Sources of indoor PM_2.5 gross α and β activities measured in 340 homes

Abstract

Particle radioactivity (PR) exposure has been linked to adverse health effects. PR refers to the presence of α- and β-emitting radioisotopes attached to fine particulate matter (PM_2.5). This study investigated sources contributing to indoor PM_2.5 gross α- and β-radioactivity levels. We measured activity from long-lived radon progeny radionuclides from archived PM_2.5 samples collected in 340 homes in Massachusetts during the period 2006–2010. We analyzed the data using linear mixed effects models and positive matrix factorization (PMF) analysis. Indoor PM_2.5 gross α-activity levels were correlated with sulfur (S), iron (Fe), bromine (Br), vanadium (V), sodium (Na), lead (Pb), potassium (K), calcium (Ca), silicon (Si), zinc (Zn), arsenic (As), titanium (Ti), radon (²²²Rn) and black carbon (BC) concentrations (p <0.05). Indoor PM_2.5 β-activity was correlated with S, As, antimony (Sb), Pb, Br and BC. We identified four indoor PM_2.5 sources: outdoor air pollution (62%), salt aerosol source (14%), fireworks and environmental tobacco smoke (7%) and indoor mixed dust (17%). Outdoor air pollution was the most significant contributor to indoor PM_2.5 α- and β-activity levels. The contributions of this source were during the summer months and when windows were open. Indoor mixed dust was also found to contribute to PM_2.5 α-activity. PM_2.5 α-activity was further associated with radon during winter months, showing radon’s important role as an indoor source of ionizing radiation.

Introduction

Annually, air pollution is accountable for 4.2 million deaths world-wide according to World Health Organization (WHO), with recent estimates indicating that this number may be substantially higher (Lelieveld et al., 2019). Of all air pollutants, fine particulate matter (PM_2.5) is the most critical environmental threat to public health, and is classified as group 1 carcinogen to humans by WHO International Agency for Research on Cancer (IARC, 2013). Epidemiological studies over the past decades suggest an association between ambient PM_2.5 and increased morbidity, mortality, and emergency hospital admissions for cardiovascular, cerebrovascular, and ischemic heart disease (Franklin et al., 2008; Zanobetti et al., 2009; Wellenius et al., 2012; Di et al., 2017).

Indoor PM_2.5 represents a key component of the overall risk associated with exposure to airborne pollution. Indoor PM_2.5 consists of out-door particles that infiltrate indoors and remain suspended, particles emitted indoors (primary), and particles formed indoors (secondary) through reactions of gas-phase precursors emitted both indoors and outdoors (Weschler and Carslaw, 2018). Many buildings rely entirely on mechanical ventilation to recirculate indoor air with a greatly reduced outdoor air dilution level, leading to the accumulation of pollutants indoors (Gonzalez-Martín et al., 2021). Typical indoor/outdoor (I/O) ratios for PM_2.5 measured in North America and Europe vary from 0.8 to 3.4 (Chen and Zhao, 2011), suggesting that indoor sources can contribute significantly to indoor air pollution. As buildings are progressively sealed against the outside environment to reduce heating and cooling energy costs, there is an increasing concern about exposures to indoor PM_2.5.

Particle radioactivity is another important characteristic of PM_2.5 and it is attributed to radionuclides that are present in the atmosphere due to anthropogenic, cosmogenic, or terrestrial sources (Papastefanou, 2010; Długosz-Lisiecka, 2016; Blomberg et al., 2020). Of these different sources, radon progeny are the most significant contributors to PM_2.5 gross α- and β-activities. Radon gas, is still highly prevalent in modern residential environments (Stanley et al., 2019) and depending on the geology, house structure and air exchange rate, it can present an important source of ionizing radiation indoors (Kolarz et al., 2009; Kang et al., 2020). Radon gas decays into radioactive progeny, which eventually attach to aerosol particles indoors forming radionuclide-attached particles. The radioactive formed aerosol particles gain the properties of the carrier aerosol particles (Reineking and Porstendorfer, 1990; Papastefanou, 2012), which upon inhalation, in vivo, continue to decay generating an internal dose of α, β, and γ radiation into the lungs, potentially causing cell damage and chronic inflammation (Nie et al., 2012). Chronic exposure to radiation also leads to variable dose-dependent effects on telomere length, neurodegenerative diseases, inflammation, impaired lung function, oxidative stress and elevated blood pressure (Scherthan et al., 2016; Nyhan et al., 2018, 2019; Santos et al., 2020).

