int. j. geogra phical information science, 1 998, vol. 12, no. 7, 651 ± 671Review ArticleGIS-based urban modelling: practices, problems, and prospectsDANIEL Z. SUIDepartment of Geography,Texas A&M University,College Station, TX 77843-3147, U SAe-mail: [email protected]. This paper reviews the practices, problems, and prospects of GIS-based urban modelling. The author argu es that current stand-alone and variousloose/tight coupling approaches for GIS-based urban modelling are essentiallytechnology-driven without adequate justi® cation and veri® cation for the urbanmodels being implemented. The absolute view of space and time embodied in thecurrent generation of GIS also imposes constraints on the type of new urba nmodels that can be developed. By reframing the future re search agenda from ageographical information science (GISci ) perspective, the author contends thatthe integration of urban modelling with GIS must proceed with the developmentof n ew models for the informational cities, the incorporation of multi-dimensionalconcepts of space and time in GIS, and the further extension of the feature-basedmodel to i mplement these new urban models and spatial-temporal conceptsaccording to t he emergin g interoperable paradigm. GISci-ba sed urban modellingwill not only espouse new computational models and implementation strategiesthat are computing p latform independent but also liberate us from the constraintsof existing urban m odels and the rigid spatial-temporal framework embedded inthe current generation of GIS, and enable us to think above a nd beyond thetechnical issues that have occupied us during the past ten years.1. IntroductionFor almost two decades in the 1960s and the 1970s, GIS and urban modellingdeveloped in parallel with few interactions. The integration of GIS with urbanmodell ing did not take place until the late 1980 s, as a part of the GIS community’se orts to improve the analytical capabilities of GIS (Goodchild et al. 1992, Anselinand Getis 1992, Fischer and Nijkamp 1992, Fotheringham and Rogerson 1994,Fischer et al. 1996). Now adays , GIS users and urban modellers have increasinglyrecogni zed the mutual bene® ts of such an integration from the preliminary successesof the past ten years. Va rio us urban modelling tech ni ques have enabled GIS usersto go beyond the data inventory and management stage to conduct sophisticatedmodell ing and simulation. For urban modeling e orts, GIS has provided modelerswith new platforms for data man agement and visua lization (Nyerges 1995). Themassive di usion of GIS in society has the potent ial to make models more transparentand to enable the communication of their operations and results to a l arge group ofusers. The growing literature on the integration of GIS wit h urban modelling atteststhe recognition of s uch mutual bene® ts ( Brail 1990, Birkin et al. 1990, Batty 1992,Brook s et a l. 1993).The objective of this paper is three-fold: (1) to review the current practices ofGIS-b ased urban modelling; (2 ) to identify the existing problems of current e ortsto link GIS with urban modelling; (3) to discuss a new research agen da from theemerging geographical information sci ence (GISci) perspective.1365-8 816 /98 $12.00Ñ1998 Taylor & Francis Ltd.D. Z. Sui652This paper is organized into ® ve sect ions. After a brief background int roductionin section one, th e current practices of GIS-based urban modeling are review ed insection two. Section 3 discusses the existing problems of coupling GIS with urbanmodell ing. Future prospects of urban modelling f rom th e perspective of geographicalinfor matio n science are covered in§4, followed by concluding remarks in§5.2. GIS-based urban modelling: current practicesBy the early 1990s, it was (and perhaps still is) a general consensus within theGIS community that the lack of sophisticated analytical and modelling capabilitieswas one of the major de® ciencies in the current generation of GIS technology(Opens haw 19 91 ). Several recent research initiatives in North America and Europefocus on the improvement of spatial analytical an d m od ellin g capa bilities of GIStechnology. The integration of GIS with urban modelling was part of these broadresearch e orts to link spatial analysis and modelli ng with GIS. Although overlappingwith many other GIS modelling e orts in terms of the general methodology, GIS-based urban modelling has a set of substantially di erent conceptual issues fromGIS-b ased environmental modelling (Goodchild et al. 1993, 1996). Current practicesof GIS-based urban modelling thus deserve a separate scrutiny.Generall y speaking, four di erent approaches have been widely used to integrateGIS with urban modelling (® gure 1). My discussions here are con® ned to method-ologi cal issues only. Those interested in the details of speci ® c models are referred toWegener ( 1994).1. Embedding GIS-like functionalitie s into urban modelling packages. Thisappro ach aims to embed GIS functionalities in urban modelling packages, and hasbeen adopted primarily by urban modellers and spatial statisticians who think ofGIS essentially as a mapping tool. Usually no commercially available GIS softwarepackag es are involved, as illu st rated by Putnam ( 1992) in the US , the Leeds groupin the UK (Clarke 1990, Birkin et al. 1996), and Haslett’s SPIDER system ( Haslett1990), etc. This approach usually gives system developers maximum freedom forsystem design. Implementation i s not constrained by any existing GIS data st ru ctures ,and usually this approach is capable of incorpora ti ng the latest development inurban modelling. The downside of this approach is that the data management andvisual ization capabilities of these urban modelling software packages are in no waycompar able to those available in commer cial GIS and programming e orts alsotend to be intensive and s ometi mes redundant. Also, we should recognize that mosturban modelling software packages were developed by individual researchers gearedtowar d speci® c projects. Although they possess certain conceptual commonalties,these urban modelling packages use a great variety of data s truct ures, programmingtool s, and hardw are platforms that make this approach extremely di cult forother users.2. Embedding urban modelling into GIS by software vendors. Although still pre-domin antly an acad emic pursuit, a few leading GIS software vendors in recent yearshave made extra e orts to