Report on infiltration and exposure model with software prototypes

(1)

Life+ EXPAH (Contract ENV/IT/000082; 2009)

Report on infiltration and exposure model with software prototypes

Pasi Lipponen, Riikka Sorjamaa and Otto Hänninen National Institute for Health and Welfare (THL),

Department of Environmental Health, Kuopio, P.O. Box 95, Finland

Abstract

Polycyclic aromatic hydrocarbons (PAHs) are known to be harmful to human health e.g. by causing lung cancer (Vineis et al, 2004). Spatial variability of the emissions, atmospheric dispersion, infiltration of the particles from outdoor to indoors, and population time-activity defines the exposures. Since the time spent indoors is typically 80-90%, exceeding times spent in traffic and outdoors, indoor exposures and thus particle infiltration from outdoors is a key issue (e.g. Hänninen et al, 2004). Infiltration from outdoors to indoors, as well as lung deposition, is largely affected by particle size (Hänninen et al, 2012). Recently, accumulation mode particle deposition to alveolar region has been noticed to be the largest contributor to the PAH uptake (Zhang et al, 2012).

Health effects of air pollutants are presumably caused by the actual molecules entering the body and causing a toxicological dose. Therefore exposure assessment is in a central role in epidemiology and risk assessment.

Exposures are modified by times spent in different microenvironments including indoors, outdoors and in traffic. The relative role of various microenvironments is determined by the sources of the pollutants in question. Thus identification of the main sources of exposures is a central element also in EXPAH.

THL actions 5.1, 5.2 and 5.3 studies the PAH indoor-outdoor relationships, infiltration and exposure in different microenvironments in different seasons by taking into account the particle size distributions. As a result, quantitative model for PAH infiltration accounting for particle size distribution and an exposure model which integrates the outdoor PAH concentrations and PAH infiltration indoors with population time-activity, including time spent indoors and in traffic to estimate the actual exposure levels generated by the emissions and atmospheric dispersion processes were developed.

In addition, particle size dependent lung deposition approach is presented and it has been used to demonstrate the effect of particle infiltration (through building envelope) on deposited mass in different parts of human respiratory tract.

(2)

EXECUTIVE SUMMARY

Urban populations spend substantial fraction of their time indoors, where the outdoor particulate matter air pollution levels are modified by the buildings. Indoor sources may potentially raise the indoor exposures substantially higher than outdoors. On the other hand, outdoor particles are partly removed from the indoor air.

The aim of this report was to investigate the role of these processes for population exposures to PAH compounds in Rome using measurement data collected within the EXPAH study and a previously developed aerosol model.

Statistical analyses of the simultaneous indoor and outdoor measurements of 11 PAH compounds showed that substantial indoor sources do not exist in Rome. The population exposure levels of both children and elderly were dominated by PAHs originating from the outdoor air during the heating season. Time spent in traffic was not associated with substantially raised exposures, suggesting that while traffic may be a minor source of PAH compounds, the exposures originate mostly from non-traffic sources.

Previous studies suggest that the infiltration of outdoor particles indoors is lower during winter. The ca. 140 simulatneous indoor and outdoor PAH measurements conducted in EXPAH study were not sufficient to capture such seasonal variation in most cases due to larger variation between individual sites and measurement weeks.

However, existence of seasonal trends can not be excluded.

Indoor exposures were 20-50% lower than outdoor concentrations. Thus time-activity, e.g. time spent outdoors, affects the individual exposures. Nevertheless, overall the population exposures were dominated by the winter season when the indoor and outdoor levels were an order of magnitude higher than in the summer.

The aerosol model allowed to estimate the particle size dependence of the infiltration as well as the respiratory tract uptake of particles and PAH compounds by region. These tools allow for interesting refinement of exposure and dose estimates in future studies.

(4)

1 Introduction

The aim of EU-LIFE+ project EXPAH is to identify and to quantify population exposure among children and elderly people to polycyclic aromatic hydrocarbons (PAHs) content in particulate matter in highly urbanized areas and to assess the impact on human health, in order to support environmental policy and regulation in this field. The project is based on an integrated approach where measurements, modelling techniques and epidemiologic investigations will be used to obtain estimation maps of population exposure to PAHs, to identify key determinants of high exposures including time-activity and locations in relation to the sources and to estimate potential health effects on the target population.

The current report describes the achievements of Actions 5.1, 5.2 and 5.3, lead by National Institute for Health and Welfare (THL). The aim of these actions is to utilize measurement data from indoor and outdoor environments to develop an infiltration model to quantitatively describe the exposure levels in indoor environments originating from outdoors. As urban populations spend substantial majority of their time indoors, the modification of exposures by buildings forms an central factor in understanding the formation of the actual exposures when accounting for actual time activities of the population members. The infiltration model is a necessary component in risk analysis and will be integrated with ambient air quality modelling in Action 5.4.

The EXPAH Actions reported here are:

Action 5.1: Statistical analysis of indoor and outdoor PM_2.5, PAH compounds and corresponding I/O ratios.

Aim of this action is a quantitative characterization of PAH indoor-outdoor relationships. Schedule: 09/2012 – 03/2013.

Action 5.2: Developing of a microenvironment infiltration model. Target is to develop a quantitative model for PAH infiltration, accounting for particle size distribution and different types of microenvironments. The model describes the expected indoor concentration of PAH compounds when the type of the building and microenvironment and the outdoor concentrations are known. Schedule: 01/2013 – 06/2013.

