Comparison between parametric and non-parametric methods to predict the annual increment of beech

Two different methods are applied in this paper for modelling diameter growth. The first is a traditional parametric method with species dependent model parameters. The second, a newly developed approach estimates the increment in a nonparametric process, the k-nearest neighbour method. To implement the nonparametric approach, a database for beech was created. The database includes the tree diameter and the ecological site, a growing-space indices. The nonparametric k-nearest neighbour (knn) method is a new approach to estimate the increment of trees. This method is based on the idea that trees with similar attributes show a similar growth behaviour. In contrast to the parametric methods the estimation of increment uses all data of the experimental areas, which are summarized in a database. A special feature of this procedure is the fact, that the data of different origins can be integrated in the database. All information of a tree can be considered in this approach, even if not all the information for each single tree is available. The estimation of increment shows a much smaller error with the knn-method. In contrast to the parametric models, the mean square error could be reduced by half, because the differences in tree age between the experimental areas play only a minor role.