Relationship Between Indoor and Outdoor Levels of Fine Particle Mass, Particle Number Concentrations and Black Smoke Under Diffe

Journal of Exposure Analysis and Environmental Epidemiology (2004) 14, 275–283 r 2004 Nature Publishing GroupAll rights reserved 1053-4245/04/$30.00

www.nature.com/jea

Relationship between indoor and outdoor levels of ﬁne particle mass, particle number concentrations and black smoke under different ventilation conditions

JOSEF CYRYS,a,b MIKE PITZ,b WOLFGANG BISCHOF,c H.-ERICH WICHMANNa,b AND JOACHIM HEINRICHa aGSF-National Research Center for Environment and Health, Institute of Epidemiology, Neuherberg, Germany bInstitute of Medical Data Management, Biometrics and Epidemiology, Ludwig-Maximilians-University of Munich, Munich, Germany cDepartment Indoor Climatology, Friedrich-Schiller-University Jena, Jena, Germany

Fine particle mass (PM2.5), black smoke (BS) and particle number concentration (NC) were measured simultaneously indoors and outdoors at an urban location in Erfurt, Germany. Measurements were conducted during 2-month periods in summer and winter. Different ventilation modes were applied during each measurement period: windows closed; windows opened widely for 15 min twice per day; windows and door across the room opened widely for 5 min twice per day and windows tilted open all day long. The lowest indoor/outdoor (I/O) ratios for all pollutants were found for closed windows, whereas the ratios for ventilated environments were higher. For closed windows, the I/O ratios for PM2.5 are larger than the corresponding values for BS and NC (0.63 vs. 0.44 or 0.33, respectively) probably due to lower penetration factors for particles sizeso500 nm and higher deposition rates for ultraﬁne particles (o100 nm). The largest differences for the I/O ratios between closed and tilted windows were found for NC (0.33 vs. 0.78). The indoor and outdoor levels of PM2.5 and BS were strongly correlated for all ventilation modes. The linear regression models showed that more than 75% of the daily indoor variation could be explained by the daily outdoor variation for those pollutants. However, the correlation between indoor and outdoor NC for ventilation twice a day was weak. It indicates that rapid changes of the air exchange rates during the day may affect the correlation and regression analysis of NC indoor and outdoor concentrations. This effect was not observed for PM2.5 or BS. This study shows the importance of the indoor air aerosol measurements for health effects studies and the need for more research on I/O transport mechanisms for NC. Journal of Exposure Analysis and Environmental Epidemiology (2004) 14, 275–283. doi:10.1038/sj.jea.7500317

Keywords: indoor/outdoor ratios, ﬁne particles, PM2.5, black smoke, particle number concentration, ventilation conditions.

Introduction Wichmann and Peters 2000; Wichmann et al., 2000; Penttinen et al., 2001). The effects of airborne particulate matter (PM) on human The results of a study conducted by Laden et al. (2000) health have been extensively examined in a large number of indicate that fine particles from mobile and coal combustion epidemiological studies (ATS, 1996; US-EPA, 1999; WHO, sources, but not crustal particles, are associated with 2000; Schneider et al., 2003). PM in those studies has usually increased mortality in six US cities. Thus, in addition to been measured as the mass of particles smaller than 10 mmin the measurements of particle number and/or mass concen- aerodynamic diameter (PM10). More recently, several studies tration the characterization of the chemical composition of have reported the health significance of fine particles smaller particles may also be important for assessment of health than 2.5 mm(PM2.5). It has been suggested that a large effects. Several studies have documented that black smoke number of ultrafine particles (particles smaller than 0.1 mm) (BS), derived from absorbance coefficients, is well correlated in urban air may be the explanation for the observed health with the concentration of elemental carbon (EC) or soot and effects of PM (Oberdo¨ rster et al., 1995; Seaton et al., 1995). can be recommended as a valid and cheapindicator in studies The adverse effect of ultrafine particles on respiratory on combustion-related air pollution and health (Kinney et al., symptoms, lung function and daily cardiopulmonary mor- 2000; Janssen et al., 2001; Gotschi et al., 2002; Cyrys et al., tality has been discussed in several studies (Peters et al., 1997; 2003). As people generally spend very little time outdoors (Jenkins et al., 1992), the correlation of health effects seems 1. Address all correspondence to: Dr. Josef Cyrys, GSF-National to suggest that outdoor pollution which is transferred indoors Research Center for Environment and Health, Institute of Epidemiology, has an important effect on health. This may be plausible only Ingolstaedter Landstr. 1, D-85764 Neuherberg, Germany. if health-relevant constituents of ambient air pollution Tel: þ 49-89-3187-4156. Fax: þ 49-89-3187-3380. E-mail: [email protected] efficiently penetrate indoors. A number of studies have been Received 16 October 2002; accepted 25 August 2003. conducted on the relationshipbetween indoor and outdoor Cyrys et al. Indoor/outdoor ratios of particulate air pollutants particles (Wallace, 1996; Monn et al., 1997; Monn, 2001). While the focus of most studies has been on particle mass, there is still very little information available on the relationshipbetween the number concentration (NC) of particlesor BS in indoor and outdoor air. However, better understanding of those relationships under different environmental conditions is of importance for improving exposure estimates and in turn for the development of efficient strategies to Side street reduce human exposure and human health risk. Outdoor measurement station To improve the understanding of the relationship between Major street indoor and outdoor particles, this study was undertaken with the following objectives: (1) to analyze the relationship between indoor and outdoor PM concentrations, NC and 40 m 2.5 Room 1 BS for an indoor environment without any indoor sources for Room 2 particles under different ventilation conditions; (2) to estimate the influence of the outdoor levels, ventilation conditions and meteorological parameters such as temperature and relative humidity on the indoor levels; and (3) to Hospital building study the diurnal variation of indoor and outdoor particle NC on an hourly basis.

