Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Infiltration of forest fire and residential wood smoke : an intervention study to asssess [sic] air cleaner… Barn, Prabjit 2006

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata

Download

Media
831-ubc_2007-0021a.pdf [ 13.32MB ]
Metadata
JSON: 831-1.0100824.json
JSON-LD: 831-1.0100824-ld.json
RDF/XML (Pretty): 831-1.0100824-rdf.xml
RDF/JSON: 831-1.0100824-rdf.json
Turtle: 831-1.0100824-turtle.txt
N-Triples: 831-1.0100824-rdf-ntriples.txt
Original Record: 831-1.0100824-source.json
Full Text
831-1.0100824-fulltext.txt
Citation
831-1.0100824.ris

Full Text

INFILTRATION OF FOREST FIRE A N D RESIDENTIAL WOOD S M O K E : A N INTERVENTION STUDY TO ASSSESS AIR C L E A N E R EFFECTIVENESS by Prabjit Barn B.Sc , The University of British Columbia, 2003 A THESIS SUBMITTED FN PARTIAL F U L F I L L M E N T OF THE REQUIREMENTS FOR THE DEGREE OF M A S T E R OF SCIENCE in THE F A C U L T Y OF G R A D U A T E STUDIES (Occupational and Environmental Hygiene) THE UNIVERSITY OF BRITISH C O L U M B I A November 2006 © Prabjit Barn, 2006 ABSTRACT Forest fires and residential wood-burning are significant sources of fine particle (PM2.5) air pollution. As PM2.5 exposure is associated with adverse health effects, populations need to be provided with exposure reduction strategies during smoke episodes that are practical, effective, and evidence-based. Public health recommendations typically include remaining indoors and use of air cleaners, yet little information is available on the effectiveness of these measures. Specific objectives of this study were to measure indoor infiltration of outdoor P M 2 5 from forest fires/residential wood smoke, to determine effectiveness of High Efficiency Particulate Air (HEPA) filter air cleaners in reducing indoor PM2.5 and to investigate determinants of infiltration and air cleaner effectiveness in homes. Winter sampling at 21 homes was conducted in 2004 in a northern Canadian community affected by residential wood-burning. Summer (2004-5) sampling at 17 homes was conducted in southern British Columbian communities impacted by vegetation fire smoke. Indoor and outdoor 1-minute PM2.5 averages and a 48-hour outdoor PM2.5 filter sample were collected at each home. A portable HEPA filter air cleaner was operated indoors with the filter removed for one of two sampling days. Infiltration (Fjnf) was calculated for each home using a recursive model (Switzer & Ott, 1992; Allen et al., 2003). Housing characteristics data were used in multivariable modeling of infiltration and air cleaner efficiency (ACE). Valid samples were obtained from 19 homes in winter and 13 homes in summer. Mean Finf ± SD values of 0.27 ±0.18 and 0.61 ± 0.27 were found for winter and summer respectively, for days when the filter was removed, with lower values of 0.10 ± 0.08 and 0.19 ± 0.20 on corresponding days with filters in place. Mean ± SD A C E , calculated as [F i nf without filter - F i n f with filter]/ Fjnf without filter, in winter and summer were 55 ± 38% and 65 ± 35% respectively. Infiltration was predicted by the number of windows in a home and the season (p<0.001) while no significant predictors of A C E were identified. Remaining indoors combined with air cleaner use was concluded to effectively reduce PM2.5 exposure during forest fire and residential wood burning episodes TABLE OF CONTENTS ABSTRACT " TABLE OF CONTENTS Ill LIST OF TABLES VI LIST OF FIGURES Vlll ACKNOWLEDGEMENTS X 1 INTRODUCTION 1 1.1 Study Rationale and Objectives 1 1.2 Literature review 3 1.2.1 Airborne particles 3 1.2.2 Particles and health 5 1.2.3 Infiltration 10 1.2.4 Air cleaners 14 1.2.5 Spatial variability 16 1.2.6 Measurement of particles 17 1.2.7 Summary 21 2 METHODOLOGY 22 2.1 Data collection 22 2.1.1 Study design 22 2.1.2 Sampling locations 27 2.1.3 Volunteer recruitment 28 2.1.4 Sampling equipment set-up 29 2.2 Sample analysis 36 2.2.1 PM 2 5 mass collection 36 2.2.2 Absorbance measurements 37 2.2.3 Levoglucosan analysis 37 2.3 Data analysis 38 2.3.1 Continuous particle measurements 38 2.3.2 Infiltration calculations 39 2.3.3 Air cleaner effectiveness 43 2.3.4 Sensitivity analysis 43 2.3.5 Multivariable modeling 45 2.3.6 Spatial variability 48 3 RESULTS 50 3.1 Data collection 50 iii 3.1.1 Summary of collected samples 50 3.1.2 Volunteer recruitment 52 3.2 Gravimetric analysis 53 3.2.1 Data quality 53 3.2.2 Collocated pDR and HI 2 5 measurements 53 3.2.3 Absorbance and levoglucosan measurements 55 3.3 Direct reading P M 2 S data 59 3.3.1 Pre-sampling pDR tests 59 3.3.2 Data clean up 60 3.4 Infiltration calculations 65 3.4.1 Removal of indoor generated peaks 66 3.4.2 Calculation of infiltration factors 66 3.5 Air cleaner efficiency 70 3.6 Sensitivity analysis 71 3.6.1 Relative humidity 71 3.6.2 Baseline drift 73 3.6.3 Forcing intercept to zero 74 3.7 Multivariate modeling 77 3.7.1 Housing characteristics 77 3.7.2 Infiltration modeling 79 3.7.3 ACE modeling 80 3.8 Spatial variability 81 4 DISCUSSION 89 4.1 Data collection 89 4.2 Infiltration 89 4.3 Air cleaner use 91 4.4 Multivariable modeling 93 4.4.1 Infiltration prediction 93 4.4.2 ACE prediction 93 4.5 Monitor evaluation 94 4.5.1 Sensitivity analysis 94 4.5.2 Direct continuous and integrated filter monitor comparison 96 4.6 Absorbance and levoglucosan analysis 96 4.7 Study strengths and limitations 98 4.8 Conclusion 99 REFERENCES 102 iv APPENDICES 117 Appendix A: Consent forms 118 Appendix B: Contact Information Sheet 125 Appendix C: Equipment log sheets 128 Appendix D: Air cleaner specifics 133 Appendix E: Housing characteristics questionnaire 134 Appendix F : Activity log 137 Appendix G: Introductory letters 139 Appendix H : Media package 146 Appendix I: Absorbance protocol 150 Appendix J: Levoglucosan protocol 151 Appendix K: GC/MS Instrument Parameters for Levoglucosan Analysis 156 Appendix L: Infiltration calculations 157 Appendix M : Concentration graphs 159 Appendix N: Relative humidity scatter plots 167 Appendix O: Infiltration factors (F i „ f ) by home 170 Appendix P: Air cleaner efficiency values by home 171 Appendix Q: Infiltration modeling variable associations 172 Appendix R: Air cleaner efficiency modeling variable associations 173 v LIST OF TABLES Table 1.1 Summary of particle deposition in different locations of respiratory tract based on particle size (Spengler et al., 1996) 6 Table 3.1 Summary of location, date and primary smoke source of homes sampled 51 Table 3.2 Summary of PM2.5 concentration and ABS coefficients calculated for field blanks collected during winter and 2005 summer sampling periods 53 Table 3.3 Summary of absorbance (ABS) coefficients and levoglucosan concentrations for filter samples collected in winter and summer sampling periods 56 Table 3.4 Summary of relative difference (average difference between monitor readings/average concentration) in concentrations by pDR monitor and test.. 60 Table 3.5 Summary statistics of correlations between indoor and outdoor pDR monitors for pre-mid and mid-post collocation test periods 61 Table 3.6 Summary statistics of relative PM2.5 concentration difference values between indoor and outdoor pDR monitors for pre-mid and mid-post collocation test periods 62 Table 3.7 Concentration percentiles of homes identified as having low indoor or outdoor PM2.5 concentrations 64 Table 3.8 Summary of homes excluded from data analysis 65 Table 3.9 Summary of infiltration factors (Fjnf) calculated for homes sampled in winter and summer 67 Table 3.10 Summary of Air Cleaner Efficiency for summer and winter 71 Table 3.11 Comparison of Fjnf values calculated for winter and summer sampling periods for all RH and RH<85% data 72 Table 3.12 Comparison of A C E values calculated for winter (n=16) and summer (n=10) sampling periods for all RH and RH<85% data 73 v i Table 3.13 Comparison of Fj„f values calculated for winter and summer sampling periods for all data and baseline drift data removed 74 Table 3.14 Summary of difference between Fjnf values with and without an intercept term 75 Table 3.15 Summary of continuous independent variables used for infiltration and air cleaner efficiency prediction modeling 78 Table 3.16 Summary of categorical independent variables used for infiltration and air cleaner efficiency prediction modeling 78 Table 3.17 Results of regression analysis of effects of housing characteristics on infiltration in the summer and winter seasons 80 Table 3.18 Summary of correlations between outdoor home and BC Ministry of Environment monitor hourly PM2.5 concentrations over a 48-hour sampling period in the communities of Williams Lake, Surrey, Prince George, Lytton and Spence's Bridge 87 vn LIST OF FIGURES Figure 1-1 Source and removal processes of infiltration (modified from Thatcher & Layton, 1995) 11 Figure 1-2 Filter samplers, Harvard impactors (left), and direct reading monitors, personal DataRAMs (right), used in PM2.5 sampling during data collection... 18 Figure 2-1 HEPA filter room air cleaner (top) with HEPA filter (bottom left) and pre-filter (bottom right) removed 24 Figure 2-2 Summary of data collected for each home during winter and summer sampling periods 27 Figure 2-3 Set up of the pDR (left) and Q-Trak (right) indoors during winter and summer sampling periods 30 Figure 2-4 Set up of HI 2. S (top), pDR (beneath HI2.s) and HOBO (bottom right) monitors outdoors during winter sampling period 32 Figure 2-5 Set up of pDR monitors outdoors during winter sampling period 33 Figure 2-6 Set up of sampling equipment outdoors during summer sampling period with HI (top), HOBO (bottom left), and pDR (bottom right) 35 Figure 2-7 Comparison of indoor peaks of indoor (a) and outdoor (b) origin 41 Figure 2-8 Removal of PM data points for rising edge of identified indoor generated PM peaks 42 Figure 3-1 Map of Sleetsis Creek fire (active from July 29 to August 9, 2005) in relation to the communities of Spence's Bridge and Lytton (Ministry of Forests and Range, 2006) 51 Figure 3-2 Relationship between collocated pDR and HI2.5 concentrations measured in homes in winter (n=18) and summer (n=10) sampling periods 55 Figure 3-3 Relationship between ABS and concentration of levoglucosan recovered from winter (n=26) and summer (n=10) filter samples 57 Figure 3-4 Relationship between HI PM2.5 concentration and concentration of levoglucosan recovered from winter and summer filter samples 58 viii Figure 3-5 Distribution of correlations between indoor and outdoor pDR monitors for all collocation test periods 62 Figure 3-6 Distribution of relative differences (average difference between indoor and outdoor monitor readings/average concentration measured during test period) between indoor and outdoor pDR monitors for all collocation test periods 63 Figure 3-7 Relationship between relative humidity (RH) and outdoor pDR PM2.5 concentration measurements for home WL-05 66 Figure 3-8 Distribution of F i nf values calculated during summer and winter sampling seasons for filter and no filter days 68 Figure 3-9 Infiltration values (F;nf) for filter and no filter days for homes sampled in the winter sampling period 69 Figure 3-10 Infiltration values (Fjnf) for fdter and no filter days for homes sampled in the summer sampling period 70 Figure 3-11 Distribution of differences in F;nf values between all data and data with the intercept term forced through zero 76 Figure 3-12 Relationship between Fjnf (with intercept) and F i n f (0 intercept) data for a. all data (n=58) and b. outlier removed data (n=53) 77 Figure 3-13 Location of BC Ministry of Environment monitoring network and sampled homes in Williams Lake in August 2004 82 Figure 3-14 Location of BC Ministry of Environment monitoring network and sampled homes in Prince George in January to March 2005 83 Figure 3-15 Location of BC Ministry of Environment monitoring network in Whistler and Kamloops and location of sampling conducted in Lytton and Spence's Bridge in August 2005 (note: scale) 84 Figure 3-16 Location of BC Ministry of Environment monitoring network and sampled home in Surrey in September 2005 85 Figure 3-17 Relationship between 24 hour averaged PM2.5 data for outdoor sampled home and BC Ministry of Environment monitoring network for the communities of Williams Lake, Prince George, and Surrey 88 ix ACKNOWLEDGEMENTS I would like to thank my supervisor, Michael Brauer, for all his support and guidance throughout the last two years. I would also like to thank Timothy Larson, Susan Kennedy, and Ray Copes, my committee members. Thank you to Melanie Noullett and the members of her U N B C air pollution class for conducting the winter sampling for this project, Tim Ma for performing the laboratory analysis and well as Kay Teschke for offering her statistical expertise. I am extremely thankful to all study participants who allowed us to come into their homes for our sampling, many of whom were willing to participate within a few hours notice. Finally, I would like to thank my friends at SOEH, my sister, and Aman for being there through the "dark" times normally associated with being a graduate student. This study was funded by the British Columbia Centre for Disease Control. 1 INTRODUCTION 1.1 Study Rationale and Objectives Smoke from forest fires and residential wood burning is a significant source of PM2.5. Over one third of total Canadian PM2.;J concentrations are attributed to forest fire emissions (Rittmaster et al., 2006) while over one quarter of total PM2.5 concentrations are attributed to residential wood burning (Environment Canada, 2006). The release of such pollution can have negative health impacts on exposed populations. Studies investigating smoke affected communities have found increased chronic obstructive pulmonary disease (COPD) and asthma related emergency room visits (Duclos et al., 1990) and increased prevalence of various respiratory symptoms such as upper respiratory tract illness, asthma, and rhinitis (Brauer, 2000) linked to smoke exposure. Additionally both the occurrence of forest fires and the use of wood stoves are expected to increase due to climate change which will translate to greater exposures in the future. Although forest fires are a natural phenomenon and play an integral role in the renewal and maintenance of the ecosystem, changes in weather due to climate change may alter both the frequency and severity of fires (Flannigan et al., 2000). Increases in greenhouse gases, including carbon dioxide (CO2), are expected to influence weather patterns globally for which evidence is already being seen. Increases in fire emissions, including CO2 concentrations, were found to be elevated for Southeast Asia, southern and northern Africa, as well as South America due to El-Nino induced droughts from 1997 to 2001 (van der Werf et al., 2004). During this time, 66% of the observed C 0 2 growth rate was attributed to forest fires alone. Changes in weather, including temperature, precipitation, and wind, are expected to vary spatially leading to increases of up to 25-50% in the occurrence of forest fires in some areas of the United States (Flannigan et al., 2000). Additionally, the use of wood stoves will continue to increase as a result of increasing fuel costs (Health Canada, 2006) which will consequently lead to increased ambient PM2.5 concentrations. 1 For those residing in or near communities where woodstove use is frequent and/or nearby areas where forest fires are common, exposure can increase the risk for individuals to experience negative health effects. It is therefore necessary to provide appropriate information to affected populations on exposure levels as well as possible exposure reduction strategies. Residents of communities affected by residential wood burning and forest fire smoke are generally provided with recommendations of remaining indoors, restricting activities and using air cleaners during smoke events (Emmanuel, 2000; Government of the Northwest Territories, 2005; United States Environmental Protection Agency, 2005). It is assumed that remaining indoors during times of high smoke levels will reduce residents' exposure to PM 2 .5 levels. While the home is believed to provide a protective barrier against particulate air pollution, few studies have investigated actual levels inside homes during episodes of high air pollution such as those created by nearby forest fires. Furthermore, studies investigating infiltration suggest that for smaller particles including PM2.5, buildings may not provide as protective a barrier as believed, as these particles can easily penetrate through building surfaces, such as open doors and windows or through building cracks (Lai, 2002). In order to appropriately recommend that people remain indoors during episodes of high outdoor air pollution, it is necessary to characterize potential indoor exposure. Additionally, little information is available on the effectiveness of air cleaners in reducing indoor levels of PM2.5 in homes, and no studies have investigated the role of home characteristics on air cleaner effectiveness. Studies that have looked at the use of air cleaners during forest fire events in either reducing exposure (Henderson et al., 2005) or mitigating health effects (Mott et al., 2002) have lacked accompanying exposure measurements, large sample sizes, or appropriate controls to assess the effectiveness of such an intervention. The infiltration of outdoor PM2.5 from forest fires during summer and residential wood burning during winter in indoor home environments was investigated in order to better characterize the PM2.5 exposure of residents in smoke-affected communities. The effectiveness of HEPA filter room air cleaners in reducing indoor levels of PM2.5 was also examined in order to determine whether their use is an appropriate recommendation. Finally, the role of certain housing characteristics in affecting both infiltration and air 2 cleaner efficiency for homes was examined. The specific objectives of this study were the following: i . To measure indoor infiltration of outdoor PM2.5 from forest fires and residential wood smoke i i . To determine effectiveness of HEPA filter air cleaners in reducing indoor levels of PM2.5 i i i . To analyze the determinants of infiltration and air cleaner effectiveness in homes 1.2 Literature review 1.2.1 Airborne particles Smoke from forest fires and residential wood burning is a complex mixture of airborne particles, polycyclic aromatic hydrocarbons, aldehydes, carbon monoxide and volatile organic compounds (Sapkota et al., 2005). Electron microscopic sizing of haze particles collected in Singapore during 1997 Indonesian fires revealed that 94% of these particles were comprised of PM2.5 with the largest fraction of particles being below 1 pm in aerodynamic diameter (Emmanuel, 2000). Wood smoke particles have been shown to comprise of fine particulate following a bimodal distribution with peaks at 0.15 pm and 0.4 pm in aerodynamic diameters (Larson & Koenig, 1994). It is estimated that globally 35% of particulate emissions into the atmosphere are released from the combination of wildfires, prescribed burning, and other biomass burning (Robinson et al., 2004) while 80% of wintertime P M can be attributed to wood burning in some areas of the United States where woodstoves are commonplace (Larson & Koenig, 1994). Long range transport is an important issue as P M emissions from fires not only affect populations located near the burning source, but also have the potential to affect those residing in communities further away. For instance, forest fires in Quebec, Canada have been shown to cause increased PM2.5 levels in areas thousands of kilometers away in the United States (Sapkota et al., 2005). During fires, emitted particles rise within hot gas plumes to a 3 layer of the atmosphere known as the mixed layer, where they are available to be advected to the next layer, the troposphere. If particles remain within the mixed layer, turbulence of air ensures that particles are distributed efficiently and dilution of P M results. Once in the troposphere however, less air turbulence combined with stronger winds ensure that both less horizontal mixing, resulting in poor dilution, as well as greater transport of P M over long distances occur. Levoglucosan, a product of the breakdown of cellulose, is produced in large quantities during wood burning and has been suggested as a tracer for wood smoke (Nolte et al., 2001). In a study conducted by Nolte et al., 2001, levoglucosan was found to be present in the highest concentrations in analyzed wood smoke filter samples compared with other common cellulose breakdown products such as monomethylinositol, galactosan, and mannosan. Additionally, levoglucosan has no known non-combustion sources and is a relatively stable molecule in the atmosphere. These properties make levoglucosan an important tracer for studies investigating wood/forest fire smoke not only from local sources, but from long range transport as well (Fraser & Lakshmanan, 2000). Airborne particles can be defined as solid or liquid particles suspended in gas, such as air (Ruzer & Harley, 2005). Numerous sources and formation processes can lead to the generation of a complex mixture of particles with different chemical and physical properties. These particles can be classified into one of three categories based on particle size. The smallest size fraction mode is the nuclei mode and includes all particles less than 0.1 um in aerodynamic diameter (Ruzer & Harley, 2005). Nuclei mode particles are formed through combustion processes as well as through atmospheric reactions such as gas-to-particle conversions (Ruzer & Harley, 2005). These particles have short atmospheric life times and for this reason are not transported over large distances. Nuclei mode particles typically coagulate to form larger particles thus entering the accumulation mode. Accumulation mode particles fall within a size range of 0.1-2 um in aerodynamic diameter (Ruzer & Harley, 2005). Particles in this mode typically result from coagulation of smaller particles as well as from combustion and photochemical processes. 4 Accumulation mode particles have the longest residence time in the atmosphere relative to nuclei mode particles and the largest particle mode, coarse particles. Coarse particles, which include all particles greater than 2.5 um in aerodynamic diameter, are generally formed through physical processes such as mechanical grinding or re-suspension of dust and other large particles (Ruzer & Harley, 2005). 1.2.2 Part icles and health The exact mechanism by which particulate matter affects health is not well understood. It is recognized that particles of different size ranges deposit in different areas of the respiratory tract, which may therefore initiate different biological responses. Two size distributions need to be addressed when investigating particle exposure and health; fine particles which include all particles less than 2.5 um in aerodynamic diameter ( P M 2 5 ) and inhalable particles which include all particles less than 10 um in aerodynamic diameter (PMio). This size distinction is made according to particle deposition locations within the respiratory tract which are a function of P M source and composition. Inhalable particles, also known as thoracic fraction particles, can deposit in locations throughout the respiratory system. While larger particles (>2.5 and <10 um) deposit in the upper regions of the respiratory tract including the nose and nasopharyngeal regions, smaller particles (<2.5 um) deposit in the lower regions. Fine particulate matter is able to follow the air stream throughout the upper airways and deposit exclusively in the deeper airways of the respiratory tract including the bronchioles and alveoli. In the alveoli, particle clearance is both difficult and slow relative to upper airways and for this reason in conjunction with their longer atmospheric residence time as well as their long range transport potential, fine particles are considered to be potentially more harmful to health (Ruzer & Harley, 2005). Inhalation of smoke particles from forest fires specifically have been studied with the majority of deposition occurring in the alveolar interstitial region (Schollnberger et al., 2002). Table 1.1 below summarizes the various locations within the respiratory tract where different particle size ranges deposit. 5 Table 1.1 Summary of particle deposition in different locations of respiratory tract based on particle size (Spengler et al., 1996) Particle size (range) Penetration of particles 11 pm and up particles do not penetrate 7 - 11 um and up particles penetrate nasal passages 4.7-7 pm particles penetrate pliarynx 3.3 -4.7 pm particles penetrate trachea and primary bronchi 2,1 -3.3 Jim particles penetrate secondary bronchi 1.1 - 2.1 p m particles penetrate terminal bronchi 0.65 - 1.1 jim particles penetrate bronchioli 0.43 -0.65 um particles penetrate alveoli Once inhaled, particles are thought to cause various inflammatory and oxidative stress processes that lead to cardiovascular and respiratory disease (Risom et al., 2005). The generation of reactive oxygen species is suggested to be due to reactions between various transition metals or organic compounds adsorbed onto particle surfaces and respiratory tract surfaces, which in turn lead to altered cell function and activation of inflammatory cells (Risom et al., 2005). Inflammation may then lead to increased blood pressure, the promotion of atherosclerosis and myocardial infarction (Delfino et al., 2005). The most health damaging characteristic of smaller particles is believed to be their large surface area which allows for the transportation of adsorbed or condensed pollutants into the body (Delfino et al., 2005). For larger particles however, the composition of the particles themselves is suggested to be a major determinant in the resulting health effects (Risom et al., 2005). As different chemical components play different roles in the body, varying health effects can be experienced as a result of inhalation of these particles. Particles of different composition result from their formation sources. Coal combustion, for example leads to the formation of particles high in the elements arsenic and selenium while re-suspended soil particles may be high in silicon and aluminum (Thurston et al., 2005). Greater knowledge of the effects of inhaled particles on biological responses within the body is needed in order to have a better understanding of the resulting health effects. 6 Countless epidemiological studies have found evidence of associations between health and exposure to particulate matter (Goldberg et al., 2003; Samet et al., 2000; Schwartz et al., 1996). Studies have found positive associations between respiratory and cardiovascular mortality and exposure to urban ambient particulate matter around the world showing that both short term exposure as well as chronic exposure can increase mortality rates in a given population. In a review study by Pope & Dockery, 2006, increases in relative risk of mortality are shown to increase by a range of 0.6-1.5%, 0.6-1.8%, and 0.6-2.2% for all cause, cardiovascular and respiratory related deaths respectively for increases in PM2.5 or PM10 in the range of 10-40 Ug/m". Pope & Dockery, 2006 also summarize studies investigating long term P M exposure finding increases in relative risk of mortality values of 1-32%, 0.6-95%, and 0.8-81% for all cause, cardiovascular, and lung cancer related deaths respectively for increases in PM2.5 or PM10 of 10-20 ug/m. In addition to mortality, various morbidity outcomes have been associated with particulate exposure. These outcomes include aggravation of asthma, increased respiratory related hospital admissions, decreased lung function, and increased risk of lung cancer, all of which show stronger associations with exposure to PM2.5 versus P M , 0 (Bruce, 2002). Acute health effects and exposure to both PM2.5 and PM10 has been shown to have a linear relationship when investigating concentrations below 100 Ug/m with health effects occurring even at low concentrations (Schwela, 2000). Currently there is no evidence to support a threshold concentration of P M exposure. This is an important finding as sensitive subpopulations such as children, the elderly and those with pre-existing disease such as COPD or asthma, can experience adverse health effects even at low concentrations. 