energy consumption trends of multi-unit … = office of energy efficiency ohc ... consumption trends...
TRANSCRIPT
ENERGY CONSUMPTION TRENDS OF MULTI-UNIT RESIDENTIAL BUILDINGS IN THE CITY OF TORONTO
by
Clarissa Binkley
A thesis submitted in conformity with the requirements for the degree of Master of Applied Science
Department of Civil Engineering University of Toronto
© Copyright by Clarissa Binkley 2012
ii
Energy Consumption Trends of Multi-Unit Residential Buildings in
the City of Toronto
Clarissa Binkley
Master of Applied Science
Department of Civil Engineering
University of Toronto
2012
Abstract
The purpose of this research is to determine the average energy intensity of multi-unit residential
buildings (MURBs) in Toronto, and evaluate whether certain building characteristics influence
energy intensity. This information is particularly important in the Toronto market. Relative to
the city’s population, Toronto has an unusually high proportion of MURBs with more than half
of residential dwellings in apartment buildings. Additionally, Toronto MURBs are significant
consumers of energy and produce an estimated 1.3M tonnes of CO2e each year. The ultimate
goal is to assess the most efficient building retrofit measures.
Energy consumption data for Toronto MURBs were collected and weather normalized.
Correlations between the energy data and the building characteristics were examined. Window
characteristics and heating system type were found to have the most significant influence on
energy intensity. Establishing energy consumption characteristics of MURBs is the first step
towards improving the energy efficiency of Toronto’s MURBs stock.
iii
Acknowledgments
There are many people who have provided invaluable support toward the completion of this
thesis. First and foremost, I would like to thank my supervisor, Professor Kim Pressnail, for his
assistance and patience throughout the process. I would also like to thank Marianne Touchie for
her major contribution of time, effort and ever-thoughtful analysis. Both Professor Pressnail and
Marianne were involved throughout the proposal, research, analysis and writing stages of the two
papers that went toward forming this thesis. I am grateful to them both for their hard work and
brilliant ideas.
This thesis would not have been possible without the generous support of the Toronto
Atmospheric Fund. I would like to thank Bryan Purcell for his thoughtful feedback, consistent
attention to the project and ongoing assistance with data collection.
I would also like to thank Woytek Kujawski at the Canadian Mortgage and Housing Corporation
and David Ethier at Toronto Hydro for their assistance in collecting and providing data for this
project. Thank you also to Warren Knowles and Graham Finch at RDH for their support in the
early stages of this thesis.
iv
Table of Contents
Acknowledgments.......................................................................................................................... iii
Table of Contents ........................................................................................................................... iv
List of Tables ................................................................................................................................ vii
List of Figures .............................................................................................................................. viii
List of Appendices .......................................................................................................................... x
List of Abbreviations and Acronyms ............................................................................................. xi
Chapter 1 Introduction .................................................................................................................... 1
1.1 Background and motivation ............................................................................................. 1
1.2 Research significance ....................................................................................................... 3
1.3 Scope of work................................................................................................................... 3
1.4 Thesis structure ................................................................................................................ 4
1.5 Terminology ..................................................................................................................... 4
Chapter 2 Literature Review ........................................................................................................... 6
2.1 Consumer-side studies...................................................................................................... 6
2.2 Supplier-side studies ...................................................................................................... 12
2.3 Comparison of consumer-side and supplier-side data.................................................... 16
Chapter 3 Data Collection and Data Sources ................................................................................ 17
3.1 Meta-Analysis data sources ............................................................................................ 17
3.2 Refined Data Set data collection .................................................................................... 19
Chapter 4 Data Characteristics and Limitations ........................................................................... 21
4.1 Summary of data ............................................................................................................ 21
4.2 Frequency distributions .................................................................................................. 24
4.3 Data Limitations ............................................................................................................. 27
v
Chapter 5 Methodology ................................................................................................................ 30
5.1 Meta-Analysis methodology .......................................................................................... 30
5.1.1 Normalization analysis decisions and assumptions ................................................ 31
5.1.2 Weather data ........................................................................................................... 33
5.1.3 Coefficient of determination (R2 values) ................................................................ 34
5.1.4 Determination of base load ..................................................................................... 34
5.1.5 Shared gas meters ................................................................................................... 36
5.1.6 Thermal energy consumption ................................................................................. 36
5.2 Refined data set methodology ........................................................................................ 37
5.2.1 Normalization procedure and assumptions ............................................................. 37
5.2.2 Determination of energy load types ........................................................................ 39
5.2.3 Rejection of outliers ................................................................................................ 39
Chapter 6 Meta-Analysis Results and Discussion ........................................................................ 41
6.1 Meta-Analysis energy consumption data ....................................................................... 41
6.1.1 Summary of energy intensity data .......................................................................... 41
6.1.2 Summary of greenhouse gas emissions .................................................................. 44
6.1.3 Base load determination .......................................................................................... 45
6.1.4 Summary of weather normalization ........................................................................ 47
6.2 Correlation results and discussion .................................................................................. 48
6.2.1 Date of construction ................................................................................................ 48
6.2.2 Weather-dependant loads ........................................................................................ 49
6.2.3 Building height........................................................................................................ 50
6.2.4 Number and size of suites ....................................................................................... 51
6.2.5 Ownership type ....................................................................................................... 53
6.2.6 Statistical Significance ............................................................................................ 55
6.3 Preliminary conclusions ................................................................................................. 56
Chapter 7 Refined Data Set Results .............................................................................................. 58
7.1 Energy intensity.............................................................................................................. 58
7.2 Selection of variables ..................................................................................................... 60
vi
7.2.1 Variables related to heating loads ........................................................................... 61
7.2.2 Variables related to cooling loads ........................................................................... 62
7.2.3 Variables related to base loads ................................................................................ 62
7.3 Separation of building types ........................................................................................... 63
7.4 Correlation analysis ........................................................................................................ 64
7.4.1 Window characteristics ........................................................................................... 65
7.4.2 Heating efficiency and cooling equipment ............................................................. 69
7.4.3 Number of occupants .............................................................................................. 71
7.4.4 Base electricity intensity ......................................................................................... 72
7.5 Multi-variable regression analysis ................................................................................. 73
7.5.1 Methodology ........................................................................................................... 73
7.5.2 Summary of all analyses ......................................................................................... 75
7.5.3 Discussion of selected results ................................................................................. 76
Chapter 8 Discussion .................................................................................................................... 79
8.1 Energy intensity.............................................................................................................. 79
8.2 Variable natural gas intensity ......................................................................................... 80
8.3 Base natural gas intensity ............................................................................................... 82
8.4 Alternative heating systems ........................................................................................... 83
8.5 Additional facilities ........................................................................................................ 84
Chapter 9 Conclusions and Further Research ............................................................................... 86
9.1 Variables affecting energy consumption ........................................................................ 86
9.2 Future work .................................................................................................................... 88
References ..................................................................................................................................... 91
vii
List of Tables
Table 1: Comparison of energy intensities from consumer-side studies and supplier-side studies 2
Table 2: Summary of characteristics influencing energy intensity ................................................. 9
Table 3: Summary of studies ........................................................................................................ 10
Table 4: Summary of energy consumption statistics from supplier-side data sources ................. 14
Table 5: Summary of Meta-Analysis data sources ....................................................................... 19
Table 6: Summary of Meta-Analysis and Refined Data Set building characteristics ................... 21
Table 7: Limitations of Meta-Analysis energy consumption data ................................................ 28
Table 8: Summary of energy consumption data normalization .................................................... 30
Table 9: Summary of building space conditioning categories ...................................................... 64
Table 10: Summary of forward-selection and logical regression analyses undertaken ................ 76
Table 11: Variables in order of selection for variable natural gas intensity systematic forward-
selection ........................................................................................................................................ 77
Table 12: Variables in order of selection for variable natural gas intensity logical method ........ 78
Table 13: Conclusions and data collection needs related to the variables examined ................... 86
viii
List of Figures
Figure 1: Flow of information for residential energy use data provided by government agencies
....................................................................................................................................................... 13
Figure 2: Distribution of building height in the sample compared with the population ............... 22
Figure 3: Distribution of building construction date in the sample compared with the population
....................................................................................................................................................... 23
Figure 4: Number of suites per building ....................................................................................... 25
Figure 5: Gross floor area per building ......................................................................................... 26
Figure 6: Frequency distribution of the average attributed suite floor area .................................. 27
Figure 7: Sample graph used for weather normalization .............................................................. 33
Figure 8: Natural gas consumption of sample MURB showing variable load and base load ...... 35
Figure 9: Total annual energy intensity ........................................................................................ 42
Figure 10: Annual natural gas intensity of each building ............................................................. 43
Figure 11: Total annual energy use per suite ................................................................................ 44
Figure 12: Total annual GHG emissions intensity ........................................................................ 45
Figure 13: Comparison of base load calculation methods for electricity ..................................... 46
Figure 14: Comparison of base load calculation methods for natural gas .................................... 47
Figure 15: Coefficient of determination values from weather normalization............................... 48
Figure 16: Energy intensity versus building vintage .................................................................... 49
Figure 17: Weather-dependant loads versus building vintage ...................................................... 50
Figure 17: Total energy use versus number of storeys ................................................................. 51
Figure 18: Energy intensity versus building height ...................................................................... 51
Figure 19: Total energy use versus number of suites ................................................................... 52
Figure 20: Energy intensity versus number of suites .................................................................... 52
Figure 21: Attributed suite size versus date of construction ......................................................... 53
Figure 22: Total energy intensity by area for each type of building ownership ........................... 54
Figure 23: Total energy intensity by suite for each type of building ownership .......................... 55
Figure 24: Suite size versus building ownership type .................................................................. 55
Figure 26: Natural gas and electricity components of total annual energy per gross floor area ... 58
Figure 27: Natural gas and electricity components of total annual energy use per suite .............. 59
ix
Figure 28: Variables categorized by ease of measurement and predicted importance ................. 61
Figure 29: Variable dominating building heating requirements ................................................... 62
Figure 30: Method used to separate buildings by primary heating systems ................................. 64
Figure 31: Variable natural gas intensity versus fenestration ratio for double-glazed windows .. 66
Figure 32: Variable natural gas intensity versus fenestration ratio for single-glazed windows ... 66
Figure 33: Electricity intensity versus fenestration ratio for air conditioned buildings ............... 67
Figure 34: Variable natural gas intensity versus glazing U-value ................................................ 68
Figure 35: Electrical intensity versus glazing U-value for air conditioned buildings .................. 69
Figure 36: Variable natural gas intensity versus boiler efficiency ............................................... 69
Figure 37: Variable natural gas intensity versus boiler age .......................................................... 70
Figure 38: Electrical intensity versus cooling capacity ................................................................ 71
Figure 39: Base annual natural gas consumption versus estimated number of occupants ........... 72
Figure 40: Base electricity intensity versus building vintage ....................................................... 73
x
List of Appendices
Appendix A Calculation of Statistics for Toronto Dwellings ....................................................... 97
Appendix B Sample Letter of Request ....................................................................................... 101
Appendix C Individual Building Characteristics ........................................................................ 104
Appendix D Characteristics of Toronto MURB Population ....................................................... 109
Appendix E Heat Equations Related to Energy Flows ............................................................... 111
Appendix F Refined Data Set Building Pictures ........................................................................ 113
Appendix G Additional Refined Data Set Correlations .............................................................. 120
Appendix H Multi-Variable Linear Regression Results ............................................................. 126
xi
List of Abbreviations and Acronyms
BC = British Columbia
CDD = Cooling degree days
CEUD = Comprehensive Energy Use Database
CWEC = Canadian Weather for Energy Calculations
CMHC = Canadian Mortgage and Housing Corporation
CO2e = Carbon dioxide equivalent
DHW = Domestic Hot Water
EC = Environment Canada
EI = Energy intensity
EUDH = Energy Use Data Handbook
GHG = Greenhouse gas
GTA = Greater Toronto Area
HDD = Heating degree days
HiSTAR = High-Rise Statistically Representative database
MAU = Make-up air unit
MURBs = Multi-unit residential buildings
NRC = National Research Council of Canada
NRCan = Natural Resources Canada
OEE = Office of Energy Efficiency
OHC = Ontario Housing Corporation
RESD = Report on Energy Supply and Demand in Canada
SHEU = Survey of Household Energy Use
SHGC = Solar heat gain coefficient
TAF = Toronto Atmospheric Fund
1
Chapter 1 Introduction
This thesis is a compilation of two research papers originally commissioned by the Toronto
Atmospheric Fund (TAF). The two papers are: Meta-Analysis of Energy Consumption in Multi-
Unit Residential Buildings in the Greater Toronto Area1 and Energy Consumption Trends of
Multi-Unit Residential Buildings in the City of Toronto.2 The papers were co-authored by
Clarissa Binkley, Marianne Touchie and Professor Kim Pressnail. For the purpose of this thesis,
the two papers have been combined and supplemented with additional research findings to form
a comprehensive document.
1.1 Background and motivation
Multi-unit residential buildings (MURBs) represent the most significant component of the
Toronto residential building inventory. Over half (55%) of the dwellings in the City of Toronto
consist of apartment buildings. The majority of all Toronto dwellings (39%) are either mid-rise
or high-rise apartment buildings of five or more storeys. Low-rise apartments of four or fewer
storeys represent 16% of the dwellings in Toronto.3 Calculations for these statistics are provided
in Appendix A, Section 1.
Consistent with MURBs comprising the most common form of dwelling in Toronto, they are
also a significant source of greenhouse gas (GHG) emissions. On an annual basis, combined
electricity and natural gas consumption of Toronto MURBs result in an estimated 1.3M tonnes of
CO2e.4 Mid and high-rise MURBs are responsible for 68% of these emissions and low-rise
MURBs are responsible for 32%5 (see Appendix A, Section 2). These figures are slightly lower
than another published estimate that Toronto MURBs erected between 1945 and 1984 are
responsible for between 2.0M and 2.2M tonnes of CO2e.6
There is a wide range of literature values for MURB energy-efficiency as expressed by average
energy intensity (EI). Energy intensity is a building’s annual energy consumption per unit of
gross floor area. Table 1 summarizes average MURB EI values from select studies. The average
energy intensity reported by each study is based on various sample groups of MURBs in Canada.
Each sample includes a different number of MURBs that represent different vintages, heights,
2
occupancy types, tenure and geographic locations. Additionally, the methodology to obtain
energy consumption data for each study was different and the collected data have a varying level
of reliability and completeness.
Energy intensity data derived primarily from aggregate supplier-side information appears to
provide significantly lower values than energy intensity data derived from disaggregate
consumer-side information. The Comprehensive Energy Use Database (CEUD) reports energy
intensity statistics for apartment buildings based primarily on supplier-side information provided
by the Report on Energy Supply and Demand in Canada and processed using the Residential
End-Use Model.7 Table 1 compares the CEUD EIs with three different studies that are based on
consumer-side information. Supplier-side and consumer-side energy studies are discussed in
more detail in the literature review in Chapter 2 of this thesis.
Table 1: Comparison of energy intensities from consumer-side studies and supplier-side studies
Study
reference
# of MURBs
in study
Data
year
Province EI
(ekWh/m2)
CUED EI for same year
and province (ekWh/m2)
CMHC8 31 2002 ON 400 211
9
HiSTAR –
Ontario10
55 1999 ON 275 20611
RDH12
39 2005 BC 213 16913
The total number of MURBs in Toronto is a matter of some speculation. It has been estimated
that as of 2006, there were more than 2,100 mid and high-rise MURBs in Toronto and more than
4,500 low-rise MURBs14
(see Appendix A, Section 3). Toronto has the 10th
largest number of
high-rise buildings of any city in the world15
and MURBs make up the majority of these high-
rise buildings. Interestingly, among the list of cities with a large number of high-rise buildings,
Toronto only ranks 44th
in terms of population.16
This means that Toronto has an unusually large
proportion of high-rise buildings for its population. A list of the 20 cities with the most high-rise
buildings in the world is provided in Appendix A, Section 4. This data is based on a definition
of high-rise buildings having 12 or more storeys, which is different from the eight or more
storeys definition used throughout the rest of this thesis.
3
1.2 Research significance
Given the large number of MURBs in Toronto and the uncertainty with the data on energy
intensity, determining an accurate estimate of energy intensity and developing an understanding
of how to reduce energy consumption is an important step in decreasing GHG emissions
associated with this sector. The Toronto City Council unanimously adopted greenhouse gas
emission targets in 2007 and the sustainable energy strategy released in 2009 indicates that the
residential sector is responsible for more than 40% of Toronto’s GHG emissions due to
electricity and natural gas consumption.17
As the largest component of the residential housing
stock, energy conservation in MURBs will play an important role in meeting GHG emission
targets in Toronto.
Additionally, more than half (51%) of all mid and high-rise MURBs in Toronto were built
between 1960 and 197918
(see Appendix A, Section 3). These buildings are aging and are
notoriously energy-inefficient.19
On a relative basis, apartments across Canada built from 1946
to 1969 have an average energy intensity that is 14% higher than the average of Canadian
apartments in all vintage categories. When compared with the most energy-efficient vintage
category (1980-1989), apartments built from 1946 to 1969 have a 45% higher EI than apartments
built from 1980 to 1989.20
Gathering and analysing energy consumption information that is representative of Toronto
MURBs is the first step toward the goal of reducing energy use. The energy consumption
information can be used to generate benchmarks that characterize current energy use profiles.
Buildings can also be categorized based on the characteristics that have the most significant
effect on energy consumption. Recognizing that there are limited economic and human
resources, the categorization can eventually be used to set priorities for energy retrofits in
MURBs.
1.3 Scope of work
The focus of this thesis is to examine trends between MURB energy consumption and building
characteristics for Toronto buildings. An analysis was first carried out on a group of 108
buildings that came from three different data sets. This stage will be referred to as the “Meta-
4
Analysis”. The Meta-Analysis data set was then distilled and supplemented with new data to
include 40 buildings with more complete energy consumption information as well as more
detailed building characteristics information. This will be referred to as the “Refined Data Set”.
Examination of both the Meta-Analysis data set and the Refined Data Set included the following
steps: data collection, data normalization, and a correlation analysis. A multi-variable regression
analysis and an examination of outliers were also completed for the Refined Data Set.
The findings from this research are intended to provide a basis on which to understand Toronto
MURB energy use with the ultimate aim to support policy and programmatic interventions to
achieve significant energy-use and greenhouse gas emissions reductions in existing buildings.