Knowing that people spend on average 90% of their time indoors, 60–70% of which is inside houses, and given that about 70% of a per-son’s yearly exposure to radiation comes from inhalation (NCRP, 2006), it is essential to understand the sources affecting indoor PM_2.5 radio-activity. Here, we investigated the association of indoor PM_2.5 with particle radioactivity in 340 homes in Massachusetts, which were sampled as part of the Normative Aging Study (Bell et al., 1972). We investigated the relationship between PM_2.5 elemental composition and radioactivity using linear mixed effects models. In addition, we used positive matrix factorization (PMF) and multiple linear regression (MLR) analysis to investigate sources of PM_2.5 gross α- and β-activity. To the best of our knowledge, this is the first study that investigates the sources of indoor PM_2.5 gross α- and β-activities.

1. Materials and methods

1.1. Measurement of indoor PM_2.5 and its composition

Indoor PM_2.5 were collected during the period July 18, 2006 through December 8, 2010 inside the main activity room (typically the family room or living room) in 340 homes in Massachusetts as part of the Normative Aging Study (NAS). Indoor weekly PM_2.5 samples were collected on Teflon filters at a flow of 1.8 L/min using the Harvard PEM (Demokritou et al., 2001). About 88% of the houses were only sampled once while the remaining 12% of houses had 2 – 3 samples during different seasons (summer, winter or transient). Outdoor PM_2.5 daily samples were collected on Teflon filters at a flow of 16.7 L/min using the Harvard Impactor (Koutrakis et al., 1993) at the Harvard Boston Supersite. Indoor and outdoor samples were compared by matching the weekly indoor samples to the corresponding outdoor samples. The filters were pre- and post-weighed using an electronic microbalance (Mettler MT-5, Columbus, OH) at controlled temperature and relative humidity room to determine PM_2.5 mass concentration. For elemental composition, Energy Dispersive X-ray Fluorescence (EDXRF) spectrometery (Epsilon 5, PANalytical, Netherlands) was used, following EPA method IO-3.3 (Kang et al., 2014). Black carbon (BC) concentrations were determined by using the reflectance method (Janssen et al., 2001), where the blackness of the particles collected on the filter was measured with a Smokestain reflectometer (model EEL M43D, Diffusion System Ltd., United Kingdom).

1.2. Particle radioactivity and radon measurements

Particle gross α- and β-activities from long-lived ²²²Rn progeny were counted on PM_2.5 filters after storage for 10–14 years. During sampling, several Rn progeny attached to PM were collected on the filters. After 3 years of storage, a transient equilibrium between ²¹⁰Pb and ²¹⁰Po is reached and the total ⍺ activity in the filter can be estimated by ²¹⁰Po decay, which is the progeny of ²¹⁰Pb (longest half-life = 22.3 years). Therefore, ⍺ activity in the air at the time of sampling can be estimated by C = A_α(t)e^λt/Q,where A_α(t) is the ⍺ activity measurement (mBq/m³), λ is the decay constant (8.51 × 10⁻⁵ d⁻¹) for ²¹⁰Pb, Q is the air sample volume (m³) and t is the duration in days between sampling and measurement (Sheets and Thompson, 1994). The measurements were made using a low background gas proportional counter (Model LB4200, Canberra Industries, Inc., Meriden, CT) with a P10 carrier gas (10% methane balanced with argon). The counter was calibrated every two weeks with a 0.0518 μCi NIST traceable ²¹⁰Po α source and a 0.00516 μCi NIST traceable ⁹⁰Sr β source on 5.7 cm planchets. The background level was below 0.1 counts per minute and for a counting time of 600 min in a sample of 14.4 m³, the detection limit for ⍺ and β activity was 0.219 and 0.355 mBq/m³, respectively. The long counting time was applied due to the low concentrations of α- and β-emitters collected on the Teflon filters. Details about the procedure can be found elsewhere (Liu et al., 2020; Kang et al., 2020).

