Recruitment models for different tree species in Norway

The objective of the study was to develop recruitment models for application in a large-scale forestry scenario model. Recruitment models were developed for all tree species of interest in Norway based on the permanent sample plots of the National Forest Inventory. Recruitment is a stochastic event with two outcomes, and a two-stage modelling approach was applied. The probability of recruitment was estimated in a first stage, and a conditional model for the number of recruits was developed in a second. Variables describing location, site conditions and stand characteristics all had significant influence, either on the probability of recruitment or on the conditional number of recruits. Evaluation with independent data revealed few significant deviations between predicted and observed number of recruits for Norway spruce and Scots pine, while the models underestimated the number of recruits for birch and other broadleaves. The relatively large random errors were probably due to lack of independent variables describing the stochastic process of seedling establishment, rather than to inappropriate model specification. Provided, the generally high level of uncertainty connected to large-scale forestry scenario analyses and the stochastic nature of recruitment, the models seems to give satisfactory levels of accuracy at the national, regional or forest-estate level. The models are applicable to even-aged and uneven aged forests as well as forests with mixed and pure species composition, and they include independent variables directly or indirectly available from forest inventories for practical forest management planning.