Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

A mobile sensor network to map CO₂ emissions in urban environments Lee, Joseph K. 2016

You don't seem to have a PDF reader installed, try download the pdf

Item Metadata

Download

Media
24-ubc_2016_may_lee_joseph.pdf [ 103.44MB ]
Metadata
JSON: 24-1.0300248.json
JSON-LD: 24-1.0300248-ld.json
RDF/XML (Pretty): 24-1.0300248-rdf.xml
RDF/JSON: 24-1.0300248-rdf.json
Turtle: 24-1.0300248-turtle.txt
N-Triples: 24-1.0300248-rdf-ntriples.txt
Original Record: 24-1.0300248-source.json
Full Text
24-1.0300248-fulltext.txt
Citation
24-1.0300248.ris

Full Text

A mobile sensor network to map CO2 emissions in urbanenvironmentsbyJoseph K. LeeBA, Geography-Environmental Studies, University of California - Los Angeles, 2012a thesis submitted in partial fulfillmentof the requirements for the degree ofMaster of Scienceinthe faculty of graduate and postdoctoral studies(Geography)The University of British Columbia(Vancouver)April 2016c© Joseph K. Lee, 2016AbstractThere are a variety methods that can characterize carbon dioxide (CO2) mixing ratios in theatmosphere at coarse scales, however mapping CO2 emissions and sequestration at fine scalesremains a challenge. In this research, a new method for mapping microscale CO2 emissions incities was developed. First, a compact, mobile CO2 sensor system was built using open sourcehardware and an infrared gas analyzer. Second, a measurement campaign was carried out inwhich 5 mobile sensors were deployed within a 12.7 km2 study area in the City of Vancouver,BC, Canada for 3.5 hours to map CO2 mixing ratios to a grid resolution of 100 m × 100 m. TheCO2 mixing ratios ranged from 382 ppm to 518 ppm and averaged 417.1 ppm and were highestin the downtown and arterial roads and lowest in well vegetated and residential areas. Third,an aerodynamic resistance approach to calculating emissions was used to derive CO2 emissionsfrom the mobile CO2 mixing ratio measurements in conjunction with data collected from a24 m tall meteorological tower in the study area. The measured emissions showed a rangeof -12 to 225 kg CO2 ha−1 hr−1 and averaged 36.39 kg CO2 ha−1 hr−1. Fourth, an emissionsinventory was developed for the study area using emissions estimates derived from buildingsenergy use and traffic counts. The emissions inventory averaged 25.88 kg CO2 ha−1 hr−1 and wasused to compare against the measured emissions. The results showed strong linearity betweenmedian CO2 mixing ratios and the total emissions inventory (R2 >0.9) binned at equal intervals.The results also indicated that the measured emissions and the total emissions inventory werepositively correlated by 78.43% with 99.43% of the measured emissions within ± 1 order ofmagnitude of the emissions inventory. Ultimately, this research demonstrates the possibility ofusing a network of mobile sensors and an aerodynamic resistance approach to map emissions athigh spatial resolution across a city. While further research is necessary, microscale emissionsmaps may be used to better inform urban policy and design as well as engage citizens aboutemissions reductions strategies.iiPrefaceThis thesis is original work completed by the author. Guidance was given by the supervisorycommittee and field assistance was provided by Rick Ketler, Zoran Nesic, Alex McMahon,Andreas Christen, Mark Richardson, Thea Rodgers, Nick (Sung Ching) Lee, Natasha Picone,Wesley Skeeter, Ryan Buchanan, Julie Van de Valk, and Yimei Li.The sensor development described in Section 2.2 was completed in collaboration with ZoranNesic (lead developer), Andreas Christen, and Rick Ketler.Photos were provided by Andreas Christen and Ryan Young.A version of the work in Section 2.2 and Section 3.1 has been published as a poster [Lee,J., Christen, A., Nesic, Z., Ketler, R. Mobile Sensing Approach to Measuring the Urban CO2Dome]. The author acted as lead investigator, composing and presenting the poster at theAmerican Geophysical Union (AGU) 2014 meeting, San Francisco, Ca.A version of the work in Section 2.2, Section 3.1, and Section 3.2 was presented as a talk[Lee, J., Christen, A., Nesic, Z., Ketler, R. A Mobile Sensor Network to Map CO2 in UrbanEnvironments]. The author acted as lead investigator, composing and presenting the talk atthe 9th International Conference on Urban Climate (AGU) 2015 meeting, Toulouse, France.iiiTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiList of Symbols and Acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xAcknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiv1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31.1.1 Limitations of Emissions Inventories Over Cities . . . . . . . . . . . . . . 31.1.2 Need for Measurements in Cities . . . . . . . . . . . . . . . . . . . . . . . 41.2 Research Question and Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . 92 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112.1 Conceptual Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112.1.1 Trace Gas Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112.1.2 Relating CO2 Concentrations to CO2 Flux . . . . . . . . . . . . . . . . . 112.2 Sensor Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152.2.1 Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152.2.2 Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162.2.3 Characterizing System Performance . . . . . . . . . . . . . . . . . . . . . 182.3 Applying the Mobile Sensor System to Map Emissions . . . . . . . . . . . . . . . 232.3.1 Study Area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242.3.2 Vancouver-Sunset Tower . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252.3.3 Grid Averaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262.3.4 Deployment Transects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26iv2.3.5 Sensor Installation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282.3.6 Sensor Deployment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302.3.7 Data Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312.4 Emissions Inventory Validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322.4.1 Emissions Inventory Development . . . . . . . . . . . . . . . . . . . . . . 332.4.2 Measured CO2 Emissions . . . . . . . . . . . . . . . . . . . . . . . . . . . 362.4.3 Emissions Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373.1 Sensor Specifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373.2 Results from the Mobile Sensor Deployment . . . . . . . . . . . . . . . . . . . . . 463.2.1 Meteorology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 473.2.2 Distribution of CO2 Mixing Ratios . . . . . . . . . . . . . . . . . . . . . . 473.2.3 Gridded Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 493.2.4 Measured Emissions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 613.3 Methodology Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 653.3.1 Emissions Inventory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 653.3.2 Relating Measured CO2 Concentrations and the Emissions Inventory . . . 743.3.3 Relating Calculated Emissions to the Emissions Inventory . . . . . . . . . 814 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 914.1 Evaluting the Mobile Sensor Specifications . . . . . . . . . . . . . . . . . . . . . . 914.2 Evaluating the Measurement Campaign . . . . . . . . . . . . . . . . . . . . . . . 944.3 Addressing Issues with the Aerodynamic Resistance Approach . . . . . . . . . . 964.4 Measuring Emissions with Mobile Sensors . . . . . . . . . . . . . . . . . . . . . . 975 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1035.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1035.2 Practical Significance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1045.3 Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106vList of TablesTable 1.1 Table of measurement methods at various urban scales . . . . . . . . . . . . . 5Table 2.1 List of sensor components and materials . . . . . . . . . . . . . . . . . . . . . 18Table 2.2 Table of CDML / NOAA standard calibration gas tank concentration andaccuracy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20Table 2.3 Summary table specifying data filtering parameters . . . . . . . . . . . . . . . 32Table 3.1 Calibration gas tank mixing ratio and accuracy and average observed valuesmeasured by the DIYSCO2 system for each mixing ratio of gas. . . . . . . . . 37Table 3.2 Accuracy test showing RMSE for each DIYSCO2 system at 1 min averages . . 39Table 3.3 Table of percentages data in the grouped inlet with specified error . . . . . . . 45Table 3.4 Table of the percentage of data in the ungrouped inlet test with specified errorin r . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46Table 3.5 Summary table of the DIYSCO2 system evaluation . . . . . . . . . . . . . . . 46Table 3.6 Summary data of the gridded r measurements . . . . . . . . . . . . . . . . . . 52Table 3.7 Summary data of the gridded average r measurements for all grid sizes . . . . 52Table 3.8 Table of values to calculate aerodynamic resistance of heat . . . . . . . . . . . 62Table 3.9 Summary data of the gridded building emissions for all grid sizes . . . . . . . 65Table 3.10 Summary data of the gridded traffic emissions for all grid sizes . . . . . . . . . 68Table 3.11 Summary data of the gridded total emissions for all grid sizes . . . . . . . . . 71Table 3.12 Table of average measured emissions and average building, traffic, and totalemissions by grid size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74Table 3.13 Measured emissions mean vs. emissions inventory mean of all grid sizes forthe study area subset . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86viList of FiguresFigure 1.1 CO2 flux box model in an urban ecosystem . . . . . . . . . . . . . . . . . . . 2Figure 1.2 Relevant atmospheric scales in an urban setting . . . . . . . . . . . . . . . . . 6Figure 2.1 Illustration of the aerodynamic resistance approach to calculating emissions . 14Figure 2.2 Annotated image of “DIYSCO2” system components . . . . . . . . . . . . . . 17Figure 2.3 Illustration of the response time experiment setup . . . . . . . . . . . . . . . 21Figure 2.5 Map of the transect used to perform the inlet tests . . . . . . . . . . . . . . . 23Figure 2.6 Map of the Vancouver study area . . . . . . . . . . . . . . . . . . . . . . . . . 24Figure 2.7 Example images of the local climate zones at street scale . . . . . . . . . . . 25Figure 2.8 100 m vector grid overlaid onto the study area . . . . . . . . . . . . . . . . . 26Figure 2.9 Maps of the 5 predefined routes to ensure comprehensive sampling of thestudy area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27Figure 2.10 Graph of the traffic volumes of the study area . . . . . . . . . . . . . . . . . . 28Figure 2.11 Sensor installation in vehicle . . . . . . . . . . . . . . . . . . . . . . . . . . . 29Figure 2.12 Sample inlet tube run out through the window during vehicle installation . . 29Figure 2.13 Sensor installation on a bike . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30Figure 2.14 Photo showing calibration period . . . . . . . . . . . . . . . . . . . . . . . . . 31Figure 2.15 Workflow diagram illustrating the process of deriving emissions from trafficcounts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34Figure 2.16 Workflow illustration showing the process of driving building emissions . . . 35Figure 3.1 Graphs illustrating the precision of the DIYSCO2 . . . . . . . . . . . . . . . . 38Figure 3.2 Graph illustrating the sensor accuracy over a seven-day measurement period 39Figure 3.3 Graph illustrating the sensor drift from the mean over 7 days . . . . . . . . . 40Figure 3.4 Graph illustrating the response time results . . . . . . . . . . . . . . . . . . . 41Figure 3.5 Grouped inlet test: r - average of all 5 DIYSCO2 systems . . . . . . . . . . . 42Figure 3.6 Grouped inlet test: r - average of all 5 DIYSCO2 systems in well mixed andlow emissions environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42Figure 3.7 Ungrouped inlet test: r - average of all 5 DIYSCO2 systems . . . . . . . . . . 43Figure 3.8 Unrouped inlet test: r - average of all 5 DIYSCO2 systems in well mixed andlow emissions environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43viiFigure 3.9 Error plot: grouped inlet test . . . . . . . . . . . . . . . . . . . . . . . . . . . 44Figure 3.10 Error plot: ungrouped inlet test . . . . . . . . . . . . . . . . . . . . . . . . . 45Figure 3.11 Frequency distribution of measured r sorted by ppm . . . . . . . . . . . . . . 47Figure 3.12 CO2 mixing ratios sorted by latitude for each point measured . . . . . . . . . 48Figure 3.14 Grid averaged r colored by neighborhood and ordered by latitude for the 100m grid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49Figure 3.15 Choropleth map of the grid averaged r for the 100 m grid . . . . . . . . . . . 50Figure 3.16 Choropleth maps of grid averaged r for the 50 m, 200 m, and 400 m grids . . 51Figure 3.17 Map of the skewness per 100 m grid cell . . . . . . . . . . . . . . . . . . . . . 53Figure 3.18 Skewness maps for the 50 m, 200 m, and 400 m grids . . . . . . . . . . . . . 54Figure 3.19 Frequency distribution of the number of samples in each 100 m grid cell. . . . 55Figure 3.20 Frequency distribution of the sample counts in each 50 m grid cell . . . . . . 56Figure 3.21 Frequency distribution of the sample counts in each 200 m grid cell . . . . . . 56Figure 3.22 Frequency distribution of the sample counts in each 400 m grid cell . . . . . . 56Figure 3.23 Map of the number of samples within each 100 x 100 m grid cell . . . . . . . 57Figure 3.24 Maps of the number of samples within each 50 m, 200 m, and 400 m grid cell 58Figure 3.25 Map of the standard deviation from the mean value of each 100 m × 100 mgrid cell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59Figure 3.26 Choropleth maps of the grid standard deviations of the mean r for each 50m, 200 m, and 400 m grid cell . . . . . . . . . . . . . . . . . . . . . . . . . . . 60Figure 3.27 Scatterplot of the measured emissions based on the aerodynamic resistanceapproach ordered by latitude and colored by neighborhood. . . . . . . . . . . 61Figure 3.28 Choropleth map of the grid measured emissions for the 100 m grid . . . . . . 63Figure 3.29 Choropleth maps of the grid measured emissions for the 50 m, 200 m, and400 m grids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64Figure 3.30 Map of building emissions inventory for a 100 m grid . . . . . . . . . . . . . . 66Figure 3.31 Choropleth maps of the gridded building emissions inventory for the 50 m,200 m, and 400 m grids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67Figure 3.32 Map of traffic emissions inventory for the 100 m grid . . . . . . . . . . . . . . 69Figure 3.33 Choropleth maps of the gridded traffic emissions inventory for the 50 m, 200m, and 400 m grids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70Figure 3.34 Map of total emissions inventory for the 100 m grid . . . . . . . . . . . . . . 72Figure 3.35 Choropleth maps of the gridded total emissions inventory for the 50 m, 200m, and 400 m grids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73Figure 3.36 Scatter plot and box-and-whisker plot showing the relationship between mea-sured r and the building, traffic, and total emissions inventories for the 50 mgrid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75viiiFigure 3.37 Scatter plot and box-and-whisker plot showing the relationship between mea-sured r and the building, traffic, and total emissions inventories for the 100m grid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76Figure 3.38 Scatter plot and box-and-whisker plot showing the relationship between mea-sured r and the building, traffic, and total emissions inventories for the 200m grid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77Figure 3.39 Scatter plot and box-and-whisker plot showing the relationship between mea-sured r and the building, traffic, and total emissions inventories for the 400m grid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78Figure 3.40 Graphs showing the linear fits of the r and the binned emissions inventoriesfor building, traffic, and total emissions data for the 100 m grid . . . . . . . . 79Figure 3.41 Graph showing the linear fits of the r and the binned emissions inventoriesfor building, traffic, and total emissions data for (a) the 50 m grid, (b) the200 m grid, and (c) the 400 m grid. . . . . . . . . . . . . . . . . . . . . . . . 80Figure 3.42 Absolute differences between the measured emissions and the emissions in-ventories for all grid sizes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82Figure 3.43 Map of the absolute differences between between the total emissions inventoryand the measured emissions for the 100 m grid . . . . . . . . . . . . . . . . . 83Figure 3.44 Choropleth maps of the gridded absolute differences in the measured emis-sions minus the total emissions inventory for the 50 m, 200 m, and 400 mgrids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84Figure 3.45 Frequency distribution of the relative differences for all the grids . . . . . . . 85Figure 3.46 Double logarithmic plot of measured emissions versus building emissions . . . 88Figure 3.47 Correlation plot of measured emissions versus traffic emissions for all grid sizes 89Figure 3.48 Correlation plot of measured emissions versus total emissions inventory . . . 90Figure 4.1 Linear fit of the gridded median r measurements and the binned emissionsinventory for buildings for (a) the 50 m grid, (b) the 100 m grid, and (c) the200 m grid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98Figure 4.2 Perspective map of the emissions per 100 m grid . . . . . . . . . . . . . . . . 101ixList of Symbols and AcronymsSymbol Definition Unitsba factor for converting m−2 to ha−1 i.e.ba = 104 m2 ha−1bm factor for converting g to kg i.e.bm = 10−3 kg g−1bo factor for converting µmol to mol i.e. bm =10−6 µmol mol−1bt factor for converting s−1 to hr−1 i.e.bt = 3600 s hr−1C Rate of combustion µmol m−2 s−1Cair heat capacity of air J kg−1 m−3c molar concentration of CO2 µmol m−3ctower c at the height of the climate tower(24 m)µmol m−3cmobile c at screen level height (2 m) µmol m−3EC eddy covarianceFc the mass flux of CO2 for a givenarea and timekg CO2 ha−1 hr−1GHG greenhouse gasxGIS geographic information systemsH sensible heat flux W m−2HVAC heating hentilation and airconditioningIRGA infrared gas analyzerLCZ local climate zoneLIDAR light detection and rangingMa molecular mass of dry air 28.97 g mol−1Mc molecular mass of CO2 44.01 g mol−−1OSM OpenStreetMapP Rate of photosytnesis µmol m−2 s−1PDT Pacific Daylight Savings TimePST Pacific Standard TimeR Rate of respiration µmol m−2 s−1r the molar density of CO2; mole ofa CO2 per mole dry air or “ molarmixing ratio”µmol m−1 orppmRSL roughness sublayerraC aerodynamic resistance of CO2 s m−1raH aerodynamic resistance of sensibleheats m−1xiSc volumetric source or sink strength,describes the mass of CO2 emittedper volume and time - the c that isinjected or removed; normalizedper ground area in this studyµmol m−3 s−1T0 surface temperature (0 m)◦CTmobile temperature at the height of themobile sensors (2 m)◦CTtower temperature at the height of theclimate tower (24 m)◦Ct time su horizontal wind speed in West-Eastdirectionm s−1UBL urban boundary layerUCL urban canopy layerv horizontal wind speed inSouth-North directionm s−1w vertical wind speed m s−1w′c′ molar flux µmol m−2 s−1x horizontal distance (Easting) my horizontal distance (Northing) mz vertical distance (Northing) mρa (dry) air density g m−3ρCO2 mass density of CO2 g CO2 m−3xiiθ potential air temperature K... average over a given time′ indicates a turblent deviation fromthe averagexiiiAcknowledgmentsThis work is dedicated to my supervisor and mentor, Dr. Andreas Christen who has alwaysinspired my creativity, supported my curiosity, and shown me the meaning of “good” science. Iwould be hard-pressed to find another mentor more capable and more enjoyable to learn fromand work with.This work would not have been possible without Zoran Nesic and Rick Ketler from whom Ilearned that there is “no small task”. I will be forever grateful for the experience to learn howto make just about anything. I also thank Professor Ron Kellett, Professor Les Lavkulich, Dr.Julia Dordel, and Kaitlin Thaney for their guidance, support, and valuable perspectives.Special thanks to my friend and mentor, Benedikt Groß without whom I would have neverdiscovered the joys of computation, collaboration, and learning through making.I would like to express my appreciation for the UBC Department of Geography’s staff, in par-ticular Sandy Lapsky, Alex Pysklywec, Mimi Yu, Jeanne Yang, Connie Cheung, Julie Ranada,Vincent Kujala, Suzanne Lawrence, and Stefanie Ickert for helping support my graduate schoolexperience.I would like to acknowledge the Terrestrial Ecosystems Research and World-wide Educationand Broadcast (TerreWEB) program for their generous support. I have benefitted tremendouslyfrom the opportunity to integrate science communication into my research efforts and to be partof a community that cares so deeply about the communication of science to the public.I express my gratitude to my family for teaching me the value of hard work and for theirnever ending love and support. I have been privileged to have such strong role models in mylife. Special thanks to Hong, Chris, and Orion for supporting my creativity and nurturing mythirst for adventure. I also thank my friends at home and abroad for being constant sources ofwarmth, adventure, and inspiration. I’m lucky to be surrounded by the most passionate anddedicated researchers, skateboarders, designers, artists, scientists, technologists, advocates, andeverything in between. I particularly acknowledge Ryan Y., Zach P., Sam M., Seigi K., Gab O.,Elanna N., Kelley H., Colleen M., Deirdre L., Sarah MP., Adam M., Craig J., Bart L., FionaH., Jed W., Catherine K., Brian J., and Christopher P. for their special impact on my life overthe years.I thank the faculty and staff at UCLA’s Department of Geography in particular Dr. DanielaCusack, Dr. Jida Wang, Dr. Judith Carney, and Dr. Susanna Hecht for their support andencouragement to pursue research in my undergraduate degree and beyond. I also thank Dr.xivCarlo Ratti and Dr. Dietmar Offenhuber for fostering my interest in visualization and cities.I extend my appreciation for the open- source and science communities and everyone workingto make science, technology, and the arts more inclusive, diverse, and accessible. In particularI extend my thanks to the Mozilla Science Lab–Kaitlin Thaney, Abby Cabunoc, Arliss Collins,Aurelia Moser, Stephanie Wright, Zannah Marsh, Christie Bahlai, Jason Bobe, and RichardSmith-Unna–for providing a transformative fellowship experience. The future of science will bea great one because of these dedicated, passionate, and clever group of open science advocates.Last, I thank the members and friends of the UBC Micromet Lab, in particular ScottKrayenhoff, Carmen Emmel, Nick Lee, Wes Skeeter, Caitlin Semmens, Kesley Everard, NatashaPicone, Julie Van de Valk, Ryan Buchannan, Yimei Li, Thea Rodgers, Mark Richardson, andAlex McMahon for their support, insightful conversations, and help to make my research pos-sible. Many thanks to Andy Black and the Biometeorology Group for use of the lab and sensorequipment.Essential funding for materials and equipment was provided by an NSERC Discovery Grant(Direct measurement of greenhouse gas exchange in urban ecosystems, PI: Christen). Schol-arship funding and training was provided by UBC Geography, NSERC CREATE-TerrestrialEcosystems Research and World-wide Education and Broadcast (TerreWEB), and the MozillaScience Lab. Internship support in 2014 was given by Studio Benedikt Groß and 47Nord.Experiment vehicles were sponsored by Moovel Labs.xvChapter 1IntroductionCities and the cumulative processes of urbanization are key drivers of local and global envi-ronmental change (Kaye et al., 2006; Grimmond, 2007). As cities are the centers of increasingpopulation growth and resource consumption, they are also the dominant source of greenhousegas (GHG) emissions - in particular carbon dioxide (CO2) - into the atmosphere (Rosenzweiget al., 2010). On the global scale, urban emissions are estimated to contribute up to 20%directly and 80% indirectly to the total anthropogenic CO2 emissions footprint (Stern, 2006;Satterthwaite, 2008) and are thus responsible for a major proportion of the greenhouse gasesintensifying positive atmospheric radiative forcing of the troposphere (IPCC, 2007) contributingto global climate change.With more than 50% of the global population now living in cities (United Nations, 2014),addressing global environmental change must happen at the local and urban scale. At theurban scale, the sources of CO2 can be attributed to the combustion of fossil fuels for heat-ing, ventilation, and air conditioning (HVAC), transportation, industrial processes, and powergeneration (Kennedy et al., 2009) along with biological sources, namely soil, plant and humanrespiration; CO2 is also taken up by photosynthesis. An illustration of box-budget model ofCO2 flux - emissions and uptake - is shown in Figure 1.1.1FCΔSCO2 Box Model AtmospherePRCRespirationPhotosynthesisCombustionStoragechangeFigure 1.1: Conceptual representation of CO2 mass flux in an urban ecosystem. The il-lustration shows that the sum of the CO2 flux (Fc) and the change in CO2 storage(∆S) is equal to the total CO2 emitted from