1.2.2.1 Vegetation source particulate exposure Exposure to vegetative biomass smoke including both wood and forest fire smoke has been shown to be associated with negative health impacts among exposed populations. Three main categories of studies examining health outcomes in exposed groups have been investigated. These represent studies of populations exposed to products of biomass 7 fuel burning during cooking in developing countries, studies of wild land fire fighters, and studies of communities affected by vegetation fires or wood smoke. Individuals living in developing countries, primarily women and children, are exposed to extremely high concentrations of PM2.5 due to biomass fuel burning during cooking. In many parts of the world, including rural India, Africa, and South America, cooking over open fires or on un-vented stoves coupled with poorly ventilated rooms results in prolonged exposure to high PM2.5 concentrations (Ezzati, 2005; Hood, 2002). A review study conducted by Brauer, 2000 showed linkages between the development of COPD in non-smoking women in various areas of the world including Mexico and Columbia and the use of biomass as a primary cooking fuel. Similar linkages have also been found in areas throughout Africa and Asia (Chan-Yeung et al., 2004). Acute respiratory infections (ARI) are a leading cause of child morbidity and mortality in developing countries and have been found to be associated with biomass burning exposures in several review studies investigating relationships between health and cooking smoke (Bruce, 2002; Smith et al., 2000). Children living in homes where the primary source of energy is solid fuel have a two to three times greater risk of developing ARI than children living in homes where cleaner fuel, such as liquid petroleum gas, is used (Smith, 2002). Several studies have found that exposure to smoke from forest fires in wild land fire fighters can lead to adverse health impacts in this population. One such study investigated forced expiratory volumes and airway responsiveness during fire fighting among 63 seasonal and full time US wildland fire fighters (Liu et al., 1992). Significant mean decreases in forced vital capacity (FVC) and forced expiratory volume (FEV|) values were found compared to pre-season values. A significant increase was also found for airway responsiveness compared to the pre-season as indicated by metacholine dose-response slopes. Differences were found to be greatest among fire fighters with asthma but not among those with a smoking history. Simiarily, Chia et al., 1990 reported an increase in airway responsiveness, as indicated by reactivity to inhaled histamine, among eight of ten fire fighters exposed to smoke in a laboratory chamber for a period of six hours. Researchers found that duration of service was associated with the increased 8 airway responsiveness. Another study found evidence of decreased lung function with short term exposure to smoke among 65 fire fighters (Slaughter et al., 2004). A mean cross-shift decrease in F E V i of 0.030L was associated with an increase of 1000 ug/m3 in PM3 .5. No such association was found for other smoke pollutants including acrolein, formaldehyde, and carbon monoxide. Despite the fact that fire fighters are typically healthier individuals compared to the general population, exposure to smoke causes negative health effects in the former group. This further indicates that a potential for negative health effects as a result of exposure to forest fire and wood smoke exposure, exists among the general public. Studies investigating the health impacts in communities due to smoke exposure from nearby forest fires and wood burning have also found evidence for negative health impacts. Emergency room visits were increased by 40% and 30% for COPD and asthma related visits respectively during a two week period during which forest fire activity was high in California (Duclos et al., 1990). Another study conducted in Southern California during 2003 wildfires found an association between increased eye and respiratory symptoms as well as both medication use and physician visits with wildfire smoke exposure (Kuenzli et al., in press). Risks for symptoms, medication use and physician visits were found to increase with the number of exposure days. In 1997-98 when forest fires were rampant across Southeast Asia, a P M i 0 increase of 100 Ug/m3 was found to cause a 12% increase in upper respiratory tract illness, a 19% increase in asthma, and a 26% increase in rhinitis in the exposed population (Aditama, 2000). Additionally in 2003, British Columbia experienced an active forest fire season with PM2.5 concentrations in the most affected cities of Kelowna and Kamloops reaching as high as 200 Ug/m3 and 140 Ug/m3 respectively (Moore et al., 2006). During a 3 week period of fire activity, physician visits for respiratory diseases in Kelowna increased by 46-78% compared to the 10 year mean. It has been suggested that emergency room visits may not be representative of populations who may not have access to medical care or of individuals who underestimate their need for medical attention and therefore actual health effects may be underestimated (Duclos et-al., 1990). 9 The majority of health studies looking at wood smoke exposure have investigated heath outcomes in children (Boman et al., 2003; Koenig et al., 1993; Larson & Koenig, 1994). Koeing et al., 1993 found a 20 Ug/m' increase in P M 2 5 to be associated with declines in lung function, as measured by forced vital capacity (FVC) values, among asthmatic school children in Seattle. Six out of eight studies investigating wood smoke exposure found the presence of wood stoves to be associated with increased symptoms, reduced spirometric measures or increased emergency room visits among children, in a review by Larson & Koenig, 1994. In another study, a positive association was found between rates of respiratory emergency department visits in Seattle and biomarkers of the vegetative burning source markers total carbon and arsenic indicating exposure to wood smoke (Schreuder et al., 2006). PM2.5 is believed to be the main pollutant in wood smoke to cause detrimental health effects (Larson & Koenig, 1994). Similar to ambient particulate matter exposure, wood smoke particles are suggested to cause inflammatory responses which then lead to the development of health effects. A clinical study conducted by Barregard et al., 2006 investigated the relationship between wood smoke exposure and inflammatory responses. Thirteen subjects were exposed to wood smoke and to filtered air for a period of 4 hours each. Analysis of blood and urine samples collected before and after each exposure showed increased levels in inflammation biomarkers such as serum amyloid (a cardiovascular risk factor), and in coagulation factors such as factor VIII in plasma and lipid peroxidation as evidenced through increased urinary excretion of a major F2-isoprostance in 9 wood smoke exposed subjects without similar increases after exposure to filtered air. As with ambient particles, more research on wood smoke particles with regards to the disease mechanisms they cause upon inhalation is needed. 1.2.3 Infiltration Since there is compelling evidence of the harmful effects of outdoor air pollution, including that produced in vegetation fires, on health it is important to accurately characterize exposure levels in order to assess possible health effects among affected communities. A major factor in determining levels of exposure for any given ambient concentration is the efficiency of indoor infiltration. Individuals spend up to 90% of their 10 time indoors and consequently, exposure to P M indoors comprises a major portion of total exposure. Infiltration is not only important to address when assigning levels of exposure but also when investigating possible exposure reduction strategies. 1.2.3.1 Infiltration parameters Infiltration is defined as the fraction of outdoor particles that penetrate indoors and remain suspended (Wilson et al., 2000). This definition takes into account the different processes that lead to the addition and removal of particle concentrations indoors. Processes that can increase indoor P M levels include penetration, indoor generation, and re-suspension while processes that can decrease indoor concentrations include deposition and ex-filtration (Ruzer & Harley, 2005). A simplified diagram of these processes can be seen below in Figure 1.1. Rcsuspension Figure 1-1 Source and removal processes of infiltration (modified from Thatcher & Layton, 1995) Penetration is defined as the fraction of particles carried by air entering a building from outdoors (Ruzer & Harley, 2005). The penetration of different particle sizes into homes and buildings has been extensively studied and it has been determined that penetration of PM2.5 is approximately 100%, translating to a penetration factor of one (Thatcher & Layton, 1995; Wallace, 1996) while a range of values have been found for larger particle sizes. Although some studies have found a penetration factor of one for particles >2.5 pm in aerodynamic diameter (Thatcher & Layton, 1995; Wallace, 1996) other studies 11 have found decreasing penetration values with increasing particle size. These penetration factors range from values of 0.9 for 0.02 um sized particles and 0.3 for 6 um sized particles (Long et al., 2001). Deposition of particles has also been studied extensively in both laboratory and residential/building settings. Indoor particles deposit onto surfaces either by Brownian motion or gravitational settling (Tung et al., 1999). While fine particles deposit by Brownian motion, larger particles will typically settle onto indoor surfaces via gravitational settling. Deposition rate of particles is also dependent on other factors including particle stability, indoor surface to volume ratio and airflow (Thatcher et al., 2002). Air exchange within a home promotes the replacement of indoor air with fresh air from outdoors. Low air exchange rates (AERs) within a home generally result in higher indoor concentrations due to the build up of particles (Abt et al., 2000). At low air exchange rates, the residence time of particles increases and deposition becomes the major indoor particle removal process. Alternatively, when A E R is high, deposition is secondary to air exchange as a removal process (Vette et al., 2001). Indoor P M generation from activities such as cooking and cleaning can lead to indoor to outdoor P M ratios of greater than one. Various indoor activities have been found to generate different particle size fractions. Abt et al., 2000 found that particle concentrations in the range of 2.5-10 um were highest during the morning and evening due to the presence of people in the home which led to re-suspension of coarse particles. Alternatively, fine fraction particles were found to be highest specifically during cooking activities. 1.2.3.2 Measuring infiltration Little work has been done on quantifying infiltration of P M in residential settings. Mass balance models can be valuable tools in predicting exposure levels. These models follow 12 the law of conservation and model exposure by taking into account the different factors important in particle exposure. A recursive model developed by Switzer & Ott, 1992 describes indoor particle concentration as the relationship between penetration of outdoor particles, current indoor particle generation and indoor concentration from a period of time earlier. This relationship is summarized below in equation 1 where ai, a2, S i n and C represent penetration of outdoor particles, decay of indoor particles, indoor generation and airborne concentration of particles respectively. C i n = a l ( C o u t ) t + a 2 ( C i r J t - 1 + S i n (1) Use of the recursive model coupled with the use of censoring algorithms (Allen et al., 2003) to quantify infiltration is useful in expanding knowledge of infiltration to a number of homes. Two other methods of quantifying infiltration include comparison of indoor and outdoor concentrations of a chemical tracer, such as sulphur, that has no indoor source (Suh et al., 1992) and direct comparison of indoor and outdoor concentrations when no indoor sources are present (Vette et al., 2001). These methods are associated with different limitations. When sampling under conditions when no indoor sources are present, the environment must either be unoccupied or sampling must occur during nighttime when indoor activity is low. The amount of data collected is limited when only nighttime hours can be used which may necessitate a longer sampling time. Although tracer methods can be used to quantify infiltration in occupied homes, they cannot be used in homes where indoor sulphur sources such as kerosenes and humidifiers exist (Allen et al., 2003; Leaderer et al., 1999). Furthermore, if sulphur particles are not representative in terms of size of the particles of concern, over- or under-estimations of infiltration can result. For example, Wallace et al., 2006 argue that the sulphur tracer method quantifies infiltration particles in the size range of 0.06 to 0.5 pm and would provide overestimates of infiltration rates for particles in a larger size range. In contrast, the recursive model can be used to quantify infiltration for any particle size. The recursive model is further advantageous due to the use of continuous concentration data in infiltration factor calculations which may be more representative compared to methods 13 that only use an average measurement of concentration. P M measurements with the use of continuous reading monitors are relatively easy to collect and can provide important time-resolved data on P M concentration fluctuations during smoke episodes caused by both forest fire and wood burning. Studies have used other mass balance models to assign personal as well as indoor exposure levels. A personal exposure model includes not only home exposures but also other relevant microenvironments, such as automobile and work exposures (Mage, 2001) Mass balance models can be further modified to include other terms of interest, including a filtration term to represent air cleaner operation (Wilson et al., 2000). Air cleaners will increase the decay rate of particles by drawing air into the cleaner and passing it through the air filter. As these particles are being drawn into the air cleaner, their residence time in the air is decreased and therefore the use of an air cleaner is suggested to be associated with a decrease in infiltration. 1.2.4 Air cleaners 1.2.4.1 Lowering of indoor concentrations Studies have indicated that portable room air cleaners are effective at reducing indoor particle concentrations although many of these studies have evaluated larger particle sizes than those relevant to this work such as house dust mites (Dermatophagoides sp.) (Antonicelli et al., 1991), dog allergens (Green et al., 1999) and cat allergens (Wood et al., 1998). Ward et al., 2005 found that the use of HEPA filter air cleaners in residential dwellings reduced indoor PM2.5 concentrations by 50-90% when compared to not using air cleaners. In this study, one to three air cleaners were run in a residential dwelling as a simulation of a shelter-in-place during the release of a possible biological agent in the size range of 0.1-2 um aerodynamic diameter. Different combinations of two heating, ventilating, and air-conditioning (HVAC) filter efficiencies, three H V A C flow rates and three air exchange rates were modeled with the release of particles outdoors and a relative air cleaner efficiency (ACE) was calculated for each scenario. A C E was highest when 3 14 versus one to two air cleaners were operated. Additionally, when air exchange was high, air cleaner efficiency was lowered due to the higher influx of outdoor particles. Offerman et al., 1985 tested the use of different air cleaner types, including HEPA filters, in reducing levels of environmental tobacco smoke (ETS) particulate with a median particle diameter of 0.15 pm. Researchers found that both HEPA filter and electrostatic air cleaners were efficient at removing particles (calculated as observed cleaning rate divided by measured flow rate) while ion-generating air cleaners were not. While HEPA filter air cleaners and electrostatic air cleaners had effective cleaning rates of 86 ± 9% and 57 ± 1 1 % respectively, effective cleaning rates for ion-generators were found to be less than 25%. Currently, only one study has assessed the use of air cleaners in lowering indoor P M exposures during forest fires while no studies have looked at P M exposure during wood burning smoke episodes. Henderson et al., 2005 collected indoor and outdoor PM2.5 measurements for 24 to 48 hours in four pairs of homes during a forest fire episode and investigated the effectiveness of using air cleaners versus only keeping windows closed. One home out of each pair was equipped with portable electrostatic air cleaners that were operated continuously during the sampling period in addition to keeping windows closed, while windows were closed in the other home and no cleaners were operated. Differences in indoor concentrations were investigated and intervention homes (i.e. the home with air cleaners) were found to have 63-88% lower indoor PM2.5 concentrations compared to non-intervention homes. Although homes were matched as closely as possible in terms of housing characteristics some differences, such as in rates of air exchange, did exist among the homes which potentially affected indoor to outdoor ratios. Additionally, the effects of housing characteristics, such as age and size of home, on infiltration were not investigated. 15 1.2.4.2 Air cleaner use and health Varied results have been found by researchers investigating the use of air cleaners and health outcomes. A number of studies have looked at the operation of various types of air cleaners in mitigating health symptoms in general settings. Bascom et al., 1996 found that use of electrostatic air cleaners was associated with decreases in headaches and rhinonorrhea compared to use of sham air cleaners (i.e. non-functional cleaners) during exposure to environmental tobacco smoke. Other symptoms such as nasal irritation and congestion were not affected by air cleaner use. MacDonald et al., 2002 performed a systematic review of randomized controlled trails with HEPA filter air cleaners. Researchers reviewed studies published from 1976 to 2000 that used a combination of HEPA filter and sham air cleaners and looked at various health outcomes including peak expiratory flow (PEF), medication use, sleep disturbance and symptoms such as nasal congestion, itchy eyes, and sneezing. The use of air cleaners was found to be associated with a small but significant decrease in both experiencing of total symptoms and sleep disturbance but was not associated with medication use, PEF and nasal symptoms such as nasal irritation and congestion. There is currently only one study that has investigated the role of air cleaners in mitigating health symptoms during a forest fire event. Mott et al., 2002 found that operation of portable HEPA filter air cleaners was linked to decreased frequency of respiratory symptom reporting by affected community members while other interventions such as mask use and evacuation from the area were not associated with any such decrease. Although this study attempted to measure effectiveness of interventions, no exposure measurements were collected and consequently the effectiveness of the air cleaners could not be quantified. 1.2.5 Spatial variability When outdoor PM2.5 concentrations are elevated, such as during forest fire or wood burning events, it is important for health authorities to recommend exposure reduction strategies to affected community members. Typically monitoring networks recording P M 16 concentrations are used to determine exposure levels within a community. Although outdoor home concentrations have been shown to be well correlated to central site monitoring sites with correlations as high as 0.96 (Clayton et al., 1993), absolute outdoor home concentrations may be over or underestimated. Many factors can affect the degree of agreement between outdoor home concentrations and monitoring networks. Correlations between the two are expected to decrease with increasing distance between the home and monitoring network. Additionally, primary particle concentrations may be more concentrated in close proximity to the source while secondary particles may be more uniformly distributed throughout an area (Monn, 2001). In the case of forest fires and residential wood burning, homes localized in areas of burning may be greatly affected by particle concentrations that may not register on monitoring networks located further away. Furthermore, in communities where a high degree of geographical variability may exist, the use of central site monitoring as the sole representation of exposure may lead to severe exposure misclassification (Ozkaynak et al., 1996). Therefore it is important to determine whether monitoring networks provide an appropriate description of community exposure to ensure that community health is being protected. 1.2.6 Measurement of particles Particulate matter can be measured using either integrated filter samplers or continuous reading monitors. Different advantages are associated with the use of these monitors. Filter samples can be collected over a period of time and an average P M concentration can be obtained for this period. Collected P M can additionally be analyzed for specific chemicals and elements which can then be attributed to particular sources. Continuous monitors can provide the same information on average concentration during a given period but can also provide more time-resolved information in terms of short-term differences in concentrations. These short term differences, when coupled with other information such as meteorological or time activity data, can also be used to characterize specific sources. Both integrated filter monitors (Harvard impactors) and continuous 17 reading monitors (Personal DataRAMs) will be used for P M 2 5 sampling in this work (see Figure 1.2) and are discussed in greater detail below. Figure 1-2 Filter samplers, Harvard impactors (left), and direct reading monitors, personal DataRAMs (right), used in PM2.5 sampling during data collection 1.2.6.1 Integrated filter monitors Integrated filter sampling in this study was conducted with the use of Harvard Impactors (His) (Air Diagnostics and Engineering, Inc., Naples, ME). His have been used for many P M exposure studies (Allen et al., 2003; Janssen et al., 1998; Koutrakis et al., 1992; Long et al., 2000) and have been evaluated against the P M 2 s Federal Reference Method (FRM) with results showing high correlation between the two sampling methods (r=0.99) (Babich et al., 2000). The HI is often used as a reference method for the evaluation of other instruments including the T E O M (R&P Inc., Albant, NY) (Turner et al., 2000) nephelometers (Radiance Research)(Allen et al., 2003) and the personal dataRAM (pDR-1000, MIE Inc, Bedford, M A ) (Liu et al., 2002; Quintana et al., 2000). The HI consists of a round metal base which contains a filter and a round nozzle which holds a series of two impactor plates. For this study, impactors with a 50% cutoff point of 2.5 urn aerodynamic diameter were used. The HI? 5 is connected to a pump which draws air through the sampler at a set volume. Particles 2.5 um or smaller flow past the impactor plates while larger particles are unable to follow the air stream and impact onto 18 the greased plates. The air stream continues to flow through the filter located at the base of the sampler onto which particles are collected. The design of the stacked impactor plates has been shown to be advantageous during sampling over a long duration due to the prevention of overloading of the plates (Turner et al., 2000). 1.2.6.2 Continuous reading monitors Due to recent evidence to support that short-term exposures in P M may greatly influence health (Delfino et al., 1998; Fischer & Koshland, in press) continuous reading monitors have been used extensively in P M health exposure studies. Additionally these monitors are useful in the understanding of P M generation and removal sources. Continuous monitors can be divided into two categories of direct and indirect reading monitors, both of which will be used in the current study. 1.2.6.2.1 Direct reading monitor Direct monitors, such as the tapered element oscillating microbalance (TEOM), measure particulate mass directly. The T E O M has been used in numerous P M exposure studies as a cost effective method of P M measurement (Ebelt et al., 2000; Evans et al., 2000; Long et al., 2000; Long et al., 2001; Quintana et al., 2000; Sapkota et al., 2005). The T E O M is a filter-based monitor that collects particles on a heated filter located at the end of a vibrating tube connected to a downstream pump (Ruzer & Harley, 2005). As P M is collected on the monitor, vibration of the filter/tube assembly changes accordingly such that the mass collected is directly related to its vibrational frequency. The filter is heated in order to remove any water that may be attached to particles therefore preventing false high readings. Babich et al., 2000 found that when compared to the Federal Reference Method (FRM), the T E O M showed good agreement in East Coast urban areas with correlations of 0.87-0.97. Lower correlations of 0.64-0.78 were found in other urban areas with high ammonium nitrate and/or organic components of P M composition. 19 1.2.6.2.2 Indirect reading monitor The second category of continuous monitors includes monitors that measure particulate mass indirectly. In the current study, the personal DataRAM (pDR) was used to collect measurements of P M 2 5 indoors and outdoors of each home. This instrument is a passive monitor that indirectly measures P M mass by detecting the amount of light scatter caused by particles diffusing into an open sensing chamber located at the top of the monitor. In the sensing chamber, infrared light is shone onto particles and the amount of scatter is reported in mass concentration (mg/m3) within a measurement range of 0.001 mg/m3 to 400 mg/m3 (Thermo Anderson, 2003). Although the manufacturer reports that the pDR has a maximum response to particles within the range of 0.1 tolO um, studies have shown that this instrument is a better indicator of PM2.5 with maximum collection efficiency at a particle size of 0.6 um (Quintana et al., 2000; Switzer, 2001). There are three modes in which the pDR is available; passive, active, and active heated. The passive mode pDR was used for data collection in this work. The three modes have been tested by researchers for potential interference from different factors such as high humidity, indoor sources, and low and high particle concentration levels causing positive or negative drift. Specifically, the effects of high humidity, and the occurrence of drift are discussed. Although many studies have concluded that the pDR is a valuable tool in epidemiological studies, concerns regarding the use of this instrument in environments with high humidity have been raised. Since the pDR measures concentration in terms of light scatter, measurements are greatly affected by relative humidity (RH). High R H leads to increased water vapor in the air leading to increased particle size for hygroscopic particles, which therefore alters light scattering properties (Poschl, 2005). It has been found that when the pDR is used in environments with R H above 85%, the data collected is highly unreliable (Quintana et al., 2000; Wu et a l , 2005). Different methods to manage this data have been suggested including statistical approaches and use of heaters or diffusion dryers to remove particle-bound water. Wu et al., 2005 found that the use of a heated monitor did not provide much protection against R H >85% (Wu et al., 2005). 20 Quintana et al., 2000 concluded that removal of any data with concurrent RH>85% measurements and interpolation of the missing data points using splines was the optimum way to handle such data. Alternatively, Wu et al., 2005 used statistical correction equations to correct pDR data collected when R H values were between 85-94% and found that data collected when RH>95% could not be accurately corrected and was removed from analysis. Another potential limitation associated with use of pDRs is negative drift. Negative drift occurs when the pDR underestimates the true concentration of P M due to zeroing errors (Wu et al., 2005). Negative drift can be identified when the time weighed average concentration (TWACman) calculated manually from the data output is not equal to the TWACpDR recorded by the instrument (Wu et al., 2005). Discrepancies result because the instrument memory which records the time series measurements records only non-negative values and assigns anything detected as a negative concentration a value of "0" mg/m. The internal memory which calculates the T W A C does record negative concentrations. When negative drift occurs, the TWACpDR will be lower than the TWACman (Wu et al., 2005). Since negative drift data cannot be corrected in anyway due to the uncertainty of the specific data points experiencing negative drift, it is suggested that the entire sample be removed from analysis (Wu et al., 2005). 1.2.7 Summary Gaps in the knowledge of the contribution of outdoor PM2.5 into homes during high P M events caused by forest fires and residential wood burning as well as the effectiveness of air cleaners during this time in reducing indoor PM2.5 exposure exist. Residents may experience a number of health effects due to exposure and in order to provide public health recommendations to these individuals, it is necessary to gather more evidence based information. In this study, indoor and outdoor sampling was conducted for homes affected by smoke from nearby forest fires or from neighborhood residential wood burning. By quantifying particle infiltration and air cleaner efficiency in these homes, an attempt was made to increase the knowledge base for these two phenomena. 