Radon measurements were obtained from Spruce Environmental Technologies, Inc., which measured radon levels in 89,076 residential buildings in New England (not in NAS homes) during our study period. We selected radon measurements detected by activated charcoal adsorption devices in the lowest livable level of each building under closed-house conditions. In total, 76,767 radon measurements from 58,214 buildings were used to calculate 1173 monthly ZIP Code radon levels. Valid estimates were considered when over five residential measurements were available for a ZIP Code during a month. For each NAS residence, we assigned a radon value based on its ZIP Code.

2.3. Statistical analysis

A linear mixed-effect regression model was used to examine relationships between particle α- and β-activities and house characteristics (i.e. windows opening, AC use), indoor PM_2.5 mass and composition during sampling. Window opening and AC use were used as categorical variables. The model is useful to analyze data including repeated measurement of a subject, such as multiple measurements for each home. The model used a random intercept for each home, while the covariate was included as a fixed effect. In Eq (1) Y_ij is the log of the PM_2.5 α- or β-activity of sample i in house j, X_ij is the average species concentration or windows/AC use during sampling period i in house j, H_j is the random intercept of house j and ε_ij is the random error.

$$Y_{\mathit{ij}}\mathit{=}{\beta }_{\mathit{0}}\mathit{+}\beta X_{ij}\mathit{+}{H}_j\mathit{+}{\epsilon }_{ij}$$ (1)

Positive matrix factorization (PMF) in conjunction with multiple linear regression (MLR) were used to identify PM_2.5 sources and examine the associations between their contribution and indoor PM_2.5 gross α-and β-activities. BC measurements were also included in PMF to aid towards the characterization of sources. PMF is a receptor model that uses a multivariate factor analysis technique, which is based on weighted least square fits. The uncertainty input data matrix followed the approaches described by Norris et al. (2014) and Polissar et al. (1998). PMF uses realistic error estimates to weight data values by enforcing non-negative constraints in the factor computational process and is widely used to identify and quantify the main sources of atmo-spheric pollutants (Belis et al., 2014; Crilley et al., 2017; Bari et al., 2015; Bourtsoukidis et al., 2020). Here, the United States Environmental Protection Agency (USEPA) PMF 5.0 model was used (Norris et al., 2014).

PMF is a descriptive model and there are no objective standards for choosing the right number of factors (Belis et al., 2013). However, in order to acquire realistic source profiles and an optimum number of factors, a multi-criterion was applied which included investigation of: 1) signal to noise ratio, 2) symmetric distribution of scaled residuals (±3σ), 3) the loss function, and 4) the interrelationship between the predicted and observed concentrations. Details of in-depth investigation of PMF optimal solution can be found in supplementary material. To evaluate the reproducibility of the PMF solution and the adequate number of PMF factors, a bootstrap technique was applied, where random blocks of observations from the original data were sampled until reaching the size of the original input data. The bootstrap model method executed 100 iterations by using a random start and a minimum Pearson correlation coefficient (R-value) of 0.6 (Bressi et al., 2014; Baudic et al., 2016; Bourtsoukidis et al., 2020). All the bootstrap modelled factors were well reproduced over at least 90% of runs, indicating that the model un-certainties can be interpreted and that the number of factors is appropriate. The R² between the PM_2.5 concentrations predicted using the PMF model and those measured PM_2.5 measured was 0.92, indicating good agreement (Figure S1). To investigate the significance of sources, we used MLR analysis on PM_2.5 gross α- or β-activities and PM_2.5 source profiles produced by PMF.

