Near Infrared Reflectance Spectroscopy for Quantification of Crop Residue, Green Manure and Catch Crop C and N Fractions Governing Decomposition Dynamics in Soil

Abstract

For environmental, as well as agronomic reasons, the turnover of carbon (C) and nitrogen (N) from crop residues, catch crops and green manures incorporated into agricultural soils has attracted much attention. It has previously been found that the C and N content in fractions from stepwise chemical digestion of plant materials constitutes an adequate basis for describing a priori the degradability of both C and N in soil. However, the analyses involved are costly and, therefore, unlikely to be used routinely. The aim of the present work was to develop near infrared (NIR) calibrations for C and N fractions governing decomposition dynamics. Within the five Nordic countries, we sampled a uniquely broad-ranged collection representing most of the fresh and mature plant materials that may be incorporated into agricultural soils from temperate regions. The specific objectives of the current study were (1) to produce NIR calibrations with data on C and N in fractions obtained by stepwise chemical digestion (SCD); (2) to validate these calibrations on independent plant samples and (3) to compare the precision and robustness of these broad-based calibrations with calibrations derived from materials within a narrower quality range. According to an internal validation set, plant N, soluble N, cellulose C, holocellulose (hemicellulose + cellulose) C, soluble C and neutral detergent fibre (NDF) dry matter were the parameters best predicted ( r2 = 0.97, 0.95, 0.94, 0.91, 0.90 and 0.94, respectively). However, the calibrations for soluble C and NDF were regarded as unstable, as their validation statistics were substantially poorer than the calibration statistics. The calibrations for all structural N fractions and lignin C were considered poor ( r2 = 0.47–0.70). By comparing our broad-based calibrations for plant N and NDF with similar calibrations for a sample set representing a commercial forage database, it was evident that the broad-based calibrations predicted a narrow-based sample set better than vice versa. For plant N, the residual mean squared error of prediction ( RMSEP), when testing the broad-based calibration with the narrow-based validation set, was substantially smaller than the RMSEP obtained when validating the broad-based calibration internally (1.8 vs 2.7 mg Ng−1 dry matter). Overall, the calibrations that performed best were those concerning the parameters most strongly influencing C and N mineralisation from plant materials.