UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Building energy modelling and mapping using airborne LiDAR Tooke, Thoreau Rory 2014

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

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

Item Metadata


24-ubc_2014_spring_tooke_thoreau.pdf [ 4.95MB ]
JSON: 24-1.0165890.json
JSON-LD: 24-1.0165890-ld.json
RDF/XML (Pretty): 24-1.0165890-rdf.xml
RDF/JSON: 24-1.0165890-rdf.json
Turtle: 24-1.0165890-turtle.txt
N-Triples: 24-1.0165890-rdf-ntriples.txt
Original Record: 24-1.0165890-source.json
Full Text

Full Text

   Building Energy Modelling and Mapping using Airborne LiDAR   by  Thoreau Rory Tooke  BA, University of British Columbia, 2007 MSc, University of British Columbia, 2009     A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  DOCTOR OF PHILOSOPHY   in   The Faculty of Graduate and Postdoctoral Studies  (Forestry)       THE UNIVERSITY OF BRITISH COLUMBIA  (Vancouver)     February 2014     ? Thoreau Rory Tooke, 2014   ii Abstract  Globally, buildings are responsible for more than 40% of energy demand and contribute more than 30% of CO2 emissions.  Various strategies and policies have been developed to reduce the negative of effects of energy use in the building sector, specifically targeting energy conservation and energy supply from renewable resources.  As a basis for these strategies, decision-makers require estimates of existing energy demand.  Traditionally, broad building sector energy estimates are derived using top-down modelling approaches that establish relations between energy use and variables such as income, fuel prices and gross domestic product.  In contrast, individual building energy modelling has evolved sophisticated physically based simulations, populated by an abundance of variables related to building construction materials and components.  However, for governments and decision-makers tasked with developing local strategies, techniques are needed to provide a detailed itemization of the building and environmental attributes that impact energy demand, as offered in building simulations, while maintaining the scalability to large areas provided in top-down models.  Advances to geospatial technologies and datasets offer novel opportunities to satisfy these two conditions.  Of particular interest is light detection and ranging (LiDAR), since it provides spatially contiguous measurements of urban form, otherwise unattainable across large areas.  This dissertation presents a novel approach that integrates LiDAR data with building energy models to provide detailed and spatially contiguous estimates of energy demand in the residential building sector.  LiDAR is used to augment building energy models by relating measured building form to internal energy components including envelope resistivity, fenestration and air leakage, and by assessing building envelope solar gains after accounting for local occlusions.  Outcomes demonstrate  iii that a LiDAR-based approach to building energy assessment is able to produce results that closely match those from manually informed building simulation software, thus offering a time and cost effective option for extensive and detailed analysis of energy demand.  By presenting methods to decompose building energy demand into the site-specific components that influence energy end-use, this dissertation offers innovative opportunities to analyze and design spatially targeted building energy policies and strategies.     iv Preface  In this dissertation I established the research objective and questions, conducted the analysis and wrote the papers and chapters. In Chapter 4, co-authors Gurtuna and Prevot processed GOES imagery to generate the cloud index values that were included for further analysis and verification.  Co-author Christen provided post-processed measurements of environmental variables relating to temperature and radiation. All co-authors also provided mentorship and editorial comments as required.  The Environmental Prediction in Canadian Cities project funded by the Canadian Foundation for Climate and Atmospheric Sciences procured and provided the airborne LiDAR data.  A version of Chapter 3 has been published in: ? Tooke, T.R., Coops, N.C., Webster, J. (2014). Predicting building ages from LiDAR data with random forests for building energy modelling. Energy and Buildings 68, 603-610.  A version of the first section of Chapter 4 has been published in: ? Tooke, T.R., Coops, N.C., Christen, A., Gurtuna, O., Prevot, A. (2012).  Integrated irradiance modeling in the urban environment based on remotely sensed data.  Solar Energy 86, 2923-2934.  A version of the second section of Chapter 4 has been published in: ? Tooke, T.R., Coops, N.C., Christen, A. (2013). A point obstruction stacking (POSt) approach to wall irradiance modeling across urban environments. Building and Environment 60, 234-242.  A version of Chapter 5 has been submitted for publication as: ? Tooke, T.R., vanderLaan, M., Coops, N.C. Mapping demand for building thermal energy services using airborne LiDAR.   v Table of Contents  Abstract ..................................................................................................................................... ii Preface ...................................................................................................................................... iv Table of Contents ...................................................................................................................... v List of Tables ......................................................................................................................... viii List of Figures .......................................................................................................................... ix Glossary of Acronyms ............................................................................................................ xii Acknowledgements ................................................................................................................ xiii Dedication .............................................................................................................................. xiv 1 Introduction ........................................................................................................................ 1 1.1 General background, objectives and chapter overview................................................ 1 1.2 Energy and emissions policy in British Columbia, Canada ......................................... 5 1.3 Fundamentals of LiDAR-based energy modelling ...................................................... 7 1.3.1 An overview of light detection and ranging (LiDAR) ........................................... 8 1.3.2 Building energy demand modelling ..................................................................... 12 1.3.3 Solar radiation in the built environment .............................................................. 15 2 Study Area and Data ......................................................................................................... 22 2.1 Light detection and ranging (LiDAR) data ................................................................ 24 2.2 Shuttle Radar Topographic Mission (SRTM) data .................................................... 28 2.3 Geostationary Operational Environmental Satellite (GOES) .................................... 28 2.4 Statistics Canada census data ..................................................................................... 29 2.5 British Columbia Assessment data ............................................................................ 29 2.6 Residential building energy audit data ....................................................................... 30 2.7 Cadaster, zoning and building footprint data ............................................................. 31 2.8 Ground instrumentation ............................................................................................. 32 3 Prediction of Building Ages using LiDAR....................................................................... 33 3.1 Introduction ................................................................................................................ 33 3.2 Method ....................................................................................................................... 36 3.2.1 Study area ............................................................................................................. 36 3.2.2 Data ...................................................................................................................... 37 3.2.3 Model predictor variables .................................................................................... 38 3.2.4 Random forests .................................................................................................... 45 3.3 Results ........................................................................................................................ 47 3.4 Implications for building energy models and simulations ......................................... 51 4 Modelling Building Envelope Irradiance ......................................................................... 53 4.1 Introduction ................................................................................................................ 53 4.2 Section 1: Integrating remote sensing data to estimate irradiance on building horizontal surfaces ............................................................................................................... 54 4.3 Study area and data .................................................................................................... 54  vi 4.4 Method ....................................................................................................................... 57 4.4.1 Atmospheric transmission .................................................................................... 58 4.4.2 Surface viewshed calculation ............................................................................... 60 4.4.3 Vegetation transmission ....................................................................................... 61 4.4.4 Urban irradiance estimation ................................................................................. 65 4.4.5 Case study ............................................................................................................ 68 4.5 Results ........................................................................................................................ 69 4.6 Discussion .................................................................................................................. 75 4.7 Section 2. An irradiance model for building vertical surfaces ................................... 78 4.8 Data ............................................................................................................................ 78 4.9 Methods ...................................................................................................................... 79 4.9.1 POSt topology ...................................................................................................... 79 4.9.2 Wall attributes ...................................................................................................... 81 4.9.3 Point spacing configurations ................................................................................ 82 4.9.4 Ray casting procedure .......................................................................................... 83 4.9.5 Irradiance modeling ............................................................................................. 85 4.9.6 Validation ............................................................................................................. 88 4.10 Results ...................................................................................................................... 90 4.11 Discussion ................................................................................................................ 94 4.12 Implications for building energy models and simulations ....................................... 97 5 Modelling and Mapping Demand for Building Thermal Energy Services ...................... 98 5.1 Introduction ................................................................................................................ 98 5.2 Methods ...................................................................................................................... 99 5.2.1 Data ...................................................................................................................... 99 5.2.2 Study area ............................................................................................................. 99 5.2.3 Modeling approach ............................................................................................ 100 5.2.4 Validation ........................................................................................................... 112 5.3 Results ...................................................................................................................... 113 5.4 Discussion ................................................................................................................ 118 5.4.1 Limitations and future work ............................................................................... 119 5.4.2 Policy applications ............................................................................................. 121 6 Local Government Building Energy Policy ................................................................... 122 6.1 Introduction .............................................................................................................. 122 6.2 Background .............................................................................................................. 124 6.2.1 The role and responsibilities of local governments ........................................... 124 6.2.2 Building energy services and the local environment ......................................... 125 6.3 Local energy policy instruments .............................................................................. 126 6.3.1 Voluntary policies .............................................................................................. 127 6.3.2 Fiscal tools ......................................................................................................... 134 6.3.3 Regulatory policies ............................................................................................ 137 6.3.4 Capacity building ............................................................................................... 142 6.4 Informing building energy policy from spatial estimates of building energy performance ....................................................................................................................... 146 6.4.1 Information campaigns and behavioural interventions ...................................... 147 6.4.2 Developer cost charges ...................................................................................... 148 6.4.3 Land use planning and rezoning ........................................................................ 149  vii 6.4.4 Local energy utilities, technology siting, and service area bylaws .................... 149 6.5 Conclusion ............................................................................................................... 151 7 Conclusion ...................................................................................................................... 153 7.1 Innovation ................................................................................................................ 158 7.2 Applications of LiDAR ............................................................................................ 158 7.2.1 Local planning.................................................................................................... 159 7.2.2 Arboriculture and urban forestry ........................................................................ 159 7.2.3 Urban climate science and architecture ............................................................. 160 7.2.4 Rooftop solar technology placement.................................................................. 161 7.3 Limitations ............................................................................................................... 162 7.4 Directions for future work ........................................................................................ 163 7.4.1 LiDAR flight planning ....................................................................................... 164 7.4.2 Optimization of solar panel placement .............................................................. 164 7.4.3 ?Big data? integration ........................................................................................ 165 References ............................................................................................................................. 167       viii List of Tables Table 3.1. Mean, range and coefficient of variation (CV) of attributes used as predictor variables. .......................................................................................................................... 42 Table 3.2. Variance explained and error (years) associated with each random forest regression model for all 3282 buildings. ......................................................................... 48 Table 4.1. Vegetation height information for study sites (m). ................................................ 71 Table 4.2. Geometric attributes of the sampled building footprints. ...................................... 89 Table 4.3. Model error assessed as total daily irradiance integrated over the entire year for each point spacing configuration (RMSE, MJ m-2 day-1). ............................................... 91 Table 4.4. Computational time for the main steps in the POSt algorithm for a typical single-family dwelling (s). .......................................................................................................... 93 Table 4.5. Computational time for the main steps in the POSt algorithm for a typical multistory building (s). .................................................................................................... 93 Table 5.1. Mean monthly measured environmental data. ....................................................... 99 Table 5.2. Building energy performance parameters for multiple unit residential buildings (MURBs). ...................................................................................................................... 105 Table 5.3. Building energy parameter regression models and associated coefficient of determination (R2), significance (p) and standard error (SE). ....................................... 115 Table 6.1. List of building energy policy instruments and key considerations for local implementation. ............................................................................................................. 128    ix List of Figures Figure 1.1. Interconnection of dissertation chapters (red) and key data sources (grey). .......... 5 Figure 1.2. Map of the Pacific Northwest region of North America showing British Columbia and heating degree days. .................................................................................................... 6 Figure 1.3. Schematic showing LiDAR data collection and basic navigational and laser return components of an aerial acquisition. ............................................................................... 10 Figure 1.4. A three-dimensional view of a LiDAR point cloud (centered at 491450, 5456065 UTM10N). ....................................................................................................................... 12 Figure 1.5. Schematic of radiation modelling where bolded text and coloured elements indicate identified gaps in geographically explicit algorithms. ....................................... 17 Figure 2.1. Map of LiDAR coverage in the City of Vancouver. ............................................ 25 Figure 2.2. Depiction of LiDAR derived products where a) is the digital terrain model (DTM) in meters, b) is the normalized digital surface model (nDSM) in meters, c) is the intensity values as 8-bit integer values, and d) is the number of laser returns. ............... 27 Figure 3.1. Study area showing residential building footprints (black) and property parcels. 37 Figure 3.2. Example of LiDAR datasets showing (from left to right) LiDAR point returns, digital terrain model (DTM) and the digital surface model (DSM). ............................... 38 Figure 3.3. Variable importance plotted from Random Forests regression using variables derived from various municipal spatial datasets. ............................................................. 48 Figure 3.4. Partial dependence plots of select predictor variables. ......................................... 51 Figure 4.1. Location of study sites, Fairview (A), Mount Pleasant (B), Marpole (C) and the meteorological tower within the City of Vancouver. ...................................................... 56 Figure 4.2. Process workflow for integrating remotely sensed data across spatial scales to model irradiance. ............................................................................................................. 58 Figure 4.3. Representation of the key steps for estimating the vegetation extinction coefficient. ....................................................................................................................... 65 Figure 4.4. Scatterplot of observed atmospheric clearness index versus predicted clearness from GOES satellite-derived cloud cover. ...................................................................... 70 Figure 4.5. Average hourly clearness index (left) and standard deviation (right) for each month derived from the GOES cloud index for the years 2000-2010. ............................ 71  x Figure 4.6. Vertical distribution of vegetation transmission representative of each study site (a), and daily vegetation transmission on the equinox(es) (b), summer solstice (c) and winter solstice (d). ........................................................................................................... 73 Figure 4.7. Average azimuthal obstruction angles (top) and distances to obstructing features (bottom) for the study sites of Fairview (a), Mount Pleasant (b) and Marpole (c). Green wedges show results for each azimuthal direction when trees are included, while grey wedges show results using buildings and solid obstructions only. .................................. 74 Figure 4.8. Direct (top) and diffuse (middle) irradiance model outputs and the difference in irradiance (%) between opaque and semi-transparent representations of vegetation (bottom) for the study sites of Fairview (a), Mount Pleasant (b) and Marpole (c).......... 77 Figure 4.9. Post data structure and mapping within a relational database where each box represents a class with associated attribute values. ......................................................... 80 Figure 4.10. Representations of each point spacing configuration over a basic cubic building geometry where (a) is the validation case, (b) the centroid approach, (c) the random approach and (d) the vertical stacking approach. ............................................................ 82 Figure 4.11. Representation of the ray casting procedure where (a) depicts example locations of cast rays, (b) depicts the azimuthal angles at which rays are cast and (c) depicts the subtended obstruction angle for a given azimuth. ........................................................... 85 Figure 4.12. Example of the surface height raster layers representing (a) the digital elevation model and (b) the normalized digital surface model, in addition to (c) the 2D geometry of buildings and (d) a 3D representation of the selected building from panel (c). .......... 90 Figure 4.13. Modelled total daily building wall irradiance over a year (MJ m-2 day-1) for all unique wall orientations comparing the point spacing obstruction techniques with the validation case. Mean daily irradiance values for single-family dwelling and multistory buildings are presented in panels a) and c), while the standard deviation of mean daily irradiance are presented in panels b) and d) respectively. ............................................... 94 Figure 5.1. Study area located in a residential area of Vancouver. The dark grey regions shows the extend of the area used for analysis to maintain entire census dissemination areas, the light grey areas represent park space, and the buildings with residential units used for the analysis are shown in black. ...................................................................... 100 Figure 5.2. Trends in building energy components compared to the year of building construction. .................................................................................................................. 114 Figure 5.3. Mapped building energy parameters averages over 25 m by 25 m grids across the study area. ...................................................................................................................... 116 Figure 5.4. Mapped (a) hot water demand and (b) space heating demand averaged over 25 m by 25 m grids across the study area. .............................................................................. 117  xi Figure 5.5. Comparison of annual energy use intensity (EUI) values from the LiDAR-based approach and the building simulation software. ............................................................ 118     xii Glossary of Acronyms 2D   Two Dimensional 3D  Three Dimensional ACH50 Air Changes per Hour at 50 pascals CCP  Cities for Climate Protection CV  Coefficient of Variation DCC  Developer Cost Charge DPA  Development Permit Area DSM  Digital Surface Model DTM  Digital Terrain Model EGH  EnerGuide for Houses EPiCC  Environmental Prediction in Canadian Cities EUI  Energy Use Intensity ETR  Extraterrestrial Radiation GHG  Greenhouse Gas GIS  Geographic Information System GOES  Geostationary Operational Environmental Satellite GPS  Global Positioning Satellite ICLEI  International Council for Local Environmental Initiatives INU  Inertial Navigation Unit LiDAR Light Detection and Ranging LCZ  Local Climate Zone MURB Multi Unit Residential Building nDSM  Normalized Digital Surface Model OOB  Out of Bag POSt  Point Obstruction and Stacking RMSE  Root Mean Squared Error RSI  Resistivity (International System of Units) SE  Standard Error SFD  Single Family Dwelling SRTM  Shuttle Radar Topographic Mission          xiii Acknowledgements  First and foremost, I would like to acknowledge the support of my supervisor, Dr. Nicholas Coops.  Nicholas provided the resources, guidance and freedom necessary to ensure success and happiness during my graduate studies at the University of British Columbia.  I would also like to acknowledge the wide-ranging support that each of my committee members have contributed during this process.  Dr. Christen helped to provide me with a deeper understanding of the fundamental physical laws and numerical representations of urban processes.  Dr. Sheppard provided critical links and insights into local energy initiatives, which helped to situate my research within the context of local sustainability planning.  Finally, Dr. Hoberg helped me in developing a nuanced understanding of the political and policy considerations facing sustainable energy initiatives.  The Natural Sciences and Engineering Research Council of Canada, through their Doctoral Postgraduate Scholarship and Engage Grant, provided major financial support for this research.    I would also like to thank all my fellow lab members, collaborators and co-authors who volunteered their advice, code, data or insights throughout my doctoral research.  Finally, I would like to thank Katie Armitage for her continuous love and support.    xiv Dedication       To Scuffy  ?All knowledge, the totality of all questions and all answers, is contained in the dog? - Franz Kafka    1 1 Introduction  1.1 General background, objectives and chapter overview  Residential and commercial sectors are estimated to contribute more than 40% of the global demand for energy, with buildings being responsible for the majority of this energy use (IEA, 2013).  Moreover, approximately one third of CO2 emissions are attributed to the combustion of fossil fuels to service buildings and their occupants (Lee & Yik, 2004).  As a result, the building sector remains a critical candidate in efforts to reduce energy demand and to offset traditional energy use with more sustainable supply options.   Existing policy and planning strategies targeting energy-use tend be concentrated at the state, national or global level (Calvert et al., 2013; Kanellakis et al., 2013; Jaccard, 2005).  However, emerging opportunities for small-scale distributed energy generation (Keirstead et al., 2012), the promotion of building energy efficiency (Kavgic et al., 2010) and support for community energy systems (St. Denis & Parker, 2009) have increased the role and responsibility of local governments in energy-related management and planning.  In addition, local governments are situated in a unique position to implement various building energy strategies as a result of their direct connection to stakeholders and their experiential capacity providing the public with utility services (Betsill & Bulkeley, 2006; Burch, 2010; Collier, 1997; Guy & Marvin, 1996).  The services and functional responsibilities of local governments are also often explicitly spatial in nature.  In other words, local management and planning matters tend be located at select areas of the city, and decisions to address these   2 issues are different from place-to-place based on the spatial manifestation of the urban environment (Batey & Brown, 2007).    To manage the spatial character of urban services and infrastructure, Geographic Information Systems (GIS) are widely used by local planning and engineering practitioners as a computational resource for storing, analyzing and communicating baseline information on the existing condition of the city (Calvert et al., 2013; Harris, 1989; Howard et al., 2012; Webster, 1993, 1994).  This baseline information provides the opportunity for descriptive analyses of the location of physical and social infrastructure and the existing rules that apply to given spaces (i.e. zoning bylaws) (Webster, 1993). GIS also enables prescriptive analyses, which tend to focus on the outcomes of potential processes and decisions as reflected for different locations across the landscape (Webster, 1993).  Although descriptive and prescriptive spatial analyses serve as fundamental planning tools, their application to urban energy management remains under-examined both in practice and in the academic community.    Horner et al. (2011) suggest that there is a disconnection between geographic information scientists and decision makers dealing with energy issues.  By extension, Horner et al. (2011) describe three critical energy-related issues that offer immediate opportunities for integration with geographic information science: 1) carbon estimation and inventory, 2) energy infrastructure placement and transmission, and 3) household energy conservation and efficiency.  Although each of these topics is applicable across spatial scales and political   3 jurisdictions, energy use in the building sector is of particular interest to local governments, given the interactions between city officials and stakeholders in this sector.   Existing efforts to support synergies between geographic information science and local building energy issues have been centered on populating spatial databases with geographic details related to the local environment and building energy components (Rylatt et al., 2003; Heiple & Sailor, 2008).  Because these metrics tend to be generated by disaggregating broad economic and utility data (Heiple & Sailor, 2008), opportunities remain to better integrate bottom-up modelling approaches using three-dimensional (3D) datasets of the urban environment.  Specifically, airborne light detection and ranging (LiDAR) data offers a contiguous representation of urban form at a spatial scale suitable to help populate individual building energy simulation and modelling.  Moreover, LiDAR is becoming an increasingly common resource available to local and regional planners.  The primary objective of this dissertation is therefore to investigate approaches that integrate airborne LiDAR for informing spatially detailed and contiguous estimates of energy demand in the residential building sector.  From this general objective, four underlying research questions are formulated (discussed in more depth later in this Chapter):  1) How is urban and building form, as derived from airborne LiDAR, related to building energy performance? 2) How can airborne LiDAR data be used to assess building solar energy gains?   4 3) How can airborne LiDAR be used to provide spatially contiguous estimates of building energy demand? 4) Which local planning and policy strategies can be informed from citywide estimates of building energy demand?  These research questions have been addressed in the context of several case studies in Vancouver, British Columbia, and selected to represent a range of urban form and vegetation types common in many North American cities. Following a description of the general study area and data sources in Chapter 2, questions 1, 2, 3 and 4 are then addressed in Chapters 3, 4, 5 and 6 respectively.  Specifically, Chapter 3 examines the use of airborne LiDAR data to augment the prediction of building age and energy performance using a machine learning approach.  Chapter 4 presents LiDAR-based approaches for estimating solar irradiance on building rooftops and building walls.  Chapter 5 then combines the findings from Chapters 3 and 4 to develop estimates of individual building energy demand across entire urban regions.  Chapter 6 provides a comprehensive review of known policy instruments available to local governments to address building energy, while highlighting those policies with unique spatial considerations that could benefit from a site-specific contiguous inventory of energy demand factors.  Finally, Chapter 7 concludes this dissertation by summarizing the major findings, potential applications and possible future research directions.  The interconnection between these chapters and the integration of relevant data sources are shown in Figure 1.1.   5  Figure 1.1. Interconnection of dissertation chapters (red) and key data sources (grey).   The remainder of this introduction will provide the policy context for developing better building energy analytics followed by a review of the background literature and modelling considerations relevant to the geographic assessment of energy use in the building sector.   1.2 Energy and emissions policy in British Columbia, Canada  In Canada, provincial governments have jurisdiction over the majority of energy-related activities, including the development and management of non-renewable resources and electricity.  Provinces also grant authority to local governments, who do not have constitutional jurisdiction.  The provincial context therefore provides the dominant arena in which Canadian energy policies are framed.  As a result, this section touches briefly on energy policy initiatives in the Province of British Columbia (Figure 1.2) relevant to the building sector and local governments.    6  Figure 1.2. Map of the Pacific Northwest region of North America showing British Columbia and heating degree days.  In 2007 the British Columbia provincial government introduced aggressive greenhouse gas (GHG) reduction targets in the Greenhouse Gas Reduction Targets Act (Bill 44) of 33% and 80% below 2007 levels for the years 2020 and 2050 respectively.  