2. Results and discussion

2.1. Characteristics of PM_2.5 samples

Table 1 shows the summary statistics. Sulfur (S) was the most abundant element collected on the indoor PM_2.5 filters. Second in abundance was Sodium (Na). Indoor PM_2.5 samples were also relatively rich in Chloride (Cl), Potassium (K), Silicon (Si), Calcium (Ca) and Iron (Fe). The observed PM_2.5 mass concentrations are comparable to those from previous studies in North American and European houses (i.e. Chen and Zhao, 2011; Bari et al., 2015; Habre et al., 2014). Mean PM_2.5 α-activity levels were similar to those reported for archived filters collected by indoor and outdoor studies (Liu et al., 2020; Kang et al., 2020), while mean β-activity levels were greater than those reported in Boston (Liu et al., 2020).

Table 1.

Descriptive statistics of PM_2.5 α- and β-activities, BC and PM_2.5 mass and elemental composition; N.D: Not detected.

	Mean	SD	Min	Max	Median
Units: mBq/m 3
PM2.5 α-activity	1.25	1.11	N.D	19.98	1.03
PM2.5 β-activity Units: μg/m3	0.82	0.56	N.D	9.42	0.63
PM2.5	7.64	4.45	0.22	37.89	6.47
BC	0.62	0.34	0.15	3.38	0.57
Units: ng/m 3
Na	108.32	191.83	N.D	2358.50	70.08
Mg	8.61	23.26	N.D	390.11	3.20
Al	12.40	13.34	N.D	126.50	9.73
Si	29.70	53.78	N.D	725.09	20.04
S	447.76	320.01	N.D	2288.25	349.17
Cl	63.44	328.53	N.D	3924.53	11.33
K	39.66	42.69	N.D	378.99	26.52
Ca	25.56	44.76	N.D	661.02	18.44
Ti	2.72	9.23	N.D	151.10	1.70
V	0.57	0.62	N.D	4.32	0.44
Cr	0.61	1.94	N.D	33.30	0.33
Mn	0.73	0.84	N.D	6.31	0.48
Fe	20.03	16.75	0.06	225.50	16.39
Ni	0.36	0.40	N.D	3.52	0.27
Cu	2.56	3.83	N.D	42.71	1.48
Zn	7.39	34.73	N.D	650.00	4.35
As	0.19	0.88	N.D	12.54	N.D
Se	0.36	0.64	N.D	4.15	N.D
Br	1.08	1.57	N.D	19.50	0.84
Sr	0.35	0.73	N.D	7.81	0.06
Sn	0.31	0.56	N.D	3.97	N.D
Sb	1.61	2.36	N.D	17.06	0.76
Ba	0.64	1.64	N.D	13.02	N.D
Pb	1.64	5.29	N.D	98.20	1.07

2.2. Elements contributing to PM_2.5 gross α- and β-activity

Using linear mixed effects models, PM_2.5 α- and β-activities were regressed on PM_2.5 elemental composition, BC, radon and house ventilation characteristics. The results of this analysis are summarized in Table 2. The distributions of PM_2.5 α- and β-activities were skewed to the right and were log-transformed. Indoor particle PM_2.5 α-activity was strongly associated (p <0.01) with indoor concentrations of PM_2.5, Sulfur (S), Iron (Fe), Bromine (Br), Vanadium (V), Sodium (Na), Lead (Pb), Potassium (K), Calcium (Ca), Silicon (Si), and Zinc (Zn) elements and significantly (p <0.05) with Radon (Rn), black carbon (BC), Arsenic (As) and Titanium (Ti). Indoor PM_2.5 β-activity was strongly associated (p <0.01) with indoor PM_2.5 and S concentrations, and significantly (p<0.05) associated with As, BC, Br, Sb and Pb concentrations. PM_2.5 α- and β-activities are strongly associated with window opening and with elements which predominantly originate from outdoors and penetrate indoors. Similarly, Kang et al. (2020) found significant associations between S (a tracer of outdoor particles) and PM_2.5 α-activity in 26 Boston homes.

Table 2.

Regression analysis of indoor PM_2.5 α- and β-activity concentrations (N = 326).