fossil fuel combustion (C) and respira-tion (R) of humans, vegetation, and soil minus what is taken up by photosynthesis(P ). Adapted from Christen (2014)2These collective processes result in increased concentrations of CO2 in the urban boundarylayer (UBL) relative to their rural surroundings, a phenomenon termed the “urban CO2 dome”(Idso et al., 2001). The development of the urban CO2 dome is significant because it is directlylinks CO2 emissions in cities to the interactions between the urban form, functions, and climate.Mitigation strategies driven by policy, design, and bottom up citizen engagement at theurban scale have the highest potential for agile and sustained implementation (IPCC, 2014).Central to the reduction of urban CO2 emissions is the availability of reliable emissions invento-ries and methods of validating city-scale emissions estimates. While there are a growing numberof models and methods of quantifying emissions in urban areas, there are disconnects betweenthe current spatial and temporal resolution of emissions models, the ever-evolving urban formand function, and city scale measurements which inform and validate emissions models (Patakiet al., 2009; Kellett et al., 2013).The overall research goal of this thesis is to develop a methodology to map CO2 emissionsin complex environments using a network of mobile sensors, data from an urban climate tower,and an aerodynamic resistance approach to calculating emissions.1.1 State of the Art1.1.1 Limitations of Emissions Inventories Over CitiesWhile there are a variety models and methods that can characterize the contributions of in-dividual emission sources or sectors of CO2 enrichment in the atmosphere, quantifying urbanemissions remains a challenge.Current urban emissions inventories are based on the down-scaling and the spatiotemporaldisaggregation of national and regional fuel consumption statistics (Parshall et al., 2010; Gurneyet al., 2009, 2012). The attribution of emissions to urban areas relies heavily on a variety ofproxies such as population density and land use and built density derived from national censusdata and geographic information systems geographic information systems (GIS) (e.g. Parshallet al., 2010).The uncertainties associated with such downscaling have thus far limited top-down ap-proaches to spatial scales of 100 km × 100 km down to 10 km × 10 km and temporal scales ofmonths to years rendering their results less effective for developing targeted emissions reduc-tions strategies at neighborhood or block-scale where most decisions take place (Kellett et al.,2013).Alternative emissions inventories are based on process-based bottom-up models which usebuilding energy models (BEMs) to upscale emissions profiles from the building to the city scale(e.g. Heiple and Sailor, 2008). These models estimate the emissions per building over aggregatedtime intervals (e.g. month to year) given a set of heuristics such as building age, building type(e.g. commercial, single-family, mixed-use, etc.), number of residents, and meteorology. Severalmodel-validated studies show that such bottom-up approaches are promising when upscaling3to the neighborhood level and when complemented with geostatistical databases such as lightdetection and ranging light detection and ranging (LIDAR) datasets (e.g. Kellett et al., 2013;Christen et al., 2011.The interdependent effects of urban environmental factors such as urban morphology, veg-etation cover and density, and urban function are difficult to integrate and can influence to agreat extent local conditions. While bottom-up models help to build a higher resolution mapof emissions, initial assumptions or model simplifications at a small scale can lead to greatuncertainties over larger areas and time scales (e.g. Marland, 2008).1.1.2 Need for Measurements in CitiesDirect measurements of emissions are necessary for validating emissions models and addressingthe uncertainties associated with model up- and down- scaling. Table 1.1 shows the methodsappropriate for monitoring CO2 at urban scales. In-situ monitoring methods from micro- toregional- scales offer solutions for helping to build informative emissions maps, but these tooface limitations related to cost and spatial and temporal resolution.4Table 1.1: Relevant scales and monitoring methods in urban settings. Adapted from(Christen, 2014)Scale(Systemsstudies)SpatialDi-men-sionAtmosphericLayerStudiedTemporalResolutionCommonMeasurementApproachesMicro-scale(buildings,roads,industry,greenspace)1-100murbancanopylayer,roughnesssublayerOne-timemeasurements atspecific locationsor alongtransects, in somecases long-termobservations (5min to years).Traverse andvertical profilemeasurements incanyons,ecophysiologicalmeasurementsusingclosed-chambers(vegetation,soils).Local-scale(neighbor-hoods,land-usezones)100 -10kminternalsublayerContinuousmeasurementsthat resovlediurnal andseasonaldynamics (30 minto years).Directeddy-covariancefluxmeasurements ontowers above thecity.Meso-scale(cities,urbanregions)10 -100kmurbanboundarylayerShort-termcampaigns orcontinuousmeasurements atselected sites thatresolve day-todayvariations andseasonaldifferences.Boundary-layerbudgets, upwind-downwind mixingratio differences,regional inversemodelling,isotopic ratios.Monitoring at the MicroscaleAt the micro-scale, emissions monitoring efforts include direct measurements of the urbancanopy layer (UCL) – the height from the surface extending to the height of the urban ele-ments – and roughness sublayer (RSL) - which extends from the surface to two to five timesthe height of the UCL – using mobile sampling traverses, vertical profile measurements, andclosed-chamber measurements. Figure 1.2 shows a diagram of the relevant urban atmosphericscales.5Figure 1.2: Illustration of the relevant atmospheric scales in an urban setting. Adaptedfrom Crawford (2014).Driving these UCL studies is the assumption that CO2 mixing ratios, given the absence ofadvection, will be higher closer to the emissions sources, allowing for attribution within the UCL.Studies by Idso et al. (2001), Henninger and Kuttler (2007), Crawford and Christen (2014), andPhillips et al. (2013) have demonstrated that mobile transect based monitoring across cities canbe used to not only capture the dynamics of the urban CO2 dome and plume, but also establisha relationship between the enhancement of CO2 and the strength of emissions. For example, thestudy by Idso et al. (2001) in Phoenix, Arizona USA helped establish the concept of the urbanCO2 dome by attributing peak CO2 mixing ratios as high as 650 ppm in the city center comparedto the 369 ppm in the rural surroundings. Higher mean mixing ratios on the order of 38-43%were measured when moving from the city center to rural areas outside of Phoenix (Idso et al.,2001). Phillips et al. (2013) notably used a mobile monitoring platform to map 3356 natural gasleaks (methane) in Boston, MA USA revealing a promising potential for such mobile systemsto resolve micro-scale emissions signatures and point sources. Recently, Crawford and Christen(2014) used mobile transect measurements to validate emissions measured at an urban eddycovariance tower and demonstrated the potential accumulation of CO2 with cold air poolingin the urban microtopography. Stationary UCL studies, such as those performed by Vogtet al. (2006), Moriwaki et al. (2006), and Lietzke and Vogt (2013) show similar results of CO2variability and spatial significance. In these studies, CO2 mixing ratios were consistently higherat the street level compared to the measured mixing ratios at or above the UCL height. Theadvantage to such studies is that it enables the simultaneous examination of exchange, effectivelyestablishing a control between emission strength, UCL concentrations, and turbulent mixing(Vogt et al., 2006). They furthermore provide the ability to examine temporal changes of CO26plume dynamics which have been found to vary between day and night - higher CO2 mixingratios are observed during the night despite lower emissions because less turbulent mixing fromsurface heating (Salmond et al., 2005). Kandias (2008) provides an extensive list of the urbanstudies of CO2 using fixed and mobile monitoring campaigns.These studies indicate a potential for fixed and mobile sensors to detect CO2 in the UCL,however they still capture a limited part of a city’s CO2 dynamics. First, the aforementionedmethods are limited spatially and by the costs of instrumentation and maintenance. Transectmeasures are project based and for climate applications have thus far been limited to mea-surement campaigns by single vehicles. Second, while mobile monitoring platforms are ableto measure mixing ratios and concentrations, they have not yet advanced to measuring fluxes(emissions and uptake). As a result, micro-scale measures such as eddy covariance and closed-chamber soil and vegetation measurements at the plot to leaf to canopy level are still necessaryfor full emissions monitoring of urban systems with varying land use and land cover typesChristen (2014).Monitoring at the Local ScaleAt the local scale (1 km to 10 km), CO2 fluxes over relatively homogenous local climate zone(LCZ) can be quantified using direct eddy covariance (EC) flux tower approaches. Urban ECstudies are an elegant approach to monitoring CO2 fluxes at the local scale and have becomeincreasingly used worldwide (Velasco and Roth, 2010). The EC approach can act as a validationfor urban emissions models and inventories and resolve the magnitude of specific GHG emissionsfrom known sources over space and time (Crawford and Christen, 2014). The urban metabolismplays a significant role in altering the measured CO2 fluxes of otherwise similar land use and landcover types. The human and biological processes affecting the measured CO2 fluxes can includethe variability and frequency of road use for transportation emissions, socio-economic differencesin energy use for HVAC systems, and regional differences in energy supply (e.g. remote electricpower generation from coal versus local natural gas combustion) and vegetation phenology andtype, respectively. EC systems do not differentiate between local emissions sources and thoseentering from outside the turbulent source area - the area that is contributing to the turbulentexchange processes as measured from the EC system; EC systems cannot resolve the emissionsbeing generated at remote locations from local scale energy use, for example. Furthermore, ECsystems cannot locate emissions sources or differentiate between sources within the turbulentsource area, but rather are better suited for quantifying the total fluxes of a given area. Themost prominent limitations of EC systems lies in the stringent requirements for siting EC towersand equipment and maintenance costs (Feigenwinter et al., 2012).Monitoring at the MesoscaleAt the meso scale, determining emissions sources from cities and urban regions are estimatedby comparing measured urban-rural CO2 mixing ratio differences, performing boundary layer7mass budgeting, inverse modeling, and analyzing stable isotope ratios (Christen, 2014). Urban-rural comparisons are a common method to determine the strength of urban emissions relativeto background concentrations (e.g. Idso et al. 2001; George et al. 2007). In such studies, aseries of emissions monitoring stations or sampling sites are established along an urban-ruralgradient. These studies give insight into the general trends of CO2 emissions intensities, butdo not provide a way to differentiate between emissions sources. Furthermore, these point-based measures do not fully describe the characteristics of the entire boundary layer which issubstantially affected by atmospheric turbulence and stability and mixing layer height (Rigbyet al., 2008). The use of stable isotopes is another way to attributing emissions to urban andrural sources, and is used to differentiate between emission sources in the UBL, mainly betweenbiogenic and anthropogenic sources. While these studies are useful for explaining fluctuationsin UBL mixing ratios, caution must be taken when interpreting the results since the isotopicsignature of carbon fuels are highly dependent on their geographical origin (Bush et al., 2007).Lastly, developments in remote sensing technology are opening up new, space-based oppor-tunities for monitoring CO2 emissions in the UBL. Current remote sensing platforms such asENVISAT or the Greenhouse Gases Observing Satellite (GOSAT) use the absorption behaviorof CO2 in the infrared region of the electromagnetic spectrum to measure the abundance ofCO2 molecules from the ground to the top of the atmosphere. The existing platforms howeveroperate at coarser spatial resolutions (280 km × 280 km) and are useful for global emissionsmonitoring, but are limited in their ability to resolve emissions at finer urban scales (Buchwitzet al., 2005; Kort et al., 2012). As anthropogenic CO2 emissions from urban fossil fuel emissionsand industrial point sources have become increasingly accepted as a major global change factor,attention has been turned toward developing spectrometers that will be able to resolve emissionsat the urban scale (2 km × 2 km) such as CarbonSAT and OCO-2 (Velasco and Roth, 2010).Remote sensing of UBL mixing ratios is exciting because it offers global coverage of emissionsat high spatial and temporal resolutions. While these satellite systems may be able to bettercapture the dynamism of urban CO2 emissions, validation of the data processing algorithms arealways necessary. Furthermore, meterological events such as clouds and rain can limit usabilityof the measurements, posing problems for example for equatorial areas (Kort et al., 2012). Inany case, these satellites will offer a powerful mechanism for measuring emissions at the urbanto regional scales and can be instrumental in influencing regional scale emissions mitigationstrategies. These space-based methods however in their early stages of data collection, and willneed methods of validation when implemented.Innovation in Urban MeasurementsThere are increasing efforts to develop integrative top-down and bottom-up methods for mon-itoring the urban climate and air pollution using pervasive computing and distributed sensornetworks (e.g. Liu et al., 2013; Mead et al., 2013; Castell et al., 2015).Top-down data mining approaches using crowdsourced smartphone data have shown the8advantage of scalability and data density. For example, Overeem et al. (2011) derived measuresof rainfall for the entire country of the Netherlands using the attenuation of a cell phonesender signal to its receiver station. In another example, Overeem et al. (2013) developeda methodology to derive fine-scale air temperature measurements using cell phone batterytemperatures to examine the urban heat island. These top down approaches will undoubtedlyimprove as smartphone penetration increases and the sensor technology improves and becomessmaller. However, it is important that despite the huge sample sizes, caution must be taken toproperly filter out faulty or misleading data points using algorithms that are adapted to localconditions. Furthermore, these top-down methods should also be validated using alternativemodels and measurements.Bottom-up approaches using distributed sensor networks have become possible in recentyears with the increasing availability of low cost climate and air pollution sensors, open sourceprogrammable microcontrollers, and improvements in networking infrastructure such as publicwi-fi networks. For example, Meier et al. (2015) used sensor data from the NetATMO weatherstation network to examine fine-scale urban heat island effects in the city of Berlin. In anotherexample, Chapman et al. (2015) developed a road sensor network to monitor road surfacetemperatures to optimally salt roads during the winter months in Birmingham, UK. Lastly,Google, Aclima, the US EPA, and UC Berkeley’s Lawrence lab together have developed amobile monitoring platform to measure pollutants and meteorology at a high spatial temporalresolution and accuracy (Aclima, 2015). A pilot monitoring campaign showed the ability oftheir mobile sensors to capture microscale variability of air pollutants (NO2, NO, and O3) inthe city of Denver, CO over the course of 1 month. These bottom-up approaches show promisingresults for mapping the urban environment at high spatial and temporal resolutions unlike everbefore. As more opportunities arise from the increasing ubiquity of sensors, considerations mustbe made to minimize problems associated with biased and poorly designed sampling and lowprecision and accuracy sensors; the long-term maintenance of these distributed sensor networkshas yet to be seen.1.2 Research Question and ObjectivesThe past studies on UCL could identify the variablity of CO2 concentrations either at or be-tween fixed sites or across a transect, yet none have actually determined the CO2 emissions orcompared the mixing ratio measurements directly to the local emissions across a city. Giventhe growing interest in distributed and mobile sensing systems and the advances in open- andmicro technologies, could there be new opportunities for the fine-scale mapping of CO2 emis-sions in cities? Furthermore, could new methods be developed that are scalable and flexibleenough to be integrated into existing infrastructure such as bikes, cars, or even autonomousflying vehicles?9Research QuestionThis research attempts to address a methodological gap in our ability to map emissions at finescales, by asking the question: Is it possible to map GHG emissions, specifically CO2, from thebottom up at a spatial resolution of neighborhoods / blocks across the city with a network ofmobile sensors?ObjectivesIn order to address the research question, five major objectives were outlined and developed:1. Sensor Development: Develop and test a compact, mobile, and multi-modal CO2 sensorfor bikes and cars.2. Measurement Campaign: Deploy the sensors in a targeted measurement campaign.3. Physical Concept: Develop a methodology to calculate emissions from measurements ofCO2 mixing ratios using knowledge about atmospheric conditions.4. Analysis and Evaluation: Compare the mixing ratio measurements and measured emis-sions to traffic and building emissions inventories.10Chapter 2Methods2.1 Conceptual Framework2.1.1 Trace Gas ConceptsCO2 is a GHG that is released in the process of respiration from plants, animals, and humans.However, in the process of fossil fuel combustion in which heat and light are produced as aresult of chemical reactions between hydrocarbons and oxygen, additional CO2 (and watervapor) which was once sequestered in the form of fossil fuels is released, or emitted. WhileCO2 is also removed from the atmosphere in the process of photosynthesis and by the world’soceans, the mass of CO2 in the atmosphere increases when the rate of emissions from fossilfuel combustion is larger than its uptake. The amount of CO2 in the air can be quantifiedas a concentration c, moles per volume air (µmol m−3). However, because concentrations aredensity dependent and therefore not conserved as a result of thermodynamic changes, measuresof pollutants, in this case CO2, can also be expressed as a molar mixing ratio r, mole of apollutant per mole air (µmol m−1, same as “ppm”).Using the ideal gas law, c can be converted to r, expressed as parts per million (ppmv) orparts per billon (ppbv), byr =cMaρa(2.1)where c is the molar concentration of CO2, Ma is the molecular mass of dry air (28.96g mol−1), and ρa is current (dry) air density (g m−3).In this thesis, CO2 mixing ratios r are used to describe the amount of CO2 in the air.2.1.2 Relating CO2 Concentrations to CO2 FluxThe underlying physical basis of this work lies in the aerodynamic resistance approach tocalculating emissions. This approach posits that the flux of CO2 for a given area and time Fc11(kg CO2 ha−1 hr−1) is equal to the difference of c at the height above the RSL (ctower) andscreen level at 2 m height (cmobile) divided by the aerodynamic resistance of CO2. This conceptis illustrated in Figure 1.1 in which the Fc is the mass of CO2 moving out of the area of thebox.To link Fc (emissions) and r at the height above the RSL, a general conservation equation,written in Einstein notation1 (Stull, 1988, p. 92, 3.4.6b), is outlined for CO2 for a grid box:∂c∂t+ uj∂c∂xj= vc∂2c∂x2j+ Sc −∂(u′jc′)∂xi(2.2)All of the terms are molar flux divergences expressed in the units µmol m−3 s−1 where Scis a volumetric source or sink strength and describes the mass of CO2 emitted per volume andtime. All other terms are described in Stull, (1988, p. 92, 3.4.6b).According to equation 2.2, the c observed for a given area and time (∂c∂t ) can be explainedby the CO2 that is advected (uj∂c∂xj), molecularly diffused (vc∂2c∂x2j), injected or removed (Sc),or turbulently mixed without advection (∂(u′jc′)∂xi). The Sc is later normalized per ground area(µmol m−2 s−1) for the purposes of this study. In order to infer the source or sink strength,Sc within each cubic meter grid box (in µmol m−3 s−1), several terms are neglected to relateemissions to actual c observed.First, the mean molecular diffusion is omitted because the first term on r.h.s, vc∂2c∂x2jis verysmall compared to all the other terms. The result is the equation in discrete form, which assumesw = 0 and is valid for an air volume of actual dimensions for example of ∆x = ∆y = 100 mand ∆z is the layer thickness vertically (in this case from ground up to ≈24 m).∆z∆c∆t+ ∆z u∆c∆x+ ∆z v∆c∆y= Sc −∆z(∆(u′c′)∆x+∆(v′c′)∆y+∆(w′c′)∆z)(2.3)Next, because w′c′ >> w′c′ and v′c′, it is possible to neglect the first two terms in theparentheses on the r.h.s, meaning that the turbulent mixing in the horizontal directions out ofthe box is negligible compared to what is moving upward. Furthermore, given that in absenceof respiration there are no sources at ground level and because the sources of CO2 from carsand buildings are all injected into the box, not at the bottom of the box, this means at z = 0,w′c′ = 0. The turbulent mixing term (∆z(∆(u′c′)∆x +∆(v′c′)∆y +∆(w′c′)∆z)) can then be shown as∆(w′c′)/∆z = w′c′z (2.4)which then inserted into the box model equation yields∆c∆t+ u∆c∆x=Sc∆z− w′c′ (2.5)1 Einstein notation is a notational convention that implies summation over a set of indexed terms in a formula,thus achieving notational brevity12The last step is to isolate the emission strength per area Sc∆z . The equation currently showsthe rate of change of CO2 over time (term 1 on the l.h.s), the advection of CO2 (term 2 on thel.h.s), the emission strength (term 1 on the r.h.s.) and the turbulent flux of CO2 (term 2 on ther.h.s.) at top of the box. In the case of a stationary situation where ∆c∆t = 0, the box can be saidto be in a steady-state; there is no accumulation or depletion of CO2 over time in the box. Thisis justified by the findings in Crawford et al. (2016) in which tower-based measurements of CO2concentrations in a well-mixed atmosphere were found to be representative of the entire UBL.This is the case during mid-day, but not so in the early morning or night in which higher CO2mixing ratios during stable night time conditions could be attributed to micro-topographicalcold air pooling (Crawford and Christen, 2014).The advection of CO2 certainly exists in each grid cell, however if it is assumed that statis-tically over the entire city, advection can be negative or positive it will thus cancel itself out.Therefore it can be seen that,Sc∆z= w′c′ (2.6)The result of w′c′ can be expressed in resistance form asw′c′ = −ctower − cmobileraC(2.7)where ctower and cmobile are the molar concentrations (in µmol m−3) of CO2 measured onthe tower and by the mobile systems in the UCL respectively. To convert mixing ratios r (inµmol mol−1, same as ppm) to µmol m−3, Equation 2.7 is used to solve for c, i.e.c =r ρaMa(2.8)For these specific conditions, we can use the data available from a eddy covariance tower inconjunction with a fleet of mobile sensors (e.g. 2.1) to map emissions in a scalable manner.The proposed methodology relies on the availability of atmospheric measurements from anurban climate tower above the height of the RSL - including measurements of sensible heat fluxH (W m−2), air temperature at 24 m height (Ta), surface temperatures (T0), and r - coupledwith screen level r from mobile sensors. From the tower measurements of air temperature (Ta)and surface temperature (T0), it is possible to calculate the aerodynamic resistance of heat raH(Kanda et al., 2007), but not CO2. As a result, the proposed methodology replaces the raCterm with raH ,raC = raH = ρcpTtower − T0H(2.9)where Ttower is the air temperature (◦C) at the height of the tower (24 m), T0 is the surfacebrightness temperature (◦C), and H is the sensible heat flux (W m−2) from ground to the top13of the tower and thereby rewrites the Equation 2.7 asw′c′ = −ctower − cmobileraH(2.10)and operates under several key assumptions:1. CO2 concentrations in a well mixed atmosphere will not change dramatically over a shorttime period (e.g. over 30 min time periods are long enough where urban fluxes are wellrepresented) given the same meteorological conditions and are therefore in an equilibrium.2. The flux is directly related to the source, see Equation 2.4.3. raC is equal to the aerodynamic resistance of raH , see Equation 2.7.4. raH and therefore raC is constant across all the urban densities/local climate zones (LCZs)in the study area/city. Despite the fact that there are varying urban densities throughouta city, the idea is that turbulence will scale with bigger