Impact of Wildfire Smoke Events on Indoor Air Quality and Evaluation of a Low-cost Filtration Method

Received: March 3, 2021
Revised: April 30, 2021
Accepted: May 9, 2021

 Copyright The Author(s). This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are cited.

Cite this article:

May, N.W., Dixon, C., Jaffe, D.A. (2021). Impact of Wildfire Smoke Events on Indoor Air Quality and Evaluation of a Low-cost Filtration Method. Aerosol Air Qual. Res. 21, 210046. https://doi.org/10.4209/aaqr.210046

HIGHLIGHTS 

• Wildfire PM_2.5 infiltration lower in residential than commercial/school buildings.
• Wildfire PM_2.5 infiltration reduced in residences with multiple filter units.
• MERV-13 and box fan a low-cost alternative to commercial filter units.
• Low-cost method effectively filters wildfire PM_2.5 and submicron particles.

Keywords: PM2.5, Indoor air pollution, Filtration, Biomass burning, Laser particle counter

1 INTRODUCTION

The area burned by wildland fires in the United States (US) increased substantially over the past two decades (Jaffe et al., 2020), exceeding 3 million hectares burned in 6 of the past 10 years (National Interagency Fire Center, 2020). This is part of a larger US trend of increasing frequency and area burned, as well as fire season length, since the 1980s (Abatzoglou and Williams, 2016; Westerling, 2016). Health impacts from exposure to smoke include cardiovascular and respiratory mortality and morbidity, adverse reproductive, developmental, and neurodegenerative diseases, as well as eye irritation, wheezing, and coughing (Reid et al., 2016; Thurston et al., 2017; Cascio, 2018). High risk populations include those with respiratory and cardiovascular disease, pregnant women and their fetuses, young children, and older adults (Rappold et al., 2017; Hooper and Kaufman, 2018). Fires in 2017 and 2018 led to the highest observed fine particulate matter (PM_2.5) concentrations ever observed at regulatory monitors in the US at the time, with 24-hour averaged values over 600 µg m^–3 in some locations (Laing and Jaffe, 2019). However, summer 2020 fires shattered those records. In high fire years, extreme PM_2.5 concentrations impact millions of people, especially in the western US. Recommendations for limiting exposure to PM_2.5 and other pollutants during wildfire smoke events focus on staying indoors, reducing infiltration, and increasing filtration (Laumbach, 2019; Davison et al., 2021).

Infiltration of PM_2.5 occurs through unintentional cracks in the building envelope, open windows or doors, and HVAC systems with outdoor intakes. The fraction of ambient PM_2.5 that penetrates indoors and remains airborne (F_in), as well as indoor source contributions (C_s), can be assessed from PM_2.5 concentrations indoors (C_i) and outdoors (C_o) with the following equation:

Barn et al. (2008) examined F_in with and without high-efficiency particulate air (HEPA) filtration in 21 British Columbia homes, with values during summer wildfires ranging from 0.19 with filtration to 0.61 without. Kirk et al. (2018) examined two homes in the Pacific Northwest during a high fire occurrence summer and observed good correlations between indoor (I) and outdoor (O) PM_2.5 concentrations but relatively low I/O ratios (0.10–0.26). F_in and I/O ratios during smoke events are influenced by road proximity, building envelope, occupant behaviors, and indoor PM_2.5 sources (Reisen et al., 2019; Shrestha et al., 2019).

Some in the air quality community have suggested that low-cost methods using Minimum Efficiency Reporting Value-13 (MERV-13) filters could reduce indoor smoke (https://pscleanair.gov/525/DIY-Air-Filter). It is not clear how well low-cost filtration methods will remove ultrafine aerosol particles in smoke (Laing et al., 2016) because the MERV-13 rating only requires a filter remove at least 50% of particles in the 0.3–1.0 µm size range (https://www.nafahq.org/understanding-merv-nafa-users-guide-to-ansi-ashrae-52-2/). However, a low-cost MERV-13 fan filter unit (FFU) achieved a > 75% reduction in residential and school PM₁ particle mass concentrations indoors relative to ambient smoke haze concentrations (Tham et al., 2018, 2020). The health/economic benefits of filtration interventions are likely to exceed purchase and operation costs, particularly if targeted to vulnerable populations (Fisk and Chan, 2017; Allen and Barn, 2020).

