Observational constraints on air quality and greenhouse gases in the Greater Toronto Area

by

Stephanie C. Pugliese

A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy Graduate Department of Chemistry University of Toronto

© Copyright by Stephanie C. Pugliese (2017)

Observational constraints on air quality and greenhouse gases in the Greater Toronto Area

Stephanie C. Pugliese

Doctor of Philosophy

Graduate Department of Chemistry University of Toronto

2017 Abstract

Urban areas are highly complex environments that are large sources of a variety of atmospheric pollutants. To better understand the spatial and temporal variability of carbon dioxide (CO2) emissions in the GTA, we developed the Southern Ontario CO2 Emissions (SOCE) inventory, a fine resolution inventory based on emissions from seven source sectors for the year 2010. When the SOCE inventory was used in combination with a chemistry transport model, agreement with mixing ratios measured in southern Ontario was strong. We found that contributions from area sources (primarily natural gas combustion) played a dominant role overnight (contributing >80 %) while on-road sources played a significant role during the day (contributing >70 %).

13 When a fine-resolution CO2 inventory is not available, stable carbon isotopes (δ CO2) are often used as tools to source apportion CO2. We combined model output CO2 mixing ratios with GTA-

13 specific δ CO2 signatures of the dominant anthropogenic sources of CO2, determined as part of

13 13 this work, to simulate hourly δ CO2. When the simulated δ CO2 was compared to measurements at the Downsview site, the agreement was generally good. To evaluate the use of Keeling and

13 chemical mass balance (CMB) analyses, we estimated the δ CO2 signature of the local

ii enhancement from both model output and measurements and found that it was heavier (~3 per mil) than that calculated by the SOCE inventory.

The emission of primary pollutants in urban areas (e.g. nitrogen oxides (NOx) and carbon monoxide (CO)) can also contribute to the production of secondary pollutants such as ozone (O3).

Data from a governmental monitoring network in the GTA from 2000-2012 were used to explore the impact of O3 precursors on local O3 levels. Non-linear chemistry and influence of meteorology explained why reductions in precursor levels did not lead to improvements in O3, particularly for

2012.

iii

Acknowledgments

This degree has been such an incredible experience for me. I grew immensely as an individual during these past five years and have learned and achieved more than I ever thought possible. There are numerous people who have given me a great deal of support along the way.

The first person I would like to thank is my supervisor, Jennifer Murphy. Jen, you have been such an incredible role model for me, particularly as a young female in chemistry. For the past five years, you have allowed me to take my research in a unique direction, continually challenging me and teaching both inside and outside the walls of the Murphy group. You are an exceptional scientist and mentor, and I am extremely grateful to have had you as my supervisor. I would also like to express my gratitude to Jon Abbatt and James Donaldson, my internal committee members, for following my progress throughout my PhD and offering advice and encouragement along the way.

I would like to thank all the members of the Murphy group, past and present, for all their collaboration and good times over the past five years. To all the Murphsters that I have shared an office with, thank you for making every day fun and positive and, of course, full of weird conversations. Thank you to Jeffery Geddes who originally taught me to code in IGOR. Finally, I would like to give a very large thank you specifically to Amy Hrdina and Pegah Baratzadeh, two incredible colleagues who have taught me so much and given me so many reasons to smile over the past few years.

I would like to thank the Laboratoire des Sciences du Climat et l’Environnement (LSCE) in Paris, France for facilitating my exchanges. Specifically, I would like to thank my host supervisor Felix Vogel for sparking my interest in isotope chemistry. Felix, you have been an incredible teacher and mentor for the past 3 years, thank you for all that you have taught me.

I would like to acknowledge all the various funding sources (NSERC, OGS, QEII-GSST, CGCS, SGS, University of Toronto, and Department of Chemistry) without which, all of the exchanges and conferences would not have been possible.

Lastly, I would like to acknowledge the support of my family. To my parents, you have always provided me with every opportunity to follow my academic dreams. Your unconditional support and love has provided me with the drive to push forward these past five years. I cannot truly iv

express my gratitude for all that you have done. To my two brothers, our need to compete against each other surely contributed to my pursuit of a PhD – thank you for all the support you have given. Specifically to Anthony, for all of our chats about atmospheric research on the drive into campus and your invaluable help with Octave, I cannot thank you enough. To my grandparents, thank you for always being so proud to have a granddaughter who is a scientist, and continually asking me to explain exactly what I do. To my aunts and uncles and cousins, thank you for all your support and for making every little award and publication a big deal. And finally, to my fiancé Angelo, thank you for being an incredible partner during this crazy period of my life. You have helped me through all the disappointments and accomplishments that come with graduate school. For all your love and support over these past years, I am truly grateful.

v

Table of Contents

Acknowledgments...... iv

Table of ContentsList of Tables ...... vi

List of Tables ...... xi

List of Figures ...... xii

List of Appendices ...... xvi

Chapter 1 Introduction...... 1

1.1 Introduction to primary and secondary pollutants in urban environments ...... 1

1.2 Sources of CO2 in urban environments ...... 2

1.3 Bottom-up estimates of urban emissions ...... 3

1.3.1 The EDGAR v4.2 CO2 Inventory ...... 4

1.3.2 The FFDAS v2 CO2 Inventory ...... 5

1.3.3 Environment and Climate Change Canada’s CO Inventory ...... 6

1.4 Measurements of CO2 ...... 8

1.4.1 High precision, low-density measurements ...... 9

1.4.2 Low precision, high-density measurements ...... 12

1.4.3 Column measurements ...... 14

1.5 Use of tracers to source apportion CO2...... 16

12 13 1.5.1 Stable carbon isotopes, CO2 and CO2 ...... 16

1.5.2 Combustion products, CO and NOx ...... 21

1.6 Urban NOx – VOC – O3 chemistry and trends ...... 22

1.6.1 Urban NOx – VOC – O3 chemistry ...... 22

1.6.2 Trends in CO and NOx emissions in Ontario ...... 22

1.6.3 Role of meteorology and transport ...... 23

1.6.4 Observed trends in ground-level O3 concentrations through time ...... 25

1.7 Synthesis and Research Objectives ...... 27 vi

1.8 References ...... 29

Chapter 2 High-resolution quantification of atmospheric CO2 mixing ratios in the Greater Toronto Area, Canada ...... 42

2.1 Abstract ...... 42

2.2 Introduction ...... 43

2.3 Methods...... 45

2.3.1 Study region ...... 45

2.3.2 The Southern Ontario Greenhouse Gas Network ...... 46

2.3.3 GEM-MACH chemistry-transport model ...... 47

2.3.4 High-resolution SOCE inventory development ...... 49

2.3.4.1 Area emissions ...... 50

2.3.4.2 Point emissions ...... 51

2.3.4.3 On-road emissions ...... 51

2.3.4.4 Off-road mobile emissions ...... 51

2.3.4.5 Marine emissions ...... 52

2.3.4.6 Residential and commercial emissions ...... 52

2.3.5 Biogenic Flux ...... 53

2.4 Results and Discussion ...... 53

2.4.1 The SOCE Inventory...... 53

2.4.2 Comparison of the SOCE inventory with other inventories ...... 56

2.4.3 Preliminary analyses using the SOCE, FFDAS v2 and EDGAR v4.2 inventories with FLEXPART ...... 58

2.4.4 Simulation of CO2 mixing ratios in the Greater Toronto Area ...... 59

2.4.5 Quantifying the model-measurement agreement ...... 62

2.4.6 Sectoral contributions to simulated CO2 mixing ratios ...... 64

2.5 Conclusions ...... 65

2.6 Contributions...... 66 vii

2.7 References ...... 67

2.8 Appendix A ...... 73

13 Chapter 3 Characterization of the δ C signatures of anthropogenic CO2 emissions in the Greater Toronto Area, Canada ...... 83

3.1 Abstract ...... 83

3.2 Introduction ...... 84

3.3 Methods...... 86

3.3.1 Study region ...... 86

3.3.2 Transportation fuels sample collection ...... 87

3.3.3 Heating and electricity generation fuels sample collection ...... 88

13 3.3.4 Ambient CO2 and δ CO2 measurements ...... 88

3.3.5 Isotope analyses ...... 89

3.3.6 Mass balance approximations ...... 90

3.3.7 Statistical analyses ...... 91

3.4 Results and Discussion ...... 91

3.4.1 The carbon isotopic composition of transportation fuels ...... 91

3.4.2 The carbon isotopic composition of heating and electricity generation fuels .97

13 3.4.3 Sensitivity case study using the GTA δ CO2 signatures ...... 99

3.4.4 Simplified isotope tracer based CO2 partitioning ...... 100

3.5 Conclusion ...... 104

3.6 Contributions...... 105

3.7 References ...... 105

13 Chapter 4 Using high-resolution simulations of atmospheric δ CO2 signatures to understand variability in source contributions of CO2 in the Greater Toronto Area, Canada ...... 110

4.1 Abstract ...... 110

4.2 Introduction ...... 111

viii

4.3 Methods...... 113

4.3.1 Geographic Domain ...... 113

13 4.3.2 Ambient measurements of CO2 and δ CO2 ...... 113

4.3.3 GEM-MACH chemistry-transport model and the SOCE inventory ...... 114

13 4.3.4 Conversion of CO2 mixing ratios to δ CO2 ...... 114

4.3.4.1 Area emissions ...... 115

4.3.4.2 Point emissions ...... 116

4.3.4.3 On-road emissions ...... 116

4.3.4.4 Off-road mobile emissions ...... 116

4.3.4.5 Marine emissions ...... 116

4.3.4.6 Residential and commercial emissions ...... 117

4.3.4.7 USA and Quebec emissions ...... 117

4.3.5 The chemical mass balance approximation and Keeling plot analysis ...... 117

4.4 Results and Discussion ...... 118

13 4.4.1 Simulation of δ CO2 in the Greater Toronto Area...... 118

13 4.4.2 Diurnal variability of measured and modelled δ CO2 ...... 120

4.4.3 Assessment of the chemical mass balance approximation to source apportion CO2 ...... 121

4.4.4 Assessment of the Keeling linear mixing model to source apportion CO2....125

4.4.5 Emission sources of C1 and C2...... 130

4.5 Conclusions ...... 131

4.6 Contributions...... 132

4.7 References ...... 132

4.8 Appendix B ...... 137

Chapter 5 The impacts of precursor reduction and meteorology on ground-level ozone in the Greater Toronto Area ...... 138

5.1 Abstract ...... 138 ix

5.2 Introduction ...... 139

5.3 Methods...... 142

5.3.1 Study region and data collection ...... 142

5.3.2 O3, NOx and VOC analyses ...... 144

5.3.3 Meteorological analyses...... 145

5.4 Results and Discussion ...... 145

5.4.1 Long term precursors and O3 levels (2000-2012) ...... 145

5.4.2 O3 levels from 2008-2012 ...... 149

5.4.3 Meteorological influences on O3 ...... 149

5.4.4 VOC – or NOx-limited ozone production ...... 154

5.4.5 GTA OH radical concentration ...... 155

5.4.6 Assessing the success of the Climate Change, Clear Air and Sustainable Energy Action Plan ...... 157

5.5 Conclusion ...... 158

5.6 Contributions...... 159

5.7 References ...... 159

Chapter 6 Conclusions and Future Research ...... 163

6.1 Conclusions ...... 163

6.2 Updated trends in ozone and its precursors across the GTA ...... 166

6.3 Future work ...... 170

6.3.1 Direct measurements of anthropogenic CO:CO2 in the GTA ...... 170

6.3.2 Development of a low precision, high-density network of CO2 sensors in the GTA ...... 171

6.3.3 Ground level ozone production in the GTA ...... 171

6.3.4 Column observations of O3 and NOx ...... 172

6.3.5 Column observations of CO2 ...... 173

6.3.6 References ...... 174 x

List of Tables

Table 2.1: Summary of atmospheric measurement programs in Southern Canada operated by Environment and Climate Change Canada

Table 2.2: Anthropogenic CO2 emissions for the year 2010 in the black-box area (shown in Figure 2a) by sector. Values in parentheses indicate the percentage contribution of the sector to the total CO2 emissions in the black-box area.

13 Table 4.1: δ CO2 signatures and contributions of each source category

13 Table 4.2: Calculated median δ Cs and contributions of C1 and C2 using the SOCE inventory and chemical mass balance

13 Table 4.3: Calculated median δ Cs and contributions of C1 and C2 using the SOCE inventory and the Keeling linear mixing model

Table 5.1: Percent of days and exceedances (at Downtown2) each summer (2008-2012) that were affected by air transport from the W-NE, W-SE or local/stagnant air.

Table 5.2: The percent difference between 2004 and 2012 NO2 and VOC concentrations (ppb) at all stations monitored (Junction was not included as monitoring stopped in 2005).

xi

List of Figures

Figure 1.1: Schematic of cavity ring down spectroscopy (CRDS). Diagram is not to scale.

Figure 1.2: The location of the 22 in situ CO2 and CH4 observational sites in Canada, as of 2016 (Worthy, 2017 (unpublished work)).

13 Figure 1.3: Example Keeling plot of δ CO2 vs. 1/CO2 for an overnight period in January in the GTA Figure 1.4: Overview of end-member δ13C variability at various location globally

Figure 1.5: Ontario annual average emission estimates of CO and NOx (tonnes) from 2000- 2015

Figure 2.1: Total anthropogenic CO2 emissions for a weekday in February 2010 estimated by the SOCE inventory for the province of Ontario and by the FFDAS v2 inventory for the remainder of the GEM-MACH PanAm domain. Locations of in-situ measurements of CO2 in the Southern Ontario GHG Network are shown in the inset (Downsview = square, Egbert = circle, Hanlan’s Point = triangle, Turkey Point = diamond). The Downsview and Hanlan’s Point sites are both located in the GTA. Units: g CO2/second/grid cell.

Figure 2.2: Anthropogenic CO2 emissions for a weekday in February 2010 in southern Ontario. Emissions are estimated by the SOCE inventory for the (a) Area sector; (b) sum of the Residential and Commercial sectors; (c) Point sector; (d) Marine sector; (e) On- road sector; (f) Off-road sector. Units: log10(g CO2/second/grid cell).

Figure 2.3: Comparison of spatial distribution of annual CO2 emissions inventories for the black-box area (shown in Figure 2a) at 0.1o x 0.1o resolution. Panel a) shows the FFDAS v2 inventory estimate, Panel b) shows the EDGAR v4.2 inventory estimate and Panel c) shows the SOCE inventory estimate. Units: log10(tonne CO2/year/grid cell). Domain totals are shown on top of each panel and locations of in-situ measurements of CO2 for three stations in the Southern Ontario GHG Network are shown on Panel a (Downsview = square, Hanlan’s Point = triangle, TAO = pentagon). The other two stations, Egbert and Turkey Point, are located outside this area.

Figure 2.4: Time series of measured (blue) and modelled February afternoon (12:00-16:00 EST) CO2 mixing ratios for the four sites used in this study. The red and gold markers are the modelled mixing ratios when using the SOCE inventory and the FFDAS v2 inventory, respectively.

Figure 2.5: Time series of mean measured (blue) and modelled diurnal CO2 mixing ratios at the four sites considered in this study for January – March 2016. The red and gold markers are the modelled diurnal mixing ratios when using the SOCE inventory and the FFDAS v2 inventory, respectively. Note the difference in scale for urban and rural sites. xii

Figure 2.6: Scatter plot of the modelled and measured afternoon (12:00-16:00 EST) CO2 mixing ratios from January-March, 2016 at the four monitoring stations used in this study. The top and bottom panels show measurement-model correlation when the SOCE inventory and the FFDAS v2 inventory were used, respectively. The model vs. measurement Correlation Coefficient (R), root mean square error (RMSE) and mean bias (MB) (units: ppm) are provided within each panel. Solid lines are the standard major axis regression lines and dashed lines are 1:1 lines shown for reference.

Figure 2.7: Modelled sectoral percent contributions to diurnal local CO2 enhancement for February 2016 at Downsview averaged by day of week. Note: Area = Area + Residential natural gas combustion + Commercial natural gas combustion. (Time zone is EST).

13 Figure 3.1: Summary of δ C signatures of all fuels. The boxes show the inter-quartile range and median values of each sample and the whisker top and bottom show the 90th and 10th percentiles, respectively. Median values are also reported in parentheses on the right of the figure. Error bars on non-boxed data are 1 sigma of standards, 0.20 ‰.

13 Figure 3.2: Map of gasoline (grade 87) (left) and diesel (middle) δ CO2 signatures and their respective histogram (right) for the Winter 2015-A, Summer 2015 and Winter 2015-B sampling campaigns.

13 Figure 3.3: Comparing δ CO2 signatures of gasoline with varying octane contents. The boxes show the inter-quartile range and median values of each gasoline sample and the whisker top and bottom show the 90th and 10th percentiles, respectively.

13 Figure 3.4: Comparing δ CO2 signatures of gasoline (top) and diesel (bottom) from different distributors in the Winter 2015-A campaign. The boxes show the inter-quartile range and median values of each sample and the whisker top and bottom show the 90th and 10th percentiles, respectively.

13 Figure 3.5: Daily carbon isotopic signatures of CO2 source (δ Cs) in the GTA for 2014. The top and bottom shaded regions on the graph show the approximate range of signatures for gasoline and/or soil respiration and natural gas combustion, respectively. Error bars represent the standard deviation of the y-intercept of the Keeling curve.

Figure 3.6: Histograms of 2014 winter (top) and summer (bottom) contributions of CO2 from End-member #1 and End-member #2 in the GTA

13 Figure 4.1: Time series of measured (blue) and modelled (red) CO2 signatures at the Downsview site for January 2016.

xiii

13 Figure 4.2: Scatter plot of the modelled and measured CO2 signatures from January-March 2016 at the Downsview station. The model vs. measurement Correlation Coefficient (R), root mean square error (RMSE) and mean bias (MB) are provided within the plot. The solid line is the standard major axis regression and dashed line is 1:1 shown for reference.

13 Figure 4.3: Time series of mean measured (blue) and modelled (red) diurnal δ CO2 signatures at the Downsview site. Error bars show standard deviation of modelled signatures.

13 12 13 Figure 4.4: Plot of δ Cs vs. Cs using a) the SOCE inventory, b) modelled CO2 and δ CO2 th 12 13 12 with 5 percentile estimates of background CO2 and δ CO2, c) measured CO2 13 and δ CO2

13 Figure 4.5: a) Time series of overnight (22:00-06:00) δ Cs calculated using modelled (red) and 12 13 measured (blue) CO2 and CO2 (with Keeling analysis) at Downsview in January-March 2016. The black curve shows the SOCE inventory sector-weighted 13 overnight averaged δ Cs. b) Box-plots of the distribution of overnight (22:00- 13 12 13 06:00) δ Cs calculated using modelled and measured CO2 and CO2 (with the Keeling analysis) and the SOCE inventory. The boxes show the inter-quartile range and median values of each sample and the whisker top and bottom show the 90th and 10th percentiles, respectively.

13 Figure 4.6: a) Time series of hourly δ Cs calculated using modelled (red) and measured (blue) 12 13 CO2 and CO2 (from Keeling analysis) at Downsview in January-March 2016. 13 The black curve shows the SOCE inventory sector-weighted hourly averaged δ Cs. Error bars indicate standard deviation in the intercept extracted from the Keeling 13 curves. b) Box-plots of the distribution of daily δ Cs calculated using modelled and 12 13 measured CO2 and CO2 (with the Keeling theory) and the SOCE inventory. The boxes show the inter-quartile range and median values of each sample and the whisker top and bottom show the 90th and 10th percentiles, respectively.

Figure 5.1: Monitoring stations in the Greater Toronto Area used for the collection of NOx and O3 data (subscript “a”), VOC data (subscript “b”), and meteorological data (subscript “c”).

Figure 5.2: Annual summer midday NO2 concentrations (ppb) in GTA urban (a) and suburban areas (b); slopes in ppb year-1. Annual summer VOC reactivity (s-1) (c); slopes in s- 1 year-1.

Figure 5.3: Annual summer VOC reactivity (s-1); slopes in s-1 year-1.

Figure 5.4: Annual summer average maximum 8-hr O3 concentrations (ppb) for Toronto urban (left) and suburban (right) areas; slopes in ppb year-1-.

Figure 5.5: Annual summer average maximum 8-hr Ox concentrations (ppb) for Toronto urban (left) and suburban (right) areas; slopes in ppb year-1-.

xiv

Figure 5.6: The number of days exceeding 65 ppb O3 Canada-wide Standard (solid lines and triangles) and the number of days exceeding 30 °C (dashed lines and squares) at the Toronto North station. The number above each marker is the Design Value for that year.

Figure 5.7: Ox hourly summer averages at North Toronto.

2 Figure 5.8: Regression of Downtown2 8-hr O3 versus incoming solar radiation (W/m ) for 2012 (black lines are the thresholds of incoming solar radiation where 65 ppb of O3 is exceeded).

Figure 5.9: Cumulative distributions of the midday average solar radiation (W/m2) experienced on each summer day (2008-2012).

-1 Figure 5.10: OH reactivity (s ) to NO2 and VOCs in the “early” period (2002-2003) and “late” period (2011-2012) at the Downtown sites.

Figure 5.11: Histograms of 1-butene:2-butene (a) and 1,2,3-trimethylbenzene:ethylbenzene (b) in the early (2002-2003) and late (2011-2012) periods.

Figure 6.1: Updated annual summer midday NO2 concentrations (ppb) in GTA urban (left) and suburban (right) areas. Dashed lines plot the original trend as reported in Chapter 5 (2000-2012); solid lines, slopes (ppb year-1) and p-values refer to updated trends (2000-2016).

Figure 6.2: Updated annual summer average maximum 8 h O3 concentrations (ppb) in GTA urban (left) and suburban (right) areas. Dashed lines plot the original trend as reported in Chapter 5 (2000-2012); solid lines, slopes (ppb year-1) and p-values refer to updated trends (2000-2016).

Figure 6.3: Updated annual summer average maximum 8 h Ox concentrations (ppb) in GTA urban (left) and suburban (right) areas. Dashed lines plot the original trend as reported in Chapter 5 (2000-2012); solid lines, slopes (ppb year-1) and p-values refer to updated trends (2000-2016).

Figure 6.4: Updated plot of the number of Canada-Wide Standard (CWS) exceedances (green) and the number of days exceeding 30oC (red) at the Toronto Downtown2 station. The number above each marker is the Design Value for that year.

xv

List of Appendices

Appendix A – Chapter 2 supplement

Appendix B – Chapter 4 supplement

xvi

Chapter 1 Introduction Chapter 1 1.1 Introduction to primary and secondary pollutants in urban environments

While occupying only 3 % of the land, urban areas are locations of residence for > 50 % of the global population and this figure is projected to increase to ~70 % by 2030 (WHO, 2015). When considering Canada alone, the urban population accounts for an even larger fraction of the total (81 % in 2011) (Statistics Canada, 2011) while urban land coverage is only 0.25 % (Statistics Canada, 2009d). This high urban population density translates into a concentrated demand for fossil fuel combustion for transportation, power and heat generation, and therefore produces a complex mixture of pollutants emitted into the urban atmosphere. The relative contribution of different pollutants in this environment is dependent on the fuel source, as well as a variety of economic, social and technological factors. Primary pollutants, such as nitrogen oxides (NOx), carbon monoxide (CO) and carbon dioxide (CO2), are emitted directly from combustion sources while secondary pollutants, such as ozone (O3), are formed in the atmosphere, typically as a product of chemical reactions between primary pollutants and sunlight.

Recent research has provided evidence that urban air pollution has implications for human and environmental health, atmospheric chemistry, as well as global climate change. Studies have shown that exposure to urban air pollution is associated with a broad range of acute and chronic health effects, ranging from minor physiological issues to premature death from respiratory and cardiovascular disease; it is estimated to kill ~1.2 million people annually (WHO, 2016). In the Greater Toronto Area (GTA) specifically, air pollution is estimated to give rise to 1300 premature deaths and 3550 hospitalizations each year, with ground-level O3 estimated to be responsible for 13-29 % of these incidences (Toronto Public Health., 2014). Additionally, agricultural losses associated with urban air pollution exposure result in annual economic losses as high as CAD 9.6 million in the province of Ontario (MOE, 2005).

Primary emissions and subsequent secondary pollutants derived from urban emissions also play a key role in affecting atmospheric chemistry processes that occur regionally and globally. In particular, tropospheric O3 can affect the oxidative capacity of the atmosphere as it is the primary

1 2 source of the hydroxyl radical (OH), which determines the lifetimes of other trace gases affected by oxidation. Similarly, NOx can take part in reactions that consume or produce OH, affecting its own atmospheric lifetime and the lifetime of other gases. Because O3 and NOx have lifetimes sufficient to travel long distances, these impacts on atmospheric chemistry and air quality in general are reflected on regional scales.

A global scale impact of urban pollutant emissions is the ability to contribute to the issue of climate change. It has been reported that ~53-87 % of global anthropogenic CO2 emissions are from urban areas (IPCC-WG3, 2014). The additional CO2 in Earth’s atmosphere, associated with fossil fuel combustion and land use change, has a globally averaged radiative forcing of 1.82 ± 0.19 W m-2 (IPCC-WGI, 2013) and a lifetime of >100 years, suggesting that its impact on the Earth’s radiative budget will continue for decades.

Acknowledging the impacts of urban pollutants, many local governments are implementing programs aimed at reducing their emissions; for example the GTA made commitments in 2007 in the Climate Change Clean Air and Sustainable Energy Action Plan to reduce emissions of primary, smog-causing pollutants to 20 % below 2004 levels by 2012 (City of Toronto., 2007) and in Change is in the air: Toronto’s Commitment to an Environmentally Sustainable Future to reduce

CO2 emissions 30 % below 1990 levels by 2020 (Framework for Public Review and Engagement, 2007). However, in order to effectively implement and evaluate urban air pollution mitigation strategies, reliable inventories are required, particularly at high spatial and temporal resolution, as well as an understanding of historical trends of pollutants in response to past mitigation efforts. This thesis aims to address several of these issues in the GTA, specifically by using in situ observations and modelling of primary and secondary pollutants, particularly NOx, volatile organic compounds (VOCs), O3 and CO2, to better understand their spatial and temporal variability.

1.2 Sources of CO2 in urban environments

Globally, atmospheric CO2 concentrations have increased from 280 ppm in pre-industrial times to greater than 400 ppm presently, largely due to the changes in fossil fuel combustion and deforestation (Hartmann et al., 2013). In 2013, two sectors were responsible for almost two-thirds of global anthropogenic CO2 emissions: electricity and heat generation (42 %), primarily a result of the worldwide reliance on coal, the most carbon-intensive fossil fuel, and transportation (23 %)

(IEA, 2015). The remainder of anthropogenic CO2 emissions is primarily from the manufacturing

3 industry (19 %) and the residential sector (6 %) (IEA, 2015). When considering the regional scale of the GTA, anthropogenic CO2 emissions are derived from similar sources, however absolute sectoral contributions are not well-known. Currently, sectoral estimates of anthropogenic CO2 contributions in the GTA are made available by the EDGAR v.4.2 inventory (EDGAR, 2010) (estimating ~41 % and ~32 % for public electricity/heat production and transportation, respectively) which is very different from a locally derived estimate (~31 % and ~48 % for public electricity/heat production and transportation, respectively) (Greening Greater Toronto, 2012).

With the importance of urban environments to global CO2 emissions and the need for implementation of effective mitigation strategies, an accurate inventory of sectoral contributions is of ever-growing importance.

1.3 Bottom-up estimates of urban emissions

Inventories provide information on the spatial and temporal variability of emissions at a variety of scales (e.g. municipal, national and/or global). Bottom-up efforts often use an accounting-type approach to estimate source-specific emissions; typically activity data is collected (e.g. amount of coal burned by a power plant or the number of vehicle kilometers traveled) followed by an estimation of emissions factors for each activity (e.g. CO2 emissions per kg of coal burned or CO emissions per kilometer travelled by passenger cars). The product of the two provides a bottom- up estimate of emissions for a given source sector. Past efforts for compiling inventories of urban pollutants have been motivated from both a science and policymaking standpoint (Gurney et al.,

2009). From a scientific perspective, focusing on CO2 as an example, carbon cycle science requires more accurate and finely resolved quantification of emissions to inform atmospheric CO2 inversions, which are used to better understand the feedbacks between the carbon cycle and climate change (Gurney et al., 2009; Patarasuk et al., 2016). From a policy perspective, an accurate inventory of CO2 is required to verify anthropogenic mitigation efforts associated with policy agreements as well as in response to the emerging requirements of proposed carbon trading systems (Gurney et al., 2009; Patarasuk et al., 2016). A number of academic and government efforts have produced bottom-up emissions estimates at local/regional (Airparif, 2010; Gately et al., 2013; Jeong et al., 2014; Moran et al., 2015), national (Gurney et al., 2009; Kurokawa et al., 2013; Petron et al., 2008) and global scales (NCAR/UCAR, 2016; Oda and Maksyutov, 2011; Olivier et al., 2005; Rayner et al., 2010). In this thesis, work by Olivier et al. (2005) and Rayner et al. (2010) to develop the EDGAR v4.2 and the Fossil Fuel Data Assimilation System (FFDAS) v2

4 inventories, respectively, were primarily used (see Chapter 2) and therefore will be discussed in detail in Sections 1.3.1 and 1.3.2. Furthermore, work by Moran et al. (2015) to develop a fine resolution CO inventory for southern Ontario provided the basis of our work to develop the

Southern Ontario CO2 Emissions (SOCE) inventory (see Chapter 2) and therefore will be discussed in detail in Section 1.3.3.

1.3.1 The EDGAR v4.2 CO2 Inventory

The EDGAR v4.2 CO2 inventory (publically available at http://edgar.jrc.ec.europa.eu/) provides annual estimates of anthropogenic emissions of GHGs and other air quality pollutants by country as well as on a 0.1o x 0.1o spatial grid from 1990-2010. The EDGAR inventory is a joint project of the European Commission Joint Research Center (JRC) and the Netherlands Environmental Assessment Agency (PBL) (EDGAR, 2010). Emissions for each compound are calculated on an annual basis and sector wise by considering factors such as country-specific activity data for each sector (or population density if no specific activity data is available), the variety of technologies in each sector, and the reduction of emissions by installed abatement measures (EDGAR, 2010). Emission factors are either defined uniformly for all countries or evaluated for individual groups of countries; in special cases, such as road traffic, emissions estimates for individual countries and independently defined activity levels are used to derive country-specific emission factors. Sources in the EDGAR inventory include fossil-fuel related emissions, biofuel combustion, industrial production and consumption processes (including solvent use), land use-related sources (including waste handling) and natural emissions. While EDGAR v4.2 provides estimates of CO2 emissions at the global scale, it is limited by its annual temporal resolution. Although it is known that CO2 emissions exhibit diurnal, weekly and seasonal cycles, emissions in the inventory are assumed constant throughout the year, an assumption that has been demonstrated to be problematic to inverse modelling simulations in previous work (Gurney et al., 2005).

An additional limitation to the EDGAR inventory is the choice of proxies used to spatially disaggregate the national totals onto the 0.1o x 0.1o grid. A study done by McDonald et al. (2014) compared a locally derived inventory of vehicle emissions (the Fuel-based Inventory for Vehicle Emissions, FIVE) against on-road emissions estimated by EDGAR v4.2 for the United States. Although national totals for the two inventories agree to within 5 %, differences among emission estimates increase as the domain of interest becomes smaller (McDonald et al., 2014). At the state

5

level, on-road CO2 emissions in EDGAR are 30 % higher for California and 20 % lower in Texas compared to McDonald et al. (2014). At the urban level, EDGAR overestimates emissions in California cities by 40 – 80 % and in Texas cities by 20 – 30 % (McDonald et al., 2014). Since the national estimates align, the differences between the inventories must be a result of differences in the ways that motor vehicle emissions are disaggregated from national totals to finer spatial scales. EDGAR uses road density as a spatial proxy to disaggregate national data, which may cause an overestimation of emissions in population centers (McDonald et al., 2014). Conversely, the FIVE inventory estimates vehicle emissions at a county level, projects the emissions onto a road atlas and then aggregates to 10 km grid cells (McDonald et al., 2014). Furthermore, EDGAR assigns most of the on-road vehicle emissions to be within cities whereas the FIVE inventory has a more realistic urban to rural emission distribution. Therefore, the use of proxies such as road density or population density that are often used to spatially disaggregate national totals in the EDGAR inventory may introduce uncertainties in the emission estimates at scales finer than the national level.

1.3.2 The FFDAS v2 CO2 Inventory

In response to the shortcomings of the EDGAR v4.2 inventory, the FFDAS v2 inventory was developed via a new methodology for estimating emissions at a fine spatial and temporal scale, assimilating multiple sources of data into a simple model for emission estimates. The FFDAS v2 inventory (publically available at http://hpcg.purdue.edu/FFDAS/) estimates CO2 emissions at the global scale at 0.1o x 0.1o spatial and hourly temporal resolution from 1997-2010. As an assimilation system, FFDAS v2 generates both an estimate of emissions and an accompanying uncertainty estimate for each grid cell. Four primary data sources were used in the FFDAS v2 inventory: (1) nationally aggregated data on CO2 emissions from the International Energy Agency (http://www.iea.org/), (2) gridded population density data from LandScan Global Population Database via East View Geospatial (www.geospatial.com), available from 2000-2010, (3) satellite observations of nightlights from human settlements from the Defense Meteorological Satellite Program Operational Linescan System (DMSP-OLS), and (4) a global power plant emission data product (from a combination of publically disclosed databases and the World Electric Power Plant (WEPP) database) that includes location information and individual power plant emission uncertainties (Asefi-Najafabady et al., 2014; Rayner et al., 2010). In a preliminary analysis, Asefi- Najafabady et al. (2014) carried out three different assimilation experiments to determine the

6 impact of spatial disaggregation on country-level emissions. In the first experiment, data were disaggregated solely by the distribution of population; in the second experiment, the data were disaggregated using the distribution of nightlights; in the third experiment, disaggregation was achieved using both population and nightlights, the approach ultimately used in the FFDAS v2 inventory (Asefi-Najafabady et al., 2014; Rayner et al., 2010). Results of the preliminary analysis showed that the population-based estimated is the most spatially variable, the nightlights-based estimate is the smoothest, and the FFDAS-based estimate is an intermediate. When the three experiments were compared against the Vulcan inventory (a fossil-fuel CO2 inventory for the United States) (Gurney et al., 2009), the FFDAS v2 product was found to correlate to a higher degree than the other two cases, particularly at areas with high population densities such as the USA eastern seaboard and the Los Angeles area, where the population-based experiment overestimates emissions considerably (Rayner et al., 2010). When the FFDAS v2 was compared against state/provincial emission totals for China, USA and Australia, correlations of 0.57, 0.60 and 0.60 respectively were measured, suggesting that FFDAS v2 has skill in distributing emissions at the subnational level (Asefi-Najafabady et al., 2014).

While FFDAS v2 has improved the estimation of temporal variability of CO2 emissions relative to the EDGAR v4.2 inventory, as well as the use of increasingly appropriate spatial proxies, there is still the issue of the relatively coarse spatial scale (0.1o x 0.1o). When trying to understand the emissions of CO2 in an urban area, a finer spatial resolution (on the order of a few kilometers or less) is required to spatially resolve independent emission sources. In the GTA, 0.1o x 0.1o corresponds to 8 km in the E-W direction and 11 km in the N-S direction.

1.3.3 Environment and Climate Change Canada’s CO Inventory

Environment and Climate Change Canada (ECCC) provides an estimate of the sectoral breakdown of CO2 on a province- or nation-wide spatial and annual temporal scale, however no finer level information is readily available. However, CO, as a Criteria Air Contaminant (CAC), is inventoried nation-wide as part of the 2010 Criteria Air Contaminants (CAC) emissions inventory. The Canadian Air Pollutant Emissions Inventory (APEI) is a comprehensive national anthropogenic emissions inventory developed by the Pollutant Inventories and Reporting Division (PIRD) of ECCC. The APEI inventory includes emissions from criteria air contaminants, such as CO, nitrogen oxides, heavy metals, polycyclic aromatic hydrocarbons and persistent organic

7 pollutants (Environment Canada, 2013). The inventory includes emissions from point sources, area sources, on-road mobile sources, and off-road mobile/aircraft/locomotive/marine sources for the base year 2010 compiled using both top-down and bottom-up approaches (Environment Canada, 2013). A summary (but not exhaustive) list of data sources that is used to compile the national CO inventory is included in sub-sections below.

1.3.3.1 Area Sources

Data to compile emissions from area sources were estimated primarily using the Report on Energy Supply and Demand (RESD) in Canada (Statistics Canada, 2012). This report was used to determine emissions from commercial fuel combustion, electric power generation, residential fuel combustion and residential fuel wood combustion, combined with corresponding emission factors (Environment Canada, 2013).

1.3.3.2 Point Sources

Data to compile emissions from point sources were estimated from national reports based on facility level activity information (e.g. the RESD for asphalt production values (Statistics Canada, 2012), Natural Resources Canada for mineral and cement production (Natural Resources Canada, 2017), and Statistics Canada for non-metallic and mineral mining and quarrying (Statistics Canada, 2017).

1.3.3.3 On-road Sources

Data to compile emissions from on-road vehicles was estimated using the Canadian Vehicle Survey (CVS) (Statistics Canada, 2009b). This resource provides information on the number of vehicles registered in Canada, the number of vehicle-kilometers and passenger-kilometers travelled (VKT and PKT, respectively) by province annually, and fuel consumption by vehicle type annually (Statistics Canada, 2009b).

1.3.3.4 Off-road Sources

Data to compile emissions from off-road vehicles was estimated from a variety of sources. To estimate emissions from off-road engines, the CVS was used to determine the number of vehicles, the VKT and PKT and fuel consumption annually (Statistics Canada, 2009b). Emissions from commercial marine vessels was estimated using information collected on the number of vessels

8 travelling on Canadian waters, tonnage (loaded and unloaded) and number of movements (Statistics Canada, 2009c). Emissions from locomotives was estimated using information collected on number of freight and commuter transports and fuel consumed (Railway Association of Canada, 2009). Finally, emissions from aircraft was estimated using information collected on aircraft movement statistics made available by NAV Canada towers and flight service stations (Statistics Canada, 2009a).

1.3.3.5 Sparse Matrix Operator Kernel Emissions (SMOKE)

The annual national level CO inventory described above was transformed using the Sparse Matrix Operator Kernel Emissions (SMOKE, https://www.cmascenter.org/smoke/) emissions processing system for spatial allocation (distribution of non-point source emissions to 2.5 km x 2.5 km resolution (roughly 0.2o x 0.2o resolution)) and temporal allocation (conversion of inventory annual emission rates into hourly values) (Moran et al., 2015). To supplement SMOKE, ~60 southern- Ontario spatial surrogate fields were developed by ECCC to spatially allocate the non-point source emissions to the fine resolution grid; surrogates have the location of emissions sources such as urban dwellings, rural dwellings, location of wood combustion, urban and rural primary and secondary roads, etc. Temporal allocation is primarily provided by SMOKE and facility level data for point sources, however temporal variability of the mobile sector was supplemented using data collected in the FEVER (Fast Evolution of Vehicle Emissions from Roadways) campaign performed in the GTA in 2010 (Gordon et al., 2012a; Gordon et al., 2012b; Zhang et al., 2012).

1.4 Measurements of CO2

In an effort to evaluate and confirm bottom-up inventories, such as those outlined in Sections 1.3.1-

1.3.3, observations of atmospheric pollutants are often made. Estimates of CO2 sectoral contributions can be achieved via ambient CO2 measurements combined with source-specific 13 tracers, such as stable carbon isotopes (δ C, Section 1.5.1) or CO and NOx (Section 1.5.2).

However multiple methods are available to make measurements of CO2, each with specific pros and cons, and therefore thoughtful consideration of the methodology is required. The three dominant methods by which CO2 mixing ratios are currently measured in urban environments is discussed in the following three sub-sections.

9

1.4.1 High precision, low-density measurements

Traditional methods of quantifying atmospheric concentrations of CO2 involves making in situ collections of air using flasks and quantifying their CO2 concentration in the laboratory. While this technique allowed the use of precise and accurate instrumentation, the primary drawback associated is low time resolution. More recently, instruments have been developed that continue to make in situ measurements of CO2 at high accuracy and precision but at a time resolution of a second. These instruments are most commonly based on cavity ring down spectroscopy (CRDS) which takes advantage of a long effective path length (tens of kilometers) to increase instrument sensitivity. Since CRDS is the principal measurement technique used for several studies in this thesis (see Chapters 2, 3 & 4), it is described here in detail. As shown in Figure 1.1, CRDS is a technique where a sample gas is continuously introduced into an optical cavity and the optical absorbance of the sample is determined, allowing for an estimate of the concentration based on a known absorption cross section (Crosson, 2008). Light from a laser is injected into the cavity through a partially reflective mirror; the light intensity in the cavity builds up over time and is measured via a second partially reflective mirror using a photo-detector located outside the cavity (Crosson, 2008). The “ring-down” time is measured by rapidly turning off the laser and measuring the light intensity in the cavity as it decays exponentially with time; decay is a result of losses due to cavity mirror and the absorption and/or scattering of the sample being measured (Crosson, 2008). The intensity (I) of light transmitted through the exit mirror will decrease as a function of time (t), according to Equation 1.1.

−푡 ⁄휏 퐼(푡) = 퐼(0)푒 (E1.1)

I(0) is the initial detected intensity, I(t) is the intensity at time t, and τ is the ring-down time (the time it takes for the intensity to reach 1/e of I(0)). The empty ring-down time is dependent on mirror separation (L), mirror reflectivity (R) and the speed of light (c), Equation 1.2. With an absorber present, the ring-down time will decrease as a result of the additional decay process of absorption, Equations 1.3a + 1.3b.

퐿 휏 = (E1.2) 푐(1−푅)

퐿 휏 = (E1.3a) 푐(1−푅+ 훼퐿)

10

훼 = 휀(휆) ∗ 퐶 (E1.3b)

Where α is the molecular absorption coefficient and is dependent on ε(λ), the absorption cross section of the analyte at wavelength λ, and C, the concentration of the analyte. Wavelengths of -1 -1 12 13 6251.8 cm and 6251.3 cm are used to characterize CO2 and CO2 respectively in instruments used by ECCC in southern Ontario.

Figure 1.1 Schematic of cavity ring down spectroscopy (CRDS). Diagram is not to scale.

There are many advantages of the use of CRDS to quantify CO2. Firstly the ring-down time is independent of laser intensity and therefore fluctuations in the light source are insignificant. A second advantage is its high sensitivity, achieved due to its long path length. Third, multiple analytes can be quantified simultaneously, allowing the use of a single analyzer in studies. Finally, with an appropriate calibration protocol, excellent precision and accuracy can be achieved; at stations operated by ECCC across Canada, Figure 1.2, precision and accuracy can be achieved as low as 0.1 ppm (based on one-minute averaging) and 0.2 ppm respectively. At ECCC, CO2 measurements follow a set of established principles and protocols outlined by a number of international agencies through recommendations of the Meeting on Carbon Dioxide, Other Greenhouse Gases, and Related Measurement Techniques, coordinated by the World Meteorological Organization (WMO) every 2 years. Amongst other procedures, this protocol

11 includes determining the response function of the analyzer against three calibrated standards tanks (Low, Mid, High) with concentrations in the same range as ambient values.

