This study predicts the establishment of regeneration of Norway spruce (Picea abies (L.) Karst.) in 3-, 4- and 5-year-old regeneration areas planted, direct-seeded and naturally regenerated, respectively, in southern Finland. Establishment of regeneration was described by the number of planted, direct-seeded and natural tree seedlings, as well as by the heights of spruces and broadleaves. For each regeneration method, establishment of regeneration was modelled by fitting a multilevel multivariate model to regeneration survey data. Site and soil quality and the method of site preparation were used as explanatory variables. Stochastic simulations that accounted for the random between-stand and within-stand variation in the regeneration result were conducted to compare the uncertainty and the need for removal of broadleaves (i.e. early cleaning) involved in different methods of regeneration for spruce.

Using disc trenching as soil preparation method, the probability of good regeneration result for planting, direct seeding and natural regeneration was 58 %, 34 % and 38 %, respectively. Using mounding, the probability of good regeneration result for planting increased to 73 %. The risk of failed or poor regeneration in planting, direct seeding and natural regeneration was 3 %, 37 % and 32 %, respectively. When considering only those simulated stands in which at least half of the plots required cleaning, 40 % of planted stands, 99 % of direct-seeded stands and 89 % of naturally regenerated stands were deemed to require treatment. Planting was the most effective and guaranteed method of regeneration. Direct seeding and natural regeneration of spruce entail a high risk of failure.

Key words: site preparation, planting, direct seeding, stand tending, stochastic simulation.