A multivariate approach to the analysis of pine needle samples using NIR

Near infrared reflectance (NIR) spectroscopy was used to the determine the concentrations of nitrogen, starch and various carbohydrates in milled Scot s pine needle samples. A multivariate calibration using partial least squar e regression (PLS) was used. To remove variation due to light scattering, w hich is particularly difficult to handle in diffuse reflectance spectroscop y, multiplicative scatter correction (MSC) was in the case of nitrogen and carbohydrate analyses and the second-order derivation of the spectra for st arch was used. The NIR prediction was good for nitrogen and somewhat less f or starch. The predictability (Q(2)) was 0.83 for nitrogen and 0.86 for sta rch, and the root mean square error of prediction (RMSEP) was 0.06 for nitr ogen and 0.91 for starch. The preliminary results of the carbohydrate model indicated that MR spectroscopy has a potentially useful role in the measur ement of the carbohydrate content of the pine needle samples; however, furt her development of the method is still necessary to reach acceptable accura cy. (C) 1998 Elsevier Science B.V. All rights reserved.