Geosmin and 2-methylisoborneol monitoring versus a combination of sensory and chemical analyses in drinking water odor evaluation Open Access

Odor occurrences in Summit Water Treatment Plant (SWTP) drinking water were evaluated by two methods. One method was the common practice of monitoring geosmin (GEO) and 2-methylisborneol (2-MIB) only. It led to the conclusion that the odor events primarily originated from Silverwood Lake (a water source for SWTP) and took place during late summer and mid-fall to early winter. The other method combined flavor profile analysis (FPA), gas chromatography/mass spectrometry (GC/MS) and Sensory GC. FPA captured a recurring moldy odor, unlike the earthy/musty odor expected from GEO and 2-MIB. 2-isopropyl-3-methoxypyrazine (IPMP) was identified as its cause by GC/MS and Sensory GC. More importantly, IPMP and its moldy odor were mainly produced in Lytle Creek (the second water source for SWTP) from late fall to mid-winter. Thus, monitoring only GEO and 2-MIB led to incomplete understanding of the chemical causes as well as the spatial and temporal patterns of SWTP's odor events. The case study serves as a precaution against equating the presence of GEO and 2-MIB with the overall drinking water odor occurrences despite the popularity of the approach. Instead, a combination of FPA and, as needed, subsequent GC/MS and Sensory GC is necessary for complete drinking water odor evaluations.

Geosmin (GEO) and 2-methylisoborneol (2-MIB) are earthy/musty compounds often caused by algal or cyanobacterial blooms and are the most commonly targeted biogenic drinking water odorants by regulations, monitoring, and studies due to their omnipresence and low odor threshold concentrations (OTCs) (Paulino et al. 2023). By comparison, other biogenic odorous compounds rarely receive the same attention from water utilities, and the sole attention to GEO and 2-MIB could lead to unexpected and unexplained odor events resulting from other odor compounds (Zhu et al. 2022). The incomplete knowledge of the chemical causes of odor events could be particularly troublesome if the overlooked odorants cannot be removed by treatments effective for GEO and 2-MIB. For example, activated carbon adsorption technology can remove GEO and 2-MIB, but it cannot remove odorous compounds of thiols and thioethers effectively (Du et al. 2024). Thus, if the thiols and thioethers were responsible for odor events in drinking water together with GEO and 2-MIB, monitoring and thus finding only GEO and 2-MIB may lead to the decision to use activated carbon for odor mitigation. As a result, unexpected odor events may take place due to unremoved thiols and thioethers. Additionally, monitoring GEO and 2-MIB alone may result in a misunderstanding of the spatial and temporal patterns of odor events if the omitted odorous compounds occur in a different pattern from that of GEO and 2-MIB. This could lead to problematic decisions regarding the selection of water sources. For instance, in a study of taste and odor (T&O) occurrences involving multiple water sources, the supplementary reservoir, as one of the sources, was the primary source of 2-MIB, while GEO concentrations rarely exceeded its OTC in any source (Zhu et al. 2023). Yet, it was found in the study that an earthy/musty odor was present in the mixed raw water regardless of supply levels from the supplementary reservoir due to earthy/musty compounds other than 2-MIB and GEO. In this case, concentrations of GEO and 2-MIB should not be the only basis for water source selection.

In order to attribute odor characteristics identified by FPA to their chemical causes, gas chromatography/mass spectrometry (GC/MS) and Sensory GC analysis can be applied. Sensory GC separates odorous compounds within a mixture with the GC column and then presents the odorants to a human panelist so that the odorants can be sensorially characterized one by one, while parallel GC/MS analysis reveals the identities of perceived compounds (Hayes et al. 2023). In this way, odor characters determined by FPA can be linked with specific compounds identified by GC/MS analysis with matching odor characters perceived by Sensory GC. Moreover, an odorant within a mixture can be discovered by Sensory GC even if its odor was masked by other odors in the sample during the FPA. The combination of Sensory GC with GC/MS and/or its derivatives, such as two-dimensional GC/MS, has been used to identify odorants responsible for specific odor characters in drinking water (Khiari et al. 1992, 1995; Young et al. 1999; Hochereau & Bruchet 2004; Yu et al. 2009; Guo et al. 2016, 2020, 2021a, 2021b, 2021c, 2023, 2024; Quintana et al. 2016; Kalweit et al. 2019; Wang et al. 2023). Finally, analytical methods, such as GC/MS selected ion monitoring (SIM), can be applied to quantitate all odorants perceived by FPA and identified by Sensory GC with parallel GC/MS in drinking water instead of GEO and 2-MIB only. The occurrences of all odorants in drinking water determined by their routine analysis should, in theory, better reflect overall odor occurrences than those of GEO and 2-MIB alone.

