DECISION SUPPORT, GIS, AND URBAN PLANNING
by
Michael Batty & Paul J. Densham
January 1996
Centre for Advanced Spatial Analysis
University College London, 1-19 Torrington Place
London WC1E 6BT, UK
ABSTRACT
Computers have been applied in urban planning almost since their inception, but only recently with the development of graphics, distributed processing, and network communications has software emerged which can now be used routinely and effectively. At the basis of these developments are geographic information systems (GIS) but gradually, these are being adapted to the kind of decision and management functions that lie at the heart of the planning process. In this brief note, we describe current developments showing what is now possible in the development of spatial decision support systems (SDSS), and planning support systems (PSS), and we then speculate on future developments in decentralised decision-making which will dominate the field in the next decade.
1. Computers in Urban Planning and Management
Computing devices have been used in public planning for 100 years. Hermann Hollerith invented the punched card machine at the turn of the century for the US Population Census, and this eventually led to the formation of the world's largest computer company, IBM. Once the digital computer was developed half century later, applications in public planning and management became widespread. By the mid 1950s, population and transportation data were being processed by computers and these were quickly followed by various simulation modelling efforts. By the late 1960s, urban data management systems were being widely implemented by public agencies for a variety of routine and less routine management and strategic planning functions. This experience has been well documented (Edralin, 1986) but in the last 10 years, applications of computers in planning have changed dramatically (Batty, 1995). The top-down approach based on remote, large-scale, database computing has been replaced by a much more personal computing style in which graphical display of urban data now provides the focus.
This bottom-up style is largely a consequence of changes in computing technology. Once the microprocessor was invented, the path to miniaturisation and personalisation was set and as the cost of memory fell dramatically, more and more applications involved graphical computing. Geographic information systems (GIS) are an obvious application but the way computers are being accessed and results displayed is now largely graphic - witness the widespread dissemination of recent Windows-based software - and this has led to a sea change in the way computers are being applied in planning. There has also been a change in types of application over the last 20 years. There is now much more emphasis on data than on modelling, on routine applications for management rather than the more grandiose applications to strategic planning which dominated the 1950s and 1960s. This is reflected as much in the way planning is now perceived in its current role in advanced (post) industrial societies, as in the way the technology has changed.
These changes can even be detected in the development of GIS. 20 years ago, the early beginnings of GIS were as an adjunct to strategic planning, particularly in landscape and resources management. Software vendors such as ESRI and ERDAS began this way while companies such as Intergraph came directly from computer-aided design (CAD). In the last 10 years, the emphasis has shifted to graphic display, the representation of spatial data, and its manipulation in quite straightforward ways. In terms of planning and problem-solving processes, to date there has been very little emphasis on formal analysis, simulation and modelling and hardly any at all on design and decision-making aids. However this picture is changing and new functions are being slowly added. In the next 10 years however, the use of computers in planning will clearly be affected by developments in computer use in general - across networks based on decentralised interaction between users - and it is likely that we will see a much greater emphasis on informal decision-making using computers interactively. We will return to these issues by way of conclusion.
2. Planning and Spatial Decision Support
Planning and management are based on a generic problem solving process which begins with problem definition and description, involves various forms of analysis which might include simulation and modelling, moves to prediction and thence to prescription or design which often involves the evaluation of alternative solutions to the problem. Decision characterises every stage of this process while the process of implementation of the chosen plan or policy involves this sequence once again. The process takes place across many scales and is clearly `iterative' or `cyclic' in form. Processes may be nested within one another while the extent to which different professionals, managers and other decision-making interests are involved through the various stages, depends upon the nature of specific applications and their context. In practice, the process is often partial and much diluted from this more formal characterisation. The typical process illustrated in Figure 1, however, remains a basis for action.
Figure 1: The Planning Process as a Sequence of Computable Methods Enabling Decision Support
Click here for a full-size .GIF image (12 KB)
Figure 1 shows how GIS and relate modelling technologies fit within this process. Indeed, this is the kind of structure that Harris (1989, 1991) refers to as a planning support system (PSS) which links a variety of computer-based software supporting decisions at different stages of the planning process (Batty, 1995). As we have implied, this kind of process is rarely executed in a comprehensive fashion and usually, only a few elements of it exist, often to the exclusion of others. For example, many involved in public planning have access to GIS but few are involved in linking GIS to modelling and forecasting while the development of formal design methods using GIS is in its infancy (Manheim, 1986). In fact, a major research program involves the use and adaptation of GIS through embedding and linking various types of predictive and prescriptive models in formal terms. Strategies for such linking range from weak to strong coupling (Batty, 1994). Models can be linked to GIS simply through the import and export of data - weak coupling - while much stronger coupling exists where models are embedded within GIS or GIS functions within models. We will demonstrate, albeit rather briefly, examples of such coupling in the next section where we show how urban spatial interaction (predictive) models and location-allocation (prescriptive or optimising) models can be linked to GIS.
