prilagođeno pretraživanje po punom tekstu

ŠUMARSKI LIST 1-2/2021 str. 38     <-- 38 -->        PDF

Taking into account the above mentioned facts, it can be noted that the prediction of the CHR model parameters proved to be the most reliable, given the positive and relatively high value of the coefficient of determination. In addition, the CHR models for both ŽA and RU have parameters (b0, b1, c0, and c1) of the same sign, indicating the same shape of the model curves. This is also the case with the Korf function, but not with the second-order polynomials that fitted the parameters of the Korsun function (b0, b1, b2, c0, c1, and c2).
Nevertheless, the site index curves obtained by the CHR function are best fitted into the data obtained by the stem analysis procedure (table 4). The empirical data of the CAIh culmination ages per site class and regions indicated that the beech in the area of RU, on average, reached the culmination of the CAIh earlier than in the area of ŽA. For example, the culmination age of the CAIh of beech trees in the RU area in the first site class (22 year) took place 5 years earlier than on the territory of ŽA (27 year).
Table 5 show results of PRESS and RSS statistics for each SC in both research regions. It can be noticed that PRESS values are in all cases higher than RSS values. Differences between the mentioned statistics (PRESS vs. RSS) are higher in the region of RU in comparison to region of ŽA, especially for the first two SCs (I, II). In generally, no model satisfies the statistics completely. However, if we observe the first two lowest values of the PRESS (gray cells in the table) for each site class (SC), we can notice that considering all site classes in both regions, the CHR model has the first two lowest values in 8 cases, and the Korsun and Korf models in 6 cases each. So, it can be concluded that, on average, the CHR model proved to be satisfactory despite the fact that in some cases it slightly underestimates or overestimates the data, especially in the first two SCs in the RU region.
The analyzed Korf, Korsun and CHR growth functions to be used for the purpose of site index curve construction are often found to be reliable and superior to many other functions used to model height–age relationships. Generally, the CHR function has been regarded as probably the most