Urbanisation is one of the most complex and dynamic processes of landscape change. Although only about 4% of the world’s land area is urbanised and densly populated (Ramankutty et al., 2006), we claim “the millennium of the cities,” since more than half of the currently 6.6 billion world population is living in urban areas (United Nations, 2008, 2009; PRB, 2007; EEA, 2006*; Kasanko et al., 2006). Projections for the future show that urbanisation – in terms of an increasing share of population living in urban areas – is very likely to continue (Batty et al., 2003; EEA, 2006; Lutz et al., 2001). Urbanisation is not only a societal problem, but also an environmental one, because it contradicts a normative ideal of “a natural or un-spoiled landscape” in spatial planning (Nuissl et al., 2008). In a multitude of studies it has been shown that land consumption is usually detrimental to the environment in different regards (e.g., Johnson, 2001; Antrop, 2004). Its impact reduces the ability of landscapes to fulfil human requirements and thus impairs ecosystem services and landscape functions in various ways (de Groot et al., 2002; MEA, 2005; Curran and de Sherbinin, 2004). Individual ecosystem services and quality of life aspects that are affected by urbanisation include the production of food, the regulation of energy and matter flows, water supply, the provision of biodiversity and of health and recreation, and the supply of green space and natural aesthetic values (Alberti, 1999). Suburbanisation and urban sprawl were the dominating land consumption processes in North America and Europe after WW II (Batty, 2008). Recently, high growth rates in developing countries have led to enormous environmental loads as discussed above (Heinrichs and Kabisch, 2006). As urban systems are very densely populated and their land use components highly interlinked (Liu et al., 2007*), developing views about their future is both a major concern in landscape research and a complex task. Modelling land use relationships helps to understand underlying drivers of land use change, to create future land use scenarios and assess possible environmental impacts (Lambin and Geist, 2006; Ravetz, 2000).
A variety of land use change models, particularly for urban landscapes, already exist, ranging from specific case studies to generic tools for a variety of urban regions. These models differ largely in terms of their structure, their representation of both space and human decisions, and their methodological implementation. Compared to land use change models in open landscapes, urban areas are shaped particularly by human activities, societal processes and human–nature interactions (Couclelis, 1997). In addition to implemented simulation models, a number of articles and book chapters elaborate on the “ideal” integrated model, theoretically necessary causal feedback loops etc. These “ideal” models shall serve as analytical frameworks to better understand the systems under study. Often, authors use frameworks like the DPSIR-framework (drivers, pressures, state, impact, responses) of the European Environment Agency (EEA) to conceptualise these conceptual models. According to Verburg, “the main drawback of using these analytical frameworks is the assumption of one-directional processes between driving factors and impacts” (Verburg, 2006*, p. 1173), because in reality, it is difficult to differentiate between impacts and drivers in a system. Bürgi et al. (2004*) distinguish five major types of driving forces: socioeconomic, political, technological, natural and cultural. Furthermore, they differentiate between primary, secondary and tertiary driving forces, as well as between intrinsic and extrinsic driving forces (Bürgi et al., 2004). In their introduction to urban simulation, Waddell and Ulfarsson (2004) sketched urban markets and agents, choices and interactions in an “ideal” urban land use model. Timmermans (2006*) criticizes that present urban models focus on functional chains like the following: demand causes allocation across space, which in turn causes traffic flows, based upon which a transportation model calculates travel times, which in turn explain residential choice. Timmermans votes to include other aspects of integration in urban land use models, such as task allocation within households, residential choice, job choice, vehicle ownership, scheduling of activities, competition and agglomeration of land uses and actors, co-evolutionary development of demographics, employment sectors, land use and activity profiles and a more thorough treatment of varying time horizons, including anticipatory and reactive behaviour. According to Miller et al. (2004*), an integrated urban systems model with a focus on transport should include socio-demographic components (evolution of population), demographics (demographic change and migration into and out of a region), decision-making (location choices of households and firms), economic variables (labour market, import/export of goods and services), transportation (activity and travel patterns of population, goods and services, depending upon urban structure and economic interchanges, performance of road and transit systems) and respective effects on land use (evolution of the built environment) and environment (atmospheric emissions generated by transportation and industry; Miller et al., 2004*). Moreover, Hunt et al. (2005*) stated eleven modelling axioms for such an “ideal urban land use model”:
- Representation of an urban system should focus on those elements that interact with the transportation system.
