New production pathway of musty 2,4,6-tribromoanisole during raw water disinfection processes at a surface water treatment plant

Taste and odor (T&O) issues are common in water utilities that rely on source water vulnerable to algal, cyanobacterial, and diatom blooms. Research has indicated that T&O compounds can also be formed during treatment processes and within the distribution system (Peter & Von Gunten 2007). These compounds impact the aesthetic quality of water by imparting taste and smell, thus causing issues with the public perception of water quality. The most commonly reported odors are earthy/musty followed by others such as grassy, swampy, and chlorinous odors (Khiari et al. 1999; Suffet et al. 2004; Peter & Von Gunten 2007). When T&O events occur at drinking water utilities, the major challenges are the unpredictability of such an event, including access to analytical methods and/or instruments that can identify the compounds causing the event and knowledge gap with the mitigation of T&O compounds that have not been studied enough. Recent advancements in analytical methods and treatability studies have explored compounds beyond 2-methylisoborneol (MIB) and geosmin and demonstrated the occurrence of several other compounds such as 2,4,6-tribromoanisole (2,4,6-TBA), dimethyl sulfides, β-cyclocitral and β-ionone in source and tap water samples at drinking water treatment plants (Pochiraju et al. 2021, 2022; Adams et al. 2022).

The City of Wichita Falls (CWF), Texas, has a surface water system with adjudicated rights to five water sources, four of which are used primarily for treatment: Lake Arrowhead, Lake Kickapoo, and the Lake Kemp/Lake Diversion system. The system has two water treatment facilities (WTFs) – Cypress WTF, which has one advanced and three conventional treatment plants, and Jasper WTF, which has two conventional treatment plants – and includes a water distribution system. After the highly successful production of odor-free drinking water through a direct potable reuse (DPR) project in 2014–2015 (Nix et al. 2021), where a blend of 50% conventionally treated raw water was mixed with 50% wastewater effluent treated by microfiltration/reverse osmosis, zero customer complaints were received during this time. It was determined that an odor-free drinking water was capable of being produced and that the Water Purification Division needed to make significant changes to its T&O monitoring and response programs.

In January 2023, a new odor event was detected by the CEL staff using sensory analysis that was believed to be the same compound that caused an event in winter 2022 that staff were unable to identify. In both cases, the odor was described as ‘musty/moldy’ and ‘fishy’ in a few cases. The production of 2,4,6-TCA was suspected but subsequently ruled out as the cause of this event after analyzing the water samples using the expanded T&O method from 2021. Instead, the production of another halogenated anisole, 2,4,6-TBA (C₇H₅Br₃O, molecular weight: 344.8 g/mol, Figure 2, right) was found to be responsible. 2,4,6-TBA has an OTC of 0.08 ng/L (Malleret & Bruchet 2002), which is below the detection limit (5 ng/L) by GC–MS.

This paper investigates the production of 2,4,6-TBA within the treatment process, the T&O event timeline, and the steps taken by CEL staff to determine the unknown compound to be 2,4,6-TBA, including treatment changes and laboratory analyses, as well as mitigation to lessen the impact of the T&O event. These efforts resulted in a reduction of 2,4,6-TBA to the below analytical detection limits by GC–MS and demonstrated the possibility of this compound being produced in treatment plants as a byproduct of the disinfection process, which has not been reported previously.

Standard Methods: 2150 ‘Threshold Odor Test’, Standard Methods: 2170 ‘Flavor Profile Analysis (FPA)’ (APHA et al. 2022), and the T&O wheel described therein were utilized to investigate the unknown odorant. These sensory analyses allowed analysts to describe T&O compounds in a systematic way, whereby samples are dechlorinated with a 5.0 g/L ascorbic acid stock solution, stirred and heated to 60 °C, and analyzed by olfaction to assign an odor number and categorize the type of odor (muddy, musty, grassy, etc.). The strength of the odor was expressed by the threshold odor number (TON). Many compounds have lower OTCs than what can be detected by analytical means, such as GC–MS.