21 2 METHODOLOGY 2.1 Data collection 2.1.1 Study design Indoor and outdoor P M 2 5 sampling was conducted in homes affected by residential wood smoke in the winter and in homes affected by forest fire smoke in the summer. Sampling was conducted in the winter of 2004 and in the summers of 2004 and 2005. Each home was sampled for a minimum of 48 hours during which time a portable HEPA filter room air cleaner was introduced into the home as a possible intervention for lowering indoor levels of PM2.5. The study was approved by the U B C Behavioural Research Ethics Committee (B03-0602). For each sampling site, a volunteer's home was visited three times. On the first day, any questions the resident had regarding the study were answered after which point they were asked to read and sign an informed consent form (see Appendix A). Volunteers were informed about the sampling protocol with regards to sampling equipment to be used, the parameters to be measured, as well as the sampling timeline. Next, sampling equipment was set up. Two direct reading monitors, the Q-Trak (8551, TSI Incorporated, USA) and the pDR (pDR-1000, Thermo Andersen, Smyrna, GA, U S A ) , were set up indoors. The Q-Trak was used to log 1-minute averages of carbon dioxide (CO2), temperature, and relative humidity (RH) while the pDR logged 1-minute averages of PM2.5 concentration. Another pDR monitor was set up outdoors with a HOBO monitor (H08-032-08, Onset Computer Corp, Pocasset, M A ) , which logged 1-minute intervals of temperature and R H . An integrated PM2.5 sample was also collected outside each home using the Harvard Impactor, HI, (Air Diagnostics and Engineering, Inc., Harrison, ME). The sampling equipment is described in detail below under section "2.1.4 Sampling equipment setup." Once set up was completed, the resident was encouraged to resume their activities as normal. At this time, any additional questions were answered, the volunteer was left with a contact sheet with all study contacts' names and information (see Appendix B), and a time to visit the home the next day, after approximately 24 hours, was agreed upon. On 22 the second day, current tests for the direct reading monitors were stopped and downloaded and mid-flow was measured for the HI (see Appendix C for all equipment information log sheets). At the end of the sampling period, the home was visited again at which point all tests were ended, sampling equipment was taken down and the volunteer was thanked for their participation in the study. A similar protocol was followed for both summer and winter sampling periods although the winter protocol was modified slightly. A l l winter sampling was conducted by Melanie Noullett and the students of her fourth year air pollution class at the University of Northern British Columbia, Prince George. One modification included a longer sampling time of approximately 60 hours for each home. Set up of two homes was normally conducted in the evening time to accommodate for students' class schedules. For each home, two students, accompanied by either Melanie Noullett or a research assistant were responsible for the set up and take down of equipment. The two students were responsible for the operation of the pDR monitors while all housing questionnaires were administered by either Melanie or the research assistant. Additionally, one student was assigned the responsibility for the operation of the HI 2 5 while the other student was assigned the responsibility for the operation of the HOBO, Q-Trak and the air cleaner. A checklist was followed to ensure all monitors were set up correctly and set up of equipment was then reviewed. 2.1.1.1 Air cleaner operation In order to compare the effect on PM2.5 infiltration with and without air cleaner (18150, Honeywell, Tennesse, US) operation within each home, the HEPA filter was installed in the air cleaner for only one of the two sampling days. The resident of each home was kept blind to the fact that the air cleaner was run without the filter for any time during the sampling period. Consequently, when the filter was put in or removed, the air cleaner was taken outside for "maintenance" to ensure that the resident did not suspect that it was modified in any way. A pre-filter located at the bottom of the air cleaner was also removed when the HEPA filter was removed. The pre-filter functioned to remove any 23 large debris from the air. The design of the air cleaner used for the study allowed for easy removal and installation of both filters. Figure 2.1 below shows both the HEPA filter and pre-filter removed from the air cleaner. A coin was flipped on the first day in order to randomly assign which of the two days the filter was run in the air cleaner. Figure 2-1 HEPA filter room air cleaner (top) with HEPA filter (bottom left) and pre-filter (bottom right) removed Air cleaner operation is dependent on both particle removal efficiency and the airflow of the cleaner. HEPA filters are designed to remove 99.97% of the particles sized 0.3 pm with greater efficiency at smaller or larger particles. The CADR (Clean Air Delivery Rate) provides information on air flow as well as appropriate room sizing. Air cleaners are typically given a C A D R for each of the following; tobacco smoke, dust, and pollen particles within the size range of 0.1 to 11 pm. For the air cleaner used in this study, a CADR of 150 for tobacco smoke was specified by the manufacturer (see Appendix D for air cleaner specifics) and based on room size, different efficiencies exist. An air cleaner with a CADR of 150 will remove 89% of the particles in a room size of 9 x 12 ft2, 74% of the particles in a room size of 12 x 18 ft2, and 51% of the particles in a room size of 18 x 24 ft2 (Office of Air and Radiation, 2006). For the particular model used, the maximum room size under which use of the cleaner was recommended by the manufacturer was 216 square feet (18 x 12 ft2). A l l rooms in which the air cleaner was used met this 24 specification. The air cleaner was run on the "high" setting on both days unless the resident was particularly sensitive to the noise in which case the "medium" setting was used. Data on chronic respiratory health as well as acute respiratory symptoms experienced throughout the sampling period was collected from, the volunteer and in order to prevent symptom reporting bias, the residents were not told that the operation of the air cleaner would be modified in any way. This data was not analyzed due to limited collection of data from the target population of volunteers suffering from pre-existing respiratory disease such as chronic obstructive respiratory disease (COPD) or asthma and will therefore not be discussed further in this work. 2.1.1.2 Supplementary data collection 2.1.1.2.1 Housing characteristics information Data on each home sampled was gathered on the first day of sampling in the form of a housing characteristics questionnaire (see Appendix E). This questionnaire was used to gather basic information on the dwelling including parameters such as age and square footage of the home, the type of stove and heating system, as well as use of fireplaces, air conditioners, and windows. This information was later used for infiltration and air cleaner efficiency prediction modeling (see section 2.3.5 Multivariable modeling). 2.1.1.2.2 Activity within the home For each home, information on activity within the home was collected. A time activity log was divided into 30 minute intervals and had headings such as "location in home", "cooking", "cleaning", and "windows open" under which residents could check off the relevant activities occurring within the home (see Appendix F). At the start of each sampling period, the log was left with the resident for completion over the two sampling days. This allowed for information on particle generating activities within the home to be tracked. The collected information was then used to determine whether large indoor PM2.5 peaks measured by the indoor pDR had possible indoor generation sources. 25 2.1.1.2.3 Air exchange rate Finally, the residents of each home were asked to completely vacate the home for any 2 hour period during the sampling period and to log this period on the time activity log. During this time, the Q-Trak recorded the decay in CO2 levels due to the absence of a source in the unoccupied home. This decay was then used to calculate an average air exchange rate (AER) within the home using equation 2 below. A E R values were then used in the infiltration and air cleaner efficiency prediction modeling mentioned earlier. Air exchange rate [ In C ( t 2 ) - I n C f t J ] (AER) = : ( V V ( 2 ) Where: Air exchange rate (AER) = number of air exchanges per hour (h"1) In C = log normal concentration (ug/m3) ti = time at start of measurement period (hours) t? = time at end of measurement period (hours) 2.1.1.3 Summary of data collection Figure 2.2 below summarizes the data collection for each home for day 1 and 2 of the sampling period. 26 Day 1 24 hours Indoor (bedroom) / outdoor monitoring P M 2 5 , RH, C 0 2 Air Cleaner +/- HEPA filter Activity Log Housing Characteristics Survey Day 2 24 hours Indoor (bedroom) / outdoor monitoring P M 2 5 , RH, C Q 2 Air Cleaner -/+ HEPA filter Activity Log Figure 2-2 Summary of data collected for each home during winter and summer sampling periods 2.1.2 Sampl ing locat ions 2.1.2.1 Winter A l l wintertime sampling was conducted in Prince George. Prince George was chosen as the winter time sampling site due to the presence of multiple sources of PM 2.5 including wintertime residential wood burning. 2.1.2.2 Summer In the summer, sampling was conducted throughout Southern British Columbia during the forest fire season in 2004 and 2005. Selection of communities impacted by forest fire smoke was based on information gathered from fire maps, wind direction, satellite imagery and where available, PM2.5 monitoring data. These resources were available on-line and were monitored regularly to ensure that communities affected by forest fire smoke were identified as quickly as possible. Once such a community was identified, 27 logistics such as transportation to the community as well as contacting possible volunteers were worked out. 2.1.3 Volunteer recruitment Initially volunteer recruitment was primarily directed towards individuals with pre-existing respiratory disease such as COPD or asthma. This was done through the BC Lung Association, regional health authority, local physicians and local support groups. When COPD or asthma volunteers could not be recruited during forest fire or wood burning events, healthy volunteers were also considered for participation in the study. An introductory letter explaining the objectives of the study was given to potential volunteers for summer and winter recruitment (see Appendix G). A media package (see Appendix H) was also used for recruitment purposes and was sent the local newspaper in Williams Lake and Prince George and posted on community bulletin boards in Lytton, Kamloops, and Spence's Bridge. The media package explained the potential negative health impacts of wood and forest fire smoke exposure, explained the study being conducted, as well as gave contact information for those interested in volunteering. Additionally, a radio interview was conducted at the local radio station in Williams Lake in August 2004. Again, the purpose of the study was explained and contact information for those wishing to participate in the study was given. The only exclusion criteria were that all volunteers had to be non-smokers and no smoking could occur in the home during sampling. Residents in homes with wood stoves were also asked to refrain from operating them during the sampling period. Names received from various contacts of potential volunteers were contacted and informed of the study. If contacts appeared interested, they were visited at their home at which point the introductory letter was given to them. Any questions or concerns were addressed at this point as well. For wintertime sampling, all volunteer homes were visited at least one time prior to sampling and volunteers were contacted at a later time to schedule a time at which sampling could begin. In Prince George the majority of the volunteer recruitment was conducted through a local COPD support group. A presentation about the study was done by Melanie Noullett at a group meeting held at the 28 local hospital. Group members were told about the study and a sign up sheet was passed around for those who were interesting in volunteering to fill out. Collected names were followed up on and all interested participants were enrolled. Enrolled participants were also asked to give names of friends and family members who would also be interested in volunteering and this information was subsequently followed up on. Prior to the 2005 summer sampling period, potential volunteers in areas that could possibly be affected by forest fires were contacted. They were informed about the study and were asked if in the case of a nearby forest fire event, they would be willing to volunteer with only a few hours notice. Volunteers were contacted in Williams Lake, Kelowna, Quesnel, Vernon, Merrit and Penticton. 2.1.4 Sampl ing equipment set-up 2.1.4.1 Indoor set up A l l indoor equipment was placed in the main bedroom of the home. This area of the home was considered to be occupied for the largest portion of time relative to any other room in the home. For two homes, monitors were set up in the living room. For one home, the resident was resistant to place sampling monitors in the main bedroom due to noise while for the other home, the resident believed that more time was spent in the living room compared to the bedroom. The indoor setup was aimed to maximize the collection of representative measurements while minimizing disruption to the residents as well as to the sampling equipment. Figure 2.3 below shows the indoor set up in homes. The HEPA filter portable room air cleaner (not shown) was placed on the opposite side of the room for each home. 29 Figure 2-3 Set up of the pDR (left) and Q-Trak (right) indoors during winter and summer sampling periods 2.1.4.1.1 Q-Trak monitor set up The Q-Trak monitor was placed on a step ladder in the main bedroom of the home at approximately breathing level to collect measurements on C 0 2 , temperature, and relative humidity. 2.1.4.1.2 Personal DataRAM monitor set up Indoor and outdoor P M 2 s sampling was conducted for each home using a continuous light-scattering monitor, the pDR. The indoor pDR monitor was placed next to the Q-Trak on a step ladder. 2.1.4.1.2.1 Collocation of indoor and outdoor monitors In order to ensure valid data was collected with the pDR monitors, 10 minute collocation data was collected by placing both monitors indoors and outdoors before each sampling session. The indoor and outdoor pDR monitors were placed side by side at the indoor sampling location for 10 minutes and then side by side at the outdoor sampling location for 10 minutes. This test was then conducted again at the end of the first 24 hours of sampling and once again at the end of 48 hours of sampling. Additionally, the indoor and outdoor pDR monitors were zeroed before and at the end of each 24 hour sampling period for each home. Each pDR instrument is supplied with a 30 zeroing pouch and a hand-held pump equipped with a HEPA filter by the manufacturer. The pDR monitor was first turned on and allowed to perform internal calibrations. Once the prompt for zeroing was displayed on the screen, the monitor was placed in the zeroing pouch. The pump was used to fill the pouch with particle free air approximately half way after which point zeroing was initiated by pressing the "enter" button on the monitor. The pouch was then filled with air until full. If zeroing was successful, "CALIBRATION: O K " was displayed on the monitor, after which it was taken out of the bag. The zeroing process took approximately 2 minutes. Each pDR monitor was factory calibrated prior to the winter sampling period as a measure to ensure high quality data collection. Monitors are recommended to be factory calibrated once a year by the manufacturer for optimum monitor performance. The same four pDR monitors were used in the one winter and two summer sampling periods. Prior to each sampling period, 2 monitors were labeled as "Indoor" and 2 monitors were labeled as "Outdoor" to ensure that within each sampling period the same monitors were used for the indoor and outdoor set up of each home. 2.1.4.1.2.2 Pre-sampling pDR tests For wintertime sampling since the pDR monitors were operated in temperature conditions outside those specified by the manufacturer, it was necessary to test the pDRs prior to sampling in order to ensure that they operated reliably. The four pDR monitors used for all sampling were tested in three different scenarios. First, all four monitors were run side by side in the SOEH lab for approximately 2 hours. This test was performed to ensure that the readings from all four monitors were in agreement as well as to assess the degree of between-monitor variability. Next, all monitors were run side by side with two of the four monitors encased in heating pads. This test was carried out in order to determine if heating pads affected the measurements collected by the monitors. Again, monitors were run for approximately 2 hours. The third test was carried out to test the pDRs in a cold environment. Three monitors encased in heating pads were run in the lab freezer for approximately 1 hour. Due to the absence of any particulate source in the freezer, in order to take meaningful measurements, it was necessary to create a P M source. This was done by holding a burning candle in the lab freezer. Data from all tests 31 were downloaded and collocated measurements were compared. Relative differences in PM2.5 concentration between monitors as well as monitor variability between tests were investigated. 2.1.4.2 Outdoor set up Outdoor set up ensured that equipment was at least 1 m away from walls as well as ensured that minimum disruption to the residents occurred. Figure 2.4 below shows the outdoor set up of equipment in the winter sampling period. Figure 2-4 Set up of HI2.5 (top), pDR (beneath HI2.5) and HOBO (bottom right) monitors outdoors during winter sampling period 2.1.4.2.1 HOBO monitor set up One-minute averages of temperature and relative humidity were also collected outdoors using a HOBO monitor. The HOBO monitor test was initiated prior to visiting each home and the measurements from the appropriate time period were extracted after downloading test data. 32 2.1.4.2.2 Personal DataRAM monitor set up The outdoor monitor was set up directly outside in the backyard of each home. For winter sampling it was decided that the outdoor pDR needed an external heat source so each outdoor monitor and battery adapter were encased in a heating pad and placed in a camera bag. It was ensured that this set up did not obstruct airflow into the sensing chamber at the top of the pDR and therefore did not affect the measurements collected. The pDR was placed in a milk crate underneath a tripod and a tarp was placed around the tripod to protect the monitor from snow, wind, and rain. Again, it was ensured that the tarp did not obstruct air flow to the pDR monitor. Figure 2.5 below shows the outdoor winter set up of the pDR. Figure 2-5 Set up of pDR monitors outdoors during winter sampling period 2.1.4.2.3 HI monitor set up For both seasons, a HI 2 5 was used to collect integrated 48-60 hour filter samples of P M 2 5 at each home. Each H I 2 5 was loaded with a 37-mm Teflon filter (R2PJ037, Pall Corporation), a cellulose supporting backing pad (64747, Gelman Sciences), and two oiled impactor plates. After sampling at one home was completed, visible impacted particles were scraped from the impactor plates with a razor and the plates were used in the HI2.5 set up at the next home. After each pair of plates had been used twice (i.e. for 2 homes), they were placed in a bag labeled "used impactor plates" and were brought back to the SOEH lab for cleaning before they were used in any additional sampling. 33 Prior to sampling at each home, the air flow rate was set on each HI 2 5 . This was done by setting the pump (224-PCXR8, SKC, PA) flow rate to approximately 4.0 L/min using a BIOS DryCal frictionless piston pump calibrator (DC-2, BIOS International Corp., New Jersey). Ten flow measurements were taken and an average value was recorded. Flow adjustment on the pump was made by turning the screw on the front panel of the pump. Once the pump flow was set, a leak test was performed by connecting the pump to the HI2.5. A flow adapter was attached to the top of the nozzle of the H I 2 5 which was then connected to the calibrator. Flow was measured again and any flow measurements collected outside the range of 3.6-4.4 L/min (off by greater than 10% of the desired flow of 4.0 L/min) were indicative of a leak. If this occurred, the HI2.5 was disconnected from the pump and calibrator, unloaded, inspected for any visual damage, and loaded again. Once the leak test was successful, the HI 2 . 5 and pump set up was taken apart and the HI2.5 nozzle was covered with a cap. At the sampling site, the HI2.5 was attached to a tripod under which the pDR set up was placed. The air pump was attached to the H L s and flow was measured again. Flow was required to be within the range of 3.9-4.1 L/min in order to be considered acceptable. If the flow rate was outside of this range but still within the range of 3.6-4.4 L/min, flow was adjusted again by turning the small screw on the front panel of the pump. The flow adapter was removed, and an inlet and a rain shield were placed on top of the H I 2 5 . The "hold" button on the air pump was pressed and the pump was temporarily stopped until the outdoor pDR test was initiated. On the second day of sampling a flow check was performed. If flow was found to be outside the range of 4.0 ± 0.1 L/min, the flow was adjusted to a value closer to 4.0 L/min. If the flow was found to be outside the range of 3.6-4.4 L/min however, the sample was discarded. At the end of the sampling period, a post flow of the HI2.5 and pump set up as well as post flow of the pump was taken. If the pre and post flows were found to differ by more than 10% then these samples were considered invalid. During a practice test run in Prince George prior to sampling, it was found that the air pump could not operate in the cold temperatures. For field sampling, the HI2.5 and pump were connected with a long piece of tubing and the pump was placed in a room in the home through an open window. The window was left only a tiny bit open to ensure the 34 tubing was not kinked and air flow was not obstructed. In order to prevent the home from losing heat during sampling, a towel was stuffed into the opening of the window. 2.1.4.3 Summer sampling set up modifications In the summer, three modifications were made in the outdoor set up. First, the pDR monitor was not encased in a camera bag with a heating pad as an external heat source was not necessary in the summer season. Second, it was not necessary to have the air pump running inside the home since it could operate reliably under the environmental conditions. It was instead placed on a milk crate underneath the tripod with the pDR and HOBO monitor. Finally, the pDR was only protected with a tarp if rain was suspected during the sampling period. Figure 2.6 below shows the outdoor summer set up. Figure 2-6 Set up of sampling equipment outdoors during summer sampling period with HI (top), HOBO (bottom left), and pDR (bottom right) 35 2.2 Sample analysis 2.2.1 PM 2 .5 mass col lect ion A l l filters were weighed prior to and after sampling at the SOEH laboratory using the Sartorius MP3 microbalance (Sartorius, Goettingen, Germany). Filters were placed in a temperature and humidity controlled (35 ±5% RH, 22 ± 3 °C) balance room in the laboratory for a minimum of 48 hours to allow for equilibration to the environment after which point weighing was conducted. The same procedure was used for pre and post weighing of filters. For pre-weighing, a clean filter was placed over an alpha emitter for approximately 2 seconds to remove any static from the filter before its placement onto the balance. Each filter was measured in triplicate and an average mass value was taken. If any measurement was found to differ by more than 10 pg, the measurements were repeated. After the third measurement, the filter was placed in a clean labeled Petri dish. The temperature and humidity at the start and end of each weighing session were recorded to ensure that no changes had occurred in the balance room that would affect the measurements obtained. For post-weighing, a similar procedure was followed. The filters were first allowed to equilibrate in the balance room for a minimum of 48 hours and each filter was then weighed in triplicate. Temperature and humidity was again recorded at the start and end of each weighing session. After post-weighing, the filters were placed back in their Petri dishes for the next step in filter analysis. Pre and post weighing was conducted for filters at the beginning and end of the winter and 2005 summer sampling. No filter samples were collected for the 2004 summer sampling period. As a quality control measure, five lab blanks were weighed at the pre and post weighing sessions for the filters used in the 2005 summer sampling. Unfortunately, information on lab blanks for the winter sampling was not collected. 36 2.2.2 Absorbance measurements Reflectance measurements of each filter were taken using the M43D Smoke Stain Reflectometer (Diffusion Systems Ltd., London, UK). Reflectance operates under the principle that filters with more impacted combustion particles will appear "blacker" and therefore will reflect less light than filters with fewer or no impacted particles. A standard operating procedure to measure reflectance and to calculate an absorbance (ABS) coefficient for each filter was followed (see Appendix I). The relationship between absorbance, PM2.5 concentration as well as levoglucosan concentration were examined in order to investigate the levels of traffic related PM2.5 collected as well as to determine whether ABS could be used as an indicator of PM2.5 from vegetation smoke. 2.2.3 Levoglucosan analysis Finally, all sampled filters were analyzed for levoglucosan, a wood tracer. Levoglucosan analysis was the last step in the analysis of the filters due to its destructive process. A l l analysis was conducted in the SOEH laboratory according to the U B C School of Occupational and Environmental Hygiene, Determination of levoglucosan in atmospheric fine particulate matter by GC/MS SOP# A.00.10 (see Appendix J). This analysis method had a limit of detection (LOD) of 0.1 pg/filter. Briefly, analysis first involved the preparation of each filter. This included the cutting of each filter to remove the outer plastic ring using a size appropriate cutting tube. The filter was then transferred, using clean forceps, to a 5mL extraction tube. Ethyl acetate (2mL) was added to each tube and the resulting solution was ultrasonicated for 30 minutes. 100 uL of this solution was transferred into GC vials, and was mixed with lOuL of pyridine and 30uL of MSTFA +1% TMCS solution. The final solution was vortexed for 10-20 seconds and placed in a dark space for 6 hours in order for complete derivation to take place. A surrogate stock solution and an internal GC standard were also prepared. See Appendix K for GC/MS conditions used. 37 2.3 Data analysis 2.3.1 Cont inuous particle measurements Data clean up steps were performed prior to data analysis. This involved the removal of incomplete data and the removal of poor data. The first step involved the removal of any homes with incomplete data collection. Homes with complete sampling for at least one of the two days, with a minimum of 24 hours worth of data, were retained in the dataset. The second step involved the removal of data with poor agreement between the indoor and outdoor pDR monitors. This was done by examining the data from the pre, mid, and post indoor and outdoor monitor collocation data collected for each home. Relative differences and correlations between the indoor and outdoor monitors for each test period (i.e. pre-mid and mid-post) were calculated. Relative difference was calculated by first calculating the average difference between the indoor and outdoor monitor reading for each test period. This average difference was then divided by the mean indoor and outdoor concentration over the test period. Histograms of both relative difference and correlation were examined and outliers in the distributions for both of these variables were identified. Outliers were defined as values that were not continuous with the rest of the data (i.e. those values that occurred before or after gaps where no observations were seen between outlying values and the rest of the observations in the distribution). These values were considered to represent extreme values and were consequently used as the exclusion values where data with relative differences greater than and correlations less than these values were considered to have poor monitor agreement. These homes were highlighted and test concentrations from these individual homes were examined to determine whether any outliers were strongly influencing any high relative difference and correlation value. If outliers were found, they were removed and relative differences and correlations were calculated again for these homes. Any homes with remaining relative differences above or correlations below the outlying values were removed from analysis. 38 Finally, line graphs of all homes were plotted using the indoor and outdoor concentrations over the sampling period. Each graph was examined visually for the appearance of unusually high or low concentrations. For homes identified as "unusual" data, time activity logs were examined to determine the source of high indoor concentrations (e.g. smoking). 2.3.2 Infiltration calculat ions Infiltration was quantified for each home using the indoor and outdoor pDR PM2.5 concentrations collected over the sampling period. Two infiltration factors (F inf), one representing the intervention day (when the filter was running in the air cleaner) and one representing the non-intervention day (when no filter was running in the air cleaner), were calculated for each home. Each infiltration calculation involved two steps; the removal of indoor generated PM2.5 peaks followed by calculation of infiltration factors. A l l analysis was conducted using Microsoft E X C E L . 2.3.2.1 Censoring of indoor generated P M 2 . 