Facile detection of illicit wastewater discharge into a water source using a kinetic-based optical fingerprinting strategy.

Abstract

Extensive monitoring of water sources necessitates the development of inexpensive and effective methods for monitoring their pollution. A particularly challenging task is detecting a sudden release of contaminated effluents into a water supply. To solve this issue, we employ a reaction-based fingerprinting technique that is based on conducting an indicator reaction of oxidation of carbocyanine dyes in the presence of a sample. The absorbance and fluorescence intensity are measured periodically using cameras, and the obtained data are processed using machine learning techniques. Monitoring of clean tap or river water was simulated by sampling every few days. Artificial contamination of this water was modeled by adding diluted sewage water (4 different samples). As a result, the contaminated samples were displayed as outliers in the score plots. In both tap and river water, 0.1% vol of wastewater (1000-fold dilution) was detected. The accuracy of discrimination between polluted and unpolluted samples exceeded 90% using linear discriminant analysis (LDA) or softmax regression (SR). Thereby, an unexpected discharge of wastewater into a water source could be rapidly detected with simple instruments. Development of this approach will contribute to improving the accuracy and ease of detection of water source contamination, making environmental monitoring methods more reliable for the benefit of public health.