Methods Figure 1. Diagram of the measurement locations for simultaneous outdoor and indoor particle collection in Erfurt. Measurement Site and Description of Sampling Period The study was conducted in Erfurt, a city with approximately 200,000 inhabitants in eastern Germany. Indoor and open (tilted means partially opened) for the entire sampling outdoor particle concentrations were measured simulta- period. In the two last weeks, the room was ventilated twice a neously in order to be able to investigate how outdoor day, at 8 pm and 2 am, for (a) 15 min with the windows wide conditions influence indoor air quality. The measurement site open during the third week, and (b) 5 min with the windows was located approximately 1 km south of the Erfurt city and the door across the room wide open (draughty center and 30 m from the nearest major road (Figure 1). At ventilation) during the fourth week. In September–October this location, ambient air was sampled from a height of 4 m 2002, the air exchange rates were measured for each above the ground using a custom inlet. The indoor ventilation mode. measurements were carried out in a large, natural-ventilated hospital building alternating between two empty rooms: one Instrumentation on this side of the building where the outdoor measurement The indoor and outdoor measurements of particle NCs were site was located (room 1), the other one on the opposite side conducted simultaneously with two identical instruments. We of the building (room 2). No indoor sources of particles were used two TSI Model 3022A Condensation Particle Counters present in either room and no human activities occurred (CPC). This instrument counts particles within 0.007 and during the measurement periods, with the exception of the 3.0 mm geometric diameter. Particles outdoors were counted ventilation activities and the changing of the filter for PM2.5 at a height of 4.0 m at a flow rate of 5 ml/s without size- and BS measurement. Both the rooms were located on the selection every half a minute. However, as PM2.5 and BS first floor of the building at 5 m above the ground. The samples are available only as 24-h samples, the comparison windows had wood frames, were of old construction and of indoor/outdoor (I/O) relationshipfor all pollutants were not airtight. The rooms were not air-conditioned. The (PM2.5, BS and NC) was possible only on a daily basis. measurements were conducted during two 8-week periods: Fine particle mass PM2.5 was collected with Harvard May–July, 2001 (summer season) and January–March, impactors (Marple et al., 1987) according to a standard 2002 (winter season). During each measuring period, operating procedure SOP TRAPCA2.0 (Hoek et al., 2002). the indoor particle concentrations were measured for 4 The sampling duration was from midnight to midnight weeks in room 1 and for 4 weeks in room 2 on the opposite (24 h). The samples were collected at a flow rate of 10 l/min, side of the building. For each week, different ventilation on Anderson 37 mm 2 mm pore size Teflon filters with a poly modes were chosen. In the first week, the windows were support ring. Before and after sampling, filters were closed. In the second week, the windows were kept tilted conditioned at 21721Cand3575% relative humidity for

