Using near-infrared reflectance spectroscopy to predict carbon, nitrogen and phosphorus content in heterogeneous plant material

Abstract

The aim of this study was to produce calibration equations between near-infrared reflectance (NIR) spectra and the concentrations of carbon, nitrogen, and phosphorus in heterogeneous material: from living needles to litter in Pinus halepensis stands subjected to prescribed burnings. The aim was to determine whether calibrations should be conducted within each stage in the transformation of needles (local calibrations), giving relationships that were accurate but valid only for each particular stage, or whether it was possible to integrate the various forms of variation in needles (global calibrations) while retaining an acceptable accuracy. A principal component analysis calculated from the sample spectral data was used to distinguish three different sets, each sharing spectral characteristics and corresponding to three categories of needle: needles collected on the pines (N), falling needles (F), and litter (L), and each containing samples collected from the burnt sites and a control site. Samples representative of all the forms of variation in spectral properties were selected from within each category and their carbon, nitrogen, and phosphorus concentrations were measured using standard wet chemistry methods; these constituted the calibration sets n, f, and l. Calibrations were produced between the nutrient concentrations and the NIR spectra of the calibration sets n, f, and l and the grouped sets (n+f, f+l, n+f+l). The results of local calibrations made from each individual category showed that the carbon, nitrogen, and phosphorus concentrations were accurately predictable by NIR spectra. The global calibrations made by lumping together several categories were valid for a wider range of concentrations and for spectrally heterogeneous materials and in most cases were just as accurate as the local calibrations produced from each individual category.