Stochastic, empirically‐informed model of landscape dynamics and its application to deforestation scenarios

Jakub Nowosad , Tomasz Stepinski

Land change including deforestation undermines the sustainability of the environment. Using data on 1992‐2015 pattern change in over 1.7 million meso‐scale landscapes worldwide we developed a stochastic model of long‐term landscape dynamics. The model suggests that observed heterogeneous landscapes are short‐lived stages in a transition between quasi‐stable homogeneous landscapes of different themes. As a case study we used Monte Carlo simulations based on our model to derive a probability distribution for evolutionary scenarios of landscapes that undergo a forest‐to‐agriculture transit, a prevalent element of deforestation. Results of simulations show that most likely and the fastest deforestation scenario is through the sequence of highly aggregated forest/agriculture mosaics with a decreasing share of the forest. Simulations also show that once forest share drops below 50% the remainder of the transit is rapid. This suggests that possible conservation policy is to protect meso‐scale tracts of land before the forest share drops below 50%.