	α-activity			β-activity
Dependent Variables	Coefficient	Standard error	p-value	Coefficient	Standard error	p-value
PM2.5	0.328	0.073	<0.001	0.202	0.066	<0.001
BC	0.325	0.038	<0.001	0.398	0.175	0.023
Na	0.053	0.021	0.013	0.016	0.011	0.127
Mg	0.044	0.024	0.086	0.009	0.013	0.507
Al	0.026	0.013	0.381	0.002	0.015	0.910
Si	0.091	0.016	0.004	0.019	0.018	0.300
S	0.133	0.017	<0.001	0.206	0.042	<0.001
Cl	−0.072	0.023	0.002	−0.091	0.012	0.003
K	0.109	0.046	0.018	0.037	0.023	0.107
Ca	0.118	0.051	0.023	0.033	0.026	0.205
Ti	0.120	0.059	0.046	0.007	0.032	0.830
V	0.360	0.098	0.001	0.063	0.052	0.233
Cr	0.036	0.042	0.401	0.043	0.049	0.385
Mn	0.044	0.038	0.255	−0.056	0.041	0.177
Fe	0.203	0.056	<0.001	0.037	0.028	0.187
Ni	0.303	0.135	0.023	0.061	0.071	0.390
Cu	0.011	0.025	0.675	0.001	0.027	0.981
Zn	0.116	0.059	0.049	0.020	0.031	0.511
As	0.301	0.121	0.012	0.306	0.106	0.004
Se	−0.010	0.046	0.833	0.035	0.049	0.479
Br	0.325	0.078	<0.001	0.161	0.069	0.028
Sr	0.028	0.049	0.571	0.025	0.052	0.626
Sn	−0.041	0.048	0.402	−0.025	0.051	0.621
Sb	0.022	0.022	0.326	0.048	0.024	0.046
Ba	−0.007	0.029	0.806	0.015	0.031	0.626
Pb	0.162	0.059	0.007	0.109	0.052	0.041
Rn	0.096	0.046	0.021	0.020	0.096	0.750
windows open	0.190	0.077	<0.001	0.213	0.067	0.009
AC	0.021	0.022	0.352	0.006	0.021	0.767

2.3. Sources of indoor PM_2.5

Fig. 1 shows the source factors resolved by the PMF model, in which the four identified sources are indoor mixed dust, salt aerosol, fireworks and environmental tobacco smoke and outdoor air pollution. Figure S2 shows the time series plots of PMF derived source contributions.

Fig. 1.

Source profiles (species percentages and concentrations) contributing to indoor PM_2.5 mass.

Indoor mixed dust.

This factor is associated with Si, Al, Ca, Mg, and Fe, which reflects a mixture of mineral indoor and outdoor re-suspended dust including natural and anthropogenic dust sources. This factor is also the highest contributor to BC (70%). BC inside houses originates from indoor cooking (Long et al., 2000; Barraza et al., 2014), and out-door vehicle tailpipe emissions (Balasubramanian and Lee, 2007). Ti, Cr, Mn, Ni and Zn, which are tracers of suspended road dust and tire wear (Pant and Harrison, 2013) have percentages greater than 60% in this factor, suggesting their high infiltration rates (Bari et al., 2015; Huang et al., 2018). In addition to outdoor sources, Mn can be emitted during vacuuming and use of cleaning products (Habre et al., 2014). Cu (54%) can have both outdoor (i.e. vehicle brake tracer) and indoor sources (i.e. vacuum cleaners, blenders and hair dryers that use copper commutators for motor rotation, Szymczak et al., 2007). The high percentage of Sb (67%) is most likely due to the mix of outdoor road dust and resuspension of dust embedded in carpeting (Majestic et al., 2012). Indoor mixed dust contributes 17% to total indoor PM_2.5 mass.

Salt aerosol.

This source is distinguished by the high Cl and relatively high Na, and has a clear winter origin (Figure S2). The contribution of this source throughout the year is consistently low, aside from a few spikes. The winter spikes of Cl are likely related to road de-icing, while the relatively lower peaks during summer months are related to sea salt particles. The indoor ratio of Na/Cl was 0.52, which is representative for outdoor salt particles indoors (Crilley et al., 2017). Br in this factor is due to outdoor sea spray aerosols (Belis et al., 2013). This source contributes 14% to indoor PM_2.5 mass.

Fireworks and environmental tobacco smoke.