276 Journal of Exposure Analysis and Environmental Epidemiology (2004) 14(4) Indoor/outdoor ratios of particulate air pollutants Cyrys et al. at least 24 h and then weighed using a Sartorius M5P V001 Statistical Analysis microbalance with a resolution of 0.0001 mg. As the data distribution was positively skewed, nonpara- After weighing, the reflectance of all Teflon filters was metric methods were used to characterize the distribution of measured using the M43D Smoke Stain Reflectometer air pollutants. Instead of arithmetic means, medians were (Diffusion Systems Ltd), which measures the percent calculated because they were believed to better characterize reflection of light incidence. The blackness of the PM2.5 filter air pollutants levels. The Pearson correlation coefficients of deposit was measured according to a procedure described log-transformed data were calculated to assess the association previously (Fischer et al., 2000). The exact measurement between indoor and outdoor levels of the measured procedure is described at the internet site for the EU multi- parameter. To assess the impact of different ventilation and center study ULTRA (http://www.ktl.fi/ultra). Briefly, the meteorological conditions on indoor concentrations, multiple average reflectance was transformed into an absorption linear regression models were developed. The logarithmically coefficient according to ISO 9835 (ISO, 1993): transformed (ln) indoor concentration was chosen as the dependent variable and logarithmically transformed outdoor a ¼ððA=2Þ=VÞ lnðR0=Rf Þ concentration, ventilation conditions, temperature and humidity were considered as potential predictors. The models where a is the absorption coefficient (mÀ1 Â 10À5), A the were adjusted for indoor location (room 1 or 2). The variable loaded filter area (0.00078 m2), V the sampled volume (m3), ‘‘season’’ was not included in the models because tempera- R0 the reflectance of field blank (%), and Rf the reflectance of ture and humidity were believed to better describe the effect sampled filter (%). The absorbance coefficient was then of season. All computations were performed using the multiplied by 105 to make the readings more comprehensible. statistical analysis package SAS for Windows version 6.12 Reflectance applied in the current work was not converted (SAS Institute, Cary, NC, USA). into units of mass concentration since the transformation used to calculate mass concentration is filter specific and no transformation is available for the filters used in this study. Results Nevertheless, in this work, the expression BS is used synonymously with reflectance, measured at light absorption I/O Relationship on a Daily Basis coefficient per meter. The measured air exchange rate was 0.91 hÀ1 for closed Air exchange rates have been determined by the constant- windows, 1.12 hÀ1 for 15 min ventilation twice a day with injection-method (VDI 4300/7, 2001) using Perfluoromethyl- windows wide open, 1.33 hÀ1 for 5 min draughty ventilation cyclohexane (CAS 355-02-2) as tracer and 1-week passive twice a day and finally 3.44 hÀ1 for windows tilted open. As sampling. the ventilation twice a day was carried out only on weekdays, we restricted the further analysis only to weekdays. Quality Control Table 1 shows the air pollution levels for all data and Before and after each I/O measurement period, both CPC stratified by ventilation mode. The particle mass was on instruments were run 2 weeks in parallel to control for average 6.9 mg/m3 indoors and 9.2 mg/m3 outdoors, the NCs interinstrument variation. The correlation coefficients were 5,076 particles/cm3 indoors and 11,450 particles/cm3 out- always larger than 0.90, most were larger than 0.95. The doors. All median I/O ratios are o1. The PM2.5 I/O ratios differences between the instruments were less than 10% (the for all ventilation modes were larger than the corresponding uncertainty range given by the manufacturer is 710%). values for BS and NC. The lowest I/O ratios for all The detection limits for PM2.5 were calculated as three pollutants were found for closed windows, whereas the ratios times the average SD, where SD is the standard deviation of for ventilated environments were higher. The largest the field blanks. The filters for field blanks measurements differences for the I/O ratios were found between closed weretakentothemeasurementsiteandplacedinthe windows and tilted windows (Figure 2). collocated samplers, but the samplers were not turned on. Multiple linear regression models to predict indoor 2 During sampling periods, we collected eight field blanks for concentrations of PM2.5, BS and NC show high R for all the indoor PM2.5/BS measurements and six for the outdoor pollutants (84% for PM2.5, 86% for BS and 83% for NC 3 PM2.5/BS measurements. The detection limit was 1.2 mg/m (Table 2)). Besides outdoor levels of the specific pollutants, 3 for PM2.5 indoors, 3.2 mg/m for PM2.5 outdoors, only ventilation mode and temperature are significantly 0.05 Â 10À5 mÀ1 for BS indoors and 0.09 Â 10À5 mÀ1 for associated with the indoor levels of pollutants. No other BS outdoors. There were three PM2.5 indoor (ntot ¼ 111), 10 parameters, such as humidity or location, are associated with PM2.5 outdoor (ntot ¼ 109), seven BS indoor (ntot ¼ 111) and the indoor pollutant levels. Outdoor levels explained most of three BS outdoor samples (ntot ¼ 109) under the detection the indoor variation of PM2.5 and BS (76% and 80%, limit. The coefficient of variance was 3.6% and 1.9% for respectively), but only 43% of the indoor variation of NC. PM2.5 and BS, respectively. The ventilation mode (in particular windows tilted open all

Journal of Exposure Analysis and Environmental Epidemiology (2004) 14(4) 277 Cyrys et al. Indoor/outdoor ratios of particulate air pollutants

Ta bl e 1 . Daily indoor and outdoor levels for all data and stratiﬁed by ventilation mode and location and the corresponding I/O ratios (for the calculation of I/O ratios only days with both indoor and outdoor concentrations higher than detection limit were used).

Pollutant N Median indoor Median outdoor N Median of I/O ratios

PM2.5 All data 77 6.9 9.2 70 0.79 BS All data 77 0.6 1.2 73 0.53 NC All data 77 5,076 11,450 77 0.42

À1 PM2.5 Windows closed (a ¼ 0.91 h ) 21 7.1 10.7 20 0.63 À1 PM2.5 2 Â 15 min ventilation (a ¼ 1.12 h ) 19 6.9 9.4 16 0.83 À1 PM2.5 2 Â 5 min ventilation (a ¼ 1.33 h ) 17 7.8 9.0 17 0.85 À1 PM2.5 Windows tilted (a ¼ 3.44 h ) 20 6.1 7.8 17 0.83 BS Windows closed (a ¼ 0.91 hÀ1) 21 0.7 1.6 18 0.44 BS 2 Â 15 min ventilation (a ¼ 1.12 hÀ1) 19 0.6 1.2 18 0.59 BS 2 Â 5 min ventilation (a ¼ 1.33 hÀ1) 17 0.6 1.3 17 0.46 BS Windows tilted (a ¼ 3.44 hÀ1) 20 0.7 1.1 20 0.61 NC Windows closed (a ¼ 0.91 hÀ1) 21 4,261 13,032 21 0.33 NC 2 Â 15 min ventilation (a ¼ 1.12 hÀ1) 19 5,018 10,577 19 0.43 NC 2 Â 5 min ventilation (a ¼ 1.33 hÀ1) 19 4,513 12,046 19 0.42 NC Windows tilted (a ¼ 3.44 hÀ1) 18 8,027 11,223 18 0.78

2.0 for NC was somewhat lower (r ¼ 0.64). The stratiﬁcation by

PM2.5 BS NC ventilation mode makes clear that the weaker correlation for 1.6 NC is caused by the lower correlation between NC indoors and outdoors when the room was ventilated twice a day. Obviously, the immediate increase of NC after windows are 1.2 opened introduces an artifact into the correlation analysis (see also Figure 3). It is noteworthy that a similar effect was

I/O Ratio 0.8 not observed for PM2.5 or BS.

