Abstract
Variance component models describing the number of new branches, branch angle, branch diameter, branch status (living, dead) and self-pruning of Scots pine (Pinus sylvestris L.) were developed in previous studies. In this study, the models were evaluated on the basis of an independent data set that differed from the data set used for model construction with respect to stand age, site fertility, geographical location, thinning treatment and fertilisation.
The residuals of the individual models, as well as the recursive system of models, were examined by dividing them into inter-tree and intra-tree components. In addition, the models were used stochastically by generating variation in the variance component describing individual trees and random errors of individual branch characteristics.
The use of the models in the independent data set did not cause systematic bias, apart from the model for the number of new branches. This was probably caused by the older age of the sample trees in the independent evaluation data set.
The other models gave a relatively good prediction of the average level of branch characteristics, but the predicted and observed characteristics differed at different heights along the stem.
The most important bias was caused by the model for self-pruning. Single branches remained attached to the trunk for a longer time than the model predicted. However, it was concluded that the models are good enough to be included in growth simulation systems.
The residuals of the individual models, as well as the recursive system of models, were examined by dividing them into inter-tree and intra-tree components. In addition, the models were used stochastically by generating variation in the variance component describing individual trees and random errors of individual branch characteristics.
The use of the models in the independent data set did not cause systematic bias, apart from the model for the number of new branches. This was probably caused by the older age of the sample trees in the independent evaluation data set.
The other models gave a relatively good prediction of the average level of branch characteristics, but the predicted and observed characteristics differed at different heights along the stem.
The most important bias was caused by the model for self-pruning. Single branches remained attached to the trunk for a longer time than the model predicted. However, it was concluded that the models are good enough to be included in growth simulation systems.
| Original language | English |
|---|---|
| Pages (from-to) | 25-39 |
| Journal | Forest Ecology and Management |
| Volume | 158 |
| Issue number | 1-3 |
| DOIs | |
| Publication status | Published - 2002 |
| MoE publication type | A1 Journal article-refereed |
Keywords
- branchiness
- model evaluation
- Pinus sylvestris
- timber quality