Drinking water aromaticity and treatability is predicted by dissolved organic matter fluorescence

Samples from fifty-five surface water resources and twenty-five drinking water treatment plants in Europe, Africa, Asia, and USA were used to analyse the fluorescence composition of global surface waters and predict aromaticity and treatability from fluorescence excitation emission matrices. Nine underlying fluorescence components were identified in the dataset using parallel factor analysis (PARAFAC) and differences in aromaticity and treatability could be predicted from ratios between components H-ii (lambda(ex)/lambda(em)= 395/521), H-iii (lambda(ex)/lambda(em)= 330/404), P-i, (lambda(ex)/lambda(em)=290/365) and P-ii (lambda(ex)/lambda(em)= 275/302). Component Hii tracked humic acids of primarily plant origin, Hiii tracked weathered/oxidised humics and the "building block" fraction measured by LC-OCD, while Pi and Pii tracked amino acids in the "low molecular weight neutrals" LC-OCD fraction. Ratios between PARAFAC components predicted DOC removal at lab scale for French rivers in standardized tests involving coagulation, powdered activated carbon (PAC), chlorination, ion exchange (IEX), and ozonation, alone and in combination. The ratio Hii/Hiii, for convenience named "PARIX" standing for "PARAFAC index", predicted SUVA according to a simple relationship: SUVA = 4.0 x PARIX (RMSEp=0.55) Lmg(-1)m(-1). These results expand the utility of fluorescence spectroscopy in water treatment applications, by demonstrating the existence of previously unknown relationships between fluorescence composition, aromaticity and treatability that appear to hold across diverse surface waters at various stages of drinking water treatment.