COMPARATIVE ANALYSIS OF MULTILEVEL REMOTE SENSING DATA FOR MODELING SOIL ATTRIBUTES

Understanding soil attributes is essential for sustainable land management in an era where environmental monitoring is paramount. This study explores the use of terrestrial and orbital remote sensing data to enhance the modeling of soil attributes. We analyzed thirteen surface horizons of Rhodic Ferralsols (FR.ro) from Brazil, collecting reflectance data using the Fieldspec Pro hyperspectral sensor across the visible, near-infrared, and shortwave infrared ranges. The hyperspectral data were resampled to align with Sentinel-2 (S-2) band intervals. Orbital spectra were derived from the median of georeferenced pixels from the S-2 time series via Google Earth Engine. Statistical tests evaluated dataset similarity, and soil attributes were modeled using Multiple Linear Regression (MLR). Results showed that resampled hyperspectral data provided higher accuracy for model soil attributes like particle size and organic matter (OM), while processed S-2 multispectral data offered promising predictions with lower precision. Overall, integrating multispectral orbital and terrestrial data can be a potential for modeling and mapping soil attributes.