Action 5.3: Developing of an exposure model. Target is to develop an exposure model which integrates the outdoor PAH concentrations and PAH infiltration indoors with population time-activity, including time spent indoors and in traffic to estimate the actual exposure levels generated by the emissions and atmospheric dispersion processes. Schedule: 03/2013 – 09/2013.

(5)

1.1 Related EXPAH Actions and input data

1.1.1 PAH Emission inventories for identification of major sources

Emission inventories were collected and updated in actions 4.1 and 4.3-4.4 and were used here for the preliminary identification of main sources of ambient PAH compounds. The inventories were reported in

Radice P, Smith P, Costa MP, D’Allura A, Pozzi C, Nanni A, Finardi S, 2012. EXPAH Actions 4.3 - 4.4:

Calculation and integration of traffic emissions with the updated Lazio Region inventory. Spatial, temporal and chemical disaggregation of the emission inventory. 41 pp.

http://www.ispesl.it/expah/documenti/R2012-05_ARIANET_EXPAH_A4.3-4_final.pdf (accessed 2013- 07-04)

Radice P, Finardi S, 2011. EXPAH - -- - ACTION 4.1: Collection of raw emission inventories and their upgrading emission inventories and their upgrading emission inventories and their upgrading emission inventories and their upgrading. 15 pp. http://www.ispesl.it/expah/documenti/R2011-

13_ARIANET_EXPAH_A4.1.pdf (accessed 2013-07-04)

1.1.2 Indoor and outdoor PAH measurements

PM_2.5 and polycyclic aromatic hydrocarbon (PAH) mass concentration and particle size distribution measurements were carried out in Rome, Italy, within timeframe of summer 2011 – summer 2012 in EXPAH Action 3.3 lead by CNR. Measurements were reported in:

Cecinato A, Romagnoli P, Balducci C, Guerriero E, Perilli M, Vichi F Imperiali A, Perrino C, Tofful L, Sargolini T, Catrambone M, Dalla Torre S, Rantica E, Gherardi M, Gatto MP, Gordiani N, L’Episcopo N, Gariazzo C, Sacco F, Sozzi R, Troiano F, Barbini F, Gargaruti C, Bolignano A., 2013. EXPAH Action 3.3: Extended Technical Report on Indoor/Outdoor monitoring of PAHs, PM_2.5 and its chemical components with ancillary measurements of gaseous toxicants in the frame of the

EXPAH Project. 214 pp.

http://www.ispesl.it/expah/documenti/Technical_Report_CNR_INAIL_2012h%20finale.pdf (accessed 2013-07-04)

1.1.3 Population time-activity measurement diaries

Time-activity data was collected in Action 3.1 from 483 children of 7-8 years and 707 elderly subjects (65-85 years old). The results were reported in:

EXPAH Action 3.1 Time-activity pattern of children and elderly in Rome. 37 pp.

http://www.ispesl.it/expah/documenti/Survey on children and elderly people time activity in Rome action 3.1. finalreport.pdf (accessed 2013-07-04)

(6)

1.2 Grouping of PAH compounds

Polycyclic aromatic hydrocarbons (PAHs) are one of the most widespread organic pollutants. They are readily present in fossil fuels, but also formed by incomplete combustion of carbon-containing fuels that do not contain PAH compounds themselves, such as wood, fat, tobacco, and incense. Different types of combustion yield different distributions of PAHs in both relative amounts of individual PAHs. Coal burning produces a different mixture than motor-fuel combustion or a forest fire, and thus the composition of PAH mixture may contain information allowing for identification of the sources. This was also one of the objectives of the current work, where positive matrix factorization (PMF) was attempted for identification of PAH sources.

PAH compounds consist of fused aromatic rings and do not contain heteroatoms or carry substituents. are lipophilic, meaning they mix more easily with oil than water. The larger compounds are less water-soluble and less volatile. Because of these properties, PAHs in the environment are found primarily in soil, sediment, and oily substances, as opposed to in water or air. However, they are also a component of concern in particulate matter suspended in air.

The toxicity of PAHs is structure-dependent. Isomers (PAHs with the same formula and number of rings) can vary from being nontoxic to extremely toxic. One PAH compound, benzo[a]pyrene, is notable for being the first chemical carcinogen to be discovered (and is one of many carcinogens found in cigarette smoke). Eight PAH compounds have been identified as carcinogenic. In various studies a range of toxicity equivalency factors have been proposed to characterize the carcinogenicity of individual com pounds. Benzo(a)pyrene has been selected as the reference point (TEF(BaP)=1.0). Table 1 lists TEF values proposed in a number of assessments (EC, 2001)

Although the health effects of individual PAHs are not exactly alike, the selected 18 PAHs listed in Table 2 are commonly considered (e.g. by Agency for Toxic Substances and Disease Registry (ATSDR)).

Table 1, Carcinogenic toxicity equivalent factors proposed for 8 PAH compounds in various assessments.

