UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Monitoring residential woodsmoke in British Columbia communities. Wagstaff, Matthew 2018

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

Item Metadata

Download

Media
24-ubc_2018_september_wagstaff_matthew.pdf [ 9.18MB ]
Metadata
JSON: 24-1.0371217.json
JSON-LD: 24-1.0371217-ld.json
RDF/XML (Pretty): 24-1.0371217-rdf.xml
RDF/JSON: 24-1.0371217-rdf.json
Turtle: 24-1.0371217-turtle.txt
N-Triples: 24-1.0371217-rdf-ntriples.txt
Original Record: 24-1.0371217-source.json
Full Text
24-1.0371217-fulltext.txt
Citation
24-1.0371217.ris

Full Text

Monitoring Residential Woodsmoke in British Columbia Communities by  Matthew Wagstaff  B.Sc., The University of British Columbia, 2014  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  MASTER OF SCIENCE in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Occupational and Environmental Hygiene)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  August 2018  © Matthew Wagstaff, 2018 ii  The following individuals certify that they have read, and recommend to the Faculty of Graduate and Postdoctoral Studies for acceptance, a thesis/dissertation entitled:  Monitoring Residential Woodsmoke in British Columbia Communities  submitted by Matthew Wagstaff in partial fulfillment of the requirements for the degree of Master of Science in Occupational and Environmental Hygiene  Examining Committee: Dr. Michael Brauer      (University of British Columbia) Co-supervisor Dr. Sarah Henderson   (University of British Columbia and BC Centre for Disease Control) Co-supervisor  Dr. Arvind Saraswat     (BC Ministry of Environment and Climate Change Strategy) Supervisory Committee Member Dr. Hugh Davies            (University of British Columbia) Additional Examiner  iii  Abstract Wood burning is a common home heating method in many communities in British Columbia and an important source of fine particulate matter (PM2.5) air pollution. During winter months communities impacted by residential woodsmoke experience high concentrations of PM2.5, at levels that have been associated with a wide range of health effects. Characterising levels of woodsmoke within and between communities can support air quality management and reduction of exposures.  This project tested novel methods to measure the relative levels and spatial variability of residential woodsmoke PM2.5 using fixed and mobile optical instruments. The methods were applied during the winter heating season (January 5th to March 2nd, 2017) across three communities identified to be impacted by residential woodsmoke from fixed-site monitoring data, and three paired communities without routine monitoring. Continuous monitoring was performed for two weeks at fixed monitoring stations in each monitored community to compare the optical instruments with established methods used to measure PM2.5 and woodsmoke. This was combined with nightly mobile monitoring using the same optical instruments, alternating between driving routes around the paired monitored and unmonitored communities to create detailed maps describing woodsmoke levels and variability. The nephelometer (Bsp) and aethalometer (delta C) tested at the fixed-site were strongly correlated with conventional methods of measuring PM2.5 (beta attenuation monitor and filter-based) and woodsmoke (levoglucosan). Comparisons between the instruments during mobile monitoring clearly identified times and areas where woodsmoke was dominating PM2.5 concentrations. Mobile monitoring indicated considerable spatial variation across all communities and identified hotspot areas with consistently elevated concentrations of both PM2.5 and woodsmoke. The spatial variance of PM2.5 concentrations was significantly greater than the temporal variance during 71% of the runs, demonstrating the importance of understanding iv  spatial variability when monitoring the air quality impacts of woodsmoke. Strong woodsmoke impacts were found in each community. In general, the unmonitored communities had PM2.5 concentrations that were similar to or higher than their partnered monitored communities, despite having smaller sizes and populations. The development of this approach allows for detailed and cost-effective characterisation of woodsmoke in monitored and unmonitored communities, which could inform source control efforts in many Canadian communities.     v  Lay Summary Wood burning is used to heat many homes in British Columbia (BC). However, residential wood burning can degrade community air quality, and cause negative health effects. This project tested a mobile monitoring method to map air quality around BC communities, and to understand the contributions of residential woodsmoke. Testing was conducted between January and March 2017 in three pairs of communities on the coast, in the mountains, and in the northern interior. The instruments used in the mobile monitoring compared well to conventional methods for measuring woodsmoke, while mobile measurements identified specific times and areas within each community where woodsmoke caused degraded air quality. The maps showed clear variation in the air quality within each community, including smoke hotspots. This mobile monitoring method will allow for detailed and cost-effective understanding of woodsmoke in communities concerned about the air quality impacts of residential wood burning. vi  Preface The testing of the mobile monitoring method using both a nephelometer and aethalometer, along with temporal comparison of these instruments with filter-based measurements of PM2.5 and levoglucosan concentrations at BC Ministry of Environment and Climate Change Strategy (ENV) monitoring stations was originally proposed to me by my supervisors and collaborators. However, I contributed to writing a proposal which led to project funding from the Clean Air Research Agenda at Health Canada. Details of the study such as locations and timelines of the fixed monitoring campaign, along with the specific methods used in mobile monitoring were planned and decided upon by myself and my co-supervisors Drs. Michael Brauer and Sarah Henderson.  Field data collection was almost exclusively carried out by myself, apart from some support from ENV technicians in accessing the monitoring stations in each of the three locations and the initial placing of the fixed site equipment. I was fully responsible for instrument preparation and calibration prior to field monitoring in each region, daily filter changes and instrument checks at the monitoring stations, and I performed all mobile monitoring alone. Data collected by ENV monitors at these stations were also accessed and used in this study.  Gravimetric filters collected in the field campaign were analysed for PM2.5 and levoglucosan concentrations at the UBC Occupational and Environmental Hygiene lab by its manager, Matty Jeronimo. All further data analysis was conducted by myself, with specific decisions on analysis supported by my supervisors.  Results of this research have been made public in community-specific reports and will be published in the peer-reviewed literature in the future.    vii  Table of Contents  Abstract ............................................................................................................................... iii Lay Summary ......................................................................................................................... v Preface ..................................................................................................................................vi Table of Contents ................................................................................................................. vii List of Tables ......................................................................................................................... xi List of Figures ....................................................................................................................... xii List of Abbreviations ............................................................................................................ xiv Acknowledgements .............................................................................................................. xv Chapter 1: Introduction .................................................................................................................. 1 1.1 Residential Woodsmoke in British Columbia .............................................................. 1 1.2 Ambient Air Quality Impacts ....................................................................................... 2 1.3 Health Impacts ............................................................................................................ 3 1.4 Limitations of Existing Monitoring .............................................................................. 5 1.5 Measurement Strategies for Residential Woodsmoke............................................... 6 1.5.1 Surveys and Emissions Inventories ......................................................................... 6 1.5.2 Studying Temporal Patterns ................................................................................... 7 1.5.3 Source Apportionment Using Tracer Compounds .................................................. 7 1.5.4 Source Apportionment Using Optical Properties ................................................... 7 1.5.5 Measuring Spatial Patterns ..................................................................................... 8 1.6 Project Overview ......................................................................................................... 9 viii  1.6.1 Project Objectives ................................................................................................... 9 1.6.2 Research Questions .............................................................................................. 10 Chapter 2: Methods ..................................................................................................................... 11 2.1 Site Selection ............................................................................................................. 11 2.2 Timeline of Field Monitoring .................................................................................... 12 2.3 Equipment Overview ................................................................................................ 13 2.3.1 ENV PM2.5 Monitoring Instruments ...................................................................... 13 2.3.2 Nephelometers ..................................................................................................... 13 2.3.3 Aethalometers ...................................................................................................... 14 2.3.4 Harvard Impactors ................................................................................................ 15 2.4 Fixed Site Monitoring ................................................................................................ 16 2.4.1 Instrument Data Collection ................................................................................... 16 2.4.2 Filters and Levoglucosan Analysis ......................................................................... 19 2.5 Mobile Monitoring .................................................................................................... 19 2.5.1 Route Creation and Monitoring Schedule ............................................................ 19 2.5.2 Equipment Set-Up and Operation ........................................................................ 20 2.6 Data Cleaning ............................................................................................................ 23 2.7 Fixed Site Analyses .................................................................................................... 24 2.7.1 Temporal Matching Between Fixed Location Instruments .................................. 24 2.7.2 Exploration of Temporal Patterns ......................................................................... 25 2.7.3 Instrument Comparisons ...................................................................................... 25 2.8 Mobile Data Analysis and Map Production .............................................................. 26 ix  2.8.1 Comparison of Spatial and Temporal Variance .................................................... 28 Chapter 3: Results ........................................................................................................................ 30 3.1 Summary of Temporal Patterns in the Monitored Communities ............................. 30 3.2 Comparison of PM2.5 Measures at the Fixed Sites .................................................... 34 3.3 Comparison of Woodsmoke-Specific Measures at the Fixed Sites .......................... 35 3.4 Comparison of Woodsmoke and Total PM2.5 Measures at the Fixed Sites .............. 36 3.4.1 Daily Differences in Woodsmoke Contribution .................................................... 38 3.5 Comparison of Woodsmoke and Total PM2.5 Measures during Mobile       Monitoring ................................................................................................................ 40 3.6 Average Conditions During Mobile Monitoring Runs ............................................... 43 3.7 Route Average Maps ................................................................................................. 46 3.7.1 Whistler and Pemberton Routes .......................................................................... 46 3.7.2 Courtenay-Cumberland and Courtenay-Comox Routes ....................................... 47 3.7.3 Vanderhoof and Fraser Lake Routes ..................................................................... 47 3.8 Comparison of Spatial and Temporal Variance ........................................................ 55 Chapter 4: Discussion ................................................................................................................... 57 4.1 Summary of Key Findings .......................................................................................... 57 4.1.1 Primary Objective - Testing the Mobile Monitoring Method ............................... 57 4.1.2 Secondary Objective - Community-Specific Findings ........................................... 57 4.2 Can a Single-Channel Nephelometer be Used to Estimate PM2.5             Concentrations? ........................................................................................................ 58 4.3 Can a Dual-Channel Aethalometer be Used to Estimate Woodsmoke Concentrations? ........................................................................................................ 59 x  4.4 Can the Two Optical Instruments be Used Together to Estimate Woodsmoke Contribution to PM2.5 Concentrations? .................................................................... 61 4.5 What Information Can Mobile Monitoring Add to the Understanding of     Residential Woodsmoke in Communities? ............................................................... 63 4.6 Is There Value in Using an Aethalometer in Addition to a Nephelometer           During Mobile Monitoring? ...................................................................................... 63 4.7 What Did We Learn About Woodsmoke in the Monitored Communities? ............. 64 4.8 What Did We Learn About the Spatial Patterns Across Each Community? ............. 66 4.8.1 Whistler and Pemberton....................................................................................... 66 4.8.2 Comox Valley Communities .................................................................................. 67 4.8.3 Vanderhoof and Fraser Lake ................................................................................. 68 4.9 What Did We Learn About Woodsmoke in the Unmonitored Communities? ......... 70 4.10 Conclusions on Mobile Method ................................................................................ 71 4.10.1 Strengths of the Method .................................................................................. 71 4.10.2 Limitations of the Method ................................................................................ 72 Chapter 5: Conclusion .................................................................................................................. 75 References .......................................................................................................................... 77 Appendices ......................................................................................................................... 84 Appendix A - Woodsmoke Mobile Monitoring: Full Protocol ........................................................ 84 Appendix B - List of Mobile Monitoring Runs .............................................................................. 140 Appendix C - Levoglucosan Analysis Procedure ........................................................................... 143 Appendix D - R Functions ............................................................................................................. 149  xi  List of Tables Table 2-1: Field monitoring timeline. ........................................................................................... 13 Table 3-1: Average conditions at the three monitoring stations during the field monitoring campaign. ...................................................................................................................................... 30 Table 3-2: Average conditions at the monitoring stations during the nighttime monitoring runs on each route. ............................................................................................................................... 44 Table 3-3: Results of the Fligner-Killeen test for homogeneity of variance used to compare the spatial variance measured during the nighttime mobile monitoring runs, and the temporal variance measured at the fixed location during the same time period. ...................................... 56  xii  List of Figures Figure 2-1: Locations of study communities in British Columbia. ................................................ 12 Figure 2-2: Nephelometer function diagram. ............................................................................... 14 Figure 2-3: Study fixed-site equipment in Whistler. ..................................................................... 17 Figure 2-4: Harvard Impactor........................................................................................................ 18 Figure 2-5: Monitoring instruments installed in mobile vehicle. ................................................. 21 Figure 2-6: Vehicle prepared for mobile monitoring. ................................................................... 22 Figure 2-7: Orientation of sample inlets. ...................................................................................... 22 Figure 3-1: Time series plot showing 24-hour average values of the five monitoring methods operated at the monitoring stations. ........................................................................................... 32 Figure 3-2: Time series plots showing the average daily patterns measured by the fixed site instruments during the monitoring campaign in each of the three monitored communities. .... 33 Figure 3-3: Scatterplot comparison of calculated hourly averages of Bsp as measured by the nephelometer and the PM2.5 concentrations reported by the beta attenuation monitor      (BAM) at each monitoring station. ............................................................................................... 34 Figure 3-4: Scatterplot comparison of the calculated 24-hour averages of Bsp as measured by the nephelometer and the 24-hour average PM2.5 concentrations measured by the filter samples at each monitoring station. ............................................................................................ 35 Figure 3-5: Scatterplot comparison of the two woodsmoke-specific measures, calculated         24-hour averages of delta C as measured by the aethalometer, and the 24-hour average levoglucosan concentrations measured by the filter samples at each monitoring station. ........ 36 Figure 3-6: Scatterplot comparison of the two measurements made from the daily filter samples; woodsmoke tracer levoglucosan concentrations against total PM2.5 concentrations. 37 xiii  Figure 3-7: Scatterplot comparison of 1-hour averages of the two optical instruments     installed at the fixed locations; woodsmoke delta C as measured by the aethalometer,      against the Bsp measured by the nephelometer. .......................................................................... 38 Figure 3-8: Scatterplot comparison of 1-hour averages of delta C and Bsp measured at the monitoring stations during: a) Day Hours between 09:00 and 17:00, and b) Night Hours between 17:00 to 09:00. ............................................................................................................... 39 Figure 3-9: Scatterplot comparisons of 1-minute averages of the delta C woodsmoke      indicator and the Bsp total PM2.5 indicator measured by the mobile instruments during the monitoring runs. ........................................................................................................................... 41 Figure 3-10: Time series comparing responses of the mobile aethalometer and nephelometer during two monitoring runs of the Fraser Lake route. ................................................................. 42 Figure 3-11: Wind roses during the nighttime runs on each route. ............................................. 45 Figure 3-12: Route average maps of the seven nighttime runs across the Whistler route. ........ 49 Figure 3-13: Route average maps of the seven nighttime runs across the Pemberton route. .... 50 Figure 3-14: Route average maps of the seven nighttime runs across the Courtenay- Cumberland route. ........................................................................................................................ 51 Figure 3-15: Route average maps of the seven nighttime runs across the Courtenay-Comox route. ............................................................................................................................................. 52 Figure 3-16: Route average maps of the six nighttime runs across the Vanderhoof route. ........ 53 Figure 3-17: Route average maps of the seven nighttime runs across the Fraser Lake route. .... 54  xiv  List of Abbreviations BAM = Beta attenuation monitor BC = British Columbia Bsp = Particle light scattering (measurement reported by a nephelometer)  °C = Degrees Celsius, measure of temperature CARA = Clean Air Research Agenda at Health Canada CCD = Courtenay and Cumberland monitoring route CCX = Courtenay and Comox monitoring route CSA = Canadian Standards Agency ENV = The British Columbia Ministry of Environment and Climate Change Strategy EPA = United States Environmental Protection Agency Mm-1 = Inverse mega-meters, reporting units of Bsp m/s = Metres per second, measure of wind speed PM = Particulate matter PM2.5 = Fine particulate matter (particulate matter with an aerodynamic diameter of less than 2.5 microns) R2 = Coefficient of determination SD = Arithmetic standard deviation UBC = The University of British Columbia Z = Mean z-score μg/m3 = Micrograms per cubic meter, measure of concentration xv  Acknowledgements This thesis project was made possible with funding from the Clean Air Research Agenda (CARA) at Health Canada, and equipment funding provided by The British Columbia (BC) Lung Association. My work on this project was supported by my UBC supervisors, along with partners from Health Canada, The BC Ministry of Environment and Climate Change Strategy (ENV), The BC Lung Association, and the BC Centre for Disease Control (BCCDC). I am also very grateful for the support of the Elizabeth Henry Scholarship for Environmental Health Research (managed by the Fraser Basin Council) which I received as part of this work.  I would like to first thank my co-supervisors Dr. Michael Brauer (UBC) and Dr. Sarah Henderson (UBC and the BCCDC) and express my sincere gratitude for their support and guidance throughout this project; I could not have asked for better supervisors.  In addition to my supervisors, I would also like to express my thanks to the following people: • My committee member Arvind Saraswat (ENV). • The main collaborators on the CARA funded project: Nina Dobbin (Health Canada), Natalie Suzuki (ENV), and Dr. Scott Weichenthal (McGill). • The UBC Occupational and Environmental Hygiene Lab Manager Matty Jeronimo for lab analysis, support and instrument troubleshooting.  • The BC Ministry of Environment staff in each region for the invaluable support I received during my field work: Chris Marsh, Graham Veale, Lorne Nicklason, Earle Plain, Lee Loran, Ben Weinstein and Gail Roth. • Community members in each region who provided feedback on monitoring routes.  • Members of the Brauer/Henderson lab group for troubleshooting support. • Kathleen McLean (BCCDC) for developing an automated app that will allow future users of this mobile method to easily map their data. Finally, and most importantly I would like to thank both my parents for their never-ending support, and especially Dallas who I would not have been able to complete this project without! From digging through snowbanks to help set up the equipment in Whistler, right through to reading this thesis for clarity, she was there every step of the way.  1  Chapter 1: Introduction  Residential Woodsmoke in British Columbia Approximately 10% of British Columbia (BC) homes outside of the Metro Vancouver area use wood as their primary source of fuel during the heating season, with a further 20% burning wood in the home at least some of the time (1). Wood is plentiful and inexpensive in many parts of the province, which makes it an attractive heating option when compared with alternatives such as natural gas, electricity, or heating oil. However, air emissions from burning wood are much higher and more variable than emissions from other types of fuels. Woodsmoke is a complex mixture of gases and particles, the specific composition of which depends on multiple factors including the wood species, its water content, and the combustion temperature (2). From a human health perspective, the most important products of incomplete combustion are fine particulate matter (PM2.5), oxygenated organics (e.g. aldehydes), hydrocarbons (e.g. polycyclic aromatic hydrocarbons), and carbon monoxide (3).   There are three main types of wood burning appliances used for residential heating in BC: indoor stoves; indoor fireplaces; and outdoor boilers. Each type has a range of emission profiles depending on its design and the behaviour of its users. A BC Ministry of Environment and Climate Change Strategy (ENV) report from 2012 on the use of wood burning appliances in the province found that woodstoves were most common (58% of users), followed by open fireplaces (42% of users), which typically generate more smoke than burning in a closed stove (1,4). By regulation, smoke emissions from modern stoves are lower than emissions from older stoves when the newer technology is correctly used (5). However, comparisons of emissions between modern and older stoves have been mixed during indoor air quality studies because other factors affect the smoke emitted into the indoor environment, such as the user behaviour and fuel properties (6,7). The ENV has been operating a woodstove changeout program over the past decade to encourage the use of cleaner burning technology, and a recent provincial survey found that 40% of respondents using fireplaces and 71% of respondents using wood stoves were aware that their appliances were certified as low emissions (1). While not explicitly 2  stated, the certification of these appliances is likely based on the standards set by the Canadian Standards Association (CSA) or the US Environmental Protection Agency (EPA) before 1994. The provincial 1994 Solid Fuel Burning Domestic Appliance Regulation prevented the sale or re-sale of any appliances that did not meet these standards after November 1, 1994 (8). This regulation was updated in November 2016, to improve the standards so that nearly all wood burning appliances sold in BC must now be certified to meet either the new lower emissions standards set by the EPA in 2015, or the equivalent standards set by the CSA in 2010, which represent a 40% reduction to the standards set in 1992 (9). The updated regulation also specifies what fuels may be burnt in wood burning appliances and contains provisions regarding the sale and installation of outdoor wood boilers. These regulations only apply to the sale of new appliances; existing wood stoves and fireplaces are not affected.  Further regulation of wood burning appliance use falls to individual municipal governments, where the scope and strength of regulations can vary considerably from very limited, to more involved regulation. One example of municipal regulations is the City of Duncan’s Wood Burning Appliances and Air Quality Bylaw No.3089, 2013 which requires the removal of all existing non-certified woodburning appliances during the transfer of any property, restricts the use of all wood burning appliances when a provincial Air Quality Advisory is in effect (with the exception of homes with no alternative heating method) and requires new constructions with wood burning appliances to also install a secondary space heating method (10).   Ambient Air Quality Impacts Woodsmoke is complex, but its impacts on ambient air quality are typically seen in routine monitoring of PM2.5 concentrations during the heating season. Although PM2.5 measurements reflect contributions from all sources, the impacts of woodsmoke are made evident by a diurnal pattern with higher concentrations in the mornings and evenings when residents are more likely to be at home and when the atmospheric mixing height is reduced (11,12). Residential woodsmoke can lead to episodes of severely degraded air quality, especially in valleys where nighttime temperature inversions are common and woodsmoke can be trapped (13).  Estimates 3  suggest that smoke from residential wood burning is the second largest anthropogenic contributor to PM2.5 emissions in BC following road dust, contributing approximately 15% of total emissions and exceeding emissions from all transportation and industry sectors (14). Smaller, more rural communities are more affected than the larger urban areas (1). As a result of residential woodsmoke, several communities in BC fail to meet the annual PM2.5 air quality objective of 8 µg/m3 set by the provincial government (15).  The location of residential woodsmoke emissions is an important consideration with respect to community exposures. Residential woodsmoke is emitted from many small point sources directly within communities where people spend the majority of their time. The distribution and relative size of the point sources across a community combined with topographical features and meteorological conditions can lead to high spatial variability across small areas. These factors can create hotspots where PM2.5 concentrations are elevated above background levels. Previous mobile monitoring campaigns in northern and coastal BC have measured and mapped this spatial variability across many communities (16–19). Despite being an important source of PM2.5 in BC, residential woodsmoke is relatively under-regulated compared with mobile and industrial emissions. While evidence such as emissions inventories and temporal patterns in air quality monitoring data suggests that woodsmoke degrades air quality in many areas during the winter, municipalities that hold the power to pass new or stricter regulations are often hesitant to take further action without empirical evidence that smoke is affecting local air quality (20).  Health Impacts Exposure to ambient PM2.5 has been associated with a wide range of acute and chronic health outcomes, particularly cardiovascular and respiratory morbidity and mortality (21,22). Globally, an estimated 3.2 million deaths can be attributed to ambient PM2.5 exposures, of which 2.1 million would be avoided if the World Health Organisation guideline for annual average PM2.5 of 10 µg/m3 was met worldwide (23). There is currently no evidence for a threshold level below which no adverse health effects occur; even at relatively low concentrations PM2.5 pollution has 4  a significant burden on health (24). Therefore, regions that have relatively clean air can still benefit when PM2.5 emissions are further reduced (23).   Although much of the evidence on the health effects of PM2.5 exposure comes from urban environments where industrial and mobile sources may dominate, there is a growing body of literature specific to woodsmoke. The most comprehensive review of available literature concluded that woodsmoke PM2.5 appears to have acute and chronic health effects that are similar to PM2.5 from other combustion sources (3). Evidence for the respiratory outcomes has been stronger and more consistent than evidence for the cardiovascular outcomes, but the most recent review highlighted multiple studies that show improvements in cardiovascular indicators when woodsmoke exposure was reduced (25).  