A Comparison of Model Forms for the Development of Height-Diameter Relationships in Even-Aged Stands

Abstract

[EN] Several model forms were analyzed to develop a height-diameter (h-d) relationship for even-aged, birch-dominated stands (8,454 h-d pairs of 198 plot-inventory combinations) in northwestern Spain. A basic model (which includes only d as predictor) and a generalized model (which also considers stand variables as regressors) were selected from among available models. Fixed and mixed forms of these models were evaluated. An approximate Bayesian estimator was used to obtain a calibrated response (prediction with fixed parameters and random effects) of the mixed models from prior height measurements of trees selected by different strategies. From a practical point of view, the calibrated basic mixed model is recommended if a randomly selected sample of 4-11 trees is available or if only the 3 trees that are the diameter quartiles are measured. If the random sample includes more than 11 trees, the basic fixed model should be locally fitted, and the generalized fixed model is recommended when only dominant height is known. Finally, the predictions of the latter model may be outperformed by the calibrated generalized mixed model, which also requires inclusion of the height of the smaller diameter trees