Scenario development in land-use change model: An empirical comparison between deductive and inductive approaches

There are bidirectional interactions between land-use and environmental systems: the way land is used has impacts on the environment, while changes in the environment have impacts on the way land is used. Land use change models enable policy makers to gain insights into possible developments of the land-use system, the possible causes of these developments, and their consequences. When creating such a model, the driving forces behind land-use change must be identified. Due to assumptions, simplifications, or lack of data many uncertainties remain concerning these forces and how they could play out in the future. Currently, a deductive scenario approach is the main method used to deal with these uncertainties. Using this approach, normally two to six scenarios are developed, for which values of the uncertain driving forces vary in accordance with the different scenarios. However, such a deductive approach only characterizes a small part of the uncertainty space. This thus increases the chance of overlooking possibly relevant regions of the uncertainty space. This research aims to explore an inductive scenario approach, namely scenario discovery, for land use change models. In scenario discovery, one generates a large number of experiments by sampling the relevant uncertain factors. Next, clusters are identified based on similar land use patterns and their conditions for occurring. Each cluster can then be translated back into a comprehensible scenario. We use the national land use change model of the Netherlands as a case study and compare the deductive scenario approach, as is currently used in practice, with the results from the inductive scenario approach. The result shows that the scenario discovery approach gives a richer overview of possible future developments of land use patterns compared to the outcomes when simulating land use based on pre-defined scenarios when using a deductive scenario approach.