The obtained NIR spectra (Figure 2) were used to create prediction models for the different reference parameters. An external validation set was used to verify the predictive performance of the calculated prediction models if sufficient samples were available. Otherwise an internal cross-validation was conducted.

Correlation diagrams which display the relation between the NIR prediction and the reference values are shown in Figures 3–12 together with the respective figures of merit (FOM). For the alpha acid parameter with reference data obtained according to EBC 7.5, a model was created covering the full range (2.40–21.74%). Furthermore, a model for only the lower range (2.43–4.26%) was created to highlight the possibility of reducing the prediction error by creating sub-models for narrower concentration ranges. 

This Application Note demonstrates the feasibility of using NIR spectroscopy for the analysis of freshly harvested hops and ground hop pellets to determine cohumulone, moisture, and total oil content, HSI, and alpha and beta acids.

Using alpha acids as an example, the difference between a prediction model covering the full concentration range and a model specific to a narrower range was highlighted. The latter reduced the prediction error by a factor of two. Users can apply this approach to maximize accuracy. In the OMNIS Software, both models can be combined to automatically perform the following sequence. First, the alpha acid concentration is estimated using the broad-range model; then, a more precise prediction is made using the restricted-range model.

Since NIRS measurements are rapid and require no chemicals, the method offers significant time- and cost-savings, as shown in Table 1.

Furthermore, the simplicity of the analysis and the availability of a pre-calibrated instrument, using the pre-calibration, hops, solid (6.06008.022), make both the implementation and the application of NIR spectroscopy with OMNIS NIRS straightforward.