One hundred μg/mL (±)-GEO and 2-MIB methanol (MeOH) solutions and water (suitable for high-performance liquid chromatography (HPLC)) were purchased from Sigma-Aldrich Inc. (St. Louis, MO, USA). p-Bromofluorobenzene (p-BFB) (>98.0%), sodium omadine (>97.0%), and 2-isopropyl-3-methoxypyrazine (IPMP) (>98.0%) were obtained from TCI America (Portland, OR, USA). HPLC-grade MeOH (≥99.9%), sodium bisulfite (NaHSO₃) (≥58.5%), and sodium chloride (NaCl) (≥99.0%) were purchased from Fisher Scientific (Hanover Park, IL, USA). Seventy per cent isopropyl alcohol (iPOH) solution was purchased from Kroger Co. (Cincinnati, OH, USA). The water suitable for HPLC was only intended for Sensory GC humidifier refills (Section 2.5.2). In other cases, water produced by a Milli-Q water purification system (Millipore-Sigma, Burlington, MA, USA) was used and was referred to as Milli-Q water.

Mixed cellulose esters (MCE) membrane filters (pore size – 0.22 μm) were obtained from Millipore-Sigma (Burlington, MA, USA). Teflon-faced silicone septa of 22 mm, 1 cm Stableflex™ 50/30 μm DVB/CAR/PDMS solid-phase microextraction (SPME) fiber assemblies, and their manual holders were purchased from Supelco (Bellefonte, PA, USA). Teflon-coated octagon magnetic stirrer bars of 5/8 × 5/16″ were bought from VWR International, LLC (Radnor, PA, USA).

Prior to use, all 1 L amber glass bottles with Teflon-lined caps were washed with a Liqui-Nox^® critical-cleaning liquid detergent (Jersey City, NJ, USA), tap water, and Milli-Q water. All magnetic stirrer bars and 65 mL amber glass sample vials together with their caps were rinsed with HPLC-grade MeOH, tap water, and Milli-Q water, and all 250 mL Erlenmeyer flasks and watch glasses were rinsed with a 70% iPOH solution, tap water, and Milli-Q water.

Clarified water before rapid sand filtration was sampled from 25/07/2022 to 09/01/2023. The sampling frequency was twice a week from 25/07/2022 to 14/08/2022. For the rest of the study, sampling was conducted weekly. SWTP filtration effluent was also sampled weekly during 17/10/2022–09/01/2023 in the hope that it better represented the product water distributed to the consumers. However, the evaluation of SWTP filtration effluent by FPA was hindered by high residual chlorine concentration (see Section 2.4.2). During the week of 22/08/2022 and that of 28/11/2022, no sample was collected due to the availability of limited staff. During sample collection, 1 L and 65 mL amber glass bottles with Teflon-faced cap liners were used to hold the original sample and preserved sample using 90 μL of 3.2% sodium omadine solution, respectively, with minimal headspace. All samples were sent to UCLA overnight in a cooler box with ice bags. Upon receipt, the samples were filtered through 0.22 μm MCE membrane filters and then stored in the dark at ∼4 °C until analysis.

Before training, all candidates for the panel were screened by the University of Pennsylvania Smell Identification Test^® (UPSIT^®) (Sensonics International, Haddon Heights, NJ, USA). Potential panelists with severe microsmia or total anosmia were excluded from the sensory panel.

