Home selection

Homes in urban and rural areas were selected across eastern Massachusetts and Rhode Island using a combination of quota sampling [26] and snowball sampling [27]. Quota sampling was used to achieve a relatively balanced set of households of different types and ages (single-family, multi-family, and multi-unit apartments). Snowball sampling was used to recruit households via word of mouth and social networks. We recruited a total of 195 homes for this study, 175 of which had gas service and 20 of which had no gas service. Household addresses, including addresses provided by the funders of this research, were not shared beyond the investigators. A limitation of this study is that of the 195 households, only two were rentals. Indoor [CH₄] data from homes was collected on 40 dates during the period of February 24, 2024 to June 20, 2024.

Participants were asked to keep doors and windows closed prior to the visit, typically a day prior to visits, but normal traffic in and out of the homes was not curtailed. Participants were not asked to limit the use of gas cooking stoves or ensure that heating was on prior to arrival, as the study was meant to capture a moment of typical gas use inside a home. This study did not collect information on whether residents or researchers could smell gas upon our arrival to the home.

Measurements were taken during the day from 7am to 5 pm. Outdoor [CH₄] was taken just prior to entering the home with a stable reading for 5 seconds. Indoors, measurements were taken at approximately the middle of the floor level. Indoor and outdoor temperature were not tracked. No homes had repeated visits.

Methane concentration data collection:

We utilized Cavity Ringdown Spectrometry (Gas Scouter G4301 Mobile Gas Concentration Analyzer, Picarro, Inc., Santa Clara, CA USA) to visit homes to measure indoor air samples, in basements and first, second, and third floors as applicable/available. This instrument, with a raw precision of 0.003 ppm [CH₄], is suitable for detecting elevated [CH₄] in ambient room air at greater precision than the 1 ppm resolution instruments used in Rowland, et al. [17] for the ambient room air leak survey part of that study. Methane measurements were made in the center of rooms, without regard to proximity to gas appliances. Checks on the analyzer calibration were performed three times during the course of this study; near when sampling began (February 29, 2024), approximately midway during sampling (April 23, 2024), and near the conclusion of home sampling (June 19, 2024). Calibration check results are presented in Appendix 1.

Because outdoor ambient [CH₄] values can vary by up to hundreds of parts per billion on a diurnal basis, we sampled outside air immediately prior to entering, and immediately after exiting test homes. With a rising (10% −90%) or falling (90% − 10%) instrument response time of 5 seconds, we waited at least 30 seconds for readings to stabilize in an indoor or outdoor environment before recording values. We recorded the difference between indoor and outdoor methane so that each home studied had its own outdoor [CH₄] comparison value. This work was completed during late winter to late spring 2024, when New England homes typically keep windows and exterior-facing doors closed. To ensure comparable indoor conditions, householders were asked to minimize ventilation through windows or doors for at least a day prior to measurements. We did not request any further special weather sealing or insulation.

When possible and convenient for the householder, we conducted brief searches for sources of elevated [CH₄] inside homes, typically on pipes, pipe connections and gas appliances including stoves, water heaters and gas boilers. A combination of close-up [CH₄] readings and bubble observations from a soap solution applied to suspected leak points were used to pinpoint leaks that could contribute to elevated [CH₄] in the ambient room air.

Data Analysis

Using RStudio© Version 2024.12.1 + 563, the Mann-Whitney U test was used to evaluate the comparison between no gas homes (n = 20) and gas homes (n = 175) because of the small sample size of no gas homes and the long tail distribution of leaks in the gas homes. We used the Wilcoxon Sign-Rank test for paired samples of outdoor and indoor air for no gas homes and a paired t-test was used to compare mean indoor and outdoor [CH₄] concentrations in the gas group. Linear regression was used to evaluate the association between elevated methane and home age and size.

Housing characteristics

A breakdown of the housing types studied was single-family 63.5%, multi-family (2–3 units) 26.6%, and multi-family (4 + units) 9.9%. There were two row houses and we categorized them as multi-family (2–3 units) because they shared a wall with another dwelling. For context, the US Census Bureau estimates that in 2023, 57.2% of Massachusetts and 61.1% of Rhode Island households are single-family. Natural gas is used for heating and cooking in about half the homes in Massachusetts [28] and Rhode Island [29]. In this study, among gas service households, 87% reported using gas stoves, 80% gas water heating, 75% gas space heating, 39% gas dryers, and 2% gas fireplaces.

