prilagođeno pretraživanje po punom tekstu

ŠUMARSKI LIST 5-6/2016 str. 47     <-- 47 -->        PDF

the purpose of determining site indices. A set of 109 temporary sample plots (circular sample plots of 500 m2 in size) was established subjectively to cover sites of different productivity quality and age classes throughout the analyzed beech forest complex. Mean age and mean height of 10% of the thickest trees were determined in each experimental plot. The measured height-age data were used to create an average height growth curve (guide-curve method) by seven growth functions (Table 2). The parameters of Korsun’s, Schumacher’s and Prodan’s functions were calculated by the least squares method, and the parameters of Todorović’s and the other models were calculated by using Levenberg-Marquardt’ and Gauss-Newton’ algorithms, respectively.
The appropriate criteria for the selection of the best model are the following:
a) Coefficient of determination:
b) Sum of relative squared errors: c) Relative mean squared error (%): where: hi is the measured dominant height,  is the esti­mated dominant height,  is the average dominant height and N represents the number of observations in a sample plot.
Since site index (SI) is the dominant tree height at a reference-observed age, previously mentioned functions should go through the given point, i.e. the height at the reference age in order to get the desired site index. This is achieved by calculating a parameter that represents the asymptote of the mentioned functions for each site class separately. The base age for the calculation of SI was 100 (SI100). By implementing the above mentioned functions, the so-called anamorphic site index curves were created.
The main statistical characteristics of the sample plots are presented in Table 1. The table shows that the age and the mean dominant height of the temporary sample plots range from 30 to over 120 and from 9,35 m to 38,5 m, respectively. The oldest tree is 141 years old.
The seven growth functions, their parameter values and the statistics of the fitting to the empirical dominant height-growth data are illustrated in Table 2. The obtained results indicate that Korsun’s and Schumacher’s functions show the best statistics. They have the smallest SRSE (0,3 and 0,4, respectively) and RMSE (0,31and 0,34, respectively). Functions by Todorović, Chapman-Richards and Hossfeld IV have