An expanded analytical method employing solid-phase microextraction (SPME, Standard Methods: 6040D) followed by GC–MS was developed for 2,4,6-TBA based on the previously described method for 18 T&O compounds by GC–tandem MS (GC–MS/MS) (Pochiraju et al. 2021). Water samples were analyzed by selected ion monitoring (SIM) using a Thermo Scientific (Waltham, MA) Trace 1310 ISQ LT GC–MS with a TraceGOLD-5MS column, with sample injection by a Supelco SPME fiber [divinylbenzene/carboxen/polydimethylsiloxane (DVB/CAR/PDMS); Sigma #57329-U]. A Thermo Scientific TriPlus RSH autosampler was used with a three-position sample tool changer. High purity standards and SPME fibers are sourced from Sigma Aldrich (St. Louis, MO), and a calibration range of 5–200 ng/L was used and a method detection limit (MDL) of 2.2 ng/L was achieved for 2,4,6-TBA. Headspace vials (20 mL) with magnetic screw caps were used, with 3 g of sodium chloride, a 10 mL sample, and a 300 μL internal standard (2-isobutyl-3-methoxypyrazine [IBMP]). Instrument conditions include a 5 min incubation time and a 30 min SPME extraction time at 65 °C, an 0.8 min SPME desorption time at 270 °C in a 50 mL/min splitless injection module using helium, a 1.5 mL/min column flow rate at 50 °C/1 min, 200 °C at 10 °C/min, 250 °C at 20 °C/min using helium as a carrier gas, 250 °C transfer line, 300 °C ion source, and 0.5 min SIM retention time window at 0.30 s SIM time.

On January 9th, 2023, CEL staff noticed a slight musty odor in the tap water by sensory analysis, while all the routinely monitored compounds namely geosmin, MIB, 2-isopropyl-3-methoxypyrazine, and 2,4,6-TCA by the SPME GC–MS analysis were non-detectable. Following sensory analyses during the January 2023 T&O event, samples were analyzed for the following week. Standards were chosen for the investigation of the event including 2,3,6-trichloroanisole and 2,4,6-TBA, which has been a known T&O compound in tap water since first detected in France (Malleret & Bruchet 2002), and has caused T&O issues in water systems previously (Bruchet et al. 2007). After confirming the odor of 2,4,6-TBA to be a positive match with the sensory analyses, general acquisition scans were run by GC–MS using a mass range of 50–400 amu with 0.2 s dwell time. Upon confirmation, a 2,4,6-TBA calibration curve was prepared (5–200 ng/L) and analyzed by SIM. Tap samples from the previous 2 weeks were then reanalyzed.

A temperature-controlled jar test was conducted to verify the 2,4,6-TBA formation where a recirculating water bath was used to simulate raw water temperature at 11.5 °C. Type 1 ultrapure water was used and added to three 2-L test jars. Each jar was spiked with 0.30 mg/L bromide (Br⁻, Ricca #R1180000-120A) to simulate the raw water concentration, two jars were spiked with 0.10 and 0.20 mg/L Cl₂ to simulate excess dosages in the D1C, and one jar was not spiked with Cl₂ to simulate no Cl₂ addition. Jars were spiked with anisole at a concentration of 1.0 μg/L (Sigma Aldrich #8014520500) to simulate humic anisoles in the raw water. Samples were rapidly mixed for 30 s and then slowly mixed for 1 h to simulate the detention time in the raw water pipe up to the D1C. Each jar was subsampled and analyzed by TON/FPA and SPME GC–MS.

To confirm that 2,4,6-TBA could be produced in the D1C raw water with naturally occurring concentrations of bromide and anisole and with the addition of Cl₂, a second temperature-controlled jar test was conducted to verify compound formation. Raw water from the D1C sample point was used and added to three 2-L test jars. Two jars were spiked with 0.10 and 0.20 mg/L Cl₂ to simulate excess dosages in the D1C, and one jar was not spiked to simulate no Cl₂ addition. Samples were rapidly mixed for 30 s and then slowly mixed for 1 h to simulate the D1C detention time. Each jar was subsampled and analyzed by TON/FPA and SPME GC–MS.

