This research examined the use and advancement of fluorescence spectroscopy as an organic characterization method in drinking water treatment, providing novel insight into the performance of and fundamental mechanisms of water treatment processes. Using fluorescence spectroscopy coupled with analysis techniques including parallel factors analysis (PARAFAC) and peak shifts, biofiltration was found to have variable impact on individual fluorophores. The fluorescence method identified production of humic-like matter by the microbial communities, ultimately resulting in a unique treated organic character of the treated water. Through correlations with formation potentials of halogenated furanones, polysaccharides were identified as possible precursors. Pre-oxidation, was suggested to result in increased proportionality of carbonyl-containing functional groups and greater carbon oxidative state.

A continuous fluorescence system was developed as part of this research and implemented in two studies focused on fouling mitigation of ultrafiltration (UF) membranes. A full-scale study was conducted that continuously monitored membrane feed water organic character. Utilizing the continuous fluorescence, improved prediction accuracy of membrane fouling was found using a neural network approach. A second study, conducted at bench-scale focused on understanding the role of organic surface changes and irreversible fouling potential. Low coagulant doses (< 0.043 mg/L Al³⁺) were found to optimally reduce irreversible fouling rates using two surface waters. At low coagulant doses, surface fluorescence of fouled membranes indicated increased tryptophan adsorption and possible greater reversibility of protein-like fouling. Coagulant doses < 0.043 mg/L Al³⁺ represent a significant reduction to those typically used in drinking water operations and presents a means for reducing operational cost while improving membrane production capacity.

Research was also carried out identifying possible methods for detecting low levels (< 1% by volume) of wastewater impact on source waters. Comparison of several analysis approaches for interpreting fluorescence data were compared. Use of the full fluorescence spectra as input to a support vector machine resulted in the greatest prediction accuracy (66.7% over 4 classes). This suggests loss of important distinguishing features by pre-processing with dimensionality reduction methods. Use of fluorescence spectroscopy is shown to be a promising method for future application in pollution detection and source water impact assessment.