Applicability of NIR spectroscopy to determine oil and other physicochemical parameters in Rosa mosqueta and Chilean hazelnut

Due to the increasing interest in certain components, specially the oil, from non-conventional seeds as Rosa mosqueta (Rosa rubiginosa) and Chilean hazelnut (Gevuina avellana), quick determinations of oil and other parameters were carried out by using near-infrared (NIR) spectroscopy. Moisture, oil, fiber (as acid detergent fiber) and protein from solid samples of the seeds as mentioned, along with those of soybean (Glycine max), already analyzed by NIR and for serving as control for the variability of the method, were studied. Sample interactions to NIR radiations were processed using the multivariate regression algorithm Partial Least Squared (PLS) to build a calibration model. Standard error of cross-validation (SECV) was used to estimate the prediction error. Moisture of Rosa mosqueta, Chilean hazelnut presscake and soybean meal (in the ranges 10–15, 10–15, 8–10%, respectively), acid detergent fiber (60–68, 12–16, 10–15%, respectively), oil (1–4, 14–20, 5–13%, respectively) and protein (1–5, 8–15, 27–45%, respectively) were previously determined by wet analysis using standard methods, so creating a library. The possibility to analyze parameters from very different oilseeds with an acceptable uncertainty was also established. Standard errors of cross-validation were between 1.25 and 2.99%, being the oil content the best predicted parameter.