EFICIêNCIA DE UTILIZA??O DE MACRONUTRIENTES EM EUCALIPTO POR MéTODO N?O DESTRUTIVO ESTIMADOS POR REDES NEURAIS ARTIFICIAIS

摘要

The Non-Destructive Sampling (NDS) provides an efficient, simple and safe characterization of chemical properties of the plant, as the Coefficient of Biological Use (CBU). The association of NDS with the technique of Artificial Neural Networks (ANN) can be a potential alternative to replace the regression equations and the traditional methods of interpolation. Therefore, this work aimed to evaluate the efficiency of ANN and non-destructive sampling for the efficiency of nutrient use in the trunk. The research plot was installed in a randomized block being studied, in three blocks, the effect of five planting spacing: T1?–?3,0?m?x?0,5?m, T2?–?3,0?m?x?1,0?m, T3?–?3,0?m?x?1,5?m, T4?–?3,0?m?x?2,0?m e T5?–?3,0?m?x?3,0?m. A sample-tree was felled to make the cubage and quantify the dry bark and wood per experimental plot, totaling 15 trees. The sample-trees were weighed in the field and subsamples of bark and wood were collected along the stem to form a composite sample per tree. Also removed was a single sample of each component obtained with the aid of a chisel and hammer in DBH in the same sample-trees. The samples were dried at 65°C until constant weight. The material was ground and subjected chemical analysis. Adjusted regression models and application of ANN to estimation of CBU Trunk from the CBU DBH Bark and CBU DBH Wood . The ANN had a higher accuracy and reliability of the regression. Modeling by artificial neural networks using only sample in the DBH region proved to be adequate for estimating the coefficient of biological use of stem.