A spatially explicit stochastic model to simulate boreal forest cover transitions: general structure and properties

We describe a spatially explicit simulation model of large-scale forest cover transition in the North American boreal biome. This model is a time-dependent Markov chain with discrete states corresponding to dominant tree species in the forest canopy. To parameterize the model, we used three temporal variables, extracted from published data from field studies: period of persistence of a given cover type, the time interval of forest cover replacement by another, and the time of complete replacement. Environmental domains based on climate, soil moisture regime, and soil nutrient status stratified all forest cover transitions. Probability matrices of forest cover transition were derived for 15 discrete states at 20-year intervals for each of the environmental domains. Five spatial databases from boreal North America were used as input to the model: forest cover composition, time since last forest cover change, climate zone, soil moisture regime, and soil nutrient status. The model output consists of spatially explicit prediction of (a) time since last disturbance, (b) transition age, (c) forest cover composition, and (d) canopy age in 20-year time steps at 1 ha resolution. As a case study, we simulated forest cover transition in a 3.7 million ha area in boreal Canada. These 200-year simulations show that spatio-temporal transition of forest cover type is significant even in the absence of catastrophic disturbances.