1.4 Thesis structure
This thesis is structured into three parts. The first part consists of Chapters 2, 3, 4, and 5 and
provides background information including a literature review, data collection procedures, the
data characteristics and the methodology. Information in each of these chapters pertains to both
the Meta-Analysis and the Refined Data Set. The second part, which is Chapter 6, is focused on
the Meta-Analysis and provides the results of the examined correlations and the preliminary
conclusions from the Meta-Analysis. The results of the Meta-Analysis represent preliminary
findings that lead to the more detailed analysis used in the Refined Data Set from which the
discussion and conclusions of this thesis are drawn. Therefore the third part, which includes
Chapters 7, 8, and 9, provides the results, discussion and conclusion of the Refined Data Set.
1.5 Terminology
A number of similar terms with subtle distinctions are used throughout this thesis. These terms
have been described below for clarity and consistency. Other terms are specific to certain
sections and will be addressed as they appear.
“Apartment” and “MURB” are used interchangeably throughout this thesis. Apartment is
generally used in reference to statistics from other sources in order to stay consistent with the
terminology of the source. However, since MURB is a broader term, it is used in all other cases.
5
The terms “dwelling”, “suite” and “household” are used in similar contexts, but are all defined
slightly differently. A “dwelling” is a self-contained living unit that could be a single detached
home or part of a single attached home, apartment building, or MURB. A “suite” is a more
specific type of dwelling found only in apartments or MURBs. A “household” is a person or a
group of people occupying one dwelling. Supplier-side energy statistics are often provided on a
household basis whereas consumer-side energy studies tend to quote energy use per household,
per dwelling, or per suite.
The “City of Toronto” has been shortened to “Toronto” throughout this thesis. The “Greater
Toronto Area” (GTA) is also referred to. The GTA consists of the larger Toronto metropolitan
region which includes the central City of Toronto as well as the municipalities of Durham,
Halton, Peel and York.
“Total energy consumption” refers to the total natural gas and electricity consumption expressed
in equivalent kilowatt hours (ekWh). Total energy consumption does not include other fuel
sources such as coal, propane, heating oil or wood. Energy consumption statistics from other
sources that are provided in gigajoules (GJ) have been converted into kWh when cited in this
thesis.
6
Chapter 2 Literature Review
A large range of values for the average energy intensity of MURBs in Canada and Ontario have
been reported in different studies. These studies will be reviewed and evaluated in order to
understand why such a large range of values exist and to determine an appropriate point of
reference with which to compare the energy intensity data presented in this thesis.
Most MURB energy studies can be classified as either consumer-side studies or supplier-side
studies. In a supplier-side study, aggregate energy consumption data is collected from energy
providers such as natural gas utility companies or electricity suppliers. Analysis techniques are
then applied to process the aggregate data and split it into more useful categories. In a
consumer-side study, energy consumption data is collected from individual households or
MURBs. The challenge of this process is to standardize the data collected and combine it into
more meaningful data categories.
2.1 Consumer-side studies
There are a number of studies that have analyzed energy consumption in MURBs by examining
electricity and natural gas meter readings for individual buildings. Because accessing this data
can be time consuming and permission is often difficult to obtain, these studies have reasonably
small sample sizes of less than 100 buildings each. Summaries of five consumer-side studies are
provided below including three studies of MURBs in Ontario, one study of MURBs across
Canada and one study of MURBs in British Columbia.
Canadian Mortgage and Housing Corporation (CMHC), 200521
Natural gas and electricity meter readings taken at 15-minute intervals for 34 MURBs over the
course of two years were used for this study. The buildings were concentrated in the GTA and
ranged in size from 69 suites to 473 suites.22
Three of the buildings were electrically heated and
had an average energy intensity of 344ekWh/m2. The remaining 31 buildings were heated with
central gas-fired boilers. The energy intensity for these buildings ranged from 281ekWh/m2 to
581ekWh/m2 and averaged 400ekWh/m
2. The average end-use distribution for the gas heated
buildings was 27% electricity, 49% natural gas space heating, and 24% domestic hot water.23
7
The focus of the study was to determine whether variations in energy use from one building to
another were related to building characteristics including age, floor area, height and facilities
such as pools. No reliable relationships between energy use and building characteristics were
found in this study and it was concluded that other factors related to ‘insulation, air leakage rates,
combustion efficiencies, stand-by losses, controls and distribution inefficiencies… or poor
energy management’24
may yield stronger correlations.
Ontario Housing Corporation (OHC), 200025
Energy audits conducted on 88 MURBs formed the basis of this study. These buildings were
randomly selected from a portfolio of public housing MURBs managed by OHC. The buildings
ranged height from one to 23 storeys, contained between 10 and 489 units, and were constructed
between 1960 and 1989. Two years of energy consumption data were collected for each
building.26
The average energy intensity for the 88 MURBs was found to be 232ekWh/m2.27
This study also examined how occupant type, fuel type, degree days, year of construction,
number of stories, gross floor area and region influenced energy use in MURBs. It was found
that the energy intensity of MURBs housing primarily senior citizens was almost half that of
MURBs housing primarily families. Additionally, it was observed that energy intensity
increases as height and gross floor area increase; however, the strength of this relationship was
not revealed. Finally, it was stated that building age affects energy intensity, but no further
discussion on this trend was provided.
National Research Council of Canada (NRC), 198228
In 1982, the Standing Committee on Energy Conservation conducted a study to define
appropriate energy consumption targets and guide requirements for the National Building Code
This was accomplished by examining energy consumption information from five different data
sets with a total of 216 MURBs located across Ontario and Quebec. One data set consisted of a
group of 50 Toronto MURBs and was collected by the OHC over a four year period (from 1975
to 1978).29
The buildings ranged height from 7 to 42 storeys, contained between 81 and 461
units, and were constructed between 1961 and 1973.30
The energy intensity for these buildings
ranged from 200ekWh/m2 to 730ekWh/m
2 and averaged 407ekWh/m
2.31
8
The data set of 50 Toronto MURBs showed the trend of decreasing energy intensity with
increasing floor area;32
however, this trend was not reflected in the other four data sets that were
included in the study.33
The second data set provided by the OHC consisted of one year of
energy consumption data for a group of 47 MURBs located in Ontario, but outside Toronto. The
energy intensity for these buildings ranged from 134ekWh/m2 to 672ekWh/m
2 and averaged
287ekWh/m2.34
The third data set of Ontario MURBs was provided by Ontario Hydro for a
group of 52 buildings each with one year of data. The energy intensity for these buildings
ranged from 126ekWh/m2 to 964ekWh/m
2 and averaged 320ekWh/m
2.35
High-rise statistically representative (HiSTAR) database, 200736
An analysis of 81 MURBs from across Canada was conducted based on information from the
HiSTAR database. The buildings ranged in height from four to 35 storeys, contained between 27
and 510 suites, had a gross floor area of between 2,200m2 and 43,000m
2, and were constructed
from 1920 to 1994.37
The average energy intensity for all 81 building ranged from 136ekWh/m2
up to 495ekWh/m2 and averaged 267ekWh/m
2.38
The majority of the buildings (55) were
located in Ontario and had an average energy intensity of 275ekWh/m2.39
Based on the analysis of the data, it was determined that the energy intensity of privately owned
MURBs was 35% higher on average than the energy intensity of MURBS providing public
housing.40
Additionally, gas heated buildings had a 40% higher energy intensity than electrically
heated buildings on average.41
Finally, the analysis showed that as energy intensity increased,
gross floor area decreased; however, as energy consumption per suite increased, gross floor area
also increased.42
This relationship was explained by the fact that larger buildings have larger
suites.
RDH Building Engineering, 201243
A study of energy consumption was conducted on 39 mid and high-rise MURBs located in
British Columbia (BC). A minimum of two years of energy consumption information for each
building was used for the analysis and whole building energy models were created for 11 of the
buildings. The buildings ranged in height from five to 33 storeys, contained between 16 and 212
suites, had a gross floor area of between 2,150m2 and 20,000m
2, and were constructed from 1974
to 2002.44
The energy intensity of these buildings ranged from 144ekWh/m2 to 299ekWh/m
2 and
9
averaged 213ekWh/m2. The average end-use distribution for the buildings was 38% electricity,
37% space heating (30% electricity and 70% natural gas), and 25% domestic hot water.45
It was observed that buildings constructed more recently (1990s to present) had higher energy
intensity than older buildings (1970s and 1980s). Additionally, buildings that derive more than
55% of their heating from electricity tended to have lower energy intensity than buildings heated
primarily with natural gas.
The five studies briefly summarized above provide a wide range of values for MURB energy
intensity based on consumer-side data with individual MURB energy intensities ranging from
136ekWh/m2 to 964ekWh/m
2 and data set averages ranging from 213ekWh/m
2 to 407ekWh/m
2.
Some of the studies examined how certain building characteristics could contribute to the
variation in energy intensity from one building to another. As summarized in Table 2, the trends
identified for occupancy type, building age and size were not strong and often showed
contradictory results from one study to another. Trends related to the type of heating appeared to
be the most consistent.
Table 2: Summary of characteristics influencing energy intensity
Characteristic Study Observed trend
Occupancy
type
OHC For public housing, the energy intensity of MURBs housing seniors is
almost half the energy intensity of MURBs housing families.
HiSTAR For public and private housing, seniors consumed only 10% less than
families on a per suite basis.
HiSTAR Privately owned MURBs have 35% higher energy intensity than MURBs
providing public housing.
Age CMHC No relationship between age and energy intensity was observed.
OHC Age affects energy intensity.
HiSTAR Buildings constructed most recently (1981 to 1994) have a 33% lower
energy intensity than older MURBs (1961 – 1980); however, these
MURBs have similar energy consumption per suite.
RDH Buildings constructed most recently (1990 to 2002) have a higher energy
intensity than older MURBs (1970s and 1980s).
Size CMHC No relationship between gross floor area and energy intensity was
observed.
OHC MURBs with a larger gross floor area have higher energy intensity.
NRC MURBs with a larger gross floor area have lower energy intensity.
HiSTAR MURBs with a larger gross floor area have lower energy intensity.
10
Type of
heating
CMHC MURBs heated with natural gas have 16% higher energy intensity than
buildings heated with electricity.
HiSTAR MURBs heated with natural gas have 40% higher energy intensity than
buildings heated with electricity.
RDH MURBs heated with natural gas have higher energy intensity than
buildings heated with electricity.
The table above examines reasons for variations in energy intensity from one building to another;
however, there are also factors contributing to the variation in average energy intensity from one
study to another. The average energy intensity reported by each study is based on MURBs in
different geographic locations, energy data from different years, MURBs representing various
vintages and heights, and different sample sizes. Additionally, although the methodology used to
obtain energy consumption data for each study was not specified, it likely varied from study to
study and yielded data with different levels of reliability and completeness. Table 3 provides a
summary of the energy intensity results from each of the studies.
Table 3: Summary of studies
Study Sample
size
MURB
location
Energy
consumption data
year
Average energy
intensity
(ekWh/m2)
Range of energy
intensities
(ekWh/m2)
CMHC 31 GTA May 2001 – May
2003
400 281 – 581
OHC 88 Ontario Unknown. 2 years
prior to 2000
232 Unknown
NRC – Toronto
data set
50 Toronto 1975 – 1978 407 200 – 730
NRC – Ontario
data set
47 Ontario
(except
Toronto)
1977 287 134 – 672
NRC – Ontario
Hydro data set
52 Ontario 1977
(estimated)
320 126 – 964
HiSTAR – all 81 Canada 1998 or 1999 267 136 – 495
HiSTAR –
Ontario
55 Ontario 1998 or 1999 275 Unknown
RHD 39 BC Minimum 2 years
from 2000 - 2010
213 144 - 299
The study with the lowest average energy intensity was the RDH study. This is likely because
the MURBs from this study were located in BC where the weather is mild compared with
Ontario. Heating degree days (HDDs) are a common measure used to estimate the heating needs
for a particular climate. The average annual HDDs for Vancouver are 2,062,46
whereas the
11
average annual HDDs for Toronto are almost twice as large at 4,066.47
Therefore, less heating
energy is required for BC MURBs. Additionally, electricity is more commonly used for heating
in BC because it is less expensive and in good supply. Both the CHMC and HiSTAR studies
showed that electrically heated buildings use less energy than gas heated buildings, so it follows
that BC MURBs would have a lower energy intensity since a higher proportion are electrically
heated.
The study with the highest average energy intensity was the NRC data set for Toronto buildings.
This may be because the data is from about 25 years earlier than the other studies examined. The
average number of annual HDDs in Toronto from 1975 to 1978 was 3,675,48
whereas the
average number of annual HDDs in Toronto from 1999 to 2002 was more than 10% lower at
3,265.49
Therefore, more heating was required in the MURBs from 1975 to 1978, leading to
higher energy intensity. Another reason that older data may yield higher average energy
intensity is because of a change in attitude toward energy efficiency. Occupants and property
managers may have been less concerned with energy conservation because of the lower cost of
energy and general public awareness for energy conservation.
The other two NRC data sets from 1977 did not have similarly high average energy intensities as
the Toronto data set did. This may be explained by the range in energy intensities. The lowest
energy intensities are 134ekWh/m2 and 126ekWh/m
2 for the two data sets. The data sets could
contain many highly energy-efficient buildings; however it is more likely that the energy
consumption data was incomplete and the quoted energy intensities did not account for all of the
energy consumption within the buildings. This may have occurred if fewer than twelve months
of data were reported for the annual total or if the building was sub-metered and not all of meter
readings were available. This commonly occurs when suites are sub-metered for electricity. The
electricity consumption of the common areas is available, but the electricity consumption for the
suites is not.
The CMHC study had the second highest average energy intensity. This is likely because
electrically heated buildings and buildings with incomplete electricity data were removed from
the sample before the average was taken. None of the other studies mention removing these two
groups, which intrinsically have lower energy intensities and would bring the average down.
12
The average energy intensity for all the consumer-side studies on Ontario MURBs reviewed here
is 305ekWh/m2. This average was weighted based on the number of buildings in each data set.
No other weightings based on the quality of data in in each of the study’s data sets were
considered since information on the reliability and completeness of the data was not readily
available. The average energy intensity of 305ekWh/m2 will be used for comparison purposes
throughout this thesis.
2.2 Supplier-side studies
Statistics on energy consumption in Canada are collected by a number of government agencies
including Statistics Canada and Natural Resources Canada’s Office of Energy Efficiency.
Information about residential energy use can be obtained from reports and data published by
these agencies. However, before this data can be used as a point of reference, it is important to
investigate how the data was collected so that the accuracy, completeness and relevance of the
data can be considered. The discussion presented here will be limited to energy data for the
residential sector with a focus on MURBs, although many of the sources discussed are concerned
with a much broader scope of energy production and consumption.
The three main sources for residential energy use data are: The Survey of Household Energy Use
(SHEU), the Energy Use Data Handbook (EUDH) and the Comprehensive Energy Use Database
(CEUD). Figure 1 provides a simplified diagram of the flow of information showing where data
comes from for these sources. As shown in the diagram, information collected in the Report on
Energy Supply and Demand in Canada (RESD), the SHEU as well as other surveys and data
sources are combined as inputs into the Residential End-Use Model. The Residential End-Use
Model processes the data and disaggregates it into categories as reported in the EUDH and the
CEUD.
13
Figure 1: Flow of information for residential energy use data provided by government agencies
Report on Energy Supply and Demand in Canada (RESD)50
This report is produced annually and consists of a series of data tables that quantify all energy
production, import, export and consumption for different sources of energy in Canada. Statistics
on the total amount of energy used in the residential sector within each province and fuel type is
derived from this report for input into the Residential End-Use Model.
The main strength of the RESD data is that it represents a survey of the universe, not a survey of
a sample. This means that the data reflects the actual total energy supply and demand. The
limitation of the RESD occurs when the data is split into sectors such as residential, commercial,
or industrial. The residential sector is concerned primarily with natural gas and electricity
consumption. This data is based on information provided by natural gas utility companies and
electric supply and distribution companies. These companies are responsible for accurately
coding customers based on their end-use sector. According to the RESD, “many reporting
companies, especially in the… natural gas & electricity areas, have difficulty in determining the
final consumers of the product.”51
This means that energy consumption is sometimes allocated
to the incorrect sector. Although this mistake would occur more frequently in other sectors, it
could occur in the residential sector. For example, a dwelling on farmland could be confused
with the agricultural sector or a MURB associated with a commercial or educational
establishment could be misallocated.
14
Survey of Household Energy Use (2007)52
The SHEU is based on both consumer-side and supplier-side information, but has been included
in the supplier-side section of this chapter because it is closely related to the other supplier-side
studies. Unlike the RESD, the SHEU is based on a three stage survey of a sample. The first
phase consisted of a computer-assisted telephone interview with dwelling owners and renters to
collect information on household energy use characteristics. The second phase was a paper mail-
back questionnaire completed by the same dwelling owners and renters to collect information on
factors affecting energy use in households. In the third phase, 2007 energy consumption data
was collected directly from households or energy suppliers of the same sample of dwellings as in
phase one and two.53
A total of 9,773 dwellings were surveyed, which represents a sample of
about 0.07%.54
Inconsistencies in the responses between the three phases show that respondents often lack the
knowledge required to provide accurate responses.55
The small sample size and the inaccurate
survey responses are the main limitations of the SHEU data.
Residential End-Use Model
This is a stock accounting model that combines the RESD and SHEU data with sales data, usage
profiles and other stock characteristics to allocate energy use to different categories. These
categories include: four building types, eight vintages, six fuel types, ten provinces, and five end-
uses.56
The results of the Residential End-Use Model are reported in the EUDH and the CEUD.
Relevant energy use data from the sources outlined above are provided in Table 4. Although the
SHEU 2007 data supersedes the SHEU 2003 data, the SHEU 2003 data has still been provided
since its statistics are often quoted as a point of reference for energy intensity in literature.