Figure 1.2: The location of the 22 in situ CO2 and CH4 observational sites in Canada, as of 2016 (Worthy, 2017 (unpublished work))

Many studies have used high precision instruments to quantify CO2 mixing ratios in urban areas in order to derive city-scale fluxes and quantify emission trends (Breon et al., 2015; Lavaux et al., 2013; McKain et al., 2012; Newman et al., 2013; Turnbull et al., 2015). Lavaux et al. (2013) used a network of two CRDS instruments in Davos, Switzerland in an attempt to confirm local emissions inventories as well as quantify the impact of the World Economic Forum 2012 Meeting (WEF-2012) on local atmospheric composition. Using observational data in combination with the

WRF-Chem atmospheric model, the authors were able to quantify CO2 emissions as 40 % higher than the initial annual inventory estimate, likely due to the use of heating fuel in the wintertime

(Lavaux et al., 2013). Furthermore, during the WEF-2012, emissions of CO2 were observed to be 35 % lower than observations made at an earlier period, despite similar temperatures and precipitation patterns (Lavaux et al., 2013). Simultaneous CO mixing ratio measurements showed an absence of traffic peaks during the meeting, indicating that a change in the contribution of on-

12

road emissions is likely responsible for the decrease in measured CO2 mixing ratio (Lavaux et al., 2013).

A study performed by McKain et al. (2012) used a network of five instruments measuring CO2 mixing ratios in Salt Lake City, Utah and the WRF-STILT atmospheric transport model to develop an observation-model framework. Their results show that their framework was capable of detecting a change in ambient CO2 emissions of 15 % or more from an urban region (McKain et al., 2012), setting a benchmark for the precision required to detect changes in atmospheric CO2 following implemented mitigation strategies.

Despite the high precision and accuracy that these instruments offer, as well as lab and/or field applicability, the primary disadvantage is the associated cost (~USD 60,000) which limits the size of the networks in urban areas. A study by Kort et al. (2013) used an inversion method to determine if a network of 3, 4 or 8 instruments is required in the Los Angeles basin to provide sufficient observational constraints on emission behavior. Their study found that a 3- and 4-site network provided insufficient spatial coverage and lacked sensitivity to the entire basin (Kort et al., 2013). The 8-site network was the minimum design required to achieve sensitivity to emissions throughout the basin and reduce predicted uncertainty (Kort et al., 2013). Although this work was specific for the Los Angeles basin, it addresses the key issue of network size and density that most urban studies often overlook.

1.4.2 Low precision, high-density measurements

While the high-precision instruments discussed in Section 1.4.1 are able to monitor CO2 concentrations with high accuracy and precision, they are often purposefully located away from urban signals to better represent a regional average. This approach cannot provide spatially resolved information in areas that many mitigation strategies are targeting (Shusterman et al., 2016). Traditional urban monitoring networks have typically relied on a small number of monitoring stations (2-15) (Breon et al., 2015; Lavaux et al., 2013; McKain et al., 2012; Newman et al., 2013; Turnbull et al., 2015) however, in order to resolve CO2 being emitted from individual sources or sectors and confirm inventories, a much higher spatial resolution is required (Kort et al., 2013). More recently, alternative approaches are being used that leverage low-cost CO2 sensors, enabling a spatially dense network of observation sites allowing for the detection of pollution hotspots, inventory development and real-time exposure assessment (Shusterman et al.,

13

2016). Low-cost high-density measurements are being made at various locations globally, such as at London Heathrow Airport as a part of Sensor Networks for Air Quality (SNAQ, http://www.snaq.org/), however the most prominent example currently is the BErkeley

Atmospheric CO2 Observation Network (BEACO2N) in the city of Oakland, California

(Shusterman et al., 2016; Turner et al., 2016). BEACO2N is a large-scale network of 28 CO2- sensing “nodes” at a spatial resolution of approximately 2 km x 2 km. Each BEACO2N node contains a non-dispersive infrared Vaisala CarboCap GMP343 sensor for CO2 measurement as well as a variety of additional sensors for measurement of other species (CO, O3, NO and NO2) and meteorological parameters (humidity and temperature) (Shusterman et al., 2016). Accounting for these sensors and ancillary hardware, each sensor has a total cost of ~USD 5500, which is approximately an order of magnitude cheaper than traditional CRDS instruments (Shusterman et al., 2016).

The use of these low-cost sensors has high potential for robust environmental surveillance.

Preliminary analyses by Shusterman et al. (2016) demonstrated that after 3 years of CO2 monitoring via the BEACO2N network, the low-cost sensors are able to accurately capture expected diurnal and seasonal cycles, compared against local measurements made using CRDS instruments (Shusterman et al., 2016). Furthermore, significant quantitative differences among the diurnal cycles at individual nodes are captured at spatial scales that many atmospheric transport models cannot represent, confirming the need for high-density monitoring to accurately represent the complex variability in an urban environment. The use of high-density networks allows the achievement of spatial resolution that matches that of local inventories, improving the ability to identify source contributions and inform policy on the development of effective mitigation strategies.

While the potential for high spatial density and the low cost of these sensors is advantageous, there is a trade-off with measurement precision and accuracy that must be considered. Typical Picarro

CRDS systems offer a precision in CO2 measurements better than 0.1 ppm; laboratory characterization of the CarboCap low-cost sensors by Shusterman et al. (2016) evaluate the average precision to be between 1.2 and 2.0 ppm. At this precision, evaluating an urban environment where anthropogenic CO2 emissions enhance atmospheric concentrations by as little as 0-2 ppm above the background in the afternoon may be complicated. Additionally, the use of high-density networks limits the access to validation and calibration infrastructure (re-calibration

14

is achieved every 12-18 months in the BEACO2N network (Shusterman et al., 2016)). Therefore, another possible complication with the use of these low-cost sensors is systematic uncertainty resulting from a combination of gradual temporal drift and constant biases from the true value; post hoc corrections need to be implemented before considering the data (Shusterman et al., 2016).

1.4.3 Column measurements

The ground-based in situ monitoring efforts outlined in Sections 1.4.1 and 1.4.2 are complemented by column measurements made either by space-based observations from a number of satellites, including SCHIAMACHY (Burrows et al., 1995), GOSAT (Yokota et al., 2009) and most recently OCO-2 (Crisp et al., 2008; Hakkarainen et al., 2016), or from ground-based instruments, such as in the TCCON network (Wunch et al., 2011). Unlike the column-integrated observation of short- lived atmospheric pollutants, such as NO2, which has allowed estimation of point sources, long- term trends and emission inventory verification, the interpretation of column CO2 observations is complicated by its long lifetime, trends, seasonality, and large atmospheric background (Hakkarainen et al., 2016). However, work by Kort et al. (2012) used GOSAT column-averaged dry air mole fraction of CO2 (XCO2) to differentiate emissions in two megacities, the Los Angeles basin and Mumbai, with nearby ‘clean’ observations representative of background air, isolating the CO2 enhancement from anthropogenic emissions. Using 10-day block averages (to isolate for times when measurements were made in both the megacity and background) a persistent enhancement of 3.2 ± 1.5 ppm and 2.4 ± 1.2 ppm were quantified for Los Angeles and Mumbai, respectively (Kort et al., 2012). In the case of Los Angeles, this enhancement is consistent with ground-based XCO2 measurements attributed to anthropogenic emissions (Kort et al., 2012; Wunch et al., 2009). This research demonstrated that satellite observations of this precision can identify enhanced XCO2 due to megacity emissions.

More recently, Hakkarainen et al. (2016) used OCO-2 XCO2 measurements to identify significant polluted regions over eastern USA, central Europe and East Asia. To isolate the pollution areas, the authors calculated XCO2 anomalies by subtracting the daily median from individual observations allowing them to simultaneously deseasonalize and detrend the data (Hakkarainen et o o al., 2016). Using a resolution of 1 x 1 and a clustering method in combination with OMI NO2 to identify four groups of increasing NO2 tropospheric column values, they were able to observe elevated XCO2 anomalies over eastern USA, central Europe, the Middle East, China, India and

15

Japan, with highest values over eastern China (as expected from emission inventories)

(Hakkarainen et al., 2016). XCO2 anomalies over Iraq were found to be comparable to those in central Europe, despite emission inventories showing relatively lower emissions in the Middle

East, suggesting possible incomplete inventories of CO2 (Hakkarainen et al., 2016). Like the study performed by Hakkarainen et al. (2016), many studies have shown the utility in quantifying CO2 in urban areas from column measurements (Bovensmann et al., 2010; Deng et al., 2016; Janardanan et al., 2016; Kort et al., 2012; Schneising et al., 2013).

Quantifying CO2 using total column measurements over in situ surface sites has many advantages.

Column measurements of CO2 are particularly useful to separate the effects of atmospheric mixing from the surface exchange because they are insensitive to variations in surface pressure and atmospheric water vapour (Wunch et al., 2011). Furthermore, because the column vertically integrates the concentration of CO2 above the surface, it is much less affected by vertical transport (Wunch et al., 2011). Additionally, column measurements have strong measurement precision and accuracy, ~0.25 % (Wunch et al., 2011). When considering column measurements specifically from spaceborne instruments, this technique offers global coverage and long timescales of observation. However, a common disadvantage for both space- and ground-based instruments is poor spatial resolution/coverage. For ground-based measurements, column observations of CO2 are limited to a few stations per continent, limiting the representativeness of the measurements to regional areas. For space-based measurements, Hakkarainen et al. (2016) showed that when OCO- 2 data was spatially disaggregated to 0.25o x 0.25o, most grid cells were not populated with data, rendering the results less robust. Urban emission inventories can have spatial resolution of 0.01o x 0.01o (e.g. the Airparif Inventory for the Ile de France region, publically available at http://www.airparif.asso.fr/en/index/index), and therefore a higher degree of spatial resolution in the observations is desired. In addition to the limitation of high spatial resolution, spaceborne instruments are also limited in their temporal resolution; sun-synchronous orbits (as the case of

SCHIAMACHY, GOSAT and OCO-2) only quantify XCO2 at a particular location on Earth at a fixed time of day, making them insensitive to diurnal variations of urban emissions that are necessary for some aspects of inventory validation.

16

1.5 Use of tracers to source apportion CO2

The methods available to measure CO2 outlined in Sections 1.4.1-1.4.3 demonstrate that accurate and precise quantification of mixing ratios can be achieved at varying spatial and temporal resolutions, however CO2 mixing ratios alone give minimal insight into the relative contributions of emission sources. With international treaties (e.g. the Kyoto Protocol and the Paris agreement) and national agreements targeting emission reductions of CO2 and other GHGs, an understanding of the major anthropogenic source contributions is imperative to effectively implement mitigation strategies. Traditionally, GHG inventory compilation solely relies on bottom-up emission statistics provided by individual facilities; at the national level, these estimates are reported to have an uncertainty of approximately 10 – 40 % (Peylin et al., 2011; Vogel et al., 2012). To be able to independently verify the emissions inventories, observation-based source-specific tracers, such as 12 13 14 carbon isotopes ( CO2, CO2 and CO2), CO and NOx are required.

12 13 1.5.1 Stable carbon isotopes, CO2 and CO2

12 13 Stable carbon isotopes, CO2 and CO2, are emitted in unique ratios (expressed relative to a known standard, Equation 1.4) and therefore can be used to identify the source as long as unique end-member ratios (hereafter referred to as “signatures”) are known.

13 퐶푠푎푚푝푙푒 12 13 퐶푠푎푚푝푙푒 훿 퐶 = ( 13 ) − 1 푥 1000 ‰ (E1.4) 퐶푠푡푎푛푑푎푟푑 12 퐶푠푡푎푛푑푎푟푑 ( )

12 13 Most commonly, studies apply the Keeling plot approach to observations of CO2 and CO2 to directly compare ambient signatures with known end-member signatures. Further, chemical mass balance (CMB) approximations are used to partition total ambient CO2 into source components. The theory of the Keeling plot approach as well as the CMB approximation are discussed in detail in Section 1.5.1.1. Applications demonstrating the utility of these approaches with stable carbon isotopes in CO2 source apportionment studies are discussed in Section 1.5.1.2.

17

1.5.1.1 Keeling plot and chemical mass balance approaches

The Keeling plot method, first proposed by Keeling (Keeling, 1958; Keeling, 1961), uses the conservation of mass to estimate source contributions of a gaseous compound (Pataki, Ehleringer et al., 2003), for example, CO2. The Keeling plot approach is based on the assumption that the atmospheric concentration of CO2 (Ct) is a combination of a background contribution (Cb) and variable amounts of CO2 contributed by sources (Cs) in the domain, Equation 1.5.

Ct = Cb + Cs (E1.5)

Given the conservation of mass, we apply the δ13C signature of each component, Equation 1.6, and combine with Equation 1.5 to produce Equation 1.7.

13 13 13 δ CtCt = δ CbCb + δ CsCs (E1.6)

13 13 13 13 δ Ct= Cb(δ Cb - δ Cs)(1/Ct) + δ Cs (E1.7)

13 13 13 δ Cs is the signature of the CO2 source(s) in the domain. In a plot of δ Ct versus 1/Ct, δ Cs can be calculated via the y-intercept, e.g. Figure 1.3, and therefore allows you to identify the signature of the local source of the CO2 as -40.1 ‰.

13 Figure 1.3: Example Keeling plot of δ CO2 vs. 1/CO2 for an overnight period in January in the GTA

18

This system makes two important assumptions, (1) there is a simple mixing of only two CO2 13 components (the background and a source), and (2) we assume that the δ C of these components does not vary over the course of the observation (Pataki et al., 2003). It is often the case that these assumptions are violated, as CO2 can be derived from a variety of anthropogenic and biogenic sources; however this is not a concern as long as the δ13C of each component is the same or if their contributions to the total flux does not change over the time period of sampling (Pataki et al., 13 2003). In this situation, the δ Cs would be the flux-weighted mean value of combined sources.

The Keeling plot approach has been applied widely to estimate sources of CO2, particularly in urban environments (Clark-Thorne and Yapp, 2003; Gorka et al., 2011; Jasek et al., 2014; Pataki et al., 2003; Pataki, Bowling et al., 2003; Widory and Javoy, 2003). An advantage of the Keeling 13 13 plot approach is that Cb, δ Cb, and Cs need not be specified in order to solve for δ Cs. However, a key disadvantage is that it requires extrapolation of the data far beyond the measured range to identify the intercept, and therefore small uncertainties in the regression slope may lead to large uncertainties in the intercept (Pataki et al., 2003).

In combination with the Keeling plot approach, the CMB approach aims to partition total CO2 into distinct components, according to Equations 1.8 and 1.9.

Ct = Cb + C1 + C2 (E1.8)

13 13 13 13 δ Ct·Ct = δ Cb·Cb + δ C1·C1 + δ C2·C2 (E1.9)

Equation 1.8 is a modified form of Equation 1.5, where Cs is now partitioned into multiple source components, C1 and C2. Additional source components can be estimated (e.g. C3, C4…Cn) using CMB if more isotope data are available (such as 14C or 18O), increasing the number of equations that need to be solved in the mathematical expression. Like the Keeling plot method, an underlying assumption of the CMB approach is that Cb, C1 and C2, and their respective signatures, remain constant during the sampling period.

1.5.1.2 Application of the Keeling plot and chemical mass balance approaches in CO2 source apportionment studies

Djuricin et al. (2010) used carbon isotopes in combination with stable oxygen isotopes (δ18O) to source apportion CO2 in the Los Angeles basin into four components: natural gas combustion,

19 gasoline combustion, aboveground respiration and belowground respiration during four measurement campaigns in 2007-2008. The CMB approach was applied, and differences in the relative contributions of sources between seasons was apparent. In the fall season, aboveground and belowground respiration dominated the local sources, contributing 43.6 ± 3.8 % and 26.8 ± 3.3 % respectively, while natural gas and gasoline combustion only accounted for 9.2 ± 1.9 % and 20.4 ± 3.1 % respectively (Djuricin et al., 2010). In December, a more uniform contribution was observed: 8.9 % (natural gas), 29.1 ± 13.0 % (gasoline), 24.2 ± 10.9 % (belowground respiration) and 27.9 ± 12.0 % (aboveground respiration) (Djuricin et al., 2010). In February and April, contributions of natural gas combustion, gasoline combustion and belowground respiration were fairly consistent and minimal, with aboveground respiration accounting for >50 % of measured

CO2 (Djuricin et al., 2010). According to the California Air Resources Board Greenhouse Gas Inventory report for 1990-2004, gasoline combustion should contribute approximately 3 times more CO2 to the atmosphere than natural gas combustion (146.1 million tons and 45.7 million tons, respectively) (Djuricin et al., 2010). However, the results from the study by Djuricin et al.

(2010) suggest that gasoline and natural gas combustion contribute similarly to overall local CO2. Furthermore, inventories typically provide information on an annual basis however results of this study suggest clear seasonal differences in source contributions, again highlighting the necessity for observation-based confirmation of inventories. This study demonstrates the utility of using isotopes to source apportion CO2 however, the results are reliant on the uncertainty in estimating isotope end-member signatures and the possibility of additional sources not accounted for in the CMB. The October results were also complicated by the presence of wildfires occurring ~15 km from the measurement station, a source not included in the CMB approach and having a δ13C similar to that of gasoline combustion or respiration (Djuricin et al., 2010).

Similar results were found in Salt Lake City, Utah where carbon and oxygen isotopes were used to source apportion CO2 in a three-component CMB approximation, natural gas combustion, gasoline combustion and biogenic respiration (Pataki et al., 2003). Approximately 60 % of CO2 in the winter nighttime was derived from natural gas combustion, a result consistent with the use of natural gas as the primary heating source, while the contribution from gasoline combustion remained consistently high for majority of 2002 (Pataki et al., 2003). The contribution from biogenic respiration maximized in the spring and late summer/fall with negligible contributions during the winter (Pataki et al., 2003). Like the results from the Djuricin et al. (2010) paper, this

20 works highlights the utility of stable carbon isotopes to quantify relative contributions of sources to overall CO2 concentrations and therefore potential inventory verification.

1.5.1.3 Spatial variability of carbon isotopic signatures

An important consideration in the use of carbon isotopes to source apportion CO2 is the geographic variability of the end-member signatures. Figure 1.4 displays the results from six independent studies measuring δ13C for four traditional end-members (natural gas, gasoline, diesel and coal combustion) at six locations globally (Bush et al., 2007; Clark-Thorne and Yapp, 2003; Masalaite et al., 2012; Semmens et al., 2015; Widory and Javoy, 2003; Zimnoch, 2009). With the precision of CRDS instruments typically ~0.15 ‰, even the seemingly small differences between signatures (particularly gasoline and diesel) renders CMB approaches less robust and makes understanding the resulting contributions complicated.

Figure 1.4 Overview of end-member δ13C variability at various location globally

12 13 A sensitivity analysis using CO2 and CO2 data collected in Gif-sur-Yvette, France was 13 performed to determine the influence changing the source signature of a component (e.g. δ C1 or 13 δ C2 in Equation 1.9) has in a CMB approximation (Pugliese, unpublished work). A two 13 component CMB approach was applied (Equation 1.9) and three common signatures for δ C1

21

13 were applied: -39 ‰, -40 ‰ and -46 ‰. Holding all other parameters constant, changing the δ C1 13 from -39 ‰ to -40 ‰ changed the relative contribution of C1 versus C2 by 2 % and changing δ C1 from -40 ‰ to -46 ‰ changed the relative contribution by 14 %. Although relatively simplistic, this sensitivity analysis highlights the need for accurate and local characterization of source signatures to ensure accurate source apportionment of CO2.

1.5.2 Combustion products, CO and NOx

13 When continuous measurements of CO2 are not available, CO and NOx are often used as alternative tracers that can identify anthropogenic sources of CO2. CO and NOx, like CO2, are products of combustion where the primary sources of CO and NOx in Canada are vehicle emissions (accounting for 53 % and 54 %, respectively) followed by commercial and residential combustion for CO (22 %) and the oil and gas industry for NOx (25 %) (Environment Canada, 2017). Different combustion processes have characteristic CO:CO2 or NOx:CO2 emission ratios, which are dependent on the combustion efficiency and the fuel type. For example, low efficiency industrial sources can have CO:CO2 around 17.5 ppb/ppm while traffic can have CO:CO2 around 5.3 ppb/ppm (Lopez et al., 2013). Lopez et al. (2013) quantified the fossil fuel CO2 component of CO2 14 14 (FFCO2) in Paris, France using CO2, CO and NOx as proxies. From their measurements of CO2, they estimated that of the total measured CO2 in a station in Paris, 77 % was of fossil fuel origin and 23 % was of biogenic origin (Lopez et al., 2013). Of that 77 %, they were able to use CO and

NOx to identify the diurnal profile of FFCO2 where peaks are observed in the early morning and late afternoon rush hours, with mean mole fractions of FFCO2 estimated at 20.6 and 18.7 ppm using CO and NOx respectively (Lopez et al., 2013). The differences between estimates were attributed to the short lifetime of NOx relative to the lifetime of CO. This study highlights the utility of the use of CO and NOx as proxies to estimate the temporal variability of FFCO2 in an urban region.

In an urban region, CO and NOx themselves are important players in affecting air quality. Both

CO and NOx are precursors to ground-level ozone production, which is one of the major components of photochemical smog. The phenomenon and the chemistry that governs its formation is discussed in the following sections.

22

1.6 Urban NOx – VOC – O3 chemistry and trends

1.6.1 Urban NOx – VOC – O3 chemistry

The chemistry of ozone production in the urban environment will be discussed in detail in Chapter 5. A brief introduction is provided here.

In the sunlit atmosphere, NO and NO2 rapidly interconvert on the order of minutes, producing and subsequently consuming O3 in a null cycle (i.e. there is no net production or consumption of NOx or O3). However net production of O3 can occur through the oxidation of CO and volatile organic compounds (VOCs) by the hydroxyl radical, OH. This reaction produces a peroxy radical which reacts with NO to form NO2, subsequently photolyzing to produce O3. A single cycle of oxidation of a VOC typically results in the production of two O3 molecules, and the regeneration of NOx and

OH radicals which act as catalysts to further O3 production (Reaction 1.1).

VOC + 4O2  2O3 + oxidized VOC + H2O (R1.1)

This catalytic O3 production chain is terminated by the loss of HOx radicals (HOx = OH + HO2 +

RO2 + RO), which can occur through multiple pathways: (1) the formation of nitric acid (through the reaction of NO2 and OH), (2) reactions that form organic nitrates, and (3) peroxy radical self- reaction to produce peroxides or other oxygenated compounds.

In an environment with a low NOx:VOC ratio, O3 production is essentially independent of VOC concentration and increases linearly with increasing NOx. This is a result of VOCs outcompeting

NOx for reactions with OH, allowing the peroxy self-reaction to dominate radical termination. In an environment with a high NOx:VOC ratio, O3 production becomes inversely proportional to changes in NOx and more sensitive to VOC reactivity. This is a result of NOx outcompeting VOCs for reactions with OH, suppressing the formation of new peroxy radicals.

1.6.2 Trends in CO and NOx emissions in Ontario

As introduced in Sections 1.5.2 and 1.6.1, CO and NOx play key roles in the formation of ground- level O3. While emissions of CO and NOx are from a variety of anthropogenic and, to a lesser degree, biogenic sources, emissions from the transportation sector are the largest contributor (>50 %) (Environment Canada, 2017). Governmental initiatives, such as the “Drive Clean” vehicle test program in Ontario (MOE., 2013), have resulted in large reductions in the emission of both of

23 these compounds over the past decade, both nationally and regionally. At the national scale, decreases from 2005 to 2015 have been observed for both CO (29 %, 7.9 to 5.6 Mt, respectively) as well as NOx (24 %, 2.5 to 1.9 Mt, respectively) (Environment Canada, 2017). On the provincial scale in Ontario specifically, similar decreases have been measured, with a 44 % decrease and a

41 % decrease in total CO and NOx emissions from 2000-2015, respectively, Figure 1.5

(Environment Canada, 2017). Similar to the decrease of CO of NOx emissions observed in the GTA and Canada, decreases have also been observed in other areas of Europe and North America (Cooper et al., 2014). As CO is often considered a surrogate for anthropogenic VOC trends, the observed decreases in both CO and NOx indicates the O3 production regime may be shifting in many regions globally.

Figure 1.5: Ontario annual average emission estimates of CO and NOx (tonnes) from 2000-2015

1.6.3 Role of meteorology and transport

Many studies have reported evidence of a correlation existing between ground-level O3 concentrations and meteorological conditions (Cox and Chu, 1993; Jacob et al., 1993; Sillman et al., 1990; Tarasova and Karpetchko, 2003; Wolff and Lioy, 1978). In particular, a strong positive correlation has been observed between O3 and temperature in regions of North America, Europe and Asia (Agudelo-Castaneda et al., 2013; Lacour et al., 2006; Lee et al., 2014; Pekey and Ozaslan,

2013; Psiloglou et al., 2013). The correlation between O3 and temperature may be a result of the

24 influence temperature has on other meteorological conditions; for example, high temperatures are often correlated with high-pressure systems, minimizing vertical mixing and enhancing photolysis rates due to high solar radiation. This effect was observed in Houston during the TexAQS II

Radical and Aerosol Measurement Project (TRAMP) campaign where the highest O3 levels (>100 ppb) correlated with high temperatures but also low wind speed and low sky cover (Lefer et al., 2010).

In addition to the mechanistic reasons why higher temperatures may correlate with enhanced O3 production rates, there are also chemical explanations. Higher temperatures may lead to the suppression of radicals being stored as peroxyacetylnitrate (PAN), an important reservoir for NOx

(Jacob et al., 1993). Additionally, warmer temperatures may also increase HOx production or local biogenic or fugitive anthropogenic VOC emissions; in environments with a high NOx:VOC ratio, this may significantly contribute to local O3 production.

Other meteorological parameters have been found to influence the production rate of O3. Relative humidity is commonly reported to be negatively correlated with O3 production (Duenas et al., 2002; Elminir, 2005; Shan et al., 2008; Shan et al., 2009; Tu et al., 2007), however the factors that drive this correlation are not well understood. Some explanations include: the photolysis of O3 and 1 subsequent loss of O( D) to H2O; the association of humid days with enhanced cloud cover and therefore reduced photochemistry; and, the association of humid days with precipitation events and therefore decreased atmospheric precursor concentrations. More recently, the inclusion of a vapor pressure deficit (VPD)-dependent deposition scheme of O3 in model simulations helped to explain the net O3 production decreases under increasing relative humidity (Kavassalis and Murphy, 2017).

In addition to meteorological influences, regional transport can also play a significant role on O3 production due to the ability of O3 and NOx to travel long distances (a few hundred kilometers) (Ge et al., 2012; Hidy and Blanchard, 2015; Li et al., 2015; Xu et al., 2011). Regional transport may play a role in urban O3 production in two ways: (1) transport precursor compounds from upwind sites to the urban area and promote local O3 production, or (2) transport local O3 production downwind into rural areas. A case of the former is commonly seen in Atlanta, Georgia and Birmingham, Alabama, two environments surrounded by areas with significant natural VOC emissions (Hidy and Blanchard, 2015). On days when air is transported from these regions,

25 biogenic VOCs can contribute significantly to VOC reactivity and therefore may play a role in urban O3 production (Hidy and Blanchard, 2015). A case of the latter is commonly seen in Beijing,

China where a study by Xu et al. (2011) measured O3 concentrations at four sites along a north- east transect away from Beijing following the prevailing wind patterns. O3 diurnal patterns at the sites indicated the O3 diurnal peak appearing sequentially at 13:00, 14:00, 15:00 and 17:00 (Beijing time) (Xu et al., 2011). A series of other wind vector analyses helped identify that the concentrations of O3 at the site furthest along the transect were profoundly affected by transport of polluted air masses from Beijing (Xu et al., 2011).

Identifying and understanding the influence of meteorology and transport on observed O3 production in urban areas is critical to assessing the impact of changing local emissions of NOx and VOCs. It is possible that annual variability in weather may be a stronger driver in O3 production than changes in precursor emission rates. This was particularly seen in a study in the GTA where mean summer temperatures and transport from the USA were key contributors to early morning O3 levels, despite continued decreases in measured ambient precursor concentrations (Geddes et al., 2009). An subsequent study found that lake-breeze circulations in the GTA contributed to elevated ozone concentrations (~13.6-14.8 ppb higher average daytime maxima) relative to days that did not experience lake breezes (Wentworth et al., 2015). Therefore, it is important to decouple the production of O3 driven by local chemistry and the influence of regional scale transport of polluted air.

1.6.4 Observed trends in ground-level O3 concentrations through time

There is evidence that background concentrations of O3 have been increasing in North America and elsewhere in the northern hemisphere (NH) (Cooper et al., 2014). Cooper et al. (2014) compiled data on surface ozone trends at stations globally during three decadal periods (1950- 1979 through 2000-2010, 1980-1989 through 2000-2010, and 1990-1999 through 2000-2010). In the first decadal period, 13 sites were reporting O3 concentrations in the NH and all sites indicated -1 increases in O3, with 11 having statistically significant trends of 1-5 ppb decade (Cooper et al., 2014). In the second decadal period, change was evident, with 15 of the available 20 sites indicating increases in O3 (15 significant) and 5 showing decreases (2 significant) (Cooper et al.,

2014). In the last decadal period, O3 trends are regionally specific: O3 generally increased in Europe through much of the 1990s but has recently leveled off or decreased at rural and

26

mountaintop sites (Derwent et al., 2013; Gilge et al., 2010); O3 has increased in North America in eastern and arctic Canada, but has remain unchanged in central and western Canada (Oltmans et al., 2013); in eastern USA, O3 has decreased strongly in the summer but is unchanged in the spring and increased in the winter (Cooper et al., 2012); O3 in East Asian countries is generally increasing

(Oltmans et al., 2013); in the southern hemisphere, O3 has increased at eight of the nine sites (although only 4 are significant decreases) (Cooper et al., 2014). The recent decreases in observed

O3 regionally are likely a direct result in the mitigation policies aimed at reducing the emission of

O3 precursor compounds. For example, in El Paso, Texas, a policy restricting gasoline use to only that with a low Reid Vapor Pressure (RVP) have helped to decrease the O3 8-hour ozone design value from 89 ppb in 1994-1996 to 72 ppb in 2013-2015 (Sather and Cavender, 2016).

Despite regional success stories, the trends reported by Cooper et al. (2014) indicate that controlling ground-level O3 formation continues to be a challenge today in many locations globally, even where stringent restrictions on primary emissions have reduced the atmospheric concentrations of precursor compounds. These challenges stem from the complexity and non- linearity of the chemistry governing the formation of O3. The complexity is exacerbated by influences of meteorology and transport (Section 1.6.2), affecting both the rate of formation and local accumulation of O3, as well as the increases in background O3 discussed previously. In Canada, a Canadian Ambient Air Quality Standard (CAAQS) for ozone is implemented, which states that the 3-year average of the 4th-highest daily maximum 8 h average should not exceed 63 ppb (updated from 65 ppb as of 2015). Despite major provincial and municipal efforts to control emissions of precursor compounds, summers in the GTA continue to experience many days with

O3 levels above the standard (Geddes et al., 2009). Therefore, research questions that need to be addressed include determining what is driving local O3 production in the GTA and other areas in Canada where exceedances are being measured, what are the anticipated outcomes for policy efforts to reduce emissions, and what are the potential meteorological and transport impacts? These questions can be investigated by taking advantage of long-term monitoring efforts, short-term intensive research campaigns, satellite data (such as that from NASA’s Ozone Monitoring Instrument (OMI)), or chemical transport modelling.

27

1.7 Synthesis and Research Objectives

Although representing a small fraction of the Earth’s surface, urban areas contribute significantly to increased global concentrations of primary and secondary atmospheric pollutants. With a focus on CO2 as a primary pollutant, this introduction has reviewed the global and regional sources as well as common techniques by which its mixing ratio is measured, highlighting in particular the advantages and disadvantages of each method. In an effort to develop robust methods to source apportion CO2, this introduction highlighted the utility of stable carbon isotopes to identify sources with unique signatures. With a focus on O3 as a secondary air quality pollutant, this introduction has reviewed the chemistry and meteorology that governs its production and concentrations as well as observed global trends over the past decades. The work presented in this thesis represents a detailed evaluation of the sources and historical trends of primary and secondary pollutants in Canada’s largest urban area, the GTA.

Chapter 2 of this thesis aims to understand the sources of CO2 in central- and southern-Ontario. A sector-specific inventory (Southern Ontario Carbon Dioxide Emissions or SOCE) was developed at fine spatial and temporal resolution. A chemistry transport model (GEM-MACH) was driven by both SOCE and FFDAS v2 inventories to simulate CO2 concentrations in eastern North

America. The output from the model was compared to CO2 measurements made at four sites in Southern Ontario. Research questions that will be addressed include:

 What sources of data can be used to build a fine spatial and temporal CO2 inventory that

can accurately quantify contributions of CO2 emissions from a variety of sectors (e.g. area, point, marine, on-road, off-road, etc.)?

 What is the spatial and temporal structure of CO2 emissions in southern Ontario?

 Are there quantifiable differences between our developed CO2 inventory and the publically available FFDAS v2 inventory for modelling southern Ontario?

 Can observation-based estimates support policy making, by monitoring emission trends and the impact of mitigation measures?

The use of stable carbon isotopes is explored in Chapter 3 to further understand the relative importance of various sources to CO2 emissions in the GTA. A variety of anthropogenic CO2

28 sources were sampled for δ13C characterization and used in a CMB approximation to estimate source contributions. The following research questions will be addressed:

13  Are the δ C signatures for anthropogenic sources of CO2 in the GTA significantly different from other areas globally?

 Is there temporal and spatial variability associated with the δ13C signatures?  What is the minimum difference in δ13C signatures that is required to separately identify 13 CO2 sources using δ C as the sole tracer?

 Can the measured δ13C signatures be used in combination with the Keeling Plot approach

and the CMB approach to infer source contributions of CO2?

Subsequently, we combine the outputs from Chapter 2 and Chapter 3 to explore the ability to 13 13 model δ CO2. Chapter 4 uses the GTA-specific δ CO2 signatures to convert model-simulated 13 CO2 concentrations into simulated δ CO2 signatures for comparison with observational data. This developed model framework is used to identify source contributions of CO2 on an hourly timescale. Research questions that will be addressed in Chapter 4 include:

13  Can model outputs of CO2 in combination with local δ CO2 signatures be used to model 13 δ CO2 and infer the source contributions to the local CO2 enhancement on an hourly timescale? 13  Does the δ CO2 signature of the local CO2 enhancement estimated by the Keeling analysis agree with that estimated by the SOCE inventory?

 Does the contribution of two end-members to local CO2 estimated by the CMB approach agree with that estimated by the SOCE inventory?

Chapter 5 of this thesis explores long term trends in O3 and its precursor compounds across the

GTA. The chemistry of O3 formation is introduced as well as the meteorological and regional atmospheric transport considerations required to understand long-term ozone trends. Research questions that will be addressed include:

 What is the relationship between concentrations of O3 and its precursor compounds in the GTA, and can that be used to inform future mitigation policies for improving air quality?

29

 How does meteorology and atmospheric transport impact the frequency of high ozone episodes?

 Has the oxidative environment (effective OH concentration) experienced by primary pollutants in the GTA changed over the past decade?

1.8 References

Agudelo-Castaneda, D., Teixeira, E., Rolim, S., Pereira, F., and Wiegand, F. (2013). Measurement of Particle Number and Related Pollutant Concentrations in an Urban Area in South Brazil. Atmospheric Environment, 70, 254-262.

Airparif. (2010). Airparif. Retrieved 12/05, 2016, from http://www.airparif.asso.fr/en/

Asefi-Najafabady, S., Rayer, P. J., Gurney, K. R., McRobert, A., Song, Y., Coltin, K., Huang, J.,

Elvidge, C., and Baugh, K. (2014). A Mulityear, Global Gridded Fossil Fuel CO2Emission Data Product: Evaluation and Analysis of Results. Journal of Geophysical Research: Atmospheres, 119, 10213-10231.

Bovensmann, H., Buchwitz, M., Burrows, J. P., Reuter, M., Krings, T., Gerilowski, K., Schneising, O., Heymann, J., Tretner, A., and Erzinger, J. (2010). A Remote Sensing Technique for Global Monitoring of Power Plant CO2 Emissions from Space and Related Applications. Atmospheric Measurement Techniques, 3, 781-811.

Breon, F. M., Broquet, G., Puygrenier, V., Chevallier, F., Xueref-Remy, I., Ramonet, M., Dieudonne, E., Lopez, M., Schmidt, M., Perrussel, O., and Cias, P. (2015). An Attempt at

Estimating Paris Area CO2 Emissions from Atmospheric Concentration Measurements. Atmospheric Chemistry and Physics, 15(1707), 1724.

Burrows, J. P., Holzle, E., Goede, A. P. H., Visser, H., and Frike, W. (1995). SCHIAMACHY - Scanning Imaging Absorption Spectrometer for Atmopsheric Chartography. Acta Astroanautica, 35, 445-451.

30

Bush, S. E., Pataki, D. E., and Ehleringer, J. R. (2007). Sources of Variation in D13C of Fossil Fuel Emissions in Salt Lake City, USA. Applied Geochemistry, 22, 715-723.

City of Toronto. (2007). Climate Change, Clean Air and Sustainable Energy Action Plan: Moving from Framework to Action. Retrieved 12/14, 2013, from http://www.toronto.ca/legdocs/mmis/2007/pe/bgrd/backgroundfile-4982.pdf

Clark-Thorne, S. T., and Yapp, C. J. (2003). Stable Carbon Isotope Constraints on Mixing and

Mass Balance of CO2 in an Urban Atmosphere: Dallas Metropolitan Area, Texas, USA. Applied Geochemistry, 18, 75-95.

Cooper, O. R., Gao, R., Tarasick, D., Leblanc, T., and Sweeney, C. (2012). Long-Term Ozone Trends at Rural Ozone Monitoring Sites Across the United States, 1990–2010. Journal of Geophysical Research, 117(D22307) doi:10.1029/2012JD018261

Cooper, O. R., Parrish, D. D., Ziemke, J., Balashov, N. V., Cupeiro, M., Galbally, I. E., Gilge, S., Horowitz, L., Jensen, N. R., Lamarque, J. F., Naik, V., Oltmans, S. J., Schwab, J., Shindell, D. T., Thompson, A. M., Thouret, V., Wang, Y., and Zbinden, R. M. (2014). Global Distribution and Trends of Tropospheric Ozone: An Observation-Based Review. Elements: Science of the Anthropocene, 2(000029) doi:10.12952/journal.elementa.000029

Cox, W. M., and Chu, S. (1993). Meteorologically Adjusted Ozone Trends in Urban Areas: A Probabilistic Approach. Atmospheric Environment, 27(4), 425-434.

Crisp, D., Miller, C. E., and DeCola, P. L. (2008). NASA Orbiting Carbon Observatory: Measuring the Column Averaged Carbon Dioxide Mole Fraction from Space. Journal of Applied Remote Sensing, 2, 1-14.

Crosson, E. R. (2008). A Cavity Ring-Down Analyzer for Measuring Atmospheric Levels of Methane, Carbon Dioxide and Water Vapor. Applied Physics B, 92, 403-408.

Deng, F., Jones, D. B. A., O'Dell, C. W., Nassar, R., and Parazoo, N. C. (2016). Combining

GOSAT XCO2 Observations Over Land and Ocean to Improve Regional CO2 Flux Estimates. Journal of Geophysical Research: Atmospheres, 121(4), 1896-1913.

31

Derwent, R. G., Manning, A. J., Simmonds, P. G., Gerard Spain, T., and O'Doherty, S. (2013). Analysis and Interpretation of 25 Years of Ozone Observations at the Mace Head Atmospheric Research Station on the Atlantic Ocean Coast of Ireland from 1987 to 2012. Atmospheric Environment, 80, 361-368.

Djuricin, S., Pataki, D. E., and Xiaomei, X. (2010). A Comparison of Tracer Methods for

Quantifying CO2 Sources in an Urban Region. Journal of Geophysical Research, 115(D11303), 1-13.

Duenas, C., Fernandez, M. C., Canete, S., Carretero, J., and Liger, E. (2002). Assessment of Ozone Variations and Meteorological Effects in an Urban Area in the Mediterranean Coast. Science of the Total Environment, 299(1-3), 97-113.

EDGAR. (2010). Emission Database for Global Atmospheric Research Release Version 4.2 of the European Commission. Joint Research Centre (JRC)/Netherlands Environmental Assessment Agency (PBL). Retrieved 06/29, 2016, from http://edgar.jrc.ec.europa.eu

Elminir, H. K. (2005). Dependence of Urban Air Pollutants on Meteorology. Science of the Total Environment, 350, 225-237.

Environment Canada. (2013). Manual for the Compilation of Canada's 2010 Air Pollutant Emissions. Catalogue no. En14-87/2013E-PDF. Retrieved 10/07, 2016, from http://www.ec.gc.ca/Publications/default.asp?lang=En&xml=8A08B403-C85E-4612-A958- 3F24446A61DB

Environment Canada. (2017). Air Pollutant Emission Inventory (APEI) Report. Retrieved 06/15, 2017, from http://www.ec.gc.ca/pollution/default.asp?lang=En&n=A17452DA- 1&offset=7&toc=show

Framework for Public Review and Engagement. (2007). Change is in the Air: Toronto’s Commitment to an Environmentally Sustainable Future. Retrieved 10/12, 2016, from www.toronto.ca/legdocs/mmis/2007/ex/bgrd/backgroundfile-2428.pdf

32

Gately, C. K., Hutyra, L. R., Sue Wing, I., and Brondfield, M. N. (2013). A Bottom-Up

Approach to on-Road CO2 Emissions Estimates: Improved Spatial Accuracy and Applications for Regional Planning. Environmental Science and Technology, 47, 2423-2430.

Ge, B. Z., Xu, X. B., Lin, W. L., Li, J., and Wang, Z. F. (2012). Impact of the Regional Transport of Urban Beijing Pollutants on the Downwind Areas in Summer: Ozone Production Efficiency Analysis. Tellus, 64, 17348. doi:10.3402/tellusb.v64i0.17348

Geddes, J., Murphy, J., and Wang, D. (2009). Long Term Changes in Nitrogen Oxides and Volatile Organic Compounds in Toronto and the Challenges Facing Local Ozone Control. Atmospheric Environment, 43, 3407-3415.

Gilge, S., Plass-Duelmer, C., Fricke, W., Kaiser, A., Ries, L., Buchmann, B., and Steinbacher, M. (2010). Ozone, Carbon Monoxide and Nitrogen Oxides Time Series at Four Alpine GAW Mountain Stations in Central Europe. Atmospheric Chemistry and Physics, 10, 12295-12316.

Gordon, M., Staebler, R. M., Liggio, J., Li, S., Wentzell, J., Lu, G., Lee, P., and Brook, J. R. (2012a). Measured and Modeled Variation in Pollutant Concentration Near Roadways. Atmopsheric Environment, 57, 138-145.

