From the Inside Out – Application of the Mass Balance ...

From the Inside Out – Applicationof the Mass Balance Model for PM

Exposure Assessment in ResidentialSettings Under the Influencesof Indoor and Outdoor Factors

The Harvard community has made thisarticle openly available. Please share howthis access benefits you. Your story matters

Citation Lee, Wan-Chen. 2015. From the Inside Out – Application of the MassBalance Model for PM Exposure Assessment in Residential SettingsUnder the Influences of Indoor and Outdoor Factors. Doctoraldissertation, Harvard T.H. Chan School of Public Health.

Citable link http://nrs.harvard.edu/urn-3:HUL.InstRepos:23205179

Terms of Use This article was downloaded from Harvard University’s DASHrepository, and is made available under the terms and conditionsapplicable to Other Posted Material, as set forth at http://nrs.harvard.edu/urn-3:HUL.InstRepos:dash.current.terms-of-use#LAA

http://osc.hul.harvard.edu/dash/open-access-feedback?handle=&title=From%20the%20Inside%20Out%20%E2%80%93%20Application%20of%20the%20Mass%20Balance%20Model%20for%20PM%20Exposure%20Assessment%20in%20Residential%20Settings%20Under%20the%20Influences%20of%20Indoor%20and%20Outdoor%20Factors&community=1/4454687&collection=1/13398961&owningCollection1/13398961&harvardAuthors=d5981b12f1ca9288facd1cd954812982&department

http://nrs.harvard.edu/urn-3:HUL.InstRepos:23205179

http://nrs.harvard.edu/urn-3:HUL.InstRepos:dash.current.terms-of-use#LAA



FROM THE INSIDE OUT – APPLICATION OF THE MASS BALANCE MODEL FOR

PM EXPOSURE ASSESSMENT IN RESIDENTIAL SETTINGS UNDER THE

INFLUENCES OF

INDOOR AND OUTDOOR FACTORS

WAN-CHEN LEE

A Dissertation Submitted to the Faculty of

The Harvard T.H. Chan School of Public Health

in Partial Fulfillment of the Requirements

for the Degree of Doctor of Science

in the Department of Environmental Health

Harvard University

Boston, Massachusetts

November, 2015

ii

Dissertation Advisor: Dr. Petros Koutrakis Wan-Chen Lee

From the Inside Out – Application of the Mass Balance Model for PM Exposure

Assessment in Residential Settings under the Influences of Indoor and Outdoor Factors

Abstract

The application of the widely used mass balance model in determining portable air purifier

(PAP) effectiveness in particulate matter (PM) removal was not validated in occupied residential

settings. The corresponding size-resolved information and measurements for the model

parameters and PAP effectiveness were also limited to better characterize human exposure to

indoor PM. Additionally, effects of ambient factors, such as meteorology, and their long-term

impacts on occupant indoor exposure to outdoor PM was unclear.

We achieved well-mixed environment and steady state of PM concentrations that met the

mass balance model assumptions. Size-resolved particle deposition rate was determined using

non-linear mixed effects model, whereas linear mixed effects model was used to estimate the

slope between the measured and modeled effectiveness for validation purpose.

To evaluate the impact of ambient factors on PM exposure, we assembled data from two

cohorts in the greater Boston area, assessing the monthly and long-term effect of temperature and

other meteorology on Sr. Long-term meteorology was projected using 15 weather models for the

past and future 20 years to estimate Sr for the corresponding periods with mixed effects models.

Both particle deposition rate and portable air purifier effectiveness were highly particle size-

dependent. Filtration was found to be the dominant removal mechanism for submicrometer

particles, whereas deposition could play a more important role in ultrafine particle removal.

iii

There was reasonable agreement between measured and modeled effectiveness with size-

resolved slopes ranging from 1.11±0.06 to 1.25±0.07 (mean±SE), except for particles <35 nm.

Sr was found to be a robust measure of indoor exposure to outdoor PM, and temperature was

its significant predictor. Seasonal effect of temperature was much more dominant when

compared to long-term effect on Sr, which differed in the whole population and the

subpopulation of naturally ventilated homes. However, long-term temperature effect was small,

with maximum of <10% for summer Sr compared to the past.

Findings from the studies improved characterization of indoor PM exposure. The study

design and methods can be used in the future to better understand exposure scenarios and their

correlation to health effects in other homes or populations.

iv

TABLE OF CONTENTS

LIST OF FIGURES ............................................................................................................ vi

LIST OF TABLES .............................................................................................................. ix

ACKNOWLEDGEMENTS ............................................................................................... x

INTRODUCTION ............................................................................................................... 1

Bibliography ..................................................................................... 10

CHAPTER 1 Size-Resolved Deposition Rates for Ultrafine and Submicrometer

Particles in a Residential Housing Unit ........................................ 16

Environmental Science & Technology. 2014, 48 (17), 10282–10290

Abstract ............................................................................................. 17

Introduction ....................................................................................... 18

Materials and methods ...................................................................... 20

Results ............................................................................................... 27

Discussion ......................................................................................... 32

Bibliography ..................................................................................... 42

Supporting information (SI) .............................................................. 46

CHAPTER 2 Validation and Application of the Mass Balance Model to Determine

the Effectiveness of Portable Air Purifiers in Removing Ultrafine and

Submicrometer Particles in an Apartment ................................... 55

Environmental Science & Technology. (Published online, July 24th, 2015)

Abstract ............................................................................................. 56

Introduction ....................................................................................... 57

Materials and methods ...................................................................... 58

v

Results ............................................................................................... 66

Discussion ......................................................................................... 73

Bibliography ..................................................................................... 78

Supporting information (SI) .............................................................. 83

CHAPTER 3 Effects of Monthly and Long-term Temperature Change on Indoor

Exposure to Outdoor PM2.5 in the Greater Boston Area ............ 102

(Working paper)

Abstract ............................................................................................. 103

Introduction ....................................................................................... 104

Materials and Methods ...................................................................... 106

Results ............................................................................................... 113

Discussion ......................................................................................... 126

Bibliography ..................................................................................... 132

CONCLUSIONS ............................................................................................................ 138

vi

LIST OF FIGURES

Figure 0.1 A schematic illustrating the scope of the overall dissertation.

Figure 0.2 A schematic illustrating the connections between Chapter 1, 2, and 3 based on the mass balance model application.

Figure 1.1 Comparison of the predicted and measured particle concentrations during the decay periods for the 11 particle size categories, using data from one sampling day (0.61 ACH) as an example. The solid markers represent the actual measurements while the solid lines are the predicted decay curves from the NLIN procedure.

Figure 1.2 Comparison of deposition rates of particles less than 1 µm in occupied houses between previous and the current studies. The shaded area represents the 95% confidence interval for the estimated mean deposition rate by particle size in this study.

Figure 2.1 Layout of the devices and instruments in the apartment.

Figure 2.2 Average size-resolved filtration efficiencies of the 2 PAPs under 3 flow settings.

Figure 2.3 An example of the continuous measurements and device operation profiles over one sampling day at g=0.91 h-1 for the total particle concentration (系岫建岻), the total particle generation rate (荊継痛墜痛銚鎮), the total flow rate of the 2 PAPs (芸捗), and the 鯨繋滞 concentration.

Figure 2.4 Size-resolved particle removal rates: filtration by PAPs, deposition, and air exchange. 系畦迎迎沈 (h-1) is the size-resolved clean air replacement rate, equal to 系畦経迎沈【撃. 系畦迎迎な, 系畦迎迎に and 系畦迎迎ぬ corresponded to the flow rates of 195, 387, and 540 m3/h, respectively. The air exchange rates were the average values under three target air exchange rates: ACH1, ACH2, and ACH3 are the target air exchange rate of 0.60, 0.90, and 1.20 h-1, respectively. k1, k2 and k3 are the average particle deposition rates at ACH1, ACH2, and ACH3, respectively.

Figure 2.5 Measured size-resolved effectiveness for the three PAP flow rates (芸捗= 195, 387, and 540 m3/h) under three target air exchange rates (A=0.60, 0.90 and1.20 h-1).

Figure 2.6 The slopes and their 95% confidence intervals by particle size obtained from the mixed effects model. The shaded area represents ±10% of the ideal coefficient of 1 (0.90-1.10).

Figure 2.7 Relationship between effectiveness and CARR.

Figure 3.1 Boxplots for Sr measurements by month for (a) the whole population with mixed AC usage, and (b) the subpopulation of naturally ventilated homes (AC=0). The

vii

solid points represent the Sr observations; whereas the filled diamonds in red represent the monthly mean of Sr.

Figure 3.2 Sr measurement for (a) the subpopulation of naturally ventilated homes (AC=0),

and (b) homes that used AC during the sampling period (AC=1). Measurements

from the two cohorts are marked in different colors.

Figure 3.3 Comparisons between projected monthly mean temperature, RH, wind speed and precipitation from CMIP5 models and NARR database for the period of 1981-2000.

Figure 3.4 Mean temperature by month for the past and the future based on paired years

(N=1 to 20). The solid and dashed lines are projections for the future and the past,

respectively. The light-colored lines represent projections from the CMIP5

models, whereas the dark-colored lines describe the multi-model means across the

CMIP5 models.

Figure 3.5 Mean estimated Sr by month for the past and the future based on paired years

(N=1 to 20) for the whole population with mixed AC usage (AC=mixed). The

solid and dashed lines are projections for the future and the past, respectively. The

light-colored lines represent projections from the CMIP5 models, whereas the

dark-colored lines describe the multi-model means across the CMIP5 models.

Figure 3.6 Mean estimated Sr by month for the past and the future based on paired years (N=1 to 20) for the subpopulation of naturally ventilated homes (AC=0). The solid and dashed lines are projections for the future and the past, respectively. The light-colored lines represent projections from the CMIP5 models, whereas the dark-colored lines describe the multi-model means across the CMIP5 models.

Figure 3.7 Projected monthly mean temperature for the past (1981-2000) and the future (2046-2065) by 15 CMIP5 models (dashed lines). The solid line is the overall monthly mean across the CMIP5 models.

Figure 3.8 Predicted monthly mean Sr for the past and the future by the two populations. The

solid lines were the overall monthly mean across the CMIP5 models while the

dashed lines were 罰 1 SD from the overall mean. AC=0 represented the

subpopulation of naturally ventilated homes and AC=mixed was referred to the

whole population.

Figure 3.9 Difference in predicted monthly mean Sr between the two populations for the past

and the future. The solid lines are the overall monthly mean difference across the

CMIP5 models while the dashed lines are 罰 1 SD from the overall mean. AC=0

viii

represents the subpopulation of naturally ventilated homes and AC=mixed is

referred to the whole population.

Figure 3.10 Monthly mean differences in estimates between the future and the past for (a) temperature, and (b) Sr. The solid lines are the overall monthly mean difference across the CMIP5 models while the dashed lines are 罰 1 SD from the overall mean. AC=0 represents the subpopulation of naturally ventilated homes and AC=mixed is referred to the whole population.

ix

LIST OF TABLES

Table 0.1 Comparisons of the study design and methods between Chapter 1, 2, and 3.

Table 1.1 Measured parameters (mean±standard deviation) during decay tests in the

apartment unit.

Table 1.2 Estimated deposition rate by particle size, categorized based on the midpoint of

mobility diameter (dm).

Table 1.3 Experimental characteristics from selected studies on deposition rates for particles

of less than 1µm in houses.

Table 2.1 Summary of measured parameters.

Table 3.1 Summary of the sampling parameters, meteorology, and the day-to-day variability

of the meteorology parameters within the sampling duration.

x

ACKNOWLEDGEMENTS

First of all, I would like to thank my advisor Dr. Petros Koutrakis for his patience and for

granting me the freedom to pursue research ideas and explore different approaches. Through his

guidance I have learned and equipped myself with skills and strength to mature as an

independent researcher.

I would also like to express my deepest gratitude to my research committee members, Dr.

Stephen Rudnick and Dr. Paul Catalano, who have provided invaluable insights and knowledge

in steering me towards the right direction for my study, and from whom I have gained friendship

and tremendous support.

My extended appreciation goes to the Taiwan Ministry of Education for the financial support

in the first three years of my doctoral program, and Coway Ltd. for the partial sponsorship of the

air purifier study in this dissertation. I am also thankful to Dr. Loretta Mickley and her team for

the effective collaboration in the climate change study.

Furthermore, I would like to acknowledge Mike Wolfson, Joy Lawrence, and Steve Ferguson

for their technical support in research design and methods, which helped establish a solid

foundation for successfully carrying out the field work. Mike has been a great mentor, and it was

such a pleasure and rewarding experience to work with him.

Last but not least, I’m profoundly indebted to my family and friends for their generous and

firm support through this journey, especially my parents and grandparents who have been so

strong and selfless for their sacrifices to help me reach this milestone.

Wan-Chen Lee

Boston, Massachusetts

1

INTRODUCTION

2

Epidemiological studies have shown significance of ambient particulate matter (PM)

exposure on health, most notably cardiovascular illness and mortality (1-3). Ambient PM can

penetrate indoor environment where people spend the majority of their time (4). Consequently,

indoor exposure to outdoor PM has been one of the central fields where researchers invest

themselves to unravel the relationship between the exposure and its resulting health risks from

the complex interactions of indoor, outdoor and human factor contributions.

Indoor PM concentration is a result of a dynamic process, featuring competing factors of

particle source emission (input) and removal (output) via various pathways, which can be

categorized into PM properties-related mechanisms, building characteristics, and occupant

behavior (5). One pollutant property-driven removal mechanism is size-dependent particle

deposition onto indoor surfaces. Both large and small particles have higher rates because larger

particles can settle by gravitation, whereas diffusion is the dominant mechanism for small

particles, such as ultrafine particles (6). Particles that have sizes between the two, mostly

submicrometer particles, tend to stay airborne because they are too large to diffuse and too small

to settle by gravitation. The size-resolved deposition rate can also be affected by air mixing and

the indoor surface area, with generally positive relationships (7, 8). Given proper experimental

design and instruments, particle deposition rates can be measured readily on site.

Similarly, penetration coefficient for outdoor PM entering indoors through building cracks is

also size-dependent (9, 10). During infiltration process, small particles are removed by diffusion,

whereas larger particles are removed both by gravitational settling and inertial impaction at the

crack entry (11). Penetration coefficient is affected by the geometry and roughness of cracks

(10). However, when penetration occurs through an open window, the coefficient is close to

unity.

3

Air through building cracks and windows, together with mechanical ventilation by fans, are

major mechanisms for building ventilation. Air exchange rate is used as measure to quantify

ventilation when there is exchange of air between indoors and outdoors. Its effect is

bidirectional, and thus it is commonly used to adjust the indoor PM concentrations. For example,

occupants close windows to minimize air exchange and the penetration of ambient particle in

high outdoor pollution episodes (e.g., forest fire) (12, 13). Contrarily, windows are open to vent

pollutants from significant indoor sources such as smoking or cooking (14). For naturally

ventilated buildings, wind speed, and indoor-outdoor temperature difference are also factors

influencing ACH, in addition to the cracks and window opening (15-17). On the other hand, AC

usage has been reported to decrease ACH through closed windows and removal by deposition

and/or additional filters inside the AC system (15, 18, 19).

Various indoor sources have been characterized for their contributions to indoor PM

concentrations, particle composition and size distribution. Emission from these sources is often

intermittent/episodic, highly variable, and tightly related to occupant activities, such as cooking,

cleaning (e.g., vacuuming), smoking, and incense burning (20-22). When present, indoor sources

tend to result in PM concentrations much higher than that from outdoors, partially due to intense

source strength and small air dilution volume of indoor space (20, 23).

In view of the omnipresent indoor particle exposure from various sources with relatively

modest natural removal mechanisms, some intervention strategies aim to more effectively reduce

overall indoor PM concentration, regardless of the sources. Houses built more recently are often

equipped with the central air system, inside which some houses install filters for particle

removal. For naturally ventilated homes, portable air purifiers can be useful (13, 24, 25). The

purifiers are designed with different flow rates and various technologies, such as filtration,

4

electrostatic precipitation, and ion generation (26, 27). Studies have shown that those equipped

with high efficiency particulate air (HEPA) filter and electrostatic precipitators are the most

effective, depending on the particle size (27-29). However, electrostatic precipitators were found

to generate ozone which is a harmful pollutant (26).

Given diverse residential indoor environment and varying occupant activity pattern, a

systematic and comprehensive understanding of the aforementioned mechanisms not only help

separate the contribution of indoor and outdoor sources, but also assist in epidemiological studies

to link PM exposure and interventions to observed health outcomes or benefits on the population

level. The mass balance equation is one such means to describe the relationship between indoor

PM concentration and the source contribution. The resulting mass balance model (MBM), often

referred as the box model, has been a common approach for determining the indoor factors that

influence indoor PM, and sometimes the PM concentration itself in homes (16, 20, 30-32). These

factors include the ones previously discussed.

Indoor-outdoor PM ratio (I/O) derived from the mass balance equation in the absence of

indoor sources is often used as an index to quantitatively characterize particle infiltration and is

used to correlate personal exposure to outdoor PM concentrations (33). It is often referred to as

the infiltration factor. In epidemiological studies where indoor PM measurements of individual

homes are often unavailable, infiltration factor can be used to estimate the indoor PM

concentration of outdoor fraction given the ambient PM concentration from the central site.

Building tightness is one important factor for infiltration factor variability. With regard to

ambient factors, mainly meteorology, studies have shown that ambient temperature is associated

with occupant window opening behavior, which in turn could affect the air exchange rate and

increase particle infiltration (34, 35).

5

Substantial amount of research have been conducted to better understand the roles of indoor

factors and mechanisms as they contribute to indoor PM concentration through interactive,

competing or enhancive effects under the broad, yet complex framework of MBM application.

Gaps remain, however, not preventing from the development of knowledge, but to attenuate

generalizability or confidence in results interpretation due to uncertainties in findings or scarce

information in newly explored study areas. For example, the application of MBM in estimating

either the indoor PM or the other model parameters was not validated experimentally in

residential settings. Violation of mass balance assumptions in the homes could potentially lead to

biased predictions of PM related parameters. Additionally, size-resolved information is limited

for both model validation and the determined parameters such as deposition rate. While intensive

efforts have been placed in understanding how factors and mechanisms indoors influence human

exposure to PM in residences, ambient factors such as meteorology were less studied. Although

the change in specific meteorological parameters due to climate change was found to influence

ambient PM2.5 concentrations (36), there was very little information in the association between

meteorology and PM penetration to indoor environment.

Answers to these pending questions would strengthen the mass balance application in indoor

PM exposure assessment and improve the understanding of size-resolved behavior for those

particles. The main motivation of this dissertation is, therefore, to fill these existing gaps with

specific aspects on the protective removal mechanisms for indoor PM, and how the ambient

meteorology comes into play. Figure 0.1 and Figure 0.2 are provided to illustrate the overall

scope of this dissertation and the connections between the chapters.

Chapter 1 and 2 are embedded in the same controlled study which was conducted in an

apartment. Chapter 1 adopted a modified experimental approach, in conjunction with a MBM, to

6

determine the size-resolved deposition rates and assess the effect of air exchange rate under

enhanced mixing conditions. Chapter 2 aimed to validate the MBM through the assessment of

the size-resolved effectiveness of PAP(s) equipped with HEPA filters in removing ultrafine

particles (UFPs) (<0.1たm) and submicrometer particles (0.10−0.53 たm). Validation was done by

comparing experimentally determined size-resolved PAP effectiveness using directly measured

particle concentrations with and without the operation of PAPs, to the modeled effectiveness

using individually measured model input parameters during the same test periods.

In Chapter 3, the research focus was extended to factors outside the mass balance box (e.g.,

homes) to explore the impact of outdoor temperature and other meteorology on particle

infiltration factor on a monthly and long-term climate change basis, using archived samples from

two observational studies featuring 340 homes in the greater Boston area. Indoor-outdoor sulfur

ratio was used as a surrogate of infiltration factor for PM2.5 to associate with the main effect of

temperature with exposure models (mixed effects models). Weekly indoor-outdoor sulfur ratio

for the future and past 20 years were estimated using projected meteorology from 15 weather

forecast models, and were summarized into monthly averages. The predicted sulfur ratio were

also examined and compared across two population scenarios to reflect the influence of AC

usage: the whole population with mixed AC usage, and the subpopulation of naturally ventilated

homes.

Although studies in the three chapters all revolve around the mass balance equation, there are

some distinct differences in the study design, experimental development and statistical methods.

Table 0.1 shows a list of major comparisons.

7

Figure 0.1. A schematic illustrating the overall scope of the dissertation.

8

Figure 0.2. A schematic illustrating the connections between Chapter 1, 2, and 3 based on the

mass balance model application.

To sum up, this dissertation presents studies on how indoor and outdoor factors influence

indoor PM levels with pioneering and interdisciplinary approaches that were not pursued

previously, including the achievement of steady state in an actual apartment for model validation

and PM measurements, and quantitatively linking the association of monthly and long-term

climate change to PM infiltration factor on a population basis. I hope the novelty of the study

design and findings in this dissertation will serve as a basis and lend ideas to researchers for

future investigation of related topics, altogether contributing to the integration of knowledge for

mass balance application in indoor PM studies; while not simply spanning the scope from

indoors to outdoors, but essentially from the inside out.

9

Table 0.1. Comparisons of the study design and methods between Chapter 1, 2, and 3.

Chapter 1 Chapter 2 Chapter 3

Location Single-home in Boston area 340 homes in Boston area

Objective Determination of PM

deposition rates in a home

MBM validation in a

home; Determination of

PAP effectiveness in a

home (intervention)

Modeling of indoor PM exposure

from ambient sources under the

influence of monthly and long-

term climate change

Particle size

range Size-resolved UFP and submicrometer PM PM2.5

PM conc. unit Number concentration Mass concentration

Data source Controlled field study

Statistical modeling - Two

observational studies in

conjunction with projected

weather data

MBM approach Concentration decay Steady state Steady state

Modeling Type Mechanistic Phenomenological

PM source Non-sourced period

following sourced period

Constant indoor PM

generation No indoor PM source

Statistical

analysis

Non-linear mixed effects

model

Linear mixed effects

model Linear mixed effects model

10

Bibliography

(1) Pope, C.; Dockery, D.; Schwartz, J. Review of Epidemiological Evidence of Health-Effects

of Particulate Air-Pollution. Inhal. Toxicol. 1995, 7 (1), 1-18; 10.3109/08958379509014267.

(2) Peters, A.; Dockery, D.; Muller, J.; Mittleman, M. Increased particulate air pollution and the

triggering of myocardial infarction. Circulation 2001, 103 (23), 2810-2815.