Furthermore, the Local Government (Green Communities) Statutes Amendment Act (Bill 27) was released in 2008 stipulating that local governments must include GHG reduction targets in all official community plans and regional growth strategies.  In addition to greenhouse gas reduction   7 targets, the provincial government enacted the Clean Energy Act in April 2010, addressing electricity self-sufficiency, conservation, and the use of clean and renewable energy resources.  This act also expresses an objective ?to encourage communities to reduce greenhouse gas emissions and use energy efficiently?.  Most recently, in 2013, the Province of British Columbia adopted new building code requirements, which added the option for performance-based indicators of regulatory compliance; to be evaluated using building energy simulation.  As a result of these energy initiatives, local governments across the province require new and innovative planning tools to address building energy.   1.3 Fundamentals of LiDAR-based energy modelling  As denoted in Figure 1.1, the fundamental approach of this dissertation focuses on the use of airborne LiDAR to 1) assess building age and relate it to building energy performance, 2) determine building envelope solar gains, and 3) integrate the geographic variability of building performance and solar energy gains to better inform estimates of the demand for building energy services.  The novel aspects of this dissertation relate specifically to the development of techniques to systematically incorporate the effects of local urban form on energy demand for the purpose of producing spatially contiguous estimates of energy-use in the building sector.  Furthermore, this research facilitates the identification of fundamental components of energy demand to inform site-specific strategies focused on more sustainable energy practices in the building sector.  Throughout this dissertation the problem of local building energy assessment is approached from a geographic perspective with an emphasis on geographic analysis techniques.   8  The following sections provide an overview of three fundamental concepts to this dissertation: 1) LiDAR, 2) building energy modelling, and 3) solar radiation modelling.  In each section, the novel research opportunities are revealed in light of the current state of the existing geographic technologies and models.  1.3.1 An overview of light detection and ranging (LiDAR)  In its essence light detection and ranging (LiDAR) is a technology used to measure the distance to objects.  To calculate distances the LiDAR system emits a laser pulse and records the time taken for that same pulse to return to a sensor.  Using the time of the return, the technology is able to accurately compute the distance to an object.  Given that the location of the sensor is known, the LiDAR data can then be used to map the position and height of objects in its scanning range.  Much of the value of LiDAR is realized when the sensor is mounted to a moving object such as a vehicle or airplane (see Figure 1.3).  In these cases, large areas can be scanned to produce very accurate 3D representations of an environment.  Discrete-return LiDAR emits and records individual laser pulses, thus three-dimensional surfaces and objects are collected and represented as a set of points, each typically containing geographic coordinates and a measurement of vertical height.  The group of LiDAR points used to represent a 3D surface is called a point-cloud.  An example showing a LiDAR point cloud representation of Vancouver?s Fairview neighbourhood is depicted in Figure 1.4.   9  LiDAR science, defined here as the analysis and processing of laser scanning measurements, dates back about half a century, shortly after the discovery of laser technology at Bell Labs in 1958 (Nelson, 2013).  Beginning in the mid 1960s, scientists started adapting laser technology to applications in the natural resource sector by designing laser-profiling systems mounted to aerial platforms.  Much of the early analysis of airborne LiDAR data was focused on producing accurate topographic and bathymetric models.  In these models, features above the surface, such as trees and buildings, were considered noise and it was not until the 1980s that substantial scientific efforts were put towards information generation from the above-ground LiDAR returns.  Together, the evolution of LiDAR processing algorithms and advances to geospatial technologies have contributed to the operationalization of LiDAR for natural resource management (Eid, 2004), and increasingly for urban planning purposes.    10  Figure 1.3. Schematic showing LiDAR data collection and basic navigational and laser return components of an aerial acquisition.  In addition to the laser scanning technology, airborne LiDAR systems have been enabled by Global Positioning System (GPS) availability and accuracy, better aircraft Inertial Navigation Systems (INU) and advances to computing power and the accompanying weight and size reduction of computers.  In combination, these technologies enable modern LiDAR systems to locate the surface position of laser pulses to within 1 m horizontal and 0.15 m vertical accuracy (Davenport et al., 2004) at a cost of around CDN$3 ? 10 ha-1 (Wulder et al., 2008).  Accurate and detailed 3D surface reconstruction over large areas has also been facilitated by   11 increases to the pulse repetition frequency of the scanning device, with modern systems specified to rates greater than 350 kHz.  In comparison, the first pulsed-laser system generated less than 100 pulses per second (0.1 kHz) (Hickman & Hogg, 1969).   The application of airborne LiDAR to natural resource applications has been well described (Eid, 2004).  In contrast, LiDAR science for urban applications presents a nascent field of inquiry, with much of the existing literature addressing 3D building reconstruction (Wang, 2013).  Although no empirical data is available, anecdotal evidence also suggests that the operational opportunities of LiDAR are more evident to practitioners in the natural resource sector than those in urban planning related roles.  However, local government officials and staff are starting to recognize the value of LiDAR for purpose-specific applications, such as floodplain mapping, landslide hazard assessment and stormwater management.  Although it remains to be seen whether airborne LiDAR will achieve the ubiquity and utility of aerial imagery in the realm of urban management and planning, there continues to be a wealth of opportunities to advance LiDAR science for improving the understanding of spatial patterns and processes in the urban environment.     12  Figure 1.4. A three-dimensional view of a LiDAR point cloud (centered at 491450, 5456065 UTM10N).   1.3.2 Building energy demand modelling  According to the first law of thermodynamics, energy cannot be created nor destroyed.  As a result, energy demand is most appropriately conceptualized in terms of the services it provides.  Energy services in the building sector include the provision of space conditioning, water heating, lighting, and the powering of devices (including home and commercial   13 appliances, electronics and industrial machinery).  Each of these services demonstrate variation across space and time, with space heating and water heating contributing more than 80% of typical Canadian residential household energy demand (Aguilar et al., 2005).   Existing efforts to develop extensive energy estimates for the building sector can be divided into two fundamental modelling approaches: top-down and bottom-up (Kavgic et al., 2010; Swan & Ugursal, 2009).  Top-down econometric models are based on historic macroeconomic trends between broad ? often national ? economic variables and energy use (Kavgic et al., 2010).  In contrast, bottom-up approaches, which are typically used in building energy simulation and modelling programs, incorporate disaggregated physical and empirical data related to individual building energy components (Kavgic et al., 2010; Keirstead et al., 2012; Heiple & Sailor, 2008).  In addition, a variety of hybrid models are also available that couple aspects from both top-down and bottom-up approaches (Jaccard, 2006).  Since a physical-based approach allows for the capture of detailed spatial variations and local environmental variables necessary to differentiate the energy demand of individual buildings (Robinson, 2006), the methodological framework for this dissertation is framed within the context of bottom-up modeling.   Due to the data intensive nature of bottom-up models and the complexity of urban environments, building energy simulations remain difficult to scale across an entire city (Swan & Ugursal, 2009).  In a geographic context, this is evident when considering the local physical environment in which a building is located.  For example, local parameters including radiation, wind, convective exchange, humidity and temperature demonstrate   14 substantial variations across built environments (Oke, 1987), although they tend to be represented as spatially stationary variables in building energy models (Crawley et al., 2008).  As a result, techniques are required to incorporate the spatial variation of environmental variables into building energy modelling.  The second category of variables critical for building energy modeling includes those parameters that represent the efficiency of a building?s envelope.  While these parameters also vary spatially, statistical relationships between the efficiency of the building envelope and its age are typically used to inform bottom-up building energy models (Kavgic et al., 2010; Swan & Ugursal, 2009; Heiple & Sailor, 2008; Aydinalp et al., 2004; Farahbakhsh et al., 1998; Rylatt et al., 2003; Parekh, 2004).  These existing models tend to group building vintages into categories based on similar structural or technological characteristics and provide an average estimate of the energy parameters associated with system components for each age category.  Examples of energy components include insulation, glazing, ventilation, construction material, and the efficiency of devices such as water heaters and furnaces.  However, unlike building geometry and design, these parameters are not possible to detect directly from active remote sensing technologies and must be inferred from empirical relations with other variables or from in-situ measurements.  To facilitate the automation of citywide energy use estimates, geographic information systems (GIS) have been attributed with details related to the local environment and building energy factors (e.g. Rylatt et al. (2003)). However, these models remain dependent on manual inputs from trained experts.  As a result, automating model parameterization using   15 remote sensing presents a sensible advancement to spatially explicit building energy assessment procedures.  Of specific interest is the integration of LiDAR, which offers unprecedented measurement details of the three-dimensional form of the urban environment and which has become an increasingly common resource in local planning and engineering departments (Goodwin et al., 2008).  While Neidhart and Sester (2004) have applied airborne LiDAR measurements of volume for the purpose of predicting building energy demand, more sophisticated LiDAR-based approaches remain otherwise unexplored within the literature. Specifically, LiDAR offers the opportunity to generate a wealth of building morphological metrics, which have yet to be examined for their ability to predict the envelope and system efficiencies of buildings.  Furthermore, LiDAR enables the quantification of solar occluding features and the spatial orientation of urban facets that are critical for determining solar gains for building energy models.  1.3.3 Solar radiation in the built environment  The solar irradiance received on building envelopes is a critical driver of building energy performance.  In addition to the material and morphological components of buildings that regulate the transfer of energy between the inside and outside of the building envelope, the external environment itself acts as an important determinant of building performance, resulting from the influence of urban form on passive solar heat gains.  Notably, physical objects, such as buildings, vegetation and terrain occlude the direct shortwave radiation emitted from the sun and the diffuse radiation reflected from the sky, and impact the reflected radiation from surrounding objects.   16  The application of remote sensing technology to urban areas affords opportunities for a new generation of solar radiation models that specifically address the unique spatial composition of the complex urban environment.  LiDAR technologies are of particular interest as they enable highly automated, accurate and precise surface reconstruction.  Various studies have employed digital surface models produced using LiDAR data for estimating incoming radiation (Arboit et al., 2008; Hofierka & Ka?uk, 2009; Levinson et al., 2009; Yu et al., 2009).  However, to advance radiation models, further research is required to determine appropriate methods for integrating spatial and temporal variations in atmospheric conditions, adapt the models for the unique surface geometry of urban environments and predict the transmission of radiation through tree canopies.  These knowledge gaps are detailed in the remaining section of this introduction and illustrated in Figure 1.5.    17  Figure 1.5. Schematic of radiation modelling where bolded text and coloured elements indicate identified gaps in geographically explicit algorithms. Measuring and modelling surface irradiance  Solar radiation received at the top of the Earth?s atmosphere, referred to as extraterrestrial radiation (ETR), varies slightly in intensity (3.4%) throughout the year as a result of the eccentricity of the Earth?s orbit, with an average ETR of 1367 W m-2 (Iqbal, 1983).  Once this radiation enters the Earth?s atmosphere it is separated into several components depending on a variety of physical processes that scatter, absorb or transmit the photons.  As a result of these processes, solar radiation received at the Earth?s surface can be classified into direct beam radiation, diffuse sky radiation and reflected radiation.  Direct beam radiation is the radiation that reaches the Earth?s surface without being absorbed or scattered by particles in the atmosphere or features on the ground.  Diffuse sky radiation is the down-welling   18 radiation that is scattered within the atmosphere.  Finally, reflected radiation is the radiation that is reflected onto a surface from the terrain or features on the Earth?s surface such as buildings and trees.  Actinometers and radiometers have been used since the early 1800s to measure electromagnetic radiative fluxes at the Earth?s surface (Stull, 2011).  By strategically placing these sensors across various environments scientists have been able to measure the spatial variation in both incoming and outgoing radiation.  Spatial interpolation of radiation measured at (or near) the ground has facilitated the development of coarse resolution continuous maps of solar radiation for the entire Earth surface (Pelland et al., 2006).  However, many meteorological ground stations are not equipped with actinometers and simple interpolation methods fail to capture the variance in radiation over long distances or in complex terrain.  As a result, a diversity of physical and empirical radiation models have been developed to better account for the geographic variation in radiation at the Earth surface (Hay & McKay, 1985; Ulgen & Hepbasli, 2004).  To classify and assess the application and utility of individual solar radiation models Gueymard & Myers (2008) identify nine criteria for evaluation: type of input data, type of output data, spatial resolution, temporal resolution, spectral resolution, type of methodology, type of algorithm, surface geometry and type of sky.  Each of these criteria must be carefully addressed when modelling radiation in urban areas, with specific attention given to available input data and associated spatial resolution.  An effective model also requires an approach that integrates available model input data across spatial scales, since the collection of   19 radiation modelling parameters are gathered from a disparate set of instruments and sensors (Polo et al., 2008), most of which do not resolve individual facets of the urban surface.  Remotely sensed datasets offer unique opportunities for integration with solar modelling approaches by providing spatially contiguous measurement of the atmosphere and the surface of the Earth. Radiation models in the geospatial sciences  Early applications of remote sensing technologies in radiation models utilized data from geostationary satellites to estimate incoming radiation by measuring reflected radiation from the ground and atmosphere.  The HELIOSAT method provides an example of an extensively tested model that is still widely used for estimating irradiance at a broad spatial resolution (1 km) (Hammer et al., 2003).  Although the heterogeneous character of the urban environment has traditionally discouraged spatially detailed solar model parameterization using remote sensing technology, increases to sensor resolution and active sensors ? such as LiDAR ? offer new opportunities for providing suitable details of urban form and land cover.  The gridded (raster) nature of much of the spatial data collected by aircraft of satellite sensors has led to array-based techniques for estimating irradiance on horizontally oriented surfaces.  These techniques have been described for a variety of environments (Duguay, 1993; Fu & Rich, 1999; Kumar et al., 1997), including urban areas (Gros et al., 2011; Hofierka & Kanuk, 2009).  The computational approach specific to radiation models in geographic information   20 systems typically adopt a hemispherical viewshed algorithm that involves the calculation of unobstructed coordinates of the local sky for a given point on the Earth?s surface (Rich et al., 1994; Kumar et al., 1997).  Critical to this approach is the availability of a digital representation of surface topography.  This digital surface model provides the heights of surface objects from which sky occluding features and their local azimuthal and zenithal coordinates are calculated.  To determine solar radiation for a given location, the viewshed is compared to the position of the sun as it moves across the sky and fraction of radiation transmitted through the atmosphere.    Solar position algorithms use established astronomic formulae to determine the topocentric coordinates of the sun for a given date and time, and can provide solar positional accuracies with less than 0.0003? of precision (Reda & Andreas, 2004).  In contrast, spatially contiguous estimates of atmospheric effects are more difficult to model given the spatial and temporal variability of clouds and aerosols (Kaufman et al, 2002).  In geographic radiation models, atmospheric processes tend to be represented using a static global value of radiation transmission (Rich et al., 1994; Fu & Rich, 1999; Kumar et al., 1997).    In addition to atmospheric effects, radiation modelling in urban environments is largely dependent on representations of urban form including elevation, surface slope and aspect, and nearby obstructing features (Duguay, 1993; Robinson & Stone, 2004).  While LiDAR has been used to develop digital surface models at a fine spatial resolution, vertically oriented surfaces such as building walls are omitted in these planimetric representations of urban topography.  Therefore, the adaptation of geographically explicit radiation models to building   21 walls remains an open area for research.  Furthermore, the transmission of radiation through vegetation in the urban canopy has not been well represented in efforts to develop surface radiation estimates at high spatial and temporal resolutions (Christen & Vogt, 2004).  As a result, new methods are required to better represent the semi-transparent nature of objects such as trees. Measuring vegetation transmission using LiDAR  In forestry applications LiDAR data point clouds have been examined to determine vegetation structural information (Lovell et al., 2003; Ria?o et al., 2003; Ria?o et al., 2004; Coops et al., 2007), although minimal research has taken place for adapting these approaches to urban environments.  Shape-based metrics of vegetation structure are of particular interest when examining radiation transmittance since they enable assessment of the dynamic interaction between radiation, tree structure and changing sun angles.  One well-developed approach to transmission curve-fitting involves the derivation of gap probability models, which determine canopy structure by correlating the attenuation of the returned laser pulses with the density, size, and distribution of foliage and woody elements (Ni-Meister et al., 2001), producing a cumulative vertically projected profile of light tranmissivity for an entire canopy.  However, these approaches have been developed for entire forest stands, suggesting the need for adaptive techniques that can characterize urban environments where trees exhibit a diversity of spatial patterns and where trees are more likely to be found in isolation.      22 2 Study Area and Data  The following chapter outlines the general study area and key datasets that were used throughout this dissertation.  All studies were conducted within the City of Vancouver, Canada.  With a population of around 600 000 people, the City of Vancouver is the largest municipality in the Province of British Columbia and the eighth largest municipality in Canada (Statistics Canada, 2012).  The City of Vancouver is located in the Pacific Maritime Ecozone, which is characterized by the warmest and wettest climate in Canada with mean monthly temperatures ranging between 4 ? 6? C in January to 12 ? 18? C in July.  As a result, the buildings in the City of Vancouver have less space conditioning requirements than other Canadian cities, with 2948 heating degree days (see Figure 1.2) and 47 cooling degree days per year (Wahlgren, 2010).  The low cooling requirement in the City of Vancouver is unique in contrast to many North American cities that require substantial electricity to operate air conditioning units.  Nonetheless, the building sector in Vancouver contributes to approximately 49% of community GHG emissions, with residential, commercial and small to medium industrial buildings emitting approximately 1 133 000 T CO2e per year (CEEI, 2010).  Natural gas in the residential building sector, used for space and water heating, is responsible for over 90% of building energy related emissions, although it accounts for only 60% of operational building energy use (CEEI, 2010).   Urban climatological and meteorological research has a long history in the City of Vancouver, dating back to pioneering research by Dr. Timothy Oke at the University of British Columbia, who helped to uncover the urban heat island phenomena (Oke, 1982).    23 Since 1978, Dr. Oke and a variety of researchers have been taking urban weather measurements from a tower located at the intersection of 49th Avenue and Knight Street in a suburban area of Vancouver.  From 2006 to 2009, Dr. Oke and Dr. James Voogt led the Environmental Prediction in Canadian Cities (EPiCC) project to better understand the urban atmosphere in Canada through a program of observation, modelling and remote sensing (Voogt et al., 2007).  As part of this project, LiDAR data was acquired for various locations across the City of Vancouver.  As a result of these measurements, along with a range of additional data that have been collected to support urban climate research, the City of Vancouver presents an ideal candidate study area for ongoing research examining the physical geographic processes of urban areas.  In addition to a long history of urban research, progressive urban planning initiatives at the City of Vancouver have also helped to establish a set of local policies and planning priorities addressing sustainability issues.  The City of Vancouver?s current guiding plan for sustainability, the Greenest City 2020 Action Plan, sets a variety of goals and targets for carbon, waste and ecosystems.  Central to this plan is a focus on green buildings, requiring all new buildings to be operationally carbon neutral by 2020, and reducing the GHG emissions in existing buildings by 20% (City of Vancouver, 2012).  While the City of Vancouver provides the general study area for this dissertation, restrictions on data availability excluded the possibility of citywide analysis.  However, areas were selected in Chapters 3 ? 5 to best represent the variety of environments and built conditions found across Vancouver and in many other North American cities.  As a result, in each of   24 these chapters a brief explanation is provided to clarify the relevant attributes for analysis, which is intended to supplement the general description of the City of Vancouver presented above.  The remainder of this chapter is dedicated to introducing the various sources of data used explicitly for the analysis throughout this dissertation.  2.1 Light detection and ranging (LiDAR) data  In Chapter 1, a general description of LiDAR is provided to familiarize the reader with the basic principles and functionality of the technology.  Here the LiDAR specifications are provided for the dataset used in the analyses in Chapters 3, 4 and 5.  The LiDAR flight path covers an area centered on Cambie Street in the middle of Vancouver, spanning from the North Arm of the Fraser River to False Creek (See Figure 2.1).  The LiDAR flight path was designed to cover a range of urban form, building types and vegetation cover.      25  Figure 2.1. Map of LiDAR coverage in the City of Vancouver.  The LiDAR was captured on August 20, 2011 using a Leica ALS60 laser scanner mounted to a fixed-wing platform with a pulse rate of 113.7 kHz, field of view of 45?, wavelength of 1064 nm and output beam divergence angle of 0.22 mr.  The sensor was flown at an altitude of 754 m, resulting in an average ground point spacing of 0.57 m and point density of 4 points m-2 (Figure 2.2d).  Ground and non-ground laser return signals were classified in-house by the LiDAR provider.  The classified LiDAR points were then gridded to create 1 m  spatial resolution images of first returns, which form a digital surface model (DSM) representing the highest feature elevations across the study area, and ground returns, which form a digital terrain model (DTM) representing the bare-ground surface (Figure 2.2a).  Both  the DSM and DTM were produced using a natural neighbour interpolation based on Delaunay triangulation (Goodwin et al., 2009).  The DTM was then subtracted from the   26 DSM to generate a normalized digital surface model (nDSM) (Figure 2.2b), which provided a representation of the heights of above ground features.    In addition to feature elevation information, the LiDAR sensor also records the return strength of the laser pulse, known as intensity, which was also gridded (Figure 2.2c) and included for analysis using the same technique described to produce the digital elevation models.   27  Figure 2.2. Depiction of LiDAR derived products where a) is the digital terrain model (DTM) in meters, b) is the normalized digital surface model (nDSM) in meters, c) is the intensity values as 8-bit integer values, and d) is the number of laser returns.    28 2.2 Shuttle Radar Topographic Mission (SRTM) data In addition to the LiDAR, which provided local elevation data, it was important to capture the vertical relief of the highly mountain terrain around the City of Vancouver.  This information was captured at a 90 m spatial resolution from NASA?s Shuttle Radar Topographic Mission (SRTM) for terrain located up to 100 km beyond the extent of the LiDAR coverage (Rodr?guez et al., 2006).  2.3 Geostationary Operational Environmental Satellite (GOES)  The archive of GOES satellite data dates back to the mid 1970s and provides one of the oldest repositories of satellite weather data for North America (Menzel & Purdom, 1994).  The GOES system uses geosynchronous satellites that track the speed of the Earth?s rotation and therefore remain stationary with respect to a point on the Earth?s surface.  This geostationary orbit is located approximately 36 000 km above the ground resulting in a spatial resolution of 1 km at the equator from the GOES satellite sensors.  The GOES system of satellites is designed for weather forecasting and meteorology research, and one of its primary uses is to measure clouds over North America at regular intervals throughout the day.  In this dissertation GOES-West Continental US images were collected in the visible channel at hourly intervals for the period from 2000 to 2010.  Spatial resolution of the data at the latitude of the study sites equaled approximately 1.7 km2 (0.925 km by 1.85 km).  The GOES   29 data is used in chapter 4 to provide a spatially contiguous estimate of cloud cover for integration into a geographically explicit model of shortwave irradiance at the urban surface.  2.4 Statistics Canada census data  Statistics Canada conducts a nationwide census every five years and typically distributes census results one or two years after the census is completed.  The last Canadian census was conducted in May 2011.  Census data is disseminated using various geographic areas and boundaries.  The finest spatial resolution data is provided as Dissemination Areas that are comprised of one or more neighbouring blocks with a population of 400 to 700 people.  While numerous variables are available from census data, the data used in this dissertation include only information related to population counts and dwelling type.  The census population data is used in Chapter 5 to inform anthropogenic heat contributions in buildings and the demand for hot water.   2.5 British Columbia Assessment data  Data on building age (construction date) was provided by BC Assessment, an agency of the government of British Columbia responsible for maintaining property tax information in accordance with the Assessment Act.  The mandate of BC Assessment is to establish and maintain uniform real property assessments throughout the Province of British Columbia.  One of the main attributes used to assess real property market value is the age of the building.  As a result, building construction date and the date associated with any major renovation are   30 collected for the majority of properties in British Columbia.  This non-confidential information was provided for all land parcels in the City of Vancouver for the year 2009 from Landcor Data Corporation.  The BC Assessment Property Database is also accessible by purchase from BC Online, which provides access to a variety of provincial government computer systems over the Internet.  2.6 Residential building energy audit data  Indicators of building energy performance were derived from Canadian residential building energy audits (http://oee.nrcan.gc.ca/residential/6551).  Data were collected as part of Natural Resources Canada?s mandate to promote energy efficiency in order to reduce the environmental impact of energy use within Canada?s residential building stock.  As part of this mandate, Natural Resources Canada has established the EnerGuide for Houses (EGH) program.  Central to this program has been the development of a database designed to track building energy efficiency evaluations across Canada.    To qualify for building energy grants as part of the recent ecoENERGY Retrofit program (2007 ? 2012), homeowners were required to have an energy audit conducted by a certified inspector.  Results of this audit data were added to the EGH database, and these data used in Chapter 5 of this dissertation to relate building age to energy use.  The building energy audit information of particular interest to this dissertation relates to those building energy components directly measured or evaluated for pre-retrofitted homes, and includes: ? the year of construction   31 ? type of house ? ceiling insulation RSI value ? foundation insulation RSI value ? main wall insulation RSI value ? air leakage at 50 pascals ? predominant window code ? wall area ? window area  To protect the privacy of ecoENERGY program participants, the audit records were located to three digit postal code areas (forward sorting areas).  The data provided information on 7253 buildings in the City of Vancouver.  2.7 Cadaster, zoning and building footprint data  The City of Vancouver provided data related to land parcels, zoning districts and building two-dimensional footprints.  In 2009 the City of Vancouver decided to embrace principles of open and accessible data, and through this initiative emerged the Vancouver Open Data catalogue, a repository of over 130 municipal datasets.  Land parcel and zoning districts were provided through this repository.  Building footprints are not available on the Open Data catalogue, but were provided by the City of Vancouver Geographical Information Systems department.  According to city staff, these building footprints were generated using LiDAR acquired and analyzed by the US Central Intelligence Agency for security purposes as part of   32 the 2010 Vancouver Olympics. A total of 93 355 primary building units were included in the footprint data and approximately 8% of these delineated buildings were within the area covered by the LiDAR.  2.8 Ground instrumentation  From January 2009 to December 2011 data was collected on air temperature (iHMP-35A, Vaisala Inc., Vantaa, Finland, operated at 26 m on a tower), soil temperature (thermocouples, between 5 and 10 cm depth at residential yard sites) and incoming longwave and shortwave radiation (CNR-1, Kipp & Zonen, Delft, Netherlands, operated at 29m on a tower).  The tower where air temperature and radiation were observed is located at 49.226?N, -123.078?W, approximately 1 km East of the LiDAR coverage.  Finally, the City of Vancouver provided information on piped water temperature from water sampling locations across the city (City of Vancouver, 2009).     33 3 Prediction of Building Ages using LiDAR  3.1 Introduction  Specific building energy efficiency retrofits and building energy supply technologies have been developed and well described (Waide et al., 2007).  However, assessing the existing base energy conditions and performance of multiple individual buildings, across entire cities for example, remains a challenge.  Research opportunities therefore exist to better inform and automate existing energy performance and efficiency parameterizations for integration with building energy modelling and simulation efforts.  The basic factors that determine building energy performance include geometry, building design, systems efficiency and occupant behaviour (Ratti et al., 2005).  For an individual building, the first three of these factors are generally static over time, while occupant behavior can be highly variable.  Although geometry, building design and systems efficiency typically remain unchanged for an individual building over several decades, patterns and trends in these attributes are observed across neighborhoods and cities due to changes in building codes, architectural preferences and technological advances.  As a result, these criteria can be described as demonstrating spatial and temporal variability.  However, because these parameters are difficult to assess for many individual buildings, spatially contiguous building energy modelling efforts typically establish relations between a building?s vintage, represented as a range of ages or construction dates, and key performance   34 attributes (Caputo et al., 2013; Aydinalp et al., 2004; Farahbakhsh et al., 1998; Neidhart & Sester, 2004).  