Gordon, M., Staebler, R. M., Liggio, J., Makar, P., Li, S., Wentzell, J., Lu, G., Lee, P., and Brook, J. R. (2012b). Measurements of Enhanced Turbulent Mixing Near Highways. Journal of Applied Meteorology and Climatology, 51, 1618-1632.

Gorka, M., Sauer, P. E., Lewicka-Szezebak, D. L., and Jedrysek, M. (2011). Carbon Isotope

Signature of Dissolved Inorganic Carbon (DIC) in Precipitation and Atmospheric CO2. Environmental Pollution, 159, 294-301.

Greening Greater Toronto. (2012). Living City Report Card 2012. Retrieved 12/04, 2015, from http://civicaction.ca/

Gurney, K. R., Chen, Y., Maki, T., Randy Kawa, S., Andrews, A., and Zhu, Z. (2005).

Sensitivity of Atmospheric CO2 Inversions to Seasonal and Interannual Variations in Fossil Fuel Emissions. Journal of Geophysical Research, 110(D10308) doi:10.1029/2004JD005373

33

Gurney, K. R., Mendoza, D. L., Zhou, Y., Fischer, M. L., Miller, C. C., Geethakumar, S., and De

La Rue du Can, S. (2009). High Resolution Fossil Fuel Combustion CO2 Emission Fluxes for the United States. Environmental Science and Technology, 43, 5535-5541.

Hakkarainen, J., Ialongo, I., and Tamminen, J. (2016). Direct Space-Based Observations of

Anthropogenic CO2 Emission Areas from OCO-2. Geophysical Research Letters, 43 doi:10.1002/2016GL070885

Hartmann, D. L., Klein Tank, A. M. G., Rusticucci, M., Alexander, L. V., Brönnimann, S., Charabi, Y., Dentener, F. J., Dlugokencky, E. J., Easterling, D. R., Kaplan, A., Soden, B. J., Thorne, P. W., Wild, M., & Zhai, P. M. (2013). In Stocker T. F., Qin D., G.-K. P., Tignor M., Allen S. K., Boschung J., Nauels A., Xia Y., Bex V. and Midgley P. M. (Eds.), Observations: Atmosphere and Surface. in: Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge, United Kingdom and New York, NY, USA: Cambridge University Press.

Hidy, G. M., and Blanchard, C. L. (2015). Precursor Reductions and Ground-Level Ozone in the Continental United States. Journal of the Air and Waste Management Association, 65(10), 1261- 1282.

IEA. (2015). CO2 Emissions from Fuel Combustion. Retrieved 11/09, 2016, from https://www.iea.org/.../CO2EmissionsFromFuelCombustionHighlights2015.pdf

IPCC-WG3. (2014). Climate Change 2014: Mitigation of Climate Change.Working Group III Contribution to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Retrieved 10/12, 2016, from http://www.ipcc.ch/report/ar5/wg3/

IPCC-WGI. (2013). Climate Change 2013: The Physical Science Basis.Working Group I Contribution to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Retrieved 11/16, 2016, from http://www.ipcc.ch/report/ar5/wg1/

Jacob, D., Logan, J., Gardner, G., Yevich, R., Spivakovsky, C., and Wofky, S. (1993). Factors Regulating Ozone Over the United States and its Export to the Global Atmosphere. Journal of Geophysical Research, 98(No. D8), 14817-14826.

34

Janardanan, R., Maksyutov, S., Oda, T., Saito, M., Kaiser, H. W., Ganshin, A., Stohl, A., Matsunaga, T., Yoshida, Y., and Yokota, T. (2016). Comparing GOSAT Observations of

Localized CO2 Enhancements by Large Emitters with Inventory-Based Estimates. Geophysical Research Letters, 43, 3486-3493.

Jasek, A., Zimnoch, M., Gorczyca, Z., Smula, E., and Rozanski, K. (2014). Seasonal Variability of Soil CO2 Flux and its Carbon Isotope Composition in Krakow Urban Area, Southern Poland. Isotopes in Environmental and Health Studies, 50(2), 143-155.

Jeong, S., Millstein, D., and Fischer, M. L. (2014). Spatially Explicit Methane Emissions from Petroleum Production and the Natural Gas Systme in California. Environmental Science and Technology, 48, 5982-5990.

Kavassalis, S. C., and Murphy, J. G. (2017). Understanding Ozone-Meteorology Correlations: A Role for Dry Deposition. Geophysical Research Letters, doi:10.1002/2016GL071791

Keeling, C. D. (1958). The Concentration and Isotopic Abundance of Carbon Dioxide in Rural Areas. Geochimica Et Cosmochimica Acta, 13, 322-334.

Keeling, C. D. (1961). The Concentration and Isotopic Abundance of Carbon Dioxide in Rural and Marine Air. Geochimica Et Cosmochimica Acta, 24, 277-298.

Kort, E. A., Angevine, W. M., Duren, R., and Miller, C. E. (2013). Surface Observations for

Monitoring Urban Fossil Fuel CO2 Emissions: Minimum Site Location Requirements for the Los Angeles Megacity. Journal of Geophysical Research: Atmospheres, 118, 1-8.

Kort, E. A., Frankenberg, C., Miller, C. E., and Oda, T. (2012). Space-Based Observations of Megacity Carbon Dioxide. Geophysical Research Letters, 39(L17806), 1-5.

Kurokawa, J., Ohara, T., Morikawa, T., Hanayama, S., Janssens-Maenhout, G., Fukui, T., Kawashima, K., and Akimoto, H. (2013). Emissions of Air Pollutants and Greenhouse Gases Over Asian Regions during 2000-2008: Regional Emission Inventory in ASia (REAS) Version 2. Atmospheric Chemistry and Physics, 13, 11019-11058.

35

Lacour, S. A., de Monte, M., Diot, P., Brocca, J., Veron, N., Colin, P., and Leblond, V. (2006). Relationship between Ozone and Temperature during the 2003 Heat Wave in France: Consequences for Health Data Analysis. BMC Public Health, 6(26) doi:0.1186/1471-2458-6-26

Lavaux, T., Miles, N. L., Richardson, S. J., Deng, A., Stauffer, D. R., Davis, K. J., Jacobson, G.,

Rella, C., Calonder, G., and DeCola, P. L. (2013). Urban Emissions of CO2 from Davos, Switzerland: The First Real-Time Monitoring System using an Atmospheric Inversion Tehcnique. Journal of Applied Meteorology and Climatology, 52, 2654-2668.

Lee, Y. C., Shindell, D. T., Faluvegi, G., Wenig, M., Lam, Y. F., Ning, Z., Hao, S., and Lai, C. S. (2014). Increase of Ozone Concentrations, its Temperature Sensitivity and the Precusor Factor in South China. Tellus, 66(23455), 1-16.

Lefer, B., Rappenglueck, B., Flynn, J., and Haman, C. (2010). Photochemical and Meteorological Relationships during the Texas-II Radical and Aerosol Measurement Project (TRAMP). Atmospheric Environment, 44, 4005-4013.

Li, J., Georgescu, M., Hyde, P., Mahalov, A., and Moustaoui, M. (2015). Regional-Scale Transport of Air Pollutants: Impacts of Southern California Emissions on Phoenix Ground-Level Ozone Concentrations. Atmospheric Chemistry and Physics, 15, 9345-9360.

Lopez, M., Schmidt, M., Delmotte, M., Colomb, A., Gros, V., Janssen, C., Lehman, S. J., Mondelain, D., Perrusel, O., Ramonet, M., Xueref-Remy, I., and Pousquet, P. (2013). CO, NOx, and 13CO2 as Tracers for Fossil Fuel CO2: Results from a Pilot Study in Paris during Winter 2010. Atmospheric Chemistry and Physics, 13, 7343-7358.

Masalaite, A., Garbaras, A., and Remeikis, V. (2012). Stable Isotopes in Environmental Investigations. Lithuanian Journal of Physics, 52(3), 261-268.

McDonald, B. C., McBride, Z. C., Martin, E. W., and Harley, R. A. (2014). High-Resolution Mapping of Motor Vehicle Carbon Dioxide Emissions. Journal of Geophysical Research: Atmospheres, 119, 5283-5298.

36

McKain, K., Wofsy, S. C., Nehrkorn, T., Eluszkiewicz, J., Erleringer, J. R., and Stephens, B. B. (2012). Assessment of Ground-Based Atmospheric Observations for Verification of Greenhouse Gas Emissions from an Urban Region. Pnas, 109(22), 8423-8428.

Miller, S. M., and Michalak, A. M. (2016). Constraining Sector-Specific CO2 and CH4 Emissions in the United States. Atmospheric Chemistry and Physics Dicsussions, doi:doi:10.5194/acp-2016-643

MOE. (2005). Transboundary Air Pollution. Retrieved 8/20, 2014, from http://www.airqualityontario.com/downloads/TransboundaryAirPollutionInOntario2005.pdf

MOE. (2013). Drive Clean. Retrieved 03/05, 2014, from http://www.ene.gov.on.ca/environment/en/category/drive_clean/index.htm

Moran, M., Zheng, Q., Zhang, J. and Pavlovic, R. (2015). RAQDPS Version 013: Upgrades to the CMC Operational Regional Air Quality Deterministic Prediction System Released in June 2015. Retrieved 10/07, 2016, from http://collaboration.cmc.ec.gc.ca/cmc/cmoi/product_guide/docs/lib/op_systems/doc_opchanges/ Technical_Note_GEM-MACH10_v1.5.3+SET2.1.1_Emissions_9Nov2015.pdf

Natural Resources Canada. (2017). Canadian Mineral Production. Retrieved 06/14, 2017, from https://www.nrcan.gc.ca/mining-materials/statistics/8848

NCAR/UCAR. (2016). Model of Emissions of Gases and Aerosols from Nature (MEGAN). Retrieved 12/05, 2016, from https://www2.acom.ucar.edu/modeling/model-emissions-gases-and- aerosols-nature-megan

Newman, S., Jeong, S., Fischer, M. L., Xu, X., Haman, C. L., Lefer, B., Alvarez, S., Rappenglueck, B., Kort, E. A., Andres, A. E., Peischl, J., Gurney, K. R., Miller, C. E., and Yung,

Y. L. (2013). Diurnal Tracking of Anthropogenic CO2 Emissions in the Los Angeles Basin Megacity during Spring 2010. Atmospheric Chemsitry and Physics, 13, 4359-4372.

Oda, T., and Maksyutov, S. (2011). A very High-Resolution (1 km X 1 km) Global Fossil Fuel

CO2 Emission Inventory Derived using a Point Source Database and Satellite Observations of Nighttime Lights. Atmospheric Chemistry and Physics, 11, 543-556.

37

Olivier, J. G. J., Van Aardenne, J. A., Dentener, F., Pagliari, V., Ganzeveld, L. N., and Peters, J. A. H. W. (2005). Recent Trends in Global Greenhouse Gas Emissions: Regional Trnds 1970- 2000 and Spatial Distribution of Key Sources in 2000. Environmental Sciences, 2(2-3), 81-99.

Oltmans, S. J., Lefohn, A. S., Shadwick, D., Harris, J. M., Scheel, H. E., Galbally, I., Tarasick, D. W., Johnson, B. J., Brunke, E. -., Claude, H., Zeng, G., Nichol, S., Schmidlin, F., Davies, J., Cuevas, E., Redondas, A., Naoe, H., Nakano, T., and Kawasato, T. (2013). Recent Tropospheric Ozone Changes - A Pattern Dominated by Slow Or no Growth. Atmospheric Environment, 67, 331-351.

Pataki, D. E., Bowling, D. R., and Ehleringer, J. R. (2003a). Seasonal Cycle of Carbon Dioxide and its Isotopic Composition in an Urban Atmosphere: Anthropogenic and Biogenic Effects. Journal of Geophysical Research, 108(D23) doi:10.1029/2003JD003865

Pataki, D. E., Ehleringer, J. R., Flanagan, L. B., Yakir, D., Bowling, D. R., Still, C. J., Buchmann, N., Kaplan, J. O., and Berry, J. A. (2003b). The Application and Interpretation of Keeling Plots in Terrestrial Carbon Research. Global Biogeochemical Cycles, 17(1), 1022-1036.

Patarasuk, R., Gurkey, K. R., O'Keeffe, D., Song, Y., Huang, J., Rao, P., Buchert, M., Lin, J. C.,

Mendoza, D., and Ehleringer, J. R. (2016). Urban High-Resolution Fossil Fuel CO2 Emissions Quantification and Exploration of Emission Drivers for Potential Policy Applications. Urban Ecosytems, 19(3), 1013-1039.

Pekey, B., and Ozaslan, U. (2013). Spatial Distribution of SO2, NO2 and O3 Concentrations in an Industrial City of Turkey using a Passive Sampling Method. Clean-Soil, Air, Water, 41(5), 423-428.

Petron, G., Tans, P., Frost, G., Chao, D., and Trainer, M. (2008). High-Resolution Emissions of

CO2 from Power Generation in the USA. Journal of Geophysical Research, 113(G04008) doi:10.1029/2007JG000602

Peylin, P., Houweling, S., Krol, M. C., Karstens, U., Rodenbeck, C., Geels, C., Vermeulen, A., Badawy, B., Aulagnier, C., Pregger, T., Delage, F., Pieterse, G., Ciais, P., and Heimann, M.

(2011). Importance of Fossil Fuel Emission Uncertainties Over Europe for CO2 Modeling: Model Intercomparison. Atmospheric Chemistry and Physics, 11(13), 6607-6622.

38

Psiloglou, B., Larissi, I., Petrakis, M., Paliatsos, A., Antoniou, A., and Viras, L. (2013). Case Studies on Summertime Measurements of O3, NO2, and SO2 with a DOAS System in an Urban Semi-Industrial Region in Athens, Greece. Environmental Monitoring and Assessment, 185, 7763-7774.

Railway Association of Canada. (2009). Rail Trends. Retrieved 06/14, 2017, from http://www.railcan.ca/publications/trends

Rayner, P. J., Raupach, M. R., Paget, M., Peylin, P., and Koffi, E. (2010). A New Global

Gridded Data Set of CO2 Emissions from Fossil Fuel Combustion: Methodology and Evaluation. Journal of Geophysical Research, 115(D19306) doi:10.1029/2009JD013439

Sather, M. E., and Cavender, K. (2016). Trends Analyses of 30 Years of Ambient 8 Hour Ozone and Precursor Monitoring Data in the South Central U.S.: Progress and Challenges. Environmental Science: Processes and Impacts, 18, 819-831.

Schneising, O., Heymann, J., Buchwitz, M., Reuter, M., Bovensmann, H., and Burrows, J. P. (2013). Anthropogenic Carbon Dioxide Source Areas Observed from Space - Assessment of Regional Enhancements and Trends. Atmospheric Chemistry and Physics, 13, 2445-2454.

Semmens, C., Ketler, R., Schwendenmann, L., Nesic, Z., & Christen, A. (2015). Isotopic composition of CO2 in gaosline, diesel and natural gas combustion exhaust in vancouver, BC, canada. Unpublished manuscript.

Shan, W., Yin, Y., Zhang, J., and Ding, Y. (2008). Observational Study of Surface Ozone at an Urban Site in East China. Atmopsheric Research, 89(3), 252-261.

Shan, W., Yin, Y., Zhang, J., Ji, X., and Deng, X. (2009). Surfance Ozone and Meteorological Condition in a Single Year at an Urban Site in Central-Eastern China. Environmental Monitoring and Assessment, 151, 127-141.

Shusterman, A. A., Teige, V. E., Turner, A. J., Newman, C., Kim, J., and Cohen, R. C. (2016).

The BErkeley Atmospheric CO2 Observation Network: Initial Evaluation. Atmospheric Chemistry and Physics, 16, 13449-13463.

39

Sillman, S., Logan, J. A., and Wofsy, S. C. (1990). The Sensitivity of Ozone to Nitrogen Oxides and Hydrocarbons in Regional Ozone Episodes. Journal of Geophysical Research, 95(D2), 1837. doi:10.1029/JD095iD02p01837

Statistics Canada. (2009a). Aircraft Movement Statistics: NAV CANADA Towers and Flight Service Stations (TP 141). Retrieved 06/14, 2017, from http://www.statcan.gc.ca/pub/51-007- x/51-007-x2014010-eng.htm

Statistics Canada. (2009b). Canadian Vehicle Survey: Annual. Retrieved 06/14, 2017, from http://www.statcan.gc.ca/pub/53-223-x/53-223-x2009000- eng.htm?contentType=application%2Fpdf

Statistics Canada. (2009c). Shipping in Canada. Retrieved 06/14, 2017, from http://www.statcan.gc.ca/pub/54-205-x/54-205-x2009000-eng.htm

Statistics Canada. (2009d). Table 3.4: Total and Urban Land Area, 1996, 2001 (Modified) and 2006. Retrieved 10/12, 2016, from http://www.statcan.gc.ca/pub/92f0138m/2008001/t/4054949- eng.htm

Statistics Canada. (2011). Population, Urban and Rural, by Province and Territory (Canada). Retrieved 10/12, 2016, from http://www.statcan.gc.ca/tables-tableaux/sum- som/l01/cst01/demo62a-eng.htm

Statistics Canada. (2012a). Report on Energy Supply and Demand. Retrieved 12/04, 2015, from http://www.statcan.gc.ca/pub/57-003-x/2015001/t037-eng.pdf

Statistics Canada. (2017). Non-Metallic Mineral Mining and Quarrying. Retrieved 06/14, 2017, from http://publications.gc.ca/site/archivee- archived.html?url=http://publications.gc.ca/Collection/Statcan/26-226-X/26-226- XIB2001000.pdf

Tarasova, O. A., and Karpetchko, A. Y. (2003). Accounting for Local Meteorological Effects in the Ozone Time-Series of Lovozero (Kola Peninsula). Atmospheric Chemistry and Physics, 3(4), 941-949.

40

Toronto Public Health. (2014). Path to Healthier Air: Toronto Air Pollution Burden of Illness Update. Retrieved 01/22, 2014, from www.toronto.ca/health/reports

Tu, J., Xia, Z., Wang, H., and Li, W. (2007). Temporal Variations in Surface Ozone and its Precursors and Meteorological Effects at an Urban Site in China. Atmospheric Research, 85, 310-337.

Turnbull, J. C., Sweeney, C., Karion, A., Newberger, T., Lehman, S. J., Tans, P. P., David, K. J., Lavaux, T., Miles, N. L., Richardson, S. J., Obiminda Cambaliza, M., Shepson, P. B., Gurney, K., Patarasuk, R., and Razlivanov, I. (2015). Toward Quantification and Source Sector

Identification of Fossil Fuel CO2 Emissions from an Urban Area: Results from the INFLUX Experiment. Journal of Geophysical Research: Atmospheres, 120, 292-312.

Turner, A. J., Shusterman, A. A., McDonald, B. C., Teige, V., Harley, R. A., and Cohen, R. C.

(2016). Network Design for Quantifying Urban CO2 Emissions: Assessing Trade-Offs between Precision and Network Density. Atmospheric Chemistry and Physics, 16, 13465-13475.

Vogel, F. R., Ishizawa, M., Chan, E., Chan, D., Hammer, S., and Levin, I. & Worthy, D.E.J.

(2012). Regional Non-CO2 Greenhouse Gas Fluxes Inferred from Atmospheric Measurements in Ontario, Canada. Journal of Integrative Environmental Sciences, 9(1), 41-55.

Wentworth, G. R., Murphy, J. G., and Sills, D. M. L. (2015). Impact of Lake Breezes on Ozone and Nitrogen Ozides in the Greater Toronto Area. Atmospheric Environment, 109, 52-60.

WHO. (2015). Global Health Observatory (GHO) Data: Urban Population Growth. Retrieved 2/3, 2015, from http://www.who.int/gho/urban_health/situation_trends/urban_population_growth_text/en/

WHO. (2016). The Urban Environment. Retrieved 11/08, 2016, from http://www.who.int/heli/risks/urban/urbanenv/en/

Widory, D., and Javoy, M. (2003). The Carbon Isotope Composition of Atmospheric CO2 in Paris. Earth and Planetary Science Letters, 215, 289-298.

Wolff, G. T., and Lioy, P. J. (1978). An Empirical Model for Forecasting Daily Ozone Levels in Northeastern U.S. Journal of the Air Pollution Control Association, 28(10), 1034-1038.

41

Wunch, D., Wennberg, P.O., Toon, G.C., Keppel-Aleks, G., Yavin, Y.G. (2009). Emissions of greenhouse gases from a North American megacity. Geophysical Research Letters, 36(15), doi:10.1029/2009GL039825.

Wunch, D., Toon, G. C., Blavier, J., Washenfelder, R. A., Notholt, J., Connor, B. J., Griffith, D. W. T., Sherlock, V., and Wennberg, P. O. (2011). The Total Carbon Column Observing Network. Philosophical Transactions of the Royal Society A, 369, 2087-2112.

Xu, J., Ma, J. Z., Zhang, X. L., Xu, X. B., Xu, X. F., Lin, W. L., Wang, Y., Meng, W., and Ma, Q. (2011). Measurements of Ozone and its Precursors in Beijing during Summertime: Impact of Urban Plumes on Ozone Pollution in Downwind Rural Areas. Atmospheric Chemistry and Physics, 11, 12241-12252.

Yokota, T., Yoshia, Y., Eguchi, N., Ota, Y., Tanaka, T., Watanabe, H., and Maksyutov, S.

(2009). Global Concentrations of CO2 and CH4 Retrieved from GOSAT: First Preliminary Results. Sola, 5, 160-163.

Zhang, J., Zheng, Q., Moran, M. D., Gordon, M., Liggio, J., Pakar, P., & Stroud, C. (2012). Improvements to SMOKE processing of canadian on-road mobile emissions. (). Toronto, Canada:

Zimnoch, M. (2009). Stable Isotope Composition of Carbon Dioxide Emitted from Anthropogenic Sources in the Krakow Region, Southern Poland. Nukleonika, 54(4), 291-295.

Chapter 2 High-resolution quantification of atmospheric CO2 mixing ratios in the Greater Toronto Area, Canada

Sections of this chapter are under review in Atmospheric Chemistry and Physics Discussions:

Pugliese, S. C., Murphy, J. G., Vogel, F., Moran, M., Zhang, J., Zheng, Q., Stroud, C., Ren, S.,

Worthy, D., Broquet, G. (2017). High-resolution quantification of atmospheric CO2 mixing ratios in the Greater Toronto Area, Canada, Atmospheric Chemistry and Physics Discussions, https://doi.org.10.5194/acp-2017-678

2 Chapter 2 2.1 Abstract

Many stakeholders are seeking methods to reduce carbon dioxide (CO2) emissions in urban areas, however reliable, high-resolution inventories are required to guide these efforts. We present the development of a high-resolution CO2 inventory available for the Greater Toronto Area and surrounding region in southern Ontario, Canada (an area of ~2.8 x 105 km2, 26 % of the province of Ontario). The new SOCE (Southern Ontario CO2 Emissions) inventory is available at the 2.5 x 2.5 km spatial and hourly temporal resolution and characterizes emissions from seven sectors: Area, Residential natural gas combustion, Commercial natural gas combustion, Point, Marine, On- road and Off-road. To assess the accuracy of the SOCE inventory, we developed an observation- model framework using the GEM-MACH chemistry-transport model run on a high-resolution grid with 2.5 km grid spacing with biogenic fluxes simulated using the European Center for Medium- Range Weather Forecasts (ECMWF) land carbon model C-TESSEL. A run using Fossil Fuel Data Assimilation System (FFDAS) v2 for the southern Ontario region was compared to a run in which its emissions were replaced by the SOCE inventory. Simulated CO2 mixing ratios were compared against in situ measurements made at four sites in southern Ontario, Downsview, Hanlan’s Point, Egbert and Turkey Point, in three winter months, January-March, 2016. Model simulations showed better agreement with measurements when using the SOCE inventory emissions versus other inventories, quantified using a variety of statistics such as correlation coefficient, root mean square

42 43 error and mean bias. Furthermore, when run with the SOCE inventory, the model had improved ability to capture the typical diurnal pattern of CO2 mixing ratios, particularly at the Downsview, Hanlan’s Point and Egbert sites. In addition to improved model-measurement agreement, the SOCE inventory offers a sectoral breakdown of emissions, allowing estimation of average time- of-day and day-of-week contributions of different sectors. Our results show that at night, emissions from Residential and Commercial natural gas combustion and other Area sources can contribute

> 80 % of the CO2 enhancement while during the day emissions from the On-road sector dominate, accounting for >70 % of the enhancement.

2.2 Introduction

Urban areas are sites of dense population and the intensity of human activities (such as transportation, industry and residential and commercial development) makes them hot-spots for anthropogenic carbon dioxide (CO2) emissions. While occupying only 3 % of the total land area, urban areas are locations of residence for 54 % of the global population and are the source of 53 –

87 % of anthropogenic CO2 emissions globally (IPCC-WG3, 2014; WHO, 2015). When considering Canada alone, the urban population accounts for an even larger fraction of the total (81 % in 2011) (Statistics Canada, 2011) while urban areas cover only 0.25 % of the land area (Statistics Canada, 2009). Recognizing their influence on the global carbon budget, many urban areas are seeking methods to reduce their anthropogenic CO2 emissions. The Greater Toronto Area (GTA) in southeastern Canada, for example, has committed to the Change is in the Air initiative as well as being a part of the C40 Cities Climate Leadership Group, both of which call to reduce

CO2 emissions 30 % below 1990 levels by 2020 (C40 Cities, 2016; Framework for Public Review and Engagement, 2007). However, in order to effectively guide anthropogenic CO2 mitigation strategies, reliable inventories are needed, particularly at high spatial and temporal resolution, to gain a better understanding of the carbon cycle (Gurney et al., 2009; Patarasuk et al., 2016). To our knowledge, the only spatially disaggregated CO2 inventories available for use in the GTA are the EDGAR v.4.2 (Emission Database for Global Atmospheric Research) CO2 inventory (available sector-wise at annual, 0.1o x 0.1o resolution) (EDGAR, 2010), the FFDAS v2 (Fossil Fuel Data o o Assimilation System) CO2 inventory (available at hourly, 0.1 x 0.1 resolution) (FFDAS, 2010), and the ODIAC (Open Source Data Inventory for Anthropogenic CO2) CO2 inventory (available at annual, 1 km resolution) (Oda and Maksyutov, 2011), all of which are limited in either their spatial and/or temporal resolution or their sectoral breakdown of emissions and therefore are not

44

well-suited for the quantification and understanding of CO2 emissions at the urban scale. The

Canadian national CO2 inventory, on the other hand, is only available at the provincial level (Environment Canada, 2012).

Efforts to develop emission inventories at the fine spatial and temporal resolution required for urban-scale understanding of CO2 emissions has been driven both by policy- and science-related questions (Gurney et al., 2009; Patarasuk et al., 2016). From a policy perspective, improving CO2 emission quantification is essential to independently evaluate whether anthropogenic mitigation regulations are being effectively implemented. From a scientific perspective, gaining information about anthropogenic CO2 emissions from urban areas has been primarily motivated by atmospheric

CO2 inversions, which are used to better understand the global carbon cycle (Gurney et al., 2009;

Patarasuk et al., 2016). Regardless of the motivation, quantification of CO2 source/sink processes currently uses two techniques: the bottom-up approach and the top-down approach. In the bottom- up approach, local-scale activity level information is combined with appropriate emission factors to infer emission rates. This method has been used widely to develop many inventories (EDGAR, 2010; FFDAS, 2010; Gurney et al., 2009) but is limited by the accuracy of the input parameters. Conversely, in the top-down approach, inverse modelling is used to exploit the variability in atmospheric mixing ratios of CO2 to identify the source/sink distributions and magnitudes; this method is limited by insufficient mixing ratio data and uncertainties in simulating atmospheric transport (Pillai et al., 2011). Given current policy needs, a strategy using solely bottom-up or top- down approaches is likely insufficient to evaluate CO2 emissions but rather a synthesis of the two methodologies is required (Miller and Michalak, 2016). Successful examples of high-resolution

CO2 inventory development are available on the urban scale, such as the Airparif inventory in Ile- de-France (publicly available at http://www.airparif.asso.fr/en/index/index) and in Indianapolis, Los Angeles, Salt Lake City and Phoenix through the Hestia project (Gurney et al., 2012), on the national scale, such as in China (Zhao et al., 2012), and on the global scale (Wang et al., 2013).

However, to our knowledge, there are currently no studies that have quantified Canadian CO2 emissions at the fine spatial and temporal resolution required for urban analyses in Canada.

In an effort to address this gap, this study was focused on quantifying CO2 emissions at a fine spatial and temporal resolution in the GTA and southern Ontario (we expanded the inventory beyond the urban area of the GTA so we could exploit information on CO2 mixing ratios collected at rural areas in central and south-western Ontario, proving additional sites for inventory

45

validation). We present the new high-resolution Southern Ontario CO2 Emissions (SOCE) inventory, which quantifies CO2 emissions from seven source sectors (On-road, Off-road, Area, Point, Marine, Residential, and Commercial natural gas combustion) at 2.5 km x 2.5 km spatial and hourly temporal resolution for an area covering ~26 % of the province of Ontario (~2.8 x 105 km2). The SOCE inventory was used in combination with the Environment and Climate Change

Canada (ECCC) GEM-MACH chemistry-transport model to simulate CO2 mixing ratios in a domain including south-eastern Canada and the northeastern USA (hereafter referred to as the “PanAm domain”) for comparison with in situ measurements made by the Southern Ontario

Greenhouse Gas Network. Until now, estimates of anthropogenic CO2 emissions in the GTA were available only from the EDGAR v.4.2 (EDGAR, 2010) and the FFDAS v2 (FFDAS, 2010) inventories, which have very different annual totals for this region (1.36 x 108 vs. 1.05x 108 tonnes

CO2, respectively). Therefore, we expect the results of this work will improve our ability to quantify the emissions of CO2 in the entire domain as well as the relative contributions of different sectors, providing a more detailed characterization of the carbon budget in the GTA.

2.3 Methods

2.3.1 Study region

The geographic focus of this study was the GTA in southern Ontario, Canada. The GTA is the largest urban area in Canada; it comprises five municipalities, Toronto, Halton, Durham, Peel and York, which together have a population exceeding 6 million (Statistics Canada., 2012b). Although the GTA comprises only 0.07 % of Canadian land area, it represents over 17 % of the total population as a result of rapid urbanization over the past few decades (Statistics Canada., 2012b).

Therefore, high-resolution characterization of CO2 emissions can help integrate climate policy with urban planning. This region is home to a network of measurement sites providing long-term, publicly available datasets of atmospheric CO2 mixing ratio measurements, Section 2.3.2 (Environment Canada, 2011), which can be used to evaluate model outputs and inventory estimates. In 2016 the government of Ontario released a Climate Change Action Plan, which includes an endowment given to the Toronto Atmospheric Fund of $17 million to invest in strategies to reduce greenhouse gas pollution in the GTA (Ontario, 2016). Therefore this research can provide timely information on the carbon budget in the GTA and help to implement effective reduction strategies.

46

2.3.2 The Southern Ontario Greenhouse Gas Network

Measurements of ambient CO2 dry air mixing ratios began in 2005 in southern Ontario at the Egbert station followed by the Downsview station (2007), Turkey Point station (2012) and Hanlan’s Point station (2014), Figure 2.1. Egbert is located ~75 km north-northwest of Toronto in a rural area, Downsview is located ~20 km north of downtown core of the city of Toronto in a populated suburban area, Turkey Point is located to the south-west of the GTA in a rural area on the north shore of Lake Erie, and Hanlan’s Point is located on Toronto Island, just south of the city of Toronto on the shore of Lake Ontario. Site details and instrument types used can be found in

Table 2.1. CO2 measurements are collected as a part of ECCC’s Greenhouse Gas Observational Program. The measurement procedure follows a set of established principles and protocols outlined by a number of international agencies through recommendations of the Meeting on Carbon Dioxide, Other Greenhouse Gases, and Related Measurement Techniques, coordinated by the World Meteorological Organization (WMO) every 2 years.

Table 2.1: Summary of atmospheric measurement programs in Southern Canada operated by Environment and Climate Change Canada

Elevation Intake In-situ Start Date Site Name Coordinates (asl) Height Instrumentation

44.231037N, 3m, March, 2005 Egbert 251m NDIR 79.783834W 25m*

November, 43.780491N, Downsview 198m 20m CRDS 2010 79.468010W

November, 42.635368N, Turkey Point 231m 35m CRDS 2012 80.557659W

43.612201N

June, 2014 Hanlan's Point 79.388705W 87m 10m CRDS

* At Egbert, a 25 m tower was installed in March 9, 2009 NDIR = Non-dispersive infrared CRDS = cavity ring-down spectroscopy

47

The atmospheric CO2 observational program at Egbert is based on non-dispersive infrared (NDIR) methodology and fine-tuned for high precision measurements (Worthy et al., 2005). The atmospheric CO2 observational programs at Turkey Point, Downsview, and Hanlan’s Point are based on Cavity Ring-Down Spectrometer (CRDS). Each Picarro CRDS system is calibrated in the ECCC central calibration facility in Toronto before deployment to the field. The response function of the analyzer is determined against 3 calibrated standards tanks (Low, Mid, High). The working (W) and target (T) tanks assigned to the system are also included in the injection sequence and calibrated. At each site, ambient measurements are made using two sample lines placed at the same level. Each sample line has separate dedicated sample pumps and driers (~ -30oC). Pressurized aluminum 30 L gas cylinders are used for the working and target tanks. The sample flow rate of the ambient and standard tank gases is set at ~300 cc/min. The injection sequence consists of a target and working tanks sequentially passed through the analyzer for 10 minutes each every 2 days. The ambient data from line1 is passed through the analyzer for 18 hours followed by Line2 for 6 hours. The Line1/Line2 sequence repeats one time before the target and working tanks are again passed through the system. The working and target tanks are calibrated on site at least once per year against a single transfer standard transported between the sites and the central laboratory facility in Toronto. The CO2 measurements from both the NDIR and CRDS analytical systems have a precision of around 0.1 ppm based on one-minute averages and are accurate to within 0.2 ppm.

2.3.3 GEM-MACH chemistry-transport model

In this project, we used the GEM-MACH (Global Environmental Multi-scale‒Modelling Air quality and CHemistry) chemistry‒transport model (CTM) (Gong et al., 2015; M. D. Moran et al., 2013; Pavlovic et al., 2016; Talbot et al., 2008) to link surface emission estimates and atmospheric mixing ratios. GEM-MACH is an on‒line CTM embedded within the Canadian weather forecast model GEM (Côté et al., 1998a; Côté et al., 1998b). The configuration of GEM-MACH used in our study has 62 vertical levels from the surface to ~1.45 hPa on a terrain-following staggered vertical grid for a log-hydrostatic pressure coordinate. The thickness of the lowest layer was 40 m. The PanAm domain used in our simulations, which includes central and southern Ontario, as well as western Quebec and the northeastern USA, is shown in Figure 2.1. The PanAm domain has 524 x 424 grid cells in the horizontal on a rotated latitude-longitude grid with 2.5-km grid spacing and covers an area of approximately 1310 km x 1060 km (total domain area is 1.39 x 106

48 km2). A 24-hour forecasting period was used with a 60-second time step for each integration cycle. Meteorological fields (wind, temperature, humidity, etc.) were re-initialized every 24 hours (i.e., after each model integration cycle); chemical fields were carried forward from the previous integration cycle (i.e., perpetual forecast). Hourly meteorological and chemical boundary conditions were provided by the ECCC operational 10 km GEM-MACH air quality forecast model (Moran et al., 2015).

Turkey Point

Figure 2.1: Total anthropogenic CO2 emissions for a weekday in February 2010 estimated by the SOCE inventory for the province of Ontario and by the FFDAS v2 inventory for the remainder of the GEM-MACH PanAm domain. Locations of in-situ measurements of CO2 in the Southern Ontario GHG Network are shown in the inset (Downsview = square, Egbert = circle, Hanlan’s Point = triangle, Turkey Point = diamond). The Downsview and Hanlan’s Point sites are both located in the GTA. Units: g CO2/second/grid cell.

49

In our study, we simulated two scenarios of CO2 surface fluxes, indicated by the sum of the following:

Scenario 1:

 Anthropogenic fossil fuel CO2 emissions within the province of Ontario estimated by the SOCE inventory, available at 2.5 km x 2.5 km spatial and hourly temporal resolution, as described in Section 2.3.4

 Anthropogenic fossil fuel CO2 emissions estimated by the FFDAS v2 inventory (FFDAS, 2010) outside of the province of Ontario (province of Quebec and USA), available at 0.1o x 0.1o spatial and hourly temporal resolution

 Biogenic CO2 fluxes from the C-TESSEL land surface model, as described in Section 2.3.5

Scenario 2:

 Anthropogenic fossil fuel CO2 emissions estimated by the FFDAS v2 inventory (FFDAS, 2010) for the entire domain, available at 0.1o x 0.1o spatial and hourly temporal resolution

 Biogenic CO2 fluxes from the C-TESSEL land surface model, as described in Section 2.3.5

CO2 is not a usual chemical species considered by GEM-MACH but a set of special inert tracer fields were added to GEM-MACH for this project to account for CO2 concentration fields associated with different source sectors and the lateral boundaries. The CO2 boundary conditions set at the lateral and top edges of the domain were obtained from the Monitoring Atmospheric Composition and Climate (MACC) global inversion, v.10.2 (http://www.copernicus- atmosphere.eu/). Model simulated specific humidity (q, kg/kg) was used to convert estimated CO2 mixing ratios to dry air mixing ratios. CO2 dry air mixing ratios are hereafter referred to CO2 mixing ratios.

2.3.4 High-resolution SOCE inventory development

The high-resolution SOCE inventory was constructed primarily from a pre-existing carbon monoxide (CO) inventory developed by the Pollutant Inventories and Reporting Division (PIRD) of ECCC as part of the 2010 Canadian Air Pollutant Emissions Inventory (APEI). The CO inventory is a comprehensive national anthropogenic inventory that includes emissions from area sources, point sources, on-road mobile sources and off-road mobile sources, including aircraft, locomotive and marine emissions for base year 2010 (M. Moran et al., 2015). This annual inventory at the provincial level compiled by PIRD was transformed into model-ready emissions

50 files using the Sparse Matrix Operator Kernel Emissions (SMOKE, https://www.cmascenter.org/smoke/) emissions processing system for spatial allocation (distribution of non-point source emissions to 2.5 km x 2.5 km (roughly 0.02o x 0.02o resolution) using spatial surrogate fields) and temporal allocation (conversion of inventory annual emission rates into hourly values) (Moran et al., 2015). More detailed information about the CO inventory compilation and subsequent processing has been provided elsewhere (Environment Canada, 2013; Moran et al., 2015; PIRD, 2016).

The objective of our work was to calculate CO2 emissions based on this processed, model-ready CO inventory for Ontario grid cells using sector-specific emission ratios estimated by the Canadian National Inventory Report (NIR) (Environment Canada, 2012). Emission sources within each sector of the CO inventory are classified by SCC (Source Classification Code) and were converted to NFR (Nomenclature for Reporting) for accurate cross-reference with the NIR CO2 and CO estimates. Provincial totals for CO2 and CO are estimated based on the NFR sources that are included in the sector, producing the following sector-averaged CO2:CO ratio:

퐶푂2(푂푛푡푎푟푖표 푡표푡푎푙,푘푡) 퐶푂2 = 퐶푂(푠푒푐푡표푟,푘푡) ∗ (E2.1) (푠푒푐푡표푟,푘푡) 퐶푂(푂푛푡푎푟푖표 푡표푡푎푙,푘푡)

This sector-averaged CO2:CO ratio is used to convert the APEI-based CO model-ready gridded emissions fields into CO2 emissions fields at the same spatial and temporal resolution. A detailed outline of this conversion is presented for each sector in the following subsections. Unless otherwise noted, temporal allocation of emissions in each sector is based on estimates made available by SMOKE.

2.3.4.1 Area emissions

Area emissions are mostly small stationary sources that represent diffuse emissions that are not inventoried at the facility level. In the APEI CO inventory, the major emission sources in the Area sector include emissions from public electricity and heat production (1A1a), residential and commercial plants (1A4a and 1A4b), stationary agriculture/forestry/fishing (1A4c), iron and steel production (2C1), and pulp and paper (2D1). The NIR estimates an Ontario total from these (and other minor sources) of 23,455 kt CO2 and 218.8 kt CO, producing a CO2:CO ratio of 107.2 kt

CO2/kt CO. This ratio was applied to every Area sector grid cell belonging to Ontario in the domain to convert sector CO emissions to CO2 emissions.

51

2.3.4.2 Point emissions

Point emissions are stationary sources in which emissions exit through a stack or identified exhaust. In the APEI CO inventory, the major emission sources in the Point sector include public electricity and heat production (1A1a), stationary combustion in manufacturing industries and construction (1A2f), chemical industry (2B5a), pulp and paper (2D1), iron and steel production (2C1) and other metal production (2C5). Unlike the Area sector, we found that applying a single

CO2:CO ratio to every facility did not produce realistic CO2 emissions due to the significant variability in combustion efficiency (and thus CO2:CO ratio). Therefore, we used ECCC Facility Reported Data (Environment Canada, 2015) to identify the geocoded location and annual average

CO2:CO for 48 individual facilities in Ontario (Table A1) and applied the specific CO2:CO ratios to the grid cells where the facilities were located. In addition, stack height of individual facilities were included in the emission model to optimize plume rise. All other point sources (minor facilities) were scaled by a sector average CO2:CO ratio of 313.1 kt CO2/kt CO, calculated from

Ontario total CO2 and CO point-source emissions from the NIR. Temporal allocation of emissions in the Point sector are based on facility level operating schedule data collected by ECCC.

2.3.4.3 On-road emissions

On-road emissions include the emissions from any on-road vehicles (quantified by the Statistics Canada Canadian Vehicle Survey) (Environment Canada, 2013). In the APEI CO inventory, the major emission sources in the On-road sector includes gasoline and diesel-powered light- and heavy-duty vehicles (1A3b). The NIR estimates an Ontario total from these (and other minor on- road sources) of 44,458 kt CO2 and 1508.3 kt CO, producing a CO2:CO ratio of 29.5 kt CO2/kt CO. This ratio was applied to every On-road grid cell belonging to Ontario in the domain to convert sector CO emissions to CO2. Temporal allocation of emissions in the On-road sector is estimated using data collected in the FEVER (Fast Evolution of Vehicle Emissions from Roadways) campaign in 2010 (Gordon et al., 2012a; Gordon et al., 2012b; Zhang et al., 2012).

