Growth models based on historical growth data gave reliable growth predictions up to the century shift; Detailed single tree growth models had lower precision for estimation of total growth than one single stand-based model; The prediction error was in average about 15% and did not increase with extended prediction period.
Full text in HTML
Full text in PDF
The performance of growth models implemented in the Swedish Forest Planning System Heureka was evaluated. Four basal area growth models were evaluated by comparing their predictions to data from five-year growth records for 1711 permanent sample plots of the National Forest Inventory (NFI-data). Also, two alternative implementations of Heureka, including a combined stand- and tree-level basal area growth model and a single tree-level model, respectively, were evaluated using data from 57 blocks in a thinning experiment (GG-data) involving Scots pine (Pinus sylvestris L.) and Norway spruce (Picea abies (L.) Karst) in which the trees were monitored for 30 years after the first thinning. The predicted volume growth was also compared to observed values. Growth models based on data from 1970’s and 1980’s overestimated growth in the NFI test plots from the early 2000’s by about 3%. Stand-level models had larger precision than tree-level models. Basal area growth was underestimated in dense NFI-plots and overestimated in non-thinned GG-plots, illustrating an un-solved modelling problem. Basal area growth was overestimated by 2–5% also in the GG-plots over the whole observation period. Volume growth was however accurately predicted for pine and underestimated by 2% for spruce. The relative prediction error did not increase with increasing length of prediction period. Thinning response models calibrated with GG-data worked well in the total application and produced growth levels for different thinning alternatives in line with observations.