ADJUSTMENT OF VOLUMETRIC EQUATIONS AND ARTIFICIAL NEURAL NETWORKS IN THE ESTIMATION OF THE VOLUME OF TAUARI IN THE TAPAJOS NATIONAL FOREST

Abstract

The objective of this study was to compare the volume estimates obtained by regression equations with artificial neural networks (ANNs) for the species Couratari stellata, from the data of rigorous cubing of 1351 trees with DBH > 50 cm from 04 (four) Annual Production Units (APUs), namely: APUs 06, 07, 08 and 09, managed respectively in 2011, 2012, 2013 and 2014, of the forest management area of the Mixed Cooperative of the Tapajos National Forest, in an area of terra firme dense ombrophylous forest. Data processing aimed to select the best regression model considering four APUs in the area of management. The best-performing equation was chosen according to the root mean square error in percent (RMSE%), Pearson's correlation and residuals graph. For the selection of the best network and its comparison with the best regression equation adjusted, the statistics used were: RMSE%, Pearson’s correlation between observed and estimated volume and bias. The best-performing equation for all APUs was the Schumacher-Hall equation, which was then compared to the best ANN obtained from data training. It was found that both methods presented acceptable adjustment and precision statistics, with potential use to estimate the volume of the species Couratari stellata. However, ANN has been shown to be slightly superior by the ability to learn and generalize the acquired knowledge and is therefore recommended for this purpose.KEYWORDS: Forest management, Romaneio, Volume modeling.