HOW TO GROW? - MODELING LAND USE CHANGE TO DEVELOP SUSTAINABLE PATHWAYS FOR SETTLEMENT GROWTH IN THE HINTERLAND OF COLOGNE, GERMANY
- 1Department of Urban Planning and Real Estate Management, University of Bonn, Germany
- 2GIScience, Institute of Geography, Heidelberg University, Germany
- 3HeiGIT at Heidelberg University, Germany
- 4empirica ag, Bonn, Germany
- 5Gaiac e.V. Forschungsinstitut für Ökosystemanalyse und -bewertung, Aachen, Germany
Keywords: Land use modelling, random forest, urban growth, Python, PostGis, urban sprawl, scenario analysis
Abstract. Urban sprawl, with its negative consequences for nature and agriculture, is one of the great challenges of the 21st century. A particular problem is the strong population growth in metropolitan regions and the associated need for new residential areas. For the study area presented in this article, the western part of the Cologne metropolitan region, an increase of up to 200,000 inhabitants (30% of the population in 2040) is expected by 2040. If planning practices remain unchanged, this would result in a need for up to 4,460 hectares of new residential areas by 2040. Much of this development would take place at the edge of existing residential areas which frequently are of high value for agricultural production. To simulate different scenarios for the allocation of the necessary residential areas in the study area, a land use model was developed based on an open software stack. To evaluate the results of the simulations, a set of 20 indicators covering aspects of agriculture, ecology, housing and economy was used to assess both the sustainability and quality of possible new residential areas. In this article, two scenarios were used to show the effects of different building densities and the protection of particularly high-quality agricultural land. While the increase in building density is expected to reduce the area needed for new residential areas, the protection of high value agricultural land has negative effects on species protection. Using two species as examples, the limitations of an indicator-based assessment of the model results are shown.