Applied Urban Modeling: New Types of Spatial Data Provide a Catalyst for New Models

Abstract

New sources of data originating from new ways of sensing and collecting data are fast emerging as sensors and software are being rapidly embedded into physical and social environments. The data that is being generated usually incorporates time as well as space in much greater locational and temporal precision than anything we have had access to hitherto. This is ‘big data’, so called because it is often orders of magnitude larger in volume than that collected from conventional ‘manual’ sources such as the traditional population census and household interview. This is stretching our notion of the system and problems of cities that we might model and it is shortening our attention span, forcing us to concentrate on more immediate problems in both a spatial and temporal sense than what has been the norm in the past. Our tools for urban planning and design are likely to be transformed by such developments as emerging sources of big data associated with spatial behavior in terms of location and transport interactions are revealing new horizons for urban modeling. Since its inception nearly 50 years ago, urban modeling has had an unapologetically practical and long-term focus on physical interventions in urban land use, accessibility, building location and infrastructure. Its primary purpose was and continues to be to understand the city and then to reshape it to meet long-term goals associated with equity and efficiency – the longevity of urban infrastructure remains an anchor for the need of longer term predictions of the lasting effects of major interventions and cumulative causation over decades. However at the other extreme, much narrower time horizons have been associated with immediate problems and actions in cities which have pushed shorttermism to the top of the policy agenda. Suddenly, models are being tasked to inform both shortand long-term issues with respect to urban policy, on matters that pertain to minutes and hours as well as those that pertain to years and decades. The opportunities for urban modeling to fulfill both such immediate and longer term tasks have become much more significant in the last 20 years: more rounded understanding of human behavior and institutions, an explosion of new data sources, and new means to monitor urban activities (through crowd sourcing for example) are providing the context for using models in more immediate applications. This is particularly the case with respect to transport planning. Moreover, widespread availability of fast computing is heralding a new surge in activity monitoring, research, model-building and policy applications. Slowly but surely our land use and transport models are being adapted to these new tasks. Almost 20 years ago, Wegener (1994) foresaw some of these new directions for the field of land use and transport modeling, in terms of both the necessity and opportunities for traditional models (which applied static, cross-sectional and aggregate methods) to incorporate different temporal dynamics at increasingly disaggregate spatial scales. A new generation of models such as cellular automata (CA) based models dealing with urban growth followed by more generic agent-based models (ABM) lie in the vanguard of a new concern for temporal