Classificação e estimativa de Estoque de Carbono nas Fitofisionomias do Pantanal: Análise da Sub-região do Abobral

Abstract

The Pantanal, a vast floodplain with a rich diversity of phytophysiognomies, plays a crucial role in carbon capture and storage, which is essential for climate change mitigation. This study focused on the Abobral sub-region, using machine learning algorithms implemented on the Google Earth Engine (GEE) platform, with emphasis on Machine Vector Support (SVM) and Random Forests (FA), to classify and estimate the carbon stock in three main phytophysiognomies: Forest Formation, Savannah Formation and Grassland Formation. The results indicated that the SVM presented superior performance, with a kappa coefficient of 0.81 and class accuracy above 70%, standing out in the classification of more complex and heterogeneous formations. In comparison, the FA classifier obtained a kappa coefficient of 0.72 and lower accuracy in the classification of the savannah formation when compared to the SVM. Despite occupying a smaller area, the Forest Formation revealed a higher average carbon stock per hectare due to the density of its vegetation. In contrast, the Savanna Formation, which covers the largest area, has the largest total carbon stock. The Grassland Formation showed the lowest carbon storage capacity per hectare. This study highlights the importance of seasonal analyses for a thorough understanding of carbon dynamics and suggests that the continued use of technologies such as GHG can enhance the monitoring and conservation of the Pantanal.