PurpleAir sensors (PAS) detect PM_2.5 using the Plantower PMS5003 laser scattering sensor. Several studies have found these to be precise, but not accurate (Ardon-Dryer et al., 2020; Li et al., 2020). For accurate measurements, calibration by a PM_2.5 reference is required. PAS are now widely used, with data available in real-time (https://www2.purpleair.com/). Recently, the US EPA developed a single correction equation for PAS and incorporated ambient outdoor data into its real-time AirNow smoke website (https://www.airnow.gov/fires/). In addition, increased availability of indoor PAS measurements presents the opportunity to significantly improve our understanding of indoor air quality (IAQ). For this study our goals are:

1. Evaluate the infiltration of smoke in different building types and the impact of interventions to improve IAQ using available data from the PurpleAir sensor network.
   
   
2. Evaluate a low-cost PM₅ filtration method that might be able to significantly improve IAQ during smoke events.

The results presented herein are highly relevant to millions of people impacted by wildland fire smoke each year and motivate further study on the impacts of smoke infiltration on IAQ and associated adverse health impacts. The dissemination of methods that the general public can easily adopt to reduce PM_2.5 exposure is of critical need in response to increasingly common wildland fires.

 
2 EXPERIMENTAL

First, we examine the relationship between indoor and outdoor PM_2.5 concentrations during a major smoke event. Hourly PM_2.5 (µg m^–3) measurements were obtained from the publicly available PAS network for 42 sites across the western US during a period of heavy wildfire smoke influence from September 5–15, 2020. The 42 sites used are a mix of 26 “residential”, 6 “school”, and 10 “commercial” buildings. Sites were categorized by a combination of publicly listed PAS name and location characteristics on Google Maps (maps.google.com). Locations were selected from existing PAS network voluntary participants and may not represent the general population. Building occupancy, indoor air quality interventions, opening of windows and doors and confirmation of building classification and location were provided by PAS users who responded to a request for information. Selected responding locations, which included five residential, one school, and two commercial buildings, served as case studies. In comparison to previous surveys of wildfire impacts on IAQ (Barn et al., 2008; Kirk et al., 2018; Reisen et al., 2019; Barkjohn et al., 2021), we examine a larger sample and more building types, but with less information on building characteristics and occupant behavior, and no examination of chemical composition of pollutants.

PAS data, designated by the user as outdoors or indoors, were downloaded as hourly averages from the PurpleAir website (www2.purpleair.com). Data were corrected with the EPA national correction equation (http:/fire.airnow.gov), which was developed by comparing sensor data across the US to measurements by federal reference and equivalent methods (Barkjohn et al., 2020):

The slope of the EPA national correction equation is within 3% of the slope of a smoke specific PAS correction equation based on multiple types of fires in the US (Holder et al., 2020) and within the range of PAS smoke adjustment factors (0.44–0.53), linear adjustments with zero intercepts, for wildfires in California and Utah (Delp and Singer, 2020). We focused on the strong wildfire episodes that occurred in the western US during the period of September 5–15, 2020. We identified smoke events and paired PAS sensors (one indoors and one outdoors) as follows:

A: Outdoor PAS corrected PM_2.5 values exceeded 80 µg m^–3 for 12 hours and smoke was identified using the NOAA HMS product.

B: PAS outdoor and indoor sensors were located within 0.5 km of each other.

Infiltration factor (F_in), and concentration of particles from indoor sources (C_S), as well as correlation coefficients (R²), were calculated from a linear regression of indoor and outdoor PAS PM_2.5 concentrations (Eq. (1)). The ratio of indoor to outdoor PM_2.5 (I/O) was calculated from respective means. Individual PAS locations are detailed in Table S1.

Next, we evaluate the utility of a low-cost filtration method. This method consisted of a 20” × 20” MERV-13 air filter, attached to a standard home box fan with tape. Photographs of the MERV-13 FFU, which can be assembled at home for ~30–50$, are shown in Fig. S1. Average air speed, measured with a VT140 Thermo-Anemometer directly in front of the fan outflow, was multiplied by filter area to estimate an air supply rate (Q_p) of 730 CFM, which was comparable to the FFU presented in Tham et al. (2018) (Q_p = 735 CFM). Attaching the MERV-13 filter closely to the box fan did not create excess heat over 8 hours of operation. However, airflow resistance by the filter contributed to a 33% reduction in airflow.

