Modelling diameter distributions of Betula alba L. stands in northwest Spain with the two-parameter Weibull function

Abstract

The diameter distributions of 125 permanent plots installed in birch dominated (Betula alba L.) stands in Galicia were modelled with the two-parameter Weibull distribution. Four different fitting methods were used: that based on percentiles of the distribution, non linear regression, maximum likelihood and the method of moments. The most accurate fit was obtained with the non linear regression (NLR) approach, considering the following statistics in the comparison: bias, mean absolute error (MAE), mean square error (MSE) and number of plots rejected by the Kolmogoroff-Smirnoff (KS) test. The scale parameter (b) and the shape parameter (c) obtained with the most accurate method (non linear regression), were first modelled with simple linear models and then related to commonly measured prediction variables (quadratic mean diameter, dominant height and stand density) with the parameter prediction model (PPM). The parameters fitted with the method of moments were recovered with the parameter recovery model (PRM) from the first and the second moments of the distribution (mean diameter and variance, respectively). Results indicated that both methods were successful in predicting the diameter frequency distributions. The PRM was more accurate than the PPM method.