Uma nova abordagem para inventário de biomassa em plantações de eucalipto no Brasil

Abstract

Traditionally, forest inventories assist in determining volumetric stocks and their structure in commercial plantations. Volume has always been the primary studied variable, representing a scientific paradigm to be changed. In the present moment, considering the manifold applications of biomass, Industry 4.0 has started to seek more comprehensive data regarding forest resources, extending beyond mere volume. Consequently, the arrangement of inventories and the management of wood consumption are directly affected by this necessity. With this concept in consideration, fundamental density emerges as the primary variable to be thoroughly investigated and modeled, owing to its favorable correlation with biomass generated through tree growth. Thus, this research proposes a new approach to biomass inventory through a hybrid methodology that employs non-destructive methods to obtain basic wood density, genetic programming for variable selection and linear model generation, and mixed-effects modeling to enhance the accuracy of dependent estimated variables. The experiment involved using 55 sample plots distributed across three commercial clones of Eucalyptus located in São Paulo, Brazil. Classic tree-level variables were adopted and combined with hardness (Pilodyn) in the extracted samples. In a subsequent step, density (derived from the previous model) was modeled using only variables collected in the traditional inventory. Finally, in the last stage, biomass forest was estimated by multiplying basic wood density and individual volume. The applied methodology has proven its efficacy, demonstrating a substantial correlation with laboratorial density. In conclusion, it can be inferred that the methodological proposition renders the implementation on an enterprise scale viable, thereby facilitating the provision of accurate tree-level biomass data. Additionally, the application of genetic programming demonstrated its robust and practical modeling capabilities.