Further processing was performed using the Unscrambler X software (V.10.2, CAMO PROCESS AS, Oslo Norway). The standard normal variate (SNV) transformation [22] was applied to spectra from both types of measurements. Subsequently, regression models were developed using Partial Least Squares (PLS) regression [23], with the dry matter values determined by oven drying used as reference values. The optimal number of PLS factors was determined by leave-one-out cross validation. Root Mean Square Error of Cross Validation (RMSECV) was estimated by the following Eq