Table 4: Summary of energy consumption statistics from supplier-side data sources
Source Sample Type Energy Use Statistic
SHEU 200757
Canada low-rise apartments 12,200kWh/household
Canada high-rise apartments 12,200kWh/household
Ontario low-rise apartments 14,500kWh/household
Ontario high-rise apartments 12,400kWh/household
15
SHEU 200358
Canada apartments 305kWh/m2
Canada apartments (built 1946-1960) 347kWh/m2 (“Use with caution”)
Canada apartments (built 1980-1989) 239kWh/m2 (“Use with caution”)
Canadian apartments (built 1990-2003) 283kWh/m2 (“Use with caution”)
Ontario apartments Statistic unavailable
British Columbia apartments 239kWh/m2
EUDH59
Canadian Apartments – 2009 197kWh/m2
17,200kWh/household
CEUD
Canadian Apartments - 200960 197kWh/m
2
17,200kWh/household
Ontario Apartments - 200961
203kWh/m2
16,100kWh/household
BC Apartments - 200962
153kWh/m2
13,900kWh/household
In the SHEU 2007, energy use per household was provided; however, energy intensity on a floor
area basis was not provided. This may be because much of the energy intensity data collected
for the SHEU 2003 was unavailable in certain categories and marked “use with caution” in other
categories. Therefore, the energy intensity data may have been considered too unreliable to be
published in the SHEU 2007.
The EUDH and the CEUD both obtain their data from the same sources as discussed previously.
This is confirmed by both sources providing the same value for energy intensity and energy use
per household for Canadian apartments in 2009.
The SHEU 2007 energy consumption per household for Canadian apartments is 40% lower than
the EUDH energy consumption per household. However, the opposite tendency is shown in the
energy intensity data. For Canadian apartments, the EUDH data is more than 50% lower than
the SHEU 2003 data. One explanation for this may be the way that common area energy use is
treated in each of the two studies. In the SHEU, energy data is obtained on a household basis.
Therefore the owner or renter being surveyed would probably not include a portion of the
common area energy use in their household consumption. This would result in a lower energy
use per household and higher energy intensity (assuming that the energy intensity of common
areas is lower than the energy intensity of the suites or households). The EUDH energy data is
based on more aggregate energy consumption information from the RESD that would probably
include the common area energy and assign a portion of it to each household. This would
increase the energy use per household and decrease the energy intensity.
16
2.3 Comparison of consumer-side and supplier-side data
Consumer-side and supplier-side studies are conduced differently and are based on different sets
of assumptions. The average energy intensity of a consumer-side data set is often compared with
data from the EUDH or the CEUD. However, absolute comparisons between these data sets
should not be relied upon. The supplier-side studies provide lower average energy intensities for
all MURB categories than the consumer-side studies do (see Table 1, page 2). Therefore,
comparing a consumer-side data set with supplier-side statistics will cause the consumer-side
data set to appear more energy inefficient than it is.
However, comparisons can be made on a relative basis between supplier-side and consumer-side
data sets. For example, the fact that MURBs located in British Columbia have a lower energy
intensity than MURBs located in Ontario was shown in the consumer-side studies and can be
reinforced by the supplier-side statistics. Additionally, the energy mix proportions of natural gas
and electricity found in consumer-side data can be compared with supplier-side data since this is
a relative comparison. Since the SHEU data is a combination of consumer-side and supplier-side
data, absolute comparisons should not be made with consumer-side or supplier-side data sets.
Finally, if comparisons are drawn between consumer-side data sets, allowances for different data
set characteristics must be considered and discussed.
17
Chapter 3 Data Collection and Data Sources
Many parties such as public housing organizations, property management companies, and
engineering consulting firms collect MURB energy consumption data. However, this data is not
generally shared between parties because a certain level of effort or investment of resources was
required to collect the data. Logically, the parties who collected the data want to realize the
benefit of the information themselves rather than sharing it with their competitors. Additionally,
MURB energy consumption data can be difficult to collect or share because of barriers due to
privacy. For these reasons, data sets used in published consumer-side energy consumption
studies, like the ones outlined in the literature review, are usually of a limited size. In most
cases, funding from a third party is required to provide the necessary motivation to disclose
MURB energy use data.
Similar obstacles related to confidentiality and opposition to sharing data were faced in
collecting information for the Meta-Analysis and the Refined Data Set. The goal of the Meta-
Analysis was to collect and examine information for as many MURBs in Toronto and the
surrounding area as possible. Following the completion of the Meta-Analysis, the Refined Data
Set was formed from a combination of newly collected data and data selected from the Meta-
Analysis. The data sources and the data collection process for the two phases are described
below.
3.1 Meta-Analysis data sources
The Meta-Analysis data set, which included 108 buildings, was formed from three sources: the
HiSTAR database developed by CMHC; the Green Condo Champions Project developed by
TAF; and the Tower Renewal Benchmarking Initiative, also developed by TAF. Combining data
from different sources provides the advantage of a larger sample for analysis; however,
variations and limitations of the data from different sources must be considered in the analysis
procedure.
18
The first data source for the Meta-Analysis was the HiSTAR database. The database was
developed by CMHC in partnership with Natural Resources Canada to gain a better
understanding of energy and water use in apartment buildings. Working with consulting firms
across Canada, they gathered data from 81 buildings and assembled the information in a
Microsoft Access database.63
CMHC generously permitted access to a modified version of this
database that protects the anonymity of the building owners. Although the database contains
information from buildings nationwide, only the 55 buildings located in Southern Ontario were
included in the Meta-Analysis since these buildings share a similar climate. The building
construction dates range from 1941 to 1994, and building heights range from four to 35 storeys.
The second data set, TAF’s Green Condo Champions (“Condo Champs”) data set, consists of
energy audits of 50 condominium buildings in the Toronto area collected by Mann Engineering.
Building directors from the participating condominiums attended training and engagement
programs focused on cutting energy costs and reducing GHG emissions and received an audit
report and recommendations for strategies to reduce energy use in their buildings. Of the 50
buildings, energy audits from 42 buildings were provided for use in the Meta-Analysis. The
building construction dates range from 1971 to 2007 and the building heights range from four to
46 storeys.
The third data set originated from the Tower Renewal Benchmarking Initiative. It was run by
the City Manager’s Office and was partially funded by TAF. In this project, monthly energy
consumption data were collected from 29 high-rise rental apartments by Enerlife Consulting.
The purpose of the project was to begin an energy benchmarking system that could lead to
improvements in energy performance. Of the 29 buildings, the annual energy consumption
information for use in this meta-analysis was provided for 11 of the buildings. The building
construction dates range from 1961 to 1978 and the building heights range from nine to 33
storeys.
Table 5 presents a summary of the Meta-Analysis data sources and the type of energy
consumption information that was provided by each source. The HiSTAR database provided
both electricity and natural gas information for each building consisting of an annual total as well
as between two and twelve months of data for the same year. The Condo Champs data set did
19
not provide any electricity consumption information; however, monthly natural gas consumption
over a period of four years was provided for each building. Finally, the Tower Renewal
Benchmarking Initiative data provided the total consumption over one year for both natural gas
and electricity.
Table 5: Summary of Meta-Analysis data sources
Source of data Number of
buildings
Building
locations
Data
year
Electricity
data?
Natural
Gas
data?
Comments
HiSTAR from
CMHC
55 34 - Toronto
11 - GTA
10 - Ontario
1998
OR
1999
YES YES 1 year of data:
annual and monthly
for most buildings,
annual and only a
few months for some
Condo Champs
from TAF, Mann
Engineering
42 42 - Toronto 2006
to
2010
NO YES 3 to 4 years of data:
monthly
Tower Renewal
Benchmarking
Initiative
11 11 - Toronto 2009 YES YES 1 year of data:
annual
3.2 Refined Data Set data collection
The Meta-Analysis was limited in part because of the extent and completeness of the energy
consumption data. The Refined Data Set was formed to address the data limitations of the Meta-
Analysis by examining a modified data set composed of buildings with more complete energy
consumption and building characteristics data. The Refined Data Set consists of 40 buildings
and is composed of both newly acquired data as well as select data from the Meta-Analysis. The
methods used to collect the new data and select from existing data are outlined below.
There were 20 new buildings added to the Refined Data Set. Two of the newly added MURBs
were the focus of a study by Tzekova et al.64
and three were the subject of a community energy
plan for the City of Toronto.65
Information on the 15 other newly added MURBs was obtained
from energy audit reports conducted by engineering consulting firms for projects being carried
out by the City of Toronto’s Tower Renewal Office.
Twenty buildings from the original Meta-Analysis data set were used in the Refined Data Set.
The Meta-Analysis data source that showed the greatest potential for inclusion was the Condo
20
Champs data set since it included four years of monthly natural gas consumption information as
well as energy audit reports for each of the buildings. However, electricity consumption
information was not contained within the original data set. To obtain this electricity data,
contacts at each of the 42 buildings were sent a letter asking for permission to acquire electricity
consumption information directly from Toronto Hydro. A sample of the letter seeking
permission to access the data as well as a Frequently Asked Questions (FAQs) sheet provided
later in the process can be found in Appendix B. Many of the contacts chose not to complete the
data release form and others did not respond, even after multiple follow-up phone calls and
emails. Permission to access electricity consumption information for nine of the 42 buildings
was eventually obtained. In four of these nine buildings, residents were metered individually and
only permission for access to common electricity consumption could be obtained. Therefore,
only five buildings in the Refined Data Set come from the Green Condo Champions Project.
To obtain more buildings for the Refined Data Set, some of the buildings from the HiSTAR
database originally included in Meta-Analysis were selected. Although the HiSTAR data
contained electricity and natural gas consumption information for 55 Ontario buildings, the
quality and completeness of the data were found to be variable. Additionally, not all of the
buildings were located in the City of Toronto. Only HiSTAR buildings that met the following
criteria were included in the Refined Data Set:
• the building had to be located in the City of Toronto;
• more than eight months of natural gas and electricity consumption data had to be
available and;
• when weather normalization was carried out, a coefficient of determination (R2) greater
than 0.8 had to be achieved for the energy source providing the primary heating (usually
natural gas).
Upon examining the HiSTAR data, 15 of the 55 buildings used in the Meta-Analysis were
considered adequate for inclusion in the Refined Data Set.
21
Chapter 4 Data Characteristics and Limitations
This chapter summarizes the characteristics of the buildings in the Meta-Analysis data set and
the Refined Data Set and compares them with characteristics of Toronto MURBs. These
comparisons are drawn to determine whether the two data sets (samples) are representative of the
actual population within various categories and whether certain categories are under-represented.
Other limitations of the data sets are also discussed in this Chapter.
4.1 Summary of data
The Meta-Analysis data set contains 108 buildings: 90 buildings are located in the City of
Toronto and 8 are located around Toronto in the cities of Thornhill, Mississauga, Scarborough
and Brampton. The remaining 10 buildings are located in other cities around Southern Ontario
including Whitby, Peterborough, Lindsay, Windsor and Ottawa.
The City of Toronto contains an estimated 2,100 mid and high-rise MURBs of five or more
storeys and 4,500 low-rise MURBs of four or fewer storeys (see Appendix A, Section 3). The
Meta-Analysis sample represents approximately 1.6% of the total number of MURBs in Toronto
and 4.8% of mid and high-rise MURBs.
The Refined Data Set of 40 MURBs includes only mid and high-rise MURBs. Therefore, the
Refined Data Set represents 1.9% of the mid and high-rise population and 0.7% of the total
MURB stock in Toronto. Table 6 contains a summary of the size and date of construction of the
buildings examined in the Meta-Analysis and the Refined Data Set. More detailed information
about the characteristics of individual buildings can be found in Appendix C.
Table 6: Summary of Meta-Analysis and Refined Data Set building characteristics
Meta-Analysis Refined Data Set
Min. Max. Median Min Max Median
Number of storeys 4 46 13 5 28 13
Number of suites 19 719 188 68 339 156
Date of construction 1941 2009 1984 1960 2003 1979
Gross floor area (m2) 2,000 102,000 18,400 3,340 35,900 13,600
Attributed suite area (m2) 42 279 104 45 250 96
22
In order to understand whether the Meta-Analysis data set and the Refined Data Set are
representative of the MURB population in Toronto, these data sets have been compared with
information taken from an online database called TObuilt. The database is focused on high-rise
buildings and there are entries for a total of 1,609 high-rise MURBs, and 131 mid-rise MURBs.66
The raw data taken from TObuilt is provided in Appendix D as well as an explanation as to how
the TObuilt data has been weighted to account for its limited data on mid-rise buildings.
In Figure 2 and Figure 3, “% of Meta-Analysis Sample” and “% of Refined Data Set Sample”
mean the number of buildings in the Meta-Analysis data set and the Refined Data Set that fall
within a given category divided by the total number of buildings in either the Meta-Analysis data
set or the Refined Data Set. Similarly, “% of population” means the number of buildings in
Toronto that fall within a given category divided by the total number of buildings in Toronto.
The data for the number of Toronto buildings in each height and age category has been
calculated by adjusting data derived from the TObuilt database. If the percentage of sample is a
similar number to the percentage of the population, the category is represented in the sample in
equal proportion to that of the actual population. If the % of sample is larger, then that category
is over-represented in the sample and if the % of sample is smaller, then that category is under-
represented.
Figure 2: Distribution of building height in the sample compared with the population
0%
5%
10%
15%
20%
25%
30%
35%
0%
5%
10%
15%
20%
25%
30%
35%
5 – 8 9 – 12 13 – 16 17 – 20 21 – 24 25 – 28 > 28
% o
f P
op
ula
tio
n
% o
f S
am
ple
Number of Storeys
Distribution of Building Height in the Sample
Compared with the Population% of Meta-Analysis Sample % of Refined Data Set Sample % of Population
23
Within the height categories shown in Figure 2, the Meta-Analysis data set is represented in all
the groupings and the Refined Data Set is represented in all but the tallest. The height ranges
that are over-represented or under-represented in both the Meta-Analysis and the Refined Data
Set are generally in neighbouring categories and would balance one another out if the ranges
considered were larger. For example, the nine to 12 storey category is over-represented, but is
balanced by the 13 to 16 storey category, which is under-represented. Similarly, the 17 to 20
storey category is under-represented, but is balanced out by the 21 to 24 storey category.
Therefore, the Meta-Analysis data set and represents an adequate distribution of building heights
for mid and high-rise MURBs and the Refined Data Set also represents an adequate distribution
of MURBs up to 28 storeys.
Figure 3: Distribution of building construction date in the sample compared with the population
Within the vintage categories shown in Figure 3, the Meta-Analysis data set is represented in all
the date ranges, and the Refined Data Set is represented in all but the oldest. Both the Meta-
Analysis data set and the Refined Data Set are most significantly over-represented in the 1991 to
2000 time period, while they are both under-represented in the 2001 to 2010 time period. The
two categories with the highest proportion of the MURB population, 1961 to 1970 and 1971 to
1980, are both reasonably well-represented by the two data sets. For the period from 1946 to
1990, the distribution of buildings in the Refined Data Set matches the population fairly well.
0%
5%
10%
15%
20%
25%
30%
35%
0%
5%
10%
15%
20%
25%
30%
35%
Before
1946
1946 –
1960
1961 –
1970
1971 –
1980
1981 –
1990
1991 –
2000
2001 –
2010
% o
f P
op
ula
tio
n
% o
f S
am
ple
Time Period
Distribution of Building Construction Date in the
Sample Compared with the Population
% of Meta-Analysis Sample % of Refined Data Set Sample % of Population
24
This is important since it shows that the sample reflects the population that is being characterized
by this study.
Overall, the Meta-Analysis data set represents buildings in all the vintage and height categories
examined. Ideally, the proportions of the buildings’ distributions among the categories would
match the population’s more closely. The Refined Data Set matches the population’s
distribution of buildings slightly better than the Meta-Analysis data set does; however, two of the
examined categories are not represented. These categories are buildings with more than 28
storeys and buildings constructed prior to 1946.
4.2 Frequency distributions
Frequency distributions have been provided for the building characteristics where data for
comparison to the population were not readily available. These characteristics include number of
suites, gross floor area, and average attributed suite size. Each frequency distribution shows the
percent of the sample represented in a particular category within the Meta-Analysis data set or
the Refined Data Set. The data label at the top of each bar denotes the actual number of
buildings within the category for each of the data sets.
The number of suites in each building, categorized into ranges by hundreds of suites, is shown in
Figure 4. Within the Meta-Analysis data set, the number of suites in each building ranged from
19 to 719 with a median of 188. The number of suites was unspecified for one of the buildings
and was estimated to fall into the one to 100 suites range based on the building’s gross floor area.
Within the Refined Data Set, the number of suites in each building ranged from 68 to 339 with a
median of 156. For both data sets, buildings that had 101 to 200 suites were the best represented.
The Refined Data Set did not represent any buildings with more than 400 suites, whereas the
Meta-Analysis data set did. Generally, both data sets represent a similar distribution of buildings
based on the number of suites per building.
25
Figure 4: Number of suites per building
Figure 5 shows the gross floor areas of the buildings split up in increments of ten thousand
square metres. The floor area of each building was specified in the original data sources. In
most cases, a definition of what was included in the floor area was not provided, however it is
assumed that the gross floor area represents the total area containing the residential suites, the
lobby, the common areas, and any conditioned recreational areas. It typically does not include
underground parking areas even though this space is conditioned in many buildings. The
implications of this assumption are discussed further in the data limitations section of this
chapter.
Within the Meta-Analysis data set, the gross floor areas ranged from 2,000m2 to 102,000m
2 with
a median of 18,400m2. Within the Refined Data Set, the gross floor areas ranged from 3,340m
2
to 35,900m2 with a median of 13,600m
2. For both data sets, buildings that fell into the range of
10,000m2 to 20,000m
2 were the best represented. The Refined Data Set did not represent any
buildings with a gross floor area greater than 40,000m2, whereas the Meta-Analysis data set did.
Generally, both data sets represent a similar distribution of buildings based on the gross floor
area.
24
40
25
12
3 4
1315
9
3
0%
10%
20%
30%
40%
50%
60%
70%
1-100 101-200 201-300 301-400 401-500 >500
Pe
rce
nt
of
Sa
mp
le
Number of Suites
Number of Suites per BuildingMeta-Analysis data set Refined Data Set
Number of
Buildings
26
Figure 5: Gross floor area per building
Figure 6 shows the average attributed suite area of the buildings split up in increments of 50m2.
The average attributed suite area provides an approximation of the average actual suite size by
dividing the gross floor area by the number of suites. Note that this average attributed suite area
will be larger than the actual suite size because it also includes a proportion of the common
areas.
Within the Meta-Analysis data set, the average attributed suite areas ranged from 42m2 to 279m
2
with a median of 104m2. Within the Refined Data Set, the gross floor areas ranged from 45m
2 to
250m2 with a median of 96m
2. For both data sets, buildings that fell into the range of 51m
2 to
100m2 average attributed suite area were the best represented. Both data sets represented
buildings in all the suite size categories and had a similar distribution of buildings based on
attributed suite sizes.