5 A procedure similar to that developed by Allen et al., 2003 was used to identify and censor from the dataset the rising edge of indoor generated PM2.5 peaks. Censoring algorithms were used to separate indoor peaks into those resulting from indoor generated PM2.5 and those resulting from infiltration from outdoor PM2.5. Specifically, we identified peaks for 30 minute averages of every half hour concentration value collected. Any increase in the indoor concentration greater than 50% of the concentration from the previous 30 minute average, without a subsequent increase in the outdoor concentration was considered to be of indoor origin (equation 3). ^ i n ) t " ^ i n ) t 1 Indoor generated peak = > 50% ( C i n ) t (3) 39 Since outdoor concentrations are generally higher than indoor concentrations, an increase of 50% in these concentrations was considered a strict definition of an outdoor peak so outdoor peaks were identified when a minimum 10% increase from the previous 30 minute outdoor average was seen. Equation 4 was used to identify outdoor peaks. Additionally, indoor and outdoor concentrations were plotted in a simple line graph to allow for visual inspection of the data to ensure that necessary data was being removed. Figure 2.7 below illustrates the difference between indoor and outdoor generated peaks. The first graph, a, is an example of an indoor peak that would be identified and removed. As can be seen from figure 2.7a, a large indoor peak exists without a comparable increase in the outdoor concentrations. Alternatively, in graph b, an indoor peak can be seen along with a similar peak in outdoor concentrations indicating that this increase in indoor concentration is due to infiltrated P M 2 S f r o m outdoors. Outdoor peak = > 50% , 30%, 10% (C o u t )t (4) 40 a. Indoor peak due to indoor origin 1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 Time f min) b. Indoor peak of outdoor origin 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 Time (min) Figure 2-7 Comparison of indoor peaks of indoor (a) and outdoor (b) origin The rising edge of identified indoor peaks was removed (i.e. those rows in the dataset with increasing indoor concentrations). Figure 2.8 below shows the removal of the rising edge of an indoor generated peak. The decaying portions of indoor peaks represent important information regarding particle decay rates in the home and for this reason were included in the data. 41 • i 3 450 350 300 250 "S 200 g 150 8 1 0 0 50 in 4^— 1 3 5 7 9 1 1 13 15 17 19 21 23 25 2? 29 31 33 35 37 39 41 43 45 Time (min) ( " ) Data point removed Figure 2-8 Removal of PM data points for rising edge of identified indoor generated PM peaks 2.3.2.2 Infiltration calculations Multiple linear regression was used to estimate the penetration of outdoor particles indoors and the decay of indoor particles for the recursive mass balance model developed by Switzer & Ott, 1992. The LINEST function in E X C E L was used to estimate a value for ai, the penetration parameter and a2, the decay parameter in the recursive mass balance shown in equation 5 below. The Sjn term represents the indoor generation term which was removed in the previous step by censoring indoor generated peaks. C i n ~ a l ( C o u t ) t + a 2 ( C i n ) t - 1 + S i n ^ The LINEST function in E X C E L uses the least squares method to calculate a straight line that best fits the input data where known y values are the indoor concentrations at time t and x values are outdoor concentrations at time t and indoor concentrations at time t-1. The output provides coefficient values for the outdoor data, representing the penetration parameter ai, and for the indoor t-1 data, representing the decay parameter a2. Once the penetration and decay parameters were calculated, infiltration was quantified using equation 6 below. The detailed protocol followed for F i n f calculations can be seen in Appendix L. 42 A 95% confidence interval was also calculated for each Fjnf value using equation 7 below. x n x (7) Standard error = sfrT x 1.96 2.3.3 Air cleaner effectiveness Air cleaner effectiveness was calculated for each home in order to quantify the effect of running the air cleaner on PM2.5 infiltration. Equation 8 was used. 2.3.4 Sensit ivity analys is Although the pDR monitors have been designed for personal sampling of particulate exposure, they have been used for numerous outdoor air studies. Researchers have addressed two main concerns of using the pDR for environmental sampling; a sampling environment with high relative humidity resulting in false high pDR readings and the occurrence of baseline drift. The effects of both of these issues on the obtained results were addressed. We further assessed the results of our modeling by investigating the degree of agreement between F i n f values when the intercept of the regression equation was forced through zero and the original F i n f values calculated when an intercept was allowed. The purpose of this sensitivity analysis was to compare our results against those of another study (Allen et al., in press) in which Fjnr values calculated with and without F f (nofilter) - F . (filter) A C E = x100% F n f (no filter) (8) 43 an intercept term in the regression model showed good agreement. When these F i n f values where no intercept term was allowed, were regressed against F i n f values calculated using the suphur tracer method for the same homes, a strong relationship was found further validating the recursive model approach to calculating infiltration. In order to see if this same relationship between Fj nf with and without an intercept term existed in our data, these two sets of data were compared. 2.3.4.1 Relative humidity Studies have found that the pDR tends to give high readings when the relative humidity (RH) of the sampling environment is above 85%. In order to determine whether or not such data needed to be removed from our dataset, concentration data was examined in two ways. First in order to determine if higher concentration values were observed at R H values above 85%, a scatter plot of R H readings collected by the HOBO monitors placed outside each home against outdoor PM2.5 concentrations measured by the pDR was generated for each home. Only outdoor data was examined as all indoor environments had R H <85%. Next, Fjnf and A C E values were re-calculated for each home with R H >85% data removed. The mean, maximum, and minimum values were compared to those calculated from the whole data set (all R H data) to determine if the removal of this data resulted in significant differences in these values and therefore whether the removal of high R H data was warranted. 2.3.4.2 Baseline drift Data showing the presence of negative drift was identified by comparing manually calculated average concentrations with those recorded by the monitor for each home. Data where the two values were not in agreement represented data affected by negative drift. Discrepancies result because the pDR memory which records the time series measurements records only non-negative values and assigns anything detected as a negative concentration a value of "0" mg/m3. The internal memory which calculates the time weighted average concentration (TWAC) however does record negative 44 concentrations. When negative drift occurs, the T W A C recorded by the monitor will be lower than the T W A C calculated manually. Mean, minimum and maximum Fjnf and A C E values were calculated with and without these homes to determine if overall values were affected and whether removal of this data was warranted. 2.3.4.3 Forcing Intercept to zero In previous work, Fjnf values calculated using the recursive model show good agreement with the sulphur tracer method when the intercept of the linear regression equation was forced through zero (Allen et al., in press). The sulphur tracer method involves the measurement of indoor and outdoor sulphur concentrations. Since there are few indoor sources of sulphur this method assumes that all sulphur particles measured indoors represent those penetrating from outdoors. Allen et al., in press found that in general, there was good agreement between the Fjnf values calculated when an intercept was allowed and when the intercept was forced to zero. In cases where there was disagreement between these two values, setting the intercept to zero had the effect of decreasing the influence of individual outliers on the Fjnf value for the given dataset. The intercept term in the recursive model represents indoor generation of P M . Although indoor generated P M was removed using censoring algorithms, these algorithms only identified P M from sources that resulted in high P M peaks. Setting the intercept to zero would only be appropriate if all indoor generated P M was removed. Since P M that may be generated from constant low-level indoor sources was not accounted for, infiltration was calculated with the indoor generation term. In order to compare results to the relationship found by Allen et al., in press, Fjnf values were calculated with the intercept forced to zero and were compared to the original Fi„f values calculated. Data was tested for significant differences and summary statistics for both datasets were compared. 2.3.5 Multivariable modeling Information from the housing questionnaire was used to develop both an infiltration prediction model and an air cleaner effectiveness prediction model. Certain housing characteristics can be important in explaining both infiltration rate of P M into a home and 45 the effectiveness of an air cleaner in that home. Some of these parameters were included in a forward step-wise multiple regression model developed to predict both of these phenomena. 2.3.5.1 Housing characteristics Data collected from each housing questionnaire (see Appendix E) completed was compiled and summary statistics were performed on the dataset. 2.3.5.2 Air exchange measurements Carbon dioxide ( C O 2 ) concentration readings collected from the Q-Trak monitors were graphed for each home. C O 2 decay rates during times when homes were unoccupied (as recorded on time activity logs by volunteers) were used as indicators of air exchange. These values were added in the infiltration and air cleaner efficiency modeling as the variable "air exchange rate (AER)." Equation 2 below was used to quantify air exchange for each home. Air exchange rate [ In C ( t 2 ) - I n C f t J ] (AER) = — Where: Air exchange rate (AER) = number of air exchanges per hour (h"1) In C = natural log concentration (ug/m3) ti = time at start of measurement period (hours) t2 = time at end of measurement period (hours) (2) 2.3.5.3 Infiltration and air cleaner effectiveness modeling The following characteristics were considered potentially relevant in affecting infiltration and air cleaner effectiveness (ACE): age of home, square footage of home, number of windows, percentage of carpeting in home, type of stove, use of range hood, use of air conditioning, type of heating system, use of fireplaces, and use of windows. Analysis 46 was performed separately on summer and winter data and due to the small sample size, was also combined with the creation of a "season" variable. Although differences in infiltration will exist between winter and summer, they will also differ due to differences in other variables including window use, use of air conditioning, and fireplace use. These variables were all included in the initial stages of modeling if they met the inclusion criteria. A l l analysis was conducted using Stata statistical software (Stata 9, StataCorp LP, Texas, US). The first inclusion criterion was a p-value with a maximum value of 0.25 for simple regressions between infiltration (and air cleaner efficiency) and each independent variable. Any variables regressed with infiltration (and air cleaner effectiveness) having a p-value >0.25 were therefore not considered for inclusion into the model. Associations between the remaining independent variables were then investigated. In order to test association among continuous variables cross-tabulations were performed and any pair with a p-value <0.05 was considered highly correlated. Correlations between continuous variables were examined and any pair of variables showing a Spearman correlation coefficient of >0.50 were considered highly correlated. Finally, simple regressions, using the continuous variable as the dependent, and the categorical variable as the independent variable, were performed for each pair of continuous and categorical variable. Any pair of variables exhibiting a model p-value <0.05 were considered highly correlated. If any variable was found to be correlated with two or more variables, that single variable was excluded. For any variables that were only associated with one other variable, the variable considered to be more important in predicting infiltration, as determined by the p-value and R from the simple regressions with infiltration, was included. Next the remaining variables were ordered by increasing model p-values from the initial simple regressions with infiltration (and air cleaner effectiveness) and included into the model in this order. With the addition of each new variable, the p-value and adjusted R 2 were examined. If the addition of the new variable was found to increase the adjusted R 2 47 and produce only a small change in the model p-value, then this addition was considered to be useful. Individual p-values were also examined to determine if each variable was significant at the 0.25 level. If the addition of a new variable made any other variable non-significant (i.e. increased the individual p-value to >0.25), the non-significant variable was then removed. 2.3.6 Spatial variability The relationship between monitoring network PM 2 . 5 concentrations and sampled backyard home PM2.5 concentrations for the communities of Williams Lake, Prince George, Surrey, Lytton, and Spence's Bridge was examined in order to determine whether the existing B C Ministry of Environment monitoring networks provide an appropriate description of community PM2.5 exposure. Hourly averages for 48-hour sampling periods for each home were calculated and matched with hourly PM2.5 concentrations recorded by Ministry of Environment monitors. Correlations between the two data sources were calculated for each home. In order to investigate whether the monitoring networks provide reliable information on exposure levels during high P M 2 5 episodes that are of greatest concern, a scatter plot was generated for 24 hour averaged data for the communities of Williams Lake, Prince George, and Surrey. Due to the large distance (75-120 km) between Ministry of Environment monitors and the communities of Lytton and Spence's Bridge, these data were not included in the analysis. For winter sampling conducted in Prince George, home measurements were compared with P M measurements collected by two monitors located within the city. Prince George has six air quality monitors, only two of which measure PM2.5. For summer sampling collected in Lytton and Spence's Bridge, correlations were examined for two monitors located in Kamloops and Whistler which are the closest geographical locations with A Q monitors. One monitor in Whistler is approximately 100 km and 120 away from Lytton and Spence's Bridge respectively, while the two monitors in Kamloops are located approximately 100km and 75 km from Lytton and Spence's Bridge respectively. In Williams Lake, samples were compared to three monitors measuring PM2.5 concentrations in the city. Finally, maps showing the location of sampled homes and A Q 48 monitors were generated using ArcGIS 9 (ESRI, California) in order to examine the distances between sampled homes and ministry A Q monitors. 49 3 RESULTS 3.1 Data collection 3.1.1 Summary of col lected samples Field sampling was conducted in winter and summer seasons. A l l winter sampling was conducted in Prince George, BC from January 7 to March 4, 2005. Sampling for the summer season was conducted in the summers of 2004 and 2005. In 2004, samples were collected in Kamloops, BC from August 3 to August 9 and in Williams Lake from August 15 to 25. In 2005, sampling was conducted in Lytton from August 4 to 19, in Spence's Bridge from August 16 to 23, and in Surrey from September 13-15. Sampling in Prince George occurred during a time of expected high occurrence of residential wood burning due to cold winter temperatures. In Kamloops and Williams Lake, small forest fires burning in surrounding areas led to days with poor air quality as indicated by "poor" ratings on the Ministry of Environment Air Quality website (Ministry of Water, Land, and Air Protection, 2006) as well as visible smoke in the city during the sampling period. In 2005, the majority of sampling was conducted in the two communities of Lytton and Spence's Bridge. A large forest fire along Sleetsis Creek, south of Spence's Bridge and north of Lytton caused poor air quality and visibility in both communities during the fire event. The severity of the smoke depended on many factors including wind direction, weather, and intensity of the fire. The fire lasted from July 29 to August 9, 2005 burning a total of 5,560 hectares (Ministry of Forests and Range, 2006). A map of the fire can be seen below in figure 3.1. The final sample was collected in the summer of 2005 in Surrey during a fire at Burns Bog, a raised peat bog located in Delta, BC. The fire was active between September 11 and 22, 2005 and burned approximately 200 hectares (Ministry of Forests and Range, 2006) as well as caused visible smoke in surrounding areas, including the city of Surrey. 50 Figure 3-1 Map of Sleetsis Creek fire (active from July 29 to August 9, 2005) in relation to the communities of Spence's Bridge and Lytton (Ministry of Forests and Range, 2006) Table 3.1 below summarizes the number of homes sampled in each city/town, the sampling period, as well as the primary smoke source for which sampling was conducted. For the summer period, 7 samples were collected in 2004 with 2 homes sampled in the city of Kamloops and 5 homes sampled in Williams Lake. In the summer of 2005, 7 homes were sampled in Lytton, 2 homes were sampled in Spence's Bridge, and one home was sampled in Surrey. In the winter, 21 homes were sampled in Prince George. Samples for 38 homes were collected over the entire sampling period. Table 3.1 Summary of location, date and primary smoke source of homes sampled Season Location Year Homes Primary Smoke Source Summer Kamloops 2004 2 Small fires burning in surrounding areas Williams Lake 2004 5 Small fires burning in surrounding areas Lytton 2005 7 Sleetsis creek fire (5,560 hectares)burning between Lytton and Spence's Bridge Spence's Bridge 2005 2 Sleetsis creek fire (5,560 hectares)) burning between Lytton and Spence's Bridge Surrey 2005 1 Burn's Bog fire (200 hectares) Winter Prince George 2004 21 Wintertime residential wood-burning Total 38 51 3.1.2 Volunteer recruitment In Prince George, 19 of the 26 people initially contacted about the study were enrolled for the study period. A l l contacts' information was received either from the initial presentation conducted at the hospital COPD support group or from friends and family members of enrolled volunteers. A variety of methods were used to recruit volunteers in the summer sampling periods. One volunteer in Kamloops was enrolled through contacts at the Kamloops Better Breathers Club, a COPD support group run throughout British Columbia by the B C Lung Association. The second volunteer was contacted as a result of sending out a mass email to friends, family and school contacts. In Williams Lake the first volunteer's contact information was acquired through the respiratory nurse at the local hospital. The information provided by the nurse was followed up on and the volunteer was enrolled. A l l subsequent volunteers in the area were acquaintances of this first volunteer. In Lytton, three townspeople were approached and told about the study on the first day of arriving into town. A l l three people were interested in volunteering and upon request passed on names of other people in town as potential participants. This information was followed up on and volunteers were lined up for the sampling period. In Spence's Bridge, the owners of a local inn were approached who gave the contact name of a potential volunteer who once contacted was willing to participate. The second volunteer in the area as well as the volunteer in Surrey were contacted through family, friends, and school contacts. In general, the initial aim to recruit COPD/asthma patients as volunteers was not successful for the summer sampling period. A l l winter sampling consisted of asthma and COPD patients, while the majority of summer volunteers were healthy, and consequently, participant type was confounded by season. Due to the recruitment methods used, this sample of study subjects does not represent a random but rather a convenience sample. . Consequently, the results obtained may not be representative of the general population. 52 3.2 Gravimetric analysis 3.2.1 Data quality Fifteen field blanks were collected for the winter sampling season and three field blanks were collected for the 2005 summer sampling season. Table 3.2 below presents P M 2 5 concentration and absorbance (ABS) coefficient summary statistics for these filters. As can be seen from the table, a slightly higher mean value of 0.9 pg/m3 was found for the winter field blanks compared to in the summer where the mean value for field blanks was 0.7 pg/m3. ABS coefficients were the same for both winter and summer sampling periods with a slightly wider range of coefficients calculated for winter field blanks. Table 3.2 Summary of PM2.5 concentration and ABS coefficients calculated for field blanks collected during winter and 2005 summer sampling periods P M 2 5 concentration (ucj/m3) ABS coetficient (10"b m"1) Season (n) Mean (SD) Min Max Mean (SD) Min Max Winter (15) 0.9 (0.56) 0.0 1.9 0.02 (0.01) 0.00 0.05 Summer (3) 0.7 (0.41) 0.3 1.1 0.02 (0.01) 0.01 0.03 For levoglucosan analysis, all field blanks were below the limit of detection (LOD) of 0.10 pg/filter which corresponds to a concentration of 0.01 pg/m 3 for both sampling seasons (calculated as 0.10 pg/mean sampling volume for each season). An LOD of 0.1 pg/m 3 was calculated for PM2.5 (calculated as 3*SD of mass of blanks/mean sampling volume) for both winter and summer filter samples and an LOD of 0.30 x 10"5 m"1 was calculated for ABS coefficient measurements (calculated as 3*SD of ABS coefficient measurements). A l l measurements for lab blanks were within ± 4 pg of one another with a maximum change of 0.01% between pre and post weighing sessions for these filters. 3.2.2 Col located pDR and H l 2 5 measurements The relationship between collocated pDR and HI2.5 data collected from each home was examined. As both instruments are indicators of PM2.5 exposure, it was necessary to ensure that readings obtained from the pDR and HI2.5 showed reasonable agreement in 53 order to increase validity of our results. Separate analysis was performed on winter and summer analysis as well as on pooled seasonal data. Eighteen collocated samples were collected in the winter while ten collocated samples were collected in the summer. Six winter samples were removed due the absence of collocated pDR measurements and one summer sample was removed due to invalid pDR measurements where the majority of the monitor readings were "0 pg/m 3" as a result of the monitor being rained on. Figure 3.2 below shows a scatter plot of collocated pDR and HI2.5 PM2.5 concentration measurements collected in winter and summer. As can be seen from the graphs, a strong relationship was found between both of these instruments with a slightly stronger relationship in winter (R2=0.85) versus summer (R2=0.78). Similar slopes were found for both summer and winter seasons which further indicate strong agreement between these monitors. For all measurements, pDR concentrations were found to overestimate the HI2.5 concentrations. For infiltration calculations as well as spatial variation analysis, pDR concentration data was used. For the latter analysis, all pDR concentrations were corrected using the the HI/pDR regression equations found from the relationships seen in figure 3.2. For infiltration calculations, no corrections for concentration values were made as the relative differences in indoor and outdoor concentrations were relevant as opposed to absolute concentrations. 54 Winter ~ 10.00 = o.oo H 1 1 1 1 1 1 1 0.00 10.00 20.00 30.00 40.00 50.00 60.00 70.00 pDR concentration (ug/m3) S u m m e r « 4.00 8 2.00 E 0.00 0.00 5.00 10.00 15.00 20.00 25.00 30.00 35.00 40.00 pDR concentration (ug/m3) Figure 3-2 Relationship between collocated pDR and HI 2. 5 concentrations measured in homes in winter (n=18) and summer (n=10) sampling periods 3.2.3 Absorbance and levoglucosan measurements Filter absorbance is often used as an indicator of traffic related P M exposure. In this work, absorbance (ABS) was measured in order to determine whether it could be used as a reliable indicator of P M from combustion of wood resulting from forest fires and residential wood burning as well as to determine the degree of PM2.5 from traffic sources in collected samples. Calculated ABS coefficients were compared to levoglucosan concentrations recovered from each filter in order to examine the degree of agreement between these two variables. Absorbance coefficients and levoglucosan concentrations are presented below in table 3.3 for 26 winter and 10 summer filter samples. As can be seen from the table, a higher 55 mean ABS coefficient of 1.16 ± 0.64 x 10'5 m~' was found in the winter versus a mean ABS of 0.51 ± 0.33 x 10 5 m"1 in the summer. A mean levoglucosan concentration of 0.4 ± 0.4 Ug/nr was calculated for winter and a lower mean concentration of 0.1 ± 0.1 pg/m was recovered from summer samples. Table 3.3 Summary of absorbance (ABS) coefficients and levoglucosan concentrations for filter samples collected in winter and summer sampling periods ABS coefficient (x 10"b m"1) Levoglucosan concentration (pg/m3) Season (n) Mean (SD) Range Mean (SD) Range Winter (26) 1.16 (0.64) 0.36-2.60 0.4 (0.4) <LOD*-1.3 Summer (10) 0.51(0.33) w „„-5 „-1 <LODn-1.16 0.1 (0.1) <LOD J-0.4 1 sample below LOD of 0.30 X 10"b m 2 4 samples below LOD of 0.1 ug/m3 3 2 samples below LOD of 0.1|ig/m3 Analysis of correlation between ABS coefficients and levoglucosan concentration data revealed Spearman correlation coefficient values of 0.59 and 0.63 for winter and summer respectively. The relationship between levoglucosan and ABS was also examined by regression of the two variables. As can be seen from figure 3.3, the relationship between levoglucosan and ABS was similar in both seasons with R 2 values of 0.35 and 0.40 for winter and summer respectively. Similar slopes of 0.33 and 0.25 were found for winter and summer regression equations respectively which further indicates a similar relationship between levoglucosan and absorbance for both seasons. These results indicate that a relatively weak relationship exists between these two variables. 56 Winter 1 " r a • 1.400 1.200 1.000 0.800 0.600 0.400 0.200 0.000 y = 0.33x + 0.09 R2 = 0.35 s2- i — • i i i 0.00 0.50 1.00 1.50 2.00 ABS coefficient (10A-5 nV-l) 2.50 3.00 0.500 | 0.400 ra 3, 0.300 0.200 0.100 S u m m e r ra o ^ 0.000 -0.100°: y = 0.25x -0.03 • R2=0.40 .A. A. H 0.20 0.40 0.60 0.80 1.00 1.20 1.. ABS coefficent(10A-5 m*-!) Figure 3-3 Relationship between ABS and concentration of levoglucosan recovered from winter (n=26) and summer (n=10) filter samples PM2.5 concentration, for both pDR and HI2.5 data, and ABS coefficient values were regressed in order to determine the strength of the relationship between these two variables. A slightly weaker relationship was found between ABS and H I 2 5 concentration for filter samples collected in the summer (R2=0.53) compared to winter 2 2 (R=0.77). Winter and summer data was pooled and an R value of 0.73 was calculated. Similar slopes of 0.05 and 0.06 were found for winter and summer respectively indicating a similar relationship between PM2.5 and ABS concentration in both seasons. Spearman correlation coefficients of 0.84 and 0.36 for winter and summer respectively for HI2.5 and ABS concentrations and 0.71 and 0.40 for winter and summer respectively for pDR PM2.5 and ABS concentrations were found. Lower R 2 values of 0.52, 0.15, and 0.52 for winter, summer, and pooled data respectively for ABS and pDR concentration data were found. 57 Similarly, PM2.5 concentration, for both pDR and HI2.5 data, and levoglucosan concentration values were regressed in order to determine the relationship between these two variables. R values of 0.42 and 0.66 for winter and summer data respectively for levoglucosan and pDR concentration data were found (see figure 3.4). Similar slopes of 0.03 and 0.02 were found for winter and summer respectively indicating a similar relationship between PM2.5 and levoglucosan concentration in both seasons. Spearman correlation coefficients of 0.65 and 0.81 for winter and summer respectively for HI 2 . 5 and levoglucosan concentrations and 0.75 and 0.76 for winter and summer respectively for pDR PM2.5 and levoglucosan concentrations further indicate a similar relationship between PM2.5 concentration and levoglucosan in winter and summer. Winter CD c o 3 1.400 1.200 1.000 0.800 0.600 0.400 0.200 0.000 y = 0.03x + 0.06 R2= 0.42 • v^ -^— I 1 » 1 1 0.00 5.00 1 0.00 1 5.00 20.00 25.00 30.00 35.00 40.00 HI PM2.5 concentration (ua'm3) S u m m e r 0.250 -i 1 HI PM2.5 concentration (ug.'m3) Figure 3-4 Relationship between HI PM2.s concentration and concentration of levoglucosan recovered from winter and summer filter samples 58 A weak relationship between ABS and levoglucosan indicates that ABS is not a reliable indicator of PM2.5 from forest fire/wood burning emissions. In contrast, similar relationships between levoglucosan and PM2.5 concentration from both monitors in both summer and winter seasons indicates that levoglucosan is an important indicator of forest fire and residential wood smoke emissions. Additionally, the stronger relationship, as indicated by R 2 and correlation values, between PM 2 . 5 and levoglucosan concentrations versus PM2.5 concentration and ABS coefficients suggest that particulate matter from forest fire/wood smoke emissions constituted a larger portion of total PM 2 .5 compared with particulate matter from traffic-related emissions for collected samples. 3.3 Direct reading PM25 data 3.3.1 Pre-sampling pDR tests In order to ensure that the pDR monitors would operate reliably in both the cold conditions during the winter period as well as that the pDR measurements were not affected by the use of the external heat source, it was necessary to perform tests prior to winter sampling. Table 3.4 below summarizes the relative difference (average difference between monitor readings/average concentration) of PM2.5 concentration for each test. As can be seen from the table, the relative difference between mean PM2.5 concentration values were in good agreement among the monitors for tests 1, 2 and 3. Paired t-tests were performed between heated and non-heated monitors for test 2. No significant differences existed among the monitors (p >0.10) indicating that the use of heaters did not affect monitor performance. 