Estimador geoestatístico e modelos de efeito misto para modelagem hipsométrica e volumétrica de povoamentos florestais

Abstract

Volume information in a forest stand is essential for the forest management, taking into consideration that most of the productive processes in a company are dependent on the productivity data that are obtained through forest inventories. In this context, the measurement procedures and data processing are fundamental elements and deserve attention, since they are the basis for determining the volume in an area. The forest sector is constantly searching for techniques that enable it to achieve excellence in the estimation of timber production, optimizing processes and maximizing profits. In this work was proposed and evaluated the performance of a model extracted from the training of an artificial neural network and of new models to estimate the total height of eucalyptus trees; of mixed models with adoption of structure of the variance and covariance matrix for hypsometric and volumetric estimates; and a geostatistical estimator for volume prediction per hectare, considering different sample intensities. The techniques presented satisfactory performances, resulting in the gain in accuracy in the estimates of total height, individual volume and stand volume. The results of this study allow the reduction of the sample intensity in pre-cut forest inventories, indicating the potential and applicability of these techniques in the solution of problems of measurement and forest management.