Sumit
Sankhyan
a,
Julia K.
Witteman
b,
Steven
Coyan
a,
Sameer
Patel‡
a and
Marina E.
Vance
*ab
aDepartment of Mechanical Engineering, University of Colorado Boulder, 1111 Engineering Drive, 427 UCB, Boulder, CO 80309, USA. E-mail: marina.vance@colorado.edu
bEnvironmental Engineering Program, University of Colorado Boulder, 4001 Discovery Drive, Boulder, CO 80303, USA
First published on 4th May 2022
In this study, we deployed multiple low-cost air quality monitors (AQMs) to investigate the transport of kitchen-generated fine particulate matter (PM2.5) into the bedrooms of four homes of different sizes over a period of more than nine weeks at each home. We also estimated the human exposure to PM2.5 associated with each microenvironment and evaluated the effects of using a portable air cleaner (PAC) to reduce those exposures. To select the best AQM for these analyses, we compared the field response of five commercially available models with that of a research-grade optical particle spectrometer. The AirVisual AQM showed the best correlation during collocation phases with R2 values in the range of 0.5–0.9 during cooking and background periods for all locations. The bedroom monitors picked up cooking emissions from the kitchen area within 1–45 min depending on the layout of each home, and median PM2.5 concentrations in the bedroom were up to 30% lower than those in the kitchen. Results from the exposure analysis suggest that PAC use is an important intervention strategy for reducing personal PM2.5 exposure, especially in indoor environments where cooking is the main source of PM2.5 concentrations. In three of the four homes using PAC consistently in the kitchen or bedroom area during intensive cooking periods reduced overall exposure values by 30–90%. Moreover, during nighttime periods, PAC usage in the bedroom area yielded the lowest levels of PM2.5 exposure for all the homes.
Environmental significanceThis article describes a comprehensive study of the PM2.5 response from a low-cost air quality monitor to study the transport between the kitchen and bedroom areas of four different houses and the resulting exposures at these fixed locations. We also quantified the benefits of using a portable air cleaner (PAC) in the kitchen and bedroom areas to reduce the resulting exposures due to indoor cooking and outdoor penetration. This study brings forth multiple results of interest to the science community as well as the general public, such as the effect of different control strategies such as window opening, extracting range hood use over the stove, and PAC use to reduce the overall PM exposure values in built environments. |
For indoor environments, basic strategies adopted for improving IAQ (mainly by reducing PM2.5 concentrations) include source control, increased ventilation, and pollutant removal.19,20 Source control measures include using improved cookstoves for lower emissions, switching to cleaner fuels for residential heating and cooking purposes.21,22 PM2.5 levels indoors can also be managed by using a mechanical ventilation system or opening the windows to increase the ventilation rates. However, the latter is only applicable in areas with low levels of ambient pollutants.23,24 In cases where source control and natural ventilation are not effective control strategies, using range hoods over the stove and portable air cleaners near the receptor can be a good alternative to reduce PM2.5 exposure.25–30 In terms of control strategies, AQMs provide an opportunity to alert consumers about degrading levels of air quality in their homes and enable them to perform some of these mitigation strategies. AQMs can also be deployed in different areas of a household, and their real-time data can be used to decide on the best placement of air cleaners for effective particle removal in multizone indoor environments.31 In terms of occupant exposure, people spend about 70% of their time in a residence.10 Of the time spent at home, people are estimated to spend about 10% of it in the kitchen microenvironment and 53% in the bedroom microenvironment.32 Although the time spent in the kitchen is significantly smaller than that in the bedroom, higher total PM2.5 exposures may take place in the kitchen depending on cooking habits, control strategies used, and outdoor penetration of ambient PM2.5.