(3) Schwartz, J.; Dockery, D.; Neas, L. Is daily mortality associated specifically with fine

particles? J. Air Waste Manage. Assoc. 1996, 46 (10), 927-939.

(4) Klepeis, N.; Nelson, W.; Ott, W.; Robinson, J.; Tsang, A.; Switzer, P.; Behar, J.; Hern, S.;

Engelmann, W. The National Human Activity Pattern Survey (NHAPS): a resource for assessing

exposure to environmental pollutants. J. Expo. Anal. Environ. Epidemiol. 2001, 11 (3), 231-252;

10.1038/sj.jea.7500165.

(5) Liu D. and Nazaroff, W.W. Modeling pollutant penetration across building envelopes.

Atmospheric Environment 2001, 35, 4451-4462.

(6) Hinds, W.C. Aerosol Technology: Properties, Behavior, and Measurement of Airborne

Particles. Wiley: New York, 1999; .

(7) Lai, A.C.K.; Byrne, M.A.; Goddard, A.J.H. Experimental studies of the effect of rough

surfaces and air speed on aerosol deposition in a test chamber. Aerosol Science and Technology

2002, 36 (10), 973-982; 10.1080/02786820290092249.

11

(8) Thatcher, T.L.; Lai, A.C.K.; Moreno-Jackson, R.; Sextro, R.G.; Nazaroff, W.W. Effects of

room furnishings and air speed on particle deposition rates indoors. Atmos. Environ. 2002, 36

(11), 1811-1819; 10.1016/S1352-2310(02)00157-7.

(9) Liu, D. and Nazaroff, W.W. Modeling pollutant penetration across building envelopes.

Atmospheric Environment 2001, , 4451-4462.

(10) Liu, D. and Nazaroff, W.W. Particle Penetration Through Building Cracks. Aerosol Science

and Technology 2003, 37, 565-573.

(11) Chen, C. and Zhao, B. Review of relationship between indoor and outdoor particles: I/O

ratio, infiltration factor and penetration factor. Atmospheric Environment 2011, 45 (2), 275-288.

(12) Wallace, L. Indoor Particles: A Review. J. Air Waste Manage. Assoc. 1996, 46, 98-126.

(13) Barn, P.; Larson, T.; Noullett, M.; Kennedy, S.; Copes, R.; Brauer, M. Infiltration of forest

fire and residential wood smoke: an evaluation of air cleaner effectiveness. Journal of Exposure

Science and Environmental Epidemiology 2008, 18 (5), 503-511.

(14) Wallace, L.A.; Mitchell, H.; T O'Connor, G.; Neas, L.; Lippmann, M.; Kattan, M. Particle

concentrations in inner-city homes of children with asthma: the effect of smoking, cooking, and

outdoor pollution. Environmental health perspectives 2003, 111 (9), 1265.

(15) Wallace, L.A.; Emmerich, S.J.; Howard-Reed, C. Continuous measurements of air change

rates in an occupied house for 1 year: the effect of temperature, wind, fans, and windows.

Journal of exposure analysis and environmental epidemiology 2002, 12 (4), 296-306.

12

(16) Wallace, L. and Williams, R. Use of personal-indoor-outdoor sulfur concentrations to

estimate the infiltration factor and outdoor exposure factor for individual homes and persons.

Environmental science & technology 2005, 39 (6), 1707-1714.

(17) Haghighat, F.; Brohus, H.; Rao, J. Modelling air infiltration due to wind fluctuations—A

review. Building and Environment, 2000, 35 (5), 377-385.

(18) Sarnat, J.A.; Koutrakis, P.; Suh, H.H. Assessing the relationship between personal

particulate and gaseous exposures of senior citizens living in Baltimore, MD. Journal of the Air

& Waste Management Association 2000, 50 (7), 1184-1198.

(19) Howard-Reed, C.; Wallace, L.A.; Emmerich, S.J. Effect of ventilation systems and air

filters on decay rates of particles produced by indoor sources in an occupied townhouse. Atmos.

Environ. 2003, 37 (38), 5295-5306; 10.1016/j.atmosenv.2003.09.012.

(20) Abt, E.; Suh, H.H.; Catalano, P.; Koutrakis, P. Relative contribution of outdoor and indoor

particle sources to indoor concentrations. Environ. Sci. Technol. 2000, 34 (17), 3579-3587;

10.1021/es990348y.

(21) Nazaroff, W.W. Indoor particle dynamics. Indoor air 2004, 14 (s7), 175-183.

(22) Abt, E.; Suh, H.H.; Allen, G.; Koutrakis, P. Characterization of indoor particle sources: A

study conducted in the metropolitan Boston area. Environmental Health Perspectives 2000, 108

(1), 35.

13

(23) Wallace, L.A.; Emmerich, S.J.; Howard-Reed, C. Source strengths of ultrafine and fine

particles due to cooking with a gas stove. Environmental Science & Technology 2004, 38 (8),

2304-2311.

(24) Shaughnessy, R. and Sextro, R. What is an effective portable air cleaning device? A review.

Journal of Occupational and Environmental Hygiene 2006, 3 (4), 169-181;

10.1080/15459620600580129.

(25) Sublett, J.L. Effectiveness of Air Filters and Air Cleaners in Allergic Respiratory Diseases:

A Review of the Recent Literature. Current Allergy and Asthma Reports 2011, 11 (5), 395-402;

10.1007/s11882-011-0208-5.

(26) Waring, M.S.; Siegel, J.A.; Corsi, R.L. Ultrafine particle removal and generation by

portable air cleaners. Atmos. Environ. 2008, 42 (20), 5003-5014;

10.1016/j.atmosenv.2008.02.011.

(27) Sultan, Z.M.; Nilsson, G.J.; Magee, R.J. Removal of ultrafine particles in indoor air:

Performance of various portable air cleaner technologies. Hvac&R Research 2011, 17 (4), 513-

525; 10.1080/10789669.2011.579219.

(28) Offermann, F.; Sextro, R.; Fisk, W.; Grimsrud, D.; Nazaroff, W.; Nero, A.; Revzan, K.;

Yater, J. Control of Respirable Particles in Indoor Air with Portable Air Cleaners. Atmos.

Environ. 1985, 19 (11), 1761-1771; 10.1016/0004-6981(85)90003-4.

14

(29) Shaughnessy, R.; Levetin, E.; Blocker, J.; Subkette, K. Effectiveness of Portable Indoor Air

Cleaners - Sensory Testing Results. Indoor Air-International Journal of Indoor Air Quality and

Climate 1994, 4 (3), 179-188; 10.1111/j.1600-0668.1994.t01-1-00006.x.

(30) Long, C.M.; Suh, H.H.; Catalano, P.J.; Koutrakis, P. Using time- and size-resolved

particulate data to quantify indoor penetration and deposition behavior. Environ. Sci. Technol.

2001, 35 (10), 2089-2099; 10.1021/es001477d.

(31) Hodas, N.; Meng, Q.; Lunden, M.M.; Turpin, B.J. Toward refined estimates of ambient PM

2.5 exposure: Evaluation of a physical outdoor-to-indoor transport model. Atmospheric

Environment, 2014, 83, 229-236.

(32) Sarnat, J.A.; Long, C.M.; Koutrakis, P.; Coull, B.A.; Schwartz, J.; Suh, H.H. Using sulfur

as a tracer of outdoor fine particulate matter. Environmental Science & Technology 2002, 36

(24), 5305-5314.

(33) Chen, C. and Zhao, B. Review of relationship between indoor and outdoor particles: I/O

ratio, infiltration factor and penetration factor. Atmospheric Environment 2011, 45 (2), 275-288.

(34) Wallace, L. and Howard-Reed, C. Continuous monitoring of ultrafine, fine, and coarse

particles in a residence for 18 months in 1999-2000. . Journal of the Air & Waste Management

Association 2002, 52 (7), 828-844.

(35) Kearney, J.; Wallace, L.; MacNeill, M.; Héroux, M.E.; Kindzierski, W.; Wheeler, A.

Residential infiltration of fine and ultrafine particles in Edmonton. Atmospheric Environment

2014, 94, 793-805.

15

(36) Dawson, J.P.; Adams, P.J.; Pandis, S.N. Sensitivity of PM 2.5 to climate in the Eastern US:

a modeling case study. Atmospheric chemistry and physics 2007, 7 (16), 4295-4309.

16

CHAPTER 1

Size-Resolved Deposition Rates for Ultrafine and Submicrometer Particles in a Residential

Housing Unit

Environmental Science & Technology. 2014, 48 (17), 10282–10290

17

Abstract

We estimated the size-resolved particle deposition rates for the ultrafine and submicron particles

using a nonlinear regression method with unknown particle background concentrations during

non-sourced period following a controlled sourced period in a well-mixed residential

environment. A dynamic adjustment method in conjunction with the constant injection of tracer

gas was used to maintain the air exchange rate at three target levels across the range of 0.61-1.24

air change per hour (ACH). Particle deposition was found to be highly size dependent with rates

ranging from 0.68±0.10 to 5.03±0.20 h-1 (mean±s.e). Our findings also suggest that the effect of

air exchange on the particle deposition under enhanced air mixing was relatively small when

compared to both the strong influence of size-dependent deposition mechanisms and the effects

of mechanical air mixing by fans. Nonetheless, the significant association between air exchange

and particle deposition rates for a few size categories indicated potential influence of air

exchange on particle deposition. In the future, the proposed approach can be used to explore the

separate or composite effects between air exchange and air mixing on particle deposition rates,

which will contribute to improved assessment of human exposure to ultrafine and submicron

particles.

Key words: Deposition; Ultrafine particles; Submicron particles; Residential indoor air quality;

Air exchange; Nonlinear regression

18

Introduction

Epidemiological studies have shown association between exposure to ambient fine

particulate matters and adverse health effects such as mortality and onset of cardiovascular

events (1-3). Exposure to indoor fine particles is of particular concern, as people spend more

than 85% of their time in enclosed buildings with the majority of that time in residences (4). An

important parameter in assessing residential particle exposures is indoor particle deposition rate,

a natural removal mechanism that contributes to the reduction of airborne particle levels indoors.

Studies conducted in occupied houses have demonstrated size-specific characteristics for

deposition rates: elevated levels for larger (>1 µm) and ultrafine particles (<100 nm), but

comparatively lower rates for submicron particles (0.1-1.0 µm) (5-11).

Despite the consistent size-dependent trend of the deposition rates, reported values of the

deposition estimates vary by more than an order of magnitude for submicron and ultrafine

particles of similar sizes. Such wide variability could possibly be due to differences in the study

designs and data analysis methods, building characteristics, furnishings, and many other factors.

Although no standard method has been established, most studies adopted the mass balance model

in conjunction with on-site measurements to determine the size-resolved deposition rates in

residential environments and assumed that the air was well-mixed (5, 7-14). However, results

from these studies suggest high uncertainties in the estimated deposition rates, partially due to

limitations in acquiring reliable measurements on particle background concentrations during

tests, potential bias from using average daily air exchange rate to represent real time air exchange

measurements, or difficulty in decoupling the particle deposition rate from another model

parameter (penetration coefficient). As a result, more research is still needed to adequately assess

the size-resolved deposition rates in real-life situations.

19

The deposition rates of submicron and ultrafine particles have been found to be strongly

influenced by indoor air mixing. Factors commonly known to affect indoor air mixing include

mechanical mixing by fans, ventilation, and air exchange between indoor and outdoor

environments. Positive associations between mechanical air mixing and particle deposition rates

have been reported from previous studies during the operation of either portable or central fans

in the ventilation system (8, 10, 15-18). On the other hand, studies conducted in occupied houses

have found inconsistent associations between particle deposition rates and air exchange (8, 11,

19, 20). As the amount of mechanical mixing increases, the consequent increase in air movement

could mask the effect of air exchange on particle deposition rate, especially on closed window

days. However, it remains unclear how much mechanical air mixing is sufficient that the effect

of air exchange becomes negligible.

We expected that achieving well-maintained test conditions in a home would minimize

uncertainties of the estimated size-resolved deposition rates, and allow us to examine the effect

of air exchange rate on the estimated rates under the enhanced mechanical air mixing using

portable fans. We therefore used a modified approach based on the mass balance model to

determine the size-resolved deposition rates. The main features of this approach are: (1) the

achievement of a well-mixed indoor environment using portable fans; (2) the generation of

artificial particles at a constant rate to substantially elevate indoor concentration prior to the

particle decay measurement; (3) the use of a dynamic method to maintain air exchange rates at

constant levels throughout the sampling period; and (4) the use of the NLIN procedure, which

does not require knowledge of the magnitude of the particle background concentration, to

estimate the size-resolved deposition rates along with their uncertainties. The results from this

20

study would provide indications for future application of the revised approach, and can be used

to assess the human exposure to submicron and ultrafine particles in homes.

Materials and methods

This study was conducted in a fully furnished, non-carpeted and occupied concrete floor

apartment unit in Cambridge, Massachusetts, during November 2011. The apartment consists of

two bedrooms, a kitchen, a living room, a bathroom, and a hallway that connects the living room

and kitchen. There was no air conditioning system in the apartment, and heating was provided by

hydronic radiant heating system. Ventilation in the house depended on the opening of windows,

doors and two small vents that exhausted the air from indoors. These openings remained closed

and taped throughout all sampling periods (with minor adjustments on the taping to maintain

constant air exchange rates). Our study design based on the mass balance approach required the

air in the room to be well-mixed. And after a series of preliminary tests, we selected the kitchen,

living room, and the hallway as the study area (approximately 34.8 m2) to ensure a well-mixed

condition using portable fans. The fans were all at their highest speed settings and were placed in

the same location and orientation for all the tests. No significant indoor sources were present

except for the artificial particle generation system. The layout of the instruments and devices in

the apartment unit is shown in Figure S1.1 in Supporting Information (SI).

Model description. We used the mass balance approach to model the concentrations of particles

in an indoor space over time. A general form of the model is

21

鳥寵日岫痛岻鳥痛噺糠鶏沈系墜沈髪彫帳日蝶伐糠系沈岫建岻伐倦沈系沈岫建岻 (1.1)

Where: 系沈(t) is the indoor concentration of particles in the 件痛朕 size category at time t (particles/cm3 or

#/cm3)

g is the air exchange rate (h-1) 鶏沈 is the penetration coefficient of particles in the 件痛朕 size category (dimensionless) 系墜沈 is the outdoor concentration of particles in the 件痛朕 size category (#/cm3) 荊継沈 is the indoor generation rate of particles in the 件痛朕 size category (#/h)

V is the effective room volume of the study zone (cm3) 倦沈 is the deposition rate of particles in the 件痛朕 size category (h-1)

The parameters 系沈(t), 鶏沈 , 系墜沈 and 倦沈 are size-dependent and can thus be expressed as size-

resolved values. The first and second terms on the right side of eq 1.1 describe the entry of

particles into the indoor space through infiltration from outdoors and the indoor generation of

particles (both are assumed to be constant during the tests). The third and the fourth terms

represent the removal of particles through exfiltration to the outdoor space and deposition onto

surfaces in the room, respectively. Assuming that: (1) the indoor air is well mixed; (2) the indoor

generation rate (荊継沈) is negligible; and (3) the outdoor concentration (系墜沈) is constant over time,

the cumulative particle concentration at any time t, 系沈(t), can be expressed by integrating eq 1.1

over time as follows:

系沈岫建岻噺底牒日寵任日碇日盤な伐結貸碇日痛匪髪結貸碇日痛系沈岫ど岻 (1.2)

22

Where: 膏沈噺糠髪倦沈 (1.3) 系沈(0) is the initial concentration at the start of measurement for particles in the 件痛朕 size category

(#/cm3)

eq 1.2 shows that the observed decrease in particle concentration during the decay period is

due to the combination of air exchange between indoors and outdoors and the deposition of

particles onto surfaces within the indoor space. It was employed as the final model in this study

where 系沈岫建岻 and 糠 were measured independently after the generation of artificial particles was

stopped. The measured parameters were subsequently used to determine 倦沈 in the data analysis

phase.

Measurement and adjustment of air exchange rate. The two features involving the air

exchange component required by the model were to maintain a constant 糠 for each test to meet

the model assumption and to determine the actual rate.

Three target air exchange rates (A) were set to determine the size-resolved particle deposition

rates with repeated tests: A=0.60, 0.90, and 1.20 ACH. To achieve this, we applied the sulfur

hexafluoride (SF6) tracer gas method and measured SF6 in two consecutive phases: (1) the

steady-state phase where we used the real-time measurement of steady-state SF6 concentration as

a proxy of 糠, and (2) the exponential decay phase where 糠 was determined (Figure S1.2 in SI)

(21). The former required a constant injection rate of SF6 ( 芸聴庁展 ), under which the SF6

concentration in the room would reach 95% of its steady state (系鎚鎚) in about 3/g hours when 糠

was held constant. The steady-state relationship between the parameters is shown in eq 1.4 (21):

23

系鎚鎚岫喧喧兼岻噺町縄鈍展盤陳典ゲ朕貼迭匪底岫朕貼迭岻蝶岫陳典岻抜など滞岫喧喧兼岻 (1.4)

At steady state, change in 系鎚鎚 directly reflected the variability of 糠 when 芸聴庁展 and V were

both fixed values. Under this circumstance, we could maintain 糠 by monitoring 系鎚鎚 and adjusting

it to a constant level. In practice, an SF6 generation system was set up to provide a constant

injection rate of 54.20 罰 0.57 cm3/min (mean±s.d.). The SF6 concentration was measured

continuously in the kitchen and living room by two SF6 monitors (Brüel & Kjær model 1302).

For most of the experiments the differences between the SF6 steady state concentrations in

the living room and the kitchen at any time were less than 10%. Prior to the SF6 release, we

closed all the windows, doors and vents and partially sealed the gaps around these closed

openings with masking tape to establish the initial air exchange conditions. When the windows

were closed during the tests, the air exchange in the home was mainly driven by the pressure and

temperature differences between indoor and outdoor environments, leading to infiltration and

exfiltration of air through gaps of doors and windows. The initially established 糠 was adjusted

downward from the highest air exchange condition by partially sealing the gaps with masking

tape to achieve the target level.

Due to the rapid mixing inside of the house, the SF6 steady state concentration reflected the

drift in 糠 within 1 or 2 measurements (about 1 or 2 minutes). To shorten the time to achieve

steady state, a “blast” of SF6 was released over a short period to elevate the SF6 concentration

close to the target steady state level before starting the injection of SF6 at a constant rate. The SF6

24

concentration would then approach 系鎚鎚 which corresponded to a specific 糠 at that time. By

adjusting the sealing of the gaps with masking tape throughout the sampling day, we were able to

keep 系鎚鎚 within 5% of the target value for the specific ACH required based on the relationship in

eq 1.4.

At the end of each sampling day, we stopped the SF6 generation and continued to measure

the SF6 concentration which then followed an exponential decay over time. The decay constant,

which represented 糠 over the decay period, corresponded to 系鎚鎚 of the sampling day and was

determined by fitting a simple linear regression curve between log-transformed SF6

concentration and time (21). We determined the daily average 糠 based on the relationship

between 糠 from the end-of-the-day measurements and the 系鎚鎚 using data from both the nine test

days and the preliminary tests (n=4), given the calculated arithmetic mean of the determined V

(撃博噺ひぱ兼戴, n=13).

Generation and measurement of particles. The High-output Extended Aerosol Respiratory

Therapy (HEART®) nebulizers (Westmed, Inc., Tucson, Arizona) were used to aerosolize

aqueous sodium chloride (NaCl) solution (0.0375% by mass) to generate NaCl particles. To

achieve sufficiently high concentrations for the decay tests, nebulizers that generated aerosols

with comparable rates and size distributions were selected and used in single, double and triple

combinations to achieve similar steady state particle concentrations for each of the three target

air exchange rates. These combinations were tested in the laboratory and had NaCl aerosolization

rates of 16.58±0.32, 32.29±0.98, and 46.24±0.58 g/h for one, two and three nebulizers,

respectively.

25

The TSI Model 3936 scanning mobility particle sizer (SMPS) equipped with a Model 3785

Water-based Condensation Particle Counter (WCPC) was used to measure particles between

0.015 to 0.533 µm (mobility diameter) for size distribution and count concentrations over

consecutive 5-min intervals.

Sampling plan. Steady state particle concentration indoors was achieved in about 2 hours

(“sourced period”), after which we turned off the particle generation system and monitored

particle concentration at consecutive 5-min intervals over approximately one hour of particle

decay (“non-sourced period”). This decay period provided a sufficient number of measurements

for an adequate analysis of the size-resolved particle deposition rates. In total, there were nine

test days with measurements under three target air exchange rates in triplicate (one test per day).

An example of the simultaneous measurements of the total particle and the SF6 concentrations

throughout the test day is depicted in Figure S1.2 in SI.

Data analysis. Particle data were divided into 11 size categories: <25, 25-35, 35-45, 45-55, 55-

65, 65-80, 80-100, 100-150, 150-200, 200-300, and >300 nm. Additionally, total particle

concentrations were also used in the analysis. We estimated 倦沈 based on eq 1.2 using

measurements from the non-sourced period subsequent to the sourced period, which provided

sufficiently high initial concentration prior to decay, 系沈岫ど岻 . The non-sourced period was

carefully defined based on the actual time that we shut off the particle generation system. In the

data analysis phase, 系沈岫ど岻 was assumed to be unknown and allowed to vary around the measured

value to minimize the effect of instrument uncertainty. Similarly, 鶏沈系墜沈 and 倦沈 were treated as

unknown parameters with values greater than zero. The nonlinear approximation procedure,

26

PROC NLIN (SAS Inc. Cary, NC), was used to determine the values of the unknown parameters

(鶏沈系墜沈, 倦沈 and 系沈岫ど岻) through an iterative process to find the best combination of the parameters

which yielded the minimum value of the residual sum of squares (the sum of the squared

differences between the modeled and measured 系沈岫建岻). One important feature of the NLIN

procedure is the selection of “good” starting values for the unknown parameters to prevent the

approximation process from converging to local minima. In this study, multiple starting points

were introduced in the NLIN procedure by specifying their physically feasible ranges with

investigator-defined intervals for each unknown parameter: 100-5,000 by 500 (particles/cm3) for 鶏沈系墜沈 , 0.1-6 by 1 (h-1) for 倦沈 , and 10-5,000 by 100 (particles/cm3) for 系沈岫ど岻. More detailed

description of the procedure can be found in the SI.

