Simulation of boreal forest landscape dynamics: modeling approaches and applications

Finland Academic dissertation To be presented, with the permission of the Faculty of Agriculture and Forestry of the University of Helsinki, for public criticism in the Walter auditorium of the EE-building (Agnes Sjöbergin k. 2) on December 10 th 2004 at 12 o'clock noon. Abstract Boreal forests are dynamic, spatially heterogeneous ecosystems, and important reserves of biological diversity and timber resources. Spatially explicit simulation models of long-term and large scale forest dynamics are needed to make inferences about the historical structure of forest ecosystems, to project ecosystem dynamics under future conditions, and to simulate populations of forest-dwelling species in dynamic landscapes. In this work two simulation approaches to forest landscape dynamics are described and tested, following the developmental line from the LANDIS model developed in Wisconsin, USA. The models describe the forest landscape as a square grid. On each grid cell, the tree layer is represented by cohorts, each cohort including trees of one species that belong to the same 10-year age class. In addition to patch-level processes of regeneration, growth and mortality, the models include landscape-level processes of tree seed dispersal and disturbance propagation and spread. In the FIN-LANDIS model version, tree cohorts do not have quantitative attributes, such as basal area or volume, but are divided into two density classes, to facilitate simulation of partial disturbance and multi-layered stands. A new sub-model of forest fire initiation and spread was also developed. FIN-LANDIS was tested by attempting to reconstruct the current forest structure of the Ulvinsalo nature reserve in eastern Finland. A sub-model of stand-level forest dynamics was developed to track quantitative attributes of tree cohorts, and linked to the landscape simulation framework. The resulting Q-LAND model was parameterized for the mixedwood boreal forests of Quebec, Canada. Simulations demonstrate that the stand dynamics submodel can be calibrated to produce realistic stand succession in accordance with specific targets. However, predictive applications requiring quantitative accuracy require more thorough parameterization and model testing. FIN-LANDIS was used to examine how landscape-level forest age-class distribution is formed under mixed-severity fire regimes. When most fires are not stand-replacing, the theoretical steady state age-class distribution is either bell-shaped or bimodal, and dominated by old age-classes. Based on simulation results and the available empirical information, it is argued that, in middle boreal Fennoscandia, old-growth forests have dominated the historical unmanaged forest landscapes regardless of fire frequency. FIN-LANDIS was coupled with a metapopulation model of an epiphytic lichen …