59 Table 3.4 Summary of relative difference (average difference between monitor readings/average concentration) in concentrations by pDR monitor and test Test Test description Monitors Heater Relative difference in P M 2 5 concentration 1 All monitors placed side by side on laboratory bench 1 and 2 Neither 0.46 1 and 3 Neither 0.35 1 and 4 Neither 0.38 2 and 3 Neither 0.46 2 and 4 Neither 0.55 3 and 4 Neither 0.32 2 2 heated and 2 unheated monitors placed side by side on laboratory bench 1 and 2 Both 0.16 1 and 3 One 0.13 1 and 4 One 0.22 2 and 3 One 0.17 2 and 4 One 0.18 3 and 4 Neither 0.20 3 3 heated monitors placed side by side in laboratory freezer 1 and 2 Both 0.19 1 and 3 Both 0.14 2 and 3 Both 0.18 In order to test variability among the monitor readings between tests, standard error of each test was calculated. Standard error (SE) values of 1.93, 3.31, and 5.10 were calculated for tests 1, 2, and 3 respectively. Although the variability, as seen with SE, did increase with test modifications (i.e. heating monitors and placing monitors in cold temperatures), it did not increase by a large degree. These results indicate that the use of heaters is appropriate in cold sampling environments as their use did not significantly affect pDR performance. 3.3.2 Data clean up 3.3.2.1 Removal of incomplete pDR data For field sampling, any homes with greater than 50% of data missing from a 24 hour period was removed from analysis. Home KM-01 had incomplete sampling both days and was removed entirely from analysis. Two homes, PG-01, and PG-09 had incomplete 60 sampling for one of the two sampling days and both of these incomplete sampling days were removed from the dataset. 3.3.2.2 Monitor agreement Correlations between indoor and outdoor monitors for test periods of "pre-mid" and "mid-post" were calculated for each home to examine pDR monitor agreement before and after every 24 hour sampling period. Table 3.5 below presents summary statistics of the Spearman correlation coefficients for each test period. As can be seen from the table, similar mean correlations of 0.79 ± 0.20 and 0.75 ± 0.30 were found for both test periods of pre-mid and mid-post respectively. Table 3.5 Summary statistics of correlations between indoor and outdoor pDR monitors for pre-mid and mid-post collocation test periods Test Mean (SD) Minimum Maximum Pre-mid 0.79 (0.20) 0.11 0.99 Mid-post 0.75 (0.30) -0.34 0.99 Figure 3.5 shows the distribution of correlations from these tests. As can be seen from the plot, the test correlations follow a left skewed distribution with extreme values representing those values less than 0.40. From this plot, any correlation <0.40 was determined to be an outlier indicative of poor agreement between the indoor and outdoor pDR monitors. Relative differences in concentrations between monitors (average difference between monitor readings/average concentration) were examined in the next step and homes with both low correlations and high relative differences were removed from analysis. Homes KM-02 and LY-01 were identified as having low poor monitor agreement for both pre-mid and mid-post test periods representing poor agreement for both sampling days. Home LY-01 had correlations of 0.18 and -0.33 for pre-mid and mid-post test periods respectively while home KM-02 had correlations of 0.11 and -0.34 for pre and post test periods respectively. 61 correlations Figure 3-5 Distribution of correlations between indoor and outdoor pDR monitors for all collocation test periods Relative differences between the indoor and outdoor pDR monitors for pre-mid and mid-post test periods were also calculated. Table 3.6 presents summary statistics of relative difference values for all homes by test period. As can be seen from the table, a mean relative difference 0.4 ±0 .3 was calculated for both pre-mid and mid-post test periods. Table 3.6 Summary statistics of relative PM 2 . 5 concentration difference values between indoor and outdoor pDR monitors for pre-mid and mid-post collocation test periods Test Mean(SD) Minimum Maximum Pre-mid 0.4 (0.3) 0.0 1.5 Mid-post 0.4 (0.3) 0.1 1.8 Figure 3.6 shows the distribution of relative differences for these tests. As can be seen from the plot, the relative difference values follow a right skewed distribution with extreme values representing those values greater than 1.0. Therefore, any test showing a relative difference greater than 1.0 was determined to be indicative of low agreement 62 between the indoor and outdoor pDR monitors. Home LY-01 was identified as having poor monitor agreement for the pre-mid test period representing the first day of sampling while home KM-02 was identified as having poor monitor agreement between both pre-mid and mid-post test periods representing both days of sampling. Home LY-01 had a relative difference value of 1.2 for pre-mid test period, while home KM-02 had relative difference values of 1.5 and 1.8 for pre-mid and mid-post test periods respectively. Relative difference (ug/m3) Figure 3-6 Distribution of relative differences (average difference between indoor and outdoor monitor readings/average concentration measured during test period) between indoor and outdoor pDR monitors for all collocation test periods Scatter plots of the indoor and outdoor concentrations during the pre-mid and mid-post test periods were plotted to identify any outliers that could have been driving the low correlations and high relative differences between the indoor and outdoor pDR monitors for homes KM-02 and LY-01. Due to the absence of any outliers driving both low concentrations and high relative differences between the indoor and outdoor pDR monitor, homes LY-01 and KM-02 were removed from subsequent data analysis. 63 3.3.2.3 Monitor concentrations Line graphs of indoor and outdoor concentrations for day 1 and day 2 of the sampling period for each home were plotted. Any homes showing unusually low concentrations or consistently higher indoor versus outdoor concentrations were identified. Low concentrations were possibly indicative of monitor problems and were defined as homes where a minimum of 50% of the data readings were "0" ug/m3. Homes with a very high number of indoor peaks were removed as the presence of these peaks did not leave enough data for infiltration calculations. Any homes where the presence of indoor peaks warranted the removal of greater than 50% of the data were removed from analysis. These graphs can be seen in Appendix M . Low concentrations were found for day 1 and day 2 indoor concentrations for homes WL-02 and LY-03, for day 1 and 2 indoor concentrations for WL-04, for day 1 indoor concentrations for homes LY-05, LY-06, and PG-21, for day 2 outdoor concentrations for LY-01, and finally for day for day 2 outdoor concentrations for home PG-11. Concentration percentiles for these identified homes were calculated and are presented below in table 3.7. As can be seen from the table homes WL-02 and PG-11 have greater than 50% of concentration value readings of 0 Ug/m . This data was removed from analysis. Table 3.7 Concentration percentiles of homes identified as having low indoor or outdoor PM2.s concentrations PM 2 . 5 concentration percentile (un/m3) Home Day In/out 5th 50th 75th 99th WL-02 1 In 0 0 0 3 WL-02 2 In 0 0 0 45 WL-04 1 In 0 3 5 9 WL-04 2 In 0 10 16 56 LY-03 1 In 3 7 11 20 LY-03 2 In 1 13 21 69 LY-04 2 Out 0 4 61 129 LY05 1 In 0 8 11 34 LY-06 1 In 0 1 4 29 PG-05 1 In 11 20 25 5 PG-11 2 Out 0 0 1 30 PG-21 1 In 4 6 9 22 64 Line graphs of homes SB-02, PG-07 and PG-17 revealed a high frequency of indoor PM2.5 peaks. Time activity data showed that a nebulizer, used to deliver aerolized medication, was operated in homes PG-07 and PG-17, which was the cause of the high indoor PM2.5 concentrations. The nebulizer was used a total of 12 and 6 times in homes PG-07 and PG-17 respectively. These homes were removed from analysis since the majority of the data comprised of large peaks. Analysis of the time activity log for SB-02 revealed smoking occurring within the home for day 1. As the presence of smoking was one of the original exclusion criteria, this day of sampling was removed from analysis. In total, 5 homes (KM-01, KM-02, LY-01, PG-07, PG-17) were completely removed from analysis, while one day of sampling was removed from 4 homes (LY-04, SB-02, PG-01, and PG-09). Table 3.8 summarizes the removal of these homes. Table 3.8 Summary of homes excluded from data analysis Home Data removed Reason for removal KM-01 Day 1 and 2 Incomplete sampling for both days KM-02 Day 1 and 2 Poor agreement between indoor and outdoor pDR monitors (i.e. low correlations and high relative differences between monitors) WL-02 Day 1 and 2 > 5 0 % "0" pg/m 3 outdoor readings LY-01 Day 1 and 2 Poor agreement between indoor and outdoor pDR monitors (i.e. low correlations and high relative differences between monitors) LY-04 Day 1 outdoor > 5 0 % "0" ug/m3 outdoor readings SB-02 Day 1 Smoking in home PG-07 Day 1 and 2 High frequency of indoor peaks PG-09 Day 1 Incomplete outdoor sampling for day 1 PG-11 Day 2 outdoor > 5 0 % "0" pg/m3 outdoor readings PG-17 Day 1 and 2 High frequency of indoor peaks 3.4 Infiltration calculations The effects of relative humidity on outdoor pDR concentrations measured were first examined. Scatter plots were generated and can be seen in Appendix N . The trend of higher concentration readings for R H >85% was only seen for home WL-05 which can be seen below in Figure 3.7. One-minute averaged PM2.5 concentrations collected when R H was >85% were removed from the data and all calculations based on this home were conducted only for R H <85% data. In total, 51% of the data was removed (1441 out of 65 2837 total 1-min averaged data points). Further analysis on the effects of R H on pDR function was conducted and is discussed in the Sensitivity Analysis section. No other corrections were made for R H for the main analysis. WL-05 70 , Figure 3-7 Relationship between relative humidity (RH) and outdoor pDR PM2.s concentration measurements for home WL-05 3.4.1 Removal of indoor generated peaks In total, indoor peaks were identified for 11 out of 26 filter days and 10 out of 28 no filter days. Averages of 49 minutes and 54 minutes worth of data were removed from filter and no filter days respectively for summer and winter homes representing 1.5% and 1.7% respectively of total data for each home. Overall mean indoor and outdoor filter and no filter day concentrations were not affected greatly. On average, all mean concentrations for 24 hour and 30 minute averaged data were within ± 3 Ug/m3 of the original data. 3.4.2 Calculat ion of infiltration factors Table 3.9 summarizes the infiltration factors (Fjnf) calculated for homes in winter, summer and pooled sampling periods for filter and no filter days. As can be seen from the table, the highest mean Fj„f of 0.61 ± 0.27 was calculated in the summer for no filter 66 days. In both seasons as well as for pooled seasonal data, no filter days have lowered mean infiltration factors. Table 3.9 Summary of infiltration factors (Fi„r) calculated for homes sampled in winter and summer Infiltration factor (Finf) Season Day AM (SD) GM (GSD) Range Summer Filter (n=10) 0.19 (0.20) 0.10 (3.82) 0.01-0.61 No filter (n=13) 0.61 (0.27) 0.55 (1.55) 0.30-1.10 Winter Filter (n=19) 0.10 (0.08) 0.07 (2.59) 0.01-0.30 No filter (n=16) 0.28 (0.18) 0.23 (1.89) 0.10-0.68 Both Filter (n=29) 0.13 (0.14) 0.08 (2.98) 0.01-0.61 No filter (n=29) 0.42 (0.27) 0.23 (2.02) 0.10-1.1 Figure 3.8 shows the distribution of F i n f values calculated for winter and summer sampling seasons over filter and no filter days. As can be seen from the graph, lower infiltration occurred for filter days for both winter and summer. Infiltration calculations include the penetration and decay of particles. While the penetration of particles is not expected to alter with air cleaner use, the decay rate is expected to increase with air cleaner use. The air cleaner draws air into the filter and expels filtered air. Consequently, particles remain in the air for a shorter period of time, increasing decay rate and decreasing infiltration when air cleaners are operated. When comparing between winter and summer, higher infiltration was measured for the summer season with values exceeding 1 for no filter days. With the removal of all particles of indoor origin, Fjnf values greater than one represent greater than 100% outdoor particle penetration. As this is not physically possible, Fjnf values greater than 1 indicate that indoor generated particles were not fully removed. 67 Summer Winter 1 1 — FUter No filter Figure 3-8 Distribution of F i n f values calculated during summer and winter sampling seasons for filter and no filter days Figure 3.9 shows the winter F;nf values for homes for which data for both filter and no filter days was available. As can be seen from the graph, a range of infiltration values was calculated for both filter and no filter days. Infiltration was lowered for most homes on days when the air cleaner was run with the filter. For homes PG-04 and PG-06, similar infiltration values were calculated for both filter and no filter days. Individual Fjnf values by home can be seen in Appendix O. 68 fr <f> V> \V" C? \ * & & cp & N ^ c£ \ N Home Figure 3-9 Infiltration values (Finf) for filter and no filter days for homes sampled in the winter sampling period Figure 3.10 below shows the summer Fj„f values for homes for which data for both filter and no filter days was available. As can be seen from the graph, a range of infiltration values was calculated for both filter and no filter days. For all homes, excluding WL-03, infiltration was lowered for days when the air cleaner was run with the filter. 69 1.60 1.40 S T 1.20 c it, LOO .2 0.80 £ 0.60 I 0.40 0.20 0.00 -r -r • NO FILTER • FILTER 1 II y I  t i l III II t LY-03 LY-05 LY-02 LY-07 WL-05 LY-06 WL-03 WL-04 Home Figure 3-10 Infiltration values (Finf) for filter and no filter days for homes sampled in the summer sampling period A group mean t -test showed a significant difference for Fj„f between seasons with significantly higher infiltration in summer versus winter for both filter and no filter days (p <0.05). Two sample mean t-tests showed that Fj„f between filter and no filter days was significantly different for summer and winter (p <0.03). F i n f values were also significantly lower for filter compared to no filter days (p <0.05). 3.5 Air cleaner efficiency Air cleaner efficiency (ACE) was calculated for each home using the Fj„f values calculated for filter and no filter days. Table 3.10 below presents summary statistics of calculated A C E values. A higher mean A C E of 65% was found in the summer versus winter with a mean A C E of 55% although no significant difference between seasons existed (p >0.10). In both seasons, negative efficiencies were calculated for some homes, with an A C E value of -3% for one home (WL-03) in the summer and -25% and -18% for two homes (PG-04 and PG-06) in winter. Maximum A C E values of 99% were found in both seasons. Individual A C E values by home are presented in Appendix P. 70 Table 3.10 Summary of Air Cleaner Efficiency for summer and winter Air cleaner efficiency (%) Season (n) Mean (SD) Min Max Summer (8) 64.5 (35.0) -3.2 98.9 Winter (16) 54.5 (37.6) -25.0 98.5 Both (24) 57.7 (36.3) -25.0 98.9 3.6 Sensitivity analysis Due to the issues of high relative humidity and baseline drift possibly influencing pDR monitor performance, sensitivity analysis on data were performed in order to further investigate these two phenomena. Finally, results of Fj„f calculations, when an intercept was allowed, were compared to those calculated when the intercept was forced through zero. 3.6.1 Relative humidity An increase in water content in the air can lead to altering of the light scattering properties of particles due to hygroscopic growth (Poschl, 2005). In order to determine if overestimation of P M 2 5 concentrations at high R H (>85%) occurred for the data collected for this work, separate analysis was performed on the data. The main analysis for infiltration (presented in section 3.4) and A C E (presented in section 3.5) calculations, was conducted using all data (i.e. data collected at all R H values) as initial analysis of the relationship between R H and P M 2 5 concentration showed no such trend of high concentrations readings above 85% (see R H and outdoor pDR P M 2 s concentration scatter plots in Appendix N and figure 3.7). For summer samples, a total of 16 days out of the total 31 days sampling days had at least one 1-min average data point collected at RH>85%. For affected homes, an average of 24% of the data was removed, with a minimum of 1% (18 1-min averaged data points out of 1297 data points removed) and a maximum of 61% (992 1-min averaged data points out of 1621 data points removed) for individual home samples. For winter, 15 days out of the total 40 days of sampling had at least one 1-min average data point collected at R H >85%. For affected homes, a mean value of 29% of the data was removed with a minimum of 2% (22 1-min averaged data points out of 1396 points removed) and a 71 maximum of 56% (677 1-min averaged data points out 2419 points removed) for individual home samples. In order to further test the sensitivity of the infiltration and air cleaner efficiency results to treatment of data at elevated R H (>85%), F i n f , and A C E were re-calculated with concentration data collected with R H >85% removed. Table 3.11 summarizes the Fjnf values calculated for winter and summer with and without R H >85% data. As can be seen from the table, mean differences, in Fjnf values calculated without excluding points with R H >85% and when applying this exclusion, were 0.04 and 0.02 for summer filter and no filter days respectively. For winter, differences of 0.01 and 0.02 were found for filter and no filter days. Based on this comparison, including data in which R H >85% did not have a significant impact on results. F i n f values for all R H data and those for high R H censored data were not significantly different (p<0.0001). Table 3.11 Comparison of F i n r values calculated for winter and summer sampling periods for all RH and RH<85% data Summer Finf Winter F i n f Data Day Mean (SD) (n) Min Max Mean (SD) (n) Min Max All RH Filter 0.19(0.20) (10) 0.01 0.61 0.10(0.08) (19) 0.01 0.30 No filter 0.60 (0.27) (13) 0.30 1.10 0.28(0.18) (16) 0.10 0.68 High RH censored Filter 0.21 (0.23) (10) 0.03 0.75 0.09 (0.08) (19) 0.01 0.30 No filter 0.63(0.30) (13) 0.30 1.08 0.28(0.18) (16) 0.10 0.68 Difference Filter 0.04(0.11) (10) 0.00 0.41 0.01 (0.03) (19) 0.00 0.16 No filter 0.02(0.04) (13) 0.00 0.14 0.02(0.03) (16) 0.00 0.10 Table 3.12 summarizes the A C E values calculated for winter and summer with and without R H >85% data. As can be seen from the table, mean differences between all R H and R H <85% only were 12.2% and 2.0% for winter and summer respectively. Minimum and maximum A C E values were not affected by R H for either summer or winter sampling periods. A C E values for all R H data and those for high R H censored data were not significantly different (p <0.0001). 72 Table 3.12 Comparison of A C E values calculated for winter (n=16) and summer (n=10) sampling periods for all RH and RH<85% data Summer ACE (%) (n=8) Winter ACE (%) (n=16) Data Mean (SD) Min Max Mean (SD) Min Max All RH 64.5 (35.0) -3.2 98.9 54.5 (37.6) -25.0 99.0 High RH censored 62.6 (38.5) -3.2 99.0 58.6 (36.4) -25.0 99.0 Difference 2.0 (4.4) 0.0 14.0 12.2 19.9) 0.0 56.3 A large difference between A C E values for all R H and high R H censored data were not found for winter versus summer and in both seasons, removal of high R H data resulted in non-significant differences between the original data. These results indicate that removal of high R H data points were not necessary for this data. 3.6.2 Basel ine drift The presence of baseline drift indicates possible contamination of the pDR sensing chamber and therefore poor pDR performance. Since only a small subset of the data was affected by baseline drift, the main analysis was conducted using all data. In order to determine whether the removal of this data had a significant effect on the results obtained, Fjnf and A C E values were recalculated for data where baseline drift data was removed and compared for original values calculated. The only way to treat negative drift data is to remove it from the dataset (Wu et al., 2005). A total of 3 days in the summer (10% of data) and 2 days in the winter (6% of data) 3 3 displayed baseline drift. For summer data, the baseline drift was 2 Ug/m", 5 ug/m and 3 3 49.3 Ug/m . For the winter samples, there was an average baseline drift of 2 Ug/m'. Table 3.13 below summarizes the F i n f values calculated for all data and those re-calculated from baseline drift removed data. As can be seen from the data, small differences exist between the two datasets with mean summer Fjnf values of 0.19 and 0.21 calculated for filter days and values of 0.60 and 0.63 for no filter days for all data and baseline drift removed data respectively. Similarly, values of 0.10 and 0.09 for all data 73 and baseline drift data respectively were calculated for filter days while mean Fi„f for no filter days was not affected by the removal of baseline drift. F i n f values for all data and those for baseline drift censored data were not significantly different (p<0.05). Table 3.13 Comparison of F i n f values calculated for winter and summer sampling periods for all data and baseline drift data removed Summer F i n f Winter F i n f Data Day Mean (SD) (n) Min Max Mean (SD) (n) Min Max All data Filter 0.19(0.20) (10) 0.01 0.61 0.10(0.08) (19) 0.01 0.30 No filter 0.60(0.27) (13) 0.30 1.10 0.28(0.18) (16) 0.10 0.68 Drift censored Filter 0.21 (0.23) (9) 0.03 0.75 0.09(0.08) (18) 0.01 0.30 No filter 0.63(0.30) (12) 0.3 1.08 0.28 (0.18) (15) 0.10 0.68 A C E values for both seasons were not affected. A l l homes identified with baseline drift in the summer were not used in the original calculations of A C E (due to absence of complete data for both filter and no filter days) and consequently their removal did not affect mean A C E values. In the winter, baseline drift was identified for day 1 and day 2 in the same home and consequently A C E could not be calculated for this home so mean A C E values for the season were not affected. 3.6.3 Forcing intercept to zero The intercept in the recursive model represents the indoor generation term. Although indoor generated P M was removed using censoring algorithms, these algorithms only identified P M from sources that resulted in high P M peaks and did not account for constant low level sources. In previous work, Allen et al., in press found good agreement between the F;nf values calculated when an intercept was allowed and when the intercept was forced to zero. In the main analysis, an intercept was allowed as this variable would represent an error term which included any indoor generated P M that was not identified and removed by the censoring algorithms. In order to compare results to the relationship found by Allen et al., in press, Fjnf values were calculated with the intercept forced to zero 74 and were compared to the original Fjnf values calculated. Data was tested for significant differences as well as summary statistics for both datasets were compared. A significant difference was found for F j n f between values calculated when an intercept was allowed and values when the intercept was forced to zero (p <0.005) for pooled seasonal data as well as winter data. The differences between these two dataset are summarized below in table 3.14. Table 3.14 Summary of difference between F i n f values with and without an intercept term Summer F i n f Winter F i n f Mean (SD) Min Max Mean (SD) Min Max All data Filter 0.19 (0.20) 0.01 0.61 0.10 (0.08) 0.01 0.30 No filter 0.60 (0.27) 0.30 1.10 0.28 (0.18) 0.10 0.68 0 intercept Filter 0.23 (0.19) 0.04 0.67 0.18 (0.13) 0.43 0.30 No filter 0.69 (0.60) 0.25 2.53 0.32 (0.21) 0.08 0.87 Difference Filter 0.12 (0.09) 0.01 0.3 0.10 (0.09) 0.00 0.3 No filter 0.12 (0.14) 0.00 0.44 0.11 (0.08) 0.02 0.32 A histogram of the difference between the two datasets was plotted in order to highlight outliers driving the large differences. This plot can be seen below in figure 3.11. 75 D i f f e r e n c e Figure 3-11 Distribution of differences in F i n f values between all data and data with the intercept term forced through zero It was determined that any difference between Fjnf values >0.30 represented extreme values and any home with these large differences were excluded. Five homes were removed from analysis. Scatter plots of all data (a) and outlier removed data (b) can be seen below in figure 3.12. As can be seen from the plots, the removal of outlier data increased the strength of the relationship between Fjnf calculated with an intercept and Fjnf values calculated when the intercept was forced through zero as R 2 increased from 0.55 to 0.78. Significant differences still remained for the two sets of data after outlier removal (p<0.05). Given that inclusion of the intercept term is preferred from the perspective of the accuracy of the mass balance model describing infiltration, in the main analysis we report results from multiple regression models that include an intercept term. 76 a. A l l da ta 3 2.5 2 | 1 5 o 1 E 0.5 0 -0.5 y = 1.06x + 0.05 R2 = 0.55 * 0.2 i i 0.4 0.6 0.8 1 1 F i r f (all da ta ) b. Out l ier removed da ta 1.2 1 0 0.4 1 0.2 0 -0.2 Finf (al l da ta ) y = 0.81 x + 0.09 - R2 = 0.78 • * * * I , • i 0.2 0.4 0.6 i 0.8 i 1 1 Figure 3-12 Relationship between F i n f (with intercept) and F i n f (0 intercept) data for a. all data (n=58) and b. outlier removed data (n=53) 3.7 Multivariable modeling 3.7.1 Housing characterist ics A summary of the dependent variables included in the infiltration and A C E modeling are summarized below in tables 3.15 and 3.16. Five continuous variables air exchange rate (AER), age of home, square footage, number of windows, and carpeting were used (table 3.15) and 7 categorical variables season, stove type, range hood use, air conditioning (AC), heating system, fireplace use, and window use were used (table 3.16). 77 Table 3.15 Summary of continuous independent variables used for infiltration and air cleaner efficiency prediction modeling Winter (n=16) Summer (n=13) Variable Mean (SD) Min Max Mean (SD) Min Max Air exchange rate (AER) (h"1) 0.14 (0.07) 0.04 0.27 0.26 (0.20) 0.07 0.69 Age of home (yrs) 37(11) 15 60 28 (13) 5 50 Square footage (fn 1668 (704) 640 2960 1784 (1105) 928 4300 Number of windows 10(3) 4 13 11 (4) 4 21 Carpeting (%) 53 (26) 0 90 57 (29) 12 93 Table 3.16 Summary of categorical independent variables used for infiltration and air cleaner efficiency prediction modeling Variable Category Winter (n) Summer (n) Season Winter or Summer 16 13 Stove type Electric 15 3 Wood + Gas 1 10 Range hood use Never 6 3 Sometimes 8 4 Always 2 5 Air conditioning No 16 9 (AC) use Yes 0 4 Heating system Gas + Wood + Furnace 14 10 Gas + Electric 2 3 Fireplace use Sometimes 3 0 Never 13 13 Windows use Always 2 4 Sometimes 2 9 Never 12 0 78 3.7.2 Infiltration model ing 3.7.2.1 Winter The following variables were included in analysis of infiltration during winter months: AER, age of home, square footage, number of windows, carpeting, range hood use, heating system, fireplace use, and window use. The variable stove type was not included in modeling due to the lack of variability in the samples. No variables were found to be significantly related (at p=0.25) to infiltration for simple regressions. 3.7.2.2 Summer The following variables were included in infiltration modeling for the summer months: AER, age of home, square footage, number of windows, carpeting, range hood use, AC use, heating system, and window use. Both number of windows and AC use were significant at the p=0.25 level and not significantly associated with each other (see Appendix Q for correlations). The addition of the AC use variable did not increase the R 2 value but was found to increase the model p-value and was therefore excluded from the model. The final model included number of windows which was found to be marginally significant, explaining 13% of the variability in infiltration (p=0.13) with increasing infiltration with increasing number of windows in a home. 3.7.2.3 Pooled seasonal data Due to the small sample size, winter and summer data was pooled. A season variable was created and the sample size was increased to n=29. The following variables were included in the analysis: season, AER, age of home, square footage, number of windows, carpeting, range hood use, AC use, heating system, window use, stove type, and fireplace use. When seasonal data was combined the following variables were found to be significantly associated (p<0.25) with infiltration: season, age of home, AC, heating system, number of windows, window use, and AER. Analysis showed associations among some of the variables. Age of home was correlated with season and window use, while 79 window use was found to be correlated with all of season, AC use, and number of windows. Finally, AER was correlated with both season and window use. The variables window use and AC use were removed from the model due to these correlations (see Appendix Q for all correlations). The final model included only two variables, season and number of windows, and is summarized below in Table 3.18. Together these variables explained approximately 41% of the variability seen in Fj n f values (p <0.0001). Table 3.17 Results of regression analysis of effects of housing characteristics on infiltration in the summer and winter seasons Housing characteristic Estimate (SE) p-value model p-value model R 2 Fjnf Intercept 0.60 (0.20) <0.0001 <0.0001 0.41 Winter Season -0.28 (0.08) <0.0001 Number of windows 0.02 (0.01) 0.04 3.7.3 A C E model ing 3.7.3.1 Winter Simple regressions on A C E values revealed that both age of home and carpeting were significant at the p=0.25 level and were not significantly associated with each other (see Appendix R for correlations). The addition of the carpeting variable did not increase the R 2 value but did increase the model p-value and was therefore not included in the final model. Only age of home was included in the final model and was found to be marginally significant (p=0.24, R =0.24) with decreasing A C E with increasing age of home. 3.7.3.2 Summer No variables were found to be significantly related (at p=0.25) to A C E for the summer sampling period. 