Based on the air exchange rate, 倦沈 were determined by the NLIN procedure at three levels:

(1) 倦沈┸底 as 倦沈 for each test, (2) 倦沈┸凋 as the average 倦沈 by the three target air exchange rates

(A=0.60, 0.90 and 1.20 ACH), and (3) 倦沈┸銚鎮鎮 as the average 倦沈 across all the nine tests (Figure

S1.3 in SI). Analysis at the first level was to determine 倦沈 for each test day separately so we

could examine the goodness of fit of the predicted indoor particle concentrations from the NLIN

procedure versus the measured values. This visual examination was necessary because the r-

squared value is generally not a meaningful measure in nonlinear regression analysis. The second

level was used to evaluate the effect of air exchange on the deposition estimates while the third

was to summarize the estimates in comparison with those from the previous studies. In the

analyses for the last two levels, 鶏沈系墜沈 and 系沈岫ど岻 were allowed to have different constant values

for each test day to account for the daily variation of particle concentrations both outdoors and

indoors, while 倦沈 was assumed to be constant under each target air exchange rate for 倦沈┸凋 and

27

constant across all ACH for 倦沈┸銚鎮鎮. Uncertainties for the estimates from the NLIN procedure were

reported as standard errors.

To evaluate the effect of air exchange rate on 倦沈, the estimates determined from the second

level (倦沈┸凋退待┻滞待, 倦沈┸凋退待┻苔待, 倦沈┸凋退怠┻態待) were first examined using a global F test with a significance

level of 0.05 for each particle size category, to see whether the 倦沈 from at least one target ACH

were different from the 倦沈 from the others. Subsequently, pairwise comparisons (n=3) of the 倦沈┸凋

values were made with the level of significance of 0.0167. The F statistics used for the pairwise

comparison procedure were based on the paired data, except that we used the same mean squared

error (MSE) from the global test. By doing so, we were able to make three comparisons on the

same basis (using the same denominator for the F-statistics), and the higher degrees of freedom

from the MSE would contribute to more stable results in the analysis.

Results

Measured parameters. The measured parameters and indoor conditions in the apartment for the

three target air exchange rates are shown in Table 1.1. The relative humidity (RH) was below the

deliquescence point of NaCl of 75.3% (at 25°C); thus, the particle-associated water was expected

to evaporate completely, leaving cubic crystals of NaCl as the aerosol (22). The variability of

indoor temperature and RH across the test days was small with coefficients of variation of less

than 5%, except for the RH at A=0.90 ACH (coefficient of variation = 10.16%).

Estimated size-resolved deposition rate. Comparisons between the measured particle

concentrations by particle size versus time and the fitted (predicted) values from the NLIN

procedure were made for all nine test days to evaluate the goodness of fit of the model. Figure

28

1.1 shows an example of the fitted plot using measurements from one test day under 0.61 ACH.

There was good agreement between measured and model-predicted particle concentrations for all

particle sizes across all air exchange rates. The steepness of descent for the decay curves

increased with increase in deposition rates, which varied substantially by particle size. The

estimated 倦沈 from the NLIN procedure are presented in Table 1.2 for each of three target air

exchange rates. The results showed strong size dependence of the estimated deposition rates

(Figure 1.1 and Figure S1.4 in SI). As a general trend, the deposition rate decreased as the size

increased for the ultrafine particles (<100 nm) and subsequently remained lower for particle sizes

between 100-550 nm. The highest estimated deposition rate (mean±s.e.) was found for particles

<25 nm with 4.45±0.14 h-1 at A=0.60 ACH, 4.73±0.22 h-1 at A=0.90 ACH, and 5.03±0.20 h-1 at

A=1.20 ACH.

Table 1.1. Measured parameters (mean±standard deviation) during decay tests in the apartment unit

Target air exchange

rate (h-1)

n Measured air exchange rate

(h-1)

Aerosolization rate of NaCl

solution# (g/h)

C(0)##

(#/cm3)

Elapsed time

(min)

Indoors

Temp.

(°C)

RH

(%)

0.60 3 0.61±0.00 16.6±0.3 12,129±1,253 66.7±5.8 24.0±1.2 33.5±1.5

0.90 3 0.91±0.01 32.3±1.0 21,270±1,235 50.0±0.0 24.8±0.5 44.3±4.5

1.20 3 1.22±0.01 46.2±0.6 18,074±3,829 53.3±2.9 23.5±0.4 34.4±12.2

# Particle generation rate is expressed as the aerosolization rate of NaCl solution (0.0375%) from a single or multiple

nebulizers with repeated tests (n>3) in the laboratory. It included water evaporation rates of 3.57±1.33, 8.02±1.49

and 13.19±1.54 g/h for one, two and three nebulizers combined, respectively. ## C(0) is the measured total initial particle concentrations during decay tests.

29

Effect of air exchange rate under enhanced air mixing. In general, estimates of 倦沈 at A=1.20

ACH (倦沈┸凋退怠┻態待岻 were lower than those from the other two target air exchange rates, except for

the smallest and the largest particle size categories (Table 1.2). Results from the pairwise

comparisons showed no statistically significant differences between 倦沈┸凋退待┻滞待 and 倦沈┸凋退待┻苔待 for

particles of all sizes (Table S1.1 in SI). Nevertheless, significant differences were observed in a

few size categories between the highest target air exchange rate (A=1.20 ACH) and the other two

lower target ACH. Specifically, 倦沈┸凋退怠┻態待 values were significantly lower than 倦沈┸凋退待┻苔待 for

particles of 35-45 nm, 65-80 nm and 80-100 nm. A similar trend was shown between 倦沈┸凋退怠┻態待

and 倦沈┸凋退待┻滞待 for particles of 35-45 nm, 80-100 nm and 150-200 nm. Overall, given the enhanced

air mixing conditions, our findings have only found sporadic statistically significant differences,

but not a consistent and relatively meaningful trend in the effect of air exchange rate on

deposition rate across all particle sizes.

30

Figure 1.1. Comparison of the predicted and measured particle concentrations during the decay

periods for the 11 particle size categories, using data from one sampling day (0.61 ACH) as an

example. The solid markers represent the actual measurements while the solid lines are the

predicted decay curves from the NLIN procedure.

31

Table 1.2. Estimated deposition rate by particle size, categorized based on the midpoint of mobility diameter (dm).

Target air exchange rate

(h-1)

Particle size

(nm) n

Estimated 倦沈 (h-

1)

Approx. s.e.#

(h-1)

Approx. 95% C.I.##

Lower Upper

0.60

<25 42 4.45 0.14 4.16 4.74

25-35 42 3.33 0.07 3.18 3.47

35-45 42 2.61 0.06 2.48 2.73

45-55 42 2.13 0.08 1.97 2.29

55-65 42 1.82 0.07 1.67 1.96

65-80 42 1.45 0.07 1.32 1.59

80-100 42 1.40 0.09 1.21 1.58

100-150 42 1.13 0.09 0.96 1.30

150-200 42 1.12 0.11 0.89 1.36

200-300 42 1.13 0.13 0.87 1.39

>300 41 1.11 0.17 0.77 1.46

Total### 42 2.23 0.05 2.14 2.33

0.90

<25 32 4.73 0.22 4.28 5.17

25-35 32 3.36 0.11 3.14 3.57

35-45 32 2.59 0.08 2.42 2.75

45-55 32 2.05 0.08 1.88 2.21

55-65 32 1.80 0.07 1.66 1.94

65-80 32 1.62 0.06 1.50 1.74

80-100 32 1.37 0.08 1.21 1.54

100-150 32 1.00 0.06 0.87 1.12

150-200 32 1.01 0.11 0.78 1.24

200-300 32 0.95 0.16 0.62 1.27

>300 32 1.03 0.26 0.49 1.56

Total### 32 2.30 0.05 2.19 2.41

32

(Table 1.2 continued)

1.20

<25 34 5.03 0.20 4.61 5.44

25-35 34 3.25 0.12 3.00 3.50

35-45 34 2.25 0.09 2.07 2.43

45-55 34 1.83 0.12 1.58 2.09

55-65 34 1.51 0.12 1.25 1.76

65-80 34 1.24 0.09 1.05 1.43

80-100 34 0.96 0.11 0.75 1.18

100-150 34 0.96 0.10 0.76 1.16

150-200 34 0.68 0.10 0.46 0.89

200-300 34 0.91 0.14 0.62 1.20

>300 34 1.42 0.24 0.94 1.91

Total### 34 2.06 0.06 1.93 2.19

# Approx. s.e. is the approximate standard error of the estimated 倦沈 from the NLIN procedure. ## Approx. 95% C.I. is the approximate 95% confidence interval for the estimated 倦沈 from the NLIN procedure. ### It represents the total particle concentration which is the sum of size-resolved particle concentrations.

Discussion

The estimated particle deposition rates in this study were found to be strongly dependent on

size, as suggested by previous studies (5, 7, 8, 10). The partially V-shaped curve of deposition

rate by size is due to the size-dependent deposition mechanisms where ultrafine particles are

removed by indoor surfaces due to diffusion while submicron particles are too large to diffuse

and too small to settle effectively by gravitation (23, 24). As the size increases for submicron

particles, gravitation gradually becomes a more dominant mechanism for deposition.

Comparisons of size-resolved particle deposition rates for particles less than 1 µm in

occupied houses are depicted in Figure 1.2. The numeric values of 倦沈┸銚鎮鎮 used for comparison can

be found in Table S1.2 in SI. The size-resolved deposition rates presented by Long et al. were

33

remarkably lower than those from the others possibly because the deposition rate was determined

using a physical-statistical model that assumes steady state and depends on the value of the

penetration coefficient, whereas the other studies adopted a non-steady state (concentration

decay) approach (5, 7, 8, 10). While acknowledging the large variability in the size-resolved

deposition rates reported from the previous studies and considering the relatively small 95%

confidence intervals for the mean estimates of 倦沈 in this study, we found that our estimates for

submicron particles were in close agreement with some of these studies (5, 8, 10). However, the

mean estimates of 倦沈 for the ultrafine particles in this study were considerably higher than the

reported deposition rates from the other studies, which could possibly be explained by the effect

of enhanced air mixing by the operation of portable fans.

Mechanical air mixing by fans is positively associated with indoor particle deposition rate (8,

10, 15-18). Thatcher et al. investigated the effect of airspeed on particle deposition rate for

particles of 0.5-20.0 µm under no fan and three fan speeds in an experimental room (17). They

found that the deposition rate of particles smaller than 1 µm at the highest fan speed was on

average 1.5 times the rate when the fan was off, and the deposition onto fan blades was not high

enough to explain the increased deposition rate by fan speed. Wallace et al. evaluated the

deposition rates for a broader range of particle sizes, including the ultrafine and submicron

particles (10). The authors compared deposition rates with and without the use of the central fan

in a townhouse and found a general trend of elevated deposition rate when the fan was on. This

enhanced deposition loss was a joint contribution of deposition, particle loss in the heating and

air conditioning system and possibly the increased air velocity.

34

Figure 1.2. Comparison of deposition rates of particles less than 1 µm in occupied houses

between previous and the current studies (5, 7, 8, 10). The shaded area represents the 95%

confidence interval for the estimated mean deposition rate by particle size in this study.

35

When provided air mixing, the particles are brought from the bulk air to the boundary layer

near the indoor surfaces via advection, through which they deposit onto the surfaces by diffusion

(24). The use of central or portable fans thus leads to a substantial increase in the amount of air

mixing indoors and contributes to increased particle deposition rates by facilitating the transport

of particles to the boundary layer and by reducing the thickness of the boundary layer (24, 25).

Such influence was more pronounced for ultrafine particles where diffusion is the dominant

mechanism for deposition. Since the level of mechanical air mixing in the current study was

thought to be much stronger than that in the other studies, the boundary layer processes explain

largely why our deposition estimates are higher, especially for ultrafine particles.

Air exchange rate is an important factor in determining particle deposition rate not only

because it is a parameter in the mass balance model but also for its relation to indoor air mixing.

Increase in air exchange rate can result in increased particle deposition by facilitating indoor air

movement while it has the parallel effect of lowering the residence time of particles indoors.

However, evaluation of these effects can be challenging, especially when complicated by

mechanical air mixing that is positively associated with particle deposition rate. Previous studies

conducted in occupied houses have shown large disparities in the effect of air exchange on

deposition rate (8, 11, 19, 20). Nevertheless, evaluations of the different findings were difficult

to do due to the varying ranges of air exchange rates along with their corresponding ventilation

or air mixing conditions. Rim et al. investigated the functional relationship between air exchange

and particle deposition for ultrafine particles based on two different levels of air mixing in an

uninhabited test house (18). They discovered that the difference in the deposition rates due to air

mixing by central fan became smaller when the air exchange rate increased (from all windows

closed to 2 windows open 7.5 cm each). Compared to Rim et al., the small effect of air exchange

36

rate on size-resolved particle deposition rates in the present study could be explained by the

masking effect of mechanical air mixing over the relatively narrow range of natural air exchange

in the apartment on closed window days (18). However, it remains unclear to what extent the

variation in one factor would mask the effect from the other.

In this study, we also estimated the average particle deposition rates based on integrated

measurements in the attempt to allow comparison with previous studies that used integrated

measurements (Table 1.2). Table 1.3 presents a summary of the experimental conditions for the

selected home studies which estimated size-resolved or integrated particle deposition rates for

particles smaller than 1 µm. The wide variability of estimates across studies could be attributed

not only to differences in chemical and physical characteristics of particles, interiors of the

houses (e.g., furnishing), use of air cleaners or air furnaces, ventilation system, surface-to-

volume ratios, indoor air mixing levels, air exchange rate, occupancy, but also to differences in

experimental and analytical methods. The integrated particle deposition rates in our study ranged

from 2.06±0.06 to 2.30±0.05 h-1 across the three target air exchange rates and were comparable

with findings from the others (13, 14). Overall, smaller uncertainties were observed in this study

for both of the size-resolved and the integrated estimates largely due to well-maintained

experimental conditions.

High indoor particle concentrations have been reported to result in coagulation which was

considered as an important mechanism of particle loss indoors, especially for ultrafine particles

(24, 26). Rim et al. investigated the coagulation of ultrafine particles during indoor episodes

resulting from various indoor sources and concluded that coagulation should be accounted for

when the number concentration for ultrafine particles exceeds 20,000 particles/cm3 (26). In this

study, the total particle number concentrations at the steady state from which the decay started

37

were approximately 20,000 particles/cm3 under various test conditions and was no more than

10% higher than this threshold for all sampling days. Therefore, the influence of coagulation on

the estimated deposition rate was considered to be negligible.

One strength of the NLIN procedure is that measurement of particle background

concentration (shown as 底牒日寵任日碇日 in eq 1.2) during each decay period is not required to determine

the particle deposition rate. Instead, the unknown value of 鶏沈系墜沈 can be estimated simultaneously

in the same procedure (Figure S1.5 in SI). In an attempt to see how deposition estimates might

have differed had we adopted the commonly used linear regression method without knowing the

particle background concentrations, we conducted analysis using the same data but with iterated

background values to estimate 倦沈 by fitting a simple linear regression curve between log-

transformed 系沈岫建岻 and time. Estimates of 倦沈 from the linear regression model showed wide

variability across different background values; however, only slight variation was observed in

the corresponding r-squared values. As an extreme example, ignoring the background (底牒日寵任日碇日噺ど) could lead to underestimation of deposition rate over a factor of 2 or more. It also resulted in

negative estimates of 倦沈 for particles >100 nm at 1.21 ACH, indicating that the accuracy of these

estimates was questionable. Our analysis suggests that three potential concerns can rise from the

linear regression method when the background level is unknown: (1) estimates of 倦沈 can be

increasingly sensitive to the variation in the background level as the sampling duration increases;

(2) neglecting the background concentration typically results in the underestimation of 倦沈; and

(3) using r-squared value as a criterion for best estimates of 倦沈 can lead to inaccuracies, as slight

changes in the r-squared value correspond to a wide range of values of estimated 倦沈. Given these

38

concerns, the nonlinear regression approach was considered to be more reliable than the linear

regression approach for this study.

Table 1.3. Experimental characteristics from selected studies on deposition rates for particles of less than 1µm in

houses (5, 8-10, 12-14).

Study House type Main

particle source

Particle size

range (µm)##

Sample type

Mixing mechanism

Particle monitor#

Air exchange

rate (h-1)

Deposition rate for <1 µm (h-1)

Abt et al.

(2000)

4 homes (occupied)

Cooking 0.02-

10 Size-

resolved Natural

convection SMPS, APS

0.16- 0.66

0.02- 1.70

Long et al.

(2001)

9 nonsmoking

homes (occupied)

Ambient 0.02-

10 Size-

resolved Natural

convection SMPS, APS

0.89 (winter);

2.1 (summer)

0.004-0.35 (winter)

0.15-0.59 (summer)

Chao et al.

(2003)

6 homes (occupied)

Ambient 0.02-9.65

Integrated, Size-

resolved

Natural convection

P-Trak, APS

1.28± 0.54

0.27 (APS) 0.52

(P-Trak) Howard-Reed et

al. (2003)

A townhouse (occupied)

Cooking, candle,

kitty litter

0.30-10

Size-resolved

Central fan (on/off)

OPC 0.64± 0.56###

0.29-0.47 (fan off) 0.66-1.0 (fan on)

Wallace et al.

(2004)

A townhouse (occupied)

Cooking, candle,

kitty litter

0.011-5.43

Size-resolved

Central fan (on/off)

SMPS, APS, OPC

0.64± 0.56

0.70-4.10 (fan off) 0.90-3.92 (fan on)

Kearney et al.

(2011)

94 homes (occupied)

Ambient 0.02-1.0

Integrated Various

types P-Trak

0.12- 0.37

(inter-quatile)

####

0.68- 0.87

Stephens and

Siegel (2012)

18 homes (unoccupied)

Ambient 0.02-1.0

Integrated

Central fan plus two box fans

(on)

P-Trak 0.13- 0.95

(GM)*

0.31- 3.24

Wallace et al.

(2013)

74 homes (occupied)

Cooking 0.02-1.0,

PM2.5 Integrated

Various types

P-Trak, DustTrak

0.35± 0.30

1.17± 0.94

This study (2013)

A home (occupied)

NaCl nebuli-zation

0.015-0.533

Size-resolved

Portable fans (on)

SMPS 0.61- 1.24

0.68- 5.03

# SMPS: scanning mobility particle sizer; APS: aerodynamic particle sizer; OPC: optical particle counter. ## Particles measured by SMPS were reported with mobility diameter. APS measured particles based on aerodynamic diameter, while OPS measured particles by light-scattering. ### The values were reported by He et al.11. #### The study was conducted in one winter and two summer seasons.* GM=geometric mean.

39

One of the limitations of this study is the lack of a standard method to validate the accuracy

of the estimates from the NLIN procedure because there is no standard method to determine size-

resolved particle deposition rate. Generally, applications of the nonlinear approach in estimating

particle deposition rates in residential environment have reportedly suffered from issues such as

low confidence in decoupling the unknown parameters and high uncertainties in the estimates of

the parameters which arise from different study designs (12, 19, 27). In comparison, the

measurements for the current study were taken in a well-mixed environment with well-fit

exponential decay curves under nearly constant air exchange rates, which was expected to

generate more reliable results. The estimated 系沈岫ど岻 was highly comparable to the actual

measurement with differences within 10% for all the analyses. Furthermore, the estimates of 倦沈 returned from the NLIN procedure were within tight 95% confidence intervals, indicating low

uncertainties in the mean deposition rates. Higher uncertainties were observed for particles larger

than 300 nm, possibly due to fewer particle numbers in this size category. The uncertainties,

when expressed as standard deviation over the mean, were highly comparable to those estimated

by Rim et al. in a manufactured test house using a technique equivalent to the NLIN procedure

(20). Consequently, the present study design in conjunction with the nonlinear analytical

approach was regarded as adequate to generate robust estimates of 倦沈. Another limitation of this study is the consequent enhanced air mixing condition from

operating a number of portable fans in the apartment, which is rare in real life situations during

closed-window days. The extent of the fan-driven air mixing in the apartment was expected to

exceed the mixing level from the central fan operation in the previous studies, leading to elevated

levels of 倦沈 (8, 10). As the level of mechanical air mixing increases to a certain extent, it can

mask the effect of air exchange rate, as suggested by the results in the current study. Therefore,

40

caution should be taken when generalizing the findings from this study to predict particle

deposition rates in normal housing conditions. However, the use of fans to facilitate air mixing

helped achieve a well-mixed environment and brought about several advantages. First, it fulfilled

the requirements of the two mass balance models which we used to determine g and 倦沈 ,

respectively, and made it possible to maintain constant air exchange rates. Second, it reduced the

uncertainty in the estimation of 倦沈 from the NLIN procedure because particle concentration

decay followed the mass balance model more closely when the particles are uniformly

distributed indoors, which in turn led to better fitting of the model to the data. It is noteworthy

that the use of the NLIN procedure also helped relieve the constraint of making assumptions on

the particle background levels indoors during data analysis, especially when the outdoor particle

concentrations were unavailable. Third, it allowed us to evaluate the effect of air exchange rate

on 倦沈 as well as the relative level of particle deposition by particle size when provided the same

amount of air mixing indoors. Last but not least, the enhanced deposition loss due to reinforced

air mixing would contribute to higher reduction of human exposures, especially to ultrafine and

submicron particles. Nevertheless, more conservative rates (lower values) should be considered

for exposure assessment in homes without significant air mixing.

The study design in conjunction with the NLIN procedure provided a feasible and alternative

method for estimating particle deposition rates when the background concentration cannot be

measured (assuming that the background concentration is relatively constant over the sampling

period). The same approach can further be applied to understand other particle behaviors in the

future. For example, it can be used to determine simultaneously the size-resolved particle

deposition rates and penetration coefficients when given the measured outdoor concentrations. In

addition, this approach can be used to study the effect of air exchange on particle deposition

41

under varying levels of air mixing by adjusting fans at different speeds. The understanding of

both the particle deposition and penetration rates in a typical home and the factors affecting them

contributes to improved assessment and prediction of human exposure to particles. Subsequent

precautionary measures or actions of intervention can then be taken to reduce particle exposure

in homes.

42

Bibliography

(1) Pope, C.; Thun, M.; Namboodiri, M.; Dockery, D.; Evans, J.; Speizer, F.; Heath, C.

Particulate Air-Pollution as a Predictor of Mortality in a Prospective-Study of Us Adults.

American Journal of Respiratory and Critical Care Medicine 1995, 151 (3), 669-674.








10.1038/sj.jea.7500165.



10.1021/es990348y.

(6) Vette, A.F.; Rea, A.W.; Lawless, P.A.; Rodes, C.E.; Evans, G.; Highsmith, V.R.; Sheldon, L.

