| Article ID | Journal | Published Year | Pages | File Type |
|---|---|---|---|---|
| 89611 | Forest Ecology and Management | 2008 | 10 Pages |
Crown dimensions are important for the quantification of tree interactions in some growth models. This study investigates the potential for structural indices and other spatial measures to improve the prediction of crown radius and crown length for birch (Betula spp.) and Sitka spruce (Picea sitchensis (Bong.) Carr.) in forests in Wales. Crown dimensions were measured for 125 birch and 154 spruce in six fully stem-mapped research plots. These data were used to test the performance of a crown radius model and a crown length model which estimated crown dimensions on the basis of allometric relationships with stem dimensions. Spatial data from the six plots were used to calculate the structural indices mean directional index, diameter correlation index, species mingling, dbh and height dominance, and dbh differentiation, as well as the Hegyi competition index, and basal area of neighbours and larger neighbours, for each crown measurement sample tree, using various numbers of nearest neighbours. Two non-spatial indices, BAL and BALMOD, were also calculated for all sample trees for comparison. These spatial and non-spatial variables were then incorporated into modified crown dimension models. Model performances, in terms of efficiency and relative bias, were compared to determine whether the inclusion of spatial or non-spatial variables resulted in any improvements over models using tree dimensions alone. Crown length and radius were found to be correlated with most of the spatial measures studied. Models incorporating spatial variables gave improvements in performance over allometric models for every data set, and performed more consistently than models containing non-spatial variables. The greatest improvements were achieved for suppressed birch in unthinned forests which had irregularly shaped and strongly displaced crowns. The spatial variable contributing to the most efficient model for each data set varied widely. This points to the complexity of tree spatial interactions and indicates that there is a great deal of scope for investigating other structural indices and crown dimension model forms.