Basic spatial predictors of building energy demand have been previously identified and exploited to facilitate automation over large areas using geographic information systems (GIS).  Examples include municipal zoning data, census data and simple building morphology characteristics (Rylatt et al., 2001).  While the option to integrate spatial data into current building energy models is underway (Crawley et al., 2008), the direct application of remote sensing products remains comparatively underutilized.  Specifically, advances to active sensors, such as light detection and ranging (LiDAR) technology, offer the potential to generate detailed information related to building geometry and design.  Building morphological characteristics are of interest to building energy modelling and simulation since a direct relation exists between energy end-use in a building and its shape.  This relation is associated largely with volume, since a larger building typically requires more energy for space conditioning, but also relates to shape parameters that influence interactions between the indoor and outdoor environments such as the amount of sunlight and shade received on the building at different times of the day and year.  While these relations are well understood, additional building morphological attributes have also been suggested as potential predictors of building age that can help to determine building energy performance.  For example, classification techniques have been used to determine categories of building age using basic building morphological characteristics and related to space heating demand (Neidhart & Sester, 2004).  Exploration of the relation between building form, age and   35 energy performance presents a nascent area of research, while novel classification techniques offer opportunities to identify the variables that are most important to energy models of the existing building stock, ultimately helping to improve model accuracies.  Machine learning algorithms have emerged as an effective set of classification techniques because of their ability to facilitate the automation of complex pattern recognition with large datasets.  One such approach is random forests (Breiman, 2001; Cutler & Stevens, 2006; Cutler et al., 2007), which has been demonstrated to provide superior performance compared to other machine learning algorithms when evaluated for numerous criteria (Caruana & Niculescu-Mizil, 2006).  The random forests learning method operates by creating many individual decision trees and predicting outcomes based on the average results from all the trees.  In urban remote sensing classification, random forests have been successfully applied at various spatial scales to estimate urban land cover from satellite imagery (Walton, 2008) and laser-scanning data (Guo et al., 2011).  Random forests have also been used to determine the importance of individual predictor variables (Guo et al., 2011; Cutler at al., 2007), offering insights into model interpretation and development.    The purpose of this Chapter is to examine the contribution of LiDAR-derived attributes to predictions of building age for integration with a building energy model presented in Chapter 5.  To do this, three random forest models are established that include variables corresponding to the increase in sophistication of available municipal spatial data.  Variables in the first model represent traditional cadastral and zoning information, while the second model includes variables representing two-dimensional building attributes.  The third model   36 provides the main focus of this study and includes a suite of LiDAR-derived variables that inform a building?s morphology.  Finally, a fourth model is generated that combines all the variables from the first three models.  The predictive capacity and error associated with each model is compared, while the fourth model is also used to inform variable importance and to help interpret the role that select variables play in predicting age.  3.2 Method  3.2.1 Study area  The specific study area examined in this Chapter covers a 2 km by 2 km primarily residential neighbourhood in Vancouver, Canada (Figure 3.1).  A range of buildings exist across the neighbourhood including residential single-family detached dwellings, townhouses, low-rise apartments of less than 5 stories, high-rise apartments of 5 stories or more, mixed-use buildings and buildings entirely devoted to commercial and institutional uses.  In total 4289 buildings were contained within the study area boundary with 3282 classified as residential and used for analysis.  The age range of buildings spanned 108 years with construction dates between the years of 1901 and 2009.     37  Figure 3.1. Study area showing residential building footprints (black) and property parcels.  3.2.2 Data  In this chapter, the LiDAR derived digital surface model (DSM), digital terrain model (DTM), normalized digital surface model (nDSM) and intensity values were used to help predict building age (Figure 3.2).  Data on building age (construction date) was provided   38 using the BC Assessment data.  While the determination of building age can be made through examining property assessment data, this data is not always available for all jurisdictions or all properties in a jurisdiction, and in certain cases the data may be prohibitively expensive to purchase for many individual properties.  Furthermore, an additional aim of this chapter is to explore how building design and geometry elements are related to its age.  Building footprints, parcel and zoning data were also used to attribute buildings with a diversity of potential determinants of age.    Figure 3.2. Example of LiDAR datasets showing (from left to right) LiDAR point returns, digital terrain model (DTM) and the digital surface model (DSM).    3.2.3 Model predictor variables  Three broad levels of sophistication in municipal spatial data were examined as general categories of variables for input to random forests regression algorithms (see Section 3.2.4).  The levels of detail were selected to represent general technological advancements that allow increases in the characterization of spatial attributes of urban features.  Since the objective of this chapter is to relate building age to its shape, the common spatial dominator for analysis   39 was that of an individual building.  Nonetheless, the majority of variables were generated from finer spatial resolution data and summarized as a single metric for each building in the study area.  The first level of detail (model 1) accounted for traditional municipal information as represented in cadastral data and zoning districts.  Variables in model 2 represented two-dimensional building attributes that can be characterized by the spatial detail available from basic remote sensing technologies.  Model 3 included the unique three-dimensional building form-based attributes that can be captured from LiDAR.  Finally, model 4 incorporated all the variables from models 1-3.  Description of the specific variables used in each model and the techniques used for their computation is provided below and listed in Table 3.1. Model 1: Traditional municipal information  Model 1 provided basic attributes from municipal spatial data related to property information and zoning districts.  Zoning data were used to help identify common development patterns within the study area.  Although buildings were pre-selected to include only those within residential areas, additional planning and development requirements and restrictions were associated with the specific zone classifier.  Therefore the zone classifier was assessed for each building and provided the only categorical predictor variable in this study.  Property information was derived from parcel polygons for each building.  Parcels enabled additional potential indicators of development patterns over time, both in terms of the dimensions of the lots themselves (lot area) and in the relation of portion of the lot occupied by the building (see Section     40 Model 2: Two-dimensional building attributes  Two-dimensional information related to individual buildings was derived from the building outlines.  In this study, building outlines were manually delineated from the LiDAR, however in many cases aerial photo interpretation is used to generate building outlines across an entire city (Shufelt, 1999). As a result, two-dimensional building attributes are considered independent of the unique three-dimensional attributes that can be extracted from LiDAR datasets.  The building outlines enabled the generation of the following two-dimensional building attributes: area, perimeter, number of walls ? determined as the number of vertices defining the two-dimensional shape of the building, primary axis ? determined as an angle between 0 and 90 degrees formed by the two consecutive vertices with the greatest distance between them, and the fractal (Herold et al., 2002) ? determined as the building perimeter divided by its area. Model 3: LiDAR-derived attributes of building morphology  The relation between energy use and a building?s geometry is typically associated with volume and physical interactions with outdoor environments.  However, building morphology has also been suggested as a potential indicator of age (Neidhart & Sester, 2004), which has been used in previous studies to help determine important energy performance parameters (Aydinalp et al., 2004; Farahbakhsh, 1998).  Building morphology was therefore further investigated to better understand the relation to building age, including spatial variations and implications for building energy modelling efforts.   41  In this chapter, all building morphological characteristics were generated from the nDSM.  A total of 16 attributes were generated (Table 3.1) for the 3282 residential buildings assessed in the study area.  Basic morphological features derived from the nDSM were height (mean, maximum and standard deviation) and volume (v), both determined by summarizing the cells contained within the building footprints.  Similar to the heights derived from the nDSM, measures of intensity included the mean, maximum and standard deviation also summarized for those cells within the building footprint.  Despite limitations for deriving actual surface reflectance (Yan et al., 2012), LiDAR intensity values were included to provide a basic indication of roofing material or roof conditions.    42 Table 3.1. Mean, range and coefficient of variation (CV) of attributes used as predictor variables. Attribute Mean Range CV Model 1 Zone  -  -  - Lot area 557.7 138.5 ? 10018.3 1.38 Model 2 Area (m2)  137.9  51.8 ? 1444.9  0.77 Perimeter (m) 47.0 29.9 ? 262.1 0.31 Number of walls 4.2 4 ? 16 0.21 Primary axis (?) 70.0 0 ? 90 0.65 Fractal 0.38 0.12 ? 0.58 0.16 Model 3 Height ? mean (m)  6.3  3.2 ? 32.8  0.22 Height ? max (m) 8.8 4.2 ? 39.3 0.27 Height ? stdev  1.7 0.27 ? 6.9 0.36 Volume (m3) 924.8 267.8 ? 16423 1.11 Slope ? mean (?) 33.0 5.8 ? 53.8 0.22 Slope ? max (?) 68.2 27.4 ? 86.2 0.11 Slope ? stdev  16.2 4.6 ? 31.8 0.23 Complexity 10.1 0.41 ? 23 0.33 Surface area 366.9 145.9 ? 5496.9 0.78 Compactness 3.9 2.5 ? 10.5 0.18 Wall area 207.8 61.5 ? 1672 0.52 Intensity ? mean 32.5 2.0 ? 198.9 0.71 Intensity ? max 141.6 6.0 ? 255 0.38 Intensity ? stdev 23.7 0.9 ? 90.4 0.53 Model 4 Building-to-lot area*  0.28  0.02 ? 0.87  0.33 *Included in addition to all attributes from models 1-3   The LiDAR analysis also included a suite of more sophisticated measures of building morphology.  The first set of these attributes related specifically to the building roof geometry and included measures of slope and complexity.  Using the DSM as input, the slope of each roof cell was calculated using the third order finite distance algorithm presented in Zhou and Liu (2004).  Measures of slope were then summarized for all cells contained   43 within the building footprint and included measures of the mean, maximum and standard deviation.  Slope was also used to provide an indication of roof complexity.  To capture variations in roof shape a spatial filter was passed over each cell within the building footprint and the standard deviation of the slope was extracted based on a 3 by 3 cell window.  Complexity was then determined as the median filtered value for all roof cells.  In urban morphological studies, building shape factors have been developed as a descriptor of texture and employed to assess potential heat loss (Ratti et al., 2005).  The shape metric compactness was therefore also computed here, which is used as an indication of the amount of exposed building per unit volume.  To assess compactness the surface area of the building must first be calculated.  Surface area (SA) was calculated from the array of slope values as:                           (3.1)  where c is the cell size and s is the slope of the pixel i of n cells contained in the building footprint.  External surface area provides a measure of the building envelope by correcting for the three-dimensional form of a pixel represented in the two-dimensional space of the digital surface model and was included as a variable for our analysis of building morphology.  Once surface area was calculated we then use equation 3.2 to determine compactness (C) as:                    (3.2)    44 where v is volume.  The final attribute of building morphology included as part of the LiDAR analysis was the area of walls.  To compute wall area for an individual building both the DSM and DTM were integrated with the building footprints.  As described above, a wall is defined as two consecutive vertices from the building footprint.  At both vertices defining a wall, heights were derived from the DSM and the DTM.  A wall can then be defined in three-dimensional space by the 4 edges (a, b, c, d) connecting the new vertices drawn from the digital surface models.  Wall area (Kw) is then computed using Bretschneider?s formula as:                                            (3.3)  where s is the semiperimeter of the wall and ? and ? are two opposite angles formed by da and bc respectively. Model 4: Combined attributes  The final model combined all the attributes generated above for input as predictor variables into the random forests regression machine learning algorithm.  In addition, a ratio of the building area to that of the lot was added and calculated as the building area divided by the lot area ? and therefore necessitated on the earlier generation of attributes in models 1 and 2.       45 3.2.4 Random forests  Random forests is an ensemble machine learning classification and regression method introduced by Breiman (2001).  The approach used in random forests extends bootstrap aggregation methods (Breiman, 1996) to function with random feature selection.  The random forests procedure operates by first establishing a new set of values equal to the size of the original observed data, with each value randomly selected from the original dataset with replacement (bootstrapping).  Each new set of values then goes through a sequence of binary splits that results in a decision tree.  At each node in the decision tree the split is determined by selecting the variable and value that minimizes the error.  In the case of random forests, a random subset of predictor variables is used to determine the split at each node.  In the case of regression, predicted values are provided as an aggregate of all trees by computing the mean value.  By using bootstrap aggregation techniques, random forests enhance prediction accuracy and enable estimation of error rates and variable importance (Breiman, 2001).  Error and variable importance are both determined using the omitted values from each bootstrap sample, referred to as out-of-bag (OOB) data (Guo et al., 2011; Breiman, 2001).  To derive an error rate, the OOB data is pushed down each tree and compared to the predicted value.  This process therefore acts as an unbiased surrogate for cross-validation (Cutler et al., 2007), and is generated internally as part of the random forests procedure.  The OOB is also used to determine variable importance.    46 Understanding the role of individual variables within random forests of hundreds of trees presents a challenging task (Walton, 2008).  Nonetheless, several approaches have been developed to facilitate the communication of important outcomes from machine learning algorithms.  The first technique, variable importance, is used to gain an understanding of the relative influence of a predictor value within the overall prediction model from the random forests.  To do this, the OOB values for each variable in a bootstrap sample are randomly permuted.  The importance is then measured as the difference between the prediction accuracy before and after the permutation and averaged over all the trees (Guo et al., 2011).  In this case, the magnitude of the decrease in prediction accuracy indicates the importance of a predictor variable.   A second option for interpreting the role of individual variables in random forests is to graphically depict their marginal effects using a partial dependence plot (Friedman, 2001).  Partial dependence is determined as the expected random forests result with respect to all variables except that being tested.  To produce an estimate of partial dependence, all the values of the variable in question are fixed while the prediction function is averaged for all the combinations of observed values for the other predictor variables in the dataset (Cutler et al., 2007).  The resulting output of the partial dependence plot enables interpretation of the relative range and average trend of output values for a given predictor variable.   In this study, the random forests regression, model error, variable importance, and partial dependence plots were all generated using the randomForest R package (Liaw & Wiener, 2002).  Three random forests regressions were prepared to represent the increase in   47 sophistication of municipal spatial data, while a fourth model was then added that combined all variables from models 1-3 in addition to building-to-lot area.  Default values were used to select the number of randomly selected variables at each split (total number of variables / 3) and the number of trees to grow (500).  A selection of partial dependence plots were prepared for selected predictor variables from model 4 that demonstrated noteworthy trends over time.  Finally, the y-axis of the partial dependence plots was scaled to the mean building year to facilitate interpretation.  3.3 Results  The range of values for the predictor variables extracted from the 3282 buildings (Table 3.1) demonstrates the diversity of residential building geometry and design elements in the city.  For most variables, the mean value tends to be located towards the lower end of the range.  The main exception to this trend is the maximum slope of the roof, where the mean is located closer to the upper end of the range.  The dispersion of the variable values is also depicted in Table 3.1 by comparing the coefficient of variation, and shows that the variables with the greatest dispersion include lot area, volume, area, and surface area, while the more homogeneous variables include the maximum roof slope, fractal, compactness, mean height and mean roof slope.  Comparison between the variance explained and error (in years) associated will all 3282 building for each of the models is provided in Table 3.2.  As expected, increases in the sophistication of predictor variables results in a better ability of the random forests regression to explain the variance, while model error is also decreased.  Comparison of the first two models shows only small differences between explained variance   48 and error.  In comparison to models 1 and 2, the use of LiDAR-derived attributes demonstrates a much-improved model outcome with 33.5% of the variance explained and a model error (RMSE) of 16.8 years.  Combining all the variables provides the best prediction of building age with 40.9% of variance explained and a model error of 15.8 years.  Table 3.2. Variance explained and error (years) associated with each random forest regression model for all 3282 buildings.  R2 RMSE Model 1 8.8 19.6 Model 2 10.3 19.5 Model 3 33.5 16.8 Model 4 40.9 15.8    Figure 3.3. Variable importance plotted from Random Forests regression using variables derived from various municipal spatial datasets.   49 The random forests regression technique also enables interpretation of the role of each variable within the model using variable importance measure and partial dependency plots.  Variable importance provides an indication of the relative dependence of the model results for each predictor variable.  Variable importance is depicted in Figure 3.3 for model 4, which includes the entire suite of predictor variables derived from the various data sources.  The most important predictor of building age was mean height, which was derived from the LiDAR data.  Following mean height in importance were the building-to-lot area and lot area variables, each of which is related to less sophisticated two-dimensional building or lot attributes.  It is useful to also highlight the importance of the zone classifier that follows building-to-lot area and lot area in rank, as it provides a simple categorical variable generated from more traditional municipal spatial data.  Finally, it is noted that the majority of the predictor variables resulted in an increase in model error greater than 10% after permutations, with the exception of building primary axis and the number of walls, both of which provided little contribution to the explained variance in the model.  Partial dependence plots for select variables were prepared to aid in identifying the role and trends of individual predictor variables in the random forests (Figure 3.4).  The variables selected included mean building height, mean building volume, mean roof slope, compactness, the building-to-lot area and the mean LiDAR intensity.  Both the mean building height and building volume show strong positive linear trends when predicting age.  This suggests that over time buildings have become taller and larger.  Mean roof slope shows a much different trend.  Approximately 60% of the values fall between 30 and 40 degrees and demonstrate a negative linear trend, indicating that the steeper roofs in this range are   50 associated with older buildings.  Patterns outside this range are less discernable, although there is a general overall trend of steeper roofs with older buildings.  Compactness demonstrates an interesting split at a threshold of about 3.5.  Approximately half of the values are less than this threshold and show a positive linear trend, while the second half of the values greater than 3.5 show a negative linear trend.  The compactness plot therefore indicates that newer buildings are associated with a compactness value of about 3.5 and as the compactness moves further from this threshold the variable predicts an increase in building age.  Of the additional attributes not directly associated with building morphology we examined building-to-lot area and the mean LiDAR intensity.  Building-to-lot area demonstrated high importance in predicting age and shows a strong positive linear trend in the partial dependence plot, indicating that newer buildings tend to maximize the space available on the lot.  Finally, the mean LiDAR intensity provides an example of a weak predictor variable, showing a small positive linear trend with intensity values between 20 and 75, which captures about 80% of the range of values.    51  Figure 3.4. Partial dependence plots of select predictor variables.   3.4 Implications for building energy models and simulations  Results from this research reveal interesting relations and trends between building age and various spatial attributes generated from a diversity of municipal planning data.  The major strength of utilizing building age to guide the development of building energy models is the ability to inform assumptions around the thermal performance and mechanical system   52 attributes of a building.  Examples may include insulation, glazing, ventilation, construction material, and the efficiency of devices such as water heaters and furnaces.  Existing building energy models and simulations tend to group building vintages into categories based on similar structural or technological characteristics and provide an average estimate of the energy parameters associated with the typical system components for a given age range  (Aydinalp et al., 2004; Crawley et al., 2008; Farahbakhsh et al., 1998; Rylatt et al., 2003).  While this method has proved useful with common spatial data, the technique still requires knowledge of building age and must reduce the initial set of data, neglecting informative variations in energy parameters.  The prediction of building age therefore offers a potential advantage over more simple classification schemes, in that more of the variance of the energy parameters may be derived from continuous estimates of building age.  From these outcomes, the capacity to assess a large number of variables, and the resulting error and prediction assessments from the model presented in this chapter, it is concluded that random forest regression provides substantial advantage over more typical classification routines when determining building age, and ultimately building performance.  As a result, predicted building ages derived using this approach will be used to supplement building assessment data for parcels missing age assessments and integrated into a comprehensive building energy demand model discussed in Chapter 5.      53 4 Modelling Building Envelope Irradiance  4.1 Introduction  Radiation received at the urban surface is highly variable in space and time resulting from the complex form and land cover of urban environments.  Understanding this variation in intercepted solar radiation is fundamental for determining building envelope solar gains.  The development of models that accurately capture the spatial and temporal variation of surface radiation on different facets of the urban surface is therefore an important step in the advancement of building energy modelling and simulation.    The general objective of this chapter is to present approaches to estimate the irradiance received on the building envelope.  To accomplish this objective, buildings are separated into two surfaces, roofs and walls.  In the first section of this chapter, a shortwave radiation model is presented that integrates various remote sensing datasets to establish irradiance on horizontally oriented surfaces (roofs) in the urban environment.  Focus is given to LiDAR-based assessment of self-shading, solar obstructions and shortwave radiation transmission through trees.  In the second section of this chapter, shortwave irradiance is calculated for building vertical surfaces (walls) by integrating LiDAR with two-dimensional building footprints.    54 4.2 Section 1: Integrating remote sensing data to estimate irradiance on building horizontal surfaces  The complexity of built form and land cover has often limited detailed model estimates of solar irradiance in the urban canopy.  These limitations arise largely from difficulties in generating contiguous detailed spatial representations of solar obstructing features such as buildings and trees.  This section presents a method that integrates remotely sensed datasets across spatial scales for estimating irradiance on building roofs.  The method draws largely on airborne LiDAR technology with specific focus on estimating radiation transmission through urban vegetation.  The incoming irradiance above the urban canopy is determined using atmospheric transmission derived from geostationary satellite imagery to provide a long-term record of seasonal fluctuations in cloud cover, and the associated direct, diffuse and reflected irradiance.  Individual model components are analyzed for three urban study areas to assess the interactions between input model parameters and patterns of different urban form.    4.3 Study area and data  To assess the model predictions, three urban areas representing a range of building form and tree cover were selected in the City of Vancouver, Canada (Figure 4.1).  The study sites each comprise an area 500 m by 500 m, with available LiDAR coverage extending 200 m beyond the edges of the site.  Each site is located in a unique Vancouver neighbourhood and selected to represent zones of urban form found in many North American cities.  The Fairview site   55 (site A in Figure 4.1) represents a medium density mixed-use neighbourhood with many multi-unit residential buildings, and a few tall institutional buildings including Vancouver General Hospital (Local Climate Zones (LCZ) 1 and 2).  Trees in this study site vary in size, including many shrubs and street trees, with heights typically lower or equal to the building stock.  The Mount Pleasant site (site B in Figure 4.1) represents a low-to-medium-density heritage neighbourhood with large, primarily single family residential dwellings (LCZ 3).  Trees in the Mount Pleasant site are abundant on both public streets and on private lots, many of which reach heights greater than 25 m, substantially taller than existing building stock, with subsequently large canopies.  The Marpole site (site C in Figure 4.1) represents a low-density suburban neighbourhood with small, primarily single family residential dwellings (LCZ 6). Trees in the Marpole site are sparse, located predominantly along streets, reaching heights of around 5-10 m, similar in size to the buildings.    56  Figure 4.1. Location of study sites, Fairview (A), Mount Pleasant (B), Marpole (C) and the meteorological tower within the City of Vancouver.   Topographic information was provided using the LiDAR derived digital surface model (DSM) for the local environment, and using Shuttle Radar Topographic Mission elevation data to represent the mountainous terrain in the region.  Geostationary Operational Environmental Satellite (GOES) data were used to assess atmospheric conditions.  Finally, ground instrumentation was used to train satellite-based estimates of atmospheric transmission.  Measurements of shortwave global irradiance were taken at continuous 30 s intervals from January 2009 to December 2010.    57 4.4 Method  The following section describes the unique components of the solar modeling algorithm employed in this portion of the Chapter, and follows the workflow presented in Figure 4.2.  The first step examines established remote sensing-based approaches for quantifying the effects of shortwave irradiance attenuation as it passes through the atmosphere, followed by a description of the obstructing feature algorithm employed to determine shading in the second step.  The third step presents a novel technique for assessing radiation transmission through the urban vegetation using LiDAR returns classified as vegetation.  Finally, the solar position algorithm is presented followed by an explanation of how each component parameter is integrated into the overall model to provide estimates of direct, diffuse and reflected irradiance.      58  Figure 4.2. Process workflow for integrating remotely sensed data across spatial scales to model irradiance.   4.4.1 Atmospheric transmission  Radiation transmission through the atmosphere is typically separated into turbidity and cloud cover effects.  Turbidity effects include absorption due to aerosols and atmospheric gases (Kasten, 1996) that are highly variable over time and space and not often measured at the desired resolution for application to urban solar modeling.  Nonetheless, broad scale maps (50 km) of turbidity (TL) have been developed by combining estimates of global irradiance   59 with satellite measurements of aerosol data (Remund et al., 2003).  This data provides current best estimates of atmospheric turbidity at a global coverage, and are available as mean monthly values.  The greatest influence on radiation attenuation in the atmosphere generally results from cloud cover (Hammer et al., 2003).  Estimates of cloud cover have been correlated to cloud index measurements acquired from geostationary satellites (Ineichen & Perez, 1999).  Cloud index is calculated as the difference in brightness value at the satellite sensor in relation to the ground reflectance brightness derived from the local minimum value over an established period of time.  A relationship is then calculated between cloud index and cloud transmission (Hammer et al., 2003) to provide the atmospheric clearness index (Kt). Estimates of diffuse radiation fraction from an isotropic sky (Muneer, 1990) can be further extracted using an identified relationship with Kt (Jacovides et al., 2006).     Statistically characterizing the cloud cover over a specific geographical location is important for accurate solar resource assessment. It is possible to generate time series datasets for cloud cover by analyzing visible range GOES satellite imagery and extracting the cloud index. The raw satellite images are first preprocessed to calibrate brightness, adjust geolocalisation and minimize satellite measurement instrument artifacts.  Cloud indices are then computed for each image based on the variations of brightness of each pixel over time, accounting for ground albedo and handling snow cover (Gurtuna & Pr?vot, 2011).    60 Lastly, instrument shortwave irradiance measurements collected from the local meteorological tower and averaged over 5-minute intervals are used to correlate the satellite derived cloud index with measured atmospheric clearness.   4.4.2 Surface viewshed calculation  The use of digital elevation models to identify radiation obstructions has been described in previous studies (Rich et al. 1994; Kumar et al., 1997).  A common approach uses a hemispherical viewshed algorithm to determine the apparent horizon for a set of azimuthal directions around a location on the surface.  The algorithm calculates the elevation angle between the location and the tallest obstructing feature for each direction using surface orientation and slope generated from a digital surface model.  This information is then used in conjunction with solar zenith and azimuthal position to determine shading of direct radiation.  Selecting appropriate search radius and number of axes for viewshed generation is critical for accurate shading estimates in complex urban terrain (Sander & Manson, 2007).    The suite of elevation angles generated at each location can also be used to provide estimates of diffuse radiation. Sky view factor is a measure of the total hemispherical sky unobstructed by surrounding features and is commonly used to estimate the diffuse component of radiation at a location on the surface (Grimmond et al., 2001).  Incorporating the inclination of the surface with the set of elevation angles produced from the hemispherical viewshed analysis, the surface sky view can be calculated according to Oke (1987) as:    61                             (4.1)  where ?o is the elevation angle of the obstruction and N(?o) is the total number of aspects for which an elevation angle is calculated.  In this study, the LiDAR generated DSM was used to calculate surface orientation, slope and apparent horizon, setting the search radius at 200 m for 36 axes, while both the LiDAR and SRTM surface models are used to determine solid solar obstruction angles and sky view factor.  The algorithm assumes no growth of building volume with height and no gaps below the building surface exposed to the sensor.  4.4.3 Vegetation transmission  The ability of discrete return LiDAR to produce multiple signal returns from a single emitted laser pulse facilitates the extraction and representation of vegetation structure.  In vegetation canopies only a fraction of the laser pulse is intercepted, allowing the laser to penetrate deep into the canopy, returning several signals until the remainder of the initial pulse reaches the ground.  