Characterization of indoor-outdoor aerosol concentration relationships during the Fresno PM

exposure studies. Aerosol Science and Technology 2001, 34 (1), 118-126;

10.1080/027868201300082120.

43



2001, 35 (10), 2089-2099; 10.1021/es001477d.

(8) Howard-Reed, C.; Wallace, L.A.; Emmerich, S.J. Effect of ventilation systems and air filters

on decay rates of particles produced by indoor sources in an occupied townhouse. Atmos.


(9) Chao, C.; Wan, M.; Cheng, E. Penetration coefficient and deposition rate as a function of

particle size in non-smoking naturally ventilated residences. Atmos. Environ. 2003, 37 (30),

4233-4241; 10.1016/S1352-2310(03)00560-0.

(10) Wallace, L.A.; Emmerich, S.J.; Howard-Reed, C. Effect of central fans and in-duct filters on

deposition rates of ultrafine and fine particles in an occupied townhouse. Atmos. Environ. 2004,

38 (3), 405-413; 10.1016/j.atmosenv.2003.10.003.

(11) He, C.R.; Morawska, L.; Gilbert, D. Particle deposition rates in residential houses. Atmos.


(12) Kearney, J.; Wallace, L.; MacNeill, M.; Xu, X.; VanRyswyk, K.; You, H.; Kulka, R.;

Wheeler, A.J. Residential indoor and outdoor ultrafine particles in Windsor, Ontario. Atmos.


(13) Stephens, B. and Siegel, J.A. Penetration of ambient submicron particles into single-family

residences and associations with building characteristics. Indoor Air 2012, 22 (6), 501-513;

10.1111/j.1600-0668.2012.00779.x.

44

(14) Wallace, L.; Kindzierski, W.; Kearney, J.; MacNeill, M.; Heroux, M.; Wheeler, A.J. Fine

and Ultrafine Particle Decay Rates in Multiple Homes. Environ. Sci. Technol. 2013, 47 (22),

12929-12937; 10.1021/es4025809t.

(15) Mosley, R.B.; Greenwell, D.J.; Sparks, L.E.; Guo, Z.; Tucker, W.G.; Fortmann, R.;

Whitfield, C. Penetration of ambient fine particles into the indoor environment. Aerosol Science

and Technology 2001, 34 (1), 127-136; 10.1080/02786820117449.

(16) Lai, A.C.K.; Byrne, M.A.; Goddard, A.J.H. Experimental studies of the effect of rough

surfaces and air speed on aerosol deposition in a test chamber. Aerosol Science and Technology

2002, 36 (10), 973-982; 10.1080/02786820290092249.



(11), 1811-1819; 10.1016/S1352-2310(02)00157-7.

(18) Rim, D.; Wallace, L.A.; Persily, A.K. Indoor Ultrafine Particles of Outdoor Origin:

Importance of Window Opening Area and Fan Operation Condition. Environ. Sci. Technol.

2013, 47 (4), 1922-1929; 10.1021/es303613e.

(19) Allen, R.; Larson, T.; Sheppard, L.; Wallace, L.; Liu, L. Use of real-time light scattering

data to estimate the contribution of infiltrated and indoor-generated particles to indoor air.

Environ. Sci. Technol. 2003, 37 (16), 3484-3492; 10.1021/es021007e.

(20) Rim, D.; Wallace, L.; Persily, A. Infiltration of Outdoor Ultrafine Panicles into a Test

House. Environ. Sci. Technol. 2010, 44 (15), 5908-5913; 10.1021/es101202a.

45

(21) ASTM Standard Test Method for Determining Air Change in a Single Zone by Means of a

Tracer Gas Dilution. E741-00. American Society for Testing and Materials International: West

Conshohocken, Pennsylvania, 2001; .

(22) Tang, I. and Munkelwitz, H. Composition and Temperature-Dependence of the

Deliquescence Properties of Hygroscopic Aerosols. Atmospheric Environment Part A-General

Topics 1993, 27 (4), 467-473; 10.1016/0960-1686(93)90204-C.

(23) Hinds, W.C. Aerosol Technology: Properties, Behavior, and Measurement of Airborne

Particles. Wiley: New York, 1999; .

(24) Nazaroff, W.W. Indoor particle dynamics. Indoor Air 2004, 14, 175-183; 10.1111/j.1600-

0668.2004.00286.x.

(25) Lai, A.C.K. and Nazaroff, W.W. Modeling indoor particle deposition from turbulent flow

onto smooth surfaces. J. Aerosol Sci. 2000, 31 (4), 463-476.

(26) Rim, D.; Green, M.; Wallace, L.; Persily, A.; Choi, J. Evolution of Ultrafine Particle Size

Distributions Following Indoor Episodic Releases: Relative Importance of Coagulation,

Deposition and Ventilation. Aerosol. Sci. Technol. 2012, 46 (5), 494-503;

10.1080/02786826.2011.639317.

(27) Bennett, D. and Koutrakis, P. Determining the infiltration of outdoor particles in the indoor

environment using a dynamic model. J. Aerosol Sci. 2006, 37 (6), 766-785;

10.1016/j.jaerosci.2005.05.020.

46

Supporting information (SI)

NLIN procedure for data analysis. PROC NLIN is a well-established procedure in SAS. It

utilizes an iterative process to estimate the specified unknown parameters which in this study are 鶏沈系墜沈, 倦沈 and 系沈岫ど岻. The procedure starts from initially guessed values (starting values) for the

parameters and subsequently searches for the best solution (combination of parameter values)

that yields the minimum value of the residual sum of squares. In theory, there can be multiple

solutions that give rise to convergence to end the procedure. However, the solutions might not

always be meaningful or close to the actual values. For example, if the procedure starts from

initially guessed values that are far from the actual values, it could converge to a local minimum

and result in biased estimates. To prevent this situation from occurring, we set 鶏沈系墜沈, 倦沈 and 系沈岫ど岻 to be greater than zero because the particle concentrations and deposition rates should

theoretically be positive values. Additionally, we specified the ranges and the intervals to allow

multiple starting values for all three parameters so that the NLIN procedure could determine the

best combination of the starting values. In fact, the NLN procedure converged to the same

estimated values even without using multiple starting values for 鶏沈系墜沈, 倦沈 and 系沈岫ど岻. This was

most likely due to the distinct profiles of the decay curves which contributed to more robust

estimation in this study.

In the analysis, we treated 系沈岫ど岻 as an unknown parameter instead of using the actual

measurements. The major reason for that was because the determination of 倦沈 from the decay

curve was sensitive to the initial concentration. We therefore specified it as an unknown

parameter and allowed the NLIN procedure to estimate its value based on the data. This

approach was advantageous in three aspects. First, we minimized the influence of the instrument

47

uncertainty for more robust 倦沈 estimation by not using the actual measurements, given that 倦沈 was sensitive to 系沈岫ど岻 . Secondly, we could still compare the estimated values from NLIN

procedure to the actual measurements to check for consistency. Thirdly, consistency between the

estimated and measured 系沈岫ど岻 indicated reliable estimation for 鶏沈系墜沈 and 倦沈 because all three

parameters were approximated simultaneously in the same procedure.

48

Table S1.1. Pairwise comparisons of size-resolved deposition rates for three target air exchange

rates (A=0.60, 0.90 and 1.20 ACH).

A=0.60 vs.0.90 (h-1)

A=0.90 vs. 1.20 (h-1)

A=0.60 vs. 1.20 (h-1)

Particle size

(nm)

F Value Prob>F

F Value Prob>F

F Value Prob>F

<25 1.07 0.3040

1.29 0.2589

3.84 0.0530

25-35 0.05 0.8277

0.58 0.4468

0.23 0.6327

35-45 0.02 0.8775

9.19 0.0032*

8.44 0.0046*

45-55 0.36 0.5525

2.60 0.1101

4.15 0.0445

55-65 0.01 0.9167

5.51 0.0211

4.80 0.0311

65-80 2.34 0.1292

12.28 0.0007*

3.44 0.0670

80-100 0.02 0.8774

9.37 0.0029*

8.79 0.0039*

100-150 1.35 0.2476

0.12 0.7252

2.08 0.1525

150-200 0.49 0.4850

4.22 0.0428

7.48 0.0075*

200-300 0.89 0.3480

0.04 0.8472

1.25 0.2668

>300 0.08 0.7722

1.47 0.2289

0.96 0.3289

Total 0.60 0.4396

9.01 0.0035*

4.18 0.0437

* p < 0.0167.

49

Table S1.2. The avearage size-resolved particle deposition rates across the nine tests (0.61-1.24

ACH).

Particle size

(nm)

n

Estimated 倦沈

(h-1)

Approx. s.e.

(h-1)

Approx. 95% C.I.

Lower Upper

<25 110 4.75 0.11 4.53 4.97

25-35 110 3.32 0.06 3.20 3.43

35-45 110 2.48* 0.05 2.39 2.58

45-55 110 2.00 0.06 1.89 2.11

55-65 110 1.71* 0.05 1.60 1.81

65-80 110 1.45* 0.04 1.36 1.54

80-100 110 1.24* 0.06 1.13 1.36

100-150 110 1.03 0.05 0.93 1.12

150-200 110 0.95* 0.07 0.82 1.08

200-300 110 1.00 0.08 0.84 1.17

>300 110 1.17 0.13 0.92 1.42

Total 110 2.20* 0.03 2.14 2.27

* p < 0.05.

50

Figure S1.1. The layout of the instruments and devices in the apartment unit. The area enclosed

by the red line denotes the study zone.

51

Figure S1.2. An example of the continuous measurements of total particle and 鯨繋滞concentrations

over one sampling day (under 0.91 ACH). The shaded area included the data used to determine

the size-resolved particle deposition rates in the present study.

52

Figure S1.3. Three levels of analyses for 倦沈: (1) 倦沈┸底 as the size-resolved deposition rates for each

test, (2) 倦沈┸凋 as the average size-resolved deposition rates under three target air exchange rates

(A=0.60, 0.90 and 1.20 ACH), and (3) 倦沈┸銚鎮鎮 as the average size-resolved deposition rates across

all the nine tests (0.61-1.24 ACH).

53

Figure S1.4. Estimated particle deposition rates corresponding to the three target air exchange rates. The error bars represent one standard error from the mean.

0.00

1.00

2.00

3.00

4.00

5.00

6.00

19 30 40 50 60 73 90 125 175 250 412

Ave

rage

dep

osit

ion

rate

(pe

r ho

ur)

Diameter midpoint (nm)

0.60 ACH

0.90 ACH

1.20 ACH

54

Figure S1.5. Estimated size-resolved 鶏沈系墜沈 from the NLIN procedure by test day. Each day corresponded to measurements under one constant air exchange rate. The first three days (depicted in red) were for A=0.60 ACH while the middle (in dark green) and the last (in blue) three days were for A=0.90 and 1.20 ACH, respectively.

Particle size (nm)

PiC

oi (

#/c

m3)

0

500

1000

1500

2000

2500

3000

100 200 300 400

Day 1

Day 2

Day 3

Day 4

Day 5

Day 6

Day 7

Day 8

Day 9

55

CHAPTER 2

Validation and Application of the Mass Balance Model to Determine the Effectiveness of

Portable Air Purifiers in Removing Ultrafine and Submicrometer Particles in an

Apartment

Environmental Science & Techonology. (Published online on July 24th, 2015)

56

Abstract

We validated the use of the mass balance model to determine the effectiveness of portable air

purifiers in removing ultrafine (<0.10 µm) and submicrometer particles (0.10-0.53 µm) in an

apartment. We evaluated two identical portable air purifiers, equipped with high efficiency

particulate air filters, for their performance under three different air flow settings and three target

air exchange rates: 0.60, 0.90 and 1.20 h-1. We subsequently used a mixed effects model to

estimate the slope between the measured and modeled effectiveness by particle size. Our study

showed that effectiveness was highly particle size-dependent. For example, at the lowest target

air exchange rate, it ranged from 0.33 to 0.56, 0.51 to 0.75, and 0.60 to 0.81 for the three air

purifier flow settings, respectively. Our findings suggested that filtration was the dominant

removal mechanism for submicrometer particles, whereas deposition could play a more

important role in ultrafine particle removal. We found reasonable agreement between measured

and modeled effectiveness with size-resolved slopes ranging from 1.11±0.06 to 1.25±0.07

(mean±s.e.), except for particles <35 nm. Our study design can be applied to investigate the

performances of other portable air purifiers as well as the influences of various parameters on

effectiveness in different residential settings.

Key words: Portable air purifier; Mass balance model; Size-resolved effectiveness; Ultrafine

particles; Submicrometer particles; Residential indoor air quality

57

Introduction

There is compelling evidence of the association between exposure to ambient fine particulate

matter (<2.5 µm) and adverse health effects, such as heart disease and mortality (1-3). People

spend the majority of their time indoors (4), where ambient fine particles penetrate and

contribute to the indoor particle level along with particles generated through cooking and other

indoor activities (5-7). A potentially effective way to remove fine particles indoors is to use

portable air purifiers (PAPs), especially in homes without in-duct filtration systems (8-10).

The effectiveness of PAPs has been defined as the proportion of particles they remove in the

indoor environment (9). It can be determined either by direct measurements of indoor particle

concentrations with and without the use of PAP(s), or by model predictions with known model

parameter values. The latter method is often conducted using the mass balance model (MBM),

where effectiveness can be expressed as a function of the Clean Air Delivery Rate (CADR). The

CADR is a metric that denotes the clean air that a PAP provides based on different particle sizes

and can be expressed as the product of the flow rate and the filtration efficiency of the portable

device (9, 11). Chamber studies have shown that PAPs equipped with high-efficiency particulate

air (HEPA) filters have the highest CADRs compared to many other PAP cleaning technologies

(12-14).

The highest modeled effectiveness for PAPs equipped with HEPA filters reported from

previous studies ranged from 0.60 to 0.95 (14-17). However, the accuracy of these values is

unknown, due to the lack of quantitative validation of the modeled results with empirical data in

a residential setting. For example, some studies have set the CADR values in scenarios that

assumed the PAP(s) were 100% efficient in removing particles of all sizes, neglecting air bypass

58

and short-circuiting situations that decrease CADRs (15). Others have predicted the effectiveness

using experimentally measured CADRs, but cited values from separate studies for the other

model parameters that were determined under different experimental conditions from each other

(14, 16, 17).

The overall objective of this study was to more comprehensively assess the effectiveness of

PAP(s) equipped with HEPA filters in removing ultrafine particles (UFPs) (<0.1 µm) and

submicrometer particles (0.10-0.53 µm) in an apartment, based on the commonly used mass

balance approach with the support of empirical data. Our four specific aims were to: (1)

experimentally determine the size-resolved PAP effectiveness using directly measured particle

concentrations with and without the operation of PAPs; (2) model and predict the size-resolved

effectiveness using individually measured model input parameters during the same test periods

as the first aim; (3) compare the modeled and experimentally-determined effectiveness values by

particle size; and (4) examine the effect of particle size, air exchange and PAP flow rate on PAP

effectiveness. To minimize measurement uncertainty, we conducted the study under well-

maintained test conditions in the apartment. We expect our findings to provide information on

the reliability of the MBM predictions, which contribute to more accurate assessment of the

reduction of residents’ exposure to indoor fine particles due to the use of PAP.


This study was conducted in a fully furnished, non-carpeted and occupied concrete floor

apartment unit in Cambridge, Massachusetts, during November 2011. The selected study area,

including the living room and the kitchen connected with a hallway, is approximately 34.8 m2.

59

During the tests, only the investigator was inside the house to attend to the experimental

operations, and none of the residents were present. The test conditions and placement of the

instruments and devices in the apartment unit were previously described (18). Two PAPs were

placed, one in the kitchen and the living room (Figure 2.1). More information regarding the

placement of the devices is in the supporting information.

Figure 2.1. Layout of the devices and instruments in the apartment.

Indoor Air Quality Model. We used the mass balance approach to model the concentration of

particles in the apartment with PAPs. We kept the indoor air well mixed using portable fans and

60

maintained a constant generation rate of the test particles (荊継沈). Penetration coefficient (鶏沈) was

assumed to be constant. We considered the concentrations of outdoor particles (系墜沈) constant

during the test periods, since they were much lower than those indoors. Based on these

conditions, the cumulative indoor particle concentration at any time t, 系沈岫建岻, can be expressed as

follows (17):

系沈岫建岻噺底牒日寵任日袋内曇日楠碇婆日盤な伐結貸碇婆日痛匪髪結貸碇婆日痛系沈岫 (2.1)

Where: 膏廷沈噺糠髪倦沈髪町肉勅日蝶 (2.2)

系沈(t) is the indoor concentration for particles in the 件痛朕 size category at time t (particles/cm3 or

#/cm3), g is the air exchange rate (h-1), 鶏沈 is the penetration coefficient of particles in the 件痛朕 size

category (dimensionless), 系墜沈 is the outdoor concentration for particles in the 件痛朕 size category

(#/cm3), 荊継沈 is the indoor generation rate of particles in the 件痛朕 size category (#/h), 撃 is the

effective room volume of the study zone (m3), 倦沈 is the deposition rate of particles in the 件痛朕 size

category (h-1), 芸捗 is the air flow rate through the HEPA filter (m3/h), 結沈 is the filtration efficiency

of PAP for particles in the 件痛朕 size category (dimensionless), and 系沈岫ど岻 is the initial

concentration at the start of measurement in the 件痛朕 size category.

When the generation of test particles and the concentrations of particles penetrating from

outdoors are constant, the indoor particle concentration at steady state can be estimated by the

following equation:

61

系沈岫タ岻噺底牒日寵任日袋内曇日楠碇婆日噺底牒日寵任日袋内曇日楠底袋賃日袋楢肉賑日楠 (2.3)

Using this equation, we can evaluate the effectiveness of PAPs in removing indoor particles

based on measured particle concentrations. Let us consider the parameter, 考沈, which is defined as

the ratio of steady state concentration in the presence of PAP (系沈岫タ岻牒凋牒) to that in the absence

of PAP (系沈岫タ岻牒凋牒博博博博博博). Then, the fraction of particles removed by the PAP compared to all three

removal mechanisms (air exchange, deposition and filtration) can be expressed as the relative

effectiveness (継沈) as follows (19):

継沈噺な伐考沈噺な伐寵日岫著岻鍋豚鍋寵日岫著岻鍋豚鍋博博博博博博博噺楢肉賑日楠底袋賃日袋楢肉賑日楠 (2.4)

One can thus compare the “experimental effectiveness” determined by direct measurement of 系沈岫タ岻牒凋牒 and 系沈岫タ岻牒凋牒博博博博博博 to the “modeled effectiveness” predicted using the experimentally-

determined input parameters (in the same apartment) shown on the far right hand side of eq

2.4.

Measurement and adjustment of air exchange rate. Three target air exchange rates (A) were

set to determine 継沈 of the PAP with repeated tests: A=0.60, 0.90, and 1.20 h-1. We used the sulfur

62

hexafluoride (SF6) tracer gas method in two consecutive phases: (1) the steady-state phase to

maintain constant air exchange rate, and (2) the exponential decay phase where the air exchange

rate was determined (18, 20). The SF6 concentration was measured in the kitchen and living

room by two SF6 monitors (Brüel & Kjær model 1302). The effective room volume (撃) can be

calculated from eq 5 as a quotient of the SF6 generation rate (芸聴庁展) and the product of SF6 steady

state concentration ( 系鎚鎚 ) (from phase 1) and the air exchange rate (from phase 2). Four

measurements from other preliminary tests under the same air mixing condition were

incorporated in this calculation to minimize uncertainties of the mean 撃 (total n=13). The

method was described more in detail in the supporting information.

撃岫兼戴岻噺町縄鈍展盤陳典ゲ朕貼迭匪底岫朕貼迭岻寵濡濡岫椎椎陳岻抜など滞岫喧喧兼岻 (2.5)

Particle generation and measurements. The particle generation system was the only particle

source indoors. The constant generation of a relatively high concentration of particles

accomplished three purposes: (1) to meet the model assumption (eq 2.1); (2) to ensure that the

generated particle concentrations were high enough to minimize the variability in (糠鶏沈系墜沈髪彫帳日蝶 )

resulted from the temporal variation of 鶏沈系墜沈 in eq 2.3; and (3) to achieve distinct difference in

steady state concentrations with and without the use of PAPs, thus minimizing the potential

effect of instrument error on particle measurements, especially when concentrations of certain

particle sizes were low.

The High-output Extended Aerosol Respiratory Therapy (HEART®) nebulizers (Westmed,

Inc., Tucson, Arizona) were used to aerosolize aqueous sodium chloride (NaCl) solution

63

(0.0375% by mass) to generate NaCl particles at three different rates of 16.58±0.32, 32.29±0.98,

and 46.24±0.58 g/h when using one, two and three nebulizers, respectively. These aerosol

generation rates were used to achieve similar steady state particle concentrations in the study

zone for A=0.06, 0.90, and 1.20 h-1, respectively. The nebulizers were refilled every two hours

with diluted solution to compensate for water loss due to evaporation during the nebulization

process.

The TSI Model 3936 scanning mobility particle sizer (SMPS) equipped with a Model 3785

Water-based Condensation Particle Counter (WCPC) was used to measure particles between

0.015 to 0.533 µm (mobility diameter) throughout the sampling days. Size-resolved count

concentrations were determined for consecutive 5-min intervals. NaCl crystals have a density of

2.2 g/cm3, and the count concentrations were converted to mass concentrations using the density

of the particles.

TSI Model 8520 DustTrak Aerosol Monitor and gravimetric sampling methods were also

used in the study. The DustTrak was used to test whether particles were uniformly distributed

across the study zone. The gravimetric sampling was performed to collect total airborne particle

mass under multiple particle steady states and was considered to be the “gold standard”.

Effectiveness for total particle mass, determined from gravimetric sampling, was subsequently

used to evaluate the accuracy of that from SMPS measurements. More detailed description of

these sampling methods can be found in the supporting information.

PAP characterization. Two PAPs equipped with HEPA filters (AP1008CH model, Coway Co.,

Ltd.) were tested in the laboratory for their size-resolved filtration efficiencies and flow rates

64

under four fan speeds (S1, S2, S3, and ST) before being deployed in the apartment. A sampling

system (SS) was set up in the laboratory, including three assemblies: (1) SS1, the PAP, (2) SS2,

the PAP with downstream ducting, and (3) SS3, the PAP with upstream and downstream ducting,

as well as the particle generation system and SMPS (Figure S2.1). The first two assemblies were

used to measure the flow rate inside the duct under different fan speeds, with compensation for

static pressure loss. The third was used to determine the size-resolved filtration efficiencies based

on direct measurements of particle concentrations both upstream and downstream of the PAP.

