Effects of sample preparation and measurement standardization on the NIRS calibration quality of nitrogen, ash and NDFom content in extensive experimental grassland biomass

Near infrared spectroscopy (NIRS) is a common method to analyze the quality of grassland biomass. However, the effort required for sample preparation and measurement can restrict the sampling rate and prediction quality. To assess suitable approaches for animal feeding and bioenergy recovery of grassland biomass, NIRS calibrations for N, ash and ash-free neutral detergent fiber content (NDFom) of hay, silage and standing swards of botanically diverse grassland communities were developed. Seven methods, representing different combinations of measurement conditions and sample preparation, were applied on silages, hay and standing sward and compared using a bench-top system (FOSS XDS Rapid Content Analyzer) and a field spectrometer system (ASD Fieldspec 3). Considering standard error of cross validation (SECV) and its inverse ratio to sample standard deviation (RPD) as the main parameters for comparison, best results were acquired for all constituents at laboratory measurement conditions with dried and ground samples on the bench-top system (RPD >3) with a coefficient of determination R-2 <0.9. The lower degree of standardization of measurement conditions on the field spectrometric methods lead to lower RPD values and a higher SECV compared to the laboratory approaches. Predictive accuracy of calibrations for nitrogen and NDFom using hay of greater particle size were still acceptable for a farm-scale application in combination with a field spectrometer (RPD >2). Calibration of ash content was challenged by high ash contamination during sampling and open canopies in the field which resulted in lower calibration quality especially on silages and standing sward. Predictions for silages and standing sward were poor probably due to moisture and structural heterogeneity and, thus, only suitable for a crude high-low differentiation. (C) 2013 Elsevier B.V. All rights reserved.