5 Conclusions

The main purpose of this review was to analyse causalities and feedback loops in current urban land use change models. Therefore, we analysed 19 simulation models stemming from four different simulation methodologies: spatial economics, system dynamics, cellular automata, and agent-based modelling. The main conclusion of this review is that there is a range of comprehensive urban land use change models but no unique approach to represent urban landscapes and human–nature interactions. Each author or working group has its own view and focuses on other parts of the urban system and the relationships within that system. Thus, the landscape aspect is of minor importance. Most of the approaches bear the potential to model local and regional urban processes, as they provide a multitude of components and variables. However, currently only a few models integrate direct or indirect feedback loops from environmental and landscape-related impacts of urban land use change on environment to the respective driving forces in the human sphere of the systems. We see the reason for this in the gap between social science methods and findings, and computational models (cf. Geist and Lambin, 2004, 2002). The former comprehensively cover behavioural heuristics on decision making but are often qualitative in nature. The latter need quantitative (sometimes spatially explicit) input data or at least simple rules to be coded and thus incorporated into the models. To bring both approaches together and to better incorporate qualitative, social science data into quantitative models is still one of the major challenges of urban land use and landscape modelling. This is a challenge, not only for modellers, since empirical data for formulating a resilient feedback loop, resulting from environmental impacts on human quality of life and decision making, is rarely available (Haase and Haase, 2008). As urban systems are open systems which do not depend on local or regional natural resources and ecosystem services, neither individual nor policy decisions strongly depend on the availability and state of nature of the surroundings (cf. Haase and Nuissl, 2007). This makes it more difficult to elicit and formalise resilient feedbacks from the environment or landscape back to the driver. Another challenge is to express urban land use relationships, and in particular the aforementioned decision making in a spatially explicit way, as most of the CA models under review do. Finally, relationships between the local and regional scale are realised only with respect to housing markets, as single choices on the local scale are able to influence regional markets and vice versa. None of the models deals with all possible linkages between “the built-up urban” and “the rural” landscape within an urban region, although CA models such as MOLAND cover both types of land use, at least in terms of land use types. Current “hotspots” of the worldwide agri-environmental discussion, such as biofuels and organic farming, should also be partially incorporated into urban models. Here, we see another way to introduce more landscape aspects into urban land use modelling.