Average 結沈 from the six repeated sampling cycles were determined. Only S1, S3, and ST were

eventually used for the apartment tests. More description of the method is included in the

supporting information.

Particle deposition rates. Size-resolved particle deposition rates were reported in a companion

study (18). In brief, they were determined from the particle concentration decay curves during

the non-sourced period without PAP operation on the test days. The nonlinear approximation

procedure, PROC NLIN (SAS Inc. Cary, NC), was used in conjunction with the measurements to

estimate the values of 鶏沈系墜沈, 倦沈, and 系沈岫ど岻 based on eq 2.1.

Sampling plan. We conducted tests in nine days for three target air exchange rates in triplicate.

Each test day featured one stable 糠 and three PAP flow rates, and included two consecutive

sampling phases: (1) experimental determination of size-resolved particle deposition rates, using

a method we reported previously (18), and (2) measurements of size-resolved particle

concentrations at steady state without and with employing PAPs. In phase two, the particle

65

generation system was re-started and operated for the rest of the day (“sourced period”). After

reaching the steady state for particles, we turned on the PAP at the lowest flow setting and

monitored the particle concentration decay until it reached a new steady state. Subsequently, we

adjusted the PAP flow rate to the next highest setting and waited for the concentration to

decrease further to its corresponding steady state. Finally, we repeated the same cycle for the

highest flow setting. At each steady state, both before and during the use of PAPs, we allowed

for sufficient sampling time to perform several SMPS measurements and to collect enough mass

for gravimetric analysis. Daily indoor temperature and relative humidity (RH) were measured

using the Dwyer Series 485 Digital Hygrometer.

Data Analyses. Particle data were divided into 11 size categories: <25, 25-35, 35-45, 45-55, 55-

65, 65-80, 80-100, 100-150, 150-200, 200-300, and >300 nm. Data were presented using the

midpoint diameter for each category. The measured size-resolved effectiveness (な伐寵日岫著岻鍋豚鍋寵日岫著岻鍋豚鍋博博博博博博博) was determined from eq 2.4, using the average concentrations from each size category at steady

states (before and after PAPs use under three flow rates). The modeled effectiveness (

楢肉賑日楠底袋賃日袋楢肉賑日楠 )

was estimated using the values of the experimentally-determined parameters: 結沈 for each size

category is the average size-resolved filtration efficiency within that category and 芸捗結沈 represents the 系畦経迎沈 (m3/h) provided by two PAPs. The ratio of 系畦経迎沈 to the effective room

volume (撃), is defined as the “clean air replacement rate” (系畦迎迎沈 , h-1). Subsequently, this

parameter can be compared to particle deposition and air exchange rates.

66

We validated the MBM by comparing the measured and modeled effectiveness. On each test

day, effectiveness levels were measured for each of the three PAP flow rates. To account for the

day-to-day variability, we used a mixed effects model with a random intercept for test day to

compare the agreement (linear relationship) between the measured and modeled effectiveness.

The former was specified as the dependent variable while the latter was set to be the independent

variable. A slope of 1 from the mixed effects model is considered as the perfect agreement,

which indicates the MBM predicts consummately the actual effectiveness. We reported the 95%

confidence intervals of the slopes and compared them to ±10% of the ideal value (coefficient of

0.90-1.10). The analysis was performed in the statistical package SAS (SAS Inc. Cary, NC).

Results

Summary of measured parameters. Table 2.1 summarizes the measured parameters. Overall,

their variability was small, with a coefficient of variation (CV) of less than 5%, except for the

effective room volume, 9%, and the RH, 22%. The average effective room volume, 98 m3, was

used to calculate the modeled effectiveness. The total flow rate through the two PAPs under

three flow settings were 195, 387, and 540 m3/h for S1, S3, and ST, respectively. The filtration

efficiency of the PAPs was size-dependent and decreased with flow rate. Based on size-resolved

particle count concentration, it ranged from 0.77-0.90 for S1, 0.73-0.89 for S3, and 0.62-0.81 for

ST (Figure 2.2). Under the air exchange rates tested, the average size-resolved CADRs of the

two PAPs were 75-88, 128-159, and 167-219 m3/h for S1, S3, and ST, respectively.

Figure 2.3 shows an example of the real-time measurement and records over one sampling

day for particle and SF6 concentrations, the flow rate of PAPs, and the operation of particle

67

generation system. As shown, particle levels plateaued five times, and the corresponding steady

state concentrations reflected the changes in PAP flow rate. The first particle concentration decay

was due to the combined effect of particle removal both by deposition and air exchange while the

others reflected additional removal by PAPs.

During the test runs, the air exchange rate was kept constant. This was achieved by adjusting

the sealing (masking tape) of the gaps around the windows and doors throughout the sampling

day, keeping the SF6 steady state concentration within 5% of the average value. For most of the

experiments the differences between the SF6 steady state concentrations in the living room and

the kitchen at any time were less than 10%, suggesting that the air inside the apartment was well

mixed.

Figure 2.2. Average size-resolved filtration efficiencies of the 2 PAPs under 3 flow settings.

Particle size (nm)

Filt

ratio

n e

ffic

ien

cy (

un

itle

ss)

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

100 200 300 400

S1S3ST

68

Table 2.1. Summary of measured parameters

Parameter N Mean s.d.* Min Max

Target air exchange rate (h-1)

0.60 3 0.61 0.00 0.60 0.62

0.90 3 0.91 0.01 0.90 0.92

1.20 3 1.22 0.01 1.21 1.24

Effective room volume (m3) 13 97.7 9.02 85.9 122

Relative humidity (RH, %) 9 37.4 8.32 25.0 49.1

Temperature (°C) 9 24.1 0.87 22.6 25.3

PAP flow rate (m3/h)**

S1 PAP#1 6 95.1 1.14 94.2 97.3

PAP#2 4 99.6 1.09 98.8 101

S3 PAP#1 6 200 0.63 199 201

PAP#2 4 186 0.77 185 187

ST PAP#1 6 274 1.59 271 276

PAP#2 4 266 0.89 265 267

Average deposition rate (h-1)***

Count 110 2.20 0.03 (s.e.) - -

* s.d.= standard deviation; s.e.=standard error

** Total flow rate provided by the two PAPs

*** Average particle deposition rate across all 9 test days (18).

In addition to air exchange rate, the MBM assumptions also required the overall levels of

particle sources and the resulting indoor concentrations to be constant. In general, CV for particle

concentrations exiting the generation system was <10%, except for particles in the smallest and

the largest size bins (CV<20%). On the other hand, particles penetrating from outdoors

consisted of 6.5-26% of the size-resolved steady state concentrations at the three target air

69

exchange rates without PAP operation. Thus, variability from the outdoor particle sources was

weakened when viewed together with that from the dominant indoor source. More details can be

found in the supporting information with examples of CV and size distribution for generated

particles and the resulting indoor steady state concentrations (Figure S2.2-2.5).

Figure 2.3. An example of the continuous measurements and device operation profiles over one

sampling day at g=0.91 h-1 for the total particle concentration (系岫建岻), the total particle generation

rate (荊継痛墜痛銚鎮), the total flow rate of the 2 PAPs (芸捗), and the 鯨繋滞 concentration.

Particle removal rate by different mechanisms. The size-resolved particle removal rates (h-1)

due to particle deposition (倦沈), air exchange (ACH) and use of PAPs (系畦迎迎沈) were compared

(Figure 2.4). These quantities were used to determine the modeled effectiveness. 系畦迎迎沈

Elapsed time (h)

SF

6 (

pp

m)

Qf (m

3/h

)IE

tota

l (g

/h)

C(t

) (#

/cm

3)

20

40

60

2 4 6 8 10

10

03

00

50

0

2 4 6 8 10

10

20

30

40

2 4 6 8 10

50

00

20

00

0

2 4 6 8 10

70

increased substantially with flow rate, and its size-resolved curve shape was similar to that

observed for filtration efficiencies (Figure 2.2).

Figure 2.4. Size-resolved particle removal rates: filtration by PAPs, deposition, and air exchange. 系畦迎迎沈 (h-1) is the size-resolved clean air replacement rate, equal to 系畦経迎沈【撃. 系畦迎迎な, 系畦迎迎に

and 系畦迎迎ぬ corresponded to the flow rates of 195, 387, and 540 m3/h, respectively. The air

exchange rates were the average values under three target air exchange rates: ACH1, ACH2, and

ACH3 are the target air exchange rate of 0.60, 0.90, and 1.20 h-1, respectively. k1, k2 and k3 are

the average particle deposition rates at ACH1, ACH2, and ACH3, respectively.

The relative contribution of the three removal mechanisms differed by particle size. For

submicrometer particles, 系畦迎迎沈 was larger than the removal rates due to particle deposition and

air exchange, even at the lowest flow setting. For UFPs, removal by particle deposition became

more dominant as particle size decreased, which exceeded the highest 系畦迎迎沈 observed for

Particle size (nm)

Re

mo

va

l ra

te (

h-1)

1

2

3

4

5

100 200 300 400

ACH1

ACH2

ACH3

CARR1

CARR2

CARR3k1k2k3

71

particles <25 nm. This trend was reflected in the modeled size-resolved effectiveness of PAPs

which is defined as the ratio of 系畦迎迎沈 to the total particle removal rates.

Measured effectiveness. Figure 2.5 presents the size-resolved effectiveness measured by SMPS

for three PAP flow rates at three target air exchange rates. For all particle sizes, the effectiveness

increased with flow rate. At A=0.60 h-1, the effectiveness was 0.33-0.56 for S1, 0.51-0.75 for S3,

and 0.60-0.81 for ST, and decreased slightly when g increased to A=0.90 h-1. At A=1.20 h-1, the

effectiveness decreased to 0.20-0.54, 0.30-0.71, and 0.49-0.77 for the three flow rates,

respectively. Air exchange rate appeared to influence the effectiveness, as indicated by the

observed higher variability for A=1.20 h-1, compared to that of the two lower air exchange rates.

Measured effectiveness for the total particles was also determined using both the count and

mass concentrations from SMPS across the 9 test days. By count, it was 0.39-0.53 for S1, 0.56-

0.64 for S3, and 0.67-0.72 for ST. By mass, the respective values were 0.49-0.59, 0.68-0.57, and

0.73-0.82.

Comparison between measured and modeled effectiveness. Overall, there is good agreement

between measured and modeled effectiveness for all particle sizes, except for those <35 nm for

which the largest discrepancy was found (Figure S2.6). For most particle sizes, the 95%

confidence intervals of the slopes contained the upper bound of ±10% of the ideal situation

(coefficient=1.10) (Figure 2.6 and Table S1.1). Exceptions were seen for particles <35 nm and

150-300 nm; however, the slope for the latter was only marginally above 1.1, suggesting

reasonable agreement.

72

Figure 2.5. Measured size-resolved effectiveness for the three PAP flow rates (芸捗= 195, 387, and

540 m3/h) under three target air exchange rates (A=0.60, 0.90 and1.20 h-1).

Particle size (nm)

Me

asu

red

effe

ctiv

en

ess

0.2

0.4

0.6

0.8

1.0

100 200 300 400

A=0.60 /h

100 200 300 400

A=0.90 /h

100 200 300 400

A=1.20 /h

Qf=195 m3

Qf=387 m3

Qf=540 m3

73

Figure 2.6. The slopes and their 95% confidence intervals by particle size obtained from the

mixed effects model. The shaded area represents ±10% of the ideal coefficient of 1 (0.90-1.10).

Discussion

Our findings showed that the PAP effectiveness was highly size-dependent and was well

predicted by the MBM under well-mixed air conditions in an apartment. Thus we can use the

measured model parameters to assess the influence of each particle removal mechanism on

effectiveness by size. Overall, effectiveness followed the trend of 系畦迎迎沈, except for UFPs where

the trend was dominated by elevated deposition rates. The considerably high deposition rates

may be explained by the enhanced air mixing conditions due to the use of portable fans.

Removal was higher for UFPs whose deposition is governed mostly by Brownian diffusion (18).

However, the average deposition rate based on total particle count concentrations in our study

74

(倦=1.17 h-1) was comparable to the average rate of 1.29 for 0.3-1 µm particles estimated in four

homes with smokers (21). Furnishing may also contribute to the elevated deposition rates by

providing additional surface area for deposition (22).

Using the MBM in residential spaces for direct effectiveness measurements is challenging

because of the difficulty to achieve steady state. Most studies measured CADRs in a test

chamber and subsequently used the steady state MBM to estimate the effectiveness for

hypothetical residential indoor spaces (14, 16, 17). Table S2.2 (in the supporting information)

summarizes some of this information, but it is not limited to studies evaluating PAPs with HEPA

filters. In general, studies that modeled effectiveness used the size-resolved deposition rates from

Riley et al. and an air exchange rate ranging from 0.2 to 0.5 h-1, but with varying CADRs and

room volumes (23).

For A=0.60 h-1 the maximum size-resolved effectiveness was observed in submicrometer size

region, and was in general agreement with, but lower than, the maximum effectiveness reported

from other studies (0.60-0.95) (Figure 2.7). The lower effectiveness was mainly due to the

elevated particle deposition rate measured in our study, which was generally at least one order of

magnitude higher than that reported by Riley et al. (23). The overall comparison, on the other

hand, suggested that effectiveness is more comparable across studies for submicrometer particles

where size-resolved effectiveness profile is more flat than that of UFPs.

Because of the size-dependent nature of the effectiveness, the overall effectiveness (all sizes

together) is sensitive to typically wide variations in both particle count and mass size

distributions, which may limit its generalizability and make comparisons more difficult. For

example, using different particle size distributions for dust-mite allergens, Fisk et al. found a

75

20% difference in the estimated mass-based effectiveness (15). For the test particles used in this

study, the measured effectiveness based on the total particle count concentration was lower than

that from the total mass concentration (using SMPS data). This was due to the relatively high

numbers of UFPs which were removed less efficiently by PAPs. The effectiveness determined

using the SMPS total mass concentration was reliable and in good agreement with that from the

gravimetric analysis (Figure S2.7 and supporting information).

Figure 2.7. Relationship between effectiveness and CARR (14-17).

The experimentally determined CADRs were comparable to those reported by manufacturers

for commercially available PAPs (Figure S2.8) (24). Our 結沈 exhibited a similar size-dependent

curve to that reported from two recent studies; however, our values were higher than theirs (0.2-

0 5 10 15 20

0.0

0.2

0.4

0.6

0.8

1.0

CARR (h-1

)

Eff

ec

tiv

en

es

s

Fisk et al. (2002)

Ward et al. (2005)

Waring et al.(2008)

Sultan et al. (2011)

This study (2015)

76

0.8) (14, 16). As expected that HEPA filters have efficiencies >99.97% (25), the observed lower

efficiencies could be explained by air bypassing the filter inside of the PAP and/or the short

circuiting of filtered air (9).

To our knowledge, this is the first study to validate the steady state MBM for predicting the

size-resolved effectiveness of PAPs in a residential setting. Matson validated the MBM using the

non-steady state approach by comparing the difference between calculated and measured mean

values of indoor ultrafine particle concentrations (26). The difference ranged from -16 to 32%

when using deposition rates determined from two residential buildings and two offices in the

model. Nevertheless, no PAP was included in this validation; thus, no direct comparison between

the measured and modeled effectiveness could be inferred from the study. In the current study,

we regressed the measured effectiveness on the modeled one using a mixed effects model. The

slopes from the regression (n=27) by particle size ranged from 1.11±0.06 (80-100 nm) to

1.50±0.14 (<25 nm), providing a quantitative and intuitive way in validating the steady state

MBM.

One of the limitations of our study was the enhanced air mixing condition. The mechanical

mixing enabled achieving a well-mixed condition, but at the same time it might have enhanced

particle deposition, resulting in lower effectiveness compared to other studies. Conversely, the

enhanced air mixing may have increased effectiveness by minimizing short circuiting of filtered

air, compared to conditions without using portable fans. The net effect of these opposite

influences in effectiveness is unknown.

Another limitation was that the findings were based on short-term performance of the PAPs.

One study reported a 25% decrease in CADR after approximately a month of use in a residential

77

bedroom due to the reduction of the PAP air flow rate (13). Another study showed a drop of 7-

14% in air flow after 2 months of continuous operation in smokers’ homes, which was equal to

blocking 33% of the filter area based on simulated tests. The drop was a result of the heavily

loaded pre-filter with fibrous gray dust (21). Nevertheless, decrease in flow rate can be

minimized by routine maintenance such as cleaning the pre-filters regularly and changing the

loaded HEPA filters.

Last but not least, care should be taken when selecting the values of particle deposition rates

for predicting effectiveness because studies have demonstrated possible influences of g on

deposition rates, either on their levels or size-resolved profiles (27-29). The deposition rate we

used in prediction scenarios was the average value determined under a range of 糠 (0.61-1.24 h-1)

which was more or less representative of the overall rates.

In this study, we validated the steady state MBM based on the reasonably good agreement

between the measured and modeled size-resolved effectiveness. We found that effectiveness was

highly size-dependent, and PAP was the dominant removal mechanism for submicrometer

particles, whereas deposition could play a more important role in UFPs removal. The study

design can be applied to investigate the performances of other PAPs as well as the influences of

various parameters on their effectiveness in different residential settings.

78

Bibliography



(2) Pope, C.; Thun, M.; Namboodiri, M.; Dockery, D.; Evans, J.; Speizer, F.; Heath, C.

Particulate Air-Pollution as a Predictor of Mortality in a Prospective-Study of Us Adults.

American Journal of Respiratory and Critical Care Medicine 1995, 151 (3), 669-674.






10.1038/sj.jea.7500165.



10.1021/es990348y.

(6) Wallace, L.; Emmerich, S.; Howard-Reed, C. Source strengths of ultrafine and fine particles

due to cooking with a gas stove. Environ. Sci. Technol. 2004, 38 (8), 2304-2311;

10.1021/es0306260.

79


particles in a residence for 18 months in 1999-2000. J. Air Waste Manage. Assoc. 2002, 52 (7),

828-844.

(8) Barn, P.; Larson, T.; Noullett, M.; Kennedy, S.; Copes, R.; Brauer, M. Infiltration of forest

fire and residential wood smoke: an evaluation of air cleaner effectiveness. Journal of Exposure

Science and Environmental Epidemiology 2008, 18 (5), 503-511; 10.1038/sj.jes.7500640.

(9) Shaughnessy, R. and Sextro, R. What is an effective portable air cleaning device? A review.

Journal of Occupational and Environmental Hygiene 2006, 3 (4), 169-181;

10.1080/15459620600580129.

(10) Sublett, J.L. Effectiveness of Air Filters and Air Cleaners in Allergic Respiratory Diseases:

A Review of the Recent Literature. Current Allergy and Asthma Reports 2011, 11 (5), 395-402;

10.1007/s11882-011-0208-5.

(11) Association of Home Appliance Manufacturers (AHAM) <br />Standard ANSI/AHAM AC-

1-2006, method for measuring performance of portable household electric room<br />air

cleaners. Association of Home Appliance Manufacturers: Washington, DC, 2006; .

(12) Offermann, F.; Sextro, R.; Fisk, W.; Grimsrud, D.; Nazaroff, W.; Nero, A.; Revzan, K.;

Yater, J. Control of Respirable Particles in Indoor Air with Portable Air Cleaners. Atmos.

Environ. 1985, 19 (11), 1761-1771; 10.1016/0004-6981(85)90003-4.

80

(13) Shaughnessy, R.; Levetin, E.; Blocker, J.; Sublette, K. Effectiveness of Portable Indoor Air

Cleaners - Sensory Testing Results. Indoor Air-International Journal of Indoor Air Quality and

Climate 1994, 4 (3), 179-188; 10.1111/j.1600-0668.1994.t01-1-00006.x.

(14) Sultan, Z.M.; Nilsson, G.J.; Magee, R.J. Removal of ultrafine particles in indoor air:

Performance of various portable air cleaner technologies. Hvac&R Research 2011, 17 (4), 513-

525; 10.1080/10789669.2011.579219.

(15) Fisk, W.; Faulkner, D.; Palonen, J.; Seppanen, O. Performance and costs of particle air

filtration technologies. Indoor Air 2002, 12 (4), 223-234; 10.1034/j.1600-0668.2002.01136.x.

(16) Waring, M.S.; Siegel, J.A.; Corsi, R.L. Ultrafine particle removal and generation by

portable air cleaners. Atmos. Environ. 2008, 42 (20), 5003-5014;

10.1016/j.atmosenv.2008.02.011.

(17) Ward, M.; Siegel, J.; Corsi, R. The effectiveness of stand alone air cleaners for shelter-in-

place. Indoor Air 2005, 15 (2), 127-134; 10.1111/j.1600-0668.2004.00326.x.

(18) Lee, W.; Wolfson, J.M.; Catalano, P.J.; Rudnick, S.N.; Koutrakis, P. Size-Resolved

Deposition Rates for Ultrafine and Submicrometer Particles in a Residential Housing Unit.

Environ. Sci. Technol. 2014, 48 (17), 10282-10290; 10.1021/es502278k.

(19) Nazaroff, W.W. Effectiveness of air cleaning technologies, In Anonymous ; SIY Indoor Air

Information Oy: Helsinki, Finland., 2000; Vol.2 pp. 49-72.

81

(20) ASTM Standard Test Method for Determining Air Change in a Single Zone by Means of a

Tracer Gas Dilution. E741-00. American Society for Testing and Materials International: West

Conshohocken, Pennsylvania, 2001; .

(21) Batterman, S.; Godwin, C.; Jia, C. Long duration tests of room air filters in cigarette

smokers' homes. Environ. Sci. Technol. 2005, 39 (18), 7260-7268; 10.1021/es048951q.



(11), 1811-1819; 10.1016/S1352-2310(02)00157-7.

(23) Riley, W.; McKone, T.; Lai, A.; Nazaroff, W. Indoor particulate matter of outdoor origin:

Importance of size-dependent removal mechanisms. Environ. Sci. Technol. 2002, 36 (2), 200-

207; 10.1021/es010723y.

(24) Anonymous Association of Home Appliance Manufacturers (AHAM) Verification Program.

(25) Department of Energy (DOE) Specification for HEPA Filters Used by DOE Contractors.

DOE-STD-3020-97 <br /> . U.S. Department of Energy: Washington, DC, 1997; .

(26) Matson, U. Comparison of the modelling and the experimental results on concentrations of

ultra-fine particles indoors. Build. Environ. 2005, 40 (7), 996-1002;

10.1016/j.buildenv.2004.09.001.

(27) Allen, R.; Larson, T.; Sheppard, L.; Wallace, L.; Liu, L. Use of real-time light scattering

data to estimate the contribution of infiltrated and indoor-generated particles to indoor air.

