Skip to main content
ARS Home » Research » Publications at this Location » Publication #164584
Title: QUANTITATIVE MID INFRARED SPECTROSCOPY AS A VIABLE ALTERNATIVE TO NEAR INFRARED SPECTROSCOPY FOR THE DETERMINATION OF PLANT TISSUE CHEMISTRY: A TEST USING FOLIAGE SAMPLES FROM TWO CONIFER SPECIES

Author
item RICHARDSON, ANDREW - UNIV OF NEW HAMPSHIRE
item Reeves Iii, James

Submitted to: Canadian Journal of Forest Research
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 11/30/2004
Publication Date: 1/20/2005
Citation: Richardson, A.D., Reeves III, J.B. 2005. Quantitative mid infrared spectroscopy as a viable alternative to near infrared spectroscopy for the determination of plant tissue chemistry: a test using foliage samples from two conifer species. Canadian Journal of Forest Research. 35:1122-1130.

Interpretive Summary: Spectroscopy is a process by which light at various wavelengths can be used to predict/determine the composition of materials such as forages, grains, etc. The objective of this study was to use reflectance spectroscopy to determine the chemical composition (nutrients, carbon, and fiber constituents: NDF, ADF, cellulose and lignin) and gross sample characteristics (species, crown position, elevation, location) of dried conifer foliage samples (e.g., pine needles) and to compare the accuracy and robustness (how well the equations can predict new samples) of calibrations based on three different spectral regions: visible and shortwave near infrared (VIS/sNIR) wavelengths), near infrared (NIR) and mid infrared (MIR). Calibrations based on NIR and MIR spectra both performed better than those based on VIS/sNIR spectra. With the exception of magnesium, at least one acceptable calibration was obtained for each analyte. Universally strong calibrations for all three spectral regions were obtained for total nitrogen, NDF, ADF and cellulose. For 11 out of 15 analytes, the MIR spectra produced the best calibration, but in most cases the difference in accuracy of the MIR vs. NIR calibrations was small. Against an independent validation set, the MIR calibrations were somewhat less robust than the NIR calibrations for some (e.g. NDF), but not all, analytes. None of the calibrations for lignin were robust. Quantitative MIR spectroscopy is at a more primitive state of development than NIR spectroscopy, but these results suggest that MIR spectroscopy is not only a viable, but also potentially superior, alternative.

Technical Abstract: The objective of this study was to use reflectance spectroscopy to determine the chemical composition (nutrients, carbon, and fiber constituents) and gross sample characteristics (species, crown position, elevation, location) of dried conifer foliage samples (N = 72), and to compare the accuracy and robustness of calibrations based on three different spectral regions: visible and shortwave near infrared (VIS/sNIR, 400'1100 nm wavelengths), near infrared (NIR, 1100'2500 nm) and mid infrared (MIR, 2500'25,000 nm). Calibrations based on NIR and MIR spectra both performed better than those based on VIS/sNIR spectra. With the exception of Mg (best calibration Rsquare = 0.46), at least one acceptable calibration Rsquare > or = to 0.80) was obtained for each analyte. Universally strong calibrations (Rsquare > or = to 0.85 for all three spectral regions) were obtained for N, NDF, ADF and cellulose. For 11 out of 15 analytes, the MIR spectra produced the best calibration, but in most cases the difference in accuracy of the MIR vs. NIR calibrations was small. Against an independent validation set, the MIR calibrations were somewhat less robust than the NIR calibrations for some (e.g. NDF), but not all, analytes. None of the calibrations for lignin were robust. Quantitative MIR spectroscopy is at a more primitive state of development than NIR spectroscopy, but these results suggest that MIR spectroscopy is not only a viable, but also potentially superior, alternative.