A Rapid Detection Method for Atrazine, Diquat, and Glyphosate in Water Using Radar Fingerprint Spectra of Algal Photosynthetic Fluorescence Parameters

As typical pesticide pollutants in water environments, herbicides pose a serious threat to the safety of aquatic ecosystems. However, effective methods for rapid screening are still lacking at present. This study developed a stepwise prediction method for atrazine, diquat, and glyphosate using radar fingerprint spectra derived from algal photosynthetic fluorescence responses. Microcystis aeruginosa was selected as the indicator alga from four freshwater algae. Using a strategy of “clustering dimension reduction and comprehensive scoring”, 11 parameters were selected from 26 JIP test parameters, and the optimal exposure time was determined to be 60   min. Radar fingerprint spectra were constructed from the selected parameters, and radial, geometric, and template features were extracted. Four classification and four regression algorithms were evaluated across seven feature sets using nested cross validation. The selected combinations were integrated into a stepwise prediction method following an “identification first, then quantification” strategy. Nested cross validation yielded an identification accuracy of 91.11% and R² values of 0.979, 0.790, and 0.982 for atrazine, diquat, and glyphosate, respectively. The final method achieved an independent validation accuracy of 96.67%, and the corresponding LOQs were 1   μg∙L -1 , 1   mg∙L -1 , and 6   mg∙L -1 , respectively. In filtered actual water samples individually spiked with one target herbicide, mean recoveries ranged from 94.20% to 105.65%. This study provides a novel method for translating algal physiological stress responses into structured spectral features, thereby providing methodological support for rapid detection of herbicides and for emergency monitoring and early warning of water pollution.