Use of near infrared reflectance spectroscopy to predict nitrogen uptake by winter wheat within fields with high variability in organic matter

In this study, the ability to predict N-uptake in winter wheat crops using NIR-spectroscopy on soil samples was evaluated. Soil samples were taken from unfertilized plots in one winter wheat field for three years (1997-1999) and in another winter wheat field nearby for one year (2000). Soil samples were analyzed for organic C content and their NIR-spectra. N-uptake was measured as total N-content in aboveground plant materials at harvest. Models calibrated to predict N-uptake were internally cross-validated and validated across years and across fields. Cross-validated calibrations predicted N-uptake with an average error of 12.1 to 15.4 kg N ha(-1). The standard deviation divided by this error (RPD) ranged between 1.9 and 2.5. In comparison, the corresponding calibrations based on organic C alone had an error from 11.7 to 28.2 kg N ha(-1) and RPDs from 1.3 to 2.5. In three of four annual calibrations within a field, the NIR based calibrations worked better than the organic C based calibrations. The prediction of N-uptake across years, but within a field, worked slightly better with an organic C based calibration than with a NIR based one, RPD, = 1.9 and 1.7, respectively. Across fields, the corresponding difference was large in favour of the NIR-calibration, RPD = 2.5 for the NIR-calibration and 1.5 for the organic C calibration. It was concluded that NIR-spectroscopy integrates information about organic C with other relevant soil components and therefore has a good potential to predict complex functions of soils such as N-mineralization. A relatively good agreement of spectral relationships to parameters related to the N-mineralization of datasets across the world suggests that more general models can be calibrated.