buildings and the resistance willnot change significantly.Cmobile   CtowerraHFc  = w’c’  =Ctower  - raHCmobile Figure 2.1: Sensible heat flux and temperature are used to calculate the aerodynamicresistance for heat. T0 is calculated from a net radiometer, inverting Stefan Boltz-mann Law. Assuming that the aerodynamic resistance of CO2 and heat are thesame, the flux is computed.14Figure 2.1 illustrates the necessary components of the proposed methodology. In order toconvert the molar flux w′c′ (in µmol m−2 s−1) to a mass flux Fc (in kg CO2 ha−1 hr−1),Fc = Mc ba bt bo bmw′c′ (2.11)where Mc is the molar mass of CO2 (44.01 g mol−1), ba is a factor for converting m−2 to ha−1(i.e. ba = 104 m2 ha−1), bt is a factor for converting s−1 to hr−1 (i.e. bt = 3600 s hr−1), bo is thefactor for converting µmol to mol (i.e. bm = 10−6 µmol mol−1) and bm is the factor for convertingg to kg (i.e. bm = 10−3 kg g−1). Note that Mc ba bt bo bs = 0.00158436 kg ha−1 hr−1 mol−1 s m2The conversion factor (Mc ba bt bo bs), once defined, can then be simply applied later, as seenin Section 2.4.2.2.2 Sensor DesignMobile measurement methods have played an important role in mapping the spatial variabilityof GHGs and air pollutants and in particular for identifying point sources and levels of exposure.In general, the mobile monitoring methods for GHGs such as CO2 and methane (CH4) rely eitheron one-to-few high cost, high precision and accuracy, and bulky sensor systems using industrystandard, but proprietary software for system programming and data logging (e.g. Brantleyet al., 2014) or a dozen or more low cost and less accurate sensors (often electrochemicalsensors) using semi-open or closed- source software for hardware interfacing and data logging.While there are advantages and disadvantages of both approaches, the key considerations fordeveloping a mobile CO2 monitoring system must be around scalability (how many can be builtand for what cost), system extendibility (can the system be built upon), accuracy and precision,temporal resolution, accessibility (e.g open source or proprietary), and the mobile platform onwhich the sensor is to be mounted.2.2.1 RequirementsA mobile CO2 monitoring system was required to address the project’s need for multiple,low cost, high accuracy sensors capable of measuring at high frequency (>1 Hz) and easilydeployable on bikes, cars, and people. A mobile monitoring system with such specificationsis necessary to cover large geographic areas (1000 km2) within limited time scales (hours) atsufficiently fine resolutions that are representative of typical urban emission patterns undersimilar atmospheric measurement conditions. Sensor systems with many of these specificationsdo already exist, but few, if any, were designed to be carried on and easily interface withvarious types of mobile platforms; all studies using high accuracy CO2 sensors either have beenstationary or have primarily used cars because of the weight and size of the sensors being usedand are highly costly (> $5000.00 USD).This proposed system would need to be capable of mapping CO2 mixing ratios (r) or den-sities (c) and would require an integrated global positioning system (GPS) to return measures15of geoposition (latitude, longitude), altitude, time, and speed - speed data are useful for datafiltering because it can be used to identify when the vehicle is stopped and may be capturingplumes of CO2 directly from vehicle exhaust (Crawford and Christen, 2014). The mobile sys-tem would be required to support flexible hardware interfacing to add functionality throughadditional environmental sensors (e.g. air temperature sensors) and networking devices (e.g.internet modem) such that the system can be developed modularly. Finally, the system shouldbe built with as much free and open source hardware and software to promote ongoing systemdevelopment, reproducibility, accessibility for communities of need.2.2.2 MaterialsComponents from the Arduino platform (Arduino CC, Ivrea, Italy), an open source pro-grammable microcontroller, were coupled with Licor’s proprietary Li-820 infrared gas ana-lyzer (IRGA) (Licor Inc., Lincoln, NB, USA) - a compact (23.23 cm x 15.25 cm x 7.62 cm, 1kg), low maintenance (approx. 2 years of continuous use) and high accuracy CO2 (± 1 ppm)single-path IRGA built for continuous CO2 monitoring applications (Li-Cor, 2015) at 1-2 Hz- to prototype a portable CO2 analyzer. The IRGA uses infrared light to determine the CO2mixing ratio within a closed path by detecting the amount of absorption of the light from thepath. With low cost compact components, open code base, and flexible hardware interfacing,the Arduino platform provided a lightweight and modular prototyping environment capableof communicating digitally with the IRGA, a GPS (Adafruit Ultimate GPS Logger Shieldwith GPS Module, Manhattan, New York, USA) unit, and digital temperature thermometers(Maxim Integrated One Wire Digital Temperature Sensor - DS18B20, San Jose, CA, USA).A custom hardware board (developed by our team’s research engineer) was used to connectall of the hardware components together in a way that: 1. appropriately distributes the cor-rect amount of power to each of the hardware components, 2. allows for additional hardwareand sensor input, and 3. keeps the sensor hardware centralized, organized, and compact. Theportable CO2 analyzer was named the “Do-It-Yourself-Sensor-CO2”, or “DIYSCO2” system.16Figure 2.2: The “DIYSCO2” system components: IRGA, Pump, Needle Valve, Filter,Internal Temperature Sensor, Fan, Sensor Mount, GPS, Custom Hardware Board,Microcontroller. The system’s dimensions are 35.8 cm x 27.8 cm x 11.8 cm andweighs 2.6 kg.The DIYSCO2 system in total requires 12 volts (V) of power which can be supplied by bat-tery or via car cigarette lighter socket and measures r in ppm, geoposition (latitude/longitude,speed, altitude, and satellite strength), and internal and external air temperature which arelogged onto a micro-Secure Digital (SD) card at 1-second intervals located on the GPS unit.Air is drawn into the DIYSCO2 system through a 3 m long inlet tube (6.35mm diameter, DekronBendable Tubing, Mt. Pleasant, Texas, USA) using a small KNF NMP015 Micro-DiaphragmPump (KNF Neuberger, Inc., Trenton, NJ, USA) first passing through a mesh filter at the sam-ple inlet head to prevent large particles from entering the DIYSCO2 system (e.g. insects) andthen through a Balston disposable filter unit (DFU) (Parker Hannifin Corporation, Lancaster,NY, USA) at the end of the 3 m tube. The flow rate is regulated by a Swagelok needle valve at700 cc min−1 as recommended by Licor to minimize the effect of internal cell pressure changeson the CO2 measurements. The entire DIYSCO2 system is 35.8 cm x 27.8 cm x 11.8 cm, weighs2.6 kg and is contained in a weather-proof case (NANUK 910, Plasticase, Terrebonne, Ca). Thematerials for the system are summarized in Table 2.1.17Table 2.1: Sensor Materials List for Building the DIYSCO2, mobile CO2 SensorItem Model CompanyIRGA Licor Li-820 Licor Inc., Lincoln, NB,USADigital TemperatureThermometerOne Wire DigitalTemperature Sensor -DS18B20Maxim Integrated, SanJose, CA, USAGPS Adafruit Ultimate GPSLogger Shield withGPS ModuleAdafruit IndustriesManhattan, New York,USAMicrocontroller Arduino Mega Arduino CC, Ivrea,ItalyCustom HardwareBoardnot applicable UBC BiometeorologyGroupmicro-Secure Digital(SD) Memory CardLexar 8 gb High SpeedSDXC CardLexarMicro-DiaphragmPumpKNF NMP015 KNF Neuberger, Inc.,Trenton, NJ, USATubing 6.35mm diameter,Dekron BendableTubingDekron, Mt. Pleasant,Texas, USAFilter Balston disposablefilter unit (DFU)Parker HannifinCorporation,Lancaster, NY, USAWeatherproof case NANUK 910 Plasticase, Terrebonne,Ca2.2.3 Characterizing System PerformanceSeveral key system specifications were evaluated, namely: precision, accuracy and drift, thesystem measurement lag time, and the effects of inlet location on measurement variability. Theprecision of the DIYSCO2 must be evaluated to properly compare the r measured betweenDIYSCO2 systems as well as with the tower. Accuracy is assessed to determine the DIYSCO2’sability to properly resolve the variability of mixing ratios. The DIYSCO2’s drift is tested toavoid changes in precision. The DIYSCO2’s measurement lag time is important to determinethe system’s temporal resolution - its ability to resolve prominent features (e.g. peaks, emissionplumes) at a given speed of a vehicle - while also correctly attributing measurements to theirgeographic space.Sensor CalibrationBefore determining the sensor specifications, the entire DIYSCO2 system - including the 3 msample tube and KNF pump drawing air at 700 cc min−1 - was first calibrated using the sensorcalibration protocol and the Licor Li-820 software for instrument configuration. The sensor18calibration protocol outlines a procedure as follows:1. The sensor system must be running for 2 hours prior to calibration to ensure the sen-sor’s cell temperature - the temperature of the light source - is warmed to the optimaltemperature of 50◦C as outlined in the Licor Li-820 operating manual (Li-Cor, 2009).2. Second, the sensor system which has been kept on from the warm up period is connectedusing a Union Tee connector to calibration gases (Nitrogen Zero CO2 gas and 414.02CO2 ppm span gas calibrated against a tunable gas diode analyzer - TGA200, CampbellScintific Inc., Logan, UT, USA) equipped with CO2 isotope analysis capabilities with anaccuracy of typically 0.1 ppm against WMO tracable standards from NOAA Table 2.2.The Union Tee connector ensures that the effect of cell pressure changes from the flow ofthe calibration gases is minimized; the changes to cell pressure affect the measurementvalues of the Licor Li-820 IRGA. As such, the flow of the calibration gases must besignficantly high enough to guarantee that ambient air is not being drawn in through theUnion Tee connector but rather that calibration gas is either flowing out of the Union Teeconnector or into the sample tube.3. Third, the calibration starts by first flushing the system with each of calibration gases -the zero and span - for two minutes each. During each flushing of the system, the value ofCO2 ppm are noted in a calibration documentation sheet to keep track of the drift overtime. Additional notes of the sensor system components - e.g. the tube length and flowrate of the pump - are added to the document.4. Fourth, to zero the system, the system is flushed again for 2 minutes and the system iscalibrated using the Licor Li-820 CO2 Gas Analyzer Software (Li-820 Software, version2.0.1, Lincoln, Nebraska, USA). The mixing ratios of CO2 (ppm) are noted before andafter the calibration. This step is repeated for the span gas.PrecisionTo test for system precision, a 6-point calibration was performed using six tanks of knownmixing ratios (see Table 2.2) of CO2 standard tanks which have been calibrated against CDML/ NOAA standard tanks whose accuracy was measured using the TGA2000 mentioned above2. To perform this test, the sensor system was first left running for 2 hours to ensure significantwarm up time. After the warm up period, the DIYSCO2 was connected to a calibration gasusing a Union Tee connector. For each of the six gases, the calibration protocol called for aninitial two minute system flush and then a recording of the values for at least 1 minute each. Aminimum of 60 points per gas sample were used to calculate the average mixing ratios measuredby the system. The data were recorded to the sensor sytem’s data logger.2All tanks are tested using a TGA200 tunable gas diode analyzer.19Table 2.2: CDML / NOAA standard calibration gas tank concentration and accuracyused to test the precision of the mobile sensor system.Tank Name Tank MixingRatio (ppm)Accuracy (ppm)CO2-007 399.079 ± 0.047CO2-010 400.340 ± 0.042BIOMET-UBC73 412.714 ± 0.112CO2-011 456.912 ± 0.107CO2-008 457.756 ± 0.131CO2-001 503.767 ± 0.025AccuracySensor accuracy and drift were tested over the course of 7 non-rainy days, each sensor drawing inair from the same point outdoors at ≈3 m height outside room 130, H.R. MacMillan 2357 MainMall, Vancouver, BC V6T 1Z4, Canada. The 5 DIYSCO2 systems were left on continuouslyduring the 7 days.System Measurement Lag TimeWhen doing environmental measures, there is a need to match the sampling rate of a sensorto the time scale of the process being resolved. The rate of response of a sensor system toenvironmental changes will depend on a number of parameters such as the rate at which thesamples enter the sensor, the molecular composition of the sample, the geometry of the system(e.g. tubing, internal structure, material composition etc), temperature, humidity, cell pressure,and the computing time of the sensor system itself, to name a few (McDermitt et al., 1995).The system measurement lag time is the time delay from when a measurement first entersthe sample inlet of the system to when the signal is registered by the sensor. For a giventube length and flow rate, the lag time will differ and therefore affect the system responsecharacteristics. In order to determine the lag time, an experiment using lab measurements wasperformed in which a solenoid switch was used to pass nitrogen gas (”zero gas”) into the sampletube inlet while simultaneously logging the exact second in which the solenoid was triggered.The data was “flagged” as long as the zero gas was being drawn into the DIYSCO2 and used tosubset only the relevant data in post processing. To calculate the lag time value for the system,the number of seconds were counted from when the sample enters the sample tube until 50%of the peak. An illustration of the experiment setup is shown in Figure 2.3.20Nitrogen GasSolenoid SwitchOutletCO2 Sensor SystemFigure 2.3: A solenoid switch is connected on one end to a tank of pure N2 (“zero gas”)and on the other to the inlet tube of the CO2 system. The solenoid switch is trig-gered when plugged into a power source allowing the nitrogen gas to pass throughto the CO2 system inlet and simultaneously sending an input voltage to the data-logger, logging the exact second at which the solenoid was triggered.Sample Inlet LocationAn experiment was performed to examine the effect of sample inlet location on the variability ofthe mobile r measurements. As part of the experiment, two specific tests were performed. First,a test was done with all of the inlet tubes bundled together measuring at the same location ofthe vehicle (now referred to as “Grouped Inlet Test”). A second test was done with each of theinlet tubes located at different locations of the vehicle (now referred to as “Ungrouped InletTest”). Images of the two sampling methodologies are shown in 2.4a and 2.4b.21(a) Grouped inlet test: All 5 inlets are grouped together at the right-front passengerside window.Inlet 1Inlet 2 Inlet 3 Inlet 4Inlet 5(b) Ungrouped inlet test: Each of the 5 sensor inlets are located at different loca-tions of the vehicle.22Both test were performed in the City of Vancouver. Figure 2.5 shows the transect of theGrouped Inlet Test which was performed between 15:00 PST and 17:00 PST on 2015-05-16(weekend) and the Ungrouped Inlet Test which was performed between 11:00 PST and 13:00PST on 2015-05-23 (weekend).0 1 2 3kilometersBURRARD INLETVANCOUVERStartEndFigure 2.5: The inlet tests were performed along the transect shown above.Both transects cover a range of local climate zones including - Scattered Trees (LCZ B),compact low-rise (LCZ 3), compact mid-rise (LCZ 2),open high-rise (LCZ 4), open mid-rise(LCZ 5) (Stewart and Oke, 2012) - with traffic volumes ranging from 300 to 850 vehicles perhour during the study periods; the traffic volume on average slightly lower during the ungroupedinlet test (182 vehicles per hour) than the grouped inlet test (158 vehicles per hour). The goalwas to examine the differences in CO2 mixing ratios observed by the sensors based on theirrespective location on the same moving mobile platform.2.3 Applying the Mobile Sensor System to Map EmissionsFive DIYSCO2 systems were mounted on cars and a bike and driven through the City ofVancouver in a planned experiment to assess the potential for the mobile monitoring andmapping of CO2 emissions in cities.Section 2.3.1 describes the study area, Section 2.3.3 describes the method for averaging23data points to a grid, Section 2.3.4 describes the transects measured in the experiment, Section2.3.6 describes the procedures of the sensor deployment, and Section 2.3.7 describes the waysin which the data from the experiment were processed.2.3.1 Study AreaThe study area is 12.7 km x 1 km quadrangle of varying landuses and local climate zoneswithin the City of Vancouver. The study area begins at the northern-most tip of the city(UTM 10,488510 E, 5451513 N) and extends close to the city’s south eastern border (UTM 10,495410E, 5462213N) and includes most of Vancouver’s major urban land uses - the downtowncore, medium density residential, single detached residential, industrial development, and forest- which are comprised of the following local climate zones (LCZ) (See Figure 2.7) - compacthighrise (LCZ1), compact lowrise (LCZ3), compact midrise (LCZ2), open lowrise (LCZ6), andforest (LCZB) (Stewart, 2011; Stewart and Oke, 2012). The study area is 1270 hectares (12.7km2) in size, encompassing approximately 11.1% of the total area of the city, and was selectedin large part, because of the provision of high resolution geospatial data on building emissionsand traffic counts for the study area, the availability of an urban climate station (Vancouver-Sunset tower) within the study area bounds, and based on the representativeness of the urbanland uses for the city (van der Laan, 2011).0 1 2 3kilometersBURRARD INLETVANCOUVERNORTH VANCOUVERBURNABYYVRRICHMONDl SUNSET CLIMATE TOWERDowntownWest EndFairview, Mt. PleasantKensington−Cedar CottageSunset, FraserviewStanley ParkFigure 2.6: 12 km2 transect study area in Vancouver, BC. The transect is 1 km × 12.7km covering the major land cover types in the city. Sunset Urban Climate toweris shown in orange.24Figure 2.7: The major local climate zone types within the study area by neighborhoodsubset include compact highrise (LCZ1), compact lowrise (LCZ3), compact midrise(LCZ2), open lowrise (LCZ6), and forest (LCZB) (Stewart, 2011; Stewart and Oke,2012)2.3.2 Vancouver-Sunset TowerThe meterological data including - air temperature (Ttower) , sensible heat flux (H), latentheat flux, wind direction, wind velocity, turbulent kinetic energy, carbon dioxide molar mixingratio r, and barometric pressure - were measured at a 24 m high eddy covariance tower, called“Vancovuer-Sunset” (Fluxnet, 2016), at the south east corner of the study area (UTM 10,494273 E, 5452641 E). The availability of Vancouver-Sunset made it possible to calculate theraH and therefore to calculate the measured emissions.252.3.3 Grid AveragingVector matrix grids of 50 m × 50 m, 100 m × 100 m, 200 m × 200 m, and 400 m × 400 m weremapped onto the study area in a GIS to spatially aggregate and attribute the r measured bythe DIYSCO2 systems. The data analysis on the 50 m, 100 m, 200 m, and 400 m grids provideda way to determine the effects of grid size on emissions estimates. While the 50 m, 200 m, and400 m grids were also used in the analysis, the 100 m grid was selected as the focus of analysisbecause the 100 m grid cell size was determined to be significantly large enough to captureany horizontal advection of emissions while also still attributing emissions at the “micro-scale”(see: Figure 1.2). Furthermore, the 100 m grid was used in Section 2.3.4 to map out predefineddriving routes through the transect for the field campaign.BURRARD INLETVANCOUVERNORTH VANCOUVERBURNABYYVRRICHMOND0 1 2 3kilometersFigure 2.8: The study area grid cells that can be traversed by car or bike based on a 100m × 100 m vector grid.2.3.4 Deployment TransectsIn order to ensure that the study area was comprehensively sampled during the duration ofthe measurement campaign, predefined transects (see Figure 2.9) were mapped for each of theDIYSCO2 systems (total: 5). Taken together, the routes were drawn such that the DIYSCO2would not only sample some of the same street segments at different times throughout the26campaign, but also planned in a way that a majority of the roads in the study area would besampled at least once in the 3.5 hour time period.Route − All Route − 108Route − 150 Route − 151Route − 1641 Route − 205Figure 2.9: The image shows 5 planned transect routes for the measurement campaignoverlaid onto the study area.The predefined routes were evaluated using the 100 m × 100 m study area grid, confirmingthat nearly all of the grid cells would be traversed if the routes were successfully completed.Finally, the measurement period was set between 9:30 to 13:30 PST because this off-peak timeperiod was identified to show relatively consistent traffic counts throughout the transect (seeFigure 2.10) as well as support the traveling efficiency of the sensors.27llll lllllllllllllllll lll5 10 15 200.020.040.060.080.10Hours of the DayTraffic Volume PercentagesFigure 2.10: The graph shows the traffic volume represented as a percentage across thetransect. The red dotted lines indicate the start and end time of the measurementcampaign (in PDT) and show that the traffic patterns remain relatively unchangedduring those times (City of Vancouver, 2015).Each vehicle was assigned to travel approximately 70 km during the study period (achiev-ing an optimal sampling density of about 3.5 km2 hr−1), traversing nearly all navigable roads(excluding some laneways). Furthermore, a bike was used to traverse the trails in the forestedarea of Stanley Park to sample in the densely forested ecosystem away from traffic, buildings,and humans.2.3.5 Sensor InstallationVehicle InstallationThe DIYSCO2 system is installed in a vehicle (see Figure 2.11) by laying the DIYSCO2 systemon the passenger side seat (or in the backseat) and connecting the sensor’s 12 V power adapterinto the car cigarette lighter port. The sample inlet tube is then run out through the windowof the vehicle to the height of 2 m as shown in Figure 2.12.28Figure 2.11: The image shows the DIYSCO2 installed in a vehicle. The inlet tubesare run out of the window. The system is connected to the vehicle power sourcethrough the 12 v provided through the cigarette lighter socket. It can also be usefulto attach an extra battery to the system so that the sensor can run continuouslywithout the vehicle power.Figure 2.12: The image shows the sample inlet tube run out through the window of thevehicle. Enough pressure is applied from the window to secure it in place, butnot enough to crush the dekron tubing or damage the window.29Bike InstallationIn order to deploy the DIYSCO2 on a bike, the setup requires at least a 40 ` backpack to carrythe sensor and 12 V gel cell battery with molex fittings and a 1.5 m long rigid mounting tube(6 mm diameter) to mount the inlet tube. The sensor is placed in the backpack with the 12 Vgel cell battery and worn on the back of the cyclist to reduce vibrations to the sensor system.The 1.5 m rigid mounting tube is attached to the back rack of the bike with 2 stainless steelpipe clamps. The sample inlet tube is run from the backpack along the rigid mounting tubeand fastened with electrical tape such that the inlet is at screen level height. Figure 2.13 showsthe bike-based DIYSCO2 system setup.Figure 2.13: The image shows the DIYSCO2 installed on a bike. The DIYSCO2 is placedin the backpack with inlet tube raised to 2 m.2.3.6 Sensor DeploymentThe measurement campaign started and ended at David Thomson Secondary School (UTM10, 494860 E, 5452010 N) located at the southeast corner of the transect. Prior to the sensordeployment, the DIYSCO2 systems were installed and turned on for a 15 minute warm upperiod. The sensors were then run together side-by-side in their respective vehicles for 5 minutesin order to log the drift before the field campaign; this is now called a “in-situ accuracy test”.The vehicles were parked away from roads in the school parking lot, each door within 1 m30from each other as shown in Figure 2.14. The prevailing winds were blowing across the park(Gordon Park) to ensure mixing. All people moved away and downwind of the vehicles to avoidcontamination from human exhaust. The data collected in the in-situ accuracy test is typicallyused to adjust the drift of the sensors in post-processing3. During this 5 mimute period, thevehicles’ engines were turned off to avoid vehicle exhaust contamination of the measurements.After the 5 minutes, the volunteers began driving along their predefined routes through thecity (see Figure 2.9); volunteers were instructed to follow the paths as closely as possible, butwere given the freedom to reconfigure the route if necessary given the possibility of unforseentraffic conditions or obstructions. After the 3.5 hour deployment, the vehicles returned to DavidThomson Secondary School where the DIYSCO2 systems were run next to each other in theirrespective vehicles for another 5 minutes for a final in-situ accuracy test.Figure 2.14: Photo showing vehicles lined up at the end of the measurement campaignfor a 5 minute calibration period.2.3.7 Data ProcessingThe data from the measurement campaign were processed and analyzed using the Python pro-gramming language (including the Pandas, GeoPandas, Numpy, Scipy, and Shapely libraries),the R Programming Language (including the RGDAL and GISTools libraries), and QuantumGIS (QGIS).3Prior to the field campaign, the sensors were lab calibrated and therefore these collected drift data were notused.31Table 2.3: Data filtered in post-processing according to the following criteria. Based on(Crawford and Christen, 2014)Data Attribute Filter if WhyGPS Speed <5 km h−1 To remove the sampling bias fromidlingr r <380 or r>1000To measure within thespecifications of the LI-820 andrealistic rIRGA CellTemperaturet <45 ◦C The ideal IRGA operatingtemperature is approx. 