Characteristics and Relationships between Indoor and Outdoor PM 2 . 5 in Beijing : A Residential Apartment Case Study

In order to understand the characteristics and relationships between indoor and outdoor PM2.5 during the heating period of 2014 in Beijing, the investigation of PM2.5 and associated species including organic and elemental carbon (OC/EC), water soluble ions, metal elements and trace organic matter (OM) were undertaken at a residential apartment. The average PM2.5 concentration was 55.2 ± 47.3 μg m for indoor and 100.4 ± 82.1 μg m for outdoor, and the indoor PM2.5 was found to be mainly from outdoors. OM and (NH4)2SO4 were the dominated components of PM2.5, accounted for 71.5% in indoor PM2.5 and 52.4% in outdoor PM2.5, followed by fine soil and NH4NO3 (23.7% and 27.9%). The polycyclic aromatic hydrocarbons (PAHs) concentration was 187.3 ng m and 387.0 ng m, and the phthalic acid esters (PAEs) concentration reached 1054.2 ng m and 515.3 ng m, for indoor and outdoor, respectively. Hexachlorobenzene (HCB) only existed indoors (5.5 ng m). HCB and most PAEs in indoor PM2.5 were dominated by indoor sources whereas other species were greatly influenced by outdoor sources especially during the pollution period.


INTRODUCTION
Fine particle (PM 2.5 ) has attracted worldwide attention in the past few decades.Previous epidemiological and toxicological studies has linked PM 2.5 to cardiovascularand respiratory-based disease and premature mortality on a global scale (Pope III and Dockery, 2006;Russell and Bert, 2009;Apte et al., 2015;Lelieveld et al., 2015).Beijing, the capital of China, is one of the cities experiencing serious atmospheric PM 2.5 pollution.In 2014, the annual average concentration of PM 2.5 in Beijing reached 85.9 µg m -3 , 2.5 times of the China National Ambient Air Quality Standard (35 µg m -3 ).At present, the serious atmospheric PM 2.5 pollution has become one of the factors influencing population health and restricting sustainable economic development, social harmony, and stability for Chinese society.
In order to control the serious PM 2.5 pollution, Chinese government has implemented a variety of polices and measures (Zhang et al., 2012).The characteristics, sources and control measures of PM 2.5 pollution have also been intensively researched in recent years (Zheng et al., 2005;Yang et al., 2011;Wang et al., 2013a;Zhang et al., 2013;Zhao et al., 2013;Huang et al., 2014).In contrary to the extensive attention towards outdoor PM 2.5 pollution, only a little attention has been paid to PM 2.5 pollution in indoor air.Considering the general facts that there exists a significant difference in air quality between indoor and outdoor, and people spend the majority of their time indoors (Jahn et al., 2013), monitoring and improving the indoor air quality is even more important.
In general, indoor PM 2.5 is influenced by direct emissions from indoor sources (e.g., cooking, smoking, and candle burning) (Jones et al., 2000;Dacunto et al., 2013), ventilation supplies from outdoor sources (Zhao et al., 2015), and transport or transformation processes within indoor environments (e.g., mixing, inter-zonal transport, re-suspension, coagulation, and phase change) (Nazaroff, 2004).The indoor/outdoor relationship studies of PM species showed that major chemical components, including organic and elemental carbon (OC/EC), water soluble ions, and metal elements in indoor PM 2.5 were greatly influenced by outdoor sources (Lim et al., 2011;Cao et al., 2012;Hassanvand et al., 2014).For trace organic matter (OM), however, polycyclic aromatic hydrocarbons (PAHs) were affected by both outdoor sources and heating processes inside apartments (Romagnoli et al., 2014), and phthalic acid esters (PAEs) were dominated by indoor sources (Zhang et al., 2014).Although the pollution of PM 2.5 in indoor environment has obtained increasing concern in recent years, few studies have been conducted to investigate the characteristics and relationship between indoor and outdoor PM 2.5 during different pollution periods.Besides, the simultaneous analysis of different kinds of species in PM 2.5 , including OC/EC, water soluble ions, metal elements and trace OM is scarce.
In the present study, PM 2.5 mass concentrations were continuously monitored inside and outside an apartment in Beijing's urban area to understand the temporal variations and indoor/outdoor relationships during the different pollution periods.On the other hand, the contents of OC/EC, water soluble ions, metal elements and trace OM including 16 kinds of PAHs, 6 kinds of PAEs and hexachlorobenzene (HCB) in PM 2.5 were investigated based on filter-based sampling and subsequent analysis, and the indoor/outdoor relationships of these species were established.