All panelists were trained according to the Standard Method 2170 B (APHA et al. 2017). Odor reference standards and an intensity scale defined by sugar solutions in the method were applied to standardize panelists' descriptions of an odor's character and its intensity value, respectively. Vocabulary modifications were made to the original method based on panel consensus during training. The odor characters for GEO, 2-MIB, and IPMP were set as earthy, musty, and moldy, respectively. Trainees familiarized themselves with the standards until they were able to give a correct and reproducible characterization of random odor references. All references were available upon request during the FPA of a sample.

FPA was applied to every delivered sample without preservative within 48 hours after its filtration by MCE filters according to the Standard Method 2170 B (APHA et al. 2017) with slight modifications. Specifically, a 50 mL sample was added to a 250 mL Erlenmeyer flask. Then the flask was covered by a watch glass and heated to 45 °C before panel analysis. Though a panel was supposed to consist of four or five panelists, a few samples were processed by a smaller panel due to the availability of limited personnel. Each analyst was reminded not to apply any perfume on the day of analysis and not to eat or drink at least 30 min before sample evaluation. If an odor character was reported by half or more of the panel from a sample, the odor was considered confirmed. Its character and average intensity value were recorded. When an odor was perceived by fewer than half of all examiners, the odor was only regarded as an ‘odor note’ and recorded as such for reference without an intensity value. It should be noted that analysis at a higher temperature of 45 °C could yield higher odor intensities and thus better odor sensitivities than the room temperature of 25 °C (APHA et al. 2023). Taste analysis was not conducted because the samples were deemed still undrinkable without adequate disinfection at the points of sampling. Final disinfection of the drinking water before its distribution to consumers was conducted outside SWTP.

A strong chlorine odor was noticed in SWTP filtration effluent samples due to extensive chlorination applied right after filters. As a result, other odors were mostly masked, and FPA could not be effectively executed. Therefore, clarified water before filtration was chosen for odor evaluation and direct comparison of the combination of FPA, Sensory GC, and GC/MS with the common practice of monitoring GEO and 2-MIB only. Nevertheless, in cases where odor characters other than chlorine were confirmed (Section 3.2.2), the results were recorded for reference.

Headspace SPME specified by the Standard Method 6040D (APHA et al. 2017) was modified and applied to all clarified water samples in the study. For GC/MS analysis, 5.0 μL of the 500 μg/L p-BFB MeOH solution was added to each sample before extraction as the internal standard. Meanwhile, no internal standard was applied for Sensory GC analysis. 2-isobutyl-3-methoxypyrazine and IPMP were detected in drinking water samples (Zahraei et al. 2021). Therefore, they were not adopted as internal standards and surrogates as per the Standard Method 6040D (APHA et al. 2017). The StableFlex™ 1 cm 50/30 μm DVB/CAR/PDMS SPME fiber used in the study was conditioned at 270 °C for 30 min. Sample volume, headspace volume, temperature, extraction time, stirring speed, and NaCl addition were 45 mL, 20 mL, 65.0 ± 1.0 °C, 35 min, 180 ± 10 rpm, and 15.00 g, respectively. Unpreserved samples were extracted and analyzed by GC methods within a week after filtration. However, due to the time-consuming nature of headspace SPME and GC processes, a few samples preserved by sodium omadine were extracted and analyzed by GC methods within a month after their filtration in case of a backlog. The SWTP filtration effluent collected on 25/10/2022 was extracted and analyzed by GC methods in the same manner (Section 3.2.2) except that 2.0 μL of the 145 g/L NaHSO₃ solution was injected into the sample before extraction as a dechlorination agent.

After headspace SPME, compounds sorbed to the SPME fiber were introduced into one of the two parallel 60 m DB-5MS GC columns (diameter – 0.25 mm, film thickness – 0.25 μm) (Agilent Technologies, Santa Clara, CA, USA) within a Varian 450-GC system (Varian, Inc., Walnut Creek, CA, USA) through an 1177 split/splitless capillary injector; 99.9999% helium (Airgas, Inc., Radnor, PA, USA) was used as the carrier gas. The injector temperature was 260 °C, the extracted compound's desorption time was 15 min, and the split vent was off for the first minute after injection before being turned on with a split ratio of 100 to minimize peak tailing. The column flow rate was 1 mL/min while the column oven temperature was held at 50.0 °C for 2 min after injection before being raised at a rate of 8.0 °C/min to 260.0 °C, where it was held for 1.75 min.