50 ◦CIRGA CellPressurep <96 kPa Typically atmospheric pressurefluctuates, however a cell pressurebelow 96 kPa might indicate sensorsystem errors.The data were first filtered from the dataset following the methods in Crawford and Christen(2014) (see: Table 2.3). Data were omitted if the GPS speeds were below 5 km h−1, the r weregreater than 1000 ppm or less than 380 ppm, and if the IRGA cell temperature and pressurewere below 45 ◦C and 96 kPa in order to remove the sampling bias from idling, to measurewithin the specifications of the Li-820 and realistic CO2 mixing ratios, and to ensure consistencybetween the IRGAs operating conditions, respectively.The data were then spatially aggregated to 50 m, 100 m, 200 m, and 400 m vector grids;for each cell, the summary statistics were computed for all the data points intersecting it. Thesummary statistics included the mean, median, maximum, minimum, range, skewness, andvariance. The data were also classified by neighborhood to enable interneighborhood compar-isons of r, measured CO2 emissions, and the emissions inventories. Only the grid cells withintersecting measurements were retained for the analysis. Last, all of the grid cells that didnot fall “completely within” the boundaries of the study area were withheld from the analysis.This was to ensure that only grid cells with complete emissions inventory data (rather thanthose that were only partially covered) were compared to the grid cells where measurementswere made. Therefore the result was that the total area covered by the 400 m grid was smallerthan the area covered by the 50 m grid because many of the larger 400 m grid cells did not fallcompeletely within the study area boundaries.2.4 Emissions Inventory ValidationA building and traffic emissions inventory were compared against the measured CO2 data andconsequently the measured CO2 emissions from the case study. The following sections describethe process of deriving the emissions inventories and the methods of analysis.322.4.1 Emissions Inventory DevelopmentTraffic Emissions InventoryUsing hourly averaged directional traffic count data from 2008 - 2013 collected by the Cityof Vancouver (City of Vancouver, 2015), a traffic emissions estimate for the study area wasgenerated to examine the relationship between the sampled CO2 and the number of vehiclestraveling during the sampling period and to generate a traffic emissions inventory map of thestudy area. The traffic counts did not distinguish between different vehicle classes. Furthermore,the traffic counts were aggregated to the street level, meaning that, for this analysis, the trafficcounts did not take into account the direction of travel.First, the hourly traffic counts were spatially attributed to the OpenStreetMap (OSM) roadnetwork. The City of Vancouver provides traffic counts collected from pneumatic road tubeswhich are output to a “.csv” file with the approximate address of where the traffic counterswere located and the traffic counts. The City also provides a geospatial representation of thelocations of the traffic counters with the address, but without the count data attached. Theaddresses from the “.csv” file and the geospatial dataset were used to merge the counts with thespatial data. However, because spatial traffic counts do not align with the OSM road network,the centroids of the spatial traffic count data were computed and then “snapped” to the OSMroad network to perform the spatial join. Before joining the traffic count data by the matchinglocations of the two datasets, the OSM road network was split into segments using the 50 m ×50 m vector grid. A small (0.5 m) buffer was applied to the traffic count centroids to ensure thatthey spatially match onto the OSM road network and then were merged to the OSM dataset.A simple algorithm was used to match the street names in the traffic count dataset to thosein the OSM street network. Manual mapping of traffic counts was necessary to attribute trafficcounts to streets that were not sampled in the traffic counts. A simple rule of proximity andlocal understanding of the traffic patterns for each of the streets was used to manually map thetraffic counts to the unsampled streets. Using the OSM street classifications, traffic counts forpaths unnavigable by vehicles were given a value of “0” traffic counts, namely “steps”, “trail”,“footpath”, and “service”. Lastly, the traffic counts for the forked roads in the dataset whichwould have doubled the count for a particular street were divided in half.With a complete model of the traffic counts for the transect, it was then possible to generatea gridded traffic emissions inventory map of CO2 (now referred to as “traffic emissions inven-tory”). The length of each of the street segments which had been split in the earlier steps werecalculated and then summed up per 50 m, 100 m, 200 m, and 400 m grid cell. Next, the lengthof navigable roads per grid cell were multiplied by the hourly traffic counts along each road,resulting in an estimate of total distance of vehicle traveled per grid cell. Each grid cell’s hourlytravel distance was then multiplied by the NRCAN fleet standard fuel comsumption (NaturalResources Canada, 2014) for urban driving (12.9 ` 100 km−1 ) and after by a CO2 emissionsfactor (2.175 kg `−1 fuel burned) (B.C. Ministry of Environment, 2014) to generate the traffic33emissions estimate map of CO2. In this study, the traffic count data provided by the City ofVancouver is averaged across all of the years that the traffic count data have been collected.The data are then scaled by 1.0216 times to reflect the relative traffic volumes for May basedautomatic highway counts at 5 locations throughout Vancouver.OpenStreetMapSpatial Join traffic countsto OSM road networksplit by grid lines merge .csv to .shp get centroid of traffic countersnap points to linesnormalize street names process: traffic  emissions inventoryTraffic Counts street namepattern matchingmanual mapping of missing streettraffic countscalculate distance travelled and fuel burned per grid cellmultiply fuel burned by emissions factorper grid cellresult: traffic emissions inventoryFigure 2.15: The diagram illustrates the process for generating the traffic emissions in-ventory.Building Emissions InventoryThe building emissions inventory was developed in previous research by van der Laan (2011)and takes into account building morphology, census statistics, and building typology to estimatethe CO2 emissions attributed to building energy use; the result is a 1 m resolution buildingemissions grid estimated in carbon dioxide equivalent (CO2e), reported in kg CO2e m−2 yr−1).CO2e is a measure of the GHG warming potential of most or all of the GHGs associated with a34process in the equivalent concentration of CO2 over comparable time scales; this factors in gasspecific residency times, etc. In this research, it is assumed that CO2e and CO2 are the same.Because the inventory is an annual estimate (reported in kg CO2e m−2 yr−1), a scaling factorbased on the monthly city emissions inventory was used to account for the late spring/earlysummer building emissions fraction. In the month of May, the building emissions for a sampleof the city of Vancouver was estimated to be 63.63% of the annual average of monthly buildingemissions (Christen et al., 2011) for the sampled area. The sum of the CO2 emissions per gridcell were scaled by 63.63% then divided by the number of days in a year (365) and the numberof hours in a day (24) to get building CO2 emissions per hour. The final building emissionsinventory were reported in kg CO2 ha−1 hr−1. In this case, it is assumed that the buildingemissions are constant over the course of the day.LiDAR process: building emissions inventory, adapted from Van der Laan (2011) Building TypologyCensusEnergy UseUrban ContextBuildingMorphologyPopulation DensityResult: building emissions inventoryFigure 2.16: The diagram illustrates the process for generating the building emissionsinventory.Total CO2 Emissions InventoryThe total emissions inventory is the sum of the building emissions inventory and the trafficemissions inventory in kg CO2 ha−1 hr−1. The total emissions inventory includes only partsof the study area where data exists for both the building emissions inventory and the trafficemissions inventory. This overlapping area between the building and trafffic emissions is nowreferred to as the “study area subset”.352.4.2 Measured CO2 EmissionsCO2 emissions per grid cell were calculated using the aerodynamic resistance approach outlinedin Section 2.1.2.First, averages of ρ, Ttower, T0, and H, all measured on top of the Vancouver-Sunset tower(see: Figure 2.8) were computed over the 3.5 hours of the field campaign described in Section 2.3to determine the raH:raC = raH = ρcpTtower − T0H(2.12)The flux (or emission strength) Fc was then calculated based on Equation 2.10 as followsFc = −Mc ba bt bo bs ctower − cmobileraC(2.13)The result is a calculation of CO2 emissions reported in kg CO2 ha−1 hr−1 at 50 m, 100 m,200 m, and 400 m grid cells (now referred to as “measured emissions”) across the study areawhere measurements were made by the mobile sensors.2.4.3 Emissions AnalysisThe building and traffic emissions estimates described in Section 2.4.1 are compared to thegridded r and the measured CO2 emissions (see: Section 2.4.2) from the measurement campaigndescribed in Section 2.3. First, the relationship between the gridded r and the emissionsinventories is established using a linear model. Second, the measured emissions are comparedto the emissions inventory and evaluated on how closely aligned (or distant) the datasets areto one another. This is done by examining the absolute and relative differences between thedatasets using the summary statistics. Last, a correlation analysis is applied to the measuredemissions and the traffic, building, and total emissions estimates.36Chapter 3ResultsThe following chapter describes the results from the sensor development and testing (Sec-tion 3.1) the measurement campaign (Section 3.2), and a comparison between measured andmodelled emissions (Section 3.3).3.1 Sensor SpecificationsPrecisionThe test for the system precision using the data shown in Table 3.1 showed strong linearity (R2of 0.9999) and a root mean square error (RMSE) of 0.233 ppm for the six different CO2 mixingratios. This indicates that the IRGA is operating well within its factory specifications of 1 ppmwhen used in conjunction with the other DIYSCO2 components.Table 3.1: Calibration gas tank mixing ratio and accuracy and average observed valuesmeasured by the DIYSCO2 system for each mixing ratio of gas.TankConcentration(ppm)Accuracy (ppm) Li-820 Observed(ppm)399.079 ± 0.047 400.3819400.340 ± 0.042 401.8431412.714 ± 0.112 413.9892456.912 ± 0.107 458.3792457.756 ± 0.131 458.6729503.767 ± 0.025 504.124637l lllll400 420 440 460 480 500400440480Reference Gas (ppm)Measured Mixing Ratios (ppm)R2=0.9999All lll l400 420 440 460 480 500−2−1012Reference Gas (ppm)Measured Mixing Ratios − Reference Gas (ppm)BFigure 3.1: (A): A linear fit of the data points with the reference gas on x-axis, observedvalues on y-axis. A linear model fit through the points showed an R2 of 0.999. (B):The figure shows the reference gas on x-axis and the absolute difference betweenthe predicted and the observed values on the y-axis.Accuracy and DriftThe accuracy between the DIYSCO2 systems determined to be ± 3 ppm based on the RMSEfrom the mean of the 5 sensors and the drift over the course of a 7 day experiment. The RMSEranged between 0.2 and 3 ppm for the seven day period and is therefore time dependent. Theabsolute error ranged between -2 and 4 ppm from the average of the 5 measurements over theseven days (total drift of 6 ppm). Given that the field campaign was planned to be 3 to 3.5hours long, the drift after 3 hours was calculated and ranged between -0.31 ppm and 0.51 ppmof the mean of the 5 sensors. Table 3.2 reports the RMSE for each DIYSCO2 system from themean of the 5 DIYSCO2 systems over the 7 day experiment.38Table 3.2: RMSE for 1 min averages from the mean over a 7 day experiment for each ofthe DIYSCO2 systems.Sensor ID RMSE (ppm)LI820 0108 0.486LI820 0150 3.005LI820 0151 2.549LI820 0205 1.119LI820 1641 0.255lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll0 50 100 150380390400410420430440450hours since calibrationCO2 Mixing Ratios (ppm)llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllSensor ID (starting offset, ppm)108 ( 0.29 ) 150 ( 0.37 ) 151 ( −1.77 ) 1641 ( 0.56 ) 205 ( 0.54 ) Sensor MeanFigure 3.2: Hourly averages of the CO2 mixing ratios measured over a 7-day period. Eachline corresponds to one of the five DIYSCO2 systems. The value in brackets is theoffset in ppm needed to align the initial measurements to the same value.39lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll0 50 100 150−3−2−101234hours since calibrationCO2 Mixing Ratios Drift from Mean (ppm)llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllSensor ID (starting offset, ppm)108 ( 0.29 ) 150 ( 0.37 ) 151 ( −1.77 ) 1641 ( 0.56 ) 205 ( 0.54 ) Sensor Mean0.51−0.312.28−1.732.87−1.733.36−2.14Figure 3.3: Testing for sensor drift of the five DIYSCO2 systems over a 7-day measure-ment period. Each line corresponds to one of the five DIYSCO2 system drifts fromthe mean. The value in brackets is the offset in ppm needed to align the initialmeasurements to the same value.Measurement Delay TimeUsing the procedure in Section 2.2.3, the measurement delay time of the DIYSCO2 systemwas determined to be 18.2 s. The result of the lab experiment plotted in Figure 3.4 showsthat of the 18.2 s, it takes 16 seconds for the sample to travel from the inlet to the IRGAand approximately 2.2 seconds for the IRGA to register 50% of the step change. Therefore,the time elapsed between the sample minima and maxima was 4.4 seconds. We use a valueof 18.2 seconds1 to correct our measurements during post-processing from data gathered fromfield campaigns in order to shift the GPS and observed r time series to properly attribute themeasurements spatially and temporally.1the data can only be shifted by an integer in the data table, therefore 18 s is used in the data processing.40l l l l l l l l l l l l l l l l l lllllll l l l l l0 5 10 15 20 25 300100200300400Time (seconds)CO2 Mixing Ratios (ppm)Start EndFigure 3.4: Results from the measurement delay time test. The “start” point indicateswhen the calibration gas was injected at the sample inlet. The “end” point indicates50% of the the step change and is colored in red. The measurement delay time isthe number of seconds from “start” to “end”.Inlet locationThe results of the sample inlet location experiment revealed two primary insights. First, inareas with a well-mixed atmosphere and on roads with little traffic, the DIYSCO2 system forthe grouped inlet test (after offsetting the values to the mean) showed an accuracy within -0.5and 0.5 ppm of the mean as shown in Figure 3.5 and Figure 3.6. For the ungrouped inlet testunder those same conditions, the accuracy deteriorated to -5 and 5 ppm of the mean as shownin Figure 3.7 and Figure 3.8.4122:36 22:48 23:00 23:12 23:24 23:36−15−10−5051015time − minutes (GMT)Measured CO2 Mixing Ratio − average of all 5 sensors (ppm)Sensor ID (offset, ppm)108 ( 0.82 ) 150 ( 3.36 ) 151 ( −3.9 ) 1641 ( 0.46 ) 205 ( −0.75 )Figure 3.5: Grouped Inlet Test: The graph shows the CO2 variation from the mean acrossthe entire domain (see: Figure 2.5). The data here is presented in 1 min averages.Each line corresponds to one of the five sensors. The value in brackets is the offsetin ppm needed to align the initial measurements to the same value.32:00 34:00 36:00 38:00 40:00 42:00−0.6−0.4−0.20.00.20.40.6time − minutes (GMT)Measured CO2 Mixing Ratio − average of all 5 sensors (ppm) − subsetSensor ID (offset, ppm)108 ( 0.82 ) 150 ( 3.36 ) 151 ( −3.9 ) 1641 ( 0.46 ) 205 ( −0.75 )Figure 3.6: Grouped Inlet Test: The graph shows a sample of the CO2 variation in awell-mixed atmosphere and on roads with little traffic. The data here is presentedin 1 min averages. Each line corresponds to one of the five sensors. The value inbrackets is the offset in ppm needed to align the initial measurements to the samevalue.4218:48 19:00 19:12 19:24 19:36 19:48−100−50050time − minutes (GMT)Measured CO2 Mixing Ratio − average of all 5 sensors (ppm)Sensor ID (offset, ppm)108 − RD ( −0.65 ) 150 − FD ( −2.28 ) 151 − TC ( 5.58 ) 1641 − FP ( −2.87 ) 205 − RP ( 0.22 )Figure 3.7: Unrouped Inlet Test: The graph shows the CO2 variation from the meanacross the entire domain (see: Figure 2.5). The data here is presented in 1 minaverages. Each line corresponds to one of the five DIYSCO2 systems. The value inbrackets is the offset in ppm needed to align the initial measurements to the samevalue.56:00 58:00 00:00 02:00 04:00−10−505time − minutes (GMT)Measured CO2 Mixing Ratio − average of all 5 sensors (ppm) − subsetSensor ID (offset, ppm)108 − RD ( −0.65 ) 150 − FD ( −2.28 ) 151 − TC ( 5.58 ) 1641 − FP ( −2.87 ) 205 − RP ( 0.22 )Figure 3.8: Ungrouped Inlet Test: The graph shows a sample of the CO2 variation in awell-mixed atmosphere and on roads with little traffic. The data here is presentedin 1 min averages. Each line corresponds to one of the five DIYSCO2 systems. Thevalue in brackets is the offset in ppm needed to align the initial measurements tothe same value.Second, with observations of higher CO2 mixing ratios, the error (expressed as the standard43deviation between all five of the DIYSCO2 locations) of the observed values for each sensorincreases for the 1 s data. This is the case for both the grouped and ungrouped inlet tests asshown in Figure 3.9 and Figure 3.10. The data show that when the inlets are grouped together,48.9%, 81.16%, and 90.14% of the data have an error within 5, 15, and 25 ppm, respectively(see Table 3.3). While this indicates that more than half of the data measured by the sensorsare within 5 ppm of eachother, the test also shows that we can expect a majority of the data(>88.85%) to have errors up to 15 ppm depending on where on the car the inlet is mounted.When examining the error of the observed values for the 1 min data, we observed that 86.3%and 98.63% of the data have an error within 5 and 25 ppm.llll ll ll l lllllllllllllllll lllllllllllllllllllll llllllllllllllllllllllll llll llll ll l l llll l lllllllll lllllll ll l l l l lllllll ll llllllllllllll lllllllll lllllll lll l lllll l l l lll llllllll ll l lllll l llll ll lll llllll llllllllll llllllllll llllll l llllllllllllllllllllllll llllll llllllllllllllll lllllll ll ll lllll lll llllll llll ll lllll l lllllllll llll lllllll llllllllllllllllllllll l llllllllllllllllllllll lllllll lllllllllllllll lllll lllllllllllllllll llllllll llll lll llllllllll l ll lllllllllllllll lllllll llll llllll llllllllllllll l llll l lllllllllllllllllllll l lllllllllllllllllllllllll lll l llllllllllllllllllllll lllllllllllllllllllllllllllll l llllllllll lllll llllllllll l lllllllllllllllllllllllllll llll lllllll ll l lllllllllllllllllllllllllllll lllllll l lllll llllllllllllllllll lllll llllllllll lllllllllllllllllllllllllllll llllllllll l llllllllllll llllllll llll ll lllllllllllll llllll lllllll lllllllllllll l ll l lll llllll lllllll llll llllllll llll lllllllllll lllll ll lllllllll lllllllllllllll llllllll lllllllllllllllllllllllllllllllllllll l l llll llllllllllll llllllllllllllllllllllllllllll lllllllllllllllll llll lll lllll lllllllllllllll llllll lllllll llllll ll lllllllllllllllllllllllll lll l lllllllllll lllllllllllllllllllllllllllllllllllllllllllllll llll llllllllllll llllll llllllllllllllll llllllllllllllllllllllll llll llllllllllll lllllllllllll400 500 600 700 800 900050100150CO2 Mixing Ratios (ungrouped) (ppm)CO2 SD of 5 sensors (ungrouped) (ppm)lll llllllllllll lllllllllllllll llllllllllllllllllNA to 431 468 to 504 541 to 577 614 to 651 687 to 724 760 to 797 833 to 870 907 to 944050100200CO2 Mixing Ratios (ungrouped) (ppm)CO2 SD of 5 sensors (ungrouped) (ppm)Figure 3.9: The graph (above) shows the measured errors from the mean CO2 mixingratios of the 5 sensors at 1 second resolution. The graph (below) shows a box plotof the measured errors binned in equal intervals.44Table 3.3: The table shows the percentage of the data in the grouped inlet test that isaccounted for with the respective errors.CO2 Mixing RatioError (ppm)% of data at 1 s % of data at 1 minr<5 48.90 86.30r<10 70.35 98.63r<15 81.16 98.63r<20 86.85 98.63r<25 90.14 98.63The 1 s data show that when the inlets are ungrouped, 54.98%, 79.08%, and 87.49% of thedata have an error within 5, 15, and 25 ppm, respectively (see Table 3.4). Figure 3.10 showsthat the errors are greatest when r is high (>600 ppm). The 1 min data show that for theungrouped inlet test, 66.67%, 91.66%, and 94.44% of the data are within 5, 15, and 25 ppm,repectively.lllllll lllll lllllll llllllllll lllllll lll lll l lllllllll lllll lll llllllllllllllllllllllll llllllllllllllll lllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllll l lllllllllllllllllllllll ll l l lll l llll ll l l l l llll l lllllllllllll l ll l ll lllllllllll lllllll ll l l lll llllllll llllllllllllllll llllllllllllllll l ll llllllllllllllll llllllll llllllllllllllll ll lllllllll lll ll ll l lllllllllllllllllllllll llll ll l llll llllllll llllllllllll lllllllll lllllllll lll l lllll llllllllllllllllllllll ll llllllllll ll l lll l llll l llll l ll llllll lll llllllll l lllll l lllllllllllllllllllllllllll llll lllllllllllllllllllllllll lllllllll l l l llllllllllllll lll llllllll lllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllll lll l lllllllllllllllll lllllllllll llll llllll ll lllll l l llllll l ll lll lll llll lllllllllllll l lllllllll llll llll lllllllll llllllllll llllllllllllllll lllllllllllllllllllllllll llllllllll lllllllllllllll lll llll lllllllllllll llllllll ll lllll llllll ll lll l lllll lllllllllllllllllllllllllllllllllllllllllllll llllllll lllllllllllllllllll ll l lllllllllllll lllllllll lll lll llll lllllllll lll llllllll l l llllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllll lllllllllllllllllll l lllll l ll llllllllllll l lll llll llllllllllllllllllllllllllllllllllllllllll lllllllllllllllllllllll llllllllllllllllllllllllllll lllllllllll llll400 450 500 550 600 650 700050100150CO2 Mixing Ratios (ungrouped) (ppm)CO2 SD of 5 sensors (ungrouped) (ppm)lll lllllllllllllll lllllllllll llllll llNA to 416 437 to 458 479 to 500 521 to 542 563 to 584 604 to 625 646 to 667 688 to 7090100200CO2 Mixing Ratios (ungrouped) (ppm)CO2 SD of 5 sensors (ungrouped) (ppm)Figure 3.10: The graph (above) shows the measured errors from the mean CO2 mixingratios at 1 second resolution. The graph (below) shows a box plot of the measurederrors.45Table 3.4: The table shows the percentage of the data in the ungrouped inlet test that isaccounted for with the respective errors.CO2 Mixing RatioError (ppm)% of data at 1 s % of data at 1 min<5 54.98 66.67<10 71.43 81.94<15 79.08 91.66<20 84.04 93.05<25 87.49 94.44Last, the results from the ungrouped inlet test shown in Figure 3.10 illustrate the effect ofsample inlet location on measured values. Sensors 108 and 150 were located above the rearand front driver’s side windows, respectively, while sensors 205 and 1641 were located abovethe rear and front passenger’s side windows, respectively. Sensor 151, was located in the rearcenter of the vehicle. Depending on the side on which the sample is located, the sensors appearto report differences in measured concentrations of up to ± 50 ppm from the mean wherebythe sensors located on the same sides of the vehicle tend behave in direct contrast to those onthe other side of the vehicle at the same timestamp.Summary TableA summary of the DIYSCO2 specifications are shown in Table 3.5.Table 3.5: Summary table of the DIYSCO2 system evaluationTest ResultPrecision 0.233 ppmAccuracy 3.2 ppmDrift (3 hours) (ppm) 0.82 ppmDrift (7 days) (ppm) 6 ppmMeasurement Delay Time 18.2 sInlet location Grouped inlet test showed lowererrors than the ungrouped inlettest for both 1 s and 1 min data;Inlet location must be considered.3.2 Results from the Mobile Sensor DeploymentThe following section describes the results from the DIYSCO2 deployment and includes a de-scription of the meteorology and context in Section 3.2.1, the distribution of the CO2 mixingratios in Section 3.2.2 and the gridded data in Section 3.2.3, and the measured emissions inSection 3.2.4.463.2.1 MeteorologyThe field campaign took place on May 28th, 2015 between 9:30 and 1:30 PST. The weather con-ditions were cloudless, convective and steady. Data measured at 24 m height from Vancouver-Sunset tower (Ca-VSu) showed temperatures between 20◦C and 22◦C and weak winds (averag-ing 2.5 ± 0.5 m s−1, the prevailing wind direction at 230◦).3.2.2 Distribution of CO2 Mixing RatiosFiltered Data PointsA total of 41,027 measurements were taken from the 5 DIYSCO2 systems during the 3.5 hourmeasurement campaign after filtering (see: Table 2.3). The results of the measurement cam-paign showed r ranging from 382.1 ppm to 906.0 ppm. The lowest r (<400 ppm) were measuredin the forest at Stanley Park, in select well vegetated residential streets, and the Mountain Viewcemetery. The highest values (>800 