Haloacetic acids (HAAs) are formed during the chlorination of water containing organic matter but predictive models for their formation are not well established at present and have not been applied to UK water systems. In this paper two different modelling approaches have been applied and evaluated to a selected UK drinking water system. The first approach involved the development of statistical predictive models for the HAA formation by using multiple linear regression with appropriate water quality and operational parameters obtained from sampling, and the second approach employed the simulated distribution system (SDS) test. Statistically significant predictors were trihalomethane (THM) levels, pH, temperature, total chlorine total organic carbon, UV254, bromide concentration and residence time, but the importance of each varied with HAA species. The models that generally explained most of the variance of individual and total HAAs included THMs as a predictor variable. The use of the SDS-HAA test included quantification of individual HAA species. The concentration of total HAA and individual compounds in the SDS test and field samples were comparable immediately after chlorination, but with increasing residence time the concentration of HAAs in the selected water distribution system were greater than those found in the SDS test.
Keywords: disinfection by-products, distribution system, drinking water, haloacetic acids, multiple linear regression models