Prediction and mapping of soil attributes using diffuse reflectance spectroscopy and magnetic susceptibility
MetadataShow full item record
The development of fast, accurate and low-cost methods to quantify soil attributes is of paramount importance to enable detailed mapping, mainly in tropical regions where there is great variation of the chemical, physical and mineralogical attributes. Therefore, the aims of this paper were (i) to investigate if visible and near infrared (VIS-NIR) spectroscopy and magnetic susceptibility (MS) can be applied to determine soil attributes at the sandstone-basaltic transition and (ii) evaluate and map their spatial distribution. Calibration models based on VIS-NIR spectroscopy and MS were developed separately for each attribute. Soil samples (0-25 cm depth) were collected at 446 sites, air-dried and passed through a 2-mm sieve and analyzed in the laboratory. To develop models based on soil spectra and laboratory data, the partial least squares regression (PLSR) was used. Already, the MS-based models were calibrated by linear regression between magnetic and laboratory data. The best prediction accuracy parameters were obtained with MS, later with VIS-NIR and lastly with VIS. The more accurate results between the observed and predicted values were found for iron oxide extracted by dithionite (R2 = 0.89, RRMSE = 0.02), clay (R2 = 0.85, RRMSE = 0.76) and total carbon (R2 = 0.83, RRMSE = 1.18) estimated by MS, revealing that this is a good predictor of key properties of studied soils, even with wide chemical and mineralogical variation. Both tools are very attractive for the strategic planning of land use and occupation, mapping large areas with detailed scale, environmental monitoring and precision agriculture.