I/O Particle NCs on an Hourly Basis 0.4 AsNCwasmeasuredonanhourlybasis,amoredetailed analysis is possible for this parameter in order to provide 0.0 12 12 12 information on the diurnal variability. Figure 3 shows the diurnal patterns of 1-h averages of indoor and outdoor NC 1 - windows closed 2 - windows tilted open stratified by ventilation mode (only Monday–Friday). The diurnal pattern of indoor and outdoor NC under minimum Figure 2. Ratios between indoors and outdoors stratified by ventilation mode. ventilation conditions (windows closed) shows that indoor concentrations are strongly attenuated compared to outdoor concentrations. The range of indoor NC is much smaller day long vs. closed windows) explains a further 40% of the than that of outdoors. The indoor air concentrations follow indoor NC variation, which is in line with the large difference the diurnal pattern of the outdoor air only slightly and with a between the NC I/O ratio for closed and tilted windows (as small lag. Whereas the outdoors NC rises in the morning shown in Table 1). After excluding sampling periods with from a baseline of approximately 9500 particles/cm3 to a tilted windows, the partial R2 for outdoor concentrations maximum concentration of 21,000 cmÀ3 at 08:00, the increases from 43% to 63% and the R2 for ventilation mode increase is less pronounced for the NC indoors during this decreases from 40% to 6%, which makes the NC regression time (increase from 4000 to 5700 particles/cm3 at 09:00). For model more similar to the models for PM2.5 and BS (data not other ventilation conditions, NC indoors are clearly affected shown). by NC outdoors. When the windows was kept tilted, NC In Table 3, the correlation coefficients between indoor and indoors followed NC outdoors and showed similar diurnal outdoor levels of PM2.5, BS and NC are listed for all raw pattern (however with lower concentration levels). Also, the data and stratified by ventilation mode (after adjusting for effect of opening windows twice a day on indoor NC could temperature). The strongest correlations (for all data) were be clearly seen. Immediately after the ventilation occurred (at found for PM2.5 and BS (r40.85), whereas the correlation 08:00 and 14:00), indoor NC increased rapidly, but did not

278 Journal of Exposure Analysis and Environmental Epidemiology (2004) 14(4) Indoor/outdoor ratios of particulate air pollutants Cyrys et al.

Ta bl e 2 . Results of regression models for indoor PM2.5, BS and NC concentrations.

Variable Parameter estimate R2 full model (individual variables) P-value

PM2.5 0.84 3 PM2.5 outdoor (mg/m ) 0.79 0.76 o0.0001 Ventilation: windows closed (a ¼ 0.91 hÀ1) 0.03 0.0115 2 Â 15 min ventilation (a ¼ 1.12 hÀ1)0.28 0.0029 2 Â 5 min ventilation (a ¼ 1.33 hÀ1)0.19 0.0374 Windows tilted open all day (a ¼ 3.44 hÀ1)0.25 0.0050 Temperature ( 1C) 0.02 0.05 o0.0001 Relative humidity (%) 0.00 0.00 0.6608 Location:Room1 Room 2 À0.05 0.00 0.3877

BS 0.86 BS outdoor (mg/m3) 1.16 0.80 o0.0001 Ventilation: windows closed (a ¼ 0.91 hÀ1) 0.04 0.0017 2 Â 15 min ventilation (a ¼ 1.12 hÀ1)0.46 0.0011 2 Â 5 min ventilation (a ¼ 1.33 hÀ1)0.21 0.1351 Windows tilted open all day (a ¼ 3.44 hÀ1)0.45 0.0008 Temperature ( 1C) À0.03 0.02 0.0037 Relative humidity (%) 0.00 0.00 0.9291 Location: Indoor 1 Indoor 2 À0.07 0.00 0.4314

NC 0.83 NC outdoor (mg/m3) 0.79 0.43 o0.0001 Ventilation: windows closed (a ¼ 0.91 hÀ1) 0.40 o0.0001 2 Â 15 min ventilation (a ¼ 1.12 hÀ1)0.20 0.0041 2 Â 5 min ventilation (a ¼ 1.33 hÀ1)0.11 0.1142 Windows tilted open all day (a ¼ 3.44 hÀ1)0.78 o0.0001 Temperature ( 1C) 0.00 0.00 0.4582 Relative humidity (%) 0.00 0.00 0.9427 Location: Indoor 1 Indoor 2 À0.05 0.00 0.302

The R2 for the full model is indicated along with the R2 contribution of each individual variable for the model containing all listed variables. Based on natural log-transformed data. Bold indicate P-valuep0.005.