ppm) were measured in the downtown core and along themajor transport corridors such as Knight St. and West Georgia St. (highway 99). The medianand average of the r during the measurement campaign were 408.7 ppm and 419.0 ppm (std.dev. 28.71 ppm) respectively for the entire dataset. Figure 3.11 shows the frequency distribu-tion of the r measured by the DIYSCO2 systems minus the tower r sorted by ppm. The chartindicates that 11% of the measured r were greater than 450 ppm.Sorted data pointsMeasured CO2 Mixing Ratios − Average Tower Mixing Ratio (ppm)010020030040050089% of thedata is lessthan 450 ppm55% of thedata is lessthan 410 ppmFigure 3.11: Frequency distribution of measured r by the DIYSCO2 sensors minus thetower r sorted by ppm. The plot shows that 55% of the data is less than 410 ppmand 89.57% of the data is less than 450 ppm4749.22 49.24 49.26 49.28 49.30400500600700800900latitude (wgs84) − decimal degreesMeasured CO2 Mixing Ratios (ppm)lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllll ll llllll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllllll lllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllllllll lll lllllllllllllllllllllllllllllllllllllllllllllllllllllll lllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll l lllllllllllllll lllllllllllllllllllllllllllllllllllllllll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll l llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllll llllllll lllll llllllllll llllllllllllllll lll llll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll ll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllllllllllllllllllllllllllll ll lll llllll lllllll lllllll lllllllllllllllllllll llll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll ll l l lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllll lllllllll ll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llll llllllllll llllll lllllllllll lllllllllll l ll lllllllllllllllllllllllllll lllllll llllll l llllllllllllllllllllll llllllll lll llllllllllllll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllll llllllllllll llllll lll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll ll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllllllllll llllllllllll lll lllllllllllllll llllllllllllllllllllllllllllllllllllllllllll llllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll ll llll llll l llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllll ll l l l ll lllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll l l llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllllll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllll lllllll ll lllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllllll ll ll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llll ll l llllll llllllllllllllllllllllllllll lllll llll lllllllllllllllllllllllllllllllllll ll lllllllll l l llllllllllll lll l ll l llllllllllllllllllll l l llllllllllllllllllllllllllllllllll ll lll llllllllllllllllllllllllllllllllllllllllllllllllllllllllll lll lllllllllllllllllllll l llll ll llllllllllllllllllllllllll llllllllllllllllllllllllllllllllll ll ll llllllllllllllllllllllllllllllllllllllllllllllllllll llll ll l llll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll l ll l lllll l lll llllllllllllllllllllllllllllllllllllllllllllllllllllll llllllllllll lll l l l lll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llll llllllllllllllllllllllllllllllllll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllllllllll llllllllllllllllllllllllllllllll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllllllllll lllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllll l lllllllllllllllllllllllllllllllllllllllllllllllllllll llllllll llllllllllllllllllllllllllllllllllllllllll l lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllll ll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllllllllll llllllllllllllllllllllllllllllllllllllllllllllllll ll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllllllllllllll llllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll l llllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllllllllllllllll lllllllllllllll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll ll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllllllllllllllllllllllllllllllllllllllllll llllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll l lllllll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lll llllllllll lllll lllllllllllllllllllllllllllll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lll llll lllllllllllllllllllllllllllllllllllll llllllll lllllllllllllllllllll l l llll l llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll ll llllll llllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllllllllllllll ll l llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll ll ll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll41st AveKing EdwardBroadwayCambieBridgeDowntown StanleyParkFigure 3.12: CO2 mixing ratios sorted by latitude. Each point represents a one-seconddata point measured during the campaign. The distinct spikes in the data cor-respond to the some of the major arterial roads which run east-west through thestudy area (see labels).Figure 3.13: CO2 mixing ratios visualized on Google Earth. The red elevated bars cor-respond to high values of CO2 mixing ratios and the yellow lower bars correspondto low values of CO2 mixing ratios.483.2.3 Gridded DataGrid Averaged Mixing RatiosOf the 1169 grid cells of of the 100 m grid that could be traversed by a road, the case studycovered 965 of the grid cells, resulting in an 82.5% coverage of the study area. The grid averagedr at 100 m resolution ranged between 393.1 ppm and 518.0 ppm, averaged 417.6 ppm (standarddeviation of all 965 grid cells was 17.96 ppm), and had a median of 410.2 ppm. Table 3.6summarizes the gridded mean, median, maximum and standard deviation for r and the samplecounts over the entire study area. Figure 3.15 shows the spatial distribution of the grid averagedr for the 100 m grid. Figure 3.16 shows the grid averaged r for the 50 m, 200 m, and 400 mgrids; the data are summarized in Table 3.7.5452000 5454000 5456000 5458000 5460000 5462000400450500latitude (UTM zone 10N) − metersMeasured CO2 Mixing Ratios (ppm)llllllDowntown: LCZ1Fairview, Mount Pleasant: LCZ3Kensington−Cedar Cottage, Riley−park: LCZ3Stanley Park: LCZBSunset, Victoria−Fraserview: LCZ6West End: LCZ2Mean ConcentrationsFigure 3.14: Average CO2 mixing ratios by latitude colored by neighborhood for the 100m grid. The latitude increases from left to right. Each point represents one gridcell. The purple line indicates a moving average across the latitudinal gradient.49CO2 Mixing Ratios (ppm)        under 393393 to 408408 to 419419 to 434434 to 451451 to 474474 to 518over 5180 1 2kmFigure 3.15: Average CO2 Mixing Ratios aggregated to a 100 x 100 m grid.50CO2 Mixing Ratios (ppm)        under 393393 to 408408 to 419419 to 434434 to 452452 to 477477 to 530over 5300 1 2kmCO2 Mixing Ratios (ppm)        under 397397 to 408408 to 417417 to 428428 to 438438 to 451451 to 473over 4730 1 2kmCO2 Mixing Ratios (ppm)        under 400400 to 406406 to 413413 to 421421 to 428428 to 435435 to 448over 4480 1 2kmFigure 3.16: The grid averaged r for a grid size of 50 m (left), 200 m (center), and 400 m (right) grids.51Table 3.6: The summary data across 100 m grid. The table shows the minimum, 1stquartile, median, mean, 3rd quartile, and maximum (all values in ppm) for thegridded data.DataVariableMin 1stQuar-tileMed Mean 3rdQuar-tileMaxMean r 393.1 405.4 410.2 417.6 425.6 518.0Minimum r 382.1 399.3 401.6 403.0 404.9 518.0Median r 393.1 404.8 408.3 413.8 418.8 518.0Maximum r 393.3 410.8 432.7 460.8 486.8 906.0StandardDeviation r0.12 3.01 8.18 14.04 20.50 119.70SampleCounts0.00 9.00 26.00 32.11 50.00 279.00Table 3.7: The summary data across all grid sizes. The table shows the mean, minimum,median, maximum for the gridded data reported in ppm.GridSizeMin.(ppm)1stQu.(ppm)Median(ppm)Mean(ppm)3rdQu.(ppm)Max.(ppm)50 m 399.7 406.0 410.7 420.1 429.9 523.4100 m 399.5 406.0 410.3 418.5 427.3 500.8200 m 399.8 407.0 412.4 418.8 429.7 473.2300 m 401.3 408.7 416.4 419.6 428.0 460.2Overall, the grid cells covering major arterial roads and downtown core of Vancouver showedthe highest maximum, minimum, median and mean of CO2 mixing ratios. Conversely, the gridcells covering residential streets and forested trails of Stanley Park in general exhibited thelowest CO2 mixing ratios for the same statistics.Grid SkewnessFor the 100 m grid, over 65.69% of the cells had a positive skewness which means that morethan half of the measured 1 s mixing ratios in the grid cell were greater than the median valuesof the same grid cell. The map in Figure 3.17 shows that the cells with a skew value greaterthan 0 are positively skewed and coversely negatively skewed when the grid cells are less than0. The percentage of positively skewed grid cells decreased to 36.92% at a grid size of 50 m andincreased to 86.06% and 96.66% for the 200 m and 400 m grid cells, respectively. The skewnessindicates where there might be intra-grid peaks in CO2. Figure 3.18 shows the grid skewnessfor the 50 m, 200 m, and 400 m grids.52Skewless than 0greater than 00 1 2kmFigure 3.17: Map of the skewness per 100 m grid cell. If the skewness is less than 0, themedian r measured in that grid cell is greater than the mean. If the skewness ifgreater than 1, the median r measured in that grid cell is less than the mean.53Skewless than 0greater than 00 1 2kmSkewless than 0greater than 00 1 2kmSkewless than 0greater than 00 1 2kmFigure 3.18: From Left to Right: the grid skewness for the 50 m, 200 m, and 400 m grids. If the skewness is less than 0, themedian r measured in that grid cell is greater than the mean. If the skewness if greater than 1, the median r measured inthat grid cell is less than the mean.54Grid Sample CountsThe sample counts were recorded for the study area to examine the representativeness of thesamples of each grid cell. Figure 4.2 showed that for the 100 m grid, 91.31% of the grid cellscontained more than 10 samples per grid cell, 69.24% of cells contained more than 20 samples,and 28.32% of cell contained more than 50 samples. Grid cells with less than 10 samplesdropped from the analysis.grid cells sorted by number of samples containedNumber of valid samples per grid cell05010015020025091.31% GTE  10 Samples69.24% GTE  20 Samples28.32% GTE  50 SamplesFigure 3.19: Frequency distribution of the number of samples in each grid cell. Thedotted red lines identify the break points where cells have greater than or equalto 10, 20, and 50 samples.For the 50 m, 200 m, and 400 m, grids, the data showed that 57.73%, 98.85%, and 98.26%of the grid cells respectively had 10 samples or more per grid cell.Figure 3.25 and Figure 3.24 show where the sample counts are highest and lowest throughoutthe study area. It is clear that the sample counts are affected by the grid size. Furthermore,grid cells along major roads tended to have more sample counts as indicated in Figure 3.18 forthe 50 m grid cells.55grid cells sorted by number of samples containedNumber of valid samples per grid cell05010015057.73% GTE  10 Samples23.34% GTE  20 Samples1.06% GTE  50 SamplesFigure 3.20: Frequency distribution of the sample counts in each 50 m grid cell. 57.73%,23.34%, and 1.06% of the grid cells have greater than or equal to 10, 20, and 50samples or more.grid cells sorted by number of samples containedNumber of valid samples per grid cell010020030040098.85% GTE  10 Samples94.25% GTE  20 Samples75.57% GTE  50 SamplesFigure 3.21: Frequency distribution of the sample counts in each 200 m grid cell. 98.85%,94.25%, and 75.57% of the grid cells have greater than or equal to 10, 20, and 50samples or more.grid cells sorted by number of samples containedNumber of valid samples per grid cell020040060080098.26% GTE  10 Samples97.39% GTE  20 Samples91.3% GTE  50 SamplesFigure 3.22: Frequency distribution of the sample counts in each 400 m grid cell. 98.26%,97.39%, and 91.3% of the grid cells have greater than or equal to 10, 20, and 50samples or more.56Counts CO2 Samples        under 00 to 1111 to 3232 to 5656 to 8989 to 150150 to 279over 2790 1 2kmFigure 3.23: Map of the number of samples within each 100 x 100 m grid cell.57Counts CO2 Samples        under 00 to 33 to 1010 to 1717 to 2626 to 4040 to 70over 700 1 2kmCounts CO2 Samples        under 00 to 4949 to 102102 to 151151 to 202202 to 275275 to 412over 4120 1 2kmCounts CO2 Samples        under 7575 to 7575 to 204204 to 437437 to 586586 to 692692 to 831over 8310 1 2kmFigure 3.24: Maps of the number of samples within each 50 m (left), 200 m (center), and 400 m (right) grid cell.58Grid Standard DeviationA map of the standard deviations of the data per 100 m grid cell show that the highest standarddeviations are located along the major arterial roads and in the downtown. In contrast, theresidential areas appear to have lower standard deviations indicating less variability in r inthose areas. The maximum standard deviation decreases from 154.8 to 119.7, 65.76, and 57.12ppm for grid averaged r at 50 m, 100 m, 200 m, and 400 m grid cells, respectively. Figure 3.25and Figure 3.26 show the spatial distribution of the standard deviations for 100 m, and 50 m,200 m, and 400 m grids.Std. Dev. CO2 Concentrations (ppm)        under 00 to 77 to 1515 to 2626 to 4141 to 6464 to 120over 1200 1 2kmFigure 3.25: Map of the standard deviation from the mean value of each 100 m × 100 mgrid cell.59Std. Dev. CO2 Concentrations (ppm)        under 00 to 66 to 1414 to 2626 to 4343 to 7474 to 155over 1550 1 2kmStd. Dev. CO2 Concentrations (ppm)        under 11 to 66 to 1414 to 2424 to 3535 to 4646 to 66over 660 1 2kmStd. Dev. CO2 Concentrations (ppm)        under 44 to 99 to 1717 to 2626 to 3333 to 3838 to 48over 480 1 2kmFigure 3.26: Choropleth maps of the grid standard deviations of the mean r for each 50 m (right), 200 m (center), and 400 m(right) grid cell.603.2.4 Measured EmissionsThe relevant data measured from the Vancouver-Sunset tower during the experiment are sum-marized in Table 3.8. The raH for the period of the measurement campaign was calculated byaveraging H, averaging T0, and averaging Ttower over the three hours of the field campaign andtherefore is not time dependent. The resulting was a calculated raH of 34.14 s m−1; this valuewas used to derive the CO2 emissions from ctower and by cmobile. The measured CO2 emissionscalculated from the methods outlined in Section 2.1.2 showed a range of -12.04 kg CO2 ha−1 hr−1(net uptake) to 225.6 kg CO2 ha−1 hr−1 for the 100 m grid. The median and average emissionswere 20.54 and 34.46 kg CO2 ha−1 hr−1, respectively. The highest emissions in general werelocated in the downtown core and along the major transport corridors and intersections. Themeasured emissions for the 50 m, 200 m, and 400 m grid are reported in kg CO2 ha−1 hr−1and are shown in Figure 3.29. A table of summarizing the mean measured emissions for eachgrid size is shown in Table 3.13.5452000 5454000 5456000 5458000 5460000 5462000050100150200latitude (UTM zone 10N) − metersCalculated Emissions kg CO2ha−1 hr−1llllllDowntown: LCZ1Fairview, Mount Pleasant: LCZ3Kensington−Cedar Cottage, Riley−park: LCZ3Stanley Park: LCZBSunset, Victoria−Fraserview: LCZ6West End: LCZ2Mean ConcentrationsFigure 3.27: Scatterplot of measured emissions based on the aerodynamic resistance ap-proach ordered by latitude and colored by neighborhood. The black dotted lineindicates zero emissions. Each dot represents a 100 x 100 m grid cell average. Thepurple line is a rolling average at each latitude.61Table 3.8: Table of Values to Calculate Aerodynamic Resistance of HeatTime(PST)Long-waveup-welling(W m−2)T0(◦C)Ttower(◦C)H(W m−2)θ (K) ρa(g m−3)Cair (Jkg−1m−3)raH(s m−1)r at 24 m(ppm)ρCO2(g CO2 m−3)09:30 458.03 26.65 19.21 239.72 -7.20 1.194 1205.53 36.23 408.450 0.74510:00 466.22 27.98 19.67 228.20 -8.07 1.191 1203.31 42.56 402.486 0.73410:30 472.92 29.05 20.18 358.61 -8.64 1.189 1200.98 28.92 400.900 0.72911:00 480.11 30.20 20.57 302.73 -9.39 1.187 1199.35 37.22 401.932 0.73011:30 486.14 31.15 21.05 392.63 -9.86 1.185 1197.09 30.05 398.098 0.72212:00 488.49 31.51 20.90 423.72 -10.38 1.186 1197.57 29.33 394.869 0.71612:30 492.81 32.18 20.95 324.30 -10.99 1.185 1197.27 40.58 394.989 0.71613:00 497.39 32.89 21.38 445.03 -11.28 1.183 1195.25 30.29 393.850 0.713Average 480.26 30.20 20.49 339.37 -9.48 1.19 1199.55 34.40 399.45 0.7362Measured Emissions (kg CO2 ha−1 hr−1)          −12 to  00 to 1919 to 4141 to 6262 to 8484 to 110110 to 130130 to 150150 to 170over 1700 1 2kmFigure 3.28: Choropleth map of the grid measured emissions for the 100 m grid generatedusing the aerodynamic resistance approach using grid averaged r for the 100 mgrid. The grid cells colored in green are areas were photosynthetic uptake isstronger than emissions (net uptake).63Measured Emissions (kg CO2 ha−1 hr−1)          −12 to  00 to 2626 to 5454 to 8282 to 110110 to 140140 to 160160 to 190190 to 220over 2200 1 2kmMeasured Emissions (kg CO2 ha−1 hr−1)          −4.8 to  00 to 1717 to 3333 to 4848 to 6363 to 7979 to 9494 to 110110 to 120over 1200 1 2kmMeasured Emissions (kg CO2 ha−1 hr−1)          0.2 to  00 to 1212 to 2121 to 3030 to 3939 to 4848 to 5757 to 6666 to 75over 750 1 2kmFigure 3.29: Choropleth maps of the measured emissions for the 50 m (left), 200 m (center), and 400 m (right) grid cells. Thegrid cells colored in green are areas were photosynthetic uptake is stronger than emissions (net uptake).643.3 Methodology EvaluationThe following section compares the measured emissions to the emission inventory.3.3.1 Emissions InventoryFollowing the methods outlined in Section 2.4, the emissions estimates for buildings and trafficwere generated.Building Emissions InventoryThe building emissions inventory (1 m resolution) provided by van der Laan (2011) were aver-aged to the 50 m, 100 m, 200 m, and 400 m vector grids and scaled to their estimated hourlyvalues. Based on the calculations, the data for the 100 m grid showed a median and meanof 6.69 and 10.19 kg CO2 ha−1 hr−1, respectively. The maximum rate of building emissionswas 93.09 kg CO2 ha−1 hr−1 and was located in the downtown core. The building emissionsinventory only covers a subset of the study area as seen in Figure 3.30.Table 3.9 shows the spatial distribution of the building emissions for the 50 m grid, 200m grid, and 400 m grids. The gridded building emissions data for the varying grid sizes aresummarized in Table 3.9.Table 3.9: The summary data for the building emissions for all grid sizes. The tableshows the mean, minimum, median, maximum for the gridded data.GridSizeMin. 1stQu.Median Mean 3rdQu.Max.50 m 0.025 4.094 6.560 10.900 10.320 162.00100 m 0.027 4.457 6.69 10.19 10.31 93.09200 m 0.47 4.04 6.55 9.71 9.63 45.02400 m 2.13 3.78 5.84 7.58 8.55 31.9865Building Emissions (kg CO2 ha−1 hr−1)         no values below  00 to 33 to 88 to 1616 to 2929 to 4646 to 93over 930 1 2kmFigure 3.30: Map of building emissions inventory. The highest building emissions areconcentrated in the downtown area. The building emissions inventory only coversa subset of the study area.66Building Emissions (kg CO2 ha−1 hr−1)         no values below  00 to 44 to 1212 to 2525 to 4444 to 7676 to 162over 1620 1 2kmBuilding Emissions (kg CO2 ha−1 hr−1)         no values below  00 to 22 to 77 to 1111 to 2020 to 3535 to 45over 450 1 2kmBuilding Emissions (kg CO2 ha−1 hr−1)         no values below  00 to 00 to 44 to 88 to 1111 to 2020 to 32over 320 1 2kmFigure 3.31: Choropleth maps of the gridded building emissions inventory for the 50 m (left), 200 m (center), and 400 m (right)grids. The building emissions inventory covers only a subset of the study area.67Traffic Emissions InventoryThe 100 m gridded traffic emissions inventory showed median and mean emissions of 2.37and 12.50 kg CO2 ha−1 hr−1, respectively. As expected, the major roads and the areas withthe densest road network (e.g. downtown) exhibited the highest emissions, all of which weregreater than 18 kg CO2 ha−1 hr−1. The greatest traffic emissions in a single grid cell was 123.6kg CO2 ha−1 hr−1 and was located at the intersection of Cambie St. and West 2nd Avenue.The gridded traffic emsission data for the varying grid sizes are summarized in Table 3.10.Figure 3.33 shows the spatial distribution of the traffic emissions for the 50 m grid, 200 m grid,and 400 m grid.Table 3.10: The summary data for the traffic emissions for all grid sizes. The table showsthe mean, minimum, median, maximum for the gridded data.GridSizeMin. 1stQu.Median Mean 3rdQu.Max.50 m 0 0 1.857 13.91 10.96 212.6100 m 0 0.9281 2.374 12.5 20.81 123.6200 m 0 1.337 6.127 11.84 20 68.99400 m 0.515 5.261 8.596 9.707 12.29 39.9268Traffic Emissions (kg CO2 ha−1 hr−1)         under 00 to 66 to 1818 to 3232 to 4545 to 6464 to 124over 1240 1 2kmFigure 3.32: Map of traffic emissions inventory for the 100 m grid.69Traffic Emissions (kg CO2 ha−1 hr−1)         under 00 to 1010 to 2828 to 5454 to 8484 to 124124 to 213over 2130 1 2kmTraffic Emissions (kg CO2 ha−1 hr−1)         under 00 to 66 to 1515 to 2626 to 3636 to 5353 to 69over 690 1 2kmTraffic Emissions (kg CO2 ha−1 hr−1)         under 11 to 11 to 88 to 1111 to 1313 to 1919 to 40over 400 1 2kmFigure 3.33: Choropleth maps of the gridded traffic emissions inventory for the 50 m (left), 200 m (center), and 400 m (right)grids. The traffic emissions inventory covers the entire transect.70Total Emissions InventoryThe total emissions inventory is the sum of the building and traffic emissions estimates. Themedian and mean of the total emissions estimates were 11.11 and 23.87 kg CO2 ha−1 hr−1, re-spectively. The maximum emissions was 140.10 kg CO2 ha−1 hr−1 located at the intersectionof Cambie St. and West 2nd Avenue of which 88.25% were from traffic emissions ( 123.64kg CO2 ha−1 hr−1) and 13% were from building emissions (16.45 kg CO2 ha−1 hr−1). The grid-ded total emsission data for the varying grid sizes are summarized in Table 3.11. Figure 3.35shows the spatial distribution of the total emissions for the 50 m grid, 200 m grid, and 400 mgrid.Table 3.11: The summary data for the total emissions for all grid sizes. The table showsthe mean, minimum, median, maximum for the gridded data.GridSizeMin. 