Drinking water treatment is a combination of physical and chemical
processes to improve the quality of the finished water (i.e. treated water). In
addition to particle removal, treatment plants must reduce the dissolved organic
matter (DOM) concentration. DOM must be reduced because it contains
disinfection byproduct (DBP) precursors and creates disinfectant demand in the
distribution system. Fluorescence measurements, including those used in this
study, offer a rapid but effective technique to measure and characterize DOM in
both the source and treated water, as well as assess powdered activated carbon
(PAC) performance. PAC is a widely used material for DOM and trace compound
removal. This research focused on two different aspects using fluorescence
metrics: (1) incorporating them into models for treated water quality prediction
and (2) to assess PAC performance and competitive adsorption from DOM.
The first part of this study focused on a modeling approach of treated
water quality data collected by former coworkers1,2 obtained from a water
treatment plant (WTP) in Ohio. The dataset includes over 800 fluorescence inplant
measurements from December 2009 to December 2012. Reduction of
disinfection by-product (DBP) formation potential was studied based on UV,
dissolved organic matter (DOM) and fluorescence-derived metrics. Parameters
measuring water quality including fluorescence intensity were collected before
and after coagulation and the DBP formation potential (DBPFP) was also
measured.A comparative study including different model functions of Gaussian,
Neural Networks, Linear Regression, Simple linear regression, SMOreg,
Bagging, M5Rules and trees was performed. The purpose of this work was to
examine fluorescence excitation emission matrix spectroscopy (EEMs) measures
as predictors and compare them with more general DOM measurement methods
(DOC and UV254). These models could then be used to monitor, model, and
manage DBPFP during water treatment processes. The correlation coefficient
values obtained from cross validation process of each function/model were
consistent. For models containing fluorescence measures, UV254 and pH the r
value changed from 0.87 (Gaussian) to 0.94 (Linear, Multilayer, Begging and M5
trees). Among all functions SMOreg were chosen regarding to linear-weighted
structure.
For all targets of DOC and HAA and THM formation potential, the
proportion of the weighted coefficients were reviewed. For HAA formation
potential, the C1 value (-0.3326) has an inverse relationship, while C2 (0.307),
C3 (0.0334) and UV254 (0.907) were all positive coefficients. For THM formation
potential, the C2 value (-0.4322) has an inverse relationship, while C1 (0.714),
UV254 (0.5618) and pH (0.0287) were all positive coefficients. The modeled
THM formation potential generally increases as C1, UV254, or pH increase.
While the modeled HAA formation potential tends to increase as C2, UV254, or
pH increase. For all DBPFP, UV showed to be the most significant among other
inputs. C1 and UV showed to be almost equally important in DOC production.The second part of this research focused on PAC performance and
competitive adsorption from DOM. PAC performance is generally a function of (1)
PAC properties (i.e. Iodine number, size distribution, moisture content), (2) DOMwater
quality, and (3) the target compound(s). All three factors can vary
significantly, making overall PAC performance difficult to predict. Batch
experiments were conducted using Akron WTP (Lake Rockwell) as the water
source in 2016-2017 and treated using various commercial PACs with varying
contact times and PAC dose. Fourteen different commercial PACs were tested at
different contact times (15, 30 and 60 minutes) prior to filtration. DBP precursor
fluorescence measures from a previously published model3 were used to
estimate DBPFP reduction. Depending on PAC dose (from 30 to 50 mg/l) and
contact time, removal of THM and/or HAA formation potential increased by 6%-
31% and 10-25% respectively.
In addition to batch testing of different PAC samples, a method for quality
assurance-quality control (QA-QC) was developed to investigate the variation of
PAC shipments. To compare PAC shipments, all the performance values were
divided by a reference PAC performance value. The resulting number is called
the normalized value, based on the measured value for water quality parameters
of C1, C2, UV removal as well as DBPFP at a fixed point in time. The normalized
value for these samples fluctuates between 0.75 and 1.0 (1.0 being the
reference). An underperformance of up to 40% was observed with certain
shipments to the reference PAC performance. Also, a price-adjusted performance factor (based on Newcombe and Cook4) was used as an indicator
of how efficient the PAC is at reducing target compounds.
Finally, kinetic and equilibrium constants were also estimated for each
PAC. Kinetic and preliminary equilibrium experiments were conducted in batches
and the kinetic coefficients were fit to the Freundlich equation. A unique 1/n for
each PAC type were fitted as well as K values for each PAC shipment in each
sample date. PAC samples were tested in different contact times of 10, 30, 60,
120 and 240 minutes, and their corresponding kinetic constants were
determined. It is understood that the maximum amount of adsorbate being
adsorbed on PAC surface is almost the same for both selected PAC types.
Jacobi tends to have higher adsorption capacity for C2 and UV254. While
Ingevity tends to have higher adsorption capacity values for C1. Ingevity tends to
have higher kinetic adsorption rate constant, for C2 and UV254. This value tends
to be higher for C1 Jacobi carbon. Based on fitted/estimated adsorption value for
NOM compounds using both carbons, Jacobi Carbon showed 16-30% more
adsorption capacity.
Moreover, regarding competitive adsorption, results of an experiment
performed by OEPA of kinetic data (0-120 min) for adsorption of microcystin as
well as DOC using two Carbons of Jacobi (CB1-MW) and Standard Purification
(WC800) have been studied. Jacobi showed an overperformance of 58% and
67% for both DOC and microcystin removals compared to WC800. While for
spectroscopic NOM measurements the overperformance is 38%, 26% and 50%
in removing NOM compounds of C1, C2 and UV254.In order to investigate the effect of competitive adsorption, removal of MIB
was modeled based on the IAST equation considering the presence of equivalent
background compound (EBC). The results show that the bigger the EBC value,
due to the background competition, the lower the MIB removal which explains the
competitive adsorption effect. For future work, it is suggested that kinetic
experiments be designed to test both carbons for NOM and target compound
(MIB, Geosmin, and Microcystins) adsorption.