Most AQMs measure particulate matter (PM) concentrations using a low-cost PM sensor which either uses an optical particle counter (OPC) to count particles in various size bins based on assumptions about particle shape and refractive index or use a nephelometer to measure the amount of light scattered by each particle which is in turn converted into a mass concentration using a conversion factor based on laboratory calibration.33 Previous studies, focusing on determining correction factors for different types of aerosols and ambient locations, have reported that the sensors used in AQMs need to be calibrated according to the local conditions for better correlation with data reported by federal equivalent methods of measurement.34–36 In recent studies, the hygroscopic growth of sampling aerosols in humid conditions (relative humidity >50%) has also been shown to affect the PM response of AQMs.37,38 There are also concerns about their performance during periods of low concentrations or very high concentrations, especially in ambient environments where they tend to deviate from linear correlation with reference instruments.39–41 Even with these challenges, the data reported by low-cost air quality monitors can provide reliable results for quantifying personal exposure, especially compared to exposure values estimated from outdoor fixed monitoring stations.42 For consumers, low-cost AQMs can be especially informative in educating occupants in real-time about their own activities that generate large amounts of PM2.5 and, conversely, actions that are effective in lowering concentrations.
The main goal of this study was to assess the indoor transport of PM2.5 and its mitigation in four households of different sizes and configurations for a total duration of nine months. Specific research objectives were to: (1) compare the PM2.5 concentration measurements from four different AQMs with a research-grade instrument to select the best AQM for subsequent research objectives. (2) Study the transport of PM2.5 between the kitchen and a bedroom for each home during cooking activities by placing identical and inter-corrected AQMs in these two areas of the household. (3) Determine the effectiveness of deploying a filtering portable air cleaner (PAC) to reduce PM2.5 exposure primarily due to indoor cooking. (4) Investigate the effects of PAC placement—in the kitchen or a bedroom—in reducing personal PM2.5 exposure during different periods of the day.
We chose the OPS as a comparison instrument for this study because it also operates on the principle of single particle counting using a laser and photodetector assembly.46 It can measure particles in the size range of 0.3–10 μm (16 bins) with an inlet flow rate of 1 liter per minute. Its compact size and low pump noise make it suitable for indoor environments. Although the OPS is not a regulatory reference instrument, it has been widely used in previous studies and has shown good agreement with other research-grade instruments.46,47 The four AQM models used in this study are also listed in Table 1. Two identical units were deployed for each AQM model.
Product | Foobot, home | IQAir, AirVisual Pro | PurpleAir, PA-II-SD | PurpleAir, PA-I-Indoor |
---|---|---|---|---|
Air quality measurements | PM2.5, total volatile organic compounds (TVOCs) | PM2.5, carbon dioxide | PM1, PM2.5, PM10 | PM1, PM2.5, PM10 |
Time resolution | ∼300 s | 10 s | 80 s | 80 s |
PM sensor | Sharp GP2Y1010AU0F | AVPM25b | Plantower PMS5003 | Plantower PMS1003 |
PM detection technique | Light scattering (0.3–2.5 μm) | Light scattering (0.3–2.5 μm) | OPC (6 size bins 0.3–2.5 μm) | OPC (6 size bins 0.3–2.5 μm) |
Cost estimate | $240 | $270 | $230 | $180 |
Example studies that used or evaluated these AQMs | 48–54 | 18, 48, 49, 55 and 56 | 38, 41 and 57–61 | 33 and 62 |
Because all particle instruments used in this study are limited to particles > ∼0.3 μm in diameter, their measurements are likely to underestimate actual PM2.5 concentrations because they miss potential PM mass contributions from particles < 300 nm, which may be important indoors, especially during some indoor cooking activities. As such, all PM2.5 concentrations reported in this work should be interpreted as PM0.3–2.5.
