Prognosis on the diameter of individual trees on the eastern region of the amazon using artificial neural networks

The prognosis of forest structure along the cutting cycle, using models of individual trees, is one of the alternatives to manage tropical forests aiming at sustainability. Currently, in forest management practiced in the Amazon Region, growth and production models are not used to predict the future stock of the forest. Thus, the sustainable economic and environmental aspects of this activity remain uncertain. The aim of this present work was to model the growth of individual trees in a forest managed in the Amazon Region, by using artificial neural networks (ANN) to serve as subsidy to the wielder in obtaining future stock after logging, thus reducing uncertainty on forest management sustainability. Selective harvest was carried out in 1979 with an intensity of 72.5 m3 ha−1 in a 64 ha area in the Tapajós National Forest - PA. In 1981, 36 permanent plots (50 m × 50 m) were installed at random and inventoried. There were nine successive measurements in 1982, 1983, 1985, 1987, 1992, 1997, 2007, 2010, and 2012. In the modeling of the future diameter, training and testing of ANN were carried out, including different semi-independent competition indexes (DSICI). All ANN, with and without DSICI, presented correlation above 99%, RMSE below 11%, and EF above 0.98. Based on the prognosis of tree growth, we were able to conclude that ANN can be effectively used to assist in the management of tropical forests and, thus, allow for the most suitable cutting intensity and cutting cycle per species, ensuring environmental and economic sustainability of forest management.