Modeling the commercial volume of pure and mixed stands of beech trees using non-parametric algorithms in the educational-research Forest of Darabkola, Sari, Iran

Commercial volume trees are an important variable that contributes to economic decision-making and analysis in forest management. In this regard, Commercial Volume modeling in Hyrcanian forests is the key to implementing sustainable forest management plans. Due to time and cost constraints and the lack of local or public volumetric measurements in the equations, most forest managers still use traditional expansion factors to estimate volume. Therefore, using independent variables in volumetric modeling is an essential step in fitting models to representation. Therefore, the present study aims to model and predict business volume with minimal error using two ANN and CART algorithms. The study area was parcels 14, 16 and 24 of the educational-research forest of Sari Faculty of Natural Resources located in Darabkola, one of the functions of Sari city. After rotating the forest, masses with different forest types were studied according to the purpose of the study: Trees of pure beech, pure hornbeam and beech-hornbeam. For this purpose, at least 20 trees fell in each type and all standing trees in each sample parcel were measured. In each plot, quantitative characteristics of all trees including total tree height, trunk height (length), diameter per chest and qualitative characteristics including the degree of tree rot were measured. After measuring all the required characteristics of fallen trees at the level of Study forest types by species, to accurately estimate the volume was measured. Then the actual volume of the tree was calculated in the form of different trunk parts based on the ESmalian relationship. Finally, ANN and CART algorithms were used for modeling in STATISTICA12.0 software environment. The results of modeling the commercial volume of three masses of pure beech, pure hornbeam and beech-hornbeam with two ANN and CART algorithms showed the values of R2 explanation coefficient (0.82; 0.77), (0.44; 0.72) respectively and (0.91; 0.84). The results of modeling the commercial volume of three masses of pure beech, pure hornbeam and beech-hornbeam with two algorithms ANN and CART showed that the ANN algorithm with a R2=0.91and the percentage of RMSE%= 10.51 is more precision. Finally, the findings showed that the ANN algorithm leads to better prediction than the CART algorithm. Also, the performance of this algorithm for beech-border mixed mass is higher than any of the pure beech and hornbeam masses.