1stQu.Median Mean 3rdQu.Max.50 m 0.1069 6.429 9.864 24.98 31.96 185.5100 m 0.5482 6.863 11.11 23.87 39.38 140.1200 m 0.4652 7.776 16.05 23.11 29.94 94.46400 m 3.915 10.76 14.11 18.26 22.75 71.971Total Emissions (kg CO2 ha−1 hr−1)         under 00 to 66 to 2020 to 3838 to 6161 to 9393 to 140over 1400 1 2kmFigure 3.34: Map of total emissions inventory for the 100 m grid. The total emissionsinventory is the sum of the building and traffic emissions estimates.72Total Emissions (kg CO2 ha−1 hr−1)         under 00 to 99 to 2727 to 5555 to 8585 to 122122 to 213over 2130 1 2kmTotal Emissions (kg CO2 ha−1 hr−1)         under 00 to 1010 to 2222 to 3434 to 5454 to 7575 to 94over 940 1 2kmTotal Emissions (kg CO2 ha−1 hr−1)         under 11 to 77 to 1313 to 1818 to 2727 to 3737 to 72over 720 1 2kmFigure 3.35: Choropleth maps of the gridded total emissions inventory for the 50 m (left), 200 m (center), and 400 m (right)grids. The total emissions inventory is the sum of the traffic and building emissions inventories and covers only the EPiCCtransect.73Table 3.12: Average measured emissions and average building, traffic, and total emissionsof the entire study area by grid size. The emissions and uptake are reported inkg CO2 ha−1 hr−1Grid Size Avg.MeasuredEmissionsAvg.ModelledTotalEmissionsAvg.ModelledBuildingEmissionsAvg.ModelledTrafficEmissions50 m 34.01 24.98 10.90 13.91100 m 35.62 23.87 10.19 12.50200 m 37.32 23.11 9.71 11.84400 m 33.92 18.26 7.58 9.71The average measured emissions and average building, traffic, and total emissions of theentire study area are shown in Table 3.12.3.3.2 Relating Measured CO2 Concentrations and the Emissions InventoryThe measured CO2 mixing ratios were first compared to the emissions estimates to identifyif higher measured mixing ratios were related to higher hourly emissions estimates from theemissions inventory. Figure 3.37, Figure 3.36, Figure 3.38, Figure 3.39, show that as the CO2emissions inventory increases, the measured r also tends to increase for the 50 m, 100 m, 200 m,and 400 m grids. The relationship between measured r and traffic appears to show the strongestincreasing trend, while the relationship between measured r and the building emissions is lessclear. It is also observed that as the emission inventory increases, the range of the measured rbecomes greater.74llllllllllllll lllllllllllllllllllllllllll lllllllllllllll lllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllll lllllllllllllllllllllll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllll ll llllllllllllllllllllllllllllllllll llllllll lllllllll llllllllllllllllllllllllllllllllllllllll llllll lllll ll lllllllll llll lllllllllllllllllllllll llllllllllllllllllll llllll llllllllllllllllllllllllllllll lll l lllllllll lllllllllllllllllllllllll lllllllll llllll lllllllllllllllllllll lllll lllllllllllllllllllllllllllllllllllll lllllllllllllllllllll lllll llllllllllllllllllllll llllllllllllllllllllllllllllllllll llllll l l lllllllllllllllll llllll ll llll l ll llllll lllllllllll lllll lllll llll l ll lll lllllllllllllllll ll ll lllllllllllllllllll llll llllllllll lllllllllll lllllllllllllllllllll llllllllllll llllll lllllll lllllllll l llllll lllllll llllllllllllll ll lllllllllllllll llllll llllllllllllll llllllllllllllllllllll llllllllllllllll llllllllllll lllllllllllllllllllllll l llllllllll llllll llllllll llll lllllll lll llllll lll lllllllllll lll ll llllll llll llllll l0.0 0.5 1.0 1.5 2.0400420440460480500520Binned Building Emissions (log of kg CO2 ha−1 hr−1) Measured Concentrations (CO2 ppm)1 to 21.1 41.2 to 61.4 102 to 122 142 to 162400420440460480Binned Building Emissions (log of kg CO2 ha−1 hr−1) Measured Concentrations (CO2 ppm)llllllllllllll lllllllllll lllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllll l lllllllllllllllllllllllllllllllllllllll llll lllllllllllll l lllllllllllllll lllllllllll lllllllllllllllllll lllllllllllllllll llll lllllllllllllllllllllllllllllllllllll llllllllll llllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllll lllllllllllllllllllllllll llllll llllllllllllllllllllllllllllllllllllllll llllllllllllllllll lllllllllllllllllllllllllllllllllllll lllllllllll llllll lllllllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllll llllllll llllllllllllllllllllllllllllllll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllllllllll llllllllllllllllllllllllllllllllllll llllll llllllllll ll llllllllllllllllllll lllll ll llllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllllllllllllllllllllllllll llllllllllllll llllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllll llllllllllll llllllllllllllllllllllllllllllllllllllllllllllll0.0 0.5 1.0 1.5 2.0400420440460480500520Binned Traffic Emissions (log of kg CO2 ha−1 hr−1) Measured Concentrations (CO2 ppm)1 to 22.2 64.5 to 85.6 128 to 149 191 to 213400420440460480500Binned Traffic Emissions (log of kg CO2 ha−1 hr−1) Measured Concentrations (CO2 ppm)lllllllllllll llllllllll lllllllllllllllllllllllllllllllllll lllllllllllllllllllllllllllllllllllll llllllllllllllllllllllll lll lllllllllllll lllllllllllll lllllllllll ll lllllllllllllll llll llllllllll ll l lllllllllllllllllllll lllllllllllllll llllllllllllllllllll lllllllllllll llllllllllllllllllllllllllllll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll ll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllllllllllllllllllllllllllllllllllllllllllllllllllll lllll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllllll lllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllll lllllllll llllllllllllllllllllllllllllllllllllllll lllllllllllllllllllllll llllllll llllllllllllllllllllllllllllllllllllllllllllllll lll lllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllll llll llllllllllllllllllllllllll llllllll ll l llllllllllllllllll lllllllllllllllllllll llllllllllllllll lllllllllllllllllllll lllllll llllllllllllllll l lllllll llllll llllllllllllllllllllllllll lllll llll lllllllllllll l llllllllllllllllllllllllllllllllllllll lllllllllll llllllllll lllllllllllllllllllllllllllll llllllllllllll lllllllllllllllllllllllll l lllllllll lllllllllllllllllllll lllllllllll llllllllll ll lll lll llllllllllll lllllllllllllllllllllllllllllllllllllllllllllllllllll lllll llll lllllllllllll llllllllllllllll llllll lllllllllllllllllllll llllllllllllll0.5 1.0 1.5 2.0400420440460480500520Binned Total Emissions (log of kg CO2 ha−1 hr−1) Measured Concentrations (CO2 ppm)2 to 23.1 65.2 to 86.2 128 to 149 192 to 213400420440460480500Total Emissions Inventory (kgCO2ha−1)Measured Concentrations (CO2 ppm)Figure 3.36: Scatter plot and box-and-whisker plot showing the relationship betweenmeasured r and the building, traffic, and total emissions inventories for the 50 mgrid.75llllllllllllllllllllllllllllllllllllllllllll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lll lllll llllllll llllll llll lllllllllllll lllllllllll llll lllllllllllll lllllllllllllllllll lllllll lllllllllllll lllllllll llllll llllllll llllllllll lllllllllllll llllllllll lllllllllllllllllllllllllll llllllllllllllll lllllllllllll lllllll llllllllllll llllllllllllllllllllllllllllllllllllllllllllllll l l llllllllllllllllllllllllllllllllllllllllllllllllll lllllllll ll llll lll llllllll lll0.0 0.5 1.0 1.5 2.0400420440460480500520Binned Building Emissions (log of kg CO2 ha−1 hr−1) Measured Concentrations (CO2 ppm)1.04 to 16.4 31.7 to 47.1 62.4 to 77.8400420440460480Binned Building Emissions (log of kg CO2 ha−1 hr−1) Measured Concentrations (CO2 ppm)l llll llllllllllllllllllllllllllllllllllllllllllllllllllll lllllllll llllllll l llllllllllll llllllllllllll lllllllllllllll llllll lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllllllllllllllll ll lllll llllllll lllllll llllllllll llllllllll llllllllllllllll llllllllllllll llllllllllll llllllllllllllllllllllllllllllllllll llllllllllllllllllllllllll lllllllllllllllllllllllllllllllll lllllllllll llllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllll llll lllllllllllllllllllllllllllll0.0 0.5 1.0 1.5 2.0400420440460480500520Binned Traffic Emissions (log of kg CO2 ha−1 hr−1) Measured Concentrations (CO2 ppm)1.03 to 21.5 41.9 to 62.3 82.8 to 103400420440460480Binned Traffic Emissions (log of kg CO2 ha−1 hr−1) Measured Concentrations (CO2 ppm)l lllll lllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllll l llllllllll llllllllllllll lllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllllll llllllllll llllllllllllllllll ll llllllllllll llllllllllllllllllllllllllllllll lllllll llllllllll lllllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllll lllllllllllllllllllllllllllllllllllllllllllllllllllll lllllllllllllllllllllllllllllllll l llllllllllllllllllllllll llllllllllllllllllllllllllllllllllllll llllllllllllllllllll lllllllllllll0.5 1.0 1.5 2.0400420440460480500520Binned Total Emissions (log of kg CO2 ha−1 hr−1) Measured Concentrations (CO2 ppm)2.02 to 25 48 to 71.1 94.1 to 117400420440460480Total Emissions Inventory (kgCO2ha−1)Measured Concentrations (CO2 ppm)Figure 3.37: Scatter plot and box-and-whisker plot showing the relationship betweenmeasured r and the building, traffic, and total emissions inventories for the 100m grid.76lllllll lllllllllllllllllllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllll llllllllllllllllllllllllllllllllllllllllllllllll0.0 0.5 1.0 1.5400420440460Binned Building Emissions (log of kg CO2 ha−1 hr−1) Measured Concentrations (CO2 ppm)1.08 to 6.57 17.6 to 23.1 34 to 39.5400410420430440450Binned Building Emissions (log of kg CO2 ha−1 hr−1) Measured Concentrations (CO2 ppm)llllllll llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll lllllllllllllll lllllll0.0 0.5 1.0 1.5400420440460Binned Traffic Emissions (log of kg CO2 ha−1 hr−1) Measured Concentrations (CO2 ppm)1.01 to 9.5 18 to 26.5 35 to 43.5 52 to 60.5400410420430440450460470Binned Traffic Emissions (log of kg CO2 ha−1 hr−1) Measured Concentrations (CO2 ppm)llllllll llllllllllllllllllllllllllllllll lllllllllllllllllllllllllllllllllllllllllllllllllll lllllllllllllllllllllllllllllllllllll lllllllllllllllllllllllllllllllllllllllllllllllllll0.5 1.0 1.5 2.0400420440460Binned Total Emissions (log of kg CO2 ha−1 hr−1) Measured Concentrations (CO2 ppm)2.03 to 13.6 36.7 to 48.2 71.4 to 82.9400410420430440450460470Total Emissions Inventory (kgCO2ha−1)Measured Concentrations (CO2 ppm)Figure 3.38: Scatter plot and box-and-whisker plot showing the relationship betweenmeasured r and the building, traffic, and total emissions inventories for the 200m grid.77lllllllllllllllllll lll0.4 0.6 0.8 1.0 1.2 1.4400410420430440Binned Building Emissions (log of kg CO2 ha−1 hr−1) Measured Concentrations (CO2 ppm)2.13 to 9.59 17.1 to 24.5400410420430440Binned Building Emissions (log of kg CO2 ha−1 hr−1) Measured Concentrations (CO2 ppm)lllllllllllllllllllllllll lll0.0 0.5 1.0 1.5410420430440Binned Traffic Emissions (log of kg CO2 ha−1 hr−1) Measured Concentrations (CO2 ppm)1.04 to 14 14 to 27 27 to 40410420430440Binned Traffic Emissions (log of kg CO2 ha−1 hr−1) Measured Concentrations (CO2 ppm)llllllllllllllllllllllllll lll0.6 0.8 1.0 1.2 1.4 1.6 1.8400410420430440Binned Total Emissions (log of kg CO2 ha−1 hr−1) Measured Concentrations (CO2 ppm)3.92 to 26.6 26.6 to 49.2 49.2 to 72400410420430440Total Emissions Inventory (kgCO2ha−1)Measured Concentrations (CO2 ppm)Figure 3.39: Scatter plot and box-and-whisker plot showing the relationship betweenmeasured r and the building, traffic, and total emissions inventories for the 400m grid.78A linear model was used to quantify each of the the relationships between the medianmeasured r and emissions inventories 2. The linear model was fit through the the grid medianr (y-axis) binned into equal intervals by the emissions inventories for buildings, traffic, and thetotal (log10 x-axis). The emissions inventories were log10 transformed to better distinguish thedifferences in the grid median r, given that most of the r tended to be left skewed. The datashowed strong linearity for all of the relationships between the gridded median r and the binnedemissions inventories. For the 100 m grid, the R2 between measured r and building, traffic, andtotal emissions was 0.9867, 0.884, and 0.9231, respectively. Building emissions were least ableto predict the variability in the gridded median r for the 50 m grid and the most able using the100 m grid. For all the grid sizes, the total emissions were able to predict at least 89% of thevariability in gridded median r according to the linear model.llllll0 10 20 30 40 50 60152025303540Binned Building Emissions (log of kg CO2 ha−1 hr−1) CO2 Mixing Ratio Enhancement (ppm) y = 0.5028x + 10.248R2 = 0.9867 llllll0 20 40 60 80101520253035Binned Traffic Emissions (log of kg CO2 ha−1 hr−1) CO2 Mixing Ratio Enhancement (ppm) y = 0.3727x + 9.447R2 = 0.884 llllll0 20 40 60 80 1005101520253035Total Emissions Inventory Binned kgCO2ha−1hr−1 CO2 Mixing Ratio Enhancement (ppm) y = 0.3344x + 6.5412R2 = 0.9231Figure 3.40: Linear fits of the cmobile and the binned emissions inventories for building,traffic, and total emissions data for the 100 m grid. The R2 between r andbuilding, traffic, and total emissions was 0.9867, 0.884, and 0.9231, respectively2399.45 ppm which was the average r measured at Vancouver-Sunset tower for the measurement campaign,see Table 3.8 was subtracted from the median measured r by the DIYSCO2 systems79llllll ll0 20 40 60 80 100 12015202530Binned Building Emissions (log of kg CO2 ha−1 hr−1) CO2 Mixing Ratio Enhancement (ppm) y = 0.107x + 16.3212R2 = 0.304llllllllll0 50 100 15010152025303540Binned Traffic Emissions (log of kg CO2 ha−1 hr−1) CO2 Mixing Ratio Enhancement (ppm) y = 0.2264x + 11.0873R2 = 0.9123llllllllll0 50 100 15010152025303540Total Emissions Inventory Binned kgCO2ha−1hr−1 CO2 Mixing Ratio Enhancement (ppm) y = 0.2235x + 9.5391R2 = 0.9009(a) The R2 between r and building, traffic, and total emissions was 0.304, 0.9123,and 0.9009, respectivelyllllllll0 5 10 15 20 25 30 35101520253035Binned Building Emissions (log of kg CO2 ha−1 hr−1) CO2 Mixing Ratio Enhancement (ppm) y = 0.7385x + 7.4915R2 = 0.7831 llllllll0 10 20 30 40 505101520253035Binned Traffic Emissions (log of kg CO2 ha−1 hr−1) CO2 Mixing Ratio Enhancement (ppm) y = 0.5809x + 7.0038R2 = 0.9682 llllllll0 20 40 601015202530Total Emissions Inventory Binned kgCO2ha−1hr−1 CO2 Mixing Ratio Enhancement (ppm) y = 0.3952x + 6.3886R2 = 0.9192(b) The R2 between r and building, traffic, and total emissions was 0.7831, 0.9682,and 0.9192, respectivelyllll5 10 15 2010152025Binned Building Emissions (log of kg CO2 ha−1 hr−1) CO2 Mixing Ratio Enhancement (ppm) y = 0.9285x + 3.7332R2 = 0.9806 lll5 10 15 20810121416Binned Traffic Emissions (log of kg CO2 ha−1 hr−1) CO2 Mixing Ratio Enhancement (ppm) y = 0.4277x + 7.0533R2 = 0.8732lll10 20 30 401012141618202224Total Emissions Inventory Binned kgCO2ha−1hr−1 CO2 Mixing Ratio Enhancement (ppm) y = 0.3735x + 7.0855R2 = 0.8924(c) The R2 between r and building, traffic, and total emissions was 0.9806, 0.8732,and 8924, respectivelyFigure 3.41: Graph showing the linear fits of the r and the binned emissions inventoriesfor building, traffic, and total emissions data for (a) the 50 m grid, (b) the 200 mgrid, and (c) the 400 m grid.803.3.3 Relating Calculated Emissions to the Emissions InventoryAbsolute and Relative DifferencesFirst, the measured emissions and emissions inventories are evaluated for their absolute differ-ences. Figure 3.42 shows the frequency distribution of the absolute differences for the 100 mgrid; the positive values indicate that the measured emissions are greater than the total emis-sions inventory whereas negative values indicate the converse. The graph shows that 73.76%of the measured emissions are greater than the total emissions inventory. The data show that53.94% of the measured emissions are within ±10 kg CO2 ha−1 hr−1 of the total emissions in-ventory estimates and 76.51% of the measured emissions are within 20 kg CO2 ha−1 hr−1. Themeasured emissions therefore are overestimating emissions reported in the emissions inventory.Most of the overestimation is occuring along the major roads whereas the underestimation isoccuring in the residential areas and in the downtown area as seen in Figure 3.43.The grid cell that shows the greatest overestimation of the emissions inventory by the mea-sured emissions is at Cambie Street, between 13th and 14th Avenue. The data showed thatthe difference was 156.36 kg CO2 ha−1 hr−1. The traffic emissions and building emissions esti-mates together are below 12.41 kg CO2 ha−1 hr−1, however, the measured emissions was 168.77kg CO2 ha−1 hr−1. Where the emissions estimates are most underestimated by the measuredemissions is at Howe St. and Robson St. with a difference of 68.00 kg CO2 ha−1 hr−1. In thislocation, the emissions inventory are almost equally contributed by building emissions (55.023kg CO2 ha−1 hr−1) and by traffic emissions (49.29 kg CO2 ha−1 hr−1), but the measured emis-sions showed a value of 36.31 kg CO2 ha−1 hr−1. Figure 3.44 shows the spatial distribution ofthe absolute differences for the 100 m grid, 50 m grid, 200 m grid, and 400 m grid.81Grid cells sorted by differenceMeasured Emissions − Inventory (kgCO2ha−1 hr−1 )−200204060Zero Line+/− 10+/− 2050mGrid cells sorted by differenceMeasured Emissions − Inventory (kgCO2ha−1 hr−1 )−50050100Zero Line+/− 10+/− 20100mGrid cells sorted by differenceMeasured Emissions − Inventory (kgCO2ha−1 hr−1 )0100200300 Zero Line+/− 10+/− 20200mGrid cells sorted by differenceMeasured Emissions − Inventory (kgCO2ha−1 hr−1 )0100200300400500600Zero Line+/− 10+/− 20400mFigure 3.42: Absolute differences: Measured Emissions minus the emissions inventoryfor the 50 m grid, 100 m grid, 200 m grid, and 400 m grid from topleft, tobottom right. The black dashed line indicates where there is zero over- or -under estimation. The red and purple dashed lines indicate where the measuredemissions are within ±10 and ±20 kg CO2 ha−1 hr−1 of the emissions estimate.Units are in kg CO2 ha−1 hr−182Absolute Emissions Differences: (kg CO2 ha−1 hr−1)         under −120−120 to −50−50 to −20−20 to 00 to 2020 to 5050 to 120over 1200 1 2kmFigure 3.43: Map of the absolute differences between the emissions between the emissionsestimates and the calculated emissions for the 100 m grid.83Absolute Emissions Differences: (kg CO2 ha−1 hr−1)         under −120−120 to −50−50 to −20−20 to 00 to 2020 to 5050 to 120over 1200 1 2kmAbsolute Emissions Differences: (kg CO2 ha−1 hr−1)         under −120−120 to −50−50 to −20−20 to 00 to 2020 to 5050 to 120over 1200 1 2kmAbsolute Emissions Differences: (kg CO2 ha−1 hr−1)         under −120−120 to −50−50 to −20−20 to 00 to 2020 to 5050 to 120over 1200 1 2kmFigure 3.44: Choropleth maps of the gridded absolute differences in the measured emissions minus the total emissions inventoryfor the 50 m (left), 200 m (center), and 400 m (right) grids.84Sorted Grid CellsRelative difference: Measured Emissions − Inventory (%)010203040506050mSorted Grid CellsRelative difference: Measured Emissions − Inventory (%)01020304050100mSorted Grid CellsRelative difference: Measured Emissions − Inventory (%)0246200mSorted Grid CellsRelative difference: Measured Emissions − Inventory (%)0.00.51.01.52.02.5400mFigure 3.45: Frequency distribution of the relative differences for all the grids.85Table 3.13: Measured Emissions mean vs. emissions inventory mean for all grid sizes inthe study area subset. Note that the mean values for the emissions inventory hereare different from those found in Table 3.12 due to the study area subset.Grid Size MeasuredEmissions Meankg CO2 ha−1 hr−1EmissionsInventory Meankg CO2 ha−1 hr−1Differencekg CO2 ha−1 hr−150 m 38.18 34.46 3.72100 m 35.67 25.41 10.26200 m 37.32 23.19 14.13300 m 33.920 18.26 15.66The relative emissions difference is defined as the difference between the measured emissionsand emissions inventory divided by the emissions inventory. In other words, it is the absolutedifference divided by the measured emissions. The data for the 100 m grid show that 42.20% ofdata have a difference within a factor of ±2 and 66.61% of data have a difference within a factorof 10. The locations with the highest relative differences were both locations in which the build-ing and traffic emissions inventories estimated almost no emissions (<0.5 kg CO2 ha−1 hr−1),but where the calculated emissions were greater than 20 kg CO2 ha−1 hr−1. Figure 3.45 showshistograms of the frequency distribution of the relative differences for the 50 m, 100 m, 200 m,and 400 m grids.Last, in comparing the means across the study area subset between the measured emis-sions and the emissions inventory, the data for the 100 m grid show that the mean of themeasured emissions was 35.67 kg CO2 ha−1 hr−1 as compared to the 25.41 kg CO2 ha−1 hr−1 ofthe emissions inventory; a difference of 10.26 kg CO2 ha−1 hr−1. The difference between themeans increases as the grid size increases as seen in Table 3.13. The least amount of differenceis seen in the 50 m grid at 3.72 kg CO2 ha−1 hr−1. The RMSE between the measured emis-sions and the total emissions inventory for the 50 m grid, 100 m grid, 200 m grid, and 400 mgrid were 29.50 kg CO2 ha−1 hr−1, 24.12 kg CO2 ha−1 hr−1, 22.22 kg CO2 ha−1 hr−1, and 19.20kg CO2 ha−1 hr−1, respectively.Correlating Measured Emissions and the Emissions InventoriesThe measured emissions and the emissions inventories were compared directly, plotted withdouble logarithmic axes. Figure 3.46 shows the calculated emissions as a function of the buildingemissions estimates. Data less than 1 kg CO2 ha−1 hr−1 are withheld from the analysis. Thedata show that 80.04%, 87.25%, 86.01%, and 95.45% of the measured emissions are within afactor of ± 10 of the building emissions estimates for the 50 m, 100 m, 200 m, and 400 mgrids, respectively. The building emissions appear to be clustered by neighborhood with thelowest urban density (LCZ6) of Sunset, Victoria-Fraserview exhibiting the lowest emissions and86Downtown with the highest urban density (LCZ1) exhibiting the highest building emissions.The measured emissions and the building emissions estimates were found to be correlatedpositively by 19.38% for the 50 m grid, 36.70% for the 100 m grid, 51.46% for the 200 m grid,and 60.67% for the 400 m grid.Figure 3.47 shows the measured emissions as a function of the traffic emissions inventorywith. Data less than 1 kg CO2 ha−1 hr−1 are withheld from the analysis. The data show that89.53%, 85.81%, 88.89%, and 90.48% of the measured emissions are within a factor of ± 10of the traffic emissions estimates for the 50 m, 100 m, 200 m, and 400 m grids, respectively.The data show that the measured emissions are more closely aligned with the areas with highertraffic emissions but performs less well in the areas with lower traffic emissions and where theurban density is lower (e.g. Sunset and Victoria-Fraserview). As a result, there appears to bea greater spread in the relationship between the measured emissions and traffic emissions. Themeasured emissions and the traffic emissions inventory were found to be correlated positivelyby 73.68% for the 50 m grid, 74.10% for the 100 m grid, 83.92% for the 200 m grid, and 89.37%for the 400 m grid.Last, Figure 3.48 shows the measured emissions as a function of the total emissions inventory.Data less than 2 kg CO2 ha−1 hr−1 are withheld from the analysis. The data show that 99.31%,99.43%, 100%, and 100% of the measured emissions are within a factor of ± 10 of the trafficemissions estimates for the 50 m, 100 m, 200 m, and 400 m grids, respectively. The measuredemissions and the total emissions inventory were found to be correlated positively by 77.27%for the 50 m grid, 78.43% for the 100 m grid, 84.24% for the 200 m grid, and 89.27% for the400 m grid.87−0.5 0.5 1.0 1.5 2.0 2.5 3.0−0.50.51.52.5Building Emissions (log of kg CO2 ha−1 hr−1)Measured Emissions (log of kgCO2ha−1hr−1)50m−0.5 0.5 1.0 1.5 2.0 2.5 3.0−0.50.51.52.5Building Emissions (log of kg CO2 ha−1 hr−1)Measured Emissions (log of kgCO2ha−1hr−1)100m−0.5 0.5 1.0 1.5 2.0 2.5 3.0−0.50.51.52.5Building Emissions (log of kg CO2 ha−1 hr−1)Measured Emissions (log of kgCO2ha−1hr−1)200m−0.5 0.5 1.0 1.5 2.0 2.5 3.0−0.50.51.52.5Building Emissions (log of kg CO2 ha−1 hr−1)Measured Emissions (log of kgCO2ha−1hr−1)400mDowntown: LCZ1Fairview, Mount Pleasant: LCZ3Kensington−Cedar Cottage, Riley−park: LCZ3Stanley Park: LCZBSunset, Victoria−Fraserview: LCZ6West End: LCZ2Figure 3.46: Double logarithmic plot of measured emissions versus building emissionsestimates. Both the x and y axes are log10 transformed. Three red lines are shownon the plot; the topmost and bottommost lines indicate 1 order of magnitudegreater or less than the zero line which indicates a perfect correlation. Data lessthan 1 kg CO2 ha−1 hr−1 are withheld from the analysis.88−0.5 0.5 1.0 1.5 2.0 2.5 3.0−0.50.51.52.5Traffic Emissions (log of kg CO2 ha−1 hr−1)Measured Emissions (log of kgCO2ha−1hr−1)50m−0.5 0.5 1.0 1.5 2.0 2.5 3.0−0.50.51.52.5Traffic Emissions (log of kg CO2 ha−1 hr−1)Measured Emissions (log of kgCO2ha−1hr−1)100m−0.5 0.5 1.0 1.5 2.0 2.5 3.0−0.50.51.52.5Traffic Emissions (log of kg CO2 ha−1 hr−1)Measured Emissions (log of kgCO2ha−1hr−1)200m−0.5 0.5 1.0 1.5 2.0 2.5 3.0−0.50.51.52.5Traffic Emissions (log of kg CO2 ha−1 hr−1)Measured Emissions (log of kgCO2ha−1hr−1)400mDowntown: LCZ1Fairview, Mount Pleasant: LCZ3Kensington−Cedar Cottage, Riley−park: LCZ3Stanley Park: LCZBSunset, Victoria−Fraserview: LCZ6West End: LCZ2Figure 3.47: Correlation plot of measured emissions versus traffic emissions inventory.Both the x and y axes are log10 transformed. Three red lines are shown on theplot;the topmost and bottommost lines indicate 1 order of magnitude greater orless than the zero line which indicates a perfect correlation. Data less than 1kg CO2 ha−1 hr−1 are withheld from the analysis.89−0.5 0.5 1.0 1.5 2.0 2.5 3.0−0.50.51.52.5Total Emissions (log of kg CO2 ha−1 hr−1)Measured Emissions (log of kgCO2ha−1hr−1)50m−0.5 0.5 1.0 1.5 2.0 2.5 3.0−0.50.51.52.5Total Emissions (log of kg CO2 ha−1 hr−1)Measured Emissions (log of kgCO2ha−1hr−1)100m−0.5 0.5 1.0 1.5 2.0 2.5 3.0−0.50.51.52.5Total Emissions (log of kg CO2 ha−1 hr−1)Measured Emissions (log of kgCO2ha−1hr−1)200m−0.5 0.5 1.0 1.5 2.0 2.5 3.0−0.50.51.52.5Total Emissions (log of kg CO2 ha−1 hr−1)Measured Emissions (log of kgCO2ha−1hr−1)400mDowntown: LCZ1Fairview, Mount Pleasant: LCZ3Kensington−Cedar Cottage, Riley−park: LCZ3Stanley Park: LCZBSunset, Victoria−Fraserview: LCZ6West End: LCZ2Figure 3.48: Correlation plot of measured emissions versus total emissions inventory.Both the x and y axes are log10 transformed. Three red lines are shown on theplot; the topmost and bottommost lines indicate 1 order of magnitude greateror less than the zero line which indicates a perfect correlation. Data less than 2kg CO2 ha−1 hr−1 are withheld from the analysis.90Chapter 4Discussion4.1 Evaluting the Mobile Sensor SpecificationsHigh Precision Mobile SensorThe test for the DIYSCO2 system precision was determined to be 0.233 ppm. The precisionof the DIYSCO2 system is therefore well within the IRGAs 1 ppm precision as specified bythe manufacturer. By calibrating the system following the procedures listed in Section 2.2.3and testing the precision of the DIYSCO2 system as a whole, it was possible to quantify thatthe other components of the system such as the microcontroller, GPS, filter, and sample inletand outlet tubes were not affecting the performance of the IRGA. The results of the precisiontest therefore demonstrated that low-cost and open source microcontrollers and their relatedcomponents, can be used together with high precision instruments without any major affectson the precision of the sensor.Time Dependent AccuracyThe accuracy of the system between the five DIYSCO2 systems was determined to be within±3 ppm over a 7 day measurement period; the system drift was 6 ppm. Figure 3.3 shows theDIYSCO2 drift from the mean as well as the initial offset needed to “pull” the sensors to acommon initial starting value. Despite calibrating each of the five DIYSCO2 systems prior tothe 7 day lab experiment, the sensors needed to be offset within some range (e.g. between 0.29ppm and -1.77 ppm). This might be explained by the varying ages and wear of the IRGAsprior to building them into the DIYSCO2 system. Other factors such as slight variations in thetube length (e.g inlet tube, outlet tube, or both), the flow rate of the pump, or input voltagesof the power source may be affecting the measurement accuracy as well. While the inlet tubelengths were all measured to be 3 m, there is still some tubing that is internal to the mobilesensor system that is used connect the inlet tube to the filter, pump, IRGA, and outlet tube.Together with slight changes in the flow rate, a small increase or decrease in tube length could91affect the measurement delay time and therefore the measured accuracy.Figure 3.3 shows that each