The different AQMs chosen for this study had been extensively tested in previous studies and were readily available in the market. PM sensors used in the PurpleAir AQMs (Plantower) have been deployed on large scales in different studies for various applications, primarily in ambient environments.41,57–59 Foobot AQMs have been used to quantify personal exposures in different indoor settings.51–53 In recent studies, the AirVisual Pro has been shown to be a reliable AQM with ease of access and better accuracy in indoor environments.48,55
The PAC used in this study (EJ120, Oransi, Raleigh, US) uses a combination of an activated carbon filter and a MERV 17 (Minimum Efficiency Reporting Value) filter to provide a maximum air flow of 330 cubic feet per min (0.16 m3 s−1). It is recommended for rooms sizes up to 116 m2 with 2 air changes per hour.63 The manufacturer recommends changing the filter every 12 months, so we used the same filter for all the homes during the entire study period of nine months. Filter loading effects were not quantified for this study and were assumed to be negligible due to the generally low background PM concentrations found in all four homes.
The study spanned nine months in 2019 and 2020, comprising fall, winter, spring, and summer seasons in Boulder, CO. The ambient daily average PM2.5 data provided by United States Environmental and Protection Agency (US EPA) for Boulder County location during these nine months is also shown in Fig. S3.† All the four homes used for this study were within a 10 km radius from the monitoring station. In Homes 1–3, the windows and doors were kept shut to maintain comfortable living conditions (T = ∼21–25 °C, RH = 30–50%) inside the homes through mechanical HVAC systems. Home 4 did not have a cooling system, so the windows were kept open continuously throughout the deployment because that coincided with the peak summer season.
Homes 1 and 3 had recirculated microwave range hoods over the stove, whereas Home 2 had an extracting wall mounted range hood (Vent-a-hood dual blower 600 CFM). Extracting range hood usually have a higher capture performance.64 Home 4 did not have a range hood over the stove. The bedroom areas in Homes 2 and 3 were mostly unoccupied throughout the day whereas in the case of other two homes, the bedroom areas were inhabited. The occupants also maintained a time log with information about the start and end times of all cooking activities. This study was exempted from Institutional Review Board review for lacking participant information or risk. Additional details regarding each location are also given in Table S1.†
Phase | 0A | 1 | 2 | 3 | 0B |
---|---|---|---|---|---|
Activity | Collocation (no PAC) | No PAC | PAC in kitchen | PAC in bedroom | Collocation (no PAC) |
AQM location | Kitchen | Kitchen + bedroom | Kitchen + bedroom | Kitchen + bedroom | Kitchen |
Duration | 2 days (min) | 3 weeks | 3 weeks | 3 weeks | 2 days (min) |
Next, we combined the collocation data from all the homes into two distinct periods—cooking and background. The resulting distributions of CAQM/COPS values (in 1 min time resolution) are shown in Fig. 2. All AQMs presented PM2.5 concentrations that were higher—sometimes >10× higher—than that of the OPS. Both AirVisual AQMs had the highest coefficient of determination (R2) values with the OPS PM2.5 data as shown in Tables S3 and S4.†
During cooking periods, the median CAQM/COPS values for the two AirVisual sensors (AV1 and AV2) were 2.2 and 1.7. The corresponding values for the PurpleAir and PurpleAir (I) AQMs ranged from 1.6–1.9. These results agree well with previous studies that also reported overestimation of PM2.5 concentrations within a factor of 2 by different low-cost AQMs compared to mass-based measurements from different reference instruments.33,49,51,58,65 Moreover, low-cost OPCs and nephelometers have been shown to exhibit greater amounts of error in mass loading values compared to the high end OPCs in these conditions.66 The R2 values for all AQMs were higher during background periods than cooking periods. During background periods, particles are likely to have penetrated from outdoors and are more likely to match low-cost sensor calibration inputs. Infiltrated particles are also less unlikely to suffer strong temporal and spatial gradients. During cooking periods, particle concentration, size distribution, optical properties, and chemical composition are likely to change quickly, creating strong temporal and spatial gradients. Sudden changes in these parameters may have led to deviation of response linearity.