This phenomenon facilitates the classification of vegetation features and enables the direct measurement of the vegetation extinction coefficient. LiDAR vegetation extraction  Since many features in an urban environment are solid objects, analyzing the secondary non-ground returns from the LiDAR dataset provides an effective technique for classifying vegetated features.  However, secondary returns are also produced when the pulse is partially   62 intercepted by a solid feature, such as the edge of a building (Goodwin et al., 2009).  Therefore, additional analysis is required to accurately separate vegetation using the secondary LiDAR returns.  One method for separating vegetated and non-vegetated features from the secondary LiDAR returns is to use spatial filtering techniques.  These techniques exploit the horizontal linear trajectory of secondary returns to separate solid objects, such as building edges, from vegetation.  In previous work, Goodwin et al. (2009) used a spatial filter to identify secondary returns along a series of radially projected vectors.  Linear features were determined based on the number of intercepted returns along a single vector as a fraction of the total secondary returns intercepted for all vectors in a given filter.  Identified linear features were then removed and a morphological filter was applied to fill gaps based on the surrounding area identified as vegetation.  Results of this vegetation extraction technique produce a binary array representing vegetated (above-ground) and non-vegetated cells.   In this section, all the selected parameters are based on previous approaches for similar locations within the City of Vancouver (Goodwin et al., 2009), the results of which demonstrated both strong statistical and visual representations of urban vegetated features.  The linear feature spatial filtering technique projected 16 vectors through a 7 by 7 cell window applied to the 1 m gridded surface layers generated from the LiDAR.  Linear features were determined when more than 30% of a given vector intersected a secondary return cell and when the same vector intersected over 50% of all the secondary returns in the filter window.  Extracted linear features were then removed from the initial secondary return   63 grid and a 5 by 5 cell morphological filter was used to erode and grow vegetation classified cells using population thresholds of less than 25% and greater than 56% respectively.  The resulting binary layer was used to identify LiDAR vegetation returns for modelling vegetation transmission using the gap probability approach presented in the following section. LiDAR measures of vegetation attenuation  Analysis of the non-ground LiDAR data points to determine structural information of vegetation has been well described in forestry applications (Lovell et al., 2003; Ria?o et al., 2003; Ria?o et al., 2004; Coops et al., 2007).  These techniques use gap probability models to determine canopy structure by correlating the attenuation of the laser pulse with the density, size, and distribution of foliage and woody elements (Ni-Meister et al., 2001) and are used to derive a cumulative vertically projected profile of the foliage in an entire canopy.    In the approach here, a modified version of the gap probability model is presented to provide an estimate of radiation attenuation as a function of the intercepted extinction profile for individual cells (Figure 4.3).  This method involves first estimating an extinction coefficient (Kv) as the proportion of LiDAR returns at each height interval by:                     (4.2)    64 where N(z) is the number of LiDAR returns at a height z above the ground, and N is the total number of independent LiDAR pulses.  Due to the range of vegetation types and structure in urban areas, we assume a random leaf orientation, resulting in a static extinction coefficient for each cell regardless of angle.  Various distributions can be fitted to the vertical profile to smooth gaps and provide a summary of the extinction coefficient as a function of height through the vegetation canopy.  A Weibull distribution function is commonly used for its ability to characterize vegetation structure in a variety of tree species (Coops et al., 2007).  A second step therefore was used to fit a two parameter Weibull probability density function (Wpdf) to the extinction profile as:                                           (4.3)  where ? is a vertical scaling parameter, ? is a shape parameter that alters the breadth of the distribution and       is the extinction coefficient for each height interval (1 m intervals up to the maximum tree height in this study).  The ? and ? parameters were estimated using a Levenberg-Marquardt least-squares analysis for LiDAR returns within the subset of cells identified as vegetation.  The Weibull parameters were then used as input to the radiation model along with the local solar path length to determine the direct beam transmission through the vegetation. The gap probability approach presented here applied a 3 by 3 cell window and a minimum height threshold of 2.5 m for generating vegetation extinction probability density functions for individual cells.    65  Figure 4.3. Representation of the key steps for estimating the vegetation extinction coefficient.    4.4.4 Urban irradiance estimation Solar position algorithm  The ENEA solar position algorithm (Grena, 2008) provides for the computation of solar position with a maximum precision error of 0.0027?, valid between the period from 2003-2022, and with computational time comparable to recent ?fast algorithms? (Michalsky, 1988; Blanco-Muriel et al., 2001).  Inputs to the algorithm include fractional Universal Time, date, difference between Universal Time and Terrestrial Time (?t), longitude, latitude, air pressure, and air temperature.  Output of the ENEA algorithm produces local topocentric coordinates of solar altitude (?s) and solar azimuth (?s) angles.    66 Integrated model implementation  Each of the modeling components described above were integrated to provide estimates of direct and diffuse shortwave irradiance.  Analysis was conducted for each cell in the LiDAR generated grid of surface elevation values.  Additional radiation parameters internal to the model presented in this section were computed following the technique presented in Hofierka & Suri (2002).  These parameters included calculation of extraterrestrial radiation, corrected atmospheric turbidity factor (Kasten, 1996), Rayleigh optical thickness (Kasten, 1996), approximated relative optical air mass (Kasten & Young, 1989), and a diffuse solar altitude function (Scharmer & Grief, 2000).  Integrating the various model parameters to predict direct shortwave irradiance on an inclined surface (Ibt) was computed as:                                                   (4.4)  where ?i in the incidence angle between the surface and sun (accounting for self-shading), I0 is the extraterrestrial radiation in W m-2, m is the relative optical air mass (see Hofierka & Suri (2002) equation 5), Rm is the Rayleigh optical thickness at air mass m (see Hofierka & Suri (2002) equations 8 and 9), Kt  is the atmospheric clearness index, and    is the vegetation attenuation determined along the path ?s  as:                 for ?s > ?o          (4.5)   67  at each intercepted vegetation cell i, j with path length   calculated as:                    (4.6)  where c is the size of cell (1 m in this study).  Along with predetermined Weibull ? and ? parameters (section 3.2) the z value intercepted at each cell i, j was used to populate Wpdf (equation 4.3) and determined as:                           (4.7)  where di,j is the horizontal distance to intercepted vegetation cell i, j and ?z is the ground-normalized relative difference in height between the cell being analyzed and intercepted cell i, j calculated as:             ?                  (4.8)  where z(DSM) is the surface height at the analyzed cell and  z(DEM)i,j is the ground elevation of the intercepted cell i, j.  In this study the maximum distance at which intercepting vegetation cells were assessed to derive    was limited to 200 m along a horizontal plane.  Diffuse radiation on a horizontal surface (Idh) was computed as:   68                               (4.9)   where    is a diffuse tranmissivity function (see Hofierka & Suri (2002) equation 22), and Fd is a solar altitude function (see Hofierka & Suri (2002) equation 23).  To minimize computational processing time this approach assumed that the diffuse radiant intensity is isotropic across the entire sky hemisphere.   Finally, reflected irradiance from surrounding objects Irw, integrated the sky view factor calculated in equation 4.1 with a generalized system albedo factor for the urban environment of ?u = 0.15 (Oke, 1987, 1988) and was computed for a wall surface as:                           (4.10)  where Ih is the direct irradiance on a horizontal surface, which was calculated by setting ?i in equation 4.4 to the local horizontal elevation angle of the sun.  4.4.5 Case study  The integrated irradiance model was applied for each study site at 30-minute time intervals on the 15th day of each month.  Model outputs of direct, diffuse and reflected irradiance were stored separately, in addition to information related to obstruction angles and vegetation   69 features.  Model components were analyzed independently to assess the interaction of input parameters with urban form characteristics at each study site.    Study sites were contrasted with respect to the key components of the model parameterization.  This assessment included a temporal and spatial comparison of the vegetation-shading impacts, obstruction angles and distances, and the estimated irradiance output.  Since the coarse spatial resolution of the GOES imagery does not distinguish between the proximity of the study sites, atmospheric effects were assessed for temporal variations exclusively.  The vegetation transmission was analyzed by comparing the transmission results of the modified gap probability algorithm with opaque representations of all obstructing features.  Vegetation was also assessed alongside the solid features in the surface viewshed calculation to compare the morphological characteristics of each site.  These results were also used to help identify the appropriate search radius and number of axes for determining obstruction metrics.  Finally, the direct and diffuse components of modeled irradiance were compared to further reveal differences in the influence of urban form between sites.   4.5 Results  Comparison between the clearness index as calculated from ground-based measurements of shortwave irradiance for the years 2009 and 2010 and subsequent cloud index generated from GOES satellite imagery demonstrated a strong correlation (r2 = 0.73, p < 0.001, RMSE = 0.13 (Figure 4.4)).  Applying the regression coefficients to the entire 10-year archive of hourly   70 GOES imagery thereby provides an indication of representative atmospheric transmission for the study area.  Figure 4.5 shows monthly mean clearness index for hourly intervals and the associated variance.  Clearness index values range up to 0.7 at midday in the summer to a low of 0.2 during winter months and mornings.  The greatest variance in clearness is observed during the months of April, May and September and generally before noon, while winter months (from November to February) display more consistency in cloud cover across the 10-year period from 2000-2010.    Figure 4.4. Scatterplot of observed atmospheric clearness index versus predicted clearness from GOES satellite-derived cloud cover.   Cells classified as trees from the LiDAR-based vegetation extraction technique varied across sites.  Greatest tree coverage was apparent at the Mount Pleasant site with ?v,T = 33% of the total plan area classified as tree, followed by Fairview at ?v,T = 15%, and Marpole at ?v,T = 12%.  Differences were also observed with respect to vegetation structural metrics.  Table 4.1   71 lists the structural information for each study area.  Again, Mount Pleasant has the greatest tree height and tree height variation, followed by Fairview and Marpole.   Figure 4.5. Average hourly clearness index (left) and standard deviation (right) for each month derived from the GOES cloud index for the years 2000-2010.    Table 4.1. Vegetation height information for study sites (m).  Fairview Mount Pleasant Marpole Vegetation Height (mean) 6.77 7.85 4.40 Vegetation Height (med) 6.27 6.45 3.91 Vegetation Height (stdev) 5.09 6.22 3.80   The two-parameter Weibull function was used to estimate radiation attenuation for each classified vegetation cell and demonstrated good fit between all study sites with a mean RMSE for all analyzed cells of 0.047 (standard deviation = 0.032).  To demonstrate the   72 typical transmission (1 -     a representative value for the alpha parameter along with associated beta value are calculated for each site.  An ? and ? value of 8 and 7, 9 and 8, and 6 and 8 was assessed for Fairview, Mount Pleasant, and Marpole, representing 10%, 6%, and 19% of the total vegetation cells respectively.  Figure 4.6 demonstrates the characteristic Weibull function for each site, in addition to diurnal variations in vegetation transmission for the summer solstice, winter solstice and equinox calculated at a 10 m distance from the cell (approximately equal to the vegetation height).  Results indicate a substantial difference in radiation absorption close to the equinox(es), more concentrated vegetated influences in the morning and afternoon for the summer solstice, and little difference between sites for the winter solstice.  Irradiance models incorporating vegetation transmission show an increase in irradiance of 6.4% for Fairview, 18.1% for Mount Pleasant, and 2.2% for Marpole compared to the representation of vegetation as opaque features.    73  Figure 4.6. Vertical distribution of vegetation transmission representative of each study site (a), and daily vegetation transmission on the equinox(es) (b), summer solstice (c) and winter solstice (d).      74  Figure 4.7. Average azimuthal obstruction angles (top) and distances to obstructing features (bottom) for the study sites of Fairview (a), Mount Pleasant (b) and Marpole (c). Green wedges show results for each azimuthal direction when trees are included, while grey wedges show results using buildings and solid obstructions only.  The vegetation was also compared with non-vegetated obstructing features for the outcomes of the surface viewshed calculation.  Figure 4.7 shows the average obstruction angle and average distance to the obstructing feature for each axis.  Maximum obstruction angles for vegetation are found at the Mount Pleasant site, while non-vegetated obstructions are greatest at the Fairview site. Average obstruction angles are also generally equal in all directions.  However, in contrast to the symmetry of obstruction angles, the study sites show a pronounced increase in distance to the obstructing feature in the East-West direction, while nearest distances typically occur North-South.  The mean distance to the obstructing feature is 35 m, 32 m, and 26 m for Fairview, Mount Pleasant, and Marpole, with standard deviations of 45 m, 41 m, and 35 m, respectively.  Finally, at each site trees represent the   75 closest average obstructing feature for every analyzed axis, except for a few directions in Marpole where non-vegetated features are slightly closer than trees.  4.6 Discussion  Representing the transmission of radiation through the atmosphere and the vegetation are demonstrated as essential components for generating accurate irradiance estimates in urban areas.  Due to the wealth of literature already examining satellite-based radiation and atmospheric transmission estimates (Perez et al., 2002; Hammer et al., 2003; Rigollier et al., 2004; Zarzalejo et al., 2009), and the much greater control that humans possess in altering the physical components of urban form, this discussion is focused primarily on the vegetation component of urban irradiance modelling.  Nonetheless, the spatial scale of atmospheric transmission estimates and the unique influence of turbidity on the urban shortwave irradiance attenuation (J?uregui & Luyando, 1999) remain important considerations for further study.  Results from Figure 4.8 suggest that urban irradiance modelling requires integrating assessments of the semi-transparent nature of vegetation in order to provide better irradiance estimates.  The analysis in this section indicates that representations of vegetation as opaque features result in substantial overestimates of solar obstruction angles.  The overestimation of solid obstructions is associated with an 18% underestimate of total irradiance in the treed residential neighbourhood of Mount Pleasant.  Neglecting vegetation transmission also results in an underestimate of total irradiance for the study sites of Fairview (6.4%) and   76 Marpole (2.2%); areas selected to represent the general range of North American urban form types.    The implications of urban form are further revealed when considering the result of the surface viewshed analysis presented in Figure 4.6.  Examining the spatial details of the radially intercepted features provides insight into the morphological character of each study site.  The non-symmetrical average distance to obstruction objects highlights the gridiron plan of the sites, with the major axis oriented North-South.  This result can be attributed to obstructing features aligning with property orientation, while also indicating more homogeneous heights in the East-West direction.  It is also important to note that these outcomes will likely differ in non grid-aligned cities, as is the case in many non-North American cities, providing for an interesting comparison in future studies.    77  Figure 4.8. Direct (top) and diffuse (middle) irradiance model outputs and the difference in irradiance (%) between opaque and semi-transparent representations of vegetation (bottom) for the study sites of Fairview (a), Mount Pleasant (b) and Marpole (c).  Results from the analysis of obstruction angles also reveal the importance of selecting an appropriate parameter for the scope of the surface viewshed calculation. The greatest distances to the obstruction features are observed for a high density, traditionally urban neighbourhood (Figure 4.7).  This is likely a result of the substantial area of rooftops that are elevated to a height above many obstructions.  To reduce computational time for future modeling endeavors, the results presented here suggest that the selection of a 100 m search   78 radius would accommodate over 90% of the variance in obstruction angles across all examined urban form types.   4.7 Section 2. An irradiance model for building vertical surfaces  The objective of this section is to introduce a novel technique to automate irradiance estimates on building walls with computational efficiency suitable for citywide building energy performance modelling applications.  To accomplish this objective, commonly available spatial data representing 2D building geometry (building footprints) are integrated with array-based urban form data (digital surface models).  The new technique utilizes a point obstruction stacking (POSt) algorithm that optimizes the placement of calculation points across building walls, which are then used to determine solar and sky occluding features from the local environment.  To demonstrate the effectiveness of the approach, three unique configurations of point spacing are examined including a single node located at the centroid of each wall, numerous nodes randomly distributed across each wall, and lastly, nodes stacked at regular intervals along the vertical edges of the wall.  To validate this approach, irradiance is compared to a continuous square-meter grid over all vertical surfaces.  4.8 Data  Data used in this section included the LiDAR-derived DSM, DEM and nDSM products and building footprints.  A total of 93 355 primary building units were included in the footprint data and approximately 8% of the buildings were within the area covered by the LiDAR.    79    4.9 Methods  4.9.1 POSt topology   Zlatanova et al. (2004) describe the absence of a universal 3D model for spatial objects and elaborate that the design of any such model must be closely related to the application.  The following section therefore provides a description of the POSt model structure and sets the computational framework for modeling direct, diffuse and reflected irradiance on building walls.  To facilitate this irradiance modelling, a 3D model structure is established based on point obstruction stacking (POSt), in which points were attributed with coordinates in 3D space and positioned across building walls to determine the obstruction angles of neighbouring features.    The basic primitive for storing spatial feature geometry is a node (point), and in the case of building footprints, these nodes depict the vertices that determine the ultimate 2D shape of the building.  Each node initially contains easting and northing coordinates, which provided the foundational base geometry for establishing the POSt model structure.  In addition to these nodes, the POSt topology distinguished edge and face primitives.  As a result, the structure resembles that of other 3D models (Zlatanova et al., 2004) such as the 3D Formal Data Structure (Molenaar, 1990) and the Urban Data Model (Coors, 2003).  Before an edge was created the nodes inherited height (z) values from the geographically intersecting cell in the surface and terrain raster images.  Edges were then generated horizontally between two consecutive nodes, and vertically between the terrain and surface feature.  The face primitive   80 was then used to define building wall objects based on four bounding edges.  Lastly, nodes defined not only the 2D outline of buildings, but were established as point objects located inside a face (wall) and became the principle element used to determine the obstructing features from the environment outside the wall.  For computational efficiency the POSt topology was mapped to a relational database as depicted in Figure 4.9.   Figure 4.9. Post data structure and mapping within a relational database where each box represents a class with associated attribute values.     81 4.9.2 Wall attributes  From the initial polygon geometry defining the 2D outline of each building, a table was created within the relational database to store wall area, orientation and irradiance values for predetermined time intervals.  All walls were assumed to represent a 90? inclined surface, therefore determination of the wall orientation did not require additional height information for calculation.  A wall orientation ?w was computed from two consecutive nodes from the original 2D building geometry and converted from Cartesian to polar coordinates as:                                  (4.11)   where atan2 is a variation of the arctangent operation common in computer languages to distinguish between diametrically opposite angles, Y0 and Y1 are the northing coordinates and X0 and X1 are the easting coordinates for an initial and subsequent node respectively.  Equation 4.11 assumes all nodes forming the 2D geometry were represented consecutively in a clockwise pattern as is consistent with the Open GIS Consortium Simple Feature Specifications.       Unlike orientation, wall area requires information on heights (z), which were calculated following equation 3.3 (in Chapter 3).    82 4.9.3 Point spacing configurations  Critical to the computational efficiency of the POSt approach is the number and location of the nodes over a face.  These nodes were used as the surface intersecting points for a ray casting procedure that determined the obstruction angle of nearby features.  To test and then validate the spatial arrangement of points in the POSt procedure, three basic node spacing techniques were developed in addition to a validation case.  Each technique is described below and represented in Figure 4.10.    Figure 4.10. Representations of each point spacing configuration over a basic cubic building geometry where (a) is the validation case, (b) the centroid approach, (c) the random approach and (d) the vertical stacking approach.     83 The first technique, centroid, provided the simplest spacing configuration, in which a single node was positioned in the center of a face based on the x, y and z coordinates of the nodes that defined the 4 edges of the wall.  The second technique, random, established a set of nodes positioned randomly across each wall.  The random technique required the user to set either an absolute number of random points, or a number of nodes relative to the area of the wall.  To ensure all walls were adequately covered regardless of area, points were allocated randomly at a ratio of 1 point for each 10 square-meters of wall surface.  The third technique, vertical, stacked nodes at equal intervals along the two vertical edges that define each wall.  In this study, nodes were stacked at each meter from the ground to the maximum height of the vertical edge.  In addition to the three point spacing configurations, a validation case was also generated in which a fine regular grid of nodes was positioned across each wall.  To provide the most consistent and comprehensive validation case, nodes were spaced at the same distance as the cell size of the height array.  As a result, a node was established at every square meter across the wall surface.   4.9.4 Ray casting procedure  Once the nodes were distributed across each wall in all four configurations, obstruction angles were calculated.  To do so a standard hemispherical viewshed algorithm (Rich et al., 1994) was modified to determine the sky-obstructing features based on the elevation values from the raster image of surface heights.  However, unlike standard horizontally oriented   84 models, the POSt algorithm used the 3D coordinates of the nodes as opposed to array cells to project a vector between the wall surface and nearby obstructions.  At each node on a wall the POSt obstruction angle algorithm performed a simple ray casting procedure (Figure 4.11).  Using the point?s x, y and z coordinates a vector was projected incrementally to determine the angle subtended between the location on the wall face and the elevation value at the intersecting cell of the surface height raster.  The maximum angle along the vector was then stored in a table in the POSt relational database.  In our development of POSt we utilized 7 unique rays for an individual node.  The first ray was projected along the same azimuth as the orientation of the wall, while the 6 subsequent rays were projected at ? 22.5? azimuthal increments from the initial vector, avoiding self-shading from the wall itself.  A similar approach of ray casting from a base point on a vertical surface has been previously employed to estimate daylighting and thermal gain through windows (Ratti et al., 2005), however the technique has not been applied to entire vertical surfaces, nor have alternative point spacing configurations been validated to assess efficient model parameterization.  Using the surface height raster derived from LiDAR also provided the advantage of inherently capturing solar occlusions due to trees and terrain since the dataset does not discriminate between surface feature objects but simply represents the maximum elevation across the study area.  Such an approach therefore enabled a highly accurate representation of the local environment surface effects without the need for additional datasets.    85  Figure 4.11. Representation of the ray casting procedure where (a) depicts example locations of cast rays, (b) depicts the azimuthal angles at which rays are cast and (c) depicts the subtended obstruction angle for a given azimuth.  4.9.5 Irradiance modeling Irradiance Model for Point Spacing Configurations  The fundamental processes involved in solar modeling are discussed in Section above.    86 The major difference in the POSt algorithm was the calculation of sky-obstructing features using a shading coefficient (SC).  For any given topocentric solar position where ?s is the solar altitude and ?s the solar azimuth, a shading coefficient was defined as the fraction of points on a wall with an obstruction angle greater than the solar elevation angle and computed as:                                     (4.12)  where {P0-n}? is the set of points defining the obstruction angles as selected from the relational database using the vector orientation with the minimum angular distance to the solar azimuth.  After the calculation of the shading coefficient, direct-beam wall irradiance (Ibw) was computed as:                               (4.13)  where I0c is the extraterrestrial radiation corrected for aerosols (Kasten, 1996) and ?i is the incident angle between the wall orientation and the sun.    The diffuse irradiance was modeled using equation 4.9 where the surface effect component was determined using sky view factor of the wall as a measure of the total (semi)   87 hemispherical sky unobstructed by surrounding features.  Sky view factor       for a wall was computed as:                                                  (4.14)  where {P0-n}?w??/2 is the set of points defining the obstruction angles for all vector orientations within ? 90? of the wall orientation.  Reflected irradiance was calculated using the same method presented in equation 4.10, substituting      for the vertically adjusted skyview       and multiplying by the area of the wall (equation 3.3). Irradiance model for the validation procedure  To assess the accuracy and performance of the point spacing configurations in estimating wall irradiance, a separate irradiance model was developed for the validation case.  In addition to the configuration of point spacing at regular spatial intervals discussed in Section 4.9.3, the validation model differs primarily in regards to the integration of the ray casting procedure within the irradiance model.  To provide a robust representation of the actual irradiance received on a wall, obstruction angles were assessed at each iteration during the solar modeling procedure.  Therefore, instead of selecting stored obstruction values at preset orientations from the relational database, an angle was subtended between each point on the wall surface and the topocentric solar coordinates at a given time interval.  This approach   88 eliminated the need for the integration of the shading coefficient in equation 4 but maintained the same calculation of diffuse and reflected irradiance presented in equations 4.9 and 4.10.  However, performing a ray casting procedure for each time step adds substantial computational expense, discouraging the use of the validation approach for applications requiring vertical surface irradiance estimates for multiple buildings.   4.9.6 Validation   Fifty building footprint polygons were randomly selected from the City of Vancouver building footprints dataset in areas with diverse topographic relief and stratified into single-family dwellings and multistory buildings.  The geometric attributes of the sampled building footprints are listed in Table 4.2.  Due to the grid-aligned street network of Vancouver, the primary axis of the majority of buildings is oriented North-South or East-West.  To facilitate the comparison of irradiance with varying wall orientations all primary axes, and subsequently all walls, were re-oriented to 0?, 45? or 90? polar coordinates.  Furthermore, each building was restricted to 4 unique wall orientations dependent on the building?s primary axis.  For the single-family dwellings these generalizations resulted in a sample split of 18 buildings orientated at 0?, 4 buildings oriented at 90? and 3 buildings oriented at 45?.  Multistory buildings were split between 0? and 90? with 11 and 14 buildings respectively.  An example of the building footprints with their original orientation overlaying the digital elevation model and a 3D depiction of a building as represented in the database is depicted in Figure 4.12.       89 Each of the point spacing configurations as well as the validation case were generated for the sampled building footprints.  The model was then run at 60-minute intervals for the 15th day of each month.  Daily total global irradiance estimates for vertical surfaces were stored in the relational database and compared to the validation procedure to determine overall model accuracy and irradiance values for unique wall orientations.  Furthermore, each computation was timed to enable further comparison of the computational expense between point spacing techniques.    Table 4.2. Geometric attributes of the sampled building footprints.  Single Family Dwellings Multistory Buildings Count 25 25 Height (m) - ? 6.9 17.3 Height (m) - ?  1.4 9.3 Footprint area (m2) - ? 144.4 1435.1 Footprint area (m2) - ? 53.1 973.7 Number of walls - ? 4.5 7.8 Number of walls - ? 0.9 3.7 ? = mean ? = standard deviation    90  Figure 4.12. Example of the surface height raster layers representing (a) the digital elevation model and (b) the normalized digital surface model, in addition to (c) the 2D geometry of buildings and (d) a 3D representation of the selected building from panel (c).   4.10 Results  Results of the validation procedure predicted clear-sky mean daily wall irradiance at 5.04 MJ m-2 day-1 for single-family dwellings and 5.28 MJ m-2 day-1 for multistory buildings.  Model   91 error for all the point spacing configurations is provided in Table 4.3.  Error was greatest with the centroid point spacing configuration with an RMSE of 1.35 MJ m-2 day-1 (27%) for single family dwellings and 1.75 MJ m-2 day-1 (33%) for multistory buildings.  The random spacing configuration showed improvement from the centroid approach with and RMSE of 1.31 MJ m-2 day-1 (26%) and 1.50 MJ m-2 day-1 (28%) for single family dwellings and multistory buildings respectively.  Lastly, when compared to the results of the validation case, the vertical approach, in which nodes were spaced at 1 m intervals along the vertical edges of each wall, demonstrated the lowest residual error for both building types, with an RMSE of 0.84 MJ m-2 day-1 (17%) for single-family dwellings and 1.16 MJ m-2 day-1 (22%) for multistory buildings.  Table 4.3. Model error assessed as total daily irradiance integrated over the entire year for each point spacing configuration (RMSE, MJ m-2 day-1).  Single Family Dwellings Multistory Buildings Centroid 1.35 1.75 Random 1.31 1.50 Vertical 0.84 1.16   Modeled irradiance values for all unique wall orientations are provided in Figure 4.13. As expected, highest mean irradiance values were obtained on walls facing directly south, ranging between 10 ? 11 MJ m-2 day-1.  Comparison of irradiance estimates for unique wall orientations indicated greater variation in mean values for single-family dwellings than multistory buildings.  Furthermore, these differences were more pronounced for wall orientations likely to receiving direct solar radiation (90? ? 270?).  The closest mean   92 irradiance estimates to the validation case were generally obtained from the vertical point configuration, except for single-family dwelling walls oriented at 135? and 270?, where both the centroid and random approaches provided better approximations of mean wall irradiance.  In contrast, examination of the variance in irradiation for wall orientations revealed a greater diversity of results between each point spacing configuration and the validation case.  For single-family dwellings, no clear trend was apparent for determining a preferred configuration to match the variance in irradiance for all unique orientations.  For multistory buildings however, the random approach closely matched the validation values at all 4 modeled wall orientations.   Computational time was also calculated for each of the basic steps in the POSt algorithm (computation resource = MAC OS X, 2.66 GHz Intel Core i5, 8 GB 1067 MHz DDR3).  