2.3.4.4 Off-road mobile emissions

Off-road emissions include the emissions from any off-road vehicles that do not travel on designated roadways, including aircraft, all off-road engines, and locomotives. In the APEI CO inventory, the major emission sources in the Off-road sector include civil aviation (1A3a), railways (1A3c), and agriculture/forestry/fishing: off-road vehicles and other machinery (1A4c). Similar to

52

the Point sector, we found that applying a single CO2:CO ratio to every grid cell did not produce realistic CO2 emissions for the two airports in the GTA, Pearson International Airport (PIA) and Billy Bishop Toronto City Airport (BBTCA). Therefore, we used air quality assessment reports compiled for each airport (RWDI AIR Inc., 2009; RWDI AIR Inc., 2013) to identify the geocoded location and facility-specific annual average CO2:CO ratio. Sources of emissions from each airport include aircraft (landing and take-off cycles), auxiliary power units, ground support equipment, roadways, airside vehicles, parking lots, stationary sources and training fires; note that emissions from aircrafts in-transit between airports, which are injected in the free troposphere, have not been included in this inventory (M. Moran et al., 2015; RWDI AIR Inc., 2009). Based on these two reports, we applied a ratio of 175 kt CO2/kt CO to the grid cell containing PIA and a ratio of 20 kt CO2/kt CO to the grid cell containing BBTCA. All other off-road sources belonging to Ontario grid cells were scaled by a sector average CO2:CO ratio of 7.2 kt CO2/kt CO, calculated from NIR-reported Ontario total CO2 and CO emissions.

2.3.4.5 Marine emissions

Commercial marine emissions include the emissions from any marine vessels travelling on the Great Lakes (quantified by the Statistics Canada Shipping in Canada) (Environment Canada, 2013). In the APEI CO inventory, the major emission source in the Marine sector is national navigation (1A3d). The NIR estimates an Ontario total from this source of 729.2 CO2 and 0.86 kt

CO, producing a CO2:CO ratio of 844.2 kt CO2/kt CO. This ratio was applied to every marine grid cell in the domain to convert sector CO emissions to CO2.

2.3.4.6 Residential and commercial emissions

Residential and commercial CO2 emissions reflect on-site combustion of natural gas for electricity and heating, a source that we found was not included in the APEI CO inventory because of the high efficiency of the furnaces and resulting low CO emissions. To include the CO2 emissions from these on-site furnaces, we used the Statistics Canada 2012 Report on Energy Supply and Demand to quantify the amount of natural gas consumed by residential and commercial buildings in Ontario, 7969.6 gigalitres (Gl) and 4895.7 Gl respectively (Statistics Canada, 2012a). We used 3 an emission factor of 1879 g CO2/m natural gas combustion (Environment Canada, 2012) to estimate CO2 emissions from residential and commercial on-site furnaces in Ontario to be 1.5 x 107 tonnes and 9.2 x 106 tonnes, respectively. These two emission totals were spatially allocated

53 using a “capped-total dwelling” spatial surrogate developed by ECCC and temporally allocated using the SMOKE emissions processing system (Moran et al., 2015).

2.3.5 Biogenic Flux

The net ecosystem exchange (NEE) fluxes used in our simulations were provided by the land surface component of the ECMWF forecasting system, C-TESSEL (Bousetta et al., 2013). Fluxes are extracted at the highest available resolutions, 15 x 15 km and 3 hour for January and February 2016 and 9 x 9 km and 3 hour for March. These data are interpolated in space and time to be consistent with our model resolution. With our main priority being understanding anthropogenic emissions in the GTA, we chose to analyze a period where the biogenic CO2 flux is minimized and therefore this paper focuses on three winter months in 2016, January to March inclusive.

2.4 Results and Discussion 2.4.1 The SOCE Inventory

Figure 2.1 displays the PanAm domain total anthropogenic CO2 emissions estimated by the SOCE inventory for the province of Ontario portion (~0.02o x 0.02o) and by the FFDAS v2 inventory (0.1o x 0.1o) (FFDAS, 2010) for the remainder of the domain. Regions of high emissions typically correspond to population centers, for example the GTA in Ontario, Montreal and Quebec City in Quebec, and Chicago, Boston and New York City (amongst others) in the USA. Emissions from highways and major roadways are only clear in the province of Ontario (at higher spatial resolution) but industrial and large scale area sources are evident across the entire domain.

The total CO2 emissions can be separated into contributions from the seven sectors in the province of Ontario described in Section 2.3.4. Figure 2.2 shows the anthropogenic CO2 contributions from the Area sector, Residential and Commercial sector, Point sector, Marine sector, On-road sector and Off-road sector, focused on southern Ontario and the GTA. If we consider emissions from a domain including the area solely around the GTA (indicated by the black-box in Figure 2.2a), the total CO2 emissions estimated by the SOCE inventory is 94.8 Mt CO2 per year, Table 2.2. Figures

2.2a and b display the CO2 emissions from the Area sector and from Residential and Commercial natural gas combustion in southern Ontario. These two sectors combined represent the largest source of CO2 in the black-box area (41.6 Mt CO2/year, contributing 43.9 % of the total). The majority of these emissions are concentrated in the GTA and surrounding urban areas as a result

54 of a significant portion of the population (64 %) being reliant on natural gas for heat production (Statistics Canada, 2007; Statistics Canada, 2012a). Figure 2.2c represents emissions from the

Point sector, contributing 24.4 Mt CO2/year, 25.7 % of the total. The largest point source emitters are located on the western shore of Lake Ontario (Hamilton and surrounding areas) as this area is one of the most industrialized regions of the country with intensive metal production activities.

Figures 2.2d, e and f display CO2 emissions from various transportation sectors, Marine, On-road, and Off-road respectively, which together contribute more than 30 % of total CO2 emissions in the area within the black box. While emissions from marine activity are minimal, those from On-road and Off-road sources are significant (25.0 % and 5.3 %, respectively), concentrating on the major highways connecting the various population centres of the GTA to the downtown core, as well as at PIA located within the city.

55

a) b) a) b

)

a) )

c) d)

d

)

)

e) f)

e) f)

) )

Figure 2.2: Anthropogenic CO2 emissions for a weekday in February 2010 in southern Ontario. Emissions are estimated by the SOCE inventory for the (a) Area sector; (b) sum of the Residential and Commercial sectors; (c) Point sector; (d) Marine sector; (e) On-road sector; (f) Off-road sector. Units: log10(g CO2/second/grid cell).

56

2.4.2 Comparison of the SOCE inventory with other inventories

The EDGAR v4.2 inventory estimates CO2 emissions on an annual basis and by sector based on Selected Nomenclature for Air Pollution (SNAP) sub-sectors while FFDAS v2 provides hourly

mean grid cell totals. Table 2.2 shows a comparison between the sectoral CO2 estimates of the SOCE and EDGAR v4.2 inventories (SNAP sectors were grouped to correspond to SOCE sectors, Table A2) as well as the domain total estimated by the FFDAS v2 inventory for the area surrounding the GTA (the black-box area outlined in Figure 2.2a). There is a significant

discrepancy between the CO2 emissions estimated by the SOCE and EDGAR v4.2, inventories both in the relative sectoral contributions as well as domain total (difference >35 %). The largest sectoral discrepancies are in the Point and the On-road sectors, where the EDGAR v4.2 inventory estimates a contribution 1.9 and 1.7 times larger than that of the SOCE inventory, respectively.

Figure 2.3 shows a comparison of the spatial distribution of the CO2 inventory predicted by a) FFDAS v2, b) EDGAR v4.2, and c) SOCE (aggregated to 0.1o x 0.1o to match the resolution of EDGAR v4.2 and FFDAS v2) for the GTA area. Figure 2.3 reveals that the largest differences

between the SOCE inventory and the EDGAR v4.2 inventory is the CO2 emissions in the GTA; EDGAR v4.2 predicts much higher emissions in the GTA (in some grid cells, differences are an order of magnitude), particularly in the downtown core relative to the SOCE inventory.

Table 2.2: Anthropogenic CO2 emissions for the year 2010 in the black-box area (shown in Figure 2a) by sector. Values in parentheses indicate the percentage contribution of the sector to the total CO2 emissions in the black-box area. ‡ # Sector FFDAS v2 CO2 Inventory EDGAR v4.2 CO2 SOCE CO2 Inventory (Mt (Mt CO2/year) Inventory (Mt CO2/year) CO2/year) Area* - 46.2 (33.9 %) 41.6 (43.9 %) Point - 45.9 (33.7 %) 24.4 (25.7 %) Marine - 0.12 (0.10 %) 0.10 (0.10 %) On-road - 41.2 (30.2 %) 23.7 (25.0 %) Off-road - 2.95 (2.2 %) 5.01 (5.3 %) Total 104.8 136.4 94.8 *Area sector represents the summation of Area + Residential + Commercial natural gas combustion. #The EDGAR inventory v4.2 can be found at http://edgar.jrc.ec.europa.eu. ‡The FFDAS v2 inventory can be found at http://hpcg.purdue.edu/FFDAS/.

57

FFDAS v2 Domain Total: EDGAR v4.2 Domain Total: SOCE Domain Total: 8 8 1.05x10 tonne/year 1.36x10 tonne/year 9.48x107 tonne/year a) b c)

) a)

) c) ) a) c) a) ) )

)

Figure 2.3: Comparison of spatial distribution of annual CO2 emissions inventories for the black- box area (shown in Figure 2a) at 0.1o x 0.1o resolution. Panel a) shows the FFDAS v2 inventory estimate, Panel b) shows the EDGAR v4.2 inventory estimate and Panel c) shows the SOCE inventory estimate. Units: log10(tonne CO2/year/grid cell). Domain totals are shown on top of each panel and locations of in-situ measurements of CO2 for three stations in the Southern Ontario GHG Network are shown on Panel a (Downsview = square, Hanlan’s Point = triangle, TAO = pentagon). The other two stations, Egbert and Turkey Point, are located outside this area.

Although there is no sectoral breakdown in the FFDAS v2 inventory, the domain total around the GTA can be compared to that of the SOCE inventory, Table 2.2. Unlike the comparison with the EDGAR v4.2 inventory, there is a closer agreement between the FFDAS v2 inventory and the SOCE inventory (difference of ~10 %). The comparison plots in Figure 2.3 show a good agreement of the spatial variability of emissions in the GTA between the FFDAS v2 and SOCE inventories; however, the gradient between urban and rural areas is not as sharp in the SOCE inventory as it is in the FFDAS v2 inventory. Furthermore, emissions along the western shore of Lake Ontario (Hamilton and the surrounding areas) are larger in the SOCE inventory than in FFDAS v2. The FFDAS v2 inventory was interpolated to 2.5 km x 2.5 km using a mass conservative interpolation scheme to allow the production of a difference plot of the two inventories, SOCE minus FFDAS v2, shown in Figure A1. The difference plot reveals the largest divergence between the inventories

occurs in the GTA and Ottawa, with the FFDAS v2 inventory estimating >1000 g CO2/second

(~30 kt CO2/year) more than the SOCE inventory in some grid cells. In addition to similar spatial variability, the FFDAS v2 and SOCE inventories also have similar temporal variability. Figure A2

58 shows the diurnal profile of estimated emissions from January-March for both the FFDAS v2 and SOCE inventories for the black-box area in the PanAm domain. Both inventories allocate the highest emissions between 08:00 and 18:00 and the lowest emission between 00:00 and 05:00, however the amplitude of the diel cycle is higher in SOCE, and emissions in the morning are as high as in the afternoon. FFDAS allocates a relatively larger proportion of the emissions to the 15:00 – 19:00 period.

2.4.3 Preliminary analyses using the SOCE, FFDAS v2 and EDGAR v4.2 inventories with FLEXPART

To investigate the impact of the differing inventories on ambient mixing ratios, preliminary analyses were run with footprints generated by the FLEXPART model driven by ECMWF meteorology and products were compared against the measured CO2 gradient between the Downsview and TAO (43.7oN, 79.4oW (see Figure 2.3), a temporary site decommissioned in January 2016) stations in the year 2014. Observed gradients ranged from +20 to -10 ppm. Figure

A3 displays the measured and modelled CO2 gradients. These results show that when the EDGAR v4.2 inventory was used, simulated CO2 gradients were consistently overestimated by ~10-60 ppm relative to observations. Conversely, when the SOCE inventory was used, a higher level of agreement was obtained between simulated mixing ratios and measurements; however, none of the model simulations were able to capture times when the gradient was negative (CO2,TAO >

CO2,Downsview), an effect we believe to be due to the TAO inlet being ~60 m above ground level and surrounded by many high-rise buildings creating canyon flows and turbulence which are not properly accounted for in ECMWF at this resolution. These factors contributed to the decommissioning of TAO in January 2016. The poor performance of our model system when using the EDGAR v4.2 inventory to simulate CO2 mixing ratios was also found by a study quantifying on-road CO2 emissions in Massachusetts, USA (Gately et al., 2013). In this study, EDGAR emission estimates were found to be significantly larger than any other inventory by as much as 9.3 million tons, or >33 %. The difference in estimates between the EDGAR v4.2 and the SOCE inventories is likely explained by their underlying differences in methodology. Being a global product and not specifically designed for mesoscale applications, the EDGAR v4.2 inventory estimates CO2 emissions based on country-specific activity data and emission factors, however the spatial proxies used to disaggregate the data are not always optimal. A study performed by McDonald et al. (2014) showed that the use of road density as a spatial proxy for vehicle emissions

59 in EDGAR v4.2 causes an overestimation of emissions in population centers (McDonald et al.,

2014). Given the much larger emission estimates for On-road CO2 from EDGAR v4.2 (Table 2.2), this also seems to be an issue in the GTA. Based on this large discrepancy, the EDGAR v4.2 inventory was not further used in this study and we focused on the inventories developed for regional scale studies.

When similar preliminary analyses were run with FLEXPART footprints using the FFDAS v2 inventory, Figure A3, good agreement was observed with CO2 gradients measured between the Downsview and TAO stations. We are confident that the enhanced measurement agreement between the FFDAS v2 and SOCE relative to EDGAR v4.2 is due to improved methodology; spatial allocation of emissions in FFDAS v2 is achieved through the use of satellite observations of nightlights from human settlements from the U.S. Defense Meteorological Satellite Program Operational Linescan System (DMSP-OLS).

Beyond the differences in methodology for estimating and allocating emissions, it is important to note that the emissions reported in Table 2.2 by the FFDAS v2, SOCE and EDGAR v4.2 inventories also fundamentally differ in time period quantified. The emissions reported for both FFDAS v2 and the SOCE are based on emissions from three winter months (January-March 2010) extrapolated for the entire year. However, emissions from EDGAR v4.2 are annual averages of all twelve months of 2010. Since CO2 emissions in the GTA are higher in the winter months relative to the summer months because of increased building and home heating, it is likely that the average annual estimates of SOCE and FFDAS v2 are slightly overestimated. This does not affect the relative agreement between SOCE and FFDAS v2 however it does further increase the divergence between the EDGAR v4.2 and SOCE and FFDAS v2 inventories. Following this and the improved agreement with observations, the FFDAS v2 inventory was used with the SOCE inventory for all subsequent modelling analyses.

2.4.4 Simulation of CO2 mixing ratios in the Greater Toronto Area

We used the GEM-MACH chemistry-transport model and the SOCE and FFDAS v2 inventories to simulate hourly CO2 mixing ratios in the PanAm domain. The model framework was evaluated for a continuous three-month period, January-March 2016, using four sampling locations in the GTA, Figure 2.1 (note that measurements for the Hanlan’s Point site were not available until

January 14, 2016). Figure 2.4 displays afternoon (12:00-16:00 EST) measured and simulated CO2

60 mixing ratios produced with the SOCE and FFDAS v2 inventories for the two emissions scenarios described in Section 2.3.3 for the month of February (Figures A4 and A5 show the same figure for other months). We chose to present only afternoon data as this is the time of day when the mixed layer is likely to be the most well-developed; nighttime and morning data showed largest variations in observations as a result of the shallow boundary layer causing surface emissions to accumulate within the lowest atmospheric layers (Breon et al., 2015; Chan et al., 2008; Gerbig et al., 2008). During the night, atmospheric mixing ratios are most sensitive to vertical mixing, an atmospheric process that is difficult to model for stable boundary layers.

Figure 2.4: Time series of measured (blue) and modelled February afternoon (12:00-16:00 EST) CO2 mixing ratios for the four sites used in this study. The red and gold markers are the modelled mixing ratios when using the SOCE inventory and the FFDAS v2 inventory, respectively.

61

The time series comparisons at all four sites demonstrate the model’s general ability to capture variability in observations of CO2, albeit with better skill for the Downsview and Egbert sites (this is particularly clear when we look at model-measurement difference plots, Figure A6 and Table A3). The model is able to capture many extreme events of mixing ratio increases and decreases, such as February 11-14, 2016 at the Downsview site; however, some short time periods are poorly simulated, such as January 21-23, 2016 at Hanlan’s Point, when the model significantly overestimated measured CO2. Generally, mixing ratios simulated by the FFDAS v2 inventory are similar or larger than those produced when the SOCE inventory is used, with differences most noticeable at the Downsview and Hanlan’s Point sites. This was expected as the difference plot shown in Figure A1 reveals that the SOCE and FFDAS v2 inventories diverge the most in the GTA (where the Downsview and Hanlan’s Point sites are located) and are more similar in rural areas (where the Turkey Point and Egbert sites are located).

Measured CO2 mixing ratios have a typical diurnal pattern, in which mixing ratios are higher at night and lower during the day, despite higher emissions during the day. This results from the daily cycle of the mixed layer, which is shallow at night due to thermal stratification and deepens during the day due to solar heating of the surface. Figure 2.5 displays the measured and modelled mean diurnal profile of CO2 at the four sites in our network using data from January-March 2016 (note difference in y-axis scale for urban vs. rural sites). At all four sites, the shapes of the modelled and measured mixing ratios throughout the day agree very well, suggesting that the GEM meteorology in our framework is capturing the diurnal variation in emissions and the boundary layer evolution. At the Downsview site, there is a very strong agreement between the modelled and measured diurnal profiles when using the SOCE inventory, whereas the FFDAS v2 simulated profile largely overestimates mixing ratios, particularly at nighttime. This is consistent with the FFDAS inventory having larger emissions than the SOCE inventory during the night (Figure A2). At the Hanlan’s

Point site, a difference of ~ 5 ppm CO2 is observed when using the SOCE inventory relative to measurements; however, similar to the Downsview site, the FFDAS v2 simulated profile has a larger difference of ~10 ppm CO2. At both the Egbert and Turkey Point sites, the use of both inventories similarly overestimates the diurnal pattern of CO2 mixing ratios by ~3-5 ppm, again likely a result of the similarities of these two inventories at these sites, Figure A1. At all four sites, it is possible that some of the biases that are observed in simulated CO2 mixing ratios may arise

62

from inaccuracies in the boundary CO2 provided by MACC; this aspect was not, however, further explored in this study.

Figure 2.5: Time series of mean measured (blue) and modelled diurnal CO2 mixing ratios at the four sites considered in this study for January – March 2016. The red and gold markers are the modelled diurnal mixing ratios when using the SOCE CO2 inventory and the FFDAS v2 inventory, respectively. Note the difference in scale for urban and rural sites.

2.4.5 Quantifying the model-measurement agreement

Figure 2.6 shows scatter plots of afternoon (12:00-16:00 EST) modelled versus measured CO2 mixing ratios from January- March 2016 at the four sites used in this study. The top row shows the correlation between measured and modelled mixing ratios using the SOCE inventory and the bottom row shows the correlation using the FFDAS v2 inventory. We quantified the agreement between modelled and measured mixing ratios using the Pearson’s correlation coefficient (a measure of the linear correlation between the x and y variables), root mean square error (RMSE, the standard deviation of the differences between measured and modelled data) and mean bias (MB, an indication of the mean over or underestimation of modelled mixing ratios). It is

63 immediately clear that there is a stronger model-measurement correlation at the Downsview and Egbert sites (R > 0.75) relative to that of Hanlan’s Point or Turkey Point (R < 0.53). The difficulty with accurately simulating CO2 mixing ratios at Hanlan’s Point and Turkey Point may arise from their proximity to shorelines, Hanlan’s Point to Lake Ontario and Turkey Point to Lake Erie (see Figure 2.1). Differential heating of land versus water near these lakes creates pressure gradients driving unique circulation patterns (Burrows, 1991; Sills et al., 2011). These circulation patterns are difficult for models to capture and therefore may contribute to the relatively poor correlation observed at Hanlan’s Point and Turkey Point.

Figure 2.6: Scatter plot of the modelled and measured afternoon (12:00-16:00 EST) CO2 mixing ratios from January-March, 2016 at the four monitoring stations used in this study. The top and bottom panels show measurement-model correlation when the SOCE inventory and the FFDAS v2 inventory were used, respectively. The model vs. measurement Correlation Coefficient (R), root mean square error (RMSE) and mean bias (MB) (units: ppm) are provided within each panel. Solid lines are the standard major axis regression lines and dashed lines are 1:1 lines shown for reference.

64

It is also clear from Figure 2.6 that simulating CO2 mixing ratios at the Egbert and Turkey Point sites using either the FFDAS v2 or the SOCE inventory results in similar performance, likely because the emissions estimated by the two inventories are similar in the vicinity of these two rural sites (see also Figure 2.5). However at both the Downsview and Hanlan’s Point sites, using the SOCE inventory provided a slightly higher correlation and reduced RMSE and mean bias (MB) relative to using the FFDAS v2 inventory. The improvement by using the SOCE inventory is likely a result of both the improved spatial resolution (2.5 km versus 10 km), and therefore more accurate allocation of emissions to grid cells, and a better estimation of emission magnitudes, as large differences are shown in Figures 2.3 and A1.

2.4.6 Sectoral contributions to simulated CO2 mixing ratios

One of the major advantages of the SOCE inventory over the FFDAS v2 inventory is the availability of sectoral emission estimates. Figure 2.7 displays the sectoral percent contributions to diurnal CO2 mixing ratio enhancements (calculated as local CO2 mixing ratios above the MACC estimated background) for the Downsview station in February 2016 averaged by the day of week (Figures A7 and A8 displays the same for other months). This figure clearly demonstrates the importance of Area emissions (defined here as the sum of the Area + Residential natural gas combustion + Commercial natural gas combustion) to simulated CO2 mixing ratios, reaching ~80 % contribution in the early morning and late evening, consistent with times when emissions from home heating are the dominant source of CO2. Contributions from Area emissions decrease to ~35 % midday, which coincides with when emissions from other sources, such as On-road, gain importance. In the midday, emissions from the On-road sector can contribute up to ~70 %, which is consistent with transportation patterns of the times when the population is travelling to and from work and other activities. The relative contributions to CO2 mixing ratios from point source emissions is quite variable during the course of a day and week, but generally seems to increase in the early morning and evening and can contribute a significant portion of total CO2 emissions (up to ~20 %). Figure 2.7 indicates that biogenic sources of CO2 play a negligible role during January- March in the GTA. Recent studies, however, have shown the importance of the biospheric -2 -1 contribution (up to ~132-308 g CO2 km s ) to measured CO2 in urban environments during the growing season (Decina et al., 2016). Therefore, this finding supports the importance of modelling

CO2 in the wintertime in cities like the GTA to avoid complications associated with biospheric contributions. The new ability to understand the sectoral contributions to CO2 mixing ratios in the

65

GTA and southern Ontario has implications from a policy perspective; with recent initiatives to curb CO2 emissions, understanding from which sector the CO2 is being emitted could be useful to assess how effective applied mitigation efforts have been or where to target future efforts. These efforts could be complemented by running simulations with additional tracers, such as carbon 12 13 monoxide (CO), nitrogen oxides (NOx), or stable carbon isotopes ( C and C) to gain further insight.

Figure 2.7: Modelled sectoral percent contributions to diurnal local CO2 enhancement for February 2016 at Downsview averaged by day of week. Note: Area = Area + Residential natural gas combustion + Commercial natural gas combustion. (Time zone is EST).

2.5 Conclusions

We presented the SOCE inventory for southern Ontario and the GTA, the first, to our knowledge, high-resolution CO2 inventory for southern Ontario and for a Canadian metropolitan region. The

SOCE inventory provides CO2 emissions estimates at 2.5 km x 2.5 km spatial and hourly temporal resolution for seven sectors: Area, Residential natural gas combustion, Commercial natural gas combustion, Point, Marine, On-road and Off-road. When compared against two existing CO2

66 inventories available for southern Ontario, the EDGAR v4.2 and the FFDAS v2 inventories, using FLEXPART footprints, the SOCE inventory had improved model-measurement agreement; FFDAS v2 agreed well with in situ measurements, but the EDGAR v4.2 inventory systematically overestimated mixing ratios. We developed a model framework using the GEM-MACH chemistry-transport model on a high-resolution 2.5 km x 2.5 km grid coupled to the SOCE and

FFDAS v2 inventories for anthropogenic CO2 emissions and C-TESSEL for biogenic CO2 fluxes. We compared output simulations to observations made at four stations across southern Ontario and for three winter months, January – March, 2016. Model-measurement agreement was strong in the afternoon using both anthropogenic inventories, particularly at the Downsview and Egbert sites. Difficulty in capturing mixing ratios at the Hanlan’s Point and Turkey Point sites was hypothesized to be a result of their close proximity to shorelines (Lake Ontario and Lake Erie, respectively) and the model’s inability to capture the unique circulation patterns that occur in those environments. Generally, across all stations and months, simulations using the SOCE inventory resulted in higher model-measurement agreement than those using the FFDAS v2 inventory, quantified using R, RMSE and mean bias. In addition to improved agreement, the primary advantage of the SOCE inventory over the FFDAS v2 inventory is the sectoral breakdown of emissions; using average day of week diurnal mixing ratio enhancements, we were able to demonstrate that emissions from area sources can contribute >80 % of CO2 mixing ratio enhancements in the early morning and evening with on-road sources contributing >50 % midday. The applications of the SOCE inventory will likely show future utility in understanding the impacts of CO2 reduction efforts in southern Ontario and identify target areas requiring further improvement.

2.6 Contributions

The SOCE inventory was prepared by Stephanie C. Pugliese, with critical input from Felix Vogel and Jennifer Murphy. The CO inventory which the SOCE inventory is based upon was provided by Mike Moran, Junhua Zhang and Qiong Zheng. The GEM-MACH modelling analyses were performed by Shuzhan Ren and Craig Stroud. The ambient CO2 data were collected by Douglas Worthy and his team at Environment and Climate Change Canada. The MACC and C-TESSEL products used in our model simulations were provided by Gregoire Broquet. The data was analyzed and interpreted for publication by Stephanie C. Pugliese. The manuscript listed at the beginning

67 of this chapter was written by Stephanie C. Pugliese, with critical input from Jennifer Murphy, Felix Vogel and Mike Moran.

Acknowledgements

The authors are thankful to Robert Kessler, Michelle Ernst, Lauriant Giroux and Senen Racki for their efforts collecting the CO2 measurements at Environment and Climate Change Canada. They would also like to thank Elton Chan for providing the FLEXPART footprints and for Pegah Baratzadeh for help creating the SOCE inventory. Support for this project from the Natural Science and Engineering Research Council of Canada is gratefully acknowledged.

2.7 References

Bousetta, S., Balsamo, G., Beljaars, A., Panareda, A. A., Calvet, J. C., Jacobs, C., van den Hurk, B., Viterbo, P., Lafont, S., Dutra, E., Jarlan, L., Balzarolo, M., Papale, D., and van der Werf, G. (2013). Natural Land Carbon Dioxide Exchanges in the ECMWF Integrated Forecasting System: Implementation and Offline Validation. Journal of Geophysical Research: Atmospheres, 118, 5923-5946.

Breon, F. M., Broquet, G., Puygrenier, V., Chevallier, F., Xueref-Remy, I., Ramonet, M., Dieudonne, E., Lopez, M., Schmidt, M., Perrussel, O., and Cias, P. (2015). An Attempt at

Estimating Paris Area CO2 Emissions from Atmospheric Concentration Measurements. Atmospheric Chemistry and Physics, 15(1707), 1724.

Burrows, W. R. (1991). Objective Guidance for 0–24-Hour and 24–48-Hour Mesoscale Forecasts of Lake-Effect Snow using CART. American Meteorological Society, 6, 357-378.

C40 Cities. (2016). C40 Cities: Global Leadership on Climate Change. Retrieved 10/12, 2016, from http://www.c40.org/cities/toronto

Chan, D., Ishizawa, M., Higuchi, K., Maksyutov, S., and Chen, J. (2008). Seasonal CO2 Rectifier Effect and Large-Scale Extratropical Atmospheric Transport. Journal of Geophysical Research, 113(D17309) doi:doi:10.1029/2007JD009443

68

Côté, J., Desmarais, J. -., Gravel, S., Méthot, A., Patoine, A., Roch, M., and and Staniforth, A. (1998b). The Operational CMC/MRB Global Environment Multiscale (GEM) Model. Part II: Results. Monthly Weather Review, 126, 1397-1418.

Côté, J., Gravel, S., Méthot, A., Patoine, A., Roch, M., and and Staniforth, A. (1998a). The Operational CMC/MRB Global Environmental Multiscale (GEM) Model. Part 1: Design Considerations and Formulation. Monthly Weather Review, 126, 1373-1395.

Decina, S. M., Hutyra, L. R., Gately, C. K., Getson, J. M., Reinmann, A. B., Short Gianotti, A. G., and Templer, P. H. (2016). Soil Respiration Contributes Substantially to Urban Carbon Fluxes in the Greater Boston Area. Environmental Pollution, 212, 433-439.

EDGAR. (2010). Emission Database for Global Atmospheric Research Release Version 4.2 of the European Commission. Joint Research Centre (JRC)/Netherlands Environmental Assessment Agency (PBL). Retrieved 06/29, 2016, from http://edgar.jrc.ec.europa.eu

Environment Canada. (2011). Canadian Greenhouse Gas Measurement Program. Retrieved 10/07, 2016, from https://www.ec.gc.ca/mges-ghgm/

Environment Canada. (2012). National Inventory Report 1990-2010: Greenhouse Gas Sources and Sinks in Canada. Retrieved 10/07, 2016, from http://www.ec.gc.ca/ges-ghg/

Environment Canada. (2013). Manual for the Compilation of Canada's 2010 Air Pollutant Emissions. Catalogue no. En14-87/2013E-PDF. Retrieved 10/07, 2016, from http://www.ec.gc.ca/Publications/default.asp?lang=En&xml=8A08B403-C85E-4612-A958- 3F24446A61DB

Environment Canada. (2015). Reported Facility Greenhouse Gas Data. Retrieved 10/07, 2016, from http://www.ec.gc.ca/ges-ghg/donnees-data/index.cfm?lang=En

FFDAS. (2010). Fossil Fuel Data Assimilation System (FFDAS). Retrieved 06/29, 2016, from http://hpcg.purdue.edu/FFDAS/

Framework for Public Review and Engagement. (2007). Change is in the Air: Toronto’s Commitment to an Environmentally Sustainable Future. Retrieved 10/12, 2016, from www.toronto.ca/legdocs/mmis/2007/ex/bgrd/backgroundfile-2428.pdf

69

Gately, C. K., Hutyra, L. R., Sue Wing, I., and Brondfield, M. N. (2013). A Bottom-Up

Approach to on-Road CO2 Emissions Estimates: Improved Spatial Accuracy and Applications for Regional Planning. Environmental Science and Technology, 47, 2423-2430.

Gerbig, C., Korner, S., and Lin, J. C. (2008). Vertical Mixing in Atmospheric Tracer Transport Models: Error Characterization and Propagation. Atmospheric Chemistry and Physics, 8, 591- 602.

Gong, W., Makar, P. A., Zhang, J., Milbrandt, J., Gravel, S., Hayden, K. L., Macdonald, A. M., and Leaitch, W. R. (2015). Modelling Aerosol-Cloud-Meteorology Interaction: A Case Study with a Fully Coupled Air Quality Model (GEM-MACH). Atmospheric Environment, 115, 695- 715.

Gordon, M., Staebler, R. M., Liggio, J., Li, S., Wentzell, J., Lu, G., Lee, P., and Brook, J. R. (2012a). Measured and Modeled Variation in Pollutant Concentration Near Roadways. Atmopsheric Environment, 57, 138-145.

Gordon, M., Staebler, R. M., Liggio, J., Makar, P., Li, S., Wentzell, J., Lu, G., Lee, P., and Brook, J. R. (2012b). Measurements of Enhanced Turbulent Mixing Near Highways. Journal of Applied Meteorology and Climatology, 51, 1618-1632.

Gurney, K. R., Mendoza, D. L., Zhou, Y., Fischer, M. L., Miller, C. C., Geethakumar, S., and De

La Rue du Can, S. (2009). High Resolution Fossil Fuel Combustion CO2 Emission Fluxes for the United States. Environmental Science and Technology, 43, 5535-5541.

Gurney, K. R., Razlivanvov, I., Song, Y., Zhou, Y., Benes, B., and Abdul-Massih, M. (2012).

Quantification of Fossil Fuel CO2 Emissions on the Building/Street Scale for a Large U.S. City. Environmental Science and Technology, 46(21), 12194-12202.

IPCC-WG3. (2014). Climate Change 2014: Mitigation of Climate Change.Working Group III Contribution to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Retrieved 10/12, 2016, from http://www.ipcc.ch/report/ar5/wg3/

70

McDonald, B. C., McBride, Z. C., Martin, E. W., and Harley, R. A. (2014). High-Resolution Mapping of Motor Vehicle Carbon Dioxide Emissions. Journal of Geophysical Research: Atmospheres, 119, 5283-5298.

Miller, S. M., and Michalak, A. M. (2016). Constraining Sector-Specific CO2 and CH4 Emissions in the United States. Atmospheric Chemistry and Physics Dicsussions, doi:doi:10.5194/acp-2016-643

Moran, M., Zheng, Q., Zhang, J. and Pavlovic, R. (2015). RAQDPS Version 013: Upgrades to the CMC Operational Regional Air Quality Deterministic Prediction System Released in June 2015. Retrieved 10/07, 2016, from http://collaboration.cmc.ec.gc.ca/cmc/cmoi/product_guide/docs/lib/op_systems/doc_opchanges/ Technical_Note_GEM-MACH10_v1.5.3+SET2.1.1_Emissions_9Nov2015.pdf

Moran, M. D., Dastoor, A., & Morneau, G. (2013). Long-range transport of air pollutants and regional and global AQ modelling. In E. Taylor, & A. McMillan (Eds.), Air Quality Management - Canadian Perspectives on a Global Issue (pp. 69-98) Springer.

Oda, T. and Maksyutov, S. (2011). A very high-resolution (1 km x 1 km) global fossil fuel CO2 emission inventory derived using a point source database and satellite observations of nighttime lights. Atmospheric Chemistry and Physics, 11, 543-556.

Ontario. (2016). Ontario's Five Year Climate Change Action Plan 2016-2020. Retrieved 09/14, 2016, from https://www.ontario.ca/page/climate-change-action-plan

Patarasuk, R., Gurkey, K. R., O'Keeffe, D., Song, Y., Huang, J., Rao, P., Buchert, M., Lin, J. C.,

Mendoza, D., and Ehleringer, J. R. (2016). Urban High-Resolution Fossil Fuel CO2 Emissions Quantification and Exploration of Emission Drivers for Potential Policy Applications. Urban Ecosytems, 19(3), 1013-1039.

Pavlovic, R., Chen, J., Anderson, K., Moran, M. D., Beaulieu, P., Davignon, D., and Cousineau, S. (2016). The FireWork Air Quality Forecast System with Near-Real-Time Biomass Burning Emissions: Recent Developments and Evaluation of Performance for the 2015 North American Wildfire Season. Journal of the Air and Waste Management Association, 66(9), 819-841.

71

Pillai, D., Gerbig, C., Ahmadov, R., Rodenbeck, C., Kretschmer, R., Koch, T., Thompson, R.,

Neininger, B., and Lavrie, J. V. (2011). High Resolution Simulations of Atmospheric CO2 Over Complex Terrain - Representing the Ochsenkopf Mountain Tall Tower. Atmospheric Chemistry and Physics, 11, 7445-7464.

PIRD. (2016). Air Pollutant Emission Inventory Report (APEIR) 1990-2014. Retrieved 10/07, 2016, from http://www.ec.gc.ca/Air/89ED82E9-CA92-424E-9FD1- C57235A04CE0/Air%20Pollutant%20Emission%20Inventory%20Report%201990-2014.pdf

RWDI AIR Inc. (2009). 2007 Emissions Inventory Toronto Pearson International Airport, Toronto, Ontario.

RWDI AIR Inc. (2013). Billy Bishop Toronto City Airport (BBTCA) final report: Air quality review.

Sills, D., Brook, J., Levy, I., Makar, P., Zhang, J., and Taylor, P. (2011). Lake Breezes in the Southern Great Lakes Region and their Influence during BAQS-Met 2007. Atmospheric Chemistry and Physics, 11, 7955-7973.

Statistics Canada. (2007). Households and the Environment: Energy Use. Retrieved 12/04, 2012, from http://www.statcan.gc.ca/pub/11-526-s/2010001/t002-eng.pdf

Statistics Canada. (2009). Table 3.4: Total and Urban Land Area, 1996, 2001 (Modified) and 2006. Retrieved 10/12, 2016, from http://www.statcan.gc.ca/pub/92f0138m/2008001/t/4054949- eng.htm

Statistics Canada. (2011). Population, Urban and Rural, by Province and Territory (Canada). Retrieved 10/12, 2016, from http://www.statcan.gc.ca/tables-tableaux/sum- som/l01/cst01/demo62a-eng.htm

Statistics Canada. (2012a). Report on Energy Supply and Demand. Retrieved 12/04, 2015, from http://www.statcan.gc.ca/pub/57-003-x/2015001/t037-eng.pdf

Statistics Canada. (2012b). Focus on Geography Series, 2011 Census. Statistics Canada Catalogue no. 98-310-XWE2011004. Retrieved 08/13, 2013, from

72 https://www12.statcan.gc.ca/census-recensement/2011/as-sa/fogs-spg/Facts-cma- eng.cfm?LANG=Eng&GK=CMA&GC=535

Talbot, D., Moran, M. D., Bouchet, V., Creview, L. -., Menard, S., Kallur, A., & and the GEM- MACH Team. (2008). Development of a new canadian operational air quality forecast model. In C. Borrego, & A. I. Miranda (Eds.), Air Pollution Modeling and its Application XIX (pp. 470- 478) Springer Science and Business Media.

Wang, R., Tao, S., Ciais, P., Shen, H. Z., Huang, Y., Chen, H., Shen, G. F., Wang, B., Li, W., Zhang, Y. Y., Lu, Y., Zhu, D., Chen, Y. C., Liu, X. P., Wang, W. T., Wang, X. L., Liu, W. X., Li, B. G., and and Piao, S. L. (2013). High-Resolution Mapping of Combustion Processes and

Implications for CO2 Emissions. Atmospheric Chemistry and Physics, 13, 5189-5203.

WHO. (2015). Global Health Observatory (GHO) Data: Urban Population Growth. Retrieved 2/3, 2015, from http://www.who.int/gho/urban_health/situation_trends/urban_population_growth_text/en/

Worthy, D. E., Platt, A., Kessler, R., Ernst, M., Audette, C., & Racki, S. (2005). An update on the Canadian GHG measurement program, in: Report of the 12th WMO/IAEA meeting of experts on carbon dioxide concentration related tracer measurement techniques. (No. 162). Toronto, Canada: World Meteorological Organization Atmosphere Watch.

Zhang, J., Zheng, Q., Moran, M. D., Gordon, M., Liggio, J., Pakar, P., & Stroud, C. (2012). Improvements to SMOKE processing of Canadian on-road mobile emissions. Toronto, Canada.

Zhao, Y., Nielsen, C. P., and McElroy, M. B. (2012). China's CO2 Emissions Estimated from the Bottom Up: Recent Trends, Spatial Distributions and Quantification of Uncertainties. Atmospheric Environment, 59, 214-223.

2.8 Appendix A

Table A1: Facility-specific CO2:CO ratios for the point sector Specific Point Sources from 1A1a (NAICS Code: 2211) CO2 CO CO2:CO Name (kt) (kt) (kt CO2/kt CO) Brighton Beach 233.3 0.28 842 Bruce Power 69.0 0.03 2091 Cardinal Power 562.1 0.06 9527 Cochrane Generating Station 1113.3 0.34 3323 East Windsor Cogeneration Center 124.3 0.05 2345 Essar Power Canada ltd. 1502.4 0.48 3137 Goreway Power Station 1209.8 0.72 1680 Greenfield Energy Center 937.9 2.18 431 Halton Hills Generating Station 397.0 0.07 5839 Iroquois Falls Generating Station 321.0 0.07 4864 Kapuskasing Power Plant 136.6 0.03 4268 Kingston Cogen 332.0 0.05 6775 Kirkland Lake Generating Station 321.9 0.63 513 Lake Superior Power 203.1 0.06 3692 Lambton Generating Station 3289.9 1.41 2337 Lennox Generating Station 120.1 0.07 1819 Mississauga Cogeneration Plant 363.8 0.13 2756 Nanticoke Generating Station 8544.1 3.40 2513 North Bay 133.5 0.03 4170 Ottawa Health Sciences Center Cogeneration Facility 220.0 0.07 3013 Plasco Trail Road 3.2 0.02 154 Portlands Energy Centre 457.3 0.34 1333 Sarnia Cogen Plant 362.4 0.22 1618 Sarnia Regional Cogeneration Facility 1194.2 0.40 2971 St. Clair Energy Center 357.6 0.29 1216 Thorold Cogen Generating Station 305.5 0.20 1543 Tunis Power Plant 109.3 0.08 1401 Whitby Cogeneration 189.3 0.09 2175 Windsor Essex Cogeneration Plant 202.1 0.07 2846 Haggersville Mine 52.0 0.04 1406 Beachville Operation 487.6 0.24 2008

73 74

Dundas Operation 406.4 0.23 1744 Northern Lime Limited 90.8 0.09 1032 Copper Cliff Nickel Refinery 51.3 0.03 1654 Copper Cliff Smelter 201.1 0.10 1971

Specific Point Sources from 1A2f (NAICS Code: 3241) CO2 CO CO2:CO Name (kt) (kt) (kt CO2/kt CO) Mississauga Lubricants Center - Petro-Canada 365.04 0.08 4680 Nanticoke Refinery - Imperial Oil 1058.08 7.83 135 Sarnia Manufacturing Center 759.47 0.34 2234 Sarnia Refinery- Suncor 628.00 4.69 134 Sarnia Refinery Plant - Imperial Oil 1270.44 2.43 523

Specific Point Sources from 2C1 (NAICS Code: 3311) CO2 CO CO2:CO Name (kt) (kt) (kt CO2/kt CO) Dofasco Hamilton 4939.9 9.23 535 Essar Steel Algoma Inc. 2219.1 0.90 2468 Gerdau Whitby 83.7 0.28 295 Hamilton Works - U.S. Steel Canada 2674.7 2.45 1093 Ivaco Rolling Mills 74.5 0.20 369 Lake Erie Works - U.S. Steel Canada 1387.6 1.49 931

Table A2: Grouping of EDGAR v4.2 SNAP sectors to SOCE inventory sectors SOCE Sector EDGAR v4.2 SNAP Sectors Area 1A4 + (6A+6C) + 1A2 + 1A1a + 2C + 2A Point 1B2a + (4C+4D) + (2B+3) + 1A2 + 1A1a + 2C + 2A Marine 1A3d On-road 1A3b Off-road 1A3a + 1A3c + 1A3e + 1A3d

75

Downsview

Hanlan’s Point

Egbert

Turkey Point

Figure A1: Annual mean difference in anthropogenic CO2 emissions (units: g CO2/second/grid cell) between the SOCE and the FFDAS v2 inventories (SOCE – FFDAS v2) for a weekday in February 2010. Locations of in-situ measurements of CO2 are shown.

