Bayesian and Classical Models to Predict Aboveground Tree Biomass Allometry

Total dry aboveground tree biomass (M) allometric equations were derived from 25 destructively sampled Hungarian oak (Quercus frainetto Ten.) trees, growing in the Chalkidiki peninsula (Northern Greece). The regression models were developed under Bayesian and classical statistical approaches. All approaches captured equally well the variability in the recorded M values across the diameter (D^sub 1.3^) range of the sampled trees. The informative Bayesian approach based on prior distributions about the allometric parameters and the size-dependent error variance provided quite accurate M predictions. The log-linear regression, nonlinear regression, and noninformative Bayesian approaches failed to predict M distribution when validation was performed against an independent data set previously collected from the study area. It was concluded that adaptation of the Bayesian theorem in tree allometry research is strongly supported. A published simplified Bayesian approach, based on a six-tree sample, was applied in our data set and was found to be a promising tool for allometric relationships. Further research is needed to robustly support or reject such an approach. Theoretical values of the slope in M - D^sub 1.3^ allometry were validated against the derived parameter. A number of scaling parameters, rather than a unique pair of them, may quite accurately describe tree biomass allometry.