According to supplier-side studies, the average suite size of apartments across Canada is 84m2.67
The average attributed suite areas in Meta-Analysis data set and the Refined Data Set were
114m2 and 102m
2 respectively. Since the attributed suite size includes common areas it is
expected that the attributed suite size would be larger than the actual suite size. The average
attributed suite sizes from the Meta-Analysis data set and the Refined Data Set indicate that
common areas occupy between 17% and 26% of the gross floor area in the average building.
22
38
23
15
4 6
10
16
10
4
0%
10%
20%
30%
40%
50%
60%
70%
0-10 10-20 20-30 30-40 40-50 >50
Pe
rce
nt
of
Sa
mp
le
Gross Floor Area (1,000s of m2)
Gross Floor AreaMeta-Analysis data set Refined Data Set
Number of
Buildings
27
Figure 6: Frequency distribution of the average attributed suite floor area
4.3 Data Limitations
The Meta-Analysis data set and the Refined Data Set deal with two types of data: energy
consumption data and building characteristics data. Imperfections in the energy consumption
data are the primary limitation of the Meta-Analysis data set and imperfections in the building
characteristics information are the primary limitation of the Refined Data Set; however,
imperfections in both types of data do contribute to the limitations in both data sets. The
majority of the limitations of both the Meta-Analysis data set and the Refined Data Set stem
from both being composed of data from multiple sources. This data has been collected by
various parties for different purposes at varying levels of detail and completeness.
Table 7 contains a summary of the deficiencies in the energy consumption data that limited the
usefulness of the Meta-Analysis data set. Since only simple building characteristics were
examined in the Meta-Analysis data set, most of the limitations of the data were due to concerns
with the energy consumption data. Primary factors are concerns that were expected to have the
greatest effect on the data analysis, while secondary factors, though important, were expected to
have less of an effect on the outcome of the analysis. The outlined imperfections are inherent in
the data. Since they cannot be removed, the limitations were considered in the approach chosen
to process the data, which is described in the next chapter.
3
49
33
16
61
23
14
1 1
0%
10%
20%
30%
40%
50%
60%
70%
0-50 51-100 101-150 151-200 > 200
Pe
rce
nt
of
Sa
mp
le
Attributed Suite Area (m2)
Average Attributed Floor AreaMeta-Analysis data set Refined Data Set
Number of
Buildings
28
Table 7: Limitations of Meta-Analysis energy consumption data
One of the limitations of the building characteristics data that was common to both the Meta-
Analysis data set and the Refined Data Set was related to the gross floor area. The method used
to obtain the gross floor area reported by each data source is not known and the parts of the
building included in the gross floor area measurement were never explicitly stated. Often, the
gross floor area is estimated from floor plans of the building, from physical measurements of the
building, or from values obtained from real estate information. These methods may not net
highly accurate measurements of total floor area. The most accurate gross floor area information
is likely to exist for condominiums since this information is used to subdivide ownership of the
building.
Errors in the estimated gross floor area have a significant effect on the results of the data
analysis, especially the energy intensity values and the average attributed suite floor area. If the
underground parking area is conditioned but not included in the gross floor area, the energy
intensity based on the occupied space of the building will be overestimated. Conversely, if the
underground parking area is not conditioned but is included in the gross floor area, the energy
intensity value would be underestimated. The effect of an inaccurate estimate of the floor area
will be diminished with a larger data set as errors off-set one another.
Since buildings for the Refined Data Set were selected expressly for the completeness of their
energy consumption data, the limitations of this data set are primarily related to the building
characteristics data. Most of the building characteristics used in the correlation analysis come
Name of data
set
Primary factor Secondary factor
HiSTAR Actual monthly use may not match the
month to which it had been assigned.
This affected the weather normalization
(especially the R2 values).
The data year was uncertain. Electricity and
natural gas information was often incomplete.
The data overall were not reliable.
Green Condo
Champions
The data set only included natural gas
data. No electricity data were
available.
No mechanical efficiencies have been applied
so the amount of useful thermal energy was
not determined. Values for mechanical
efficiencies were generally available in the
audit reports.
Tower
Renewal
Only one year of annual data was
available. This was challenging to
properly weather normalize.
No mechanical efficiencies have been applied,
so the amount of useful thermal energy was
not determined No values for mechanical
efficiencies were available.
29
from information collected in building energy audit reports for each building. Building
information was collected by at least 12 different individuals from at least seven different
engineering consulting firms. The practices and the assumptions made by each firm vary, and
the judgment of each individual may contribute to inconsistencies and variability in the data.
Since the energy audits were not collected specifically for this study and since information was
not necessarily recorded with the aim of being directly comparable with other data sources, the
data had be scrutinized for inconsistencies. It appeared that data were often estimated; however,
no distinction is made between actually observed data and estimated data in the audit reports.
Therefore it is difficult to distinguish the reliable data from the less reliable data.
Photographs of the buildings obtained from the audit reports and internet searches were used to
improve the consistency of information such as the fenestration ratio, the presence of balconies
and through-wall slabs, the general type of wall construction and the number of floors.
Photographs were also used to confirm the presence of window unit air conditioners and roof
equipment such as make-up air units. Searches on the building address were used to obtain
supplementary information about the ownership type (seniors’ home, hospice, co-operative
housing organization) and the presence of amenities such as fitness facilities or a pool. Finally,
census information combined with the number of suites was used to estimate the number of
occupants. The effect of the limitations of the data will be discussed further as each variable is
examined in the correlations analysis in Chapter 7.
30
Chapter 5 Methodology
Both the Meta-Analysis data set and the Refined Data Set were examined by first normalizing
the energy consumption data and then determining correlations between the energy consumption
data and the building characteristics. The energy consumption data was normalized so that direct
comparisons between buildings could be made. The main building characteristics examined in
the Meta-Analysis were building age, building size and ownership type. The main building
characteristics examined in the Refined Data Set were fenestration-to-wall ratio, window
conductance, boiler efficiency, cooling capacity of air conditioners and number of occupants.
5.1 Meta-Analysis methodology
From building to building, the energy data varied according to location, billing periods and
billing years. Therefore normalization was required to account for these variations. The
normalization process applied depended on the data characteristics and was different for each of
the three data sources. The normalization methods applied to each data source have been
summarized in Table 8 and a description of the process and purpose of each of these
normalizations follows.
Table 8: Summary of energy consumption data normalization
Processes applied HiSTAR Green Condo
Champions
Tower
Renewal
Calendarization �
Month-length
normalization � �
Total consumption
normalization �
Linear regression
weather normalization � �
Annual weather
normalization �
Once the data were normalized, buildings were tabulated according to characteristics such as
age, ownership type (rental or condominium), gross floor area, number of suites and building
height. Normalized energy consumption data were then plotted against the building
characteristics to examine correlations and determine any trends.
31
5.1.1 Normalization analysis decisions and assumptions
This section describes the normalization procedures listed in Table 8 and identifies the data sets
to which the normalization was applied.
Calendarization
Typically, energy billing cycles do not correspond to the beginning and end of the calendar
month. These billing cycles also differ from building to building. In order to compare buildings
directly, the associated energy use for each calendar month must be determined.
Fortunately, the Green Condo Champions data contained meter reading dates. Meter reading
dates were used to apportion energy use to the appropriate calendar month based on the average
daily energy use during a billing period. The HiSTAR data were already allocated to particular
months and did not contain billing dates. As such, it was assumed that the given consumption
corresponded exactly with the month to which it was assigned. The Tower Renewal data were
not calendarized because only annual consumption data were available.
Month-length normalization
As the number of days in each month varies throughout the year, the consumption data and
heating degree days (HDDs) were normalized to a standard month with a length of 30.42 days,
which corresponds to a non-leap year. This was done so that each data point would have equal
weighting in the linear regression used for weather normalization. The total annual energy
consumption and total HDDs were kept the same, but divided proportionately among the months
of standard length. This process was completed for the HiSTAR and Green Condo Champions
data sets since both contained monthly data. It could not be applied to the annual Tower
Renewal energy consumption data.
Total consumption normalization
This process was only completed on the HiSTAR data. The process was necessary because
energy consumption was provided for the individual months along with the total annual energy
consumption. The sum of the monthly consumption data did not always equal the annual energy
consumption provided. In some cases, this discrepancy arose because energy consumption
information was missing for one or more of the months. In addition, it is possible that the billing
32
period did not match the actual month to which the energy was assigned. If, for example, some
of the energy billed in January should have been assigned to the previous year, this would be
reflected in the annual total, but not in the monthly data. A third possible source of discrepancies
is the meter-reading practices of the utility. The utility companies will sometimes estimate the
energy use based on historical consumption rather than reading the meter and then apply a
correction to the estimate later. The correction would be reflected in the annual total, but not in
the monthly data. To account these discrepancies, normalization of the total energy consumption
was carried out by determining the difference in the sum of the monthly consumption data and
the annual data. This difference was then distributed evenly among the months before the
weather normalization was completed.
Weather normalization
The energy consumption data have been collected from a range of years and also for buildings in
different locations. Weather conditions vary from year to year and city to city thereby
influencing energy use. Thus, energy consumption data had to be weather normalized to a
common year and location to allow comparison between buildings. Following completion of the
normalizations above, weather normalization was carried out on the HiSTAR and Green Condo
Champions data sets. Weather normalization involved the following steps:
1. The monthly HDDs (x-axis) were plotted against the monthly energy consumption (y-
axis) for all of the available months of data as shown in Figure 7.
2. Linear regression was then used to determine a line of best fit for the data.
3. The coefficient of determination or the R2 value of the linear regression was determined.
The significance of this term is described below.
4. The equation of the line of best fit was used to determine the energy consumption over a
standard weather year as determined from Canadian Weather for Energy Calculations
(CWEC) data. This was done by inputting the monthly HDDs calculated from the
CWEC to determine the resulting standard monthly energy consumption. The sum of the
monthly energy consumption became the weather normalized annual consumption value.
Figure 7 shows an example of the plots created to perform the weather normalization for
each building. This graph is for one of the buildings in the Green Condo Champions data set.
33
Figure 7: Sample graph used for weather normalization
Since no monthly data were available, this process could not be applied to weather normalize the
Tower Renewal data. Instead the data were normalized on an annual basis. The normalization
assumed the same relationship as used in the monthly weather normalization:
���������������� �
���������������∝
������������������� �
�����������������
5.1.2 Weather data
Historical weather data were retrieved and associated with the energy consumption data.
Monthly HDD data were obtained from Environment Canada for the applicable cities and years.
For buildings located in and around Toronto including those in Mississauga, Scarborough,
Brampton and Thornhill, data from the weather station in Toronto near the University of Toronto
were used. For locations further afield such as Lindsay, Whitby, Ottawa, Peterborough and
Windsor, city-specific weather data were applied.
The standard weather year to which all of the energy consumption data was normalized was
based on CWEC data. A standard reference year such as this captures the average weather over
a long period of time. The CWEC data used in this report were based on Toronto mean
temperatures from 1960-1989. The HDDs for both historical weather and the standard reference
y = 740x + 55,000
R² = 0.99
0
100,000
200,000
300,000
400,000
500,000
600,000
700,000
0 200 400 600 800
Na
tura
l G
as
Co
nsu
mp
tio
n
(ek
Wh
/mo
nth
)
Actual Monthly Heating Degree Days
Weather Normalization For Building C20
34
year were calculated with an 18°C base, established on the assumption that heating is not
required until the exterior temperature falls below this base.68
5.1.3 Coefficient of determination (R2 values)
The coefficient of determination indicates how well the line of best fit calculated from the linear
regression explains the variation in the data. For example, how closely the HDDs correlate with
the energy consumption data of a building in the weather normalization process. An R2 value
close to one means that most of the variation is explained; however, a number close to zero
means that most of the variation is unexplained.
R2 values above 0.8 are considered acceptable in the context of weather normalization. If the R
2
value is between 0.6 and 0.8, the data were lacking a good correlation but the weather
normalization is often still done. In practice, if the R2 value is below 0.6 the weather
normalization is typically not completed. In the Meta-Analysis data set, 50% of the R2 values
were greater than 0.6, but only 20% of the buildings had a good correlation for both electricity
and natural gas consumption with weather. The HiSTAR data yielded R2 values that ranged
from 0.0001 up to 0.99. The reasons for values below 0.6 will be discussed in Chapter 6.
Nevertheless, the weather normalization was still completed so that the HiSTAR data could be
compared with the other data sets.
5.1.4 Determination of base load
When energy consumption is examined on a monthly basis, it can be separated into two
components as illustrated in Figure 8: the base loads (weather independent) and the variable
loads (weather dependent). Although lighting use, domestic hot water use and plug-loads can
show minor seasonal fluctuations and contribute to heating the building, these are considered
negligible compared with the seasonal variations produced by the heating and cooling systems.
35
Figure 8: Natural gas consumption of sample MURB showing variable load and base load
For the HiSTAR and Green Condo Champions data sets, two methods were used to determine
the base load. In the first method, the average of the two months of lowest energy consumption
of each year was to estimate the base load. For example, if three years of data were available the
base load would be determined from the average of the six months of relevant data. If the
building was air conditioned, the lowest consumption was typically in the shoulder months
(April or May and September or October), while buildings without air conditioning had the
lowest consumption during the summer months.
The second method for determining the based load used the y-intercept value from the weather
normalization procedure outlined above. The y-intercept represents the energy consumed by the
building when there are no HDDs or, in other words, when no heating is required.
The majority of the Meta-Analysis correlations were conducted on the total energy consumed in
the MURBs rather than the base or variable components of electricity and natural gas.
Therefore, using these two methods to determine the base load was primarily done for
investigative purposes to determine the best method for the Refined Data Set. A comparison of
the two methods and their respective results are examined in Chapter 6.
When the base and natural gas loads were used in the Meta-Analysis a combination of the two
methods were used depending on the energy consumption data. If the weather normalization
0
20,000
40,000
60,000
80,000
100,000
120,000
140,000
Ma
r-…
Ma
y-…
Jul-
07
Sep
-07
No
v-0
7
Jan
-08
Ma
r-…
Ma
y-…
Jul-
08
Sep
-08
No
v-0
8
Jan
-09
Ma
r-…
Ma
y-…
Jul-
09
Sep
-09
No
v-0
9
Jan
-10
Ma
r-…
Ma
y-…
Jul-
10
Sep
-10
Mo
nth
ly N
atu
ral
Ga
s C
on
sum
pti
on
(m
3)
Natural Gas Consumption of Sample MURB
Base Load
Variable Load
36
yielded an R2 value greater than 0.6 and more than six months of energy consumption data were
available, then the y-intercept method was used. Otherwise, the low average method was used.
As the Tower Renewal data contained no monthly data, the base load determination methods
outlined above were not applied. Fortunately, this data set already included separate base and
variable loads. The method used to determine these loads is unknown but they are likely based
on the average energy consumption in the summer months.
5.1.5 Shared gas meters
Some of the buildings in the Green Condo Champions data set shared a gas meter between two
buildings. In each case, the “twin” buildings had very similar physical characteristics. Therefore
the natural gas consumption data were split proportionally by the gross area of each building.
Though a reasonable approximation, it is likely that building operation and occupant behaviour
would make these numbers different.
5.1.6 Thermal energy consumption
When natural gas is consumed and converted into useful thermal energy to heat the MURB and
the domestic hot water, some energy is lost based on the efficiency of the equipment (usually the
boilers). The focus of the Meta-Analysis is on energy consumption; however it is also important
to recognize that energy consumption is a function of both a building’s thermal energy
requirements and the efficiency of the equipment converting natural gas into useful energy.
Cubic metres of natural gas supplied were converted to equivalent kilowatt-hours without
application of an efficiency factor since the efficiencies of the natural gas boilers in each
building were not known for all of the data in the Meta-Analysis. The actual efficiency and
therefore the useful thermal energy made available to the building depend on the age, type and
inherent efficiency of the mechanical equipment. Efficiencies typically range between 60-90%,
but can sometimes be higher or lower. The thermal energy requirements of a building are a
function of many parameters including type of building envelope, air tightness, building
orientation, occupancy and electrical load.
The conversion from cubic metres of natural gas supplied to equivalent kilowatt-hours of energy
was based on a factor of 37.08MJ/m3 or 10.3kWh/m
3.69
37
5.2 Refined data set methodology
Many of the methods and assumptions applied in the Meta-Analysis were also employed for the
Refined Data Set. This section discusses the aspects of the Refined Data Set methodology that
differ from the Meta-Analysis and how extreme outliers have been considered and resolved. For
each of the 40 buildings, monthly natural gas and electricity data were weather normalized using
a standard weather year as determined from the CWEC. Following this normalization, the base
component (weather independent) and the variable component (weather dependent) of the
natural gas and electricity use were identified. Using the normalized energy data, functional
relationships between the variables relating to the mechanical and the electrical system, the
building envelope, and the occupancy characteristics of the building were sought. These
variables were tested against various measures of energy consumption to determine where
correlations existed. After the correlations were examined, a multi-variable regression analysis
was completed to determine whether correlations could be improved when the interaction
between the building characteristics were considered. The methodology for this portion of the
analysis is described in Chapter 7 in combination with the results of the multi-variable regression
analysis. Finally, buildings that appeared to be anomalies from the identified trends were
examined in greater detail.
5.2.1 Normalization procedure and assumptions
The normalization procedure used for the Refined Data Set differed slightly from the procedure
used in the Meta-Analysis because of the different characteristics of the energy consumption
data. Some aspects of the procedure were also improved based on experience from the Meta-
Analysis.
The same calendarization procedure as described for the Meta-Analysis was applied for the
Refined Data Set. The month-length normalization was removed from the normalization
procedure. A sensitivity analysis revealed that month-length normalization has an undetectable
effect on the final weather normalization and was deemed unnecessary. No annual weather
normalization was required for any of the buildings since all of the energy consumption data
used in the Refined Data Set was provided on a monthly basis. The modifications applied to
38
total consumption normalization and the weather normalization for the Refined Data Set are
described below.
Total consumption normalization
As in the Meta-Analysis, the total consumption normalization was applied to any buildings
where the sum of the monthly data was not equal to the given annual sum. In the Meta-Analysis,
this difference was distributed evenly among the months. For the Refined Data Set however, this
difference was distributed proportionally to the amount of energy consumption already provided
for a given month. This means that for months with lower energy consumption, a smaller
portion of the difference was added, while for months with higher energy consumption, a larger
portion of the difference was added. By distributing the difference proportionally, the total
consumption normalization contributes to both the base load and the variable load. In the Meta-
Analysis, the total consumption normalization only affected the base load since the same amount
of energy was being added or subtracted from every month.