A more targeted review of the physiochemical properties of woodsmoke PM2.5 from residential wood burning has also been conducted. It considers how the particles change under various combustion conditions and over time, and the potential significance of the observed changes in terms of health effects (26). The review concluded that particles produced during inefficient combustion (typical of fireplaces and older wood stoves) have a larger organic component and likely have higher lung deposition efficiencies. Particles produced during efficient combustion (typical of the more advanced wood and pellet stoves) tend to be dominated by the inorganic component, which deposits less efficiently in the respiratory tract due to rapid coagulation of the particles. However, the high water solubility of the inorganic fraction may play an important role in biological effects at the cellular level (26). Recent research has also shown that the lung toxicity and mutagenic potency of PM2.5 created by wood burning varies with different wood species and in different combustion conditions (27).  In addition to the health effects observed in the general population, studies have suggested that some groups may be more susceptible to the effects of ambient PM2.5 exposure (22). Individuals suffering from pre-existing conditions such as diabetes and chronic respiratory or cardiovascular diseases are more susceptible (28), while the elderly are at higher risk of hospitalisation and death. Children also face higher risk because their respiratory systems are 5  developing and they breathe more air relative to their body weight (22). These populations are also likely to spend the majority of their time within a community, potentially in close proximity to emission sources, and where hotspots of elevated PM2.5 concentrations may be present.   Limitations of Existing Monitoring Regulatory air quality monitoring networks, such as the BC network operated across the province by ENV and Metro Vancouver, collect valuable data, but have two major limitations with respect to monitoring residential woodsmoke: 1. They cannot provide source-specific information 2. They cannot provide spatially resolved information The instruments installed for regulatory assessment of air quality are chosen to monitor the criteria air pollutants (such as PM2.5), but no source-specific information is collected. Therefore, the data from these networks can identify communities that have relatively high PM2.5 concentrations, but they cannot provide direct information on which sources are responsible for the higher concentrations. Routinely measured PM2.5 cannot be used in isolation to accurately characterise the air quality impacts of residential woodsmoke. Because combined contributions from all sources are measured, more source-specific information is needed to understand the independent impacts of woodsmoke in affected communities. To date there is no systematic best practice for conducting woodsmoke monitoring in Canadian communities with or without regulatory PM2.5 stations, though many elements of such a system have been tested within the BC context.  The second limitation is spatial resolution on both regional and local scales. From an economic standpoint, the number of fixed monitoring stations that can be installed and maintained is limited. On a regional scale, this means that larger communities are prioritised, and smaller communities have limited or no data. On a local scale, this means that most monitored communities (excluding large cities) have a single monitoring station installed at a location that is chosen to be approximately representative of the average air quality.  6  This lack of spatial resolution is especially important in the case of residential woodsmoke. On the regional scale, smaller and more rural communities generally have higher rates of residential wood burning (due to a lack of alternative heating options), which may result in higher PM2.5 concentrations; however, there is no means to quantify this impact without available monitoring data. On a local scale, residential woodsmoke is emitted from many small point sources across a community. The distribution and relative size of these point sources (due to type and quality of wood burning appliance, burning habits etc.) combined with topographical features and meteorological conditions can lead to high spatial variability and dramatically different air quality, even across small areas. Because a single monitoring station will only collect data at a single point within the airshed, it can only provide a small piece of the picture in terms of the air quality experienced across an entire community. If the purpose of the station is to be representative of the community and capture the average exposure, the location may be biased high or low. Even if the location of the station is representative of average conditions across the community, the station will not be capturing certain areas which may have much higher exposures, and we will not have an accurate picture of the true exposures that the entire population experience.   Measurement Strategies for Residential Woodsmoke Ambient PM2.5 is composed of particles from multiple sources, making it challenging to estimate the contribution from woodsmoke alone.  Five approaches to this challenge are described below.  1.5.1 Surveys and Emissions Inventories Emissions inventories and surveys are commonly conducted by various levels of government to estimate total emissions of air pollutants and the contributions of various sources. This information can be used to understand the importance of different pollutant sources and prioritise areas for improvement. For example the 2016 Canadian Air Pollutant Emissions 7  Inventory estimated that approximately 15% of total PM2.5 emissions in BC were created by residential wood burning, ranking ahead of all anthropogenic sources except road dust (14).  1.5.2 Studying Temporal Patterns Identifying temporal patterns in routine air quality monitoring data that are indicative of residential wood burning is a novel approach proposed by Hong et al (20). An algorithm was developed to retrospectively classify days as smoke-impacted using hourly PM2.5 and daily temperature data. Three parameters were established for the algorithm: sufficient variability in daily 1-hour PM2.5 concentrations; daily temperatures cold enough to require home heating; and a low ratio of daytime to nighttime mean PM2.5 concentrations, which is indicative of the higher usage of woodstoves during the evening and early morning hours due to colder temperatures and more residents being at home. Values were established for each of these parameters using data from known smoke-impacted days in Courtenay, BC identified using levoglucosan measurements from another study (29). This method was then applied to rank 23 communities in BC by their annual number of smoke-impacted days, and three of the top-ranked communities were chosen as the target monitored communities for this thesis. 1.5.3 Source Apportionment Using Tracer Compounds Chemical tracers are useful tools to identify and quantify the contribution of specific pollutants. The tracer most commonly used to identify woodsmoke is levoglucosan (1, 6-anhydro-b-D-glucopyranose). This compound is specific to the source because it only forms during the combustion of cellulose, which occurs at temperatures greater than 300°C (30). It is also one of the most abundantly produced organic compounds in woodsmoke (4) and is stable when emitted (31). However, levoglucosan is not a perfect tracer because the emissions can vary with fuel type (wood species) and combustion efficiency (4). 1.5.4 Source Apportionment Using Optical Properties The optical properties of woodsmoke can also be used to differentiate it from PM2.5 generated by other sources. Studies comparing PM2.5 from woodsmoke with that from vehicle exhaust 8  suggest that woodsmoke PM absorbs more light in the UV and blue wavelengths whereas vehicle exhaust absorbs light from UV through infrared (32,33). Using these differences in absorbance, dual- or multi-wavelength aethalometers can be used to help distinguish woodsmoke PM2.5 from other sources (13).  The difference between absorption of light at 370 nm (known as UVC because it measures ultraviolet absorption) and 880 nm (known as BC because it measures absorption by black carbon) wavelengths is known as delta C and is used as a woodsmoke indicator. This difference is specific to biomass combustion because the organic aerosol components of woodsmoke absorb more light at 370 nm relative to 880 nm, which does not happen with PM2.5 generated by other sources (34,35). Delta C is also strongly correlated with levoglucosan concentrations in ambient PM2.5 (36). Using a dual- or multi-channel aethalometer in this way is more cost-effective and simpler than monitoring chemical tracers, and can provide much greater temporal and spatial resolution.  1.5.5 Measuring Spatial Patterns Residential woodsmoke emissions can have high spatial variability across BC communities because it is emitted from many small point sources that are unevenly distributed across the community. Spatial patterns can therefore be useful when evaluating the contribution of residential woodsmoke to overall PM2.5. For example, consistently elevated PM2.5 concentrations in residential areas are likely created by residential wood burning. Several studies have employed mobile monitoring as a method for mapping woodsmoke hotspots in both rural and urban environments (16,17,19,37), and some have also evaluated the utility of multi-wavelength aethalometers for this purpose (13). Mobile monitoring is a relatively economical and flexible method for measuring particle concentrations at high spatial resolution across various terrains, land use areas, and pollutant gradients.  9   Project Overview This project was designed to develop a new, cost-effective method that can be used to address the two identified limitations of existing air quality monitoring networks with respect to residential woodsmoke: (1) they cannot provide source-specific information and (2) they cannot provide spatially resolved information. This method combines mobile monitoring with a nephelometer to address the lack of spatial resolution and a multi-wavelength aethalometer to address the lack of source-specific information. The method was tested by applying it across three communities with monitoring stations known to be impacted by residential woodsmoke (referred to as the “monitored communities” throughout this thesis), and three nearby unmonitored communities (referred to as the “unmonitored communities” throughout this thesis).  To evaluate the effectiveness of these instruments and assess the contribution of residential woodsmoke to PM2.5 concentrations in the three monitored communities, the testing of the mobile method was combined with additional monitoring at the fixed site monitoring stations. The two types of optical instruments were temporarily installed and levoglucosan samples were collected to compare results with the regulatory data from the beta attenuation monitors (BAM) operated by ENV.  1.6.1 Project Objectives The primary objective of this thesis was to develop a systematic and cost-effective method for assessing woodsmoke in both monitored and unmonitored communities across Canada. This required comparison of methods for measuring total PM2.5 with methods for establishing the woodsmoke contribution to total PM2.5. It also required testing the approach in both monitored and unmonitored communities. This was important because the method could be a valuable tool to characterise woodsmoke impacts across the many smaller Canadian communities that currently have no permanent air quality monitoring. The secondary objective of this thesis was to collect total PM2.5 and woodsmoke contribution data in six specific BC communities, of which three were monitored and three were 10  unmonitored. Mobile monitoring was conducted to estimate both total PM2.5 and woodsmoke concentrations across each community with high spatial resolution to create detailed air quality maps. These maps were then used to estimate the contribution of woodsmoke to PM2.5 patterns, identify hotspot neighbourhoods, and assess the representativeness of available monitoring station locations. 1.6.2 Research Questions A series of research questions was used to support each of the two objectives: Method development questions: 1. Can a single-channel nephelometer be used to estimate total PM2.5 concentrations? 2. Can a dual-channel aethalometer be used to estimate woodsmoke concentrations? 3. Can the two optical instruments be used together to estimate woodsmoke contribution to PM2.5 concentrations? 4. What information can mobile monitoring add to the understanding of woodsmoke in communities? 5. Is there value in using an aethalometer in addition to a nephelometer during mobile monitoring? Community-specific questions: 1. What did we learn about woodsmoke in the monitored communities? 2. What did we learn about woodsmoke in the unmonitored communities? 3. What did we learn about the spatial patterns across each community?   11  Chapter 2: Methods  Site Selection Three communities with ENV monitoring stations known to be impacted by residential woodsmoke were identified and each paired with a nearby, unmonitored community. Previous work conducted by Hong et al. had ranked 23 BC communities by the number of days identified as woodsmoke-impacted, with Houston, Courtenay, Port Alberni, Vanderhoof, and Whistler ranked as the most affected (20). From these five communities, three were selected from different regions: Courtenay from the Vancouver Island region; Vanderhoof from the northern interior region; and Whistler from the coast mountain region. To demonstrate the utility of the mobile monitoring method, it was important to test the method in unmonitored communities where woodsmoke was likely to be an important source. Therefore, three nearby unmonitored communities were selected to pair with each of the monitored communities. Cumberland was paired with Courtenay, Fraser Lake was paired with Vanderhoof, and Pemberton was paired with Whistler (Figure 2-1).  One driving route for the mobile monitoring was created for each community in the Whistler / Pemberton and Vanderhoof / Fraser Lake route pairs. However, due to the proximity of communities in the Comox Valley, the town of Comox was also included in the driving routes in this region. One route covered most of Courtenay (area southwest of the Courtenay river) and the unmonitored community of Cumberland, while the second route covered Comox, the rest of Courtenay (area northeast of the Courtenay river), and parts of the Comox Valley Regional District. These two routes are referred to as the Courtenay-Cumberland and Courtenay-Comox routes respectively throughout this thesis, and together as the Courtenay-Cumberland / Courtenay-Comox route pair.   12   Figure 2-1: Locations of study communities in British Columbia.  Timeline of Field Monitoring Field monitoring was conducted over three 2-week periods from January-March 2017 (Table 2-1), when average temperatures are typically low and use of wood for home heating is high. Fixed location monitoring was conducted at the monitoring stations for the entire 2-week period in each of the monitored communities. Mobile monitoring was performed in the monitored community and the nearby unmonitored community on alternating nights, for a total of seven nighttime runs on each route. Two daytime runs were also conducted on each route for comparison (Appendix B – List of Mobile Monitoring Runs). The nighttime mobile 13  monitoring runs were started at approximately 21:00 each evening, when woodsmoke is likely to be the dominant source of PM2.5 and less traffic is present on the roads.  Table 2-1: Field monitoring timeline. Community Pair  (*indicates the monitored community) Start Date End Date Whistler* and Pemberton 5th January 2017 19th January 2017 Courtenay* and Cumberland 24th January 2017 7th February 2017 Vanderhoof* and Fraser Lake 16th February 2017 2nd March 2017   Equipment Overview Four types of monitoring equipment were used in this project and are briefly explained here. 2.3.1 ENV PM2.5 Monitoring Instruments The BC air quality monitoring network primarily uses beta attenuation monitors (BAM) to measure PM2.5 concentrations at monitoring stations across the province. These instruments collect 1-hour samples of PM2.5 on a glass filter tape and estimate the total mass by measuring the difference in beta attenuation through the filter before and after the sample collection. The airflow rate through the filter tape is recorded and used to convert the beta attenuation to an average PM2.5 concentration over the hour. 2.3.2 Nephelometers Nephelometers are single-wavelength optical instruments that estimate particulate matter concentrations in real-time by measuring light scattering by particles in sample air (Figure 2-2).  The instrument measures total light scattering (Bscat), but routine calibration allows the instrument to correct for the effect of Rayleigh scattering by gases and report only the scattering caused by particles. This is known as Bsp, where “sp” refers to light scattering by particles, and has been strongly correlated with PM2.5 concentrations (38).  This measure is reported in units of inverse Mega-meters (Mm-1). 14   Figure 2-2: Nephelometer function diagram.  Single-wavelength light is shined through the air sample, and the amount of light that is scattered by particles in the air is measured by the light detector.  During this project two nephelometers were used: (1) an Ecotech M9003 was installed alongside the ENV monitoring equipment at the fixed stations in the monitored communities; and (2) an Ecotech Aurora 1000 was used in the mobile monitoring vehicle. Each instrument used 525 nm light sources. Both instruments were operated with inlet heaters to keep the relative humidity in the sample air below 60%, which reduces the potential for light scattering by water droplets. No size selective inlets were attached to the nephelometer sample lines because light scattering is dominated by very fine particles (aerodynamic diameters between 0.1 and 1 μm) and larger particles have a minimal contribution (39). 2.3.3 Aethalometers Fast response multi-wavelength aethalometers can provide more information about the chemical composition of a PM2.5 sample than BAMs or nephelometers. To collect data on PM2.5 only, a size selective cyclone is attached to the air inlet tubing to filter out larger particles. The aethalometer then deposits PM2.5 from sample air onto a quartz filter tape and shines multiple wavelengths of light through the sample every second to measure the proportion of each wavelength that is absorbed by the sample. The previously described delta C is used to indicate woodsmoke. Because aethalometers are primarily used to measure black carbon concentrations, the instrument internally converts absorbance measurements to 15  concentrations of particles absorbing at each wavelength. Therefore, delta C is reported as the difference in concentration of particles absorbing at 370 nm and particles absorbing at 880 nm in units of micrograms per cubic metre (µg/m3). Two aethalometers were used in this project to measure delta C: a Magee Scientific AE21 (loaned to the project by ENV) was installed alongside the ENV equipment at the fixed stations in the monitored communities; and (2) a Magee Scientific AE33 (purchased for this project and future research with funds from the BC Lung Association) was used in the mobile monitoring vehicle.  As filter-based optical measurements are affected by mass loading, where instrument response decreases with increased loading on the filter (40), both instruments systematically pause measurements and advance the filter to a clean section whenever measurements reach a maximum absorbance. Because this effect is incremental, the data is impacted between tape advances. The data from the older AE21 model must be manually corrected for this issue, while the AE33 was designed to overcome this issue internally in real-time using the patented DualSpot™ method (41).  2.3.4 Harvard Impactors Impactors are used to filter air and collect particulate matter of a known size (in this case PM2.5) on a Teflon filter for further analysis. Impactors operate at a specific flow rate of air at which the inertia of particles larger than the desired size will cause them to impact on a plate when passing through the Impactor, leaving only the particles of the desired size in the airstream passing through the Teflon filter. Impactors are used in conjunction with a calibrated air pump that maintains a constant air flow rate and reports the total volume of air that passed through the filter over the operational period. By weighing the sample filters before and after sampling, the mass of PM2.5 that has been deposited on the filter can be calculated and then converted into an average concentration during sampling. These filters can also be chemically analysed for compounds of interest such as levoglucosan.  16   Fixed Site Monitoring 2.4.1 Instrument Data Collection In each of the three monitored communities (Whistler, Courtenay, and Vanderhoof) an Ecotech M9003 nephelometer and a Magee Scientific AE21 aethalometer were installed and operated on the roofs of the ENV fixed stations alongside the ENV monitoring equipment for a 2-week period. Each instrument was housed in a weatherproof Pelican case to protect it from low temperatures and wintertime precipitation (Figure 2-3). The AE21 was operated with a BGI SCC 1.829 cyclone to remove particles larger than 2.5 microns from the sample air, with a Magee water trap also connected to the sample inlet tubing to reduce humidity in the sample airflow and protect the instrument from water damage. Both instruments were set to record data at the highest possible temporal resolution, 1-minute averages for the M9003 nephelometer and 5-minute averages for the AE21 aethalometer. Both instruments were calibrated following instructions from their manufacturers prior to transportation to each site.   17   Figure 2-3: Study fixed-site equipment in Whistler.  The BC Ministry of Environment (ENV) monitoring instruments are stored in the white housing with the sample inlets and small weather station attached to the top. The M9003 nephelometer was installed in the black pelican case to the right, with the sample inlet protruding from the top of the case. On the left, a blue tarpaulin covered the pelican case housing the AE21 aethalometer with its sample inlet attached to the side of the ENV housing. The two Harvard Impactors with their air pumps, power supplies, and back-up batteries were in individual protective cases (Figure 2-4) installed on the wooden frame. The inlets for all instruments were configured to sit as closely to the same height as possible.  Two Harvard Impactors with attached air pumps were also used to collect 24-hour PM2.5 samples on Teflon filters (37mm with PTFE Membrane) for each day of monitoring at the fixed sites (Figure 2-3 and Figure 2-4). The OMNI 400 air pumps used with the impactors were calibrated to a flow rate of 10 L min-1 using a MesaLabs Defender 520 DryCal before each 24-hour filter sample was started, and the flow rate was re-measured after each 24-hour period. The start and stop times, average flow rate, and total air volume were recorded for each sample. The Harvard Impactors, filter cassettes, and impaction plates were cleaned between each use, and clean mineral oil was applied to the impaction plates just prior to the installation in the impactors.  18  Four impactors were rotated to expedite changeovers and minimise time lost between 24-hour periods. Two clean impactors were assembled with a clean filter indoors, so that the two used impactors could be quickly swapped with clean impactors on each day of sampling. Filters were stored in individual petri dishes and transported in the upright position between leaving from and returning to the UBC Occupational and Environmental Hygiene Lab where they were weighed and analysed. Each petri dish was labelled with a filter identification code, and a matching ID sticker was removed from the petri dish and affixed to the impactor while the filter was in use. The same sticker was used to seal the petri dish when the filter was removed from the used impactor. This ensured filters were not misidentified or misplaced. Filters were changed at approximately 16:00 each day in Whistler, but this was then changed to approximately 12:00 for the Courtenay and Vanderhoof stations.   Figure 2-4: Harvard Impactor.  Harvard Impactor installed in protective Pelican case with OMNI 400 air pump, power supply, and back-up battery during monitoring.  19  2.4.2 Filters and Levoglucosan Analysis The 37mm Teflon filters used to collect the 24-hour PM2.5 samples in the Harvard Impactors were weighed in the climate- and humidity-controlled balance room at the UBC Occupational and Environmental Hygiene Lab before and after sampling, along with two field blanks per location. The total PM2.5 mass was calculated as the weight difference of each filter pre- and post-sampling. The total volume of air sampled was calculated from the active time and average flow rate of the pump that was attached to the filter and used to calculate the average PM2.5 concentration for the approximately 24-hour sampling period. This was performed for all filters, and the average daily PM2.5 concentration was calculated by taking the average of the two filters collected on each day.  One of each daily pair of filters was also analysed for levoglucosan mass by gas chromatography–mass spectrometry (GC/MS) at the UBC Occupational and Environmental Hygiene Lab (Appendix C – Levoglucosan Analysis Procedure). Combining these results with the total volume of air sampled by the pump attached to the filter, the average levoglucosan concentration for each approximately 24-hour period was calculated. The second filter from each day was sent to the University of Toronto for analysis of the oxidative potential of the sample as part of a linked Health Canada study.   Mobile Monitoring A thorough mobile monitoring protocol was created for use in this study and future work (Appendix A – Woodsmoke Mobile Monitoring: Full Protocol). A brief review of the methods covered by the protocol is provided here.  2.5.1 Route Creation and Monitoring Schedule One driving route was created for each community that started and ended at the ENV monitoring station in that community pair. These routes were designed to cover the entire area in as much detail as possible within a reasonable driving time. Feedback from community groups was solicited and used to ensure the routes focused on the residential areas of interest 20  to each community. The routes were designed to facilitate smooth data collection by looping around blocks and avoiding U-turns to prevent stop-and-go driving. These routes were programmed into a GPS navigator device (Garmin Nuvi 2497 using Garmin Basecamp software) to give directions to the driver during monitoring runs and to ensure that routes were accurately repeated.  The mobile monitoring was performed in conjunction with the fixed site monitoring, alternating each night between the monitored community route and the paired unmonitored community route. Each route was driven on seven nights starting at approximately 21:00, and twice during daytime periods (Appendix B – List of Mobile Monitoring Runs). Routes took between two and a half, and four hours to complete. To limit the effects of repeating temporal patterns on the mobile monitoring data, routes were scheduled to be driven in alternating directions, forward and reverse, so the same sections of the route were not sampled at the same time each night. The paired community routes were alternated to increase the probability of sampling under similar weather conditions on each route. For example, the weather conditions during the Whistler / Pemberton route pair were more likely to be similar between the two routes if they were driven on alternating evenings rather than if the Whistler route was driven for seven consecutive nights, followed by the Pemberton route for seven consecutive nights.  2.5.2 Equipment Set-Up and Operation Both the Aurora 1000 nephelometer and the AE33 aethalometer were installed on the rear seat of a vehicle (Figure 2-5). They were powered using a 12V power inverter connected to the vehicle 12V outlet.  21   Figure 2-5: Monitoring instruments installed in mobile vehicle.  Magee AE33 Aethalometer (foreground) and Ecotech Aurora 1000 Nephelometer (background) installed on rear seat of vehicle for mobile monitoring. The sample inlet tubing of both instruments was passed through the rear window and attached to the side of the vehicle on the opposite side from the exhaust to limit self-contamination of data. A BGI SCC 1.829 cyclone was attached to the aethalometer inlet line to remove particles larger than PM2.5 from the sample air. This cyclone has a shield covering the air inlet, which was attached to the vehicle at the front of the rear window orientated approximately 30° from upright to prevent precipitation from falling into the inlet, and to prevent air from being forced into the inlet due to the motion of the vehicle. A plastic funnel was connected to the end of the nephelometer sample tubing and attached behind the rear window orientated approximately 30° below horizontal. Again, this was to prevent precipitation from entering the inlet tubing and airflow from being forced into the opening. The attachment location of the inlets was chosen to increase the width of bends in the sample tubing and minimise particles being removed from the airflow by impacting on the tubing walls. The window opening around the inlet tubing was sealed using foam and duct tape to prevent moisture entering the vehicle and limit heat escape (Figure 2-6 and Figure 2-7). 22   Figure 2-6: Vehicle prepared for mobile monitoring.  The aethalometer cyclone inlet is attached above the rear window and the nephelometer sample tubing is attached with a funnel behind the rear window.   Figure 2-7: Orientation of sample inlets.  Inlets were attached to the exterior of the mobile monitoring vehicle. Cyclone connected to AE33 sample tubing was attached to the rear window orientated approximately 30° above horizontal. A funnel was connected to Aurora 1000 sample tubing and attached to the rear of the vehicle at approximately 30° below horizontal.     23  The AE33 aethalometer records data at 1-second intervals, but the Aurora 1000 can only save to its internal logger at 1-minute intervals. To improve the temporal and spatial resolution of the monitoring, a laptop was used to save live 1-second data from the Aurora 1000 to a text file through a serial connection and the Windows HyperTerminal program. This allowed us to collect 1-second measurements from both instruments. The 1-second nephelometer data were lost from two runs (one nighttime run on the Vanderhoof route, and one daytime run on the Courtenay and Comox route) due to errors in the laptop connection. These runs were excluded from further analyses.  A GPS datalogger (GlobalSat DG-100) was also used in the vehicle to record its location at 1-second intervals. In order to match the instrument data to the GPS location, the instrument clocks were reset to the accurate time prior to each mobile run. While monitoring, the vehicle was driven at approximately 30 km/h whenever safe (higher speeds were necessary on most highways to avoid obstructing traffic), which equates to instrument measurements approximately every 8.3 meters. Throughout the monitoring run the laptop was also used to record relevant notes about the sampling and specific events that might affect the data. These notes files were used to keep a record of: the start and end times of the monitoring run; the driving direction of the route (forwards or reverse); and a qualitative assessment of weather conditions along with the current temperature from the vehicle readout. Notable events during the monitoring run were also recorded along with their times, such as: having to wait at a railway crossing; driving behind a large vehicle kicking up visible road dust; or driving through a visible smoke plume.  Data Cleaning All data cleaning, analysis, and map production were performed using the R statistical computing environment (42). Data collected by the AE21 aethalometer installed at the fixed monitoring stations were first corrected using the WUAQL AethDataMasher software (Version 7.1) created by George Allen et al. (43). This process removes the ‘spot loading effect’ that affects data collected by older aethalometer models, and smooths values across the gaps 24  during which the instrument is paused for tape advances. In addition, negative values can occur when the instrument is exposed to a very low concentration immediately following a very high concentration. These were replaced with the closest positive reading as per the data cleaning protocol used by ENV. If no positive values were reported within a 30-minute period of a negative reading (six measurements for the AE21, which had a 5-minute averaging period), the value was set to missing. Negative values from the mobile AE33 aethalometer were set to missing if no positive values were reported within a 30-second period (30 measurements for the AE33, which had a 1-second averaging period). This stricter limit was used on the mobile data to prevent values from being incorrectly attached to locations. Data collected by the mobile aethalometer, nephelometer, and GPS prior to the run start times (while the vehicle and instruments warmed) were removed to ensure data were consistent between runs. This was done by cropping the data from the mobile instruments using the run start and stop times recorded in the field notes file for each run. Data from all instruments were then matched using the measurement time for each record, which connected GPS coordinates to each 1-second measurement by the mobile instruments. The 1-second mobile values were then matched to the 1-hour fixed site values, providing complete data for further analyses.   