Comparative evaluation of limit-of-detection-based and K-nearest-neighbors-based classification approaches for...

Rapid and low-cost detection of wastewater treatment plant (WWTP) effluent contamination in aquatic systems protects public health and environmental quality. While optical whiteners and fluorescence-derived organic compounds in treated wastewater effluent serve as detection surrogates, few long-term studies have employed machine-learning methods to evaluate fluorescence excitation-emission matrices (EEMs) for predicting wastewater impacts. This study compares limit of detection (LOD)-based and k-nearest neighbors (KNN)-based classification methods using fluorescence-derived indicators and conventional physicochemical parameters for reliable wastewater effluent detection. For the first time in Korea, EEMs were obtained from water samples collected over 20 months from two uncontaminated upstream sites (Bukcheon and Byeongseongcheon Rivers), WWTP effluent, and an effluent-impacted downstream site in Nakdong River tributaries. The LOD-based classification achieved accuracies of 82-94 % under stable baseline conditions (Bukcheon River) but declined to 71-76 % under variable baseline conditions (Byeongseongcheon River). In contrast, the KNN approach demonstrated greater robustness, achieving optimal accuracies of 98 % and 91 % at the two sites, respectively, while maintaining stable performance across parameter combinations. Feature importance analysis identified tryptophan-like (Peak T) and microbial humic-like (Peak M), were the most important classification features, while conductivity was the most influential physicochemical parameter. Monte Carlo sensitivity analysis revealed minimum detectable effluent mixing ratios of 13-23 % required to achieve 90 % classification accuracy, with KNN demonstrating greater accuracy in challenging scenarios. These findings demonstrate the value of combining advanced fluorescence spectroscopy with machine learning for improved water quality monitoring and the detection of wastewater effluent contamination.