76

Figure A2: Diurnal profile of CO2 emissions (grams) estimated by the FFDAS v2 and SOCE inventories for January-March for the black-box domain in Figure 2.2a.

77

Figure A3: Time series of measured (blue) and modelled Downsview – TAO gradient CO2 mixing ratios. The black, red and gold lines are the modelled mixing ratios when using the EDGAR v4.2, FFDAS v2 and SOCE CO2 inventories, respectively.

78

Figure A4: Time series of measured (blue) and modelled January afternoon (12:00-16:00 EST) CO2 mixing ratios for the four sites used in this study. The red and gold markers are the modelled mixing ratios when using the SOCE CO2 inventory and the FFDAS v2 inventory, respectively.

79

Figure A5: Time series of measured (blue) and modelled March afternoon (12:00-16:00 EST) CO2 mixing ratios for the four sites used in this study. The red and gold markers are the modelled mixing ratios when using the SOCE CO2 inventory and the FFDAS v2 inventory, respectively.

80

Figure A6: Time series of the difference between modelled and measured CO2 mixing ratios in February (12:00-16:00 EST) for the four sites used in this study. The red and gold markers are the differences when using the SOCE CO2 inventory and the FFDAS v2 inventory, respectively.

Table A3: Average differences between modelled and measured CO2 mixing ratios from January – March (12:00-16:00 EST) for the four sites used in this study. Station Difference with SOCE Difference with FFDAS v2 inventory (ppm) inventory (ppm) Downsview 2.2 6.8 Egbert 4.2 4.5 Hanlan’s Point 6.2 11.1 Turkey Point 3.6 3.8

81

Figure A7: Modelled sectoral percent contributions to diurnal local CO2 enhancement for January 2016 at Downsview averaged by day of week. Note: Area = Area + Residential natural gas combustion + Commercial natural gas combustion. (Time zone is EST).

82

Figure A8: Modelled sectoral percent contributions to diurnal local CO2 enhancement for March 2016 at Downsview averaged by day of week. Note: Area = Area + Residential natural gas combustion + Commercial natural gas combustion. (Time zone is EST).

83

Chapter 3 13 Characterization of the δ C signatures of anthropogenic CO2 emissions in the Greater Toronto Area, Canada

Sections of this chapter appear in the following peer-reviewed publication:

Pugliese, S. C., Murphy, J. G., Vogel, F., and Worthy, D. (2016). Characterization of the δ13C signatures of anthropogenic CO2 emissions in the Greater Toronto Area, Canada, Applied Geochemistry, 83, 171-180.

3 Chapter 3

3.1 Abstract

In urban environments, carbon dioxide (CO2) is emitted from a variety of anthropogenic and 13 biogenic sources. The isotopic ratio δ CO2 has been widely used to source apportion CO2 as different sources have distinct isotopic fingerprints; the disadvantage of this technique is that 13 δ CO2 signatures are often spatially and temporally specific. We present a study characterizing 13 the δ CO2 signatures of the dominant anthropogenic sources of CO2 in the Greater Toronto Area (GTA). Refined gasoline and diesel fuel were sampled from various stations around the GTA in 13 three separate campaigns (April, July and November 2015) to assess the variability of their δ CO2 13 signatures. Mean winter δ CO2 signatures for refined gasoline were measured to be -27.58 ± 0.52 ‰ and -28.12 ± 0.43 ‰ (uncertainties represent the standard deviation in sample signatures) in April and November respectively while the mean summer signature was measured to be -28.09 ± 13 0.34 ‰. Diesel fuel samples δ CO2 signatures from the same campaign periods were measured to be -29.09 ± 0.34 ‰, -29.47 ± 0.29 ‰, and -29.28 ± 0.21 ‰, respectively. We hypothesize that inter-seasonal variability in signatures is likely a result of the use of different parent crude petroleum. We found no significant impacts from octane grade, fuel distributor or municipality the 13 fuel was purchased from on the measured δ CO2 signatures. Other transportation fuels that were 13 13 measured include dyed-diesel (δ CO2 signature = -29.3 ± 0.20 ‰) and jet fuel (δ CO2 signature = -29.5 ± 0.20 ‰), which were both different than measurements made for the same fuels in other locations globally. To account for emissions from residential and commercial heating and

84 electricity production, a variety of fuels were characterized in this study. The primary heating fuel 13 used in the GTA, natural gas, was measured to have a δ CH4 signature of -44.2 ± 0.20 ‰. 13 Anthracite coal (a fuel that was used prior to 2014 for electricity) was measured to have a δ CO2 signature of -23.8 ± 0.20 ‰, which is fairly consistent with coal samples measured elsewhere. With the transition in Ontario from coal to biomass-fueled power generation, softwood, hardwood 13 and mixed wood pellets were sampled from two distributors in the GTA and measured δ CO2 signatures were -25.0 ± 0.20 ‰, -26.8 ± 0.20 ‰, and -25.8 ± 0.20 ‰, respectively. Using these measured signatures, we performed a sensitivity analysis to quantify the precision required for 13 continuous ambient measurements to separately identify fuel type using δ CO2 as the sole tracer. The local signatures were also used in a mass balance calculation to quantify the relative contribution of two end-members (with the signatures of -44 ‰ and -28 ‰) to total measured CO2 in the GTA in 2014.

3.2 Introduction

Over the last three centuries, concentrations of atmospheric carbon dioxide (CO2) have been increasing, primarily as a result of fossil fuel combustion and deforestation (Hartmann et al., 2013). In response to international treaties, such as the Kyoto Protocol, Copenhagen Accord and the Paris

Agreement, there has been a recent drive to reduce or offset emissions of CO2 and other greenhouse gases (GHGs). However, even in heavily urbanized areas, CO2 is derived from a variety of biological (e.g. respiration from humans, soil and plants) and anthropogenic (e.g. residential and commercial heating, electricity production, vehicle exhaust) sources and is atmospherically transported across regional and national borders. Therefore, to independently evaluate whether regulations are being properly implemented, there is an increasing need to not only monitor ambient concentrations of CO2 but also identify the major sources and sink processes and quantify their relative contributions. Several tracers, such as carbon monoxide (CO), have been used in an effort to address this challenge, as combustion processes yield CO and CO2 in specific ratios according to fuel type and engine efficiency (Djuricin et al., 2010). Using CO as a tracer for anthropogenic CO2 is a relatively inexpensive technique, however analyses can be complicated by the fact that there are biogenic sources of CO as well as unrelated sink processes (Gamnitzer et al.,

2006; Vogel et al., 2010). An alternate approach to identify the sources of atmospheric CO2 is the

85 use of its isotopic composition. The most useful isotope to investigate the fossil fuel contribution 14 14 of atmospheric CO2 is radiocarbon, C, as fossil fuel combustion emits CO2 with no C because its half-life (5730 years) is much shorter than the age of fossil fuels (Djuricin et al., 2010). 14 Conversely, present day background and biogenic CO2 is enriched in C, enabling the ability to 14 easily trace fossil fuel-derived CO2. However, analysing even small samples for C is time- consuming and costly, making it impractical for continuous monitoring. An alternate approach is 13 12 to use the stable isotope composition of CO2 ( CO2 and CO2) where their ratio (relative to a known standard, Equation 3.1) can be used to identify the source as long as the unique end-member signatures are known.

13 퐶푠푎푚푝푙푒 12 13 퐶푠푎푚푝푙푒 훿 퐶 = ( 13 ) − 1 푥 1000 ‰ (E3.1) 퐶푠푡푎푛푑푎푟푑 12 퐶푠푡푎푛푑푎푟푑 ( )

The isotopic end-members of various CO2 sources are typically different enough that partitioning between various fossil fuel combustion activities as well as biological activities has been achieved successfully (Djuricin et al., 2010; Gorka and Lewicka-Szezebak, 2013; Kuc and Zimnoch, 1998; Pataki et al., 2003; Widory and Javoy, 2003; Zimnoch et al., 2012). A study demonstrating the utility of δ13C analyses was performed in 2013 by Gorka and Lewicka-Szczebak in Wroclaw, in southern Poland. In the two heating seasons of 2011 (January – March and October – December), 13 they were able to identify the influence of coal in the first heating season (with a source δ CO2 of -25.7 ‰) and the influence of diesel and gasoline in the second heating season, following the 13 opening of a new highway (with a source δ CO2 of –27.6 ‰) (Gorka and Lewicka-Szezebak, 13 2013). The primary difficulty associated with using δ C to source apportion CO2 is that isotopic signatures exhibit geographic variability and therefore a prerequisite of this analysis is to characterize local emissions (Bush et al., 2007; Widory and Javoy, 2003). However, there have been only a few studies globally that have considered the spatial and temporal variability of 13 anthropogenic δ CO2 signatures with even fewer, if any, exhaustive studies characterizing the 13 δ CO2 signature of sources in a Canadian metropolitan area.

86

In an effort to address this gap, this study was focused on evaluating the spatial and temporal variability of carbon isotopic signatures of fossil fuel combustion processes in the Greater Toronto Area (GTA) and southern Ontario. We characterized the carbon isotope ratio of locally supplied fuels for transportation (gasoline, diesel, dyed-diesel and jet fuel) and heating/electricity generation (natural gas, kerosene, stove oil, coal and wood pellets). Samples of gasoline and diesel fuel were collected from different municipalities within the GTA as well as in different seasons to evaluate the extent of spatial and temporal variability of the isotopic signatures. To our knowledge, 13 this is the first attempt at characterizing the δ CO2 signatures of anthropogenic CO2 sources in the 13 GTA as well as the first characterization of δ CO2 signatures of dyed diesel, kerosene, stove oil and wood pellets globally. Through a mass balance approximation, we demonstrate the utility of using these local carbon isotope signatures to quantify the contribution of different end-members to total CO2 annually. Currently, estimates of anthropogenic CO2 emissions in the GTA are made available by the EDGAR v.4.2 (EDGAR, 2010) and the FFDAS (FFDAS, 2010) inventories, 8 7 which have very different annual totals (1.42 x 10 and 6.04 x 10 tonnes CO2, respectively). Furthermore, the EDGAR inventory provides estimates of sectoral contributions in the GTA (~41 % natural gas and ~32 % gasoline + diesel) which is different from a locally derived estimate (~31 % natural gas and ~48 % gasoline + diesel, with gasoline accounting for ~73 % of that combined fraction)(Greening Greater Toronto, 2012). Therefore, we expect the results of this work will improve our ability to quantify the relative contributions of anthropogenic sources of CO2 in the GTA and help in reducing discrepancies between bottom-up inventories and top-down atmospheric calculations.

3.3 Methods

3.3.1 Study region

The GTA is located in southern Ontario on the northwest shore of Lake Ontario and is the largest urban area in Canada, as well as one of its most densely populated regions (945.4 persons per square kilometre) (Statistics Canada., 2012). The GTA comprises five municipalities, Halton, Durham, Peel, Toronto and York, which together have a population exceeding 6 million (Statistics Canada., 2012). The majority of residences in Ontario are heated by natural gas furnaces (~64 %) with the rest reliant primarily on electricity (~25 %) and minimally by propane (~3 %) and kerosene and stove oil (<1 %) (Statistics Canada, 2007; Statistics Canada, 2012). Prior to 2014,

87

~22 % of electricity in Ontario was derived from coal combustion (Ontario Ministry of Energy, 2015). When considering the GTA only, coal combustion accounted for an even larger fraction (~50 %), most of which was produced at the Nanticoke Generating Station, ~130 km south-west of Toronto (Independent Electricity System Operator (IESO), 2015). Following Ontario’s coal phase-out action plan, requiring all coal-fired power plants in the province to be decommissioned or transitioned into biomass-fueled power plants by 2014, electricity is now primarily produced via nuclear (~62 %), hydro (~22 %) and natural gas (~12 %) (Ontario Ministry of Energy, 2015).

3.3.2 Transportation fuels sample collection

Refined gasoline and diesel samples were collected from 38 different stations across the GTA (Section 3.4.1). Three campaigns were carried out to ensure both winter- and summer-grade gasoline were sampled (the transition between the two grades usually occurs in June and September of each year (U.S. Energy Information Agency, 2016)). Campaign #1 sampled winter- grade gasoline and diesel fuel from all stations on April 1 & 8, 2015 (hereafter referred to as “Winter 2015-A”), campaign #2 sampled summer-grade gasoline and diesel fuel from all stations on July 17 & 22, 2015 (hereafter referred to as “Summer 2015”), and campaign #3 sampled winter- Lake Ontario grade gasoline and diesel fuel from a limited subset of stations on November 19, 2015 (hereafter referred to as “Winter 2015-B”). During each campaign, 0.25-0.50 L of fuel was sampled into amber glass bottles from stations in Downtown Toronto as well as five suburban regions (Mississauga, Vaughan, Scarborough, Oakville and Brampton). In each region, grade 87 gasoline and diesel fuel were sampled from the three top distributors, Shell, Petro Canada and Esso, as well as three smaller retailers, Canadian Tire, Husky and Pioneer, except the Winter 2015-B campaign which sampled only the large distributors. In the Winter 2015-A campaign, grade 87, 89, 91 and 94 (where applicable) gasoline was sampled from each retailer in one of the municipalities.

13 Although in our study the refined gasoline was sampled and directly analyzed for its δ CO2 13 content, a previous study by Bush et al. in 2007 measured the δ CO2 signatures of both refined gasoline and vehicle exhaust. After analysis of six paired refined gasoline and exhaust samples, no 13 significant differences (paired t-test, p>0.05) were found in the measured δ CO2 signatures and therefore it was concluded that the combustion process imparts negligible further fractionation (Bush et al., 2007). Therefore, results from our work can be compared to studies performed

88 elsewhere analyzing either the refined fuel directly or the produced vehicle exhaust. If combustion does impart a small but consistent fractionation, as suggested by Widory (2006), the significant differences between the signatures of the samples measured in this work would remain consistent and therefore the qualitative results would still hold.

Other sampled transportation fuels include dyed-diesel (used for off-road activities, sampled from Alpha-Oil, Vaughan, Ontario) and jet fuel (used for aviation, sampled from Air Canada Express airlines at Toronto Pearson International Airport, Mississauga, Ontario). All samples of transportation fuels were transferred into 0.5 dram (ca 1.8 ml) vials.

3.3.3 Heating and electricity generation fuels sample collection

13 To characterize the δ CO2 signature from residential heating in the GTA, a 40 psi sample from a natural gas stove was collected into a metal canister (Restek TO-Can Air Sampling Canister). The inlet of the flask was positioned at the burner with the pilot extinguished. In addition, 0.5 L of stove-oil and kerosene were sampled at Alpha Oil and Lambert Oil (Vaughan, Ontario), respectively, and aliquots were transferred to 0.5 dram vials.

To characterize the emissions representative of those from the Nanticoke Generating Station when it was coal-fueled, we collected a sample of anthracite coal from Port Carbon, Pennsylvania, USA (an area where Ontario historically imported a significant portion of its coal from (Historica Canada, 2015)). Furthermore, to characterize potential future emissions that may result if Nanticoke is converted into a biomass-fueled power plant, we collected softwood, hardwood and softwood/hardwood mixed wood pellets from two retailers in the GTA.

13 3.3.4 Ambient CO2 and δ CO2 measurements

13 o o CO2 and δ CO2 measurements were collected at the Downsview facility (43.7 N, 79.4 W) as a part of Environment and Climate Change Canada (ECCC)’s Greenhouse Gas Observational Program. The measurement procedure follows a set of established principles and protocols outlined by a number of international agencies through recommendations of the Meeting on Carbon Dioxide, Other Greenhouse Gases, and Related Measurement Techniques, coordinated by the World Meteorological Organization (WMO) every 2 years. The measurement analytical setup 13 for both CO2 and δ CO2 is based on Cavity Ring-Down Spectroscopy (CRDS) (Picarro G2301 and G1101-i, respectively). Ambient measurements are made using two sample lines, both located

89 at the same height on the roof of the Downsview building and each line is dried to approximately

3 degrees. For CO2 measurements, the response function of the analyzer is determined against 3 calibrated standard tanks (low, mid and high) and the working and target standards are also included in the injection sequence. The gases are passed into the CRDS analyzer following a cascade pattern of high, working, mid, target and low, and is repeated six times for each calibration.

The CO2 measurements have a precision better than 0.1 ppm and are accurate within 0.2 ppm. For 13 δ CO2 measurements, details on specific instrument set-up, calibration traceability and measurement protocols have been outlined previously (Vogel et al., 2013). Briefly, calibration is achieved via a sampling strategy of measuring cycles every 30 minutes (target gases for 10 minutes and ambient or calibration gases for 20 minutes) (Vogel et al., 2013). To minimize long term drift 13 in the instrument, additional calibrations are run every 7 hours (Vogel et al., 2013). The δ CO2 measurements have an average precision of 0.15 ‰.

3.3.5 Isotope analyses

13 Analyses of δ CO2 in the solid and liquid fuels (gasoline, diesel, stove-oil, kerosene, coal, wood pellets) were performed using Continuous Flow-Elemental Analysis-Isotope Ratio Mass Spectrometry (CF-EA-IRMS) at the Isotope Science Lab – University of Calgary (ISL-UofC). A Finnigan Mat Delta+XL mass spectrometer is interfaced with a Costech 4010 elemental analyzer (ISL-UofC, 2016). The samples (both liquid and solid) are packed into tin cups (with the liquid samples packed specifically into smooth walled tin cups with a quartz fibre pad inside to limit capillary movement up the walls) and cold sealed. The cups are dropped onto a quartz tube combustion column maintained at 1020oC and flash-combustion is achieved by injecting a pulse of oxygen (O2) gas at the time of sample drop. Helium carries the eluent gases through a cupric o oxide (CuO) reduction column at 650 C. Separation of N2 and CO2 is achieved using gas 13 chromatography prior to entering the ion source of the mass spectrometer. δ CO2 is determined by comparing the sample peaks areas to six international reference peak areas (USGS 24, IAEA- 13 CH-6, IAEA-CH-7, NBS 22, USGS 40, and USGS 41), 4.1; δ CO2 values are referenced to the VPDB scale. The precision and accuracy of these measurements was ± 0.20 ‰ (ISL-UofC, 2016).

The flask containing the natural gas sample was diluted with zero air (roughly by a factor of 6) 13 and the δ CH4 was characterized using a cavity ring down spectroscopy (CRDS) analyzer (Picarro G2201-i) at ECCC; the signature is referenced to the VPDB scale.

90

3.3.6 Mass balance approximations

Following Equation 3.2, which equates the measured atmospheric CO2 concentration (Ct) to the sum of the background concentration (Cb) and an additional concentration produced by a local source (Cs), Keeling (Keeling, 1958; Keeling, 1961) applied isotopic mass balance and derived a linear mixing model, Equation 3.3.

Ct = Cb + Cs (E3.2)

13 13 13 13 δ Ct = Cb(δ Cb – δ Cs)(1/Ct) + δ CS (E3.3)

13 13 The advantage of this method is that Cb, δ Cb and Cs can remain unknown when solving for δ CS 13 13 as the intercept of a graph of δ Ct vs. 1/Ct. Equation 3.3 was used in this study to estimate δ CO2 of locally sourced CO2 for each day in 2014. . The Keeling plot intercept was only considered if the R2 of the regression was > 0.95.

We extended the Keeling plot analysis to partition the total source (Cs) into two component parts:

End-member #1 (C1) and End-member #2 (C2). We do not include the influence of coal in this approximation because following Ontario’s coal phase-out action plan, all coal-fired power plants in the province were decommissioned by 2014. We calculate the mixing ratio and isotopic th signature of background CO2 by averaging all measurements below the 5 percentile of CO2 over a moving window of 48 hours (Ammoura et al., 2016). Therefore, to solve for the remaining contribution of C1 and C2, we applied the following mass balance:

Ct = Cb + C1 + C2 (E3.4)

13 13 13 13 δ Ct·Ct = δ Cb·Cb + δ C1·C1 + δ C2·C2 (E3.5)

13 End-members for δ CO2 were estimated from measurements made in this study, described in Sections 3.4.1 and 3.4.2, and assumed to be -44 ‰ and -28 ‰, for natural gas and non-natural gas, respectively. To perform seasonal analyses, we analyzed historical climate data collected by ECCC and categorized May through September inclusive as summer months and October through April inclusive as winter months.

91

3.3.7 Statistical analyses

Statistical analyses were performed on Igor Pro Scientific Software v.6.34A. Statistical tests included the standard t-test and the one-way analysis of variance (ANOVA) test where the null hypothesis (Ho) being tested is that µ1 = µ2 = µn, where µ is the mean of the individual distributions. When the ANOVA test rejected Ho, a multiple comparison Tukey test and the Newman-Keuls test were used to identify the distributions that are different from each other. Results with p < 0.05 are considered significant.

3.4 Results and Discussion

3.4.1 The carbon isotopic composition of transportation fuels

Grade 87 gasoline obtained from local stations in Winter 2015-A and Winter 2015-B campaigns 13 had significantly different (ANOVA, p<0.05) mean δ CO2 signatures of -27.58 ± 0.52 ‰ and - 28.12 ± 0.43 ‰ respectively (uncertainties represent the standard deviation in the signatures from a particular group of samples), Figures 3.1 and 3.2. In comparison, Summer 2015 grade 87 gasoline 13 had a mean δ CO2 signature of -28.09 ± 0.34 ‰, Figures 3.1 and 3.2, which was significantly different to that of Winter 2015-A (ANOVA, p<0.05) but similar to that of Winter 2015-B (ANOVA, p>0.05). Inter-seasonal differences are likely explained by varying sources of crude petroleum with different geographic origin (Bush et al., 2007; Yeh and Epstein, 1981). We 13 hypothesized that the δ CO2 signatures may also be impacted by the variations in blends of summer versus winter gasoline grades, as regions with large seasonal variation in temperature often have gasoline with different seasonal compositions to alter the Reid Vapour Pressure (RVP) of the fuel (Leonard, 2007). However, the results of our study suggest that this impact on the 13 13 overall δ CO2 signature is likely small, as made evident by the statistically similar δ CO2 13 signature of Summer 2015 and Winter 2015-B fuels. The differences in the gasoline δ CO2 signature of Winter 2015-A and the other two campaigns is therefore likely due to a change in the parent crude petroleum used to produce the gasoline.

92

13 Figure 3.1: Summary of δ C signatures of all fuels. The boxes show the inter-quartile range and median values of each sample and the whisker top and bottom show the 90th and 10th percentiles, respectively. Median values are also reported in parentheses on the right of the figure. Error bars on non-boxed data are 1 sigma of standards, 0.20 ‰.

93

Winter 2015-A

Summer 2015

Winter 2015-B Lake Ontario Lake Ontario

13 Figure 3.2: Map of gasoline (grade 87) (left) and diesel (middle) δ CO2 signatures and their respective histogram (right) for the Winter 2015-A, Summer 2015 and Winter 2015-B sampling campaigns.

Our results are similar to those found by Bush et al. (2007) in Salt Lake City, Utah, USA where 13 different δ CO2 signatures of vehicle exhaust were measured in winter and summer (-27.9 ± 0.08

94

‰ and -28.5 ± 0.04 ‰, respectively). Despite this, our results differ from mean literature values 13 reported for refined gasoline elsewhere; a season unspecific mean δ CO2 signature of -31.7 ‰ was measured in Wroclaw, Poland (Gorka et al., 2011), -28.8 ‰ in Paris, France (Widory and Javoy, 2003), -30.1 ± 0.3 ‰ in Lithuania (Masalaite et al., 2012) and -26.6 ± 0.1 ‰ in Tokyo, Japan (Masalaite et al., 2012). This variability is likely a result of differences in the parent crude petroleum from different geographic source locations. The differences between the gasoline 13 δ CO2 signatures measured in the GTA and elsewhere confirms the need for local (and timely) characterization to accurately perform CO2 source apportionment analyses.

In the Winter 2015-A campaign, gasoline of different octane contents (87, 89 and 91) was sampled 13 to assess the potential impact on the δ CO2 signature, Figure 3.3. Although there is small 13 variability, mean δ CO2 signatures across octane contents were not significantly different (ANOVA, p > 0.05).

13 Figure 3.3: Comparing δ CO2 signatures of gasoline with varying octane contents. The boxes show the inter-quartile range and median values of each gasoline sample and the whisker top and bottom show the 90th and 10th percentiles, respectively.

Diesel fuel obtained from local stations in Winter 2015-A and Winter 2015-B campaigns had 13 significantly different (ANOVA, p<0.05) mean δ CO2 signatures of -29.09± 0.34 ‰ and -29.47 ± 0.29 ‰ respectively, Figures 3.1 and 3.2. In comparison, Summer 2015 diesel fuel had a mean

95

13 δ CO2 signature of -29.28 ± 0.21 ‰, Figures 3.1 and 3.2, which was significantly different to that of both Winter 2015-A and Winter 2015-B (ANOVA, p<0.05). To our knowledge, this is the first 13 account where summer and winter diesel fuel δ CO2 signatures were measured and found to be 13 significantly different. Similar to the differences in gasoline δ CO2 signatures, this is likely a 13 result in different parent crude petroleum and not due to seasonal blends of the fuel. The δ CO2 signatures measured for diesel fuel in this study differ from mean literature values reported 13 elsewhere; a season unspecific mean δ CO2 signature of -31.9 ‰ was measured in Wroclaw, Poland (Gorka et al., 2011), -28.9 ‰ in Paris, France (Widory and Javoy, 2003), -31.6 ‰ in Lithuania (Masalaite et al., 2012) and -27.2 ‰ in China (Masalaite et al., 2012).

13 When considering both gasoline and diesel, the mean δ CO2 signature of winter and summer gasoline are significantly different than those of winter and summer diesel, respectively (ANOVA, p<0.05). Figure 3.2 displays the histograms of the gasoline and diesel data in each campaign and 13 it is apparent there is no overlap in the δ CO2 signatures. To our knowledge, this is the first account 13 where the δ CO2 signatures of gasoline and diesel fuels are distinct; a similar histogram created 13 from the gasoline and diesel δ CO2 signatures measured in Paris, France show complete overlap of the data (Widory and Javoy, 2003). If an instrument with sufficiently high precision is used, it may be possible that differences in these signatures will allow estimation of CO2 derived from gasoline and diesel independently, rather than as a combined single source.

Figure 3.4 (displaying Winter 2015-A as an example) depicts small variability, but no significant 13 difference in the δ CO2 signatures of gasoline or diesel purchased at any of the distributors (ANOVA, p>0.05) for the Winter 2015-A or Summer 2015 campaigns. The analysis was run for the Winter 2015-B campaign and in this period, there were small but significant differences in the signatures of gasoline from Petro-Canada and Esso (ANOVA and Tukey Test, p<0.05). These differences may be a result of the two distributors using crude petroleum from varying geographic origin or refinery operation.

96

13 Figure 3.4: Comparing δ CO2 signatures of gasoline (top) and diesel (bottom) from different distributors in the Winter 2015-A campaign. The boxes show the inter-quartile range and median values of each sample and the whisker top and bottom show the 90th and 10th percentiles, respectively.

Consideration was also given to determine if gasoline or diesel fuel purchased from different 13 municipalities in the GTA affected the δ CO2 signatures, Figure 3.2. For all three sampling 13 campaigns, there is no difference in the δ CO2 signatures of gasoline across municipalities in the

97

GTA (ANOVA, p>0.05). However there was a small but significant difference for diesel fuel 13 across municipalities for the Winter 2015-A campaign, where the measured δ CO2 signatures in Oakville were different from those measured Downtown, Scarborough and Vaughan (ANOVA and Tukey Test, p<0.05).

In southern Ontario, CO2 from off-road activities can contribute ~5 % of total emissions (EDGAR, 13 2010) and therefore in an effort to capture this, dyed-diesel was also sampled and its δ CO2 signature was measured to be -29.3 ± 0.20 ‰. This signature is consistent with on-road diesel fuel 13 from the same campaign, Winter 2015-B δ CO2 signature of -29.47 ± 0.29 ‰ which was to be expected, as the two are essentially the same products but with an added red-dye for identification purposes. The only other known reference to the characterization of off-road diesel in Lithuania is consistent with our study where in 2009 “diesel for farmers” and on-road diesel were measured to 13 have δ CO2 signatures of -31.6 ± 0.1 ‰ and -31.4 ± 0.5 ‰, respectively (Masalaite et al., 2012).

The final transportation fuel that was characterized in this study was jet fuel, as CO2 emissions from air traffic can contribute ~10 % of total consumption-based emissions of the city (Greening Greater Toronto, 2012). Furthermore, these emissions extend beyond the borders of the GTA as Toronto Pearson International Airport has on average > 400,000 departures each day and therefore jet CO2 will be emitted all over North America and globally and can contribute to the global 13 background. Jet fuel was measured to have a δ CO2 signature of -29.5 ± 0.20 ‰ which is more depleted in 13C relative to jet fuels measured elsewhere (e.g. -27 ‰ was measured in Charleston, South Carolina (Landmeyer et al., 1996)).

3.4.2 The carbon isotopic composition of heating and electricity generation fuels

With the majority of residences in Ontario heated by natural gas combustion (~64 %), emissions from this process are estimated to contribute ~31 % of local CO2 emissions in the Greater Toronto Area (Greening Greater Toronto, 2012). Natural gas obtained from a residence in the City of 13 Toronto in February 2016 had a mean δ CH4 signature of -44.2 ± 0.20 ‰, Figure 3.1. This value 13 compares well to a natural gas δ CH4 signature of -42.7 ± 0.20 ‰ measured in Toronto in 2015 (M. Lopez, personal communication). Similar to the gasoline/diesel analyses, we assume that the combustion process imparts no further fractionation and therefore combusted natural gas would 13 have a similar δ CO2 signature. The GTA natural gas combustion signature is somewhat

98

13 consistent with mean literature values reported elsewhere; a season unspecific mean δ CO2 signature of -42 ‰ was measured in Dallas, Texas (Clark-Thorne and Yapp, 2003), -39.1 ‰ from Paris, France (Widory and Javoy, 2003) and -44 ‰ as the global average (Andres et al., 2000).

A small proportion of Ontario’s residents heat their homes using kerosene and stove oil (<1 %) 13 13 (Statistics Canada, 2012) and therefore their δ CO2 signatures were measured. The δ CO2 signature of kerosene was measured to be -29.3 ± 0.20 ‰ and that of stove oil was measured to be 13 -28.8 ± 0.20 ‰, Figure 3.1. To our knowledge, this is the first measurement of the δ CO2 signature of kerosene and stove oil. While their use in developed regions like Ontario may be limited, this information may be valuable to help understand emissions of CO2 in developing regions where their use for cooking and lighting is widespread (Lam et al., 2012).

Prior to the 2014 coal phase-out in Ontario, the majority of electricity used in the GTA was produced at the coal-fired Nanticoke Generating Station, 130 km to the southwest. While Nanticoke is not located in the GTA, the prevailing westerly winds in Ontario often meant the GTA was impacted by its emissions. Therefore, in an attempt to characterize these emissions for 13 retrospective analyses, a sample of anthracite coal was measured to have a δ CO2 signature of - 23.8 ± 0.2 ‰, Figure 3.1. This is fairly consistent with mean literature coal signatures reported elsewhere; the accepted global value for coal combustion is -24.1 ± 0.3 ‰ regardless of source, as over 80 % of coal is composed of humic material or type III kerogen which has little variability in carbon isotopic composition (Andres et al., 2000). To characterize possible future emissions from Nanticoke when it may be used as a biomass-fueled facility, we characterized softwood, hardwood 13 and a softwood/hardwood mixture of wood pellets; δ CO2 signatures were found to be -25.0 ± 0.2 ‰, -26.8 ± 0.2 ‰ and -25.8 ± 0.2 ‰ respectively. The softwood pellets are certified by the Pellet Fuels Institute (PFI) with similar guidelines to those upheld by the Ontario Power Generation, and therefore the softwood pellets are likely the closest in composition to those that would be used at the Nanticoke Station. To our knowledge, this is the first reported characterization of wood pellets and therefore may be valuable to understand sources of CO2 emissions elsewhere.

99

13 3.4.3 Sensitivity case study using the GTA δ CO2 signatures

13 We performed a sensitivity analysis to evaluate whether the measured δ CO2 signatures of the dominant anthropogenic sources of CO2 in the GTA were different enough that they could be uniquely identified through source apportionment in a theoretical plume measured at a single point 13 in time. After analyzing the variability in ambient CO2 and CO2 data collected at Downsview station, we assume in our analysis that background air entering the GTA during this plume event 13 has an average CO2 concentration of 400 ppm and an average δ CO2 signature of -9 ‰. If we observe a plume with a CO2 enhancement of 20 ppm above the background, a common occurrence 13 in the GTA, and the source is solely from natural gas combustion (with the δ CO2 signature 13 measured in this study, -44.2 ‰), we would expect to measure a δ CO2 signature of -10.68 ‰. Conversely, if the source of the plume was solely from gasoline or diesel combustion, an observed 13 δ CO2 signature of -9.88 ‰ and -9.96 ‰ respectively would be expected. This exercise 13 demonstrates that with the ability of an instrument to continuously measure atmospheric δ CO2 at a precision of ±0.15 ‰, we would be able to distinguish between a plume arising solely from natural gas combustion and one from gasoline or diesel, but we would not be able to separate plumes from gasoline and diesel. It may be useful in this case to use other tracers, such as CO to calculate CO:CO2 ratios, which are quite variable between gasoline and diesel vehicles (in Ile de

France, 37.44 and 1.41 ppb CO/ppm CO2, respectively (Ammoura et al., 2014)).

In the GTA, Picarro analyzers at the Downsview facility (Section 4.3.4) as well at other locations 13 around the city provide continuous ambient CO2 and δ CO2. We therefore want to identify whether the precision of these instruments is sufficient to distinguish different emissions sources at the city scale rather than in a single plume. Our idealized plume analysis was extended to incorporate emissions from the entire southern Ontario domain where the FFDAS CO2 inventory -2 -1 7 estimates the intensity of emissions to be ~0.30 tonnes CO2 km h (6.04x10 tonnes of CO2 is 4 2 released annually in a 2.29x10 km area). Assuming the CO2 is emitted into a well-mixed 1 km -1 deep boundary layer, and the average wind speed in the area is 10 km h , we would observe a CO2 concentration elevated by ~0.93 ppm in the center of the city after an air mass has spent 6 hours in the domain. If the source of the emissions was solely natural gas, gasoline or diesel combustion, 13 we would expect to observe δ CO2 signatures of -9.08 ‰, -9.04 ‰ and -9.05 ‰, respectively, compared to the background signature of -9 ‰. It is clear that in this analysis, the precision of the continuous monitoring instrument limits our ability to distinguish between CO2 that was produced

100

13 from natural gas, gasoline or diesel when using δ CO2 as the sole tracer. Based on our calculation, 13 to successfully use δ CO2 to source apportion CO2, an enhancement of ~2.7 ppm over the domain would be necessary to have a detectable signal in the isotopic signature with a signal-to-noise ratio of greater than 1. A more realistic analysis would emit CO2 into the box from a combination of gasoline, diesel and natural gas combustion; if we assume that in the GTA, the mixture has a 13 δ CO2 signature of -34.38 ‰ (from 40 % natural gas combustion, 50 % gasoline combustion and 13 10 % diesel combustion) the observed δ CO2 signature would be -9.06 ‰. Keeling plots and other tracers would then have to be used to identify the mix of sources.

13 These sensitivity analyses demonstrate that in order to use δ CO2 as a tool to distinguish various sources of CO2, not only do the local signatures of sources need to be known, but there has to be a large enough enhancement of CO2 in the domain to be able to overcome precision limits of instruments. Prior to commencing a study like the one performed here, analyses like these would 13 be useful to determine if the sole use of δ CO2 is sufficient to source apportion CO2 or if additional tracers need to be considered. However, this simplified case study makes two important assumptions: the CO2 input from the background is constant and the CO2 input from the biosphere is negligible in the plumes we are considering. While this assumption may be appropriate if considering a plume at a single point in time during the winter season, this study as well as previous 13 research have demonstrated that δ CO2 shows distinct seasonal cycles as a result of differential fossil fuel use in winter versus summer (Bush et al., 2007). In addition, even in heavily urbanized areas, soil respiration can contribute emissions of a comparable magnitude to those from fossil fuel -2 -1 derived CO2 (FFCO2) ~3-7 µmolCO2 m s during the growing season (Decina et al., 2016). Therefore, making these simplified assumptions of constant background and negligible biospheric

CO2 contributions when performing source apportion analyses on timescales longer than proposed here (a plume at a single point in time) could introduce significant error.

3.4.4 Simplified isotope tracer based CO2 partitioning

13 Using the Keeling plot linear mixing model, Equation 3.3, we calculated δ Cs for each day in 2014, Figure 3.5.

101

13 Figure 3.5: Daily carbon isotopic signatures of CO2 source (δ Cs) in the GTA for 2014. The top and bottom shaded regions on the graph show the approximate range of signatures for gasoline and/or soil respiration and natural gas combustion, respectively. Error bars represent the standard deviation of the y-intercept of the Keeling curve.

Of the possible 365 days in 2014, signatures were reported for 82 days (data was not available for 54 days and data did not meet the selection criteria (Section 3.3.6) for the rest of the periods). There is higher variability in wintertime signatures (-35 to -45 ‰), than those observed during the 13 summer (-25 to -33 ‰). There is a distinct seasonal cycle in δ Cs, with the median signature in the summer approximately 10-15 ‰ more enriched in 13C than that of winter. Depleted signatures in the winter likely correspond to increased consumption of natural gas for residential and commercial heating, and the relatively enriched signatures in the summer associated with increased biosphere respiration (~-24 ‰ (Searle et al., 2012)), and/or a larger proportion of fossil fuel CO2 13 being derived from transportation fuels. There are days in the summer where δ Cs is heavier than 13 the δ CO2 of gasoline or biospheric respiration. We can attribute this occurrence to either transport 13 13 from the USA (as was observed on July 6, 2014 (δ Cs = -26.2 ‰)), where coal (δ CO2 = -23.8 ‰) is used for electricity production and/or transport from northern Ontario (as was observed on 13 13 August 18, 2014 (δ Cs = -23.3 ‰)), where the biogenic δ CO2 of rural/remote regions can be ~- 24 ‰ (Searle et al., 2012).

13 To further investigate the relative contribution of different end-members to δ Cs, we used chemical mass balance (CMB) (Equations 3.4 and 3.5) to partition CO2. Signatures for “End- member #1” and “End-member #2” were assigned to be -44 ‰ and -28 ‰ to represent natural gas

102 and non-natural gas, respectively. Figure 3.6 shows histograms depicting the contribution of each end-member normalized to probability density during winter and summer 2014.

Figure 3.6: Histograms of 2014 winter (top) and summer (bottom) contributions of CO2 from End- member #1 and End-member #2 in the GTA

It is evident from the summer histogram that the contribution of End-member #1 to measured CO2 is minimized while End-member #2 contribution is maximized; End-member #2 accounted for >

103

13 50 % of the total δ Cs on 83 % of the hours in 2014. Conversely, in the winter months, the relative contribution of each end-member is fairly consistent, with End-member #2 accounting for > 50 % 13 of the total δ Cs on only 56 % of the hours. If we consider the entire year of 2014, our calculations suggest that End-member #1 contributes an average of ~38 % to total CO2. Based on our study, 13 we are confident that End-member #1 represents natural gas (δ CO2 = -44.2 ‰) and therefore our estimate agrees well with both the EDGAR v.4.2 and the Greening Greater Toronto CO2 inventories where average annual natural gas contribution is ~ 41 % and ~ 31 % respectively for the GTA (EDGAR, 2010; Greening Greater Toronto, 2012). However, the contribution of End- member #1 vs. End-member #2 in our analysis has a high degree of diurnal and intra-annual variability (e.g. overnight and winter End-member #1 contributions can reach 99.9 %) and therefore in order to accurately estimate average annual contributions, continuous long-term in- 13 situ collection of CO2 and CO2 data is necessary.

While we can confidently identify End-member #1 as natural gas, we cannot identify whether the change in contribution of End-member #2 between seasons is a result of increased biospheric activity in the summer, an increase in the relative proportion of transportation fuels, or a combination of the two. Soil respiration in other urban environments has been measured to have a 13 δ CO2 signature of -27.8 ± 1.50 ‰ (Jasek et al., 2014) and the average gasoline signature measured in this study was -27.9 ± 0.76 ‰, therefore the two are not distinguishable based on carbon isotopes alone. To make this distinction, other tracers, such as 14C, 18O or CO, would be useful. When the same calculations were performed for 2013 data, the relative contribution of each 13 end-member was fairly consistent with 2014 (End-member #2 accounted for > 50 % of total δ Cs on 78 % of the mornings in summer and 63 % of the mornings in winter). This type of analysis shows particular utility to estimate the inter-seasonal and inter-annual variability of the contribution of each end-member and how it correlates with policy regulations in the region.

Lake Ontario Lake Ontario

104

3.5 Conclusion

13 Our study performed a detailed characterization of the δ CO2 signatures of various transportation and heating and electricity generation fuels in the GTA. Winter refined gasoline was measured to 13 have a δ CO2 signature of -27.58 ± 0.52 ‰ and -28.12 ± 0.43 ‰ in Winter 2015-A and Winter 2015-B, respectively, while the Summer 2015 signature was measured to be -28.09 ± 0.34 ‰. Inter-seasonal differences are likely attributed to use of parent crude petroleum from different geographic origin and not to the differing compositions of winter and summer blends of gasoline. 13 Octane grade was not found to have a significant impact on the measured δ CO2 signatures. Diesel 13 13 fuel was relatively more depleted in C than gasoline (δ CO2 signatures measured in Winter 2015-A, Winter 2015-B and Summer 2015 were -29.09 ± 0.34 ‰, -29.47± 0.29 ‰, and -29.28 ± 0.21 ‰, respectively) and to the authors’ knowledge, this is the first account where gasoline and 13 diesel have distinctly different δ CO2 signatures, potentially allowing for independent source attribution. Based on our analyses of gasoline and diesel carbon isotopic composition in the GTA, it is clear that variability of signatures from different municipalities within the GTA or different distributors is generally small and not significant however the seasonal variability can be large and 13 significant. Therefore, if δ CO2 signatures are to be used to source apportion CO2 in other regions, our study suggests to characterize fuels in the same season the analyses will be made. To account 13 for off-road transportation emissions, dyed-diesel and jet fuel were found to have δ CO2 signatures of -29.3 ± 0.20 ‰ and -29.5 ± 0.20 ‰, respectively. In addition to transportation, residential and commercial heating and electricity production are large sources of CO2 to the atmosphere in the GTA and therefore to account for current emissions, natural gas was sampled 13 and the δ CH4 signature was measured to be -44.2 ±0.20 ‰. Since natural gas combustion was not the dominant source of electricity to the GTA prior to 2014, anthracite coal was also measured 13 and found to have a δ CO2 signature of -23.8 ± 0.20 ‰. The results of this study demonstrated 13 that mean δ CO2 signatures of transportation and heating and electricity generation fuels in the GTA are significantly different from each other and those measured in other areas. These qualitative results hold true even if the combustion process imparts further fractionation to the samples (Widory, 2006).