Fixed Site Analyses 2.7.1 Temporal Matching Between Fixed Location Instruments Following the field monitoring campaign, the 1-hour average PM2.5 data recorded by the ENV BAM instruments were retrieved from the BC Air Quality online database along with other air quality and meteorological data recorded at the fixed monitoring stations (44). The 1-minute and 5-minute averages from the M9003 nephelometer and the AE21 aethalometer, respectively, were converted to 1-hour averages to match the PM2.5 data from the BAM instruments. Along with the 1-hour data from the BAM instruments, the nephelometer and aethalometer data were also converted to 24-hour averages to match the levoglucosan and PM2.5 concentrations from the 24-hour filter samples.  25  Because the first filter samples (February 17th) in Vanderhoof only covered a 19-hour period due to delays in set up of the instruments, this day was excluded from analysis of the 24-hour average data, leaving 14 days of filter measurements for each of Whistler and Courtenay, and 13 for Vanderhoof. Vanderhoof data from this period were still included in the analysis of 1-hour averages of the other instruments. Data were also missing for the BAM and nephelometer during part of the day on February 4th in Courtenay due to a suspected power cut at the monitoring station. Therefore, analyses including 24-hour averages of those instruments only cover 13 days in Courtenay.  2.7.2 Exploration of Temporal Patterns To explore the long-term patterns over the 2-week period at each fixed monitoring site, time series plots were created showing the 24-hour averages of the data from each of the five methods: (1) PM2.5 concentrations measured by the ENV BAMs; (2) Bsp measured by the nephelometer; (3) delta C measured by the aethalometer; (4) PM2.5 measured on the filter samples; and (5) levoglucosan measured on the filter samples. To allow for comparison of instrument responses, the 24-hour averages of each instrument were plotted on separate scales from zero to the max daily average observed during the whole monitoring campaign. To explore the average daily patterns observed in each community, the average value of each of the three temporally-resolved instruments (BAM, nephelometer, and aethalometer) was calculated for each hour of the day across the two-week period in each community. These results were then plotted as time series following the same methods used for the 24-hour averages to compare the relative responses of the instruments.   2.7.3 Instrument Comparisons Relationships between the methods used at the three fixed sites were compared using scatter plots and simple linear regression. While it was possible to analyse the relationships between all of the five methods, the specific combinations below were chosen to address specific objectives.  26  2.7.3.1 Comparison of Bsp with Established PM2.5 Measures To compare the nephelometer measurements with more established and direct PM2.5 measurements, 1-hour Bsp averages from the fixed M9003 nephelometer were compared with 1-hour averages from the ENV BAMs. This linear relationship for each community was used to convert the mobile Aurora 1000 nephelometer Bsp values to PM2.5 concentration estimates for the maps of each route. In addition, 24-hour average Bsp values were compared with the PM2.5 concentrations calculated from the filter samples.  2.7.3.2 Comparison of Woodsmoke-Specific Measures To assess the ability of the dual-channel aethalometer to specifically measure woodsmoke, the 24-hour averages of the delta C measurements from the AE21 were compared with levoglucosan concentrations measured on the filter samples. 2.7.3.3 Woodsmoke Contribution Analysis To estimate the contribution of residential woodsmoke to total PM2.5 concentrations, multiple relationships were explored. The chemical tracer method was explored by examining the relationship between levoglucosan and total PM2.5 concentrations recorded on the filters. The optical method was also explored by examining the relationship between delta C as reported by the AE21 aethalometer and Bsp as reported by the M9003 nephelometer.  We also used data from the optical methods to explore whether the relationship between delta C and Bsp differed during times of day when woodsmoke was expected to be a more or less dominant source of total PM2.5. The data were split into nighttime hours (17:00 – 9:00) when residents are more likely to be at home and operating their wood burning appliances, and daytime hours (9:00 – 17:00) when residents are more likely to be away from their homes.  Mobile Data Analysis and Map Production  The relationship between the two mobile instruments was also used to explore the influence of residential woodsmoke on PM2.5 concentrations using time series plots, scatter plots, and 27  simple linear regression. Similar to the night and day comparison of the fixed site data, the relationship between the mobile instruments was also compared during the daytime (N=4) and nighttime (N=14) monitoring runs in each of the three regions to see how the relationship differed during times when woodsmoke was expected to be more or less dominant.  To address the secondary objective of this thesis, average spatial patterns of estimated Bsp and delta C were calculated and mapped across each monitoring route. A simple averaging of the raw data across the monitoring runs would be heavily impacted by temporal differences between the monitoring nights. To avoid this, the spatial patterns from each mobile run were first extracted by calculating the relative z-score for each measurement. The z-score of an observation is defined as its relative location within the distribution of all observations, expressed as multiples of the standard deviation between the observation and the mean of the distribution. Because data collected by the mobile instruments were lognormally distributed for each run, the data from each instrument were first log-transformed to more closely approximate a normal distribution prior to the z-score calculation. As such, all map legends are on an exponential scale. In addition, data from different areas connected by long, uninhabited sections of highway on the Pemberton and Fraser Lake routes were first cropped to remove these sections, ensuring that z-scores were only calculated based on the inhabited areas of the monitoring routes. Because the route to Fraser Lake passed through the smaller community of Fort Fraser and the neighbourhood of Engen, these were included in the route.  Raster grids were then created for each location covering the extent of the route, at a spatial resolution of 33.33 m2. Raster layers were calculated for each monitoring run where all z-scores that were recorded within each raster cell were averaged, and the raster grid was then focally smoothed using the surrounding 3x3 matrix of cells weighted by the number of measurements within each cell. This resulted in each cell having a weighted average representation of the surrounding 100 m2 area. Finally, to calculate the average spatial pattern across the route during the nighttime monitoring, the matching raster cells of the layers from each run were averaged, with a requirement for each cell to have measurements from at least five runs. When 28  mapping these average patterns, each raster cell was shaded according to its calculated z-score, using an eight-bin scale with equal break points from less than -1.5 to greater than +1.5. To examine the location of the fixed monitoring stations relative to the spatial pattern across each route, the calculated z-score of the closest cell to the station was extracted. This value was converted to an estimate in the original units of each instrument, and the same was done for the break points on the z-score scale. These estimates were generated by determining which instrument values would be equal to the z-score in the distribution of the data from all the runs used in the map. For easier interpretation of the map created using the nephelometer data, the estimates of equivalent Bsp were converted to a PM2.5 concentration estimate using the community-specific linear relationship established between the fixed 1-hour nephelometer Bsp and ENV BAM PM2.5 measurements. This conversion was included because users are more familiar with PM2.5 concentrations from air quality reports, objectives, and advisories than they are with Bsp values. The delta C values reported for the aethalometer map were used as relative values because there is no clear conversion to a woodsmoke PM2.5 concentration. Further reasoning behind this decision is discussed in Section 4.4. To add context to the spatial maps, average temperatures, PM2.5 concentrations, and wind speeds along with wind roses were calculated using the ENV data from the monitoring stations during each nighttime monitoring run using the 1-hour averages between 20:00 and 01:00. Meteorological data during the final night trip on the Courtenay and Comox route was not available online and so the averages for this route are based on the other six nighttime runs.  2.8.1 Comparison of Spatial and Temporal Variance To contrast spatial and temporal components of air quality variability, the spatial variance captured by the mobile aethalometer and nephelometer were compared with the temporal variance captured by the aethalometer and nephelometer at the fixed locations. The 1-second data from each mobile instrument were first averaged to match the 1-minute period for the fixed nephelometer and the 5-minute period for the fixed aethalometer. The mobile data were then adjusted for temporal variability at the fixed location over the duration of the drive to 29  extract just the spatial variability (Equation 2-1). Similar methods have been used in previous research to adjust mobile monitoring data for short-term temporal trends during monitoring runs (16,17). The spatial variance in the mobile data for each monitoring run was then compared with the temporal variance in the fixed location data during that run using a Fligner-Killeen test for homogeneity of variance (45). The Fligner-Killeen test was chosen for its robustness to non-normal data and an alpha level of p < 0.05 was used to assess significance.  Equation 2-1: Equation for adjusting mobile measurements for short-term temporal trends at the fixed location, where ‘t’ is the time-specific value of the mobile measurement (MobileRaw) or smoothed 15-minute running mean of the fixed site (FixedRunning) data, and ‘run’ is the mean of the fixed site measurements (FixedRaw) over the duration of the monitoring run. 𝑀𝑜𝑏𝑖𝑙𝑒𝐴𝑑𝑗𝑢𝑠𝑡𝑒𝑑𝑡 =  𝑀𝑜𝑏𝑖𝑙𝑒𝑅𝑎𝑤𝑡𝐹𝑖𝑥𝑒𝑑𝑅𝑢𝑛𝑛𝑖𝑛𝑔𝑡𝐹𝑖𝑥𝑒𝑑𝑅𝑎𝑤𝑟𝑢𝑛⁄ 30  Chapter 3: Results  Summary of Temporal Patterns in the Monitored Communities The average conditions measured at the three monitoring stations during the field campaigns were variable across the monitored communities (Table 3-1). Temperatures were lowest in Whistler and highest in Vanderhoof, but PM2.5 concentrations, levoglucosan concentrations, Bsp, and delta C were considerably higher in Courtenay than in the other two communities.   Table 3-1: Average conditions at the three monitoring stations during the field monitoring campaign.  Mean (SD) presented for each measure. BAM acronym refers to the beta attenuation monitors operated by the BC Ministry of Environment and Climate Change Strategy.  BAM PM2.5 Filter PM2.5 Levoglucosan Bsp delta C Temperature Wind Speed  (µg/m3) (µg/m3) (µg/m3) (Mm-1) (µg/m3) (°C) (m/s) Whistler 5th Jan 2017 –  19th Jan 2017 11.1 (8.9) 5.3 (3.3) 0.36 (0.31) 25.9 (29.5) 0.57 (0.79) -5.8 (5.1) 0.7 (0.6) Courtenay 24th Jan 2017 –  7th Feb 2017 16.4 (14) 13.4 (6.8) 1.31 (0.76) 70.6 (71) 1.2 (1.44) 3.9 (2.6) 1.1 (0.8) Vanderhoof 16th Feb 2017 – 2nd Mar 2017 9.4 (8.4) 9.3 (4.7) 0.26 (0.19) 35.9 (37) 0.46 (0.57) 4.9 (4.3) 1.3 (0.8)  High temporal correlation was observed between all measurements taken at the fixed monitoring stations in the three monitored communities (Figure 3-1). The three total PM2.5 measures (BAM, filter, and Bsp) produced similar daily patterns, as did the woodsmoke-specific measures (levoglucosan and delta C). In Whistler, the relative 24-hour averages of the BAM PM2.5 were consistently higher than the filter PM2.5 and the nephelometer Bsp. This was not observed in the other communities (Figure 3-1 and Table 3-1). In Vanderhoof, the two 31  woodsmoke-specific measures were consistently lower than the total PM2.5 measures, but still followed the same patterns across days (Figure 3-1). The highest values for each method were measured in Courtenay, where air quality advisories were issued during the sampling when the 24-hour average PM2.5 concentrations measured by the BAM surpassed the provincial objective of 25 µg/m3. The hourly data showed a diurnal pattern in each community, with smaller peak values in the morning (at 08:00 or 09:00), lower values in the afternoon (reaching daily minimums between 13:00 and 17:00), and then larger peak values at night (Figure 3-2). While this pattern was present in all three communities, the diurnal peaks in Courtenay were more pronounced compared with the other communities. Here, the average morning and night PM2.5 peaks were approximately 16 µg/m3 and 23 µg/m3 higher, respectively, than the average afternoon low of 5 µg/m3. The range of PM2.5 concentrations in both other communities was approximately 10 µg/m3. 32   Figure 3-1: Time series plot showing 24-hour average values of the five monitoring methods operated at the monitoring stations.  The two y-axes for the three total PM2.5 measures are shown on the left (Bsp measured by the nephelometer, and PM2.5 concentrations measured by the BC Ministry of Environment beta attenuation monitors (BAM) and filter samples), while the two y-axes for the woodsmoke-specific measures are shown on the right (levoglucosan concentrations from filter samples, and delta C measured by the aethalometer). Each of the four y-axes are scaled from zero to the max value measured on that scale across the six weeks of field monitoring.Bsp Filter - PM2.5 BAM - PM2.5 delta C Levoglucosan B sp (Mm-1) PM2.5 (µg/m3 ) Levo (µg/m3 ) delta C (µg/m3 ) 33   Figure 3-2: Time series plots showing the average daily patterns measured by the fixed site instruments during the monitoring campaign in each of the three monitored communities.  The two y-axes for the total PM2.5 measures are shown on the left (Bsp measured by the nephelometer, and PM2.5 concentrations measured by the BC Ministry of Environment beta attenuation monitors (BAM)), while the y-axis for the woodsmoke-specific indicator delta C measured by the aethalometer is shown on the right. Each of the four y-axes are scaled from zero to the max value measured on that scale across the six weeks of field monitoring. Bsp BAM - PM2.5 delta C  B sp (Mm-1) PM2.5 (µg/m3 ) delta C (µg/m3 ) 34   Comparison of PM2.5 Measures at the Fixed Sites Strong correlation was observed between 1-hour average Bsp measurements from the M9003 nephelometer and BAM PM2.5 concentrations, with a coefficient of determination (R2) of 0.88 across all locations (Figure 3-3). However, the linear relationship varied between communities, with a greater slope found in Whistler (0.31 μgm-3/Mm-1) than in Courtenay (0.22 μgm-3/Mm-1) and Vanderhoof (0.24 μgm-3/Mm-1). The three community-specific relationships were used to convert mobile Bsp values to PM2.5 estimates for mapping of the nephelometer results.   Figure 3-3: Scatterplot comparison of calculated hourly averages of Bsp as measured by the nephelometer and the PM2.5 concentrations reported by the beta attenuation monitor (BAM) at each monitoring station.  Linear regression lines are shown for each location with the slope, intercept and R2 values presented in the accompanying table. Colour of points and linear regression lines indicates location and the 95% confidence interval is shown around the linear relationship calculated using the overall data.   Strong correlation was also observed between the 24-hour average Bsp measurements and the filter-based PM2.5 concentrations, with an R2 of 0.94 across all locations (Figure 3-4). The relationship between the nephelometer data and the filter PM2.5 measurements was more consistent between communities than the relationship between the nephelometer and BAM instruments. Specifically, the Bsp in Whistler had stronger correlation with the filter-based PM2.5 than with the BAM measurements (R2 = 0.99 compared with 0.88). Although both the BAM and Location Slope Int R2 Whistler 0.31 3.1 0.88 Courtenay 0.22 1.1 0.95 Vanderhoof 0.24 0.82 0.91 Overall 0.22 2.7 0.88  Bsp (Mm-1) BAM - PM2.5 (µg/m3 ) 35  filter-based measurements reflect total PM2.5 concentrations, the slopes of the relationships across all sites differed slightly. Overall, there was a lower slope of 0.18 μgm-3/Mm-1 for the filter-based relationship with Bsp, compared with 0.22 μgm-3/Mm-1 for the BAM relationship.   Figure 3-4: Scatterplot comparison of the calculated 24-hour averages of Bsp as measured by the nephelometer and the 24-hour average PM2.5 concentrations measured by the filter samples at each monitoring station.  Linear regression lines are shown for each location with the slope, intercept and R2 values presented in the accompanying table. Colour of points and linear regression lines indicates location and the 95% confidence interval is shown around the linear relationship calculated using the overall data.    Comparison of Woodsmoke-Specific Measures at the Fixed Sites When comparing delta C with levoglucosan concentrations (Figure 3-5), very strong correlation was observed in each of the three communities, with community-specific R2 values of at least 0.95 and an overall R2 of 0.90. While a one-to-one linear relationship was observed when considering data across all three sites, there were differences in the slope of the relationship between communities. The measured relationships in Whistler and Vanderhoof were similar (slopes of 0.68 and 0.62 μg/m3 of levoglucosan per μg/m3 delta C, respectively), but the slope of the relationship between the methods was steeper in Courtenay (0.99 μg/m3 of levoglucosan per μg/m3 delta C).  Location Slope Int R2 Whistler 0.18 0.57 0.99 Courtenay 0.17 1.3 0.97 Vanderhoof 0.21 1.7 0.91 Overall 0.18 1.5 0.94  Bsp (Mm-1) Filter - PM2.5 (µg/m3 ) 36   Figure 3-5: Scatterplot comparison of the two woodsmoke-specific measures, calculated 24-hour averages of delta C as measured by the aethalometer, and the 24-hour average levoglucosan concentrations measured by the filter samples at each monitoring station.  Linear regression lines are shown for each location with the slope, intercept and R2 values presented in the accompanying table. Colour of points and linear regression lines indicates location and the 95% confidence interval is shown around the linear relationship calculated using the overall data.    Comparison of Woodsmoke and Total PM2.5 Measures at the Fixed Sites Strong correlation was found between woodsmoke-specific methods and total PM2.5 concentrations using both the chemical tracer levoglucosan (Figure 3-6) and the optical delta C (Figure 3-7).  Correlation between the filter measures was stronger in Whistler and Courtenay (R2 of 0.98 and 0.90, respectively) and levoglucosan accounted for approximately ten percent of the PM2.5 mass on the gravimetric filter samples in these regions. Levoglucosan accounted for approximately five percent of the PM2.5 in Vanderhoof, and the relationship had relatively weaker correlation (R2 = 0.76) suggesting that woodsmoke contributed less to PM2.5 concentrations relative to the other communities.  Location Slope Int R2 Whistler 0.68 -0.03 0.99 Courtenay 0.99 0.13 0.95 Vanderhoof 0.62 -0.02 0.95 Overall 1.00 -0.12 0.90  Levoglucosan (µg/m3 ) delta C (µg/m3) 37   Figure 3-6: Scatterplot comparison of the two measurements made from the daily filter samples; woodsmoke tracer levoglucosan concentrations against total PM2.5 concentrations.  Linear regression lines are shown for each location with the slope, intercept and R2 values presented in the accompanying table. Colour of points and linear regression lines indicates location and the 95% confidence interval is shown around the linear relationship calculated using the overall data.   Similar patterns were found in the relationships between the 1-hour averages of the optical measurements at the fixed sites (Figure 3-7). Whistler and Courtenay again showed strong correlation between the delta C and Bsp measurements (R2 of 0.91 and 0.89, respectively). However, Courtenay had a steeper slope than Whistler (44 Mm-1/μgm-3 in Courtenay compared with 36 Mm-1/μgm-3 in Whistler). In Vanderhoof, the relationship had a relatively higher slope and weaker correlation compared with the other communities (60 Mm-1/μgm-3, R2 = 0.77).  Location Slope Int R2 Whistler 10 1.50 0.98 Courtenay 8.5 2.30 0.90 Vanderhoof 21 3.80 0.76 Overall 7.6 4.30 0.73  Filter - PM2.5 (µg/m3 ) Levoglucosan (µg/m3) 38   Figure 3-7: Scatterplot comparison of 1-hour averages of the two optical instruments installed at the fixed locations; woodsmoke delta C as measured by the aethalometer, against Bsp measured by the nephelometer. Linear regression lines are shown for each location with the slope, intercept and R2 values presented in the accompanying table. Colour of points and linear regression lines indicates location and the 95% confidence interval is shown around the linear relationship calculated using the overall data.   3.4.1 Daily Differences in Woodsmoke Contribution Correlation between optical measures of woodsmoke and total PM2.5 was stronger during nighttime hours (between 17:00 and 09:00) than during daytime hours (between 09:00 and 17:00) (Figure 3-8). While the equation of the overall linear relationship remained similar, the correlation across the three fixed stations decreased from 0.87 at night to 0.66 during the day. This followed the expected pattern as residential wood burning was expected to be more prevalent during the night hours.  While all three community-specific relationships between delta C and Bsp showed weaker correlation during the day than at night, the differences in the relationship varied. The daytime relationship in Courtenay was very similar to the nighttime relationship (R2 = 0.83 compared with 0.88), while it was slightly weaker in Whistler (R2 = 0.71 compared with 0.92) and considerably weaker in Vanderhoof (R2 = 0.25 compared with 0.86). The equations of the linear relationships remained similar in each community (Figure 3-8). Location Slope Int R2 Whistler 36 5.10 0.91 Courtenay 44 17.0 0.89 Vanderhoof 60 7.90 0.77 Overall 45 9.90 0.86  B sp (Mm-1) delta C (µg/m3) 39    Figure 3-8: Scatterplot comparison of 1-hour averages of delta C and Bsp measured at the monitoring stations during: a) Day Hours between 09:00 and 17:00, and b) Night Hours between 17:00 to 09:00.  Linear regression lines are shown for each location with the slope, intercept and R2 values presented in the accompanying table. Colour of points and linear regression lines indicates location and the 95% confidence interval is shown around the linear relationship calculated using the overall data.  Location Slope Int R2 Whistler 37 4.80 0.71 Courtenay 49 11.0 0.83 Vanderhoof 56 8.70 0.25 Overall 49 7.50 0.66  Location Slope Int R2 Whistler 36 5.20 0.92 Courtenay 43 21.0 0.88 Vanderhoof 61 7.70 0.86 Overall 45 11.0 0.87  B sp (Mm-1) B sp (Mm-1) delta C (µg/m3) delta C (µg/m3) a) Day Hours b) Night Hours 40   Comparison of Woodsmoke and Total PM2.5 Measures during Mobile Monitoring Scatter plots comparing the delta C woodsmoke indicator and the Bsp PM2.5 indicator measured during mobile monitoring show variable relationships between the three regions and between the 13-14 nighttime and 3-4 daytime runs (Figure 3-9). Stronger correlation was observed during the nighttime runs than during the daytime runs in each region as expected, with R2 values of 0.66 compared with 0.49 on the Whistler / Pemberton route pair, and 0.75 compared with 0.48 on the Courtenay-Cumberland / Courtenay-Comox route pair. The relationships were weaker on the Vanderhoof / Fraser Lake route pair with an R2 value of 0.36 during the nighttime, and no correlation during the daytime (R2 = 0.0006). The scatterplots for this region show that high Bsp at times of low delta C were responsible for the weaker relationship, especially during the daytime monitoring runs (Figure 3-9). They also show that the highest values measured by both instruments were typically much lower during the daytime runs than during the nighttime for all communities. The one exception was the Bsp measurement during the daytime monitoring runs on the Vanderhoof / Fraser Lake route pair, where very high values were measured. Although the delta C and Bsp measurements were correlated, scatterplots show unexplained variability between the two instruments even during nighttime runs. For example, a time series comparison of the two instruments during two runs on the Fraser Lake route shows extreme differences in correlation (Figure 3-10). During the nighttime run there is clear correlation between the two instruments with both instruments responding to the same plumes over time. In contrast, the daytime plumes measured by the nephelometer show no similar response from the aethalometer.     41   Figure 3-9: Scatterplot comparisons of 1-minute averages of the delta C woodsmoke indicator and the Bsp total PM2.5 indicator measured by the mobile instruments during the monitoring runs.  One plot is shown for each route pair (Whistler / Pemberton, Courtenay-Cumberland (CCD) / Courtenay-Comox (CCX), and Vanderhoof / Fraser Lake) during nighttime monitoring runs in the upper row of plots, and daytime monitoring runs in the lower row of plots. All plots are shown on the same axes scales with three extreme Bsp measurements not shown during one Vanderhoof daytime run, and two extreme delta C measurements not shown during one Fraser Lake night run. These extreme measurements are included in the linear regressions shown.  R2 = 0.66 R2 = 0.75 R2 = 0.36 R2 = 0.49 R2 = 0.48 R2 = 0.0006 B sp (Mm-1) B sp (Mm-1) B sp (Mm-1) B sp (Mm-1) B sp (Mm-1) B sp (Mm-1) delta C (µg/m3) delta C (µg/m3) delta C (µg/m3) delta C (µg/m3) delta C (µg/m3) delta C (µg/m3) 42    Figure 3-10: Time series comparing responses of the mobile aethalometer and nephelometer during two monitoring runs of the Fraser Lake route. The y-axes for both instruments are plotted from zero to the maximum value observed during both runs. The first day run on the Fraser Lake route (Day Run 1, February 22nd, 2017) shows an example of low correlation between the two mobile instruments, while Night Run 4 (February 23rd, 2017) is an example of high correlation between instruments.    Nephelometer - B sp (Mm-1) Aethalometer - delta C (µg/m3 ) Fraser Lake - Day Run 1 Fraser Lake - Night Run 4 43   Average Conditions During Mobile Monitoring Runs The mobile monitoring routes were alternated each night between the monitored and unmonitored communities to capture similar conditions, but there were still differences in each community pair (Table 3-2). Conditions on the Whistler and Pemberton monitoring nights were the most similar with average temperatures of -6.6 and -6.2 °C, respectively, and calm winds. These conditions favour high use of wood burning appliances and limited venting of emissions from the area. Winds were primarily from the southeast (Figure 3-11). The PM2.5 concentrations measured by the ENV BAM on the Pemberton nights were slightly higher than those on the Whistler nights (16.3 μg/m3 compared with 14.1 μg/m3).  Large differences were observed between the nights on which the Courtenay-Cumberland and Courtenay-Comox routes were monitored. Average temperatures were similar (3.2 and 3.8 °C), but higher wind speeds and lower PM2.5 concentrations were reported by the ENV instruments during the Courtenay-Comox nights (1.3 m/s and 20.6 μg/m3) compared with the Courtenay-Cumberland nights (0.7 m/s and 31.3 μg/m3). The nighttime wind direction in this region was predominantly from the west (offshore winds). One exception to this pattern was observed during a Courtenay-Comox run, when stronger winds were measured from the southeast (Figure 3-11).  Wind speeds and temperatures were both higher on average during the monitoring of Vanderhoof and Fraser Lake, which likely contributed to the lower average PM2.5 concentrations in this region. The averages of all monitoring station measurements were slightly higher during the Vanderhoof nights compared with the Fraser Lake nights (Table 3-2). Winds during the Vanderhoof nights were predominantly from the west, while wind directions were variable on the Fraser Lake nights (Figure 3-11). There was one run on each route with high wind speeds and very low PM2.5 conditions (3.4 m/s and 0.44 µg/m3 during the first Vanderhoof nighttime run, and 2.5 m/s and 1.8 µg/m3 during the seventh Fraser Lake nighttime run).  44  These differences in conditions between route pairs must be considered when comparing maps of the paired route maps. They are especially important when comparing the Courtenay-Cumberland / Courtenay-Comox route pair, because the fixed site PM2.5 concentrations were more than 50% higher during the Courtenay-Cumberland nights compared with the Courtenay-Comox nights.   Table 3-2: Average conditions at the monitoring stations during the nighttime monitoring runs on each route. Values calculated using the 1-hour averages of the fixed nephelometer and aethalometer along with the BC Ministry of Environment measurements of PM2.5 with the beta attenuation monitors (BAM PM2.5) and meteorological data between the hours of 20:00 and 01:00.  Monitoring Route Bsp (Mm-1) delta C (µg/m3) BAM PM2.5 (µg/m3) Temp (°C) Wind Speed (m/s) Whistler 37 0.9 14.1 -6.6 0.6 Pemberton 43 1.1 16.3 -6.2 0.7 Courtenay-Cumberland 145 2.7 31.3 3.2 0.7 Courtenay-Comox 86 1.6 20.6 3.8 1.3 Vanderhoof 62 0.9 14.1 5.2 1.3 Fraser Lake 46 0.6 11.2 4.4 1.0    45   Figure 3-11: Wind roses during the nighttime runs on each route.  Wind data collected at the monitoring stations by the BC Ministry of Environment (ENV).    Wind Speed (m/s) 46   Route Average Maps The maps that were created to show the average spatial patterns measured across each community during winter nights are presented in this section. The spatial patterns measured in this study are only relevant to the winter heating season as meteorological patterns may differ by season and other sources may be more prominent in other seasons. Maps labeled ‘A’ in Figure 3-12 through Figure 3-17 show the average patterns of Bsp measured by the nephelometer as an indicator of total PM2.5 concentrations, while maps labeled ‘B’ show the average patterns of the woodsmoke indicator delta C measured by the aethalometer. In general, the average patterns measured by both instruments show similar patterns on each route, supporting the expectation that woodsmoke is the dominant PM2.5 source in these communities during winter nights.  3.7.1 Whistler and Pemberton Routes Measurements from the Whistler route show smoke hotspots to the west of the fixed monitoring station and along the east side of Alta Lake, with lower values measured on the west of Alta Lake and in the southwest of the map (Figure 3-12). Measurements from the Pemberton route show hotspots throughout the area, particularly in southeastern Pemberton, the southern parts of Mount Currie, and in the Xit’olacw area to the northeast (Figure 3-13). The Bsp measurements at the location of the fixed Whistler monitoring station were slightly lower than the average across the Whistler route with a mean z-score (Z) of -0.14. They were much lower than the average across the Pemberton route (Z = -0.87), with the PM2.5 estimate at the fixed-site around half that of the route average. The delta C measurements close to the fixed station followed a similar pattern, with mean Z of -0.16 and -0.35 on the Whistler and Pemberton routes, respectively. Substantial variability was observed across both routes. For the Whistler route, the highest averages on the nephelometer map were up to four times higher than the average at the fixed monitoring station, and for the Pemberton route they were up to ten times higher. On the aethalometer maps the highest averages were up to five and six times 47  higher than the averages at the fixed monitoring station for the Whistler and Pemberton routes, respectively. 3.7.2 Courtenay-Cumberland and Courtenay-Comox Routes The aethalometer and nephelometer route average maps for the Courtenay-Cumberland route show similar smoke hotspots, particularly in the centre and southeast of Courtenay and the northeast of Cumberland (Figure 3-14). The northeast of Courtenay and north of Royston both had higher relative delta C than Bsp values, suggesting more woodsmoke impact in these places. Meanwhile, there were relatively lower delta C values than Bsp values observed along the highway between Courtenay and Cumberland, suggesting more impact from non-woodsmoke sources. Measurements from the Courtenay-Comox route show hotspots around the Courtenay monitoring station and in the northwest of the route, with slightly higher relative values also in the main residential area of Comox (Figure 3-15). The delta C map shows similar patterns to the nephelometer map, but there appears to be less variation from the mean with large areas of the map falling in the middle of the z-score scale (Figure 3-15).  