PAS measurements were obtained from two rooms in a single-family residence in Seattle, WA, which served as a case study for the evaluation of the low-cost filtration method presented here. Room A was 200 m³, with multiple windows and large openings to the interior and two exterior doors, and Room B was 50 m³, with only one window, one interior door, and one exterior door. First, identical MERV-13 FFUs simultaneously recirculated air in two separate rooms, both unoccupied with all doors and windows closed to reduce natural ventilation, in a Seattle, WA, home on September 13, 2020. We used a mass balance approach (Shi and Li, 2019) to estimate the infiltration rate (Q_inf) of each room from the clean air delivery rate (CADR) of the FFU and simultaneously measured post-filtration steady state PM_2.5 I/O ratios based on the following equation:

Penetration (P) was set to 1 and indoor deposition rate (K) was set to 0.39 h⁻¹ (Thatcher and Layton, 1995; Wallace, 1996). To further evaluate submicron particle removal efficiency by the MERV-13 FFU, a Grainger 23V750 handheld particle counter (Reiman et al., 2018) was used to measure size-resolved particle number concentrations alongside PAS measurements during an additional test of the FFU in Room B on September 14, 2020. A Grimm Environmental Dust Monitor (EDM-180) (Grimm and Eatough, 2009) was later used to calculate a correction equation for the Grainger number concentrations (Fig. S2). Details of size-resolved particle measurements are presented in the Supporting Information.

 
3 RESULTS and DISCUSSION

 
3.1 Indoor and Outdoor PM_2.5 Relationship in Buildings in the Western United States

Corrected (Eq. (2)) hourly average concentrations of indoor and outdoor PM_2.5 measured by PAS, at 42 western US locations were used to evaluate PM_2.5 infiltration by building type. Box plots of F_in, C_S, R², and I/O by building type are presented in Fig. 1. More complete information on these 42 buildings is given in Table S1. The large range of F_in (0.01–0.87) observed in residential buildings (N = 26) is consistent with previous observations of residential F_in (0.01–1.10) during wildfire smoke periods (Barn et al., 2008). The low median F_in (0.21) in residential buildings was more similar to median F_in of summer homes with HEPA filtration (0.19) than without (0.61) (Barn et al., 2008). Commercial buildings (N = 10) exhibited a higher median F_in (0.45), with a smaller range of 0.30–0.71, and school buildings (N = 6) had the highest median F_in value (0.68), with a wide range of 0.41–0.80. F_in thus reflects ASHRAE Standards 62.1 and 62.2 recommended outdoor air ventilation rates, which are lowest for residential buildings (ASHRAE, 2019a) and increase from commercial to school buildings (ASHRAE, 2019b). However, the conclusions of this survey are limited without coinciding records of occupant activity and measurements of air exchange rate (AER), which is controlled by leakage, mechanical and natural ventilation. For example, school and commercial F_in values in this study may have been further elevated by measures to increase outdoor air exchange in shared buildings during the COVID-19 pandemic (Schoen, 2020). Substantial reductions in occupancy in commercial and school buildings due to pandemic restrictions could also have influenced F_in.

Fig. 1. Box plots by building type of infiltration factor (F_in) and concentration of particles from indoor source (C_S), as well as indoor to outdoor PM_2.5 mass concentration correlation coefficients (R²) and ratios (I/O), for the 6 school, 10 commercial, and 26 residential PurpleAir sensor locations during a September 5–15, 2020, wildfire smoke–impacted period. Lower and upper box boundaries are 25^th and 75^th percentiles, respectively, lower and upper error lines are 10^th and 90^th percentiles, respectively, lines inside box are medians, red Xs are means, and circles are outlier data.

I/O ratios were below 1 and mirrored the building dependency of F_in values, with medians increasing from residential (0.33), to commercial (0.58), to schools (0.73). Median C_S were slightly higher in residential (6.29 µg m^–3) than commercial (5.38 µg m^–3) and school (1.55 µg m^–3) buildings. The percentage of PM_2.5 indoors due to infiltration can be described by the correlation coefficient (R²) of indoor and outdoor PM_2.5 (Bucur and Danet, 2019). Residential specific indoor sources (i.e., cooking, cleaning) (Wallace, 2006) may have contributed to lower residential R² (median = 0.77), while the elevated R² observed in high AER commercial (median = 0.90) and school (median = 0.93) buildings demonstrates that infiltration by ventilation was the predominant source of PM_2.5 therein. Increased filtration can also lower R² (Deng et al., 2015), as well as F_in and I/O (Barn et al., 2008; Allen et al., 2011; Wheeler et al., 2014; Ward et al., 2017). Therefore, the differences observed in wildfire smoke PM_2.5 infiltration (Fig. 1) must be examined in relation to both ventilation and filtration.