Results of each point spacing configuration and the validation procedure are presented in Table 4.4 for the selected 25 single-family dwellings, and Table 4.5 for the 25 selected multistory buildings.  Results confirm the substantial computational expense of the validation technique.  Differences in computational expense between point spacing techniques also existed, but were much less pronounced than the comparison with the validation procedure.  For single-family dwellings, the centroid, random and vertical approaches were 294, 194 and 135 times faster than the validation procedure, respectively.  Similarly, the validation comparison for multistory buildings demonstrated a 1259, 252, and 266 times reduction in computational time for the centroid, random and vertical configurations, respectively.  Lastly, results comparing the basic modeling step listed in Tables 4.2.3 and 4.2.4 indicated   93 that the greatest computational expense was associated with the determination of obstruction angles.    Table 4.4. Computational time for the main steps in the POSt algorithm for a typical single-family dwelling (s).  Point Generation Obstruction Angles Irradiance Total Validation 0.83 - 6155.40 6156.23 Centroid 0.65 4.41 15.91 20.97 Random 0.68 15.13 15.98 31.79 Vertical 0.68 25.02 19.99 45.69  Table 4.5. Computational time for the main steps in the POSt algorithm for a typical multistory building (s).  Point Generation Obstruction Angles Irradiance Total Validation 1.75 - 31046.72 31048.47 Centroid 0.61 4.40 19.65 24.66 Random 0.72 75.92 46.64 123.28 Vertical 0.69 70.58 45.26 116.53     94  Figure 4.13. Modelled total daily building wall irradiance over a year (MJ m-2 day-1) for all unique wall orientations comparing the point spacing obstruction techniques with the validation case. Mean daily irradiance values for single-family dwelling and multistory buildings are presented in panels a) and c), while the standard deviation of mean daily irradiance are presented in panels b) and d) respectively.  4.11 Discussion  Studies of irradiance on building vertical surfaces have been conducted primarily in the disciplines of urban climate (Arnfield, 2003) and building energy simulation (Crawley et al.,   95 2008).  As a result, existing analysis and techniques tend to either aggregate vertical surface irradiance within a broader scale analysis, in the example of ?canyons? in urban climatology (Arnfield, 2003), or neglect the form of the local environment, as is the case in many energy simulation models (Crawley et al., 2008; Loutzenhiser et al., 2007).  The technique presented here enables the computation of wall irradiance at a spatial scale, and with a computational structure, that can facilitate model parameterization for both the aforementioned disciplines.  Furthermore, the technique is designed to be computationally efficient over large areas and exploits datasets commonly available to municipal authorities.       The comparison between spacing configurations has important implications for application of the POSt method requiring citywide wall irradiance estimates.  As demonstrated in this study, the validation procedure requires substantial computational time, and thus prohibits execution for all buildings within a city using standard desktop computers.  The City of Vancouver building footprint data for example, includes approximately 93 000 primary buildings, which would require over 7200 hours of processing time using the validation procedure described above.  In comparison, the vertical point spacing configuration, which provided the lowest error estimates, would require just over 50 hours of processing time.  Nonetheless, the validation procedure provides a potential modeling framework for case study subsets within the city, or when high accuracy of irradiance estimates are required, such as with models of urban climate, human comfort and health (see e.g. (Lindberg & Grimmond, 2011).  It is also important to note that processing time could be reduced in all cases by exploiting parallel processing options available with most current desktop computers.   96  The POSt topology, and storage in a relational database, facilitate future model iterations and enhance the range of urban applications.  The basis for the current study described the use of the POSt method for application to building performance and efficiency modeling, however a similar approach is suitable to urban climate modeling (Krayenhoff & Voogt, 2007) for heat-island analysis for example, or to analysis with applications in urban design (Yang et al., 2007), emergency and evacuation planning (Lee, 2007) or virtual 3D city models (Kolbe, 2007).   The primary limitation of the POSt model structure for assisting the prediction of building energy performance is that it restricts geometric representation of protruding or receding features inside a wall.  Similarly, angled or curved walls are simplified to vertically orientated planar surfaces in this approach.  Nonetheless, the intended application of the model is for citywide applications, and complete spatial representations of buildings rarely contain such detailed geometries. Under circumstances where geometric details of building walls are available (e.g. CAD or CityGML datasets), existing building energy simulation software may be a more appropriate modeling option.  Consideration of the accuracy of the existing 2D building geometry is also important when conducting an analysis of wall irradiance with POSt.  Misrepresentation of wall heights from the digital surface models is possible in situations where the defining vertex of a building outline does not align with the appropriate grid cell.       97 4.12 Implications for building energy models and simulations  In this chapter two techniques are presented to better estimate the irradiance on building envelopes.  This first technique provides a universally applicable framework based on existing remote sensing techniques while adding the calculation of vegetation transmission by exploiting the laser pulse heights from raw LiDAR point clouds.  A novel method based on gap probability analysis is used to determine the fraction of incoming radiation that is transmitted through the urban vegetation canopy, and helps to demonstrate the importance of accounting for the semi-transparent nature of vegetated features when modeling urban irradiance.  In the second section an approach is presented to model irradiance on building walls and was designed to function with common municipal datasets that represent the 2D geometry of buildings and raster images of surface heights.    The outcomes of this chapter highlight the importance of accounting for both spatial and temporal variations of individual model parameters in estimates of urban solar irradiance.  As building energy models and simulations continue to evolve, these variations must be considered to accurately inform those aspects of energy planning that require detailed irradiance estimates in the building sector.  By extension, the following chapter examines the integration of solar gains for building energy modelling using the techniques presented above.      98 5 Modelling and Mapping Demand for Building Thermal Energy Services  5.1 Introduction  The objective of this Chapter is to combine the outcomes from Chapters 3 and 4 to develop and present a robust technique for generating comprehensive estimates of individual building energy demand across entire urban regions.  To do this, LiDAR is used to inform three fundamental modeling components.  First, LiDAR is used to generate an accurate representation of building size and shape.  Once this massing information is established for each building, the LiDAR predictions of building age presented in Chapter 3 are related to building energy performance parameters including resistivity values for envelope components and air leakage.  The LiDAR based irradiance models presented in Chapter 4 are then integrated to determine solar gains.  These components are all combined in a novel modelling approach to provide spatially contiguous estimates of thermal energy demand for individual buildings.  Due to the difficulty in establishing empirical validation data, results are compared to outcomes from the Canadian building energy simulation software, HOT2000.    99 5.2 Methods  5.2.1 Data  Data used in this chapter include LiDAR, Canadian residential building energy audits, census population, City of Vancouver cadasters, and local measurement of climate variables from January 2009 to December 2011 on air temperature, soil temperature, incoming longwave radiation and piped water temperature.  The climate data were collected at various time intervals and summarized by mean monthly values as listed in Table 5.1.  Table 5.1. Mean monthly measured environmental data.  Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Air Temperature (?C) 3.9 5.4 6.5 9.0 12.6 14.8 17.4 17.4 14.7 10.3 7.1 3.5 Soil Temperature (?C) 5.7 6.3 7.9 10.7 13.7 16.9 18.7 19.5 17.7 13.5 8.6 5.2 Longwave Radiation (W m-2) 333.1 320.6 315.5 327.4 333.1 356.6 354.8 348.7 363.5 337.5 317.8 313.3 Piped Water Temperature (?C) 5.0 5.5 6.0 8.0 10.5 13.5 15.0 16.0 16.5 14.0 10.0 7.5   5.2.2 Study area  The geographic model developed in this chapter is demonstrated for a mixed-use, primarily residential neighbourhood (Figure 5.1) in the City of Vancouver, Canada, which encompasses buildings with a range of vintages and structural forms.  The study area covers a total of 4 km2 with a census-derived population of 28 167 persons, resulting in a population   100 density of 73 persons ha-1.  The total number of buildings identified within the study area was 4289 with approximately 80% of those buildings designated as containing residential units.    Figure 5.1. Study area located in a residential area of Vancouver. The dark grey regions shows the extend of the area used for analysis to maintain entire census dissemination areas, the light grey areas represent park space, and the buildings with residential units used for the analysis are shown in black.   5.2.3 Modeling approach  The technique presented here focuses on estimates of building thermal energy services for residential buildings and is separated into space and water heating.  Moreover, space heating   101 estimates are further separated into building component losses and gains, which can provide critical information for local energy conservation policies.  It is also important to note that the described technique provides an estimate of nominal energy demand as opposed to consumption, the difference being that the latter requires detailed information about building occupant behavioural patterns and the demand for lighting and additional power devices.    The following section provides a detailed description of the modeling framework followed to enable estimation of individual building thermal energy demand for numerous buildings.  The basic modeling approach is divided into two sections; the first describes the methods used to attribute individual buildings with the necessary information to populate a building heat transfer algorithm, while the second section describes the computational approach of the heat transfer algorithm itself. Building attributes Building physical attributes  Basic building geometry was derived from the building footprints.  This two-dimensional geometry was then combined with the LiDAR to derive additional three-dimensional attributes including volume, roof area and wall area.  These attributes were calculated from the LiDAR derived normalized digital surface model (nDSM).  Volume was then calculated as the sum of the elevation grid cells within each building footprint.  Similarly, roof area used the gridded data to derive the slope of each cell and the sum of the area of each sloping cell was summarized for each building.  Wall area was calculated following the method presented   102 in equation 3.3, where each consecutive node pair in the building geometry is extruded to the height of the corresponding elevation grid cell and area of the vertical wall face derived from the resulting 4 bounding nodes.  Because of the existence of residential units located in mixed-use buildings, a further step was needed to separate the volume of these residential areas.  In cases where buildings served two uses, namely residential and commercial or institutional (identified from zoning districts), residential volume was estimated by reducing the total volume by that associated with the ground-level units.  The final physical attribute required for each building was fenestration, including both the size of window openings and their location on building walls.  Total window area was assessed using a statistical relationship with building age (see Section and volume, and developed from the energy audit data with regression coefficients listed in Table 5.3.  Allocating window area to individual walls assumed a fractional split with 70% of window area located on the front and back of the building and the remainder located on the sides.  Overall building orientation was determined using the direction of the longest edge of the lot on which the building is located and walls attributed as front, back or side in relation to this orientation. Building occupancy  To determine the number of occupants in each building, census population was disaggregated according to the representative fraction of building volume within each census dissemination   103 area.  The general approach followed the dasymetric mapping technique presented in Lwin & Murayama (2009) as:                              (5.1)  where BPi is the population of building i, CP is the population for the census area, BVk is the total volume for the buildings in the census area, and BVi is the volume of building i.  However, equation 5.1 is problematic in that it assumes that occupants of single-family dwellings (SFDs) and multi-unit residential buildings (MURBs) inhabit the same relative fraction of indoor space.  In reality, residents of SFDs will occupy more space than their MURB counterparts.  Therefore, the dasymetric approach in equation 5.1 was adapted to account for this skewed split by first allocating census population as either SFD or MURB by aggregating relevant dwelling type details from the census data (vanderLaan et al., 2012).  Residential building type volume and population were then substituted into equation 5.1 for variables BV and CP respectively. Building age  Building age, or more appropriately, year built, is often used to help model building energy performance (Caputo et al., 2013; Aydinalp et al., 2004; Farahbakhsh et al., 1998; Neidhart & Sester, 2004). Here, the LiDAR data was analyzed to provide an indication of the year of building construction by deriving an empirically based relation among various parameters related to its morphology.  As described in Chapter 3, a random forests machine learning   104 approach was used to predict building age where assessment data was not available, using a suite of variables derived from various municipal spatial datasets including LiDAR. Building energy parameters                  A common approach for assessing building energy parameters determines building archetypes, categorized by vintage using various building age ranges, and which is then used to summarize similar structural and technological characteristics of those buildings.  However, as suggested from the results of Chapter 3, aggregation approaches neglect the potential annual variability in building energy performance values.  As a result, methods that relate actual year built to building energy performance can provide greater precision for building energy modelling efforts.    To predict building energy parameters, energy audit data was empirically related to year built.  The building energy audit data contained measurements of the thermal resistivity of a building?s roof, walls, foundation, and windows in addition to air leakage, window area, and window solar heat gain coefficient.  Using the year built value also included in the energy audit records, regression analysis was conducted to derive empirically based models for each building energy performance parameter for later use in the heat transfer algorithm.  However, since energy audit information was collected for only SFDs, MURBs within the study area were assigned static building energy parameters based on Kellett et al. (2013) and listed in Table 5.2.     105 Table 5.2. Building energy performance parameters for multiple unit residential buildings (MURBs). Roof Resistivity (RSI) Wall Resistivity (RSI) Foundation Resistivity (RSI) Window Resistivity (RSI) Window Solar Heat Gain Coefficient Natural Ventilation Rate 2.4 2.2 1.0 0.24 0.68 0.3 Building irradiance  The final building attribute derived in advance of the energy modeling procedure was irradiance, which was split by rooftop and individual walls.  In building energy modelling, solar radiation provides one of the primary energy gains into a building, but can be highly variable across space due to the three-dimensional structure of the local environment.  As a result, Robinson (2006) suggests that it is important to simulate irradiance explicitly to provide locally accurate and relevant results.  To assess the solar energy available to each building, roof and wall irradiance algorithms presented in Chapter 4 were computed.  In each model the solar position algorithm was executed at 1 h time intervals.  Shortwave radiation was summarized as monthly averaged cumulative daily values.    106 Energy service demand Domestic hot water  Domestic hot water demand in Canada accounts for approximately 22% of total household energy consumption (Aguilar et al., 2005).  The key determinant of this consumption is building occupancy, although advanced models have been designed to account for multiple variables relating to demographics, technology and the environment (Aguilar et al., 2005; Lutz et al., 1996; Jorgensen et al., 2009).  As expected, diurnal consumption is associated with the occupancy schedule of the building, which tends to peak in the mornings and evenings (Meyer & Tshimankinda, 1997; Lutz et al., 1996; Biaou & Bernier, 2008). However, seasonal variations are more difficult to determine since Canadian assessments are unavailable (Aguilar et al., 2005).  While some studies have indicated strong seasonal variations in hot water consumption (Meyer & Tshimankinda, 1997; Aguilar et al., 2005), a study by the Canadian Mortage and Housing Corporation (CMHC) indicates no overall change in domestic water consumption with heating degree-days.  In addition, the Canadian situation typically has water heaters and tanks stored in conditioned spaces, so the effects of seasonal air temperature fluctuations are likely to be minimal.    The variables used to model the energy demand for hot water in this chapter included building occupancy and monthly incoming water temperature.  Estimates of energy demand for domestic hot water were then calculated by integrating building population with per capita hot water consumption estimates.  Existing modelled estimates of Canadian volumetric daily personal hot water consumption range between 47 and 167 litres (Aguilar et   107 al., 2005).  Due to geographic proximity and similar climatic and socioeconomic conditions, results of a flow trace analysis of 10 homes in the City of Seattle were used to estimate hot water consumption for the City of Vancouver at 95 liters person-1 day-1 (DeOreo & Mayer, 2000).  Volumetric hot water consumption was then converted to energy demand by calculating the energy required to raise the incoming piped water temperature to a designated hot water storage temperature by:                                 (5.2)  where WD is the per capita water heating energy demand, WV is the volume of consumed hot water, C is the specific heat of water (4.187 J gK-1), Tstorage is the storage temperature, and Tpiped is the incoming piped water temperature.  In this analysis, hot water storage temperature was assumed to be 60? C (as recommended by the World Health Organization (2002) to induce bactericidal effects), while incoming piped water temperature was assessed monthly based on average temperatures measured at multiple locations for the City of Vancouver?s drinking water (City of Vancouver, 2009). Space heating  The majority of household energy use in Canada is associated with space heating (Aydinalp et al., 2004; Rylatt et al., 2003). Space heating demand (Qdemand) for an individual building can be determined using an energy balance equation (Aydinalp et al., 2004) that accounts for heat gains and losses as:    108                               (5.3)  where Qlosses includes energy lost through the building envelope and from building ventilation and infiltration, and Qgains includes energy gained through the same components in addition to anthropogenic energy gains such as human metabolic processes, lighting and sensible heat output from energy using devices.  Because energy losses are largely associated with the physical building itself, the input variables used to determine heat losses are relatively stable through time and depend largely on the construction and material components of a building.  Nonetheless, there exists substantial spatial variation in these building components as a result of changes to building codes, architectural preference and technological advances. In contrast to heat losses, which are dominated by a technological component, gains are associated primarily with environmental and behavioural components.  As such, model input parameters used in calculating energy gains demonstrate substantial variability across both time and space.  The following section describes the computational process used to calculate building thermal energy demand by integrating the earlier described building attributes within a building heat transfer algorithm.  While equation 5.3 suggests that a simple separation of losses and gains can be used to assess building energy demand, the approach here requires the incorporation of radiation gains prior to calculating losses in order to represented the temperature at the building surface.  Sol-air temperature (Tsol-air) is used to characterize this effective air   109 temperature by developing a heat transfer equivalent temperature in the absence of any radiation and was calculated as:                                         (5.4)  where To is the outside air temperature, ?s is the albedo of surface s, Qsw is the incoming shortwave radiation, ?Qlw is the balance of longwave radiation at the building surface and hc is the combined radiative and convective heat transfer coefficient.  Albedo was set to 0.1 for the entire building envelope (Oke, 1998).  Qsw was derived separately for the roof and each wall, while ?Qlw used the sky view factor computed for the roof (equation 4.1) and each wall (equation 4.14) as part of the irradiance modeling procedure to determine net longwave radiation exchange with the sky, and assumed a net zero exchange between the building envelope and surrounding urban features.  Finally, hc was set to 25 W K-1 m-2 based on empirical studies for normal building exposure (Hagishima et al., 2005; Hagishima & Tanimoto, 2003; Loveday & Taki, 1998).  Using Tsol-air for components of the building envelope, conductive energy losses for wall, roof and window surfaces were computed as:                                   (5.5)  where RSIs is the conductive resistivity of surface s, As is the area of surface s and Ti is a volume averaged daily comfort index temperature set at 16? C to account for unheated spaces   110 and temporal variations in room heating requirements.  The convective loss through the foundation walls and floor (Qbsmt) was calculated separately from the above ground building components and required the initial calculation of foundation heat transfer functions for the basement floor (hbf) and walls (hbw) as defined in Lath & Boileau (1969) as:                                                                (5.6)  and                                                           (5.7)  where Ab and Aw are the areas of the foundation floor and walls respectively, d is the depth of the foundation, Ts is the ground temperature and RSIbsmt is the conductive resistivity of the foundation.  Total foundation energy loss was then calculated as:                                           (5.8)  In addition to conductive losses through the various components of the building envelope, losses to ventilation and infiltration were also calculated.  In the context of a single family residential building, these losses are dominated by the infiltration of air from openings in the building envelope, while mechanical ventilation losses are minimal compared to larger MURBs.  Using the attributed air changes per hour at 50 Pa (ACH50) for each building, infiltration loss (Qvent) was calculated as:   111                                         (5.9)  where NC is a coefficient used to convert ACH50 to a natural ventilation rate that accounts for weather driving forces, cp is the specific heat of air (1006 J m-3 K) and V is the building volume.  In this study, NC was set to 20 based on Sherman (1987).  Window energy gains (Qshg) provide an exception to the sol-air temperature envelope gains integrated into the energy loss equation (equation 5.1) and were calculated separately as:                             (5.10)  where Awin is the area of the window and SHGC is the attributed solar heat gain coefficient for the building windows.  Lastly, anthropogenic gains were calculated.  On average humans emit approximately 116 W of energy and inhabit residential buildings at various lengths of time throughout the day.  Additionally, building occupants contribute energy gains to a building through lighting and appliance use, the total contribution of which can be estimated as:                                   (5.11)    112 where Acf is the conditioned floor area of the building.  Since building volume was calculated from the LiDAR, a conversion factor was derived to estimate Acf based on 79 modelled and measured cases within the local area (http://elementsdb.sala.ubc.ca/).  The conversion factors calculated here were 3.1 and 3.2 for SFDs and MURBs respectively.  5.2.4 Validation  Validating building energy simulation models is extremely challenging given building accessibility constraints, the confidentiality and spatial availability of utility consumption data, and the partitioning of consumption data into fuel types as opposed to service demands.  Validation therefore involved comparison with energy simulation outputs from HOT2000 (Haltrecht et al., 1999), in order to situate results in relation to those from the common tool used for predicting energy demand in Canadian low-rise residential buildings.  Do to the time required to manually populate individual building models in the energy simulation software, a total of 10 buildings with a range of local shading conditions were randomly selected and where relevant modelling parameters were entered into the simulation software matching those from the modelling procedure presented above.  To correct for potential bias associated with building volume and to facilitate comparison with other studies, absolute energy demand values were converted into building energy use intensity (EUI) as:                          (5.12)    113 5.3 Results  Regression analyses of the building construction date (year built) and the various building energy components confirmed strong correlations in the majority of cases (Figure 5.2, Table 5.3).  The strongest correlation was observed with wall resistivity (R2 = 0.95, n = 108, p < 0.001), followed by roof resistivity (R2 = 0.88, n = 108, p < 0.001) and infiltration (R2 = 0.85, n = 108, p < 0.001).  The lowest correlation was observed with foundation resistivity for buildings built before 1971 (R2 = 0.07, n = 72, p < 0.1).  Regression coefficients for each of the building energy parameters are also listed in Table 5.3, which provided modeling inputs when conducting the final building energy demand estimates.  Model outputs of building energy parameters displayed in Figure 5.3 are summarized into regularly spaced grid cells in order to facilitate interpretation of spatial patterns due to the large number of buildings.  However, it is important to note that data are ultimately stored for each individual building record.  Mapping each of the energy parameters across the study area highlights a number of spatial patterns.  Building age demonstrates a relatively stochastic pattern across the study area, resulting in substantial spatial variation in roof, wall, foundation and window resistivity values.  Alternatively, solar heat gain coefficient values show relatively little spatial variation, with most buildings in the study area attributed with values close to the upper range of 0.68.  Finally, the maps show contiguous regions of consistent energy parameter values, which are associated with MURB building types that have been assigned with static energy parameters regardless of building age.    114  Figure 5.2. Trends in building energy components compared to the year of building construction.   Combining the estimated building energy parameters enabled the execution of a building heat transfer algorithm to determine the demand for building thermal energy services.  Results of the annual space heating and water heating demands are displayed in Figure 5.4.  As expected, both hot water demand and space heating demand are highest for MURBs due primarily to higher occupancies and greater volumes respectively.  It is interesting to note however, that the dominant energy service differs between development type, where SFDs are dominated by the thermal service demand for space heating, while MURBs are dominated by hot water demand.  Figure 5.4 also helps to illustrate the difference in energy ranges between thermal service types with the maximum range of building hot water demand reaching 338 MWh yr-1 versus 95 MWh yr-1 for space heating.     115 Table 5.3. Building energy parameter regression models and associated coefficient of determination (R2), significance (p) and standard error (SE). Energy Parameter Regression Equation R2 p SE Ceiling RSI 1486.4567 - 1.5459(year) + 4.0251e-04(year)2 0.88 *** 0.03 Wall RSI 387.2749 - 0.4081(year) + 1.0778e-04(year)2 0.95 *** 0.001 Window RSI -6627.47311 + 10.2205101(year) - 5.25306837e-03(year)2 + 8.99879265e-07(year)3 0.81 *** 0.002 Window SHGC 5232.6998 - 8.0661(year) + 4.144498e-03(year)2 - 7.097355e-07 (year)3 0.73 *** 0.04 Window Area -140.2 + 7.054e-02(year) + 4.737e-02(volume) 0.80 *** 0.17 Infiltration 162.41 - 7.79e-02(year) 0.85 *** 0.1 Foundation RSI (<= 1971) -2.6022 + 1.8489e-03(year) 0.07 * 0.45 Foundation RSI (> 1971) 9.83191219e+05 - 1.4821217e+03(year) + 7.44714415e-01(year)2 - 1.24725488e-04(year)3 0.81 *** 0.08 *** significant at alpha < 0.001   * significant at alpha < 0.1     116  Figure 5.3. Mapped building energy parameters averages over 25 m by 25 m grids across the study area.   117  The comparison of annual building space heating demand from the LiDAR-based model and building energy simulation software (HOT2000) is shown in Figure 5.5.  Results show a strong correlation between both models (R2 = 0.93, p < 0.001) and a relatively uniform distribution around the identity line, indicating no discernable deviation with the approach presented here.  A T-test indicates no statistically significant difference between the LiDAR-based approach and the building simulation outcomes (t = -0.11, p = 0.91) when compared on a monthly basis.  Because of the relative simplicity of the hot water demand computation, and its stability over the year, no validation was completed for this component.   Figure 5.4. Mapped (a) hot water demand and (b) space heating demand averaged over 25 m by 25 m grids across the study area.     118   Figure 5.5. Comparison of annual energy use intensity (EUI) values from the LiDAR-based approach and the building simulation software.   5.4 Discussion  Comparison of the outputs from the LiDAR based approach with widely used building energy simulation software indicate very close building energy demand outcomes, with statistically insignificant differences in monthly space heating demand and a very high coefficient of determination when comparing annual building EUI.  A further comparison with recent studies also indicates the estimated values are closely aligned with relevant outcomes from the building energy community.  Using a furnace efficiency of 78%, the average building space heating EUI for the approach in this chapter is 64 KWh m-2 yr-2.  In comparison, the Energy Information Administration Residential Energy Consumption Survey 2009 (http://www.eia.gov/consumption/residential/data/2009) provides a space heating EUI for Pacific Northwest Single Family Detached Units of 61.3 KWh m-2 yr-1.    119 These outcomes suggest the potential applicability of LiDAR-based energy estimates for citywide building energy demand modeling.  Advantages of this approach are especially apparent when considering the high level of automation necessary to estimate thermal energy demand for multiple buildings.  Traditional building energy simulation software (e.g. Crawley et al. (2008)) requires the user to populate individual building models with a wide range of values; a time consuming task requiring detailed knowledge and measurement of specific building components.  In contrast, citywide energy estimates have traditionally relied on a single EUI for all similar building or development types (Heiple & Sailor, 2008).  Results from this chapter demonstrate the range of EUIs even within the basic category of single-family dwelling.  Understanding this rich variability in energy demand can have applications for planning and policy design.  5.4.1 Limitations and future work  The approach used in this chapter focuses on developing an estimate of nominal thermal energy demand for individual buildings, allowing the user to combine heating plant efficiencies, energy demand for appliances and occupant behavior in later steps.  The justification for this approach was to minimize the number of assumptions entered into the basic modeling procedure.  Nonetheless, several assumptions still exist within the model, requiring additional research to better quantify potential variability.  Specifically, the spatially static coefficients used to convert ACH into a natural ventilation rate (NC in equation 5.9) and to estimate radiative and convective heat transfer at the surface of the building envelope (hc in equation 5.4), are expected to demonstrate spatial variations   120 resulting from urban form effects.  LiDAR data offers a potential option for better understanding these effects, and is suggested for future research efforts.  The LiDAR acquisition in this study used data collected under full leaf-on conditions.  As a result, the approach presented here neglects the potential increase transmission through deciduous trees in winter months.  This is likely to result in an underestimate of solar gains into a building in the winter, and therefore an overestimate of nominal thermal energy demand.  Future research is therefore also recommended to compare the variation in radiation transmissivity between seasonal LiDAR acquisitions.  The relation of building age to various building energy parameters as shown in Figure 5.2 also neglects outliers resulting from substantial energy retrofits or initial building construction efforts that provided better performance than required by the building code at the time.  While these potential outliers are likely to have minimal impact on the modelled energy demand of entire neighbourhoods or cities, caution should be taken when inferring any single building demand outcome.  Similarly, the spatial variation in MURB energy parameters is not accounted for in our model due to limited energy audit information.  Collection and release of energy audit information for MURBs is therefore recommended and encouraged to facilitate similar detailed future building energy modeling endeavors.      