3. Spatial Technologies for Design and Decision Support
Before we are able to demonstrate these ideas, we must briefly review the development of spatial technologies of which GIS are central. Spatial technologies involve any kind of software which is essentially descriptive of data with an explicit spatial or geographical dimension. Mapping software is not explicitly spatial for often data is not georeferenced in a form that can be manipulated, and thus not strictly part of those technologies we define as being spatial. The conventional definition of spatial technologies are those which have integrated and explicit functions for storing, manipulating and displaying spatial data where the spatial dimension is the key to each of these functions.
Geographic information systems are the main spatial technologies to date although increasingly other systems dealing with spatial data are acquiring spatial database and display capabilities. For example, the remote sensing package Imagine from ERDAS can be used as a GIS while GIS packages such as ARC/INFO and Intergraph's MGE provide very clear and well-defined links to other mapping, CAD, and remote sensing software. Even the growing desktop GIS packages such as MapInfo and ArcView 2 contain formal links and macro languages which enable their functionality to be extended directly through new programming or indirectly through links to other software. There are many examples of extended functionalities which link GIS to analysis and modelling and we will illustrate two varieties here. We have developed some of the commonly used spatial interaction-location models based on simulating transport flows from home to work within established GIS frameworks such as ARC/INFO. In Figure 2, we show a typical screen from an application of GIS to urban population density modelling which we have developed for the Buffalo region in Western New York. We have defined a modelling process consisting of data analysis, model calibration, and prediction as a set of relations embedded within a GIS. This system uses ARC/INFO as the display medium but also uses the software as the organising frame for the sequence of analysis and modelling operations which are accessed as links to the outside world through system macros.
Figure 2: Embedding the Modelling Process within the Proprietary GIS ARC/INFO
Click here for a full-size GIF image (49 KB)
The advantages of using GIS to structure simulation modelling is in the way this software is neutral to its sources of data. Once generic data analysis functions are set up, these can be applied to observed data, model results, forecasts, designs, whatever. Data functions thus dominate the system. In short, although the GIS acts as the framework, most of its relational functions are never actually used, yet the structure of its software forms the essential organisation of the application. In the ArcPlot frame which is shown, the popup-pulldown menuing window at the top shows the sequence of modelling operations, each element of which is accessed through the Arc Macro Language link to other program modules, while the display of results of these operations is the GIS itself. In the screen which is shown, the model has already been calibrated and thematic maps of observed data, model predictions and residuals are shown. There are many other graphics features such as 3-d surfaces, scatter plots, and dynamically-linked or `hot' windows in the system, all accessible through a hierarchy of menu items (Batty and Xie, 1994).
A very different approach but one which leads to the same kind of application involves developing purpose-built GIS type functions within a specific modelling package. In short, rather than embedding less elaborate models within a comprehensive GIS, it is possible to embed a limited range of GIS functions within a more elaborate modelling framework. In Figure 3, we show a typical example of a modelling process which has been developed within a purpose-built visual environment, the model in question being based on simulating the interaction between homes and workplaces through the journey-to-work for a crude representation of the city of Melbourne, Australia (Batty, 1994). The modelling process - beginning with data exploration/analysis, model calibration, and then prediction/forecasting - is used to structure the visual layout with the main portion of the screen given over to maps of data and results which pertain to these three stages in the process. Moreover, these maps are drawn as bar charts to indicate population and employment at different locations. Interzonal and intrazonal flows can be visualised directly whereas these types of flow map are difficult to customise and display quickly.
The problems of course lie with the absence of interactive functions such as zoom and the way these are controlled with pointer devices such as the mouse. Although we have built a rudimentary zoom (aggregation) facility into this particular program, this is quite limited and were we to really attempt to replicate the efficiency of proprietary systems, we would in effect be replicating GIS. Nevertheless, the interactive modelling process implied by Figure 3 comprises less than 3000 FORTRAN statements and as the user is so close to both the graphics and the model, changes to the visualisation can be made at will. Nevertheless, the system has to be so closely tailored to the city in question that even changing the problem's size - from 8 to say 80 zones - causes major changes in the visualisation which necessitates reprogramming. Using proprietary GIS as in the Buffalo application avoids this.