PM2.5 measurements from the AirVisual units presented the highest values of R2 in the range of 0.5–0.9 for both units, and the corresponding slope values ranged between 1.3–2.2. Moreover, the mean normalized bias (MNB) and the root mean squared error (RMSE) values were lowest for the AirVisual 2 unit as compared to other AQMs during both background and cooking periods (Tables S3 and S4†). Based on these results, the PM2.5 concentrations reported by AirVisual had the best agreement with OPS among the AQMs tested in this study. For this reason, we used AirVisual results for all subsequent analyses in this work. In order to get the results of the two AirVisual units in the kitchen and bedroom areas to agree, we applied a correction factor to the AV1 PM2.5 values which was derived from a linear regression analysis obtained from Phase 0 (collocation) data as shown in Fig. S5.†
A CBedrrom/CKitchen factor was used to compare the concentrations in the kitchen and bedroom area for each home. This factor was calculated by taking the time-averaged concentrations over 90 min for both kitchen and bedroom PM2.5 concentrations, considering a starting time (t = 0) the concentration peak as reported by the bedroom monitor. Similar factor mentioned as L/K ratio in Wan et al.67 has been used to compare PM levels due to indoor cooking in the living rooms and kitchens of 12 different homes. We also calculated the time difference (Δt) between the peak as it occurred in the kitchen and in the bedroom for a given cooking activity during Phase 1. A boxplot showing the distributions of Δt and CBedroom/CKitchen factors for all four homes is shown in Fig. 3.
For Home 1, the median values of Δt were highest among all the homes (45 min). This could be because the bedroom door was usually kept closed. The corresponding median value of CBedroom/CKitchen value was close to 1 since the kitchen area was adjacent to the bedroom area in this home, so PM2.5 concentrations equalized between the two spaces whenever the bedroom door was opened. A similar median value of CBedroom/CKitchen factor was also calculated for Home 3. In this home, the bedroom was located directly across the hallway from the kitchen area. For this home, the median value of Δt was ∼5 min, which is much lower than in Home 1, probably because in Home 3 the bedroom door was always open.
The distance between the stove and bedroom AQMs was largest for Home 2 (∼10 m around two corners) and, therefore, the median value of CBedroom/CKitchen (0.7) was lowest among all the homes studied. The corresponding median value of Δt for this home was 17.5 min and the bedroom door was always open. In Home 4, due to open windows in the bedroom and kitchen area, and due to the absence of interior walls between the two units (Home 4 being a studio apartment), the bedroom AQMs picked up the concentrations from the kitchen area within a minute for each cooking activity and the median CBedroom/CKitchen was 0.8.
Overall, the most important factors governing PM2.5 transport from the kitchen to the bedroom of these four homes were the presence of physical barriers between these spaces (e.g., interior walls and whether doors were kept open or shut), different layouts of kitchen and bedroom areas in each home, distance from the stove to the kitchen and bedroom AQMs in addition to the ventilation conditions (e.g., open windows). It is also interesting to note that the median values of CBedroom/CKitchen for homes which had an extracting fume hood (Home 2) and open windows in the kitchen area during cooking periods (Home 4) were lower than that value for the other homes. This could be due to the fact that these control measures prevented the majority of the kitchen concentrations from reaching the bedroom area, thereby lowering the time-averaged concentrations (calculated for 90 min post peak). These results indicate the effectiveness of such control measures in reducing PM exposure due to indoor cooking in both the kitchen and bedroom areas. This is expanded in greater detail in the next section.
For Homes 1 and 3, median concentrations during Phase 2 (PAC in kitchen) were significantly lower than that of the other two phases indicating the effectiveness of PAC use for reducing PM2.5 exposure. The median concentrations for Phase 1 and Phase 3 in all the homes also exhibit the same diurnal pattern and an overlap to a certain extent. This shows that PAC deployment in the bedroom (Phase 3) did not affect concentrations in the kitchen in a significant manner. Homes 2 and 4 had significantly lower PM2.5 concentrations in the kitchen area, due to an efficient extracting range hood in Home 2 and open windows in Home 4, as explained in the previous section. According to the inhabitants of Home 3, the high PM2.5 concentrations observed overnight in kitchen of Home 3 during could be attributed to the infiltration of tobacco and marijuana smoke from the downstairs unit. Smoke transported into the apartment through the kitchen sink drainpipes.
