Relating organic matter character to trihalomethanes formation potential: a data mining approach

Abstract

The removal of natural organic matter (NOM) during water treatment is becoming more important for all water utilities in the UK, as a result of tightened regulatory standards for trihalomethanes (THM), disinfection by-products (DBP) formed when residual organics react with chlorine. This paper considers the spatial and temporal variability of raw and clarified water arising from 16 surface water treatment works in the Midlands region of the UK. A wide range of investigation techniques are applied in order to develop effective strategies for the treatment of NOM-rich water. For the first time, rigorous data mining techniques are applied to a major dataset in order to examine potential inter-relationships between a wide range of quality parameters including, inter alia, total organic carbon (TOC), UV254, coagulation pH, resin fractionation (hydrophilic acids (HPIA), hydrophobic acids (HPOA), hydrophilic non-acids (HPINA)) and total THM formation potential (TTHMFP). This paper focuses on the use of principal component analysis (PCA) to develop robust algorithms for the prediction of TOC removal and hence THM formation. Results show that raw water characteristics can be categorised into three main types, according to their HPOA content and specific ultraviolet absorbance (SUVA.). PCA identified possible THMFP precursors, according to raw water type verified by strong statistical relationships.