13 By using the δ CO2 signatures reported in this study, we were able to estimate the contribution of two end-members (with signatures of -44 ‰ and -28 ‰) to total CO2 measured in the GTA. We found that End-member #2 dominates in the summertime (contributing > 50 % of the total

105 signature on 83 % of hours) but End-member #1 contribution increases in the wintertime, likely following the increase in natural gas combustion for heating. Our study emphasizes the utility of using carbon isotopes to estimate the contribution of sources to measured CO2 but also highlights the possibility for bias in estimates and the necessity for local characterization to reduce the over- or under-estimation of CO2 source apportionment.

3.6 Contributions

The samples characterized for their δ13C signature were collected by Stephanie C. Pugliese and 12 13 sent for analysis to the ISL-UofC for analysis. The ambient CO2 and CO2 data were collected by Douglas Worthy and his team at Environment and Climate Change Canada. The data was analyzed and interpreted for publication by Stephanie C. Pugliese. The manuscript listed at the beginning of this chapter was written by Stephanie C. Pugliese, with critical input from Jennifer Murphy and Felix Vogel.

Acknowledgements

Support for this project from the Natural Science and Engineering Research Council of Canada is gratefully acknowledged.

3.7 References

Ammoura, L., Xueref-Remy, I., Gros, V., Baudic, A., Bonsang, B., Petit, J. -., Perrussel, O., Bonnaire, N., Sciare, J., and Chevallier, F. (2014). Atmospheric Measurements of Ratios between CO2 and Co-Emitted Species from Traffic: A Tunnel Study in the Paris Megacity. Atmospheric Chemistry and Physics, 14, 12871-12882.

Ammoura, L., Xueref-Remy, I., Vogel, F., Gros, V., Baudic, A., Bonsang, B., Delmotte, M., Te, Y., and Chevallier, F. (2016). Exploiting Stagnant Conditions to Derive Robust Emission Ratio

Estimates for CO2, CO and Volatile Organic Compounds in Paris. Atmospheric Chemistry and Physics, 16, 15653-15664.

Andres, R. J., Marland, G., Boden, T., & Bischof, S. (2000). Carbon dioxide emissions from fossil fuel composition and cement manufacture, 1751-1991, and an estimate of their isotopic

106 composition and latitudinal distribution. In T. M. L. Wigley, & D. S. Schimel (Eds.), The Carbon Cycle (pp. 53-62). Cambridge: Cambridge University Press.

Bush, S. E., Pataki, D. E., and Ehleringer, J. R. (2007). Sources of Variation in D13C of Fossil Fuel Emissions in Salt Lake City, USA. Applied Geochemistry, 22, 715-723.

Clark-Thorne, S. T., and Yapp, C. J. (2003). Stable Carbon Isotope Constraints on Mixing and

Mass Balance of CO2 in an Urban Atmosphere: Dallas Metropolitan Area, Texas, USA. Applied Geochemistry, 18, 75-95.

Decina, S. M., Hutyra, L. R., Gately, C. K., Getson, J. M., Reinmann, A. B., Short Gianotti, A. G., and Templer, P. H. (2016). Soil Respiration Contributes Substantially to Urban Carbon Fluxes in the Greater Boston Area. Environmental Pollution, 212, 433-439.

Djuricin, S., Pataki, D. E., and Xiaomei, X. (2010). A Comparison of Tracer Methods for

Quantifying CO2 Sources in an Urban Region. Journal of Geophysical Research, 115(D11303), 1-13.

EDGAR. (2010). Emission Database for Global Atmospheric Research Release Version 4.2 of the European Commission. Joint Research Centre (JRC)/Netherlands Environmental Assessment Agency (PBL). Retrieved 06/29, 2016, from http://edgar.jrc.ec.europa.eu

FFDAS. (2010). Fossil Fuel Data Assimilation System (FFDAS). Retrieved 06/29, 2016, from http://hpcg.purdue.edu/FFDAS/

Gamnitzer, U., Karstens, U., Kromer, B., Neubert, R. E. M., Meijer, H. A. J., Schroeder, H., and

Levin, I. (2006). Carbon Monoxide: A Quantitative Tracer for Fossil Fuel CO2? Journal of Geophysical Research, 111(D22302) doi:10.1029/2005JD006966

Gorka, M., and Lewicka-Szezebak, D. L. (2013). One-Year Spatial and Temporal Monitoring of

Concentration and Carbon Isotopic Composition of Atmospheric CO2 in a Wroclaw (SW Poland) City Area. Applied Geochemistry, 35, 7-13.

Gorka, M., Sauer, P. E., Lewicka-Szezebak, D. L., and Jedrysek, M. (2011). Carbon Isotope

Signature of Dissolved Inorganic Carbon (DIC) in Precipitation and Atmospheric CO2. Environmental Pollution, 159, 294-301.

107

Greening Greater Toronto. (2012). Living City Report Card 2012. Retrieved 12/04, 2015, from http://civicaction.ca/

Hartmann, D. L., Klein Tank, A. M. G., Rusticucci, M., Alexander, L. V., Brönnimann, S., Charabi, Y., Dentener, F. J., Dlugokencky, E. J., Easterling, D. R., Kaplan, A., Soden, B. J., Thorne, P. W., Wild, M., & Zhai, P. M. (2013). In Stocker T. F., Qin D., G.-K. P., Tignor M., Allen S. K., Boschung J., Nauels A., Xia Y., Bex V. and Midgley P. M. (Eds.), Observations: Atmosphere and Surface. in: Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge, United Kingdom and New York, NY, USA: Cambridge University Press.

Historica Canada. (2015). Ontario: Energy. Retrieved 12/07, 2015, from http://www.thecanadianencyclopedia.ca/en/article/ontario/

Independent Electricity System Operator (IESO). (2015). 10-Year Outlook Highlights: An Assessment of the Adequacy of Generation and Transmission Facilities to Meet Future Electricity Needs in Ontario. Retrieved 12/07, 2015, from http://www.ieso.ca/Documents/marketReports/10yearOutlook-highlights-2005jul.pdf

ISL-UofC. (2016). Carbon and Nitrogen Isotope Analyses by Continuous-Flow Isotope Ratio Mass Spectrometry. Retrieved 01/20, 2016, from http://www.ucalgary.ca/uofcisl/techniques

Jasek, A., Zimnoch, M., Gorczyca, Z., Smula, E., and Rozanski, K. (2014). Seasonal Variability of Soil CO2 Flux and its Carbon Isotope Composition in Krakow Urban Area, Southern Poland. Isotopes in Environmental and Health Studies, 50(2), 143-155.

Keeling, C. D. (1958). The Concentration and Isotopic Abundance of Carbon Dioxide in Rural Areas. Geochimica Et Cosmochimica Acta, 13, 322-334.

Keeling, C. D. (1961). The Concentration and Isotopic Abundance of Carbon Dioxide in Rural and Marine Air. Geochimica Et Cosmochimica Acta, 24, 277-298.

Kuc, T., and Zimnoch, M. (1998). Changes of the CO2 Sources and Sinks in a Polluted Urban Area (Southern Poland) Over the Last Decade, Derived from the Carbon Isotope Composition. Radiocarbon, 40(1), 417-423.

108

Lam, N. L., Smith, K. R., Gautheir, A., and Bates, M. N. (2012). Kerosene: A Review of Household Uses and their Hazards in Low- and Middle-Income Countries. Journal of Toxicology and Environmental Health, Part B, 5(6), 396-432.

Landmeyer, J. E., Vroblesky, D. A., and Chapelle, F. H. (1996). Stable Carbon Isotope Evidence of Biodegradation Zonation in a Shallow Jet-Fuel Contaminated Aquifer. Environmental Science and Technology, 30, 1120-1128.

Leonard, B. (2007). Federal Trade Commission Report on Spring/Summer 2006 Nationwide Gasoline Price Increases. Retrieved 01/14, 2016, from https://www.ftc.gov/sites/default/files/documents/reports/federal-trade-commission-report- spring/summer-2006-nationwide-gasoline-price-increases/p040101gas06increase.pdf

Masalaite, A., Garbaras, A., and Remeikis, V. (2012). Stable Isotopes in Environmental Investigations. Lithuanian Journal of Physics, 52(3), 261-268.

Ontario Ministry of Energy. (2015). Ontario Energy Report. Retrieved 12/04, 2015, from http://www.ontarioenergyreport.ca/

Pataki, D. E., Bowling, D. R., and Ehleringer, J. R. (2003). Seasonal Cycle of Carbon Dioxide and its Isotopic Composition in an Urban Atmosphere: Anthropogenic and Biogenic Effects. Journal of Geophysical Research, 108(D23) doi:10.1029/2003JD003865

Searle, S. Y., Turnbull, M. H., Boelman, N. T., Schuster, W. S. F., Yakir, D., and Griffin, K. L. (2012). Urban Environment of New York City Promotes Growth in Northern Red Oak Seedlings. Tree Physiology, 32(4), 389-400.

Statistics Canada. (2007). Households and the Environment: Energy Use. Retrieved 12/04, 2012, from http://www.statcan.gc.ca/pub/11-526-s/2010001/t002-eng.pdf

Statistics Canada. (2012). Report on Energy Supply and Demand. Retrieved 12/04, 2015, from http://www.statcan.gc.ca/pub/57-003-x/2015001/t037-eng.pdf

109

Statistics Canada. (2012). Focus on Geography Series, 2011 Census. Statistics Canada Catalogue no. 98-310-XWE2011004. Retrieved 08/13, 2013, from https://www12.statcan.gc.ca/census- recensement/2011/as-sa/fogs-spg/Facts-cma-eng.cfm?LANG=Eng&GK=CMA&GC=535

U.S. Energy Information Agency. (2016). Date of Switch to Summer-Grade Gasoline Approaches. Retrieved 06/17, 2016, from http://www.eia.gov/todayinenergy/detail.cfm?id=11031

Vogel, F. R., Hammer, S., Steinhof, A., Kromer, B., and Levin, I. (2010). Implication of Weekly 14 and Diurnal C Calibration on Hourly Estimates of CO-Based Fossil Fuel CO2 at a Moderately Polluted Site in Southwestern Germany. Tellus, 62B(512), 520.

Vogel, F. R., Huang, L., Ernst, D., Giroux, L., Racki, S., and Worthy, D. E. J. (2013). Evaluation 13 of a Cavity Ring-Down Spectrometer for in Situ Observations of CO2. Atmopsheric Measurement Techniques, 6, 301-308.

Widory, D. (2006). Combustibles, Fuels and their Combustion Products: A View through Carbon Isotopes. Combustion Theory and Modelling, 10(5), 831-841.

Widory, D., and Javoy, M. (2003). The Carbon Isotope Composition of Atmospheric CO2 in Paris. Earth and Planetary Science Letters, 215, 289-298.

Yeh, H., and Epstein, S. (1981). Hydrogen and Carbon Isotopes of Petroleum and Related Organic Matter. Geochimica Et Cosmochimica Acta, 45, 753-762.

Zimnoch, M., Jelen, D., Galkowski, M., Kuc, T., Necki, J., Chmura, L., Gorczyca, Z., Jasek, A., and Rozanski, K. (2012). Partitioning of Atmospheric Carbon Dioxide Over Central Europe:

Insights from Combined Measurements of CO2 Mixing Ratios and their Carbon Isotope Composition. Isotopes in Environmental and Health Studies, 48, 421-433.

110

Chapter 4 13 Using high-resolution simulations of atmospheric δ CO2 signatures to understand variability in source contributions of CO2 in the Greater Toronto Area, Canada

Sections of this chapter are in preparation for submission for peer-reviewed publication:

Pugliese, S. C., Murphy, J. G., Vogel, F., Moran, M., Zhang, J., Zheng, Q., Stroud, C., Ren, S., 13 Worthy, D., Broquet, G. (2017). Using high-resolution simulations of atmospheric δ CO2 signatures to understand variability in source contributions of CO2 in the Greater Toronto Area, Canada, Journal of Geophysical Research: Atmospheres, in prep.

4 Chapter 4 4.1 Abstract

13 Stable carbon isotopic signatures (δ CO2) can provide insight into the sectoral contributions of carbon dioxide (CO2) on continuous timescales in situations where signatures of unique end- 13 members are known. We present an atmospheric modelling framework to combine local δ CO2 signatures from a variety of combustion fuels with the Southern Ontario CO2 Emissions (SOCE) 13 inventory to simulate ambient δ CO2 signatures in the Greater Toronto Area (GTA). We 13 demonstrate that the modelled δ CO2 captures the variability in measured hourly signatures at the Downsview station from January to March, 2016 as well as the diurnal profile of signatures. We 13 use our ability to model δ CO2 in the GTA to identify the limitations and advantages of traditional chemical mass balance (CMB) and Keeling analyses. We performed three CMB approximations 13 13 to estimate the δ CO2 signature of the local CO2 enhancement (δ Cs) for measured and modelled data (generated using two background estimates: 5th percentile of a 48 h running average, and output from the MACC (Monitoring Atmospheric Composition and Climate) global inversion). th 13 We found that when the 5 percentile was used as an estimate of background CO2 and δ CO2, the 13 δ Cs was often unrealistic and heavier compared to that using MACC. We compared the CMB approximations with Keeling analyses based on both overnight and hour-of-day data. For both 13 approaches, when Keeling analyses were run on modelled and measured CO2 and δ CO2 data, the

111

13 13 estimated δ Cs was variable and the median δ Cs consistently heavier than that predicted by the SOCE inventory.

4.2 Introduction

The anticipated impacts of climate change have led to a number of initiatives globally and regionally to reduce the emission of carbon dioxide (CO2) and other greenhouse gases. Agreements such as that made in Paris, France in 2015 or initiatives such as the C40 Cities Climate Leadership

Group require cities and nations to reduce their emissions of CO2, particularly from their largest sectors. These commitments create a need for independent emission verification to evaluate whether reduction targets have been met and/or to assess where future mitigation efforts should be focused.

Recently, the Southern Ontario CO2 Emissions (SOCE) inventory has been made available for southern Ontario (Canada) at a fine spatial and temporal resolution (2.5 km x 2.5 km and hourly, respectively) (Pugliese et al., 2017). This inventory provides information on the emission of CO2 from nine source categories in the domain for the year 2010. While the inventory provides useful information on the spatial and temporal variability of CO2 emissions in the Greater Toronto Area (GTA), it is limited by the fact that it was compiled using data (emissions factors, activity data, etc.) from a single year, and can thus not account for the sometimes significant inter-annual variability of anthropogenic CO2 emissions due to temperature anomalies (Breon et al., 2017). Because the sectoral emission estimates may only be updated every 5-10 years, relying on inventories to evaluate the effectiveness of policy efforts is not ideal. Therefore, there is a need for an alternate method that can provide source attribution estimates on an ongoing basis to inform policy efforts in a timely manner. A tool used to trace sources of CO2 is its carbon isotopic 13 13 12 composition, δ CO2, where the ratio of C to C (relative to a known standard, Equation 1) can be used to identify the source as long as the unique end-member signatures are known.

13 퐶푠푎푚푝푙푒 12 13 퐶푠푎푚푝푙푒 훿 퐶푂2 = ( 13 ) − 1 푥 1000 ‰ (E4.1) 퐶푠푡푎푛푑푎푟푑 12 퐶푠푡푎푛푑푎푟푑 ( )

112

13 If continuous measurements of CO2 and δ CO2 are made and signatures of end-members are sufficiently different, this method can be used to evaluate the contribution of different sources to

CO2 continuously.

Isotopes have been used to understand the relative contributions of different sources of CO2 in an urban environment in previous studies. A study performed in Los Angeles in winter 2007-2008 allocated enhancements in mixing ratios of CO2 between four sources, natural gas combustion, gasoline combustion, above-ground respiration, and below-ground respiration, in a chemical mass 13 18 14 balance (CMB) approximation using δ CO2 in combination with other tracers, δC O2 and Δ C (Djuricin et al., 2010). Their results showed that despite Los Angeles being a site of dense population and significant anthropogenic CO2 emissions, above-ground respiration contributed the largest fraction to enhanced CO2 mixing ratios, reaching as high as ~80 % in April 2008 (Djuricin et al., 2010). Furthermore, when considering the anthropogenic fraction, their results showed that natural gas and gasoline combustion contribute similarly to the overall enhanced CO2 in the local atmosphere, despite a local inventory suggesting that the gasoline contribution should be three times larger than the natural gas contribution (Djuricin et al., 2010). Similarly, another CMB study 13 18 performed in Salt Lake City in 2002 used δ CO2 and δC O2 to show that natural gas and gasoline combustion can contribute 57 % and 42 %, respectively, to enhancements in CO2 mixing ratios in the wintertime (Pataki et al., 2003a). These and other studies (Widory and Javoy, 2003) 13 demonstrate the utility that δ CO2 analyses offer for CO2 source apportionment in urban centers.

13 The use of δ CO2 measurements to source apportion CO2 could provide an effective and timely solution to the independent validation of emissions reductions required in urban environments. 13 Typically, studies that use δ CO2 to source apportion CO2 do so through the use of a CMB approximation to separate the CO2 enhancement into two or more components, or a Keeling 13 analysis to solve for the δ CO2 signature of the enhancement. However, there has yet to be an in- depth evaluation of these methods to assess how altering the input parameters or the time periods that are considered can impact the interpretation of the results. To the authors’ knowledge, the best 13 example is a sensitivity study performed by Pataki et al. (2003a) to assess how changing the δ CO2 signatures of three local sectors by 1 ‰ would impact the results of their CMB analysis. Their results showed that these variations caused changes in the attributions to each sector of up to 24 13 %, indicating that local and accurate measurements of δ CO2 signatures are required for estimation of CO2 source contributions (Pataki et al., 2003a). An alternate study used three datasets

113

13 of CO2 and δ CO2 to determine the impact of uncertainty in measurements of the calculated 13 13 δ CO2 on the CO2 enhancement (δ Cs) from respiration (Zobitz et al., 2006). Their results showed 13 that uncertainty in measurements of δ CO2 have a greater effect on the uncertainty in the 13 calculated δ Cs than the uncertainty in CO2 (Zobitz et al., 2006).

In an effort to expand on the work performed by Pataki et al. (2003a) and Zobitz et al. (2006) and to evaluate the use of stable carbon isotopes to source apportion CO2, we focused on creating a 13 framework to model δ CO2 signatures using source-specific signatures and modelled hourly CO2 13 mixing ratios. We used modelled and measured δ CO2 signatures in a series of Keeling and CMB 13 analyses to compare the predicted signature of the CO2 enhancement (δ Cs) and contributions of two components, C1 and C2. We expect the results of this work to improve our understanding of the utility of stable carbon isotopes to source apportion CO2, as well as to identify some of the advantages and limitations associated with the Keeling and CMB analyses.

4.3 Methods

4.3.1 Geographic Domain

The geographic focus of this study was the GTA in southern Ontario, Canada. The GTA comprises five municipalities, Halton, Durham, Peel, Toronto and York, which together have a population exceeding 6 million (Statistics Canada, 2012). Toronto is the capital city of Ontario as well as the largest urban area in Canada. While the GTA represents only 0.07 % of land cover in Canada, it accounts for ~19 % of the total population (Statistics Canada, 2011) and is estimated to account for ~10 % of its CO2 emissions (Environment Canada, 2012).

13 4.3.2 Ambient measurements of CO2 and δ CO2

13 o o CO2 and δ CO2 were continuously measured at the Downsview site (43.7 N, 79.4 W) using cavity-ring down spectroscopy as a part of Environment and Climate Change Canada (ECCC)’s Greenhouse Gas Observational Program. Details on specific instrument set-up, calibration traceability and measurement protocols have been outlined previously (Pugliese et al., 2016;

Pugliese et al., 2017; Vogel et al., 2013). The CO2 measurements used in this work have a precision better than 0.1 ppm based on one-minute averaging intervals and are accurate within 0.2 ppm; the 13 δ CO2 measurements have an average precision of 0.2 ‰ (Vogel et al., 2013) and an accuracy of 13 0.02 ‰ (Huang et al., 2013). We chose Downsview station for measurements of CO2 and δ CO2

114 as it is situated ~15 km north of the city of Toronto with its emission footprint dominated by sources within the GTA (with minor contributions from surrounding regions).

4.3.3 GEM-MACH chemistry-transport model and the SOCE inventory

In this project, we used the GEM-MACH (Global Environmental Multi-scale‒Modelling Air quality and CHemistry) chemistry-transport model (CTM) (Gong et al., 2015; Moran et al., 2013; Pavlovic et al., 2016; Talbot et al., 2008) to link surface emission estimates (from the SOCE inventory for Ontario and the FFDAS v2 inventory for USA and Quebec) and atmospheric mixing ratios. Details about the configuration of the GEM-MACH model, the FFDAS v2 inventory and the SOCE inventory used in our study have been described elsewhere (Pugliese et al., 2017).

Briefly, the SOCE inventory estimates CO2 emissions at a fine spatial and temporal resolution (2.5 km x 2.5 km and hourly, respectively) for nine categories: Area, On-road, Off-road, Residential and Commercial natural gas combustion, Point and Marine sources as well as emissions from the

USA and Quebec. The CO2 boundary conditions set at the lateral and top edges of the domain were obtained from the Monitoring Atmospheric Composition and Climate (MACC) global inversion, v.10.2 (http://www.copernicus-atmosphere.eu/). The GEM-MACH CTM was used to simulate hourly mixing ratio enhancements of CO2 from each of the nine categories at the Downsview station for January-March, 2016. Model-measurement agreement at the Downsview station was found to be good, especially in the afternoon (Pugliese et al., 2017). Model-simulated specific humidity (q, kg/kg) was used to convert estimated CO2 mixing ratios to dry air mixing 13 ratios. In this study, these simulated dry air hourly mixing ratios of CO2 were converted to δ CO2, as outlined in the following section.

13 4.3.4 Conversion of CO2 mixing ratios to δ CO2

13 To convert simulated hourly CO2 mixing ratios to δ CO2 signatures, we followed the CMB approach which states that the total CO2 (Ct) is a sum of the contribution from individual sources 13 (Ci), Equation 4.2. Based on this, the isotopic composition of the total CO2 (δ Ct) can be calculated 13 from the sum of the weighted signatures of the contributing sources (δ Ci), Equation 4.3, where i is the index of each source.

Ct = C1 + C2 + C3 +…Ci (E4.2)

13 13 δ C t ∙ 퐶푡 = ∑푖 δ 퐶푖 ∙ 퐶푖 (E4.3)

115

The sources that are influencing Ct in our simulation framework are a background contribution,

Cb, (estimated by MACC) as well as the sum of the nine source categories. In our framework, 13 δ Cb was assumed to have a constant value of -8.9 ‰, similar to background signatures used by other studies globally (e.g. Djuricin et al., 2010) and consistent with that measured at Downsview o o 13 as well as at the background facility in Egbert, Ontario (44.2 N, 79.8 W). The δ Ci of each sector 13 was calculated from a weighted average of the δ CO2 signatures of individual fuel sources of each sector, which were characterized specifically for the GTA in a previous study, Table 4.1 (Chapter

3; Pugliese et al., 2016). CO2 emissions from the biosphere was not included as a sector in this study as previous modelling studies quantified a negligible contribution from this source in the winter months in the GTA (Chapter 2; Pugliese et al., 2017). The contribution of subsectors to overall source sectors has been discussed previously (Chapter 2; Pugliese et al., 2017), but the contribution of individual fuel sources and their corresponding signatures is discussed briefly below.

13 Table 4.1: δ CO2 signatures and contributions of each source category

13 Sector δ CO2 Signature Estimated Source (‰)* Contribution (%)‡ Area -29.6 10 Residential and Commercial natural gas -44.2 40 combustion Point -36.2 8 On-road -27.9 28 Off-road -28.8 7 Marine -30.6 0.03 USA -28 7 Quebec -28 0.1

13 * δ CO2 signatures are weighted averages of individual components in each sector ‡ Contributions are based on enhanced CO2 at Downsview station for January-March, 2016

4.3.4.1 Area emissions

Area emissions are mostly small stationary sources that represent diffuse emissions that are not inventoried at the facility level. The major fuel sources in the Area sector include: natural gas (-

116

44.2 ‰ (Pugliese et al., 2016)), propane (-27.7 ‰ (Rudolph et al., 2002)), petroleum (-25.9 ‰ (Bush et al., 2007)), kerosene and stove oil (-29.3 ‰ and -28.8 ‰, respectively (Pugliese et al., 2016)), and heavy and light fuel oil (-31 ‰ (Vardag et al., 2015)). These sources, weighted by 13 their contribution to total CO2 emissions from the Area sector, produce an average δ CO2 of -29.6 ‰.

4.3.4.2 Point emissions

Point emissions are stationary sources in which emissions exit through a stack or identified exhaust. The major fuel sources in the Point sector include: natural gas (-44.2 ‰ (Pugliese et al., 2016)), propane (-27.7 ‰ (Rudolph et al., 2002)), petroleum (-25.9 ‰ (Bush et al., 2007)), kerosene and stove oil (-29.3 ‰ and -28.8 ‰, respectively (Pugliese et al., 2016)), and heavy and light fuel oil (-31 ‰ (Vardag et al., 2015)). The sources, weighted by their contribution to total 13 CO2 emissions in the Point sector, produce an average δ CO2 of -36.2 ‰.

4.3.4.3 On-road emissions

On-road emissions include the emissions from any on-road vehicles. The major fuel sources in the On-road sector include: gasoline (-27.6 ‰ (Pugliese et al., 2016)) and diesel (-29.1 ‰ (Pugliese et al., 2016)). These sources, weighted by their contribution to total CO2 emissions from the On- 13 road sector, produce an average δ CO2 of -27.9 ‰.

4.3.4.4 Off-road mobile emissions

Off-road emissions include the emissions from any off-road vehicles that do not travel on designated roadways, including aircraft, all off-road engines, and locomotives. The major fuel sources in the Off-road sector include: gasoline (-27.6 ‰ (Pugliese et al., 2016)), diesel (-29.1 ‰ (Pugliese et al., 2016)), jet fuel (-29.5 ‰ (Pugliese et al., 2016)), and heavy fuel oil (-31 ‰ (Vardag et al., 2015)). These sources, weighted by their contribution to total CO2 emissions from the Off- 13 road sector, produce an average δ CO2 of -28.8‰.

4.3.4.5 Marine emissions

Commercial marine emissions include the emissions from any marine vessels travelling on the Great Lakes. The major fuel sources in the Marine sector include: diesel (-29.1 ‰ (Pugliese et al., 2016)) and heavy fuel oil (-31 ‰ (Vardag et al., 2015)). These sources, weighted by their

117

13 contribution to total CO2 emissions from the Marine sector, produce an average δ CO2 of -30.6 ‰.

4.3.4.6 Residential and commercial emissions

Residential and commercial CO2 emissions reflect on-site combustion of natural gas for electricity and heating. The only fuel source in this sector is natural gas (-44.2 ‰ (Pugliese et al., 2016)) and 13 therefore the sector average δ CO2 was -44.2 ‰.

4.3.4.7 USA and Quebec emissions

Emissions from the USA and province of Quebec that are within the modeling domain (Pugliese et al., 2017) are included in our inventory. Because of the difficulty associated with assigning a 13 13 single δ CO2 signature for an entire country or province, both were assigned an average δ CO2 signature of -28 ‰. We conducted two sensitivity analyses to assess the impact on the calculated 13 13 δ CO2 when the δ CO2 signature of USA and Quebec were changed to -26 ‰ and -30 ‰ for both 13 sources. When the δ CO2 signature of USA and Quebec was -26 ‰, -28 ‰ and -30 ‰, there was 13 no significant difference in the calculated monthly median δ CO2. Furthermore, when correlation 13 13 plots were made to quantify the δ CO2 model-measurement agreement, the use of a δ CO2 signature of -26 ‰, -28 ‰ and -30 ‰ for Quebec and USA had no impact on the calculated 13 correlation coefficient, root mean square error or mean bias. Therefore, a δ CO2 signature of -28 13 ‰ for both USA and Quebec were used to calculated hourly δ CO2.

4.3.5 The chemical mass balance approximation and Keeling plot analysis

We used the CMB approximation to partition the total CO2 enhancement (Cs) at Downsview into two component parts, End-member #1 (C1) and End-member #2 (C2). For this analysis, we used three sets of input data to discern the impact of the background contribution chosen: (1) measured 13 13 CO2 and δ CO2 (where background CO2 and δ CO2 were estimated by averaging all th measurements below the 5 percentile of CO2 over a moving window of 48 hours, an approach 13 13 used in (Ammoura et al., 2016); (2) modelled CO2 and δ CO2 (where background CO2 and δ CO2 th were estimated by averaging all simulated mixing ratios below the 5 percentile of CO2 over a 13 moving window of 48 hours); and (3) modelled CO2 and δ CO2 (where background CO2 was 13 provided by MACC and background δ CO2 was assumed to be constant at -8.9 ‰). To solve for the remaining contribution of C1 and C2, we applied the CMB outlined in Equations 4.2 and 4.3.

118

End-members for C1 and C2 were estimated from measurements made in Pugliese et al. (2016) and assumed to be -28 ‰ and -44 ‰, for non-natural gas and natural gas, respectively.

We complemented our CMB analyses with Keeling linear mixing model analyses, a specific emission scenario case outlined in Equation 4.2 where there are only two sources contributing to

Ct, a background component (Cb) and a local enhancement (Cs), Equation 4.4 (Keeling, 1958; Keeling, 1961). Keeling applied isotopic mass balance and derived a linear mixing model, Equation 4.5.

Ct = Cb + Cs (E4.4) 13 13 13 13 δ Ct = Cb(δ Cb – δ Cs)(1/Ct) + δ CS (E4.5)

The advantage of using the Keeling linear mixing model is that the contribution and signature of 13 the background CO2 (Cb and δ Cb) do not need to be known to solve for the signature of the 13 13 enhancement (δ CS), which is equal to the intercept in a plot of δ Ct vs. (1/Ct). We applied the 13 Keeling linear mixing model (Equation 4.5) to both measured and modelled CO2 and δ CO2. We 13 calculated δ CS for two subsets of data within our model framework, (1) one plot for every overnight (22:00-06:00) period, and (2) one plot for every hour of the day, to assess how these 13 changes in timescale impacted our interpretation of the calculated δ CS. We performed a sensitivity analysis to determine if altering the R-squared (R2) threshold we accepted to calculate 13 13 δ CS significantly impacted our interpretation of the results; we found that the δ CS was statistically similar (p > 0.05) when the R2 threshold was set at 0.50, 0.70 or 0.95. Therefore, the Keeling plot intercept was considered if R2 > 0.50 to maximize the amount of data used in the analyses.

4.4 Results and Discussion

13 4.4.1 Simulation of δ CO2 in the Greater Toronto Area

13 Figure 4.1 displays the times series of measured and modelled δ CO2 signatures at the Downsview site for January 2016 (Figures B1 and B2 show the same figure for February and March, respectively). The similarity between the time series during all three months demonstrates the 13 model’s general capability to capture the variability in observations of δ CO2, the synoptic-scale 13 changes in δ CO2, as well as daily changes. The model is generally able to simulate the variability in signatures between -9 ‰ and -12 ‰; however, some short time periods are poorly simulated,

119

13 particularly nighttime events when the model estimates a much more negative δ CO2 signature than measured. We hypothesize that this arises from the model’s limited ability to simulate vertical mixing in the nighttime under strongly stable wintertime conditions, compromising its ability to simulate accurate mixing ratio enhancements of CO2 during this time. The majority of CO2 emissions between 21:00 and 05:00 EST are from building natural gas combustion, with a very 13 light δ CO2 signature (-44.2 ‰), which can lead to significant underestimations of modelled 13 δ CO2 during this time. Typically, CO2 modelling studies focus on simulating afternoon periods 13 to minimize this issue of vertical mixing, however in the case of δ CO2, we are most confident in its use as a tracer of CO2 during the nighttime as the shallow boundary layer that forms maximizes the enhancement of CO2 above the background by trapping emissions close to the ground, thus 13 providing more dynamic range in the δ CO2 measurement. Furthermore, we are modelling 13C/12C, not a mixing ratio, and thus are less susceptible to strong biases. Therefore in this study, the entire day is modelled and included in subsequent analyses.

13 Figure 4.1: Time series of measured (blue) and modelled (red) CO2 signatures at the Downsview site for January 2016.

13 To quantify the δ CO2 model-measurement agreement, Figure 4.2 displays a scatter plot of modelled vs. measured signatures in January-March, 2016, with the points coloured by hour of day. Overall, it is clear from Figure 4.2 that there is a reasonable agreement between modelled and

120

13 measured δ CO2, with a correlation coefficient (R) of 0.58 and minimal root mean square error (RMSE) and mean bias (MB) (1.05 ‰ and 0.04 ‰ respectively). When the same plot was made using only afternoon (12:00 – 16:00 EST) data, similar model-measurements agreement was obtained (R = 0.56, RMSE = 0.50 ‰ and MB = 0.14 ‰), supporting our decision to use simulations from the entire day. Figure 4.2 also highlights that the points which deviate most from the 1:1 line are those in the early morning (between 00:00 – 10:00) and those late at night (between ~17:00- 13 23:00). This result is consistent with the hypothesis that underestimations of modelled δ CO2 result from the model’s general limited ability to simulate vertical mixing in the nighttime.

13 Figure 4.2: Scatter plot of the modelled and measured CO2 signatures from January-March 2016 at the Downsview station. The model vs. measurement Correlation Coefficient (R), root mean square error (RMSE) and mean bias (MB) are provided within the plot. The solid line is the standard major axis regression and dashed line is 1:1 shown for reference.

13 4.4.2 Diurnal variability of measured and modelled δ CO2

13 In the wintertime in an urban environment like Toronto, measured δ CO2 signatures have a typical diurnal pattern, in which signatures are lighter at night (as a result of natural gas combustion for heating) and heavier during the day (as a result of increased combustion of gasoline and diesel for transportation as well as from a smaller CO2 enhancement, resulting in the overall signature being closer to the background signature). Figure 4.3 (created using data from the entire model period,

121

13 January-March) displays the model’s strong ability to capture this diurnal pattern of δ CO2 13 signatures. The small night-day gradient of measured δ CO2, approximately -1.1 ‰, is consistent with a recent study done in Heidelberg, Germany (Vardag et al., 2016) and is characteristic of a winter environment, where most of the local CO2 enhancement is dominated by anthropogenic sources with very light signatures. However, night‒day gradients are heavily site‒dependent because of variations in the magnitude of local emissions and the hour-of-day dependence of sources with different signatures (Szaran, 1998; Szaran et al., 2002).

13 Figure 4.3: Time series of mean measured (blue) and modelled (red) diurnal δ CO2 signatures at the Downsview site. Error bars show standard deviation of signatures.

4.4.3 Assessment of the chemical mass balance approximation to source apportion CO2

In the absence of a chemical transport model, a CMB analysis is often undertaken using 13 measurements of CO2 and δ CO2 to estimate the contribution of different sectors, C1 and C2, using Equations 4.2 and 4.3 (Djuricin et al., 2010; Pataki et al., 2003a; Xu et al., 2017). With our framework, we are able to carry out the same type of analysis using model simulations of CO2 and 13 δ CO2, which, in principle, should be consistent with the inventory used in the model. Therefore,

122 as an assessment of the CMB approach, we performed a series of three different calculations using modelled and measured data to evaluate whether the predicted contributions of C1 and C2 ever 13 agreed with that calculated by the SOCE inventory. Figure 4.4a displays the hourly δ CS values as a function of Cs calculated by summing the contributions of CO2 enhancements from each of the sectors in the SOCE inventory (and Quebec and USA), whereas the subsequent panels show 13 th results from the CMB using modelled CO2 and δ CO2 (with 5 percentile estimates of background 13 13 th CO2 and δ CO2, Figure 4.4b), and measured CO2 and δ CO2 (with 5 percentile estimates of 13 13 background CO2 and δ CO2, Figure 4.4c). Because the information content of measured δ CO2 is limited by the precision of our instrument (0.2 ‰), we coloured the points in each panel of 13 13 Figure 4.4 by the difference in measured δ Ct and the background δ Cb to understand when the signature of the enhancement in CO2 we are observing is approaching the detection limit of the 13 instrument. For all panels, we calculated the median δ Cs (and report in Table 4.2) only for times 13 13 when a CO2 enhancement of greater than 20 ppm is observed and δ Ct- δ Cb < -0.2 ‰.

13 Figure 4.4a shows the results using the SOCE inventory; a very consistent δ CS is observed 13 13 (median = -36.3 ± 0.1 ‰). By definition, the δ Ct signature is always lighter than the δ Cb signature and the difference generally increases with Cs. Results from Figure 4.4b show that when 13 13 th the CMB is used to estimate δ CS using modelled CO2 and δ CO2 with the 5 percentile as a 13 background estimate, a significant number of hours are estimated to have unrealistic δ CS values, including estimates which are greater than zero and outside the known range of sources of CO2 in 13 the domain (heavier than -23 ‰ and lighter than -44 ‰). A number of estimates of δ CS are 13 13 limited by the precision of our instrument (coloured blue - yellow on the panel, δ Ct- δ Cb > -0.2

‰). Furthermore, this scenario has an additional challenge where Cs is often estimated to be less than zero (this is an outcome of using the 5th percentile over a moving 48-hour window). The 13 median δ CS calculated by this modelled scenario is -35.4 ± 0.3 ‰, slightly heavier than that predicted by the SOCE inventory. However, this estimate is based on a small portion of the data (~11 % of the simulation period) as the majority of it was lost in the filtering process. Figure 4.4c 13 th displays the results of the CMB on measured CO2 and δ CO2 with the 5 percentile as a background estimate. Similar to the results in Figure 4.4b, a number of hours are estimated to 13 13 13 have unrealistic values of δ Cs. However, the majority of these times are also when δ Ct - δ Cb > 0, which is again a challenge associated with using the 5th percentile to estimate background 13 13 values of CO2 and δ CO2. The median δ CS calculated by this measured scenario is -33.4 ± 0.4

123

‰, heavier than that predicted by the both the SOCE inventory and model using a 5th percentile background estimate. This estimate is again limited by only considering ~11 % of the data.

If we assume the local enhancement predicted by each of the three scenarios was derived from two end-members with the signatures of non-natural gas (C1, -28 ‰) and natural gas (C2, -44 ‰), the range of estimates for the contribution of each source is shown in Table 4.2. The SOCE inventory estimates a contribution of 48 ± 1 % C1 and 52 ± 1 % C2, while the scenarios using modelled and th measured data with the 5 percentile background estimates infer a larger contribution of C1 (52 ±

2 % and 66 ± 3 %, respectively) and a smaller contribution of C2 (46 ± 2 % and 34 ± 3 %, respectively). The results of these analyses therefore highlight the limitations of using the 5th 13 th percentile as an estimate of background CO2 and δ CO2. For both scenarios when the 5 percentile is used as a background estimate (Figures 4.4b and 4.4c), the CMB predicted unrealistic values of 13 δ CS for the majority of hours (>80 % of data was removed by the filtering process) and estimated contributions of C1 and C2 that are different than those calculated by the SOCE inventory. Our 13 filtering process (calculating the median δ CS only when Cs > 20 ppm) likely biases the analysis towards nighttime data, and therefore we expect that natural gas combustion (C2) would play a significant role in CO2 emissions, as estimated by the SOCE inventory, as residents are using their furnaces for heating during this simulation period (January-March). Therefore this analysis 13 highlights the need for a suitable background estimate of both CO2 and δ CO2 to source apportion

CO2 using the CMB approximation.

13 Table 4.2: Calculated median δ Cs and contributions of C1 and C2 using the SOCE inventory and chemical mass balance 13 Method Input data δ Cs Contribution Contribution of (‰) of C1, -28 ‰ C2, -44 ‰ (%) (%) SOCE Inventory -36.3* ± 0.1 48 ± 1 52 ± 1 Modelled with 5th -35.4* ± 0.3 54 ± 2 46 ± 2 Chemical Mass percentile background Balance Measured with 5th -33.4* ± 0.4 66 ± 3 34 ± 3 percentile background * 13 13 These data were filtered for times when Cs > 20 ppm and the δ Ct- δ Cb < -0.2 ‰. Note: Uncertainty range reports the standard error of the mean.

124

13 12 13 Figure 4.4: Plot of δ Cs vs. Cs using a) the SOCE inventory, b) modelled CO2 and δ CO2 with th 12 13 12 13 5 percentile estimates of background CO2 and δ CO2, c) measured CO2 and δ CO2

125

4.4.4 Assessment of the Keeling linear mixing model to source apportion CO2

13 When background estimates of CO2 and δ CO2 are not available, a Keeling analysis (Equation 13 4.5) is often used to estimate δ Cs (Pataki et al., 2003a; Pataki et al., 2003b). It is often the case that time periods are chosen when the local enhancement of CO2 is maximized, such as overnight periods or periods with low wind speed. While this selectivity might increase the robustness of the 13 analysis, the representativeness of the calculated δ Cs may be compromised. Therefore, as an assessment of this approach, we performed two different Keeling analyses using modelled and 13 measured data to evaluate (1) whether δ Cs calculated by modelled and measured data can provide reasonable agreement with the SOCE inventory, and (2) whether the time of day chosen to quantify 13 δ Cs affects the interpretation of our results.

Figure 4.5 shows the results of the Keeling analysis run with modelled (red) and measured (blue) 13 data for each overnight period (22:00 – 06:00) in the simulation; the black trace shows the δ Cs 13 calculated using the flux-weighted δ Ci from each of the nine source sectors in the SOCE inventory. Of the possible 91 days in January-March 2016, signatures were calculated for 43 overnight periods (measurement data were not available for 15 days and data did not meet the selection criteria (Section 4.3.5) for the rest of the periods). Figure 4.5a shows the individual 13 overnight calculated δ CS whereas the box and whisker plot in Figure 4.5b displays a distribution 13 of all the overnight δ CS for each scenario considered. Both Figures 4.5a and 4.5b reveal that the 13 SOCE inventory estimates a very consistent δ CS during all overnight periods, with a median 13 13 δ CS of -37.7 ± 0.8 ‰, Table 4.3. Both the δ CS calculated by measured and modelled CO2 and 13 δ CO2 have a much higher degree of variability than that calculated from the inventory, 13 particularly when using modelled data. Furthermore, the δ CS calculated from both modelled and measured data implies source signatures that are outside the range of characterized CO2 sources in 13 the GTA (Pugliese et al., 2016), values heavier than ~ -23 ‰ or lighter than -44 ‰. For the δ CS calculated from modelled data, it is likely that model errors and uncertainty are contributing to the 13 13 unrealistic values of δ CS. For the δ CS calculated from measurements, it is possible that emissions from Ashbridge’s Bay Waste Water Treatment Plant (ABTP, located to the south-east of the downtown Toronto core on the shore of Lake Ontario) are contributing to some of the light 13 δ CS being calculated, as the isotopic signature of waste treatment plants is ~-46 to -47 ‰

(Townsend-Small et al., 2012). However, the contribution of CO2 to the GTA from ABTP is

126 minimal annually (~ 0.09 %) (Toronto Water, 2009) and therefore it cannot solely explain the light signatures, or the signatures lighter than -50 ‰, that are predicted by the Keeling analysis. In 13 13 addition to having more variability than the inventory-based δ CS, the overnight median δ CS is 13 relatively heavier, -35.8 ± 5.6 ‰ and -34.4 ± 6.2 ‰ for measured and modelled δ CS respectively,

Table 4.3. If we assume this local enhancement was derived from two end-members, C1 and C2, the SOCE inventory estimates a contribution of 39 ± 1 % C1 and 61± 1 % C2, Table 4.3. As discussed in Section 4.4.3, this result is expected as natural gas combustion (C2) would likely play a significant role in CO2 emissions during overnight periods as residents are using their furnaces for heating during the winter months. However, Keeling analysis with modelled data estimates 60

± 14 % C1 and 40 ± 14 % C2, and the Keeling analysis with measured data estimates 51 ± 9 % C1 and 49 ± 9 % C2, Table 4.3. Therefore, these results reveal that using the Keeling analysis with either measured or modelled data does not produce estimates of the relative contributions of C1 and C2 that are consistent with the SOCE inventory for overnight periods.