Assessment BaP BaA BbF BjF BkF CH DBA IP

California EPA 1998 (Collins)

1 0.1 0.1 0.1 0.1 0.01 0.1

Canada 1994 (Meek)

1 0.06 0.05 0.04 0.12

Chu and Chen 1984

1 0.013 0.08 0.04 0.001 0.69 0.017

Clement 1986

1 0.145 0.14 0.061 0.066 0.0044 1.11 0.232

Larsen and Larsen

1998

1 0.005 0.1 0.05 0.05 0.03 1.1 0.1

Nisbet and LaGoy 1992

1 0.1 0.1 0.1 0.01 5 0.1

Ontario 1997 (Muller)

1 0.014 0.11 0.045 0.037 0.026 0.89 0.067

The Netherlands 1989

1 0-0.04 0.03-

0.09 0.05- 0.89

0-0.08

(7)

Table 2. List of carcinogenic and other PAH compounds. Those measured in the EXPAH study are identified with the EXPAH code.

Group EXPAH PAH Compound Net formula Melting TEF^b

Code^a

°C g/mol

Carcinogenic PAHs

1 BaP

benzo[a]pyrene C20H12 179 252 1

2 BaA

benz[a]anthracene C18H12 158 228 0.1

3 BbF

benzo[b]fluoranthene C20H12 168 252 0.1

4 BjF

benzo[j]fluoranthene C20H12 165 252 0.1

5 BkF

benzo[k]fluoranthene C20H12 217 252 0.1

6 CH

chrysene C18H12 254 228 0.01

7 DBA

dibenz(a,h)anthracene C22H14 262 278 (0.1-5)

^c

8 IP

indeno(1,2,3-cd)pyrene 0.1

Non-carcinogenic PAHs listed on ATSDR 18-substances list

9

acenaphthene C12H10 93.4 145

n/a

10

acenaphthylene C12H8 92 152

n/a

11

anthracene C14H10 218 178

n/a

12 BeP

benzo[e]pyrene C20H12 252

n/a

13 BPE

benzo[ghi]perylene C22H12 278 276

n/a

14

coronene C24H12 438 300

n/a

15

fluoranthene C16H10 111 202

n/a

16

fluorene C13H10 116 166

n/a

17

phenanthrene C14H10 101 178

n/a

18

pyrene C16H10 145 202

n/a

Other common PAH compounds

19

naphthalene C10H8 80 128

n/a

20 PE

perylene C20H12 276 252

n/a

a compounds measured in the EXPAH field campaigne

b carcinogenic toxicity equivalent factors proposed by Cal-EPA, 1998

c variable values have been proposed for DBA

(8)

2 Identification of main sources of PAH exposures

Understanding of main sources of pollutants is central in exposure assessment. Behavioral and time-activity parameters associated with high exposures are essential for accurate and meaningful exposure analysis.

According to the emission inventories reviewed and updated in EXPAH (Radice et al., 2011, 2012), the expected main sources of PAH exposures are residential heating, waste incineration and traffic (Figure 1).

Residential heating is estimated to be responsible for 80-90 % of total known PAH emissions, followed by waste inceneration (10-15%).

Figure 1. Main sources of PAH. Source: EXPAH Emission inventory (Radice et al., 2012)

This assumption of residential biomass combustion as the main source of PAH compounds in Rome area was taken as the starting hypotheses in the data analyses and was confirmed throughout the analysis. Here we just want to highlight this by showing a temporal overview of the observed outdoor and indoor total ΣPAH levels (Figure 2). Winter levels observed in December 2011-February 2012 were more than an order of magnitude higher than the summertime PAH levels.

Waste incinerator sources were attempted to be identified from the observed outdoor concentration data by using inverse distance method and combining the locations of the 10 identified point sources within the Rome metropolitan area with their estimated source strengths. The point source PAH emission strengths varied between 0.1 – 9.24 kg/a. No correlation was observed between the measurement data and these point source descriptors.

Quantification of traffic source contribution to observed outdoor PAH levels was tested by using questionnaire data from the measurement sites including traffic density estimates (very low, low, medium, high) and attempting to associate these with the observed levels. The results remained statistically insignificant.

Further, an advance statistical method, positive matrix factorization (PMF) was applied to the outdoor measurements in an attempt to identify source categories. However, the limited size of the dataset did not allow for achieving statistically significant results (Appendix 1).

(9)

Figure 2. Outdoor and indoor average PAH mass concentrations in different seasons between April 2011 and August 2012.

3 Indoor-outdoor analysis

Development of the infiltration model was conducted in two phases. The first phase was based on direct statistical analysis of the microenvironmental indoor and outdoor measurements conducted in Action 3.3. These results were then used as an input to the application of a mass-balance model to quantify the key parameters needed for exposure modelling.

3.1 Mass-balance modelling of infiltration rates

Mass-balance approach has been previously successfully applied on PM_2.5 concentrations and validated against large population-based datasets from a number of European countries (Hänninen et al., 2004, 2011). Here the same mass-balance –based approach was applied to the PAH measurements. The following section first summarizes the physical processes as mass-balance based equations and then the regression analyses used to estimate the mass-balance parameters.

In the general form, mass-balance can be presented as (Dockery and Spengler, 1981):

Eq 1

where P = penetration efficiency, a = air exchange rate, k = decay rate indoors, Q = source strength and V = interior volume of the building. The latter term in the equation can be ignored for 24 hour or longer sampling periods.

