The pressure on land, by e.g., requirements for producing food, feed, fibre and bioenergy, has led to debates about sustainable land use planning. Multi-objective spatial optimization offers opportunities for land use planning under the multiple dimensions of sustainability. It involves finding all optimal land use configurations given multiple, often conflicting, objectives (e.g., minimize GHG emissions, maximize employment, maximize mean species abundance etc.), a control variable (what can be changed), and a set of constraints. All optimal land use configurations together form the Pareto frontier. Because for all these solutions it is impossible to improve one objective without impairing another, the Pareto frontier shows trade-offs between the different objectives, and thus between the sustainability dimensions in the case of sustainable land use planning.
Land use optimization problems are complex, as the number of possible spatial configurations (the search space) is very large. To solve them, typically heuristic algorithms from Artificial Intelligence are used, for example Genetic Algorithms. Spatial input data (e.g. the current land use configuration or carbon storage) are used to feed these algorithms, and the question arises, which representation shall be used: fields (raster data), or objects (vector data), or a combined representation. For computational efficiency, raster is usually chosen. Nevertheless, the raster representation yields disadvantages compared to vector representation. For example, information regarding the topology is lost and data inaccuracies can be generated during the conversion to raster.
The aim of this thesis is to investigate the implications of the data representation in land use optimization on the algorithm structure and the resulting optimal solutions.
This thesis is performed in collaboration with Moritz Hildemann (University of Münster, Germany).
- To define a simple land-use allocation problem with two conflicting objectives.
- To set up a spatial optimization algorithm for both a vector and raster representation of input data.
- To run both optimization algorithms and compare algorithm performance as well as the resulting Pareto fronts and optimal solutions.
- Verstegen J.A., Jonker J.G.G., Karssenberg D., van der Hilst F., Schmitz O., de Jong S.M., Faaij A.P.C. (2017). How a Pareto frontier complements scenario projections in land use change impact assessment. Environmental Modelling & Software 97, 287-302. DOI: 10.1016/j.envsoft.2017.08.006.
- Xiao N., Murray A.T. (2019). Spatial optimization for land acquisition problems: A review of models, solution methods, and GIS support. Transactions in GIS 23, 645-671. DOI: 10.1111/tgis.12545.
- Ligmann-Zielinska A., Church R.L., Jankowski P. (2008). Spatial optimization as a generative technique for sustainable multiobjective land‐use allocation. International Journal of Geographical Information Science 22(6), 601-622, DOI: 10.1080/13658810701587495.
- Programming experience
- An interest in algorithms
Theme(s): Modelling & visualisation; Integrated Land Monitoring