Understanding and Modelling the Processes that Drive Urban Growth
To effectively plan and model future urban growth, we must first understand how cities develop and naturally evolve through individuals' actions. Our starting point is the recognition that cities in some sense generate themselves—they are complex, adaptive, self-organising systems (White et al., 2015). Cities are created by people, either individually or through organisations such as businesses and governments, but much of this development happens unintentionally as people just go about their daily lives. They focus typically on meeting immediate needs, such as dropping children at school, commuting to work, building a home, etc. Think about it! You don’t intend to build a city, but you are a part of it. As a result, cities emerge from a collection of individual decisions. However, the planning efforts that do exist are meant to guide the city’s development. For these efforts to be successful, they must rely on a deep understanding of the processes that drive urban growth. Good models can help us understand these processes, simulate how cities can grow, make images of future urban landscapes more tangible, assess the long-term consequences of today’s planning decisions, and ultimately enable us to create better, more sustainable urban environments.
Imagination of alternative landscapes for Flanders (Kuhk et al., 2011)
GeoDynamiX: a Cellular Automata Land-Use Model
GeoDynamiX, a cellular automata (CA) land-use model developed and continuously enhanced by VITO (White et al., 2012 , 2015; Crols et al., 2015; Crols, 2017), simulates future land-use patterns at a spatial resolution of one hectare and a yearly temporal resolution of up to 2100. The model is driven by spatial demands, such as an expected growth in residential land use or population growth which is transferred into a demand for residential space. GeoDynamiX can model various activities such as population, employment, nature, etc. depending on the type of available input data and the chosen scenario. The model allocates these demands to the most suitable locations by compromising four elements: zoning legislation, physical suitability, transportation accessibility, and the neighbourhood effect (a weighted average of attraction and repulsion between land uses, population, and employment at varying scales).
This tool is designed to create and analyse land use scenarios, simulating the autonomous development of future land-use dynamics. Our land-use model can process various types of input data to suit different user needs. It can operate with freely available remote sensing-based land-cover datasets—such as the ESA WorldCover map—which offer a solid foundation but with a limited number of land-use classes. For users with access to more detailed local land-use maps, the model can integrate a broader range of land-use classes, unlocking additional functionalities. While using more detailed input data improves model precision, the GeoDynamiX solution remains accessible to users relying on open-source datasets, offering flexibility for a wide variety of applications.
GeoDynamiX evaluates the impact of different policy decisions and identifies optimal locations for specific spatial elements, including residential areas, industries, wind turbines, and forests. Gain valuable insights into how your region could evolve by 2050 under different policy scenarios with our advanced land-use model!
Spatial Interactions Between Land Use or Activities
What makes this type of model special? Cellular automata land-use models take into account competition for land by multiple users (or land use types) by working with (distance-dependent) spatial interaction rules. These rules, often referred to as the "neighbourhood effect," are central to the model’s ability to predict how nearby land uses influence one another. The future status of a cell is determined by weights that capture the impact of surrounding land uses within its environment. These weights reflect both the distance to neighbouring cells and their current land-use status or activity.
For instance, industrial areas exhibit a dual influence on residential development: they discourage (push effect) residential growth at very close distances due to undesirable proximity, while they encourage (pull effect) development slightly farther away by providing access to employment. In general, most spatial interactions are strongly positive at short distances—such as the need for shopping facilities near residential areas—and gradually weaken as distance increases.
Graphical illustration of the functioning and quantification of the “neighbourhood effect”
In addition to neighbourhood interactions, future land-use changes also depend on factors like zoning, suitability, and accessibility. All these factors are quantified in spatial data layers feeding the model and allowing to calculate a transition potential for each cell, for each activity, every time-step of the model. The model calculates this transition potential at time-steps of one year and at a resolution of typically one hectare.
Schematic overview of the different calculation steps of the model, in a time loop
For each cell, the potential for all land uses is calculated and ranked. Then, all cells are ordered based on the highest potential for a specific land use. The CA model assigns the land use with the highest potential to the top-ranked cells, continuing this assignment until the demand for that land use is met.
