A REMOTE SENSING PROXY ON THE DEVELOPMENT OF THE BRAZILIAN SOIL TEXTURE MAP

This study presents the digital mapping of soil texture in Brazil, utilizing remote sensing techniques and machine learning, with a focus on clay, silt, and sand fractions. The research covers the entire Brazilian territory and integrates high-resolution spectral data (30 meters) with harmonized soil samples. The Random Forest algorithm was employed to create predictive models, validated by performance metrics such as the coefficient of determination (R²) and root mean square error (RMSE). Results revealed significant variability in soil fractions across Brazilian territory. The study demonstrates the relevance of remote sensing in large-scale soil property characterization, highlighting its applicability for sustainable agriculture management and natural resource conservation across different Brazilian biomes.