13 Table 4.3: Calculated median δ Cs and contributions of C1 and C2 using the SOCE inventory and the Keeling linear mixing model 13 Method Input data δ Cs Contribution of Contribution of (‰) C1, -28 ‰ C2, -44 ‰ (%) (%) Inventory - -37.7 ± 0.2 39 ± 1 61 ± 1 (Overnight) Keeling Modelled -34.4 ± 2.3 60 ± 14 40 ± 14 (Overnight) Measured -35.8 ± 1.4 51 ± 9 49 ± 9 Inventory - -34.9 ± 0.5 57 ± 3 43 ± 3 (24-hr, 0:00-23:00) Keeling Modelled -32.3 ± 1.1 73 ± 7 27 ± 7 (24-hr, Measured -32.6 ± 0.6 71 ± 4 29 ± 4 0:00-23:00)

Inventory - -38.2 ± 0.5 36 ± 3 64 ± 3 (24-hr, 22:00-06:00) Keeling Modelled -37.3 ± 1.4 42 ± 9 58 ± 9 (24-hr, Measured -34.0 ± 0.6 63 ± 4 38 ± 4 22:00-06:00) Note: Uncertainty range reports the standard error of the mean.

127

13 Figure 4.5: a) Time series of overnight (22:00-06:00 EST) δ Cs calculated using modelled (red) 12 13 and measured (blue) CO2 and CO2 (with Keeling analysis) at Downsview in January-March 13 2016. The black curve shows the SOCE inventory sector-weighted overnight averaged δ Cs. b) 13 Box-plots of the distribution of overnight (22:00-06:00 EST) δ Cs calculated using modelled and 12 13 measured CO2 and CO2 (with the Keeling analysis) and the SOCE inventory. The boxes show the inter-quartile range and median values of each sample, the whisker top and bottom show the 90th and 10th percentiles, respectively, and the circle markers show the mean value of each sample.

128

13 To improve the representativeness of the calculated δ CS, we applied the same Keeling plot linear 13 mixing model, Equation 4.5, to calculate δ CS for each hour in the simulation using both measured 13 13 and modelled δ CO2. Figure 4.6a shows the individual hourly δ CS calculated via the Keeling 13 model using modelled and measured CO2 and δ CO2 in red and blue respectively; the black curve 13 13 shows the hourly δ CS calculated using the flux-weighted δ Ci from each of the nine source categories in the SOCE inventory. The box and whisker plot in Figure 4.6b displays a distribution 13 of all the individual hours δ CS for each scenario considered. Similar to the results from the 13 overnight analyses, the hourly δ CS calculated by the SOCE inventory is fairly consistent throughout the day, ranging between ~-32 ‰ and -38 ‰ (median =-34.9 ± 2.2 ‰, Table 4.3) with 13 the lightest signatures estimated at nighttime. However, the hourly δ CS estimated by Keeling analyses with modelled and measured data is much more variable than that calculated from the inventory, particularly in the daytime when signatures are predicted to be as heavy as ~-23 ‰. It 13 is possible that the mismatch between the modelled and observed δ CS in the early morning is due to the fact that the GEM-MACH model cannot fully reflect the fast planetary boundary layer height changes and therefore the simple two-component mixing model does not hold. Similarly, the mismatch observed in the late evening could be due to the model’s inability to capture the timing of the boundary layer collapse, again affecting the assumption of a two-component source system. 13 In addition to larger variability, median values for modelled and measured δ CS using the Keeling analysis are again heavier than those calculated from the SOCE inventory, -32.3 ± 3.5 ‰ and - 32.6 ± 1.7 ‰, respectively. If we assume this local enhancement was derived from two end- members with the signatures of C1 and C2, the SOCE inventory has a contribution of 57 % C1 and

43 % C2, Keeling theory with modelled data estimates 73 % C1 and 27 % C2, and Keeling theory with measured data estimates 71 % C1 and 29 % C2, Table 4.3.

If we consider only the nighttime hours (22:00 – 06:00 EST) from this 24-hr plot, there is a much 13 stronger agreement between the δ CS calculated using the SOCE inventory and modelled data, - 38.2 ± 0.5 ‰ and -37.3 ± 1.4 ‰ respectively. Therefore, our results suggest that if nighttime is selected as the simulation period for Keeling analyses, performing an hour-of-day analysis (as 13 shown in Figure 4.6) and selecting nighttime data produces a δ CS that is more representative of the true emission scenario than selecting individual overnight periods (as shown in Figure 4.5).

129

13 Figure 4.6: a) Time series of hourly δ Cs calculated using modelled (red) and measured (blue) 12 13 CO2 and CO2 (from Keeling analysis) at Downsview in January-March 2016. The black curve 13 shows the SOCE inventory sector-weighted hourly averaged δ Cs. Error bars indicate standard deviation in the intercept extracted from the Keeling curves. b) Box-plots of the distribution of 13 12 13 daily δ Cs calculated using modelled and measured CO2 and CO2 (with the Keeling theory) and the SOCE inventory. The boxes show the inter-quartile range and median values of each sample and the whisker top and bottom show the 90th and 10th percentiles, respectively.

130

13 Generally, in both Keeling analyses performed, modelled and measured CO2 and δ CO2 13 calculated δ Cs and contributions of C1 and C2 that were very different than those estimated by the SOCE inventory. This result highlights the need for supplementing Keeling analyses with other source apportionment tools, such as measurements of carbon monoxide (CO) to calculate CO:CO2 or measurements of oxygen isotopes (δ18O), to confirm the interpretation of the Keeling results. 13 Furthermore, the values of δ Cs calculated using overnight data alone were different to those using data from the entire day (-37.7 ± 0.8 ‰ and -34.9 ± 2.2 ‰, respectively, estimated by the SOCE inventory). This emphasizes that although it is advantageous to maximize the Cs when using the 13 Keeling approach to source apportion CO2, the representativeness of the calculated δ Cs is likely compromised and needs to be accounted for when interpreting the results.

4.4.5 Emission sources of C1 and C2

In all of our analyses above, we are confident that C2 (-44 ‰) represents CO2 emissions from the combustion of natural gas (Pugliese et al., 2016). However, we are less confident on the source of

C1 (-28 ‰), because both gasoline/diesel combustion and biogenic respiration (from plants, soil 13 and humans) have δ CO2 signatures similar to -28 ‰. Using data collected at Borden forest (~200 km north of the city of Toronto) (Froelich et al., 2015), and assuming a 28 % tree cover in the GTA

(Urban Forestry, 2013), we were able to estimate that CO2 emissions from plant and soil respiration are negligible (~2 % of the anthropogenic contribution) in the winter months (January – March) in the GTA. Furthermore, a study performed in Paris, France found that human respiration resulted -1 in approximately 1100 kt yr of CO2 emissions (Lopez et al., 2013); with a population density in the GTA approximately one-fifth of that found in Paris, the contribution of CO2 from human respiration in the GTA is likely negligible (~0.2 % of the anthropogenic contribution). Therefore, these results suggest that in our study the source of C1 is likely dominated by gasoline/diesel combustion. However, in the summertime, scaling the plant and soil respiration from Borden

(Froelich et al., 2015) leads to estimates in the range of ~10 % of fossil fuel CO2 in the GTA. 13 Therefore during the growing season, the use of δ CO2 signatures to identify CO2 sources is likely complicated.

131

4.5 Conclusions

13 We presented a framework in which local sector δ CO2 signatures are combined with modelled 13 CO2 mixing ratios to simulate ambient δ CO2 in the GTA. We demonstrate a strong ability to 13 capture the variability in δ CO2 signatures in January-March, 2016, quantified by a low RMSE and MB (1.05 ‰ and 0.04 ‰, respectively). Furthermore, our model simulation was also able to 13 capture the diurnal profile of δ CO2, with an offset of approximately 0.2 ‰ in the afternoon. With 13 an ability to model δ CO2 with a high degree of model-measurement agreement, we used this unique dataset to evaluate the limitations of traditional CMB approximations and Keeling analyses that are used to source apportion CO2. We performed a series of three CMB approximations to 13 estimate δ Cs using modelled and measured data to evaluate whether agreement with that estimated by the SOCE inventory was ever achieved. Our results showed that when the 5th 13 13 percentile was used as an estimate of background CO2 and its δ CO2, estimated δ Cs was heavier compared to that predicted by the SOCE inventory. This result highlights the need for an accurate 13 estimate of background CO2 and δ CO2 in order to properly apportion enhancements of CO2 using CMB. We complemented these analyses by performing two Keeling analyses using modelled and 13 measured CO2 and δ CO2 data for overnight (22:00-06:00 EST) periods as well as for every hour of the day. Consistently, Keeling analyses using modelled and measured data predicted a heavier 13 and more variable δ Cs than that calculated from the SOCE inventory for the same time period.

Furthermore, when the CO2 enhancement was apportioned to non-natural gas and natural gas components, the Keeling analyses estimated very different relative contributions compared with those estimated by the SOCE inventory. The results from these Keeling analyses suggested that studies using carbon isotopes to source apportion CO2 should use complementary tools, such as 18 CO:CO2 or δ O, to help confirm the contributions estimated by the Keeling approaches. In the 13 absence of a full simulation of modelled δ CO2, a CMB or Keeling analysis of measurement data that yielded a very different allocation of sources compared to the bottom-up inventory might be viewed as a reason to doubt the inventory. However, our analysis shows that carrying out the CMB or Keeling analysis on the model data actually yields a source allocation more similar to the measurements than to the underlying inventory.

132

4.6 Contributions

13 The simulated CO2 mixing ratios that were used to model δ CO2 were prepared by Stephanie C. Pugliese, with critical help developing the inventory and running the GEM-MACH model from Felix Vogel, Mike Moran, Junhua Zhang, Qiong Zheng, Shuzhan Ren and Craig Stroud. The 13 ambient CO2 and δ CO2 data were collected by Douglas Worthy and his team at Environment and Climate Change Canada. The MACC and C-TESSEL products used in our model simulations were provided by Gregoire Broquet. The data was analyzed and interpreted for publication by Stephanie C. Pugliese. The manuscript listed at the beginning of this chapter was written by Stephanie C. Pugliese, with critical input from Jennifer Murphy and Felix Vogel.

Acknowledgements

The authors are thankful to Robert Kessler, Michelle Ernst, Lauriant Giroux and Senen Racki for their efforts collecting the CO2 measurements at Environment and Climate Change Canada. Support for this project from the Natural Science and Engineering Research Council of Canada is gratefully acknowledged.

4.7 References

Ammoura, L., Xueref-Remy, I., Vogel, F., Gros, V., Baudic, A., Bonsang, B., Delmotte, M., Te, Y., and Chevallier, F. (2015). Exploiting Stagnant Conditions to Derive Robust Emission Ratio

Estimates for CO2, CO and Volatile Organic Compounds in Paris. Atmospheric Chemistry and Physics, 16, 15653-15664.

Breon, F. M., Boucher, O., and Brender, P. (2017). Inter-Annual Variability in Fossil-Fuel CO2

Emissions due to Temperature Anomalies. Environmental Research Letters, 12, 074009. doi:https://doi.org/10.1088/1748-9326/aa693d

Bush, S. E., Pataki, D. E., and Ehleringer, J. R. (2007). Sources of Variation in D13C of Fossil Fuel Emissions in Salt Lake City, USA. Applied Geochemistry, 22, 715-723.

Djuricin, S., Pataki, D. E., and Xiaomei, X. (2010). A Comparison of Tracer Methods for

Quantifying CO2 Sources in an Urban Region. Journal of Geophysical Research, 115(D11303), 1-13.

133

Environment Canada. (2012). National Inventory Report 1990-2010: Greenhouse Gas Sources and Sinks in Canada. Retrieved 10/07, 2016, from http://www.ec.gc.ca/ges-ghg/

Froelich, N., Croft, H., Chen, J.M., Gonsamo, A., and Staebler, R.M. (2015) Trends of carbon fluxes and climate over a mixed temperate-boreal transition forest in southern Ontario, Canada. Agricultural and Forest Meteorology, 211-212, 72-84.

Gong, W., Makar, P. A., Zhang, J., Milbrandt, J., Gravel, S., Hayden, K. L., Macdonald, A. M., and Leaitch, W. R. (2015). Modelling Aerosol-Cloud-Meteorology Interaction: A Case Study with a Fully Coupled Air Quality Model (GEM-MACH). Atmospheric Environment, 115, 695- 715.

Huang, L., Chivulescu, A., Ernst, D., Zhang, W., Norman, A. -., and Lee, Y. -. (2013).

Maintaining Consistent Traceability in High Precision Isotope Measurements of CO2: A Way to Verify Atmospheric Trends of 13C and 18O. Atmospheric Measurement Techniques, 6, 1685- 1705.

Keeling, C. D. (1958). The Concentration and Isotopic Abundance of Carbon Dioxide in Rural Areas. Geochimica Et Cosmochimica Acta, 13, 322-334.

Keeling, C. D. (1961). The Concentration and Isotopic Abundance of Carbon Dioxide in Rural and Marine Air. Geochimica Et Cosmochimica Acta, 24, 277-298.

Lopez, M., Schmidt, M., Delmotte, M., Colomb, A., Gros, V., Janssen, C., Lehman, S. J., Mondelain, D.,

Perrusel, O., Ramonet, M., Xueref-Remy, I., and Pousquet, P. (2013). CO, NOx, and 13CO2 as Tracers for Fossil Fuel CO2: Results from a Pilot Study in Paris during Winter 2010. Atmospheric Chemistry and

Physics, 13, 7343-7358.

Moran, M. D., Dastoor, A., & Morneau, G. (2013). Long-range transport of air pollutants and regional and global AQ modelling. In E. Taylor, & A. McMillan (Eds.), Air Quality Management - Canadian Perspectives on a Global Issue (pp. 69-98) Springer.

134

Pataki, D. E., Bowling, D. R., and Ehleringer, J. R. (2003a). Seasonal Cycle of Carbon Dioxide and its Isotopic Composition in an Urban Atmosphere: Anthropogenic and Biogenic Effects. Journal of Geophysical Research, 108(D23) doi:10.1029/2003JD003865

Pataki, D. E., Ehleringer, J. R., Flanagan, L. B., Yakir, D., Bowling, D. R., Still, C. J., Buchmann, N., Kaplan, J. O., and Berry, J. A. (2003b). The Application and Interpretation of Keeling Plots in Terrestrial Carbon Research. Global Biogeochemical Cycles, 17(1), 1022-1036.

Pavlovic, R., Chen, J., Anderson, K., Moran, M. D., Beaulieu, P., Davignon, D., and Cousineau, S. (2016). The FireWork Air Quality Forecast System with Near-Real-Time Biomass Burning Emissions: Recent Developments and Evaluation of Performance for the 2015 North American Wildfire Season. Journal of the Air and Waste Management Association, 66(9), 819-841.

Pugliese, S. C., Murphy, J. G., Vogel, F., Moran, M., Zhang, Q., Zheng, Q., Stround, C., Ren, S.,

Worthy, D., and Broquet, G. (2017). High-Resolution Quantification of Atmospheric CO2 Mixing Ratios in the Greater Toronto Area, Canada. Atmospheric Chemistry and Physics, In Prep.

Pugliese, S. C., Murphy, J. G., Vogel, F., and Worthy, D. (2016). Characterization of the Δ13C

Signatures of Anthropogenic CO2 Emissions in the Greater Toronto Area, Canada. Applied Geochemistry, 83, 171-180.

Rudolph, J., Czuba, E., Norman, A. L., Huang, L., and Ernst, D. (2002). Stable Carbon Isotope Composition of Non-Methane Hydrocarbons in Emissions from Transportation Related Sources and Atmospheric Observations in an Urban Atmosphere. Atmospheric Chemistry and Physics, 36, 1173-1181.

Statistics Canada. (2011). Population, Urban and Rural, by Province and Territory (Canada). Retrieved 10/12, 2016, from http://www.statcan.gc.ca/tables-tableaux/sum- som/l01/cst01/demo62a-eng.htm

Statistics Canada. (2012b). Focus on Geography Series, 2011 Census. Statistics Canada Catalogue no. 98-310-XWE2011004. Retrieved 08/13, 2013, from https://www12.statcan.gc.ca/census-recensement/2011/as-sa/fogs-spg/Facts-cma- eng.cfm?LANG=Eng&GK=CMA&GC=535

135

13 Szaran, J. (1998). Seasonal Variations of D C Values and CO2 Concentration in the Air during Vegetation Growth. Isotopes in Environmental and Health Studies, 34, 341-348.

Szaran, J., Niezgoda, H., and Trembackzowski, A. (2002). Respiration and Assimilation Processes Reflected in the Carbon Isotopic Composition of Atmospheric Carbon Doixide. Nukleonika, 47(1), S59-S61.

Talbot, D., Moran, M. D., Bouchet, V., Creview, L. -., Menard, S., Kallur, A., & and the GEM- MACH Team. (2008). Development of a new canadian operational air quality forecast model. In C. Borrego, & A. I. Miranda (Eds.), Air Pollution Modeling and its Application XIX (pp. 470- 478) Springer Science and Business Media.

Toronto Water. (2009). Ashbridges bay treatment plant biogas cogeneration proposal from toronto hydro energy services. ( No. P:\2009\Cluster B\TW\pw09020).

Townsend-Small, A., Tyler, S. C., Pataki, D. E., Xu, X., and Christensen, L. E. (2012). Isotopic Measurements of Atmospheric Methane in Los Angeles, California, USA: Inflence of "Fugitive" Fossil Fuel Emissions. Journal of Geophysical Research, 117(D07308) doi:doi:10.1029/2011JD016826

Vardag, S. N., Gerbig, C., Janssens-Maenhout, G., and Levin, I. (2015). Estimation of 13 Continuous Anthropogenic CO2: Model-Based Evaluation of CO2, CO, D C(CO2) and 14 D C(CO2) Tracer Methods. Atmospheric Chemistry and Physics, 15, 12705-12729.

Vardag, S. N., Hammer, S., and Levin, I. (2016). Evlauation of 4 Years of Continuous

13 D C(CO2) Data using a Moving Keeling Plot Method. Biogeosciences, 13, 4237-4251.

Vogel, F. R., Huang, L., Ernst, D., Giroux, L., Racki, S., and Worthy, D. E. J. (2013). Evaluation 13 of a Cavity Ring-Down Spectrometer for in Situ Observations of CO2. Atmopsheric Measurement Techniques, 6, 301-308.

Widory, D., and Javoy, M. (2003). The Carbon Isotope Composition of Atmospheric CO2 in Paris. Earth and Planetary Science Letters, 215, 289-298.

136

Xu, J., Lee, X., Xiao, W., Cao, C., Liu, S., Wen, X., Xu, J., Zhang, Z., and Zhao, J. (2017). Interpreting the 13C/12C Ratio of Carbon Dioxide in an Urban Airshed in the Yangtze River Delta, China. Atmospheric Chemistry and Physics, 17, 3385-3399.

Zobitz, J. M., Keener, J. P., Schnyder, H., and Bowling, D. R. (2006). Sensitivity Analysis and Quantification of Uncertainty for Isotopic Mixnig Relationships in Carbon Cycle Research. Agricultural and Forest Meteorology, 136(1-2), 56-75.

4.8 Appendix B

13 Figure B1: Time series of measured (blue) and modelled (red) CO2 signatures at the Downsview site for February 2016.

13 Figure B2: Time series of measured (blue) and modelled (red) CO2 signatures at the Downsview site for March 2016.

137 138

Chapter 5 The impacts of precursor reduction and meteorology on ground- level ozone in the Greater Toronto Area

Sections of this chapter appear in the following peer-reviewed publication:

Pugliese, S. C., Murphy, J. G., Geddes, J. A., and Wang, J. M. (2014). The impacts of precursor reduction and meteorology on ground-level ozone in the Greater Toronto Area, Atmospheric Chemistry and Physics, 14, 8197-8207.

5 Chapter 5 h 5.1 Abstract

Tropospheric ozone (O3) is a major component of photochemical smog and is a known human health hazard as well as a damaging factor for vegetation. Its precursor compounds, nitrogen

oxides (NOx) and volatile organic compounds (VOCs), have a variety of anthropogenic and biogenic sources and exhibit non-linear effects on ozone production. As an update to previous

studies on ground-level ozone in the GTA, we present an analysis of NO2, VOC and O3 data from federal and provincial governmental monitoring sites in the GTA from 2000-2012. We show that

over the study period, summertime 24-hr VOC reactivity and NO2 midday (11:00-15:00) concentrations at all sites decreased significantly; since 2000, all sites experienced a decrease in

NO2 of 28-62 % and in measured VOC reactivity of at least 53-71 %. Comparing 2002-2003 to

2011-2012, the summed reactivity of OH towards NO2 and a suite of measured VOCs decreased from 8.6 to 4.6 s-1. Ratios of reactive VOC pairs indicate that the effective OH concentration experienced by primary pollutants in the GTA has increased significantly over the study period. Despite the continuous decrease in precursor levels, ozone concentrations are not following the same pattern at all stations; it was found that the Canada-Wide Standard for ozone continues to be exceeded at all monitoring stations. Additionally, while the years 2008-2011 had consistently lower ozone levels than previous years, 2012 experienced one of the highest recorded summertime ozone concentrations and a large number of smog episodes. We demonstrate that these high ozone

139 observations in 2012 may be a result of the number of days with high solar radiation, the number of stagnant periods and the transport of high ozone levels from upwind regions.

5.2 Introduction

At ground level, O3 is toxic to both humans and vegetation as a result of its ability to oxidize biological tissues (Bell et al., 2005). It is well documented that human exposure to ozone leads to respiratory symptoms and increased risk for hospital admissions (Bell et al., 2004) and as a result of these and reduced agricultural yields, yearly economic losses attributed to ozone and fine particulate matter pollution are as high as $9.6 billion in the province of Ontario alone (MOE., 2005). In response to this, many governments and regulatory agencies have imposed air quality standards to protect the population against exposure to O3 and other pollutants. In 2000, Canada adopted a Canada-Wide Standard (CWS) for ozone, which states that the 3 year average of the 4th highest daily maximum 8-hr average should not exceed 65 ppb. In 2007 the City of Toronto, Canada’s largest urban area, made a commitment in its Climate Change Clean Air and Sustainable Energy Action Plan to reduce emissions of local smog-causing pollutants 20 % below 2004 levels by 2012 (City of Toronto., 2007). This regulation is happening in the context of regional scale initiatives to control emissions of ozone precursor compounds, such as the “Drive Clean” vehicle test program in Ontario (MOE., 2013) and the phasing out of coal-fired power generating stations

(Bradley, 2013). However, as a secondary pollutant, O3 has proven to be one of the most difficult pollutants to bring into compliance with air quality standards. Its precursor compounds, VOCs and

NOx (NOx = NO + NO2), are emitted by a variety of anthropogenic and biogenic sources, the former dominated by combustion, fuel evaporation and chemical manufacturing while the latter dominated by fossil fuel combustion and transportation (ICF., 2007). The production rate of ozone depends on precursor concentrations in a non-linear fashion. Furthermore, previous studies of O3 concentrations have indicated considerable variability from day to day and from year to year as a result of not only changes in precursor emissions but also meteorology (Agudelo-Castaneda et al., 2013; Figueiredo et al., 2013; Jacob et al., 1993; Pekey and Ozaslan, 2013; Psiloglou et al., 2013).

Empirical studies have shown very strong positive correlations between O3 production and temperature (Wolff and Lioy, 1978) and weaker correlations of ozone accumulation with wind speed and direction, pressure, cloud cover and humidity (Jacob et al., 1993). All of these

140 meteorological parameters affect the photochemistry occurring in the troposphere and therefore the rate of O3 production.

In the troposphere in the summertime, ozone is produced rapidly via the photochemical oxidation of VOCs in the presence of NOx. During the day, the interconversion of NO and NO2 occurs with

O3 on the order of minutes (following R5.1a-5.2).

NO2 + hv  NO + O (R5.1a)

O + O2 + M  O3 + M (R5.1b)

NO + O3  NO2 + O2 (R5.2)

This chemistry produces a null cycle with respect to NOx and O3, there is no net production or consumption of either. However, in the presence of VOCs, net production of O3 can occur following oxidation of a hydrocarbon (RH, where R is any organic group) by the hydroxyl radical,

OH, producing an organic peroxy radical, RO2 (R5.3). The organic peroxy radical can then further react with NO to form NO2 and an organic alkoxy radical, RO (R5.4).

OH+ RH + O2  H2O + RO2 (R5.3)

RO2 + NO  RO + NO2 (R5.4)

The NO2 formed in R5.4 can photolyze during the daytime to regenerate NO and an oxygen atom

(R5.1a), which can then recombine with an oxygen molecule to form O3 (R5.1b). There are several possible fates for the RO radical, it may react with O2, thermally decompose or isomerize.

Typically, carbonyl compounds and an HO2 radical are produced (R5.5) and the net reaction (R5.1- R5.6) results in the formation of two ozone molecules (R5.7).

RO + O2  R’CHO + HO2 (R5.5)

HO2 + NO  OH + NO2 (R5.6)

RH + 4O2  R’CHO + 2O3 + H2O (R5.7)

In this study, the total oxidant, Ox, is defined as the sum of NO2 and O3 ([Ox] = [NO2] + [O3]) and therefore Ox can only increase in the presence of VOCs when O3 is formed via reactions (R5.3)

141

and (R5.4) followed by (R5.1a+b), whereas it is conserved when O3 is formed via (R5.2) followed by reactions (R5.1a+b). The advantage of analyzing Ox over O3 is that Ox is a better measure of the photochemical production of ozone as it more closely represents the total oxidant; it is not affected by the titration of O3 with NO.

This catalytic ozone production chain is terminated by the loss of HOx radicals (HOx = OH + RO

+ HO2 + RO2), which can occur by multiple pathways. In an environment with a low NOx:VOC ratio, such as a rural location, peroxy radicals may undergo a self-reaction instead of reacting with NO to produce peroxides or other oxygenated compounds (R5.8).

RO2 + R’O2  ROOR’ + O2 (R5.8)

Under this pathway, the addition of NOx increases O3 production by enhancing the rate of (R5.6) while addition of VOCs has a negligible effect as they can react with nearly every OH produced.

Thus, O3 production increases linearly with increasing NOx and is less sensitive to VOC reactivity.

However, in an environment with a high NOx:VOC ratio, the dominant sink for HOx is the oxidation of NO2 by OH, forming nitric acid (R5.9).

NO2 + OH + M  HNO3 + M (R5.9)

Under this pathway, the addition of NOx decreases O3 production because NO2 can compete with

VOCs for OH. Thus, O3 production becomes inversely proportional to NOx levels and more sensitive to VOC reactivity.

Alternatively, chain termination can also occur following the reaction between peroxy radicals and

NOx, forming peroxy acyl nitrates (PNs = RC(O)O2NO2, R5.10) or alkyl nitrates (ANs = RONO2,

R5.11) (Farmer et al., 2011). PNs serve as a temporary reservoir for NOx and suppress Ox formation in the near-field but transport and release NOx in the far-field, extending the formation of ozone (Perring et al., 2010). Conversely, ANs are considered permanent sinks for NOx, affecting only local Ox production (Perring et al., 2010).

RC(O)O2 + NO2 ↔ RC(O)O2NO2 (R5.10)

RO2 + NO + M  RONO2 + M (R5.11)

142

This analysis extends an earlier study by Geddes et al. (2009) that demonstrated how summertime

VOC reactivity and ambient concentrations of NO2 decreased from 2000 to 2007 by up to 40 % in the Greater Toronto Area (GTA), but no significant ozone reductions were observed. It was argued that decreased titration may have contributed to higher O3 mixing ratios, and that air transport from the southwest may have contributed to early morning Ox levels (Geddes et al., 2009). In this work, we discuss how GTA Ox levels now appear to be responding to decreases in precursor compounds, as significant reductions are observed from 2000 to 2012. We also identify that in the short term,

2012 marked one of the highest years in GTA Ox during the study period. We discuss the strong link between Ox levels and local meteorology, particularly solar radiation as well as the degree of regional air transport, to help explain this occurrence. We also examine changes in OH reactivity and abundance over the study period.

5.3 Methods

5.3.1 Study region and data collection

The City of Toronto (43o40’ N, 79o23’W) is located in Southern Ontario on the northwest shore of Lake Ontario, and is the largest urban area in Canada as well as one of its most densely populated regions (945.4 persons per square kilometre) (Statistics Canada., 2012). The Greater Toronto Area comprises five municipalities, Halton, Durham, Peel, Toronto and York, which together have a population exceeding 6 million (Statistics Canada., 2012). During the summer, the GTA is affected by warm southerly and south-westerly air transport, as well as local land-lake breezes from Lake

Ontario (Makar et al., 2010). The city’s NOx emissions are dominated by the transportation sector (63 %), with diesel trucks accounting for a disproportionately large percentage (36 %) (ICF., 2007). When considering only emissions made directly in the city, transportation accounts for an increasingly larger amount (73 %) of total NOx emissions, with diesel vehicles accounting for 45 % of this total (ICF., 2007). In Toronto, anthropogenic VOC emissions are almost exclusively from gasoline powered cars and light trucks (ICF., 2007). The GTA experiences frequent smog episodes despite efforts made by the city to reduce emissions of smog precursors.

NOx, O3 and VOC data used in this study were obtained from the National Air Pollution

Surveillance (NAPS) network. For the NOx and O3 analyses, eight sites across the GTA were selected, four of which are considered urban (Downtown2, Toronto North, Toronto East, and Toronto West2) and four which are considered suburban (Oshawa, Brampton2, Newmarket and

143

Oakville), see Figure 5.1. Hourly data for both NOx and O3 are available from the year 2000 onwards and are publically accessible at http://www.airqualityontario.com. At all stations, NO2 and O3 measurements were made by automated continuous chemiluminescent and UV-absorption analyzers respectively, with sampling heights varying from 4-12 m above ground level. During the same study period, VOC data were obtained from Environment Canada for four stations across the GTA (Downtown1 (sampling from 2002-present), Toronto West1 (sampling from 2000-2010), Junction (sampling from 2000-2005) and Brampton1 (sampling from 2001-2010)), also shown in Figure 5.1. At all stations, 24-h samples were collected once every 6 days by evacuated electropolished stainless steel canister and sent to Environment Canada (Ottawa, Ontario) for analysis by gas chromatography/flame ionization detection (for C2 hydrocarbons) and gas chromatography/mass spectrometric detection (for C3-C12 hydrocarbons) as described in Wang et al. (2005). In 2000, 2001, 2004 and 2005, carbonyl compounds were automatically sampled at the Junction site using 2,4-dinitrophenylhydrazine coated silica Sep-Pak cartridges for 24 h, then separated and identified using HPLC and UV DAD detection at 365 nm. Additional VOC sampling was carried out in 2011 at the Downtown Toronto sampling site (Downtown 1) using an automatic sampler (model 910PC, XonTech Inc., VanNuys, CA) with eight consecutive 3-h samples collected every 24 hours over five days during late summer. Samples were analyzed for both non- polar and polar VOCs using GC techniques described in (Wang et al., 2005).

144

Lake Ontario

Figure 5.1: Monitoring stations in the Greater Toronto Area used for the collection of NOx and O3 data (subscript “a”), VOC data (subscript “b”), and meteorological data (subscript “c”).

5.3.2 O3, NOx and VOC analyses

Daily 8-hr maximum O3 and Ox were calculated for each site throughout the thirteen year study period. The O3 diurnal cycle was examined and it was determined that the most photochemically relevant hours for O3 production were between 11:00 and 15:00. Therefore, NO2 hourly data were averaged during this time period (hereafter referred to as the “NO2 midday average”). Using this data, annual summer averages were calculated (where summer is defined as May to September inclusive) as it is the most photochemically relevant time of year.

VOC reactivity was calculated as the product of the VOC’s number density and its rate constant against the hydroxyl radical (Atkinson, 1997; Seinfeld and Pandis, 2006). Summer annual averages for total VOC reactivity were calculated as the sum of all VOC reactivity (∑ki[VOCi]) in units of inverse seconds. The influences of biogenic, anthropogenic and oxygenated VOCs (OVOCs) were distinguished by identifying the sum of isoprene, cymene, pinene, limonene and camphene as biogenic VOCs, the sum of formaldehyde, acetaldehyde, acrolein, acetone, and propionaldehyde

145 as OVOCs (applicable for only the Junction site) and the sum of the remaining 40 VOCs as anthropogenic.

For all compounds, annual trends were calculated by a linear regression analysis and their significance measured by their p-values. P-values were calculated from a standard T-test where the null hypothesis being tested is that the slope of the regression line is equal to zero. Annual trends with p-values < 0.05 are considered significant.

5.3.3 Meteorological analyses

Hourly meteorological data was continuously collected throughout the study period at Toronto’s Pearson International Airport and this data was accessed from National Climate Data and Information Archive operated by Environment Canada. Maximum daily temperatures were calculated for each summer and the number of days where this maximum exceeded 30 °C was recorded. From this same archive, hourly wind speed and direction were obtained and net wind vectors were calculated for a 12 hour period prior to each afternoon (00:00 to 12:00) to assess air mass history. This technique was adopted from Geddes et al. (2009) where the magnitude of the x- and y- component of the hourly wind vector was summed and used to calculate the resultant vector by trigonometry. The result is a single vector that represents the strength and the net direction of air transport for each day. This approach was chosen because the location of the GTA on the northern shore of Lake Ontario leads to frequent local lake breeze flow patterns that are not generally well-represented in back trajectory analyses (Sills et al., 2011).

To assess the level of photochemical activity occurring on each day, the amount of incoming solar radiation was analyzed. Data were collected using both a Net Radiometer (CNR1, Campbell Scientific Corp.) and a Pyranometer (CMP 11, Kipp and Zonen B.V.) operated at the University of Toronto Mississauga Department of Geography’s Meteorological Station (UTMMS). From the hourly data, midday (11:00 to 15:00) solar radiation averages were calculated in W m-2.

5.4 Results and Discussion

5.4.1 Long term precursors and O3 levels (2000-2012)

Annual summer midday averages of daily NO2 are shown in Figure 5.2, demonstrating that an overall decrease is present throughout the thirteen year study period (2000-2012). The urban sites

146 have the steepest slopes, between -0.64 and -0.92 ppb∙year-1 (p<0.01) while the suburban sites also have decreasing trends of -0.20 to -0.55 ppb∙year-1 (p<0.05). Overall, the summer averages in 2012 were at least 45 % lower than 2000 levels at the urban sites (the largest difference is in North Toronto with a decrease of 54 %) and at least 28 % lower at the suburban sites (the largest difference is in Brampton2 with a decrease of 62 %). The most urban station (Downtown2) and that closest to a major highway (West2) continue to report the highest NO2 levels of all the monitoring stations. Furthermore, the stations furthest removed from the urban center and major highways, Newmarket and Oakville, report the lowest NO2 levels. From 2000 to 2010, the number of registered vehicles in Ontario has increased from approximately 8.6 to 10.6 million (Statistics

Canada., 2011) and therefore the decrease in NO2 levels is likely related to improvements in vehicle catalyst technology or the phasing out of older, less-efficient vehicles (MOE., 2013). Other factors may include the closure of the Lakeview Generating Station within the GTA in 2005 (NO2 emissions in 2004 were 5000 tonnes), as a part of Ontario’s phasing out of coal-fired power generating stations (Bradley, 2013), and the large reduction of emissions from the Nanticoke

Generating Station, located less than 100 km southwest of the GTA (NO2 emissions decreased from 38000 to 3000 tonnes between 2002 and 2012).

a) b)

Figure 5.2: Annual summer midday NO2 concentrations (ppb) in GTA urban (a) and suburban areas (b); slopes in ppb year-1. Annual summer VOC reactivity (s-1) (c); slopes in s-1 year-1.

147

A similar decreasing trend is apparent for anthropogenic VOCs in the GTA. Annual summer averages of VOC reactivity are shown in Figure 5.3 and a steadily decreasing trend for anthropogenic VOCs is apparent across all sites. Monitoring at Junction stopped in 2005 and at West1 and Brampton1 in 2011; therefore only Downtown1 data are available for 2011 and 2012. All sites have statistically significant declining slopes, between -0.16 and -0.23 s-1∙year-1 (p<0.01) and overall decreases of at least 53 % from 2000 levels (excluding Junction site), with the Downtown1 site experiencing the largest decrease of 62 %. VOC reactivity from biogenic sources shows no apparent trend across the study period. It is clear that at all sites, VOC reactivity from anthropogenic emissions is approaching the reactivity from biogenic emissions, particularly at the West1 site.

Figure 5.3: Annual summer VOC reactivity (s-1); slopes in s-1 year-1.

Figures 5.4 and 5.5 show annual summer daily 8-h maximum O3 and Ox for the GTA. Following the reductions of both its precursor compounds, O3 and Ox levels have also generally decreased over the study period. Linear regression analyses show that the eight sites have negative slopes for -1 -1 Ox, ranging between -0.91 and -1.1 ppb∙year at the urban stations and -0.57 and -1.0 ppb∙year at the suburban stations, most of which are statistically significant (with the exception of Oshawa and Oakville which are both missing data at the beginning of the study period). At all sites slopes are negative for O3 but are not statistically significant (with the exception of Newmarket). Over the entire study period, decreases in Ox of 2.4-7.6 % occurred in the urban stations and 3.5-13.9 %

148

at the suburban stations. If 2012 data is removed from the figures, the slopes for Ox become significant (p < 0.01) at all sites, except for Oshawa which has only been monitoring since 2005 (p < 0.07).

Figure 5.4: Annual summer average maximum 8-hr O3 concentrations (ppb) for Toronto urban (left) and suburban (right) areas; slopes in ppb year-1-.

Figure 5.5: Annual summer average maximum 8-hr Ox concentrations (ppb) for Toronto urban (left) and suburban (right) areas; slopes in ppb year-1-.

149

5.4.2 O3 levels from 2008-2012

While over the entire study period Ox levels in GTA have decreased, from 2008 onwards the same trend is not observed. Figure 5.5 shows that at all monitored sites, between 2008 and 2011 Ox levels in the GTA were consistent lower than from 2000-2007. However, the measured average summer daily maximum 8-h Ox in 2012 was significantly higher than that of the previous four years (an increase of 3-4 ppb). The data presented above demonstrates that NO2 levels have not significantly changed from 2011 to 2012 (the largest change was a decrease of 1.7 ppb at the Toronto North site). Additionally, results show that VOC reactivity decreased from 2011 to 2012.

Therefore, the 2012 Ox increase cannot be explained as a result of a change in either of the precursor compounds. Alternatively, this variability in Ox may be explained by meteorological influences, which is explored in the following section.

5.4.3 Meteorological influences on O3

Previous studies have provided evidence of a correlation existing between ozone levels and meteorological conditions, such as wind direction and speed, temperature and relative humidity (Agudelo-Castaneda et al., 2013; Figueiredo et al., 2013; Pekey and Ozaslan, 2013; Psiloglou et al., 2013). Specifically, Jacob et al. (1993) discuss the significant dependence of O3 concentrations on temperature, indicating that at higher temperatures, local O3 production is maximized as a result of the suppression of radicals being stored as peroxyacetylnitrate (PAN). Higher temperatures may also result in enhanced local production of O3 by increasing HOx production, or increasing local biogenic VOC emissions. Figure 5.6 displays the number of days each year where the 8-hr O3 average exceeded 65 ppb, the number of days which experienced temperatures exceeding 30°C , th as well as the Design Value (the 4 highest 8-hr O3 measurement annually, averaged over three consecutive years (Environment Canada, 2013)) for that year. A relationship between ozone exceedances and high temperatures is clear during the 2000-2009 period, but less clear from 2010-

2012. Although 2012 had a large increase in the number of days exceeding the O3 standard, it did not have significantly more days experiencing warmer temperatures compared to 2010 or 2011. Therefore, based on this data, the warm temperatures experienced in the summer of 2012 did not play the dominant role in the Ox increase observed.

150

Figure 5.6: The number of days exceeding 65 ppb O3 Canada-wide Standard (solid lines and triangles) and the number of days exceeding 30 °C (dashed lines and squares) at the Toronto North station. The number above each marker is the Design Value for that year.

High temperatures in the GTA are often associated with warm southerly flow and therefore it is possible that the correlation between O3 and temperature is driven by air transport from upwind regions. As described in Section 5.3.3, net wind vectors for each summer day from midnight to noon were determined. Following Geddes et al. (2009), we defined days when the wind speed and direction resulted in a net movement of >120 km from the south and southwest (135o-270o) as days influenced by “west to south-east flow (W-SE)”; days when the net wind speed and direction resulted in a net movement of >120 km from the north (270o-45o) as days influenced by “west to northeast flow (W-NE)”; and days when the net wind movement was <120 km as “stagnant” or “local”. From this designation, days that were affected by W-SE air transport represent days most likely influenced by polluted air masses from surrounding urban areas in Canada and the US, whereas days that were affected by W-NE air transport represent days most likely influenced by transport from remote regions. Table 5.1 shows the percentage of days affected by each wind designation from 2008-2012. Throughout this period, the summer of 2012 was affected by a large number of days with air transport from the W-SE, the fewest days with air transport from the W-

151

NE and the highest number of stagnant periods. This could have contributed to the high Ox levels measured in the GTA in 2012; more days were being affected by polluted air transport from the south and air being trapped over the city (allowing precursor levels to accumulate and enhance local O3 production) as well as the fewest days with “clean” air transport from the north.