This source is determined by the high contribution of K and relatively high Sr and Ba. This factor showed consistent smaller peaks throughout the sampling period which are likely related to indoor smoking, as 72% of the subjects enrolled in NAS study were former or current smokers (Nyhan et al., 2018). Habre et al. (2014) found that 1 μg/m³ of nicotine increases indoor K levels by 35 ng/m³. The summer peaks (Figure S2) in this factor suggest regular use of barbeque and meat grilling (Schauer et al., 1999). Winter peaks in this factor occurred during the first week of July 2007, 2008 and 2010 (Figure S2) and during the first week of January 2007. These peaks are likely related to outdoor New Year’s and Independence Day fireworks concentrations penetrating indoors, where K, Sr and Ba salts are added in fireworks to create red-yellow colors (Widory et al., 2010; Seidel and Birnbaum, 2015). This source accounted for 7% of total indoor PM_2.5 mass during the sampling period.

Outdoor air pollution.

The resolved factors with largest contributions of S have been previously identified as due to regional outdoor pollution (Masri et al., 2015). This factor also includes secondary inorganic aerosols due to the relatively elevated values of S (Crilley et al., 2017). High levels of V were also evident in this factor, suggesting an outdoor source of fuel/oil combustion possibly from vehicles or other industrial activities that penetrate indoors. Outdoor air pollution source is the largest contributor to indoor PM_2.5, and accounts for 62% of their mass on average.

Fig. 2 shows the percentages of source contributions to indoor PM_2.5 mass. Outdoor air pollution accounts for 68% of indoor PM_2.5, while 17% of PM_2.5 mass is attributed to indoor mixed dust. This is in agreement with studies made in houses where approximately 62% of indoor PM_2.5 exposure was caused by the transport of outdoor PM_2.5 to the indoor environment via ventilation and 16% by indoor sources (Asikainen et al., 2016).

Fig. 2.

Percentages of source contributions to indoor PM2.5 resolved by PMF mo

3.4. Sources affecting PM_2.5 α- and β-activities

Multiple linear regression analysis was used to investigate the relationship of the four previously identified sources and PM_2.5 α- and β-activity levels. The coefficient estimates of the multiple linear regression were used to build prediction models. Table 3 shows the results of the regression analyses for indoor PM_2.5 α-activity. Outdoor air pollution and indoor mixed dust were most significantly related to PM_2.5 α-activity (p <0.01). Equation (2) is used to predict indoor PM_2.5 α-activity. The sources names represent their contributions to indoor PM_2.5 mass. Note that fireworks and environmental tobacco smoke, and salt aerosol sources were not included in the prediction model because they do not contribute to the overall PM_2.5 α-activity. To estimate the per-centage of α-activity due to outdoor air pollution and indoor mixed dust sources we multiplied their PMF slopes (Eq (2); Table 3) with the sulfur and radon mean values respectively and divide them with the mean PM_2.5 α-activity (Table 1). The result showed that outdoor air pollution and indoor mixed dust sources contribute 54.5% and 33.9% to the α-activity in the filter.

Table 3.

Multiple linear regression analysis results and source contributions to PM_2.5 α-activity levels. Estimates represent the slope of the total activity that is attributed to the source.

	Estimate	Standard Error	p-value
(Intercept)	0.488	0.048	<0.001
Indoor mixed dust	0.098	0.034	0.005
Salt aerosol	−0.001	0.001	0.784
Fireworks and environmental tobacco smoke	−0.019	0.012	0.114
Outdoor air pollution	0.152	0.022	<0.001

$$Indoor{\log }_{\mathit{10}}\mathit{(}P{M}_{\mathit{2.5}}\alpha \mathit{‐}activity\mathit{)}\mathit{=}0.488+0.152\mathit{\times }Outdoorairpollution\mathit{+}0.098\mathit{\times }Indoormixeddust$$

Table 4 shows the results of the multiple linear regression analyses for indoor PM_2.5 β-activity, which was positively associated with indoor mixed dust and outdoor air pollution. However only outdoor air pollution was significantly associated with PM_2.5 β-activity (p <0.001); therefore, indoor mixed dust, fireworks and environmental tobacco smoke and salt aerosol sources were excluded from the predictive model. Equation (3) was used to predict indoor PM_2.5 β-activity. Out-door air pollution was the only predictor of PM_2.5 β-activity and con-tributes 42.2% to the β-activity in the filter.

