Artificial intelligence in the annual increment of eucalyptus hybrid for defoliation and toe drought
DOI:
https://doi.org/10.6008/CBPC2179-6858.2021.007.0006Keywords:
Neuron, Artificial Neural Networks, ForestryAbstract
The techniques for conducting eucalyptus stands involve silvicultural treatments such as fertilization, maintenance of pruning, pruning, thinning, and the management of pests and diseases that attack the stand, so that forest pests and diseases are responsible for large losses in the development of plantations. This study evaluates the application of artificial neural networks in predicting the effect of defoliation and drying of tops of trees in reducing the volumetric productivity in commercial Eucalyptus plantations urophylla x Eucalyptus grandis hybrid clone in Pará southeastern mesoregion. The data were collected in commercial Eucalyptus plantations located in the Dom Eliseu, Pará State. A 100 Multilayer Perceptron networks were trained. The input layer used the qualitative variables, Defoliation Severity and Ponteiro Dryness Severity (SevSec) and the quantitative variable Average Annual Increase (IMA). In the intermediate layer the number of neurons for data processing varied between 3 and 11. Considering the value of the Person correlation coefficient we have different indications of the best nets in training and validation. The difference between the correlation of the observed values of the IMA7 2012 and the predicted values for the IMA7 2012 expressed by r indicates that net 6 was efficient in training, but did not show as much efficiency in validation, being the value of r in training, 0.964, while in validation the second worst result, 0.933, with net 3 showing a value of 0.979. Following a different trend than r and DMP (%), the RMSE ranks the best net at training among the best at the validation stage. Network 6 in the learning phase showed a value of 0.00, indicating accuracy in predicting the accuracy of this parameter. The overall average of the estimated values was higher than the average of the observed IMA7 2012 showing consistency with the graphical analysis of the residuals of network 5, which has a slight tendency to overestimate the results, but without statistical significance by the Chi-square test. The artificial neural networks are presented as a tool with high statistical accuracy and can be used in predicting the effect of defoliation and drying of tops of trees in reducing the volumetric productivity in commercial plantations of Eucalyptus urophylla x Eucalyptus grandis hybrid clone in the Pará southeastern mesoregion.
Downloads
Downloads
Published
Issue
Section
License
Copyright (c) 2021 Ibero-American Journal of Environmental Sciences
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.
The CBPC - Companhia Brasileira de Produção Científica (Brazil CNPJ: 11.221.422/0001-03) the material rights of the published works. The rights relate to the publication of the work anywhere in the world, including rights to renewals, expansions and dissemination of the contribution, as well as other subsidiary rights. All electronically published works may subsequently be published in printed collections under the coordination of this company and / or its partners. The authors preserve the copyright, but are not allowed to publish the contribution in another medium, printed or digital, in Portuguese or in translation.