Exploratory modelling of stochastic land use cover change (LUCC) future scenarios for spatial planning

Abstract

Land Use Cover Change (LUCC) models simulate future geospatial projections of urban and rural areas by including physical, regulatory and transportation-related factors as important drivers of change. Different sources of uncertainty in LUCC modelling such as data inputs, model parameters, model structure and future scenarios have been widely researched in the past by evaluating output variability (i.e. uncertainty analysis) and quantifying the importance of specific uncertain factors (i.e. sensitivity analysis). These efforts focused mainly on reducing uncertainty and improving model accuracy and predictions, even though LUCC models are mainly stochastic (e.g. cellular automata models) and used for highly uncertain long-term projections. With a different perspective, Exploratory Modelling embraces uncertainty and does not try to predict the future, instead, takes a model as a device for simulating potential future scenarios. This presentation shows how advances in uncertainty research can be used in LUCC models within an exploratory approach to better understand spatial planning options. Specifically, the main novelties of this piece of research are 1) the extension of Sobol indices, as a sensitivity analysis technique, to evaluate the influence of several uncertain factors including model parameters, planning decisions and the stochasticity of the model; 2) the use of scenario discovery techniques to develop sensitivity-informed uncertainty maps; and 3) the investigation of the potential of such model-based techniques for spatial planning. As a case study, we project future urban development scenarios in the Lausanne-Morges agglomeration, in Switzerland, in connection with major infrastructure development plans. We then generate a large ensemble of future scenarios and identify the most influential factors on urban development (i.e. Sobol indices) and the combination of factors that produce salient scenarios for the planning of the agglomeration (i.e. scenario discovery). This paper showcases the value of conducting spatially explicit exploratory modelling as the decision-relevant factors show important spatial variations.