Equation can be written shortly according to Hänninen et al. (2004) as Eq 2

0 5 10 15 20 25 30 35 40 45 50

04-01 05-01 06-01 07-01 08-01 08-31 10-01 11-01 12-01 01-01 01-31 03-02 04-01 05-02 06-02 07-02 08-02 09-01

PAH concentration (ng m-3)

Date in 2011-2012 (mm-dd)

ΣPAH indoor ΣPAH outdoor

(10)

where Ci is total indoor concentration, Cig is concentration of indoor generated particles and Cai is concentration of ambient particles that has infiltrated indoors. Concentration of indoor generated particles is therefore

Eq 3

and source term can be expressed as Eq 4

Concentration of outdoor particles which has infiltrated indoors is Eq 5

and infiltration factor can be then written as Eq 6

Hänninen proposed to estimate air exchange rate (a) is from the infiltration equation as Eq 7

The earlier application based on the availability of elemental sulphur, used as a marker of outdoor originating aerosol and therefore allowing for independent estimation of Finf; here the Finf is estimated as a population mean using regression analysis, and therefore also the air exchange rate estimates represent typical population values and contain larger uncertainties.

(11)

3.2 Regression analysis for mean infiltration rates

This section summarizes the regression analyses of infiltration rates by microenvironment type, season and pollutant compound. Seasonal variation in case of outdoor PM_2.5 mass concentration was expected based on previous studies. As will be seen in the detailed presentation, however, seasonal variation of indoor-to-outdoor ratio of PAH concentration is seen only in very few cases. Overall, the measured data set is not of sufficient size to support accurate estimation of the seasonal variation of PAH infiltration even though based on the theoretical framework and previous studies on PM_2.5 mass concentrations such variation could be expected.

Simultaneous indoor and outdoor measurements were conducted in 2011-2012 in four selected microenvironment types to characterize population exposures in different types of daily environments. The included microenvironments were homes (9+1 different sites), offices (3+1), schools (7) and vehicles (2 cars and selected bus line) (Table 3).

Table 3. Microenvironment types and number of sites in each category with simultaneous indoor-outdoor measurements of PAH compounds.

Microenvironment type Measurement sites

Home ⁹^a

Office ³^b

School ⁷

Vehicles ³^c

a HTR home was substituted by HPR at the summer campaign; thus total of 10 residences were involved.

b Additional measurements were conducted at institute location MLI

c Two cars and one bus route

Summary of the obtained numbers of successful concentration measurement and the observed PAH compound, PM2.5 and EC/OC concentrations are given in Table 4. Various groupings were necessary. First, in the chromatographic analyses peaks for BbF, BjF and BkF could be identified variably so that in substantial fraction of the samples 2 or 3 of these peaks were overlapping and the results were therefore obtained only for the sum of the compounds. For other samples each individual peak could be identified. From the point of view of carcinogenic toxicity this poses only a limited problem, as the TEF-values for all three compounds are similar.

For the exposure modelling conducted by INAIL, a summed ―PAHs‖ variable was constructed, including BaP, IP and sum of B(b,j,k)F. For the consideration of carcinogenic potency, cPAH-variable was created in which CAL-EPA TEF-factors have been used to weight the concentrations of individual carcinogenic PAH compounds (see Table 1 and Table 2).

(12)

Table 4. Summary of the measurement data and compound groupings used in the analyses.

Group Compound Unit Outdoor measurements Indoor measurements

mean sd n mean sd n

Carcinogenic

1 BaP ng m^-3 0.56 0.80 145 0.41 0.55 146

2 BaA ng m^-3 0.31 0.40 145 0.17 0.22 146

3 BbF ng m^-3 0.90 1.24 48 0.51 0.70 54

4 BjF ng m^-3 0.09 0.02 24 0.08 0.03 24

5 BkF ng m^-3 0.06 0.02 27 0.06 0.02 24

6 CH ng m^-3 0.71 1.04 55 0.33 0.35 56

7 DBA ng m^-3 0.10 0.14 130 0.07 0.08 132

8 IP ng m^-3 0.62 0.84 145 0.46 0.57 146

Non-carcinogenic

9 BPE ng m^-3 0.69 0.96 145 0.53 0.67 146

10 PE ng m^-3 0.35 0.55 50 0.16 0.19 48

11 BeP ng m^-3 0.68 1.00 55 0.55 0.64 56

Combined peaks

12 BbjF ng m^-3 0.16 0.10 3

13 BjkF ng m^-3 2.12 2.25 24 1.18 0.93 30

14 BbjkF ng m^-3 0.91 1.29 94 0.59 0.65 92

Summed

cPAHs^a ng m^-3 3.35 2.08

ΣC×TEF^b ng m^-3 0.76 0.54

sPAH^c ng m^-3 7.55 13.46 11 2.34 3.59 12

ΣPAH^d ng m^-3 4.15 6.50 146 2.85 3.77 146

Ratios

BaP/ΣPAH % 13 % 5 % 145 14 % 5 % 146

BaP/PAHs % 5 % 1 % 2 9 % 2 % 5

BaP/BeP % 13 % 30 % 49 15 % 31 % 50

Particulate matter air pollution

PM25 µg m-3 22.5 15.5 140 23.0 13.2 143

EC µg m-3 2.2 1.3 28 1.8 1.5 28

OC µg m-3 6.7 6.0 28 6.6 3.0 28

Calculated sums^e

ΣBbjF ng m-3 0.31 0.82 146 0.20 0.49 146

ΣBbjkF ng m-3 0.32 0.82 146 0.21 0.49 146

ΣBjkF ng m-3 0.03 0.06 146 0.02 0.05 146

a sum of carcinogenic PAHs (1-8)

b sum of Concentration × TEF for carcinogenic PAHs (using Cal-EPA TEF factors)

c sum of BaP, IP, B(b,j,k)F

d sum of compounds (1-11) (including combined peaks when individual peaks were not observed)

e calculated for samples for which individual peaks were available for comparison with the combined peaks data

Infiltration factors (Finf) for PM_2.5 and bound components were estimated from a filtered data set, where indoor- outdoor data points with Ci > Co have been excluded to avoid influence on the Finf estimated as ß1 from the regression (Figure 3).

