Mid- and near-infrared spectroscopic assessment of soil compositional parameters and structural indices in two Ferralsols

Bulk soil samples (BS) (n = 120) from long-term tillage experiments and areas covered by secondary natural forest were collected in the Brazilian savanna region (Cerrado) at Santo Antônio de Goiás, and southern Atlantic forest region at Londrina. Bulk soils were separated into 8 aggregate size fractions (ASF) by wet sieving (n = 700). Bulk soil samples (BS) and ASF were analyzed for total C, N by dry combustion. BS were also analyzed for sand, silt, and clay by densitometry, and several measures of soil aggregation were determined. All samples were scanned in the mid- and near-infrared (Mid-IR, NIR) by diffuse reflectance of ground, non-KBr diluted samples using a Digilab FTS-7000 FTIR spectrometer equipped with a Pike Auto-Diff auto-sampler. Quantitative calibrations for the various analytes were performed separately for the original set of 120 BS samples and the set of 700 ASF samples using partial least squares regression (PLSR). Results for the set of 120 BS for total C and N were excellent using either spectral range with R2 > 0.97, except for C by Mid-IR (R2 = 0.93). Similar results were obtained for determinations of sand and clay (R2 > 0.94). Silt, determined by difference (100% − %sand − %clay), gave much lower R2 (0.80 and 0.63 for Mid-IR and NIR, respectively). For aggregation indices, R2 of ∼ 0.80 were achieved in the Mid-IR, NIR results were not as good (R2 < 0.67). Efforts with the 700 ASF samples resulted in R2 of 0.98 and 0.96 for C, and 0.96 and 0.97 for N, for Mid-IR and NIR calibrations, respectively. Calibration efforts to determine the values for the set of 700 ASF samples using set of 120 BS from which they were fractionated gave R2 = 0.90 and 0.87 in the Mid-IR and NIR, respectively, when 5% of the ASF were added to the set of 120 BS. Spectral discrimination of sample source was easily and very accurately determined for all data sets examined.