121 5.4.2 Policy applications  Gaining an understanding of building energy demand for individual buildings across the city has multiple applications and can better inform various stakeholder initiatives.  Many of these applications involve mandates for energy and emissions reductions, with local governments and utilities being the primary authorities tasked with designing local energy planning and policy.  In the planning context, growth management, land use and urban form have been suggested as having a similar impact on building energy as the construction materials of individual buildings (Senbel & Church, 2010).  In this regard, planning and policy at both the scale of the entire city and of the individual building can be better informed with baseline estimates of building energy demand.  At the city scale for example, planning the development and expansion of district energy infrastructure requires detailed estimates, forecasts and understanding of the spatial arrangement of thermal energy loads (Finney et al., 2012).  At the scale of an individual building, retrofit and energy efficiency opportunities, building energy labeling and demand-side management strategies can be designed more effectively with spatially contiguous information on building energy demand (Allcott & Mullainathan, 2010). While the topics above provide only a few examples of the potential application of citywide building energy assessments, they represent some of the most prominent local energy-related actions and initiatives considered by local governments and utilities.  In the following chapter, a description of all known local building energy policies is provided, and the role that geographically explicit building energy modelling plays in policy analysis in discussed.     122 6 Local Government Building Energy Policy  6.1 Introduction  Unique obstacles to energy improvement in the building sector include the long-lived nature of building products, extended and complex supply chains, differences in market participation between owners and users, the stationary character of building components, the heterogeneity of buildings, high capital costs, and the large number of small firms acting as participants in this market (Ryghaug & S?rensen, 2009).  Consequently, a well-articulated mix of policies is desirable to tackle building energy.  This policy mix must also consider the broader objectives in which building energy is contributing to problematic market externalities or as an impediment to effective service provision that governments are tasked with addressing.  Common aims of energy efficiency and renewable supply strategies include mitigating emissions (?rge-Vorsatz et al., 2007), promoting security and resilience to fluctuations in global energy markets (Yergin, 2006), improving the profitability of the production and use of energy products (Ryghaug & S?rensen, 2009), and discouraging imbalances between low entropy (high quality) energy sources and high entropy (low quality) services (Schlueter  & Thesseling, 2009).  A growing body of literature exists examining various policy instruments and decision support mechanisms available for addressing the energy sector at regional, national, or international levels (Kannelakis et al., 2013; Jaccard, 2005).  However, specific contextual considerations for local governments have been relatively underrepresented in this body of   123 literature (Calvert, Pearce, & Mabee, 2013).  Despite this lack of attention, local governments appear to be situated in a unique position to implement various building energy strategies and policies as a result of their direct connection to stakeholders and their experiential capacity providing the public with various services (Betsill & Bulkeley, 2006; Burch, 2010; Collier, 1997; Guy & Marvin, 1996).  Moreover, both buildings and renewable energy systems tend to operate predominantly at local scales, with local environmental considerations that are sometimes difficult to capture in broader policies delivered from higher level government authorities (Calvert, Pearce, & Mabee, 2013).  Given that attention has generally been focused on policies at these higher levels of government, the objective of this chapter is to review a suite of building energy policies relevant for implementation by local governments and to then determine which of these policies can be better informed from citywide estimates of building energy demand.  Before reviewing these policy instruments, the role and responsibilities of local governments in current society and the attributes of building energy within the local environment are discussed.  Then, to compare and contrast the policy strategies available to local governments for dealing with building energy, known policy instruments are classified into the general categories of voluntary, fiscal, regulatory and capacity building.  Each policy instrument is also evaluated based on a set of considerations designed to help in the strategic design of a mix of policies that suit the contextual objectives of the local government.  Finally, the spatial considerations of relevant policy instruments are explored, especially as they relate to spatially contiguous estimates of building energy demand as has been explored in the previous chapters of this dissertation.    124 6.2 Background  6.2.1 The role and responsibilities of local governments  Local governments are the lowest tier of public administration in the majority of countries across the world.  As such, they are typically created by a higher level of government through national constitutions or legislation, state or provincial constitutions or legislation, or by executive order (Shah & Shah, 2006).  Local governments are generally responsible for delivering a range of specific services to the public within a geographic area that best represents these service markets.  Furthermore, in most developed nations, local governments adhere to ideological perspectives of fiscal federalism and new public management, which prescribe that the roles and responsibilities of local governments are to address market failures and deliver public goods efficiently and equitably (Shah & Shah, 2006; Finn, 2008).   To address the externalities associated with building energy, local governments have implemented a range of energy-focused strategies (Jones, et al. 2000; Tang et al., 2010; Betsill & Bulkeley, 2006; Brody, 2008; Gore, 2010; Lutsey & Sperling, 2008; Wood & Thomson, 2012).  The active role that local governments take in developing these initiatives has been associated with the substantial energy demand of human dominated environments, the direct relationship between local authorities, stakeholders and the public, and their established competencies in service delivery (Betsill & Bulkeley, 2006; Burch, 2010; Collier, 1997; Guy & Marvin, 1996).  One of the primary initial actions taken by local governments across developed nations has been to raise the awareness of energy and emissions issues   125 related to buildings, which has spurred initiatives incorporating a wide-range of locally relevant policy instruments (Calvert et al., 2013; Tang et al., 2010).   6.2.2 Building energy services and the local environment  Building energy is used in this Chapter as a term that encompasses both the operational energy-use of buildings and the supply of renewable sources of energy.  The demand for energy within a building is associated with a variety of services.  These include space conditioning, water heating, lighting and powering devices (including home and commercial appliances, electronics and industrial machinery).  Together these services combine to determine the operating energy demand of a building, and vary as a result of technology efficiencies, environmental conditions, and occupant behavioural patterns (Ratti et al., 2005).  Operational energy accounts for between 90 ? 95% of a building?s total energy demand over its lifetime, with the remainder (embodied energy) used in the fabrication of the building materials and components (Ravetz, 2008).   In addition to the dynamics of building energy services, variation in the physical form of built landscapes is important to consider when addressing building energy from the perspective of local governments.  For example, urban form can have indirect impacts on building energy resulting from building layout and landscaping, with consequent effects on solar energy gains (Robinson, 2006), convective heat transfer (Hagishima & Tanimoto, 2003), and local air temperatures (Oke, 1987).  Transmission and distribution losses associated with the supply of heat, and to a lesser extent electricity, are also highly dependent on the form and density of the built environment (Ewing & Rong, 2008; Finney et al., 2012).    126 However, deliberations on urban form as part of the local decision making process generally only apply to new developments.  Moreover, the reproduction of the building stock and supporting service infrastructure is typically slow, occurring at a rate of 1 ? 2% per year (Ravetz, 2008).  In the context of both short and long range planning within the mandate of local authorities, it is important to consider that this slow turnover means that the majority of the future building stock already exists.  As a result, simply targeting new developments with local energy policy will have a limited effect on overall energy demand.  Local energy policy must therefore consider both new and existing buildings if substantial alterations to current patterns of energy use are desired.  6.3 Local energy policy instruments  Given the diversity of authority granted to local governments within and between nations, certain policy instruments may be more or less feasible for administration in a given area.  Nonetheless, this section explains a suite of known policy instruments that have been used by local governments to elicit actions that influence energy use in the building sector.  To facilitate the comparison of policy options, the instruments are separated into the broad conventional categories of voluntary, fiscal, regulatory and capacity building policies, and then grouped into specific categories that indicate the fundamental planning mechanism used to elicit action.  Policies are also discussed in terms of their applicability to new developments or existing developments, their ability to target renewable energy supply or energy efficiency, potential jurisdictional constraints, geographic variations, public and administrative acceptability, and the level of certainty in the outcomes of the policy   127 instruments.  Respecting that contextual factors exist that inevitably determine the specific considerations associated with integrating policies, an initial rubric for evaluating local building energy policies is provided in Table 6.1, where policy evaluation criteria are based on whether each instrument can be explicitly gauged according to the criteria listed above.  6.3.1 Voluntary policies   Voluntary policies have been widely adopted to encourage building energy efficiency, and are actively promoted for their ability to offer flexible mechanisms for building owners to attain desired actions while being low cost, politically acceptable and administratively feasible (Jaccard, 2005)(Lee & Yik, 2004).  Voluntary policies associated with building energy tend to describe basic label-based information campaigns (Lee & Yik, 2004).   While these tools are discussed, in this section policy instruments are also interpreted from the field of behavioural psychology, where the use of various behavioural interventions are presented as an additional suite of policy and information campaign design strategies that can be adopted by local governments to elicit sustainability-focused actions.   128 Table 6.1. List of building energy policy instruments and key considerations for local implementation.  Policy Instruments New Buildings Existing Buildings Demand/ Efficiency Renewable Supply Jurisdictional Constraints Geography Acceptability Certainty of Outcomes Voluntary         Information Campaigns            Developer checklists ?  ? ?  ? ?     Energy labeling ? ? ? ?  ? ?  Behavioral Interventions            Framing ? ? ? ?  ? ?     Commitment devices ? ? ? ?   ?     Implementation intentions ? ? ? ?   ?     Default options ? ? ? ?   ?     Social norms ? ? ? ?  ? ?     Nonlinear demand curves ? ? ? ?  ? ?  Fiscal         Disincentives            Energy tax ? ? ? ? ?       Property tax ? ?  ?        Developer cost charges (DCC) ?   ?  ?   Financing            Tax-lien ? ? ? ?   ?     On-bill ? ? ? ? ?  ?  Incentives            Tax and DCC Exemptions ?  ? ?   ?     Reduced permitting fees ?  ? ?   ?     Expedited building permitting ?  ? ?   ?     Grants and rebates ? ? ? ?   ?     Density bonuses ?  ? ?   ?  Regulatory         Building Codes            Prescriptive ?  ? ? ? ?  ?    Performance-based ?  ? ? ? ?  ? Land Use Planning            Zoning ?  ? ?  ?      Development permit areas ?  ? ? ? ?  ?    Comprehensive development ?  ? ? ? ?  ?    Phased development ?  ? ? ? ?  ? Service Area ? ?  ? ? ?  ? Capacity Building            Finances ? ? ? ? ?       Administration and information ? ? ? ?   ?     Authority ? ? ? ? ?       Energy utilities ? ?  ? ? ?  ?   129 Information campaigns  Voluntary building energy policies have been primarily concerned with information campaigns that encourage the practice of energy labeling.  This policy approach functions by imposing that property owners identify building energy performance criteria when a property is traded.  A widespread energy labeling policy, where building energy performance features are made explicit during the transfer of all buildings, has been suggested as an effective approach for mitigating the externalities of imperfect information in the building sector by making energy investments more transparent and therefore a more salient factor in determining a building?s market value (Curran, 2010; Linden et al., 2006; L?tzkendorf & Speer, 2005; Brown, 2001).  An example of building energy labelling has been in place in the UK, where Energy Performance Certificates (EPCs) are required for all buildings at the time of construction or when marketed for sale or rent.    Information campaigns have also been designed by local governments to help rectify imperfect information in building markets by strategically educating developers on energy efficiency or supply opportunities during events that trigger contact with the local government.  For example, checklists of non-compulsory energy efficiency measures have been used during the application process for building permits and rezoning to ensure that developers are aware of effective energy investment opportunities (CEA, 2008; MacNab et al., 2011).  By combining developer checklists with fiscal incentives (see Section and negotiated regulatory policies (see Section 6.3.3), this approach can also encourage   130 additional building energy investments by offering compensation for the perceived economic burdens of these supplementary measures. Behavioural interventions  Despite the optimistic rhetoric that frames the energy efficiency literature, there continues to exist a low level of penetration of energy efficient technologies and behaviours in the building sector (Allcott & Mullainathan, 2010; Lee & Yik, 2004; Ryghaug & S?rensen, 2009).  That building energy technologies also have short payback periods has led to the formation of an efficiency paradox (Jaffe & Stavins, 1994), in that low costs for available technologies are not translated into economy-wide energy demand reductions, even in the presence of widespread information campaigns.  One resulting outcome of this paradox has been to suggest that governments ought to focus efforts on policies that demonstrate greater effectiveness in achieving sustainability objectives, such as mandatory regulations and market-oriented instruments (Jaccard, 2005; Richardson, 2012).  However, Allcott & Mullainathan (2010) have also suggested the need for policy makers to consider alternatives to the dominant rational choice model for understanding human behavior1, and therefore to approach information campaigns with a better understanding of behavioural science.  To encourage building energy performance and renewable supply options, it is important that local governments comprehend the various motivations that individuals possess when assuming energy-related actions and behaviours.  For example, Gifford et al. (2011) indicate                                                1 Rational choice theory states that humans will always choose the most cost-effective means for achieving a specific goal, and is often used by economists to describe behaviour and suggest policy strategies (Gifford et al., 2011)   131 that when making choices that affect the environment, individuals cognitively negotiate non-price considerations related to geophysical conditions, the local regulatory context, technologies, and psychological factors.  Gaps often also exist between attitudes and actions, therefore the provision of basic education and information on building energy will not necessarily translate into desired actions (Gifford et al., 2011).  Instead policies and information campaigns need to be strategically designed to incorporate concepts from behavioural psychology, referred to as behavioural interventions, in order to achieve their maximum effectiveness.    To address energy demand reductions, behavioural scientists have identified numerous non-price interventions that can offer effective options for changing consumer choices towards more sustainable decisions.  In many cases, these interventions can be manifested as subtle elements of broader information campaigns and policy design.  Allcott & Mullainathan (2010) describe a set of six categories of behavioural principles that have demonstrated substantial promise for achieving large-scale energy reductions: framing, commitment devices, implementation intentions, default options, social norms, and the exploitation of nonlinear demand curves.    The first of these principles, framing, suggests that when given a choice people will act differently depending on how the options are portrayed, e.g. positively or negatively.  Using the example of both checklists and labeling presented above, framing may be used by strategically drawing comparisons between the energy performance of a proposed building or development and that of a high-performing alternative.  The second principle, commitment   132 devices, considers the natural tendency for humans to procrastinate and therefore interventions in this category are designed to establish binding decisions in the present, which encourage desired actions in the future.  For example, a local government might offer energy efficiency rebates or grants after signing up for a building energy audit and completing the recommended energy improvements.  Similarly, implementation intentions describe the use of tools than can be used to help guide the decision-maker through the steps and procedures along a path to achieving the desired action.  Using the example of checklists again, local governments can design these lists not only as a descriptive set of objectives, but also as a collection of ordered procedures to attain those objectives.  Default options, in which the best-performing choice or option is automatically made as the initial decision can be designed in the anticipation that the natural tendency for procrastination will maintain the defaulted desired action.  In this case, for example, local governments can promote connecting to a local energy utility service (see Section as the default option, while allowing the option to switch to an alternative supplier at a later date.  Appeals to social norms, in which desired decisions are likened to common actions made by a broader population, provides another established behavioural principle with promise for addressing building energy.  Finally, the exploitation of nonlinear demand curves offers a strategic approach for the design of incentive policies based on the notion that demand curves often exhibit nonlinear trends when prices approach zero (Allcott & Mullainathan, 2010), and therefore much greater demand for a given energy technology may be achievable with only a small decrease in price.  Nonlinear demand curves therefore have apparent implications when designing building energy grant and rebate schemes, although optimizing these grant programs assumes a more precise understanding of the shape of the demand curve for a given   133 technology.  Many of these behavioural considerations have been successfully integrated as part of Software-as-a-Service company Opower?s (http://www.opower.com) strategies to help utilities promote energy efficiency.  Psychological factors are also important to understand in relation to local government service and infrastructure delivery (Wolsink, 2010).  For example, the social acceptance of siting an energy technology, such as a district energy plant or a group of wind turbines, should be considered within a framework that acknowledges the process in which the investment is implemented, including issues of trust, distributive and procedural equality, and familiarity or experience with the technology itself (Hujits et al., 2012).   More advanced information campaigns that integrate lessons from behavioural sciences also carry with them important spatial considerations.  Specifically, Allcott and Mullainathan (2010) explain that descriptive social norms, where energy-use is compared to others sharing similar household or demographic characteristics, can be effective at reducing building energy demand.  Of particular interest to identifying and communicating social norms, is the concept of geodemographics (Harris et al., 2005), which have been used to target management strategies based on the positive spatial autocorrelation of socio-economic and behavioural characteristics.  Druckman and Jackson (2008) have also advocated that a spatially targeted approach to the implementation of energy-efficiency strategies is an efficient means for achieving building energy reductions.      134 6.3.2 Fiscal tools  Fiscal policy instruments discussed here include the use of local government revenues through revenue collection in the form of taxes and fees, financing mechanisms, and expenditures in the form of subsidies.  In certain cases, fiscal tools can also be designed to be revenue-neutral by recycling revenues through reductions elsewhere.  However, in contrast to higher-level government fiscal policies, the use of fiscal tools by local governments is often restricted by legal requirements to balance operating budgets and constraints on borrowing to meet capital requirements (Kitchen & Slack, 2003).  Furthermore, many charges must be explicitly linked to specific service provisions (CEA, 2008).  Nonetheless, local governments have encouraged renewable energy generation and energy efficiency investments through a variety of fiscal policy instruments. Disincentives  As with the implementation of many taxations policies, the general perception is that public support for new charges is generally low, and that businesses and special interest groups tend to assemble anti-tax campaigns and lobbies, discouraging the use of taxation by government officials (Kallbekken & S?len, 2011).  However, Kallbekken and S?len (2011) challenge the assumption that voters always act in their own self-interest, and suggest the importance of communicating the effectiveness and the welfare consequences of taxes to garner support from the electorate.  Moreover, despite suggestions that local governments function as a more effective democratic institution than higher levels of government, due to in large part to their proximity to the public, Cutler & Matthews (2005) explain that local politicians are   135 partly insulated from public accountability.  Justification for this contradiction has been based on the observation that local elections are not well aligned with partisan politics, and information provided to the local voter tends to be harder to find (Cutler & Matthews, 2005).  As a result, local governments may offer the preferred institution for implementing taxes to influence energy use in the building sector.  The most direct disincentive to address building energy is the development of an energy tax, where energy generated offsite and delivered to a building is subject to a charge.  However, unless energy is provided as a service by the local government, options for direct taxation of energy are likely to be limited as a result of jurisdictional constraints over local taxation authority.  Therefore, disincentive mechanisms to address building energy must often be strategically designed to function within existing local tax law, namely property taxes, sales taxes or charges for the provision of a service or set of services.  For example, taxes can be structured as local improvement charges that increase property tax bills or local sales taxes in return for a specific neighbourhood energy service (CEA, 2008; MacNab et al., 2011; Merk et al., 2012).  Developer cost charges (DCCs) may also be used as a form of disincentive, which operate by explicitly recuperating the capital costs accrued to the local government when extending services to a new development (CEA, 2008; Jaccard et al., 1997; MacNab et al., 2011; Merk et al., 2012).  The State of Utah provides an example of local government energy taxation opportunities, where opt-in legislation enables municipalities to add a tax to the sale of gas and electricity distributed within their jurisdiction.    136 Financing   Fiscal tools in the form of an optional tax to may also be used as a financing strategy to overcome the initial capital cost associated with investments in building energy efficiency or supply technologies.  In these cases, the local government can use tax-lien financing to enable building owners to purchase a range of energy technologies, which are repaid through an increase to property taxes (White, 2010).  In circumstances where the municipality is also the energy service provider, the repayment may be structured through increases to the utility bill (on-bill financing), which has the advantage of placing the financial burden on the building occupant(s) who receive the energy savings benefits associated with the technology (White, 2010).  In the US, the states of California, Oregon, Minnesota, Michigan, Ohio, Connecticut and Florida have enabling legislation and active programs allowing tax-lien energy financing, commonly referred to in the US as Property Assessed Clean Energy (PACE) financing. Incentives  A range of tax exemptions and profit enticements, broadly referred to here as incentives, are also available as potential fiscal tools that local governments can use to address building energy.  Exemptions to property taxes and DCCs have been used to support developments that include environmental features such as renewable energy technologies and energy efficient technologies and design (CEA, 2008; Jaccard et al., 1997; MacNab et al., 2011).  As an example, from 2009 to 2013 New York City provided property tax abatements for building owners who installed solar panels of 5 to 8.75% of panel-related expenditures.     137 Similarly, expedited approvals of building permits or reduced permitting fees can be prioritized for developments that include identified energy technologies (CEA, 2008; MacNab et al., 2011).  More generally, explicit subsidies in the form of grants or rebates on qualified energy technologies offer transparent support mechanisms to encourage building energy objectives.  Finally, density bonuses, which grant exclusions from density restrictions in zoning bylaws, can be exchanged for the provision of environmental amenities, and may be extended to encourage building energy investments (CEA, 2008; MacNab et al., 2011).  Using a broad definition of subsidies, the latter instrument of density bonuses offers the most certain incentive-based policy by minimizing the potential for ?free riders?2, an issue that can arise from property tax and DCC exemptions, grants and rebates, and reduced permitting fees.  However, density bonuses are a popular tool used by various municipal departments, and thus proposed implementation of the tool is likely to be challenged by other local government bureaucrats.   6.3.3 Regulatory policies  Regulatory policies describe the mandatory standards or technologies imposed on the building sector by governments.  Though policy analysts promote market-based disincentives as the preferred means to alter broad scale patterns of energy use (Jaccard, 2005), the discrepancy between occupants and owners in the building sector introduces a unique situation in which both groups of actors do not necessarily participate in the same markets.  Specifically, under rental conditions occupants are typically responsible for energy bills,                                                2 In this context, free riders indicate those people or firms who would have taken the desired action regardless of the incentive.   138 while owners are responsible for building energy technologies.  As a result, market-based incentives are less effective for encouraging owners to invest in energy efficiency or supply measures that benefit the renter.  Furthermore, the building industry has been recognized for possessing conservative values, scoring among the lowest industries in terms of research and development and innovation (Ryghaug & S?rensen, 2009).  To address these challenges governments have enacted various regulatory policies to ensure minimal building energy objectives are achieved.    In their strictest sense, regulations enforce restrictions, such as the ban on the use of a given technology or energy source (Lee & Yik, 2004).  However, regulations can also be designed to be less intrusive by allowing flexibility in the instruments and technologies used by the building sector to achieve minimum efficiency or supply objectives.  Furthermore, local governments can choose to set broad regulations that apply ubiquitously across their jurisdictional boundary, or as part of a negotiation process with single developers.  The following section discusses a range of regulatory policies as categorized into building codes, land use planning and service area bylaws. Building codes  Building codes have been the most widely used regulation for achieving building energy efficiency, and have been adopted in most developed countries (Lee & Yik, 2004).  However, it is rare that local governments are given jurisdictional responsibility for their design (Laustsen, 2008).  Rather, the role of local governments has largely been to ensure existing codes are enforced.     139  The establishment of building codes is generally separated into national building codes, which are created by a singular government, or as has recently been the case, into model building codes, which are generated as part of supranational collaborations of governments and industry groups promoting international standards (Laustsen, 2008).  In both cases, most building codes will include some basic geographic adjustments that account for broad climatic variations.    Historically, building codes have been established as a prescriptive list of building envelope components and technology efficiencies, although recently this approach has been criticized over its inflexibility in allowing innovative alternatives that achieve equal energy reductions (Ryghaug & S?rensen, 2009).  Performance based building codes have thus been introduced in various jurisdictions that stipulate maximum total energy use as opposed to specific building technologies (Coglianese et al., 2003).  In practice, performance based codes necessitate that building energy simulations are used to determine the overall energy of a building, and consequently building inspectors must be trained in the use of building modelling software.  The effectiveness of performance-based codes is therefore dependent on the quality of the simulations employed and the capacity of the inspector to ensure simulations are suitably modelled (see Section    Unique behavioural actions and responses are also difficult to assess and regulate using building codes.  For that reason, building codes have also been criticized for their inability to   140 control for rebound effects3.  Nonetheless, several empirical studies have found that energy reductions in the building sector tend to be consistent with the implementation of building codes (Jacobsen & Kotchen, 2013; Costa & Kahn, 2010).  Although the design of building codes may extend beyond the authority of local governments, the promotion and encouragement of opt-in legislation has be used to encourage desired higher-level mandates without provoking political resistance from opponents in other jurisdictions.  Future proofing strategies provide an example of potential opt-in building code directives, whereby internal building components are integrated during the construction phase of a building to reduce the future costs and burdens of retrofits that integrate renewable energy or energy efficient technologies (Roaf et al., 2005; Burch, 2010).  Such an approach can help to address the adaptation challenges associated with the long-term stock turnover of buildings.  Renewable energy requirements may also be mandated through amendments to the building code or through opt-in legislation that would require buildings to generate a certain portion of their energy on-site using various renewable energy technologies, such as the Merton Rule in the United Kingdom; a local government policy that requires that at least 10% of a new development?s energy is met by onsite renewable energy generation (Williams, 2010).                                                 3 The rebound effect refers to a counter response to the implementation of a technology designed to improve the efficiency of a resource or energy, leading to a lessened impact of that technology or in some cases an overall increase in the use of the resource or energy.   141 Land use planning  Local government regulatory policies generally focus on zoning rules and bylaws that explicitly specify maximum or minimum levels of measurable development characteristics.  Examples include floor area ratio, density, air rights and types of land uses.  In the context of energy planning, these zoning rules can be used to achieve energy efficient development patterns (Miller, 2013) and to facilitate the implementation of renewable energy technologies as part of a municipal utility.  Similarly, development permit areas (DPAs) can be specified that require specific features external to the building when granting building permits, such as landscaping to improve environmental conditions, district energy facilities, or onsite renewable energy technologies (CEA, 2008; MacNab et al., 2011).  The potential also exists for DPAs to require that a portion of a development?s energy demand be met through renewable energy (CEA, 2008).  Conversely, local authorities may negotiate exceptions to existing zoning policy with a developer for the provision of energy efficiency or supply technologies.  Land use planning has also addressed building energy use objectives by guaranteeing long term zoning stability through comprehensive development zoning, where a negotiated site plan becomes the de facto land-use plan for a given area, in exchange for energy efficient designs and technologies that extend beyond the minimum level required from the building code (CEA, 2008).  Similarly, phased development agreements can also be negotiated with developers to secure long term zoning bylaws in exchange for building energy technologies (CEA, 2008; MacNab et al., 2011).    142 Service area bylaws  In some cases local governments have been granted the authority to mandate that new developments and existing buildings utilize renewable energy by enacting local service area bylaws (CEA, 2008; MacNab et al., 2011).  