Both the Bsp and delta C levels around the Courtenay fixed station were higher than the average across the Courtenay-Cumberland route (Z = 0.41 and 0.38, respectively) and much higher than the average across the Courtenay-Comox route (Z = 1.82 and 1.04, respectively) during the monitoring period. Even still, higher values were observed across the routes with PM2.5 estimates up to 1.6 and 2.2 times higher, and delta C estimates up to 2.2 and 1.4 times higher than those at the monitoring station on the Courtenay-Cumberland and Courtenay-Comox routes, respectively.  3.7.3 Vanderhoof and Fraser Lake Routes Measurements from the Vanderhoof route show hotspots in the northwest and northeast (just south of the Nechako river), and in the neighbourhoods north of the Nechako river (Figure 3-16). The hotspots in the northwest of the map and northwest of the central area of Vanderhoof were more pronounced on the nephelometer map compared with the aethalometer map. Very high Bsp levels were measured in some sections of the Vanderhoof route, with the highest 48  estimated PM2.5 up to 18 times higher than those around the fixed site. Nephelometer measurements on the Fraser Lake route showed hotspots in the north and west of Fraser Lake, as well as the western end of Fort Fraser (Figure 3-17). When contrasting against the aethalometer map, the hotspots are less pronounced in Fraser Lake with lower relative values, but the hotspot in Fort Fraser had higher z-scores. Differences are also observed on the Vanderhoof segment of the route, with lower relative delta C values along the highway section, compared to the relative nephelometer measurements. The PM2.5 estimates calculated for the fixed Vanderhoof station were lower than the averages across both the Vanderhoof route (Z = -0.61) and the Fraser Lake route (Z = -0.65). However, the station fell closer to the centre of the delta C distributions, slightly above the average of the Vanderhoof route (Z = 0.09) and slightly below the average of the Fraser Lake route (Z = -0.22).   49   Figure 3-12: Route average maps of the seven nighttime runs across the Whistler route.  Map A shows the average spatial PM2.5 patterns estimated from Bsp measured by the nephelometer. Map B shows the average spatial patterns of delta C measured by the aethalometer. The route is shaded based on average z-score, showing the relative average values of each variable during the nighttime monitoring runs. The location of the Whistler monitoring station is identified by the blue circle with the average z-score of the closest cell. The z-score break points between shading bins along with the mean z-score measured at the fixed site location are converted to PM2.5 and delta C estimates (in units of µg/m3) for the nephelometer and aethalometer data respectively.  Mean Z  Score -1.5 -1 -0.5 0 0.5 1 1.5 PM2.5 (Estimate) 5.0 6.0 7.5 9.7 13 19 27  Whistler Station Z = -0.14 PM2.5 = 9 μg/m3 Mean Z  Score -1.5 -1 -0.5 0 0.5 1 1.5 delta C (Estimate) 0.2 0.3 0.6 1.1 2.0 3.8 7.0  Whistler Station Z = -0.16 delta C = 0.9 μg/m3 A B 50   Figure 3-13: Route average maps of the seven nighttime runs across the Pemberton route. Map A shows the average spatial PM2.5 patterns estimated from Bsp measured by the nephelometer. Map B shows the average spatial patterns of delta C measured by the aethalometer. The route is shaded based on average z-score, showing the relative average values of each variable during the nighttime monitoring runs. The uninhabited highway section of the route between Whistler and Pemberton was removed prior to calculation of z-scores and is also cropped from this map. The location of the Whistler monitoring station is identified by the blue circle with the average z-score of the closest cell. The z-score break points between shading bins along with the mean z-score measured at the fixed site location are converted to PM2.5 and delta C estimates (in units of µg/m3) for the nephelometer and aethalometer data respectively. Mean Z  Score -1.5 -1 -0.5 0 0.5 1 1.5 PM2.5 (Estimate) 6.0 7.8 11 16 24 38 61  Whistler Station Z = -0.87 PM2.5 = 8.4 μg/m3 Mean Z  Score -1.5 -1 -0.5 0 0.5 1 1.5 delta C (Estimate) 0.2 0.4 0.8 1.5 2.7 5.1 9.5  Whistler Station Z = -0.35 delta C = 0.95 μg/m3 Pemberton  Mount Currie  Pemberton  Mount Currie  A B 51   Figure 3-14: Route average maps of the seven nighttime runs across the Courtenay-Cumberland route. Map A shows the average spatial PM2.5 patterns estimated from Bsp measured by the nephelometer. Map B shows the average spatial patterns of delta C measured by the aethalometer. The route is shaded based on average z-score, showing the relative average values of each variable during the nighttime monitoring runs. The location of the Courtenay monitoring station is identified by the blue circle with the average z-score of the closest cell. The z-score break points between shading bins along with the mean z-score measured at the fixed site location are converted to PM2.5 and delta C estimates (in units of µg/m3) for the nephelometer and aethalometer data respectively. Courtenay  Cumberland  Mean Z  Score -1.5 -1 -0.5 0 0.5 1 1.5 delta C (Estimate) 0.3 0.5 1.1 2.0 4.1 8.0 16  Courtenay Station Z = 0.38 delta C = 3.4 μg/m3 Mean Z  Score -1.5 -1 -0.5 0 0.5 1 1.5 PM2.5 (Estimate) 5.7 8.3 12 18 28 43 66  Courtenay Station Z = 0.41 PM2.5 = 26 μg/m3 Courtenay  Cumberland  Royston  Royston  A B 52   Figure 3-15: Route average maps of the seven nighttime runs across the Courtenay-Comox route. Map A shows the average spatial PM2.5 patterns estimated from Bsp measured by the nephelometer. Map B shows the average spatial patterns of delta C measured by the aethalometer. The route is shaded based on average z-score, showing the relative average values of each variable during the nighttime monitoring runs. The location of the Courtenay monitoring station is identified by the blue circle with the average z-score of the closest cell. The z-score break points between shading bins along with the mean z-score measured at the fixed site location are converted to PM2.5 and delta C estimates (in units of µg/m3) for the nephelometer and aethalometer data respectively. Mean Z  Score -1.5 -1 -0.5 0 0.5 1 1.5 delta C (Estimate) 0.1 0.2 0.4 0.7 1.3 2.5 4.7  Courtenay Station Z = 1.04 delta C = 2.6 μg/m3 Courtenay  Comox  Mean Z  Score -1.5 -1 -0.5 0 0.5 1 1.5 PM2.5 (Estimate) 2.5 3.3 4.7 6.8 10 15 24  Courtenay Station Z = 1.82 PM2.5 = 32 μg/m3 Courtenay  Comox  A B 53   Figure 3-16: Route average maps of the six nighttime runs across the Vanderhoof route. Map A shows the average spatial PM2.5 patterns estimated from Bsp measured by the nephelometer. Map B shows the average spatial patterns of delta C measured by the aethalometer. The route is shaded based on average z-score, showing the relative average values of each variable during the nighttime monitoring runs. The location of the Vanderhoof monitoring station is identified by the blue circle with the average z-score of the closest cell. The z-score break points between shading bins along with the mean z-score measured at the fixed site location are converted to PM2.5 and delta C estimates (in units of µg/m3) for the nephelometer and aethalometer data respectively. Mean Z  Score -1.5 -1 -0.5 0 0.5 1 1.5 delta C (Estimate) 0.2 0.4 0.6 1.1 1.9 3.4 6.0  Vanderhoof Station Z = 0.09 delta C = 1.2 μg/m3 Mean Z  Score -1.5 -1 -0.5 0 0.5 1 1.5 PM2.5 (Estimate) 3.0 4.8 8.4 15 28 51 96  Vanderhoof Station Z = -0.61 PM2.5 = 7.4 μg/m3 A B 54   Figure 3-17: Route average maps of the seven nighttime runs across the Fraser Lake route. Maps A1-4 show the average spatial PM2.5 patterns estimated from Bsp measured by the nephelometer. Maps B1-4 show the average spatial patterns of delta C measured by the aethalometer. The route is shaded based on average z-score, showing the relative average values of each variable. Segments 1-4 show the four inhabited areas of the route used to calculate z-scores (mostly uninhabited sections of highway between these areas were removed prior to analysis). Segment 4 is presented on a smaller scale than segments 1-3 due to the relative sizes of communities. The Vanderhoof monitoring station is identified by the blue circle with the average z-score of the closest cell. The z-score break points between shading bins along with the mean z-score measured at the fixed site location are converted to PM2.5 and delta C estimates (in units of µg/m3) for the nephelometer and aethalometer data respectively. Mean Z  Score -1.5 -1 -0.5 0 0.5 1 1.5 PM2.5 (Estimate) 3.5 5.1 7.9 12 20 32 51  Vanderhoof Station Z = -0.65     PM2.5 = 6.9 μg/m3 A1 – Fraser Lake  A2 – Fort Fraser  A3 – Engen  A4 – Vanderhoof  Mean Z  Score -1.5 -1 -0.5 0 0.5 1 1.5 delta C (Estimate) 0.2 0.3 0.6 1.0 1.9 3.5 6.5  Vanderhoof Station Z = -0.22     delta C = 0.8 μg/m3 B1 – Fraser Lake  B2 – Fort Fraser  B3 – Engen  B4 – Vanderhoof  55   Comparison of Spatial and Temporal Variance The spatial variance measured during the nighttime mobile monitoring runs was significantly greater than the temporal variance measured by the fixed site instruments over the same time period in 71% of cases for Bsp and in 83% of cases for delta C (Table 3-3). Temporal variance was significantly greater than spatial variance in the Bsp data in 15% of runs, while this was true for delta C during only one of the 41 runs across all routes. No significant difference in variance was found in 15% of the runs when comparing the data from both instruments.  The results differed by monitoring route. The spatial variance across Whistler and across Pemberton was significantly greater than the temporal variance in 3 of 7 and 6 of 7 runs, respectively, for Bsp. For delta C this was true in 5 of 7 cases for each route. The temporal variance was significantly greater than the spatial variance in Bsp for 2 of 7 and 1 of 7 of the Whistler and Pemberton routes, respectively, while the temporal variance in delta C was never significantly greater than spatial variance on either route. The spatial variance across the Courtenay-Cumberland and Courtenay-Comox runs was significantly greater than the temporal variance in 5 of 7 and 3 of 7 runs, respectively, for Bsp and in 7 of 7 and 4 of 7 runs, respectively, for delta C. The temporal variance was not significantly greater than the spatial variance on any Courtenay-Cumberland runs when comparing either instrument. For the Courtney-Comox route the temporal variance was significantly greater during 2 of 7 and 1 of 7 runs for Bsp and delta C, respectively. Across the Vanderhoof and Fraser Lake routes, all routes showed significant differences between temporal variance and spatial variance. Across both routes, the temporal variance was only greater than the spatial variance during 1 of 6 Vanderhoof runs for Bsp, and all other comparisons showed significantly greater spatial variance.    56  Table 3-3: Results of the Fligner-Killeen test for homogeneity of variance used to compare the spatial variance measured during the nighttime mobile monitoring runs, and the temporal variance measured at the fixed location during the same time period. Results show the number of monitoring runs per route that were found to have significantly greater spatial variance, no statistically significant difference in variance, or significantly greater temporal variance for each instrument comparison. An alpha level of p = 0.05 was used to determine significance.    Nephelometer PM2.5 Indicator Aethalometer Woodsmoke Indicator Route No. Runs Greater Spatial No Significance Greater Temporal Greater Spatial No Significance Greater Temporal Whistler 7 3 2 2 5 2 - Pemberton 7 6 - 1 5 2 - Courtenay-Cumberland 7 5 2 - 7 - - Courtenay-Comox 7 3 2 2 4 2 1 Vanderhoof 6 5 - 1 6 - - Fraser Lake 7 7 - - 7 - - SUM 41 29 6 6 34 6 1 FRACTION 0.71 0.15 0.15 0.83 0.15 0.02      57  Chapter 4: Discussion The primary objective of this study was to test a cost-effective method designed to monitor levels and spatial variability of residential woodsmoke across communities. The secondary objective was to apply the method in woodsmoke-impacted communities to explore spatial patterns of total PM2.5 and woodsmoke across these areas. The optical instruments used to measure total PM2.5 and woodsmoke were compared with more established approaches at three fixed locations. The mobile monitoring method was applied across three monitored communities and three nearby unmonitored communities. This chapter first discusses research questions pertaining to the primary objective of testing this method, before discussing the results across the communities and finally making conclusions on the ability of the method to monitor residential woodsmoke.   Summary of Key Findings 4.1.1 Primary Objective - Testing the Mobile Monitoring Method The two optical instruments performed well when compared with more established methods of monitoring PM2.5 and woodsmoke concentrations at the fixed sites. Mobile monitoring using these instruments was able to capture considerable spatial variation across the communities. Comparisons of the mobile instruments show the woodsmoke-specific delta C measurements by the aethalometer added clarity to the total PM2.5 patterns measured by the nephelometer, and identified areas impacted by woodsmoke. 4.1.2 Secondary Objective - Community-Specific Findings High correlation was observed between all measures of both woodsmoke and total PM2.5 concentrations supporting the a priori expectation that the PM2.5 in the monitored communities was driven by residential wood burning during winter nights. This conclusion was strengthened by the strong diurnal patterns observed, which is typical of woodsmoke-impacted communities. Spatial variability was significantly greater than temporal variability in 29 and 34 out of 41 runs for Bsp and delta C respectively, highlighting the need for spatial monitoring of 58  woodsmoke. Despite their smaller sizes and populations, concentrations of total PM2.5 and woodsmoke in the three unmonitored communities were similar to, and in some areas higher than, those measured in the nearby monitored communities.     Can a Single-Channel Nephelometer be Used to Estimate PM2.5 Concentrations? A single-channel nephelometer was used in the mobile monitoring to estimate total PM2.5 concentrations in real-time with high temporal resolution. Because the nephelometer does not measure PM2.5 concentrations directly, the performance of this instrument was compared with PM2.5 measurements from BAMs and gravimetric filter samples at fixed locations in the three monitored communities. In general, the nephelometer performed well as a proxy for PM2.5 measurements, with strong correlation observed at all three locations between the 1-hour Bsp and BAM PM2.5 (Figure 3-3), and between the 24-hour Bsp and filter-based PM2.5 (Figure 3-4). These results suggest that the nephelometer light scattering measurements can be used to estimate total PM2.5 concentrations during mobile monitoring.  However, the relationship between Bsp and BAM PM2.5 was not entirely consistent, and the linear relationship between the instruments in Whistler had a 50% steeper slope than observed in Courtenay and Vanderhoof (Figure 3-3). Previous research has shown that the relationship between particle light scattering and PM2.5 concentrations should be established on a site- and season-specific basis (38,46). Chow et al. compared nephelometers (a similar model to the M9003 used in this research) with PM2.5 filter measurements at sites across California and found the relationship varied between sites during the winter, with average light scattering coefficients (Bsp divided by PM2.5 concentrations) similar to the relationships observed between the Bsp and filter-based PM2.5 measurements reported here (46).  While the relationship between the nephelometer and BAM differed in Whistler, the relationship between the nephelometer and filter-based measurements of PM2.5 were more consistent between the communities. As seen in Table 3-1, the average PM2.5 concentrations measured by the BAM in Whistler were double the average filter-based PM2.5 concentrations, while these methods were more similar in Courtenay and Vanderhoof. 59  One possible explanation for differences seen between communities could be the average size fraction of the particles observed in each region. Light scattering measured by the nephelometer is dominated by the smallest particles (~0.1 – 1.0 µm) (39) and so the instrument is less responsive to larger particles in the PM2.5 size range. In contrast, BAM measurements may be slightly affected by particle properties such as density, but more accurately respond to the full range of particles smaller than 2.5 µm. Filter-based measurements are indiscriminate of specific size range, because they simply measure the total mass of all particles collected below the size cut point of the collection device (in this case the Harvard Impactor removed any particles larger than 2.5 µm). Differences in the size range or density of particles observed in each airshed could therefore impact the observed relationships between the measurement methods. However, there is no clear reason why the average size of particles would be larger or have higher density in Whistler than in the other communities.  While the nephelometer performed well as a proxy measurement for PM2.5 overall, because of the observed differences between communities in the relationship between the nephelometer and BAM hourly measurements, the region-specific relationships were used to convert the results of the spatial maps created using the mobile nephelometer measurements into PM2.5 estimates during this project. In future research we would also suggest using region-specific relationships wherever possible when converting Bsp to equivalent PM2.5 concentrations to account for the potential differences between regions.    Can a Dual-Channel Aethalometer be Used to Estimate Woodsmoke Concentrations? The mobile method tested in this thesis used an aethalometer to collect source-specific information about total PM2.5, specifically by measuring the woodsmoke signal known as delta C. Strong correlation was observed between the optical and chemical tracer methods of measuring woodsmoke at the monitoring stations in each of the three airsheds, showing that delta C measurements were comparable with the more established method of measuring levoglucosan concentrations (Figure 3-5). Once again, the relationship between the two measures was not consistent across the three regions. While the relationships measured in 60  Whistler and Vanderhoof were similar (slopes of 0.68 and 0.62 respectively), the relationship between the two methods in Courtenay was steeper (slope of 0.99). Although levoglucosan is commonly used as a chemical tracer for woodsmoke, it is not necessarily released at a consistent emission factor, and the amount of levoglucosan formed during combustion can vary based on factors such as combustion temperature, the wood type, and moisture content (4,47). If consistent differences in average combustion factors exist between Courtenay and the other two locations this could be responsible for the different relationship observed between the two woodsmoke methods.  Previous research comparing the use of delta C and levoglucosan has also shown differences in this relationship by location. Wang et al. measured both during winter months in Rochester, NY and reported a slope of 0.17 (levoglucosan over delta C) with strong correlation (R2 = 0.89) (36), compared with the overall slope of 1.0 (R2 = 0.90) reported here. Harrison et al. found the slope of relationships to be 0.22 at a rural site and 0.15 at an urban site in the United Kingdom, but with much weaker correlation in both cases (R2 of 0.25 for both) (48). While both studies reported lower slopes than found in the three BC communities, they also measured much lower delta C and levoglucosan values. The mean levoglucosan and delta C concentrations measured at the three locations in this study ranged from 0.26 to 1.3 µg/m3, and 0.46 to 1.2 µg/m3 respectively. In comparison, the maximum 24-hour average measurements of levoglucosan and delta C by Wang et al. were approximately 0.1 and 0.5 µg/m3 respectively, with most measurements much lower (36), and the 90th percentiles of levoglucosan measurements reported by Harrison et al. were only 0.14 at the rural site and 0.07 at the urban site (delta C values were not reported by this study) (48). These differences in the delta C to levoglucosan relationship between locations and studies could be caused by the varying emission factor of levoglucosan based on average combustion conditions.  During the monitoring in Vanderhoof and Fraser Lake, high levels of road dust were visibly observed, adding another source of PM2.5. This is an annual occurrence in parts of the province where the winters are typically colder, and snow blankets the ground for longer periods. When the winter snow first melts after being on the ground for a long period, traction material that 61  was laid on the roads during the winter is exposed and can then be aerosolised by road traffic. The presence of road dust in the Vanderhoof / Fraser Lake route pair presented the opportunity to examine the specificity of the delta C measurement for woodsmoke. During the monitoring runs conducted on these routes, passing vehicles visibly aerosolised road dust and the nephelometer Bsp readings spiked while the aethalometer readings did not respond. This is evident in the comparison of a daytime and nighttime run on the Fraser Lake route (Figure 3-10). Because there were considerably more vehicles on the road during the daytime run, many spikes are observed in the Bsp time series that can be identified as non-woodsmoke due to the lack of relative delta C response. In contrast, very limited traffic was present on the following nighttime run and both instruments respond to most of the observed high concentrations, suggesting the PM2.5 in these areas was primarily from a biomass combustion source. This demonstrated the specificity of the delta C measure to capture woodsmoke only.    Can the Two Optical Instruments be Used Together to Estimate Woodsmoke Contribution to PM2.5 Concentrations? In previous research, chemical tracers such as levoglucosan have been monitored and compared with PM2.5 concentrations to estimate the contribution of residential woodsmoke (16,17,49). Some studies have also used the aethalometer delta C indicator (13,49,50). A number of studies have attempted to establish delta C (or other absorption measures by an aethalometer) conversion factors to estimate concentrations of woodsmoke PM2.5 (13,49,50). Conversions between levoglucosan concentrations and woodsmoke PM2.5 have also been proposed (4,49). However, these conversion factors have ranged from study to study and between locations, possibility due to aforementioned variability in levoglucosan production during combustion. Therefore, the delta C measurements made here were left unadjusted and simply used as relative measures of woodsmoke when creating maps of the mobile monitoring data and comparing relationships between measures at the fixed sites.  To explore the contribution of woodsmoke to total PM2.5 concentrations, the woodsmoke-specific and total PM2.5 measures at the fixed sites were compared. First, using the relationship 62  between the 24-hour filter-based samples (levoglucosan against total PM2.5, Figure 3-6), and then using the relationship between the 1-hour averages of the optical measures used in the mobile method (delta C against Bsp, Figure 3-7). Similar linear correlation was found for both relationships in each community and both relationships also highlighted similar differences between the communities. The relationship in Vanderhoof had a steeper slope and weaker correlation than the other two communities, which was expected given the visible observations of road dust discussed in the previous section.  The higher temporal resolution of the optical instruments compared with the filter-based measurements allowed for the comparison of this woodsmoke to total PM2.5 during different periods of the day (Figure 3-8). As expected, correlation between delta C and Bsp was stronger during nighttime hours, when residents are more likely to be at home and using wood-burning appliances. In general, it was weaker during the daytime hours, when fewer wood-burning appliances are expected to be operating (more residents away from their homes, and higher temperatures) and other PM sources such as traffic emissions may have an increased contribution. This finding was similar when comparing the mobile instruments during daytime and nighttime runs (Figure 3-9).  One notable difference between daytime and nighttime periods was observed in Vanderhoof, where many measurements were recorded with high Bsp and relatively low delta C, especially during the daytime hours at the fixed site (R2 = 0.25) and during the mobile runs (R2 = 0.0006). During daytime it is expected that woodsmoke would be reduced and road dust would be increased based on physical observations and higher traffic levels. Therefore, these findings support the ability of the comparison between the optical methods to identify times when woodsmoke is driving PM2.5 concentrations.  While the percent of woodsmoke contribution to PM2.5 concentrations was not specifically calculated using the delta C and Bsp measurements reported here, the study findings demonstrate that pairing of these two optical methods can provide useful semi-quantitative information on the influence of woodsmoke in a region. The similar relationships between delta 63  C, Bsp, levoglucosan, and PM2.5 measured at the fixed sites also shows that the combined optical methods can estimate woodsmoke contributions with similar results to the chemical tracer method that has been more extensively used in previous research. The measurement of delta C also has many advantages over the chemical tracer method, including real-time measurements with high temporal resolution, and elimination of the need for expensive and lengthy laboratory analysis of samples. While there is a large initial cost to purchase an aethalometer, operating costs are minimal.   What Information Can Mobile Monitoring Add to the Understanding of Residential Woodsmoke in Communities? A consistent finding between the route average maps (Figure 3-12 through Figure 3-17) was that substantial spatial variation was captured by this mobile monitoring method. Spatial variance captured by the mobile monitoring method was significantly greater than temporal variance captured at the fixed monitoring stations during 71% of the nighttime mobile monitoring runs when comparing Bsp, and 83% when comparing delta C (Table 3-3). This indicates that spatial variation is often more important than temporal variation when assessing community woodsmoke impacts and justifies the need for spatial monitoring to add context to the temporal data collected by routine air quality monitoring. Monitoring at a single location captures only a small piece of the picture in terms of population exposure within a region. Data collected by well-designed mobile monitoring campaigns can be used to: (1) map the average spatial patterns across a region (such as the maps presented in this thesis Figure 3-12 through Figure 3-17); (2) assess the representativeness of current fixed monitoring locations; and (3) identify hotspots of consistently elevated PM2.5 and woodsmoke concentrations.     Is There Value in Using an Aethalometer in Addition to a Nephelometer During Mobile Monitoring? This study was conducted across regions previously identified as being impacted by residential woodsmoke (20), and during times when woodsmoke was expected to be the dominant PM2.5 64  source. As a result, few differences were found between the average spatial patterns of total PM2.5 (as estimated by Bsp) and woodsmoke (as estimated by delta C) on each of the six monitoring routes (Figure 3-12 through Figure 3-17). Strong correlation was also found between the 1-minute averages of the two mobile instruments during nighttime runs on the Courtenay-Cumberland / Courtenay-Comox and Whistler / Pemberton (Figure 3-9) route pairs. This may suggest that using a nephelometer or aethalometer alone during mobile monitoring may be sufficient to measure spatial patterns during these times when woodsmoke is expected to be the dominant source. However, even in these areas and during these time periods, there was still unexplained variation in this relationship between Bsp and delta C, supporting the necessity of using both instruments in tandem. This need is more obvious during times when woodsmoke is less prevalent, such as the road dust conditions experienced in Vanderhoof. During the daytime monitoring runs, correlation between the instruments was much lower (Figure 3-9), and without the use of an aethalometer it would not be possible to identify that the high Bsp measured in Figure 3-10 was primarily not caused by PM2.5 from a woodsmoke source. This specificity of the delta C measure is valuable when attempting to measure the impact of residential woodsmoke on air quality, and it should be used in future mobile monitoring of communities where woodsmoke is expected to be a major source.   What Did We Learn About Woodsmoke in the Monitored Communities? Results of monitoring at the ENV stations in the three monitored communities show clear patterns supporting the work done by Hong et al. to identify these communities as heavily woodsmoke-impacted during the winter months (20). The diurnal pattern of BAM PM2.5 concentrations typically observed during winter months in communities impacted by residential woodsmoke were evident in each community during the sampling (Figure 3-2). In addition, the aethalometer delta C closely followed the patterns of the overall PM2.5 concentrations. This supports the theory that these diurnal patterns are a result of woodsmoke contributions.  Daily averages of all measures at the monitoring stations (both total PM2.5 and woodsmoke-specific) followed the same relative patterns over the monitoring campaign (Figure 3-1). The 65  individual comparisons between them were highly correlated, supporting the assumption that woodsmoke was the dominant source of PM2.5 concentrations in these communities during the winter. In Whistler and Courtenay specifically, linear relationships between levoglucosan and PM2.5 on the 24-hour filter-based samples, and between the 1-hour averages of Bsp and delta C had strong correlation (R2 ≥ 0.89) (Figure 3-5 and Figure 3-6). The relationships in Vanderhoof followed steeper slopes in both cases (i.e. lower ratio of woodsmoke to PM2.5) and had somewhat weaker correlation (R2 ~ 0.76), suggesting woodsmoke was a less dominant source in this community during the monitoring period. Visual observations and reduced correlation during daytime patterns between the fixed and mobile instruments suggested road dust was an important PM2.5 source in Vanderhoof during this period (Figure 3-8 and Figure 3-9). Periods earlier in the winter when the ground is snow covered may be less impacted by this source.   The levoglucosan measurements reported here were consistent with those made in other woodsmoke-impacted communities in BC. The average and standard deviation (SD) of levoglucosan measurements in Courtenay (mean = 1.31 µg/m3, SD = 0.76 µg/m3) were slightly lower than those measured by Weichenthal et al. in the same community during the winter of 2013/14 (mean = 1.6 µg/m3, SD = 1.3 µg/m3) (51). This same study also measured much lower levoglucosan concentrations in Prince George, BC (the closest large community to Vanderhoof) during the same winter (mean = 0.1 µg/m3, SD = 0.1 µg/m3). Another study by Millar et al. measured levoglucosan during the winter heating season across a number of small BC communities to the northwest of the Vanderhoof / Fraser Lake route pair, with mean concentrations in these communities ranging from 0.27 µg/m3 to 1.29 µg/m3 (16). The average concentrations in Vanderhoof (mean = 0.26 µg/m3, SD = 0.19 µg/m3) were double those measured by Weichenthal et al. (2017) in Prince George, which is a larger and more urban centre, but were similar to the nearby communities measured by Millar et al. (2012). The average concentrations in Whistler (mean = 0.36 µg/m3, SD = 0.31 µg/m3) were also within this range. All three averages were above the levoglucosan concentrations measured in more urban and less woodsmoke-impacted areas in the cities around greater Vancouver by Larson et al. in the winter of 2004/05 (17).  