Site and Sampling
Samples and data were collected on the 9th floor of a residential apartment building in Haidian district, Beijing, approximately 50 meters from a two-way road.The sampling site is located in the cultural and educational zone, and there is not any big industrial air pollution source around.The apartment's three occupants are non-smokers and cook two times daily, which represents typical working families in urban Beijing.The indoor sampling site was in the living room (approximately 1.5 m above the floor), and the outdoor sampling site was located outside the north bedroom window.During the sampling period, the apartment windows were only opened on clean air days for approximately half an hour a day.
Real-time mass concentrations of indoor and outdoor PM 2.5 were continuously monitored from January 30 to February 25, 2015 using SDM 805 laser particle monitors (Jinan Nuofang electronic technology Ltd, China) with 5 minute counting intervals.A parallel test between the laser particle monitor and Tapered Element Oscillating Microbalance (TEMO) (Thermal Fisher) was conducted before monitoring.The regression equation obtained from the test was y = 1.02x + 1.33 (R 2 = 0.999).In this expression, x and y is the realtime 1-h PM 2.5 mass concentration based on TEMO and laser particle monitor, respectively.
Filter-based PM 2.5 samples at each sampling site were collected by using two parallel Mini-volume samplers (Airmetrics, USA) on Whatman 47 mm quartz and Teflon filters from December 15, 2014 to March 7, 2015.Each sampling cycle lasted 48 h, and the sampling flow rate was 5 L min -1 .Before sampling, the quartz filters were pre-heated at 500°C for 4 h to remove carbonaceous contaminants.After 48-hour sampling period, the loaded filters were stored in a refrigerator below 0°C before the gravimetric and chemical analysis.
Meteorological data including wind speed (WS) and wind direction (WD) were obtained from China Meteorological Data Sharing Service System.The real-time PM 2.5 mass concentration monitoring data was communicated from the sampling devices to the computer over wireless.

Gravimetric and Chemical Analysis
Before and after sampling, Teflon filters were equilibrated in a controlled chamber (T = 20 ± 1°C, RH = 50 ± 5%) for 24 h, and then weighed by an electronic balance with a detection limit of 2 µg (CPA-26P, Sartorius, German).
Teflon filters were analyzed for both water soluble ions and metal elements.Half of each filter was extracted once in 10 mL Milli-Q water using an ultrasonic bath for 30 min, and then filtered through a 0.45 µm polytetrafluoroethylene (PTFE) syringe filter.Part of the filtrate was analyzed for anions of SO 4 2-, NO 3 -and Cl -using ion chromatography (IC) (792 Basic IC, Metrohm, Switzerland), and part was determined for NH 4 + according to Nessler's reagent colorimetric method formulated by Ministry of Environmental Protection of People's Republic of China (2009).The rest of Teflon filter was firstly digested in a Teflon vessel using a mixture of HNO 3 , HClO 4 , and HF at 170°C for 4.5 h.After the mixed acids in the extract were evaporated, the extract was diluted to 10 mL using Milli-Q water.The 13 metal elements (Na, Mg, Al, Ca, Ti, Mn, Fe, Ni, Cu, Zn, Sr, Mo, and Pb) were analyzed using inductively coupled plasma mass spectrometry (ICP-MS) (Agilent 7500a, Agilent Technologies, USA).Field blank values were subtracted from sample concentrations.
The recovery rates of trace OM were evaluated by injecting standard mixtures formulated to mimic real sample concentrations into three blank filters respectively and conducting the entire pretreatment and analysis process.The recovery rates of most trace OM were in the range of 80-125%, whereas that of NaP only reached 60% because of its strong volatility.

Data Analysis
PM 2.5 mass concentrations were reconstructed using the method adopted by Li et al. (2013).The PM 2.5 chemical compositions were classified into six major chemical components: NH 4 NO 3 , (NH 4 ) 2 SO 4 , OM, EC, fine soil and others in this research.The NH 4 NO 3 mass was estimated by multiplying the NO 3 -mass by a factor of 1.29 and the (NH 4 ) 2 SO 4 mass was estimated by multiplying the SO 4 2mass by a factor of 1.38.The OM concentration was derived from multiplying the OC concentrations by a factor of 1.4 to account for unmeasured atoms (Xing et al., 2013).
The fine soil mass was estimated using factors associated with common oxides of soil elements (Al 2 O 3 , SiO 2 , CaO, FeO, Fe 2 O 3 , and TiO 2 ), and calculated as follows (Chan et al., 1997): (1) Given a lack of data for Si, the Si mass was estimated using the Al mass, multiplied by a factor of 3.36 (Rudnick and Gao, 2003).