This approach allows local governments to charge for locally provided energy services, offering potential additional revenues sources, or price signaling to encourage reductions in energy demand (Merk et al., 2012).  Service area bylaws are a valuable policy instrument when establishing municipal energy utilities (see Section  As an example, the City of Surrey in British Columbia has developed its own energy utility and introduced a district energy bylaw that mandates the compulsory use of district energy and compulsory hydronic heating systems depending on the location and building floor area ratio of new construction and substantially renovated buildings.  6.3.4 Capacity building  An initial step in identifying the capacity for local governments to attend to building energy is to acknowledge and address recognized barriers to action (Burch, 2010; Robinson & Gore, 2005; Senbel et al., 2012).  Empirically identified barriers include limited revenues and budget constraints (Robinson & Gore, 2005; Schwartz, 2012; Senbel et al., 2012), staff competences and lack of information (Burch, 2010; Schwartz, 2012; Robinson & Gore, 2005; Senbel et al., 2012), and limited local authority (Robinson & Gore, 2005).  Acknowledging these barriers, local governments can seek to overcome them, increasing their capacity to address building energy, but where such barriers are politically, administratively, or financially insurmountable, local governments must facilitate the   143 effective use of existing resources and capacities (Burch, 2010).  While energy policy instruments that have the potential to operate within the existing institutional framework of local governments have been discussed above, institutional capacity building strategies are also worth noting. Finances  Despite cities being the primary drivers of economic prosperity, the fiscal environment in which local governments operate is generally limited to property tax revenues, sales taxes, or higher-level government transfers in lieu of income, corporate and sales taxes (Kitchen & Slack, 2003; Merk et al., 2012).  Raising revenue to manage building energy is therefore traditionally limited to increasing tax or encouraging new development, which can often be at odds with issues of social or environmental sustainability.  In the context of building energy policy and revenue generation, property tax increases, developer cost charges, service area bylaws and municipally owned energy utilities may offer an effective option for the removal of local government fiscal barriers to government actions addressing building energy (Merk et al., 2012). Institutional competences and information  Numerous studies have indicated that staff competences, including time, knowledge, and leadership, are critical for developing local capacity to address environmental concerns (Burch, 2010, Robinson & Gore, 2005; Senbel et al., 2012; Schwartz, 2012).  Institutional change to improve the competences of local bureaucrats can therefore involve the   144 establishment of a dedicated branch of local government to coordinate and support the development of sustainability issues such as building energy policy (Burch, 2010; Jaccard et al., 2011; Schwartz, 2012).  Schwartz (2012) suggests that such an endeavor is likely to attract knowledgeable staff, establish a supportive environment for knowledge transfer and generation, and ensure sufficient time is committed to environmental priorities.  The process of local public administration also presents circumstances where unintentional restrictions are imposed on new developments that inhibit investment in building energy technologies.  These restrictions typically relate to permitting, standards, and planning (CEA, 2008).  For example, subtle disparities between standards and new technologies have been shown to augment the liability exposure of building inspectors (Boswall, 2005), resulting in a reluctance to administer and issue permits (CEA, 2008).  To overcome these barriers, internal information campaigns targeting local officials must also be considered as a capacity building strategy.  Lastly, in relation to information transfer, it is important to note potential power that organizes horizontally, as is observed in organizations such as the International Council for Local Environmental Initiatives (ICLEI) (Wood & Thompson, 2012).  Using the example of ICLEI?s Cities for Climate Protection (CCP) program, Betstill and Bulkeley (2006) suggest that new environmental governance regimes are emerging where local governments are not necessarily preparing environmental strategies in isolation, but where knowledge and leadership are transferred horizontally though international networks of local authorities.  Participation in horizontally structured institutions may therefore augment a local   145 government?s ability to address building energy by drawing on example best practices or existing policy guidelines. Local authority  Institutional adjustments to increase local government authority in relation to building energy policy must usually be negotiated with the central power granting administration.  Furthermore, because higher-level legislation and directives tend to apply to numerous local governments, the pursuit of legislative change by any singular local administration is politically and administratively challenging.  Collective negotiation and interest entities acting on behalf of multiple local governments provides a potential avenue for removing local authority barriers.  Similarly, local governments can encourage and promote opt-in clauses, typically related to building code amendments, within higher-level legislation, giving choice to individual governments to enact new energy-specific mandates (MacNab et al., 2011). Local energy utilities  Local energy utilities offer a unique option for addressing building energy.  The trend in the decentralization of power generation across developed countries presents new opportunities for municipalities to extend building service provisions for heat and electricity, encouraging local and global environmental benefits when using renewable or more energy-efficient technologies as the means of energy production (Karger & Hennings, 2008; Jones et al., 2000).  In addition to natural renewable energy resources such as the sun, wind, biomass and   146 water, local governments often also encompass various waste energy resources within their jurisdictional boundaries that can be used to supply energy including landfills, wastewater, drinking water reservoirs, and large energy consumers such as pools, ice rinks, hospitals and industries (Jaccard et al., 1997).  It should also be emphasized that the strategic design of pricing and rate structures may be used to promote effective delivery of an energy service and to encourage energy demand reductions (Merk et al., 2012).  In the case of local energy generation, land use planning ought to also be considered to encourage appropriate developments that facilitate the harmonization of local energy demand and supply profiles, and to ensure energy technologies can be effectively sited within a neighbourhood.  Lastly, the public unfamiliarity with new energy technologies requires that their concerns be considered and where possible ameliorated by local governments. Information campaigns designed with respect to principles from behavioural science and psychology (see Section offer opportunities to encourage public acceptance of technology siting (Gifford et al., 2011; Wolsink, 2010).  6.4 Informing building energy policy from spatial estimates of building energy performance  Although not all the policies listed in this chapter have explicit spatial considerations, a set of policies stand out as strong candidates for integration with geographic information sciences.  The following section presents those policies that can benefit from detailed baseline building energy demand estimates, as explored throughout this dissertation.  These policies include 1) information campaigns and behavioural interventions, 2) developer cost charges, 3) land use   147 planning, and 4) local energy utility planning, energy infrastructure siting and service area bylaws.  6.4.1 Information campaigns and behavioural interventions  Since this dissertation has focused on developing better baseline estimates of building energy demand, it follows that information campaigns that adopt this level of detail can be used to educate building owners, occupants and developers about the various opportunities for energy savings, or simply the estimated existing energy performance of a building.  Touching again on the concept of building energy labeling, estimates of performance parameters can be used to establish an initial database of building performance metrics, which could be updated by building owners, inspectors or trained auditors to reflect the actual conditions of the building.  Sharing and communicating this information through labeling and offers an opportunity to establish a social dialogue on building energy issues, helping to make it a more salient issue.  This may also help to internalize the value of energy efficiency and renewable supply investments during building sector market transactions.  To extend the concept of priming a social dialogue, baseline building energy performance information can also be strategically incorporated with behavioural intervention principles. Framing and social norms exhibit clear links with spatially explicit building energy performance metrics.  For example, framing techniques can be used to compare building energy performance to an ideal case, current technology options or cost effectiveness and cost saving opportunities tailored specifically to the local context.  As another example, social norms can be used by comparing estimated building energy performance for an   148 individual building to neighbouring buildings within a strategically bounded geographic area.  These options are also available by conducting energy audits, although this requires prior engagement or curiosity in energy demand; therefore, spatially contiguous and universally available energy performance automatically generated across an entire city allows more opportunity for local comparisons and is a much more convenient option for helping to make building energy a salient consideration for the public as a whole.  6.4.2 Developer cost charges  Requiring payment from developers to help fund the cost of energy service delivery through DCCs has inherent spatial considerations.  Specifically, designing effective DCC rates must begin with an accurate assessment of the costs associated with various development types and the proximity to existing services and infrastructure.  Spatial analysis is therefore important to locate existing services and patterns of current urban form to determine the physical pathways and associated cost of extending infrastructure.  Land use and density of new developments will also establish the quantity of energy service requirement, therefore accurately estimating costs associated with proposed development types is critical to ensure that DCCs are fair and effective.  For example, designing higher density developments with consistent and moderately high thermal loads can trigger the financial viability of district energy systems, and potentially eliminate the need for gas distribution infrastructure. Spatially contiguous baseline estimates of energy demand in the building sector serve as a reliable scenario for examining the existing cost of service delivery, and can be used as stage to base investigations around the future development proposals and potential options for integration with existing infrastructure. However, as noted earlier, DCCs targeting building   149 energy issues are effectively implemented if the local government is acting as the energy service provider, and therefore endures the financial cost of extending energy infrastructure to new developments.  6.4.3 Land use planning and rezoning  To ensure local energy targets and objectives are achieved and maintained, an initial estimate of building energy demand must be established to gauge improvements and changes over time.  However, providing geographic details of existing building energy performance provides local government officials with added capacity to improve the efficacy of policy design.  Since zoning bylaws are familiar to local planners, they offer unique and easily adaptable options for integration with spatially explicit estimates of building energy demand.  For example, net zero energy zoning bylaws can ensure that energy impacts accompanying new developments are minimized by requiring that those developments not exceed existing energy demand.  However, effectively designing such a bylaw requires knowledge of the existing demand within the spatial constraints of the new zone.  The framework provided in this dissertation therefore presents one approach for providing local planners with the necessary information to prepare novel zoning bylaws.  6.4.4 Local energy utilities, technology siting, and service area bylaws  When local governments act as the energy service provider, with responsibilities for procuring and delivering energy to buildings and developments, a variety of additional spatial considerations also emerge.  These include descriptive assessments of local energy   150 resources, existing infrastructure and service locations and physical limitations for technology and infrastructure siting (Howard et al., 2012; Calvert et al., 2013).  Integrating the aforementioned descriptive elements in a decision-making process also requires prescriptive spatial analyses that address the socio-political limitations and opportunities of energy service provision for various locations.  In this regard, geographic information systems have acted as a solution space that allows local citizens, planning staff and elected officials to engage and interact with local energy planning scenarios, with the intended purpose of leading to more socially acceptable decisions (Calvert et al., 2013; Higgs, Berry, Kidner, & Langford, 2008; Horner, Zhao, & Chapin, 2011; Lind?n et al., 2006; Webster, 1994).  By extension, the descriptive analysis of building performance as presented in this dissertation offers baseline conditions that can be employed to inform local government infrastructure planning and establish the stage for planning alternate energy scenarios.  In addition to infrastructure planning, technology siting and planning scenarios, service area bylaws that mandate the connection to a local energy service provider, must determine the spatial extent where the bylaw is applied while recognizing the economic and social limitations or opportunities in altering these geographic boundaries.  As a result, the availability of spatially explicit building energy conditions is also critical to these planning initiatives, and can be incorporated to limit the assumptions that are typically necessary for establishing service area bylaws.    151 6.5 Conclusion  Direct contact with stakeholders and citizens and the experiential capacity in delivering services and goods to the public place local governments in a unique position to address building energy issues.  Whereas local governments face jurisdictional barriers that may limit the extent to which actions can be taken to address building energy, there nonetheless exist a wide range of policy instruments already available, which can have a substantial contribution to reductions in energy demand or to increases in renewable energy supply.    This Chapter described these building energy policy tools and catalogued them into general categories of voluntary, fiscal, regulatory or capacity building and according to the fundamental strategic mechanisms being employed to elicit actions targeting energy in the building sector.  Voluntary policy instruments encompass information campaigns, with a specific focus on building energy labeling, in addition to subtle behavioural interventions that can be incorporated into existing initiatives to maximize their effectiveness in achieving desired outcomes.  Fiscal policy tools include disincentives, financing and incentives, and present a wide-range of options that local governments can adopt to better address building energy.  Regulatory policies cover land use planning, which focuses on the dominant local strategic mechanism of zoning, in addition to building codes and service area bylaws.  Lastly, empirically identified capacity building policies were identified that offer local governments opportunities to incorporate a wider range of mechanisms for managing reductions in building energy demand or increases in the supply of renewable energy.     152 Given the unique barriers to energy improvement in the building sector, a mix of policy instruments must be evaluated to effectively address building energy.  Local context is critical when strategically identifying this set of policies.  As a result, the potential for informing building energy policies were discussed in the context of detailed, spatially contiguous estimates of baseload energy demand, as offered by the research in this dissertation.  Because the majority of existing literature has tended to focus of policies from higher levels of government, the total effect that local governments can have on building energy remains to be seen.  However, the strategies presented in this review assert that local governments are equipped with a suite of instruments that enable them to tackle a range of concerns and objectives associated with energy in the building sector.  Advancing these strategies requires ongoing efforts to augment local capacity to address building energy and to provide robust descriptive analysis of the existing condition of energy in the building sector at a spatial resolution that targets the site-specific planning and policy mandate of local governments.    153 7 Conclusion  Motivated by the increased requirements and opportunities for local governments to integrate energy into their planning mandate, and the growing popularity of airborne LiDAR as a local planning resource, this dissertation has investigated novel techniques for LiDAR-based assessments of energy demand in the building sector.  Specifically, airborne LiDAR was examined for its ability to:  ? predict building age, which was related to building energy performance parameters,  ? estimate the incoming shortwave radiation on building envelopes, with a focus on the radiation received on building walls and transmitted through trees, and  ? populate a spatially contiguous model of building energy demand by integrating LiDAR-informed building energy parameters and solar gains.    The culmination of these approaches presents a significant contribution to the disciplines of urban geography, geospatial sciences and energy planning by offering techniques for spatially explicit identification of building energy use with applications to site-specific building energy strategies.    Given the primary objective of investigating LiDAR for spatially contiguous building energy modelling, four underlying research questions were posed and addressed in the body of this dissertation.  These research questions were: 1) How is urban and building form, as derived from airborne LiDAR, related to building energy performance?   154 2) How can airborne LiDAR data be used to assess building solar energy gains? 3) How can airborne LiDAR be used to provide spatially contiguous estimates of building energy demand? 4) Which local planning and policy strategies can be informed from citywide estimates of building energy demand?  The findings from each of these research topics are summarized below.  1) How is urban and building form, as derived from airborne LiDAR, related to building energy performance?  To determine how urban and building form can be related to energy performance, three-dimensional representations of the urban surface were generated using airborne LiDAR data and then compared to building age, which was subsequently related to building energy performance parameters.  This research employed a random forests machine learning approach to develop a robust prediction of building age using spatial data common to urban planning agencies and building shape attributes derived from LiDAR.  Results showed that LiDAR derived attributes alone were able to explain 33.5% of the variance of building age with an average error (RMSE) of 16.8 years.  By combining all spatial attributes, the prediction accuracy was increased by 20% (R2 = 40.9) and error reduced to 15.8 years.  Of all the variables included in the model, building height, which was derived from the LiDAR, proved to be the most important metric for predicting age, while the three subsequently ranked variables were the ratio of building area to that of the lot, the lot area itself, and the   155 zone classifier.  LiDAR predictions of building age were then statistically related to building energy performance parameters.    Household energy audit information from over 7000 houses in the City of Vancouver was used to determine the relation between age and building energy components that included envelope resistivity, air leakage and fenestration.  Building age demonstrated a strong correlation to most building energy metrics, with more than 80% of variance explained by age for infiltration, window area, window resistivity, wall resistivity, ceiling resistivity and foundation resistivity for buildings built after 1971.  Relating LiDAR to age, and then relating age to building energy components, was necessary due to the data confidentiality of the energy audit information that restricted the spatial identification of individual homes.  Future research priorities thus include assessment of the direct relation between building form and energy component performance.  2) How can airborne LiDAR data be used to assess building solar energy gains?  LiDAR was used to augment the spatial assessment of building envelope solar energy gains.  This research was comprised of two related research themes: 1) the assessment of radiation on horizontally oriented surfaces, and 2) the assessment of radiation on vertically oriented surfaces.  In the first theme, existing radiation modelling techniques were adapted to include LiDAR-based measurements of radiation transmission through trees and GIS-based algorithms were improved to include remotely sensed spatial and temporal variations in atmospheric conditions.  In the second theme, a novel approach based on airborne LiDAR   156 and two-dimensional building footprints was presented to allow for the computation of irradiance on building walls.    The outcomes of the research revealed that trees tend to be the primary surface feature obstructing building envelopes, and that in the study areas they are more likely to be closer in proximity than solid occluding features, such as buildings.  Moreover, the LiDAR analysis of the impact of trees on incoming radiation showed that representing trees as solid objects results in up to an 18% underestimate of building radiation in treed residential neighbourhoods.  In a second component, the expensive computational cost of estimating irradiance on building walls was revealed leading to the development of generalization techniques.  The point obstruction stacking technique (POSt), which assesses solar occlusions along the vertical edge of buildings, was demonstrated as the most computationally efficient approach for predicting wall irradiance.  The error associated with the POSt approach was 0.84 MJ m-2 day-1 for single-family detached dwellings and 1.16 MJ m-2 day-1 for multi unit buildings, and when  compared to the validation case, provided slight overestimates of shortwave irradiance.  3) How can airborne LiDAR be used to provide spatially contiguous estimates of building energy demand?  To provide spatially contiguous estimates of building energy demand, the outcomes of LiDAR informed building energy performance parameters and solar gains were integrated and used as input to a novel thermal energy demand model.  This integrated approach   157 enabled a novel assessment of both the building-specific and environment-specific variables necessary to provide accurate estimates of building energy demand.  Results from the model aligned closely with outcomes from building energy simulation software (t = -0.11, p=0.91) and with broad national efforts to determine building energy use typologies.  However, in comparison to traditional approaches for citywide building energy modelling, the developed approach does not require the classification of buildings into typologies, and therefore does not suffer from the loss of data variability that accompanies aggregation techniques.  By enabling the identification of energy performance parameters for individual buildings while enabling scaling across large areas, the technique offers a significant contribution to the automation of baseline energy information for staging robust planning and policy strategies.  4) Which local planning and policy strategies can be informed from citywide estimates of building energy demand?  A comprehensive overview of building energy policies for local governments was provided.  Discussion then focused on those policies and planning strategies that can be informed by spatial disaggregated building energy estimates.  While explicit application of citywide individual building energy demand were not apparent for all policies, those of interest included information campaigns and behavioural interventions, developer cost charges (DCC), land use planning, capacity building through locally-owned energy utilities, and service area bylaws; due in large part to their inherent geographic nature.      158 7.1 Innovation  The key innovative elements and novelty developed in this dissertation are associated largely with the identification of site-specific determinant of building energy use.  In particular, within this dissertation: ? Statistical relations were established between building age and building energy performance components ? A new model was developed that accurately estimates solar radiation on all facets of the urban surface and through trees in the urban canopy ? For the first time, the site specific local environment and the internal building energy components were integrated to assess building energy demand for numerous buildings ? An innovative policy review was undertaken, synthesizing all known strategies for local governments to address building energy  7.2 Applications of LiDAR  Given the continued availability and advancement in airborne LiDAR technology and processing techniques, both the operational and scientific applications of this research are important to consider.      159 7.2.1 Local planning  The operational applications of this research are most relevant to the professional practice of planning.  However, although LiDAR science is well established in the natural resource sector, it remains a relatively niche technology for urban planning purposes.  Research efforts focused on expanding LiDAR processing and analysis techniques for a range of planning purposes therefore have the opportunity to encourage more nuanced planning decisions by accounting for spatial variations and phenomena that are otherwise difficult to assess.  The structure and techniques presented in this dissertation provide a framework to develop energy planning scenarios to skilled geospatial scientists, analysts and technicians.  Specifically, the methods and outcomes of this research provide a robust baseline condition of energy demand on which to examine the impacts of various building energy conservation policies, and for siting energy generating technologies.  7.2.2 Arboriculture and urban forestry  Results of simulations by Simpson and McPherson (2006) indicate that the strategic planting of several trees can produce up to a 50% decrease in cooling energy use and a 23% decrease in peak electricity use.  Similarly, Huang et al. (1987) show that a 25% increase in tree cover results in a 25 ? 40% decrease in building cooling energy use, and Laband and Sophocleus (2009) demonstrate that a house with full exposure, uses 2.6 times the cooling energy as an identical building in close to complete shade.  This potential to reduce electricity   160 consumption also encourages the added secondary benefit of using shade trees to mitigate carbon emissions.  Akbari (2002), for example, explains that a shade tree can offset the carbon emissions associated with power plants at a rate of up to 4 times its physiological ability to sequester carbon.  In contrast to potential energy demand reductions from tree shading strategies, some jurisdictions have regulated the planting and trimming of trees to ensure solar rooftop technology investments are maximized.    Although it may be possible to design a tree planting strategy that can satisfy both an energy conservation initiative while maximizing sunlight on local and neighbouring rooftops, such a situation requires a detailed analysis of the local environment and form.  LiDAR based radiation and building energy modelling approaches therefore offer opportunities to better understand strategic planting strategies for individual trees in built environments as part of local energy and emissions goals.  7.2.3 Urban climate science and architecture  The application of LiDAR for detailed irradiance modelling in complex environments is also relevant for advancing scientific knowledge in the disciplines of urban climatology, urban meteorology (Arnfield, 2003) and building architectural sciences (Crawley et al., 2008).  Existing analysis and techniques in these fields tend to either aggregate surface irradiance within a broader scale analysis, as in the example of directionally averaged urban canyons in urban climatology and meteorology (Arnfield, 2003; Martilli et al., 2002; Masson, 2000), or neglect the spatial variation in the form of the local environment altogether, as is the case in   161 many energy simulation models (Crawley et al., 2008; Loutzenhiser et al., 2007).  As a result, LiDAR technology and processing techniques, and specifically the spatial irradiance model presented in this dissertation enable advances in active scientific fields focused on urban environments.  7.2.4 Rooftop solar technology placement  Solar energy generating technologies can be divided into two basic categories: those that produce electricity, and those that produce heat.  The electricity generating technologies are photovoltaic (PV) cells that convert solar radiation into direct current electricity.  Solar heating systems are comprised of solar thermal collectors that capture heat energy from the sun for use in a variety of heating applications.  On a small-scale, solar heating systems are typically used in water heating for domestic purposes and in some cases for indoor space heating.  Given the approaches developed in this dissertation, ideal rooftop solar technology placement can be assessed for individual homes.  A common outlet for communicating rooftop solar energy potential and placement has been to develop online mapping tools (e.g. Solar Boston (http://gis.cityofboston.gov/solarboston), the District of North Vancouver?s Solar Calculator (http://geoweb.dnv.org/applications/solarapp), and the New York City Solar Map (http://www.nycsolarmap.com/)), which is also facilitated by generating contiguous spatial estimates of rooftop received solar irradiance.    162 7.3 Limitations  While this dissertation presents novel techniques for building energy assessment and provides significant contributions to spatial irradiance modelling by accounting for light transmission through trees in the urban canopy and shortwave radiation received on building walls, limitations are important to note.   In the model-focused research presented in Chapters 3, 4 and 5, error was determined using best available validation data, which ranged from local empirical ground measurements to comparative results from scientifically robust studies.  While error is present in any model that abstracts complex natural and human processes, the limitations on the applications of this dissertation due to error relate to the decision-making certainty and confidence for a single building.  Although the methods presented here account for spatial variation not generally available in traditional building simulation software, selection of a single cost-effective energy retrofit for an existing individual building is still best achieved using manual audits.  Approached from a geographic perspective, one of the fundamental goals of this dissertation was to develop and validate procedures to better account for the spatial variation of environmental and building construction determinants of energy use.  This general objective was achieved by integrating accurate three-dimensional measurements of the urban surface using airborne LiDAR.  However, there remain several variables that are assumed as static across the landscape, despite known spatial variation.  These variables include wind,   163 temperature, humidity, air pressure, coefficients determining the natural ventilation rate and convective heat transfer at the building surface, albedo, emissivity and the anisotropic quality of the diffuse sky.  However, the sensitivity of the models and the degree to which these variables fluctuate across space are minimal compared to radiation fluxes and building energy components.   The research presented here focused on technological and environmental conditions of building energy; however, behaviour remains a critical determinant of energy end use that is beyond the scope of this dissertation.  While occupant behaviour has been suggested to have a similar contribution to end building energy use as building design and environmental conditions (Ratti et al., 2005), it is a stochastic process that is prohibitively difficult to empirically model and apply at the level of individual buildings.  Nonetheless, one potential option for future modelling efforts is to compare utility data collected from ?smart meters? to predicted energy use.  This comparison could allow insights to be gathered regarding the spatial nature of deviations from nominal or expected energy use, with subsequent policy implications.  7.4 Directions for future work  Benefits of a LiDAR-based approach to building energy modelling include the detailed representation of the local urban form, the high potential for automation and a more detailed identification of the spatial variation in energy parameters that are important to better inform   164 local building energy planning and policy.  Based on the research findings in the thesis, a number of areas of future research have emerged of particular interest.    7.4.1 LiDAR flight planning   One area that has been underexplored is the sensitivity of LiDAR-based models resulting from a range of LiDAR specifications and accuracy.  Issues to consider include leaf-off and leaf-on vegetation conditions, point density, flight patterns, and the vertical and horizontal accuracy of the LiDAR returns.  To further investigate the influence of these technology and acquisition focused research questions on model accuracy, a combination of multiple LiDAR flights with statistics-based data reduction and filtering approaches are recommended to provide a diversity of dataset upon which to test the outcomes of the LiDAR-informed models.  The consequences of this research would have direct implication for LiDAR procurement proposals and vendor flight planning.  7.4.2 Optimization of solar panel placement  As mentioned above, rooftop solar placement strategies can be informed using LiDAR-based solar modelling approaches.  However, all known examinations of technology placement strategies focus on locating solar panels on unobstructed surfaces oriented to the equator.  While this strategy maximizes the energy output of the panel, it neglects a more holistic view of energy system management.  