66  While the average measurements in Courtenay reported here and by Weichenthal et al (52). have been relatively higher than other measurements across BC, there are still areas where much higher average values have been measured. Bergauff et al. measured average levoglucosan concentrations of 3.0 µg/m3 in Libby, Montana in 2004 before a comprehensive woodstove exchange program where approximately 1200 older stoves (in a community of 2700 people) were replaced with certified new stoves or alternative heating options. After the exchange, levoglucosan concentrations had been reduced by 50% with average concentrations of 1.5 µg/m3 by the winter of 2006/07 (53). Even following this large air quality improvement, concentrations in Libby, Montana were still similar to the average concentrations that have been measured in Courtenay. Results of the mobile monitoring showed that residential woodsmoke dominated the spatial variability in PM2.5 concentrations as well as the temporal variability. The average route maps calculated for each instrument showed similar patterns for the PM2.5 estimates from the Bsp measurements and for the delta C measurements (Figure 3-12 through Figure 3-17). Representativeness of the monitoring station locations is discussed in the following section.  What Did We Learn About the Spatial Patterns Across Each Community? 4.8.1 Whistler and Pemberton The average patterns calculated across Whistler (Figure 3-12) mostly followed the residential areas on both maps, with higher levels observed in the denser area of Whistler to the east of Alta Lake, and lower levels on the west of the lake where there are fewer homes. The location of the Whistler monitoring station was slightly below the average across the Whistler route using both instruments. This was despite being located next to the neighbourhood of Alpine Meadows, which was identified as the clearest hotspot on both maps. However, the wind roses for this monitoring station (Figure 3-11) show winds mainly came from the south and not from the direction of this hotspot. This hotspot is a neighbourhood of relatively older homes and likely has a higher concentration of wood burning appliances. This contrasts with the loop in the far southwest of the route, where the lowest levels were measured in the Cheakamus Crossing 67  neighbourhood. This neighbourhood was developed as a sustainable community for the Vancouver 2010 Winter Olympics and consists of new homes heated by a District Energy System (54). As such, it is unsurprising to find better air quality in this area during these winter evenings. While the winds were relatively calm during monitoring, they were predominantly from the southerly direction, which likely contributed to these lower levels at the south of the map.  The Whistler monitoring station was not representative of the conditions in Pemberton during the monitoring campaign, with higher levels observed across the Pemberton valley during the nighttime monitoring runs (Figure 3-13). As with the Whistler route, substantial variation was observed across the Pemberton route, and three hotspots were identified by both instruments. The first was within the densest part of Pemberton on the west of the maps, while the others were within the Lil’wat First Nation communities of Mt. Currie and Xit’olacw. At the time of writing there are no gas lines connected to the Pemberton valley, so residential heating options in this area are restricted to electric heat, imported propane or fuel oil, and wood burning appliances. This likely leads to higher rates of wood burning for home heating. 4.8.2 Comox Valley Communities The nephelometer and aethalometer maps showed similar smoke hotspots throughout the Courtenay-Cumberland route, particularly in the centre and southeast of Courtenay and the northeast of Cumberland, both of which are quite dense residential areas (Figure 3-14). Winds during this route were predominantly from the west (Figure 3-11) and may have contributed to the Cumberland values being higher in the eastern part of the community. Similar hotspots were again identified by both instruments on the Courtenay-Comox route, with high concentrations in downtown Courtenay and areas around the monitoring station, along with a small residential area in the northwest of the map (Figure 3-15). A previous mobile monitoring campaign in the Comox Valley in 2009 did not produce average patterns across their monitoring runs, but individual route maps appear to highlight similar hotspots in the centre and southeast 68  of Courtenay, along with the northeast of Cumberland, as well as a similar trend of higher concentrations in Courtenay compared with Comox (18).  The Courtenay monitoring station fell above the averages for both instruments on both routes, especially on the Courtenay-Comox route where the z-scores were 1.82 and 1.04 for the PM2.5 and delta C maps, respectively. This result suggests that the location of the monitoring station in Courtenay may read higher than many areas across the airshed during winter nights. Despite the relatively high z-scores around the monitoring station, other areas of both routes had even higher average values. Comparing these routes is difficult as they were monitored on different nights with quite different average conditions. However, the higher z-score of the monitoring station on the eastern Courtenay-Comox map relative to the Courtenay-Cumberland map suggests that many areas of the Courtenay-Comox route have lower average concentrations than the areas of the Courtenay-Cumberland route that have relative PM2.5 and woodsmoke concentrations greater than at the monitoring station. This highlights the importance of route mapping in the overall method. Because it is difficult to compare directly between routes, monitoring should be designed to cover an entire community of interest as thoroughly as possible.  The high z-scores of the Courtenay monitoring station raise the question of how such locations are identified. Is the objective to provide information about the average exposure of the community? Or is the objective to measure conditions in highly-impacted areas to inform air quality improvements? This method cannot answer these questions, but it highlights the amount of additional information that spatial monitoring can add to temporal monitoring at a single site, which can be invaluable when making decisions regarding regulatory air quality monitoring in an airshed.  4.8.3 Vanderhoof and Fraser Lake The Vanderhoof maps also showed spatial patterns mostly following the population density in the area, with the less populated outlying areas to the south showing lower z-scores for both instruments (Figure 3-16). Areas north of the highway had higher concentrations than those to 69  the south, and hotspots were visible in the northeast and northwest of the route section south of the Nechako river. The hotspot in the far northwest of the PM2.5 map was much less extreme on the delta C map. Although woodsmoke from the surrounding homes contributed to the PM2.5 measured in this area (as it is slightly elevated on the delta C map), this section of road was unpaved, and dust was likely responsible for the PM2.5 hotspot. In addition to Fraser Lake, this route covered the smaller community of Fort Fraser and the very small neighbourhood of Engen. The maps from this route show small hotspots in the north and west of Fraser Lake, as well as the western end of Fort Fraser (Figure 3-17). The section of Highway 16 entering and leaving Vanderhoof shows lower z-scores on the aethalometer map, which may indicate that higher values on the nephelometer map were caused by road dust on this section.  Average concentrations at the Vanderhoof station were below the average PM2.5 concentrations on both routes, but similar to the average delta C values. This difference in z-scores calculated around the monitoring station by the two instruments is not as surprising given the lower correlation between the two instruments during mobile monitoring in this region (Figure 3-9). The PM2.5 levels around the station could be lower than the route averages due to very high values calculated in specific areas of the route increasing the route mean, even though the data were log-transformed. The maximum cell values on the map were up to 18 times the cell value at the fixed site on the Vanderhoof route, and five times higher on the Fraser Lake route. The Vanderhoof monitoring station is surrounded by commercial buildings and a school with a large open field, with few residential buildings and wood burning appliances nearby. Even so, the monitoring station fell near the mean of the delta C route average map. This finding is likely influenced by the areas included in the route, with the low delta C averages of the more outlying areas to the south acting to decrease the route mean. When examining the spatial patterns around the monitoring station more closely, the station location had some of the lowest average delta C values in the densest section of the community that is between Highway 16 and the Nechako river. This matches the expectation of lower values around the monitoring station (due to limited wood burning nearby, as described above), and the pattern of concentrations following residential areas as the rest of this area is mostly residential. This 70  again highlights the importance of planning routes that are as representative as possible of the entire community when displaying results on a relative scale.  What Did We Learn About Woodsmoke in the Unmonitored Communities? In general, the unmonitored communities had concentrations comparable with and, in some areas, greater than the monitored communities. The average of the PM2.5 and delta C estimates generated for the maps of the Pemberton and Fraser Lake routes were all higher than the relative values at the location of the monitoring stations in the paired monitored community (Figure 3-13 and Figure 3-17). While Cumberland was not covered by an independent route, the Bsp and delta C values of the hotspot from the center to the northeast of the community were higher than the average values measured around the Courtenay monitoring station and were comparable with identified hotspots in Courtenay.  Considerable spatial variability in both Bsp and delta C was also measured across the unmonitored communities. Spatial variance was considerably greater than temporal variance at the nearest monitoring station during six of the seven Pemberton runs when comparing Bsp measurements, and five of seven when comparing the delta C measurements. The same comparison on the Fraser Lake route found significantly greater spatial variance than temporal variance during all runs for both mobile instruments (Table 3-3). The spatial patterns across the PM2.5 and delta C maps for each of these unmonitored communities were similar, suggesting woodsmoke was driving PM2.5 during these winter nights.  These results show that residential woodsmoke is an air quality concern in unmonitored communities, and that the mobile methods tested here can be used to assess the impacts. Data on the range of values across these communities, the spatial variability, and the PM2.5 and woodsmoke hotspots can be particularly valuable for communities without other forms of monitoring.   71   Conclusions on Mobile Method The mobile monitoring method developed in this thesis using both a nephelometer and aethalometer effectively captured valuable spatial data on approximate PM2.5 concentrations and the contributions of residential woodsmoke across a region. The testing of this method in three community pairs showed that it can be applied to complement existing fixed site monitors or to quickly characterise approximate conditions in otherwise unmonitored areas. Strengths and limitations of the method are discussed below. 4.10.1 Strengths of the Method This mobile monitoring method has many strengths. First, it was able to capture significant spatial variability with high resolution across the testing communities, such that the average patterns of relative PM2.5 and woodsmoke could be mapped. This allowed for the identification of hotspot areas with consistently elevated values. The relatively novel use of a dual-channel aethalometer in mobile monitoring (previously tested by Allen et al. (13)) and the measurement of the woodsmoke indicator delta C in addition to Bsp can provide strong evidence that woodsmoke is a significant source in a region. Both instruments used in the mobile method were well-correlated with more established measurement methods and the Bsp measured by the nephelometer could be converted to estimated PM2.5 concentrations based on site-specific relationships.   In addition to providing spatial context around established monitoring stations, the testing of this method in three unmonitored communities showed that it is a good option for characterising PM2.5 and residential woodsmoke in such areas. By itself, this method can give an overview of spatial variability across a community, comparing relative patterns of total PM2.5 and woodsmoke, and it can provide estimates of the true exposures within these communities. While measurements can only be considered semi-quantitative (as they are only measured over a short time period), data collected in this way could be compared with nearby monitored communities and adjusted using temporal patterns in those areas.  72  Two other strengths of this method are its lower cost and ease of use compared with chemical analysis and fixed site monitoring. While the initial cost of an aethalometer similar to that used in this mobile method is considerable at approximately $30,000, operating costs of this instrument are negligible. In contrast, analysis of each filter sample for levoglucosan costs $95 per filter. This led to a total cost of approximately $45,000 over the course of this study, greater than the initial cost of the aethalometer (which can continue to be used in future research). Therefore, the aethalometer is a cheaper option to monitor woodsmoke concentrations than levoglucosan, and as shown in this research, can provide comparable results with much greater resolution. Personnel costs are higher for mobile monitoring than fixed site monitoring as a longer time commitment is required each day (approximately three hours per mobile monitoring run, compared to one hour to prepare impactors and change filters at a fixed site). However, the ease of use of the instruments used in this method limited this cost as only one operator was necessary for mobile monitoring runs (in this study I conducted all mobile monitoring alone) and the added benefit of high spatial resolution can be invaluable. The ease of use of the method also presents the opportunity to reduce the cost of the method further by making the method and equipment available for community volunteers to use as part of a citizen science project. Involving community members with air quality monitoring in this way will also increase engagement with air quality issues on a local scale. These strengths make the method an option moving forward to apply in small communities lacking monitoring across Canada. 4.10.2 Limitations of the Method Due to several limitations of this mobile monitoring method, the values associated with the resulting route average maps are semi-quantitative. Primarily, the method is limited in its inability to monitor whole communities simultaneously. This introduces error into the measurements, because temporal variation cannot be measured by mobile instruments and many other unmeasured variables may affect the data. In the most extreme example, all of the variability measured by the mobile instruments could actually be due to temporal variation rather than spatial differences. Ideally, any such air quality assessment would incorporate both 73  high spatial and high temporal resolution by having many continuous monitors installed across a region. However, this is generally not feasible due to the associated costs of purchasing and operating so many instruments. While the method used here was designed to provide a more cost-efficient option, it is limited by its inability to control for the temporal variation at all measured locations.  Steps were taken during the development of this method to minimise limitations related to temporal variation. First, the monitoring schedule was focused on a specific time of day, and seven monitoring runs were driven on each route to assess average spatial patterns rather than short-term patterns. Second, the driving direction of each route was alternated to reduce the impact of any consistent temporal patterns by averaging these out across multiple trips.  While high spatial resolution can be achieved with the 1-second measurements recorded by the mobile aethalometer, the light scattering estimation of total PM2.5 recorded by the nephelometer is limited by instrumental smoothing. The nephelometer used in the mobile monitoring measured light scattering of the current air sample, but it took approximately 23 seconds for the air to be completely replaced. This lag reduces the immediacy of response and naturally smooths the 1-second measurements. However, by recording the raw 1-second readings using a serial connection to a laptop (rather than using the 1-minute averages reported by the instrument), the data were more accurately assigned to the locations at which they were measured before the spatial averaging.  The difference in independence of the individual measurements between the instruments is visible when comparing the final route average maps. Increased variation from cell-to-cell is observed in the aethalometer maps (most clearly seen in straight road sections) compared with the nephelometer maps. Additionally, areas that are high or low on both maps were often more extreme in either direction on the nephelometer maps due to the instrumental smoothing of Bsp values. While driving direction was alternated primarily to avoid monitoring the same areas at the same time of day, it also served to reduce the impact of this instrumental smoothing on 74  the nephelometer maps because values were smoothed in each direction from areas with consistently higher or lower readings.  The methods chosen to calculate and represent the route average maps also have a number of limitations. Efforts were initially made to adjust the mobile data for temporal variation at the fixed locations (using methods similar to Equation 2-1), but this approach led to over exaggeration of heavily impacted areas in the route average maps, and there is no evidence to show that the temporal variation at the fixed locations can be extrapolated across the region for this purpose. The method of calculating z-scores for the data recorded by the two instruments during each monitoring run was chosen to extract the relative value of each measurement. Because z-scores are relative values, the route average maps show how each area compares with the rest of the route, which increases the importance of the monitoring route design. Using the z-scores method also required log-transforming the mobile data, which meant the legends of the route average maps are based on an exponential scale and are more difficult to interpret for community members. The calculation of the z-score equivalents for the map legends are only rough approximations, because the reverse calculation was performed on different distributions than were used to calculate the z-scores. Specifically, the z-scores were calculated for each run and then averaged, while the map legends were calculated using the data from all runs. Despite these limitations, the use of z-scores removes the need for a fixed monitor in future applications of this method, increasing the feasibility of its implementation in a wide range of communities.  75  Chapter 5: Conclusion This thesis successfully met the primary objective of testing the ability of a cost-effective mobile monitoring method to characterise residential woodsmoke within affected communities. It also met the secondary objective of measuring the average spatial patterns and influence of woodsmoke in three sets of monitored and unmonitored communities in BC.  The method was able to measure spatial variability with high resolution in both total PM2.5 using a nephelometer and in woodsmoke using a multi-channel aethalometer. These two instruments performed well when compared with more established methods of monitoring total PM2.5 (BAM and filter-based measurements) and woodsmoke (levoglucosan). When used together, they provided valuable information on the relative contribution of woodsmoke to PM2.5, both temporally at the fixed site monitoring stations and spatially as part of the mobile method. The mobile method also performed well in unmonitored communities and should be a valuable tool to quickly and cost-effectively characterise air quality for other such communities in the future.  Residential woodsmoke was the dominant driver of PM2.5 in each of the community pairs, with strong correlation between relative woodsmoke and PM2.5 concentrations both temporally and spatially. Diurnal patterns of PM2.5 commonly observed in woodsmoke-impacted communities were evident in the three monitored communities, and the delta C measurements indicated that daily peaks were caused by wood combustion. Woodsmoke was somewhat less dominant in Vanderhoof, with more unexplained variation in the woodsmoke-PM2.5 relationships than observed in Whistler and Courtenay. This was attributed to the effect of road dust as a secondary major source of PM2.5 during the monitoring period. The highest average concentrations of each measurement at the fixed sites were measured in Courtenay, where comparatively high average levoglucosan concentrations have also been measured in previous research (29).   76  Testing of the mobile method in three unmonitored communities found PM2.5 and woodsmoke concentrations comparable with those in the paired monitored communities, despite their smaller sizes and populations. In many areas these concentrations were higher than the concentrations measured at the nearest monitoring stations.  Significant spatial variation measured by the mobile method across each monitoring route showed how one fixed monitoring station in a community is unable to capture the range of air quality experienced within a community at any given time. Spatial variance in total PM2.5 concentrations was significantly greater than temporal variance at the monitoring stations during 71% of the nighttime monitoring runs across all routes. For woodsmoke this was true in 83% of cases. This demonstrates the importance of measuring spatial variability when monitoring air quality, especially in communities where residential woodsmoke is an important source, and it justifies the need for spatial monitoring using methods similar to those tested here.  Identification of woodsmoke hotspots can be valuable for communities and those responsible for air quality management. Such data can be used to inform and target the design of source-control efforts to improve air quality in a region. The highly resolved spatial maps achievable with this method could also be combined with health outcomes data in future research. Following the success of this method in the field monitoring campaign during the winter of 2017, plans were made to make the method and necessary equipment available to community groups across the province. Training materials were created along with computer programs that simplify set-up during monitoring and automatically produce route average maps based on the monitoring data. This makes the method as accessible as possible for users with limited data collection and analyses experience. These tools were successfully trialled with volunteers applying them in both Valemount and Golden during early 2018, with minimal support from project investigators. The methods will be made more widely available to other groups as of winter 2019.  77  References 1.  BC Ministry of Environment. Inventory of Wood-burning Appliance Use in British Columbia Report of Findings. 2012.  2.  McDonald JD, Zielinska B, Fujita EM, Sagebiel JC, Chow JC, Watson JG. Fine Particle and Gaseous Emission Rates from Residential Wood Combustion. Environ Sci Technol [Internet]. 2000 Jun;34(11):2080–91. Available from: http://pubs.acs.org/doi/abs/10.1021/es9909632 3.  Naeher LP, Brauer M, Lipsett M, Zelikoff JT, Simpson CD, Koenig JQ, et al. Woodsmoke Health Effects: A Review. Inhal Toxicol [Internet]. 2007;19(1):67–106. Available from: http://www.tandfonline.com/doi/full/10.1080/08958370600985875 4.  Fine PM, Cass GR, Simoneit BRT. Chemical Characterization of Fine Particle Emissions from the Fireplace Combustion of Wood Types Grown in the Midwestern and Western United States. Environ Eng Sci. 2004;21(6):705–21.  5.  Bäfver LS, Leckner B, Tullin C, Berntsen M. Particle emissions from pellets stoves and modern and old-type wood stoves. Biomass and Bioenergy. 2011;35(8):3648–55.  6.  Ward T, Noonan C. Results of a residential indoor PM2.5 sampling program before and after a woodstove changeout. Indoor Air. 2008;18(5):408–15.  7.  Allen RW, Leckie S, Millar G, Brauer M. The impact of wood stove technology upgrades on indoor residential air quality. Atmos Environ [Internet]. Elsevier Ltd; 2009;43(37):5908–15. Available from: http://dx.doi.org/10.1016/j.atmosenv.2009.08.016 8.  BC Government. Solid Fuel Burning Domestic Appliance Regulation [Internet]. 1994. Available from: http://www.bclaws.ca/Recon/document/ID/freeside/44_302_94   78  9.  BC Ministry of Environment. Factsheet on the Solid Fuel Burning Domestic Appliance Regulation [Internet]. 2017. Available from: https://www2.gov.bc.ca/assets/gov/environment/air-land-water/air/reports-pub/sfbdar_factsheet.pdf 10.  City of Duncan. City of Duncan Wood Burning Appliances and Air Quality Bylaw No . 3089, 2013 [Internet]. 2013. Available from: https://duncan.civicweb.net/document/37471 11.  Glasius M, Ketzel M, Wåhlin P, Jensen B, Mønster J, Berkowicz R, et al. Impact of wood combustion on particle levels in a residential area in Denmark. Atmos Environ. 2006;40(37):7115–24.  12.  Hellén H, Hakola H, Haaparanta S, Pietarila H, Kauhaniemi M. Influence of residential wood combustion on local air quality. Sci Total Environ. 2008;393(2):283–90.  13.  Allen GA, Miller PJ, Rector LJ, Brauer M, Su JG. Characterization of Valley Winter Woodsmoke Concentrations in Northern NY Using Highly Time-Resolved Measurements. Aerosol Air Qual Res [Internet]. 2011;11(5):519–30. Available from: http://www.aaqr.org/Doi.php?id=6_AAQR-11-03-OA-0031&v=11&i=5&m=10&y=2011 14.  Environment and Climate Change Canada. Air Pollutant Emission Inventory 2016 - Online Data Search [Internet]. 2016 [cited 2018 Jul 15]. Available from: https://pollution-waste.canada.ca/air-emission-inventory 15.  BC Lung Association. BC State of the Air Report 2015 [Internet]. 2015. Available from: https://bc.lung.ca/protect-your-lungs/air-quality-lung-health/bc-state-air-report 16.  Millar G. Characterizing residential wood smoke at the neighbourhood scale: An evaluation of five communities in British Columbia [Internet]. University of Northern British Columbia; 2012. Available from: https://unbc.arcabc.ca/islandora/object/unbc%3A16207  79  17.  Larson T, Su J, Baribeau A-M, Buzzelli M, Setton E, Brauer M. A Spatial Model of Urban Winter Woodsmoke Concentrations. Environ Sci Technol [Internet]. 2007;41(7):2429–36. Available from: http://dx.doi.org/10.1021/es0614060 18.  Poplawski K, Setton E. Monitoring of Fine Particulate Matter (PM 2.5) in the Comox Valley Regional District , BC [Internet]. Victoria, BC; 2009. Available from: http://web.uvic.ca/~ssrl01/SSRLtemp/CVRD Fine Particulates.pdf 19.  Henderson SB. BC Clean Air Research Fund - Final Report on 2013-14 Mobile Monitoring. 2014.  20.  Hong KY, Weichenthal S, Saraswat A, King GH, Henderson SB, Brauer M. Systematic identification and prioritization of communities impacted by residential woodsmoke in British Columbia , Canada. Environ Pollut [Internet]. Elsevier Ltd; 2017;220:797–806. Available from: http://dx.doi.org/10.1016/j.envpol.2016.10.056 21.  Hoek G, Krishnan RM, Beelen R, Peters A, Ostro B, Brunekreef B, et al. Long-term air pollution exposure and cardio- respiratory mortality: a review. Environ Heal. 2013;12(1):43.  22.  Bell ML, Zanobetti A, Dominici F. Evidence on vulnerability and susceptibility to health risks associated with short-term exposure to particulate matter: a systematic review and meta-analysis. Am J Epidemiol. 2013;178(6):865–76.  23.  Apte JS, Marshall JD, Cohen AJ, Brauer M. Addressing Global Mortality from Ambient PM 2.5. Environ Sci Technol. 2015;49:8057–66.  24.  World Health Organization. Health Effects of Particulate Matter - Policy Implications for Countries in Eastern Europe, Caucasus and Central Asia [Internet]. 2013. Available from: http://www.euro.who.int/__data/assets/pdf_file/0006/189051/Health-effects- of-particulate-matter-final-Eng.pdf   80  25.  Sigsgaard T, Forsberg B, Annesi-Maesano I, Blomberg A, Bølling A, Boman C, et al. Health impacts of anthropogenic biomass burning in the developed world. Eur Respir J [Internet]. 2015;46(6):1577–88. Available from: http://erj.ersjournals.com/lookup/doi/10.1183/13993003.01865-2014 26.  Kocbach Bølling A, Pagels J, Yttri KE, Barregard L, Sallsten G, Schwarze PE, et al. Health effects of residential wood smoke particles: the importance of combustion conditions and physicochemical particle properties. Part Fibre Toxicol. 2009;6(29):1–20.  27.  Kim YH, Warren SH, Krantz QT, King C, Jaskot R, Preston WT, et al. Mutagenicity and Lung Toxicity of Smoldering vs. Flaming Emissions from Various Biomass Fuels: Implications for Health Effects from Wildland Fires. Environ Health Perspect. 2018;126(1):1–14.  28.  Alessandrini ER, Stafoggia M, Faustini A, Berti G, Canova C, Togni A De, et al. Association Between Short-Term Exposure to PM2.5 and PM10 and Mortality in Susceptible Subgroups : A Multisite Case-Crossover Analysis of Individual Effect Modifiers. Am J Epidemiol. 2016;184(10):744–54.  29.  Weichenthal S, Lavigne E, Evans G, Pollitt K, Burnett RT. Ambient PM2.5 and risk of emergency room visits for myocardial infarction: impact of regional PM2.5 oxidative potential: a case-crossover study. Environ Heal [Internet]. Environmental Health; 2016;15(1):46. Available from: http://ehjournal.biomedcentral.com/articles/10.1186/s12940-016-0129-9 30.  Shafizadeh F, Sekiguchi Y. Oxidation of chars during smoldering combustion of cellulosic materials. Combust Flame. 1984;55(2):171–9.  31.  Fraser MP, Lakshmanan K. Using Levoglucosan as a Molecular Marker for the Long-Range Transport of Biomass Combustion Aerosols. Environ Sci Technol. 2000;34(21):4560–4.  32.  Jeong CH, Hopke PK, Kim E, Lee DW. The comparison between thermal-optical transmittance elemental carbon and Aethalometer black carbon measured at multiple monitoring sites. Atmos Environ. 2004;38(31):5193–204.  81  33.  Park K, Chow JC, Watson JG, Trimble DL, Doraiswamy P, Park K, et al. Comparison of Continuous and Filter-Based Carbon Measurements at the Fresno Supersite. J Air Waste Manage Assoc [Internet]. 2006;56(4):474–91. Available from: http://www.tandfonline.com/doi/abs/10.1080/10473289.2006.10464521 34.  Allen G, Babich P, Poirot RL. Evaluation of a New Approach for Real Time Assessment of Wood Smoke PM. Air Waste Manag Assoc Visibility Spec Conf Reg Glob Perspect Haze Causes, Consequences Controv. 2004;1–11.  35.  Zhang KM, Allen G, Yang B, Chen G, Gu J, Schwab J, et al. Joint measurements of PM 2.5 and light-absorptive PM in woodsmoke-dominated ambient and plume environments. Atmos Chem Phys. 2017;17:11441–52.  36.  Wang Y, Hopke PK, Rattigan O V., Xia X, Chalupa DC, Utell MJ. Characterization of residential wood combustion particles using the two-wavelength aethalometer. Environ Sci Technol. 2011;45(17):7387–93.  37.  Robinson DL, Monro JM, Campbell EA. Spatial variability and population exposure to PM2.5 pollution from woodsmoke in a New South Wales country town. Atmos Environ. 2007;41(26):5464–78.  38.  Keywood MD, Ayers GP, Gras JL, Gillett RW, Cohen D. Size Distribution and Sources of Aerosol in Launceston, Australia, during Winter 1997. J Air Waste Manage Assoc. 2000;50(3):418–27.  39.  Molnár A, Mészáros E. On the relation between the size and chemical composition of aerosol particles and their optical properties. Atmos Environ. 2001;35:5053–8.  40.  Turner J, Allen G, Brown S, Frank N. A Refined Methodology to Adjust Aethalometer Black Carbon Data for Measurement Artifacts. In: 104th Annual Meeting of the AIr & Waste Managment Association. Orlando, FL; 2011. p. 1–7.  41.  Magee Scientific. Aethalometer ® Model AE33 User Manual. 2017.   82  42.  R Foundation for Statistical Computing. R: A Language and Environment for Statistical Computing [Internet]. Vienna, Austria; 2017. Available from: www.R-project.org/ 43.  Allen G, Turner J, Frank N. WUAQL AethDataMasher 7.1 [Internet]. 2012. Available from: https://www3.epa.gov/ttnamti1/files/2012conference/3C01Allen.pdf 44.  BC Ministry of Environment and Climate Change Strategy. BC Air Quality Data [Internet]. Available from: https://www2.gov.bc.ca/gov/content/environment/air-land-water/air/air-quality/current-air-quality-data 45.  University of Bergen BioStats. Test for homogeneity of variances - Levene’s test and the Fligner Killeen test [Internet]. 2016 [cited 2018 Aug 13]. Available from: https://biostats.w.uib.no/test-for-homogeneity-of-variances-levenes-test/ 46.  Chow JC, Watson JG, Park K, Robinson NF, Lowenthal DH, Park K, et al. Comparison of Particle Light Scattering and Fine Particulate Matter Mass in Central California Comparison of Particle Light Scattering and Fine Particulate Matter Mass in Central California. J Air Waste Manage Assoc. 2006;56(4):398–410.  47.  Schmidl C, Marr I, Caseiro A, Kotianova P, Berner A, Bauer H, et al. Chemical characterisation of fine particle emissions from wood stove combustion of common woods growing in mid-European Alpine regions. Atmos Environ. 2008;42(1):126–41.  48.  