Weather normalization
All of the buildings in the Refined Data set were located in Toronto, but since the energy
consumption data came from different years ranging from 1999 to 2011, the data still had to be
weather normalized. Two modifications were made to the weather normalization procedure
applied to the Refined Data Set: cooling degree days (CDDs) were taken into account in the
normalization of the electricity data, and base values other than an 18°C set point for the HDD
data were considered.
The only two weather variables considered in the weather normalization were HDDs and CDDs.
Other weather conditions such as temperature maximum and minimum values, humidity and
cloud cover may influence energy consumption; however, these weather conditions are not
commonly considered in studies involving weather normalization of building energy
consumption data.70
The weather dependency of natural gas use is only related to heating and was therefore weather
normalized using HDDs only. However, electricity consumption can be related to air
conditioning loads as well as heating loads depending on the heating energy type. As such,
electricity consumption data were weather normalized considering both HDDs and CDDs.
39
Therefore, three weather normalization processes were completed on the electricity data. One
normalization process used HDDs only, one normalization process used CDDs only and one
normalization process used HDDs for the winter months (October to March) and CDDs for the
summer months (April to September). The normalization process that yielded the highest R2
value was the normalization process that was finally applied. The same weather normalization
process using linear regression that was used in the Meta-Analysis and is described in Section
5.1.1 was also used for the Refined Data Set.
In the Meta-Analysis, a base of 18°C was assumed for calculating the HDDs. The validity of
this assumption was tested in the weather normalization of the Refined Data Set by testing bases
of 16°C and 20°C in the normalization of the natural gas data for half of the buildings in the data
set. The natural gas data was normalized three times against the HDDs calculated from a
different base. The normalization that yielded the highest R2 value was the HDD base that was
considered the most appropriate for the building. For 90% of the buildings, the HDDs calculated
from an 18°C base produced the highest R2 values. For consistency, an 18°C base was used for
all buildings.
5.2.2 Determination of energy load types
As in the Meta-Analysis, both the y-intercept method and the low average method were used to
differentiate between the base and the variable loads; however, slightly different selection
criteria were used to determine which method was applied to each data set. The base electrical
load for buildings that were weather normalized using HDDs or CDDs only was determined
using the y-intercept method. For buildings that were normalized using a combination of HDDs
and CDDs, the average of the two lowest months of electricity consumption from every year of
data were used to determine the base electrical load. The base natural gas load was determined
using the y-intercept method for all of the buildings. Therefore, there are four components of
energy consumption data that will be referred as follows: variable natural gas, base natural gas,
variable electricity and base electricity.
5.2.3 Rejection of outliers
The natural gas and electricity consumption data were obtained from the billed energy use for
each of the buildings in the data set. Prior to weather normalization, outliers in the energy
40
consumption data were removed. Outliers can sometimes occur when the billed energy use does
not reflect the actual energy use in the period. A common circumstance in which billed data
does not align with actual energy use is when meter readings are estimated by the utility
companies. In such cases, outliers can arise, particularly where the actual meter reading and the
billing correction is not made for a few months. Outliers can also occur when building systems
are shut down for replacement or maintenance. In some cases, occupant behaviour can be
responsible for outliers in the energy consumption data. Finally, errors in data entry can also be
a source of outliers.
Outliers were removed when energy consumption was plotted against HDDs or CDDs in the
weather normalization process. Removal was based on the following criteria:
• When an electricity consumption datum was more than 20% different from the predicted
electricity consumption as determined by the equation for the line of best fit, it was
removed.
• When a natural gas consumption datum was more than 30% different from the predicted
natural gas consumption as determined the equation for the line of best fit, it was
removed.
• A few of the buildings had energy consumption data that appeared to contain outliers for
the same two to four months of the year for every year of data. To avoid removing too
many outliers and leaving a gap in the data, each calendar month where the data showed
the lowest error was preserved and included in the weather normalization. Therefore, a
rule was developed that a monthly datum was only removed as long as all 12 calendar
months were still represented after its removal.
41
Chapter 6 Meta-Analysis Results and Discussion
The results, discussion and conclusion for correlation analysis of the Meta-Analysis data set are
presented in this Chapter. A summary of the energy consumption data is presented and
compared with other studies. This is followed by the correlation analysis graphs combined with
a discussion of each graph. Finally, the conclusions from the Meta-Analysis are outlined.
6.1 Meta-Analysis energy consumption data
This section summarizes the energy consumption characteristics of the Meta-Analysis data in
terms of energy intensity, energy use per suite and base loads. A discussion of the R2 values for
weather normalization is also included.
6.1.1 Summary of energy intensity data
For buildings where natural gas and electricity data were available (HiSTAR and Tower
Renewal data), the annual energy intensity was based on the total annual energy consumed from
both electric and natural gas sources divided by the building’s gross floor area. This total energy
intensity split out by natural gas and electricity consumption is shown in Figure 9. The 42 Green
Condo Champions buildings did not include electricity data, but the annual natural gas energy
intensity data were presented for reference purposes in Figure 9.
The total energy intensity values vary widely from 88 ekWh/m2 to 520 ekWh/m
2 with an average
of 295ekWh/m2. This is in line with the combined average energy intensity for Ontario MURBs
of 305ekWh/m2, which was calculated from the consumer-side studies examined in the literature
review in Chapter 2. The average energy mix of the Meta-Analysis data set is 38% electricity
and 62% natural gas, which is similar to the Comprehensive Energy Use Database mix of 33%
electricity and 67% natural gas for Ontario apartments in 2009.71
Since the comparison is on a
relative basis, the comparison between this consumer-side study and supplier-side data is valid.
Natural gas appears to be the primary heat source in the majority of the buildings. However, a
few of the buildings use a significantly higher proportion of electricity. Presumably these
42
buildings have electric space heating systems and use natural gas to heat domestic hot water, but
further investigation would be required to confirm this assumption.
Figure 9: Total annual energy intensity
The annual natural gas energy intensity is shown separately from electricity use for the 108
buildings in Figure 10. The natural gas energy intensity is in equivalent kilowatt-hours per
square metre and ranges from 26 ekWh/m2 to 447 ekWh/m
2. The yellow bars included in Figure
10 show the range of potential thermal energy intensities for each building. The thermal energy
intensity represents the useful thermal energy delivered to the building and is a function of boiler
efficiency. Boiler efficiency can vary from about 60% up to almost 100%. Therefore, a 40%
range has been applied in Figure 10. It is expected that the actual thermal energy intensity
required to meet the building’s natural gas heating demand lies somewhere within the range
represented by the yellow bar.
The buildings with the highest natural gas energy intensity values may be good candidates for
new boilers with higher efficiencies. The yellow bars show the range of energy intensity
reductions which could be achieved by installing more efficient boilers, though it is important to
note the numerous other factors affecting these systems such as operating and maintenance
items. Regardless of boiler efficiency or operation, the buildings will still demand the same
amount of heat energy, but having a more efficient boiler system means that more of the energy
0
100
200
300
400
500
600
H7
H5
5
H1
0
H1
8
H2
0
H5
4
H3
2
H3
5
H3
4
H1
9
H3
7
H5
1
H4
0
H4
3
C1
7
C3
9
C1
9
C3
4
C1
6
C2
0
C3
6
C3
5
C2
1
C1
1
C4
2
T1
1
T9
En
erg
y I
nte
nsi
ty (
ek
Wh
/m2)
Total Annual Energy Intensity
Natural Gas Electricity
HiSTAR Green Condo Champions Tower
Renewal
43
delivered to the building becomes usable heat directed to occupied spaces. In other words, less
natural gas is required to produce the same amount of usable heat.
Figure 10: Annual natural gas intensity of each building
The total energy use per suite is based on the total energy provided from electric and natural gas
sources divided by the number of suites in the building (HiSTAR and Tower Renewal data only).
Figure 11 shows that the total annual energy use on a per suite basis ranges from
6,270ekWh/suite to 58,480ekWh/suite with an average of 25,100ekWh/suite. This energy use
per suite is nearly twice as high as the SHEU 2007 data, which quotes an average of
14,500ekWh/suite for low-rise apartments and 12,400ekWh/suite for high-rise apartments.72
This comparison should be taken with caution because of the limitations of comparing SHEU
data with consumer-side studies previously discussed. The NRC Toronto data set, which is a
consumer-side study, provides a much higher average energy use per suite of 35,000ekWh/suite
and a range of 23,100ekWh/suite to 47,600ekWh/suite.73
Since this study is based on data from
the 1970s, it is expected that the energy use would be higher than the Meta-Analysis data.
0
50
100
150
200
250
300
350
400
450
500
C2
6
C1
7
H2
2
H5
0
C2
5
T1
0
H5
5
H1
5
C2
2
C4
1
C1
6
C3
C2
0
C3
3
H9
C2
H3
1
C4
C2
1
H3
0
C6
H4
1
T7
C3
2
H2
6
T3
C8A
nn
ua
l N
atu
ral
Ga
s E
ne
rgy
In
ten
sity
(ek
Wh
/m2)
Building Name
Annual Natural Gas Energy Intensity
44
Figure 11: Total annual energy use per suite
6.1.2 Summary of greenhouse gas emissions
The greenhouse gas (GHG) emissions due to electricity and natural gas consumption in MURBs
were calculated by applying emissions factors. The emissions factors have been expressed in
terms of kilograms of carbon dioxide equivalent (kgCO2e). The emissions factor for natural gas
is 1.879kgCO2e/m3 and the emissions factor for electricity (2010) is 0.187kgCO2e/kWh.
74
Figure 12 shows the annual GHG emissions from each building per square metre. Similar to
Figure 10, Figure 12 splits up the GHG emissions so that emissions due to electricity and natural
gas are shown separately. The buildings from the Green Condo Champions data set only show
emissions due to natural gas because no electricity data was available. The total annual GHG
emissions intensities range from 16.6kgCO2e/m2 up to 96.6kgCO2e/m
2 and the average was
55.3kgCO2e/m2. Since the GHG emissions intensities are a directly related to the energy
intensities, the GHG emissions intensities show the same wide variation within the data set.
0
10000
20000
30000
40000
50000
60000
70000
T1
0
H5
H1
5
H2
0
H1
0
H1
4
H6
H4
H2
9
H7
T5
H2
H3
0
H3
8
H3
2
H3
1
T7
H4
4
H5
5
H1
9
T6
T8
H4
0
H4
6
T2
T4
H4
3
H3
6
T3
H5
2
H4
9
H4
7
H2
1
To
tal
An
nu
al
En
erg
y p
er
Su
ite
(ek
Wh
/su
ite
)
Building Name
Total Annual Energy Per Suite
45
Figure 12: Total annual GHG emissions intensity
6.1.3 Base load determination
As explained in Chapter 5, two methods were used to determine the base load within the
electrical and natural gas energy consumption data. The first method used the y-intercept from
the weather normalization and the second method used the average of the two months with the
lowest energy consumption. Figure 13 and Figure 14 show comparisons of the base load
calculation methods for electricity and natural gas respectively.
There is good agreement between the results of the two methods used to derive the electric base
load. This could be because most of the buildings are heated with natural gas and the electricity
load remains relatively constant throughout the year. The y-intercept method was preferred
because its value is based on a greater number of data points than the low average method.
However, for buildings exhibiting an R2 value lower than 0.6, or those with fewer than six
months of energy consumption data, the low average method was preferred.
0
20
40
60
80
100
120H
7
H5
5
H1
0
H1
8
H3
H5
4
H3
2
H3
5
H3
4
H4
7
H6
H3
0
H4
0
H4
3
C1
7
C3
9
C1
9
C3
4
C1
6
C2
0
C3
6
C3
5
C2
1
C1
1
T1
0
T2
T6GH
G E
mis
sio
ns
Inte
nsi
ty (
ek
gC
O2/m
2)
Total Annual GHG Emissions Intensity
Electricity Emissions Natural Gas Emissions
HiSTAR Green Condo ChampionsTower
Renewal
46
Figure 13: Comparison of base load calculation methods for electricity
There is more variation between the results of the two methods used for natural gas base load
derivation. This is likely because most buildings are heated with natural gas, so natural gas
consumption data show greater seasonal variation than electricity consumption data; therefore,
the annual variation makes it more difficult to strip out the base load. Once again, the y-intercept
method was generally preferred but the low average method was used for buildings exhibiting an
R2 value lower than 0.6.
0
500,000
1,000,000
1,500,000
2,000,000
2,500,000A
nn
ua
l B
ase
Lo
ad
(k
Wh
)
Buildings in Data Set
Electric Base Load Comparison
Y-Intercept Method
Low Average Method
47
Figure 14: Comparison of base load calculation methods for natural gas
6.1.4 Summary of weather normalization
Chapter 5 includes a discussion of the importance of R2 values and the typical acceptable range.
Figure 15 shows the range of R2 values achieved during the weather normalization process of the
HiSTAR building data. Only 20% were considered to have a reliable correlation between
energy use and weather for both natural gas and electricity use. However, the results of the
correlation analysis show that the buildings with poor R2 values were not typically the outliers,
so the effects of the low R2 value should not skew the correlations significantly. The Green
Condo Champions natural gas consumption data generally showed a very strong correlation in
the weather normalization and yielded an average R2 value of 0.91. The annual Tower Renewal
data did not have associated R2 values.
0
500
1,000
1,500
2,000
2,500
3,000B
ase
Lo
ad
(1
,00
0s
of
ek
Wh
)
Buildings in Data Set
Natural Gas Base Load Comparison Y-Intercept Method Low Average Method
48
Figure 15: Coefficient of determination values from weather normalization
6.2 Correlation results and discussion
Correlations between energy consumption data and building age, building height, suite size and
ownership type are presented and discussed in the following sections. Many of the graphs show
a line of best fit and an associated R2 values. This does not indicate that a significant
relationship between the two variables necessarily exists. It is only meant to illustrate whether
the correlation is weak or strong.
6.2.1 Date of construction
There is generally a negative correlation between energy intensity and building age as shown in
Figure 16. Up until the 1970s, the energy intensity of older buildings was found to be somewhat
higher than the newer buildings. The average energy intensity of buildings constructed in the
1980s and 1990s was slightly higher than those constructed in the 1970s. The higher energy
intensity of the oldest buildings in the data set could be due to the age of the mechanical systems
0
0.6
0 0.6
R2
va
lue
fro
m E
lect
rici
ty N
orm
ali
zati
on
R2 value from Natural Gas Normalization
R2 Values - HiSTAR Weather Normalization
Acceptable
Correlation
Poor
Correlation
49
and condition of the building envelope; however, the increase in energy use in newer buildings
may be due to increased fenestration-to-wall ratios typically seen in more modern buildings as
curtain wall construction became more prevalent. Further analysis of the building characteristics
of each building is required to verify this hypothesis.
Figure 16: Energy intensity versus building vintage
6.2.2 Weather-dependant loads
After determining the base loads for each building, these loads were removed from the total
energy consumption to determine the total weather-dependant load, which is also called the total
variable load. Figure 17 shows the total variable energy intensity (total variable load divided by
the building floor area) plotted against the building vintage. This correlation was plotted to test
the hypothesis that the trend shown in Figure 16 of older buildings having higher energy
intensities exists because of changes in building envelope characteristics over time.
Building envelope characteristics primarily influence the variable load component of energy
consumed because the resistance to heat transfer provided by the building envelope is one of the
factors that dictate the heating and cooling requirements of a particular building. Figure 17
shows that isolating the variable energy intensity has made the trend of older buildings having a
higher energy intensity more pronounced. This supports the suggestion that building envelope
characteristics are one of the key variables driving this trend. Many other variables also
influence this relationship and must be examined further before any conclusive statements can be
0
100
200
300
400
500
600
1930 1940 1950 1960 1970 1980 1990 2000
En
erg
y I
nte
nsi
ty (
ek
Wh
/m2)
Date of Construction
The Influence of Building Vintage on Energy Intensity
Before 1960 1961-1970 1971-1980 1981-1990 1991-2000 Average EI
50
made; however, improvements such as increased air tightness and thermal resistance would
presumably go furthest to improve energy performance in buildings with high total variable load
intensities like the ones built before 1970. The interaction between specific building envelope
characteristics and variable energy intensity are tested in the correlation analysis for the Refined
Data Set in Chapter 7.
Figure 17: Weather-dependant loads versus building vintage
6.2.3 Building height
Logically, total energy use should increase with building height and overall building size as
shown in Figure 18. Although the gradient is statistically significant at a 95% confidence level
based on a t-test, the coefficient of correlation is only 0.6. The lack of an excellent correlation in
Figure 18 demonstrates that there are other factors besides building size contributing to energy
consumption. As such, Figure 19 shows no correlation between energy intensity and number of
storeys. The building vintages were highlighted in different colours in Figure 19 to determine
whether any trends existed when building age was also considered; however, no trends were
detected.
0
50
100
150
200
250
300
350
400
1940 1950 1960 1970 1980 1990 2000
Va
ria
ble
En
erg
y I
nte
nsi
ty (
ek
Wh
/m2
)
Building Construction Date
The Influence of Building Vintage on Weather-
Dependant Energy IntensityBefore 1960 1961-1970 1971-1980 1981-1990 1991-2000 Average
51
Figure 18: Total energy use versus number of storeys
Figure 19: Energy intensity versus building height
6.2.4 Number and size of suites
Similar to building height, as the number of suites increase, energy use is expected to increase as
shown in Figure 20. The gradient in Figure 20 is statistically significant at a 95% confidence
level based on a t-test and the coefficient of determination is 0.59. However, the gradient in
Figure 21 is not statistically significant and the coefficient of correlation is less than 0.005.
Therefore, the reasonable correlation in Figure 20 and lack of correlation between energy
y = 300,000x + 430,000
R² = 0.60
0
2,000
4,000
6,000
8,000
10,000
12,000
0 10 20 30 40To
tal
An
nu
al
En
erg
y (
1,0
00
s o
f e
kW
h)
Number of Storeys
The Influence of Number of Storeys
on Total Energy
0
100
200
300
400
500
600
0 10 20 30 40
En
erg
y I
nte
nsi
ty (
ek
Wh
/m2)
Number of Storeys
The Influence of Number of Storeys on Energy
Intensity by Building VintageBefore 1960 1961-1970 1971-1980 1981-1990 1991-2000
52
intensity and number of suites in Figure 21 mean there are other parameters that need to be
considered.
