HYPSOMETRIC RELATIONSHIP OF SCHIZOLOBIUM PARAHYBA VAR. AMAZONICUM IN PLANTATIONS INTEGRATED WITH LIVESTOCK IN EASTERN AMAZONIA: APPLICATIONS OF DIFFERENT MODELING METHODS

Main Article Content

Quinny Rocha
https://orcid.org/0000-0002-0791-517X
Patricia Ferreira Muribeca
https://orcid.org/0009-0004-4209-1075
Bruno Fernandes Veras
https://orcid.org/0009-0001-3653-5344
Raylon Pereira Maciel
https://orcid.org/0000-0001-5097-2797
Lina Bufalino
https://orcid.org/0000-0002-7688-3140
Rodrigo Geroni Mendes Nascimento
https://orcid.org/0000-0003-4981-8658

Abstract

Contextualization: The study of obtaining basic variables of the forest inventory helps in the more assertive determination of planted forest production. Given the above, this work aimed to select, among the methods modeling, the one that best estimates the heights of trees in a Schizolobium parahyba in a livestock and forest integration system in the countryside of Pará state, Brazil, indicates or not the use of a general regression equation for the types of management analyzed; compare whether there is a gain in precision with the increase in the complexity of the regression models and artificial neural networks. Three hypsometric regression models were tested, namely Curtis, Stoffels & Van Soest, and Petterson as linear, mixed, non-linear, and covariate models. The neural networks were of the Multilayer Perceptron type with one and two variables in the input layer.


Results: The Stoffels & Van Soest hypsometric model, in its linear form, presented the best regression adjustment comparison metrics, followed by the Curtis model. The artificial neural networks with two input variables resulted in better estimates of tree heights.


Conclusion: The regression models in their linear form were more efficient in estimating the height of the trees of the parica plantation in a silvopastoral system, and the equations adjusted for each stratum demonstrated better results when compared with the adjustments considering the entire plantation. The increase in the complexity of the regression models did not indicate better estimates. However, in the neural networks, there were better estimates.

Article Details

Section
Article
Author Biographies

Quinny Rocha, Universidade Federal Rural da Amazônia (UFRA)

Instituto de Ciências Agrárias, Av. Tancredo Neves, nº 2501, Bairro Terra Firme. Belém-PA

Patricia Ferreira Muribeca, Universidade Federal Rural da Amazônia (UFRA)

Instituto de Ciências Agrárias, Av. Tancredo Neves, nº 2501, Bairro Terra Firme. Belém-PA

Bruno Fernandes Veras, Universidade Federal Rural da Amazônia (UFRA)

Instituto de Ciências Agrárias, Av. Tancredo Neves, nº 2501, Bairro Terra Firme. Belém-PA

Raylon Pereira Maciel, Universidade Federal Rural da Amazônia (UFRA)

Campus Parauapebas,  Rodovia PA 257, Km 07, Área Rural

Lina Bufalino, Universidade Federal Rural da Amazônia (UFRA)

Instituto de Ciências Agrárias, Av. Tancredo Neves, nº 2501, Bairro Terra Firme. Belém-PA

Rodrigo Geroni Mendes Nascimento, Universidade Federal Rural da Amazônia (UFRA)

Instituto de Ciências Agrárias, Av. Tancredo Neves, nº 2501, Bairro Terra Firme. Belém-PA