(13)

Figure 3. Regression estimate for indoor source contribution (ß0) has been estimated from the data set containing all data points (i.e. also the Ci>Co points shown in red). In the estimation of Finf (ß1) these points have been excluded.

Statistical significance is tested for

a) ß0 = 0, i.e. support for the existence of indoor sources in the given microenvironment, season, and pollutant, b) ß1 differs between microenvironment specific and combined data.

Four seasons were defined as Dec-Feb and forward, but most consistent results indicating seasonal effects were obtained using only heating vs. non-heating season. Heating season was defined as from November to March.

Numerical results from the statistical testing of differences in ß0 and ß1 between microenvironments and compounds are presented in Table 5 and between seasons in Table 6 followed by more visual graphical presentation of a subset of these analyses. The regression estimate for indoor source contribution (ß0) has been estimated from the data set containing all data points (i.e. also the Ci>Co points). In the estimation of Finf (ß1) these points have been excluded. Error of the regressions is estimated by means of Standard Error (SE) (Lane, D.M., Online Book).

0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0

0.0 20.0 40.0 60.0 80.0

Indoor concentration

Outdoor concentration

PM2.5 (µg m-3) Site=All, µEnv=All, Season=All (n=75+64)

PM2.5 1:1

Ci≤Co (for ß1) Ci>Co (in ß0) y=0.63x + 2.38 (R²=66%, n=75) Cig (ß0, all data) 95%CI (ß1=0.52) 95%CI (ß1=0.73)

(14)

Table 5. Comparison of Finf (ß1, Ci<Co dataset) and Cig (ß0, all data) for PM2.5, BaP, ΣPAH, EC and OC.

.

Table 6. Comparison of seasonal differences in Finf (ß1, Ci<Co dataset) in homes.

Run set 1: by microenvironment Regression results for Ci<Co dataset Unfiltered dataset Ci<Co -dataset

Finf(Ci<Co dataset) Cig (Full dataset) p(ß0=0) p(Finf(µe)=Finf(All))

CompoundSite µEnv Season ß1 SE1 R2 n ß0 SE0 R2 n

1 PM2.5 all all all 0.627 0.053 65.8 % 75 1.704 1.720 58.5 % 139 0.161 . 0.500 .

2 PM2.5 all home all 0.647 0.066 78.3 % 29 4.382 2.260 53.1 % 53 0.026 ** 0.432 .

3 PM2.5 all school all 0.710 0.084 76.4 % 24 7.297 1.894 49.2 % 42 0.000 *** 0.272 .

4 PM2.5 all office all 0.272 0.135 36.6 % 9 10.249 3.136 16.7 % 15 0.001 *** 0.030 **

5 PM2.5 all traffic all 0.835 0.207 59.6 % 13 -12.102 7.271 67.1 % 29 0.048 ** 0.212 .

6 PM2.5 bus all all 0.586 0.355 57.6 % 4 -11.011 5.821 89.0 % 16 0.029 ** 0.460 .

7 BaP all all all 0.561 0.024 82.3 % 115 0.066 0.027 77.1 % 145 0.008 *** 0.500 .

8 BaP all home all 0.550 0.036 85.3 % 43 0.056 0.032 79.3 % 53 0.042 ** 0.433 .

9 BaP all school all 0.629 0.057 82.8 % 27 -0.027 0.072 81.9 % 33 0.356 . 0.203 .

10 BaP all office all 0.479 0.060 75.4 % 23 0.095 0.080 71.6 % 26 0.118 . 0.165 .

11 BaP all traffic all 0.639 0.037 93.6 % 22 0.119 0.047 78.4 % 33 0.006 *** 0.102 .

12 BaP bus all all 0.934 0.027 99.2 % 12 -0.004 0.057 84.1 % 20 0.472 . 0.000 ***

13ΣPAH all all all 0.500 0.025 76.7 % 121 0.778 0.189 74.3 % 146 0.000 *** 0.500 .

14ΣPAH all home all 0.529 0.031 86.6 % 48 0.608 0.262 81.1 % 54 0.010 ** 0.305 .

15ΣPAH all school all 0.773 0.056 89.3 % 25 -0.189 0.378 89.7 % 33 0.309 . 0.000 ***

16ΣPAH all office all 0.305 0.040 72.9 % 24 1.216 0.423 70.8 % 26 0.002 *** 0.001 ***

17ΣPAH all traffic all 0.626 0.045 89.7 % 24 0.590 0.232 80.1 % 33 0.006 *** 0.038 **