Figure 20: Total energy use versus number of suites
Figure 21: Energy intensity versus number of suites
Figure 22 shows a plot of the attributed suite size versus date of construction. From the oldest
buildings in the data set to the 1980’s, the average attributed suite size increases. In recent years,
the data reveal the prevalence of smaller suite sizes. This could be influenced by the size of
common area spaces or the actual suite sizes. Larger suites could house more people which
would likely result in higher energy use; however, this is not reflected in the energy intensity by
R² = 0.5933
0
2,000
4,000
6,000
8,000
10,000
12,000
14,000
16,000
18,000
0 200 400 600 800To
tal
An
nu
al
En
erg
y (
1,0
00
s o
f e
kW
h)
Number of Suites
The Influence of Number of Suites on Total
Energy
R² = 0.0048
0
100
200
300
400
500
600
0 200 400 600 800
En
erg
y I
nte
nsi
ty (
ek
Wh
/m2
)
Number of Suites
The Influence of Number of Suites on Energy
Intesnity
53
vintage shown in Figure 16 above. Therefore, the average number of people per suite may have
remained the same over a period of increasing attributed suite sizes but occupancy information is
required to verify this.
Figure 22: Attributed suite size versus date of construction
6.2.5 Ownership type
Figure 23 and Figure 24 show the energy use per floor area and per suite for different types of
building ownership. This was investigated because energy use is often a function of
responsibility for payment of utility bills. Generally speaking, condominium owners are more
likely to be sub-metered than renters; therefore, condo owners may use less energy because they
are directly responsible for the cost. This hypothesis is supported in Figure 23, which shows that
condos have the lowest average annual energy intensity. However, Figure 24 shows that condos
have the highest energy use per suite, which simultaneously negates the hypothesis. Evidently,
other factors besides the ownership type play a role. Household income may be one of these
factors. It is reasonable to assume that condo owners would have a higher income than renters or
those in subsidized housing. With a higher income, the condo owners are more likely to have
larger suites with more appliances and electronics that consume energy. Another explanation for
having the lowest energy intensity but the highest per suite energy consumption could be due to
the common areas. Condominiums often have larger common areas including a lobby, exercise
rooms, pools, party rooms and board rooms. Since the total building energy divided by the
0
50
100
150
200
250
300
1940 1950 1960 1970 1980 1990 2000 2010
Att
rib
ute
d S
uit
e A
rea
(m
2)
Date of Construction
The Influence of Construction Date
on Suite SizeBefore 1960 1961-1970 1971-1980 1981-19901991-2000 2001-2010 Average
54
number of suites is being measured and not the actual energy consumption per suite, this could
be another explanation of the results.
The energy intensity of the subsidized rental on both the area and per suite basis are lower
compared to the other ownership types. This may be because in these particular buildings,
owners are typically working with very tight operating budgets and may see energy efficiency as
a means of controlling this budget. Owners of rental properties also typically have more control
over the types of lighting and appliances in the building perhaps allowing them the ability to
better control energy use. Finally, the common areas in these buildings are typically less
extensive than in a condominium, which could contribute to lower energy use on a per suite
basis.
Figure 23: Total energy intensity by area for each type of building ownership
0
100
200
300
400
500
600
En
erg
y I
nte
nsi
ty (
ek
wh
/m2)
Energy Intensity by Ownership TypeOwned, No Subsidy Owned, subsidized Rented, No Subsidy
Rented, Subsidized Average
55
Figure 24: Total energy intensity by suite for each type of building ownership
Suite size is also worth investigating because this affects energy intensity and occupancy.
However, without occupancy data, it is not possible to analyze this particular correlation further.
Figure 25: Suite size versus building ownership type
6.2.6 Statistical Significance
Figure 16, Figure 17 and Figure 22 to Figure 25 were all graphs that showed trends based on
categories with bars representing the average for each group. The difference between the
averages from one category to another was not statistically significant for any of the graphs. The
0
10000
20000
30000
40000
50000
60000
70000
Su
ite
En
erg
y U
se (
ek
Wh
/su
ite
)
Suite Energy Use by Ownership TypeOwned, No Subsidy Owned, subsidized Rented, No Subsidy
Rented, Subsidized Average
0
50
100
150
200
250
Su
ite
Siz
e (
m2)
Suite Size by Ownership TypeOwned, No Subsidy Owned, Subsidized Rented, No SubsidyRented, Subsidized Average
56
significance test was based on a two-tailed t-test at a 95% confidence level. Even though the
results are not statistically significant, they can still be used as a basis for observations; however
further testing and study is required before these results should be used as a basis for decision
making.
6.3 Preliminary conclusions
Generally the data analysis did not yield the strong correlations that were anticipated at the outset
of the Meta-Analysis. However, considering the limitations of the data, there are some
interesting observations and conclusions that can be drawn. The findings of the Meta-Analysis
also guided the focus of the investigations carried out on the Refined Data Set:
1. The Meta-Analysis of MURB energy consumption data represents a reasonable sample
size. Energy consumption data for approximately 1.8% of the total MURB stock and
4.8% of the mid and high-rise MURB stock in the City of Toronto were analyzed in the
Meta-Analysis across all vintages of buildings.
2. The energy intensity values determined from actual natural gas and electricity metering
were similar to other published consumer-side studies of Ontario MURBs. An average
intensity of 295ekWh/m2 was found, which is similar to the average energy intensity of
305ekWh/m2 from the consumer-side studies reviewed in Chapter 2.
3. There was a wide range in the observed energy intensities with the lowest energy
intensity less than one fifth of the highest one. Even within each vintage category, there
was significant variation in energy intensity. The range in energy intensity may reflect a
variety of factors including differences in the way the buildings are operated, differences
in the efficiency of the buildings’ major mechanical and electrical systems, and
differences in construction type.
4. Since no reliable data on the efficiencies of the natural gas burning appliances were
available for the Meta-Analysis, thermal energy intensities could only be bracketed
assuming that hypothesized efficiencies could be as low as 60% but no more than 100%.
Efficiencies as well as the operating and maintenance condition of the boilers should be
included in further investigation in the Refined Data Set so that conclusions can be drawn
about whether envelope retrofits, mechanical retrofits or both would be the best energy
reduction strategy.
57
5. An examination of the relationship between energy intensity and year of construction as
shown in Figure 17 reveals that variable energy intensity generally decreased until the
1970s. After the 1970s, intensities began to rise. This latter rise may be due to the
combined effects of better thermal insulation and air tightness measures being off-set by
higher fenestration-to-wall ratios. This hypothesis may explain the apparent paradox of
the declining energy-efficiency of the more modern MURBs in the data set.
6. In a preliminary way, the Meta-Analysis also explored the relationship between
ownership type and energy intensity. Although condominiums had the lowest average
energy intensity on a gross floor area basis, they had the highest energy intensity on a per
suite basis. This is counter to intuition unless the effect of household income and energy
used in common areas are considered in the explanation. Further study is necessary so
that the underlying reasons can be more clearly identified and effective measures taken to
reduce energy consumption in all MURBs.
58
Chapter 7 Refined Data Set Results
This chapter presents a selection of correlations between the weather-normalized energy
consumption data and a number of building characteristics. In the sections that follow, an
overview of building energy intensity has been presented, followed by a discussion of the
variables that influence energy use. To improve correlations examined in the Meta-Analysis,
buildings were grouped in general energy-use categories as discussed in Section 7.3. Next the
correlations revealed during analysis of the Refined Data Set have been presented. Finally, the
results of a multi-variable regression analysis are provided.
7.1 Energy intensity
Within the Refined Data Set, which is focused on only mid and high-rise MURBs, the total
annual energy consumption ranged from 1,125 MWh to 12,190 MWh. In order to facilitate
comparisons between buildings, it is helpful to normalize the data based on building size.
Figure 26 and Figure 27 show the annual energy consumption on a per gross floor area (energy
intensity) and on a per suite basis. As discussed in Chapter 5, the gross floor area is considered
to be the total conditioned floor area in a building, but typically does not include underground
parking, even if the parking area is conditioned.
Figure 26: Natural gas and electricity components of total annual energy per gross floor area
0
100
200
300
400
500
600
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39
En
erg
y I
nte
nsi
ty (
ek
Wh
/m2)
Building Number
Total Annual Energy Intensity
Variable Natural Gas Base Natural Gas Variable Electrical Base Electrical
59
Figure 27: Natural gas and electricity components of total annual energy use per suite
Figure 26 and Figure 27 show that even after being normalized for building size, there is still a
great deal of variation. The relative standard error is a statistical measure that indicates
dispersion of data. It is a normalized measure of standard deviation calculated by the dividing
standard deviation by the sample mean. The relative standard error in the sample of values for
total annual energy consumption is reduced from 62% to 50% when normalized by number of
suites and reduced to 31% when normalized by gross floor area. Therefore, the industry practice
of normalizing by floor area is supported by this statistic.
The average energy intensity for the data set is 290ekWh/m2, which is statistically the same as
the average intensity of the Meta-Analysis data set at 295ekWh/m2. This energy intensity value
is close to the average energy intensity of the consumer-side studies of Ontario MURBs of
305ekWh/m2, which were examined in the literature reivew. The average energy mix of the
data set is 33% electricity and 67% natural gas which is the same as CEUD mix for Ontario
apartments in 2009.75
In Figure 27, the attributed suite size has been plotted above the energy use per suite for
comparison. As in the Meta-Analysis, attributed suite size was calculated by dividing the gross
floor area by the total number of suites in the building, and is therefore an overestimate.
Generally, larger suite size can be used to explain higher per suite energy use. In most cases, the
-100
0
100
200
300
0
20
40
60
80
100
120
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39
Att
rib
ute
d S
uit
e S
ize
(m
2)
En
erg
y P
er
Su
ite
(1,0
00
s o
f e
kW
h/s
uit
e)
Building Number
Total Annual Energy Use Per Suite
Variable Natural Gas Base Natural Gas Variable Electrical
Base Electrical Suite Size
60
reason a building has larger attributed suite sizes is because the building has significant common
facilities whose floor area has been attributed to each suite.
7.2 Selection of variables
As normalization by building size does not fully explain the variation in energy use, the purpose
of this section is to determine what other variables have a correlation with the energy
consumption data and can rationalize components of the variation. There is a diverse range of
variables that could contribute to the energy consumption within a building. Variables related to
occupant behaviour, mechanical and electrical systems, control systems, building envelope, site
environment, building management or demographics could all be involved. The variables
examined in the Refined Data Set have been limited to physically measureable or observable
variables that were available in the building energy audit reports. Additionally, findings from the
Meta-Analysis were used to guide predictions on the variables that would be expected to have
the greatest effect on the variation in energy consumption. Figure 28 provides a summary of
possible variables and an indication of the data availability. Thermal glazing characteristics, the
efficiency of the space heating system and the number of occupants are the three variables with
the highest level of expected importance. In this figure, the expected importance of the variables
has been based on the initial belief that variables which affect heat loss and heat generation in
buildings should be well correlated with energy consumption.
61
Figure 28: Variables categorized by ease of measurement and predicted importance
7.2.1 Variables related to heating loads
The heating energy requirements for a building are a function of:
(a) the heat loss through the building envelope;
(b) the efficiency of heat generation.
Based on equations for conduction heat loss and air leakage provided in Appendix E, glazing
type and glazing area are the variables expected to govern heat loss. The overall conductance of
a window is denoted by its U-value, while the air tightness of a window is generally determined
by whether the window is fixed or operable; by the type of operable window (sliding, awning or
casement); and by the condition of the window.
In the case of natural gas heated buildings, the variable that is expected to affect heat generation
is the boiler plant efficiency. Although estimates of boiler efficiency and age were available in
this data set, there are many other factors such as operation and maintenance that contribute
significantly to the overall plant efficiency which have not been captured.
62
Figure 29 summarizes the main contributing factors to heat loss and generation for which data
were available.
Figure 29: Variable dominating building heating requirements
7.2.2 Variables related to cooling loads
Similar to heating, cooling loads are affected primarily by heat gain through the building
envelope and the efficiency of the cooling system. In addition to being influenced by conduction
and air leakage through the glazing, solar heat gains due to radiation through the glazing are also
expected to have a significant effect. Based on the equation for radiation heat transfer found in
Appendix E, the two variables with the greatest effect are the glazing area and the solar heat gain
coefficient (SHGC) of the glass. With a higher SHGC, more radiation can penetrate the building
and heat it up, resulting in a higher cooling load.
7.2.3 Variables related to base loads
Base loads are generally expected to be a function of the number of occupants, the gross floor
area and equipment efficiencies. Domestic hot water use and plug loads are most closely related
to the number of occupants because the activity of each occupant defines the magnitude of these
loads. Lighting, fan, and pump loads are most closely related to the gross floor area because the
size of the interior space defines the magnitude of loads. Since heating and cooling equipment
efficiencies are often difficult to obtain, building age many be used as a proxy so long as the
building has not undergone any significant renovations. Base loads, however, cannot be readily
separated into these categories based only on monthly energy consumption data. Therefore, the
following relationships were expected:
63
����������� ������ = ������������������ (eqn. 1)
������������������������ = ������� � �, ���������������� (eqn. 2)
7.3 Separation of building types
The majority of the buildings in the Refined Data Set use natural gas boilers as their primary
heating system; however, some of the buildings are heated primarily with electricity or with a
split between electricity and natural gas. To ensure that like-to-like comparisons were made,
buildings within the data set have been grouped by the type of heating system before variable
natural gas energy and variable electricity were correlated with other variables.
Figure 30 shows a plot of annual variable natural gas intensity versus annual variable electricity
intensity. Buildings with more than 100ekWh/m2 of variable natural gas intensity and less than
30kWh/m2 of variable electricity intensity, which occupy the upper left quadrant in Figure 30,
were considered to have natural gas boilers as their primary heating system. The delineations of
100ekWh/m2 and 30kWh/m
2 were chosen after the graph was plotted based on the cluster of
buildings that appeared to be heated with natural gas boilers. The three buildings in the lower
right quadrant are buildings that are most likely heated with electricity. Finally, the eight
buildings in the lower left quadrant have both low variable natural gas intensity and variable
electricity intensity. This low intensity led to the preliminary conclusion that the lowest four
buildings were anomalous. Therefore, these buildings are examined further in Chapter 8 of this
thesis. The other four buildings in the lower left quadrant were confirmed as being heated with
a combination of natural gas and electricity based on information from their building energy
audit report.
64
Figure 30: Method used to separate buildings by primary heating systems
Buildings with air-conditioning were separated based on having an electrical normalization that
related to CDDs with an R2 value of at least 0.5. Table 9 summarizes the number of buildings
assigned to each category.
Table 9: Summary of building space conditioning categories
Building space conditioning category # of buildings assigned
to category
Primarily natural gas heating 29
Primarily electrical heating 3
Combined natural gas and electrical heating 4
Air conditioned buildings 11
Buildings with undetermined heating systems 4
7.4 Correlation analysis
The correlations presented in this section were chosen for analysis based on the logic outlined in
the selection of variables section of this chapter (see Section 7.2) and the findings of the Meta-
Analysis. The building characteristics involved in these correlations include the fenestration-to-
wall ratio, window conductance, boiler efficiency, cooling capacity of air conditioners and the
number of occupants.
0
50
100
150
200
250
300
350
0 20 40 60 80 100 120
An
nu
al
Va
ria
ble
Na
tura
l G
as
Inte
nsi
ty
(ek
Wh
/m2)
Annual Variable Electricity Intensity (ekWh/m2)
Determination of Primary Heating Fuel Source
Natural Gas
Heating
Electrical Heating
Undetermined/
Combination
65
7.4.1 Window characteristics
The first variables examined in the correlation analysis of the Refined Data Set were related to
the window characteristics. As explained in Section 7.2, window area, air tightness, thermal
conductance, and SHGC of windows are expected to influence both heating and cooling loads.
Some of the plots include buildings which have been identified as outliers. These anomalies will
be examined in Chapter 8. The line of best fit plot applies only to data points that are not
considered outliers.
7.4.1.1 Fenestration-to-wall ratio
The fenestration-to-wall ratio (fenestration ratio) is the area of the building’s exterior walls
covered in glazing divided by the total wall surface area. Although the fenestration ratio for each
building was stated in many of the audit reports, it was often based on an estimate. Estimates of
the fenestration ratio were checked and modified as necessary by comparing the stated
fenestration ratio with building photographs. By comparing photographs of the various
buildings, fenestration ratios were corrected so that buildings with similar glazing configurations
had similar fenestration ratios.
Since much of the building information obtained for this study was provided on the condition
that the building identity remains confidential, full building photographs could not be provided in
this report. However, close-up photos which maintain the anonymity of the buildings have been
provided in Appendix F. In addition to the photographs, the originally estimated fenestration
ratios as well as the revised estimate of the fenestration ratios have been provided for each
building. Figure 31, Figure 32, and Figure 33 are all based on the revised estimate of
fenestration ratios.
Since the majority of heat loss and solar heat gain through the building envelope is often through
the glazing, it is expected that the larger the fenestration ratio, the higher the heating and cooling
loads will be. This relationship was shown to be stronger in buildings with double-glazed
windows and natural gas heating (Figure 31), than for buildings with single-glazed windows and
natural gas heating (Figure 32).
66
Figure 31: Variable natural gas intensity versus fenestration ratio for double-glazed windows
Figure 32: Variable natural gas intensity versus fenestration ratio for single-glazed windows
One of the reasons that the R2 value for Figure 32 is lower than the R
2 value in Figure 31 is that
the buildings with single-glazed windows are generally older and the glazing is in worse
condition. Thus, the air tightness of the glazing assemblies may be the factor that governs heat
loss for these buildings, not just the fenestration ratio. Since no information was available on the
air tightness of the glazing assemblies for the buildings in the data set, it might be helpful to field
examine the windows in question. In such a field review, the type of window and the proportion
y = 150x + 84
R² = 0.69
0
50
100
150
200
250
300
350
0% 20% 40% 60% 80%
Va
ria
ble
Na
tura
l G
as
Inte
nsi
ty
(ek
Wh
/m2)
Fenestration Ratio
The Influence of Fenestration Ratio on
Variable Natural Gas Intensity - Double Glazed
Outliers
y = 250x + 120
R² = 0.41
0
50
100
150
200
250
300
350
0% 20% 40% 60% 80%
Va
ria
ble
Na
tura
l G
as
Inte
nsi
ty
(ek
Wh
/m2)
Fenestration Ratio
The Influence of Fenestration Ratio on
Variable Natural Gas Intensity - Single Glazed
Outliers
67
of operable to fixed windows should be determined since operable windows greatly affect the air
leakage of buildings and ultimately energy losses. Recording window age and any resealing that
may have occurred would also be helpful in any future data collection efforts to determine an
estimated air leakage contribution.
