In this study, we tested a new distribution function to model a large number of permanent plots of five tree species, which represent the wide heterogeneity and complexity of forest stands in NW Spain. This is the first evaluation of the Gumbel CDF in the field of forest modelling, although the function is applied in other environmental sciences such as hydrology.We compared moments- and mode-based methods for estimating the Gumbel CDF because the lowest value in many left-truncated distributions is the minimum inventoried diameter (5 cm), which is the mode of the extreme value distributions. Minimum diameter distributions do not follow a theoretical normal or exponential model, and the Gumbel
CDF fitted by the mode yielded the best results for minimum diameters of B. pubescens and P. radiata stands in which the values of the 25th and 50th percentiles were the lowest (see Table 2 and Fig. 1d, h). Probability density functions such as Johnson’s SB were fitted using the real mode value (Hafley and Buford 1985), although poorer results were obtained than with other methods based on percentiles, moments or maximum likelihood (Zhou and McTague 1996; Zhang et al. 2003).