Estimation of Eucalyptus Tree Height in Clonal and Progeny Tests Using Artificial Neural Networks

Abstract

The goal of this study was to test the applicability of artificial neural networks for estimating tree heights in clonal tests and progenies. We used data from 8,329 clonal tests collected for six age groups, divided into six blocks and five repetitions. For the progeny tests, we used 36,793 data points, collected at age 5 and divided into ten blocks and five repetitions. The categorical input variables considered were age, treatment, and block. The diameter (dap) was used with continuous input variables. For training the networks, we used two samples. Sub-sample 1 was composed of the first tree of each block. In sub-sample 2, the tree was selected randomly within each block. This selection was made in both tests. The selected data were separated, with 70% used for training and 30% used for validation. The other unselected trees were used for generalization. For each age and treatment, we used the Kolmogorov-Smirnov (KS) test to verify the normality of the errors. The results show that ANNs can be used to estimate the heights of trees subjected to various experimental plot treatments, with no loss of accuracy or estimation precision.