Environ. Sci. Technol. 2003, 37 (16), 3484-3492; 10.1021/es021007e.

82

(28) He, C.R.; Morawska, L.; Gilbert, D. Particle deposition rates in residential houses. Atmos.


(29) Rim, D.; Wallace, L.A.; Persily, A.K. Indoor Ultrafine Particles of Outdoor Origin:

Importance of Window Opening Area and Fan Operation Condition. Environ. Sci. Technol.

2013, 47 (4), 1922-1929; 10.1021/es303613e.

83

Supporting information (SI)

Placement of the devices. The apartment layout is fairly symmetric with the bathroom in the

middle, one bedroom and the living room on one side, and the second bedroom and the kitchen

on the other side (Figure 2.1). Both the particle and SF6 generation systems were placed in the

middle of the hallway against the wall. A table fan was used to disperse the SF6 from the source

into the hallway. One box fan was set up at one end of the hallway to blow the air towards the

living room while a second one was set to blow the air towards the kitchen. This arrangement

was to facilitate the distribution of the particles and SF6 to both sides of the room from the

sources.

In the living room, another box fan was placed in the diagonal from the hallway and blew

towards the neighboring corner for air mixing. We located one SF6 monitor and the gravimetric

sampling assembly in the center of the living room, while the Scanning Mobility Particle Sizer

(SMPS) was located slightly further towards the wall. The air purifier was placed at a moderate

distance away from the box fan in the hallway and a shorter distance from the SF6 monitor. A

table fan was employed to blow at the outlet of the air purifier to move filtered air to the room

and minimize re-circulation of the filtered air back into the inlet without being distributed in the

room. The placement of instruments and the fans were identical in the kitchen as it was for the

living room, except for the devices for particle measurements. Under this arrangement, the

particle concentration was determined to be uniform in the study zone based on DustTrak

measurements in additional tests.

84

Adjustment to achieve constant air exchange rate. There was no air conditioning system in

the apartment, and heating was provided by hydronic radiant heating system. Ventilation was

normally through the opening or cracks of windows, doors, and the two small vents that

exhausted indoor air to outdoors. During the sampling period, these openings were tightly closed

and sealed using masking tape. To minimize potential bias in indoor measurements due to the

interference of air inside the space excluded from the study zone, namely the bedrooms and the

bathroom, the windows of the bedrooms were fully open to ensure the particle concentration was

as close as that from outdoors. In principle, adjustment on the taping was made on the door that

led to immediate outdoor environment, which was the patio door in the living room; whereas on

the kitchen side, the bedroom door was a better choice because of limited ventilation in the lobby

outside the entrance door. By removing the masking tape or re-sealing the gaps, we continuously

adjusted the steady state SF6 concentration to the constant level that corresponded to the desired

air exchange rate. More details of the method can be found elsewhere (18).

Measurements of particles. The DustTrak was originally considered as one of the primary

devices for particle measurements in our study due to its portability. However, we found the

calibration ratio between the DustTrak and SMPS measurements varied by day and the various

levels of steady state concentrations of the particles on site, suggesting that the DustTrak

measurement was highly sensitive to the change in particle size distribution. As a result, it was

used only to check if the particles were uniformly distributed across the study zone by measuring

particle concentrations at steady state in multiple locations in the apartment.

85

The gravimetric sampling method provides the average particle mass concentrations over the

sampling time. It was used to collect total suspended particles over the multiple particle steady

states to verify the mass-based effectiveness determined from SMPS measurements (as

integrated samples). A 37 mm 2 たm PTFE filter was placed inside the Harvard Impactor (Air

Diagnostics and Engineering, Inc., Harrison, ME) without the impaction plate, where the room

air was drawn through the filter by a vacuum pump at a post-filter flow rate of 25 L/min. Two

identical gravimetric samplers were placed side by side for all measurements. The particle

weight was determined by the pre- and post-weighing of filters using an electronic microbalance

with laboratory blanks. The filter concentration was calculated from the net mass and the

sampling air volume (which equals to flow rate multiplied by sampling time). The concentrations

acquired from the two parallel samplers were averaged to get a representative mass concentration

of the corresponding steady state.

Characterization of the portable air purifier (PAP). The air purifier had one slit (33 cm x3

cm) on each side that allowed air to be drawn in, pass through a series of particle filters (pre-

filters and the HEPA filter), and return to the room from the exit on the rear side of the device. A

removable front cover was used to access the particle filters for routine replacement. To avoid

damaging the cover for the purpose of laboratory tests, we made a replacement cover with the

same size and a hole in the middle for upstream duct connection. A sampling system (SS) which

included three subsets was set up in the laboratory for the measurements (Figure S2.1). The flow

rates were determined by first recording the pressure drop across the filter of the free-standing air

purifier (SS1), followed by measuring the centerline air velocity inside the duct of SS2 using an

air velocity meter (Velocicalc Model 8345, TSI. Inc., Shoreview, MN). A blower with adjustable

86

fan speed was installed in-line at the end of the duct to overcome the resistance and compensate

for the additional pressure loss due to the addition of the duct, using pressure drop measurements

from SS1 as reference. The measurements were made from the lowest (S1) to the highest (ST)

fan speed of the air purifier for five cycles for the pressure drop and six cycles for the centerline

air velocity in the duct where the average values were used in conjunction with the cross-

sectional area of the duct to determine the four flow rates for each air purifier. The resulting flow

rates were considered as the actual flow rates for the PAP and were set as the target flow rates

when we determined the size-resolved filtration efficiencies with SS3.

Filtration efficiency (結沈) is a measure of the proportion of particles removed by the PAP and

is dependent on particle size (eq S2.1). It was known to have an inverse relationship with air

speed (flow rate), and thus should be determined separately under each fan speed (30).

結沈噺な伐寵日┸日韮如賑禰寵日┸任祢禰如賑禰 (S2.1)

Where :

系沈┸沈津鎮勅痛 is the particle concentration in the pre-filtered air for particles in the 件痛朕 size category

系沈┸墜通痛鎮勅痛 is the particle concentration in the filtered air for particles in the 件痛朕 size category

87

To measure 系沈┸沈津鎮勅痛 and 系沈┸墜通痛鎮勅痛, we added an additional duct to the upstream of the PAP and

adjusted the downstream end-of-the-duct fan to compensate for the flow rates to the target

values. Ultrafine and submicrometer particles were generated using the same nebulization system

as the one in the apartment. The particles were dried with clean, dry air in the mixing chamber

before entering the upstream duct. Each of the upstream and downstream ducts was equipped

with an isokinetic sampling port, inside which the air velocity was nearly the same as that in the

duct, to allow measurements of particle concentrations using the SMPS. The use of isokinetic

probes assures that the size distribution of the measured particles was not distorted in the

measurement process. Each sampling cycle consisted of alternate measurement of particle

concentrations inside the ducts by switching between upstream and downstream sampling ports.

Averages of the size-resolved efficiencies from the six repeated sampling cycles were

determined for each of the four air purifier fan speeds, respectively.

Constant particle generation rate and steady state concentration. Figure S2.2 shows the

coefficient of variation (CV, %) for the size-resolved particle concentrations exiting the three

different combinations of nebulizers, which were measured in the laboratory. The CV was

generally low, except for the smallest and the largest size bins. This was because these two size

bins had lower concentrations (at the tails of the size distribution). Overall, CV <10% suggested

stable and constant generation of particles.

Figure S2.3 shows the CV of the steady state concentrations in the apartment without PAP

operation and at the PAP fan speed of S1, S3, and ST in the nine test days. The variability profile

was similar to that in Figure S2.2. Additionally, both the use of PAPs and higher air exchange

rate could possibly contribute to the variability of concentrations in the largest particle size bin

88

(particles >300 nm). Filtration efficiency was higher for large particles. The highest three values

of CV for particles >300 nm were under the target air exchange of 1.20/h with the use of air

purifiers on the highest two speeds.

Figure S2.4 shows the particle size distribution using the same data as those from Figure

S2.2. Good agreement was observed from the repeated measurements of particle concentrations

and their distribution on the same test day.

Figure S2.5 displays the particle size distribution at steady state under four PAP flow settings

(Qf= 0, 195, 387, and 540 m3/h) in the apartment. We used measurements from one test day as an

example to show the evolution of the size distribution with and without the use of PAPs. The

distribution exhibited similar trend but slightly different among the test days, partially because

we used different combinations of nebulizers for different target air exchange rates (as shown in

Figure S2.4) and of the change in relative humidity on site.

Validation of the SMPS mass data. We calibrated the effectiveness for total particle mass

concentrations from the SMPS measurements using that from the gravimetric method. Figure

S2.7-(a) shows the calibration using all measurements. However, measurements from

gravimetric sampling for one particular day were overall problematic with unreasonably low

mass concentrations. We removed those three measurements and found that the r-squared value

increased from 0.79 to 0.89, but the coefficient (slope) between the two methods remained stable

at around 0.95-0.96 (Figure S2.7-(a) and (b)). We concluded that measurements from SMPS

were reliable.

89

Table S2.1. Slopes from the mixed effects model by particle size.

Size (nm) Slope Standard Error

<25 nm 1.50 0.14

25-35 nm 1.40 0.11

35-45 nm 1.25 0.07

45-55 nm 1.19 0.06

55-65 nm 1.18 0.05

65-80 nm 1.15 0.06

80-100 nm 1.11 0.06

100-150 nm 1.12 0.05

150-200 nm 1.19 0.04

200-300 nm 1.17 0.03

>300 nm 1.14 0.05

Total 1.16 0.05

90

Table S2.2. Summary of studies that reported CADRs or effectiveness based on the steady state MBM for PAPs equipped with HEPA filters for removing ultrafine and submicrometer particles (12-17).

Study Method Room setting

Particle source and size (µm)

Particle monitor#

PM sample type

Type of PAP with HEPA

filter

Filtration efficiency#

*

Flow rate (m3/h)

CADR

(m3/h)* Effectiveness*

Offermann et al.

(1985)

Measured CADR

CADR:

35.1 m3 test room $

ETS:

0.09-1.25

OPC

CADR:

size-resolved

1

115±17 % for PM of 0.45 µm

173-343

306±14 at medium speed**

N.A.

Shaughnessy et al.

(1994)

Measured CADR

CADR: 24.8m3 AHAM test chamber $

ETS $$ LAS CADR: integrated

HEPA filter: 1;

HEPA-type filter: 2 $$$

83±1 % for HEPA

filter**; 96±1 & 82±2 %

for HEPA-

type filters**

492,306 and 342

407.4±4.8 for HEPA

filter**; 212.4±3.6

& 276.6±6.6

for HEPA-

type filters**

N.A.

Fisk et al.

(2002)

Modeled effectiveness

Modeling with and without

HVAC system in indoor space

ETS and outdoor fine PM $$

N.A. Effectiveness: integrated

1 Assume to be 100%

N.A. N.A. Overall: approx. 0.75-

0.95 for CARR of 2-10 h-1

when HVAC system was off

91

(Table S2.2 continued)

Ward et al.

(2005)

Measured CADR;


CADR: 11 m3 stainless steel chamber;

Effectiveness: 377m3 house with HVAC system

Incense burning: 0.1-2

OPC CADR:

size-resolved

Effectiveness: size-resolved

CADR: 3***;

Effectiveness: 1 type with 3

devices

N.A. N.A. Size-resolved:

271-332***

Size-resolved: maximum at 0.90 for three and at 0.75 for one activated

device(s) when HVAC system

was off;

Waring et al.

(2008)

Measured CADR;


CADR: 14.75 m3 stainless steel chamber;

Effectiveness: 50 m3 room and 392 m3 house without HVAC system

Incense burning: 0.013-0.514

SMPS CADR:

size-resolved

2 < 60 % for PM < 0.20

µm and increased slightly

for PM > 0.20 µm

309 and 571

Size-resolved: 95-259

and 203-481

Size-resolved (> 0.05 µm): 0.80-0.90 in a room; 0.40-0.60 in whole

house

92


Sultan et al.

(2011)

Measured CADR;


CADR: 55m3 stainless steel chamber;

Effectiveness: 29 m3 room without HVAC system

NaCl: 0.014-0.533

SMPS CADR: integrated &

size-resolved;

Effectiveness: integrated & size-resolved

3 Overall:

0.30-0.62;

Size-resolved

(40-100nm):

0.30-0.70

248, 454 and 970

Overall: 60, 295 and 444

Overall: 0.68-0.95

This study

(2015)

Measured CADR;

Measured and modeled effectiveness

CADR: laboratory system

Effectiveness:

98 m3 $ apartment without HVAC system

NaCl: 0.015-0.533

SMPS;

GS

CADR:

size-resolved;

Effectiveness: integrated & size-resolved

1 type with 2 devices

Size-resolved: 0.77-0.90

for S1; 0.66-0.82

for S3; 0.62-0.81

for ST

S1=98, S3=194,

and ST=270 (average of the 2 devices)

Size-resolved: 75-88 for S1, 128-159 for S3, and 167-219 for ST

Overall (count)+:

0.39-0.53 for S1, 0.56-0.64 for S3, 0.67-0.72 for ST;

Size-resolved+: maximum at 0.72 for S1,

0.81 for S3 and 0.85 for ST.

# OPC represents a general category of optical particle counters; SMPS; PTFE filters were used in conjunction with cassettes to sample total particles; LAS is the laser aerosol spectrometer; GS is gravimetric sampling method.

* The “overall” measures were based on the removal of total particulate matter (PM); whereas the size-resolved CADR was based on particle size.

93


** The range was ± 95% confidence limits from Shaughnessy et al.,13 and ± 90% confidence limits from Offermann et al. (12).

*** The CADR for each particle size was the average value from the three PAPs tested which was used to model the effectiveness (17).

$ It was the effective room volume or the net air space volume. AHAM stands for Association of Home Appliance Manufacturers.

$$ Studies also included other particle sources.

$$$ HEPA-type filters with an efficiency of 95% for particles of 0.3 µm.

+ Determined from SMPS measurements.

94

Figure S2.1. Air Purifier Testing System. All components of the sampling system were scaled

based on the actual dimension, where the duct diameter was 0.15 m (6 inches).

95

Figure S2.2. An example of the coefficient of variation (%) of the size-resolved concentrations for the generated particles using different nebulizer combinations (one=Neb016, two=Neb016+018, and three=Neb all) in the laboratory. Each nebulizer combination was sampled twice, each for a 2-hour period (24 repeated measurements). The data were from the same test day (one set of the laboratory tests).

96

Figure S2.3. Coefficient of variation (%) of indoor particle concentration by particle size at each

steady state (four PAP flow settings (Qf= 0, 195, 387, and 540 m3/h)) in the apartment for the

nine test days.

0.00

5.00

10.00

15.00

20.00

25.00

30.00

35.00

0 100 200 300 400 500

CV

(%)

Midpoint diameter (nm)

97

Figure S2.4. Size distribution of the generated particles using data from the same test as those in

Figure S2.2.

0

200

400

600

800

1000

1200

1400

1600

0 100 200 300 400 500 600

Co

nce

ntr

ati

on

(#

/cm

3)

Midpoint diameter (nm)

Neb016-1

Neb016-2

Neb016+018-1

Neb016+018-2

Neb016+017+018-1

Neb016+017+018-2

98

Figure S2.5. The particle size distribution at steady state under four PAP flow settings (Qf= 0,

195, 387, and 540 m3/h) in the apartment, using one test day (g= 0.61/h) as an example.

0

50

100

150

200

250

300

350

400

0 100 200 300 400 500 600

Ste

ad

y s

tate

co

nce

ntr

ati

on

(#

/cm

3)

Midpoitn diamter (nm)

Qf=0 (m^3/h)

Qf=195

(m^3/h)

Qf=387

(m^3/h)

99

Figure S2.6. A series of scatter plots for the size-resolved effectiveness between measured and

modeled values. The modeled effectiveness for total particles was adjusted for particle size

distribution using that from the steady state concentrations in the apartment prior to the operation

of PAPs.

Modeled Effectiveness

Measure

d E

ffectiveness

0.2

0.4

0.6

0.8

0.20.40.60.8

<25 nm 25-35 nm

0.20.40.60.8

35-45 nm 45-55 nm

55-65 nm 65-80 nm 80-100 nm

0.2

0.4

0.6

0.8

100-150 nm

0.2

0.4

0.6

0.8

150-200 nm

0.20.40.60.8

200-300 nm >300 nm

0.20.40.60.8

Total

Qf=195 m3

Qf=387 m3

Qf=540 m3

100

Figure S2.7. Validation of effectiveness based on total particle mass concentrations from SMPS

using those from gravimetric analysis. (a) Using all data (n=27), (b) removing problematic data

from gravimetric measurements (n=24).

101

Figure S2.8. Comparison of the measured size-resolved CADRs for three PAP flow settings in

the current study to the integrated CADRs for ETS from the database of Association of Home

Appliance Manufacturers (AHAM) that contained a total of 263 devices. The size-resolved

CADRs were calculated from the average flow rates and the size-resolved filtration efficiencies

of the 2 PAPs.

102

CHAPTER 3

Effects of Monthly and Long-term Temperature Change on Indoor Exposure to Outdoor

PM2.5 in the Greater Boston Area

(Working paper)

103

Abstract

In this study we assembled data from two cohorts in the greater Boston area, assessing the

monthly and long-term effect of temperature and other meteorology on Sr, a surrogate of I/O for

PM2.5 in two populations: the whole population with mixed AC usage and the subpopulation of

naturally ventilated homes. We found that Sr was independent of meteorology studied, with the

exception of temperature. Monthly effect of temperature was much more dominant when

compared to long-term effect on Sr, which differed in the two populations. In the future, the

seasonal difference (between summer and winter) in Sr was estimated to be as high as 54% for

naturally ventilated home and 30% for the whole population, using winter as the baseline.

Additionally, future Sr in naturally ventilated homes would be approximately 20% higher

compared to the whole population in summer, whereas the difference was small in winter. We

also observed monthly difference in long-term temperature effect for the naturally ventilated

homes, corresponding to an average of 2.1-2.9 襖 increase in monthly temperature. However,

difference was small with maximum of 7% for Sr in July, using values from the past as the

baseline for comparison. In the future, when given the data on future outdoor PM concentrations,

Sr can be used to independently estimate the outdoor fraction of indoor PM, regardless various

indoor sources. It can subsequently be applied to further assess the modification of Sr on the

relationship between future indoor PM exposure and public health in the greater Boston area.

Key words: Indoor-outdoor sulfur ratio; Temperature; Climate change; Particle Infiltration

104

Introduction

There is a large body of evidence implicating short- and long-term exposures to PM2.5 as a

leading contributor to the global burden of disease (1). Individual exposure to PM2.5 can vary

considerably, and is subject to modification from physical, behavioral, and socio-demographic

factors. Notably, since most individuals spend the majority of their time indoors (2), total

exposure to PM2.5 occurs for many people while indoors (3).

Home ventilation has been identified as a central driver of indoor PM2.5 levels (4, 5).

Specifically, home ventilation drives the composition of indoor PM2.5 levels through its

competing effects on increasing ambient particle infiltration and reducing the source strengths of

indoor particle contributions (6). Ventilation is commonly expressed as the air exchange rate, or

the number of times an indoor air volume is replaced by outdoor air over time (e.g., per hour).

Although limited, previous studies have suggested that air exchange rate may modify air

pollution-related short- and long-term health risk (7-12). In an important initial investigation,

Janssen et al. (11) found that city-specific air conditioning (AC) prevalence estimates in 14 US

locations were inversely associated with city-specific effect estimates of outdoor PM2.5 on

hospital admissions for heart and lung disease. That study assumed that homes with central AC

had lower air exchange rate as compared to homes that opened windows for ventilation.

Recently, Sarnat et al. (13) reported significant, positive interactions between air exchange rate

and several air pollutants, including PM2.5, on asthma emergency department visits in Atlanta.

The observed modification of outdoor PM2.5 health effects by air exchange rate may be due, in

part, to its impact on driving greater exposures to particles from outdoor sources.

105

Ambient temperature has been reported to be one of the driving forces for occupant behavior

in window opening and AC operation (14, 15), which greatly influence air exchange rate. Air

exchange rate is also associated with building envelope tightness and wind speed (16-18). Given

the influence of meteorology on air exchanger rate from physical mechanism through building

envelope or/and residential behavior, projected changes in ambient temperature associated with

monthly (or seasonal) and long-term global climate trends may affect population exposures to

PM2.5, which in turn could affect their related disease burden. No studies to our knowledge have

examined this to date.

The objective of this study was, therefore, to quantify the change of indoor exposure to

outdoor PM as it relates to ambient temperature change resulting from monthly variation and

climate change for the population in the greater Boston area. We hypothesized that variation in

ambient temperatures associated with monthly change, highlighted by summer-winter difference,

and long-term climate change would impact home air exchange rates, leading to either decreased

air exchange rates during the increasingly warmer months with AC usage, or increased air

exchange rates due to open windows, or a combination of both. These changes would, in turn,

alter the contributions of outdoor particle sources to indoor air quality, and subsequently lead to

differential effects of PM2.5 exposures by month or season, and in the future relative to the past.

To test the hypothesis, we assembled a large database to establish relationships between the

indoor-outdoor sulfur ratio, Sr, and ambient temperature for the whole study population, as well

as a subpopulation without AC usage. Sr has been widely accepted as a means of approximating

outdoor PM2.5 infiltration fraction by many studies, which are summarized in a recently

published review article from Diapouli et al. (19, 20). Sulfur is a unique tracer for studying the

origin of indoor PM2.5 particles for the following reasons: 1) in general, it does not have indoor

106

sources; 2) it is a regional pollutant, and thus it exhibits very little spatial variability. This makes

possible the use of outdoor measurements at a centrally located supersite to apply for the entire

region; 3) it is a major constituent of PM2.5 and its infiltration and deposition rates are similar to

those of PM2.5, and; 4) it is a very stable pollutant that can be measured readily and accurately

(21). We then used projected past and future temperature to predict their corresponding Sr over

twenty years, respectively. Results from this study were expected to contribute to an initial

understanding of the role of ambient temperature and the impact of monthly variation in

meteorology and climate change on residential exposure to outdoor PM2.5.


Mass balance model (MBM). We used the same mass balance equation as that in Chapter 1 and

2 for indoor PM concentration. In the presence of indoor sources, the indoor PM2.5

concentrations at steady state can be determined as

系沈津鳥噺繋墜髪繋沈津鳥噺底牒寵任底袋賃髪内曇楠底袋賃 (3.1)

where, Cind and Co are the concentrations of particles indoors and outdoors, respectively (µg/m3);

Fo and Find are the concentrations of indoor particles of indoor and outdoor origin, respectively

(µg/m3); Į is the home air exchange rate (hr-1); P is the particle penetration coefficient

(dimensionless); k is the deposition rate of particles indoors (hr-1); IE is the emission rate of

indoor particle sources (µg/hr); and 撃 is the house volume (m3).

