Application of chemometrics to identify artificial ripening in sapota (Manilkara Zapota) using visible near infrared absorbance spectra

Abstract

Artificial ripening of climacteric fruits by industrial-grade calcium carbide (CaC2) is a common practice in developing countries that introduces toxic substances (e.g. arsenic, lead, etc.) into the fruit. However, detecting artificially ripened sapota in the laboratory is destructive, time-consuming, and skill dependent. In the present study, Vis-NIR (350–2500 nm) absorbance spectra along with chemometrics was used to detect arsenic on the CaC2 induced artificial ripened sapota. The spectral data were pre-processed using standard normal variate (SNV), multiplicative scatter correction (MSC), SNV + MSC, 1st derivative and SG derivative -127 (Savitzky-Golay-1st derivative; 2nd order; window size-7). Raw data and pre-processed data were used to classify the artificially ripened sapota from naturally ripened ones by using supervised classification techniques, - Partial Least Square –Discriminant Analysis (PLS-DA); Support Vector Machine (SVM) and k-Nearest Neighbour (k-NN). Principal component analysis (PCA) was conducted on the absorbance of entire raw spectra, the first three principal components explained more than 97% variance. PC1 (68.74%) and PC3 (7.6%) denoted the product characteristics, while PC2 (21.46%) explained the ripening process. The wavelengths which have higher loadings on PC2 (716, 1066 and 1438 nm) were selected as the optimum wavelengths for the detection of arsenic residues deposited over the fruit by CaC2. The classification accuracy of artificially ripened sapota was 100% for PLS-DA, SVM and k-NN models with raw, MSC, SNV, MSC + SNV and SG derivative-127 pre-processed spectral data except for raw data classified with k-NN and MSC with SVM. The performance of all classification models decreased with 1st derivative pre-processing.