Determination of Chemical Properties in Jatropha Curcas L. Seed IP-3P by Partial Least-Squares Regression and Near-Infrared Reflectance Spectroscopy

The objective of this study was to assessnear infrared reflectance (NIR) method for predicting chemical properties of Jatrophacurcas L.in form of powderas a first step for development of a non destructive method.The chemical properties of Jatrophacurcas L. predicted were moisture, free fatty acid (FFA), and oilcontents. The reflectance of powders were measured in the wavelengths from 1000 to 2500 nm usingNIRFlex Solids Petri Apparatus. PartialLeast Squares (PLS)method isused for calibration of NIR and chemical data. Three data pre-treatment for NIR data were studied for obtaining the best calibration, namelynormalization, first derivative Savitzky–Golay 9 points, and combination both of them.The calibration model performance were inspected by precision (standard error of calibration and coefficient of variability should be as close to zero as possible); accuracy (V-set Bias should be as close to zero as possible), regression coefficient and coefficient of determination should be as close to one as possible.Relative prediction deviation (RPD) values higher than 2.0. In the general,it was found that the correlation coefficients between the reference values and NIR predicted values were higher than 0.88 for all variables (r=0.97 for moisture content, r=0.91 for oil content and 0.88 for FFA content) indicating the robustness of calibration model. The coefficient of varians(CV) of calibration model of moisture, oil and FFA content were 5.4 %, 5.1 % and 7.1 % respectively. The RPD of calibration model of moisture, oil and FFA content were 3.9 %, 2.5 % and 2.13 %, respectively. The best calibration model of NIR for moistureand oil contents using the PLS method and the data pre-treatments of the combination by normalization between 0 to 1 and first derivative Savizky–Golay 9 points, except for FFA content was the first derivative Savitzky - Golay 9 points. The PLS factor different for each propertiesi.e.7, 4 and5for moisture, oil and FFA content, respectively. This results suggest that NIRmethod could be used to predict chemical properties of powder of Jatropha curcas L.