Next, we present the PM2.5 exposure analysis in the kitchen and bedroom areas for two distinct periods: a daytime analysis using time intervals between 6:00 am to 10:00 pm and a nighttime analysis using the remainder of the day (10:00 pm to 6:00 am).
During Phase 1 (no PAC use), daytime PM2.5 exposure values were on average 17–43% lower in the bedroom compared to the kitchen of the four homes. This is likely due to PM2.5 emissions during cooking activities increasing concentrations in the kitchen, which are then diluted and lost to surface deposition and exfiltration during transport to the bedroom.70,71 During the nighttime (Fig. 6), this difference dropped to 0–23%, which further confirms the hypothesis that the differential is driven by cooking activities.
Using a PAC in the kitchen (Phase 2) or bedroom (Phase 3) reduced the mean exposure values by 30–90% in that respective area when compared to the corresponding values from Phase 1 (no PAC use) in three of the four homes. When the PAC was used in the kitchen (Phase 2), the average mean exposure values in the kitchen area dropped by 30–70% for all homes except Home 2. For Home 2, an increase in exposure was observed during PAC use compared to Phase 1. However, this phase also coincided with the holiday season, with additional guests and significantly more cooking being performed in the home. A similar comparison between the mean exposure values of the bedroom area during Phase 3 (PAC in bedroom) with Phase 1 values also yielded similar reductions in mean values: 53% for Home 1, 46% for Home 3, and 85% for Home 4.
Also noteworthy is that for both Homes 2 and 4, the mean exposure values were slightly higher in the bedroom area compared to the kitchen area during Phase 2. This could be because Home 2 had an extracting range hood and Home 4 occupants opened windows during cooking periods. These strategies may have played larger roles in governing PM2.5 exposures than the use of a PAC in the kitchen. Therefore, the exposure values calculated for the bedroom areas may be from other sources, such as outdoor infiltration.
The use of a PAC in the bedroom (Phase 3) was also very effective in reducing daytime PM2.5 exposure in that room. The mean values in the bedroom area were 53–85% lower than in the kitchen area for all four homes. This difference is much more pronounced than during Phase 1 (no PAC use), when bedroom exposures were 17–43% lower than in the kitchen. Between Phase 2 and Phase 3, the bedroom exposure values were either in the same range (Home 1 and Home 2) or slightly lower for the latter phase in the case of remaining homes. Therefore, PAC use in the kitchen area during the daytime can be an effective option for reducing personal PM2.5 exposure levels due to indoor cooking for occupants who will be spending majority of their time indoors in the kitchen and bedroom areas combined during that period. Moreover, after the period of active cooking when the occupant moves out of the kitchen to other areas, pollutants could homogenize spatially over the entire home, therefore, maximizing personal exposure which again supports this intervention strategy of reducing emissions at the source.
The overnight PM2.5 exposure values during Phase 1 were one order of magnitude lower compared to daytime periods, thereby suggesting the role of outdoor infiltration in PM2.5 on exposure levels indoors, even though its contribution is much lower than indoor cooking. This is likely due to low ambient PM2.5 levels during this study. However, it is important to mention that the time period for daytime exposure was twice that of nighttime periods (16 h vs. 8 h). During Phase 2, the mean overnight exposure values for the kitchen area of Home 3 were ∼50% lower than those of the corresponding mean values for Phase 1 (21 μg m−3 h). A moderate reduction in mean exposure (∼35%) was also observed for Home 4, where the average exposure value during nighttime periods was calculated to be 8 μg m−3 h. Similarly, when the PAC was placed in the bedroom area (Phase 3), the mean exposure values in the bedroom area were 30–90% lower than the corresponding Phase 1 values for all the homes combined. Overall, the mean bedroom exposure values during Phase 3 were lowest among all the phases for all homes. The corresponding mean exposure values were calculated to be: Home 1 (1 μg m−3 h), Home 2 (8 μg m−3 h), Home 3 (15 μg m−3 h), and Home 4 (1 μg m−3 h).
