Portable Sensor Array for On-Site Detection and Discrimination of Pesticides and Herbicides Using Multivariate Analysis

Modern agricultural practice relies heavily on pesticides and herbicides to increase crop productivity, and consequently, their residues have a negative impact on the environment and public health. Thus, keeping these issues in account, herein we developed an azodye-based chromogenic sensor array for the detection and discrimination of pesticides and herbicides in food and soil samples, utilizing machine learning approaches such as hierarchical clustering analysis, principal component analysis, linear discriminant analysis (LDA), and partial least square regression (PLSR). The azodye-based sensor array was developed in combination with various metal ions owing to their different photophysical properties, which led to distinct patterns toward various pesticides and herbicides. The obtained distinct patterns were recognized and processed through automated multivariate analysis, which enables the selective and sensitive identification and discrimination of various target analytes. Further, the qualitative and quantitative determination of target analytes were performed using LDA and PLSR; the results obtained show a linear correlation with varied concentrations of target analytes with R-2 values from 0.89 to 0.96, the limit of detection from 5.3 to 11.8 ppm with a linear working range from 1 to 30 mu M toward analytes under investigation. Further, the developed sensor array was successfully utilized for the discrimination of a binary mixture of pesticide (chlorpyrifos) and herbicide (glyphosate).

Automated summary

Modern agricultural practice heavily relies on pesticides and herbicides, which have negative impacts on the environment and public health. In this study, an azodye-based chromogenic sensor array combined with machine learning approaches was developed for the selective and sensitive detection and discrimination of pesticides and herbicides. The sensor array, utilizing the different photophysical properties of various metal ions, showed distinct patterns towards different target analytes. Automated multivariate analysis, such as hierarchical clustering analysis, principal component analysis, linear discriminant analysis, and partial least square regression, was employed for the recognition and processing of the obtained patterns. The developed sensor array was successfully used for the qualitative and quantitative determination of target analytes, showing good linear correlation and detection limits.