 
3.2 Impact of Indoor Air Quality Interventions: Residential Case Studies

Time series of indoor and outdoor PM_2.5, as well as linear regression plot of indoor and outdoor PM_2.5 used to calculate F_in and C_S, for three residences in the western US (WA, NV, ID) is presented in Fig. 2. The NV residence (F_in = 0.04) reported the continuous use of 3 HEPA filters and the WA residence (F_in = 0.09) reported the continuous use of 4 HEPA filters, with both reporting keeping all windows and doors closed to reduce natural ventilation. The ID residence reported using a furnace fan with a MERV-12, which ran 4 times per day for ½ hour, as well as a HEPA filter in the bedroom at night. The ID residence exhibited higher F_in (0.31), which was likely elevated by the combination of intermittency of filtration with a lower number of filter units and increased natural ventilation from reported opening of windows for multiple hours at night to cool the house. HVAC cooling lowers F_in in part by reducing AER from natural ventilation (Meng et al., 2009; Clark et al., 2010). IAQ of the ID and NV residences were impacted by contributions from cooking and/or cleaning (Wallace, 2006), as suggested by the elevated C_s and brief spikes in indoor PM_2.5 that are independent of and/or exceeded outdoor PM_2.5.

Fig. 2. Outdoor and indoor PM_2.5 mass concentrations measured by PurpleAir sensors (PAS) with EPA correction applied for three residential locations during a September 5–15, 2020 period of wildfire smoke. Location name and information on indoor air quality interventions is presented in the time series (left), with corresponding F_in (slope), C_S (y-intercept), and correlation coefficients (R²) presented in correlation plots (right).

An additional western US residential case study (OR), where occupants also reported closing windows and doors, is presented in Fig. S3. The addition of two HEPA filters and an upgraded HVAC filter during a 2019 wildfire period resulted in 50% lower infiltration (F_in = 0.18) compared to a 2018 wildfire period with standard HVAC filtration (F_in = 0.38). The WA, NV, and OR residences are consistent with prior observations of HEPA filters reducing indoor PM_2.5 from wood smoke in homes (Barn et al., 2008; Allen et al., 2011; Wheeler et al., 2014; Ward et al., 2017). One school and two commercial case studies, which all filtered air solely by HVAC and generally exhibited greater PM_2.5 infiltration, are presented in Fig. S4. Previously observed dependencies on filtration, ventilation, occupant activity, and room size (Du et al., 2011) potentially contributed to the wide range of PM_2.5 infiltration parameters observed in the residential, commercial, and school datasets (Fig. 1).

 
3.3 Evaluation of a Low-cost Indoor PM_2.5 Filtration Tool

The efficacy of the MERV-13 FFU in improving IAQ during a smoke event was evaluated with both PAS and submicron particle count information (Fig. 3). During the September 13, 2020 sampling period, the nearest outdoor PAS (2 km) to the Seattle, WA residence reported sustained elevated outdoor PM_2.5 (127 ± 9 µg m^–3). Prior to recirculating room air through the FFUs, the PAS measured average (± 1σ) PM_2.5 of 64 ± 2 µg m^–3 in Room A (I/O = 0.50) and 40 ± 2 µg m^–3 in Room B (I/O = 0.31). After 90 minutes of FFU recirculation in Room A, PM_2.5 reached a consistent 28 ± 2 µg m^–3 (I/O = 0.22), for a ~56% reduction. In comparison, it took less than 60 minutes of FFU recirculation in Room B to lower average indoor PM_2.5 to 0.4 ± 0.4 µg m^–3 (I/O = 0.003), a ~99% reduction. Differences in particle removal efficiency by a filter unit between rooms has previously been shown to be controlled by air exchange and position (Novoselac and Siegel, 2009). Outdoor PM_2.5 remained elevated (134 ± 5 µg m^–3) during an additional test of the FFU in Room B on September 14, 2020 (Fig. 3). 60 minutes of FFU recirculation reduced indoor PM_2.5 from 73 ± 1 µg m^–3 (I/O = 0.47) to 6.2 ± 0.6 µg m^–3 (I/O = 0.04), a ~91% reduction. Corrected particle number concentrations measured by the Grainger particle counter demonstrated the FFU filtered 0.3–1.0 µm particles at a comparable efficiency (> 90%) as PM_2.5. Substantial reductions in PM_2.5 and 0.3–1.0 µm particles after FFU recirculation further demonstrate the utility of low-cost methods for improving IAQ during smoke events (Tham et al., 2018, 2020).