Temporal Variations of Indoor and Outdoor PM 2.5
Fig. 1 provides time series of real-time 1-h PM 2.5 mass concentrations, measured both inside and outside the apartment, together with meteorological parameters.The arrow length represents WS with a range of 0-9 m s -1 and the arrow direction represents WD.Daily PM 2.5 concentrations ranged from 4.5 to 193.1 µg m -3 for indoor with an average of 55.2 ± 47.3 µg m -3 , and ranged from 5.9 to 298.9 µg m -3 for outdoor with an average of 100.4 ± 82.1 µg m -3 .Compared against World Health Organization (WHO) guidelines, the average indoor and outdoor PM 2.5 concentration was almost 2 and 4 times of the daily outdoor limit (25 µg m -3 ), respectively.In addition, approximately 41% of the indoor and 56% of the outdoor daily mean values exceeded the limit value of PM 2.5 (75 µg m -3 ) stipulated by the China National Ambient Air Quality Standards (GB 3095-2012).
As shown in Fig. 1, there were six pollution cycles during the monitoring period, with duration of 3 to 6 days for each cycle.For each cycle, the PM 2.5 concentrations kept at a low level for 1-2 days firstly, then rose slowly over a few days, and finally fell abruptly to the valley value in a few hours.This cyclical pattern has also been observed by Jia et al. (2008).Based on PM 2.5 concentration variation, each cycle can be divided into three periods: a clean air period, a PM 2.5 accumulation period, and a PM 2.5 decline period.As illustrated in Fig. 1, the clean air period came along with the wind from north, northwest, or northeast with an average speed of 3.1 m s -1 that favors the transport and diffusion of air pollutants.During the PM 2.5 accumulation period, the winds were mainly from south or southwest with a low WS of less than 1.7 m s -1 , which resulted in the accumulation and formation of secondary pollutants, as described by Wang et al. (2013b).The sudden drop during PM 2.5 decline period could be attributed to the gale from north or northwest with a speed exceeding 4.0 m s -1 .

Indoor/Outdoor Relationships of PM 2.5 during Different Periods
Indoor/Outdoor (I/O) ratios and regression lines were analyzed to compare indoor/outdoor relationships of PM 2.5 during the clean air period, PM 2.5 accumulation period and PM 2.5 decline period.In this study, the average I/O ratio was 0.60 during the clean air period and 0.51 during the PM 2.5 accumulation period, and the correlation coefficient (r) was 0.746 and 0.944, respectively, which indicated that indoor PM 2.5 was greatly influenced by outdoor sources.The concentrations of PM 2.5 between indoor and outdoor PM 2.5 can be expressed by linear regression equation as follows: In this expression, C in is the indoor PM 2.5 concentration based on both indoor and outdoor sources; C out is the outdoor PM 2.5 concentrations; F in is the infiltration factor; and C ṡ is the indoor PM 2.5 concentration contributed by indoor sources, as described in Chen and Zhao (2011).
It can be seen from Fig. 2 that during the clean air period, the PM 2.5 from indoor sources (C ṡ ) was 4.90 µg m -3 and the corresponding average indoor PM 2.5 concentration (C in ) was 10.53 µg m -3 , suggesting that approximately 47% of the indoor PM 2.5 came from indoor sources.The intercept during the PM 2.5 accumulation period was less than zero (-7.50 µg m -3 ), which might be attributed to some species' decomposition in indoor PM 2.5 .In this study, the I/O ratios for NH 4 NO 3 and NH 4 Cl were about 0.28 and 0.32, respectively, much lower than 0.67 for (NH 4 ) 2 SO 4 .The result proved that part decomposition of NH 4 NO 3 and NH 4 Cl inside the apartment did occur, which could be attributed that indoor temperature was higher than outdoor one during the heating period.
As also shown in Fig. 2, F in was estimated as 0.29 for the clean air period and 0.56 for the PM 2.5 accumulation period, showing that the infiltration of PM 2.5 during the PM 2.5 accumulation period occurred more easily.This result might be due to the fact that during the PM 2.5 accumulation period, quite a number of outside PM 2.5 were secondary aerosols (e.g., sulfate, nitrate, ammonium, secondary organic carbon) with size less than 1 µm (John et al., 1990;Pandis et al., 1993), which have better infiltration performance, compared with particles of other sizes (Long et al., 2001).
Unlike the I/O ratios (< 1) during the clean air period and PM 2.5 accumulation period, the I/O ratio during the PM 2.5 decline period reached 1.26, which was attributed to the fact that the drop in indoor PM 2.5 concentration was later than that in outdoor PM 2.5 concentration, and the PM 2.5 falling rate for indoor (15.0 µg m -3 h -1 ) was lower than that for outdoor (29.4 µg m -3 h -1 ).In this case, fine particulates might travel from inside to outside.