Ta bl e 3 . Pearson correlation coefficients between indoors and out- another 1–2 h returned to the initial levels. The brief doors for all data (raw) and stratified by ventilation mode on a daily ventilation twice a day caused an increase of the ratio during basis (adjusted for temperature), log-transformed data. the ventilation period and immediately after it. Apart from the ventilation activities, the diurnal pattern of the I/O ratio PM BS NC 2.5 for the rest of the day is very similar to the I/O ratio for All data 0.88 0.91 0.64 closed windows. Resulting from this pattern, the daily n ¼ 74 n ¼ 74 n ¼ 74 average of the NC I/O ratio with ventilation twice a day À1 Windows closed (a ¼ 0.91 h ) 0.91 0.93 0.92 remains low. With tilted windows, where the indoor NC level n 21 n 21 n 21 ¼ ¼ ¼ follows the outdoor NC level quite well, the I/O ratio varies 2 Â 15 min ventilation (a ¼ 1.12 hÀ1) 0.93 0.96 0.62 n ¼ 18 n ¼ 18 n ¼ 18 from 0.70 to 0.90. 2 Â 5 min ventilation (a ¼ 1.33 hÀ1) 0.95 0.87 0.76 n ¼ 17 n ¼ 17 n ¼ 17 À1 Windows tilted (a ¼ 3.44 h ) 0.92 0.90 0.83 Discussion n ¼ 18 n ¼ 18 n ¼ 18 Limitations reach the outdoor levels neither in the morning nor in the Several limitations should be taken into account when afternoon. After the windows were closed, indoor NC interpreting our results. All field data were taken at one decreased exponentially. At 1 h after the ventilation, NC particular location with a specific ventilation system. The decreased by more than half of the increase peak and after windows were not airtight, the rooms were not air-

Journal of Exposure Analysis and Environmental Epidemiology (2004) 14(4) 279 Cyrys et al. Indoor/outdoor ratios of particulate air pollutants

Windows closed (a=0.91 h-1) applied successively, that is, the concentration ranges for the 24000 1.2 different ventilation modes varied (the range of concentra- 20000 1.0 tions measured in the period in which windows were opened twice a day is smaller than the concentration range of the 16000 0.8

) period in which windows were closed or tilted open). Our 3 - 12000 0.6 results do not necessarily represent the I/O relationship for I/O ratio NC (cm the considered pollutants in other locations with different 8000 0.4 ventilation systems, I/O activities, geographical settings, 4000 0.2 weather conditions, etc. Nevertheless, this study provides a unique opportunity to examine I/O ratios of particle mass, 0 0.0 04 812 16 20 24 particle number and BS in the absence of any known indoor Hours of the day aerosol sources and under well-deﬁned ventilation mode. 2x15 min ventilation (a=1.12 h-1) 24000 1.2 I/O Relationship Ventilation at 8 am Ventilation at 2 pm 20000 1.0 The particle mass and number concentrations measured in our study were rather moderate in comparison with the 16000 0.8 annual averages measured in Erfurt in earlier years ) 3 - 12000 0.6 (Wichmann et al., 2000, 2002; Ebelt et al., 2001). However, I/O ratio

NC (cm this study focused on the relationshipbetween indoor and 8000 0.4 outdoor PM2.5, BS and NC in the absence of any indoor

4000 0.2 sources under controlled ventilation conditions. A number of studies on the relationshipbetween indoor and outdoor 0 0.0 04 812162024 particle mass concentration have been conducted. All of the Hours of the day studies showed that indoor particle mass concentration is a 2x15 min ventilation (a=1.33 h-1) function of a number of factors, even in the absence of any 24000 1.2 signiﬁcant indoor sources of particles. The most important of Ventilation at 8 am Ventilation at 2 pm 20000 1.0 these factors are the outdoor particle concentration, air exchange rate, particle penetration efﬁciency from the 16000 0.8 outdoor to the indoor environment, the particle deposition ) 3 - 12000 0.6 rate on indoor surfaces, and meteorological factors outdoors I/O ratio

NC (cm (Kamens et al., 1991; Thatcher and Layton, 1995; Chan, 8000 0.4 2002).

4000 0.2 To estimate the contribution of outdoor particles to indoor levels, a physical-statistical model can be used. This model is 0 0.0 04812162024 based on the indoor air mass balance equation that has been Hours of the day applied in previous exposure and indoor studies (O¨ zkaynak Windows tilted (a=3.44 h-1) et al., 1996; Abt et al., 2000a). The model assumes that the 24000 1.2 amount of pollutant that enters the home without apparent

20000 1.0 indoor sources equals that leaving the home, and is expressed as 16000 0.8

) C PaC = a k 3 IN ¼ OUT ð þ Þ - 12000 0.6 I/O ratio NC (cm where CIN and COUT are the indoor and outdoor concentra- 8000 0.4 tions, respectively (mg/m3), P is the penetration efficiency À1 4000 0.2 (dimensionless), a is the ventilation rate (h )andk is the À1 deposition rate (h ). Thus, the I/O ratio (CIN/COUT)isa 0 0.0 04 812162024 function of the penetration efficiency and of particle losses Hours of the day from exfiltration and deposition (a and k). Figure 3. Diurnal pattern of indoor and outdoor average particle The penetration rate for PM is approximately one and number concentration and I/O ratios for different ventilation modes 2.5 (weekdays only). deposition rates were found to be typically between 0.4 and 1.0 hÀ1 (Monn, 2001). As the measured air exchange rate in conditioned. The temperature and relative humidity ranges in our indoor room was 0.91 hÀ1 (when window were closed), our study were relatively small, due to the moderate climatic the I/O ratio for PM2.5 is expected to be between 0.48 and conditions in Central Europe. The ventilation modes were 0.70. The measured value of 0.63 is in line with the expected