For GC/MS full scan analysis, eluted compounds from the GC column were introduced into a Varian 220-MS system (Varian, Inc., Walnut Creek, CA, USA), where electron ionization (EI) at 70 eV was applied. Then, the produced ions with a mass-to-charge ratio (m/z) within 41–260 were scanned at a frequency of 1 Hz. Temperature setpoints of the MS system were trap – 150 °C, manifold – 40 °C, and transfer line – 270 °C. In Sensory GC analysis, odorous compounds separated by the other GC column were combined with humidified air for odor fatigue prevention and then presented to a human analyst at a sniffing port. The elution time of each odorant, its odor character, and its peak intensity value were recorded. Odorant identification was achieved by matching its retention times in both GC/MS and Sensory GC systems, its odor character at the sniffing port and its mass spectrum obtained by the MS system with those of purchased standard chemicals.

GC/MS SIM was used to quantitate selected odorants in every clarified water sample where an odor was confirmed by FPA. The procedure was the same with GC/MS full scan analysis except that the ions produced by EI were analyzed under a selected ion scan (SIS) mode during the elution windows of compounds for quantitation (Table 2). Among the target compounds, GEO and 2-MIB were quantitated to follow the common practice in the water industry to monitor the two compounds for odor evaluation. IPMP was quantitated after its identification by Sensory GC with parallel GC/MS full scan analysis as a moldy odorant in drinking water (Section 3.2.2) to evaluate its occurrence patterns. p-BFB served as the internal standard throughout the study.

External calibration was conducted for compound quantitation using standard samples containing GEO, 2-MIB, and IPMP at different levels. The linear range, limit of detection (LOD), and limit of quantitation (LOQ) for each analyte are shown in Table 3.

High earthy/musty odor intensities in late summer and mid-fall (Figure 6) were consistent with elevated GEO concentrations during the same periods, as shown in Figure 5. However, a strong earthy/musty odor was perceived by FPA in late August and early September while GEO concentration was low at the same time. The discrepancy might have been caused by the presence of 2-MIB in clarified water. The OTC value of 2-MIB has been reported to be 1.2–1.3 ng/L at 45 °C by Piriou et al. (2009). This means that the probability of perceiving the odor of 2-MIB in water at 45 °C is 50% when the 2-MIB concentration reaches 1.2–1.3 ng/L. For the same reason, 2-MIB's earthy/musty odor could be confirmed by FPA in this study when 2-MIB concentration was higher than 1.2–1.3 ng/L because FPA was conducted at 45 °C and an odor was confirmed when half or more (50% or more) of the panel perceived the same odor (Section 2.4.2). By comparison, the LOD of 2-MIB was 3.3 ng/L (Section 2.5.3), which was higher than its OTC. As a result, it was possible for 2-MIB to be confirmed sensorially by FPA yet undetected by GC/MS SIM. If this was the case, FPA demonstrated superior sensitivity than GC/MS SIM for earthy/musty events monitoring.

Another discrepancy took place in late fall (November) when GEO concentration was high but earthy/musty odor intensities were low (Figure 6). The period coincided with the presence of chlorine odor. It was recorded as an odor note (Section 2.4.2) on 01/11/2022, and then it was confirmed with average intensity values of 2.5 and 3.3 on 14/11/2022 and 21/11/2022, respectively. Therefore, a higher-than-usual chlorination dosage was likely applied before rapid sand filtration (Figure 4) in late fall, and it likely led to a higher chlorine concentration in clarified water. Piriou et al. (2009) found that chlorine can mask the earthy/musty odor caused by GEO and 2-MIB. Thus, the discrepancy was probably caused by the masking effects of a higher level of chlorine.

A strong moldy odor showed up in mid-winter (January) when the earthy/musty odor disappeared (Figure 6) due to very low GEO concentration (Figure 5). Its appearance took place at the same time as the Lytle Creek supply became 100% of SWTP's raw water. Therefore, the patterns of odor occurrences determined by the common practice of monitoring GEO and 2-MIB only were questionable. A strong drinking water odor occurred in mid-winter, not just in late summer and mid-fall to early winter. Besides, Lytle Creek was likely a major odor source, just like Silverwood Lake.