Fig. 3. (Top) PM_2.5 mass concentrations measured on September 13, 2020, by PurpleAir sensors (PAS) before and after, as denoted by dashed line, recirculating indoor air through the MERV-13 fan filter unit in Room A (200 m³) and Room B (50 m³) of a Seattle, WA, home impacted by wildfire smoke. (Bottom) PM_2.5 mass concentrations measured on September 14, 2020, by a PAS, as well as 0.3 and 0.5 µm size bin corrected particle number concentrations measured by the Grainger particle counter, before and after, as denoted by dashed line, recirculating indoor air through the MERV-13 fan filter unit in Room B.

The CADR of the MERV-13 FFU was estimated as the product of average air flow (1130 m³ h^–1) and 0.3–1.0 µm MERV-13 filtration efficiency (50%) to be 560 m³ h^–1. The operating cost effectiveness, defined as the CADR per electric power consumption of the fan (maximum 100 watts), was estimated to be 5.6 CADR/watt. However, CADR and CADR/watt are likely overestimated here because this method does not account for imperfect mixing and recirculation reducing experimental filtration efficiency. The CADR of commercial HEPA filter units ranges from 200–700 m³ h^–1, with power consumption ranging from 30—200 watts and cost effectiveness ranging from 1–7.2 CADR/watt (Waring et al., 2008; Sultan et al., 2011; Noh and Yook, 2016; Shi and Li, 2019). The MERV-13 FFU described here is in the middle of the range of CADR and CADR/watt for commercial HEPA units, but at much lower initial cost. Future work is needed to determine the MERV-13 FFU effective air cleaning ratio (EACR), which is the ratio of the experimental CADR to the theoretical CADR (Noh and Oh, 2015), as well as the filtration efficiency of ultrafine particles (Laing et al., 2016) and gas-phase components of wildfire smoke (Kirk et al., 2018; Messier et al., 2019).

Calculated Q_inf from Eq. (3) (Shi and Li, 2019), which were higher for Room A (181 m³ h^–1) than Room B (1.76 m³ h^–1), were divided by room size to estimate AER. Residential AER is a predictor of PM_2.5 infiltration and exposure (Meng et al., 2009). The estimated AER of the larger and more externally connected Room A (0.91 h^–1) was a factor of 25 higher than the smaller and less externally connected Room B (0.035 h^–1), with both values falling within the range of prior residential observations (Yamamoto et al., 2010). The reduced performance of the FFU in the high AER Room A compared to the low AER Room B (Fig. 3) highlights the impact of ventilation on filter performance. The lower F_in observed in the prior WA, NV, and OR residential case studies (Figs. 2 and S3) may also be explained by low AER, as well as high CADR from multiple filter units. Buildings and rooms with greater AER may thus require multiple filter units to increase CADR and maximize potential reductions in PM_2.5 exposure during smoke events.

 
4 CONCLUSIONS

Indoor and outdoor PM_2.5 measurements by PAS were used to evaluate PM_2.5 infiltration in western US residential, commercial, and school buildings during a period of wildfire smoke. Infiltration of ambient smoke was the predominant source of indoor PM_2.5, and contributions from indoor sources were minor (Fig. 1). Building type median F_in, I/O, and R² calculated from PAS indoor and outdoor PM_2.5 concentrations corresponded to ASHRAE-recommended ventilation rates (ASHRAE, 2019b, 2019a), with infiltration reduced in low air exchange residential buildings and elevated in high air exchange commercial and school buildings.

The use of multiple filter units in residences, in addition to closing all windows and doors, was associated with substantially lower residential F_in (Figs. 2 and S3). The low-cost filtration method presented here, which was shown to effectively remove PM_2.5 (Fig. 3), provides a cost-effective alternative to commercial filter units for improving IAQ during wildfire smoke periods. One FFU was most effective in reducing I/O in a room with lower estimated air exchange. In a larger room or with larger air exchange, additional FFUs could be installed. Widespread dissemination of information and recommendations of filtration interventions, such as the low-cost method presented here, has the potential to substantially reduce indoor PM_2.5 exposure during the western US wildfire season.