Table 4.

Multiple linear regression analysis results and source contributions to PM_2.5 β-activity levels. Estimates represent the slope of the total activity that is attributed to the source.

	Estimate	Standard Error	p-value
(Intercept)	0.413	0.057	<0.001
Indoor mixed dust	0.011	0.041	0.788
Salt aerosol	−0.001	0.003	0.734
Fireworks and environmental tobacco smoke	−0.011	0.013	0.413
Outdoor air pollution	0.084	0.025	<0.001

$$Indoor{\log }_{10}\mathit{(}P{M}_{2.5}\beta \mathit{‐}activity\mathit{)}\mathit{=}0.413+0.084\times Outdoorairpollution$$ (3)

Ambient air radionuclides exist in two forms: as “unattached clusters” with a diffusion equivalent diameter size ranging from 0.5 to 5 nm, and as “aerosol-attached clusters” with particle diameters varying be-tween 5 and 3000 nm (Grundel and Porstendorfer, 2004; Papastefanou, 2010). ²²²Rn is the source of long-lived (²¹⁰Pb, ²¹⁰Bi, ²¹⁰Po) radionuclides which are the most important contributors of ambient particle gross α- and β-activities in the absence of anthropogenic radionuclides. The behavior of the attached airborne radionuclides is determined by that of their carrier aerosol particles, thus ambient aerosol-attached clusters can be transported, mixed and deposited indoors. Kaneyasu et al. (2012) revealed that the sulfate aerosols were the transport media of radioactive cesium in the atmosphere, which supports our finding that the outdoor air pollution is the most important contributor to PM_2.5 α-and β-activities indoors associated with long-lived radionuclides.

Indoor dust includes particles generated both in the indoor and outdoor environments (see section 3.3). PMF results showed that indoor dust is the greatest contributor to BC (70%), while the remaining BC is attributed to outdoor air pollution (20%) and fireworks and environmental tobacco smoke (10%) (Fig. 1). BC originates from sources that are largely anthropogenic and related to combustion processes, motor vehicle emissions, fossil fuel burning, industrial processes (i.e. ore/ metals smelting) and biomass/wood burning. All these carbonaceous aerosol sources have a contemporary ¹⁴C signature (Dusek et al., 2017) and depending on their origin and the air exchange rates of the household, may contribute to indoor particle radioactivity. The significant association of PM_2.5 α- and β-activities with BC (Table 2) further supports that there might be direct ¹⁴ C contributions to PM_2.5 radioactivities.

Radon gas generated in soil may escape through openings in the floor or wall and can accumulate in the air of houses. ²¹⁸Po (generated by radon decay) indoors reacts in few seconds with water molecules and trace gases forming unattached progeny with diameter ranging 0.5–1 nm (Reineking and Porstendorfer, 1990). This unattached progeny may either attach onto aerosol particles producing radioactive aerosol or, alternatively, plate out on walls and furniture. Indoors, these two processes are in competition and take place on a time scale of few minutes. The rate of these interactions depends on several parameters such as aerosol concentration and surface area, room dimensions, nature of the furniture surface and air exchange rate. Long-lived Rn progeny attached to PM is directly related to short-lived progeny (Dlugosz-Lisiecka, 2016), since the consecutive decays of short-lived ²¹⁸Po, ²¹⁴Pb, ²¹⁴Bi and ²¹⁴Po produce long-lived ²¹⁰Pb radionuclides, whose PM_2.5 gross α- and β-activities were detected in our analysis conducted many years after sampling collection. In an experimental study conducted in a temperature-controlled chamber, Trassierra et al. (2016) examined the synergetic effect of radon progeny dynamics and ultrafine particles from indoor sources and suggested that the presence of a significant particle surface area concentration can reduce the unattached fraction of radon progeny. However, the authors also pointed out that this process can simultaneously strongly increase the proportion of radionuclides deposited on the particles, thus the inhalable dose carried into the lungs indoors. These studies support our finding that indoor mixed dust can be a source of radionuclides indoors, where radon progeny have an important role.