80 3.7.3.3 Pooled seasonal data When seasonal data was combined, the variables age of home and carpeting were found to be significantly related to A C E (at p=0.25) and were not significantly associated with each other (see Appendix R for correlations). The addition of the carpeting variable did not increase the R value but was found to increase the model p-value and was therefore excluded from the model. The final model included only the variable age of home which explained approximately 8% of the variability in A C E values for the combined winter and summer data and was found to be marginally significant (p=0.09) with decreasing A C E with increasing age of home. 3.8 Spatial variability In order to investigate the degree of spatial variability in terms of P M concentration, levels of PM2.5 outside sampled homes were compared to the B C Ministry of Environment monitoring network data during the winter and summer sampling periods. Maps were generated in order to show the location of Ministry of Environment monitors in relation to the homes sampled in each town/city. Figures 3.13, 3.14, 3.15, 3.16 show the location of monitors along with the location of homes sampled in Williams Lake, Prince George, Lytton and Spence's Bridge and Surrey respectively. As can be seen from the figures, a large degree of variation exists with respect to the distance between Ministry air quality (AQ) monitoring station locations and the sampled homes. For Lytton and Spence's Bridge, as no ministry A Q monitors exist in these towns, the location of these towns and the closest A Q monitors (Whistler and Kamloops) are shown. Note that the scale of figure 3.15 differs by a large degree in comparison with that of figures 3.13, 3.14, and 3.16. Additionally, only the locations of Lytton and Spence's Bridge are shown as opposed to the locations of the sampled homes in these communities. 81 Figure 3-13 Location of BC Ministry of Environment monitoring network and sampled homes in Williams Lake in August 2004 82 Figure 3-14 Location of BC Ministry of Environment monitoring network and sampled homes in Prince George in January to March 2005 83 S P A C E ' S BRIDGE SLEETSIS CREEK FIRE (2)1 L e g e n d • Sampling locations ^ | Ministry Air Quality Monitors i ! Water A 0 1 30 i 60 1 i 120 Kilometers i I Figure 3-15 Location of BC Ministry of Environment monitoring network in Whistler and Kamloops and location of sampling conducted in Lytton and Spence's Bridge in August 2005 (note: scale) 84 Figure 3-16 Location of BC Ministry of Environment monitoring network and sampled home in Surrey in September 2005 In order to determine if Ministry of Environment air quality (AQ) readings appropriately characterize outdoor levels throughout a city/town during smoke episodes, the relationship between sampled outdoor home concentrations and those concentrations measured at the nearest monitoring network site was examined. As the pDR tends to overestimate PM2.5 concentrations, all pDR concentrations collected in this work were first calibrated against the HI2.5 based upon the collocated measurements outside of each home (see equations 10 (winter, R2=0.85) and 11 (summer, R2=0.78) below as well as figure 3.2). H l c o n c = 0.47(pDR c o n c ) - 0 . 6 6 ( 1 0 ) H l c o n c =0.28(pDR c o n e ) + 3.29 f 1 1 l 85 Correlations between the outdoor home and monitoring network concentrations were calculated for 48 hour data for each sampled home as well as for each monitor in the communities of Williams Lake, Prince George, and Surrey, Lytton, and Spence's Bridge. These correlations are summarized below in table 3.18 by city/town. A l l Lytton and Spence's Bridge home concentrations were compared to both the Whistler and Kamloops monitors as no monitors were located in either Lytton or Spence's Bridge. The lowest correlations were seen between Lytton and Spence's Bridge home concentrations and Whistler and Kamloops monitoring data. This is reasonable to expect as these monitors were located much further away compared with the monitor and homes in Prince George, Surrey, and Williams Lake (see figures 3.13, 3.14, 3.15, 3.16) with monitoring networks located over 75 km away. The HI readings have been shown to overestimate the T E O M readings (Noullett, Jackson, & Brauer, 2006) and consequently a ratio of unity would not be expected. Overall, monitoring networks in William's Lake, Surrey, and Prince George showed reasonably good correlation with outdoor home concentrations. A low correlation was seen for monitor 450270 in Prince George, with a mean Spearman correlation coefficient of 0.21. From figure 3.14, it can be seen that this monitor is located further away from sampled homes compared to the other monitor, which in conjunction with other factors such as geography, may explain the lower monitor agreement. 86 Table 3.18 Summary of correlations between outdoor home and BC Ministry of Environment monitor hourly PM 2 . 5 concentrations over a 48-hour sampling period in the communities of Williams Lake, Surrey, Prince George, Lytton and Spence's Bridge Location Monitor Mean (SD) Range Williams Lake (n=4) E248797 0.48 (0.18) 0.32-0.73 605020 0.47 (0.39) -0.11-0.71 Surrey (n=1)1 E206271 0.42 -Prince George (n=19) 450270 0.21 (0.18) -0.6 450307 0.46 (0.22) 0.41-0.82 Kamloops (n=7)2 E206898 -0.05 (0.25) -0.43-0.32 605008 0.10 (0.30) -0.16-0.59 Whistler (n=7) 2 E227431 -0.16 (0.31) -0.46-0.52 1 No S D or range were calculated due to a sample size of 1 at this sampling location 2 B C Ministry of Environment monitoring data compared to data collected at homes in Lytton and Spence 's Bridge The relationship between 24 hour averaged PM 2 . 5 concentration data for each home and matched 24 hour averaged ministry data was examined for the communities of Williams Lake, Prince George, and Surrey. Due to the large distance between A Q monitors and the communities of Lytton and Spence's Bridge (>75 km), data from homes in these communities was excluded from analysis. As can be seen from figure 3.17 below, a fairly weak relationship exists between the home and A Q monitors. A high degree of scatter exists between home and A Q monitors for both low and high P M 2 5 concentrations. For high PM2.5 concentrations (>25 pg/m3), monitoring networks appear to underestimate concentrations measured at sampled outdoor home locations. This result indicates that the existing monitoring network may not accurately describe community PM2.5 exposure during events such as forest fires or residential wood burning when concentrations may be high. 87 CM •*—' E « I o o. 3 > V, w c Si, c o a) « i - V 3 O O C 8 (0 'c o m 35 30 25 20 15 10 5 0 y = 0.19x + 6.22 R2 = 0.07 • • • • • * • • .* * • • • « « 0 10 20 30 40 50 Outdoor home 24 hour averaged PM2.5 concentrations (ug/m3) Figure 3-17 Relationship between 24 hour averaged PM2.s data for outdoor sampled home and BC Ministry of Environment monitoring network for the communities of Williams Lake, Prince George, and Surrey 88 4 Discussion 4.1 Data collection Field sampling was conducted in winter and summer seasons. A l l winter-time participants resided in Prince George and were recruited through a COPD support group at the local hospital as well as through friends and family of enrolled volunteers. Sampling in the summer occurred in Kamloops and Williams Lake in 2004 and in Lytton, Spence's Bridge and Surrey in 2005. Communities affected by forest fire smoke were identified through a variety of information sources including on-line air quality readings, wildfire information websites as well as satellite imaging of smoke plumes. Since residential wood burning occurs over a predictable period of time, identifying an area in which to conduct sampling and lining up volunteers prior to the sampling period was found to be the most appropriate method of sampling in the winter. For summer sampling, identification of an affected community was made possible through the sources listed earlier but volunteer recruitment proved more challenging due to the unpredictable and relatively short-term smoke episodes resulting from forest fires. The most successful method of recruitment in the summer was found to be approaching local townspeople once in an affected community. Subsequent volunteers were found through "word of mouth" from existing participants. This as well as issues arising from equipment use resulted in a sample size of 19 homes in the winter and 13 homes in the summer which was slightly smaller than the sample size of 20 homes per season that was targeted prior to starting the study. Additionally, all study participants represent a convenience sample and the results therefore may not be representative of the general public. 4.2 Infiltration In order to assess health effects among an exposed population, it is important to accurately predict or estimate exposure levels. As the majority of people's time is spent indoors, exposure indoors plays a major role in total exposure. In the past, exposure studies have mainly relied on central site outdoor measurements as total exposure levels under the assumption that individual exposure is equivalent to these measurements. 89 Indoor exposure is not necessarily equal to outdoor exposure levels and in order to appropriately and accurately estimate exposure levels to individuals it is important to quantify infiltration. Average F;nf values of 0.27 ±0.18 and 0.61 ± 0.27 were found in homes sampled in winter and summer periods respectively. A range of infiltration factors was found with summer values ranging from 0.30 to 1.10 and winter values ranging from 0.10 to 0.68. These values were calculated on non-intervention days (i.e. on days when the air cleaner was run without the HEPA filter). The F i n f values measured in this study suggest that for summer, the home does not provide optimum protection against PM 2 . 5 exposure. While in the winter, approximately 30% of outdoor PM2.5 infiltrates indoors, in the summer this is true of over 60% of the outdoor particulate matter. High variability in infiltration was observed across homes in both seasons indicating that for some homes very little protection may in fact be offered to residents remaining indoors during times of high P M levels. Two other studies have used the recursive model to calculate infiltration of PM2.5 into homes and both studies have found infiltration rates higher than those found for this current work. As in our work, Allen et al., 2003 calculated a higher Fjnf for the non-heating season (March to September), with a mean value of 0.79 ± 0.18 compared to a mean value of 0.53 ± 0.61 for the heating season (October to February). Wu et al., 2006 used the recursive model to measure infiltration of PM2.5 from agricultural burning into homes. Sampling was conducted in thirty-three homes from September to November with F i n f values ranging from 0.25 to 0.94. The lower F;nf values seen for this work could potentially be explained by the difference in housing characteristics resulting from geographical differences. A l l winter time sampling was conducted in homes located in a Northern BC community where temperatures typically reach as low as -12 °C in the winter. Consequently, residents equip for this cold climate through more insulated, "tighter" homes. This same line of reasoning would also apply to homes sampled in the summer as the majority of homes 90 were sampled in interior BC. Additionally, for summer homes where sampling was conducted during forest fire events when residents may tend to close windows in an attempt to reduce smoke entering the home. At times when the level of smoke is not bothersome but P M concentrations may still be high, the number of windows open in a home during the summer would still be expected to be higher than in the winter. This would lead to higher infiltration as opening of more windows and doors in order to promote cooling within the home would function to allow more particulate matter from outdoors to enter the home. Air exchange rates within a home would therefore be higher in homes sampled in the summer versus the winter as reflected in our average A E R values calculated. An average AER of 0.25 ± 0.21 h"1 was found for homes sampled in the summer compared to homes sampled in the winter with an average A E R of 0.14 ± 0.06 h"1. These rates are lower than those measured for other studies (Abt et a l , 2000a; Allen et al., 2003) which can again be attributed to the "tighter" homes in this study compared with those of other studies due mainly to geographical differences. 4.3 Air cleaner use In both seasons, infiltration was lowered with air cleaner use. Infiltration factors of 0.19 ± 0.20 and 0.10 ± 0.20 were found for days when the air cleaner was run with the HEPA filter for summer and winter sampling periods respectively. Infiltration decreased with air cleaner use for 7 out of 8 homes sampled in the summer with Fjnf values ranging from 0.01 to 0.61. Infiltration decreased in 14 out of 16 homes sampled in the winter with Fi nf values ranging from 0.01 to 0.30. Air cleaner efficiency calculations showed that air cleaners were effective at decreasing infiltration for most homes in both winter and summer sampling periods. Mean air cleaner efficiency (ACE) values of 65 ± 35% and 55 ± 38 % were calculated for summer and winter sampling periods respectively. As infiltration was not decreased on filter days for some homes, negative A C E values were also calculated with values of -25% and -3.2% for winter and summer sampling seasons respectively. For both seasons, maximum A C E values of 99% were also observed. 91 Although a higher mean A C E was found for the summer, a significant difference for A C E between winter and summer seasons was not found. This indicates that air cleaners were effective in both seasons, and therefore their use is an important recommendation in winter and summer seasons. No differences in AERs, or any other housing characteristics including volume of the home, window usage or home age were found for homes where infiltration was not lowered on filter days compared to homes where infiltration was lowered. Therefore it is unclear why the air cleaner was not effective in homes WL-03, PG-04 and PG-06 where Fjnf was not lowered with air cleaner use. A number of factors could have lead to this observation. Since the conditions of the home were not controlled for, a change in other variables such as air exchange could have affected both infiltration and air cleaner effectiveness. Additionally other factors, including measurement error or the air cleaner being turned off or set to a lower setting for a portion of the sampling period, may have contributed to the ineffectiveness of the air cleaners in these homes. This indicates that a greater detail of information may be needed to be collected during the sampling period in order to capture the changes in home conditions. One such detail could include the addition of the number of windows open and the presence of indoor sources other than cooking or cleaning to the time activity log, as well as the collection of multiple A E R measurements. No other study has calculated A C E by measuring decreases in Fjnf. Henderson et al., 2005 investigated indoor PM2.5 concentrations in homes with and without air cleaners. It was found that indoor concentrations were lowered by 63-88% in unoccupied homes with 3 electrostatic air cleaners compared to unoccupied homes with no air cleaners during forest fire smoke events. Although operation of air cleaners was shown to decrease indoor PM2.5 concentrations, a similar efficiency would not be expected when the home was occupied and indoor sources increased P M within the home. Mott et al., 2002 also reported that air cleaners were useful during periods of high P M due to forest fires. No exposure measurements were taken for the study, but use of a HEPA filter was associated 92 with decreased reporting of respiratory symptoms experienced by affected residents. Longer use of the air cleaner was also associated with decreased symptom reporting. 4.4 Multivariable modeling 4.4.1 Infiltration prediction No variables were significant in explaining the variability seen in Fjnf values for the winter or summer season. The small sample size, which led to the observance of a low degree of variation within the data, is thought to partly explain this result. When seasonal data was combined, season and number of windows were significantly related to infiltration and together explained 41% of the variability seen in F i n f values (p <0.0001). The number of windows in a home was associated with an increase in infiltration, with a coefficient of 0.02 ± 0.01 while the winter season was associated with a decrease in infiltration, with a coefficient of -0.28 ± 0.08. This result is consistent with Fjnf calculations for our work which showed lower infiltration for homes sampled in winter for both filter and no filter days. An increase in infiltration with the number of windows in a home is reasonable as more windows in a home may result in greater air exchange rate. Similar to our results, in a regression analysis to predict F i nf, Allen et al., 2003 found a coefficient of 0.27 ± 0.05 for homes in the non heating season and a coefficient of 0.02 ± 0.05 for the presence of storm windows. Abt et al., 2000 found that season was a better predictor of indoor concentrations within a home compared to other variables such as home type. Home type was not included in our modeling due to the low degree of variability seen in homes sampled with the majority of homes being single dwellings. 4.4.2 A C E prediction No variables were significantly associated with A C E for winter or summer analysis. When seasonal data was combined, age of home was found to be marginally associated 93 with A C E , explaining 8% of the variability seen in A C E values (p=0.09) indicating decreased efficiency with increasing age of home. This result agrees with the literature which shows that older homes are generally less insulated and leakier, resulting in greater air exchange. This would therefore result in a greater volume of air entering and leaving the home, which would increase the volume of air for the air cleaner to filter. The variable, air exchange rate was not found to be significantly related to A C E . A smaller range of values was seen for this variable which in part could explain air exchange rate not being significantly related to A C E values. Additionally, air exchange was only measured as a point value once over the two sampling days. Although this value was believed to be representative of the air exchange within the home, multiple measurements would have been more informative. Another potential explanation for no housing characteristics predicting A C E is that air cleaners are in fact effective, regardless of the homes in which they are operated. Air cleaners were found to be effective in both summer and winter and across most of the sampled homes. For homes where air cleaners were not effective, no significant difference in housing characteristics existed in comparison to other sampled homes where air cleaners were effective. These results could therefore indicate that the use of air cleaners is appropriate in homes that differ in housing characteristics. 4.5 Monitor evaluation 4.5.1 Sensit ivity analysis Drift in monitor concentration was identified for both indoor and outdoor monitors for 3 summer and 1 winter home constituting 9% of total data. Drift data was included in the main analysis as sensitivity analysis showed that removal of this data, as recommended in other studies, did not significantly affect results. A maximum drift of 2 ug/m" was found and represents a smaller value than what other studies have found. Rea et al., 2001 found both negative and positive drift values ranging from 15-20 Ug/m3 where monitors were zeroed only once a week during data collection. Howard-Reed et al., 2000 found monitor drift of up to 6 Ug/m3 for independent lab tests prior to data collection for personal 94 monitors tested over a 48 hour period. It was not clear if zeroing was performed on the monitors prior to or during the test period. Wu et al., 2005 also identified drift in 9% of their data but after removal of this data, found overall concentrations to change by only approximately 1%. In our work, the low level of drift observed was attributed to the high frequency of monitor zeroing throughout the study as well as factory calibration of monitors prior to sampling. It was concluded that negative drift did not play a major role in overall pDR performance in our work. High R H was identified in 15 out of 32 homes and in 23 out of 64 days worth of sampling for both winter and summer sampling periods. Relative humidity was concluded to have an effect on the sampled concentration of only one home as this was the only home which showed evidence for higher PM2.5 concentrations at higher RH. No significant differences (at p=0.05) were found between the original and the R H >85% removed data for average F i n f or A C E values. Different methods to manage P M concentration data for periods of high R H have been suggested including statistical approaches (Quintana et al., 2000; Wu et al., 2005) and use of heaters or diffusion dryers to remove particle-bound water (Quintana et al., 2000). Both Quintana et al., 2000 and Wu et al., 2005 concluded that removal of R H >95% was the optimal method of handling such data. As no association between R H and high PM2.5 readings was found for data in this work, such data was not removed. This finding is different from what other researchers have found. Relative humidity is dependent on both the amount of water vapor available and the temperature of the surrounding air (Poschl, 2005). While increasing the water vapor available will lead to an increase in the actual amount of water in the air consequently increasing RH, reducing temperature will also increase RH without affecting the amount of water present in the air. Since light scattering is affected by the amount of water in the air, an increase in R H will not necessarily affect measurements collected if the increase in R H is due to a change in temperature alone. The lack of a relationship between high P M concentration readings at increasing R H seen in our work could potentially be explained 95 by this phenomenon. A measure of absolute water content in the air would allow for more appropriate comparisons between studies to determine if the environmental sampling situations are similar between studies and therefore whether a similar relationship between humidity and pDR PM2.5 concentrations measured should be observed. 4.5.2 Direct cont inuous and integrated filter monitor compar ison Collocated pDR and HI2.5 data showed good agreement with an R 2 value of 0.86 for pooled winter and summer data. Separate analysis of winter and summer data showed relationships with R 2 values of 0.85 and 0.78 respectively. These values are comparable to those found by Lui et al., 2002 with R 2 values of 0.84 for readings collected in homes with no cooking and 0.77 for days with cooking between pDR and HI2.5 concentration data. The direct reading monitors, the pDRs, were concluded as being reliable indicators of PM2.5 after analysis of collected data. However, under some conditions such as low P M environments, the pDR may not be as reliable as other monitors and this should be taken into account when planning a sampling strategy. High relative humidity and the occurrence of negative drift were not found to significantly affect pDR performance. 4.6 Absorbance and levoglucosan analysis Mean ABS coefficients were 1.2 ± 0.7 x 10"5 m"1 and 0.5 ± 0.3 x 10"5 m"1 for winter and summer respectively. The winter average is comparable to a mean ABS coefficient of 1.4 x 10 5 m"1 found by Noullett, 2004. This study was conducted in Prince George during February to March. Lower correlations between ABS and PM2.5 concentration were found for our work versus those found by Noullett, 2004. This may be due to the larger portion of P M from residential wood smoke in our study versus that of Noullett, 2004 as the latter study sampled for ambient PM? sduring a period when wood burning would not be as frequent as during our study. Consequently, the PM2.5 sampled by 96 Noullett, 2004 may have consisted of a larger portion of traffic-related particulate leading to a stronger relationship between PM2.5 concentration and ABS coefficients. For comparison purposes, ratios of mean absorbance to mean P M 2 5 concentration were calculated for samples collected at each site in a study conducted by Smargiassi et al., 2005 looking at P M concentration and ABS relationships and was compared to winter and summer ratios calculated for our work. The following ratios were calculated for the 4 sampling sites with differing levels of traffic from Smargiassi et al., 2005: 0.15, 0.10, 0.12, and 0.16. For our work, lower ratios of 0.04 and 0.08 for summer and winter home samples were calculated. The latter study investigated ABS coefficients at 4 residential sites in order to determine traffic density level. Findings showed that although PM2.5 concentrations at all 4 sites were similar, a gradient of absorption coefficients was observed with increasing ABS with increasing traffic density. Since the study analyzed traffic in an urban setting as opposed to wood/forest fire smoke in smaller communities as in our work, lower ABS coefficients are reasonable to expect. This also indicates that PM 2 5concentration and ABS will not always show a strong correlation and that ABS may be a better indicator of PM2.5 from traffic combustion sources versus vegetation combustion sources. Mean levoglucosan concentrations of 0.40 ± 0.30 pg/m 3 and 0.10 ± 0.10 pg/m3 were found for winter and summer sampling periods respectively. These values indicate that higher levels of particulate matter resulting from wood burning in the winter were collected than particulate resulting from forest fires in the summer. Regression analysis showed that similar relationships between PM2.5 and levoglucosan concentration existed for both seasons. Reasonably good agreement was found between the amount of levoglucosan recovered from each filter and P M 2 5concentration collected. Both pDR and HI2.5 data showed similar agreement with R 2 values ranging from 0.53-0.58 for winter and summer data. The fact that similar R 2 and correlation values were seen for both HI2.5 and pDR data between levoglucosan provides support for the use of pDRs as reliable PM2.5 monitors. 97 Levoglucosan concentrations were also regressed with ABS coefficient values. Relationships with R 2 values of 0.35 and 0.40 between ABS and levoglucosan concentrations for winter and summer periods respectively were found. ABS coefficient and levoglucosan concentration correlations revealed similar winter and summer correlation values of 0.59 and 0.63 respectively. The stronger relationship between levoglucosan and PM2.5 concentration and the weak relationship between levoglucosan and ABS coefficients indicate that levoglucosan analysis is important and useful when looking at wood smoke source based P M studies while ABS may only be indicative of traffic-related P M . 4.7 Study strengths and limitations A major strength of this work is that all sampling was conducted in homes during forest fire and residential wood burning smoke events versus in unoccupied homes or in laboratory settings. This was possible due to the sampling method which included continuous PM2.5 concentration sampling allowing for the calculation of infiltration in homes after the removal of indoor P M sources. Due to the complex composition of P M , it is not possible to attribute any sample of particulate matter solely to one source, even if sampling is conducted during periods of wood or forest fire smoke. Integrated filter sampling allowed for analysis of levoglucosan of collected filters, enabling us to validate that a large portion of collected PM2.5 was due to forest fire/wood smoke. Finally, the collection of housing characteristics allowed for the investigation of variables important in explaining both Fjnf and A C E although the small sample size hindered the development of informative models. Although a larger sample size was targeted, it was difficult to meet in the summer due to the sporadic nature of forest fires. In addition, data clean up steps called for the removal of invalid data that further reduced the final sample size. For this data collection period, a shorter sampling time of 48-60 hours was used compared to other studies looking at infiltration which had sampling times ranging from 24 hours to 12 month periods. Due to the nature of forest fires, it is unrealistic to sample in a home for longer than this period 98 as most forest fire smoke events are relatively short lived. For winter time sampling, because residential wood burning is more predictable and consistent, a longer sampling time would be possible. In order to compare seasonal differences however, it would be difficult to compare F i n f values when the amount of data collected per season differs by a large degree. Furthermore, the collection of more or more detailed information would have proved beneficial for all analyses. These include multiple measurements of AER, more detailed time activity diaries as well as the collection of concurrent health related information. Although this work, as well as the study conducted by Henderson et al., 2005 have shown that indoor PM2.5 concentrations as well as infiltration decrease with the use of an air cleaner, questions as to their overall effectiveness in reducing respiratory symptoms still remain. Although elevated PM2.5 concentrations can cause the experiencing of adverse health effects, the decrease in indoor PM2.5 levels associated with air cleaner use has not reliably been shown to decrease the occurrence of these symptoms. This work could further be advanced by studying both PM2.5 concentration as well as respiratory symptoms experienced. This was attempted but due to logistic problems such as a short sampling period as well as lack of sufficient number of COPD participants, it was difficult to gather such information. 