Sulfur is the major constituent of PM2.5, as reported from studies conducted in the greater

Boston area (13). We could therefore replace PM2.5 in eq 3.1 by sulfur concentration. Given the

107

lack of indoor sulfur sources, eq 3.1 can be rearranged to express the indoor-outdoor sulfur ratio

(Sr), also known as the sulfur based infiltration factor:

鯨追噺聴日韮匂聴任噺底牒底袋賃 (3.2)

where, Sind and So are the indoor and outdoor concentrations of sulfur, respectively (µg/m3). In

this study, Sr was equal to the PM2.5 infiltration factor which was the fraction of outdoor particles

present indoors, and could be linked to indoor exposure to outdoor PM2.5 on the population level.

Study population. We assembled a large database of archived (retrospective) indoor and

outdoor PM2.5 mass and sulfur concentrations collected at homes from two cohorts between 2006

and 2010 in Boston, MA烉the Diabetes, Cardiac Disease, and Pollution Vulnerability (DCDPV)

Study (20, 22) and the Normative Aging Study (NAS) (20, 23). DCDPV was conducted to assess

the relation of traffic related and transported air pollution to vascular/endothelial, inflammatory

and autonomic outcomes, and evaluated differences in effects based on particle composition. 70

homes were selected for a repeated measurements study. Exclusion criteria included factors that

might complicate estimation of pollution exposures (e.g., ambient air pollution measurements,

such as second hand smoking at home, living more than 25 km away from the central Supersite),

as well as certain highly compromising conditions or diseases (22). Integrated air pollutant

samples were collected inside 70 homes for 6 days. For most of these homes 4-5 samples were

collected at least once during each season. A total of 341 samples were collected between 2006 -

2010 and were used in our study.

The NAS was linked to the original NAS which is a longitudinal study of aging in Eastern

Massachusetts established in 1963 by the Veterans Administration (VA) with 2,280 community-

108

dwelling, healthy men enrolled between 1963 and 1968 (20, 23). During 2006-2010, 321

samples were collected from 270 homes for approximately 7 days, where most homes were

sampled once, except for 50 homes sampled twice.

Altogether, we assembled 662 samples from 340 homes collected across all months in the

greater Boston area in this study. Selected variables included home location (longitude and

latitude), air pollutant sampling duration, weather parameters recorded from the central sites

(Harvard Supersite and weather station at Boston Logan Airport), average indoor air pollutant

concentrations over sampling durations, air conditioning (AC) usage (AC=1 for yes and AC=0

for no), and window open status (yes or no). This database was used to establish relationships

between ambient temperature and indoor particle exposure to ambient particles (Sr) using the

exposure models.

Air pollution measurements. Indoor PM2.5 samples were collected using a custom-made

Harvard sampling system in both cohort studies (20, 22, 23). The sampler was sent to the

subject’s home by express shipping in a specially constructed container the week before their

health examination appointment. The subject was instructed to place the sampler in the main

activity room of the house (other than a kitchen), typically the family room or living room. The

sampler started automatically when plugged in, and after one week, the sampler was unplugged.

The subject then brought the sampler to their health exam, or called for an express shipping pick-

up. The sampler collected PM2.5 particles on a Teflon filter which was then analyzed for PM2.5

mass concentration and other particle components, including sulfur concentration.

109

As mentioned preciously, sulfur is a regional pollutant and is relatively stable. We used the

ambient sulfur measurements at the central Supersite at the Harvard University Countway

Library to represent the outdoor sulfur concentrations for each home in the greater Boston area,

corresponding to their sampling durations of indoor air pollutants. Measurements from the

Supersite were not affected by local point sources.

Meteorology data. The original data sets for the two cohorts included meteorological records

from central sites (Harvard Supersite or weather stations at Boston Logan Airport). In the

preliminary analysis, we compared the meteorological measurements from Boston Logan Airport

in the Global Summary of Days (GSOD) database to the Harvard Supersite values and found that

wind speed varied substantially due to geographic differences between the oceanside and more

inland locations of the monitoring sites. Temperature, on the other hand, was relatively

consistent.

To minimize the potential bias in meteorology records due to the varying distance of the

home locations to the central sites, we used meteorology data from the North American Regional

Reanalysis (NARR) database. NARR is conducted by National Centers for Environmental

Prediction (NCEP) and provides reanalysis data from 1979 to near present. It produces historical

high resolution (32 x32 km per grid) data for North America, which is assimilated from

observational sources including surface, rawinsonde, satellite and aircraft (24). In this study, we

matched the centroids of the grids in NARR to the home locations. Meteorology from the grid

with the shortest distance to a home was used to provide representative weather data for that

home for the same sampling period. The selected meteorological parameters included average

110

daily temperature (measured at 2 m above the surface), wind speed (measured at 10 m above the

surface), precipitation and relative humidity (measured at 2 m above the surface). Maximum and

minimum daily temperature was determined from the 3-h averages in the same database.

Negative values for precipitation were replaced by 0.

Climate forecast model. We forecasted temperature, wind speed, relative humidity, and

precipitation in Boston for two 20-year periods: 1981-2000 (the past) and 2046-2065 (the future).

Forecasts for both the past and the future were made using data archived for the Coupled Model

Inter-comparison Project Phase 5 (CMIP5), an initiative of the Intergovernmental Panel on

Climate Change Fifth Assessment Report (25). This database contains projected meteorology

generated by a suite of climate models for a range of socioeconomic scenarios at a 100x100 km

resolution. We selected 15 CMIP5 models because they provided daily averages for the selected

weather parameters. Predictions were performed for a set of scenarios which met specified

targets for anthropogenic radiative forcing, a measure of climate change (26). These scenarios,

known as the Representative Concentration Pathways (RCPs), aim for radiative forcings in the

year 2100 of 8.5 Wm-2 (RCP8.5), 6.0 Wm-2 (RCP6.0), and 4.5 Wm-2 (RCP4.5). The fourth

scenario peaks at 3 Wm-2 before declining by 2100 (RCP2.6). When applied to models, these

scenarios yielded a broad range of climate trajectories for the 21st century. The models

themselves also contain uncertainty due to the challenge in representing climate feedbacks, such

as changes in cloud cover or sea ice. Such feedbacks can either amplify or diminish the climate

response to increasing greenhouse gases. The range in the modeled climate response is especially

large for variables describing the frequency or intensity of extreme events (e.g., heat waves) that

are important to human health. However, there is no standard way to adjust for the potential bias

111

and uncertainty in model predictions. Despite this common limitation, projections for the future

would be made with better confidence if the models could accurately describe the meteorology in

the past. In this study, we therefore compared the projected meteorology to the NARR data

(averages across 9 grids) using the period of 1981-2000 to check for representativeness of the

model projections for historic data. Differences between projections for the future and the past

were either presented within individual CMIP5 model or based on CMIP5 multi-model means,

with the underlying assumption that projections had minimized uncertainties and biases within

the same CMIP5 model or as the overall multi-model means.

The projected daily meteorology was processed into weekly averages before being used to

estimate the future and past Sr in the exposure model where variables were based on weekly

average values.

Statistical analysis. An important hypothesis for the study was that Sr was a function of ambient

temperature. Sr was calculated from the indoor sulfur concentration and that from the central site

for all homes. Descriptive statistics included the distribution of the meteorological variables, use

of AC, indoor and outdoor sulfur concentrations, and Sr. The subsequent statistical analysis

consisted of two stages: (1) to construct the exposure models using data collected during 2006-

2010, and (2) to estimate the past and future Sr using the exposure models in conjunction with

projected meteorology data.

In the first stage, linear mixed effects models were constructed to evaluate the associations

between the main effect of temperature and Sr. A random home-specific intercept was fit in the

models to account for the residual correlation from the repeated measurements within the same

112

home and the heterogeneity of the overall Sr between homes. Weekly averages of the wind

speed, relative humidity and precipitation that corresponded to the home sampling duration were

evaluated to see if they had significant effects on Sr and the main effect of temperature.

However, the wide distribution of the day-to-day variability of the averaged values, especially

for precipitation, could potentially bias the observed associations. For example, one would be

less confident in the relationship between the predictors and Sr when the precipitation was

averaged from the sampling week with heavy rainfall clustering for a couple days, compared to

the same average precipitation with similar amount of daily rainfall for the week. As a result,

weighting was applied to the models based on proportionality of one over the variances of the

mean values for the meteorological variables, respectively. Finally, to assess whether the main

effect(s) differed by the two cohorts, a variable coded with the cohort names was added to the

models to see its interaction with the independent variables.

Since distribution of AC usage was expected to impact the effect of temperature on Sr on the

population basis, the aforementioned analyses were performed for two population scenarios,

respectively: all homes (the original population) with mixed AC usage and the subpopulation of

naturally ventilated homes, a more generalized scenario based on no AC usage. The final models

were considered as the exposure models and used subsequently to estimate the future and past Sr.

In the second stage, the weekly averages of projected meteorological values from the 15

CMIP5 models for 1981-2000 and 2046-2065 were used in the exposure models to predict

weekly average Sr for the past and the future 20 years. The summary statistics, such as the

predicted monthly or yearly averages of variables, were calculated from the weekly averages

from each CMIP5 model. Variability in predictions across the CMIP5 models (CMIP5 model

113

variability) was characterized using ± standard deviation (SD) from the overall mean predictions

of the CMIP5 models.

We used R (version 2.15.1; R Foundation for Statistical Computing, Vienna, Austria) and

statistical package SAS (SAS Institute Inc., Cary, NC) for the analyses. Effect estimates with p-

value ≤ 0.05 were considered significant.

Results

Summary of parameters. Table 3.1 is a summary of the sampling parameters, meteorology, and

the day-to-day variability of the meteorology values that was expressed as the variance of the

values within the sampling duration for each observation period. After removing Sr >1 that

violated MBM assumption, the data set contained 614 measurements from 321 homes, where the

average sampling duration was 6.18 ± 0.84 (mean ± SD) days. 71 (N=138) homes used AC

during the sampling week while 278 homes (N=471) did not. Information on AC usage was

unavailable for five homes. Among the meteorology parameters, precipitation had the largest

day-to-day variability within the sampling week. One observation was not used in the calculation

of summary statistics for the day-to-day variability because the home was only sampled for less

than a day. Values of one over variance were used as weighting in the exposure models based on

proportionality.

Mean Sr measured from the whole population and the subpopulation is presented by month

in Figure 3.1-(a) and (b), respectively. Mean Sr was generally higher for warmer months. The

difference was more dominant in the subpopulation, where the interquartile range was narrower

114

in June, July, and August, compared to that in the whole population. One possible explanation

was the more prevalent use of AC in summer time as shown in Figure 3.2.

Exposure models. Temperature was found to be a significant predictor for Sr in both population

scenarios, whereas the other meteorological parameters were not. The fitted exposure model for

the mean Sr for all homes (whole population) was

鯨追噺ど┻ねぱ髪ど┻どどのの劇

where, T is the weekly averaged temperature (襖). There was a positive linear relationship

between temperature and Sr, where one Celsius degree increase in temperature led to an increase

of 0.0055 in Sr.

After excluding homes that used AC, the relationship became quadratic for the sub-

population, the naturally ventilated homes. The fitted exposure model was as follows:

鯨追噺ど┻ねば髪ど┻どどにの劇髪ど┻どどどぬの劇態

It is implied in the model that the increase in Sr for every one degree increase in temperature

would be more rapid when the temperature was high.

115

Table 3.1. Summary of the sampling parameters, meteorology, and the day-to-day variability of the meteorology parameters within the sampling duration.

N Mean SD Median Min Max

Duration (d) 614 6.18 0.84 5.83 0.53 9.78

Indoor sulfur (g/m3) 614 0.47 0.31 0.37 0 1.92

Outdoor sulfur (g/m3) 614 0.84 0.45 0.73 0.22 2.69

Indoor-outdoor sulfur ratio (Sr)

All 614 0.55 0.19 0.55 0 1.00

AC=0 471 0.56 0.19 0.55 0 1.00

AC=1 138 0.52 0.21 0.54 0.00016 0.93

Meteorology

Temperature (襖岻 614 12.45 8.59 13.65 -6.25 26.65

Relative humidity (%) 614 78.24 6.82 78.87 53.48 95.50

Wind speed (m/s) 614 4.00 1.21 3.75 1.63 9.16

Precipitation (mm/day) 614 2.92 3.31 1.95 0 24.11

Day-to-day variability (variance)*

Temperature 613 8.81 8.91 6.04 0.085 67.93

Relative humidity 613 79.34 73.53 56.81 1.28 452.70

Wind speed 613 3.40 3.16 2.55 0.30 38.28

Precipitation 613 51.53 108.56 12.16 0 1009.00

*Day-to-day variability was the variance of the parameter daily values within the sampling duration for each observation period.

116

Figure 3.1. Boxplots for Sr measurements by month for (a) the whole population with mixed AC

usage, and (b) the subpopulation of naturally ventilated homes (AC=0). The solid points

represent the Sr observations; whereas the filled diamonds in red represent the monthly mean of

Sr.

117

(a) (b)

Figure 3.2. Sr measurement for (a) the subpopulation of naturally ventilated homes (AC=0), and

(b) homes that used AC during the sampling period (AC=1). Measurements from the two cohorts

are marked in different colors.

Comparisons between CMIP5 projections and NARR data. Figure 3.3 shows the

comparisons between projected monthly mean meteorology from CMIP5 models and NARR

database for the period of 1981-2000. There was excellent agreement in temperature between the

multi-model means from the CMIP5 model projections and the NARR database, but relatively

poor agreement for the other meteorological parameters. Given temperature was the only

significant predictor of Sr in the exposure models, projected results of Sr for both the past and the

future were considered to be reliable in reflecting the overall trend on the population basis.

118

Figure 3.3. Comparisons between projected monthly mean temperature, RH, wind speed and

precipitation from CMIP5 models and NARR database for the period of 1981-2000.

Effect of temperature on Sr. To understand the temporal effects of temperature for the past and

the future together with the variability of CMIP5 model projections, we first look at the projected

monthly temperature and their corresponding Sr for both periods by each of the 15 models.

Seasonal effect was highlighted based on the comparison between summer and winter, whereas

long-term effect was primarily referred to the difference between the future and the past

quantities by paired years (N=1 to 20), each of which was 65 years apart. Figure 3.4 shows the

mean temperature by month in each projected paired year. Overall, long-term effect of

temperature appeared to be more obvious in warmer months compared to cold ones. For

119

example, temperature in July was generally warmer in the future, but the increment was harder to

see for January. Conversely, the CMIP5 model variability was larger for winter. As expected, the

trend was exactly the same for the predicted Sr based on the whole population, because Sr had a

linear relationship with temperature (Figure 3.5).

On the other hand, the effect of temperature differed for the subpopulation of naturally

ventilated homes, with quadratic relationship between temperature and Sr (Figure 3.6). A more

rapid increase in Sr was observed in warmer months where CMIP5 model variability was also

higher, compared to winter time. One possible explanation is that the CMIP5 model variability in

the monthly mean temperature was magnified by the quadratic term in the exposure model by

increasingly high temperature.

Finally, the trend of temperature and Sr projections over the 20 year period was relative static

within the same CMIP5 model. We could therefore summarize the estimates into monthly

average across the whole 20 year period, which made it easier to examine the seasonal and long-

term effects of temperature on Sr with CMIP5 model variability.

Seasonal effects. Figure 3.7 shows the 20-year averages of monthly mean temperature for the

future and the past from all 15 CMIP5 models and their overall monthly averages. In general, the

CMIP5 models had good predictability for temperature, where the model-specific trends were

consistent. Variability of the monthly mean across models was more obvious for extreme

temperatures, such as summer and winter. For the future, variability increased in summer. Given

the temperature profiles, the corresponding Sr is displayed in Figure 3.8 for the two populations.

120

Figure 3.4. Mean temperature by month for the past and the future based on paired years (N=1 to

20). The solid and dashed lines are projections for the future and the past, respectively. The light-

colored lines represent projections from the CMIP5 models, whereas the dark-colored lines

describe the multi-model means across the CMIP5 models.

Seasonal effect was observed for both periods and populations. It was a lot higher for

naturally ventilated homes compared to the whole population with mixed AC usage.

Additionally, the CMIP5 model variability did not mask this trend. In the future, the seasonal

difference (between summer and winter) in the overall mean Sr was estimated to be as high as

54% (in summer) for naturally ventilated home and 30% for the whole population, using winter

as the baseline.

121

Figure 3.5. Mean estimated Sr by month for the past and the future based on paired years (N=1 to

20) for the whole population with mixed AC usage (AC=mixed). The solid and dashed lines are

projections for the future and the past, respectively. The light-colored lines represent projections

from the CMIP5 models, whereas the dark-colored lines describe the multi-model means across

the CMIP5 models.

The difference in predicted monthly Sr among the two populations in the past and the future

can be more clearly seen in Figure 3.9. Generally, the difference was smaller for the past than the

future because Sr in naturally ventilated homes was more sensitive to increasingly high

temperature in the future, especially in summer. The overall monthly difference ranged from -

0.0062 to 0.13, with the maximum value in the future summer. It suggested that the future Sr in

naturally ventilated homes would be approximately 21% higher when compared to Sr from the

122

whole population with mixed AC usage in July. A similar profile was observed for the past 20

years with the difference ranging from -0.0085 to 0.094, corresponding to a maximum of

approximately 16 % difference based on mean Sr for the whole population in the same month.

Difference in wintertime Sr, on the other hand, was not as suggestive between the two periods or

populations when accounted for the CMIP5 model variability. Overall, predicted monthly Sr was

higher in naturally ventilated homes in summer, whereas it was not much different in winter

between the populations.

Figure 3.6. Mean estimated Sr by month for the past and the future based on paired years (N=1 to

20) for the subpopulation of naturally ventilated homes (AC=0). The solid and dashed lines are

projections for the future and the past, respectively. The light-colored lines represent projections

from the CMIP5 models, whereas the dark-colored lines describe the multi-model means across

the CMIP5 models.

123

Figure 3.7. Projected monthly mean temperature for the past (1981-2000) and the future (2046-

2065) by 15 CMIP5 models (dashed lines). The solid line is the overall monthly mean across the

CMIP5 models.

Figure 3.8. Estimated monthly mean Sr for the past and the future by the two populations. The

solid lines are the overall monthly mean across the CMIP5 models while the dashed lines are 罰 1

SD from the overall mean. AC=0 represents the subpopulation of naturally ventilated homes and

AC=mixed is referred to the whole population.

124

Figure 3.9. Difference in estimated monthly mean Sr between the two populations for the past

and the future. The solid lines are the overall monthly mean difference across the CMIP5 models

while the dashed lines are 罰 1 SD from the overall mean. AC=0 represents the subpopulation of

naturally ventilated homes and AC=mixed is referred to the whole population.

Long-term effect. To evaluate the long-term effect of temperature on Sr, differences in the future

and the past quantities for temperature and predicted Sr were used as the metrics. The future

temperature would be higher than the past, ranging from 2.1-2.9 襖 based on the overall monthly

average (Figure 3.10-(a)) across all CMIP5 models. Such increases corresponded to overall

monthly mean Sr increment of 0.014-0.016 for all homes and 0.0010-0.047 for naturally

ventilated homes, respectively (Figure 3.10-(b)). The increment did not differ by season for the

whole population, where increment in winter was 1.17 times that in summer. On the contrary,

increment was much more obvious in summer and 46 times that in winter for naturally ventilated

homes. Between the two populations, increment in Sr for naturally ventilated homes was much

higher, approximately 3.4 times the amount of that when considering all homes in July.

125

Nevertheless, the percentage of the increment was small when compared to the predicted

monthly mean of Sr. For example, it was approximately 2% for the whole population and 7% for

the subpopulation in the month of July.

(a)

(b)

Figure 3.10. Monthly mean differences in estimates between the future and the past for (a)

temperature, and (b) Sr. The solid lines were the overall monthly mean difference across the

CMIP5 models while the dashed lines were 罰 1 SD from the overall mean. AC=0 represented

the subpopulation of naturally ventilated homes and AC=mixed was referred to the whole

population.

126

Discussion

In this study we examined the impact of meteorology on Sr, a surrogate of the outdoor

fraction of indoor PM concentration, with the emphasis on temperature.

Meteorology is known to have strong influences on outdoor PM concentrations. For example,

temperature and absolute humidity possess species-specific effect on PM concentration, mostly

through the competing or additive effect between sulfate, nitrates, and organic aerosols. Higher

temperature increases evaporation of nitrates and organic aerosols, leading to decreased

concentrations. Conversely, it increases sulfate formation through temperature dependent

oxidation process and with more abundant oxidants participating in the reaction (27, 28),

whereas absolute humidity is associated with increase in ammonium nitrate aerosols in summer

due to higher water vapor concentration (27). Wind and precipitation, on the other hand, have

non-species-specific effect on PM concentration. PM are diluted due to higher wind speed or

washed out by higher precipitation (29). However, the resulting temporal or spatial variability

could be high.

To acquire more representative meteorology for the homes in our study cohorts, we matched

the homes to the high resolution data from NARR based on location and sampling period. This

approach aimed to minimize the uncertainty resulting from spatial variability, as opposed to

using data from the central site, on the observed association between meteorology and Sr. We

also accounted for temporal variability by imposing weighting on the association based on the

proportionality of one over variance within the averaged values during each sampling period for

the meteorological parameters. Precipitation was found to have the highest day-to-day

variability. In the final exposure models, temperature was the only significant predictor for Sr,

127

suggesting that the influence of meteorology on Sr was probably not through species-specific PM

properties or their ambient concentrations.

