Precise assessments of forest species’ composition help analyze biodiversity patterns, estimate wood stocks, and improve carbon stock estimates. Therefore, the objective of this work was to evaluate the use of high-resolution images obtained from Unmanned Aerial Vehicle (UAV) for the identification of forest species in areas of forest regeneration in the Amazon. For this purpose, convolutional neural networks (CNN) were trained using the Keras–Tensorflow package with the faster_rcnn_inception_v2_pets model. Samples of six forest species were used to train CNN. From these, attempts were made with the number of thresholds, which is the cutoff value of the function; any value below this output is considered 0, and values above are treated as an output 1; that is, values above the value stipulated in the Threshold are considered as identified species. The results showed that the reduction in the threshold decreases the accuracy of identification, as well as the overlap of the polygons of species identification. However, in comparison with the data collected in the field, it was observed that there exists a high correlation between the trees identified by the CNN and those observed in the plots. The statistical metrics used to validate the classification results showed that CNN are able to identify species with accuracy above 90%. Based on our results, which demonstrate good accuracy and precision in the identification of species, we conclude that convolutional neural networks are an effective tool in classifying objects from UAV images.

The population growth and the climate changes impose challenges to society, especially regarding food supply and the maintenance of desirable soil conditions. Besides, it is necessary to reduce greenhouse gas emissions. Hence, there is a necessity of studies addressing the understanding of soil conditions under different land uses, as soils can become carbon pools or sources depending on management practices. This study was conducted in soil profiles of Amazon Plinthosols under different land uses: undisturbed forests, cattle ranching pastures, and mixed crops. Soil density and carbon fraction were assessed at four depths (0-5, 5-10, 10-20, and 20-40 cm), while particle size and mineralogy were examined at 0-20 and 20-40 cm. Soils under undisturbed forests presented lower densities and higher carbon fractions when compared to other land uses. These soils also presented the highest carbon stock; however, the Tukey test indicated no significant differences. Soil densities were lower in forest environments. Cattle ranching pastures presented higher carbon stocks when compared to mixed crops (at age 10 years). Our results indicated that proper soil management practices are needed to maintain the soil productive capacity after converting forested areas for other uses. Plinthosols, due to its textural and mineralogical characteristics, presented high amounts of quartz, demonstrating how sensitive these environments are to changes in the landscape due to deforestation and how unstable the maintenance of carbon stocks is because of such conditions. This study provides useful information to the understanding of the carbon stock in Amazon Plinthosols. It may be helpful to improve the Brazilian Inventory of Anthropogenic Emissions and Removals of Greenhouse Gases and to understand the impacts of land-use change.

ESTOQUE DE BIOMASSA E CARBONO EM PASTAGENS CULTIVADAS NO NORTE DE RONDÔNIA

VULNERABILIDADE NATURAL À EROSÃO NO SUDOESTE DA AMAZÔNIA ASSOCIADA AOS SEUS MODOS DE OCUPAÇÃO – O CASO DO ENTORNO DA TERRA INDÍGENA URU EU WAU WAU

Cattle ranching is the primary land-use of deforested areas in the Brazilian Amazon. Deforestation precedes pasture establishment, implying tremendous amounts of greenhouse gas emissions caused by carbon stock losses. Despite several studies addressing carbon storage in forests, there is a lack of data regarding cultivated pastures. Hence, the estimation of greenhouse gas emissions associated with land-use change becomes uncertain. In this study, we assessed the carbon stock of cultivated pastures located in Rondônia, southwestern Brazilian Amazon. A total of 50 squared plots of 1 m² were randomly allocated in cattle ranching farms covered by Oxisols (Dystrophic Yellow and Dystrophic Red-Yellow Latosols). Carbon fraction ranged from 0.36 for belowground biomass to 0.45 gC.g-1 d.m. for aboveground biomass. The average total carbon stock was 5.17 MgC.ha-1, with non-significant differences when stratifying data by soil types. Considering data from the III Brazilian Inventory of Anthropogenic Emissions and Removals of Greenhouse Gases, our results suggested that land-use change from primary forests to cultivated pastures resulted in a loss of 192.54 MgC.ha-1, which corresponds to a net emission of 705.98 MgCO2eq.ha-1 to the atmosphere. This study provides valuable information to improve the Brazilian Inventory of Anthropogenic Emissions and Removals of Greenhouse Gases.

CARBON STOCK AND REMOVAL OF CO2 IN YOUNG STANDS OF FOREST RESTORING IN RONDÔNIA