Prospecção de reservatórios de oleorresina de Copaifera L. por meio de análises não destrutivas

Abstract

Copaíba oleoresin (Copaifera L.) is a potential raw material for several industry segments, due to its multiple properties. However, identifying the location of the reservoirs of this substance in tree trunks is an obstacle to the predictability of its continued supply, affecting the sustainable commercialization of the product. In an environment of high heterogeneity among arboreal individuals, it becomes a constant challenge in the search for non-invasive methods for prospecting for oleoresin. In this work, by means of literature review and experimentation, we aimed to: i) review general aspects concerning the genus Copaifera L. and researches that reinforce its potential, the demand for technologies for non-timber forest products, the main technologies available, besides the main aspects perceived as challenges for this enterprise; ii) analyze the potential of impulse tomography (IT) for the prospection of oleoresin reservoirs in the trunk of Copaifera sp. trees; iii) verify the relationship between dendrometric, meteorological and phenological variables (presence/absence of leaves) with the speed propagation of mechanical waves (VPOM) and with the average VPOM (VmPOM); iv) evaluate different configurations of artificial neural networks (ANN) and indicate the most appropriate model for predicting the oleoresin volume of Copaifera sp., based on dendrometric, acoustic and seasonal variables. The tomographies were performed in cross sections, in 35 trees, at diameter height at breast height (DBH or 0%) and, among these, in 18 at heights of 25%, 50%, 75% and 100% (1st bifurcation), obtaining the variables: VmPOM, minimum VPOM, maximum VPOM, prospecting height in percentage (Hp%), total tree height, diameter in Hp% and percentage of area of the section affected by low velocities. At this site, we also investigated the interference of different seasonal periods on tomographic results, through acoustic, dendrometric and meteorological (minimum and maximum temperature, relative humidity), in addition to the leaf phenology condition, in two groups of trees: a) group A = 14 trees – seasonal transition to rainy period (2018); b) group B = 14 trees – seasonal dry period (2019). In addition, different artificial neural networks (ANN) configurations were tested, aiming at the prediction of oleoresin volume, in which the general Multilayer Perceptron (MLP) supervised learning architecture was employed. For all analyses, descriptive, experimental and multivariate statistical tests were used. It is possible to prospect reservoirs with a significant amount of oleoresin using IT, but mainly to indicate the exclusion of trees, necessarily without reservoirs or other discontinuities. IT is sensitive to capture changes in the trunk of trees as a function of seasonal periods. The indication of a high-accurate ANN (training correlation = 0,994 and validation = 0,996) brings oleoresin management closer to other interesting technologies for its planning and management, such as applications to improve the RNA-user interface, so as to optimize the inventory stage and, mainly, the cost-benefit analysis associated with a management area.