18ΣPAH bus all all 0.913 0.036 98.2 % 14 0.022 0.266 87.4 % 20 0.467 . 0.000 ***

19 PAH4 all all all 0.555 0.020 88.1 % 111 0.185 0.062 82.3 % 146 0.001 *** 0.500 .

20 PAH4 all home all 0.646 0.018 96.7 % 44 0.146 0.073 89.1 % 54 0.022 ** 0.008 ***

21 PAH4 all school all 0.633 0.034 91.9 % 33 -0.027 0.131 88.7 % 42 0.417 . 0.071 *

22 PAH4 all office all 0.383 0.048 81.9 % 16 0.259 0.187 82.4 % 17 0.083 * 0.005 ***

23 PAH4 all traffic all 0.642 0.047 92.0 % 18 0.202 0.094 77.3 % 33 0.015 ** 0.098 *

24 PAH4 bus all all 0.969 0.024 99.6 % 8 -0.025 0.093 89.4 % 20 0.393 . 0.000 ***

25 EC all all all 0.717 0.080 79.2 % 23 -0.345 0.238 81.2 % 28 0.074 * 0.500 .

26 EC all home all

27 EC all school all 0.961 0.086 90.5 % 15 -0.623 0.234 90.7 % 20 0.004 *** 0.071 *

28 EC all office all 0.485 0.056 92.6 % 8 0.212 0.126 92.6 % 8 0.046 ** 0.044 **

29 EC all traffic all

30 EC bus all all

31 OC all all all 0.537 0.040 95.8 % 10 3.545 0.370 82.1 % 28 0.000 *** 0.500 .

32 OC all home all

33 OC all school all 0.563 0.024 98.9 % 8 3.078 0.386 89.4 % 20 0.000 *** 0.343 .

34 OC all office all 2 5.559 0.580 45.4 % 8 0.000 *** . .

35 OC all traffic all

36 OC bus all all

Run set 2: home by seasons Regression results for Ci<Co dataset Unfiltered dataset Ci<Co -dataset

Finf(Ci<Co dataset) Cig (Full dataset) p(ß0=0) p(Finf(season)<>Finf(All))

CompoundSite µEnv Season ß1 SE1 R2 n ß0 SE0 R2 n

37 PM2.5 all home all 0.647 0.066 78.3 % 29 4.382 2.260 53.1 % 53 0.026 ** 0.500 .

38 PM2.5 all home heating 0.582 0.096 94.9 % 4 14.501 10.575 33.3 % 11 0.085 * 0.341 .

39 PM2.5 all home nonheating 0.324 0.068 49.8 % 25 12.272 2.138 5.8 % 42 0.000 *** 0.008 ***

40 PM2.5 all home winter 0.582 0.096 94.9 % 4 14.501 10.575 33.3 % 11 0.085 * 0.341 .

41 PM2.5 all home spring 0.617 0.097 67.9 % 21 8.899 2.325 16.9 % 34 0.000 *** 0.427 .

42 PM2.5 all home summer 0.216 0.113 64.8 % 4 18.227 5.707 0.2 % 8 0.001 *** 0.008 ***

43 PM2.5 all home fall 0 0

44 BaP all home all 0.550 0.036 85.3 % 43 0.056 0.032 79.3 % 53 0.042 ** 0.500 .

45 BaP all home heating 0.433 0.155 56.5 % 8 0.457 0.217 50.5 % 11 0.017 ** 0.269 .

46 BaP all home nonheating 0.638 0.050 83.3 % 35 0.016 0.007 79.0 % 42 0.011 ** 0.154 .

47 BaP all home winter 0.433 0.155 56.5 % 8 0.457 0.217 50.5 % 11 0.017 ** 0.269 .

48 BaP all home spring 0.640 0.057 81.7 % 30 0.015 0.009 78.2 % 34 0.049 ** 0.168 .

49 BaP all home summer 0.868 0.136 93.2 % 5 -0.005 0.018 67.9 % 8 0.389 . 0.032 **

50 BaP all home fall

51ΣPAH all home all 0.529 0.031 86.6 % 48 0.608 0.262 81.1 % 54 0.010 ** 0.500 .

52ΣPAH all home heating 0.348 0.099 58.1 % 11 4.899 1.410 52.9 % 12 0.000 *** 0.081 *

53ΣPAH all home nonheating 0.618 0.048 82.8 % 37 0.148 0.055 76.5 % 42 0.003 *** 0.128 .

54ΣPAH all home winter 0.348 0.099 58.1 % 11 4.899 1.410 52.9 % 12 0.000 *** 0.081 *

55ΣPAH all home spring 0.618 0.049 84.2 % 32 0.108 0.055 81.5 % 34 0.024 ** 0.131 .

56ΣPAH all home summer 0.727 0.068 97.5 % 5 0.242 0.148 73.4 % 8 0.051 * 0.022 **

57ΣPAH all home fall

58 PAH4 all home all 0.646 0.018 96.7 % 44 0.146 0.073 89.1 % 54 0.022 ** 0.500 .

59 PAH4 all home heating 0.614 0.098 86.8 % 8 1.187 0.403 76.8 % 12 0.002 *** 0.390 .

60 PAH4 all home nonheating 0.704 0.068 75.8 % 36 0.040 0.026 69.6 % 42 0.064 * 0.251 .

61 PAH4 all home winter 0.614 0.098 86.8 % 8 1.187 0.403 76.8 % 12 0.002 *** 0.390 .

62 PAH4 all home spring 0.673 0.070 76.5 % 30 0.032 0.027 72.3 % 34 0.116 . 0.382 .

63 PAH4 all home summer 0.900 0.151 89.9 % 6 0.115 0.071 67.0 % 8 0.053 * 0.066 *

64 PAH4 all home fall 0 0

(15)

The results show that indoor sources can be expected in most cases, but that the magnitude of the source is limited except for PM_2.5 (column Cig in Table 5). However, the data supports only microenvironment specific Finf for PM_2.5 in offices, BaP for bus, ΣPAH in all microenvironments (Table 5).

