Environ. Sci. Technol. 2005, 39, 9343-9350
Modeling Cryptosporidium parvum Oocyst Inactivation and Bromate Formation in a Full-Scale Ozone Contactor GEORGE TANG,† KWABENA ADU-SARKODIE,‡ DOOIL KIM,§ JAE-HONG KIM,§ SUSAN TEEFY,⊥ HIBA M. SHUKAIRY,| AND BENITO J. MARIN ˜ A S * ,† Department of Civil and Environmental Engineering, University of Illinois at Urbana-Champaign, Urbana, Illinois 61801, Damon S Williams Associates, Phoenix, Arizona 85016, School of Civil and Environmental Engineering, Georgia Institute of Technology, Atlanta, Georgia 30332, Water Quality & Treatment Solutions, Inc., Castro Valley, California 94546, and U.S. Environmental Protection Agency, Technical Support Center, Cincinnati, Ohio 45268
The inactivation of Cryptosporidium parvum oocysts and the formation of bromate were assessed simultaneously by performing experiments with a full-scale ozone bubblediffuser contactor used for drinking water disinfection. Fluorescence-dyed polystyrene microspheres were used as surrogates for C. parvum oocysts. Semi-batch ozonation experiments were performed to determine the fluorescenceintensity decay of individual microspheres, which was measured by flow cytometry. The results obtained with the microspheres were correlated to the inactivation kinetics of C. parvum oocysts by choosing an appropriate threshold fluorescence intensity below which microspheres were considered to be equivalent to nonviable oocysts. A mathematical model was then used to predict the inactivation efficiency and bromate formation. The contactor hydrodynamics were characterized by running tracer tests, and the kinetic parameters for ozone decomposition and bromate formation were obtained by performing batch experiments. Model predictions were in good agreement with full-scale experimental results. Additional model simulations revealed that ozone contactors should be designed with the lowest possible backmixing so that the target inactivation efficiency can be achieved with the lowest possible formation of bromate.
Introduction Chlorination remains the most commonly used process for drinking water disinfection in the United States and in the world. However, many water utilities are looking for alternatives because free chlorine produces halogenated disinfection * Corresponding author phone: (217) 333-6961; fax: (217) 3336968; e-mail:
[email protected]. † University of Illinois at Urbana-Champaign. ‡ Damon S Williams Associates. § Georgia Institute of Technology. ⊥ Water Quality & Treatment Solutions, Inc. | US Environmental Protection Agency. 10.1021/es050345n CCC: $30.25 Published on Web 11/03/2005
2005 American Chemical Society
byproducts (DBPs) of public health concern and chlorination is less effective for controlling emerging pathogens such as Cryptosporidium parvum oocysts under typical drinking water disinfection conditions (1, 2). Ozone is being considered more frequently for use as a primary disinfectant because it provides effective control of C. parvum oocysts (3, 4). However, a broader implementation of the ozone disinfection process may currently be hindered by the formation of bromate, a DBP resulting from the reactions of bromide, a relatively common constituent in natural waters, with ozone and its decomposition intermediate hydroxyl radical. Median bromide concentrations reported in a survey of 500 drinking water treatment plants in the United States were 27 and 65 µg/L for surface water and groundwater sources, respectively (5). These values are sufficiently high to potentially form bromate at levels greater than its current maximum contaminant level of 10 µg/L (6). However, besides bromide concentration, various other water quality parameters, temperature, and the level of exposure to ozone also affect the level of bromate formed. In general, the formation of bromate increases with increasing levels of ozone exposure, temperature, and pH, and decreasing concentrations of ammonia, and it is also affected by hydroxyl radical promoters such as hydrogen peroxide and radical scavengers such as alkalinity (7-11). Natural organic matter (NOM) can also act as hydroxyl radical scavengers and promoters and thus affect the overall formation of bromate. In contrast, the inactivation efficiency of C. parvum oocysts with ozone is only affected by ozone exposure and temperature, with increasing values of both enhancing the inactivation rate (3). Consequently, these different effects on the kinetics of bromate formation and C. parvum oocyst inactivation have allowed the development of water quality pretreatment approaches, such as pH depression and ammonia addition, that could result in lower formation of bromate for a given target level of C. parvum oocysts inactivation efficiency (12). The hydrodynamics inside an ozone contactor also affect the overall conversion of all chemical