280 Journal of Exposure Analysis and Environmental Epidemiology (2004) 14(4) Indoor/outdoor ratios of particulate air pollutants Cyrys et al. values. The estimated deposition rate is 0.53 hÀ1.Fora of people (Abt et al., 2000b). Koponen et al. (2001) studied typical home with an air exchange rate of 0.75 hÀ1, the I/O the effect of ventilation on the I/O ratios of particle number ratio would be about 0.65 for PM2.5 and 0.45 for PM10 concentrations in different particle size classes. Although this (Monn, 2001). However, such low I/O ratios were deter- study applied a different study design (the windows in the mined in only a few homes. The average I/O ratio of PM2.5 room were permanently closed and the room was ventilated (n ¼ 15 homes) reported by Morawska et al. (2001) was 1.08. by a ventilation system equipped with filters) the results The I/O ratios were measured in homes without indoor reported are similar to our results. The ventilation in the activities resulting in particle generation and under minimum Kopenen study began during office hours and was turned off ventilation condition (all doors and windows were closed). in night time. The I/O ratio for particles between 90 and As BS daily average concentrations are strongly correlated 500 nm increased from 0.20 upto 0.36 when the ventilation with the particle mass fraction between 50 and 100 nm and was turned on and remained quite stable during the day. To 100–500 nm (r ¼ 0.85 and 0.77, respectively; data not our knowledge, only one study (Morawska et al., 2001) has shown), we expected that the main fraction of BS measured shown I/O ratios of NC for conditions similar to this study in our study to be between 50 and 500 nm. As shown by (particle NC in the size range 15–685 nm and air exchange Wichmann et al. (2000), the particle NC is dominated by the rate between 0.5 and 1.0 hÀ1). The average I/O ratio found in ultrafine fraction (with diameter smaller than 100 nm). Since the Morawska study was higher (0.7870.49) than that particles smaller than 100 nm were found to have higher found in our study (0.33). However, the I/O ratios were deposition rates, more likely coarse-mode particles than measured in 15 houses selected from a mix of house types, accumulation-mode particles (Lai and Nazaroff, 2000; Long both in terms of age and design, including new and old et al., 2001), we expect lower I/O ratios for BS and NC in houses. In four houses (26%), the I/O ratios for NC were comparison with PM2.5. Moreover, the penetration factors less than 0.45, the lowest I/O ratio was 0.21. Therefore, as for all particles sizes o500 nm were determined to be less stated above, one should still keepin mind that our results than unity (Vette et al., 2001). The lower penetration factors cannot represent the general I/O relationship for the may contribute to the lower I/O ratios for those pollutants considered pollutants for other locations with different due to more efficient filtration of BS and NC than PM2.5 by ventilation systems, age, house design and construction the shell of the building. material. In fact, the I/O ratios for BS and NC measured in our In conclusion, the lower BS and NC I/O ratios suggest study were clearly lower than those for PM2.5. The I/O ratios that due to lower penetration efficiencies and/or higher for BS (from 0.44 for closed windows upto 0.61 for windows deposition rates, the building protects individuals more from tilted open) were in general lower than the I/O ratios for BS exposure to BS and NC than from exposure to PM2.5. or EC reported in other studies in which indoor activities took place. The average indoor-to-outdoor EC concentration Influence of Ventilation and Meteorological Conditions on in nonsmoking residences found by Geller et al. (2002) was the I/O Relationship of PM2.5,BSandNC 0.85. Fischer et al. (2000) measured I/O ratios of BS at The strong correlation coefficients between indoor and homes located on high and low traffic intensity streets. The I/ outdoor levels of PM2.5 and BS for all applied ventilation O ratio was 0.76 for high traffic homes and 0.69 for low modes indicate that the outdoor concentrations can be used traffic homes. The averaged I/O BS ratios determined by to predict the indoor concentrations for those pollutants in Gotschi et al. (2002) at homes in four European cities ranged the absence of prominent indoor aerosol sources. The linear from 0.79 to 0.96. However, the ratios reported in those regression models showed that more than 75% of the daily studies seem to imply the presence of indoor emission indoor variation could be explained by the daily outdoor sources. Thus, the lower I/O ratios of BS concentrations variation for those pollutants. Moreover, the I/O ratio found in our study in comparison with the other studies could increases only moderately by opening the windows: from be explained by the absence of any indoor sources. 0.63 (windows closed) to 0.83 (windows tilted open) for

The lowest I/O ratio for closed windows was found for NC PM2.5 and from 0.44 to 0.61 for BS. (0.33). Only a few studies on the indoor/outdoor relationship This increase is much more pronounced for NC (0.33 to of particle number concentrations have been conducted 0.78). Moreover, the inﬂuence of ventilation mode on the I/O previously. Tu and Knutson (1988) made real-time indoor relationshipis stronger for NC. Whereas the correlation and outdoor measurements of particle number concentra- between indoor and outdoor NC for closed and tilted tions in single-family houses with a slow air exchange rate. windows is comparable with those for PM2.5 and BS, it is The I/O ratios that varied from 0.20 to 41 indicated that in weaker for ventilation twice a day. The temporary opening of such airtight houses with a lot of human activity, indoor the windows, which is the most common ventilation behavior sources accounted for a large portion of the indoor air during the day in the winter season (Wichmann et al., 1999), particle counts. The main activities associated with particle immediately increased NC indoors (Figure 3). The very rapid generation indoors included cooking, cleaning and movement increase of NC during ventilation and the exponential