Chemical Compositions of Indoor/Outdoor PM 2.5
Fig. 3 shows the reconstructed indoor/outdoor PM 2.5 chemical compositions during all days, normal days and Chinese Spring Festival, calculated based on material balance.The Spring Festival lasted from the eve of lunar New Year (February 18, 2015) to lantern festival (March 5, 2015).The frequent burning of firework during this period has adverse effects on air quality (Wang et al., 2007).The number at the top of each column is the corresponding average PM 2.5 concentration for each period.For the total sampling period, the proportion of different components in PM 2.5 is as follows: OM (52.0% for indoor and 37.5% for outdoor), (NH 4 ) 2 SO 4 (19.5% for indoor and 14.9% for outdoor), NH 4 NO 3 (8.3% for indoor and 15.3% for outdoor), fine soil (15.4% for indoor and 12.6% for outdoor), EC (4.7% for indoor and 4.2% for outdoor).
Compared with the result for outdoor, the proportion of OM for indoor increased 14.5%, which might be attributed to the presence of more OM sources in indoor environment.The average concentration of OC was 18.3 µg m -3 for indoor and 27.9 µg m -3 for outdoor.It is noted that the percentage of OC to total carbon (TC) (OC/TC) in indoor and outdoor environment reached 90% and 87%, respectively, higher than the values (67%-86%) previously reported in Chinese urban areas (including Beijing, Hongkong, Guangzhou, Shenzhen, Zhuhai and Xi'an) (Cao et al., 2003;Cao et al., 2005;Peng et al., 2009;Cao et al., 2012).This result might be due to the intensive OC emissions from coal burning during heating period.Despite significant efforts made in Beijing, scattered coal stoves were still widely used in the suburban area.
The proportion of NH 4 NO 3 for indoor decreased 7.0%, compared to that for outdoor, which could be attributed to the decomposition of NH 4 NO 3 inside the apartment where the temperature was relatively high, compared with that of outdoor.It is noted that the NO 3 -/SO 4 2-ratio in outdoor environment was 1.1, considerably higher than 0.49 in the winter of 2001-2003in Beijing (Wang et al., 2005).In fact, NO 3 -/SO 4 2-ratio is used to indicating the relative importance of mobile and stationary sources (Yao et al., 2002), and the value suggested the dominance of mobile sources over stationary ones at present.
Compared with the results during normal days, the indoor and outdoor PM 2.5 concentrations during Spring Festival were much higher.For outdoor environment, the concentrations of fine soil, (NH 4 ) 2 SO 4 , NH 4 NO 3 during Spring Festival were 5.2, 2.3 and 1.1 times of those during normal days, respectively.For indoor environment, the corresponding times were 5.7, 1.9 and 2.1.These results were due to the firework burning during Spring Festival.According to Wang et al. (2007), firework burning will generate primary components of K, Ba, Sr, Cl -, Pb, Mg, Al, Cu, Mn and promote the formation of secondary components of SO 4 2-and NO 3 -.The outdoor concentration of Cl -, Pb, Mg, Al, Cu, Mn, SO 4 2and NO 3 -during Spring Festival in this study was found to be 2. 5, 4.5, 12.7, 7.8, 10.6, 3.5, 2.3 and 1.1 times of those during normal days, respectively.K, the most important characteristic element for firework burning, was not detected in this study because the presence of argon ion would cause a big error in the detection of K when ICP-MS uses argon as a carrier gas.On the contrary, the indoor and outdoor concentration of OM during Spring Festival decreased 14% and 24%, compared to those during normal days, respectively, which was attributed to the significant decrease in the emissions from traffic, industry and cooking during Spring Festival when population and anthropogenic activities decreased a lot in this city.

