaced plots. This indicated a different kind of growth: families were subdividing land in anticipation of future building, a process driven by local customs rather than state planning.
The third story revealed the city's past. In the older, pre-war center, the buildings followed a different rhythm, with larger plots and more regular spacing that reflected the city's original design. This area had not been completely overwritten by the crisis; instead, the old structure remained, absorbing new growth without losing its fundamental shape. The fourth and most complex story was found in the zones between the old city and the new camps. Here, the model detected a mix of formal and informal elements, where displaced families had filled in the gaps between existing structures. This "hybrid" pattern showed how vulnerable groups, including women and marginalized minorities, were creatively recombining available spaces to find safety and shelter. The computer showed that these were not just random gaps, but specific strategies of survival that blended the old city with the new needs.
To ensure these findings were real and not just an artifact of the computer, the researchers compared their results against simpler methods. When they asked the computer to look only at the size and shape of buildings without considering how they were connected, the result was a confusing mess. The buildings appeared scattered and unrelated, failing to show the clear neighborhoods and distinct growth patterns that the new method revealed. This proved that the connection between buildings was the key to understanding the city. The study also tested the stability of its findings by running the analysis multiple times. While the exact boundaries of individual buildings shifted slightly between runs, the four main stories of the city remained consistent. This suggests that the patterns are robust features of the urban landscape, not accidental glitches in the data.
The implications of this work extend far beyond the computer screen. By showing that informal expansion follows structured logics rather than chaos, the study offers a new way for aid workers and local leaders to understand crisis cities. It suggests that in areas where families are crowded together, the priority should be sanitation and water, as the density makes disease spread a high risk. In areas where land is being subdivided slowly, the focus should be on planning infrastructure before the buildings go up, to avoid cutting off future neighborhoods. And in the mixed zones, the best approach is to work with the community to upgrade what is already there, rather than trying to replace it. The research demonstrates that even in the absence of official maps, the physical shape of a city holds a detailed record of how its people have survived, adapted, and organized themselves in the face of displacement. It turns the scattered footprints of thousands of homes into a readable map of human resilience.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.