Harrison RM, Beddows DCS, Hu L, Yin J. Comparison of methods for evaluation of wood smoke and estimation of UK ambient concentrations. Atmos Chem Phys. 2012;12:8271–83.  49.  Fuller GW, Tremper AH, Baker TD, Espen K, Butter D. Contribution of wood burning to PM10 in London. Atmos Environ. 2014;87:87–94.  50.  Conneticut Department of Environmental Protection. Evaluation of Wood Smoke Contribution to Particle Matter in Connecticut [Internet]. 2011. Available from: www.nescaum.org/documents/ctdep_woodsmokefinalreport-20110207.pdf/  83  51.  Weichenthal S, Crouse DL, Pinault L, Godri-Pollitt K, Lavigne E, Evans G, et al. Oxidative burden of fine particulate air pollution and risk of cause-specific mortality in the Canadian Census Health and Environment Cohort (CanCHEC). Environ Res [Internet]. Elsevier; 2016;146:92–9. Available from: http://dx.doi.org/10.1016/j.envres.2015.12.013 52.  Weichenthal S, Kulka R, Lavigne E, Rijswijk D Van, Brauer M, Villeneuve PJ, et al. Biomass Burning as a Source of Ambient Fine Particulate Air Pollution and Acute Myocardial Infarction. Epidemiology. 2017;28(3):329–37.  53.  Bergauff MA, Ward TJ, Noonan CW, Palmer CP. The effect of a woodstove changeout on ambient levels of PM2.5 and chemical tracers for woodsmoke in Libby , Montana. Atmos Environ [Internet]. Elsevier Ltd; 2009;43(18):2938–43. Available from: http://dx.doi.org/10.1016/j.atmosenv.2009.02.055 54.  cheakamuscrossing.ca. Cheakamus Crossing - A Sustainable Community [Internet]. [cited 2018 Jun 21]. Available from: http://www.cheakamuscrossing.ca/sustainibility  84  Appendices - Woodsmoke Mobile Monitoring: Full Protocol    85 Table of Contents 1. Introduction ........................................................................................................................................ 86 2. Protocol Summary .............................................................................................................................. 87 2.1 Included Equipment .................................................................................................................... 87 2.2 Instrument Overview .................................................................................................................. 88 2.3 Monitoring Campaign Overview ................................................................................................. 90 3. Preparing for Monitoring Campaign .................................................................................................. 91 3.1 Planning Monitoring Campaign Schedule ................................................................................... 91 3.2 Planning Monitoring Routes ....................................................................................................... 91 3.3 GPS Navigator Setup – Creating Monitoring Routes .................................................................. 92 3.4 Prepare a Laptop and Data Management ................................................................................. 102 3.5 Instrument Checks, Settings and Calibration ............................................................................ 104 3.5.1 Aethalometer - Magee Scientific AE33 ............................................................................. 104 3.5.2 Nephelometer – Ecotech Aurora 1000 ............................................................................. 108 3.6 Setting Up a Vehicle for Monitoring Campaign ........................................................................ 113 4. Mobile Monitoring Run Preparation ............................................................................................... 118 4.1 Heat the Vehicle and Instruments ............................................................................................ 118 4.2 Prepare the Nephelometer ....................................................................................................... 119 4.3 Prepare the Laptop ................................................................................................................... 120 4.4 Prepare the Aethalometer and GPS .......................................................................................... 123 4.5 Start the GPS Navigator ............................................................................................................ 124 4.6 Connect the Sampling Lines ...................................................................................................... 124 5. Mobile Monitoring Run .................................................................................................................... 125 5.1 Pre-Run Checklist ...................................................................................................................... 125 5.2 Driving Notes ............................................................................................................................. 126 5.3 End of Run – Data Saving and Instrument Shutdown ............................................................... 127 6. Post-Monitoring Campaign .............................................................................................................. 129 6.1 Download Back-up Data from the Nephelometer .................................................................... 129 6.2 Using Online Shiny Application to Map Your Data ................................................................... 131 6.3 Interpreting the Maps ............................................................................................................... 134 6.4 Returning Instruments .............................................................................................................. 135 7. Appendix: Manual Set-up of Nephelometer/Laptop Connection .................................................. 136    86 1. Introduction Smoke from residential wood burning is a leading contributor to fine particulate matter (PM2.5) pollution in British Columbia (BC) and has been shown to impact respiratory and cardiovascular health. During the winter heating season, many BC communities often approach and exceed provincial and national standards for PM2.5 concentrations due to high rates of wood burning combined with geographic locations where there is a tendency for inversions to form and trap cold stagnant air.  While the BC Ministry of Environment and Climate Change Strategy (ENV) air quality monitoring network collects valuable data of air pollutant concentrations (such as PM2.5) at monitoring stations across the province, it is not feasible to install and maintain multiple stations in every community. As a result, no data is collected on air quality in smaller communities, and larger communities typically have data from a single location at most. This is an important limitation as air quality can differ considerably across communities where residential wood burning is prevalent, as there are many small point sources within the community. For example, areas within a community with higher numbers of wood-burning appliances will likely have higher PM2.5 levels than other areas with limited wood burning. This spatial variation cannot be captured by the ENV air quality monitoring network and so it is hard to know whether the data collected at each ENV monitoring station accurately represents the air quality levels across that community.  To be able to measure the spatial air quality patterns across communities in the province and add context to data from the ENV network, a mobile monitoring method was developed and tested by researchers at the University of British Columbia with support from Health Canada, the BC Lung Association, and the BC Ministry of Environment. In addition to measuring air quality patterns across a community with high spatial resolution, the method was also designed for use in small communities that currently lack monitoring stations to obtain a snapshot of air quality patterns in those communities. In these smaller and more rural BC communities, wood burning is typically more prevalent and therefore measuring air quality patterns across these communities is important.  This protocol is designed to help the user implement this method and monitor air quality patterns across their community of interest.      87 2. Protocol Summary This protocol is designed to guide the user through conducting a mobile monitoring campaign to measure residential woodsmoke patterns across a community. The protocol will guide you through: planning and preparing for the campaign, operating the instruments, conducting the monitoring itself and finally data management, analysis and interpretation of results.  2.1 Included Equipment As a registered user of this method you will receive the following equipment necessary to conduct your monitoring campaign: • Aethalometer – Magee Scientific AE33 – to measure woodsmoke in air samples o Power Cord o Air Inlet Tubing and Connectors o PM2.5 Selective Cyclone o Water Trap o GPS Receiver and Cords o USB Data Stick (containing instrument manuals and for data download) o Transportation Case  • Nephelometer – Ecotech Aurora 1000 – to estimate PM2.5 in air samples o Power Cord o Serial Communication Cord and USB Adaptor o Air Inlet Tubing and Funnel o Transportation Case  • Vehicle Power Inverter – Nexxtech – to power the instruments in a vehicle during monitoring  • GPS Navigation Device – Garmin Nuvi 2497 – to provide driver with route directions o Vehicle Charging Cable o Window Holder o USB Cable     88 2.2 Instrument Overview Two optical instruments are used in this method to collect different data on fine particulate matter (PM2.5) in an air sample. A nephelometer measures an estimate of total PM2.5 concentrations, while an aethalometer measures the relative amount of woodsmoke in the air. For further information refer to the descriptions below.   Nephelometer – Ecotech Aurora 1000 A nephelometer can provide a real-time estimate of the amount of PM2.5 in the air by measuring the amount of light scattered by these particles (which is well correlated with the concentration). This effect is what causes the haze associated with smoke and is why visibility is generally reduced on smoky days.  To measure the amount of light scattering, the instrument draws an air sample into a sealed dark container, and then shines a light beam through it with sensors set at an angle to measure scattered light. If there were no particles in the air most of the light would pass straight through, but more and more of the light is scattered in different directions as the concentration of PM2.5/number of particles in the air increases. The general concept is illustrated in Figure 1 below.       Figure 1: Nephelometer Function Diagram. Light is shined through the air sample, and the amount of light that is scattered by particles in the air is measured.   89 Aethalometer – Magee Scientific AE33 An aethalometer is used to provide more information about the chemical make-up (and therefore the potential source) of a PM2.5 sample rather than just measuring the total amount. To collect data on just PM2.5 (particles smaller than 2.5µm in diameter), a size selective cyclone is attached to the air inlet tubing to filter out any particles larger than this. The aethalometer deposits PM2.5 from sample air onto a quartz filter tape, and then shines multiple wavelengths of light through the sample to measure how much of each wavelength is absorbed by the sample. The differences in absorption at different wavelengths can tell us a lot about the chemical composition and potential source of a sample. In this case we are interested in the amount of PM2.5 that is created by residential wood combustion. While the aethalometer measures absorption of seven wavelengths, we are specifically interested in the difference between the 880nm band (which is known as BC and measures absorption by black carbon), and the 370nm wavelength band (which is known as UVC and measures the ultraviolet absorption). This difference is known as Delta C and has been shown to be a good indicator that the PM2.5 sample was likely created by wood burning. Figure 2 below demonstrates this and shows how the aethalometer responds to two different sources of PM. The plot on the right shows a test using a diesel generator as a PM source, and here there is little difference in absorption between the wavelengths (the seven wavelength bands are difficult to see as their values are so close). But when the instrument is measuring PM created by wood burning in the plot on the left, the UV band (shown in red) is significantly higher than the BC band (in grey) at the times when smoke was measured by the instrument, and this difference (shown by the blue arrow) known as Delta C can be measured.   UFigure 2: Aethalometer response over time to two different PM sources.  Adapted from: Zhang, K. M., Yang, B., Chen, G., Gu, J., Schwab, J., Felton, D., and Allen, G.: Joint Measurements of PM2.5 and light-absorptive PM in woodsmoke-dominated ambient and plume environments, Atmos. Chem. Phys. Discuss., https://doi.org/10.5194/acp-2017-213,  2017.   90 2.3 Monitoring Campaign Overview For clarity, here is an overview of the steps involved with a monitoring campaign using this method. 1. Prepare for Monitoring: 1.1. Plan monitoring campaign schedule 1.2. Plan monitoring route o Enter monitoring route into Garmin Basecamp software and add to the GPS Navigator 1.3. Prepare a laptop and data management 1.4. Set up instruments and confirm settings are correct 1.5. Check calibrations of instruments o Aethalometer – Stability Test and Clean Air Test o Nephelometer – Zero Check (and Zero Adjust) 1.6. Set up a vehicle for the monitoring campaign  2. Individual Mobile Monitoring Run: 2.1. Heat vehicle and instruments 2.2. Prepare the nephelometer 2.3. Prepare the laptop connections and notes file 2.4. Prepare aethalometer and GPS device 2.5. Start GPS Navigator  2.6. Pre-Run Checklist 2.7. Drive the route! – (the actual monitoring) 2.8. End of run – save data and shut instruments down  3. Post-Monitoring Campaign: 3.1. Data download – Nephelometer backup and Air Quality Network 3.2. Use the Shiny application to map the monitoring data 3.3. Interpret the maps 3.4. Return instruments   91 3. Preparing for Monitoring Campaign  3.1 Planning Monitoring Campaign Schedule Monitoring of residential woodsmoke should be conducted during the winter months when wood heating appliances throughout a community will be in use. Mobile monitoring runs should be scheduled for evenings when wood heating appliances have been active for a few hours and concentrations will have built up, for example starting around 8-9pm. To limit the effects of other variables, monitoring runs should be scheduled for both weekday and weekend evenings over at least one week and the weather during this period should be recorded. To avoid always monitoring the same areas at the same time, monitoring routes should be driven in alternating directions each evening. Conducting mobile monitoring runs during the day can also collect useful data to compare to the night time results (but this is optional).  To avoid unnecessary complications, inform the local RCMP / Police department of your plans to repeatedly drive slowly along residential streets, include details of your schedule and a vehicle description. Residents may be concerned if they recognise a vehicle slowly driving their small streets every night.  3.2 Planning Monitoring Routes Planning you monitoring route is a major component of the preparation for a mobile monitoring campaign. Final results will show PM2.5 and woodsmoke concentrations relative to other areas on the driving route, and therefore you want to ensure that the route represents the community as much as possible, focusing on more populated areas. To identify important areas to include in routes it is best to make use of multiple information sources such as local knowledge and satellite imagery (eg. Google Earth).  Key considerations when planning monitoring routes: • Monitoring routes should cover as much of a community as possible within a reasonable time frame to limit variation in conditions (aim for no more than 1.5 hours). • Try to make routes as continuous as possible (e.g. aim for loops and minimise U-turns) to prevent stop and go driving. Ideally data wants to be collected evenly across the route. • If there is an air quality monitoring station in the area it is advisable to start and finish the route parked as close to the monitoring instruments as possible to compare data.     92 3.3 GPS Navigator Setup – Creating Monitoring Routes Garmin Basecamp is a free software program that can be used to pre-plan routes that can be saved to a Garmin GPS navigation device to provide the mobile monitoring driver with directions throughout the routes. This software is relatively easy to use and is available for free download at: https://www.garmin.com/en-CA/shop/downloads/basecamp . Only very basic maps showing major roads are available in the Garmin Basecamp software and so to be able to see more detail with which to plan routes, it is necessary to first connect a Garmin navigation device to the computer using the included USB cable. When the device is recognized you will see it appear in the upper left-hand panel under devices (in this screenshot the nuvi 2497 within the red box). To ensure the correct map is selected, click on Maps in the menu bar and select ‘City Navigator…’ .  To navigate in the Basecamp software, select the Pan icon (image of a hand) in the toolbar to be able to click and drag the map around, or hover the mouse over the blue north arrow in the top left of the map window to bring up the zoom and pan arrow buttons (shown in the screenshot below) which you can use to move the map.         93 Maps in the Garmin Basecamp software are saved in layers and show more detail as you zoom in. If you are only seeing major roads, zoom in and local roads will appear. The two screenshots below show the difference between a single layer when zooming in (right image is more zoomed in) on Whistler, BC.           To create a new route, change the Activity Profiles button to Driving, then select New Route in the toolbar (3 connected green squares), then close the pop-up window.        94 Your mouse cursor will now change to a small pen with a plus sign and you can begin to create your route by adding waypoints to the map by clicking along the roads. The waypoints you add will appear as small black circles on the map, and the Garmin software will automatically calculate the best way between the waypoints and plot this route along with small arrows indicating direction of travel. To move and zoom the map without leaving the route creator mode, you can use the pan buttons in the top left (hover over blue north arrow). Then to leave the route creator mode, just right click.  The new route will appear in the lower left toolbar, and double clicking on the route will bring up the properties window where you can rename the route and edit the colour and other properties.       95 To edit a route and add new points or edit existing points, select the route in this menu and use the three tools in the top toolbar (highlighted by the red square): Insert, Move Point, and Erase.   Insert will allow you to add new waypoints to the start and end of the route, or between existing waypoints by hovering over the area of interest until it is highlighted and clicking. This will then re-enter the same mode you were in when you first created the route (cursor will be a small pen with a plus sign), and you can add further waypoints.             1. Location of Insert Tool 2. Selecting the end of the route to add new waypoints 3. Adding waypoints between existing points to edit the route   96 Move Point will allow you to click and drag existing waypoints to a new location.                 Finally, the Erase tool will allow you to remove existing points. When selected your cursor will change to an eraser and points will be highlighted with a red cross when you hover over them – click to erase.     1. Location of Move Point Tool 2. Selecting point to move 3. Dragging waypoint to new location 1. Location of Erase Tool 2. Selecting point to delete   97 Each time you add a waypoint or edit the route, the software will choose the route it thinks is best to connect the waypoints. Your goal is to make the software follow the route of your choice with the fewest waypoints, this simplifies the instructions the device will give the driver.  When using the device to navigate a route, the device announces directions until the vehicle reaches the next waypoint. When a waypoint is reached there is a pause in verbal directions until the device recognises that the vehicle has moved past that waypoint. For this reason, it is important when entering the route into the Basecamp software to place waypoints just after intersections in the direction of intended travel (as in the example below) rather than directly at the intersection – otherwise the driver won’t be told which way to turn until they’re already at the intersection.              CORRECT: Black waypoints just after intersections INCORRECT: Black waypoints directly at intersections   98 A good way to check your route is working as expected is to use the playback feature. These controls are located in the top left (highlighted here by the red box), and by pressing the play arrow, a large red arrow will appear on the route and follow your directions. There is also a drop-down menu to change the playback speed, and a slider to move ahead in the route. The bottom of the window also presents summary data on the length and estimated drive time of the route (highlighted here by the dark blue box).       99  During a monitoring campaign it is important to alternate the direction in which routes are driven, so that each area is not consistently sampled at the same time of night. Therefore, when you have finished designing your route, another version of the route driving in the opposite direction must be created.  To do this, duplicate the finished route in the Basecamp software by right clicking on the route name and selecting Duplicate. Then rename the original route as ROUTE NAME – FORWARD, and the duplicated route as ROUTE NAME – REVERSE.      To calculate the reverse route, now right click on the new route and select Invert Route. This will instruct the software to calculate the route in the opposite direction through the waypoints. When this is done the route may change a little and you will need to move the waypoints slightly (and potentially insert more between existing waypoints) to make the software choose the intended route. Of note here is to move waypoints placed just after an intersection to the other side of the intersection (again to ensure the driver is given directions before arriving at the intersection). If there are one-way roads on the route etc. that make it impossible to drive the route exactly in the opposite direction, try to make the forwards and reverse routes as similar as possible.      100 The final step in the Basecamp software is to drag the completed routes from the lower left menu to the GPS Navigator device in the top left menu to save them to the device.  You can then right click on the device name and select ‘Eject’. The Garmin navigator device has a limit of 29 waypoints that it can contain in one route, and so when the device is now turned on it will split the route into parts after each 29 waypoints. At the end of each part, the driver will therefore have to quickly pull over and start the next part of the route navigation.          101 When you are ready to use the route directions you access the pre-programmed trips from basecamp on the device by clicking on Apps on the main screen followed by Trip Planner.         Finally select the trip of interest to see the trip’s information including distance and estimated travel time before hitting Go! to start the directions.          102 3.4 Prepare a Laptop and Data Management A laptop will need to be used by the co-pilot (or driver if monitoring alone) during monitoring runs. To prepare a laptop, copy the included folder ‘Woodsmoke Mobile Monitoring Materials’ to the Desktop of your laptop, and check this folder contains the sub-folder ‘Laptop Programs for Monitoring’.  Data files need to be saved and named in a consistent way to work with the Shiny application. To store the data from your campaign, create an overall folder on your laptop named: YYYY_MM_LOCATION_MONITORING using the year and month of your monitoring along with the community name as the location (e.g. 2017_01_WHISTLER_MONITORING). When the monitoring is completed, this folder needs to contain: 1. A completed copy of the TripList.csv file (a template is included in the ‘Woodsmoke Monitoring Materials’ folder).  a. This file should be filled in using Microsoft Excel to enter the details of each trip, including: Date, Route Direction, Start Time and End Time.   b. The ‘Date’ column must be formatted as MM/DD/YYYY. Excel will try to reformat your date column automatically, so to prevent this, enter an apostrophe before the date: i.e. 'MM/DD/YYYY c. Route Direction should equal ‘Forwards’ or ‘Reverse’.  d. The Start Time is the time you actually begin driving the route (after all set up and preparation was completed), and the End Time is when you returned to the same location at the end of the route.  e. Be careful to save this file as a comma separated values file (.csv) by using the Save As option and choosing from the dropdown menu under the file naming bar. The normal save button will save the file in the default Excel file type (.xls). 2. A completed copy of the Instrument_Calibrations.xls file (also included in the ‘Woodsmoke Monitoring Materials’ folder) to keep record of all calibrations and cleaning performed (include date and time along with calibration results for both instruments, such as Zero Check value from the nephelometer). 3. A file containing the nephelometer 1-minute averages as a backup. Download this data at the end of your monitoring as explained in Section 6.1. 4. Most importantly, individual sub-folders for each monitoring trip as described below.     103 Individual Trip Files For each trip please save files in the following way:  1. Within the main folder create sub-folders for each trip named: TripNo_LOCATION_YYYY_MM_DD (e.g. the first trip in Whistler on Jan 5th, 2017 was named Trip1_WHISTLER_2017_01_05)  2. Each trip folder should then contain the following files (how to save these files is explained during the monitoring run instructions): • The Notes file named: TripNo_Notes_YYYYMMDD.txt • The nephelometer file named: TripNo_NEPH_YYYYMMDD.txt • The two aethalometer files from the date of the trip with the original naming:  o AE33_AE33-S04-00415_YYYYMMDD.dat o AE33_log_AE33-S04-00415_YYYYMMDD.dat  Here is an example of the folder layout for the first two trips while monitoring in Whistler in January 2017: 2017_01_WHISTLER_MONITORING Trip1_WHISTLER_2017_01_05 Trip2_WHISTLER_2017_01_06 Trip1_Notes_20170105.txt Trip1_NEPH_20170105.txt AE33_AE33-S04-00415_20170105.dat AE33_log_AE33-S04-00415_20170105.dat Trip2_Notes_20170106.txt Trip2_NEPH_20170106.txt AE33_AE33-S04-00415_20170106.dat AE33_log_AE33-S04-00415_20170106.dat      104 3.5 Instrument Checks, Settings and Calibration Each instrument needs to be acquired, set up and calibrated in accordance with the manufacturer’s procedures. If anything is unclear here or you are looking for more information, please refer to the included manufacturer’s manuals. 3.5.1 Aethalometer - Magee Scientific AE33 Instrument Set Up Included components: • Magee Scientific AE33 Aethalometer • Power Cord • Air Inlet Tubing and Connectors • PM2.5 Selective Cyclone • Water Trap • GPS Receiver and Cords • USB Data Stick (containing instrument manuals and for data download) • Pelican Transportation Case  To set up the aethalometer sampling line, the PM2.5 Selective Cyclone and Water Trap must be connected in sequence to the AE33 air inlet using the included tubing. This sequence is shown in the above photo with the air flow direction shown with red arrows.  The first component of the sampling line is the cyclone which is designed to remove large particulate matter from the air flow. The included cyclone is a BGI SCC 1.829 and is specially designed so that it will remove anything larger than PM2.5 from an air stream travelling at a flow rate of 5 liters per minute.  The second component is the water and debris trap which will remove any water from the air stream before it enters the instrument. Be careful when connecting this component as it has a specific direction. There is a raised arrow on the side of the water trap which indicates air flow direction, and so the water trap should be connected with the arrow facing the instrument.  Finally, the tubing itself is made from an anti-static material to limit any smaller particulate matter being removed from the air flow by static forces. To prevent PM2.5 from impacting the sides of the tubing and leaving the airflow, keep the tubing as straight as possible with long smooth bends rather than sharp curves.  These components need cleaning periodically so please confirm whether they were cleaned prior to you receiving the equipment. A good way to check is to inspect the silver ‘grit pot’ on the front of the cyclone, which unscrews from the rest of the cyclone. In the unlikely case that it appears dirty, these components require cleaning. To do this, first separate the sampling line, Water Trap PM2.5 Cyclone Connects to AE33 Inlet Outside Vehicle Inside Vehicle   105 and then take each component apart (be careful not to lose the o-rings from the cyclone). Each piece (cyclone parts, water trap and tubing) should then be flushed with running water and wiped with a water dampened lint free cloth. The pieces then need be allowed to dry in a warm, dust-free area before reassembly. The components should not need cleaning again during your monitoring campaign.   Instrument Operation To operate the AE33, first remove the red inlet/outlet protection screwcaps from the rear of the instrument. If these are not removed the airflow will be blocked when the pump starts, and this pressure could damage the pump (NOTE: the instrument fan is noticeably loud, but if the flow is blocked it will be VERY loud). Connect the power cord to a power source and turn the instrument on using the two power switches (one on the rear next to the power cord and one inside the door of the instrument – the door is opened by pressing the silver button on the front until it pops out to form a handle, and then pulling this handle).  The front screen of the instrument will then illuminate and detail the various electrical checks the instrument performs before starting. When these are complete the screen will switch to the home page and begin to operate. Readings will read ‘NA’ for the first few minutes as the filter tape is advanced and the instrument prepares to operate. All further control of the instrument is performed using the touch screen. The headings at the top of the screen are used to switch between settings pages, and for this guide, pages will be identified as the top row option followed by the second row option, for example ‘OPERATION/GENERAL’. The home page is the default screen and shows the live values for Black Carbon and UV Particulate Matter, along with the flow rate, the current timebase setting (which controls whether the instrument is saving 1-second measurements or 1-minute averages), the number of tape advances remaining before a new filter tape is necessary, a status indicator, and finally the current date and time. When the instrument is running, the screen is set to screensaver mode where it will turn dark after a few minutes of inactivity. The usually green status indicator will continue to show on this screen. To wake the display up, simply press the screen. The status indicator is the most important feature of this home page. A green tick and value of 0 indicates the instrument is running as expected, if the instrument is stopped a red cross will appear, and if there are other issues a yellow exclamation mark will be present. If the value is not 0, you can press the status indicator and another screen will appear explaining what the issue is (as seen to the right – a code of ‘1’ shows the instrument is advancing the filter tape. This is the most common non-zero status). Please contact us for support if this occurs and there is a status update you do not understand.    106 The 3 other tabs at the top of the screen that you can switch between are: OPERATION, DATA and ABOUT. The OPERATION screen is used to control the instrument and change settings in the 4 sub-pages, while the DATA page has two sub-pages: TABLE which shows the live measurements, and EXPORT which is used to export data to a USB stick. The final ABOUT tab simply shows instrument and user registration details.  As mentioned, the instrument will begin to operate as soon as it is switched on (unless you choose ‘skip to main menu’ on the loading screen), and no settings can be changed during operation. To be able to make changes, first stop the instrument using the ‘Stop’ button on the top-right of the ‘OPERATION - GENERAL’ tab (under the red square in the image to the right). You will notice here that the other options are all greyed out until the instrument stops, and then they will become available. To turn off the instrument, the instrument must be stopped as explained above, then press ‘Shut Down’ at the bottom right of the ‘OPERATION-GENERAL’ tab. When the instrument shuts down, use the power switch on the back to complete the process (avoid using the interior power switch, as one is sufficient). Then disconnect the sampling line from the instrument and replace the red protection caps. Whenever you are moving or transporting the instrument, always keep it level with the feet facing down.   