DIGITALNA ARHIVA ŠUMARSKOG LISTA
prilagođeno pretraživanje po punom tekstu
| ŠUMARSKI LIST 3-4/2012 str. 47 <-- 47 --> PDF |
Discussion Rasprava The lack of sample plots measured at young ages was an important restriction in the initial data set of the study. Accordingly, the inclusion of additional dominant trees selected in younger stands at various altitudes was crucial to develop a model that could predict accurately dominant height development at young ages. Some authors (Rojo and Montero 1996; Palahi et al. 2004) recommend the inclusion of tree stem analysis data instead. It is emphasized that models based on temporary plots and stem analyses data deliver higher dominant heights at young ages than models based only on temporary plots which supports the conclusion that the analysed sample trees were dominant at young ages as well. Stem analyses data was not used in the current study due to the uneven-aged structure of Castanea sativa dominated stands on the Northern slopes on Belasitsa mountain and to avoid the risk of violation of the assumption of the independence of the error term. According to West (1995), the violation of this assumption is likely to produce an estimator of the covariance matrix of the parameter estimates that is negatively biased, leading to the invalidation of the normal statistical hypothesis test about the fitted equation. The relatively large difference between the asymptotic coefficients of the three functions used in the current study is in accordance with the relevant literature. Zhang (1997) reports on greater asymptotic height predicted by Lundqvist-Korf function in comparison with the Richards one. According to Ratkowsky (1983), the asymptotic coefficient is the least stable parameter in non-linear growth functions and the least-squares fit often results in biologically unreasonable upper asymptotes, especially when there are few data observations near the asymptote. In such cases, overestimation or underestimation of the height of the large-sized trees might be expected, regardless the function fitted. Top height growth for chestnut has been modelled in other European countries where the species is present. Examples are: the curves of Everard and Christie (1995) for chestnut plantations in Great Britan, the curves of Manetti et al. (2001) for chestnut coppice and high forests based on aggregated data from different regions of Europe and the curves of Alvares et al. (2010) for chestnut plantations in Northern Spain aged up to 20 years. In comparison, current study data set is characterized by more balanced and extended age distribution than data used in the other studies attempting to derive representative height growth curves for Castanea sativa. Despite the fewer observations in the older age classes (e.g. more than 70−80 years) current study data set covers the range of 10−110 years whereas data sets of all other studies do not comprised trees over than 70 years of age. Accordingly the SIC elaborated in the current study provide basis for best up to now SIC estimation for mature (over than 80 years of age) chestnut stands as well: although most of the non-linear growth functions can adequately predict height growth, they may produce large errors when applied beyond the range of model development data (Zhang et al. 1996). According to Clutter et al. (1983), most techniques for site index curve construction can be viewed as special cases of three general methods: (1) the guide curve method; (2) the parameter |