Comparison of the seasonal differences in residential infiltration in Table 6 interestingly suggests that during non-heating season the infiltration would be lower than during the heating season; affected especially summer when PM_2.5 Finf is only 0.2. This is due to the limited data set size and outlying points. In contrast, during the summer the BaP Finf is 0.8 and statistically significantly higher than the all-season estimate. For ΣPAH estimates summer is again higher than all season combined + there is weak evidence for difference for heating season and winter (lower infiltration as could be expected).

Figure 4 reveals that there are only 8 measurements of PM_2.5 indoor –outdoor relationship for homes in the summer season and that in four cases of these there is clear indication of indoor sources (Ci>Co). When the Finf is estimated from the remaining four Ci<Co data points, we see that the slope is strongly determined by the one high outdoor level outlier and that for the remaining three other Ci<Co data points Finf is very close to unity. It has to be concluded that the summer Finf estimate for PM_2.5 is not reliable due to the limited data set size.

Figure 4. PM2.5 infiltration data for homes in summer.

Differences between all microenvironment types are suggested for the annual infiltration rate of ΣPAH and for BaP in busses.

There is limited evidence in the data on statistically significant seasonal differences in infiltration. For BaP and ΣPAH the infiltration seems to be slightly higher during the non-heating season

As a result, it might be better to use pooled estimated for Finf values and to use expert judgment adjustment for expected differences, especially seasons, when the ventilation patterns are known to vary. Too detailed parametrization of the estimates from the EXPAH measurement data set may put too much weight on limited number of measurements.

This approach suggests that seasonal differences are ignored and differences between microenvironments are included. However, the former approach is not in line with the findings presented in our previous review and meta-analysis (Hänninen et al., 2011).

0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 45.0 50.0

0.0 10.0 20.0 30.0 40.0 50.0

Indoor concentration

PM2.5 (µg m-3) Site=all, µEnv=home, Season=summer

PM2.5 1:1

Ci≤Co (for ß1) Ci>Co (in ß0)

y=0.22x + 9.42 (R²=65%, n=4) Cig (ß0, all data)

(16)

3.2.1 PAH and B(a)P seasonal differences in homes

It can be seen in Figure 5 that in the heating season case there seems to be two subsets in the data; outdoor levels around 10 ng m^-3 suggest high infiltration close to unity, while higher values around 20 and above suggest lower infiltration. Even though there are no obvious indoor sources in the data (Ci>Co), the ß0 estimate is 5 ng m^-3. In the non-heating season the data cloud is more homogeneous with the slope at almost a double value at 0.62.

Figure 5. Comparison of ΣPAH Finf by heating vs non-heating season in homes.

Recommendation for discussion: use 0.62 for summer and either bimodal 0.4+0.6 or unimodal 0.55 (estimate with ß0 forced to zero).

Similarly, comparison of heating vs non-heating season for BaP suggests that there is some evidence on lower infiltration during the 5 winter months. In this case the ß1 estimates (0.43 and 0.64) seem usable directly (winter estimate is slightly tilted down; estimate forced to origin would be 0.55).

Figure 6. Comparison of BaP Finf by heating vs non-heating season in homes.

0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0

0.0 10.0 20.0 30.0 40.0

ΣPAH (ng m-3) Site=all, µEnv=home, Season=Heating

ΣPAH

1:1

Ci≤Co (for ß1)

Ci>Co (in ß0)

0.0 0.5 1.0 1.5 2.0 2.5 3.0

0.0 1.0 2.0 3.0

Indoor concentration

ΣPAH (ng m-3) Site=all, µEnv=home, Season=Nonheating

ΣPAH

1:1

Ci≤Co (for ß1)

Ci>Co (in ß0)

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

0.0 1.0 2.0 3.0 4.0

Outdoor concentration BaP (ng m-3) Site=all, µEnv=home, Season=Heating

BaP

1:1

Ci≤Co (for ß1)

Ci>Co (in ß0)

0.0 0.1 0.1 0.2 0.2 0.3 0.3 0.4 0.4 0.5

0.0 0.1 0.2 0.3 0.4 0.5

BaP (ng m-3) Site=all, µEnv=home, Season=Nonheating

BaP

1:1

Ci≤Co (for ß1)

Ci>Co (in ß0)

(17)

3.2.2 PAH and B(a)P seasonal differences in schools

Figure 7. Comparison of ΣPAH Finf by heating vs non-heating season in schools.

In schools the regression ß0 estimates become slightly negative for both heating and non-heating seasons. The heating season slope is slightly larger, but estimate forced via origin is 0.74, identical with the non-heating season. The results indicate to use 0.74 for all seasons.

Figure 8. Comparison of BaP Finf by heating vs non-heating season in schools.

Also BaP estimates suggests higher infiltration during heating season; the difference being affected by the non- zero ß0 in non-heating season; forcing the regression through origin produces ß1 = 0.54 for the non-heating season, thus not supporting plausible seasonal differences. The results indicate to use all-season estimate for schools (0.63).

