Near infrared reflectance spectroscopy compared with soil clay and organic matter content for estimating within-field variation in N uptake in cereals

Adjusting fertiliser applications to within-field variations in nitrogen (N) mineralisation during the growing season can increase yields, improve crop quality, reduce costs and decrease nutrient losses to the environment. Predicting such variations at a reasonable cost is therefore important. In a 3-year study, Near Infrared Reflectance (NIR) spectroscopy was compared with soil organic matter (SOM) and clay content as predictors of plant N uptake using cross-validated PLS (Partial Least Squares) regression models. Plant N uptake was measured as total nitrogen in aboveground plant parts at harvest, in plots without N fertilisation within three different fields in southern Sweden. NIR spectroscopy and combined clay and SOM content resulted in equally good estimations of plant N uptake in fields with large variation in SOM content. Cross-validated NIR calibrations for plant N uptake within fields for separate years resulted in r2 values of 0.75–0.85 and average cross-validation errors of 11–16 kg N ha−1 for two fields (1 year excluded at one field because of farmyard manure application). No significant improvements were seen when NIR-spectra, clay and SOM were included in the same model, suggesting that the additional predictive capacity of NIR over SOM relates to soil texture variations. NIR calibrations also performed poorly in one field where plant N uptake could not be explained by SOM or clay content. Predictions within fields between years produced r2 values of 0.56–0.89 and prediction errors of 12–26 kg N ha−1 for one field. These results confirm that N uptake prediction accuracy can be improved by using NIR spectroscopy in fields with large SOM variations. However, good estimations could not be made between fields, indicating difficulties in creating more general calibration models for large geographical areas.