Evaluating chemometric techniques for non-destructive detection of glyphosate residues in single pulse grains by using FTIR spectroscopy

Abstract

The measurement of pesticide content in pulses at various stages of the supply chain is important in order to manage the maximum residue level (MRL) set by different government agencies. The objective of this study was to develop a non-destructive detection system to determine the glyphosate content in 6 pulses (chickpea, yellow pea, red lentil, large green lentil, French green lentil, and black beluga lentil) based on Fourier transform infrared spectroscopy (FTIR). Organically grown pulses were artificially spiked with glyphosate at 5 concentrations (0 mg/kg, 5 mg/kg, 10 mg/kg, 15 mg/kg and 20 mg/kg) and used for the development and testing of FTIR spectroscopy and associated chemometric models. Principal component analysis (PCA) led to the discrimination and clustering in the pulse samples based on the applied glyphosate levels. Various preprocessing and variable selection techniques were applied on the spectral dataset and partial least squares (PLS) regression was used to predict the glyphosate levels in pulses. The correlation coefficient for prediction (Rp²) of glyphosate was 0.93, 0.92, 0.96, 0.91, 0.96, and 0.92 for yellow pea, chickpea, large green lentil, red lentil, black beluga, and French green lentil, respectively with optimized preprocessing and variable selection techniques.