In summary, using a PAC in the kitchen and bedroom areas reduced PM2.5 exposure between 10–90% during daytime and overnight periods in most of the homes, with only a few exceptions as shown in Tables S5 and S6.† For homes that did not have an effective control strategy (Home 3 and Home 1), the reductions in mean exposure values were usually greater as compared to the other two homes with extracting range hood and open windows for higher air exchange rates, especially during Phase 2 of deployment. Moreover, the absolute values for the reduction in mean exposure compared to no PAC phase were greater in the kitchen area than the bedroom area during the daytime periods in both Homes 1 and 3, suggesting PAC placement in the kitchen areas during daytime periods. For overnight periods when people usually spend majority of the time in the bedroom area, the PM2.5 exposure values in the bedroom area during Phase 3 were lowest as compared to the other phases in all the homes.
In terms of indoor PM2.5 transport between the kitchen and bedroom areas of different homes, concentrations in the bedroom were 70–100% of those in the kitchen. The kitchen emissions peaked in the bedrooms after 1–45 minutes were elapsed from the start of a cooking event. The fastest transport was observed in Home 4 (no internal walls) and the slowest in Home 1, where the bedroom door was kept closed. Overall, both parameters varied depending upon the layout and relative location of the AQMs in the kitchen and bedroom area with regards to the stove. Baseline conditions were investigated during Phase 1, when no PAC was employed.
The exposure analysis performed in this study suggests that PAC use is an important intervention strategy for reducing personal PM2.5 exposure, especially in indoor environments where cooking is the main source of PM2.5. The bedroom exposure values were also comparable to the exposure at the kitchen location in all the homes. During daytime (6:00 am to 10:00 pm), PAC use in the bedroom or kitchen area yielded 30–90% reductions in PM2.5 exposure in three of the four homes. Daytime exposure results also suggest that using a PAC in the kitchen results in lower exposure values in both the bedroom and kitchen areas. During overnight periods, PAC use resulted in the lowest exposure values in all homes, with a reduction in mean exposure values by 30–90% or 4–25 μg m−3 h as compared to not using a PAC in the bedroom.
This study is limited to four homes located in one city during a 9 month period when ambient PM2.5 concentrations were relatively low. As such, our exposure analysis showed greater importance of cooking activities compared to outdoor PM2.5 infiltration in driving the PM2.5 exposure values indoors. Another important caveat of this study is the fact that people on average spend much less time in their kitchen areas as compared to the bedroom areas during an entire day. However, the concentrations in the kitchen are usually higher during active cooking periods, so the resulting exposure values in the kitchen area could still be comparable to the bedroom area values especially in areas where outdoor infiltration doesn't play a major role in driving the indoor PM2.5 exposure. A study like this should be performed in a more polluted period or city and in homes with varying air-tightness levels to study the importance of outdoor infiltration that might become a greater contributor to indoor PM2.5 exposure than cooking emissions.
Finally, we did not observe any drift in PM2.5 concentrations as reported by various AQMs as they were moved from one home to the other. Overall, PM2.5 data reported by AirVisual AQM showed the best correlation with the corresponding OPS data during colocation phases with R2 values in the range of 0.5–0.9 for cooking and background periods.
Footnotes |
† Electronic supplementary information (ESI) available. See https://doi.org/10.1039/d2ea00025c |
‡ Now at Discipline of Civil Engineering and Discipline of Chemical Engineering, IIT Gandhinagar, Gandhinagar, Gujarat 382355, India |
This journal is © The Royal Society of Chemistry 2022 |