Glyphosate detection by voltammetric techniques. A comparison between statistical methods and an artificial neural network

Abstract

Glyphosate quantification methods are complex and expensive, and its control in natural water bodies is getting more important year after year. In order to find a new system that facilitates the detection of glyphosate, we present a comparison between two models to predict glyphosate concentration in aqueous dissolutions. One of them is done by an artificial neural network (ANN) embedded in a microcontroller and the other one is done by statistic methods (Partial Least Squares) in a computer. From an analytical point of view, voltammetric techniques have been selected to obtain electrochemical responses to different glyphosate concentrations. In order to get them, a voltammetry/amperometry device designed by our research group and called FraPlus has been used. In this work we have selected two sensitive electrodes (cobalt and copper), 10 different glyphosate concentrations and a train pulse made by nine different voltammetric pulses to build the models. The ANN developed model is able to properly relate data obtained by FraPlus and glyphosate concentrations and the obtained value for regression coefficient (R) is 0.9998 and the P-value is 0.0. Taking into account these results, we propose this ANN model based in voltammetric techniques working with glyphosate concentrations in a buffer as an approach to glyphosate detection in natural water bodies.