The largest [CH₄] elevations were in basements, where [CH₄] concentrations were nearly double outdoor concentrations, with a mean 1.91 ppm above outdoor [CH₄], while [CH₄] in 1st, 2nd, and 3rd floors averaged about 1.29 ppm higher than outdoor ambient. In the gas group, mean indoor [CH₄] readings were higher than mean outdoor concentrations for all floors. By contrast, homes without gas service were either not significantly different from outdoor ambient (p > 0.10, Table 1), or only marginally on first floors with a [CH₄] elevation of 0.10 ppm (p = 0.04). These [CH₄] elevations in homes without gas service may have been caused by biotic factors such as human metabolism or improperly functioning plumbing.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Summary statistics of [CH₄] concentrations in parts per million (ppm) outdoors and indoors. Indoor concentrations are unadjusted for outdoor concentrations. P-values are for differences between indoor and outdoor [CH₄].

https://doi.org/10.1371/journal.pone.0348496.t001

In 195 homes studied, homes supplied with gas showed significantly higher median [CH₄] concentrations than homes not served by gas after outdoor methane concentrations were subtracted (Table 2, Fig 1).

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Summary statistics adjusted for outdoor [CH₄] concentrations. Indoor minus outdoor methane concentrations ([CH₄], parts per million) from 195 households surveyed in this study. P-values are for differences between gas and no gas homes.

https://doi.org/10.1371/journal.pone.0348496.t002

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Adjusted indoor [CH₄] concentrations of gas and no gas households are shown by floor (basements, 1st, 2nd, and 3rd floors, from left to right).

Not all homes had basements, 1st, or 2nd, and 3rd floors. The mean is denoted with an “X” and the median is shown with a line within the box. Outliers not shown extend up to 38 ppm. See Fig 2 for more information about outliers. “Gas” refers to households with gas service; “No Gas” refers to households without gas service, all-electric and without propane.

https://doi.org/10.1371/journal.pone.0348496.g001

The distribution of leak sizes in homes with gas service was not bell-shaped, as indicated by the downward shift of the median values of [CH₄] difference shown in Fig 1 (shown by the line in the box) from the mean [CH₄] values (shown with an X). As across other sectors of the fracked gas process chain [30,31], the distribution of leak sizes in homes with gas service were skewed with a “long tail” containing a small proportion of large [CH₄] values and a large proportion of homes with lower [CH₄] elevations (Fig 2).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Histogram showing numbers of homes with different levels of indoor minus outdoor methane ([CH₄]) concentrations.

[CH₄] shown is the highest [CH₄] obtained at any floor. Black bars represent homes with gas service; the gray bar represents homes without gas service. The + 1.0 ppm enhancement noted in the figure is equivalent to a 3.0 ppm total [CH₄] reading.

https://doi.org/10.1371/journal.pone.0348496.g002

Ninety-three percent of gas homes had methane concentrations higher than no gas homes and nearly half of gas homes had elevated methane concentrations greater than 1.0 ppm [CH₄] concentrations above outdoor [CH₄] levels.

We had the opportunity to conduct brief searches for suspected methane sources for elevated [CH₄] in 137 (78%) of the 175 homes with gas service. Of these homes, we positively identified leak sources in 124 homes, or in 91% of homes checked. Leak sources were found all along the distribution system within homes: pipe connection (63), appliance malfunction or incomplete burning (53), shutoff valves (13), migration (3) and corrosion (1). Eighteen (18) of the leaks came from gas company pipes. Beyond the statistical association of higher [CH₄] in homes with gas, this provides further evidence linking elevated [CH₄] in homes to incomplete combustion, and leaks from gas pipes, connections, and appliances.

There did not appear to be a relationship between home attributes or geography and elevated [CH₄] in homes served by gas. For example, the 10 homes showing the highest elevation of [CH₄] relative to outdoors were located in nine different municipalities and in both states. Of these top 10 homes, all contained basements and first floors, eight had second floors, and three had third floors, and all ten were found in single family homes. There was no relationship between indoor [CH₄] and either home age (P > 0.1) or square footage (P > 0.1). Thus, living in a relatively newer home should not be a reason to assume it is less likely to contain elevated methane. This study only included two rental units, and we weren’t able to evaluate if there were differences between methane concentrations in owner-occupied homes and rental homes.