0.0 5.0 10.0 15.0 20.0 25.0

ΣPAH (ng m-3) Site=all, µEnv=School, Season=Heating

ΣPAH

1:1

Ci≤Co (for ß1)

Ci>Co (in ß0)

y=0.84x + -1.43 (R²=84%, n=14) Cig (ß0, all data)

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0

0.0 0.5 1.0 1.5 2.0

ΣPAH (ng m-3) Site=all, µEnv=School, Season=Nonheating

ΣPAH

1:1

Ci≤Co (for ß1)

Ci>Co (in ß0)

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5

0.0 1.0 2.0 3.0 4.0

Outdoor concentration BaP (ng m-3) Site=all, µEnv=School, Season=Heating

BaP

1:1

Ci≤Co (for ß1)

Ci>Co (in ß0)

0.0 0.0 0.0 0.1 0.1 0.1 0.1 0.1 0.2

0.0 0.1 0.1 0.2 0.2

BaP (ng m-3) Site=all, µEnv=School, Season=Nonheating

BaP

1:1

Ci≤Co (for ß1)

Ci>Co (in ß0)

(18)

3.2.3 PAH and B(a)P seasonal differences in offices

Figure 9 shows that the i-o regression suggests indoor sources for both seasons and that the heating season Finf estimate is affected by the relatively larger ß0. For both heating and non-heating seasons the Finf values are much lower than in homes and schools; this supports the approach to differentiate the Finf values at least for offices.

The heating season estimate is affected by the two points at higher outdoor levels; thus forcing the ß1 down.

Forcing the both regressions via origin produces values 0.36 and 0.63, respectively; these values are recommended.

Figure 9. Comparison of ΣPAH Finf by heating vs non-heating season in offices.

The BaP estimate for heating season is tilted down due to ß0 >0 and a single higher Co point; thus the low Finf estimate at ß1 = 0.27 seems unreliable; forced via origin the slope is 0.46. The non-heating season value seems reliable at 0.63. These two latter values are recommended.

Figure 10. Comparison of BaP Finf by heating vs non-heating season in offices.

0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 45.0 50.0

0.0 10.0 20.0 30.0 40.0 50.0

ΣPAH (ng m-3) Site=all, µEnv=Office, Season=Heating

ΣPAH

1:1

Ci≤Co (for ß1)

Ci>Co (in ß0)

0.0 0.5 1.0 1.5 2.0 2.5

ΣPAH (ng m-3) Site=all, µEnv=Office, Season=Nonheating

ΣPAH

1:1

Ci≤Co (for ß1)

Ci>Co (in ß0)

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5

0.0 1.0 2.0 3.0 4.0 5.0

Outdoor concentration BaP (ng m-3) Site=all, µEnv=Office, Season=Heating

BaP

1:1

Ci≤Co (for ß1)

Ci>Co (in ß0)

0.0 0.1 0.1 0.2 0.2 0.3 0.3

0.0 0.1 0.2 0.3

BaP (ng m-3) Site=all, µEnv=Office, Season=Nonheating

BaP

1:1

Ci≤Co (for ß1)

Ci>Co (in ß0)

(19)

3.2.4 PAH and B(a)P seasonal differences in traffic

Due to limited data set size the original regressions for heating and non-heating seasons are unstable for ßs. The raw values suggest expected higher infiltration for winter, but when removing the instability by forcing the regressions to origin (second graphics below), the slopes are very close to each other (0.63 and 0.69 for heating and non-heating seasons, respectively.

Figure 11. Comparison of ΣPAH Finf by heating vs non-heating season in traffic.

0.0 2.0 4.0 6.0 8.0 10.0 12.0

0.0 5.0 10.0 15.0

ΣPAH (ng m-3) Site=All, µEnv=Traffic, Season=Heating

ΣPAH

1:1

Ci≤Co (for ß1)

Ci>Co (in ß0)

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8

0.0 0.5 1.0 1.5 2.0

ΣPAH (ng m-3) Site=All, µEnv=Traffic, Season=Nonheating

ΣPAH

1:1

Ci≤Co (for ß1)

Ci>Co (in ß0)

y = 0.7481x + 0.0595 R² = 0.8908 y = 0.6302x + 0.0034

R² = 0.8503

0.00 2.00 4.00 6.00 8.00 10.00 12.00

0.00 5.00 10.00 15.00

Indoor level (ng m-3)

Outdoor level (ng m^-3)

ΣPAH

ΣPAH

1:1

Ci≤Co (for ß1)

Ci>Co (in ß0)

y = 0.9077x + 0.0021 R² = 0.8706

y = 0.696x - 0.0073 R² = 0.8107

0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60 1.80

0.00 0.50 1.00 1.50 2.00

Indoor level (ng m-3)

Outdoor level (ng m^-3)

ΣPAH

ΣPAH

1:1

Ci≤Co (for ß1)

Ci>Co (in ß0)

(20)

Similar behaviour is seen for the heating season data in traffic for BaP. The slope 0.49 increases to 0.61 when forcing the regression via origin. The non-heating season estimate remains stable at 0.67.

Figure 12. Comparison of BaP Finf by heating vs non-heating season in traffic, original regressions and those forces via origin below.

0.0 0.5 1.0 1.5 2.0 2.5

BaP (ng m-3) Site=All, µEnv=Traffic, Season=Heating

BaP

1:1

Ci≤Co (for ß1)

Ci>Co (in ß0)

0.0 0.1 0.1 0.2 0.2 0.3 0.3 0.4

0.0 0.1 0.2 0.3 0.4

BaP (ng m-3) Site=All, µEnv=Traffic, Season=Nonheating

BaP

1:1

Ci≤Co (for ß1)

Ci>Co (in ß0)

Report on infiltration and exposure model with software prototypes

Life+ EXPAH (Contract ENV/IT/000082; 2009)