top of page

Predicting change in the building stock



In 2012 a remarkable study was published in the Journal of Planning History*. The research, carried out by Kiril Stanilov, demonstarted how historical spatial data, when employed within computerised mathematical models, could be used to predict the spatial patterns of urban growth and change. The astonishing accuracy of this method is shown in the images below.


The research throws light on the powerful role of policy and planning in determining urban growth, and explores the existence of systematic spatial relationships, resilient to change, which Stanilov defined as an ‘urban code’.


























Image courtesy Kiril Stanilov and Mike Batty


In the study the pattern of land development in a 200km2 area of West London was tracked from 1875 to 2005, with data for 60 types of land uses over seven time periods digitally traced using historical OS maps. Transport networks were also meticulously digitised. Changes in the patterns of land use from one period to another were then analysed in relation to distance to the Central Business District, major roads, railway and underground stations, and suburban centres. Specific policies such as those relating to the Green Belt and new building densities were also assessed, along with scholarly works on the history and planning of London. Patterns identified between the first three time periods (1875 to 1915) were then translated into rules. These were fed into a cellular automaton model in which the land area was subdivided into ‘cells’ on a grid. The cells changed their state (land use) through discrete time steps according to the rules extracted from the analysis. These rules were then iterated to produce computer generated predictions for 1935, 1960, 1985 and 2005. 


It is hypothesised that the Stanilov's methodology could also be used to predict inner city demolition and in doing so provide vital information on the likely success or failure of current and proposed development schemes. This hypothesis will begin to be tested at CASA in 2017.



bottom of page