Cooling loads are also affected by the fenestration ratio because of potential for solar gains
through glazing in addition to conductive and convective heat gains from the outdoors. Figure
33 shows that a higher fenestration ratio leads to greater air conditioning loads as expected. Of
the 11 buildings identified as having air conditioning, nine had double-glazed windows and were
included in Figure 33. For most of the buildings, no information on the SHGC of the windows
was available and could not be considered. However, the correlation between electrical intensity
and fenestration ratio is reasonably strong, so the effect of the SHGC may not be as important as
the fenestration ratio.
Figure 33: Electricity intensity versus fenestration ratio for air conditioned buildings
Figure 31 and Figure 33 both have statistically significant gradients at a 95% confidence level,
however Figure 32 does not. At an 80% confidence level, the gradient in Figure 32 is
statistically significant.
7.4.1.2 Window conductance
The glazing assemblies of each building have been assigned an overall thermal conductance
value (U-value) that was either provided in the audit report or is based on a physical description
y = 57x + 58
R² = 0.58
0
20
40
60
80
100
120
140
0% 20% 40% 60% 80%
To
tal
Ele
ctri
cal
Inte
nsi
ty (
ek
Wh
/m2)
Fenestration Ratio
The Influence of Fenestration Ratio on
Total Electrical Intensity in Air Conditioned
Buildings - Double Glazed
68
of the windows. Higher U-values mean more heat transfer and thus heating and cooling loads
will presumably be higher. Both Figure 34 and Figure 35 show the expected correlation. In
Figure 34, the trend is weak, so the average variable natural gas intensity of the data for each of
the three U-values has been plotted to show the trend instead of a line of best fit. At a 95%
confidence level the difference in average values for the groups were not statistically significant.
However, the gradient in Figure 35 was statistically significant at a 95% confidence level.
The considerable variation in the data shows that glazing U-value contributes to heat loss in a
building, but is not the only governing factor. The stronger correlation shown with fenestration
ratio suggests that glazing area has a more significant effect on heating intensity than window
thermal conductance does.
Figure 34: Variable natural gas intensity versus glazing U-value
0
50
100
150
200
250
300
350
0 1 2 3 4 5 6Va
ria
ble
Na
tura
l G
as
Inte
nsi
ty (
ek
Wh
/m2)
Window U-Value (W/m2K)
The Influence of Window U-Value on
Variable Natural Gas IntensityAverage Outliers
69
Figure 35: Electrical intensity versus glazing U-value for air conditioned buildings
7.4.2 Heating efficiency and cooling equipment
The second factor affecting building heating loads is the efficiency of the heating system. It is
expected that the more efficient the heating system is, the lower the variable natural gas intensity
will be. Although the R2 value is low, Figure 36 does show the expected relationship.
Additionally, the gradient is statistically insignificant, even when the confidence level is
decreased to 80%.
Figure 36: Variable natural gas intensity versus boiler efficiency
y = 14x + 54
R² = 0.46
0
20
40
60
80
100
120
140
160
0 1 2 3 4 5 6
To
tal
Ele
ctri
city
In
ten
sity
(e
kW
h/m
2)
Glazing U-Value (W/m2K)
The Influence of Glazing U-Value on
Electricity Intensity in Air Conditioned Buildings
y = -210x + 320
R² = 0.12
0
50
100
150
200
250
300
350
60% 65% 70% 75% 80% 85% 90% 95%
Va
ria
ble
Na
tura
l G
as
Inte
nsi
ty
(ek
Wh
/m2)
Boiler Efficiency
The Influence of Boiler Efficiency on
Variable Natural Gas Intensity
70
The relationship between variable natural gas intensity and boiler efficiency as shown in Figure
36 is not as strong as expected because the boiler efficiencies provided in the audit reports may
not reflect the actual efficiency of the heating system. Therefore, while the approximate
relationship is correct, it could be stronger if it were based on more accurate data. The provided
efficiencies are either rated or estimated efficiencies. The rated efficiency is the efficiency of the
boiler when it was new, but this efficiency declines as the boiler ages. The rate of decline
depends on maintenance practices, the boiler use patterns, the type of boiler, and the boiler and
pipe configuration. The only way to determine the actual efficiency of a boiler in service is to
run a diagnostic test of natural gas input versus heat output. This was not part of the energy audit
for any of the buildings in this data set.
As a proxy for actual boiler efficiency, the boiler age was estimated based on information
provided about building renovations and boiler replacements in the audit reports. Figure 37 also
shows the expected relationship that the variable natural gas intensity increases as the boiler age
increases. While the R2 value was lower, the gradient was shown to be statically significant at an
80% confidence level. Although the correlation was not as strong as desired, its statistical
significance is an indication that boiler age contributes to boiler efficiency and influences
variable natural gas intensity.
Figure 37: Variable natural gas intensity versus boiler age
y = 1.0x + 150
R² = 0.065
0
50
100
150
200
250
300
350
0 10 20 30 40 50Va
ria
ble
Na
tura
l G
as
Inte
nsi
ty
(ek
Wh
/m2)
Boiler Age (years)
The Influence of Boiler Age on
Variable Natural Gas Intensity
71
The cooling system information that was most readily available from the energy audit reports
was the cooling capacity. Not surprisingly, electricity intensity shows a strong correlation with
cooling capacity for the 11 buildings with air conditioning as seen in Figure 38. The gradient
was found to be statistically significant at a 95% confidence level.
Figure 38: Electrical intensity versus cooling capacity
Figure 38 indicates that the buildings that use air conditioning were appropriately identified in
the building selection process outlined in Section 7.3 and that the cooling capacities provided in
the audit reports are reasonably consistent.
7.4.3 Number of occupants
While the number of bedrooms in each suite was sometimes specified, and the number of suites
in the building was always specified, an estimate of the number of occupants in each building
was found in only two of the 40 audit reports. Since the number of occupants is the most
important variable relating to the base natural gas loads, the number of occupants was estimated
from census data. The average number of people per household (for all dwelling types) for each
building’s census neighbourhood was obtained. If the building had suites that included three
bedrooms or more, 0.5 people per household were added on to the census average. If the
building was a seniors’ home, 0.5 people per household were subtracted from the census
average. The number of people per household was then multiplied by the number of suites in the
building to estimate of the total number of occupants.
y = 0.10x + 71
R² = 0.84
0
20
40
60
80
100
120
140
160
0 100 200 300 400 500 600
To
tal
Ele
ctri
city
In
ten
sity
(e
kW
h/m
2)
Cooling Capacity (tonne of cooling)
The Influence of Cooling Capacity on
Total Electricity Intensity
72
Figure 39 shows a strong correlation between the base natural gas consumption and the estimated
number of occupants. Additionally, the gradient was found to be statistically significant at a 95%
confidence level. This figure supports the conclusion that domestic hot water energy is a
function of the number of occupants.
Figure 39: Base annual natural gas consumption versus estimated number of occupants
7.4.4 Base electricity intensity
There are many components that contribute to base electricity consumption and these
components vary widely from one building to another and cannot be reflected in one variable.
The hypothesis stated in Section 7.3 that base electricity intensity could be related to building
age is addressed in Figure 40. The R2 value in this correlation is very low, however; the gradient
is statically significant when tested at a 90% confidence level. Therefore, building vintage may
be an appropriate proxy, but it is not an accurate one. It is interesting to note the upward trend
indicating that new buildings tend to have a slightly higher base electrical intensity than older
buildings.
y = 2600x - 200,000
R² = 0.79
0
500,000
1,000,000
1,500,000
2,000,000
2,500,000
3,000,000
3,500,000
0 200 400 600 800 1000Ba
se N
atu
ral
Ga
s C
on
sum
pti
on
(e
kW
h)
Estimated Number of Occupants
The Influence of Estimated Number of Occupants
on Base Natural Gas consumption
Outliers
73
Figure 40: Base electricity intensity versus building vintage
The correlations that were examined in the Meta-Analysis have been reproduced for the Refined
Data Set. These graphs are provided in Appendix G since the correlations did not lead to any
new conclusions.
7.5 Multi-variable regression analysis
The correlation analysis showed that there are minor correlations between variable natural gas
intensity and glazing properties as well as boiler efficiency. The purpose of the multi-variable
regression analysis was to determine whether these correlations can be improved when more
than one explanatory variable is considered at a time.
7.5.1 Methodology
The multi-variable regression analysis was completed in Microsoft Excel using the regression
function from the Analysis ToolPak Add-In. A stepwise forward-selection approach to
maximize the adjusted R2
value was used for each regression analysis. An additional approach,
based on the logic of which variables should govern each energy consumption component, was
used to select the order in which variables were added for some of the regression analyses.
The adjusted R2 value derived from the multi-variable regression analyses is the same as the R
2
value in the single variable regression analysis except that a correction has been made to account
for the number of variables involved. As variables are added, more of the variation in the data
y = 0.64x - 1,200
R² = 0.073
0
20
40
60
80
100
120
140
160
180
1950 1960 1970 1980 1990 2000 2010Ba
se E
lect
rica
l In
ten
sity
(k
Wh
/m2)
Year of Consturction
The Influence of Building Vintage on Base
Electricity Intensity
74
should automatically be explained; therefore, to account for the advantage of having additional
variables, a reduction factor is applied to the R2 value. This means that the adjusted R
2 value is
equal to the R2 value in a single-variable regression, but is less than the R
2 value when more than
one variable is involved.
Each stepwise forward-selection approach to multi-variable regression analysis results in a linear
equation relating the selected variables to coefficients as follows:
y=c1xmax1+c2xmax2+c3xmax3+… cnxmaxn (eqn. 3)
where:
y = The component of energy use or energy intensity being examined
c = The coefficient resulting from the multi-variable regression analysis
x = The variable selected to maximize the R2 value
The multi-variable regression analysis involves the following steps:
1. All of the variables to be considered in the analysis are chosen. These variables are
called x1,x2, x3,…,xn and each analysis includes n variables. The variable “y” is the
component of energy consumption for which the regression is being completed.
2. A single-variable linear regression of y versus xi is completed for all n variables. The
variable that yields the highest adjusted R2 value for in the single-variable linear
regression is designated xmax1.
3. A double-variable linear regression of y versus xmax1 and all remaining xi is completed for
all n-1 variables. The variable that yields the highest adjusted R2 value in the double
variable linear regression is designated xmax2. If the new maximum adjusted R2 value is
smaller than the maximum R2
value in the previous step, then the regression is complete
and xmax1 is the only variable involved in the final regression results. If the new
maximum adjusted R2 value is larger than the maximum R
2 value in the previous step,
then a triple variable regression must be completed.
4. Variables continue to be added one at a time until the R2 value is maximized.
75
5. The final result is a linear equation relating the selected variables (xmax1, xmax2,…,etc.) to
y using coefficients.
7.5.2 Summary of all analyses
The variables considered in the multi-variable regression analysis were selected based on two
factors: whether data is available for most of the buildings in the Refined Data Set; and whether
the variable was expected to affect energy consumption. It should be noted that the outliers
identified in the single variable analysis provided above were not removed for the multi-variable
analysis. Table 10 provides a summary of all of the regression analyses undertaken, the
variables that were selected and the final adjusted R2 value. More detailed results of each of the
regression analyses are available in Appendix H.
76
Table 10: Summary of forward-selection and logical regression analyses undertaken
Component of
energy
consumption
Stepwise forward-selection Logical approach
Variables selected Adjusted
R2 value
Variables selected Adjusted
R2 value
Ann
ua
l en
erg
y c
onsum
ptio
n
Total annual
energy
consumption
Boiler efficiency
MAU ventilation capacity
Gross floor area
Balconies and floor slabs
0.85
Annual
variable
natural gas
consumption
Wall R-value
Gross floor area
MAU ventilation capacity
0.81
Annual base
natural gas
consumption
DHW boiler efficiency
Gross Floor Area
0.17
Annual
variable
electricity
consumption
No variables selected All R2
values
were less
than 0.1
Annual base
electricity
consumption
MAU ventilation capacity
Gross floor area
Cooling capacity
0.67
Ann
ua
l E
nerg
y In
tensity Total annual
energy
intensity
Fenestration ratio
MAU ventilation capacity
Number of suites
Year built
Wall R-value
Number of Floors
0.57 Glazing U-value 0.10
Annual
variable
natural gas
intensity
Glazing U-value
Boiler capacity
Number of suites
0.19 Glazing U-value
Boiler capacity
0.18
7.5.3 Discussion of selected results
Overall, the highest adjusted R2 value was achieved for the analysis on total annual energy use.
Once the energy use component was normalized by gross floor area to variable natural gas
intensity, the adjusted R2 value was greatly reduced. These are the same results as found in the
correlation analysis: obvious correlations were possible with total annual energy use; however,
the variation became more difficult to explain once energy was normalized to gross floor area.
Even when many variables were considered at once, the relationships between these variables
77
and variable natural gas intensity were not consistent across all buildings. This is likely because
of the insufficient quality of the data and the inherent variability of energy use in buildings.
The variables that should govern each component of energy use were identified and explained in
Section 4.3. The findings of the multi-variable regression analysis indicate that although these
variables may govern, they do not govern in equal proportions for all buildings. One building’s
energy use may be influenced far more by air leakage, while another may be more significantly
affected by an inefficient boiler.
The two results discussed below were chosen to compare the systematic forward-selection
method with the logical method examining the variable natural gas intensity, since is the
foremost component of energy use discussed in this study.
7.5.3.1 Variable natural gas intensity: systematic forward-selection
The variables considered include: the number of floors, the number of suites, the building
vintage, heating boiler capacity, heating boiler efficiency, make-up air unit (MAU) ventilation
capacity, the presence of balconies and through-wall floor slabs, wall R-value, glazing U-value,
fenestration ratio, boiler age. Although not all of these variables were expected to govern
variable natural gas intensity, any variable that was thought to have a possible affect was
included.
Table 11 shows a summary of the results from the analysis. As expected, the variable natural gas
intensity is related to heat loss in the building through the glazing U-value and is related to heat
generation in the building through the boiler capacity. Additionally, the number of suites was
negatively correlated the variable natural gas use.
The relative weighting was calculated as follows:
"#$%&'$#&()$*+#&%,&-($×/*$++,0,$12
∑�"#$%&'$#&()$*+#&%,&-($×/*$++,0,$12 +*%&((#&%,&-($4 (eqn. 4)
Table 11: Variables in order of selection for variable natural gas intensity systematic forward-selection
Order selected Variable Coefficients Relative
weighting
Adjusted R2 at time
variable was added
78
Intercept 126
1 Glazing U-value 68.5 51% 0.099
2 Boiler capacity 1.61 18% 0.181
3 # of suites -0.16 31% 0.193
7.5.3.2 Variable natural gas intensity: logical method
Based on a logical approach, the variables considered in the systematic forward-selection were
reduced to only include: glazing U-value, fenestration ratio, boiler efficiency, boiler capacity,
and boiler age.
Table 12: Variables in order of selection for variable natural gas intensity logical method
Order selected Variable Coefficients Relative
weighting
Adjusted R2 at time
variable was added
Intercept 102
1 Glazing U 67.5 81% 0.099
2 Boiler capacity 1.09 19% 0.181
79
Chapter 8 Discussion
Within the correlations analysis of the Refined Data Set, a number of buildings were identified as
“anomalies” either because they were outliers in the correlations charts or because they had
abnormally high or low components of energy use. In the sections that follow, the buildings
identified as anomalies based on total energy intensity will be discussed first, followed by the
outliers in variable natural gas intensity and then in base natural gas intensity. Finally, buildings
with different heating systems are highlighted as well as buildings with extra facilities. The
buildings will be referred to by their energy intensity ranking number as shown in Figure 26,
where Building 1 has the lowest energy intensity and Building 40 has the highest energy
intensity.
Figure 26 (reproduced from page 58): Gas and electricity components of total annual energy per gross floor area
8.1 Energy intensity
The nine buildings with the lowest energy intensity were all derived from the HiSTAR database,
which was also used in the Meta-Analysis. The energy intensity values for these nine buildings
range from 90ekWh/m2 up to 210 ekWh/m
2. Although these energy intensities are physically
possible in high performance, low energy buildings, the building envelope and mechanical
characteristics of these buildings would suggest that these are not particularly high performance
0
100
200
300
400
500
600
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39
En
erg
y I
nte
nsi
ty (
ek
Wh
/m2)
Building Number
Total Annual Energy Intensity
Variable Natural Gas Base Natural Gas Variable Electrical Base Electrical
80
buildings. Therefore, Buildings 1 through 9 were automatically accepted as outliers if they were
significantly below the expected energy use in a correlation
The building with the lowest energy intensity, Building 1, is a six storey building built in the
1980s, and has an award-winning green roof. Although the green roof can account for very
minor reductions in heat loss, other factors that were not specified in the energy audit are likely
to be responsible for the low energy intensity. The presence of a green roof may be an indicator
that other aspects of the building have been managed in an energy-conscious manner that could
contribute to the low energy intensity.
The two buildings with the highest energy intensities, Building 39 and 40, were identified prior
to this study as being energy inefficient. Both buildings have significant upgrades planned.
Therefore, these buildings are both anomalies since their energy intensity is significantly higher
than the average; however, this is likely not due to an error in the data. Building 39 and 40 were
also automatically accepted as outliers if they were significantly above the expected energy use
in a correlation.
Overall, there was not a particular factor that could account for the presence of a large group of
anomalies. The anomalies did not arise because a variable was neglected or was not available
for the correlation analysis. The anomalies are generally a result of a special circumstance that
applies to one or two buildings in the data set
8.2 Variable natural gas intensity
Buildings involved in variable natural gas intensity correlations that were designated as outliers
were identified in Figure 31 and Figure 32, which show the correlation between variable natural
gas intensity and the fenestration ratio for double glazed and single glazed windows. These
figures have been reproduced below for convenience.
In Figure 31, there were four outliers in the data. The two outliers above the line of best fit were
Buildings 34 and 35. As both of these were constructed in the 1960s, they are more likely to
have single-glazed windows rather than double-glazed windows. Upon examination of
photographs of the buildings, it was not evident that any window replacement has taken place;
therefore, these two buildings may have been incorrectly assigned to the double glazed category.