Future studies should further examine the hypothesis by long-term monitoring of Lytle Creek IPMP concentration, turbidity, flow rate, and precipitation in order to see if the sharp fluctuation of creek flow rate caused by precipitation always leads to an increase in both turbidity (sediment disturbance) and IPMP level. If so, Lytle Creek flow rate data may provide early warning to SWTP as the recipient to remove IPMP in its influent.

Another approach is to collect Lytle Creek sediment samples and stir them with filtered Lytle Creek water in a laboratory to observe if IPMP can be formed. A control group with a sterilized sediment sample can also be evaluated to understand whether IPMP production is a biotic or abiotic process. Determination of microbial constituents in Lytle Creek sediment is also warranted to identify potential IPMP producers.

The difference between the formation mechanism for IPMP (sediment disturbance by precipitation) and that for GEO and 2-MIB (phytoplankton and sediment organisms followed by fall turnover) was likely the cause of different spatial and temporal patterns between moldy odor from IPMP and earthy/musty odor from GEO and 2-MIB. In addition, the possible link between precipitation and IPMP formation has major implications for the effects of climate change on drinking water odor occurrences. Previous research has shown that heavy rainfall could decrease the concentrations of GEO and 2-MIB in reservoirs (Winston et al. 2014; Kim et al. 2021; Wu et al. 2022). Since it has been projected that California will encounter more extreme precipitation events in the future due to climate change (Swain et al. 2018; Huang et al. 2020; Feldman et al. 2021), Silverwood Lake as the origin of GEO and 2-MIB may play a lesser role in SWTP's drinking water odor occurrences during such events. On the other hand, Lytle Creek, as the source of IPMP, may contribute more to the overall odor nuisances or even become the sole odor source during precipitation extremes due to stronger sediment disturbance leading to more efficient IPMP formation. Future studies should investigate how future variations of precipitation caused by climate change will affect the relative contribution of the two water sources (Silverwood Lake and Lytle Creek) to the overall odor nuisances in SWTP's drinking water and its implications for water resource management.

GEO and 2-MIB are the most common T&O compounds in water worldwide (Devi et al. 2021). Thus, their occurrence in water sources has been exclusively evaluated in order to understand T&O events (Howard 2020; Lee et al. 2020, 2023; Chislock et al. 2021; Goodling 2021; Franklin et al. 2023; Hooper 2023; Hooper et al. 2023; Jeong Hwan et al. 2023). However, their prevalence does not necessarily rule out the importance of other T&O compounds. The point was made in real-world scenarios in this case study where both the methods, to monitor GEO and 2-MIB only and an alternative method combining FPA, Sensory GC, and GC/MS, were applied simultaneously to the same drinking water for odor evaluation. By equating the presence of GEO and 2-MIB to the overall odor occurrences, it was concluded by monitoring the two odorants alone that SWTP's drinking water odor mainly occurred during late summer and mid-fall to early winter, and the odor originated from Silverwood Lake.

However, the application of FPA to clarified water successfully captured the presence of a moldy odor that could not be attributed to earthy/musty GEO and 2-MIB. Analysis by GC/MS combined with Sensory GC identified IPMP as the culprit. More importantly, it was also found that IPMP and its moldy odor were formed in the Lytle Creek supply instead of Silverwood Lake, where GEO and 2-MIB originated. Besides, occurrences of moldy IPMP took place during late fall to mid-winter instead of late summer and mid-fall to early winter when GEO and 2-MIB occurred. Similar cases where odorants other than GEO and 2-MIB occurred in drinking water during different periods and/or at different locations compared to the two compounds have been reported by previous research (Chen et al. 2010; Guo et al. 2021a; Adams et al. 2023a, 2023b; Zhu et al. 2023). Therefore, the chemical causes as well as the spatial and temporal patterns of drinking water odors cannot be comprehensively understood by monitoring GEO and 2-MIB only, despite the popularity of the approach. By comparison, the application of FPA first to perceive the presence of all odors is needed. Then the use of Sensory GC and GC/MS is necessary to identify and monitor the odorants causing the odors. The combination of these methods is an effective alternative for the analytical evaluation of drinking water odors.