2,4,6-TBA was the only T&O compound detected, with the first confirmation in tap water at 5.57 ng/L on January 25th, 2023. Samples collected between January 9th and 23rd, 2023 were tested but showed no detection for 2,4,6-TBA. Actinomycete culturing confirmed that no colonies were present in tap water samples, and raw water samples never exceeded 100 CFU/100 mL, which is normal for CWF's raw water. Sensory analysis confirmed that the compound was still present but not detected by GC–MS. To remove the odorous compound, PAC was dosed at 10 mg/L prior to the clarifiers (Figure 1) at Cypress and Jasper WTFs on January 25th, 2023 as recommended by Pochiraju et al. (2022). Within the next 2–3 days, CEL staff noticed a marked reduction in odor by sensory analysis, indicating that the PAC was removing the compound being introduced into the system pre-clarification.

To test the hypothesis that the ClO₂ disinfection process was causing the formation of 2,4,6-TBA in the raw water, ClO₂ was shut off to the Cypress WTF on February 7th, 2023. To allow for the complete renewal of the raw water line, samples were collected after 4 h had elapsed since the ClO₂ shutdown (the raw water line detention time: approximately 60 min). Sensory analysis yielded only earthy/muddy odors, with no musty odor. Results from SPME GC–MS were non-detectable for 2,4,6-TBA at levels <5.00 ng/L, which proved the hypothesis. Additional analyses conducted on February 8th, 2023 to confirm the results from previous sampling events showed non-detectable levels for sensory analyses as well as SPME GC–MS at <5.00 ng/L. When ClO₂ was turned on again the formation of 2,4,6-TBA returned, which strongly suggests that the musty/moldy 2,4,6-TBA was a product of the ClO₂ disinfection process.

It is well known that bromide is not oxidized by ClO₂ and will not form hypobromite, hypobromous acid (HOBr), bromoform, or bromate. However, it was surmised that 2,4,6-TBA is formed through the reaction of naturally occurring bromide and humic anisoles with excess free chlorine (Cl₂), which could be present in measurable concentrations from the generation of ClO₂. When ClO₂ is generated, Cl₂ is the target excess reactant, thereby limiting unreacted chlorite. When ClO₂ was turned off, 2,4,6-TBA remained non-detectable.

It was found that 2,4,6-TBA could be produced by (1) the biochemical methylation of a bromophenol precursor by actinomycete colonies or fungus Paecilomyces variotii (Whitfield et al. 1997) or (2) in the reaction of anisoles in humic substances and bromide through disinfection (Acero et al. 2005; Melbye et al. 2006). The biological conversion of fungicide 2,4,6-tribromophenol (2,4,6-TBP) to 2,4,6-TBA (Figure 4(1)) is commonly known in packaged food and beverages (Whitfield et al. 1997; Jönsson et al. 2006; Giannikopoulos & Whitfield 2009). Also, Malleret & Bruchet (2002) hypothesized that the leaching of 2,4,6-TBP from tank coating followed by the biological conversion of 2,4,6-TBP caused the 2,4,6-TBA formation in their drinking water. However, in the current study, such biological activity was absent around D1C (Figure 3) where the 2,4,6-TBA formation was noticed.

Results of the jar test with Type 1 ultrapure water show that in Jar 1, no odor was detected and that no 2,4,6-TBA was formed, which was spiked only with 0.30 mg/L Br⁻ and 1.0 μg/L anisole. Jars 2 and 3 were each spiked with increasing concentrations of free Cl₂, and each had a noticeable increasing musty odor of 2,4,6-TBA and the compound was confirmed in Jar 3 (Table 2). These data support the production method observed during the monitoring of the treatment plant at the D1C, and that excess free Cl₂ can cause the production of 2,4,6-TBA in the presence of Br⁻ and anisoles.