and inactivation reactions (7, 15) with the effect being more pronounced for reactions with relatively higher conversion (e.g., inactivation of C. parvum oocysts in the range of 0-99.9%) compared to reactions with lower conversion (e.g., bromate formation from bromide in the range of 0-30%). Consequently, reactor design and operation could be optimized to further minimize bromate formation while achieving the target inactivation efficiency for C. parvum oocysts (12). The consideration of reactor hydrodynamics can be facilitated by using mathematical models. The axial dispersion reactor (ADR) (13) model has been applied successfully to represent the inactivation kinetics of Cryptosporidium oocysts in a pilotscale ozone bubble-diffuser contactor treating natural water (14, 15). However, the ability of the ADR model to accurately represent disinfection efficiency as well as bromate formation in full-scale ozone bubble-diffuser contactors remains to be shown. Demonstrating the applicability of the ADR model to represent the performance of full-scale ozone bubble-diffuser contactors would require performing validation tests. Unfortunately, although it is relatively easy to do for bromate formation, the inactivation efficiency of C. parvum oocysts could not be assessed directly in full-scale systems because analytical technologies do not exist at this time for assessing oocyst viability at the relatively low concentrations at which these parasites occur in natural waters. However, this impasse could be overcome with indirect methods based on the use of surrogates. Fluorescent-dyed microspheres, originally VOL. 39, NO. 23, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
9343
developed as nonbiological surrogates to assess the inactivation of Giardia cysts with ozone (16), have been demonstrated to serve as good surrogates for C. parvum oocysts in benchscale studies with semi-batch ozone disinfection reactors (17). When oocysts and microspheres are exposed to dissolved ozone, the disinfectant diffuses into the particles, reacts with internal particle components, and eventually “inactivates” the particle. Inactivation is measured by the loss of viability in the case of the oocysts, and by the internal fluorescence intensity decreasing below a certain threshold value in the case of the microspheres (16). The fluorescence-intensity threshold could be adjusted so that the inactivation kinetics of both particles would match. Furthermore, microspheres have a density similar to that of oocysts (17) and thus they are exposed to the same level of backmixing experienced by pathogens in a flow-through contactor. The objective of this study was to simultaneously assess the disinfection efficiency and bromate formation in a fullscale ozone contactor system, using microspheres as nonbiological surrogates for C. parvum oocysts. A mathematical model, previously developed and validated to represent the disinfection efficiency achieved in pilot-scale ozone bubblediffuser contactors (14, 15), was modified to predict both inactivation efficiency and bromate formation in full-scale ozone bubble-diffuser contactors with multiple chambers. The full-scale contactor model was validated with experimental data and applied to predict process performance. Validation experiments were performed with a full-scale ozone contactor at the Alameda County Water District treatment plant, located in Fremont, California.
Materials and Methods Microspheres. Fluoresbrite Plain YG microspheres (Polysciences, Warrington, PA) were used as surrogates for C. parvum oocysts. They were polystyrene-based with an average diameter of 0.94 µm and a density of 1.045 g/mL. The dye, a proprietary chemical referred to as YG by the manufacturer, was hydrophobic and matched the fluorescence filter settings for fluorescein isothiocyanate, i.e., maximum excitation and emission wavelengths of 458 and 540 nm, respectively. Pre-ozonation of Microspheres. Similar to the observation in previous studies (17), preliminary tests revealed that the fluorescence intensity of the microspheres supplied by the manufacturer was too high for optimum simulation of the kinetics of C. parvum oocysts inactivation. A trial-anderror approach revealed that an optimum kinetic match could be achieved by reducing the average microsphere fluorescence intensity by approximately 95%. This was achieved by exposure to ozone in a semi-batch reactor following the procedures described by Marin ˜ as et al. (17) with the following modifications. A total mass of microspheres of 6 g was suspended in 400 mL of 0.01 M phosphate buffer solution (PBS) at pH 7 and pretreated with ozone at 20 °C for 150 min. The ozone gas flow rate was 5.9 standard cubic meter per hour (SCMH), and the ozone gas concentration, although not directly measured, resulted in a