4.8 Conclusion Infiltration and air cleaner effectiveness results of this study have important health policy implications. First, the results indicate that infiltration differs significantly across seasons. For summer, high F i n f values indicate that remaining indoors may not be protective of exposure. A mean Fjnf value of 0.61 was found in homes, indicating a mean reduction in concentrations of only 39% relative to outdoor PM2.5 concentrations in the summer. A mean A C E of 65% was found for summer, indicating that the use of air cleaners indoors could increase the protection level to approximately 80% relative to outdoor exposure. In winter, infiltration was found to be low, especially when compared to values found in the literature. A l l homes were sampled in Prince George, a northern Canadian community. It is reasonable to expect that these homes would be "tight" in order to maximize heat retention within the home during the cold winter climate which 99 would have the additional benefit of lowered infiltration of outdoor pollutants, including PM2.5. With a mean Fjnf value of 0.27 measured in these homes, a mean reduction in indoor concentrations of 73% relative to outdoors is expected. With the use of an air cleaner, the level of protection increases to 85% relative to outdoors. In terms of housing characteristics such as home age and volume and behavioral characteristics such as window use, none could significantly predict the effectiveness of air cleaners. The fact that air cleaners were found to be effective in both summer and winter across a wide range of housing characteristics indicates that their use can be recommended in any home. Additionally, we were unable to explain much of the variability seen in infiltration suggesting that infiltration is not easily predicted from home to home and therefore providing a general recommendation of using air cleaners to the public may be appropriate. This is an important finding for community members concerned about their exposure during times of high PM2.5 concentration as our results indicate that the use of HEPA filter air cleaners can dramatically reduce indoor concentrations across different homes. As this study only evaluated HEPA filter air cleaners it is not possible to extrapolate these findings to air cleaners that operate via different techniques. Additionally, room sizing is an important consideration as air cleaners are only designed to filter a certain volume of air over a given time. By increasing the room size, and therefore the volume of air the cleaner must filter, air cleaner effectiveness can be reduced. Along with recommending the use of HEPA filter room air cleaners during smoke episodes, it is crucial to provide information on room sizing, as specified by the manufacturer, in order to ensure that these interventions will be appropriate. Fixed-location monitoring network (e.g. Ministry of Environment) air quality measurements which are used to provide a measure of community exposure, do not, in all cases, provide an accurate picture of exposure. First, in sparsely populated areas there may not be any monitors in the vicinity. Second, in communities where monitors do exist, during times of high PM2.5 exposure, these monitors may not provide appropriate descriptions of community exposure due to the high degree of spatial variability in concentrations resulting from residential wood burning or forest fires. This needs to be taken into account during forest fire and residential wood burning episodes leading to 100 high PM2.5 concentrations as some residents may be experiencing higher levels of exposure than those reflected by central site measurements. 101 REFERENCES Abt, E., H , H. , Catalano, P., & Koutrakis, P. (2000a). Relative contribution of outdoor and indoor particle sources to indoor concentrations. Environmental Science & Technology, 34, 3579-3587. Abt, E., Suh, H. H. , Allen, G., & Koutrakis, P. (2000). Characterization of indoor particle sources: A study conducted in the metropolitan Boston area. Environmental Health Perspectives, 108(1), 35-44. Aditama, T. Y . (2000). Impact of haze from forest fire to respiratory health: Indonesian experience. Respirology (Carlton, Vic), 5(2), 169-174. Allen, R., Larson, T., Sheppard, L. , Wallace, L. , J, L. , & Lui, L-J. S. (2003). Use of real-time light scattering data to estimate the contribution of infiltrated and indoor-generated particles to indoor air. Environmental Science & Technology, J7(16), 3484-3492. Allen, R., Wallace, L. , Larson, T., Sheppard, L. , & Liu, L. S. (in press). Evaluation of the recursive model approach for estimating particulate matter infiltration efficiencies using continuous light scattering data. Journal of Exposure Science & Environmental Epidemiology. Antonicelli, L. , Bilo, M . B., Pucci, S., Schou, C , & Bonifazi, F. (1991). Efficacy of an air-cleaning device equipped with a high efficiency particulate air filter in house dust mite respiratory allergy. Allergy, 46(8), 594-600. 102 Babich, P., Davey, M . , Allen, G., & Koutrakis, P. (2000). Method comparisons for particulate nitrate, elemental carbon, and PM2.5 mass in seven U.S. cities. Journal of the Air & Waste Management Association (1995), 50(1), 1095-1105. Barregard, L. , Sallsten, G., Gustafson, P., Andersson, L. , Johansson, L., & Basu, S. (2006). Experimental exposure to wood-smoke particles in healthy humans: Effects on markers of inflammation, coagulation, and lipid peroxidation. Inhalation Toxicology, 18(11), 845-853. Bascom, R., Fitzgerald, T. K. , Kesavanathan, J., & Swift, D. L . (1996). A portable air cleaner partially reduces the upper respiratory response to side stream tobacco smoke. Applied Occupational and Environmental Hygiene, 11(6), 553-559. Boman, B. C., Forsberg, A. B., & Jarvholm, B. G. (2003). Adverse health effects from ambient air pollution in relation to residential wood combustion in modern society. Scandinavian Journal of Work, Environment & Health, 29(4), 251-260. Brauer, M . (2000). Health impacts of air pollution from vegetation fires. Brunei International Medical Journal, 2, 221-236. Bruce, N . (2002). The health effects of indoor air pollution exposure in developing countries. Retrieved February 2, 2005, from w w w. who. int/peh/air/Indoor/OEH02.5 .pdf Chan-Yeung, M . , Ait-Khaled, N . , White, N . , Ip, M . S., & Tan, W. C. (2004). The burden and impact of COPD in asia and africa. The International Journal of Tuberculosis & Lung Disease, 8(1), 2-14. 103 Chia, K. S., Jeyaratnam, J., Chan, T. B., & Lim, T. K. (1990). Airway responsiveness of firefighters after smoke exposure. British Journal of Industrial Medicine, 47(8), 524-527. Clayton, C. A. , Perritt, R. L. , Pellizzari, E. D., Thomas, K. W., Whitmore, R. W., & Wallace, L. A. (1993). Particle total exposure assessment methodology (PTEAM) study: Distributions of aerosol and elemental concentrations in personal, indoor, and outdoor air samples in a southern California community. Journal of Exposure Analysis & Environmental Epidemiology, 3(2), 227-250. Delfino, R. J., Sioutas, C , & Malik, S. (2005). Potential role of ultrafine particles in associations between airborne particle mass and cardiovascular health. Environmental Health Perspectives, 113(8), 934-946. Delfino, R. J., Zeiger, R. S., Seltzer, J. M . , & Street, D. H. (1998). Symptoms in pediatric asthmatics and air pollution: Differences in effects by symptom severity, anti-inflammatory medication use and particulate averaging time. Environmental Health Perspectives, 106(11), 751-761. Duclos, P., Sanderson, L. M . , & Lipsett, M . (1990). The 1987 forest fire disaster in California: Assessment of emergency room visits. Archives of Environmental Health, 45(\), 53-58. Ebelt, S. T., Petkau, A. J., Vedal, S., Fisher, T. V. , & Brauer, M . (2000). Exposure of chronic obstructive pulmonary disease patients to particulate matter: Relationships 104 between personal and ambient air concentrations. Journal of the Air & Waste Management Association (1995), 50(1), 1081-1094. Emmanuel, S. C. (2000). Impact to lung health of haze from forest fires: The Singapore experience. Respirology (Carlton, Vic), 5(2), 175-182. Environment Canada. (2006). Main emission sources, particulate matter. Retrieved 09/19, 2006, from http://www.ec.gc.ca/cleanair-airpur/Main_Emission_Sources-WS9A24D6D1- l_En.htm Evans, G. F., Highsmith, R. V. , Sheldon, L. S., Suggs, J. C , Williams, R. W., & Zweidinger, R. B. (2000). The 1999 Fresno particulate matter exposure studies: Comparison of community, outdoor, and residential P M mass measurements. Journal of the Air & Waste Management Association (1995), 50(11), 1887-1896. Ezzati, M . (2005). Indoor air pollution and health in developing countries. Lancet, 366(94S0), 104-106. Fischer, S. L. , & Koshland, C. P. (in press). Field performance of a nephelometer in rural kitchens: Effects of high humidity excursions and correlations to gravimetric analyses. Journal of Exposure Science & Environmental Epidemiology. Flannigan, M . D., Stocks, B. J., & Wotton, B. M . (2000). Climate change and forest fires. The Science of the Total Environment, 262(3), 221-229. 105 Fraser, M . , & and Lakshmanan, K. (2000). Using levoglucosan as a molecular marker for the long-range transport of biomass combustion aerosols. Environmental Science & Technology, 34, 4560-4564. Goldberg, M . S., Burnett, R. T., & Stieb, D. (2003). A review of time-series studies used to evaluate the short-term effects of air pollution on human health. Reviews on Environmental Health, 18(4), 269-303. Government of the Northwest Territories. (2005). Health effects of smoke exposure due to forest fires. Retrieved 07/21, 2006, from http://forestmanagement.enr.gov.nt.ca/fire_management/fire_docs/forestfiresmokeJu ne2005.pdf Green, R., Simpson, A. , Custovic, A. , Faragher, B., Chapman, M . , & Woodcock, A . (1999). The effect of air filtration on airborne dog allergen. Allergy, 54(5), 484-488. Health Canada. (2006). It's your health, wood smoke. Retrieved 09/20, 2006, from http://www.hc-sc.gc.ca/iyh-vsv/environ/wood-bois_e.html Henderson, D. E., Milford, J. B., & Miller, S. L. (2005). Prescribed burns and wildfires in Colorado: Impacts of mitigation measures on indoor air particulate matter. Journal of the Air & Waste Management Association (1995), 55(10), 1516-1526. Hood, E. (2002). Cooking catastrophe: Chronic exposure to burning biomass. Environmental Health Perspectives, 110(11), A691. 106 Howard-Reed, C , Rea, A . W., Zufall, M . J., Burke, J. M . , Williams, R. W., & Suggs, J. C. (2000). Use of a continuous nephelometer to measure personal exposure to particles during the U.S. environmental protection agency Baltimore and Fresno panel studies. Journal of the Air & Waste Management Association (1995), 50(1), 1125-1132. Janssen, N . A. , Hoek, G., Brunekreef, B., Harssema, H., Mensink, I., & Zuidhof, A . (1998). Personal sampling of particles in adults: Relation among personal, indoor, and outdoor air concentrations. American Journal of Epidemiology, 147(6), 537-547. Koenig, J. Q., Larson, T. V . , Hanley, Q. S., Rebolledo, V. , Dumler, K., & Checkoway, H. (1993). Pulmonary function changes in children associated with fine particulate matter. Environmental Research, 63(1), 26-38. Koutrakis, P., Briggs, S. L . K., & Leaderer, B. P. (1992). Source appointment of indoor aerosols in Suffolk and Onondaga counties, New-York. Environmental Science & Technology, 26(3), 521-527. Kuenzli, N . , Avol, E., Wu, L , Gauderman, W. J., Rappaport, E., & Millstein, J. (in press). Health effects of the 2003 southern California wildfire on children. American Journal of Respiratory & Critical Care Medicine. Lai, A . C. (2002). Particle deposition indoors: A review. Indoor Air, 12(4), 211-214. Larson, T. V. , & Koenig, J. Q. (1994). Wood smoke: Emissions and non-cancer respiratory effects. Annual Review of Public Health, 15, 133-156. 107 Leaderer, B. P., Naeher, L. , Jankun, T., Balenger, K., Holford, T. R., & Toth, C. (1999). Indoor, outdoor, and regional summer and winter concentrations of PMio, PM2.5, S04(2)-, H+, NH4+, N03-, NH3, and nitrous acid in homes with and without kerosene space heaters. Environmental Health Perspectives, 107(3), 223-231. Liu, D., Tager, I. B., Balmes, J. R., & Harrison, R. J. (1992). The effect of smoke inhalation on lung function and airway responsiveness in wildland fire fighters. The American Review of Respiratory Disease, 146(6), 1469-1473. Liu, L . J., Slaughter, J. C , & Larson, T. V . (2002). Comparison of light scattering devices and impactors for particulate measurements in indoor, outdoor, and personal environments. Environmental Science & Technology, 56(13), 2977-2986. Long, C. M . , Suh, H . H. , & and Koutrakis, P. (2000). Characterization of indoor particle sources using continuous mass and size monitors. Journal of the Air & Waste Management Association, 50, 1236-1250. Long, C. M . , Suh, H . H. , Catalano, P. J., & Koutrakis, P. (2001). Using time- and size-resolved particulate data to quantify indoor penetration and deposition behavior. Environmental Science & Technology, 55(10), 2089-2099. Mage, D. T. (2001). A procedure for use in estimating human exposure to particulate matter of ambient origin. Journal of the Air & Waste Management Association (1995), 51(1), 7-10. 108 McDonald, E., Cook, D., Newman, T., Griffith, L. , Cox, G., & Guyatt, G. (2002). Effect of air filtration systems on asthma: A systematic review of randomized trials. Chest, 722(5), 1535-1542. Ministry of Forest and Range. (2006) Archived situation reports, Wildfire news. Retrieved 08/08, 2004, from http://www.for.gov. be. ca/pScripts/Protect/WildfireNews/Situation.asp?archive=l Ministry of Water, Land, and Air Protection. (2006). BC air quality online. Retrieved 02/02, 2005, from http://wlapwww.gov.be.ca:8000/pls/aqiis/air.summary Monn, C. (2001). Exposure assessment of air pollutants: A review on spatial heterogeneity and indoor/outdoor/personal exposure to suspended particulate matter, nitrogen dioxide and ozone. Atmospheric Environment, 35, 1-32. Moore, D., Copes, R., Fisk, R., Joy, R., Chan, K., & Brauer, M . (2006). Population health effects of air quality changes due to forest fires in British Columbia in 2003: Estimates from physician-visit billing data. Canadian Journal of Public Health. Revue Canadienne De Sante Publique, 97(2), 105-108. Mott, J. A. , Meyer, P., Mannino, D., Redd, S. C , Smith, E. M . , & Gotway-Crawford, C. (2002). Wildland forest fire smoke: Health effects and intervention evaluation, Hoopa, California, 1999. The Western Journal of Medicine, 776(3), 157-162. Nolte, C. G., Schauer, J. J., Cass, G. R., & Simoneit, B. R. (2001). Highly polar organic compounds present in wood smoke and in the ambient atmosphere. Environmental Science & Technology, 55(10), 1912-1919. 109 Noullett, M . (2004). Exposure to fine particulate air pollution in Prince George, British Columbia. Unpublished Natural Resources and Environmental Studies, University of Northern British Columbia. Noullett, M . , Jackson, P., & Brauer, M . (2006). Winter measurements of children's personal exposure and ambient fine particle mass, sulphate and light absorbing components in a northern community. Atmospheric Environment, 40, 1971-1990. Offermann, F. J., Sextro, R. G., Fisk, W. J., Grimsrud, D. T., Nasaroff, W. W., & Nero, A. V . (1985). Control of respirable particles in indoor air with portable air cleaners. Atmospheric Environment, 79(11), 1761-1771. Office of Air and Radiation. (2006). Residential air cleaning devices: a summary of available information. Retrieved November/04, 2006, from http://www.epa.gOv/iaq/pubs/residair.html#standard%20for%20portables. Ozkaynak, H. , Xue, J., Spengler, J., Wallace, L., Pellizzari, E., & Jenkins, P. (1996). Personal exposure to airborne particles and metals: Results from the particle T E A M study in riverside, California. Journal of Exposure Analysis & Environmental Epidemiology, 6(1), 57-78. Pope, C. A.,3rd, & Dockery, D. W. (2006). Health effects of fine particulate air pollution: Lines that connect. Journal of the Air & Waste Management Association (1995), 56(6), 709-742. 110 Poschl, U . (2005). Atmospheric aerosols: Composition, transformation, climate and health effects. Angewandte Chemie (International Ed. in English), 44(46), 7520-7540. Quintana, P. J., Samimi, B. S., Kleinman, M . T., Liu, L. J., Soto, K., & Warner, G. Y . (2000). Evaluation of a real-time passive personal particle monitor in fixed site residential indoor and ambient measurements. Journal of Exposure Analysis & Environmental Epidemiology, 10(5), 437-445. Rea, A. W., Zufall, M . J., Williams, R. W., Sheldon, L. , & Howard-Reed, C. (2001). The influence of human activity patterns on personal P M exposure: A comparative analysis of filter-based and continuous particle measurements. Journal of the Air & Waste Management Association (1995), 51(9), 1271-1279. Risom, L. , Moller, P., & Loft, S. (2005). Oxidative stress-induced D N A damage by particulate air pollution. Mutation Research, 592(1-2), 119-137. Rittmaster, R., Adamowicz, W. L. , Amiro, B., & Pelletier, R. T. (2006). Economic analysis of health effects from forest fires. Canadian Journal of Forest Research, 36, 868-877. Robinson, M . S., Chavez, J., Velazquez, S., & Jayanty, R. K. M . (2004). Chemical speciation of PM2.5 collected during prescribed fires of the Coconino national forest near flagstaff, Arizona. Journal of Air & Waste Management Association, 54, 1112-1123. I l l Ruzer, L. , & Harley, N . (Eds.). (2005). Aerosols handbook, measurement, dosimetry, and health effects from http://www.environetbase.com/books/2047/1161 l_fm.pdf Samet, J. M . , Zeger, S. L. , Dominici, F., Curriero, F., Coursac, I., & Dockery, D. W. (2000). The national morbidity, mortality, and air pollution study, part II: Morbidity and mortality from air pollution in the united states. Research Report (Health Effects Institute), 94(Pt 2), 5-70; discussion 71-9. Sapkota, A. , Symons, J. M . , Kleissl, J., Wang, L. , Parlange, M . B., & Ondov, J., et al. (2005). Impact of the 2002 Canadian forest fires on particulate matter air quality in Baltimore city. Environmental Science & Technology, 39(1), 24-32. Schollnberger, H. , Aden, J., & Scott, B. R. (2002). Respiratory tract deposition efficiencies: Evaluation of effects from smoke released in the Cerro Grande forest fire. Journal of Aerosol Medicine, 15(4), 387-399. Schreuder, A. B. , Larson, T. V. , Sheppard, L. , & Claiborn, C. S. (2006). Ambient woodsmoke and associated respiratory emergency department visits in Spokane, Washington. International Journal of Occupational & Environmental Health, 12(2), 147-153. Schwartz, J., Dockery, D. W., & Neas, L . M . (1996). Is daily mortality associated specifically with fine particles? Journal of the Air & Waste Management Association (1995), 46(10), 927-939. Schwela, D. (2000). Air pollution and health in urban areas. Reviews on Environmental Health, 75(1-2), 13-42. 112 Slaughter, J. C , Koenig, J. Q., & Reinhardt, T. E. (2004). Association between lung function and exposure to smoke among firefighters at prescribed burns. Journal of Occupational & Environmental Hygiene, 1(1), 45-49. Smargiassi, A. , Baldwin, M . , Pilger, C , Dugandzic, R., & Brauer, M . (2005). Small-scale spatial variability of particle concentrations and traffic levels in Montreal: A pilot study. The Science of the Total Environment, 338(3), 243-251. Smith, K. R. (2002). Indoor air pollution in developing countries: Recommendations for research. Indoor Air, 12(3), 198-207. Smith, K. R., Samet, J. M . , Romieu, I., & Bruce, N . (2000). Indoor air pollution in developing countries and acute lower respiratory infections in children. Thorax, 55(6), 518-532. Spengler, J. D., Koutrakis, P., Dockery, D. W., Raizenne, M . , & Speizer, F. E. (1996). Health effects of acid aerosols on North American children: Air pollution exposures. Environmental Health Perspectives, 104(5), 492-499. Suh, H. H. , Spengler, J. D., & Koutrakis, P. (1992). Personal exposures to acid aerosols and ammonia. Environmental Science & Technology, 26(12), 2507-2517. Switzer, P., & Ott, W. (1992). Derivation of an indoor air averaging time model from the mass balance equation for the case of independent source inputs and fixed air exchange rates. Journal of Exposure Analysis & Environmental Epidemiology, 2(Suppl. 2), 113-135. 113 Switzer, P. (2001). Estimating separately personal exposure to ambient and non-ambient particulate matter for epidemiology and risk assessment: Why and how. Journal of the Air & Waste Management Association (1995), 51(3), 322-3; author reply 329-38. Thatcher, T. L. , & Laytonb, D. W. (1995). Deposition, resuspension, and penetration of particles within a residence. Atmospheric Environment, 29(13), 1487-1497. Thatcher, T. L. , Lai, A . C. K., Moreno-Jackson, R., Sextro, R. G., & Nazaroff, W. W. (2002). Effects of room furnishings and air speed on particle deposition rates indoors. Atmospheric Environment, 36, 1811-1819. Thermo Anderson. (2003). Personal DataRAM™ Series. Retrieved 10/13, 2004, from www.wolfsense.com/pdf/miepdrl000.pdf Thurston, G. D., Ito, K., Mar, T., Christensen, W. F., Eatough, D. J., & Henry, R. C. (2005). Workgroup report: Workshop on source apportionment of particulate matter health effects—inter-comparison of results and implications. Environmental Health Perspectives, 113(12), 1768-1774. Tung, T. C. W., Chao, C. Y . H. , & Burnett, J. (1999). A methodology to investigate the particulate penetration coeffiecient through building shell. Atmospheric Environment, 33, 881-893. Turner, W., Olson, B. , & Allen, G. (2000). Calibration of sharp cut impactors for indoor and outdoor particle sampling. Journal of the Air & Waste Management Association, 50, 484-487. 114 United States Environmental Protection Agency. How smoke from fires can affect your health. Retrieved June/23, 2006, from http://depts.washington.edu/wildfire/resources/EPA2003.pdf van der Werf, G. R., Randerson, J. T., Collatz, G. J., Giglio, L., Kasibhatla, P. S., & Arellano, A. F.,Jr. (2004). Continental-scale partitioning of fire emissions during the 1997 to 2001 el Nino/La Nina period. Science, 505(5654), 73-76. Vette, A . F., Rea, A. W., Lawless, P. A. , Rodes, C. E., Evans, G., & Highsmith, V . R. (2001). Characterization of indoor-outdoor aerosol concentration relationships during the Fresno P M exposure studies. Aerosol Science & Technology, 34, 118-126. Wallace, L. , Williams, R., Rea, A. , & Croghan, C. (2006). Continuous weeklong measurements of personal exposures and indoor concentrations of fine particles for 37 health-impaired North Carolina residents for up to four seasons. Atmospheric Environment, 40, 399-414. Wallace, L. (1996). Indoor particles: A review. Journal of the Air & Waste Management Association (1995), 46(2), 98-126. Ward, M . , Siegel, J. A. , & Corsi, R. L. (2005). The effectiveness of stand alone air cleaners for shelter-in-place. Indoor A ir, 15(2), 127-134. Wilson, W. E., Mage, D. T., & Grant, L . D. (2000). Estimating separately personal exposure to ambient and non-ambient particulate matter for epidemiology and risk assessment: Why and how. Journal of the Air & Waste Management Association (1995), 50(1), 1167-1183. 115 Wood, R. A. , Johnson, E. F., Van Natta, M . L. , Chen, P. H. , & Eggleston, P. A. (1998). A placebo-controlled trial of a HEPA air cleaner in the treatment of cat allergy. American Journal of Respiratory & Critical Care Medicine, 158(1), 115-120. Wu, C. H. , Jimenez, J., Claiborn, C , Gould, T., Simpson, C. D., Larson, T., & Lui, L-J. S. (2006). Agricultural burning smoke in eastern Washington: Part II. exposure assessment. Atmospheric Environment, 40(2%), 5379-5392. Wu, C. F., Delfino, R. J., Floro, J. N . , Samimi, B. S., Quintana, P. J., & Kleinman, M . T. (2005). Evaluation and quality control of personal nephelometers in indoor, outdoor and personal environments. Journal of Exposure Analysis & Environmental Epidemiology, 15(1), 99-110. 116 APPENDICES 117 Appendix A: Consent forms effects, and residents of smoke-affected areas need information to assess the safety of their homes as a refuge from forest fire smoke. We plan to measure indoor and outdoor concentrations of smoke, to evaluate factors associated with high and low indoor concentrations and to assess the effectiveness of portable air cleaners on reducing indoor levels of smoke. Our measurements will be compared to outdoor measurements collected by regional air quality monitoring stations. I understand that this research is being conducted as part of Ms. Barn's graduate thesis in the School of Occupational and Environmental Hygiene. Study Procedures: I understand that: • The sampling period is 48-hours, and that a researcher will contact me to arrange an appointment to visit my home and to install air quality monitoring equipment in the main bedroom, and in a secure outdoor area. This equipment will measure fine particulate matter (PM 2 5 ) in the air as well as carbon dioxide (CO2) . A quiet, commercially available, portable air cleaner will also be operated during portions of the measurement period. A l l of the equipment is designed to operate quietly and unobtrusively and presents no known risks to the occupants of the home. • The researcher will need access to my home for up to two hours at the time of installation, and that I must be present to answer some questions about my home's characteristics (age, building material, etc.) and health symptoms. I will also need to be present the following day, at a time convenient to me, to answer some questions about my health symptoms and to allow the researcher to check the operation of the monitoring equipment. • If you currently suffer from any respiratory disease (emphysema, chronic bronchitis or chronic obstructive pulmonary disease), we request your permission to contact your personal physician for the purposes of determining the severity of your disease. • I and other residents should engage in normal activities during the duration of the monitoring, and that I will be asked to keep a simple log of any particle-generating activities in the household (cooking, dusting, smoking, barbequing, etc.) for this period. • For the purposes of determining the ventilation rate of our home I and other residents will be asked to vacate the premises for a period of at least two hours at a convenient time during the 48-hour study period. I will be required to note the time and length of this absence on the study log. • The researcher will need access to my home for a period of up to one hour at the end of the study period. At this time I will also need to be present to answer some questions about my health symptoms. Exclusions: I understand that smoking will affect the results of this research, and therefore households with residents who smoke must be excluded from this study. 