A New Post-Processing Technique to Improve the Use of EEMs to Predict DBP Formation Potential
- Ben Stanford, Justin Irving, Allison Reinert, Ben Wright, Bill Becker - Hazen and Sawyer
Through two recently-completed studies funded by Water Research Foundation and the New York State Energy Research and Development Authority (NYSERDA), we have investigated the use of field-deployable and laboratory-based sensors to characterize natural organic matter across a diverse set of upstate New York reservoirs. While each of the two studies have had different focus (one examining the impact of climate change on water quality and disinfection byproducts (DBPs), the other developing tools to support water use in a multi-reservoir system), the combined data sets have both been used to develop predictive models for trihalomethanes (THMs) and haloacetic acids (HAAs) from those supplies.
In past studies, 3-D fluorescence excitation-emission matrix (EEM) spectroscopy has been used as a tool to gather information about the character of natural organic matter (NOM) in water supplies and various techniques including parallel factor analysis (PARAFAC), fluorescence regional integration, peak picking, and principle component analysis (PCA) have been applied to develop predictive models for DBP formation. However, these techniques have generally been limited in their success in that they may work for one particular type of DBP (e.g., chloroform), one class of DBPs (e.g., THMs), or in one or two reservoirs only. Also, techniques such as PARAFAC require hundreds of samples to be collected from one source to develop a reliable selection of components.
In this study, we were able to capitalize on the fact that water quality data are most typically log-normally distributed, thereby allowing us to develop a new post-processing technique for EEMs that resulted in models that were effective across all of the reservoirs tested and could predict both THMs and HAAs with Pearson Correlation coefficients of greater than 0.87 HAAs and 0.90 for THMs across seventy-one (71) source water samples from seven (7) different locations. This technique was then further tested on a completely different water supply (a southern river) that resulted in an R2 value of up to 0.96 for THMs (HAAs were not tested at this second site). In this presentation we will provide details of the study design, a comparison of various models that were tested for DBP formation prediction, and the methodology developed for the post-processing technique used to develop our models for DBP formation.
For more information, please contact the author at email@example.com.
Hear about new publications with our email newsletter
Horizons showcases significant water, wastewater, reuse, and stormwater projects and innovations that help our clients to achieve their goals, and can help you achieve yours. Articles are written by top engineers and process group leaders, demonstrating and explaining the beneficial application of a variety of technologies and tools.