steady-state dissolved concentration of 1.75 mg/L when applied to 0.01 M pH 7 PBS prior to adding the microspheres. Semi-batch Ozonation Experiments. The reaction between ozone and pre-ozonated microspheres was studied using the apparatus and following the procedures described by Marin ˜ as et al. (17), except for the following modifications. The reactor consisted of a 250 mL glass cylinder immersed in a water bath with the temperature set to 20 °C, approximately the average temperature of Alameda County Water District (ACWD) water during full-scale testing. The ozone gas flow rate was kept constant at 5.9 SCMH, and, once the reactor reached steady state, pre-ozonated microspheres were dosed at 0.2 mg/L. Samples with volumes of 9344
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 23, 2005
FIGURE 1. Schematic of the full-scale ozone contactor at the ACWD Treatment Plant, Fremont, California. Note: 1′ ) 1 foot ) 30.5 cm. 1 mL were withdrawn from the reactor at ozone exposure intervals of 1 mg min/L, and quenched with 0.2 mL of a 1.67% sodium thiosulfate solution. The fluorescence-intensity distribution of microspheres was analyzed by flow cytometry as described subsequently. Batch Ozonation Experiments. The kinetics of ozone decomposition and bromate formation in ACWD water was determined with a batch reactor following the procedures developed by Kim et al. (15) with the following modifications. Experiments were performed with a contactor influent sample collected at the end of the first day of full-scale experimentation (May 28, 2002). We measured the following water quality parameters: pH 7.3; total organic carbon, 3.5 mg/L; total alkalinity, 82 mg/L as CaCO3; temperature, 19 °C; and bromide, 150 µg/L. An ozonation experiment was initiated by transferring 9.7 mL of an ozone stock solution into a 100 mL gastight syringe containing 90.3 mL of ACWD water. Two ozone decomposition experiments were performed at initial dissolved ozone concentrations, CL,0 ) 1.58 and 1.95 mg/L, both at 20 °C. The first sample for dissolved ozone concentration determination was taken after allowing the reaction to proceed for 1.33 min and several additional samples were taken at 1 min intervals for a total of 8.33 min using a 1 mL syringe. An additional sample was taken using a 5 mL syringe at approximately 10 min for bromate concentration determination. The final bromate formation was also measured in two additional batch experiments performed at CL,0 ) 2.12 and 2.36 mg/L following the same procedures. All ozone and bromate samples from batch experiments were quenched with 0.2 mL of a 9.0 g/L ethylene di-amine (EDA) solution. Full-Scale Ozone Contactor. Tracer and ozonation experiments were performed with one of two parallel ozone contactors at the ACWD treatment plant. A longitudinal crosssection of the unit is shown in Figure 1. The contactor consisted of five chambers with each further divided into down-flow and up-flow sections. In addition to the chamber dimensions shown in Figure 1, the contactor was 14.75 ft (4.50 m) wide, and the average water column height was 22.2 ft (6.77 m). Five sampling ports were distributed throughout the water column of each down-flow section with one additional port located near the outlet of each up-flow section. Ozone gas could be applied through porous bubble diffusers, located at the bottom of the first and second (1A and 2A) down-flow sections. Additional information about the ozone contactor and other treatment processes used at the ACWD treatment plant was given by Marin ˜ as et al. (17). Full-Scale Tracer Tests. Two-step input tracer tests were performed with gas being applied to the down-flow chamber section 1A at a target flow rate of 187-204 SCMH, and using fluoride as the tracer chemical. Two tracer tests were performed at a target water flow rates of 7 and 14 million gallons per day (MGD). Fluoride samples were taken at the effluent sampling port from chamber sections 1A, 2A, 3A (sampling port 5 of each section), and 5B for the test at 14 MGD, and at the effluent of chamber sections 1A (sampling port 5) and 5B for the test at 7 MGD.
TABLE 1. Experimental Matrix for Full-Scale Testing test
date
1 2 3 4 5 6
5/28/02 5/29/02 5/29/02 5/30/02 5/30/02 5/30/02
water gas feed gas ozone flow rate flow rate conc dose (MGD) (SCMH)a temp (°C) (wt %) (mg/L) factorb 9.5 9.7 9.5 15.5 15.5 15.5
180/0 214/0 211/0 204/0 204/0 204/202
18.9 18.9 19.1 19.7 19.7 20.3
1.87 2.58 2.02 3.19 2.87 2.03
2.65 4.20 3.29 3.18 2.83 4.03
1.7 1.0 1.3 1.4 1.4 1.1
a First/second chambers. b Factor by which the values D ) 0.73 mg/ 0 L, and kD ) 0.011 s-1, determined in batch ozonation experiments, were multiplied in the prediction of each full-scale test. For example, the values used for test 1 predictions were D0 ) 0.73 × 1.7 ) 1.24 mg/L, and kD ) 0.011 × 1.7 ) 0.019 s-1.
