Drunken walking through the forest? On the importance of extreme variability in spatial models of land use change

Transformative change in cities, landscapes and regions is typically anticipated through computational simulation modelling of future land use scenarios based on estimated land-use change trajectories. One of the greatest sources of uncertainty in such models is the amount of land which is expected to change (land use demand), yet this is significantly underexplored. Modellers often simply fix it at a plausible estimate, e.g. from population tendency projections or policy targets, with only slight variations across very different hypothetical futures. We propose an innovative approach to address this problem. We simulate extreme land-use demand variability in future forest land allocation in Scotland, UK, using a stochastic Markov chain known as a “drunkard’s walk”, coupled to a cellular automata land use model. The model reveals problematic assumptions in Scotland’s Forestry Strategy, which anticipates 21% of Scotland’s land area under Forestry by 2032. While the area target seems achievable, the timescale is doubtful, since only very extreme variations in the Markov function (5% year on year) allow it to be met by the specified dates, and then only