Uso do sensoriamento remoto no estudo da distribuição da matéria orgânica do solo de uma área úmida isolada de topo plano

Abstract

Soil organic matter (SOM) is an indicator of soil quality, and understanding its distribution is fundamental to the debate on carbon dynamics and greenhouse gas emissions. In recent decades, remote sensing has contributed to the study of soils and the distribution of SOM. The objective of this study was to evaluate the feasibility of using spectral data to study the prediction and spatial distribution of organic matter content in the surface horizon of an isolated flat-topped wetland in the municipality of Cordeirópolis, in the state of São Paulo. A linear statistical model of the wetland was constructed to predict the distribution of MOS, which was validated using the relationship between the spectral indices analyzed and the MOS content obtained by field and laboratory sampling. The evolution of land use around the isolated depression between 1985 and 2022 was also analyzed using data available on the Mapbiomas platform. In the results obtained with the model from Sentinel-2A satellite images, the minimum MOS value was 22 g/kg and the maximum was 47.66 g/kg, which is in agreement with the data collected in the field, which ranged from a minimum of 27.55 g/kg to a maximum of 42.07 g/kg, with an accuracy - R² of 0.88, a root mean square error - RMSE of 3.62 and a mean absolute error - MAE of 2.64. The results of applying the model to RPAS images (drones) reached a minimum value of 25.25 g/kg and a maximum value of 48.92 g/kg MOS, also close to the data collected in the field. On the RPAS scale, R² of 0.88, RMSE of 3.62, MAE of 2.85 and mean square error - MSE of 13.11 were obtained. It can be considered that the proposed method was satisfactory in predicting the MOS content of the soils in the studied wetland, demonstrating the gains in scale for using the method with satellite images and RPAS