We have adjusted two growth and yield models to temporary sample plots from across Canada, and used climate variables in lieu of phytometric indices such as site index to represent, in part, the site-level variability in growth potential. Comparison of predicted increments in plot-level height, basal area and merchantable wood volume to increments of these variables measured in permanent sample plots shows a moderate to poor predictive ability. Comparison with the performance of four operational growth and yield models from different provinces across Canada shows comparable predictive power of this new model versus that of the provincial models. Based on these results, we suggest that the simplification of regional growth and yield models may be achieved without further loss of predictive power, and that the large error in the prediction of growth increment is mostly associated with the use of temporary sample plots which, by definition, contain little information on stand dynamics. We also suggest that, because of the empirical nature of these growth and yield models, the scale of application should determine the appropriate scale of the model. National estimates of forest growth are therefore less likely to be biased if obtained from a national model only than if obtained from a combination of regional models, where those exist, gap-filled with estimates from a national model. Key words: yield model, merchantable wood volume, stand age, climatic variables, simultaneous regression, robust regression
Spatial Tree Mapping Using Photography