Degraded service lines and safety

Although it was not an objective of this study to investigate leak origins or the condition of pipes, fittings and appliances, when possible, attribution of elevated [CH₄] to gas leaks was evaluated by identifying leak locations and incomplete combustion. In doing this work, one unanticipated finding was that three of 18 (17%) bare steel service lines (observed in the opportunistic process of leak origin determination) in homes served by gas were found to be in poor condition and called in for repair. The service line material can often be determined at the outside meter coupling or at the service entrance inside the basement of the home.

Prevalence of elevated methane

We observed that in 175 homes served with gas, elevated indoor methane is not only common, but to be expected. Conversely, in gas-free homes, elevated indoor methane levels were found only marginally on first floors, with less than 1/10th the elevation of [CH₄] on average than in homes served with gas. Basements in gas homes had the highest methane concentrations compared to no gas homes and other floors in gas homes. In New England, basements are commonplace and are a typical place to locate a gas service line entrance, heating system, water heater, clothes dryer, and gas lines leading to other appliances. Incomplete combustion from basement appliances and gas leaks from service lines contributed to these higher concentrations.

Data collected in this study spanned a period from winter through late spring. Although we attempted to ensure comparable indoor conditions by requesting householders minimize ventilation at least a day prior to sampling, it is plausible that weather conditions and/or incomplete combustion in indoor gas appliances could have produced seasonally varying levels of indoor [CH₄] enhancement. Yet in general we found that when there was an elevated indoor methane reading, there was a corresponding gas leak. We found gas stoves and water heaters had fitting leaks indoors even when off. Therefore, in our opinion having the heat on or off made little difference in our ability to detect elevated methane from a leak, although it is possible that whether the appliance was on or not may contributed to the magnitude of the leak.

Prior to this research, the only categorical finding on prevalence of detectable elevated [CH₄] in indoor ambient air in gas-served homes reported 13 of 323 homes with a gas leak, or about 4% of households [17]. With the finer resolution analyzer used in this study, we estimate from the data presented in Fig 2 that 93% of homes served with gas had greater [CH₄] than homes without gas. Reconciling these vastly different estimates is obtained by comparing the number of homes found with [CH₄] of 10 ppm elevation relative to outdoors. In this study we find six homes of 175 gas-served homes, or 3.4% to compare with the 4% found in Rowland, et al. [17]. Analogous to the “super-emitter” character of gas leaks across the gas process chain, these findings indicate there could be gas leaks that are under explosive limits but go undetected by residents.

Implications for safety and methane emissions

Our preliminary finding that one sixth of steel service lines were in poor condition is a safety concern and warrants focused attention on bare steel pipe condition. The majority of leak-prone service lines are bare steel [1]. We recommend that bare steel and other leak-prone service lines be inspected on an annual basis. There is currently no standard minimum frequency of required service line inspection across states.

A practical implication of our finding of a skewed distribution of [CH₄] in homes is that homes with the largest [CH₄] can be prioritized by regulators and utilities for repair, and leak flux rates can be assessed according to a recently developed method [23]. It is important to note that just because a home exhibits relatively large [CH₄] does not mean it has a large flux rate, as it depends on the air exchange rate in the home. A small leak in a home with a tight envelope and low air exchange could show a large elevation in methane, while conversely a large leak in a home with a high level of air exchange could show a small elevation in methane.

This study was not designed to determine methane emission rates but rather stable concentrations, where steady indoor air composition is promoted by weather conditions where windows and doors are likely to be closed. To obtain rate estimates, at least two measurements must be made over a recorded time interval, with an additional informational constraint such as a known [CH₄] starting point, or a known room or home air exchange rate. This approach to estimate emissions was made in Nicholas et al. [23]. That study, while having a distinctly different objective, nevertheless is in broad consistency with findings here. For example, Nicholas et al. [23] identified leaks in 18 of 20 (90%) of homes, similar to the 93% of homes with gas documented here that had [CH₄] higher than the average of non-gas homes; and among the 137 homes in this study where we had permission to search for the methane source, we pinpointed the leak or incomplete combustion origin 90% of the time. Secondly, the suggestion of a long-tailed distribution with potential “super-emitters” in Nicholas et al. [23] is further supported by the long-tailed distribution found here (Fig 2) with almost ten times as many homes studied.