Notably, energy demand profiles do not typically track solar radiation intensity.  This has implications for grid management and local energy security.    165 For example, some European countries have required that baseload energy generating facilities shutdown, or ramp up and down, due to the requirement to accommodate the abundance of solar generated electricity.  As a result, utilities have claimed that baseload power plants are physically strained, their financial viability threatened, and that the potential long-term implications of these strategies will ultimately result in an unreliable electricity grid (?European utilities,? 2013).  Future research is therefore recommended to examine spatial optimization strategies for the placement of solar panels that compliment demand profiles.  In such a study, LiDAR-based methods would enable the accurate representation of the physical constraints for panel placement, while optimization procedures can then be implemented to examine ideal placement strategies applied to a collection of buildings across large areas.     7.4.3 ?Big data? integration and remote sensing data fusion  As faced in many studies examining physical and natural processes in complex built environments, validation techniques remain difficult to administer due to the inaccessibility of private space and the extreme heterogeneity of the landscape.  In tandem with the growing availability of airborne LiDAR data over urban landscapes, the integration of networked devices (e.g. mobile phones, utility meters, thermostats) offers exciting future research and validation opportunities to advance our understanding of building energy use.  Future research opportunities and strategies covering this topic are wide ranging, but one approach of emerging interest for scientific studies is the use of crowdsourcing, where citizens are asked to collect and share data using ubiquitous or purpose-specific sensors and devices.    166 Crowdsourcing provides an ideal compliment to remote sensing datasets in urban areas, as it enables opportunities for collecting information in private space not directly observable by traditional remote sensors.   Finally, it is important to recognize other remote sensing data products that may compliment the information provided from LiDAR technologies.  Traditionally, aerial photography has been used to support urban planning and management.  Advances to aerial imagery that enable the digital capture of visible and near-infrared spectra are now commonly available from remote sensing data vendors.  This digital data captures unique spectral ranges that can facilitate greater automation when conducting feature classifications.  Furthermore, sensing of the thermal spectra offers novel opportunities to collect measurements of the heat energy emmitted from a building, which has also been used to support local engagement and educational campaigns in support of energy efficiency (Hay et al., 2011), and may be used to help validate the LiDAR-based approach presented in this dissertation.      167 References  Aguilar C, White DJ, Ryan DL. (2005). Domestic water heating and water heater energy  consumption in Canada. Canadian Building Energy End-Use Data and Analysis Centre 2005; Available online at:  http://sedccoalition.eu/wpcontent/uploads/2011/07/ CREEDAC-Canadian-Residential-Hot-Water-Apr-2005.pdf  Akbari, H. (2002). Shade trees reduce building energy use and CO2 emissions from power  plants.  Environmental Pollution 116, 119-126.  Allcott, H., & Mullainathan, S. (2010). Behavioral science and energy policy. Science,  327(5970), 1204-1205.  Arboit, M., Diblasi, a, Fernandezllano, J., Derosa, C. (2008). Assessing the solar potential of  low density urban environments in Andean cities with desert climates: The case of the city of Mendoza, in Argentina. Renewable Energy 33(8), 1733-1748.  Arnfield AJ. (2003). Two decades of urban climate research: a review of turbulence,  exchanges of energy and water, and the urban heat island. International Journal of Climatlology 23, 1-26.  Aydinalp, M., Ugursal, V., & Fung, A. (2004). Modeling of the space and domestic hot- water heating energy-consumption in the residential sector using neural networks. Applied Energy 79(2), 159?178.   Batey, P., Brown, P. (2007).  The spatial targeting of urban policy initiatives: a  geodemographic assessment tool.  Environment and Planning A, 2774-2793.  Betsill, M. M., & Bulkeley, H. (2006). Cities and the multilevel governance of global climate  change. Global Governance: A Review of Multilateralism and International Organizations 12(2), 141-159.  Biaou A, Bernier M. (2008). Achieving total domestic hot water production with renewable  energy. Building and Environment 43(4), 651?660.  Blanco-Muriel, M., Alarcon-Padilla, D.C., Lopea-Moratalla, T., Lara-Coira, M. (2001).  Computing the solar vector. Solar Energy 70(5), 431?441.  Boswall, R.G. (2005). Municipal liability for building permits and inspections. Clark Wilson  LLP.  Breiman, L. (1996).  Bagging predictors. Machine Learning 24, 123-140.  Breiman, L. (2001).  Random forests. Machine Learning 45, 15-32.    168 Brody, S. D., Zahran, S., Grover, H., Vedlitz, A. (2008). A spatial analysis of local climate  change policy in the United States: Risk, stress, and opportunity. Landscape and Urban Planning, 87(1), 33-41.  Brown, M. A. (2001). Market failures and barriers as a basis for clean energy policies.   Energy Policy 29, 1197?1207.  Burch, S. (2010). Transforming barriers into enablers of action on climate change: Insights  from three municipal case studies in British Columbia, Canada. Global Environmental Change 20(2), 287-297.  Calvert, K., Pearce, J. M., Mabee, W. E. (2013). Toward renewable energy geo-information  infrastructures: Applications of GIScience and remote sensing that build institutional capacity. Renewable and Sustainable Energy Reviews 18, 416?429.   CMHC Canadian Mortgage and Housing Corporation. Analysis of the annual energy and  water consumption of apartment buildings in the CMHC HiSTAR database. Research Highlights, Technical Series 01-142. Available online at: http://www.cmhc-schl.gc.ca/publications/en/rh-pr/tech/tech01-142.htm  Caputo P, Costa G, Ferrari S. (2013). A supporting method for defining energy strategies in  the building sector at urban scale. Energy Policy 55, 261?270.   Caruana, R., Niculescu-Mizil, A. (2006).  An empirical comparison of supervised learning  algorithms.  Proceedings of the 23rd International Conference on Machine Learning, Pittsburgh, PA.  CEEI Community Energy and Emissions Inventory (2010).  Reports. Available online at:  http://www.env.gov.bc.ca/cas/mitigation/ceei/reports.html  Christen, A., Vogt, R. (2004). Energy and radiation balance of a central European city.   International Journal of Climatology 24, 1395-1421.  City of Vancouver (2009).  Drinking Water Quality: Annual Report. Available online at: http://vancouver.ca/engsvcs/watersewers/water/pdf/WQReport_2009.pdf  City of Vancouver (2012).  Greenest City 2020 Action Plan.  Available online at:  http://vancouver.ca/files/cov/Greenest-city-action-plan.pdf  Coglianese, C., Nash, J., Olmstead, T. (2003). Performance-based regulation: Prospects and  limitations in health, safety, and environmental protection. Administrative Law Review 55, 705-730.    169 Collier, U. (1997). Local authorities and climate protection in the European union: Putting  subsidiarity into practice? Local Environment: The International Journal of Justice and Sustainability 2(1), 39-57.  Community Energy Association CEA (2008). Policy and Governance Tools: Renewable  Energy Guide for Local Governments in British Columbia. Vancouver.  Coops, N C, Hilker, T., Wulder, M. A., St-Onge, B., Newnham, G. J., Siggins, A.,  Trofymow, J. A. T. (2007). Estimating canopy structure of Douglas-fir forest stands from discrete-return LiDAR. Trees-Structure and Function 21(3), 295?310.  Coors V. (2003).  3D GIS in networking environments.  Computers, Environment and Urban  Systems 27(4), 345-357.  Costa, D,L., Kahn, M.E. (2010).  Why has California?s residential electricity been so flat  since the 1980s? A microeconomic approach.  NBER working paper 15978.  Crawley, D., Hand, J., Kummert, M., & Griffith, B. (2008). Contrasting the capabilities of  building energy performance simulation programs. Building and Environment 43(4), 661?673.  Curran, D. (2010).  Juridiction options for energy efficiency and renewable energy in  buildings.  Report for the Pembina Institute, 43 pages.  Cutler, D.R., Edwards, T.C., Beard, K.H., Cutler, A., Hess, K.T., Gibson, J., Lawler, J.J.  (2007).  Random forests for classification in ecology.  Ecology 88(11), 2783-2792.  Cutler, F., Matthews, J. S. (2005). The challenge of municipal voting: Vancouver 2002.  Canadian Journal of Political Science 38(2), 359-382.  Cutler, A., Stevens, J.R. (2006).  Random forests for microarrays.  Methods in Enzymology  411, 422-432.  Davenport, I. J., Holden, N., Gurney, R. J. (2004). Characterizing errors in airborne laser  altimetry data to extract soil roughness. Geoscience and Remote Sensing, IEEE Transactions on 42(10), 2130-2141.  DeOreo WB, Mayer PW. (2000). The End Uses of Hot Water in Single Family Homes from  Flow Trace Analysis. Aquacraft Inc. Report.  Duguay, C. R. (1993). Radiation modeling in mountainous terrain review and status.  Mountain Research and Development 13(4), 339-357.  Eid, T., Gobakken, T., N?sset, E. (2004). Comparing stand inventories for large areas based  on photo-interpretation and laser scanning by means of cost-plus-loss analyses. Scandinavian Journal of Forest Research 19(6), 512-523.    170 European utilities: How to lose half a trillion euros. (2013, October 12).  The Economist, 27- 30.  Ewing, R., Rong, F. (2008). The impact of urban form on US residential energy use.  Housing Policy Debate 19(1), 1?30.   Farahbakhsh, H., Ugursal, V., Fung, A. (1998). A residential end-use energy consumption  model for Canada. International journal of energy research 22(13), 1133?1143.   Finn, J.G. (2008).  Building stronger local governments and regions: an action plan for the  future of local governance in New Brunswick.  Report of the Commissioner on the Future of Local Governance, 196 pages.  Finney KN, Sharifi VN, Swithenbank J, Nolan A, White S, Ogden S. (2012). Developments  to an existing city-wide district energy network ? Part I: Identification of potential expansions using heat mapping. Energy Conversion and Management 62, 165?175.  Friedman, J. H. (2001). Greedy function approximation: a gradient boosting machine. Annals  of Statistics, 1189-1232.  Fu P, Rich P. (1999). Design and implementation of the Solar Analyst: an ArcView  extension for modeling solar radiation at landscape scales. Proc. of the 19th Annual ESRI User Conference. San Diego, USA.  Gifford, R., Kormos, C., McIntyre, A. (2011). Behavioral dimensions of climate change:  drivers, responses, barriers, and interventions. Wiley Interdisciplinary Reviews: Climate Change.  Goodwin, N.R., Coops, N.C., Tooke, T.R., Christen, A., Voogt, J.A. (2009). Characterising  urban surface cover and structure with airborne LiDAR technology. Canadian Journal of Remote Sensing 35, 297-309.  Gore, C. D. (2010). The limits and opportunities of networks: Municipalities and Canadian  climate change policy. Review of Policy Research 27(1), 27-46.  Grena, R. (2008). An algorithm for the computation of the solar position. Solar Energy 82(5),  462-470.   Grimmond, C. S. B., Potter, S. K., Zutter, H. N., Souch, C. (2001). Rapid methods to  estimate sky-view factors applied to urban areas. International Journal of Climatology 21(7), 903-913.   Gros, A., Bozonnet, E., & Inard, C. (2011).  Modelling the radiative exchanges in urban  areas?: A review. Advances in Building Energy Research 5(1), 163-206.    171 Gueymard CA, Myers LB. (2008). Solar radiation measurement: progress in radiometry for  improved modeling. In: Badescu, V. (Ed.), Modeling Solar Radiation at the Earth Surface. Springer-Verlag, Berlin, pp. 479-511.   Guo, L., Chehata, N., Mallet, C., Boukir, S. (2011). Relevance of airborne lidar and  multispectral image data for urban scene classification using Random Forests.  ISPRS Journal of Photogrammetry and Remote Sensing, 56-66.  Gurtuna, O., Pr?vot, A. (2011). An Overview of Solar Resource Assessment Using  Meteorological Satellite Data. In Proceedings of the Recent Advances in Space Technologies Conference, Istanbul, Turkey.  Guy, S., Marvin, S. (1996). Disconnected policy: the shaping of local energy management.  Environment and Planning C 14, 145-158.  Hagishima, A., Tanimoto, J. (2003). Field measurements for estimating the convective heat  transfer coefficient at building surfaces. Building and Environment 38(7), 873?881.   Hagishima A, Tanimoto J, Narita K. (2005). Intercomparisons of experimental convective  heat transfer coefficients and mass transfer coefficients of urban surfaces. Bound-Layer Meteorolology 117(3), 551?576.  Haltrecht D, Zmeureanu R, Beausoleil-Morrison I. (1999). Defining the methodology for the  next-generation HOT2000 simulator. Proceedings of Building Simulation'99. Vol. 1.  Hammer, A., Heinemann, D., Hoyer, C., Kuhlemann, R., Lorenz, E., Muller, R., Beyer, H. G.  (2003). Solar energy assessment using remote sensing technologies. Remote Sensing of Environment 86(3), 423?432.   Harris, B. (1989). Beyond geographic information systems. Journal of the American  Planning Association 55(1), 85?90.   Harris, R., Sleight, P., Webber, R. (2005).  Geodemographics, GIS and Neighbourhood  Targeting. John Wiley & Sons, West Sussex.  Hay, G. J., Kyle, C., Hemachandran, B., Chen, G., Rahman, M. M., Fung, T. S., & Arvai, J.  L. (2011). Geospatial technologies to improve urban energy efficiency. Remote Sensing 3(7), 1380-1405.  Hay, J. E., McKay, D. C. (1985). Estimating solar irradiance on inclined surfaces: a review  and assessment of methodologies. International Journal of Solar Energy 3(4-5), 203-240.  Heiple S, Sailor D. (2008). Using building energy simulation and geospatial modeling  techniques to determine high resolution building sector energy consumption profiles. Energy and Buildings 40(8), 1426?1436.    172 Herold, M., Scepan, J., Clarke, K. C. (2002). The use of remote sensing and landscape  metrics to describe structures and changes in urban land uses. Environment and Planning A 34(8), 1443?1458.  Hickman, G.D., Hogg, J.E. (1969).  Application of an airborne pulsed laser for near shore  bathymetric measurements.  Remote Sensing of Environment 1, 47-58.   Higgs, G., Berry, R., Kidner, D., Langford, M. (2008). Using IT approaches to promote  public participation in renewable energy planning: Prospects and challenges. Land Use Policy 25(4), 596?607.   Hofierka, J., Ka?uk, J. (2009). Assessment of photovoltaic potential in urban areas using  open-source solar radiation tools. Renewable Energy 34(10), 2206-2214.  Hofierka, J., Suri, M. (2002). The solar radiation model for open source GIS: Implementation  and applications. International GRASS users conference, Trento, Italy (pp. 11-13).  Horner, M. W., Zhao, T., Chapin, T. S. (2011). Toward an integrated GIScience and energy  research agenda. Annals of the Association of American Geographers 101(4), 764-774.   Howard, B., Parshall, L., Thompson, J., Hammer, S., Dickinson, J., Modi, V. (2012). Spatial  distribution of urban building energy consumption by end use. Energy and Buildings 45, 141?151.    Huang, Y.J., Akbari, H., Taha, H., Rosenfeld, A.H. (1987). The potential of vegetation in  reducing summer cooling loads in residential buildings. Journal of Applied Meteorology 26(9), 1103?1116.  Huijts, N. M. A., Molin, E. J. E., Steg, L. (2012). Psychological factors influencing  sustainable energy technology acceptance: A review-based comprehensive framework. Renewable and Sustainable Energy Reviews, 525-531.  Ineichen, P., Perez, R. (1999). Derivation of cloud index from geostationary satellites and  application to the production of solar irradiance and daylight illuminance data. Theoretical and Applied Climatology 64(1-2), 119-130.   IEA International Energy Agency. (2013). World Energy Outlook 2013. Available online at  http://www.worldenergyoutlook.org  Iqbal, M. (1983). An introduction to solar radiation. Access Online via Elsevier.  Jaccard, M. (2005). Sustainable Fossil Fuels. Cambridge University Press: New York.    173 Jaccard, M., Dennis, M. (2006). Estimating home energy decision parameters for a hybrid  energy?economy policy model. Environmental Modeling and Assessment 11(2), 91-100.  Jaccard, M., Failing, L., Berry, T. (1997). From equipment to infrastructure: community  energy management and greenhouse gas emission reduction. Energy Policy 25(13), 1065-1074.  Jaccard, M., Melton, N., Nyboer, J. (2011). Institutions and processes for scaling up  renewables: Run-of-river hydropower in British Columbia. Energy Policy 39(7), 4042-4050.  Jacobsen, GD., Kotchen, MJ (2013).  Are building codes effective at saving energy?  Evidence from residential billing data in Florida.  The Review of Economics and Statistics 95, 34-49.  Jacovides, C., Tymvios, F., Assimakopoulos, V., Kaltsounides, N. (2006). Comparative study  of various correlations in estimating hourly diffuse fraction of global solar radiation. Renewable Energy 31(15), 2492-2504.  Jaffe, A. B., Stavins, R. N. (1994). The energy-efficiency gap. What does it mean? Energy  Policy 22(10), 804-810.  J?uregui E., Luyando E. (1999). Global radiation attenuation by air pollution and its effects  on the thermal climate in Mexico City. International Journal of Climatology 19(6), 683?694.  Jones, E., Leach, M., Wade, J. (2000). Local policies for DSM: the UK?s home energy  conservation act. Energy Policy 28(3), 201?211.   Jorgensen B, Graymore M, O?Toole K. (2009). Household water use behavior: an integrated  model. Journal of Environmental Management 1, 227-236.  Kallbekken, S., S?len, H. (2011). Public acceptance for environmental taxes: Self-interest,  environmental and distributional concerns. Energy Policy 39(5), 2966-2973.  Kanellakis, M., Martinopoulos, G., Zachariadis, T. (2013). European energy policy?A  review. Energy Policy 62, 1020-1030.  Karger, C. R., Hennings, W. (2009). Sustainability evaluation of decentralized electricity  generation. Renewable and Sustainable Energy Reviews 13(3), 583-593.  Kasten, F. (1996). The linke turbidity factor based on improved values of the integral  Rayleigh optical thickness. Solar Energy 56(3), 239-244.    174 Kasten, F., Young, A. T. (1989). Revised optical air mass tables and approximation formula.  Applied Optics 28(22), 4735?4738.  Kaufman, Y. J., Tanr?, D., Boucher, O. (2002). A satellite view of aerosols in the climate  system. Nature 419(6903), 215-223.  Kavgic M, Mavrogianni A, Mumovic D, Summerfield A, Stevanovic Z, Djurovic-Petrovic  M. (2010). A review of bottom-up building stock models for energy consumption in the residential sector. Building and Environment 45(7), 1683?97.  Keirstead J, Jennings M, Sivakumar A. (2012).  A review of urban energy system models:  approaches, challenges and opportunities. Renewable and Sustainable Energy Reviews 16(6), 3847?66.  Kellett R, Christen A, Coops NC, VanderLaan M, Crawford B, Tooke TR, Olchovski I.  (2013). A systems approach to carbon cycling and emissions modeling at an urban neighborhood scale. Landscape and Urban Planning 110, 48?58.  Kitchen, H. M., & Slack, E. (2003). Special study: new finance options for municipal  governments. Policy 309, 28.  Kolbe TH.  (2007). Representing and exchanging 3D city models with CityGML.  In: Lee J,  Zlatanova S, editors. 3D Geo-Information Science. Berlin: Springer-Verlag; 2007, p. 15-32.  Krayenhoff E.S., Voogt JA. (2007). A microscale three-dimensional urban energy balance  model for studying surface temperatures, Boundary-Layer Meteorology 123, 433-461.  Kumar, L., Skidmore, A. K., Knowles, E. (1997). Modelling topographic variation in solar  radiation in a GIS environment. International Journal of Geographical Information Science 11(5), 475-497.  Laband, D.N., Sophocleus, J.P. (2009). An experimental analysis of the impact of tree  shade on electricity consumption. Arboriculture and Urban Forestry 35(4), 197?202.  Lath J.K., Boileau G.G. (1969). Heat losses from house basements. Canadian Building  19(10), 39.  Laustsen, J (2008).  Energy efficiency requirements in building codes, energy efficiency  policies for new buildings. International Energy Agency Information Paper, 85 pages.  Lee J. (2007). A three-dimensional navigable data model to support emergency response in  microspatial built-environments.  Annals of the Association of American Geographers 97(3), 512-529.    175 Lee, W., & Yik, F. (2004). Regulatory and voluntary approaches for enhancing building  energy efficiency. Progress in Energy and Combustion Science 30(5), 477?499.   Levinson, R., Akbari, H., Pomerantz, M., Gupta, S. (2009). Solar access of residential  rooftops in four California cities. Solar Energy 83(12), 2120?2135.  Liaw, A., Wiener, M. (2002). Classification and Regression by randomForest. R News 2(3),  18-22.  Lindberg, F., Grimmond, C. S. B. (2011). The influence of vegetation and building  morphology on shadow patterns and mean radiant temperatures in urban areas: model development and evaluation. Theoretical and Applied Climatology 105, 1?13.  Lind?n, A.-L., Carlsson-Kanyama, A., & Eriksson, B. (2006). Efficient and inefficient  aspects of residential energy behaviour: What are the policy instruments for change? Energy Policy 34(14), 1918?1927.   Loutzenhiser P.G, Manz H, Felsmann C, Strachan PA, Frank T, Maxwell G.M. (2007).  Empirical validation of models to compute solar irradiance on inclined surfaces for building energy simulation. Solar Energy 81, 254-267.  Loveday D.L., Taki A.H. (1998). Outside surface resistance: proposed new value for building  design. Building Services Engineering Research and Technology 19(1), 23?29.  Lovell, J., Jupp, D. L., Culvenor, D., Coops, N. C. (2003). Using airborne and ground-based  ranging lidar to measure canopy structure in Australian forests. Canadian Journal of Remote Sensing 29(5), 607-622.  Lutsey, N., Sperling, D. (2008). America's bottom-up climate change mitigation policy.  Energy Policy 36(2), 673-685.  Lutz, J., Liu, X., McMahon, J., Dunham, C., Shown, L., McGrue, Q.  (1996). Modeling  patterns of hot water use in households. Lawrence Berkeley National Laboratory. Available online at: http://eetd.lbl.gov/publications/modeling-patterns-hot-water-use-households  L?tzkendorf, T., Speer, T. M. (2005). Alleviating asymmetric information in property  markets: building performance and product quality as signals for consumers. Building Research & Information 33(2), 182-195.  Lwin, K., Murayama, Y. (2009). A GIS approach to estimation of building population for  micro-analysis. Transactions in GIS 13(4), 401-414.  MacNab, J., Cobb, P., Bailie, A., Beckstead, C., Horne, M., Frappe-Seneclauze, T-P. (2011).  A renewable energy requirement for BC. The Pembina Institute.    176 Martilli, A., Clappier, A., Rotach, M. W. (2002). An urban surface exchange  parameterisation for mesoscale models. Boundary-Layer Meteorology 104(2), 261-304.  Masson, V. (2000). A physically-based scheme for the urban energy budget in atmospheric  models. Boundary-Layer Meteorology 94(3), 357-397.  Menzel, W. P., Purdom, J. F. (1994). Introducing GOES-I: The first of a new generation of  geostationary operational environmental satellites. Bulletin of the American Meteorological Society 75(5), 757-781.  Merk, O., Saussier, S., Staropoli, C., Slack, E., Kim, J.H. (2012). Financing Green Urban  Infrastructure. OECD Regional Development Working Papers 2012/10, OECD Publishing; http://dc.doi.org/10.1787/5k92p0c6j6r0-en  Meyer, J., Tshimankinda, M. (1997). Domestic hot water consumption in South African  houses for developed and developing communities. International Journal of Energy Research 21(7), 667?73.  Michalsky, J.J. (1988). The astronomical Almanac?s algorithm for approximate solar position  (1950?2050). Solar Energy 40(3), 227?235.  Miller, N. (2013). Urban form and building energy: quantifying relationships using a multi- scale approach.  PhD Dissertation available online at: https://circle.ubc.ca/handle/2429/44744.  Molenaar M. (1990). A formal data structure for 3D vector maps. Proceedings of EGIS?90  1990, 2: 770-781.  Muneer, T. (1990). Solar radiation model for Europe. Building Services Engineering  Research and Technology 11(4), 153-163.  Neidhart, H., Sester, M. (2004). Identifying building types and building clusters using 3D- laser scanning and GIS-data. Geo-Imagery Bridging Continents, XXth ISPRS Congress, Istanbul (pp. 715?720). Available online at: http://cartesia.org/geodoc/isprs2004/comm4/papers/442.pdf  Nelson, R. (2013). How did we get here? An early history of forestry lidar. Canadian Journal  of Remote Sensing 39(S1), S1-S12.  Ni-Meister, W., Jupp, D. L. B., Dubayah, R. (2001). Modeling lidar waveforms in  heterogeneous and discrete canopies. IEEE Transactions on Geoscience and Remote Sensing 39(9), 1943-1958.  Oke, T. R. (1982). The energetic basis of the urban heat island. Quarterly Journal of the  Royal Meteorological Society 108(455), 1-24.   177  Oke, T.R. (1987). Boundary Layer Climates. Methuen and Co. Ltd: London.  Oke TR. (1988). The urban energy balance. Progress in Physical Geography 12, 471-508.  Parekh, A. (2005). Development of archetypes of building characteristics libraries for  simplified energy use evaluation of houses. In Proc. 9th Int IBPSA Conf (pp. 921-928).  Perez, R., Ineichen, P., Moore, K., Kmiecik, M., Chain, C., George, R.,Vignola, F. (2002). A  new operational model for satellite-derived irradiances: description and validation. Solar Energy 73, 307?317.  Polo, J., Zarzalejo, L. F., Ram?rez, L. (2008). Solar radiation derived from satellite images. In  Modeling Solar Radiation at the Earth?s Surface (pp. 449-462). Springer Berlin Heidelberg.   Ratti, C., Baker, N., Steemers, K. (2005). Energy consumption and urban texture. Energy and  Buildings 37(7), 762?776.  Ravetz, J. (2008). State of the stock?What do we know about existing buildings and their  future prospects? Energy Policy 36(12), 4462-4470.  Reda, I., & Andreas, A. (2004). Solar position algorithm for solar radiation applications.  Solar Energy 76(5), 577-589.  Remund, J., Wald, L., Lef?vre, M., Ranchin, T., Page, J. (2003). Worldwide Linke turbidity  information. ISES Solar World Congress. Goteborg, Sweden.  Ria?o, D., Chuvieco, E., Cond?s, S., Gonz?lez-Matesanz, J., Ustin, S. L. (2004). Generation  of crown bulk density for Pinus sylvestris L. from lidar. Remote Sensing of Environment 92(3), 345-352.  Ria?o, D., Meier, E., Allg?wer, B., Chuvieco, E., Ustin, S. L. (2003). Modeling airborne  laser scanning data for the spatial generation of critical forest parameters in fire behavior modeling. Remote Sensing of Environment 86(2), 177?186.  Rich, P.M., Dubayah, R., Hetrick, W.A., Saving, S.C. (1994). Using Viewshed models to  calculate intercepted solar radiation: applications in ecology. American Society for Photogrammetry and Remote Sensing Technical Papers, 524-529.  Richardson, B. (2012). Local climate change law. In B. Richardson (Ed.), Local Climate  Change Law: Environmental Regulation in Cities and Other Localities (pp. 3-28). Cheltenham, UK: Edward Elgar Pub.    178 Rigollier C., Lef?vre M., Wald L., (2004). The method Heliosat-2 for deriving shortwave  solar radiation data from satellite images. Solar Energy 77(2), 159-169.  Roaf, S., Crichton, D., Nicol, F. (2005). Adapting buildings and cities for climate change.  Oxford: Architectural Press.  Robinson, P. J., Gore, C. D. (2005). Barriers to Canadian municipal response to climate  change. Canadian Journal of Urban Research 14(1), 102-121.  Robinson, D. (2006). Urban morphology and indicators of radiation availability. Solar  Energy 80(12), 1643?1648.   Robinson, D., Stone, A. (2004). Solar radiation modelling in the urban context. Solar Energy  77(3), 295?309.  Rodr?guez, E., Morris, C. S., Belz, J. E. (2006). A Global Assessment of the SRTM  Performance. Photogrammetric Engineering & Remote Sensing 72(3), 249-260.  Ryghaug, M., S?rensen, K. H. (2009). How energy efficiency fails in the building industry.  Energy Policy 37(3), 984?991.   Rylatt, M., Gadsden, S., Lomas, K. (2001). GIS-based decision support for solar energy  planning in urban environments. Computers, Environment and Urban Systems 25(6), 579-603.  Rylatt, R., Gadsden, S., Lomas, K. (2003). Methods of predicting urban domestic energy  demand with reduced datasets: a review and a new GIS-based approach. Building Services Engineering Research and Technology 24(2), 93?102.  Sander, H., Manson, S. M. (2007). Heights and locations of artificial structures in viewshed  calculation: How close is close enough? Landscape and Urban Planning 82(4), 257-270.  Scharmer, K., Greif, J., (2000). The European solar radiation atlas. Vol. 2: Database and  exploitation software. Paris (Les Presses de l'?cole des Mines).  Schwartz, E. (2012). Local solutions to a global problem? Climate change policy-making in  Vancouver. In B. Richardson (Ed.), Local Climate Change Law: Environmental Regulation in Cities and Other Localities (pp. 3-28). Cheltenham, UK: Edward Elgar Pub.  Senbel M, Church S. (2010). The relationship between urban form and GHG emissions. The  University of British Columbia. Available online at: https://circle.ubc.ca/handle/2429/39750    179 Senbel, M., Fergusson, D., Stevens, M. (2012). Local responses to regional mandates:  assessing municipal greenhouse gas emissions reduction targets in British Columbia. Sustainability: Science, Practice, & Policy 9(1), 1012-068.  Shah, A., Shah, S. (2006). The new vision of local governance and the evolving roles of local  governments. In: Shah, A. (Ed.), Local Governance in Developing Countries. World Bank, Washington, D.C.  Sheppard, S. R. (2012). Visualizing climate change: a guide to visual communication of  climate change and developing local solutions. Routledge.  Sherman M. (1987). Estimation of infiltration from leakage and climate indicators.  Energy  and Buildings 10, 81-86.  Shufelt, J.A. (1999). Performance evaluation and analysis of monocular building extraction  from aerial imagery.  IEEE Transactions on Pattern Analysis and Machine Intelligence 21, 311-326.  Schlueter, A., Thesseling, F. (2009). Building information model based energy/exergy  performance assessment in early design stages. Automation in Construction 18(2), 153-163.   Simpson, J.R., McPherson, E.G. (1996). Potential of tree shade for reducing residential  energy use in California. Journal of Arboriculture 22, 10?18.  Statistics Canada (2012). The Canadian population in 2011: Population counts and growth.  Statistics Canada, ISBN 978-1-100-19962-7. Retrieved 2012-05-30.  St Denis, G., Parker, P. (2009). Community energy planning in Canada: The role of  renewable energy. Renewable and Sustainable Energy Reviews 13(8), 2088-2095.  Stull, R. B. (2011). Meteorology for Scientists and Engineers (in prep.). Draft of 3rd Edition.  Chapter 2, ?Radiation? p. 27-52.  Swan L.G., Ugursal V.I. (2009). Modeling of end-use energy consumption in the residential  sector: a review of modeling techniques. Renewable and Sustainable Energy Reviews 13(8), 1819?35.   Tang, Z., Brody, S. D., Quinn, C., Chang, L., Wei, T. (2010). Moving from agenda to action:  evaluating local climate change action plans. Journal of Environmental Planning and Management 53(1), 41?62.   Ulgen, K., & Hepbasli, A. (2004). Solar radiation models. Part 1: a review. Energy Sources,  26(5), 507-520.    180 ?rge-Vorsatz, D., Danny Harvey, L. D., Mirasgedis, S., & Levine, M. D. (2007). Mitigating  CO2 emissions from energy use in the world's buildings. Building Research & Information 35(4), 379-398.  VanderLaan, M., Tooke, R., Christen, A., Coops, N., Heyman, E., Olchovski, I. (2012).  Statistics on the built infrastructure at the Vancouver EPiCC experimental sites.  EPiCC Technical Report No. 4. Available online at http://www.geog.ubc.ca/~epicc/reports/Vancouver-EPiCC-Tech-Report-4.pdf  Voogt, JA., Oke, TR., Belair, S., Benjamin, M., Christen, A., Coops, NC., Grimmond, CSB.,  Lemonsu, A., Mailhot, J., Masson, V., McKendry, I., Strachan, I., Wang, J. (2007).  The Environmental Prediction in Canadian Cities (EPiCC) Network.  Seventh Symposium on the Urban Environment, American Meteorological Society, San Diego USA, Sept 9 - 13, 2007.  Wahlgren, R.V. (2010).  Heating/Cooling Degree-day Seasonality in British Columbia.  BC  Hydro Customer Information Management ? Load Analysis, 85 pages.  Waide, P., Amann, J.T., Hinge, A. (2007).  Energy efficiency in the North American existing  building stock. IEA Information Paper, International Energy Agency: Paris, France.  Walton, J.T. (2008). Subpixel urban land cover estimation: comparing cubist, random forests,  and support vector regression.  Photogrammetric Engineering and Remote Sensing 74, 1213-1222.  Wang, R. (2013). 3D building modeling using images and LiDAR: a review. International  Journal of Image and Data Fusion 4(4), 273-292.  Webster, C. J. (1993). GIS and the scientific inputs to urban planning. Part 1: description.  Environment and Planning B: Planning and Design 20(6), 709?728.   Webster, C. J. (1994). GIS and the scientific inputs to planning. Part 2: prediction and  prescription. Environment and Planning B: Planning and Design 21(2), 145?157.   White, P. (2010).  An awakening in energy efficiency: financing private sector energy  retrofits. Johnson Control Inc. Available at: http://www.johnsoncontrols.ca/content/dam/WWW/jci/be/solutions_for_your/private_sector/Financing_PrivateSector_whitepaper_FINAL.pdf  Williams, J. (2010). The deployment of decentralised energy systems as part of the housing  growth programme in the UK. Energy Policy 38, 7604-7613.  Wolsink, M. (2010). Contested environmental policy infrastructure: Socio-political  acceptance of renewable energy, water, and waste facilities. Environmental Impact Assessment Review 30(5), 302-311.    181 Wood, S, Thompson, K. (2012) Transnational voluntary climate change initiatives for local  governments: key variables, drivers, and likely effects. In: Richardson B, editors. Local Climate Change Law: Environmental Regulation in Cities and Other Localities, Cheltenham, UK: Edward Elgar Publishing, p. 29-66.  World Health Organization (2002). Legionella. In: Guidelines for Drinking Water.  Addendum: Microbiological Agents in Drinking Water, 2nd edn.  Wulder, M. A., Bater, C. W., Coops, N. C., Hilker, T., White, J. C. (2008). The role of  LiDAR in sustainable forest management. The Forestry Chronicle 84(6), 807-826.  Yan, W. Y., Shaker, A., Habib, A., & Kersting, A. P. (2012). Improving classification  accuracy of airborne LiDAR intensity data by geometric calibration and radiometric correction. ISPRS Journal of Photogrammetry and Remote Sensing 67, 35?44.  Yang, P., Putra, S., Li, W. (2007).  Viewsphere: a GIS-based 3D visibility analysis for urban  design evaluation.  Environment and Planning B 34, 971?992.  Yergin, D. (2006). Ensuring energy security. Foreign Affairs, 69-82.  Yu, B., Liu, H., Wu, J., Lin, W.M. (2009). Investigating impacts of urban morphology on  spatio-temporal variations of solar radiation with airborne LIDAR data and a solar flux model: a case study of downtown Houston. International Journal of Remote Sensing 30(17), 4359-4385.  Zarzalejo, L. F., Polo, J., Mart?n, L., Ram?rez, L., Espinar, B. (2009). A new statistical  approach for deriving global solar radiation from satellite images. Solar Energy 83(4), 480-484.  Zhou, Q., Liu, K. (2004). Analysis of errors of derived slope and aspect related to DEM data  properties. Computers and Geosciences 30, 369-378.  Zlatanova, S., Rahman, A.A., Shi, W. (2004). Topological models and frameworks for 3D  spatial objects. Computers and Geosciences 30(4), 419-428.   


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



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"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items