Indoor/Outdoor Relationships of OC, EC, Water Soluble Ions and Total Metal Elements
Linear regression analysis was used to investigate the  indoor/outdoor relationships of OC, EC, water soluble ions and total metal elements during the sampling period.As Table 1 shows, the indoor and outdoor concentrations of OC, EC, Cl -, SO 4 2-and NH 4 + were all strongly correlated with r of 0.79-0.99,indicating that outdoor sources had a considerable impact to indoor PM 2.5 .The intercept in Table 1 indicated that approximately 27% of indoor OC was from indoor sources, and the value was 5% for indoor EC.The result was consistent with He et al. (2004) and Cao et al. (2012).It is noted that the intercepts for Cl -and NH 4 + were both below zero.In general, most of NH 4 + in PM 2.5 are in the form of (NH 4 ) 2 SO 4 and NH 4 NO 3 .According to the charge balance between NH 4 + and acidic ions, the stoichiometric ratio for [NH 4 + ] and [SO 4 2-+ NO 3 -] was 1.18, which suggested that extra NH 4 + would be in the form of NH 4 Cl.Thus the negative intercept was due to the decomposition of NH 4 Cl and NH 4 NO 3 inside the apartment where the temperature was relatively high, compared with that of outdoor (Lunden et al., 2003).
To investigate the indoor/outdoor relationship at different pollution levels, samples were divided into two categories according to the standard daily mean value of 75 µg m -3 for PM 2.5 (China National Ambient Air Quality Standard, GB3095-2012).Ones were samples of clean period with outdoor PM 2.5 concentrations less than or equal to 75 µg m -3 , and the others were samples of pollution period with outdoor PM 2.5 concentrations higher than 75 µg m -3 .I/O ratios, r and P values (through paired-sample t-tests) during the clean Table 1.Indoor/outdoor relationships of OC, EC, water soluble ions, total metal elements, PAHs with different rings and DMP (regression equation: µg m -3 for the first seven species and ng m -3 for the others).period and pollution period were computed for the target components.The concentrations of OC, EC, water soluble ions and total metal elements for indoor were significantly (P < 0.05) lower than those for outdoor.On the other hand, the indoor and outdoor concentrations of most species were in good correlation (r > 0.6).These results indicated these indoor species were greatly influenced by outdoor sources.It is also noted that the indoor/outdoor correlation of most pollutants (OC, Cl -, SO 4 2-and metal elements) were better during pollution period, which aligned with the fact that the outdoor PM 2.5 infiltration during pollution period was stronger than that during clean period.The I/O ratios of Cl -, NO 3 -, and NH 4 + were much lower than that for SO 4 2-, which could be attributed to the decomposition of NH 4 NO 3 and NH 4 Cl, as has been mentioned before.In fact, SO 4 2often serves as the characteristic component to evaluate PM 2.5 infiltration performance for its high stability at ambient temperatures (Jeremy et al., 2002).