Figure 3: Embedding Simple GIS Functions within a Visual Modelling Environment
Click here for a full-size GIF image (43 KB)
These two examples extend GIS to embrace predictive modelling but to progress the planning process, GIS must be used for prescriptive or optimising modelling which involves the design of solutions to formally structured problems. This is perhaps a more focussed form of decision or planning support although the use of GIS in any of the stages of the planning process shown in Figure 1 involves the generic concept of decision support. Densham (1991) has developed a suite of programs called LADSS (Location-Allocation Decision Support System) which link heuristic optimisation techniques for matching the supply of various facilities such as schools, shopping centres, or hospitals to the demands for these same facilities by the affected population. Various objective functions can be optimised but typically these involve functions which minimise distance, travel time or travel costs between the demand and supply points. Developing such models within GIS provides very powerful visualisation facilities for display and manipulation, giving immediate intuitive evaluation capabilities which a wide range of non-technical users and decision makers can relate to. In Figure 4, we show one such screen from the output of LADSS where the application is to school catchment districts in Iowa. Here the various flow lines show the optimal allocation of schools to area offices, and within such a visual environment for modelling, many different varieties of solution based on widely different assumptions can be explored. In short, this kind of visual environment provides the basis for informed learning about the problems in question as well as providing the obvious outputs as locational solutions.
Figure 4: Allocation of School Districts to Area Education Authority Offices in Iowa to minimise Travel Distance on the Primary Road System
Click here for a full-size GIF image (35 KB)
There are many other types of visualisation capability which GIS provides but we are not able to present those which pertain to data itself due to lack of space. Animating data is possible both in temporal terms but also through ways of exploring, hence understanding urban patterns. All we can do here is to refer readers to the literature but increasingly such applications show ways in which GIS and spatial technologies in general are coming to embrace all varieties of analysis from near-data to near-design kinds of model (Batty and Howes, 1996; Densham, 1996).
4. Digital Environments for Decision Support
Our last foray into developments in computing and GIS likely to affect urban planning in the next decade involves the ways integration between spatial representation, modelling and optimisation-design will be implemented in the coming years. Increasingly these will take place in a digital environment which itself will be integrated through networking. As an example of this, much of the data in the examples which we have shown in Figures 2 to 4 has been retrieved and manipulated over networks and over different platforms. But this has usually been by individuals or small groups and as yet the more general user has not been involved interactively with such applications. The next decade will see the development of whole groups of non-technical users in planning being directly involved in the use of this kind of information technology across distributed networks. The current growth of the Internet and the World Wide Web is clear evidence of the potential for this kind of integration and more and more data and applications software are available in this medium. For example, at University College London, we are building an interface to information about London which we refer to as `Virtual London'. As a pilot to this project, we have an online version of University College which essentially enables the user to visually wander around the college in 3-d and pick up information about buildings, rooms, and people in them. If you want to examine this and play with it yourself, it is available in the public domain. If you click on our home page whose address is http://www.geog.ucl.ac.uk/casa/ you can then download the relevant software and run the application on a Mac or PC. This is the kind of application which will become commonplace during the next decade.
Computing environments for collaborative, distributed spatial decision making have been explored by Densham and Armstrong (1994) and a version of this on which systems such as LADSS might run, is shown in Figure 5. Heterogeneous processing environments such as this can support both individual and group use of spatial decision support systems (SDSS). Each computer architecture in this scheme provides a particular mix of processing capabilities. Some architectures provide very high performance for a particular subset of processing tasks while others trade performance, and glamour, for flexibility. Similarly, the processing requirements of different kinds of software vary greatly and dictate their suitability for different computing platforms. SDSSs tend to have large computational burdens and diverse ranges of processing requirements that make it difficult for any single computer architecture to accommodate them effectively. Consequently, SDSS designers are turning to suites of computers with heterogeneous processing characteristics to support their systems (Densham and Armstrong, 1994). The processing requirements of individual elements of a SDSS are analysed and the best available host architecture is identified. Each user task is shipped to the appropriate computers for processing across very high bandwidth communications channels. Both individual users and groups working together to solve complex spatial problems can be supported in this way.
