Mid-Infrared Spectrum Analysis for Mapping Attributes of Cohesive Soils in Brazil

Diffuse reflectance spectroscopy is a promising technique for advances in soil studies. Thus, the objective of this work was to evaluate the efficiency of diffuse reflectance spectroscopy in the mid-infrared range to estimate the sand, silt, clay, pH, H+ Al, sum of bases, organic matter contents, phosphorus, and remaining phosphorus of Argissolo Amarelo distrocoeso tipico (Typic Hapludult) in the state of Maranhao, Brazil. Two areas with different characteristics were selected for sampling: Area 1 and Area 2. A square sampling grid with 121 points was used for each area. Samples were collected from the 0.0-0.2 m soil layer. The spectra were recorded in the mid-infrared range (2500-25000 nm; 4000-400 cm(-1)) at 8 cm(-1) resolution. The data of Area 1 were used for chemometric model calibrations by Partial Least Squares Regression analysis. The data of Area 2 were used for the geostatistical modeling. All attributes presented, in general, positive calibration parameters, with adjusted coefficient of determination (R-adj(2)) for sand (0.76), silt (0.51), clay (0.77), pH (0.51), H+ Al (0.45), sum of bases (0.75), organic matter contents (0.71) and remaining phosphorus (0.6), and residual prediction deviation equal to or higher than 1.4, except for phosphorus. The lowest prediction errors were found for sand (17%) and silt (19%) contents, pH (15%), and remaining phosphorus (8%). The distribution of spatial attributes - measured and predicted - presented positive correlation, confirming the potential of diffuse reflectance spectroscopy as an alternative for prediction of soil attributes.