SELECTION OF REGRESSION MODELS FOR GRAPHICALLY DETERMINED SITE CLASS CURVES FOR FIR IN UNEVEN-AGED STANDS IN BOSNIA AND HERCEGOVINA

UDK: 630*54:582.475(497.6) In expert activities of forest managements, the forest stand volume is most frequently determined by way of volume tables, the so-called management tariffs. For an evaluation of stand volume using this method, the method of volume tables, it is necessary to know the site class (rating) of the stand for present tree species that is used as an argument for the selection of suitable volume progression (management tariff). The site quality for certain species in mixed stands is determined by comparing the height of trees at certain diameters at breast height (dbh) with the height of appropriate dispositions of height site class curves. In so doing the focus is placed on the ratio between the heights of large diameter trees, and the quality of the stand is rated within an interval of more defined site quality classes. For the purposes of a more objective and simpler  assessment of site quality, there have recently been attempts to make site quality assessments mathematically, using appropriate formulas in which tree heights or average heights by diameter classes and heights determined by mathematical functions of site class curves are used. For economically important tree species in Bosnia and Herzegovina (fir, spruce, beech, sessile oak, black pine and scots pine), the dispositions of height site class curves (and classes) are constructed graphically and mathematical functions for them are not known. A large number of mathematical models that are often used to represent growth trends and that satisfy the needs of height curves is analyzed in this paper with a view to determining the most suitable regression model for simulating height site class curves for fir in high forests in Bosnia and Herzegovina. The expanded Prodan model (with an additional item in the denominator) has been selected as the most suitable one on the basis of statistical indicators of the regression model quality.