Growth and yield prognosis of Corymbia citriodora using artificial neural networks

Abstract

Because of the importance of applying refinements to prediction techniques, the application of Artificial Intelligence, such as Artificial Neural Networks (ANN), has become an advantageous alternative for modeling forest growth and production. In this sense, this work was developed with the objective of evaluating the use of ANN in wood growth and production prediction, comparing it with regression analysis. Data were collected from a Corymbia citriodora plantation through a continuous forest inventory at ages of 42, 54, and 66 months. In the evaluation of the results obtained, in a validation with dependent data, the following statistical criteria were adopted: mean of the percentage deviations, square root of the mean error, correlation, and sum of squares of the residues, in addition to the graphic analysis of the residue distribution. The biological interpretation of growth and forest production trends was also included in this evaluation. In a final step, cross-validation was performed using the chi-square test at 5% significance level. It was concluded that the prediction performed using ANN resulted in a better level of accuracy than the use of regression analysis.