- An urban system consists of physical elements, actors and processes.
- A transportation system is multimodal and involves both people and goods.
- Markets are the basic organising principle of an urban system.
- Flows of people, goods, information and money arise out of demand.
- Urban areas do not reach an equilibrium.
- System time must be explicitly dealt with.
- Feedback between short-term and long-term processes has to be integrated (e.g., travel and infrastructure).
- Some factors may be treated as exogenous for modelling purposes.
- Some activities arise in response to external demand.
- A very detailed level of representation for actors and processes is necessary.
A variety of reviews including urban land use models already exist: Agarwal et al. (2002) as well as Schaldach and Priess (2008) review integrated land use models in general, also including models that deal with non-urban land uses such as forestry, pasture and agriculture. Axhausen (2006) specialises in models on transportation demand and traffic flows. Beckmann (2006) and Iacono et al. (2008) focus on interactions between urban land use and transportation. The authors predominantly discuss modelling approaches and does not give details regarding single models. Similar to this, Berling-Wolff and Wu (2004) provide an historical overview of modelling approaches and do not discuss single models. The U.S. EPA (2000) focuses on models of urban growth and sprawl but mainly includes U.S. American approaches and – because of its publication date – does not include recently published models. Geurs and van Wee (2004) and Hunt et al. (2005) focus on models which emphasize the interaction between urban land use and the transportation system. Furthermore, Timmermans (2006) gives a historical overview and describes a large number of models but does not give a comparative description of presently developed models. With his review on modelling the urban ecosystem, Alberti (2008*) puts less emphasis on urban land use change, but rather focues on the environmental impacts and human-induced environmental stress of the urban system. The review utilises a range of evaluation criteria, of which feedback mechanisms, multiple actors and the inclusion of uncertainy are seen as the most challenging (Alberti, 2008). Finally, Verburg et al. (2004) sketch a few exemplary models, but their focus lies on discussing general modelling approaches and not on single causal feedbacks.
Set against the background summarised in Section 1.3, this review analyses economic models, system dynamics approaches, cellular automata and agent-based models developed for urban systems by systematically addressing a range of critieria such as the conceptual approach, model components and included variables. In doing so, it aims at giving an overview on the respective model structures. The main purpose of the review is to derive ideas for causal relationships within land use change in urban systems, with a special emphasis on integrating social and natural science dimensions. The innovative aspect of this review compared to existing reviews is the aim to explicitly analyse causalities and feedbacks in urban land use changes.
As Verburg (2006*) points out, an integration of social and biophysical systems could be enhanced by including feedback mechanisms in land use models, e.g., the feedback between driving factors and effects of land use change (here understood as impacts), the feedback between local and regional processes, and the feedback between agents and spatial units (Verburg, 2006*). “Less common in land use modelling is the simulation of feedbacks between impacts on socio-economic and environmental conditions and the driving factors of land use change” (Verburg, 2006, p. 1173). Therefore, the review presented here will include a glance at those feedbacks. Since urban land use models deal with spatial entities – that is, among others, the landscape itself – an important aspect of selecting modelling approaches for the review is spatial explicitness in terms of landscape property. In addition, urban landscapes are highly complex, as highlighted in several paragraphs of the introduction part of this paper; therefore, one should focus on comprehensive models that include different relationships, influences and dependencies along with their spatial representation. The paper is organised as follows. Section 2 sets up a set of evaluation criteria for conducting the model review, which follows in Section 3. Section 4 especially focuses on causalities and feedback loops of land use change, before coming to the paper’s conclusions (Section 5).