Results of the jar test with raw water show that in Jar 1, no musty odor was detected and that no 2,4,6-TBA was formed, which was not spiked with free Cl₂. Jars 2 and 3 were each spiked with increasing concentrations of free Cl₂, and each had a noticeable increasing musty odor of 2,4,6-TBA and the compound was confirmed in Jars 2 and 3 (Table 3). These data also support the production method observed during the monitoring of the treatment plant at the D1C, and that excess free Cl₂ can cause the production of 2,4,6-TBA in the presence of naturally occurring Br⁻ and anisoles. Higher production in Type 1 ultrapure water can be attributed to the lack of organic Cl₂ demand, as would be present in the raw water jar test.

The most common methods for monitoring ClO₂ are Standard Methods: 4500-C and -E, which are amperometric titration methods that determine the concentrations of chlorine species, e.g., ClO₂, free Cl₂, and chlorite (⁠⁠). ClO₂ generators must be routinely monitored to ensure the yield and purity being produced. A high yield (>95%) ensures that enough ClO₂ is being generated to meet disinfection requirements.

Some utilities do not strictly monitor the purity of ClO₂ or check the concentration of free Cl₂. However, it is important to understand because chemical feed dosages need to be adjusted to achieve a desired amount of ClO₂ while minimizing excess free Cl₂ and ⁠. In the United States, excess must be minimized because the tap water U.S. EPA Maximum Contaminant Level is 1.0 mg/L. Excess free Cl₂ must be minimized by careful operation of ClO₂ generators to avoid the formation of unwanted disinfection byproducts, such as 2,4,6-TBA, in the treated water as this study shows. Overfeeding of free Cl₂ to the generator can eliminate residual and can produce a high yield of ClO₂. However, the excess Cl₂ will result in a low purity, which is a signal for operators to increase the Cl₂ feed rates, which possibly exacerbate the disinfection byproduct formation.

To achieve a high yield (>95%), the dosage must be optimized to minimize ⁠. To achieve a high purity (>90%), dosage must be optimized to minimize and Cl₂. Too little Cl₂ would cause an increase in and a lower yield and purity, and too much Cl₂ could result in little to no excess ⁠, resulting in a high yield but potentially low purity of generated ClO₂ due to excess Cl₂.

By developing a multi-layered T&O monitoring program, CEL was able to detect a T&O event again by sensory methods prior to detection by SPME GC–MS. A target compound, 2,4,6-TBA, was chosen based on sensory results and the T&O wheel, which proved to be effective when samples were analyzed by SPME GC–MS. 2,4,6-TBA was confirmed to be present up to 9.86 ng/L and a TON of 3. To determine the likely mode of production, three possible sources of production of 2,4,6-TBA were tested. Actinomycetes and chloramination were ruled out, and it was found that 2,4,6-TBA was being produced in the raw water line post-ClO₂ addition, prior to treatment at the WTFs, through the reaction of excess free Cl₂ from the ClO₂ generation process with naturally occurring bromide and humic anisoles. PAC was effective in eliminating the T&O residual in the tap water, reduced 2,4,6-TBA to a non-detectable level of <5.00 ng/L, and achieved a TON of 1. The ClO₂ generation process was turned off and the production of 2,4,6-TBA was eliminated. To the best of our knowledge, this is the first report describing the possible formation of 2,4,6-TBA in the drinking water treatment process in this manner.

CEL's T&O monitoring program continues to prove that sensory analysis is an integral part of an effective plan to detect unknowns. Proactive utilities should include sensory analysis in routine monitoring, especially if they do not have the capability to analyze samples by SPME GC–MS in-house. Future research is needed to continue to characterize T&O compounds, their routes of production, mitigation strategies, and the most effective means to identify them early to maintain consumer confidence. This research group is currently working to meet this need and has formed a Joint Task Group to expand SM 6040D and evaluate a suite of known T&O compounds that can be problematic for water systems (Adams et al. in preparation).

H.A. and M.S. performed conceptualization, designed the methodology, administered the project, wrote and edited the manuscript. S.P. performed conceptualization, designed the methodology, validated the study, wrote and edited the manuscript. K.I. performed conceptualization, designed the methodology, wrote and edited the manuscript. S.R. and E.A. validated the study, wrote and edited the manuscript. D.N. performed conceptualization, designed methodology, administered the project, validated the study, wrote and edited the manuscript.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.