In a study to explore the impact of climate change on indoor air pollution, Nazaroff (30)

classified the influential factors into three categories: properties of pollutants, building factors

(e.g., ventilation), and occupant behavior (e.g., AC usage). Use of AC can decrease the Sr due to

closed windows or/and the filters inside the AC system (31-33). Given the findings of

differential effect of meteorology on Sr in this study, the association between temperature and Sr

was unlikely due to the first category; instead, it could be a result from the last two. We

evaluated the main effect of temperature for the original population with mixed AC usage and

the subpopulation of naturally ventilated homes in the exposure models. It is noteworthy that

adding AC usage as a binary variable (yes or no) in the exposure models was not feasible due to

unbalanced sample size in the binary categories. Additionally, AC was almost exclusively used

in summer, making it impossible to evaluate the effect of temperature that covered such a narrow

temperature range. Separating the exposure models into two populations enabled comparisons of

main effects under different distribution of AC usage. When only considering homes without AC

usage, temperature had a quadratic relationship with Sr, where more rapid increase in Sr was

observed for summer than winter. This could be explained by open windows or doors to ventilate

the house. However, after including the homes that used AC, the differential relationship by

temperature disappeared, particularly for high temperature range. This was because the high Sr

expected for homes with open window in summer were cancelled out by the competing effect

from homes using AC that decreased Sr. This finding not only addresses the impact of AC usage,

but also brings out the importance of population selection when estimating Sr on the population

basis. In our case, the difference in the future Sr can be up to 21 % in July between the

128

populations studied. Therefore, small effects of temperature on Sr in the whole population do not

rule out the health risks for the subpopulation.

The analysis conducted using the selected homes was considered to produce robust

relationships between temperature and Sr. Although the original indoor exposure studies for the

cohorts were not designed to select homes representative of the general population, collectively,

these studies encompass a wide range of homes with different characteristics, collected

throughout the year under varying weather conditions.

We assumed that individuals would react the same way to the same temperature condition

within 65 years, part of which was projected into the future. We also assumed that homes in the

future would be the same as they were 65 years ago. It is possible, however, that homes would be

different in the next 50 years due to advancement in building technology, such as better

insulation, heating, and cooling. But the rate of technology penetration is unknown and depends

on the cost and affordability. On the other hand, the age of homes can vary from a few decades

to 100 or more years, suggesting that it takes a long time to replace homes. Among the 1,151,000

homes surveyed in the greater Boston area in 2007, the median year of structure built was 1951,

where 90% were built before 1985 (34). Since the average monthly temperature was estimated to

change slowly over decades by only 2-3oC, it is reasonable to expect that people would adjust to

climate changes by opening windows more or using more air conditioning in warmer days to

ventilate the house. These changes have the potential to substantially impact home ventilation,

and as a result, the contribution of outdoor sources to total indoor PM2.5, as suggested by the

findings in this study.

129

One limitation of the study was the low temporal resolution of the data, namely weekly

averages of indoor measurements and lack of detailed records on indoor activities. Two potential

issues arise from this limitation. Unavoidably, the multi-day sampling duration included the

intermittent periods of indoor sulfur emission, use of air purifying devices, vacuuming, and other

indoor activities that could potentially violate the MBM assumption and influence the indoor PM

concentration. The interference as anticipated to be larger in the exposure model for the whole

population, where the frequency of AC usage (e.g., number of days in a week) was unknown.

Nevertheless, given some of these activities were routines and limited indoor sulfur sources, the

within home correlation and variability across homes for Sr were expected to be captured in part

by the mixed effects models. Consequently, the exposure models would still give an overall

picture of temperature effect on Sr on a multi-day basis.

The second issue was the insensitivity of the effect estimate to more extreme scenarios, for

example, days of heatwaves. In other words, the day-to-day variability was smoothed by using

integrated samples, leading to smaller effect estimates of temperature in the exposure models.

Nevertheless, we still observed seasonal difference in long-term effect (Figure 3.10-(b)), as well

as long-term difference in seasonal effect of temperature on Sr (Figure 3.9). Although the

increments of overall monthly mean Sr in the future, when based on percentage, were not very

large, they could deviate from but still center around the observed monthly mean Sr values when

daily data are available. Consequently, it remains inconclusive whether climate change would

impact Sr significantly and thus modify the effect of PM on health, especially in vulnerable

populations.

Another limitation of the study was the lack of information of to what extent did the home

environment meet the steady state assumption of the MBM over the sampling period. The multi-

130

day sampling duration could introduce certain variability in the steady state situation, as

described above. In Chapter 2, we know that under relatively well-mixed environment, the MBM

could be applied to estimate the indoor PM concentrations as a function of outdoor PM

concentration, air exchange rate, particle penetration coefficient, particle deposition rate, and

filtration parameters in a given space. Validation of the assumption is not feasible in large-scaled

home studies; nevertheless, the steady state approach has been used in many studies to determine

those parameters (6, 21, 35-37). Based on the assumption, we were able to estimate the

corresponding parameters in the MBM with measured Sr, which is of great contribution to the

modeling application in estimating PM exposure from a large number of homes.

A unique strength of the study is that we were the first to explore the relationship between

monthly and long-term temperature change and Sr on the population basis, with capitalizing on

an assembly of a large number of home measurements. Our results projecting changes in the

relative temporal contribution of temperature to indoor exposure to outdoor PM are valuable for

future assessments and providing implications in possible intervention. We found that Sr was

independent of meteorology studied, with the exception of temperature. Monthly effect of

temperature, highlighted by summer-winter difference, was much more dominant when

compared to long-term effect on Sr, which differed in the two populations. PM2.5 exposure was

high in summer, especially in naturally ventilated homes. AC, air purifying devices and/or closed

window status could be potentially effective interventions for reducing indoor PM exposure,

based on the comparison of Sr between the two populations in this study, and the findings from

the previous chapter.

Finally, as discussed in the previous paragraphs, findings from the study can be utilized to

evaluate the relevant parameters influencing indoor PM concentrations in the study homes, such

131

as particle penetration coefficient and deposition rate, which in turn helps determine the indoor

source contribution. More importantly, when given the data of future outdoor PM concentrations,

Sr can be used to independently estimate the outdoor fraction of indoor PM, regardless various

indoor sources. It can subsequently be applied to further assess the modification of Sr on the

relationship between future indoor PM exposure and public health in the greater Boston area.

In the future, the study method can be applied to assess the effect of temperature on Sr in

other cities of distinct weather conditions, building characteristics and occupant behaviors, where

Boston can be used as reference city for comparison.

132

Bibliography

(1) Lim, S.S.; Vos, T.; Flaxman, A.D.; Danaei, G.; Shibuya, K.; Adair-Rohani, H. A comparative

risk assessment of burden of disease and injury attributable to 67 risk factors and risk factor

clusters in 21 regions, 1990–2010: a systematic analysis for the Global Burden of Disease Study

2010. The lancet 2013, 380 (9859), 2224-2260.




10.1038/sj.jea.7500165.

(3) Meng, Q.Y.; Turpin, B.J.; Korn, L.; Weisel, C.P.; Morandi, M.; Colome, S. Influence of

ambient (outdoor) sources on residential indoor and personal PM2. 5 concentrations: analyses of

RIOPA data. Journal of Exposure Science and Environmental Epidemiology 2005, 15 (1), 17-28.

(4) Liu, D. and Nazaroff, W.W. Particle Penetration Through Building Cracks. Aerosol Science

and Technology 2003, 37, 565-573.

(5) Thatcher, T.L. and Layton, D.W. Deposition, resuspension, and penetration of particles

within a residence. Atmospheric Environment 1995, 29 (13), 1487-1497.



2001, 35 (10), 2089-2099; 10.1021/es001477d.

133

(7) Chen, C.; Zhao, B.; Weschler, C.J. Indoor exposure to “outdoor PM10”: assessing its

influence on the relationship between PM10 and short-term mortality in US cities. Epidemiology

2012, 23 (6), 870-878.

(8) Hodas, N.; Meng, Q.; Lunden, M.M.; Rich, D.Q.; Özkaynak, H.; Baxter, L.K. Variability in

the fraction of ambient fine particulate matter found indoors and observed heterogeneity in

health effect estimates. Journal of Exposure Science and Environmental Epidemiology 2012, 22

(5), 448-454.

(9) Bell, M.L. and Dominici, F. Effect modification by community characteristics on the short-

term effects of ozone exposure and mortality in 98 US communities. American journal of

epidemiology 2008, 167 (8), 986-997.

(10) Levy, J.I.; Chemerynski, S.M.; Sarnat, J.A. zone exposure and mortality: an empiric bayes

metaregression analysis. Epidemiology 2005, 16 (4), 458-468.

(11) Janssen, N.A.; Schwartz, J.; Zanobetti, A.; Suh, H.H. Air conditioning and source-specific

particles as modifiers of the effect of PM (10) on hospital admissions for heart and lung disease.

Environmental health perspectives 2002, 110 (1), 43.

(12) Medina-Ramón, M.; Zanobetti, A.; Schwartz, J. The effect of ozone and PM10 on hospital

admissions for pneumonia and chronic obstructive pulmonary disease: a national multicity study.

American journal of epidemiology 2006, 163 (6), 579-588.

134

(13) Sarnat, J.A.; Sarnat, S.E.; Flanders, W.D.; Chang, H.H.; Mulholland, J.; Baxter, L.

Spatiotemporally resolved air exchange rate as a modifier of acute air pollution-related morbidity

in Atlanta. Journal of Exposure Science and Environmental Epidemiology 2013, 23 (6), 606-615.

(14) Wallace, L.A.; Emmerich, S.J.; Howard-Reed, C. Continuous measurements of air change

rates in an occupied house for 1 year: the effect of temperature, wind, fans, and windows.

Journal of exposure analysis and environmental epidemiology 2002, 12 (4), 296-306.

(15) Kearney, J.; Wallace, L.; MacNeill, M.; Héroux, M.E.; Kindzierski, W.; Wheeler, A.

Residential infiltration of fine and ultrafine particles in Edmonton. Atmospheric Environment

2014, 94, 793-805.

(16) Chan, W.R.; Price, P.N.; Nazaroff, W.W.; Gadgil, A.J. Distribution of residential air

leakage: Implications for health outcome of an outdoor toxic release. Indoor air 2005, 15 (11),

1729-1734.

(17) Persily, A.; Musser, A.; Emmerich, S.J. Modeled infiltration rate distributions for US

housing. Indoor air 2010, 20 (6), 473-485.

(18) Sherman, M.H. and Chan, W.R. Building air tightness: research and practice. Building

Ventilation: the state of the Art 2006, , 137-162.

(19) Diapouli, E.; Chaloulakou, A.; Koutrakis, P. Estimating the concentration of indoor particles

of outdoor origin: A review. Journal of the Air & Waste Management Association 2013, 63 (10),

1113-1129.

135

(20) Hoxha, M.; Dioni, L.; McCracken, J.P.; Baccarelli, A.; Melly, S.J.; Coull, B.A. Annual

ambient black carbon associated with shorter telomeres in elderly men: Veterans Affairs

Normative Aging Study. Environmental health perspectives 2010, 118 (11), 1564-1570.

(21) Sarnat, J.A.; Long, C.M.; Koutrakis, P.; Coull, B.A.; Schwartz, J.; Suh, H.H. Using sulfur

as a tracer of outdoor fine particulate matter. Environmental Science & Technology 2002, 36

(24), 5305-5314.

(22) Chuang, K.J.; Coull, B.A.; Zanobetti, A.; Suh, H.; Schwartz, J.; Stone, P.H. Particulate air

pollution as a risk factor for ST-segment depression in patients with coronary artery disease.

Circulation 2008, 118 (13), 1314-1320.

(23) Baja, E.S.; Schwartz, J.D.; Wellenius, G.A.; Coull, B.A.; Zanobetti, A.; Vokonas, P.S.; Suh,

H.H. Traffic-related air pollution and QT interval: modification by diabetes, obesity, and

oxidative stress gene polymorphisms in the normative aging study. Environmental health

perspectives 2010, 118 (6), 840-846.

(24) NCEP North American Regional Reanalysis (NARR) .

(25) Stocker, T.F.; Qin, D.; Plattner, G.K.; Tignor, M.; Allen, S.K.; Boschung, J.

Intergovernmental panel on climate change, working group I contribution to the IPCC fifth

assessment report (AR5). 2013, .

(26) Moss, R.H.; Edmonds, J.A.; Hibbard, K.A.; Manning, M.R.; Rose, S.K.; Van Vuuren, D.P.

The next generation of scenarios for climate change research and assessment. Nature 2010, 463

(7282), 747-756.

136

(27) Seinfeld, J.H. and Pandis, S.N. Atmospheric chemistry and physics: from air pollution to

climate change. John Wiley & Sons 2012, .

(28) Sweet, C.W. and Gatz, D.F. Summary and analysis of available PM 2.5 measurements in

Illinois. Atmospheric Environment 1998, 32 (6), 1129-1133.

(29) Dawson, J.P.; Adams, P.J.; Pandis, S.N. Sensitivity of PM 2.5 to climate in the Eastern US:

a modeling case study. Atmospheric chemistry and physics 2007, 7 (16), 4295-4309.

(30) Nazaroff, W.W. Exploring the consequences of climate change for indoor air quality.

Environmental Research Letters 2013, 8 (1), 015022.

(31) Howard-Reed, C.; Wallace, L.A.; Emmerich, S.J. Effect of ventilation systems and air

filters on decay rates of particles produced by indoor sources in an occupied townhouse. Atmos.


(32) Sarnat, J.A.; Koutrakis, P.; Suh, H.H. Assessing the relationship between personal

particulate and gaseous exposures of senior citizens living in Baltimore, MD. Journal of the Air

& Waste Management Association 2000, 50 (7), 1184-1198.


particles in a residence for 18 months in 1999-2000. . Journal of the Air & Waste Management

Association 2002, 52 (7), 828-844.

(34) U.S. Census Bureau American Housing Survey for the. Boston Metropolitan Area: 2007.

137



10.1021/es990348y.

(36) Wallace, L. and Williams, R. Use of personal-indoor-outdoor sulfur concentrations to

estimate the infiltration factor and outdoor exposure factor for individual homes and persons.

Environmental science & technology 2005, 39 (6), 1707-1714.

(37) Hodas, N.; Meng, Q.; Lunden, M.M.; Turpin, B.J. Toward refined estimates of ambient PM

2.5 exposure: Evaluation of a physical outdoor-to-indoor transport model. Atmospheric

Environment, 2014, 83, 229-236.

138

CONCLUSIONS

139

In the preceding chapters we discussed in each study the challenges that motivated the

objectives, limitations of the study design, and implications from the findings. In this chapter, I

would like to summarize the major findings from the presented work and shed light on both the

innovation and limitations, based on which I suggested future directions.

In Chapter 1, we estimated the size-resolved particle deposition rates for the ultrafine and

submicrometer particles during non-sourced period following a controlled sourced period in a

well-mixed residential environment inside an apartment. The study design in conjunction with

the non-linear mixed effects modeling procedure provided a feasible and alternative method for

estimating particle deposition rates when the background concentration cannot be measured. A

dynamic adjustment method with constant injection of tracer gas was used to maintain the air

exchange rate at three target levels: 0.60, 0.90 and 1.20 h-1, as the sampling conditions.

Particle deposition was found to be highly size dependent with rates ranging from 0.68 ±

0.10 to 5.03 ± 0.20 h-1 (mean ± SE). While acknowledging the large variability in the size-

resolved deposition rates reported from the previous studies and considering the relatively small

95% confidence intervals for the mean estimates of 倦沈 in this study, we found that our estimates

for submicrometer particles were in close agreement with some of these studies. However, the

mean estimates of deposition rate for the ultrafine particles were considerably higher than the

reported deposition rates from the others, which could possibly be explained by the effect of

enhanced air mixing by the operation of portable fans. The effect of air exchange on the particle

deposition under enhanced air mixing was relatively small when compared to both the strong

influence of size-dependent deposition mechanisms and the effects of mechanical air mixing by

fans. Nonetheless, the significant association between air exchange and particle deposition rates

for a few size categories indicated potential influence of air exchange on particle deposition.

140

In Chapter 2, as a companion study, we validated the use of the MBM to determine the

effectiveness of portable air purifiers in removing ultrafine and submicrometer particles in the

same apartment. We evaluated two identical portable air purifiers, equipped with high efficiency

particulate air filters, for their performance under three different air flow settings and three target

air exchange rates. We subsequently used a mixed effects model to estimate the slope between

the measured and modeled effectiveness by particle size. Similar to the findings in Chapter 1,

effectiveness was highly particle size-dependent. For example, at the lowest target air exchange

rate, it ranged from 0.33 to 0.56, 0.51 to 0.75, and 0.60 to 0.81 for the three air purifier flow

settings, respectively. Our findings suggested that filtration was the dominant removal

mechanism for submicrometer particles, whereas deposition could play a more important role in

ultrafine particle removal. We found reasonable agreement between measured and modeled

effectiveness with size-resolved slopes ranging from 1.11 ± 0.06 to 1.25 ± 0.07 (mean ± SE),

except for particles <35 nm.

In Chapter 3, we assembled data from two cohorts in the greater Boston area, assessing the

monthly and long-term effect of temperature and other meteorology on Sr, a surrogate of

infiltration factor for PM2.5 in two populations: the whole population with mixed AC usage and

the subpopulation of naturally ventilated homes. We found that Sr was independent of

meteorology studied, with the exception of temperature. Monthly effect of temperature,

highlighted by summer-winter difference, was much more dominant when compared to long-

term effect on Sr, which differed in the two populations. In the future, the seasonal difference

(between summer and winter) in Sr was estimated to be as high as 54% for naturally ventilated

home and 30% for the whole population, using winter as the baseline. The overall monthly

difference ranged from -0.0062 to 0.13, with the maximum value in the future summer. It

141

suggested that the future Sr in naturally ventilated homes would be approximately 20% higher

when compared to Sr from the whole population with mixed AC usage in summer. Overall,

predicted monthly Sr was higher in naturally ventilated homes in summer, whereas there was no

population difference in winter.

Overall, the future temperature was higher than the past, ranging from 2.1-2.9 襖 based on

the monthly average. Such increases corresponded to monthly Sr elevation of 0.014-0.016 for all

homes and 0.0010-0.047 for naturally ventilated homes, respectively (Figure 3.8). The elevation

did not differ by season for the whole population, where increment in winter was 1.17 times that

in summer. On the contrary, increment was much more obvious in summer and was 46 times that

in winter for naturally ventilated homes. Between the two populations, increment in Sr for

naturally ventilated homes was much higher, approximately 3.4 times the amount of that when

considering all homes in July. Nevertheless, the percentage of the increment was small when

compared to the predicted monthly mean of Sr. For example, it was approximately 2% for the

whole population and 7% for the subpopulation in the month of July. In sum, we observed

seasonal difference in long-term effect (Figure 3.10-(b)), as well as long-term difference in

seasonal effect of temperature on Sr (Figure 3.9).

Study limitation is often a two sided sword, with one aspect of attenuating the interpretability

of the current study findings, while indicating future directions to fill the existing gaps. One

major limitation in the single-home study was the enhanced air mixing to achieve well-mixed

environment. As an initial investigation and implementation of the study approach, it was

reasonable yet unavoidable, to start from the most conservative scenario with least deviation

from the model assumptions. Results can provide us insight for future work and overcome the

existing limitations. For example, we would be interested in exploring how air mixing conditions

142

influence size-resolved particle deposition rate and PAP effectiveness in terms of spatial and

temporal concentration distribution of indoor PM, by varying fan speeds. The same approach can

be used to study the effect of air exchange on particle deposition as well.

The major limitation in the multi-home observational study was lack of high temporal

resolution data, leading to interference from intermittent indoor sources or more variable

combination of activities that violates MBM assumptions, and small effect estimate of

temperature. However, given daily data in the future, we expect to more sensitively detect effects

of meteorology on Sr and to be able to estimate particle deposition rate and penetration

coefficient with known air exchange rate values.

Both the single-home and multi-home observational studies have original components that

are pioneering and innovative in their specific fields of study. One important merit shared by

these research designs and structures, however, is that they can be replicated in various

residential settings, or in observational studies with distinct temporal or spatial features, such as

outside of the greater Boston area.

To our knowledge, the home study in Chapter 1 and 2 was the first to validate the steady

state MBM for predicting the size-resolved effectiveness of PAPs in a residential setting. The

study design and approach for establishing an environment that met the MBM assumptions were

ambitious but successful, making it possible to estimate size-resolved deposition rate, validate

the use of MBM in a home, and assess the effectiveness of PAPs with minimized uncertainties

from violation of assumption. The design and the corresponding approach featured the

achievement of a well-mixed indoor environment using portable fans, the generation of artificial

particles at a constant rate to substantially elevate indoor concentration to minimize the

143

interference from the variability of outdoor PM, and the use of a dynamic method to maintain air

exchange rates at constant levels throughout the sampling period. The same approach can further

be applied to understand other particle behaviors in the future. For example, given simultaneous

ambient PM measurement outside of the house, we can estimate the penetration coefficient via

non-linear mixed effects model. Although this type of controlled home experiments with size-

resolved data acquisition is not feasible for large-scale home studies, increasing amount of data

collection using selected homes with various characteristics can still improve generalizability

and representativeness of the collective measurements. Indoor PM exposure can be subsequently

estimated or predicted based on the known values of model parameters with improved

confidence in data interpretation. Precautionary measures or actions of intervention can then be

suggested to reduce particle exposure in homes.

Similarly, a unique strength of the study in Chapter 3 is that we were the first to explore the

relationship between monthly and long-term temperature change and Sr on the population basis,

with capitalizing on an assembly of a large number of home measurements. Discussion of

meteorology effect from other studies is often limited by incomplete cycle of seasons and the

number of samples. Our database consists a large number of measurements and homes that were

sampled through all seasons, meanwhile covering various building characteristics. Furthermore,

data on occupant behavior such as window opening was relatively balanced across the

temperature range studied. Additionally, we matched study homes to the NARR database by

location and sampling date, minimizing uncertainties due to spatial variability of meteorology

when compared to the conventional approach of using meteorology data from the central site.

We also reported monthly mean Sr with CMIP5 model variability to demonstrate potential

uncertainties in Sr prediction from weather projection. With the progressing development of the

144

CMIP models, reliability in meteorology prediction are expected to increase with uncertainties

better explained, contributing to more refined Sr predictions. In the future, the study method can

be applied to assess the effect of temperature on Sr in other cities of distinct weather conditions,

building characteristics and occupant behaviors, where Boston can be used as the reference city

for comparison.

When viewed together, understandings of the physical mechanisms in the first two chapters,

such as size-resolved particle deposition behavior, PAP effectiveness and reliability of MBM

applications, can be used to explain, in part, the observed association between meteorology and

Sr in the third chapter, especially with available daily observations in the future. As a result,

findings from this dissertation not only intertwine in the causal framework of linking human

exposure to indoor PM and the related health risks, but also contribute to more comprehensive

exposure assessment. By assessing the mass balance application inside out, we could envision

future researches in exploring roles of various mechanisms and interventions in reducing indoor

PM exposure, eventually improving our knowledge on the development of strategies to protect

public health.

From the Inside Out – Application of the Mass Balance ...

Documents

Transcript of From the Inside Out – Application of the Mass Balance ...