Journal of Exposure Analysis and Environmental Epidemiology (2004) 14(4) 281 Cyrys et al. Indoor/outdoor ratios of particulate air pollutants

decrease after ventilation (because of removal by coagulation Ebelt S., Brauer M., Cyrys J., Tuch T., Kreyling W.G., and Wichmann processes and/or diffusion to surfaces) affects the correlations H.E., et al. Air quality in postunification Erfurt, East Germany: between NC indoors and outdoors and attenuates the associating changes in pollutant concentrations with changes in emissions. Environ Health Perspect 2001: 109(4): 325–333. prediction of indoors NC using outdoors NC. This result Fischer P.H., Hoek G., van Reeuwiijk H., Briggs D.J., Lebret E., and indicates that whereas a substantial fraction of the indoor van Wijnen J.H., et al. Traffic-related differences in outdoor and PM2.5 and BS variation can be attributed to that of the indoor concentrations of particles and volatile organic compounds in outdoor concentrations, the variation in indoor NC is Amsterdam. Atmos Environ 2000: 34: 3713–3722. strongly influenced by the rapid changes of the ventilation. Geller M.D., Chang M., Sioutas C., Ostro B.D., and Lipsett M.J. We could not exclude that temperature differences between Indoor/outdoor relationshipand chemical composition of fine and coarse particles in the southern California deserts. Atmos Environ outdoors and indoors might influence the I/O ratio. 2002: 36: 1099–1110. However, since the indoor locations are not occupied, we Gotschi T., Oglesby L., Mathys P., Monn C., Manalis N., and

believe that the temperature gradient between indoor and Koistinen K., et al. Comparison of black smoke and PM2.5 levels in outdoor is not so pronounced as for occupied (comfortable indoor and outdoor environments of four European cities. Environ and more stable indoor temperature) rooms. Unfortunately, Sci Technol 2002: 36(6): 1191–1197. Hoek G., Meliefste K., Cyrys J., Lewne´ M., Bellander T., and Brauer we did not measure the indoor temperature, so that we are B., et al. Spatial variability of fine particle concentrations in three not able to verify this assumption. No statistically significant European countries. Atmos Environ 2002: 36: 4077–4088. associations were seen between other meteorologically Janssen N.A.H., van Vliet P., van Aarts F., Harssema H., and conditions, such as wind speed or relative humidity and I/ Brunekreef B. Assessment of exposure to traffic related air pollution O ratios of the considered pollutants. of children attending schools near motorways. Atmos Environ 2001: 35: 3875–3884. In conclusion, ambient concentrations of PM2.5 and BS Jenkins P.L., Philips T.J., Mulberg J.M., and Hui S.P. Activity patterns can be used as a good approximation of indoor concentra- of Californians: use of and proximity to indoor pollutant sources. tions in the absence of indoor particle sources. However, this Atmos Environ 1992: 26A: 2141–2148. result is true for NC only with restrictions. Rapid changes of ISO. Ambient air-determination of a black smoke index (ISO 9835), the air exchange rates during the day may lead to lower International Organization for Standardization, International Stan- dard 9835-1993 (E). 1993. correlations between indoor and outdoor NC concentrations. Kamens R., Lee C.T., Weiner R., and Leith D. A study to characterize This effect is less pronounced for PM2.5 or BS. This study indoor particles in three non-smoking homes. Atmos Res 1991: 25: shows the importance of the indoor air aerosol measurements 939–948. for health effects studies and the need for more research on Kinney P.L., Aggarwal M., Northridge M.A., Janssen N.A.H., mechanisms of I/O transport of NC. andShepardP.PersonalexposuretoPM2.5 and diesel exhaust particles on Harlem sidewalks. Environ Health Perspect 2000: 108: 213–218. Koponen I.K., Asmi A., Keronen P., Puhto K., and Kulmala M. Acknowledgments Indoor air measurement campaign in Helsinki, Finland 1999 F the effect of outdoor air pollution on indoor air. Atmos Environ 2001: 35: We acknowledge the Federal Environmental Agency Berlin 1465–1477. for supporting this research through Grant FE 200 42 263. A Laden F., Neas L.M., Dockery D.W., and Schwartz J. Association of fine particulate matter from different sources with daily mortality in special thank is paid to Mr. Koschine for his contribution to six U.S. cities. Environ Health Perspect 2000: 108(10): 941–947. the performance of the field and laboratory work and to Ms. Lai K., and Nazaroff W.W. Modeling indoor particle deposition Cara Carty for critical reading of the manuscript. from turbulent flow onto smooth surfaces. J Aerosol Sci 2000: 31(4): 463–476. Long C.M., Suh H.H., Catalano P.J., and Koutrakis P. Using References time- and size-resolved particulate data to quantify indoor penetration and deposition behavior. Environ Sci Technol 2001: 35(10): Abt E., Suh H.H., Catalano P., and Koutrakis P. The relative 2089–2099. contribution of outdoor and indoor particle sources to indoor Marple V., Rubow K.L., Turner W., and Spengler J.D. Low flow rate concentrations. Environ Sci Technol 2000a: 34: 3579–3587. sharp cut impactors for indoor sampling: design and calibration. Abt E., Suh H.H., Allen G., and Koutrakis P. Characterization of J Air Pollut Control Assoc 1987: 37: 1303–1307. indoor particle sources: a study conducted in the metropolitan Monn Ch, Fuchs A., Ho¨ gger D., Junker M., Kogelschatz D., and Roth Boston area. Environ Health Perspect 2000b: 108(1): 35–44. N., et al. Particulate matter less than 10 mm(PM10) and fine particles ATS, American Thorasic Society. State of the art: health effects of less than 2.5 mm(PM2.5): relationships between indoor, outdoor and outdoor air pollution. Am J Respir Crit Care Med 1996: 153: 3–50. personal concentrations. Sci Total Environ 1997: 208: 15–21. Chan A.T. Indoor–outdoor relationships of particulate matter and Monn Ch. Exposure assessment of air pollutants: a review on spatial nitrogen oxides under different outdoor meteorological conditions. heterogenity and indoor/outdoor/personal exposure to suspended Atmos Environ 2002: 36: 1543–1551. particulate matter, nitrogen dioxide and ozone. Atmos Environ 2001: Cyrys J., Heinrich J., Hoek G., Meliefste K., Lewne´ M., and Gehring 35: 1–32. U., et al. Comparison between different traffic related particle Morawska L., He C., Hitchins J., Gilbert D., and Parappukkaran S.