Table 5.1: Percent of days and exceedances (at Downtown2) each summer (2008-2012) that were affected by air transport from the W-NE, W-SE or local/stagnant air. Year Days with Exceedances Days with Exceedances Days with Exceedances transport from transport from transport from from from from W-NE W-SE Local 2008 29 0 18 17 52 83 2009 33 20 14 40 53 40 2010 29 0 26 33 45 66 2011 38 33 9 17 53 50 2012 25 6 20 29 54 65

Figure 5.7 displays the hourly summer averages at Toronto North for Ox from the three designated air directions as well as from all directions. Overall, air transport from W-SE or locally produced resulted in the highest daily maximum Ox levels. Between 2010 and 2012, the maximum daily Ox reached when air arrived from W-NE was consistent, a result expected as this air traveled over remote regions and therefore O3 production is likely NOx limited and dominated by biogenic VOC reactivity. The maximum daily Ox reached in 2012 when air was arriving from W-SE was significantly higher than in 2010 or 2011; this occurrence helps to explain the increase in Ox observed in 2012. Lastly, the maximum daily Ox reached as a result of stagnant conditions was again higher in 2012 than in 2010 or 2011. This increase in local Ox concentrations cannot be explained by wind transport or a change in precursor concentrations, as explained in Section 5.4.1, and therefore other meteorological influences must be affecting the local maximum level of Ox reached.

152

70 70

60 W-NE 2010 60 Local 2010 W-NE 2011 Local 2011

50 W-NE 2012 50 Local 2012

(ppb)

(ppb)

3 3 40 40

30 30 Downtown O 20 O Downtown 20

10 10

0 5 10 15 20 0 5 10 15 20 Hour Hour

70 70 W-SE 2010 60 W-SE 2011 60 2010 (all data) W-SE 2012 2011 (all data)

50 50 2012 (all data)

(ppb)

(ppb)

3 3 40 40

30 30

Downtown O 20 Downtown O 20

10 10

0 5 10 15 20 0 5 10 15 20 Hour Hour

Figure 5.7: Ox hourly summer averages at North Toronto.

Since the local Ox levels experienced in 2012 cannot be explained by changes in air transport or precursor concentrations, it is likely a result of changes in the photochemistry occurring in the troposphere. The level of photochemical activity occurring during each summer day was analyzed by considering the amount of incoming solar radiation in the GTA. Data collected from a net radiometer at UTMMS was analyzed for the period 2008-2012. Figure 5.8 displays a regression - between Downtown2 maximum 8-hr O3 levels (ppb) and midday incoming solar radiation (W m 2 2 ) in 2012. From the R value, it is clear that the variance in O3 levels can partly (~16 %) be explained by midday radiation levels (variances ranged from ~8-19 % from 2008-2012). Based on this figure, a midday average of ~600 W m-2 is required for an ozone exceedance. Figure 5.9 shows the annual cumulative distribution function of the midday average summer solar radiation data plotted against the number of days in each summer. For the period 2008-2012, 2012 experienced the greatest number of days affected by a midday solar radiation average >600 W m-2; 2012 experienced 103 days of >600 W m-2 levels of incoming solar radiation whereas 2011 and 2010 experienced 84, 2009 experienced 92 and 2008 experienced 85. When this analysis was performed

153 with the pyranometer data from the UTMMS site, the same qualitative result was obtained. The amount of incoming solar radiation will strongly influence the production of HOx radicals, and thus the ability for O3 to be produced; local O3 production is minimized on “cloudy” days with lower levels of incoming solar radiation and maximized on “clear” days with medium to higher levels of incoming solar radiation.

80 y = 0.024x + 30.8 (p < 0.01) 2 R =0.16

70

(ppb) 3 60

50

40 Downtown8-hr O 30

200 400 600 800 2 Incoming Solar Radiation (W/m ) 2 Figure 5.8: Regression of Downtown2 8-hr O3 versus incoming solar radiation (W/m ) for 2012 (black lines are the thresholds of incoming solar radiation where 65 ppb of O3 is exceeded).

b 1000

) )

-2 800

600 2012 2011 2010 400 2009

2008 Midday Solar Radiation (W m (W Radiation Solar Midday 200

+ + + +

0 20 40 60 80 100 120 140 160 Number of days in the summer Figure 5.9: Cumulative distributions of the midday average solar radiation (W/m2) experienced on each summer day (2008-2012).

154

5.4.4 VOC – or NOx-limited ozone production

The ozone production regime is dictated by the fate of HOx radicals, whether the RO2 radical self- reacts (denoting a NOx-sensitive regime) or the OH radical reacts with NO2 to form HNO3 (denoting a VOC-sensitive regime) (R5.8 and R5.9). Based on the data available, we used the relative reactivity of OH to NO2 and the sum of speciated VOCs at the Downtown site as an approach to approximate the ozone production regime during the study period. Figure 5.10 shows OH reactivity to each species in “early” (2002-2003) and “late” (2011-2012) periods of the data set. In the early period, NO2 accounted for a larger portion of OH reactivity relative to VOCs, 4.93 and 3.64 s-1, respectively. In the late period, although reactivity toward both compounds decreased,

NO2 reactivity to OH accounts for an increasingly larger portion of total reactivity, 2.88 vs. 1.68 s-1 for VOC reactivity. There are limited long-term CO measurements in the GTA during this period, however data from the Toronto West2 station suggests that midday reactivity with CO decreased from 1.4 s-1 to 0.77 s-1 between 2003 and 2012. These results suggest that the preferred fate of OH radical in the late period is reaction with NO2 (R5.9), thereby suggesting that ozone production has become more NOx-saturated and more sensitive to VOC reactivity over the last decade.

10

8

VOCs

) -1 6

4

VOCs OHreactivity (s NO2

2 NO2

0 2002-2003 2011-2012

-1 Figure 5.10: OH reactivity (s ) to NO2 and VOCs in the “early” period (2002-2003) and “late” period (2011-2012) at the Downtown sites.

155

This analysis of the GTA ozone production regime is limited because the ambient concentrations of NOx and VOCs were made at sites near the ground whereas ozone formation takes place in a convective layer which can extend hundreds of meters above the surface (Sillman, 1999). Surface observations likely overestimate the average concentration of primary pollutants within the boundary layer, especially for shorter-lived compounds. Another limitation to the VOC analyses performed in this study is the exclusion of OVOCs from all monitoring stations except Junction. In 2011, we performed five separate days of 3-hr canister measurements during a two week period between August 27 and September 12. The OH reactivity to non-methane hydrocarbons was calculated as two fractions: VOCs (same 40 VOCs as chosen from Environment Canada data, with the exception of propylene and isopentane) and OVOCs (23 compounds). Throughout the two- week period, the OVOCs accounted for ~60 % of the total OH reactivity to VOCs. It is possible, however, that this may be an overestimate of midday OH reactivity; for the two week sampling -1 period, the average 24-hr VOC reactivity from the traditional NAPS compounds was 1.6 s and from the OVOC compounds was 2.4 s-1-. Focusing only on samples collected between 12:00 and 15:00, the reactivity from the traditional NAPS compounds was 1.3 s-1 and from the OVOC compounds was 2.3 s-1. Therefore, based on this limited set of more detailed measurements, we can infer that 24-hr average NAPS samples may over-estimate the midday reactivity from the traditional compounds by 19 % and the total VOC reactivity (VOC + OVOCs) by 10 %. These detailed measurements also suggest that ozone production may be more NOx-sensitive than was determined using the traditional VOC measurements. Overall, these results suggest that our current understanding of VOC reactivity in the GTA is limited.

5.4.5 GTA OH radical concentration

Since the OH radical plays a key role in the production of ozone, it is important to understand how its abundance in the GTA has changed over the study period. By monitoring the ratio of two co- emitted VOCs, we can estimate how the concentration of OH has changed, assuming the distribution of distances from the emission sources to the receptor (monitoring site) has not -11 3 -1 -1 changed. Figure 5.11 displays the ratio of two alkenes, 1-butene (kOH = 3.0x10 cm molec s ) -11 3 -1 -1 and cis-2-butene (kOH = 5.6x10 cm molec s ), as well as two aromatic VOCs, 1,2,3- -11 3 -1 -1 -11 3 - trimethylbenzene (kOH = 3.3x10 cm molec s ) and ethylbenzene (kOH = 7.1x10 cm molec 1 s-1). All of these compounds have lifetimes in the atmosphere on the order of a few hours, assuming an OH concentration of 106 molec cm-3, and therefore observations made in downtown

156

Toronto should predominantly reflect oxidation rates in the GTA atmosphere. We assume that the emission ratios of these compound pairs have not changed substantially during the study period. As Figure 5.11 displays, the ratio for each pair of compounds is smaller in the early period than in the late period, with medians changing from 3.5 to 4.5 for 1-butene:2-butene and 0.19 to 0.25 for 1,2,3-trimethylbenzene:ethylbenzene. Since cis-2-butene and ethylbenzene have higher rates of reaction with OH with respect to their paired VOC, this increase in their ratio indicates that the OH concentration in the GTA has increased from the early to late period. Assuming that the rates of HOx production have stayed relatively constant, higher levels of OH in 2011-2012 are consistent with the large reduction in total reactivity, described in Section 5.4.4. An increase in OH radical abundance has implications for ozone production, the relationship between precursor emissions and concentrations, and the oxidative capacity of the troposphere.

2002-2003 (Median = 3.5) 25 2011-2012 (Median = 4.5)

20

15

10

5

0 0 1 2 3 4 5 6 7 a) 1-butene:2-butene

25 2002-2003 (Median = 0.19) 2011-2012 (Median = 0.25)

20

15

10

5

0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 b) 1,2,3-trimethylbenzene:ethylbenzene Figure 5.11: Histograms of 1-butene:2-butene (a) and 1,2,3-trimethylbenzene:ethylbenzene (b) in the early (2002-2003) and late (2011-2012) periods.

157

5.4.6 Assessing the success of the Climate Change, Clear Air and Sustainable Energy Action Plan

In 2007, the City of Toronto made a commitment to reduce emissions of local smog-causing pollutants 20 % below 2004 levels by 2012. Table 5.2 shows the percent differences between 2004 and 2012 NO2 and VOC concentrations at all sites monitored. Reductions in NO2 concentrations ranged from 27.4 – 49.9 % (Toronto North and Brampton2, respectively), providing evidence of the success of the implemented mitigation strategies during the past decade, such as catalytic converters on vehicles and new combustion technology to reduce NOx emissions from power plants (ICF., 2007). Additionally, reductions in VOC concentrations ranged from 31.7 – 52.8 % (Downtown1 and West1, respectively). Again, this result provides support for the success of the City of Toronto strategies aimed at reducing anthropogenic emissions of these compounds, such as the ChemTRAC initiative which includes a focus on reducing the VOC content in paints and wood coatings (Toronto Public Health., 2013). While these results suggest that the emission reduction target may have been achieved, there are limitations to our interpretation of the data.

Monitoring network data suggests that the concentrations of NO2 and VOCs have been reduced by more than 20 %, but the action plan commitments are actually for emissions. In Section 5.4.5 we showed that oxidation rates in the urban atmosphere appear to have accelerated, meaning that changes in emissions and concentrations will have a non-linear relationship, and that the data in Table 5.2 likely overestimate the changes in emissions. Nevertheless, the large reduction in concentrations is likely consistent with a decrease of more than 20 % in emissions, though exact quantification is difficult. Furthermore, as suggested in Section 5.4.4, the role of OVOCs as ozone precursors in the GTA is unclear due to lack of monitoring and therefore these conclusions about the success of the action plan are not necessarily comprehensive.

Table 5.2: The percent difference between 2004 and 2012 NO2 and VOC concentrations (ppb) at all stations monitored (Junction was not included as monitoring stopped in 2005). Station Downtown North East West Oshawa Brampton Newmarket Oakville NO2 31.2 27.4 29.1 42.1 34.6 49.9 26.8 30.3 (%) VOC 31.7 - - 52.8 - 33.4 - - (%)

158

5.5 Conclusion

Significant reductions in the concentrations of O3 precursor compounds have been observed since

2000; NO2 levels decreased by at least 45 % in urban sites and 28 % in suburban sites and VOC reactivity decreased by at least 53 %. These results provide evidence for the effectiveness of implemented vehicle emission clean-up technologies as well as other regulatory initiatives throughout the province and municipalities in the GTA. In response to these decreases, GTA Ox levels have also decreased about 2.4-13.9 % at all sites during the 13 year study period (with statistical significance at most sites). In 2012, however, some of the highest recorded Ox concentrations were observed following four years of consistently low levels. Since concentrations of both precursor compounds have continued decreasing since 2008, this increase must have been influenced by the meteorology experienced in 2012. We found that although the warm temperatures experienced in 2012 did not likely play a role in the Ox increase compared to 2010 and 2011, air transport from upwind regions may have. Compared to the preceding 5 years, the summer of 2012 was affected by a large number of days with air transport from the W-SE (polluted air travelling over urban areas in Canada and the US), the fewest days with air transport from the W-NE (clean air travelling over remote regions) and the most days affected by stagnant conditions

(air trapped over the GTA allowing precursor compounds to accumulate and enhance local O3 production). Additionally, since the production of O3 depends on photochemistry in the troposphere, the levels of incoming solar radiation were analyzed during the study period. It was found that 2012 experienced the largest number of days with levels of incoming solar radiation -2 exceeding 600 W m , which likely contributed to the enhancement of local O3 production.

The results of this study also demonstrate the success of the GTA in achieving the 20 % reduction in precursor emissions set by the Toronto Climate Change, Clean Air and Sustainable Energy

Action Plan in 2007. Between 2004 and 2012, NO2 reductions ranged from 27.4 – 49.9 % and VOC concentration reductions ranged from 31.7-52.8 %. These results provide evidence of the benefits of municipal and provincial regulations aimed at controlling the emission of ozone precursors throughout the past decade. Nevertheless, the ozone design value at all GTA monitoring stations has exceeded the Canada-wide Standard every year between 2002 and 2012. Reductions in precursor emissions appear to have increased in the local abundance of OH, resulting in only moderate reductions in local ozone production rates. The importance of including OVOCs in O3

159 production analyses has been demonstrated, as short-term measurements indicate that they account for a significant fraction of OH reactivity.

5.6 Contributions

The observations of NO2, O3 and VOCs used in this analysis are made publically available by Environment Canada's NAPS (National Air Pollutant Surveillance) and Environmental Monitoring and Reporting Branch at the Ontario Ministry of the Environment. Daniel K. Wang was responsible for the collection of the VOC data. The data was analyzed and interpreted for publication by Stephanie C. Pugliese. The manuscript listed at the beginning of this chapter was written by Stephanie C. Pugliese, with critical input from Jennifer Murphy.

Acknowledgements

Support for this project from the Ontario Graduate Scholarship is gratefully acknowledged.

5.7 References

Agudelo-Castaneda, D., Teixeira, E., Rolim, S., Pereira, F., and Wiegand, F. (2013). Measurement of particle number and related pollutant concentrations in an urban area in south Brazil. Atmospheric Environment, 70, 254-262.

Atkinson, R. (1997). Gas-phase tropospheric chemistry of volatile organic compounds. 1. Alkanes and alkenes. Journal of Physical and Chemical Reference Data, 26, 215-290.

Bell, M. L., Dominici, F., and Samet, J. M. (2005). A meta-analysis of time-series studies of ozone and mortality with comparison to the national morbidity, mortality and air-pollution study. Epidimiology, 16, 436-445.

160

Bell, M. L., Samet, J. M. and Dominici, F. (2004). Ozone and mortality: A meta-analysis of time-series studies and comparison to a multi-city study (the national morbidity, mortality, and air pollution study). Retrieved 11/18, 2013, from http://biostats.bepress.com/jhubiostat/paper57

Bradley, H. J. J. (2013). Bill 138, ending coal for cleaner air act, 2013. Retrieved 03/05, 2014, from http://www.ontla.on.ca/web/bills/bills_detail.do?locale=en&Intranet&BillID=2901

City of Toronto. (2007). Climate change, clean air and sustainable energy action plan: Moving from framework to action. Retrieved 12/14, 2013, from http://www.toronto.ca/legdocs/mmis/2007/pe/bgrd/backgroundfile-4982.pdf

Environment Canada,. (2013). Canada wide standards (CWS). Retrieved 12/19, 2013, from http://www.ec.gc.ca/rnspa-naps/default.asp?lang=En&n=07BC2AC0-1

Farmer, D., Perring, A., Wooldridge, P., Blake, A., Baker, A. Meinardi, S., Huey, L., Tanner, D., Vargas, O. and Cohen, R. (2011). Impact of organic nitrates on urban ozone production. Atmospheric Chemistry and Physics, 11, 4085-4094.

Figueiredo, M., Monteiro, A., Lopes, M., Ferreira, J., and Borrego, C. (2013). Air quality assessment of Estarreja, an urban industrialized area, in a coastal region of Portugal. Environmental Monitoring and Assessment, 185, 5847-5860.

Geddes, J., Murphy, J., and Wang, D. (2009). Long term changes in nitrogen oxides and volatile organic compounds in Toronto and the challenges facing local ozone control. Atmospheric Environment, 43, 3407-3415.

ICF. (2007). Greenhouse gases and air pollutants in the city of Toronto: Towards a harmonized strategy for reducing emissions. Retrieved 08/13, 2013, from http://www1.toronto.ca/city_of_toronto/environment_and_energy/key_priorities/files/pdf/ghg- aq-inventory-june2007.pdf

Jacob, D., Logan, J., Gardner, G., Yevich, R., Spivakovsky, C., and Wofky, S. (1993). Factors regulating ozone over the United States and its export to the global atmosphere. Journal of Geophysical Research, 14, 817.

161

Makar, P., Zhang, J., Gong, W., Stroud, C., Sills, D., Hayden, K., Brook, J., Levy, I., Mihele, C., Moran, M., Tarasick, D., He, H. and Plummer, D. (2010). Mass tracking for chemical analysis: The causes of ozone formation in southern Ontario during BAQS-met 2007. Atmospheric Chemistry and Physics, 10, 11151-11173.

MOE. (2005). Transboundary air pollution in Ontario. Ontario Ministry of the Environment. Retrieved 09/06, 2013, from http://www.ene.gov.on.ca/stdprodconsume/groups/lr/@ene/@resources/documents/resource/std0 1_079137.pdf

MOE. (2013). Drive clean. Retrieved 03/05, 2014, from http://www.ene.gov.on.ca/environment/en/category/drive_clean/index.htm

Pekey, B., and Ozaslan, U. (2013). Spatial distribution of SO2, NO2 and O3 concentrations in an industrial city of Turkey using a passive sampling method. Clean-Soil, Air, Water, 41(5), 423- 428.

Perring, A., Bertram, T., Farmer, D., Wooldridge, P., Dibb, J., Blake, N., Blake, D., Singh, H., Fuelberg, H., Diskin, G., Sachse, G. and Cohen, R. (2010). The production and persistence of

RONO2 in the Mexico City plume. Atmospheric Chemistry and Physics, 10, 7215-7229.

Psiloglou, B., Larissi, I., Petrakis, M., Paliatsos, A., Antoniou, A., and Viras, L. (2013). Case studies on summertime measurements of O3, NO2, and SO2 with a DOAS system in an urban semi-industrial region in Athens, Greece. Environmental Monitoring and Assessment, 185, 7763- 7774.

Seinfeld, J., and Pandis, S. (2006). Atmospheric chemistry and physics: From air pollution to climate change. New Jersey: John Wiley & Sons Inc.

Sillman, S. (1999). The relation between ozone, NOx and hydrocarbons in urban and polluted rural environments. Atmospheric Environment, 33, 1821-1845.

Sills, D., Brook, J., Levy, I., Makar, P., Zhang, J., and Taylor, P. (2011). Lake breezes in the southern great lakes region and their influence during BAQS-met 2007. Atmospheric Chemistry and Physics, 11, 7955-7973.

162

Statistics Canada. (2011). Table405-0004 - road motor vehicles, registrations, annual (number), CANSIM (database). Retrieved 08/14, 2013, from http://www.statcan.gc.ca/tables-tableaux/sum- som/l01/cst01/trade14a-eng.htm

Statistics Canada. (2012). Focus on geography series, 2011 census. Statistics Canada catalogue no. 98-310-XWE2011004. Retrieved 08/13, 2013, from https://www12.statcan.gc.ca/census- recensement/2011/as-sa/fogs-spg/Facts-cma-eng.cfm?LANG=Eng&GK=CMA&GC=535

Toronto Public Health. (2013). Tracking and reducing chemicals in Toronto: ChemTRAC annual report - 2011 reporting year. Retrieved 12/19, 2013, from http://www.toronto.ca/legdocs/mmis/2013/hl/bgrd/backgroundfile-59180.pdf

Wang, D., Fuentes, J., Travers, D., Dann, T., and Connolly, T. (2005). Non-methane hydrocarbons and carbonyls in the lower Fraser valley during PACIFIC 2001. Atmospheric Environment, 39, 5261.

Wolff, G., and Lioy, P. (1978). An empirical model for forecasting maximum daily ozone levels in the northeastern U.S. Air and Waste Management Association, 28(10), 1034.

163

Chapter 6 Conclusions and Future Research Chapter 6 ] 6.1 Conclusions

Urban areas are highly complex environments which, because of their concentrated human activity patterns, are large sources of a variety of atmospheric pollutants, both short- and long-lived. While they occupy only 3 % of the earth’s total land cover, urban areas are responsible for emitting 53 –

87 % of globally emitted CO2, amongst other pollutants (IPCC-WG3, 2014). Atmospheric scientists have become increasingly interested in investigating the emissions and chemistry that occur in the urban environment because of their impacts on (1) air quality and human health and (2) regional and global climate forcing. Globally, a number of research campaigns have been devoted to investigating urban environments, such as MEGAPOLI in European cities (MEGAPOLI, 2011) and INFLUX in Indianapolis, USA (INFLUX, 2017). The work in this thesis has taken advantage of monitoring of a variety of short- and long-lived pollutants in the Greater Toronto Area (GTA), Canada’s largest urban environment, to gain an understanding of their spatial and temporal variability as well as the influence of emissions, meteorology and transport, on their measured mixing ratios.

To better understand the spatial and temporal variability of CO2 emissions in the GTA, we

developed a fine resolution (2.5 km x 2.5 km spatial and hourly temporal) CO2 inventory for southern Ontario, the SOCE inventory. Emissions were estimated according to seven source sectors: Area, Point, Marine, On-road, Off-road, Residential natural gas combustion and Commercial natural gas combustion for the year 2010. When run with the GEM-MACH chemistry

transport model, results showed that we had improved ability to simulate measurements of CO2 at all stations in the GTA relative to a pre-existing inventory, FFDAS v2 (Rayner et al., 2010), quantified by improved R, RMSE and mean bias, particularly when representing the diurnal

profiles of mixing ratios. In addition to our improved ability to model total CO2, an important

contribution is our capability to define the sectoral contributions of CO2 at the hourly timescale.

164

From a policy perspective, this improves our ability to understand appropriate sectors to target for effective mitigation of CO2 emissions.

High resolution inventories, such as the SOCE inventory, provide information about the emissions of CO2 at a spatial and temporal resolution that has proven to be very useful to policy makers when implementing mitigation efforts. The drawback however is that they are based on activity data and emissions factors for a single year and therefore, because of the exhaustive nature of their 13 18 production, may become quickly out of date. Other tracers, such as δ CO2 or δ O, that can help infer source contributions can be used to apportion CO2 as long as the signature of unique end- members is known (Djuricin et al., 2010; Gorka and Lewicka-Szezebak, 2013; Kuc and Zimnoch, 1998; Pataki et al., 2003). The caveat associated with this technique is that signatures are location- 13 specific. Continuous measurements of CO2 and δ CO2 are made at two sites in the SOCE inventory domain, however to source apportion CO2 using this information, GTA-specific signatures of end-members needed to be characterized. Chapter 3 discussed the results of three intensive sampling campaigns performed in 2014 of the primary anthropogenic sources of CO2, 13 such as gasoline, diesel, natural gas, etc., and characterized their δ CO2 signature using isotope ratio mass spectrometry. To our knowledge, this was the first complete characterization of its kind in a Canadian metropolitan area. Our results showed that there is little spatial variability in the 13 δ CO2 of gasoline and diesel measured at various stations across the GTA and the temporal variability observed is likely attributed to changes in parent crude petroleum used, signaling the need for timely characterization of signatures. When the signatures were used in a mass balance 13 approximation with continuous measurements of CO2 and δ CO2 made at the Downsview station, we were able to estimate that natural gas combustion contributes ~38 % annually to total CO2 emitted in the GTA, which aligned well with local inventory estimates (31- 41 %) (EDGAR, 2010; Greening Greater Toronto, 2012).

Chapter 4 combined the work performed in Chapters 2 and 3 in one of the first known attempts to 13 model δ CO2 using a chemical transport model. Hourly simulations of CO2 (modelled in Chapter 13 13 2) and GTA specific δ CO2 (characterized in Chapter 3) were used to simulate hourly δ CO2 signatures. Simulations were compared against hourly measurements made at the Downsview station. Our results showed strong measurement-model agreement, both in terms of hourly 13 13 signatures and calculated monthly δ Cs. We used our ability to model δ CO2 in the GTA to perform analyses to identify the limitations and advantages of traditional Keeling analyses and

165

13 13 mass balance approximations. When we compared the estimated δ CO2 of the local source (δ Cs) from the Keeling analyses to those estimated by the SOCE inventory, we found that the Keeling 13 analyses consistently predicted relatively heavier values of δ Cs. We performed a similar analysis 13 using chemical mass balance to estimate δ Cs for measured and modelled data (generated using two different background estimates: 5th percentile and MACC (Monitoring Atmospheric Composition and Climate). We found that when the 5th percentile was used as an estimate of 12 13 13 background CO2 and δ CO2, the estimated δ Cs was often unrealistic (a positive integer or a number outside of the range of known CO2 sources in the GTA). The results of this chapter 12 13 highlighted the need for accurate estimation of background CO2 and δ CO2 as well as the need for complementary source apportion tools to confirm the contributions estimated by the Keeling and mass balance approaches.

Results from work in Chapter 2 provided evidence of the large impact meteorology, particularly regional-scale transport, can have on the mixing ratios of long-lived pollutants measured in an urban environment. When considering short-lived pollutants, such as NOx and O3, meteorology can also have a pronounced impact. While these short-lived urban pollutants do not contribute to climate change locally, they do have significant impacts on air quality. Unlike the sparse network of monitoring stations available for CO2, there is a relatively dense network already established in the GTA providing continuous measurements of NOx and O3. This information was used in

Chapter 5 to better understand the relationship between ambient concentrations of O3 and its precursor compounds, NOx and VOCs, and to investigate the influence of meteorology and transport on daily and inter-annual variability. In this work, we used continuous measurements of

NOx, VOCs and O3 from 2000-2012 collected by provincial and federal governmental monitoring efforts at various urban and suburban stations across the GTA. Our results showed that despite decreases in NO2 midday concentrations by 28-62 % and VOC reactivities by 53-71 %, O3 concentrations showed only small (and insignificant) decreases. Furthermore, 2012 marked a very high year in terms of O3 concentrations (increases in summer average maximum 8-hour averages of 5-7 ppb relative to previous years). We found that the enhanced O3 concentrations measured in

2012 are likely a result of increased HOx production from elevated levels of incoming solar radiation and influence of transport from areas with high emission of precursor compounds, relative to previous years. Using the change in relative reactivity of OH to NO2 and the sum of the

VOCs considered in our study, we were able to infer that from 2000-2012, O3 production has

166

become more NOx saturated and more sensitive to VOC reactivity, similar to a previous study performed in the region (Geddes et al., 2009). This result was however limited by the fact that measurement of oxygenated VOCs (OVOCs) were not included in the routine monitoring protocol. From a short campaign where canister samples of VOCs were made in Toronto and subjected to a broader range of analysis, we found that OVOCs accounted for ~60 % of total OH reactivity to VOCs and therefore our understanding of the VOC reactivity in the GTA is incomplete.

The work in this thesis has improved our understanding of the spatial and temporal variability of a variety of short- and long-lived compounds in the GTA. This thesis has provided an important extension to CTMs by creating a methodology to produce a fine resolution CO2 emissions inventory to yield much better model-measurement agreement for CO2 mixing ratios than available with pre-existing global inventories. It has also laid the groundwork for further studies to use 13 13 tracers, such as δ CO2, to source apportion CO2 with an inventory and characterization of δ CO2 of dominant CO2 sources in southern Ontario.

6.2 Updated trends in ozone and its precursors across the GTA

Since the data from Chapter 5 was last analyzed, there is now the opportunity to update the trends with four subsequent years of data. Figure 6.1 shows updated annual summer midday NO2 concentrations measured at urban and suburban sites across the GTA. At most stations, the annual decline in NO2 from 2000-2016 is less steep than those reported for 2000-2012 (Chapter 5). This results in a 0 % to 38 % decline in NO2 reductions based on fully updated trends (2000-2016) relative to original trends from 2000-2012 (calculated as [slope(2000-2012) – slope(2000-2016)] / slope(2000-2012)); Oakville is the only station where the original and updated trend had the same slope. This decline in NO2 reductions appears to be an indication that the rate of decline in reported

NO2 concentrations across GTA monitoring stations is slowing down.

167

Figure 6.1: Updated annual summer midday NO2 concentrations (ppb) in GTA urban (left) and suburban (right) areas. Dashed lines plot the original trend as reported in Chapter 5 (2000-2012); solid lines, slopes (ppb year-1) and p-values refer to updated trends (2000-2016).

Figure 6.2 shows updated annual summer average maximum 8 hour O3 concentration measured at urban and suburban stations across the GTA. Following 2012 where anomalously high concentrations of O3 were observed, summer annual average maximum 8 hour concentrations from 2013-2016 have returned to “normal” levels, although concentrations appear to be slightly increasing over the past couple of years. Updated trends (2000-2016) show a steeper decline relative to original trends (2000-2012) at the Downtown2 and Oakville stations but a less steep or similar decline at all other stations. Decreases at Newmarket continue to be statistically significant (p < 0.05) and decreases at the Brampton2 station are now statistically significant; all other stations continue to experience declines although remain insignificant. While concentrations of NO2 have continued to decrease since the original period, VOC monitoring is no longer routinely performed at multiple stations in the GTA therefore it is difficult to interpret the recent observed changes in

O3 concentrations.

168

Figure 6.2: Updated annual summer average maximum 8 h O3 concentrations (ppb) in GTA urban (left) and suburban (right) areas. Dashed lines plot the original trend as reported in Chapter 5 (2000-2012); solid lines, slopes (ppb year-1) and p-values refer to updated trends (2000-2016).

Figure 6.3 shows updated annual summer average maximum 8 hour Ox (the sum of NO2 + O3) concentrations measured at urban and suburban sites in the GTA. Original trends (2000-2012) showed that Ox declines were not significant at Oshawa and Newmarket however updated trends (2000-2016) show that declines are statistically significant at all urban and suburban sites across the GTA (p < 0.01). However, slopes are slightly less steep or the same from original trends at most sites across the GTA, indicating rates of decline are slowing down.

169

Figure 6.3: Updated annual summer average maximum 8 h Ox concentrations (ppb) in GTA urban (left) and suburban (right) areas. Dashed lines plot the original trend as reported in Chapter 5 (2000-2012); solid lines, slopes (ppb year-1) and p-values refer to updated trends (2000-2016).

Lastly, Figure 6.4 shows an updated plot displaying the correlation between the number of days where O3 concentrations exceeded the air quality standard (CWS of 65 ppb prior to 2015 and CAAQS of 63 ppb afterwards) and the number of days exceeding 30 degrees Celsius. Following

2012 where a large increase in O3 exceedances was observed relative to previous years, 2013-2016 reported a small number of days with O3 exceedances (< 6), despite very warm temperatures observed in summer 2016.

170

Figure 6.4: Updated plot of the number of Canada-Wide Standard (CWS) exceedances (green) and the number of days exceeding 30oC (red) at the Toronto Downtown2 station. The number above each marker is the Design Value for that year.

6.3 Future work

Important uncertainties persist in our understanding of urban environments and the complexity of pollutants that are emitted from these landscapes. Specific possibilities for future research that relate directly to the work described in this thesis are described here.

6.3.1 Direct measurements of anthropogenic CO:CO2 in the GTA

Chapter 2 of this thesis outlined the development of the SOCE inventory available for southern

Ontario. The inventory was built using sector-specific ratios of CO:CO2 to convert an already existing CO inventory built by ECCC. The ratios used for this conversion were based on provincial level data made publically available through the 2010 National Inventory Report, however it is possible that the ratios representative of fuel combustion in the GTA differ from the provincial average, leading to a possible under- or overestimation of calculated CO2. However, it is difficult to confirm 2010 CO:CO2 ratios with measurements made in 2017 because of changes to vehicle fleet and engine combustion efficiency. In 2017, an updated national CO inventory will be made available as well as an updated NPRI, reporting new emission estimates of CO and CO2 in Canada.

171

These new datasets can therefore be combined with measurements made from short-term campaigns characterizing the CO:CO2 ratio of dominant combustion sources in the GTA, allowing the comparison of provincial level estimates to local ratios.

Characterization of CO:CO2 ratios in the GTA not only allows for inventory validation but also provides an additional tracer by which CO2 can be source apportioned; as mentioned in Chapter 3, 13 δ CO2 analyses are limited by the fact that gasoline combustion and respiration have the same isotopic signatures. However these processes can be separately identified by CO:CO2 (as biogenic sources of CO are negligible in the urban environment) and therefore this new dataset could be used to further understand CO2 emissions in the GTA. The use of CO:CO2 to identify the fossil fuel component of CO2 has been performed previously (Djuricin et al., 2010; Gamnitzer et al.,

2006) and therefore a similar framework could be used in the GTA to better understand CO2 sources.

6.3.2 Development of a low precision, high-density network of CO2 sensors in the GTA

Our current ability to understand the spatial variability of CO2 and verify the SOCE inventory relies on a low-density network of high accuracy and precision CRDS instruments. While this network offers important information about regionally averaged CO2 mixing ratios, a much higher spatial resolution is required to resolve CO2 being emitted from individual sources. Similar to the work being done in Oakland, California with the BEACO2N network (Shusterman et al., 2016;

Turner et al., 2016), establishing a spatially dense network of low-cost CO2 sensors in the GTA would allow detection of pollution hotspots and improved verification of the SOCE inventory. Low precision, high-density networks have already shown high potential for robust environmental surveillance; with the BEACO2N network, the low-cost sensors are able to accurately capture expected diurnal and seasonal cycles (Shusterman et al., 2016; Turner et al., 2016). Furthermore, the network can capture differences in daily cycles between nodes, confirming the need of high- density monitoring to accurately represent the complex variability in an urban environment.

6.3.3 Ground level ozone production in the GTA

It is necessary to continue analyzing the trends in NOx, VOCs and O3 across the GTA in the coming years, as it appears that reported NO2 concentrations appear to be stabilizing while O3 concentrations continue to be quite variable. Furthermore, as reported NO2 mixing ratios continue

172 to decrease, there is increasing uncertainty in the associated measurements as interferences in the instrument response due to other oxidized nitrogen (NOz) species become important. This motivates the need to employ other instruments with a conversion technique that is more selective for NO2 at the monitoring stations if an accurate connection is to be made between emission changes and atmospheric concentrations. Moreover, as discussed in Section 6.2, routine measurements of VOCs are no longer performed in the GTA and in an environment that is NOx- saturated, understanding variability in VOC reactivity is especially important. Short-term intensive campaigns which measure VOC (and OVOC) reactivity in addition to an accurate quantification of NO2 mixing ratios may be necessary to compliment the long-term monitoring that occurs across the GTA.

6.3.4 Column observations of O3 and NOx

To complement the in situ observations of O3 and NOx that were used in Chapter 5 of this thesis, column observations from Fourier transform infrared (FTIR) instruments or satellite measurements may be relied upon in the future for understanding concentrations of air quality contaminants in the GTA. The University of Toronto Atmospheric Observatory (TAO) houses an

FTIR spectrometer which has the capability to measure O3, NO, CO as well as a variety of other species. Its use to understand ground-level O3 concentrations in Toronto has already been demonstrated by Whaley et al. (2015). In their study, the in-situ and column O3 measurements had remarkably similar sensitivity and, in some cases, the column measurements had increased sensitivity to detect transport of pollution from the west coast (Whaley et al., 2015). Because of their ability to capture a larger spatial area than in situ measurements, they may provide a better estimate of citywide pollution levels. Column measurements however can be limited in the number of measurements available per day (typically around 3-6 measurements of each species per day) and are reliant on clear skies.

In addition to column measurements from the TAO FTIR spectrometer, satellite measurements of

O3 and NOx could contribute to our understanding of air quality in the GTA. A promising contribution that should be available in 2019 is the TEMPO (Tropospheric Emissions: Monitoring of Pollution) instrument for measurements of O3, NO2, SO2 HCHO and aerosol. TEMPO will be deployed on a satellite in geostationary orbit to provide measurements across the North American

173 continent during daytime (Earth Observation Portal, 2017), providing an increased temporal resolution of measurements.

6.3.5 Column observations of CO2 To evaluate the accuracy of emissions estimated by the SOCE inventory in Chapter 2, in situ hourly observations of CO2 mixing ratios collected at stations across southern Ontario were used. For these analyses, model simulated CO2 was only extracted from the lowest vertical layer, however,

GEM-MACH simulates CO2 up to ~1.45 hPa. Therefore, further extension of this work could extract model simulated CO2 mixing ratios from all model heights to compare with observations from space-based instruments on satellites such as OCO-2 (Crisp et al., 2008). From modelling results in Chapter 2, we estimated that surface level urban enhancements in CO2 generally range from ~2-3 ppm up to ~50-60 ppm. Many studies have previously used satellite observations to identify urban enhancements (Hakkarainen et al., 2016; Kort et al., 2012); Kort et al. (2012) was able to identify persistent enhancements of XCO2 of 3.2 ± 1.5 ppm in Los Angeles using 10-day block averages of satellite observations. Hakkarainen et al. (2016) used OCO-2 XCO2 anomalies (subtracting the daily median from individual observations) to detect enhancements up to 3 ppm of CO2 in their North American domain.

To complement measurements made by OCO-2, in 2021 NASA plans to launch the Geostationary Carbon Cycle Observatory (GeoCARB). GeoCARB will measure total column concentrations of

CO2, CO and CH4 with a horizontal ground resolution of 5-10 km over North, Central and South America. As a geostationary satellite, GeoCARB will provide increased temporal resolution of measured CO2 mixing ratios over that provided by OCO-2. Therefore, in future analyses, it would be interesting to investigate whether the enhancements estimated by SOCE + GEM-MACH will be detectable by GeoCARB, OCO-2 or other space-based instruments.

174

6.3.6 References

Crisp, D., Miller, C. E., and DeCola, P. L. (2008). NASA Orbiting Carbon Observatory: Measuring the Column Averaged Carbon Dioxide Mole Fraction from Space. Journal of Applied Remote Sensing, 2, 1-14.

Djuricin, S., Pataki, D. E., and Xiaomei, X. (2010). A Comparison of Tracer Methods for

Quantifying CO2 Sources in an Urban Region. Journal of Geophysical Research, 115(D11303), 1-13.

Earth Observation Portal. (2017). Tempo. Retrieved 05/17, 2017, from https://directory.eoportal.org/web/eoportal/satellite-missions/t/tempo

EDGAR (2010). Emission Database for Global Atmospheric Research Release Version 4.2 of the European Commission. Retrieved 06/29, 2016, from: Joint Research Centre (JRC)/Netherlands Environmental Assessment Agency (PBL). http://edgar.jrc.ec.europa.eu.

Gamnitzer, U., Karstens, U., Kromer, B., Neubert, R. E. M., Meijer, H. A. J., Schroeder, H., and

Levin, I. (2006). Carbon Monoxide: A Quantitative Tracer for Fossil Fuel CO2? Journal of Geophysical Research, 111(D22302) doi:10.1029/2005JD006966

Geddes, J., Murphy, J., and Wang, D. (2009). Long Term Changes in Nitrogen Oxides and Volatile Organic Compounds in Toronto and the Challenges Facing Local Ozone Control. Atmospheric Environment, 43, 3407-3415.

Gorka, M., and Lewicka-Szezebak, D. L. (2013). One-Year Spatial and Temporal Monitoring of

Concentration and Carbon Isotopic Composition of Atmospheric CO2 in a Wroclaw (SW Poland) City Area. Applied Geochemistry, 35, 7-13.

Greening Greater Toronto, 2012. Living City Report Card 2012. Retrieved 12/04, 2015, from: http://civicaction.ca/.

175

Hakkarainen, J., Ialongo, I., and Tamminen, J. (2016). Direct Space-Based Observations of

Anthropogenic CO2 Emission Areas from OCO-2. Geophysical Research Letters, 43 doi:10.1002/2016GL070885

INFLUX. (2017). INFLUX: Measuring CO2 and CH4 Emissions in Indianapolis . Retrieved 05/11, 2017, from http://sites.psu.edu/influx/

IPCC-WG3. (2014). Climate Change 2014: Mitigation of Climate Change.Working Group III Contribution to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Retrieved 10/12, 2016, from http://www.ipcc.ch/report/ar5/wg3/

Kort, E. A., Frankenberg, C., Miller, C. E., and Oda, T. (2012). Space-Based Observations of Megacity Carbon Dioxide. Geophysical Research Letters, 39(L17806), 1-5.

Kuc, T., and Zimnoch, M. (1998). Changes of the CO2 Sources and Sinks in a Polluted Urban Area (Southern Poland) Over the Last Decade, Derived from the Carbon Isotope Composition. Radiocarbon, 40(1), 417-423.

MEGAPOLI. (2011). MEgacities: EMissions, Urban, Regional and GLobal ATmospheric POLLution and Climate Effects, and INtegrated Tools for Assessment and Mitigation. Retrieved 05/11, 2017, from http://megapoli.dmi.dk/maininfo/projsum.html

Pataki, D. E., Bowling, D. R., and Ehleringer, J. R. (2003a). Seasonal Cycle of Carbon Dioxide and its Isotopic Composition in an Urban Atmosphere: Anthropogenic and Biogenic Effects. Journal of Geophysical Research, 108(D23) doi:10.1029/2003JD003865

Rayner, P. J., Raupach, M. R., Paget, M., Peylin, P., and Koffi, E. (2010). A New Global

Gridded Data Set of CO2 Emissions from Fossil Fuel Combustion: Methodology and Evaluation. Journal of Geophysical Research, 115(D19306) doi:10.1029/2009JD013439

Shusterman, A. A., Teige, V. E., Turner, A. J., Newman, C., Kim, J., and Cohen, R. C. (2016).

The BErkeley Atmospheric CO2 Observation Network: Initial Evaluation. Atmospheric Chemistry and Physics, 16, 13449-13463.

176

Turner, A. J., Shusterman, A. A., McDonald, B. C., Teige, V., Harley, R. A., and Cohen, R. C.

(2016). Network Design for Quantifying Urban CO2 Emissions: Assessing Trade-Offs between Precision and Network Density. Atmospheric Chemistry and Physics, 16, 13465-13475.

Whaley, C. H., Strong, K., Jones, D. B. A., Walker, T. W., Jiang, Z., Henze, D. K., Cooke, M. A., McLinden, C. A., Mittermeier, R. L., Pommier, M., and Fogal, P. F. (2015). Toronto Area Ozone: Long-Term Measurements and Modeled Sources of Poor Air Quality Events. Journal of Geophysical Research: Atmospheres, 120, 11368-11390.