ABSTRACT
Precision and bias of a model designed to predict site index of Scots pine (P. sylvestris L.) from site variables in Sweden were tested using data from 1985 inventory plots. The model was biased and relatively imprecise (standard error = 3.7 m). A new model was constructed using a fitting subset of data, employing sums of mean monthly estimates of photosynthetically active radiation modified by local monthly climatic conditions as a primary independent variable. The best model used day-time temperature modifiers to calculate potential radiation-use efficiency. Modifiers for vapour pressure deficit and soil water did not improve the model. Elevation, distance to the sea, and phytometer indicators of nutritional fertility added small but significant improvements to the predictions. The final model had a standard error of 2.06 m for predictions of site index that ranged from 18 to 30 m at age 100. When applied to a validation subset of plots the model displayed a standard error of 2.09 m and very similar residual patterns to those observed during fitting. The new model represents a significant improvement over the older model, and further improvements may be feasible when historical climatic estimates and a higher resolution digital elevation model become available.
Acknowledgements
The authors wish to thank the University of Canterbury and the research program Trees and Crops for the Future (TC4F) for providing study leave for Professor Euan Mason. We also wish to acknowledge the Sveriges Meteorologiska och Hydrologiska Institute (SMHI) for access to monthly estimates of photosynthetically active radiation across Sweden.
Disclosure statement
No potential conflict of interest was reported by the authors.
ORCID
Euan G. Mason http://orcid.org/0000-0001-9024-9106