David A. Reckhow
Conventional drinking water treatment generally requires maintenance of a disinfectant residual, most commonly free or combined chlorine. Since the discovery of disinfection byproducts (DBPs), utilities have relied on a broad range of water quality parameters to help them achieve compliance with DBP regulations. Most notably, dissolved organic carbon (DOC) and ultraviolet absorbance at 254 nm (UV254) have been used to assess DBP precursor levels in near-real time. These are often compared to actual system DBP concentrations or to results from laboratory DBP precursor tests resulting in a site-specific precursor-DBP model. The DOC or UV254 models are often inaccurate, requiring re-calibration, especially in the face of changing water quality. For this reason, there is an interest in developing better real-time precursor monitoring strategies and these new methods will be compared against a DOC and UV254 predictive model from the same water source. This project focuses on two new near real-time technologies that are intended to allow better prediction of the conventional 72-hour lab based THM precursor test. The first non- conventional methodology, the AMS method, is integrated with a commercially-available THM monitoring instrument. This instrument automatically chlorinates a sample of river water and incubates the sample at an elevated temperature for approximately one hour before using a purge and trap method to collect the THMs followed by a colorimetric method to quantify and analyze THM concentrations. The second technology, the accelerated lab method, employs a reaction vessel modified for temperature fluctuations in which sample and chlorine are added. The sample is then incubated at an elevated temperature for one hour. Next, the sample is sent to a gas chromatograph for THM analysis and residual chlorine is measured. The accelerated lab method was the most successful in predicting THM concentrations formed by the standard lab method with an average absolute error of 11 ppb and a median absolute error of 9 ppb. The second most successful method of predicting standard THM formation potential was an equation that resulted from a multivariate analysis of TOC and UV254 values taken for the Mill River in Hadley, MA. This equation predicted concentrations produced by the standard lab method with an average absolute error of 15 ppb and a median absolute error of 10 ppb. The third best method of predicting standard method THM concentrations was another multivariate analysis, this time using the log of standard lab chlorine demand and UV254 measurements. This method predicted standard lab THM concentrations with an average absolute error of 15 ppb and a median absolute error of 11 ppb. The fourth most successful method for predicting standard THM concentrations was the AMS THM-100 with an average absolute error of 20 ppb and a median absolute error of 19 ppb. Each of these methods had a sample size of 34 data points and the data points used were identical across the methods to mitigate sample bias. While the absolute errors between the methods may indicate a clear choice of methodology in terms of accuracy, each method has certain aspects that are beneficial in unique ways.