Individualização de copas de Eucalyptus spp. em ambiente SIG

Abstract

Increased access to information generated by drones and the low cost of their use have presented great potential for use in precision forestry processes applied in the development of forest management in the Northeastern states of Brazil, which historically have low productivity in the sector. In the quest to explore the potential of using drones in low-cost forest management aimed at the production chain in the Northeast, the objective of this work was to quantitatively describe the precision of the metric information extracted from Eucalyptus spp crowns. existing in an experimental forest plantation represented in a high-resolution RGB image provided by a drone, based on the comparison of data from an Object-Oriented Supervised Classification (OBIA) with reference data resulting from manual vectorization (VMR). The study was carried out in the Forest Experimentation Area of the Jundiaí Agricultural School (EAJ) - Specialized Academic Unit in Agricultural Sciences (UECIA) of the Federal University of Rio Grande do Norte (UFRN). 64 canopies/trees were manually vectored (VMR) that delimited an area of interest within the entrance orthomosaic. The classified layer generated by OBIA for the orthomosaic was cropped to the area of interest, making it possible to quantify the percentage of land use and occupation for the classes of land use and occupation defined for this work, as well as the execution and methodology of canopy individualization (IC) developed in this study, which provided the measurement of the metrics of the canopy areas already sampled by VMR. The data obtained were categorized as paired and non-parametric and the medians were subjected to statistical differentiation at a 5% level of significance through the application of the Wilcoxon test, which also determined their Effect Size (TDE). The Wilcoxon test revealed a significant difference between the measurement methods of the compared areas (z= 1.1; p< 0.05), with a small effect size. However, according to observation of the results, the methodology used in this work proved to be effective in generating information relevant to low-cost forest management.