Full-Scale Ozonation Experiments. Six full-scale ozonation experiments were performed during three consecutive days in May, 2002 (Table 1). Experimental variables were ozone dose (2.65-4.20 mg/L) and water flow rate (9.5-15.5 MGD). Ozone gas was applied in the first down-flow chamber section (1A) for tests 1-5, and split evenly between the first two down-flow chamber sections (1A and 2A) for test 6. Experiments were performed at a constant gas flow of 180214 SCMH per contactor section. The applied ozone dose was varied by adjusting the gaseous ozone concentration within the range of 1.87-3.19% by weight (wt %). Dissolved ozone and bromate concentration profiles were determined by taking samples from all ports depicted in Figure 1. Water quality parameters, measured daily, were as follows: turbidity, 8.63 ( 0.43 NTU; alkalinity, 82 ( 4 mg/L as CaCO3; conductivity, 382 ( 4 µmho/cm; pH, 7.29 ( 0.02. The following protocol was used for full-scale demonstration experiments. A suspension of microspheres at a concentration of 1.7 g/L was pumped at a flow rate of 17 mL/min into an injection port located at the contactor influent water pipe. This resulted in a microsphere dose of 0.7-1.1 µg/L or (1.3-2.0) × 106 microspheres/L. The contactor was then operated for a period of approximately twice the hydraulic retention time (HRT) of the overall contactor (HRT was estimated at 12 or 19 min for the experimental water flow rate of 15.5 or 9.5-9.7 MGD), which, based on tracer data, allowed the microsphere concentration throughout the contactor to be within ∼90% of its steady-state value. Samples were then taken from all sampling ports prior to stopping the addition of microspheres. The samples were collected into two 1 L containers, each dosed with 1 mL of undiluted EDA reagent (density of 889 g/L) to quench any residual ozone, and shipped to the University of Illinois for analyses as described subsequently. Flow Cytometry Analyses. Microsphere samples were concentrated prior to analysis following the procedures outlined by Marin ˜ as et al. (17) with the only modification that the centrifugation step lasted 30 min. The distribution of microsphere fluorescence intensity in both bench-scale and full-scale experimental samples was measured with a Coulter Epics XL-MCL flow cytometer (Beckman-Coulter, Inc., Fullerton, CA), following the procedures described by Marin ˜ as et al. (17). The relative fluorescence scale, ranging from 1 to 1024, was adjusted so that the average fluorescence emitted by microspheres in the pretreated stock was 70. Analytical Methods. Ozone concentrations were determined by the indigo colorimetric method (18) using a molar absorptivity of 23 150 M-1 cm-1 at the wavelength of 600 nm (19). Bromide/bromate samples were passed through 0.45 µm syringe filters (Osmonics, Minnetonka, MN) and analyzed with an ion chromatograph Model DX-300 (Dionex Corp., Sunnyvale, CA) using conductivity detection for bromide and spectrophotometric detection based on the modified version
(20) of a postcolumn derivatization method (21) for bromate. The tracer fluoride was analyzed with an ion-selective electrode according to Standard Method 4500-F (22).
Modeling Approach Ozone Decomposition Kinetics. The kinetics of ozone decomposition in ACWD water was represented using the empirical approach developed by Kim et al. (14). Ozone was assumed to decompose through two simultaneous reactions: a second-order reaction with NOM chemical groups, such as R-CdC-R′, exerting fast ozone demand (referred to as “fast demand” in this manuscript), and a pseudo-firstorder reaction. Kim et al. (15) observed that the rate of the second-order reaction between dissolved ozone and the fast demand was too fast to be measurable with their batch reactor. Because not much C. parvum oocyst inactivation or bromate formation occurred by the time this relatively fast second-order reaction was completed, the ozone decay in the batch reactor could be represented by subtracting the ozone consumed by the fast demand from the initial ozone concentration, and using a simplified rate expression including only the pseudo-first-order term. However, this approach could not be used in a flow-through bubble column reactor because ozone is transferred gradually throughout the water column. In such case, the more general rate expression including both second-order and pseudo-firstorder terms needs to be used. Although the second-order rate constant could not be measured, an adequate approach would be to use a sufficiently high kR value so that no significant C. parvum oocyst inactivation or bromate formation would take place. The kR value used by Kim et al. (15) for the second-order rate constant for the reaction between ozone and the fast demand, kR ) 3.20 ( 0.16 (L/mg s), was also used in this study. This rate constant was found to be adequate because it resulted in C. parvum oocyst inactivation of