Apart from better odor analysis, the combination of FPA, Sensory GC, and GC/MS can lead to better water management practices than GEO and 2-MIB monitoring only. For example, based on GEO and 2-MIB analysis, drinking water at SWTP could be considered low-risk in terms of odor nuisances during mid-winter due to low levels of the two odorants. However, IPMP could be formed during the same time and lead to odor events when monitoring efforts by water suppliers might have been limited because of low expectations of nuisances. Similarly, monitoring of GEO and 2-MIB only could lead to over-reliance on Lytle Creek supply by SWTP for drinking water when possible due to the wrong conclusion that Silverwood Lake, as the other available source, was the primary odor source (Section 3.1).

In conclusion, the combination of FPA, Sensory GC, and GC/MS could help the water facility with multiple water sources to better decide which water source to use and when to use it compared to GEO and 2-MIB monitoring. The conclusion is consistent with the findings by Zhu et al. (2023). In that study, analysis of GEO and 2-MIB only pointed to the supplementary reservoir as the primary odor source among multiple water sources. However, more complaints were made when the share of the supplementary reservoir in the total water supply was low or zero due to other odorous compounds. Thus, monitoring of GEO and 2-MIB only could not be relied on for water source selection in that case either.

Finally, the combination of FPA, followed by Sensory GC and GC/MS, can better direct odor mitigation efforts. At the time of the study, a plan to control the odor problems in SWTP water using granular activated carbon (GAC) filtration was under consideration. A feasibility study was planned for the future GAC filters. Specifically, clarified water would be spiked with high levels of GEO, 2-MIB, and IPMP to simulate an odor event and then passed through bench-scale GAC columns. By testing the influent and effluent of the GAC columns for GEO, 2-MIB, and IPMP, the ability of GAC filters to remove all three compounds simultaneously could be assessed. In this way, a possible scenario where GAC filters could only remove GEO and 2-MIB efficiently but not IPMP could be identified before the GAC filters' costly full-scale application. Such a possibility could not have been explored if only GEO and 2-MIB had been monitored for odor evaluation because IPMP would not have been identified in that way.

This study validated the use of FPA followed by Sensory GC and GC/MS to be an effective method to analyze the chemical causes as well as the spatial and temporal patterns of drinking water odors. FPA can be used to characterize drinking water odor nuisances regardless of their causative compounds. While odorants cannot be identified by FPA itself, the confirmed odor characters can be used as clues to identify the odorants by Sensory GC and parallel GC/MS full scan analysis. Then, analytical methods, such as GC/MS SIM in this study, can be applied to monitor all identified odorants in drinking water and understand their patterns. By monitoring only ions of interest with specific m/z values, GC/MS SIM has higher sensitivity than GC/MS full scan analysis, but the full scan analysis is usually needed at first to identify compounds for quantitation and their corresponding ions for monitoring (Zhu et al. 2022). Finally, FPA can provide feedback to the analytical methods in terms of the list of target compounds and their sensitivities. For example, confirmation of moldy odor character not attributable to the initial target compounds GEO and 2-MIB in this study by FPA (Section 3.2.1) indicated the presence of an odorant other than the two compounds. Besides, the perception of moldy odor by FPA on 07/11/2022 and 09/01/2023 in this study while its causative compound IPMP was undetected by GC/MS SIM (Section 3.2.3) proved that the sensitivity of GC/MS SIM for IPMP was insufficient. As a result, the three methods can complement each other for successful odor evaluation despite each of their limitations.

However, areas for improvement were also noted in the methodology. First, the frequency of sampling and analysis should be higher. The fact that the first-ever confirmation of moldy odor took place in SWTP filtration effluent instead of clarified water before filtration indicates that IPMP likely passed through the clarified water collection point between two weekly sampling events. Fortunately, the IPMP was retained in the pump station's well for filtration effluent storage and was captured there. In addition, Suffet et al. (1996) found that most drinking water T&O events lasted for less than a week, which is another justification for more frequent sampling and analysis. Since the limiting factor for frequency in this study was the time-consuming sample extraction by SPME and subsequent GC analyses, the adoption of an automatic GC sampler capable of SPME can greatly improve the overall sample processing efficiency.