Implications for air quality and health

These undetected methane concentrations in indoor air have health implications because of the potential for exposure to co-pollutants. Heating and cooking with natural gas is common in the Northeast, with half of households using gas in Massachusetts and Rhode Island. The majority of natural gas in the United States is extracted through a process called hydraulic fracturing and is commonly known as “fracked” gas. In the United States, 89% of U.S. dry natural gas production in 2022 was extracted using fracking, and its production is expected to continue to grow, with estimates of it making up 93% of production by 2050 [32]. The presence of volatile organic compounds like the carcinogen benzene, have been characterized as part of the fracking process [33], found near fracking sites [34,35], and characterized in distributed gas used in residential end-uses in Massachusetts [16]. We distinguish “fracked” natural gas from conventional methods because of its significant growth in use since 2005 [32,36], known deleterious impact on water sources at the site of drilling [37], and because carcinogens used in the hydraulic fracturing process can affect human health at the well-sites, and potentially impact human health in the home where gas is used. Massachusetts gas contains carcinogenic compounds including benzene [16], of concern for human health due to both acute and chronic exposure [38,39].

In addition to carcinogens like benzene, odorants (mercaptans) in gas also have health implications. Mercaptans were documented to have acute health impacts, such as from a large gas leak in 2015 in Aliso Canyon, California [40]. While little is known about long-term, low-dose exposure to fracked gas odorants in residential settings, mercaptans are known to impact human health; chronic mercaptan exposure in occupational settings is regulated by the National Institute for Occupational Safety and Health (NIOSH) (REL C 0.5 ppm (1 mg/m3) [15-minute]) [41], and mercaptan exposure caused lung inflammation and apoptosis in rats [42].

While it is assumed people will be able to smell gas indoors, evidence suggests this assumption may not always be true. There are differences in concentration of odorants across the gas system [17], and differences in individual’s ability to smell gas, may compromise the safety of gas in homes. People with poor or compromised sense of smell may be unaware of leaking gas. Compromised sense of smell is often not detected by individuals until it is advanced and is more common among the elderly [43]. People can also become adapted or habituated to persistent odors to the point that they no longer detect them [44]. On streets and sidewalks, our previous experience (e.g., in informal observations during research published [6]) is that many people, if focused on the task, can smell gas leaks at methane levels only 1 part per million (ppm) above the ambient background of 2 ppm, or at 3 ppm). In our study, 47% of households with gas that we studied had methane concentrations at or above 3 ppm. While they are detectable, at least to the occasional visitor with a keen nose, they may not be noticed by habituated or adapted residents. We recommend that gas companies be required to use equipment with a detection threshold of 1 ppm [CH₄] above ambient, and to be required to report elevations above 1 ppm [CH₄], relative to outdoors, in interior spaces to residents and in annual reports to state regulators.

Almost all relevant studies on health impacts of gas in homes to date have focused on health impacts of combusted (not leaked) fracked gas. Studies that have looked at the health impacts of unburned fracked gas focused predominantly on the upstream gas sector, near locations of fracking [45–47]. This study complements studies that document constituents of gas found in homes (e.g., Michanowicz, et al. [16], Lebel, et al. [22]) because it describes the high prevalence of elevated methane in residential homes. This study represents, to our knowledge, the largest household survey in New England using a comparison group and documenting the presence and prevalence of elevated methane in ambient room air and demonstrating its association with gas equipment in homes.

Limitations and future research

While this study contributes knowledge that elevated [CH₄] in New England homes served by fracked gas is common, there are a number of study limitations: we did not measure fluxes of methane, concentrations of co-pollutants such as benzene, nor assess odorant detectability by residents. We did not sample comprehensively nor randomly across all housing stock and household types (including renters), household ventilation types or degree of insulation, and did not undertake a comprehensive survey of service line conditions. These limitations raise further questions for follow-on studies that are conducted in as large or larger sample sizes than the 195 homes studied here.