of the DIYSCO2 systems were behaving differently relative tothe mean over the course of the 7 day experiment. While DIYSCO2 systems with IDs 108and 1641 seemed to perform well relative to the mean, the DIYSCO2 systems with IDs 150and 205 diverged away from the mean, setting the upper and lower limits of observed drift. Itis important to highlight that the DIYSCO2 drift was not necessarily linear. The DIYSCO2with the ID of 151 behaved similarly with DIYSCO2 systems with IDs 108 and 1641 until 60hours into the experiment and then drifted almost to the same levels as the DIYSCO2 withthe ID of 205. The data measured in this drift test could be used to offset the sensors overthe course of a measurement campaign to help improve the accuracy and account for sensordrift for longer term studies. In the case of the 3.5 hour experiment performed in this study,we could expect a drift of at least ±0.5 ppm. It is possible that the impact of vibrationsand other environmental factors while driving may impact the sensor accuracy and drift andtherefore attempt to minimize those error by performing in-situ accuracy tests before and afterexperiments as described in Section 2.3.6. The results of our in-situ accuracy tests showed thatfor the 3.5 hour field campaign the absolute drift of the DIYSCO2s was 0.95 ppm. The driftduring the field campaign was 0.13 ppm greater than what was found in the lab test. Thisdifference however is less than the precision of the instrument and therefore shows that thedrift was not dramatically different than what was tested in the lab.It should be noted that the DIYSCO2 system is tested for accuracy and drift under controlledlab conditions but not in the field. What is missing from the sensor system is a method toaddress the effects of each sensor drift while in the field on a continuous basis as is done inother mobile traverses such as in Bukowiecki et al. (2002).Possible methods for in-field calibrations may include the use of manual calibrations usingTedlar sampling bags filled with a known concentration of CO2 and applying them to the sampleinlet of the sensor system. Similarly, a more precise method would be to use a small, pressurizedtank filled with a known concentration of gas as opposed to the sampling bags to reduce thepossibility of contamination when filling the sampling bags. Another more automated method ofcalibrating the sensor systems is to use a chemical scrubber which is built into the sensor systemthat, when triggered, scrubs out the CO2 from the incoming samples using chemicals such asAscarite II and Magnesium Perchlorate (Li-Cor, 2008). While the use of chemical scrubberswould require changing the chemicals once they begin to lose effectiveness, this method wouldprovide a lightweight solution for a sensor integrated calibration system.Last, it is important to also employ a “non-invasive” method to check the alignment of thevarious DIYSCO2 systems and their drift during a field campaign. One such method is thein-situ accuracy test. By running the DIYSCO2 systems together at the beginning and end ofa field campaign, it is possible to determine how much drift occurred during the measurementperiod. However a relationship must also be established from the mobile sensors to the CO2sensor at the fixed location above the height of the RSL (e.g. the tower). To do this, the92sensors must be placed at the same height as the fixed location sensor (e.g. the tower) and runtogether, side-by-side before and after an experiment or measurement campaign or periodicallythroughout a longer term campaign (e.g. weeks to months). This would help account for sensordrifts that can occur between the mobile sensor systems and between the tower sensor. Thiswould help reduce the possible errors between the sensors. It is important to note that for thefield campaign presented in this research, the DIYSCO2 systems were calibrated in the lab priorto the experiment and therefore the values were not adjusted to the values measured duringthe 5 minute sensor alignment test before and after the field campaign.Measurement Delay TimeStudies such as Ha¨b et al. (2015) and Hu¨bner et al. (2014) show the importance of correcting forsensor lag time in order to properly attribute environmental measurements in space and timefor mobile transect measurements. In our study, the measurement delay time was tested in alab experiment and determined to be 18.2 seconds for the sensor system with a 3 m long sampleinlet tube and flow rate of 700 cc min−1. If it takes 14 to 15 seconds for a vehicle traveling, 25km hr−1 (or 7 m s−1) horizontally across a 100 m × 100 m grid cell, then it is clear that thecurrent design of the DIYSCO2 system would be incorrectly attributing measurements fromthe grid cell it had previously traveled through to its current location. Thus, in this example,the spatial error would be 126 m. The spatial misrepresentation of the measurements wouldonly intensify when driving at higher speeds.Inlet LocationStudies such as those by Bukowiecki et al. (2002), Tao et al. (2015) and Crawford et al. (2011)show the importance of designing an inlet to avoid contamination of the samples by the vehi-cle’s own exhaust, the exhaust of preceeding vehicles that are brought upwards from the tailpipe through turbulence caused by the vehicle front, and by static pressure changes caused byacceleration, respectively. Attempts therefore have been made to put sample inlets in front ofthe vehicle at 2 m height where the turbluence from the vehicle front might be minimized or at3.5 m or above to sample above the height where vehicle exhaust contamination is likely to beless prevalent. While the designs for sample inlets in the sudies mentioned above are feasiblefor more permanent sensor installations, the goal of the mobile CO2 system developed for thisstudy was to maintain a lightweight and flexible system that could be easily installed and de-installed from the mobile platform on which it is traveling and therefore the sample inlets aresimply raised to 2 m height out of the passenger seat window. These considerations offer someexplanation for the large variability in the observed r for both the grouped and ungrouped inlettests in areas with many CO2 sources by suggesting that the inlets could be detecting CO2plumes unevenly as a result of the complex turbulence structures created while driving.934.2 Evaluating the Measurement CampaignMulti-Modal Measurement PlatformMany studies have measured r across transects through cities (Jimenez et al., 2000; Idso et al.,2001; Henninger and Kuttler, 2007; Crawford et al., 2011), however no study has deployedmultiple mobile CO2 sensors to monitor GHG emissions across a city until now. Studies such asthose by Tao et al. (2015) and Crawford et al. (2011) demonstrate mobile systems for monitoringGHGs such as CO2, but most of these systems are still bulky (e.g. 32 kg for the case of Taoet al. (2015)) and limited by their cost and installation needs. Therefore most urban transectstudies use sensors that are generally designed for one mode of transport (Bukowiecki et al.,2002; Crawford et al., 2011; Elen et al., 2013). While these uni-modal environmental monitoringsystems have the advantage that they can be well equipped with additional components suchas calibration tanks or computers, they do not allow for easy interfacing with other modesof transport; this however is quickly changing given the growing accessibility of opensourcehardware and source code. New research by Silvestri et al. (2014) shows the possibility ofmeasuring CO2 and other pollutants using quadcopter drones and illustrates another use casefor how open source microcontrollers and sensor components can be used to “miniaturize” andin many cases reduce the costs of bulky sensor systems and measure new characteristics ofcomplex environments. The results of the measurement campaign show the advantage of aflexible sensor system that can be quickly moved to and from a car to a bike (and vice versa)to measure in areas inaccessible by car (e.g. approximately 995 data points were collected inthe forest in Stanley Park which could not have otherwise been collected).Sampling DensityFor this study, a total of 5 mobile sensor systems were developed and deployed across a 12.7 km2study area over a period of 3.5 hours; the sampling density was 42.3 samples ha−1. In total, 9.7km2 of the 11.7 km2 that could be traversed in the 100 m grid were sampled, 91.31% of whichhad 10 samples or more and 69.24% of which had 20 samples or more. This means that excludingany data with less than 10 samples, the sampling density was roughly 0.5 km2 sensor−1 hr−1 or8.8 km2 over the 3.5 hour period for the 5 sensors using a 100 m grid. If it is assumed thatthis sampling density is appropriate for representing urban scale processes, it would require230 constantly moving mobile sensors to be deployed across the City of Vancouver (115 km2)to measure CO2 emissions across the city during the same time. Assuming that cost is nota factor or rather that the costs of GHG and pollution sensors become cheaper in the future,possibilities exist to integrate mobile sensor systems into vehicles such as taxis (e.g. 600 in theCity of Vancouver) and mobility-on-demand services (e.g. >1000 carshare vehicles in the Cityof Vancouver).As described in Section 2.3.7, a 5 km h−1 threshold on vehicle speed is used to filter themeasured r from the DIYSCO2 systems, resulting in a total of 41,027 data points. An exam-94ination of the filtered data shows that 5.6% of the raw data (2,465 data points) are less than5 km h−1 and greater than 1 km h−1. This indicates that we achieve a 1.01% increase in dataavailability per km h−1 between 5 km h−1 and 1 km h−1. If we were to keep all of the datawithout filtering for speeds below 1 km h−1, the total number of points would be 58,877, inwhich case 30.3% of the raw data would have been driven at speeds less than 5 km h−1. Byadjusting the filtering parameters, it is therefore possible to increase the data availability, butfuther investigation is needed to determine the optimal and appropriate filtering thresholds forthe specific context. Increasing the data availability would help to improve the spatial coverageof the measurements, but also help identify areas with greater emissions variability such as atroad intersections.Raw DataFigure 3.12 shows a 1-d representation of the 41,027 measurements taken during the measure-ment campaign. The graphs shows that the r is greatest along the major arterial roads anddowntown and lowest in Stanley Park, the Mountain View Cemetery, and some well vegetatedstreets in the West End. Furthermore, the graph shows that there tends to be more variabilityalong the major roads and in areas of commercial activity such as along the Broadway corridor,the Mount Pleasant neighborhood, and Downtown. Despite these peaks in the data, Figure 3.11shows that a majority of the r is less than 450 ppm (89.57%) and that 55% of the data is lessthan 410 ppm.Grid Averaged rThe raw data points were gridded to the 50 m, 100 m, 200 m, and 400 m vector grids. Thegrid size affected the study area mean by 3 ppm, but had a greater effect on the grid maximumvalues - the grid maximum was highest for the 50 m grid at 523.4 ppm compared to 500.8ppm, 473.2 ppm, and 460.2 ppm for the 100 m, 200 m, and 400 m grid cells. This is expectedbecause the more extreme measured r values would be averaged out by the larger grid cell sizes.Conversely, for the smaller grid cell sizes, the larger maximum r would be retained given thatthey would not be averaged out over a larger spatial area. Table 3.7 summarizes the summarystatistics for the gridded r for each of the grid cell sizes. A study by Crawford (2014) (seepg.86) showed similar ranges in r for a 1.9 km2 study area around Vancouver-Sunset tower aswas observed in our field campaign using a 200 m grid.954.3 Addressing Issues with the Aerodynamic ResistanceApproachThe Aerodynamic Resistance of Heat and Surface TemperatureIn the methods proposed in Section 2.4.2, raH is calculated using Ttower and T0 as opposed totemperature at screen level height (Tmobile) where the r measurements are occuring. Existingresearch shows that the surface temperatures can experience a greater range of values which canbe attributed to the effects of shading and absorption which is different from the air above itwhich is free to advect and mix (Bennett, 2005). An exploration of air temperatures measuredat 2 m height by the DIYSCO2 systems (Tmobile) that fell within 500 m of the Vancouver-Sunsettower at any time during the 3.5 hour measurement campaign showed an average air tempera-ture of 22.01 ◦C which was 8.7◦C less than T0. The results of this exploration showed an raH of2.16 s m−1 which is 16 times less than the raH that was calculated from the T0 (34.14 s m−1).What this indicates is that T0 is much warmer than Tmobile and therefore that the measuredemissions would be unrealistically higher if raH were calculated from Tmobile. Data from a reportby Christen (2013, unpublished) from the Basel Urban Boundary Layer Experiment (BUBLE)indicated typical raH values for various LCZs in the city of Basel, Switzerland using T0. Fortwo LCZ2 sites the raH was 32.6 and 43.4 s m−1. For a LCZ5/6 in the same city, the raH wasestimated to be 68.8 s m−1. These values align with the raH measured in this experiment inVancouver and suggests that the raH could be higher. Given that most of the measured emis-sions tended to overestimate the emissions inventories (e.g. 73% of measured emissions weregreater than the emissions inventory for the 100 m grid.), a higher raH may help to bring themeasured emissions and the emissions inventories into closer alignment, though this might alsomean that we might no longer detect where P is greater than C and R. A study by Adderleyet al. (2015) quantifies the error in measuring T0 with a radiometer at 28 m height during thedaytime which can be as high as 3.5 K. Further experiments should be done to determine howraH and consequently the resulting Fc change when using temperature measurements at screenlevel versus T0.Averaging the Aerodynamic Resistance of Heat Over TimeTable 3.8 shows the values from the 3.5 hour measurement campaign that were used to calculateraH . As described in Section 2.4.2, raH is calculated as the average of all the raH at 30 minintervals derived from the the average H, average T0, and average Ttower for the time duringthe field campaign. This resulted in an raH of 34.14 s m−1. However the correct methodof calculating raH would be to have taken the average of H, T0, and Ttower for the entiretime during the field campaign and then to calculate the raH from those averages. If raH isinstead calculated from this method, the result is 33.50 s m−1. While the difference here isnot substantial, these computational decisions can scale up or down the Fc, particularly if thedifferences between the minima and maxima for Cair, θ, andH are great during the measurement96campaign. Furthermore, an improved method of calculating Fc would therefore be to subset thethe measured r for each of the 30 min time periods and then calculate raH . We would thereforecalculate Fc over time time and then sum results. Given that over the measurement periodthe raH ranged from 28.92 s m−1 to 42.56 s m−1 (a difference of 13.64 s m−1), the strength ofemissions are highly dependent on the result of raH . By using raH to calculate Fc over time,Fc might be better represented.Evidence of Changing Aerodynamic Resistance Across the Study AreaThere is evidence of changing aerodynamic resistances across the study area. One featurethat is shown in Figure 3.12 is the low r measured on Cambie Bridge. While there are nomeasurements to confirm the exact values of the wind speed on the bridge, it is likely that,due to the elevated nature of the bridge (approximately 18 m), there are likely to be higherwind speeds and therefore more turbulent mixing. Consequently the aerodynamic resistanceon the bridge is much lower than the aerodynamic resistance assumed. Given that the trafficon Cambie Bridge is high, it is unlikely that the r should be low. Conversely, in the narrowstreet canyons of downtown and in Stanley Park, it is likely that the aerodynamic resistance ishigher, because of the sheltered nature of the deep canyons and forest canopy, respectively.4.4 Measuring Emissions with Mobile SensorsMeasured r and the Emissions InventoriesThe results comparing measured r and the emissions inventories are presented in Section 3.3.3.The data show that, in general, building and traffic emissions when binned into equal intervalsare strong predictors for the r measured in the city. This implies that r, from microscale toneighborhood scales, are strongly related the emissions being generated at those scales. Evenfurther, this suggests that it is possible to link r to emissions and therefore make estimates, albeitcrude ones, about emissions across a complex landscape under specific atmospheric conditions.Overall, the building emissions were less able to predict the variability in r. Two possibilitiesmay explain why we observe this. First, building emissions are most likely well blended in theprocess of downward mixing in the UCL. As a result of this blending, the emissions signalbased on r might show more spatial variability. This phenomenon is visualized on the maps inFigure 3.43 and Figure 3.44 where the measured emissions tend to underestimate the emissionsdowntown where there are a high density of tall buildings. Second, because the sensors aremostly measuring along roads, they are closer to traffic emissions sources compared to buildingemissions sources which occur off of the street. As such the highest r that is measured is lesslikely to be associated with the measured peaks in r which are more likely to be a result oftraffic emissions. Therefore, we might consider removing the higher r values (>460 ppm) fromthe analysis. If this is done, the building emissions are better able to predict the measured rsuch as seen in in Figure 4.1 in the case for the 50 m and 200 m grids, but not the 100 m grid.97llllllll0 20 40 60 80 1001012141618Binned Building Emissions (log of kg CO2 ha−1 hr−1) CO2 Mixing Ratio Enhancement (ppm) y = 0.096x + 10.8097R2 = 0.6(a) The R2 improves from 0.304 to 0.6 when thehigher r values are witheld from the comparison forthe 50 m grid.llllll0 10 20 30 408101214161820Binned Building Emissions (log of kg CO2 ha−1 hr−1) CO2 Mixing Ratio Enhancement (ppm) y = 0.2987x + 7.92R2 = 0.9583(b) The R2 degrades from 0.9867 to 0.9583 whenthe higher r values are witheld from the comparisonfor the 100 m grid.llllllll0 5 10 15 20 25 30 35101520Binned Building Emissions (log of kg CO2 ha−1 hr−1) CO2 Mixing Ratio Enhancement (ppm) y = 0.4987x + 7.1064R2 = 0.9276(c) The R2 improves from 0.7831 to 0.9276 whenthe higher r values are witheld from the comparisonfor the 200 m grid.Figure 4.1: Linear fit of the gridded median r measurements and the binned emissionsinventory for buildings for (a) the 50 m grid, (b) the 100 m grid, and (c) the 200m grid.98Measured Emissions and the Emissions InventoriesThis research demonstrates the potential to apply an aerodynamic resistance approach to mea-suring emissions using a network of mobile sensors and data from an urban climate tower. Inexamining the averaged measured emissions with the total emissions inventory for the studyarea subset as shown in Table 3.13, we see that the average measured emissions overestimatethe average total emissions inventory by 3.72, 10.26, 14.13, and 15.66 kg CO2 ha−1 hr−1 for the50 m, 100 m, 200 m, and 400 m grids, respectively.The increasing difference between the average measured emissions and the average totalemissions may be best explained by the bias towards roads in the sampling methodology. Thedifference between the average measured emissions and the total emissions inventory is relativelysmall for the 50 m grid because the measurements are made mostly along roads and thereforedo not include areas such as in the backyards of homes and within large street blocks. As aresult, when comparing the average measured emissions to the average of the total emissionsinventories for the 100 m, 200 m, and 400 m grids, we see that the sampling bias becomesmore apparent. This indicates that the 50 m grid cell size is a more appropriate resolution forgridding the point measurements collected using this methodology. Additional sampling alongalleys and laneways and more representative sampling using alternative mobility options suchas bikes or autonomous flying vehicles may help to improve the relationship between measuredemissions and the emissions inventory when gridding at lower resolutions.Another methodology to improve the grid averaging is to subsample the larger grid cellsusing a finer scale grid (e.g. 20 m× 20 m or less) and then averaging those finer grid cells to lowergrid resolutions as done in Crawford (2014). This would help to reduce some sampling biases attwo critical moments. First, it may be possible to average out some of the extreme values withina grid cell that may be contributing to an over- or under- estimation of emissions within a gridcell. Second, it offers a possibility to determine the representativeness of the grid cell sample.Because the current methodology simply spatially attributes any point(s) to the grid cell inwhich it intersects, we don’t account for situations in which point measurements misrepresentgrid cells by passing through the cell periphery or as a result of GPS error. By quantifying thesampling biases, we can therefore develop a better understanding of the relationship betweenthe point measurements and the measured emissions.However, despite these increasing differences between the average measured emissions andthe average total emissions inventory with grid size, the data for the 50 m, 100 m, and 200m grids, aligned relatively well with a study by Christen et al. (2011) that showed that forthe 1.9 km2 neighborhood around Vancouver-Sunset tower, CO2 emissions were modelled to be26.87 kg CO2 ha−1 hr−1 and validated using the EC tower data to be 25.96 kg CO2 ha−1 hr−1.Interestingly, the average total emissions inventory were within 2kg CO2 ha−1 hr−1 as thosevalues found in Christen et al. (2011) for the 100 m and 200 m grids in our study. On onehand this might suggest that in general the 1.9 km2 neighborhood around the Vancouver-Sunsettower is representative of the emissions behavior across the study area subset. On the other99hand, this might also suggest some inability of the total emissions inventory to capture thevariability of emissions that are occuring. Several factors may account for the higher rate ofmeasured emissions compared to that of the emissions inventory in addition to the samplingbias discussed above.First, the emissions inventories were not based on real-time models of the data for theperiod of the measurement campaign. The building emissions inventory presents a challengewhen comparing the grid averaged r and the measured emissions because the building emissionsinventory is downscaled to an hourly average from a yearly estimate. This hourly average isassumed to be constant over the course of the day, however, studies (e.g. Martani et al. (2012))show that most building occupancy (and therefore energy use) occurs between 9:00 and 19:00,with peaks around 13:00 and 16:00. Furthermore, this does not address the fact that spatially,building energy use changes throughout the day as people go to and from work and home.Future work might attempt to quantify the spatial ebb and flow of people using a combinationof surveys, census data, and methods using call detail records to derive home versus worklocations as shown in Holleczek et al. (2014). Building energy use intensity might be modeledby season and diurnally based on factors such as building occupancy, building age, form, andfunction. However it could be that the average measured emissions should be higher because,as mentioned earlier, building emissions–particularly those happening in the downtown wherebuildings tend to be taller–occur higher up in the UCL and therefore might just be more blendedout. As a result, building emissions and therefore the average measured emissions may actuallybe higher than what is currently reported.To explain the traffic emissions difference, we must account for the fact that the trafficemissions inventory was derived from spatially and temporally disaggregated samples of trafficcounts. As a result, the traffic emissions inventory may compound errors over time and space.Spatially, the traffic count dataset covers mostly the major roads which leaves much of theresidential areas unsampled. The method described in Section 2.4 is used to map traffic countvalues across the residential streets to overcome the missing traffic counts, however validationis necessary to determine whether this method is appropriate. Temporally, the traffic emissionsinventory is not a real-time representation of the traffic counts during the measurement cam-paign and therefore may underrepresent traffic emissions of some streets while overrepresentingthe emissions of others. Furthermore, the traffic emissions are generated using an emissionsfactor that is a fleet average for the emitted CO2 per liter of fuel burned. More precise esti-mates of emissions factor in the differences in the emissions factor by vehicle type and fuel type(Kellett et al., 2013). Last, the traffic count data does not indicate the amount of emissionsfrom idling that occur as a result of traffic jams and thus introduces