%0 Research article
%T Recruitment models for Norway spruce, Scots pine, birch and other broadleaves in young growth forests in Norway
%A Lexerød, Nils
%A Eid, Trond
%D 2005
%J Silva Fennica
%V 39
%N 3
%R doi:10.14214/sf.376
%U https://silvafennica.fi/article/376
%X The objective of the present study was to develop recruitment models for  Norway spruce, Scots pine, birch and other broadleaves in young growth  forests in Norway. The models were developed from permanent sample plots  established by the National Forest Inventory, and they will be included  in a growth simulator that is part of a large-scale forestry scenario  model. The modelling was therefore restricted to independent variables  directly or indirectly available from inventories for practical forest  management planning. A two-stage modelling approach that suited the  stochastic nature of recruitment in boreal forests was used. Models  predicting the probability of recruitment were estimated in a first  stage, and conditional models for the number of recruits were developed  in a second. The probability models as well as the conditional models  were biologically realistic and logical. The goodness of fit tests  revealed that the probability models fitted the data well, while the  coefficients of determination for the conditional models were relatively  low. No independent test data were available, but comparisons of  predicted and observed number of recruits in different sub-groups of the  data revealed few large deviations. The high level of large random  errors was probably due to the great variability observed in number of  recruits rather than inappropriate specifications of the models.  Provided the generally high level of uncertainty connected to analysis  performed with large-scale forestry scenario models and the stochastic  nature of recruitment, the presented models seem to give satisfactory  levels of accuracy.