Figure 5: The Use of Heterogeneous Processing Environments to Support Individual and Group Use of Spatial Decision Support Systems
Click here for a full-size GIF image (45 KB)
One of the benefits of this approach is that highly visual and interactive modelling and analysis environments can be implemented (Densham, 1996). An intrinsic element of such environments are new types of graphical display, such as that in Figure 4, designed explicitly to meet users' needs at various stages during problem-solving, that can be built by exploiting the computing resources available in heterogeneous computing environments (Densham and Armstrong, 1993). What is now required is much more considered analysis of the ergonomics and problem-solving capabilities of such technologies, and the next decade is also likely to see much greater emphasis on the ways in which technologies can be used to improve processes of human interaction, hence processes and methods of problem-solving.
5. The Future of GIS
GIS and related spatial technologies are suddenly expanding to embrace many traditionally separate functions. At the same time, many other types of software are beginning to add GIS-like functionality: spreadsheets and their improved graphics capabilities in handling 2-d maps and 3-d visualisations are a case in point. In one sense, software is breaking up on the desktop into basic modules which can be hooked together in diverse ways while other software is becoming increasingly generic in that all manner of textual, numerical, and graphical functions are being included under the same rubric. GIS itself is changing as more functions are embodied in hardware and as the vendors increasingly begin to specialise in applications, in data, and in specialist niche markets based on computer services. In this paper we have sketched the contemporary scene with respect to urban planning, where the future is likely to be dominated by smaller, finer scale applications at the level of urban design, the increasing development of 3-d GIS capability, and the use of urban remote sensing for data capture and generation. Many of these applications and extensions are underway at different centres around the world, as part of the European Science Foundation's GISDATA program, as part of the US National Science Foundation's National Center for Geographic Information and Analysis (NCGIA) program, and at centres like our own here at the University of London. The integration of diverse software and methods will have a major impact on what we are able to plan for and how we might develop effective planning for complex urban environments in the next decade.
References
Batty, M. (1994) Using GIS for Visual Simulation Modeling, GIS World, 7, (10), 46-48.
Batty, M. (1995) Planning Support Systems and the New Logic of Computation, Regional Development Dialogue, 16, (1), 1-17.
Batty, M., and Y. Xie (1994) Urban Analysis in a GIS Environment: Population Density Modeling Using ARC/INFO, in A.S. Fotheringham and P.A. Rogerson (Eds.) Spatial Analysis and GIS, Taylor and Francis, London, 189-219.
Batty, M. and Howes, D. (1996) Exploring Urban Development Dynamics through Visualisation and Animation in D. Parker (Ed.) Innovations in GIS 3, Taylor and Francis, London, forthcoming.
Densham, P. (1991) Spatial Decision Support Systems, in D.J. Maguire, M.F. Goodchild, and D.W. Rhind (Ed.) Geographical Information Systems: Principles and Applications, Longman, London, 403-412.
Densham, P.J. (1996) Visual Interactive Locational Analysis, in P. Longley and M. Batty (Eds.) Spatial Analysis: Modelling in a GIS Environment, GeoInformation International, Cambridge, UK, forthcoming.
Densham, P.J. and M.P. Armstrong (1993) Supporting Visual Interactive Locational Analysis using Multiple Abstracted Topological Structures, Proceedings, Auto Carto 11, Eleventh International Symposium on Computer-Assisted Cartography, 12-22.
Densham, P.J. and M.P. Armstrong (1994) A Heterogeneous Processing Approach to Spatial Decision Support Systems, in T.C. Waugh and R.G. Healey (Eds.) Advances in GIS Research, Proceedings, Sixth International Symposium on Spatial Data Handling, 1, 2945.
Edralin, J.S. (1986) Information Systems for Urban and Regional Planning: A State-of the-Art Review, in H. Sazanami (Editor) Information Systems for Urban and Regional Planning: Asian and Pacific Perspectives, United Nations Centre for Regional Development, Nagoya, Japan, 9-22.
Harris, B. (1989) Beyond Geographic Information Systems: Computers and the Planning Professional, Journal of the American Planning Association, 55, 85-90.
Harris, B. (1991) Planning Theory and the Design of Planning Support Systems, A Paper presented to the Second International Conference on Computers in Planning and Management, Oxford, 6-8 July, 1991.
Manheim, M.L. (1986) Creativity-Support Systems for Planning, Design and Decision Support, Microcomputers in Civil Engineering, 1, 14-31
.
Created 15/1/96
Last modified by Paul Densham on 11/8/2000 (or 8/11/2000 depending on your accent)
pdensham@geog.ucl.ac.uk