another aspect of possibleuncertainty within the traffic emissions inventory.Second, the emissions inventory factors only building and traffic emissions and excludesother sources of emissions such as those from human, animal, and plant and soil respiration.Additional sources of CO2 emissions could come from human activities such as landscaping (e.g.100lawnmowers and leafblowers) and construction. For example, study by Kellett et al. (2013)showed that, in a 1.9 km × 1.9 km study area around Vancouver-Sunset tower, emissions fromhuman respiration and vegetation and soils can account for 8% and 5% respectively of the totalemissions, respectively.Third, the measured emissions could possibly be higher in streets with a denser canopyregardless if the canopy is vegetation or buildings. An area with a dense urban canopy mayactually reduce mixing (Jin et al., 2014) and as a result, the measured r might be higher thanwould be expected for a well vegetated street given the presence of a fleeting CO2 source such asa CO2 plume from a passing car. It would therefore be beneficial to consider higher aerodynamicresistances and to use models that relate canopy porosity to create maps of variability in raH .Given that the study area subset (which is the area that is used to compare the measuredemissions and emissions inventory) includes more diverse land use and cover types, particularlythose with more emissions sources, it is reasonable that the measured emissions would be ashigh as 39.69 kg CO2 ha−1 hr−1. This is confirmed by the close alignment of the averagesbetween the measured emissions and the total emissions inventory at the 50 m grid size.Evidence of photosynthetic uptake in well vegetated urban forest.Moderate emissions from residential neighborhoodsEmission peaks fromarterial shipping road.High emissions from densetraffic, building energy use,and population in the downtown core.Figure 4.2: Perspective map of the emissions per 100 m grid. The map is annotated withlabels showing the areas of high emissions and also where uptake is occuring.Last, we found that in addition to measuring positive net emissions, we also observed areaswhere the mobile sensors also detected negative net ecosystem exchange (NEE), such as in theforest at Stanley Park and the lawn area at Mountain View Cemetery. It is important to notethat most grid cells have some uptake, but in many cells the emissions from C and R combinedare greater than the P . While the mobile sensors cannot detect uptake per say, the resultssuggest that the methodology under specific atmospheric and environmental conditions (e.g.low winds and no emissions sources and daytime, summer conditions) can determine where101P is greater than the C and R. In comparing our measured emissions from Stanley Park (-12 kg CO2 ha−1 hr−1) to a study by Humphreys et al. (2006) who measured NEE for a forestwith similar stand composition (Douglas Fir forest) in April and June in the same latitude, weactually find that the measured emissions were within a factor of 2 of those observed in thestudy. We might best explain these differences given that the sensor was not fixed in place andmoving rapidly throughout the understory of the forest at Stanley Park and that the forestcanopy may be increasing the aerodynamic resistance and thus the calculated emissions.Ultimately, the evaluation of the measured emissions and the emissions inventories showedwhere and why there might be close alignment or divergences between the datasets and suggestspromising new research opportunities for improving the methodology.102Chapter 5Conclusion5.1 SummaryThis research has shown the successful development of novel materials and methods for monitor-ing and mapping CO2 mixing ratios and emissions in complex environments. First, a compactand mobile CO2 sensor system called the “DIYSCO2” was built based on both open and closedsource hardware and software. The system was tested to show a precision of 0.233 ppm, a timedependent sensor accuracy of 3 ppm (over 7 days), and response time of 18.2 seconds with a3 m inlet tube length running at flow rate of 700 cc min−1. The DIYSCO2 development andtesting demonstrated that open source components can aid in the development of compact andhigh performance urban sensor systems.Second, a method to map emissions based on the aerodynamic resistance approach wasdeveloped and tested. While the method is sensitive to the measurements that are used toderive the aerodynamic resistance of heat and requires that a number of assumptions andconditions are met, this research shows that the aerodynamic resistance approach can be usedto derive emissions at a fine scale from measures of mixing ratios. To test this, 41,000 datapoints from a 3.5 hour measurement campaign were produced by 5 DIYSCO2 systems andaggregated, processed, and analyzed at varying spatial resolutions grid resolutions. The resultsof the analysis on the 100 m × 100 m grid showed that grid averaged r ranged from 382 ppm to518 ppm and average 417.1 ppm across the study area. Comparisons of the gridded median rand the equally binned total emissions inventories showed strong linearity (R2 >0.89) for the 50m, 100 m, 200 m, and 400 m grids which suggests that building and traffic emissions account formore than 89% of the variability in the measured r and therefore the emissions. The measuredemissions across the study area ranged from -12 kg CO2 ha−1 hr−1 to 225 kg CO2 ha−1 hr−1,thus showing the possibility for this methodology to detect negative emissions, or where Pis greater than C and R. The data showed the measured emissions and the total emissionsinventory were positively correlated by 78.43%, with 99.43% of the measured emissions within± 1 order of magnitude of the emissions inventory. The average of the measured emissions103aligned the closest to the emissions inventory when observing the 50 m grid resolution and isbest explained by the sampling bias towards roads. Despite the simplicity of the methodology,the research demonstrated that it is possible to measure emissions across a complex landscapewith a fleet of mobile sensors, an urban climate tower, and the use of the aerodynamic approachto calculating emissions.5.2 Practical SignificanceThe research presented in this thesis is proof of concept for a future in which environmentalsensing is integrated into urban mobility. The work demonstrates the possiblity of detectingand mapping CO2 emissions, but the concept can and should be translated to the mapping ofother trace gases (e.g. methane), pollutants and particulates, air and surface temperature, andother environmental variables that affect human health, comfort, and safety. Climate changewill be one of the greatest challenges facing humanity now and into the future and will only becompounded by the health and safety issues caused air pollution and extreme weather events.By taking advantage of mobile platforms for mapping and monitoring the environment, newpathways can be formed around the ways in which we sense, respond, and adapt to changes inthe places we live and inform the public about the impact of GHG emissions. It is only throughenhanced efforts in science communication can we begin to address the multi-faceted challengesof urban emissions and build stronger and more diverse collaborations with practitioners andthe public. More work must be done to enhance the public understanding of urban emissions bytaking advantange of the scale of outreach and communication that is facilitated by the internetand the scope of possibilities to make large scale changes by working with local communities inthe research efforts. Given the current exclusivity and inaccessibility of scientific research donein this domain (and in science in general), we must work to facilitate transparency and openand reproducible research such that these findings and technology can be further developed,tested, and explored in other cities, particularly in places where data is scarce and access toexpensive sensor systems is limited (e.g. in developing countries).This work furthermore illustrates how the development of smaller, more affordable, andhackable sensor systems can facilitate new methodological approaches to monitoring CO2 (andthe environment more generally). With a fleet of mobile sensors and the methodologies forprocessing the derived datasets, the possiblity to map and consequently validate emissionsinventories is possible. By developing these technologies, methodologies, and scenarios aroundemissions mapping in cities, this research aims to better inform policy makers, urban designers,and the public about the impact of urban form and function on the emissions across the cityand how mitigation stategies can be achieved from the top-down (policy and design) and thebottom up (citizen engagement).1045.3 Future DirectionsConsiderations for future research:1. Sensor Design and Functionality : The sensor design could be improved by: 1.) Reducingthe length of the sample inlet tube to reduce the sensor response lag time, 2.) Adding in-built calibration using a chemical scrubber of Ascarite II and Magnesium Perchlorate, 3.)Adding wireless communication for real-time data transfer and visualization via the weband sensor control, 4.) Roof-based, magnetized vehicle installation to maximize flexibilityand minimize intrusiveness.2. Aerodynamic Resistance Approach: 1.) Develop methods for mapping aerodynamic re-sistances of heat across a complex environment through modeling or measurements, 2.)Evaluate alternative methods of calculating raC from the aerodynamic resistance of mo-mentum.3. Analysis Capabilities: 1.) The code used for the analysis could be packaged into a web-based application to create a graphical user interface (GUI) to not only offer an open andaccessible platform for processing data collected from the mobile sensor system, but alsoallow for a simple way to tune the parameters of the analysis such as the grid size andfiltering variables such as sample count. The platform should be a tool for both scientistsand planners to gain insights from the data.4. Measurement Campaigns: 1.) Compare the summer measurement campaign data to thosemeasured in winter conditions to assess impact of winter time fossil fuel combustion forspace heating.5. Public Communication and Outreach: The community outreach potential for involvingthe public in this work is an important factor to address. These might be addressedby the following: 1.) Develop maps and interactive visualizations for members of theinformed public to help guide discussions around emissions reduction strategies and pol-lution mitigation. 2.) Develop community based sensing. Ideas such as “adopt-a-sensor”or “CO2-data-collection-tours” could be a way to engage the public in science as well assupport the research efforts.The materials, code, and data used in this thesis can be found at: https://github.com/ubc-micromet/DIYSCO2. Contributions and feedback are welcome and appreciated.105BibliographyAclima. Mapping how our cities live and breath, 2015. URL http://insights.aclima.io/. → pages 9Adderley, C., Christen, A., and Voogt, J. A. The effect of radiometer placement and view oninferred directional and hemispheric radiometric temperatures of an urban canopy.Atmospheric Measurement Techniques, 8(7):2699–2714, 2015. → pages 96B.C. Ministry of Environment. 2014 B.C. Best Practices Methodology for QuantifyingGreenhouse Gas Emissions. Technical report, Nov. 2014. → pages 33Bennett, A. Heat Fluxes from Street Canyons. Master’s thesis, University of Reading,Department of Meteorology, Mar. 2005. → pages 96Brantley, H. L., Hagler, G. S. W., Kimbrough, E. S., Williams, R. W., Mukerjee, S., and Neas,L. M. Mobile air monitoring data-processing strategies and effects on spatial air pollutiontrends. Atmospheric Measurement Techniques, 7(7):2169–2183, 2014. → pages 15Buchwitz, M., Burrows, J. P., Bovensmann, H., Warneke, T., Notholt, J., Meirink, J. F.,Goede, A. P. H., Bergamaschi, P., Korner, S., Heimann, M., and Schulz, A. Atmosphericmethane and carbon dioxide from SCIAMACHY satellite data: initial comparison withchemistry and transport models. Atmospheric Chemistry and Physics, pages 1–22, Mar.2005. → pages 8Bukowiecki, N., Dommen, J., Pre´voˆt, A. S. H., Richter, R., Weingartner, E., andBaltensperger, U. A mobile pollutant measurement laboratory—measuring gas phase andaerosol ambient concentrations with high spatial and temporal resolution. AtmosphericEnvironment, pages 1–11, Oct. 2002. → pages 92, 93, 94Bush, S., Pataki, D., and Ehleringer, J. Sources of variation in 13c of fossil fuel emissions insalt lake city, USA. Applied Geochemistry, 22(4):715 – 723, 2007. → pages 8Castell, N., Kobernus, M., Liu, H.-Y., Schneider, P., Lahoz, W., Berre, A. J., and Noll, J.Mobile technologies and services for environmental monitoring: The Citi-Sense-MOBapproach. Urban Climate, pages 1–13, Aug. 2015. → pages 8Chapman, L., Young, D., Muller, C. L., Rose, P., Lucas, C., and Walden, J. Winter RoadMaintenance and the Internet of Things. In International Road Weather Conference, pages1–8, Sept. 2015. → pages 9Christen, A. Atmospheric measurement techniques to quantify greenhouse gas emissions fromcities. Urban Climate, 10, Part 2:241 – 260, 2014. ICUC8: The 8th International Conferenceon Urban Climate and the 10th Symposium on the Urban Environment. → pages 2, 5, 7, 8106Christen, A., Coops, N. C., Crawford, B. R., Kellett, R., Liss, K. N., Olchovski, I., Tooke,T. R., Laan, M., van der, and Voogt, J. A. Validation of modeled carbon-dioxide emissionsfrom an urban neighborhood with direct eddy-covariance measurements. AtmosphericEnvironment, 45(33):6057–6069, 2011. → pages 4, 35, 99City of Vancouver. Open data catalogue - directional volume traffic counts data, 2015. URLhttp://data.vancouver.ca/datacatalogue/trafficCounts.htm. → pages 28, 33Crawford, B. Measurements of carbon dioxide fluxes and concentrations at multiple scales inVancouver, Canada. PhD thesis, Aug. 2014. → pages 6, 95, 99Crawford, B., Christen, A., and McKendry, I. Diurnal Course of Carbon Dioxide MixingRatios in the Urban Boundary Layer in Response to Surface Emissions. Journal of AppliedMeteorology and Climatology, 55(3):507–529, Mar. 2016. → pages 13Crawford, B. R. and Christen, A. Spatial source attribution of measured urban eddycovariance CO2 fluxes. Theoretical Applied Climatology, 119:733–755, Apr. 2014. → pages6, 7, 13, 16, 32Crawford, B. R., Grimmond, C. S. B., and Christen, A. Five years of carbon dioxide fluxesmeasurements in a highly vegetated suburban area. Atmospheric Environment, 45(4):896–905, Feb. 2011. → pages 93, 94Elen, B., Peters, J., Poppel, M., Bleux, N., Theunis, J., Reggente, M., and Standaert, A. TheAeroflex: A Bicycle for Mobile Air Quality Measurements. Sensors, 13(1):221–240, Jan.2013. → pages 94Feigenwinter, C., Vogt, R., and Christen, A. Eddy Covariance Measurements over UrbanAreas. Springer, Jan. 2012. → pages 7Fluxnet. Integrating worldwide CO2, water, and energy flux measurements, 2016. URLhttp://fluxnet.ornl.gov/site/4132. → pages 25George, K., Ziska, L., Bunce, J., and Quebedeaux, B. Elevated atmospheric CO2concentration and temperature across an urbanrural transect. Atmospheric Environment,41(35):7654 – 7665, 2007. → pages 8Grimmond, C. S. B. Urbanization and global environmental change: local effects of urbanwarming. The Geographical Journal, 173(1):83–88, Mar. 2007. → pages 1Gurney, K. R., Mendoza, D. L., Zhou, Y., Fischer, M. L., Miller, C. C., Geethakumar, S., andCan, S., de la Rue du. High Resolution Fossil Fuel Combustion CO2 Emission Fluxes forthe United States. Environmental Science & Technology, 43(14):5535–5541, July 2009. →pages 3Gurney, K. R., Razlivanov, I., Song, Y., Zhou, Y., Benes, B., and Abdul-Massih, M.Quantification of Fossil Fuel CO2 Emissions on the Building/Street Scale for a Large U.S.City. Environmental Science & Technology, 46(21):12194–12202, Nov. 2012. → pages 3Ha¨b, K., Ruddell, B. L., and Middel, A. Sensor lag correction for mobile urban microclimatemeasurements. Urban Climate, 14(Part 4):622–635, Dec. 2015. → pages 93107Heiple, S. and Sailor, D. J. Using building energy simulation and geospatial modelingtechniques to determine high resolution building sector energy consumption profiles. Energyand Buildings, 40(8):1426 – 1436, 2008. → pages 3Henninger, S. and Kuttler, W. Methodology for mobile measurements of carbon dioxidewithin the urban canopy layer. Climate Research, 34:161–167, July 2007. → pages 6, 94Holleczek, T., Yu, L., Lee, J. K., Senn, O., Ratti, C., and Jaillet, P. Detecting weak publictransport connections from cellphone and public transport data. In Proceedings of the 2014International Conference on Big Data Science and Computing, pages 1–8, New York, NewYork, USA, 2014. ACM Press. → pages 100Hu¨bner, J., Olesch, J., Falke, H., Meixner, F. X., and Foken, T. A horizontal mobilemeasuring system for atmospheric quantities. Atmospheric Measurement Techniques, 7(9):2967–2980, 2014. → pages 93Humphreys, E. R., Black, T. A., Morgenstern, K., Cai, T., Drewitt, G. B., Nesic, Z., andTrofymow, J. A. Carbon dioxide fluxes in coastal Douglas-fir stands at different stages ofdevelopment after clearcut harvesting. Agricultural and Forest Meteorology, 140(1-4):6–22,Nov. 2006. → pages 102Idso, C., Idso, S., and Balling, R. An intensive two-week study of an urban CO2 dome inPhoenix, Arizona, USA . Atmospheric Environment, 35:995–1000, Jan. 2001. → pages 3, 6,8, 94IPCC. Climate Change Mitigation: Contribution of Working Group III to the FourthAssessment Report of the Intergovernmental Panel on Climate Change. CambridgeUniversity Press, Nov. 2007. → pages 1IPCC. Climate Change Impacts, Adaptation, and Vulnerability: Contribution of WorkingGroup IV to the Fifth Assessment Report of the Intergovernmental Panel on ClimateChange. Cambridge University Press, Nov. 2014. → pages 3Jimenez, J. L., McManus, J. B., Shorter, J. H., Nelson, D. D., Zahniser, M. S., Koplow, M.,McRae, G. J., and Kolb, C. E. Cross road and mobile tunable infrared laser measurementsof nitrous oxide emissions from motor vehicles. Chemosphere - Global Change Science,pages 1–16, Dec. 2000. → pages 94Jin, S., Guo, J., Wheeler, S., Kan, L., and Che, S. Evaluation of impacts of trees on PM2.5dispersion in urban streets. Atmospheric Environment, 99(C):277–287, Dec. 2014. → pages101Kanda, M., Kanega, M., Kawai, T., Moriwaki, R., and Sugawara, H. Roughness lengths formomentum and heat derived from outdoor urban scale models. Journal of AppliedMeteorology and Climatology, 46(7):1067–1079, July 2007. → pages 13Kandias, M. R. A. A. G. Temporal and spatial patterns of near-surface CO2 concentrations inSingapore. Master’s thesis, National University of Singapore, June 2008. → pages 7Kaye, J., Grofmann, P., Grimm, N., Baker, L., and Pouyat, R. A distinct urbanbiogeochemistry? Trends in Ecology & Evolution, 21(4):192–199, Apr. 2006. → pages 1108Kellett, R., Christen, A., Coops, N. C., Laan, M., van der, Crawford, B., Tooke, T. R., andOlchovski, I. A systems approach to carbon cycling and emissions modeling at an urbanneighborhood scale. Landscape and Urban Planning, 110:48–58, Feb. 2013. → pages 3, 4,100, 101Kennedy, C., Steinberger, J., Gasson, B., Hansen, Y., Hillman, T., Havra´nek, M., Pataki, D.,Phdungsilp, A., Ramaswami, A., and Mendez, G. V. Greenhouse Gas Emissions fromGlobal Cities. Environmental Science & Technology, 43(19):7297–7302, Oct. 2009. → pages1Kort, E. A., Frankenberg, C., Miller, C. E., and Oda, T. Space-based observations of megacitycarbon dioxide. Geophysical Research Letters, 39(17):n/a–n/a, Sept. 2012. → pages 8Li-Cor. New procedures for filling internal chemical scrubber bottles in the LI-7500, LI-7000,and LI06400xt, 2008. URL http://www.licor.com/env/newsline/2008/05/new-procedures-for-filling-internal-chemical-scrubber-bottles-in-the-li-7500-li-7000-and-li-6400xt/.→ pages 92Li-Cor. LI-820 CO2 Analyzer Instruction Manual, June 2009. → pages 19Li-Cor. Advantages of the LI-820, 2015. URLhttps://www.licor.com/env/products/gas analysis/LI-820/. → pages 16Lietzke, B. and Vogt, R. Variability of CO2 concentrations and fluxes in and above an urbanstreet canyon. Atmospheric Environment, 74:60 – 72, 2013. → pages 6Liu, H.-Y., Skjetne, E., and Kobernus, M. Mobile phone tracking: in support of modellingtraffic-related air pollution contribution to individual exposure and its implications forpublic health impact assessment. Environmental Health, pages 1–12, Nov. 2013. → pages 8Marland, G. Uncertainties in accounting for CO2 from fossil fuels. Journal of IndustrialEcology, 12(2):136–139, 2008. → pages 4Martani, C., Lee, D., Robinson, P., Britter, R., and Ratti, C. ENERNET: Studying thedynamic relationship between building occupancy and energy consumption. Energy &Buildings, 47:584–591, Apr. 2012. → pages 100McDermitt, D. K., Welles, J. M., and Eckles, R. D. Effects of Temperature, Pressure andWater Vapor on Gas Phase Infrared Absorption by CO2. pages 1–5, Mar. 1995. → pages 20Mead, M. I., Popoola, O. A. M., Stewart, G. B., Landshoff, P., Calleja, M., Hayes, M.,Baldovi, J. J., McLeod, M. W., Hodgson, T. F., Dicks, J., Lewis, A., Cohen, J., Baron, R.,Saffell, J. R., and Jones, R. L. The use of electrochemical sensors for monitoring urban airquality in low-cost, high-density networks. Atmospheric Environment, 70(C):186–203, May2013. → pages 8Meier, F., Fenner, D., Grassmann, T., Janicke, B., Otto, M., and Scherer, D. Challenges andbenefits from crowdsourced atmospheric data for urban climate research using berlin,germany, as testbed. International Conference on Urban Climates, 2015. → pages 9Moriwaki, R., Kanda, M., and Nitta, H. Carbon dioxide build-up within a suburban canopylayer in winter night. Atmospheric Environment, 40(8):1394–1407, 2006. → pages 6109Natural Resources Canada. 2014 Fuel Consumption Guide. pages 1–11, Jan. 2014. → pages 33Overeem, A., Leijnse, H., and Uijlenhoet, R. Measuring urban rainfall using microwave linksfrom commercial cellular communication networks. Water Resources Research, 47(12):n/a–n/a, Dec. 2011. → pages 9Overeem, A., R Robinson, J. C., Leijnse, H., Steeneveld, G. J., P Horn, B. K., and Uijlenhoet,R. Crowdsourcing urban air temperatures from smartphone battery temperatures.Geophysical Research Letters, 40(15):4081–4085, Aug. 2013. → pages 9Parshall, L., Gurney, K., Hammer, S. A., Mendoza, D., Zhou, Y., and Geethakumar, S.Modeling energy consumption and CO2 emissions at the urban scale: Methodologicalchallenges and insights from the United States. Energy Policy, 38(9):4765–4782, May 2010.→ pages 3Pataki, D., Emmi, P., Forster, C., Mills, J., Pardyjak, E., Peterson, T., Thompson, J., andDudley-Murphy, E. An integrated approach to improving fossil fuel emissions scenarioswith urban ecosystem studies. Ecological Complexity, 6(1):1 – 14, 2009. → pages 3Phillips, N. G., Ackley, R., Crosson, E. R., Down, A., Hutyra, L. R., Brondfield, M., Karr,J. D., Zhao, K., and Jackson, R. B. Environmental Pollution. Environmental Pollution, 173(C):1–4, Feb. 2013. → pages 6Rigby, M., Toumi, R., Fisher, R., Lowry, D., and Nisbet, E. G. First continuousmeasurements of CO2 mixing ratio in central London using a compact diffusion probe.Atmospheric Environment, 42(39):8943–8953, Dec. 2008. → pages 8Rosenzweig, C., Solecki, W., and Hammer, S. A. Cities lead the way inclimate-change action.Nature, pages 1–3, Oct. 2010. → pages 1Salmond, J. A., Oke, T. R., Grimmond, C. S. B., Roberts, S. M., and Offerle, B. D. Ventingof heat and carbon dioxide from urban canyons at night. Journal of Applied Meteorology,44:1180–1193, 2005. → pages 7Satterthwaite, D. Cities’ contribution to global warming: notes on the allocation of greenhousegas emissions. Environment and Urbanization, 20(2):539–549, Oct. 2008. → pages 1Silvestri, M., Diaz, J. A., Marotta, E., Musacchio, M., Buongiorno, M. F., Sansivero, F.,Cardellini, C., Pieri, D., Amici, S., Bagnato, E., Beddini, G., Belviso, P., Carandente, A.,Colini, L., Doumaz, F., Peluso, R., and Spinetti, C. Use of multiple in situ and remotesensing instruments and techniques at solfatara field campaign for measurements of CO2,H2S, and SO2 emissions. Quaderni di Geofisica, pages 1–31, Oct. 2014. → pages 94Stern, N. Stern Review: The Economics of Climate Change. Oct. 2006. → pages 1Stewart, I. D. Redefining the urban heat island. PhD thesis, University of British Columbia,Oct 2011. → pages 24, 25Stewart, I. D. and Oke, T. R. Local Climate Zones for Urban Temperature Studies. Bulletinof the American Meteorological Society, 93(12):1879–1900, Dec. 2012. → pages 23, 24, 25Tao, L., Sun, K., Miller, D. J., Pan, D., Golston, L. M., and Zondlo, M. A. Low-power,open-path mobile sensing platform for high-resolution measurements of greenhouse gasesand air pollutants. Applied Physics B, 119(1):153–164, Mar. 2015. → pages 93, 94110United Nations. Concise Report on the World Population Situation 2014. pages 1–38, July2014. → pages 1Laan, M., van der. Scaling Urban Energy Use and Greenhouse Gas Emissions throughLiDAR. Master’s thesis, The University of British Columbia, Dec. 2011. → pages 24, 34, 65Velasco, E. and Roth, M. Cities as Net Sources of CO2: Review of Atmospheric CO2Exchange in Urban Environments Measured by Eddy Covariance Technique. GeographyCompass, 4(9):1238–1259, Aug. 2010. → pages 7, 8Vogt, R., Christen, A., Rotach, M. W., Roth, M., and Satyanarayana, A. N. V. Temporaldynamics of CO2 fluxes and profiles over a central European city. Theoretical AppliedClimatology, 84:117–126, Jan. 2006. → pages 6111

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0300248/manifest

Comment

Related Items