1.1. Temporal variability of indoor PM_2.5 α-activity

During the winter homes normally are kept closed and indoor radon levels, likely penetrating the houses through cracks in the walls and/or the underground gaps between heating pipes (Lacompte et al., 2014), increase, while infiltration of outdoor particles decrease. During the summer, home ventilation increases and indoor levels go down, whilst more outdoor particles infiltrate indoors. To show these two conditions we examined the relationship of PM_2.5 α-activity and radon in the winter and the relationship of PM_2.5 α-activity and indoor to outdoor I/O sulfur ratio in the summer. The I/O sulfur ratio has been used as a surrogate of particle infiltration of PM_2.5 indoors by previous studies (Sarnat et al., 2002; Kang et al., 2010; Habre et al., 2014). Fig. 3 shows the dependence of (a) indoor PM_2.5 α-activity and radon for November, December, January and February, and (b) indoor gross α-activity versus I/O sulfur ratio for June, July August. A significantly (p <0.05) positive association is found between indoor particle α-activity and radon during winter months. Important positive association (p <0.05) was also found be-tween particle α-activity and I/O sulfur ratio during summer months.

Fig. 3.

Relationship between (top) indoor α-radioactivity and radon during cold period (Nov, Dec, Jan, Feb), (bottom) indoor α-radioactivity and I/O sulfur ratio during warm period (June, July, Aug).

The relationship between PM_2.5 α-activity and radon was negative, but not significant during summer months (Figure S3). In agreement with our result, studies have found that radon-related air ions were higher inside houses during the heating season than in the non-heating season (Kolarz et al., 2009; Kang et al., 2020). These summer and winter re-lationships also verify the PMF results, where outdoor air pollution and indoor mixed dust sources (combined accounted for 79% of indoor PM_2.5 mass) were significantly associated with PM_2.5 α- and β-activities.

2. Conclusions

The findings of this study provide key insights about indoor PM_2.5 toxicity and also add to a very limited base of information in the scientific literature on sources of PM_2.5 radioactivity in indoor environments. Indoor PM_2.5 radioactivity was investigated using archived PM_2.5 samples collected in 340 homes during 2006–2010 in Massachusetts. Mixed effects linear models showed significant association of S, Fe, Br, V, Na, Pb, K, Ca, Si, Zn, BC and Rn with PM_2.5 α- and β-activities. The analysis of sources of indoor PM_2.5 showed that outdoor air pollution, which contributed 62% of the PM_2.5 mass indoors, was the most significant (p <0.01) source to both PM_2.5 α- and β-activities. Fine mixed dust of indoor origin, which is a source responsible for 17% of the indoor PM_2.5 mass, was also significantly associated (p <0.05) with indoor PM_2.5 α-activity. Overall, outdoor air pollution source accounted for 54.5% and 42.2% of PM_2.5 α- and β-activities respectively, while indoor mixed dust source accounted for 33.9% of PM_2.5 α-activity. A strong positive relationship between PM_2.5 α-activity and I/O sulfur ratio was found during the summer months indicating higher transport of outdoor radionuclides indoors. During winter, PM_2.5 gross α-activity was strongly linked with radon suggesting that its progeny is an important source of indoor particle radioactivity.

A number of limitations exist in our study. PM_2.5 α- and β-activities were estimated from archived filters and short-lived PM_2.5 activities could not be assessed. In characterizing indoor PM_2.5 α- and β-activities and their sources, sampling of homes were clustered in space and time, resulting in an inability to assess the spatial variability of PM_2.5 concentrations and PM_2.5 α- and β-activities in the greater Boston area. Organic carbon and secondary pollutants (sulfate, nitrate, and ammonium) were not measured in PM_2.5 mass. We acknowledge that all source factors identified by the PMF model in this study are the first and more work is needed to further resolve source contributions to PM_2.5 radio-activity in order to enhance the understanding about the impact of in-door PM_2.5 toxicity linked to radioactive products.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.envres.2021.111114.