The Science of Urban Form: Data-driven solutions to real-world problems

Intelligent analysis has the potential to inform design decisions and create more socially and environmentally sustainable places.

Indeed, this novel approach to architecture and urban design has already begun to have a profound impact on our cities and communities; how new urban spaces are designed; how they look; and how our understanding of existing places has evolved.

It is vital, therefore, that architects and city planners continue to evolve their application of this data to identify and scrutinise trends and patterns across our cities. Intelligent analysis has the potential to inform design decisions and create more socially and environmentally sustainable places that respond intuitively to public need.

The Urban Design Group has worked with tools and developed methodologies to integrate data science into all levels of urban design practice.

Pre-pandemic, the combination of smart technology and urbanism was still thought to be mainly focused on large-scale city infrastructures, like intelligent transport and smart energy grids. However, as we will demonstrate, we are seeing a shift in focus to the social needs of citizens and communities. This is the true purpose of cities: to enable a huge number of individuals to live not just in a safe but an adaptable, sustainable and resilient urban environment.

With that mind, the UDG has worked with tools and developed methodologies to integrate data science into all levels of urban design practice. Tools that not only help us inform and deliver projects at a wide range of scales but, most importantly, that help us answer questions tailored to specific design problems.

Indeed, knowing what question to ask is a fundamental part of the data science process – arguably the most important part. Urban designers and data scientists alike have been trained to understand the ‘context’ of data in order to ascertain its optimum application towards a design question, and create maximum value. 

Data that can give us insights into the functioning of urban environments are plentiful: from social media data that can help capture public perceptions about a place, to call detail record (CDR) data gathered by telecom companies that capture how people move around and use urban areas. However, these data sets all have a particular context – caveats of how the data was collected – that must be understood before any insights can be extracted. This context might be as broad as: what part of the population is being captured? Or as narrow as: what personal biases might be implicit in the data? These contexts create specific constraints on how the data can be used and how insights can be extrapolated for decision making.

Our methodology in integrating urban design and data science can be categorised into three approaches: Understanding cities and regions, master planning new cities, and predicting people’s movement behaviour. In this essay, we will frame each approach with a question, and explain the different tools – and data sets – at our disposal.

There are several ways in which we use data science and urban computational tools to analyse these elements – information that tells us where and why things happen in cities. One of the most common is the use of digital cartographic tools, known as Geographical Information Systems (GIS).

Often regarded as a mapping tool, a GIS can also be considered a tool for data science as it allows us to create, manage, analyse and visualise spatial data. Formally established by the US-based Environmental Systems Research Institute (ESRI) in the 1960s, a GIS is designed to capture, manipulate, store and manage data. They are, in simple terms, an extension of cartography – the science of making maps – and allow a person to visualise and interpret data. A GIS platform normally contains a base map where additional data has been added as necessary. There is no limit to the amount of data that can be added to a GIS map, which capitalises on analysis and visualisation of data in support of research or qualitative analysis.

By representing and analysing data using GIS techniques, the UDG works with spatial models that evaluate how the structure of the built environment performs at a geographic scale, via the hierarchies of topography, streets and architecture outlined above. Outcomes from this analysis include the optimised physical form for a city, its most efficient land-use patterns or mobility options and predictions of demographic characteristics.

Data-driven GIS analysis can help generate better pedestrian flows and social interactions that sustain economic development, improve public health, lower transportation energy expenditure, and build social awareness – all of which form a comprehensive framework of urban sustainability at a large scale.  

In a more specialised field of data science at this scale, the UDG approaches cities and regions as a network of connections in a scientifically measurable ‘object’. The accurate analysis of this urban object can predict how people are going to use the city; it can measure spatial relationships, such as how we move, interact and trade.

At this level, urban design is concerned with questions like: How can street and block design reconnect urban neighbourhoods? How does a new piece of the city impact the way people travel? What has caused a specific distribution of land uses? Or what does the spatial structure of a city tell us about its function? All these kinds of questions require that we think of space as a pattern of relationships whose structure we need to understand.

The UDG works with spatial models that evaluate how the structure of the built environment performs at a geographic scale

The UDG has tackled such questions by employing spatial science-based analytical techniques – one of which is Space Syntax. Based on extensive research in the 1980s by urbanists Bill Hillier and Julienne Hanson, Space Syntax works on the theory that we choose paths not only to reduce physical exertion but also to develop and maintain efficient mental descriptions of environments; functionality and legibility of a space are closely interlinked.

In a recent example, the UDG's analysis formed a vital part of the practice’s work on the competition-winning design for the development of Stockholm Central Station. The development brings several modes of transport together in a single, integrated hub, creating a new mixed-use district for the city of Stockholm. This approach helped inform design decisions from the outset, with the aspiration to create the world’s first carbon restorative transport interchange. We used Space Syntax analysis to model the future ‘connectedness’ of the redeveloped historic street network and the potential socio-economic impact of the designs on the station and the city. The outcome was a notable 70 per cent increase in pedestrian flows and a 50 per cent increase in public open space.