indicators: elemental carbon (EC), PM2.5 mass and absorbance. The relationshipbetween indoor and outdoor airborne particles in J Expos Anal Environ Epidemiol 2003: 13: 134–143. the residential environment. Atmos Environ 2001: 35: 3463–3473.

282 Journal of Exposure Analysis and Environmental Epidemiology (2004) 14(4) Indoor/outdoor ratios of particulate air pollutants Cyrys et al.

Oberdo¨ rster G., Gelein R.M., Ferin J., and Weiss B. Association of US-EPA. Air Quality Criteria for Particulate Matter. Vol. II. EPA 600/ particulate air pollution and acute mortality: involvement of ultrafine P-99/002bB. 1-3-2001. EPA, Research Triangle Park, NC, 1999. particles? Inhal Toxicol 1995: 7: 111–124. VDI 4300 Blatt 7. Messen von Innenraumluftverunreinigungen F O¨ zkaynak H., Xue J., Spengler J., Wallace L., Pellizzari E., and Jenkins Bestimmung der Luftwechselzahl in Innenraümen.BeuthVerlag, P. Personal exposure to airborne particles and metals: results from Berlin, 2001. the Particle TEAM study in Riverside, California. JExposAnal Vette A.F., Rea A.W., Lawless P.A., Rodes C.E., Evans G., and Environ Epidemiol 1996: 6(1): 57–78. Highsmith V.R., et al. Characterization of indoor–outdoor aerosol Penttinen P., Timonen K.L., Tiittanen P., Mirme A., Ruuskanen J., and concentration relationships during the fresno PM exposure studies. Pekkanen J. Number concentration and size of particles in urban air: Aerosol Sci Technol 2001: 34: 118–126. effects on spirometric lung function in adult asthmatic subjects. Wallace L. Indoor particles: a revive. J Air Waste Manage Assoc 1996: Environ Health Perspect 2001: 109(4): 319–323. 46: 98–128. Peters A., Wichmann H.E., Tuch T., Heinrich J., and Heyder J. WHO, World Health Organization. Air Quality Guidelines for Europe F Respiratory effects are associated with the number of Second Edition. WHO Regional Publications, European Series No. ultrafine particles. Am J Respir Crit Care Med 1997: 155: 1376– 91:1-273, 2000. 1383. Wichmann H.E., Gerken M., Wellmann J., Kreuzer M., Kreienbrock Schneider T., Sundell J., Bischof W., Bohgard M., Cherrie J.W., L., and Keller G., et al. Lungenkrebsrisiko durch Radon in der Clausen P.A., Dreborg S., Kildes J., Kjærgaard S.K., Lvik M., Bundesrepublik Deutschland (Ost) F Thu¨ringen und Sachsen.[in Pasanen P., and Skyberg K. EuroPart. Airborne particles in the German] Ecomed Verlag, Landsberg am Lech, 1999. indoor environment. A European interdisciplinary review of scientific Wichmann H.E., and Peters A. Epidemiological evidence of the effects evidence on associations between exposure to particles in buildings of ultrafine particle exposure. Philos Trans R Soc London 2000: 358: and health effects. Indoor Air 2003: 13: 38–48. 2751–2769. Seaton A., MacNee W., Donaldson K., and Godden D. Particulate air Wichmann H.E., Spix C., Tuch T., Wo¨ lke G., Peters A., and Heinrich pollution and acute health effects. Lancet 1995: 345: 176–178. J., et al. Daily mortality and fine and ultrafine particles in Erfurt, Thatcher T.L., and Layton D.W. Deposition, resuspension, and Germany Part I: role of particle number and particle mass. Health penetration of particles within a residence. Atmos Environ 1995: Effects Institute Report No. 98, 2000. 29(13): 1487–1497. Wichmann H.E., Cyrys J., Sto¨ lzel M., Spix C., Wittmaack K., and Tu K., and Knutson E.O. Indoor outdoor aerosol measurements for Tuch T., et al. Sources and Elemental Composition of Ambient two residential buildings in New Jersey. Aerosol Sci Technol 1988: 9: Particles in Erfurt, Germany. Ecomed Verlag, Landsberg am Lech, 71–82. 2002.

Journal of Exposure Analysis and Environmental Epidemiology (2004) 14(4) 283