Aside from frequency, the sensitivities of GC/MS SIM should be improved. In this study, though the odorant other than GEO and 2-MIB was identified as IPMP (Section 3.2.2), GC/MS SIM could not detect IPMP in samples with moldy odor perceived by FPA. Moreover, the LOD of GC/MS SIM for 2-MIB in this study was significantly higher than its OTC at 45 °C, which may explain samples where a strong earthy/musty odor was detected by FPA yet GEO concentration was low (Section 3.2.1). Alternative sample extraction techniques can be attempted for better GC/MS SIM sensitivities. For instance, Lian et al. (2019) achieved lower LOD values for IPMP (0.2 ng/L) and 2-MIB (0.3 ng/L) than those in this study (IPMP – 0.5 ng/L; 2-MIB – 3.3 ng/L) using online purge-and-trap GC/MS. The LOD value for 2-MIB (0.3 ng/L) was significantly lower than its OTC value of 1.2–1.3 ng/L at 45 °C (Piriou et al. 2009). Similarly, the LOD value for IPMP (0.2 ng/L) was the same as its OTC value at 40 °C (Young et al. 1996), though its OTC value at 45 °C is likely lower (Section 3.2.3). As for GEO, the LOD value was the same as the LOD in this study (0.2 ng/L), and it was lower than the OTC value of 0.4–0.86 ng/L for (-)-GEO (the naturally present isomer) at 45 °C (Piriou et al. 2009).

Finally, it should be noted that FPA and Sensory GC require extra resources that may not be readily available to water utilities. In particular, a panel with multiple individuals and regular training with specific standards is necessary for both methods. It may result in additional hiring costs, sample analysis time, and training activities compared to GC/MS SIM alone, especially if an autosampler is available for GC/MS SIM. To minimize the burden on water utilities, a future study is warranted to evaluate the feasibility of setting up a laboratory that specializes in FPA and Sensory GC. In this way, multiple water facilities can send their samples to the same laboratory for FPA and Sensory GC instead of setting up a separate panel for every facility.

In this study, a comparison was made between the common practice in the water industry to monitor GEO and 2-MIB only and the combination of FPA to characterize all odors and subsequent Sensory GC and GC/MS to identify and monitor the odorants responsible for odors perceived by the FPA in their performance on drinking water odor evaluation. While GEO and 2-MIB mainly originated from Silverwood Lake during late summer and mid-fall to early winter, IPMP was found by the combination of FPA, Sensory GC, and GC/MS to occur in Lytle Creek from late fall to mid-winter. The different occurrence patterns between IPMP and GEO, 2-MIB were likely the result of different formation mechanisms. As a result, monitoring of GEO and 2-MIB alone failed to determine the chemical causes as well as the spatial and temporal patterns of odor events. Conclusions based on GEO and 2-MIB only could lead to problematic actions by SWTP during source selection and treatment application. By comparison, the application of FPA followed by Sensory GC and GC/MS allowed for the identification of potential problems with these actions and implications of climate change for drinking water odor occurrences. The study should raise awareness of odorous compounds other than GEO and 2-MIB, their role in drinking water odor issues, and the ability of the combination of FPA, Sensory GC, and GC/MS to identify them and take them into consideration.

We thank San Gabriel Valley Water Company for the financial support of the study. It is acknowledged that Josh M. Swift of the San Gabriel Valley Water Company, 15966 Arrow Route, Fontana, CA 92335, was the extremely helpful coordinating official of the Water Company that enhanced the ability of this study.

Free and informed consent of the participants or their legal representatives was obtained and the study protocol was approved by the appropriate Committee for the Protection of Human Participants UCLA Medical Institutional Review Board (MIRB), by the UCLA, CA, USA, IRB#11-002514 from 07/12/2021 to 23/05/2024.

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

The authors declare there is no conflict.