Annual Maps for Objective Insights and Future Landscape Simulation
GeoDynamiX produces annual map layers of land use (industrial, agricultural, residential) and activities (population, employment in different economic sectors) with a resolution of one hectare.
Land use in Brussels and surroundings, present (left) versus simulated future (right), in 34 categories
In addition, you can define and calculate spatial indicators derived from land-use and other external map layers (e.g. transportation networks) to create indicator maps such as the degree of urbanisation or the fragmentation of natural and agricultural areas.
Left: degree of urbanisation (0-1) in 2010 (top) vs simulation of 2050 (bottom)
Right: fragmentation of natural and agricultural areas (0-1) in 2010 (top) vs simulation of 2050 (bottom)
Unlocking the World With Remote Sensing and Land-Use Data
In data-rich regions like Flanders, GeoDynamiX benefits from extensive, high-resolution geographical datasets. VITO has access to multiple high-resolution data layers, including remote sensing products like the Groenkaart (one-meter resolution green area data), a large in-situ geographical dataset of all commercial and public services, and a public transport dataset including time schedules and connectivity between transport nodes. This allows us to simulate a wide range of activities, which can be allocated to a comprehensive set of land-use categories on the input maps, including various industrial and commercial types (see image of land use in Brussels with 34 categories). It also enables us to generate a broad range of valuable indicators, such as the proximity to services and accessibility to public transport.
Total accessibility of services derived from a rich collection of Flemish in-situ land-use data on services (Verachtert et al., 2023)
Even in data-scarce areas the GeoDynamiX model can generate reliable insights thanks to the growing availability of remote sensing data such as the open and free Copernicus Sentinel satellite data. Especially in low-income countries urbanisation is posing serious challenges in meeting the needs of their growing urban population. By simulating urban growth and projecting the future extent of cities—such as in the example of Niamey shown in the figure below—GeoDynamiX helps assess potential impacts on land use, housing, accessibility, and more.
Historical evolution of the city Niamey in Niger and projection for 2050
based on the SSP2 scenario using the GeoDynamiX model (simulated in the u-CLIP project)
GeoDynamiX has the power to support in designing sustainable development policies. This model can explore the potential impact on future land-use scenarios such as encouraging densification or implementing a zero land-take policy. Whether you have access to detailed local datasets or rely on open-source land-cover data, GeoDynamiX provides you with valuable insights to guide informed decision-making. Get in touch with us to discover how our model can support your planning challenges and help shape a more sustainable future!
- White, R., Engelen, G., & Uljee, I. (2015). Modeling cities and regions as complex systems: From theory to planning applications. MIT Press.
- Kuhk, A., Engelen, G., E., Vandenbroeck, P., Lievois, E., Schreurs, J., & Moulaert, F. (2011). De toekomst van de Vlaamse Ruimte in een veranderende wereld. Aanzet tot scenario-analyse voor het ruimtelijk beleid in Vlaanderen vertrekkend van de studie Welvaart en Leefomgeving Nederland (2006). Kwalitatieve analyse. Studie in opdracht van Afdeling Ruimtelijke Planning Departement Ruimtelijke Ordening, Woonbeleid en Onroerend Erfgoed
- Crols, T., White, R., Uljee, I., Engelen, G., Poelmans, L., & Canters, F. (2015). A travel time-based variable grid approach for an activity-based cellular automata model. International journal of geographical information science, 29(10), 1757-1781.
- Crols, T. (2017). Integrating network distances into an activity based cellular automata land-use model–Semi-automated calibration and application to Flanders, Belgium. Doctoral dissertation
- White, R., Uljee, I., & Engelen, G. (2012). Integrated modelling of population, employment and land-use change with a multiple activity-based variable grid cellular automaton. International Journal of Geographical Information Science, 26(7), 1251-1280.
- Verachtert, E., Mayeres, I., Vermeiren, K., Van der Meulen, M., Vanhulsel, M., Vanderstraeten, G., ... & Poelmans, L. (2023). Mapping regional accessibility of public transport and services in support of spatial planning: A case study in Flanders. Land Use Policy, 133, 106873.