Check Instrument Settings When the instrument is stopped, the settings on the ‘OPERATION – GENERAL’ tab should be confirmed to match those in the image above. If there are any differences here, they can be changed by pressing on the option and selecting from the drop-down menus or typing. The TimeBase setting of 1-second is important to be able to provide high spatial resolution during the mobile monitoring, and the Flow setting of 5 liters per minute is important as this is the air flow rate at which the inlet cyclone will correctly separate PM2.5 from larger particulate in the sample air. Finally, the GPS radio button at the bottom should be selected to instruct the instrument to synchronise it’s time with the GPS unit (if this GPS option is not present, please refer to the AE33-GPS Module Implementation Note included with the instrument manuals for the steps to re-connect the GPS device).     107 The settings on the ’OPERATION – ADVANCED’ tab (seen below) should be correct when you get hold of the instrument and should not be changed. Most of these values indicate the status of various internal sensors. The ’OPERATION - LOG’ tab saves a log of every change to the instrument and may be referenced if you are unsure what the instrument is doing. Do not use the ‘OPERATION-MANUAL’ tab – this is for Magee Scientific technicians only.     Instrument Calibration Checks The AE33 is a very stable instrument and most calibration checks do not need to be performed regularly. Chapter 9 of the AE33 manual covers maintenance of the device with a suggested schedule for maintenance checks and clear step by step guides to each procedure (Note: This suggested schedule is based on the assumption the instrument is running 24/7 and not just for short periods as with our monitoring, so checks do not need to be as frequent). The procedure that you are most likely to encounter is if the filter tape runs out during your operation of the instrument. The filter tape in the instrument is held on two spools either side of the optical measurement area and is moved incrementally from left to right during operation. As particulate matter is deposited onto the filter tape, the tape gets darker, and when it gets too dark for more measurements to be taken, the instrument performs a tape advance to move a new clean section of tape under the optical components. Eventually the tape will run out and need to be replaced. The guide for this procedure on page 65 of the AE33 manual is very thorough.    108 You cannot perform flow checks, calibrations or leakage tests without a professional flow-meter so these procedures should be avoided. However, both a ‘Stability Test’ and a ‘Clean Air Test’ can be performed without any additional equipment simply by selecting the options on the ‘OPERATION – GENERAL’ screen when the instrument is stopped.  The Stability Test checks the performance of the light source and detector without air flow in the system, while the Clean Air Test is a similar test with the airflow activated but filtered through an internal filter to remove any particles from the air. These two tests should be run prior to the start of monitoring. Set up the instrument indoors when doing this, and each test takes around 20 minutes to run before reporting results to the main screen when they finish. Please record the date and time of each test in the Instrument Calibrations spreadsheet along with the result the instrument gives you (such as ‘Stability test results are acceptable’).  3.5.2 Nephelometer – Ecotech Aurora 1000  Instrument Set Up Included components: • Ecotech Aurora 1000 Nephelometer • Power Cord • Serial Communication Cord and USB Adaptor • Air Inlet Tubing and Funnel • Pelican Transportation Case  There are fewer external components needed to use the nephelometer and so the set up of the instrument is much simpler. The sampling line only consists of tubing to direct air flow from the exterior of the vehicle into the inlet of the instrument (this is the largest of the three inlets on the top of the instrument, and is labeled ‘sample’ – see above). As this tubing does not have a cyclone or similar component limiting the material entering the tubing, larger dirt and dust particles can impact onto the tubing, and it should be cleaned regularly. Due to the nature of long tubing it is difficult to clean. The most effective method is to clean the tubing either in a shower or outside using a hose. Hold the tube vertically and pour a small amount of dish soap into the higher end, then rotate the tubing to allow the soap to coat the interior walls as it runs down. Now run high-pressure water through the tubing to remove the soap, dirt and dust from the tube. Rotate the tube while doing this to ensure all residual soap is removed from the tube. Hang the tube to dry in a warm area. Compressed air can also be used to remove residual water from the tube if necessary. The tubing must be fully dry before connecting to the instrument.     109 Instrument Operation To operate the instrument, connect a power source and turn on the instrument with the switch on the right side. The nephelometer will first go through some quick checks which are shown on the screen, followed by a short warm up (during which the screen will read ‘Inst Warmup Time’). You will then hear the instrument pump start and the nephelometer will start to take measurements. Normal operation will then begin, and the screen will show the current scattering measurement at the top, along with the sample temperature, relative humidity and pressure readings from the environmental sensors (as seen in the photo below). The row with the headings ST°C, RH% and BP will alternate between headings and the current date and time.   The instrument is easy to control. The screen will illuminate when any of the buttons are pressed, and the up and down buttons control the screen contrast while on the main screen (this doesn’t need to be changed unless you accidentally turn the contrast all the way down so the values are no longer visible…). To access the menu, press the enter or select buttons, then use the up and down arrows to navigate, and the enter or select button to enter a menu branch. To go back a level press page up, or to finish and return to the main screen press exit. To change a setting, highlight the row of interest, press select and then use the up and down arrows to switch to the option of choice. Then press the enter button to confirm the highlighted choice, or if you need to cancel the change press exit. When setting the time, use the select and page up buttons to move right and left between digits, then use the up and down arrows to change the currently highlighted digit. When you are finished with the instrument simply switch it off using the power switch on the right of the case. Then disconnect the sampling lines and cover the inlets with caps or tape to prevent dirt or dust falling into the instrument. Whenever you are moving or transporting the instrument, always keep it level with the feet facing down.     110 Check Instrument Settings Prior to monitoring, verify the following settings are selected in the instrument menu: Menu Option Correct Setting Reason Under Calibration -> Cal Settings -> Auto Cal Intv Off To prevent calibrations starting during monitoring. Under Control -> Sample Heater RH High humidity can affect nephelometer measurements, and so these commands set the internal air heater to heat the sample air until relative humidity of the sample is less than 60%. Desired RH <60% Under Report Prefs -> Filtering Kalman Basic unit settings to standardise the data. Date Format Y-M-D Temp Unit °C Press. Unit Mb Normalise to 25°C Under Serial I/O -> MltDr Baud Rt 38400 These are the communication settings for the instrument’s serial ports and must match with a connected laptop for successful data transfer. Setting the service port to ‘Reading’ mode and an output of ‘1sec’ allows us to read and save the live 1 second data directly from the instrument onto a laptop through a serial cord connected to that port.  Mlt Parity None SvcPt BaudRt 9600 SvcPt Parity None SvcPort Mode Reading Reading Outp 1 sec Under Datalogging -> Log Period 1 min 1 minute is the shortest time averaging period available. This data serves as a backup if there are any issues with the 1 second data captured on the laptop.     111 Calibration Checks and Adjustments The nephelometer works by comparing optical measurements to two calibrated values. The first calibration is a ‘Span’ calibration and involves testing the instrument with a standard certified gas - usually high purity carbon dioxide. The second calibration is a ‘Zero’ calibration and refers to a measurement of normal air that has been internally filtered to contain zero particles.  The zero calibration is important as the gases that make up air naturally scatter light, and so having this value allows the instrument to remove the impact of these gases from the measurement of the sample. There are two options for each type of calibration: a check, and an adjust. The span and zero checks are used to check the instrument is still measuring expected values when exposed to these controlled samples, and an adjust is to reset the calibration of the instrument if the checks show that this is necessary. Checking and adjusting the span point requires testing the instrument’s response to high purity carbon dioxide and so will be performed between monitoring periods. Please confirm with us that this has been checked recently (In the very unlikely case we ask you to perform a span point calibration yourself, refer to the set up on page 20 of the Aurora manual, and the procedure in the calibration chapter on page 39). Zero checks are easy to perform as they use a filter built into the instrument to create clean air for the calibration. This check should be performed after every few monitoring runs to ensure the instrument is still working as expected. To perform a zero check, first run the instrument for around 10 minutes to allow the instrument to stabilise, then remove the red cap from the inlet labeled ‘zero’ on the top of the instrument and use the following menu path: CALIBRATION -> ACTIVATE CAL -> DO ZERO CHK. If you accidentally start a different calibration procedure, turn off the instrument using the power switch and this will prevent settings being changed.  The test will then begin (you will hear the pump start and the screen will change) and continue until both the minimum calibration time has been reached (default is 15 minutes) and the instrument has achieved 95% stability. When the zero check is finished the instrument will return to normal operation. To check the result of the test, you will have to navigate to CALIBRATION -> PARAMETER and check the updated value alongside LAST ZERO. Please record the date/time and new results of any calibration procedures in the Instrument Calibrations Log.      112 After any check is performed the results should be compared to Table 5 from the Aurora manual included below. If the LAST ZERO value is above 2, or below -2, then a Zero Adjust should be performed. To do this is similar to the zero check but instead navigate to ‘CALIBRATION -> ACTIVATE CAL -> DO ZERO ADJ.’. Again, please record the results from the CALIBRATION -> PARAMETER -> LAST ZERO page. If the value from the zero check is outside 0 ± 4, perform a second zero check before proceeding to confirm this value before contacting us for support. In the rare case you are testing the Span point, the correct span value for CO2 that should be used for calculating the calibration tolerance is 22.7Mm-1.    Repeat zero check then contact us for support.    113 3.6 Setting Up a Vehicle for Monitoring Campaign This section covers how to install and connect the instruments in a vehicle ready for mobile monitoring. To prepare a vehicle for mobile monitoring, the two instruments will be placed on the back seat of the vehicle with their air intakes passed through a rear window and attached to the outside of the vehicle (on the opposite side to the vehicle exhaust). Please perform this using the following steps:  1. Place the AE33 in the rear of the vehicle. • Place the instrument on the same side of the seat as the vehicle exhaust with the screen facing outwards as shown.  • If the rear seat can fold down flat this works well to place the aethalometer on, otherwise place the AE33 directly on the seat. • Secure the instrument using the seat belt or bungee cords.      2. Connect the cords to the rear of the AE33. • Connect the black power cord at the bottom right. • The power cord should be fed to an available 12V socket (use an outlet in the trunk area if possible to keep the front outlet available to power the GPS navigator). • Connect the serial and USB cords from the GPS device to the USB and COM1 ports in the bottom left. • The GPS receiver (lower right photo) attached to the cords should be passed through to the front of the vehicle and placed on the vehicle dash. • Do not connect the black inlet tubing (top) to the AE33 yet until starting monitoring.      114   3. Place the Aurora 1000 in the rear of the vehicle. • The Aurora 1000 should be placed on the opposite side of the seat to the AE33. • This instrument generally fits well with the seat in the normal upright position as the seat prevents the instrument falling over. • Again secure using the vehicle seat belts or bungee cords.     4. Connect the Aurora 1000 cords. • Carefully attach the power cord to the power socket on the right side of the Aurora 1000 (this is a 4-pin plug that you will need to line up and then use the outside screw section to secure).  • Feed the power cord to the 12V power socket. • Attach the communication cord with attached USB converter to the top-right port labeled RS232 Service.  • Feed the communication cord to the front of the vehicle where it will be connected to a laptop during operation.  5. Place the power inverter by a 12V socket. • Place the power inverter in the rear of the vehicle if possible by a 12V power socket. • Plug in the power cords from both instruments.  • The power inverter should NOT be connected to the 12V socket unless the vehicle is running, as it may drain the vehicle battery.      115 6. Prepare the window for attaching the inlets. • Start by half-lowering the rear window on the opposite side to the vehicle exhaust.  • Line the edge of the window with the foam pipe covering.  • Close the vehicle door.     7. Attach the AE33 cyclone to the side of the vehicle. • Due to the shape of the cyclone, it should be positioned on foam at the front of the window facing into the direction of airflow when the vehicle is moving, and slightly upwards (to prevent moisture and other debris entering the air inlet). The smaller silver cylinder should be facing down (held by the hand in this photo). • First attach a foam pad to the vehicle (to protect the vehicle from the cyclone) at the front of the rear window using duct tape (seen under the cylone in the photo). • Pass the water trap and end of the tubing through the window opening and into the vehicle (Do not connect the air inlet tubing to the AE33 yet).   • Now tape the cyclone to the vehicle (this is easier with a second person to hold the cyclone while you attach it). To safely secure the cyclone, use longer strips of tape first in a cross pattern across the main body.  • Then seal over all the foam in the area to limit water, but make sure to not block the area around the inlet.     116  8. Attach the Aurora 1000 Inlet.  • As there is no cyclone on this inlet and the tubing enters the instrument directly, this inlet should be orientated in the opposite way to the AE33 inlet, by placing it at the rear of the vehicle facing slightly down (approximately 30 degrees below horizontal). This orientation along with the cone attached to the end of the inlet minimises the risk of water and other debris entering the instrument. • Pass the air inlet tubing for the Aurora through the window into the vehicle.  • Attach some foam to the vehicle to protect the paint, and then attach the cone on top of the foam. • Make sure to also secure the tubing with tape to prevent it moving around while driving.   9. Seal the remaining window opening.  • When both inlets are secured carefully raise the window to minimise the opening but do not squeeze the inlet tubes. • Seal the remaining window opening and around the two tubes to keep water out of, and heat inside the vehicle. Long strips of tape horizontally across the opening work best, starting at the bottom and working up before sealing around the tubing.  • Do not open this door or adjust this window for the remainder of the monitoring to prevent tearing the inlets from the vehicle or damaging the tubing (a good way to prevent accidentally doing this is to place tape over both the door handle and the window control button).    117  10. Attach the air inlets to the instruments. • The Aurora inlet tubing simply slides onto the metal sample inlet opening pipe on the top of the instrument. • The AE33 inlet tubing screws into the opening at the top on the rear of the instrument, and the water trap can sit on the top of the Aurora (as seen in this photo). To screw the inlet in without twisting the inlet tubing, screw the white connector into the instrument, while holding the grey connector (between the tubing and the white threaded connector) still.      Between monitoring sessions disconnect the inlet tubing from both instruments and cover the inlets with tape to minimise any risk of moisture or dust entering the instruments. If temperatures are going to be negative overnight, take the instruments out of the car and bring them inside, otherwise they take a very long time to warm back up when you are ready to monitor. Also if the vehicle is parked in an unsecure location it is advisable to bring the instruments inside overnight. When doing this just disconnect the power and communication cords along with the inlets and leave these cords in place in the vehicle ready for monitoring. Whenever transporting the instruments keep them upright with the feet facing down.           118 4. Mobile Monitoring Run Preparation The following is a guide to the individual steps to perform a mobile monitoring run. Before starting, check there are no reasons to stop the monitoring run until the end (check vehicle gas, laptop is charged, washroom breaks etc). 4.1 Heat the Vehicle and Instruments 1. Turn on the vehicle and heaters to get the interior to a warm temperature (~20C, this can be done while driving to the start of the route).  IMPORTANT: If the instruments are turned on right away when cold, humid air may condense on the cold surfaces within the instrument and cause damage.  For this reason, on cold evenings, the instruments should be removed from the vehicle and brought inside to prevent them getting too cold. This should also be done if the vehicle is not parked in a secure location such as a garage overnight. When doing this, you can disconnect all of the cords from the instruments and leave those in place in the vehicle.  2. If the instruments were removed from the vehicle, place them back in the car as described in Section 3.6 and connect the cords as below:   • Connect the power cord to the rear of the AE33 and connect the GPS Serial and USB cords to the COM1 and USB ports also on the rear. DO NOT CONNECT THE INLET LINE YET.  • Connect the power cord to the Aurora 1000, and the serial cord to the ‘RS232 Service’ port.  DO NOT CONNECT THE INLET LINE YET.  3. Plug the power inverter into the 12V vehicle socket and plug the two instrument power cords into it.  4. When the instruments are no longer cold to touch, check the inlet lines are still disconnected from the instruments and turn the instruments on to allow them to sample warm interior air from the vehicle. • Switch the power switch at the back of the AE33 – the screen will turn on and perform checks, after about 5 minutes it will start recording data. • Switch the power switch on the right side of the Aurora 1000. Perform the rest of the set-up work while the vehicle and instruments warm up.   119 4.2 Prepare the Nephelometer 5. The first step while the nephelometer is heating is to reset the clock to match the aethalometer (AE33). To match the nephelometer (Aurora 1000) data to the GPS location data recorded by the aethalometer, the clocks on the two instruments must match. The aethalometer clock is automatically reset to match the GPS time and so the nephelometer clock needs to be manually set to match this time.  This process is easier with two people. One person will watch the clock on the home screen of the AE33 (through the rear side door), while the other sets the Aurora clock to match. To adjust the nephelometer clock press ‘enter’ to access the menu, and then scroll down to ‘ADJUST CLOCK’. Press ‘enter’ to enter this option. Change the date and time to one minute ahead of the AE33 (e.g. if AE33 reads 19:45:27, set Aurora to 19:46:00) by pressing ‘enter’ on an option, then using the ‘Select’ and 'Page Up’ buttons to move left and right between digits respectively, and the up and down arrows to adjust the currently highlighted digit. When you have the date and time correct, press ’enter’ and move down to highlight the ‘Save Time’ option at the bottom of the screen. When this is ready, the person watching the AE33 screen can count down to the prepared time, and just before the times match, the person operating the Aurora presses ‘enter’ on the ‘Save Time’ option. The screen will now say ‘Setting Clock…’ for a few seconds while the change is saved before returning to the home screen. When this is complete, confirm the two clocks match within a second of one another by each reading out the seconds together (the Aurora only shows the clock on the screen for a few seconds at a time). If you are not happy with the match, repeat this process as many times as necessary.  All other settings on the nephelometer should still match the settings described in Section 3.5.2 and the measurements will begin automatically.   6. Connect the ‘RS232 service’ port (top right) on the side of the nephelometer to a laptop USB port using the serial cord and USB adaptor.     120 4.3 Prepare the Laptop A laptop is used for two things during the trip: saving 1-second data from the nephelometer to a text file using a program called Windows HyperTerminal, and recording and saving important notes about the trip. Two programs have been created to automate both the process of setting up a HyperTerminal connection on the laptop and preparing a notes file in a standard format.  7. When the USB/Serial cord from the nephelometer is connected, open the ‘Woodsmoke Mobile Monitoring Materials’ folder, and the ‘Laptop Programs for Monitoring’ sub-folder.  8. If using a Windows 10 laptop, run the program called ‘AutomateHyperTerminal-Win10.exe’ by double-clicking on it, then do not touch the mouse while the laptop automatically goes through the HyperTerminal set up screens and starts the connection for you. If the program is successful you will see live nephelometer data begin to appear in a large window with a smaller window prompting you to ‘Capture Text’ as seen here. If this does not work, try the second program called ‘AutomateHyperTerminal-Win7.exe’, and if neither work you will have to manually set up the connection yourself as described in the of this guide (this is not a complicated process).  9. If the program was successful, click on ‘Browse’ and navigate to the trip folder where you will save the data (this was set up in Section 3.4). Name the file using the format: TripNo_NEPH_YYYYMMDD.txt (e.g. Trip1_NEPH_20170105.txt) and click save. This will take you back to the ‘Capture Text’ window where you will click ‘Start’.   121 10. To confirm HyperTerminal is now correctly saving the nephelometer data, check the word Capture in the bottom border is bolded (shown here in the red box – compared to the light grey in the previous screenshot).   The live data contains the following columns: Date, Time, Bscat, Sample Temperature (temperature of the air sample), Enclosure Temperature (temperature of the instrument, Relative Humidity and Pressure (then placeholders).  Bscat is the back-scattering value measured by the instrument and is well correlated with the PM2.5 concentration – therefore this first column following the time contains the values that you are interested in.  11. To start a new notes file for the trip, run the program called ‘OpenNewNotesFile.exe’ in the ‘Laptop Programs for Monitoring’ folder. This program will ask you for the location name in all capitals (e.g. WHISTLER), the trip number (e.g. 1), and the direction you will be driving the route this time (FORWARDS or REVERSE). Then do not touch the mouse while a new notes file opens and types in these values before prompting you to save the file. Browse to the trip folder and name the file using the format: TripNo_Notes_YYYYMMDD.txt (e.g. Trip1_Notes_20170105.txt), and choose ‘Save’.   12. Enter the driver and co-pilot names and a short description of the weather (e.g. include the current temperature from the vehicle read out, whether skies are currently clear or cloudy, current precipitation type such as light rain or heavy snow, and windy or calm). Throughout the monitoring run, the co-pilot can record notes on things that impact the trip such as trip delays (due to traffic, having to wait at a train crossing etc.) as well as things that impact PM2.5 levels. Things to record here include: driving through a visible smoke plume, driving behind a truck kicking up a lot of road dust etc. Make sure to include the current time from the HyperTerminal window (to the closest 10 seconds) with each note. The keyboard shortcut ‘ctrl-S’ can be used throughout the run to save the file.  IMPORTANT: ALWAYS RECORD THE TRIP START AND END TIMES IN THIS NOTES FILE.    122 Summary This screenshot shows how the laptop screen should look during a monitoring run. The HyperTerminal window is showing live 1-second data from the nephelometer, and the Capture icon in the bar at the bottom of the window is black (rather than grey) which confirms this data is being saved. The Notes file is saved (you can tell by the name in the top bar), and contains the date, trip number, direction (Forward here), weather summary and most importantly the start time. There is also an example of something you should take note of here with “visible smoke drifting across road” at 16:45:20, as notes like these can help with data interpretation.       123 4.4 Prepare the Aethalometer and GPS  13. Place the GPS receiver that is connected to the AE33 (as seen on the right) on the vehicle dashboard where it has a clear view of the sky and will receive the best signal. A very small flashing red LED (shown here by the red arrow) will indicate that it is successfully recording the current location.   14. The instrument will read NA’s on the home screen when it is initially turned on as the filter tape is advanced. After about 5 minutes when normal values do appear, check the instrument is running correctly by: • On the home screen check there are values appearing next to BC and UVPM, and there is a green tick next to Status and a code of ‘0’. • Press ‘DATA’ on the top row of the screen, and then ‘TABLE’ on the second row. This opens the Data Table (seen here). Confirm all channel rows are now reading values greater than 0 (during the warm up stage every box on this page will read 0) to check the instrument is operating correctly.  • Also on this page check there is a row present at the bottom labeled ‘External device 1’ which contains GPS data (as shown to the right). This confirms the GPS is recording a location (GPS data in this field shows UTC time first, the latitude position followed by N, the longitude position followed by W and then other values that show the accuracy of the signal).   AE33 ready for monitoring – DATA/TABLE page showing readings in all cells and valid GPS values in the ‘External Device 1’ row.   124 4.5 Start the GPS Navigator The final device to prepare is the GPS navigator with pre-loaded route directions.  15. If not already in place, attach the GPS holder to the windshield by first pressing it firmly against the glass so the rubber is flush and then using the switch to lock it in place.  16. Turn on the Garmin Navigator device and plug it into the 12V outlet. 17. Start the route directions by choosing Apps and then scrolling down with the arrows to choose Trip Planner. Select the pre-loaded route in the correct direction (FORWARDS or REVERSE) and choose Go! to start the directions.    4.6 Connect the Sampling Lines 18. Now the instruments and laptop are ready for monitoring, check the instruments are no longer cold (you can check if they are cold to the touch and also check the temperature readings on the nephelometer in column 5 of the HyperTerminal read out), then you can connect the sample lines to the instrument inlets: • Slide the plastic tubing onto the metal inlet on the top of the Aurora nephelometer labeled ‘Sample’. • Screw the white threaded connector on the AE33 aethalometer sample line into the inlet at the top of the rear of the AE33. To remove kinks from the tubing and have the tubing sit as straight as possible, hold the dark grey tubing connector still while screwing in the white connector to the AE33 (this dark grey piece will rotate within the white connector).  19. Go through the pre-run checklist (Section 5.1), record the start time in the notes file (read the time from the HyperTerminal window) and then begin the run! Note: This start time is important as it is used to delete the data collected while the instruments warm up inside the vehicle.       125 5. Mobile Monitoring Run During the monitoring run the field technician will follow the monitoring route designed for the community as entered in the GPS navigator while occasionally checking all instruments are working as expected. Brief notes will also be taken on special circumstances observed such as heavy traffic or visible smoke that may affect instrument readings  5.1 Pre-Run Checklist Before beginning a monitoring run please use this checklist to ensure everything is set up correctly for a smooth monitoring run:  1. There are no reasons to stop the monitoring run until the end (check gas level, that the laptop is charged, no need for a washroom break etc). 2. The two instruments were heated prior to set up (with sampling lines disconnected). 3. Nephelometer: o Clock has been reset to match the aethalometer o Values are appearing as expected on the home screen o Sampling line has been reattached to inlet 4. Laptop: o HyperTerminal program is running and 1 second nephelometer data is appearing live on the screen o Capture text has been set up, and the capture text icon is bold in the lower bar o Notes file has been opened and saved ready for notes to be taken throughout the run - Trip number, driving direction and weather summary entered 5. Aethalometer and GPS: o Green tick and status code of ‘0’ are present on the home screen o GPS receiver is placed on the vehicle dashboard and small red LED is flashing o Both aethalometer and GPS data are as expected on the ‘DATA/TABLE’ page o Sampling line has been reattached to the inlet 6. GPS Navigator: o Monitoring route has been selected in the right direction and ready to start 7. Now you can enter the trip start time on the notes file and begin the monitoring route!     126 5.2 Driving Notes While driving the monitoring route we want to aim for even coverage around the route as much as possible. To achieve this drive at slower than normal speeds (~25-30 km/h in residential areas) but without becoming an obstruction to traffic (i.e. on highways you will have to drive at normal speeds). Measurements every second at 30 km/h equals approximately one measurement every 8.4 m. If there are long pauses in the run for whatever reason (traffic, train crossing, road construction, forgotten bathroom or gas break etc.) be sure to record these in the notes file. Ideally just keep the instruments running, but if you are alone and need to turn off the vehicle you will have to restart the instruments. If the instruments have to be restarted, you will need to start a new HyperTerminal connection between the nephelometer and laptop. Save the new file with the same name format but ending in NEPH-YYYYMMDD-PART2.txt.    127 5.3 End of Run – Data Saving and Instrument Shutdown At the end of the trip – DO NOT turn off the engine before shutting down the instruments: 1. Record the stop time and a current weather summary in the notes file – check this file is saved before closing. 2. Save the HyperTerminal file and end the connection to the nephelometer (see next page for guide). 3. Turn the nephelometer off with the power switch on the side. 4. Download data from the aethalometer by inserting the included USB stick into the front panel, pressing DATA at the top and then EXPORT on the second row down. Touch the date next to from and it should change to the current date, then choose ‘ExportToUSB’ for today’s files to be copied to the USB. 5. Shut the aethalometer down using the ‘Stop’ and then ‘Shut Down’ commands on the OPERATION/GENERAL page, before using the power switch on the back.  6. Disconnect the sampling lines from both instrument air inlets. 