Development of an individual tree volume model for Irish Sitka spruce and comparison with existing UK Forestry Commission and Irish GROWFOR models.
Keywords: Stepwise regression, 2k-fold cross validation, subset comparison, Box-Cox transformation, mean squared predictive error.
AbstractThe two most common models used to estimate tree volume in Ireland are the British Forestry Commission (FC) models, developed using British data and the Irish Growfor models developed from Irish stand data. Here, a model analogous to the British FC model for Sitka spruce (Picea sitchensis (Bong.) Carr.) is developed using individual Irish tree data and all models are compared. Sitka spruce data from Coillte Teo (Irish Forestry Board) stands in the Wexford and Waterford regions, consisting of 81 thinned plots with a total of 4,315 volume-sampled trees (5,419 trees including repeated measurements), were used for the study. Stepwise regression was carried out to select the model and a Box-Cox transformation was used to improve model fit. A 2k-fold cross validation was used to check the validity of the selected model across different subsets of the data. The selected model slightly over-predicts volume in its mean volume per ha estimate by 0.38%, whereas the GROWFOR estimate under-predicts by 4.6% and the British FC model under-predicts by 8.6%.
How to Cite
O’Rourke, S., Mac Siúrtáin, M., & Kelly, G. (2013). Development of an individual tree volume model for Irish Sitka spruce and comparison with existing UK Forestry Commission and Irish GROWFOR models. Irish Forestry, 184-199. Retrieved from https://journal.societyofirishforesters.ie/index.php/forestry/article/view/10109