120 symptoms. Eligible residents are being invited to participate because recent events in this province have identified a need for further research into this relatively unstudied problem. Exposure to particles in smoke has been associated with significant health effects, and residents of smoke-affected areas need information to assess the safety of their homes as a refuge from forest fire smoke. Those who live in colder winter climates also need information to help deal with smoke from wood burning in their neighbourhood. We plan to measure indoor and outdoor concentrations of smoke, to evaluate factors associated with high and low indoor concentrations and to assess the effectiveness of portable air cleaners on reducing indoor levels of smoke. Our measurements will be compared to outdoor measurements collected by regional air quality monitoring stations. I understand that this research is being conducted as part of Ms. Barn's graduate thesis in the School of Occupational and Environmental Hygiene. Sampling in Prince George will also be part of a class project and students of the Environmental Science Program ENSC 412 Air Pollution class will aid in data collection. Study Procedures: I understand that: • The sampling period is 48-hours, and that a researcher will contact me to arrange an appointment to visit my home and to install air quality monitoring equipment in the main bedroom, and in a secure outdoor area. This equipment will measure fine particulate matter (PM2.5) in the air as well as carbon dioxide (CO2). A quiet, commercially available, portable air cleaner will also be operated during portions of the measurement period. A l l of the equipment is designed to operate quietly and unobtrusively and presents no known risks to the occupants of the home. • The researcher will need access to my home for up to two hours at the time of installation, and that I must be present to answer some questions about my home's characteristics (age, building material, etc.) and health symptoms. I will also need to be present the following day, at a time convenient to me, to answer some questions about my health symptoms and to allow the researcher to check the operation of the monitoring equipment. • If you currently suffer from any respiratory disease (emphysema, chronic bronchitis or chronic obstructive pulmonary disease), we request your permission to contact your personal physician for the purposes of determining the severity of your disease. • I and other residents should engage in normal activities during the duration of the monitoring, and I will be asked to keep a simple log of any particle-generating activities in the household (cooking, dusting, smoking, barbequing, etc.) for this period. • For the purposes of determining the ventilation rate of our home I and other residents will be asked to vacate the premises for a period of at least two hours at a convenient time during the 48-hour study period. I will be required to note the time and length of this absence on the study log. 123 Appendix B: Contact Information Sheet Appendix C: Equipment log sheets D A T A R A M L O G Operator N a m e s : Locat ion ID: Start Date: Indoor D a t a R A M : Outdoor D a t a R A M : DAY 1 Pre-measurement indoor calibration start time Indoor measurement start time Indoor measurement end time Post-measurement indoor calibration end time Outdoor measurement start time Outdoor measurement end time Post-measurement outdoor calibration end time FILE NAMES DAY 2 P re-measurement indoor calibration start time Indoor measurement start time Indoor measurement end time Post-measurement indoor calibration end time Outdoor measurement start time Outdoor measurement end time Post-measurement outdoor calibration end time FILE NAMES PESCJPTION OF INDOOR LOCATION DESCRIPTION OF OUTDOOR LOCATION 129 HOBO LOG O P E R A T O R N A M E S : L O C A T I O N : . S T A R T D A T E : . M O N I T O R : START TIME AND DATE END TIME AND DATE FILE NAME NOTES 130 Q-TRAK LOG OPERATOR NAMES: LOCATION:_ START DATE: Q-TRAK:; DAY 1 S T A R T TIME END TIME FILE N A M E DAY 2 S T A R T TIME END TIME FILE N A M E NOTES Day 1: HEPA and Pre-Filter IN OUT Day 2: HEPA and Pre- Filter IN OUT 131 PARTICULATE SAMPLING L O G SITE OPERATORS: : AIR PUMP ID: LOCATION: F L O W M E T E R ID: FILTER ID: AIR P U M P ON OFF NOTES D A T E A N D TIME MONITOR TIME PRE-M E A S U R E M E N T FLOW R A T E (L/MIN) D A T E A N D TIME MONTOR TIME POST-M E A S U R E M E N T F L O W R A T E (L/MIN) I M P A C T O R O N MID CHECK O F F D A T E A N D TIME FLOW (L/MIN) D A T E A N D TIME F L O W (L/MIN) D A T E A N D TIME FLOW (L/MIN) INITIAL ADJUSTED INITIAL ADJUSTED INITIAL t o Appendix D: Air cleaner specifics • Product name: SilentComfort, Ultra Quiet HEPA Air Cleaner, Model 18150, Honeywell Enviracaire • Filters: HEPA filter (replacement filter: HRF-14), carbon pre-filter (replacement pre-filter: 38002) • A H A M Certified • Recommended Room Size: 15'xl5' ( 235 sq.ft.) • A H A M CADR Rating: 150 133 Appendix E: Housing characteristics questionnaire D W E L L I N G I N F O R M A T I O N F O R M ( T O B E C O M P L E T E D B Y F I E L D T E C H N I C I A N ) S ITE ID: D A T E : 1. Address: 2. Postal Code: 3. GPS Accuracy: 4. Latitude: 5. Longitude: 6. Elevation: 7. Age of home: 8. Proximity to major roads (major road = 4 lanes): On a major road? Yes No If not, within 50m of a major road? Yes No 9. Type of building: single house townhouse apartment building other: 10. Building location: Street canyon (street for which the ratio of the distance from the buildings to the axis of the street and the height of the building was less than 1.5 - this can be estimated by field worker) Yes No 11. Apartment location in building (if applicable): Floor number: Corner unit? Yes No Side of building: North South East West 12. Size of home: Square footage: Ceiling height: Approximate volume: 135 13. Number of rooms: 14. Number of windows: Number of windows that open: 7. Estimate percentage of floor space covered with carpets (for the entire house): 8. Kitchen: Stove type? Range hood? If yes, is it used? If yes, how often? 9. Type of Ventilation: Natural only System: Gas Yes Yes Always Electric No No Sometimes Never 10. Air conditioning: Present? Yes No If yes, what kind? Window Central 11. If the home does not have air conditioning, what strategies do the residents use to keep cool: Open windows Ceiling fans Floor/table fans 12. Heating System: Electrical Gas Forced Air/Furnace Hot Water/Radiator Other: 13. Fireplace: Present? Yes No Type and number: Wood Gas How often is it used? 14. Independent air filter/cleaner: Yes No Type: Location: How often is it used? 15. Windows: Description of windows that are opened (size, quantity, average use per summer day) Always: Sometimes: 16. Attached garage: Yes No Never: Appendix F: Activity log ACTIVITY LOG LOCATION: DATE: FILTER: START TIME: ^ Location of resident Home unoccu-pied? Cooking? Dusting or Vacuuming? Tobacco Smoke? Windows Open? A/C On? Oxygen use? Medication? Notes/Comments Time KITCHEN LIVING RM BEDRM YES YES YES YES YES YES YES YES 5:00-5:30 A M 5:30-6:00 A M 6:00-6:30 A M 6:30-7:00 A M 7:00-7:30 A M 7:30-8:00 A M 8:00-8:30 A M 8:30-9:00 A M 9:00-9:30 A M 9:30-10:00 A M 10:00-10:30 A M 10:30-11:00 A M 11:00-11:30 A M 11:30-12:00 P M 12:00-12:30 P M 12:30-1:00 P M 1:00-1:30 P M 1:30-2:00 P M 2:00-2:30 P M oo Appendix G: Introductory letters UBCF" T H E U N I V E R S I T Y O F B R I T I S H C O L U M B I A School of Occupational & Environmental Hygiene 3 r d Floor - 2206 East Mall Vancouver, BC Canada V6T 1Z3 www.soeh.ubc.ca (604) 822-9595 te/ (604) 822-9588 fax U B C Forest Fire Smoke Study May 4, 2005 Dear Resident: We are a team of researchers from the University of British Columbia (UBC) School of Occupational and Environmental Hygiene and the BC Centre for Disease Control who are measuring indoor concentrations of airborne particles in B C communities affected by forest fire smoke. We are writing to ask you to allow us to monitor the air inside and outside your home for a 48-hour period. We are interested in learning more about indoor levels of smoke in communities affected by forest fire events and the relationship between indoor smoke levels and health symptoms. We regret to point out that those homes with residents who smoke will not be considered for participation. Airborne particles are present in outdoor air at elevated concentrations during forest fire events. High concentrations have been linked to increased respiratory health symptoms, especially in persons with pre-existing respiratory conditions. While regional air quality monitoring stations measure outdoor concentrations of smoke, indoor levels are difficult to predict from this information and there has been insufficient research carried out to know what the appropriate advice to give to community residents regarding strategies to reduce exposure to smoke during forest fire periods is. Given recent events in British Columbia, the goal of this research is to determine what range of particle concentrations residents of communities impacted by forest fire smoke can expect to be exposed to while indoors. We will also evaluate the effectiveness of portable air cleaners on reducing indoor levels of smoke. Although this information will likely be too late for making recommendations for your community this year, we hope that this study will allow us to provide better advice in the future about ways that people can best reduce exposures in such situations. The study will take place between May 2005 and September 2005. We will monitor the air in participating homes for 48 hours with three small sampling devices located in the main bedroom of the house, and two samplers located outdoors. Sampling can begin at a time convenient for you. We will also operate a quiet, commercially available, portable air cleaner in the main bedroom during portions of the measurement period. The researcher will heed to be in your home for up to 2 hours to set up the 140 T H E U N I V E R S I T Y O F B R I T I S H C O L U M B I A School of Occupational & Environmental Hygiene 3rd Floor - 2206 East Mall Vancouver, BC Canada V6T 1Z3 www.soeh.ubc.ca (604) 822-9595 tel (604) 822-9588 fax UBC/UNBC Forest Fire and Residential Wood Smoke Study Research Information Letter November 4, 2004 Dear Resident: We are a team of researchers from the University of British Columbia (UBC) School of Occupational and Environmental Hygiene, University of Northern British Columbia (UNBC) Environmental Science Program and the BC Centre for Disease Control who are measuring indoor concentrations of airborne particles in BC communities affected by forest fire smoke and smoke from residential wood burning. We would like you to allow us to monitor the air inside and outside your home for a 48-hour period. We are interested in learning more about the relationship between indoor smoke levels and health symptoms. We regret to point out that those homes with residents who smoke will not be considered for participation. High concentrations of airborne particles have been linked to increased respiratory health symptoms, especially in persons with pre-existing respiratory conditions. While regional air quality monitoring stations measure outdoor concentrations of smoke, indoor levels are difficult to predict from this information. There has not been sufficient research carried out to know what is the appropriate advice to give to community residents regarding strategies to reduce exposure to smoke during periods of poor air quality. During the winter in Prince George there are often temperature inversions that act to trap pollutants such as wood smoke and other combustion particles within the city "bowl". During these times there is very little circulation of pollutants and it is suspected that concentrations can build-up to high levels in neighbourhoods with residential wood burning. Given recent events in British Columbia, the initial goal of this research was to determine what range of particle concentrations residents of communities impacted by forest fire smoke could expect to be exposed to while indoors. We have expanded the scope of this project to also assess indoor exposure to residential wood-smoke in Prince George. We will also evaluate the effectiveness of portable air cleaners on reducing indoor levels of smoke. We hope that this study will allow us to provide better advice in the future about ways that people can best reduce exposures in such situations. 143 Appendix H: Media package , UBC C E N T R E FOR H E A L T H & ENVIROMNENT R E S E A R C H C H E R MEDIA RELEASE | AUGUST 2004 U B C Researchers Measure Home Air Quality During Forest Fires How much smoke gets inside your home during a forest fire? That's what researchers at UBC's Centre for Health and Environment Research are currently working to discover in the Williams Lake area. According to Dr Mike Brauer, who is leading the research, "Williams Lake has three air quality monitoring stations that measure outdoor concentrations of smoke, but it's difficult to predict indoor levels from this information. Our research will enable us to tell people what size particulates and how much of them people can expect to be exposed to in their own homes". Forest fire smoke is a mixture of gases, organic compounds, and tiny particles of matter. The particulate matter, often small enough to make it into the tiny air sacs in the lungs, is the greatest health concern during the short term exposure typical of wildfire events. Dr. Brauer continues "Standard advice to protect lung health during times of reduced air quality is to stay indoors and do as little as possible. But until now, very little research has been done on the concentrations of particulate matter that can occur inside homes during a forest fire event. We need to increase our understanding of the public health impacts of forest fire smoke and especially particulate matter so that we can be sure people are getting the best advice to protect their health". Although the weather conditions over the weekend have brought fires in Lonesome Lake, Klinaklini Valley and Hobson Lake under control and air quality has improved, the fires are still burning and smoke levels are still elevated. "As long as there are smoke particles in the air, people, especially those vulnerable to low air quality, will continue to experience health effects" says researcher Prabjit Barn who has already visited five homes in the Williams Lake area to measure indoor air quality, and interview residents about any health effects of smoke. The team is planning to visit a further 15 homes in Williams and is including people who are already living with respiratory problems in the test group to gain insight into the effects of smoke on this vulnerable population. They are also measuring the effectiveness of portable air cleaners on reducing indoor levels of smoke. 147 Appendix I: Absorbance protocol Absorbance Briefly, reflectance measurements were first taken on 5 blank filters. Each filter was taken out of its individual Petri dish and placed on the white reference area on the SSR and reflectance was measured at five different points on the filter using the 5 point measurement system. The sequence and location of these 5 measurement points can be seen below. The average value and standard deviation of the 5 measurements for each filter were calculated, and the filter with the median average value and a standard deviation less than 0.5 was selected as the primary control filter. This filter was again placed on the monitor and the reflectance of the SSR monitor was set to 100. Next each of the sampled filters were removed from their individual Petri dishes, placed on the white reference area and measured using the 5 point measuring system. After measurement, each filter was returned to its Petri dish. The reflectance of the primary control filter was measured every 5 filters to ensure that the monitor did not experience drift and therefore did not affect the measurements obtained. After every 25 filter measurements, the SSR was recalibrated by setting the primary control filter reflectance to 100. A l l filters were handled with tweezers and the monitor was wiped with alcohol prior to and after all measurements. At the end of each session, a quality assurance step was performed by re-measuring a random 10% of the filters. If these re-measurement values were found to differ by more than 3% from the original measurements all filters were measured again. Once all reflectance measurements were taken, the values were entered into a spreadsheet and the mean and standard deviation reflectance for each filter were calculated. An absorbance coefficient was then calculated for each filter using the equation below. A = (A/2V ) n l n (R F /Rs) where a= absorption coefficient A=area of the stain of the filter (m2) V=volume sampled (nr) Rs= reflectance of the sample filter as percentage of Ro (100.0) RF= average reflectance of field blank filters 150 Appendix J: Levoglucosan protocol UBC School of Occupational and Environmental Hygiene (SOEH) Determination of Levoglucosan in Atmospheric Fine Particulate Matter by GC/MS Creation Date: 07/14/05 Method Version: SOEH-SOP# A.00.10 Introduction Levoglucosan (figure 1) is a sugar anhydride and is used as a molecular marker for the presence of wood smoke in air. The components detected in wood smoke are numerous, PAH'S , aldehydres, free radicals and methoxylated phenols, but the detection of levoglucosan has proven to be a reliable indicator for wood combustion from residential fireplaces or forest fires. Solvent extraction of 37 m.m.or 41 m.m. eflon filters (PTFE Membrane W/PMP Ring 2.0 um) with ethyl acetate, derivatization of levoglucosan and subsequent GC/MS analysis is a very selective and sensitive quantitative method. Figure 1: Levoglucosan 1,6-Anhydro-beta-D-glucopyranose (498-07-7) O H . C 6 H , 0 O 5 M.W. = 162.142 Apparatus: GC/MS System - Varian Saturn 2000 Hitachi HiMac centrifuge (CT5DL model) Gelman Teflo™ W/Ring - PTFE Membrane W/PMP Ring: 2.0 um 37 m.m. P/N R2PJ037 Filter Cutter (see Figure 2) Chemicals 1.6- Anhydro-beta-beta-D-glucopyranose - Sigma-Aldrich P/N 316555-1G (99.9% purity) 1,3,5-Tri-Isopropylbenzene (internal standard) - Fluka P/N 92075 (97% purity) 2.7- Anhydro-beta-D-alto heptulopyranose - Sigma-Aldrich P/N S3375 MSTFA + 1% TMCS (N-Methyl-N-trimethylsilyltri-fluoroacetamide + 1% Trimethylchlorosilane) 10 x 1 mL ampules - Pierce Chemicals P/N 48915 152 Pyridine (ACS grade) - Fluka # 82702 - 99.8% purity Ethyl Acetate - Fisher analytical grade Procedure Removal of Teflon portion of the filter Each teflon filter has an outside plastic ring that maintains the teflon filter's round shape. Removing the telfon filter material requires a special tool designed to position and cut out the teflon portion. For 37 m.m. teflon filters place the filter inside a G P M cassette holder and install the support ring. Snug down the support ring to prevent the filter from rotating during the cutting step. Insert the cutting tube and rotate with a downward force. This will cut out the teflon portion of the filter. Using clean forceps transfer the filter to a 5 mL extraction vessel. Prior to the extraction/derivatization procedure spike 50 uL of the stock 7-dehydrocholesterol to each vessel (surrogate standard). Extraction and Derivatization Levoglucosan is light sensitive so take precautions to not expose the sample vials to intense direct light. Transfer 2 mL of ethyl acetate into the extraction vessel and ultrasonicate for 30 mins. Centrifuge only if the samples have high suspended particulate matter. Transfer exactly 100 uL of the final extract into GC vials that have 300 uL inserts installed. Try not to re-suspend the particulates. Add 10 uL of pyridine and 30 uL of MSTFA + 1% TMCS solution. Vortex for 10-20 sees and place the samples in a dark location for a minimum of 6 hours to complete the derivatization. Prior to GC/MS analysis spike 5 uL of trimethyl isopropyl benzene internal standard into each vial. 153 Preparation of Levoglucosan Stock Solution Weigh about 0.010 to 0.030 grams an amount of levoglucosan into an aluminium boat and record precisely the final weight. Transfer to a 50 mL volumetric flask and top up with HPLC grade ethyl acetate. Mix vigourously to dissolve all the crystals and to aid solubilization, ultrasonication can assist in this process. Make sure no solid crystals remain undissolved. The stock solution can be stored at -80 °C. Calculate the final concentration in nanograms per microliter (ng/uL) and record the date of preparation. Preparation of 7-Dehydrocholesterol Surrogate Stock Solution Weigh about 0.010 to 0.030 grams an amount of 7-Dehydrocholesterol into an aluminium boat and record precisely the final weight. Transfer to a 50 mL volumetric flask and top up with HPLC grade ethyl acetate. Mix vigourously to dissolve all the crystals and to aid solubilization, ultrasonication can assist in this process. Make sure no solid crystals remain undissolved. The stock solution can be stored at -80 °C. Calculate the final concentration in nanograms per microliter (ng/uL) and record the date of preparation. Preparation of Trimethylisopropylbenzene Internal Standard Transfer 30 uL of trimethylisopropylbenzene into 25 mL volumetric flask and top up with ethyl acetate. Spike 5 uL of this solution into each GC vial after derivatization is completed. Figure 2 Mass Spectrum of the Trimethylsilyl derivative of Levoglucosan Spectrum 1A BP 73 (8742=100%) levo rt fa 204 min.Scan:769 Channel:! Ion:4792 usRIC: i 200 300 500 iOO i quired Range m/z| Figure 3 Limited Mass Chromatogram of Quantitaion Ion of Levoglucosan-TMS (m/z 333) 154 Ions: 333.0 Merged leva rt dil.sms 2000 CENTROID RAW Limited Moss Chromatogram (m/z 333) 1 Levoglucosan-TMS Bi &\ * i | u • -H*Ri i -IstoA s -j ffl -i i L | * | m i B\S\ Ions: 333.0 Merged rt chk SIS 0.1 .sms 2000 CENTROID RAW Levoglucosan-TMS Limited Mass Chromatogram (m/z 333) Method Revisions Revision Number Author Date Description SOEH-SOP# A.00.16 Timothy Ma 07/14/05 1st Version Appendix K: GC/MS Instrument Parameters for Levoglucosan Analysis GC/MS Instrument Parameters: Injection Vol (uL): 1.0 uL Column: HP-5 30 meters x 0.25 m.m. I.D. (0.25 um film thickness) Gas Type: Helium (UHP Grade) Temperature Program: 65 °C (1 min hold) to 310 °C @ 20 °C/min Final Hold Time (mins): 4.2 min Total Run time(mins): 15 mins Injection Port Splittless Injection Time: 0.5 minutes Temperature: 290 C Single Ion Monitoring Mode (S.I.M.) Delay Time (mins): 4.5 mins S.I.M. Table Start Time (mins) Monitored Ions Dwell Time (msec) Group #1 - Diisopropyl benzene (Internal Standard) 4.5 mins (Istd) 161:189:204: 50 Group #2 - Levoglucosan 6.65 mins 204:217:333 50 Group #3 - 7-Dehydrochlolesterol (Surrogate) 10.55 325:351:456 50 156 Appendix L: Infiltration calculations 1. Import day 1 and day 2 indoor and outdoor pDR data into E X C E L files. Save files as "Filter" indoor/outdoor and "No filter" indoor/outdoor into a "raw data" folder. 2. For "Filter" data, convert indoor and outdoor data from mg/m3 to ug/m3 (multiply values by 1000) 3. Calculate 30 min running averages for indoor and outdoor data: • Create two new columns and label "30 min avg indoor" and "30 min avg outdoor" • Calculate an average for every 30 rows of data 4. Identify indoor generated peaks: • Create new "sort" column to assign "1" for every 30 t h value (ie. every half hour) and "0" for every other row (use calculation: =fF(D34=0,l,IF(D34=30,l,0)) where column D is column with minutes) • Copy and paste "30 min avg indoor," "30 min avg outdoor" and "sort" column into new worksheet. Label worksheet as "Peaks 1", • Highlight entire worksheet and sort "sort" column by descending order. Delete all rows of data with "0" value in "sort" column. • Delete all values in "sort" column • Copy and paste values from first column (indoor concentrations) into this column starting from the second row. Label this column as "t-1 indoor data" • Delete values in first row of first two columns and label them as "t indoor" and "t outdoor" • Create 4 new columns next to "t-1 indoor" and label them in the following order: "in (50%)," "out (50%)," "out (30%)," and "out (10%)." For "in (50%)" column insert following formula into cell and fill down: =IF(((A3-A2)/A3)>0.5,1,0). For "out" columns fill in following formula and change to 0.3 and 0.10 for 30% and 10% columns respectively: =IF(((B3-B2)/B3)>0.5,1,0). This will assign any row where the concentration is a certain percentage higher than the value before as a "1." a. For a given row, if a "1" is assigned to the "in (50%)" column and any of the "out" columns, it will be considered as a peak due to infiltration. b. For a given row, if a "1" is assigned to only the "in (50%)" column and not for any of the "out" columns, it will be considered as in indoor generated peak. • Highlight these rows identified as indoor generated peaks as well as subsequent rows with rising concentrations (ie. the rising edge of the peak) • Copy and paste "t indoor," "t outdoor," and "t-1 indoor" columns into a new worksheet and label it as "Peaks 1-B." 5. Remove indoor generated peaks: • Delete highlighted rows and make note in worksheet of rows deleted. 6. Calculate Fjnf: Linear regression will be used to calculate Fjnf. The penetration and decay terms (ai and &2 respectively) will be estimated using the LINEST function in E X C E L . The equation Fjnf= ai/(l-a2) will then be used to calculate infiltration from the regression output. 157 Create 3 new columns in "Peaks 1-B" worksheet and label in following order: a2, ai, and intercept • In the a 2 column enter the following equation: LINEST(A2:A77,B2:C77,TRUE,TRUE) • Highlight 5 rows in each of the 3 new columns and press alt, control and enter keys simultaneously. The following output will be generated: a 2 A intercept S E a 2 S E a 1 SEjnterceDt r2 SEjndoor F df SS r e a • In a row below the output type Fjnf and next column over in the same row, enter the following equation: F2/(l-E2) • In the row underneath, type in 95% CI and enter in the next column over in the same row enter the following equation: SQRT(F3A2*76*(1/(1-E2))A2+ (F2/(l-E2)A2)A2*E3A2*76)/SQRT(76)* 1.96. 7. Repeat steps 2-6 for "No filter" measurements. Label worksheets as "Peaks 2" and "Peaks 2-B" 158 Appendix M: Concentration graphs 9 11 13 15 17 19 21 23 25 27 29 31 33 Time (30 min intervals) SB-01 NO FILTER 1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 Time (30 min intervals) SB-02 FILTER 5 7 9 11 13 15 17 19 21 Time (30 minute intervals) SR-01 FILTER 1 4 7 10 13 16 19 22 25 28 31 34 37 40 Time (30 min intervals) 200 PG-01 FILTER 1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 Time (30min intervals) SB-02 NO RLTER 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 Time (30 min intervals) SR-01 NO HLTER 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 Time (30 min intervals) PG-03 FILTER 11 13 15 17 19 21 23 25 27 29 31 Time (30 min intervals) | 120 100 80 60 40 20 0 PG-03 NO FILTER - Indoor - Outdoor 1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 Time (30 min intervals) PG-04 FILTER 200 . ~ 150 c p> c •& 100 8 3-50 - Indoor - Outdoor J 4 7 10 13 16 19 22 25 28 31 34 37 40 43 Time (30 min intervals) 8 3 PG-04 NO FILTER 5 7 9 11 13 15 17 19 21 23 25 27 29 31 Time (30 min intervals) PG-05 HLTER 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 Time (30 min intervals) PG-05 NO RLTER - Indoor - Outdoor 1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 Time (30 min intervals) c o " E PG-06 FILTER 1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 Time (30 min intervals) PG-06 NO RLTER 10 13 16 19 22 25 28 31 34 37 40 43 Time (30 min intervals) PG-07 RLTER 250 1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 Time (30 min intervals) c CD Q> C 250 PG-07 NO RLTER - Indoor ~ Outdoor 1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 Time (30 min intervals) O 3 PG-12 R L T E R Indoor Outdoor 1 6 11 16 21 26 31 36 41 46 61 56 61 66 71 76 81 Time (30 min intervals) PG-12 N O R L T E R 8 3 1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 Time (30 min intervals) 164 PG-13 HLTER 6 11 16 21 2 6 31 36 41 4 6 51 5 6 61 6 6 71 7 6 Time (30 min intervals) c CO « c u S PG-13 NO FILTER 70 60 5 0 4 0 g ? 3 0 ei 10 4 - Indoor - Outdoor 4 7 10 13 16 19 22 2 5 2 8 31 3 4 3 7 4 0 Time (30 min intervals) 100 80 60 40 20 0 PG-14 FILTER - Indoor - Outdoor r ' _ # i_jWr: 6 11 16 21 2 6 31 3 6 41 4 6 51 5 6 61 6 6 71 7 6 Time (30 min intervals) 8 I 8 3 in 100 80 60 40 20 0 PG-14 NO FILTER Indoor Outdoor • r f 7 10 13 16 1 9 22 2 5 2 8 31 34 3 7 4 0 4 3 Time (30min intervals) PG-15 FILTER - Indoor - Outdoor 1 6 11 16 21 2 6 31 3 6 41 4 6 51 5 6 61 6 6 71 76 81 Time (30 min intervals) c CO o £ PG-15 NO FILTER - Indoor - Outdoor j r \tf*" 1 4 7 1 0 13 16 1 9 2 2 2 5 2 8 31 3 4 3 7 4 0 4 3 Time (30 m in intervals) PG-01A FILTER - Indoor - Outdoor I U 1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 Time (30 min intervals) 8 i PG20 FILTER - Indoor - Outdoor 4 7 10 13 16 19 22 25 28 31 34 37 40 43 Time (30 min intervals) c CO o c o < 5 °> S 3 in oj E 250 200 150 100 50 0 PG-09A HLTER - kidoor Outdoor T / ' * M « U « 4 7 10 13 16 19 22 25 28 31 34 37 40 43 Time (30 min intervals) PG-01 A NO FILTER - Indoor - Outdoor c 200 o I 150 c m 8 3 in 50 n i o 4 7 10 13 16 19 22 25 28 31 34 37 40 43 Time (30 min intervals) PG-20 NO FILTER 1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 Time (30 min intervals) u -5 8 3. 200 150 100 50 0 PG-09A NO FILTER Indoor Outdoor I 5 _ J 1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 Tim e (30 m in intervals) 166 Appendix N: Relative humidity scatter plots ON oo PM2.5 concentration (ug/m3) ro ui o ui o o o o o o o -I ! ! 1 1 PM2.5 concentration (ug/m3) ro -t* CO Co o ro o o o o o o o m J i — i — i i i i O 1 PM2.5 concentration (ug/m3) - ' - ' M M OI O OI O OI o o o o o o o PM2.5 concentration (ug/m3) ^ ro ro oi o oi o ui o o o o o o ° 1 Appendix O: Infiltration factors (Finf) by home Home Season Finf with filter SE Finf with no filter SE WL-01 Summer 0.66 0.53 WL-03 Summer 0.32 0.16 0.31 0.36 WL-04 Summer 0.03 0.04 0.30 0.15 WL-05 Summer 0.07 0.28 0.41 0.26 LY-02 Summer 0.37 0.92 0.77 0.28 LY-03 Summer 0.07 0.07 0.98 0.46 LY-04 Summer 0.67 0.08 LY-05 Summer 0.01 0.15 0.92 5.51 LY-06 Summer 0.32 0.20 0.40 0.38 LY-07 Summer 0.03 0.23 0.43 0.34 SB-01 Summer 0.61 0.34 1.10 0.45 SB-02 Summer 0.44 0.38 SR-01 Summer 0.11 0.20 0.43 0.15 PG-01 Winter 0.04 0.06 PG-02 Winter 0.17 0.09 0.20 0.10 PG-03 Winter 0.17 0.07 0.36 0.11 PG-04 Winter 0.30 0.07 0.24 0.11 PG-05 Winter 0.03 0.03 0.20 0.13 PG-06 Winter 0.13 0.05 0.11 0.16 PG-08 Winter 0.04 0.08 0.15 0.08 PG-09 Winter 0.03 0.24 PG-10 Winter 0.02 0.34 0.10 0.45 PG-11 Winter 0.12 0.08 PG-12 Winter 0.01 0.04 0.68 1.00 PG-13 Winter 0.15 0.06 0.50 2.49 PG-14 Winter 0.05 0.06 0.10 0.09 PG-15 Winter 0.22 0.08 0.28 0.17 PG-16 Winter 0.03 0.08 0.12 0.12 PG-18 Winter 0.16 0.98 0.25 0.52 PG-01 A Winter 0.07 0.01 0.47 0.66 PG-20 Winter 0.13 0.38 0.55 0.28 Appendix P: Air cleaner efficiency values by home Home Season Air cleaner efficiency (%) WL-03 Summer -3.23 WL-04 Summer 90.00 WL-05 Summer 82.93 LY-02 Summer 51.95 LY-03 Summer 92.86 LY-05 Summer 98.91 LY-06 Summer 20.00 LY-07 Summer 93.02 SB-01 Summer 44.55 SR-01 Summer 74.42 PG-02 Winter 15.00 PG-03 Winter 52.78 PG-04 Winter -25.00 PG-05 Winter 85.00 PG-06 Winter -18.18 PG-08 Winter 73.33 PG-10 Winter 80.00 PG-12 Winter 98.53 PG-13 Winter 70.00 PG-14 Winter 50.00 PG-15 Winter 21.43 PG-16 Winter 75.00 PG-18 Winter 36.00 PG-01A Winter 85.11 PG-20 Winter 76.36 PG-09A Winter 80.00 Appendix Q: Infiltration modeling variable associations Summer data: Categorical and continuous variables Dependent variable Independent variable p-value Number of windows AC use 0.3485 Pooled data: Continuous variables Air exchange rate Age of home Number of windows Air exchange rate 1.0000 Age of home -0.1085 . 1.0000 Number of windows 0.0747 -0.0822 1.0000 Categorical and continuous variables: Dependent variable Independent variable p-value Age of home season 0.0424 ac use 0.8735 window use 0.0353 Number of windows season 0.2022 ac use 0.1772 window use 0.1171 Air exchange rate season 0.0304 ac use 0.5317 window use 0.2241 Categorical variables Variables p-value Season 0.1641 Ac use Season 0.0001 Window use Ac use 0.1971 Window use 172 Appendix R: Air cleaner efficiency modeling variable associations Winter data: Continuous and categorical variables Dependent variable Independent variable p-value Age of home Carpeting 0.3574 Pooled data: Continuous and categorical variables Dependent variable Independent variable p-value Age of home Carpeting 0.4763 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            data-media="{[{embed.selectedMedia}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.831.1-0100824/manifest

Comment

Related Items