Generic Prediction of Soil Organic Carbon in Alfisols Using Diffuse Reflectance Fourier-Transform Mid-Infrared Spectroscopy uri icon

abstract

  • Partial least squares regression (PLSR) analysis was used for the generic prediction of soil organic C (C-org) content in Alfisols using diffuse reflectance Fourier-transform mid-infrared spectroscopy (DRIFT-MIRS). For prediction improvement, we tested the potential of C-org removal by heating and spectral subtraction of mineralogic signals in DRIFT-MIRS spectra. Eighty-seven Alfisol samples (C-org range 1-32 g kg(-1)) of contrasting geographic origin were heated (550 degrees C) or left untreated (NT). Spectra from DRIFT-MIRS were recorded and spectral subtraction was performed by taking the difference between DRIFT-MIRS spectra of NT and heated soils. Prediction of C-org by PLSR was acceptable for NT (R-2 = 0.93; root mean square error of cross validation = 0.20; residual prediction deviation = 3.7) but was not improved by heating and spectral subtraction. Additional studies are recommended to substantiate the potential of PLSR-based, generic quantification of C-org in unheated soils for soil groups other than Alfisols and to assess if heating is suited to advance generic prediction of C-org across soil orders with widely different mineralogic signatures.
  • Partial least squares regression (PLSR) analysis was used for the generic prediction of soil organic C (Corg) content in Alfisols using diffuse reflectance Fourier-transform mid-infrared spectroscopy (DRIFT-MIRS). For prediction improvement, we tested the potential of Corg removal by heating and spectral subtraction of mineralogic signals in DRIFT-MIRS spectra. Eighty-seven Alfisol samples (Corg range 1â??32 g kgâ??1) of contrasting geographic origin were heated (550°C) or left untreated (NT). Spectra from DRIFT-MIRS were recorded and spectral subtraction was performed by taking the difference between DRIFT-MIRS spectra of NT and heated soils. Prediction of Corg by PLSR was acceptable for NT (R2 = 0.93; root mean square error of cross validation = 0.20; residual prediction deviation = 3.7) but was not improved by heating and spectral subtraction. Additional studies are recommended to substantiate the potential of PLSR-based, generic quantification of Corg in unheated soils for soil groups other than Alfisols and to assess if heating is suited to advance generic prediction of Corg across soil orders with widely different mineralogic signatures

publication date

  • 2011
  • 2011
  • 2011