7. Remove the power transformer from the 12V inlet to prevent it draining the vehicle battery. 8. You’re finished!  IMPORTANT NOTE 1: Make sure to charge the laptop between runs. As there are only two outlets in the power transformer, the laptop must run from battery power during monitoring.  IMPORTANT NOTE 2: On nights where negative temperatures are expected or if you are parking in an unsecure location, remove all connections from the turned off instruments and cover the inlets. Then bring them inside for the evening (keep the instruments upright and level when moving and storing). This is both more secure and saves you time when waiting for the instruments to heat up the following evening.     128  Closing HyperTerminal Connection at End of Run 1. Stop Capturing Text. To stop the text capture, in the top bar select ‘Transfer’ -> ‘Capture Text…’ -> ‘Stop’ as seen to the right.          2. Check the file.  Navigate to the file where the captured text was saved and open it to confirm it was saved as expected.        3. Close the program to end the connection. When you have ended the text capture and confirmed the file is correct, you can end the connection to the nephelometer simply by closing the program. The program will ask if you want to save the connection – choose NO as this is not possible with this copied version of HyperTerminal.     129 6. Post-Monitoring Campaign 6.1 Download Back-up Data from the Nephelometer At the end of the monitoring campaign, download the 1-minute average data from the nephelometer onto the laptop as a backup for the 1-second data collected through HyperTerminal during the monitoring. To do this, turn the nephelometer on, and connect the serial cord to the lower ‘RS232 multi-drop’ port (rather than the ‘RS232 service’ port used during monitoring – shown here on the right). Open the Ecotech Nephelometer Data Downloader program on the laptop (shown below with numbered steps indicated on the screenshot) and check the following settings: 1. Click on the small box with 3 dots next to the output file to browse for the main monitoring folder and then set the file name as: YYYY_MM_Location_NEPH_1Min.csv  2. Below the file name, choose Append to file, and check both the COM port matches the port the USB is attached to, and the Baud rate matches the nephelometer settings (should be 38400).  3. To the right under Output file preferences select:  • Date format = YYYY-MM-DD hh:mm:ss  • Temp unit = Celsius • Field separator = Comma delimited 4. Under Start Date/Time, enter the first day of the monitoring campaign. The End Date/Time should already be set to the end of the current day. 5. Finally click, Collect Data in the top right corner for the laptop to download this data from the nephelometer.   130 6. As data is downloaded, it will appear in the lower grid, with every 10 1-minute averages switching between yellow and white highlighting. When the download is complete, you can open the file to check it was created correctly and then close the program.            131 6.2 Using Online Shiny Application to Map Your Data  An online application has been created using the Shiny platform to automatically create maps using your data. These maps will show the average spatial patterns across your monitoring route measured with each instrument. To use this application, the data has to be saved in a specific way:  1. Within your overall YYYY_MM_LOCATION_MONITORING folder, create a new folder called SHINY_LOCATION.  2. Copy and paste the completed TripList.csv file into this folder.   3. Copy and paste all of your individual trip folders into this folder (do not edit or make any changes to the original data files – these should always be kept as is for backup).  4. For each trip, copy and paste the nephelometer file into the SHINY_LOCATION folder.  If there were any trips where problems led to multiple nephelometer files being created (for example if you had to stop mid-route), these files have to be combined into a single file for that trip before transferring to the SHINY_LOCATION folder. To do this: • First make sure all the original files are named ending in -PART1, -PART2 etc. • Open the PART1 file in Notepad (this should be the default program if you just double click on a .txt file to open it). Click ‘File’ in the top bar then ‘Save As’ to save a copy of this file with the standard naming (TripNo_NEPH_YYYYMMDD.txt) with no PART1 etc. into the sub-folder for that trip. • Now open all further files for that trip in Notepad and manually copy and paste the data from these secondary files into the new file that you just saved. Do this by scrolling to the end of the first file and deleting the last row if it is incomplete, then switch to the PART2 file and delete the first incomplete row there as well. In the PART2 file press ctrl-A to select all of the data (all rows will highlight blue), and ctrl-C to copy it.  Now return to the first file, place the cursor on a new line at the end of the data and press ctrl-V to paste the copied data at the end of the first file. Repeat this process if there are more than two files, copying the data from each file to the end of the newly created overall file.      132 • The result is shown in the example below (which shows PART1, PART2 and the combined data in the windows from left to right), the final file on the right now looks like a normal file apart from the time break between two rows where no data was recorded (the last row of the original file and the first row of the second file are highlighted here).    5. Also for each trip copy and paste the aethalometer file (just the larger data file, within the log file) into the SHINY_LOCATION folder and rename them in a similar format to the other files as: TripNo_AE33_YYYYMMDD.dat.  If you happened to have any trips that ended after midnight, the aethalometer and GPS data from those trips will be saved in two files as the AE33 instrument saves a new file for each date. Therefore you will have to combine these files in the same way as described above for the nephelometer files, by opening the first file in Notepad, saving a copy named with the format TripNo_AE33_YYYYMMDD.dat, and then copy and pasting the data from the second date file into this new file.   6. When this is complete your SHINY_LOCATION folder and file names should look like this:    133 7. To use the Shiny Application, go to: https://kathleenmclean.shinyapps.io/woodsmoke_mobile_monitoring/  and follow the instructions on the right side of the screen.   8. The instructions will guide you through the options and uploading your files on the left side of the screen. As files are uploaded, the lower right side of the screen will inform you whether the files were uploaded in the correct format. If any errors are reported here, first check the naming of your files. If there is an error with the TripList.csv file, also check the format of the Dates within this Excel file are MM/DD/YYYY (otherwise the app won’t be able to match this file to the instrument data files).  9. If you choose to add the location of a fixed site monitor to the map, the coordinates of your local monitor can be found using Google Maps. If you switch to satellite view in the lower left corner, you can left click on the location of the monitor (the location of the Whistler monitoring station is shown here), and the coordinates will be shown in the pop-up box at the bottom of the page (here in the red box you can see the Latitude = 50.144260, and the Longitude = -122.960356).                 134 10. When you have uploaded your data, switch to the Maps tab at the top of the screen and the maps will load. You can switch between the data from the two instruments (PM2.5 estimates from the nephelometer, and woodsmoke specific Delta C from the aethalometer), and adjust the zoom level of the map (11-13 are zoom values determined by Google maps). You can also export the maps as PDFs using the Download Map option.                6.3 Interpreting the Maps The maps created by the Shiny application show the average patterns across the monitoring route during your monitoring runs. As the data you have collected is just a snapshot in time, these maps are only semi-quantitative and are intended to identify general patterns and possible hotspot areas.  To create the maps, the app first standardises the data from each trip by converting the measurements to Z scores (also known as ‘standard scores’), and then calculates the spatial pattern across the monitoring route during that trip by taking the average of all measurements that fall within each cell of a grid with 100m2 cells. Finally, the app calculates the overall average pattern during your monitoring by taking the average of the Z score patterns calculated for each trip, producing a map of average Z scores across the region.     135 Z scores are a statistical method used to compare values across a range. On this scale a Z score of 0 is equal to the average across the entire monitoring route, and the other values indicate how many standard deviations (a standard measure of variation in the data) each box was away from the total average across the route. For example, an area of the map with a Z score equal to +1 means that box was on average one standard deviation greater than the average during each trip (and so likely has higher PM2.5 values on average), while a box with a Z score equal to  -1.5 means that area of the map was on average one and a half standard deviations lower than the mean during each trip (i.e. that area had ‘cleaner’ than average air during the monitoring).  These Z scores are calculated on an exponential scale and so the conversions to estimates of standard units alongside the legend are included to help you interpret how much one shade is greater than the shade before. These values are included to add context to the shading and make it easier to understand, but remember that these are estimates and should not be used quantitatively, rather they are included for reference to be able to compare two areas of the map.  6.4 Returning Instruments When you have completed your monitoring campaign, please repackage the instruments in the packaging you received them in, ensuring that they are well protected and that the box is labelled for the correct orientation, so that the instruments are kept upright during transportation. All supporting materials including the instrument sampling lines, the aethalometer cyclone, GPS receiver, Garmin navigation device, power cords etc. should be carefully packaged together in the third box. All packages should then be returned to the following address: Dr. Michael Brauer - Room 370B School of Population and Public Health The University of British Columbia 2206 East Mall Vancouver, BC  V6T 1Z3     136 7. Appendix: Manual Set-up of Nephelometer/Laptop Connection • Connect the ‘RS232 service’ port on the side of the nephelometer to the laptop using the serial cord and USB adaptor. • Follow these steps to save live nephelometer data to the laptop using Windows HyperTerminal:  1. Open Windows HyperTerminal Windows HyperTerminal is a program that came standard on older versions of Windows. The folder you have should contain two files: one with type ‘DLL File’ and one with type ‘Application’. Open the application file to run the program.   2. Close the pop-up window asking for location information and then confirm.                  137   3. Create a new connection. A new connection box will now pop up. Give an arbitrary name such as ‘1’ to the connection and select ok.            4. The program will again ask for location information. Cancel again (same as step 2).       5. Start a new connection to the nephelometer. This ‘Connect To’ pop-up will appear following the cancellation of the previous windows. The important part here is choosing the correct COM port in the lower drop-down. Typically, the program will automatically select the correct port, but if the next step doesn’t work. You will need to go to the Device Manager to find which USB port is actively connected to the nephelometer and repeat this step.     138  6. Choose the correct settings for the connection. For HyperTerminal to receive data from the nephelometer, the connection settings here must match the settings on the instrument. The correct settings are shown here – typically the only value that needs to be changed is the ‘Bits per second’ to 9600.   7. The connection will now start. If the connection was successful, live data from the nephelometer will now start appearing on the screen. The columns present are: Date, Time, Bscat, Sample Temperature, Enclosure Temperature, Relative Humidity and Pressure followed by placeholders. Bscat is the back-scattering value measured by the instrument and is well correlated with the PM2.5 concentration.     8. IMPORTANT: Set HyperTerminal to save the incoming data to a text file. Data appearing on the screen is not saved unless you go to the ‘Transfer’ menu and select ‘Capture Text…’. If this step is missed there will be no data saved from the monitoring trip.         139  9. Create a file and start saving data. After selecting ‘Capture Text…’  you need to browse to the location you want to save the file and enter the file name as:   YY_MM_DD_Location_TripNo_NEPH.txt   Then click ‘Start’ to begin capturing the 1-second data from the nephelometer.      10. Confirm data is being saved. To confirm HyperTerminal is now saving data, the Capture word in the bottom border will now be bolded (shown here in the red box – light grey in previous screenshots). You should also confirm the file has been created by navigating to the trip folder in your file explorer.      140  - List of Mobile Monitoring Runs Table A-1: List of mobile monitoring runs in the Whistler / Pemberton route pair. Run # Date Driving Route Night/Day Code Route Start Route End 1 2017-01-05 Whistler Night WN1 21:15 00:51 2 2017-01-06 Pemberton Night PN1 20:48 00:13 3 2017-01-07 Whistler Day WD2 12:08 14:59 4 2017-01-07 Whistler Night WN2 20:32 23:45 5 2017-01-08 Whistler Night WN3 19:55 23:26 6 2017-01-09 Pemberton Night PN2 19:53 23:23 7 2017-01-10 Pemberton Night PN3 20:22 23:41 8 2017-01-11 Whistler Night WN4 20:48 00:07 9 2017-01-12 Pemberton Night PN4 20:19 23:33 10 2017-01-13 Pemberton Day PD1 12:00 14:54 11 2017-01-13 Whistler Night WN5 20:52 00:11 12 2017-01-14 Pemberton Day PD2 11:59 15:09 13 2017-01-14 Pemberton Night PN5 20:57 00:28 14 2017-01-15 Whistler Night WN6 21:06 00:45 15 2017-01-16 Whistler Day WD3 12:17 15:11 16 2017-01-16 Pemberton Night PN6 21:03 00:39 17 2017-01-17 Whistler Night WN7 21:18 01:02 18 2017-01-18 Pemberton Night PN7 21:05 01:05    141  Table A-2: List of mobile monitoring runs in the Courtenay-Cumberland / Courtenay-Comox route pair. Run # Date Driving Route Night/Day Code Route Start Route End 19 2017-01-24 Courtenay-Comox Night CCX-N1 21:33 01:08 20 2017-01-25 Courtenay-Cumberland Night CCD-N1 21:07 00:50 21 2017-01-26 Courtenay-Comox Night CCX-N2 21:06 00:50 22 2017-01-27 Courtenay-Cumberland Night CCD-N2 21:10 00:45 23 2017-01-28 Courtenay-Comox Night CCX-N3 21:15 00:46 24 2017-01-29 Courtenay-Cumberland Night CCD-N3 21:22 00:57 25 2017-01-30 Courtenay-Comox Day CCX-D1 12:05 15:32 26 2017-01-30 Courtenay-Comox Night CCX-N4 21:10 00:38 27 2017-01-31 Courtenay-Cumberland Night CCD-N4 21:18 01:07 28 2017-02-01 Courtenay-Comox Night CCX-N5 21:11 01:17 29 2017-02-02 Courtenay-Cumberland Day CCD-D1 12:27 15:37 30 2017-02-02 Courtenay-Cumberland Night CCD-N5 21:20 00:51 31 2017-02-03 Courtenay-Comox Day CCX-D2 12:27 16:26 32 2017-02-03 Courtenay-Comox Night CCX-N6 20:58 01:07 33 2017-02-04 Courtenay-Cumberland Day CCD-D2 12:26 16:20 34 2017-02-04 Courtenay-Cumberland Night CCD-N6 21:35 02:01 35 2017-02-05 Courtenay-Comox Night CCX-N7 21:09 01:25 36 2017-02-06 Courtenay-Cumberland Night CCD-N7 21:12 01:17    142  Table A-3: List of mobile monitoring runs in the Vanderhoof / Fraser Lake route pair. Run # Date Driving Route Night/Day Code Route Start Route End 37 2017-02-16 Vanderhoof Night VN1 21:18 23:58 38 2017-02-17 Fraser Lake Night FLN1 21:06 00:00 39 2017-02-18 Vanderhoof Night VN2 21:14 23:48 40 2017-02-19 Fraser Lake Night FLN2 21:18 00:17 41 2017-02-20 Vanderhoof Night VN3 21:37 00:19 42 2017-02-21 Vanderhoof Day VD1 12:33 15:03 43 2017-02-21 Fraser Lake Night FLN3 21:17 00:11 44 2017-02-22 Fraser Lake Day FLD1 12:38 15:18 45 2017-02-22 Vanderhoof Night VN4 21:16 00:04 46 2017-02-23 Fraser Lake Night FLN4 21:28 00:34 47 2017-02-24 Vanderhoof Night VN5 21:35 00:22 48 2017-02-25 Fraser Lake Night FLN5 21:18 00:15 49 2017-02-26 Vanderhoof Night VN6 21:37 00:22 50 2017-02-27 Vanderhoof Day VD2 12:36 15:03 51 2017-02-27 Fraser Lake Night FLN6 21:24 00:28 52 2017-02-28 Vanderhoof Night VN7 21:24 00:14 53 2017-03-01 Fraser Lake Day FLD2 12:36 15:06 54 2017-03-01 Fraser Lake Night FLN7 21:23 00:17    143  - Levoglucosan Analysis Procedure 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.00.18 Last Update: 01/09/13    Introduction  Levoglucosan (Figure 1) is a sugar anhydride and is used as a molecular marker for the detection and evaluation for the presence of wood smoke in air.  The components detected in wood smoke are numerous: PAH’S, aldehydes, 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 mmor 41 mm teflon filters 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 (CAS #: 498-07-7)   C6H10O5        M.W. = 162.142 Apparatus and Chemicals A. Apparatus: Analytical Instrument: GC/MS System - Agilent Technologies 5973 GC/MSD Centrifuge: Hitachi HiMac centrifuge (CT5DL model) Sampling Medium: Gelman TefloTM W/Ring – PTFE Membrane W/PMP Ring: 2.0 um 37 mm  P/N R2PJ037  Filter Cutting Tool: [Method 1] Teflon filter cutting tool (see Figure 2) [Method 2] Scissors, scalpel, forceps and Petri dish. 144  Figure 2: Teflon filter cutting tool   B. Chemicals – Supplier Details 1,6-Anhydro-beta-beta-D-glucopyranose (Levoglucosan)  – Sigma-Aldrich P/N 316555-1G  (99.9% purity)  1,3,5-Tri-isopropylbenzene (Internal Standard) – Fluka P/N 92075 (97% purity)  7-Dehydrocholesterol (Surrogate) - Sigma-Aldrich  Ethyl Acetate – Fisher analytical grade  MSTFA + 1% TMCS (N-Methyl-N-trimethylsilyltrifluoroacetamide + 1% Trimethylchlorosilane 10 x 1 mL ampules - Pierce Chemicals P/N 48915  Pyridine (ACS grade) – Fluka # 82702 - 99.8% purity  C. Chemicals – Preparation of Stock Solutions 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 25 mL volumetric flask and top up with ethyl acetate (depending on sample matrix, this dilution can be altered).  Mix vigorously to dissolve all the crystals and to aid solubilization, ultrasonication can assist in this process.  Make sure no solid crystals are undissolved.  The stock solution can be stored at -80 oC.  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 0.1 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 vigorously to dissolve all the crystals and to aid solubilizing, ultrasonication can assist in this process.  Make sure no solid crystals are undissolved.  The stock solution can be stored at -80 oC.  Calculate the final concentration in nanograms per microliter (ng/uL) and record the date of preparation. 145   Preparation of Tri-isopropylbenzene Internal Standard: Transfer 5 uL of tri-isopropylbenzene into 25 mL volumetric flask and top up with ethyl acetate.  Dilute to an intermediate stock at an appropriate level (5-20 times). Spike 10 uL of this solution into each GC vial after derivatization has been completed.   Sample Preparation Procedure  A. Removal of Teflon portion of the filter If the entire filter is to be analyzed for levoglucosan, Method 1 should be used.  If the filter is to be cut in half (e.g. to analyze one half for levoglucosan and one half for a different analyte), Method 2 should be used.  Method 1 Each teflon filter has an outer plastic ring that maintains the teflon filter’s round shape.    Removing the telfon filter material is conducted with a special tool (Figure 2) designed to position and cut out the teflon portion and remove the outer plastic ring.  For 37 mm teflon filters place the filter inside a GPM 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 4 mL extraction vessel.  Prior to the extraction and derivatization procedures spike each sample with 20 uL of the stock 7-dehydrocholesterol standard (surrogate).  Method 2 If the filter is to be cut in half, this must be done BEFORE removing the outer plastic ring.  Using the forceps, grip the filter by the outer plastic ring and hold it above a Petri dish.  Use clean scissors to cut the filter in half as accurately as possible.  Place the two filter halves into separate Petri dishes.  Using the forceps to brace the outer ring, use the scalpel to carefully cut the filter material away from the ring.  Transfer the filter to a 4 mL extraction vessel.  Prior to the extraction and derivatization procedures spike each sample with 20 uL of the stock 7-dehydrocholesterol standard (surrogate). 146  B. 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 15 uL of pyridine and 30 uL of MSTFA + 1% TMCS solution.  Vortex for 10-20 secs and place the samples in a dark location for a minimum of 6 hours to complete the derivatization.  Prior to GC/MS analysis spike 10 uL of tri-isopropyl benzene internal standard into each vial.  GC/MS Conditions  DB-5 (5% phenyl) capillary column 30 meters x 0.25 mm I.D. with 0.25 um film thickness  Temperature Program: 65 oC (1 min hold) to 310 oC @ 25 oC (5 min hold)  Run time (mins): 15 mins  Injection Port Temperature:   290 oC  Injection Vol (uL): 1 uL  Splitless Injection Time (min):  0.50 min  Inlet Pressure (psi):  10 psi with constant flow  Initial Flow (mL/min):   1.1 mL/min         147  TABLE 1:  Single Ion Monitoring (S.I.M.) for Levoglucosan, Internal Standard and Surrogate   Figure 3: S.I.M. Chromatogram of Levoglucosan-TMS, Tri-isopropyl benzene (Istd) and 7-Dehydrocholesterol (Surrogate)     Figure 4:  Full Scan Mass Spectrum of the Trimethylsilyl derivative of Levoglucosan       Component Retention Time (min) Quan Mass (Q1) Quan Mass (Q2) Quan Mass (Q3) Dwell Time (msec) Istd 5.681 161.00 189.00 204.00 50 Levoglucosan 7.527 204.00 217.00 333.00 50 Surrogate 13.522 325.00 351.00 456.00 50 148  Figure 5: Limited Mass Chromatogram of Quan Ion of Levoglucosan-TMS (m/z 333)        References  Determination of Levoglucosan in Atmospheric Fine Particulate Matter – Christopher Simpson, Russell L.Dills, Bethany S. Katz, and David A. Kalman, Dept of Environmental and Occupational Health Sciences, University of Washington, Technical paper ISSN 1047-3289 J. Air and Waste Management Association 54:689-694   Method Revisions Revision Number  Author   Date  Description SOEH-SOP# A.00.16    Timothy Ma      07/14/05          1st Version SOEH-SOP# A.00.17  Timothy Ma  09/24/08 2nd Version SOEH-SOP# A.00.18  Jeff Nichol  01/08/13 3rd Version    149  - R Functions This appendix includes important functions created in r for the spatial analysis and mapping performed in this thesis. Required r packages for these functions: • raster • sp • rgdal • ggmap  1. Example script to create a raster layer to cover a region of interest. ############################################################################# # Script to create a raster layer to cover a community of interest  #   to be used for spatial averaging and smoothing of mobile data ############################################################################# # Steps taken from example at:  #  http://stackoverflow.com/questions/9542039/resolution-values-for-rasters-in-r # CREATE WHISTLER / PEMBERTON RASTER LAYER # #  define latitude and longitude boundary box  xtll <- matrix(nrow = 2, ncol = 2)  xtll[1,1] <- c(-123.060) # x min  xtll[1,2] <- c(50.076)   # y min  xtll[2,1] <- c(-122.636) # x max  xtll[2,2] <- c(50.351)   # y max  # Convert to SpatialPoints with world epsg:4326  xtll=SpatialPoints(xtll,CRS("+init=epsg:4326"))  # convert CRS to EPSG:3005 (NAD83/BC Albers)  spTransform(xtll,CRS("+init=epsg:3005")) # Take the extent from the previous result and round to nearest 10m # extent: 1210374, 1239239, 566606.9, 598519.8  (xmin, xmax, ymin, ymax)  ext = extent(1210370, 1239240, 566600, 598520)  # Determine ncol and nrow by counting the number of rows and columns to make # each side of a 3x3 square = approx 100m  length(1210370:1239239)/(100/3) # ncol = 866  length(566600:598520)/(100/3)   # nrow = 958 # create raster with these extents, calculated # of columns and under same  # projection  r.WP = raster(ext, ncol=866, nrow=958, crs="+init=epsg:3005")   rm(ext, xtll)  150  2. Function to spatially average and smooth mobile monitoring results across a raster grid for each monitoring run.  ############################################################################# # FUNCTION - Focal.Smoothing.RasterStackTrips(shp, r, variable) #     - creates a raster stack to hold all individual trip rasters #     - loops through individual trips in shp and: #         - overlays raster layer onto SpatialPointsDataFrame with rasterize   #   function #             - creates 2 rasters, one with raw counts and one with means of #       all records from within each cell #             - sets cells with counts less than 1 to NA #             - multiplies the 2 rasters to obtain a mean*count raster layer #         - performs focal smoothing on raster layers with focal function #   using an equally weighted 3x3 box #    (to create overall effect of 100mx100m) #         - divides focally smoothed mean*count layer by the focally  #   smoothed count layer to effectively create a raster layer of #   means focally smoothed using a 3x3 grid weighted by cell counts #         - adds the trip raster to the raster stack #     - returns the completed raster stack # # ## INPUTS: # shp = SpatialPointsDataFrame # r = blank raster layer created to cover the area of interest with cells #      approximately 33m x 33m # variable = column name of variable to use for cell means ############################################################################  Focal.Smoothing.RasterStackTrips <- function(shp,r,variable){      # create empty raster stack    s <- stack()      # loop through individual trips    for(trip in unique(shp$Trip)){          # subset trip data      trip.shp <- shp[which(shp$Trip == trip),]          # Project spatial data frame to match raster projection of BC Albers      trip.shp <- spTransform(trip.shp,CRS("+init=epsg:3005"))     151           ## Use rasterize function to overlay dataframe and produce 2 layers:     #     1 - r.count - counts the number of records in each cell     #     2 - r.mean  - takes the mean in each cell      r.count <- rasterize(x = trip.shp, y = r, field =  trip.shp@data[,variable], fun = 'count', na.rm = T)      r.mean <- rasterize(x = trip.shp, y = r, field =  trip.shp@data[,variable], fun = mean, na.rm = T)          # Set the cells where there are less than 1 record to NA      r.mean[which(r.count@data@values < 1)] <- NA      r.count[which(r.count@data@values < 1)] <- NA          # multiply the 2 rasters to create a mean*count layer = r.MxC      r.MxC <- r.mean * r.count          # Use focal function to perform focal smoothing on the count and  mean*count layers     #   (Weighting is with an equally weighted 3x3 grid)      r.count.SMOOTHED <- focal(r.count, w=matrix(1,3,3), fun=sum, na.rm = T)      r.MxC.SMOOTHED <- focal(r.MxC, w=matrix(1,3,3), fun=sum, na.rm = T)          # Divide the smoothed mean*count by the smoothed count layer to  effectively create a mean layer focally smoothed using a 3x3 grid  weighted by the cell counts      r.mean.SMOOTHED <- r.MxC.SMOOTHED / r.count.SMOOTHED          # add trip name to layer      r.mean.SMOOTHED@data@names <- trip          # Add trip layer to the raster stack        s <- stack(s, r.mean.SMOOTHED)    }       return(s)  }   152  3. Function to extract the average spatial pattern across the individual run raster layers.  ############################################################################# # FUNCTION - RasterStack.Average.Pattern(stack, min.trips) #     - calculates the average pattern of the raster stack #     - counts how many trips were not NA in each cell and sets the cells in #   the average layer to NA if this is below the min.trips number #  (this is to remove cells that were only monitored in 1 of 10 #  trips for example) #     - returns the completed average pattern raster layer # # ## INPUTS: # stack = raster stack with the indiv trip patterns # min.trips = minimum number of trips to not equal NA in each cell ############################################################################  RasterStack.Average.Pattern <- function(stack, min.trips){      # Calculate the average pattern of the raster stack    r.avg.pattern <- mean(stack, na.rm = T)      # Count the number of none NA values in each cell in the raster brick    rNA <- sum(!is.na(stack))      # Set the values of cells in the average pattern layer to NA if there were not at least the min.trips number of none NA values in the raster brick    r.avg.pattern[which(rNA@data@values < min.trips)] <- NA       return(r.avg.pattern)  }   153  4. Function to convert the final raster layers to polygons ready for mapping.  ############################################################################# # FUNCTION - convert.raster.to.polygons(Rast) #           - converts raster to polygons #           - transforms to WGS84 # # ## INPUTS: # Rast = raster to be converted ############################################################################  # (Based on: http://stackoverflow.com/questions/33530055/add-raster-to-ggmap-base-map-set-alpha-transparency-and-fill-to-inset-raster)  convert.raster.to.polygons <- function(Rast){      # Convert to polygons     rtp <- rasterToPolygons(Rast)    rtp@data$id <- 1:nrow(rtp@data)    # add id column for join data after the fortify      # project    projection_wgs84 = CRS("+proj=longlat +datum=WGS84")    rtp = spTransform(rtp, projection_wgs84)      # convert to normal dataframe and merge the data to it    rtpFort <- fortify(rtp, data = rtp@data)    rtpFortMer <- merge(rtpFort, rtp@data, by.x = 'id', by.y = 'id')         return(rtpFortMer)  }   154  5. Function to download base maps from Google. ############################################################################# # FUNCTION - Download.Base.Map(Auto, Data, Center, Zoom) # #           - downloads google base map #           - if auto = T, centers the map on the center of the data #           - if auto = F, centers map based on variable ‘Center’ # # ## INPUTS: # Auto = T or F - whether to center based on data or manual # Data = Data that will be mapped # Center = Manually defined center to use if Auto = F  # Zoom = zoom level for get_googlemap (usually 11-14) ############################################################################  Download.Base.Map <- function(Auto, Data, Center, Zoom){      # If Auto = T, define the center for the map request    if(Auto == T){      Center <- c(lon = mean(c(min(Data$long, na.rm = T) - 0.01,                               max(Data$long, na.rm = T) + 0.01)),                  lat = mean(c(min(Data$lat, na.rm = T) - 0.01,                               max(Data$lat, na.rm = T) + 0.01)))    }    # Download the map with style command to instruct get_googlemap to:   #     1. Hide road labels   #     2. Hide administrative labels   #     3. Hide all points of interest   #     4. Plot landscape features in a simplified form   #     5. Hide landscape labels   #     6. Hide water body labels   #     7. Hide transit line labels (ferries etc)      Map <- get_googlemap(center = Center,                         zoom = Zoom,                         maptype = "roadmap",                         color = "bw",                         scale = 2,                         style = "feature:road|element:labels|visibility:off&style=feature:administrative|element:labels|visibility:off&style=feature:poi|element:all|visibility:off&style=feature:landscape|element:all|visibility:simplified&style=feature:landscape|element:labels|visibility:off&style=feature:water|element:labels|visibility:off&style=feature:transit|element:labels|visibility:off")       return(Map)    } 155  6. Function to create maps of final polygons.  ############################################################################# # FUNCTION - create.map(Base.Map, Polygons, Pal) #           - maps and fills polygons that have been pre-binned based  #    on z-score # # ## INPUTS: # Base.Map = Base map that has already been downloaded # Polygons = Polygon layer to add to the map # Pal = Colour palette with hex codes for colour fills ############################################################################  create.map <- function(Base.Map, Polygons, Pal){      # create map with base layer    m <- ggmap(Base.Map)      # add polygon layer values coloured by Z.bin column with  transparency (alpha) = 90%, and size = 0 to remove the polygon outlines    m <- m + geom_polygon(data = Polygons,                           aes(x = long, y = lat, group = group, fill = Z.Bin),                           alpha = 0.9,                           size = 0,                          show.legend = F)       # Add colour scale and legend    m <- m + scale_fill_manual(values = Pal)      # add themes to simplify the appearance of the map    m <- m + theme(axis.title=element_blank(),                   axis.text=element_blank(),                   axis.ticks=element_blank())      # print map    m   }  

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}]}"
                            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.24.1-0371217/manifest

Comment

Related Items