Characteristics and Indoor/Outdoor Relationships of Trace Organic Matter
During the sampling period, the PAHs concentration was 187.3 ng m -3 for indoor and 387.0 ng m -3 for outdoor, and the outdoor value was around 1.7-fold of that detected in 2001 winter (228.9 ng m -3 ) and comparable to that detected in 2008 winter in Beijing (407.6 ng m -3 ) (Huang et al., 2006;Wang et al., 2008).In indoor environment, 4-ring PAHs accounted for 38.0% of the total PAHs mass, followed by 5-ring PAHs (36.2%), 6-ring PAHs (12.1%), 3-ring PAHs (11.2%) and 2-ring PAH (2.4%).In outdoor environment, 4-ring PAHs made up 52.2% of PM 2.5 mass, followed by 5ring PAHs (29.3%), 6-ring PAHs (9.4%), 3-ring PAHs (7.9%) and 2-ring PAH (1.2%).Fig. 4 shows the distributions of PAHs and PAEs in indoor and outdoor samples during different periods.The number at the top of each column was corresponded to the concentration of PAHs or PAEs.Compared with the results during normal days, the PAHs concentration in indoor and outdoor during Spring Festival decreased about 50% and 58%, respectively, which could also be attributed to the significant decrease in the emission from traffic, industry and cooking, as has been mentioned before.
The PAEs concentrations ranged from 301.6 to 2606.8 ng m -3 for indoor with an average of 1054.2 ng m -3 , and ranged from 122.3 to 1658.3 ng m -3 for outdoor with an average of 515.3 ng m -3 , which was comparable with those collected from 13 homes in Tianjin (14.9-1591.3ng m -3 for indoor and 7.3-1244.2ng m -3 for outdoor) (Zhang et al., 2014) and those collected from 10 offices in Hangzhou (1375.8ng m -3 for indoor) (Min et al., 2015).Among the 6 kinds of PAEs, DBP and DEHP were the dominant species, accounting for 46% and 43% of the total indoor PAEs mass, and 12% and 81% of the outdoor PAEs mass, which could be attributed to the wide use of DBP and DEHP as additives in China's industrial production (up to 90% of total output) (Wang et al., 2008).Unlike PAHs, the indoor concentration of DEP, DBP, BBP, DEHP and DNOP was approximately 2.0, 7.9, 70.9, 1.1 and 4.8 times of the outdoor concentrations, respectively.Clearly, there existed considerable PAEs sources in the apartment, such as cosmetics, personal care products and resin products (e.g., polyvinyl chloride plastics, polyvinyl acetate, polyurethane) (Gómez-Hens and Aguilar-Caballos, 2003).HCB was detected only in the indoor samples and the average concentration was 5.5 ng m -3 , which indicated the predominant contribution of indoor sources (e.g., rubber, aluminum products, dyes, wood preservatives) (Kümmling et al., 2001).
The indoor/outdoor relationships of PAHs with different rings and DMP were also investigated (shown in Table 1).PAHs with rings of 3, 4, 5 and 6 were in good positive correlation with r of 0.88-0.93,suggesting outdoor sources had a great influence on these species.Whereas, the r of 2ring PAH and DMP was only 0.606 and 0.371, which suggested the definite effect from indoor sources.According to the research conducted by Batterman et al. (2012) and Guerrero and Corsi (2012), the indoor sources of 2-ring PAH include the mothball in closets and deodorizer in toilet bowls.As discussed by Gómez-Hens and Aguilar-Caballos (2003), cosmetics, personal care products and resin products are the indoor sources of DMP.Compared with the result during the clean period, the indoor and outdoor concentrations of 2-ring PAH and DMP during the pollution period were better correlated, which could be attributed to the stronger infiltration of outdoor PM 2.5 during the pollution period as mentioned before.

CONCLUSIONS
This study presented the results of real-time PM 2.5 concentrations and chemical composition of PM 2.5 inside and outside an apartment in urban Beijing during heating period.The average PM 2.5 concentration from January 30 to February 25, 2015 was 55.2 ± 47.3 µg m -3 for indoor and 100.4 ± 82.1 µg m -3 for outdoor, revealing there existed a serious PM 2.5 pollution in Beijing and indoor PM 2.5 was mainly from outdoor sources.OM and (NH 4 ) 2 SO 4 were the dominated components of PM 2.5 , accounted for 71.5% in indoor PM 2.5 and 52.4% in outdoor PM 2.5 , followed by fine soil and NH 4 NO 3 (23.7%and 27.9%).The PAHs concentration was 187.3 ng m -3 for indoor and 387.0 ng m -3 for outdoor, and PAEs concentration was 1054.2 ng m -3 and 515.3 ng m -3 for indoor and outdoor, respectively.HCB was detected only in the indoor samples (5.5 ng m -3 ).Higher concentrations of Cl -, Pb, Mg, Al, Cu, Mn, SO 4 2-and NO 3 were observed during Spring Festival, compared with normal days, which could be attributed to the firework burning.HCB and most PAEs in indoor PM 2.5 were dominated by indoor sources whereas other species were greatly influenced by outdoor sources especially during the pollution period.
Effective measures should be taken to control the infiltration of outdoor pollutants especially during heavy pollution period.

Fig. 1 .
Fig. 1.Time series of real-time mass concentrations of indoor and outdoor PM 2.5 and meteorological parameters.

Fig. 2 .
Fig. 2. I/O ratios and regression lines during the clean air period, PM 2.5 accumulation period and PM 2.5 decline period.

Fig. 3 .
Fig. 3. Reconstructed chemical compositions of indoor and outdoor PM 2.5 organized by material balance during different periods.
P values were calculated from paired-sample t-test, comparing the indoor and outdoor concentrations.
b Correlation coefficients.c Indoor/outdoor concentration ratios.d