However, the tools and methods we employ are not only applied to new projects, but also to improve our understanding of existing cities and regions. We do this by taking advantage of open data sets and combining them in novel ways to secure insights into different urban environments that would not otherwise be available.

We presented an example of this type of research to the International Seminar on Urban Form (ISUF). We combined data on street networks and building footprints across five different cities to create a database of street cross-sections and a set of measures for assessing their visual complexity. This allowed us to reveal a taxonomy of street typologies that we can use as a vital new reference when working on urban environments.

Using these methods and tools across different projects has allowed us to build a library of building and block typologies for different land-use types and climate regions; an invaluable resource that we can continuously reuse and refine. This has significantly cut down the amount of time spent in three-dimensional modelling, allowing us to explore more design options and evaluate their effectiveness.

Today, human behaviour simulation tools, like MassMotion, integrate LoS and other metrics to operate in a three-dimensionally modelled environment. The tool simulates how pedestrians, presented as digital avatars, use a given space within a certain time frame – how they enter or exit, for example – and how they move within a dynamically changing crowd.

When we test a design proposal or existing scenario with MassMotion, the avatars are programmed to behave with ‘social forces’ (the multiple interactions between individuals and their surroundings), helping us reveal where crossflows and counter-flows could happen and the resulting LoS a space can have.

Being able to translate this learning into a scientific understanding of movement is key. While it is impossible to replicate human activity in a completely precise way, Foster + Partners and the UDG are providing innovation through the addition of supplementary data – often produced in-house – to replicate human behaviour in the most accurate way possible.

The types of environments we have applied our data-based pedestrian modelling to range from the scale of architectural design, like workplaces, retail and cultural venues, to infrastructural environments, such as airports and train stations. We have even used this modelling up to the scale of residential neighbourhoods and district masterplanning.

 One of the team’s recent projects in this area is the use of proximity analysis in pedestrian simulations – essentially modelling social distancing – to help manage and phase the safe reopening of facilities after Covid-19 lockdowns. We are re-visiting designs to assess the lessons learnt from the pandemic. Our work for the South Sabah Al-Ahmed Masterplan in Kuwait is an example of this.

Accelerated by the unique conditions imparted on communities during to the pandemic, the trend for neighbourhoods and built-up areas increasing their open space has seen a resurgence of interest across the globe. The 61.5 square-kilometre city of South Sabah is formed of ten neighbourhood clusters and, for the client and Foster + Partners team, it was important to revisit the planning of these neighbourhoods with regards to new occupancy levels and pedestrian routes between residential clusters.

Accelerated by the pandemic, the trend for neighbourhoods and built-up areas increasing their open space has seen a resurgence of interest across the globe.

To explore the impact of social distancing and help test various mitigation scenarios, the UDG ran a proximity analysis report, which reveals how people would move within a set distance – in this case within a two-metre distance of each other. This analysis helped us test residential density against the amount of public open space within a typical neighbourhood in South Sabah. The baseline condition had shaded canopies only in main public areas, such as playgrounds and small pocket parks and we found that the areas that reached a maximum capacity while maintaining social distancing were these areas. However, providing shaded sidewalks with walkable spaces of at least three metres in width, reduced the amount of time users spent in close proximity by 50 per cent compared to the baseline.

By testing multiple scenarios, and working with additional data of neighbourhood population provided by the client, the team was able to best replicate occupant behaviour to quantify mitigation strategies; we helped both designers and client to simulate the pros and cons of an adaptable and flexible public realm.

It is urban data science that has emerged to help us better understand … the growth and evolution of urban settlements. Today, we have an array of powerful data-based urban and statistical methodologies but, at Foster + Partners, we are always looking to the future.

In the future, we can imagine being able to generate our own data, such as population estimates or classification of building typologies, for places where it is not readily available. Existing analytical models that can take days or weeks to compute, such as those related to environmental analysis, can be replaced in early stages with surrogate models using machine learning approaches. This allows us to quickly understand constraints and iterate different creative options at the very earliest design stages. Additionally, such models can help reverse engineer typical design workflows, allowing designers to start with the outcomes they would like and generate the appropriate geometry.

Lastly, taking advantage of recent advances in artificial intelligence and leveraging these with feedback from designers (an approach known as ‘human in the loop’), machine learning can be used to explore design options and arrive at novel design solutions that are both technically sound and incorporate subjective criteria. As these tools and methods mature, they will form an integral part of our urban designers’ toolset. 

Harnessing data-driven urban planning and design is the new present. Foster + Partners and the UDG are using advanced computational tools to incorporate big data into urban design; championing watertight urban science and allowing for theories to be challenged with alternative hypotheses and rigorously tested. Tomorrow, utilising machine learning and artificial intelligence-driven algorithms to create plans and predict their impacts will make architectural and urban planning professions adapt to new perspectives and will help to further address the climate emergency in our cities.