Estimativa espacial do estoque de carbono orgânico do solo no município de São José do Norte - RS

Abstract

The growing demand for information on soil attributes is evident in the global need to mitigate the effects of climate change. Studies involving the determination of soil attributes at the municipal level are scarce and are becoming increasingly important tools to aid decision-making in the face of the current climate and environmental reality. The spatial and vertical quantification of soil organic carbon stock (ECOS) in natural landscapes receives special attention, as it is closely related to climate change. Thus, digital soil mapping (DSM) emerges as an efficient tool, enabling the processing of information and results in a fast and satisfactory manner. This study performed spatial and depth estimates established by Global Soil Mapping of ECOS in the municipality of São José do Norte - RS, using machine learning techniques, Random Forest (RF) and cLHS (Conditioned Latin Hypercube Sampling) to optimize the sampling grid. For modeling, the digital elevation model (DEM) from the Alos Palsar DEM mission and remote sensing images from LANDSAT-OLI 8 were used, both with a spatial resolution of 30m. In general, among the set of covariates that contributed to the improved accuracy of the proposed model and reduced uncertainty in the spatialization performed, those originating from geology were most significant, due to the fact that the municipality has flat areas and wetlands throughout practically its entire extension. These geomorphological characteristics made in-depth collection difficult, in addition to reducing the capacity for spatialization of the soil organic carbon stock.