Development and validation of a method for determining pesticides in groundwater from complex overlapped HPLC signals and multivariate curve resolution

A method is presented to determine pesticides in groundwater by way of multivariate curve resolution alternating least squares (MCR-ALS) mixed bilinear–trilinear models of overlapped chromatographic second-order data (HPLC-DAD). Four of the pesticides in the samples showed highly overlapped spectra, which made the quantification especially difficult. Different calibration and validation sets were used to test the proper number and distribution of the samples. Performance characteristics (figures of merit) such as sensitivity, precision and limit of detection were assessed from the calibration mixtures. Accuracy in the predictions was estimated in terms of the root mean square of percentage deviation (RMSPD) and explained variance (Q2). We obtained values of RMSPD of 7%, 10% and 8% and Q2-values of 99%, 93% and 99% for the pesticides vinclozolin, chlorfenvinphos and parathion-ethyl, respectively. A set of natural groundwater samples was also used to validate the model and to assess the fitness of the method for quantification of the pesticides in natural samples. Three of the pesticides under study were satisfactorily resolved and quantified in the groundwater samples by the proposed procedure achieving 88%, 96% and 94% of the explained variance for the pesticides vinclozolin, chlorfenvinphos and parathion-ethyl, respectively.