estimating air pollution infiltration efficiencies for

Estimating Air Pollution Infiltration Efficiencies for Exposure

Assessment and Epidemiology

Ryan Allen, PhD Faculty of Health Sciences

Simon Fraser University Burnaby, BC

Overview• Why estimate residential infiltration efficiencies?• How?

–Tracer–Recursive model

• Description, validation, examples

• Application to epi–Panel studies–Model building for other study designs

• Windsor results–And possible future directions

AmbientPM2.5

“Leaky”

AmbientPM2.5

Infiltration Efficiency (Finf )

AmbientPM2.5

“Tightly Sealed”

–Total exposure = ambient + nonambient

• The fraction of the ambient concentration that penetrates indoors and remains suspended

• Function of AER, penetration, deposition

• Exposure to ambient pollution depends on Finf & time spent outdoors

Why?• To better interpret epi. results from different

locations and/or from different seasons

• Reduce exposure misclassification in epi studies

• To tease apart health impacts of ambient vs. nonambient pollution

• To better interpret epi. results from different locations and/or from different seasons

• Reduce exposure misclassification in epi studies

• To tease apart health impacts of ambient vs. nonambient pollution

Hystad et al., JESEE, in press

Why?

72%

8%

18%2%

Vegetative BurningMobile

SecondaryCrustal

58%

4%

14%

24%

Why?• To better interpret epi. results from different locations and/or from different

seasons

• Reduce exposure misclassification in epi studies

• To tease apart health impacts of ambient vs. nonambient pollution

Larson et al., JAWMA, 2004

Outdoors (12.2 ug/m3) Indoors (9.4 ug/m3)

How?• Tracer (most commonly sulfur or sulfate)

–Requires that there be no (or few) indoor or personal sources

–Indoor/outdoor ratio or slope gives Finf

• Recursive model–Requires continuous indoor/outdoor

measurements (e.g. nephelometer, DustTrak)–Does not require absence of indoor sources–Shows promise for estimating Finf of pollutants

without good tracers (e.g. ultrafines)

Average indoor concentration during the hour

=

Some fraction of the average outdoor concentrationduring the hour

+

Some fraction of the indoor concentrationthat remains from the PREVIOUS hour

+Contribution from indoor sources during the hour

int

int

outt

int SCCC 121 ++= −ββ

~0

2

1inf 1 β

β−

=+

=ka

PaF

Recursive Model Technique

Allen et al., ES&T, 2003

0

1

2

3

4

5

6

7

8

22-Feb 24-Feb 26-Feb 28-Feb 1-Mar 3-Mar

Ligh

t Sca

tterin

g C

oeffi

cien

t (x1

0 -5 m-1

)

OutdoorsIndoorsIndoor Source Hour

Recursive Model Technique

Allen et al., ES&T, 2003

Recursive Model Technique Validation

Allen et al., ES&T, 2003

Recursive Model Technique Validation

Allen et al., JESEE, 2007

Recursive Model Technique Examples

Barn et al., JESEE, 2008

• Air cleaner effectiveness in woodsmoke and forest fire impacted community

Recursive Model Technique Examples

Polidori et al., JAWMA, 2007

• Contributions of ambient and nonambient sources at retirement facilities in southern CA

Application of Finf to Epidemiology Panel Studies

Allen et al., Inhal. Tox., 2008

Application of Finf to Epidemiology Model Development for Large Studies

Koenig et al., EHP., 2005

R2 = 0.60

R2 = 0.66

Application of Finf to Epidemiology Model Development for Large Studies

Hystad et al., JESEE., In Press

• Victoria, BC

• Spatial property assessment data (SPAD)

• Model including season predicts 54% of total variance

• Potentially allows Finf to be estimated in many homes without I/O monitoring

Windsor PM2.5 Finf

Mean: 0.36IQR: 0.26 – 0.44

Mean: 0.33IQR: 0.25 – 0.40

52 homes monitored in both seasons

Windsor PM2.5 Finf Models

Season Predictor(s) R2

Summer I-O Temp Diff, Window Opening 0.36

Winter Building age, air cleaner use 0.20

Windsor UFP Finf

Mean: 0.34IQR: 0.19 – 0.50

Mean: 0.21IQR: 0.13 – 0.27

27 homes monitored in both seasons

Windsor UFP Finf Models

Season Predictor(s) R2

Summer Outdoor Temp, Window Opening 0.56

Winter ----- -----

PM2.5 Finf vs. UFP Finf

r = 0.44

r = 0.68

Other Possibilities for Windsor Data• Continue development of Finf models• Comparison of Finf for different PM species

–PM2.5–UFP–LAC

• Health effects–Ambient / nonambient PM vs. lung function