Determining organophosphorus pesticides in agriculture: A combined approach of ion-mobility spectrometry with robust principal component analysis and multivariate adaptive regression splines

Abstract

The high toxicity and widespread use of organophosphorus pesticides (OPPs) in agriculture make their accurate detection and quantification a critical challenge. Traditional analytical techniques like gas chromatography (GC) and high-performance liquid chromatography (HPLC) face limitations due to their cost, time-consuming procedures, and labor intensity. This study explores a novel analytical approach that utilizes robust principal component analysis (rPCA) and multivariate adaptive regression splines (MARS) to enable the determination of OPPs using ion mobility spectrometry (IMS). IMS data were compressed using rPCA to identify the principal components (PCs) that best capture the relevant information.

Kenard Stone algorithm was employed to create the calibration and test sets for model development and validation, respectively. The calibration set (containing 35 samples and 6 PCs) was used to train the rPCA-MARS model. Principal Component Regression (PCR) and Partial Least Squares Regression (PLS-R) models were compared for their ability to predict the quantitative values of OPPs to the rPCA-MARS model. The efficiency of the rPCA-MARS model was evaluated using several metrics: R-squared (R²), R² estimated by generalized cross-validation (R²_GCV), adjusted R-squared (R²_adj), sum of squared errors (SSE), and mean square error (MSE). The optimal rPCA-MARS model utilized 7 basis functions to effectively characterize the OPPs values.

The linear rPCA-MARS model for Ethion performs well on both the calibration and test sets. The piecewise-cubic rPCA-MARS model achieved excellent performance on the calibration set, with R² = 0.995, R²_adj = 0.994, and SSE = 0.368. The test set results were equally impressive, showing R² = 0.993, R²_adj = 0.993, and SSE = 0.220. The cubic rPCA-MARS model exhibited exceptional predictive performance and generalizability, achieving a low MSE of 0.012 and a high R²_GCV of 0.992. These results underscore the superior predictive capability of the rPCA-MARS framework for Ethion determination in this study. Building on the success with Ethion, the rPCA-MARS model shows promise for predicting concentrations of Malathion and Phosalone.This finding highlights the model's potential for broader applications in OPPs analysis in agriculture. This paves the way for developing rapid, cost-effective, and environmentally friendly methods for monitoring and managing OPPs within agricultural ecosystems.