81
Buildings 6 and 7 were both below the best fit line, but they have already been identified as
anomalous based on their total energy intensity.
Figure 31 (reproduced from page 66): Variable natural gas intensity versus fenestration ratio for double glazed windows
In Figure 32, a cluster of five buildings with lower than expected variable natural gas intensities
were considered outliers. Three of these buildings were Buildings 5, 8, and 9 and were rejected
due to their low total energy intensity, as discussed previously. The other two buildings were
Buildings 18 and 20. Building 18 had lower variable natural gas energy intensity because it is
one of the few buildings with a high efficiency boiler and a new make-up air unit system.
Building 20 is a subsidized housing project. It does not have a particularly efficient heating
system. However the audit report does acknowledge that there is a very high occupancy density
with six to eight people living in a two bedroom apartment; therefore, internal gains may account
for the relatively low variable natural gas use relative to the fenestration ratio. The hypothesis
that Building 20 has a very high occupancy density is consistent with its higher than average
base natural gas consumption, both on an area basis and on a per suite basis.
y = 150x + 84
R² = 0.69
0
50
100
150
200
250
300
350
0% 20% 40% 60% 80%
Va
ria
ble
Na
tura
l G
as
Inte
nsi
ty
(ek
Wh
/m2)
Fenestration Ratio
The Influence of Fenestration Ratio on
Variable Natural Gas Intensity - Double Glazed
Outliers
82
Figure 32 (reproduced from page 66): Variable natural gas intensity versus fenestration ratio for single glazed windows
Figure 34 (see page 68) will not be re-introduced in this section because the anomalies in Figure
34 are the same buildings as those identified and discussed for Figure 31 and Figure 32.
8.3 Base natural gas intensity
In Figure 39, Buildings 2, 4, and 6 are identified as being below the line of best fit and Buildings
27 and 40 are above the line of best fit. Buildings 2, 4, 6 and 40 were already discussed as being
anomalies because of their high or low energy intensities. Building 27 includes an indoor pool
as well as two laundry rooms with gas-fired dryers, which could explain the higher outlying
natural gas base loads.
y = 250x + 120
R² = 0.41
0
50
100
150
200
250
300
350
0% 20% 40% 60% 80%
Va
ria
ble
Na
tura
l G
as
Inte
nsi
ty
(ek
Wh
/m2)
Fenestration Ratio
The Influence of Fenestration Ratio on
Variable Natural Gas Intensity - Single Glazed
Outliers
83
Figure 39 (reproduced from page 72): Base annual natural gas consumption versus estimated number of occupants
The MURBs designated as seniors’ homes tend to exhibit lower base natural gas use and most of
the MURBs designated as subsidized rental housing were above the line of best fit as above in
Figure 39. The reason for this trend is probably because the estimate of the number of occupants
in the building was not accurate. The number of occupants in each apartment is probably higher
than the census average in subsidized housing and below the census average in senior housing.
An attempt to account for the lower occupancy in seniors’ housing was made by subtracting 0.5
from the average number of people per household leaving about 1.8 people per household for
most neighborhoods where the seniors’ housing was located. In actual fact, the average number
of people per household is probably just above 1. No adjustment to the census average was
made for MURBs providing subsidized housing to account for their higher than average
occupant density.
8.4 Alternative heating systems
Buildings 10 and 11 had below average energy intensity and buildings 37 and 38 had above
average energy intensities. This may have been because these buildings had a unique aspect to
their heating system compared with the other buildings in the data set.
Building 10 has a radiant heating system consisting of electrical resistance coils in the floor
slabs. The radiant heating system is more efficient than other systems because radiant heating
y = 2600x - 200,000
R² = 0.79
0
500,000
1,000,000
1,500,000
2,000,000
2,500,000
3,000,000
3,500,000
0 200 400 600 800 1000Ba
se N
atu
ral
Ga
s C
on
sum
pti
on
(e
kW
h)
Estimated Number of Occupants
The Influence of Estimated Number of Occupants
on Base Natural Gas consumption
Outliers
84
systems can heat the room to a lower temperature while still allowing the occupants feel
comfortable. This is because heat delivered through radiation feels warmer than heat delivered
through convection. The building has also had new windows installed, which is an additional
factor that accounts for the building’s relatively low energy intensity since the new widows
reduce the conductive heat losses and losses through air leakage .
Building 11 does not have heating provided by a central boiler. Instead, each suite has a separate
boiler that provides domestic hot water and heating. This type of configuration can lead to
reduced energy use for a few reasons. First, less heat is wasted as it is being distributed from the
boiler around the building. Second, it is likely that occupants are billed individually for their hot
water and heating so they are more conscious of wasting hot water and have a stronger
motivation to set the thermostat to a lower temperature. Finally, the smaller individual boilers
may manage demand more efficiently than central boiler that has to keep large quantities of
water heated even if they are not being used.
In Buildings 37 and 38, the domestic hot water is generated by the heating boilers. Generally,
this configuration should be more efficient. However, in the case of Buildings 37 and 38, there
are two large heating boilers in each building which both have a much greater capacity than a
domestic hot water boiler would. Even with just one boiler running, the heating output could be
much greater than required and therefore, the heating system may be operating inefficiently and
using extra energy.
8.5 Additional facilities
Building 38 has the highest per suite energy use as shown in Figure 27(see page 59). This is
likely because the building contains a child care centre that is not billed separately. The
attributed suite size appears to be abnormally large because a significant portion of the building
is taken up by the child care centre. In combination with the inefficient domestic hot water
system, the child care centre may also contribute to the high energy intensity of Building 38.
The swimming pools included in the building facilities probably contribute to the above average
energy intensity in Buildings 29 and 39. Buildings 28, 29, 35 and 36 all have heated
underground parking garages that could contribute to their above average energy intensity. Since
85
the parking garage has not been included in the gross floor area for each of the buildings, any
energy consumed by heating, ventilation or lighting of the parking garage is added directly to the
building energy use without any differentiation that the energy is actually being consumed
outside the gross floor area.
86
Chapter 9 Conclusions and Further Research
The Refined Data Set was an improvement on the data set used in the Meta-Analysis: it not only
contained more complete energy consumption information and detailed building characteristics
for all buildings, it also more closely reflected the population characteristics of Toronto mid and
high-rise MURBs in terms of the distribution of building ages, heights sampled and the split in
the electricity to natural gas use.
Although the Refined Data Set was an improvement, there were still limitations in the building
characteristics data. Firstly, the data was collected by many different parties whose collection
practices varied. This introduced inconsistencies in the data, which are reflected by some of the
weaker than expected correlations, such as correlations with the fenestration ratio and the boiler
efficiency. Additionally, no information was available for a few of the variables that were
identified as having a potentially significant impact on energy use such as the glazing type and
the actual number of occupants.
9.1 Variables affecting energy consumption
In Section 7.3, the variables expected to affect the components of energy consumption were
predicted. Some of these hypotheses were confirmed by the correlation analysis and the multi-
variable regression results and some were negated. A summary of these conclusions is provided
in Table 13.
Table 13: Conclusions and data collection needs related to the variables examined
Variable Conclusion Data collection needs
Fenestration
ratio
The fenestration ratio was shown to affect energy
use related to heating and cooling as predicted.
The correlation was shown to be stronger in
buildings with double-glazed windows than in
buildings with single-glazed windows leading to a
hypothesis that window air leakage may also be a
significant contributor.
Air leakage Air leakage is predicted to have an impact on
energy use related to heating and cooling,
especially in buildings with single-glazing. This
suggestion could not be tested since no data on
window air tightness were available.
Information on window type, age,
condition (including sealant), and
ratio of operable to fixed windows.
87
Variable Conclusion Data collection needs
Fenestration
ratio
The fenestration ratio was shown to affect energy
use related to heating and cooling as predicted.
The correlation was shown to be stronger in
buildings with double-glazed windows than in
buildings with single-glazed windows leading to a
hypothesis that window air leakage may also be a
significant contributor.
Air leakage Air leakage is predicted to have an impact on
energy use related to heating and cooling,
especially in buildings with single-glazing. This
suggestion could not be tested since no data on
window air tightness were available.
Information on window type, age,
condition (including sealant), and
ratio of operable to fixed windows.
Glazing U-
value
The considerable variation in the data correlating
the glazing U-value with heating and cooling
loads suggest that glazing U-value affects
heating and cooling loads but may not be a
governing variable.
Boiler
efficiency
The expected relationship between boiler
efficiency and variable natural gas intensity was
found. However, the R2 value was lower than
expected. This was because the data on boiler
efficiency was not comprehensive enough to
accurately project actual boiler plant efficiencies.
Information that will indicate the
boiler’s actual efficiency, not the
rated efficiency, should be
collected. This information includes
the age, maintenance practices,
boiler use patterns, the type of
boiler, and the boiler and pipe
configuration.
Cooling
capacity
The relationship between cooling capacity and
electrical intensity in air conditioned buildings
provided the strongest relationship in the
analysis.
Number of
occupants
The number of occupants was shown to be the
governing variable related to base natural gas
consumption; however, this result was based on
an estimate of the number of occupants rather
than the actual number, which was not available.
The multi-variable regression analysis also
suggested that the number of occupants may
influence variable natural gas intensity since the
number of occupants can contribute to internal
heat gain.
The number of occupants in each
building is an important variable for
which data should be collected.
Alternatively, census data at a more
detailed level could be used to
determine the number of occupants
in each building.
Common
areas and
special
facilities
The analysis of anomalies revealed that although
there was not one particular factor that could
explain a large group of the anomalies,
information on the special facilities included in the
buildings aided in the explanation of a number of
the anomalies.
Quantify the floor area dedicated to
common area facilities and note
high energy use amenities like
pools.
88
Based on the variables found to be most influential, it is clear that window characteristics and
heating system type are the two main factors that should be considered when determining
building typologies.
9.2 Future work
The Meta-Analysis and the Refined Data Set were the first two phases of the research
commissioned by TAF. The final phase includes creating a database, generating suggestions for
typology-specific building upgrades, and developing energy models of four buildings to assess
the effect of certain upgrades.
The database will contain the 40 buildings from the Refined Data Set and will also include the
functionality to add new buildings. The analysis provided by the database will follow the same
method for weather normalization, determination of base and variable loads, and groupings for
the type of building heating system used for the Refined Data Set. Additionally, the correlations
found to be strong will be built into the database.
The suggestions for typology-specific building upgrades will be based on the variables found to
be most strongly linked to energy consumption in the correlation analysis such as the window
characteristics and the heating system efficiencies. Additionally, the research included in the
Meta-Analysis and the Refined Data Set on the distribution of MURB heights and ages in the
Toronto population will allow large groups of buildings to be targeted.
Beyond the research commissioned by TAF, this research is relevant to current planning and
research at the Office of Energy Efficiency. The main challenge of this thesis and of many other
consumer-side energy studies of MURBs was gaining access to accurate energy consumption
information for a large enough sample of buildings. Many organizations collect energy
consumption information for MURBs and maintain private databases for their own purposes;
however, most are unwilling to share this information for privacy reasons or due to industry
competition.
In the past, there has been more emphasis on improving the energy efficiency of commercial
sector buildings through government incentives, policies, and planning.76
Single detached and
singled attached homes as well as schools have also been common targets for programs to
89
improve energy efficiency. More recently however, the need for benchmarking energy
consumption in MURBs has moved to the forefront. The United States Environmental
Protection Agency uses the ENERGY STAR programme to benchmark and rate buildings.
Natural Resources Canada’s Office of Energy Efficiency has recommended that a benchmarking
tool based on ENERGY STAR be launched in Canada.77
When such an energy benchmarking tool is launched for MURBs, lessons learned from the
Meta-Analysis and the Refined Data Set studies should be taken into account. The following is a
list of considerations based on the results of this thesis:
1. Energy consumption data must be weather normalized. This is necessary in order to
compare buildings from across the country and use energy consumption data from
different years. Additionally, energy consumption data must be provided on a monthly
basis so that it can be properly weather normalized.
2. Data must be entered on a whole building basis rather than on a household basis.
Although data can be entered on a household basis for single detached homes, this
approach is not appropriate for MURBs. Individual suite owners usually only have
access to information that will yield the energy intensity of their suite and not the entire
building energy intensity.
3. Buildings must be categorized based on the type of heating system. Electrically heated
buildings should not be compared directly to natural gas heated buildings.
4. In addition to the monthly energy consumption data, essential information on building
characteristics that must be collected includes: gross floor area, number of storeys,
number of suites, number of occupants, date of construction, and building location. The
gross floor area should be clearly defined.
5. Important information on building characteristics that should be collected includes:
fenestration ratio, glazing characteristics (see Table 13), boiler efficiencies, air
conditioning cooling capacity, percent of gross floor area allocated to common areas, and
type of ownership (including responsibility for paying utility bills).
90
6. Additional information on building characteristics that could be collected includes: a
building photograph, heating boiler capacity, MAU ventilation capacities, wall R-value,
presence of balconies and through-wall floor slabs, the type of amenities and common
areas.
Understanding actual baseline energy intensities of MURBs is the first step to improving the
energy efficiency of these buildings. The Meta-Analysis and the Refined Data Set represent
the start of the benchmarking process required for Toronto MURBs. However, it is essential
that this benchmarking process take place on a much larger scale to include MURBs in cities
across Canada. Benchmarking data can then be combined with improvements to building
codes, and rehabilitation of existing buildings to result in a building stock with better energy
efficiency.
91
References
[1] 1 Clarissa Binkley, Marianne Touchie, and Kim Pressnial, Meta-Analysis of Energy
Consumption in Multi-Unit Residential Buildings in the Greater Toronto Area (Report
submitted to Toronto Atmospheric Fund, 2011).
[2] 2 Clarissa Binkley, Marianne Touchie, and Kim Pressnail, Energy Consumption Trends of
Multi-Unit Residential Buildings in the City of Toronto (Report submitted to Toronto
Atmospheric Fund, 2012).
[3] 3 City of Toronto: Policy and Research, City Planning Division, and Social Policy
Analysis & Research, Release of 2006 Census Results Marital Status, Families, Households
and Dwelling Characteristics (September 2007) 8
<http://www.toronto.ca/demographics/pdf/2006_familieshouseholds_backgrounder.pdf> at
21 July 2012.
[4] 4 Natural Resources Canada, Survey of Household Energy Use 2007 Detailed Statistical
Report (2010) Office of Energy Efficiency, 168 (Table 11.3), 180 (Table 11.5)
<http://oee.nrcan.gc.ca/Publications/statistics/sheu07/pdf/sheu07.pdf> at 21 July 2012.
[5] 5 Ibid.
[6] 6 Graeme Stewart and Jason Thorne, Tower Neighbourhood Renewal in the Greater
Golden Horseshoe (2010) Centre for Urban Growth and Renewal, 48
<http://www.cugr.ca/tnrggh> at 21 July 2012.
[7] 7 Natural Resources Canada, Comprehensive Energy Use Database – Sources (11 July
2012) Office of Energy Efficiency,
<http://oee.nrcan.gc.ca/corporate/statistics/neud/dpa/comprehensive_tables/sources.cfm?attr=
0> at 7 August 2012.
[8] 8 Douglas Hart, ‘Energy and Water Consumption Load Profiles in Multi-Unit Residential
Buildings’ (2005) Technical Series 05-119, Research Highlights Canadian Mortgage and
Housing Corporation, 2, <http://www.cmhc-schl.gc.ca/odpub/pdf/64940.pdf> at 22 July
2012.
[9] 9 Natural Resources Canada, Comprehensive Energy Use Database – Residential Sector
Ontario Table 38 (16 April 2012) Office of Energy Efficiency,
<http://oee.nrcan.gc.ca/corporate/statistics/neud/dpa/tablestrends2/res_on_38_e_4.cfm?attr=0
> at 9 August 2012.
[10] 10 Ronggui Liu, ‘Energy Consumption and Energy Intensity in Multi-Unit Residential
Buildings (MURBs) in Canada’ (2007) 04, Canadian Building Energy End-Use Data and
Analysis Centre, 6 (Table 3.1.2) <
http://www.cbeedac.com/publications/documents/MURBsrp04.pdf> at 22 July 2012.
92
[11] 11 Natural Resources Canada, Comprehensive Energy Use Database – Residential Sector
Ontario Table 38, above n 9.
[12] 12 RDH Building Engineering Limited, ‘Energy Consumption and Conservation in Mid-
and High-Rise Residential Buildings in British Columbia’ (2012) Canadian Mortgage and
Housing Corporation Research Report, 1, <ftp://ftp.cmhc-schl.gc.ca/chic-
ccdh/Research_Reports-
Rapports_de_recherche/eng_unilingual/CA1%20MH%2012E51_w.pdf> at 22 July 2012.
[13] 13 Natural Resources Canada, Comprehensive Energy Use Database – Residential Sector
British Columbia Table 38 (16 April 2012) Office of Energy Efficiency,
<http://oee.nrcan.gc.ca/corporate/statistics/neud/dpa/tablestrends2/res_bc_38_e_4.cfm?attr=0
> at 9 August 2012.
[14] 14 City of Toronto: Policy and Research, City Planning Division, and Social Policy
Analysis & Research, above n 3.
[15] 15 Emporis, Toronto Buildings, < http://www.emporis.com/> at 10 July 2012.
[16] 16 United Nations Statistics Division, Demographic Yearbook 2005 (2011) Table 8,
<http://unstats.un.org/unsd/demographic/products/dyb/dyb2005.htm> at 10 July 2012.
[17] 17Live Green Toronto, The Power to Live Green: Toronto’s Sustainable Living Plan
(October 2009) 11 (Table 2) < http://www.toronto.ca/livegreen/downloads/2009-
10_report.pdf> at 9 August 2012.
[18] 18Bob Krawczyk, A Database of Buildings in Toronto, Canada (2010) TOBuilt <
http://www.tobuilt.ca/> at 5 May 2012.
[19] 19 Stewart and Thorne, above n 6.
[20] 20 Natural Resources Canada, 2003 Survey of Household Energy Use (SHEU) Detailed
Statistical Report, Office of Energy Efficiency (2006) 121 (Table 8.2) <
http://oee.nrcan.gc.ca/publications/statistics/sheu03 /pdf/sheu03.pdf > at 21 July 2012.
[21] 21 Douglas Hart, above n 8.
[22] 22 Ibid 1.
[23] 23 Ibid 2.
[24] 24 Ibid 3.
[25] 25 Enermodal Engineering Limited, ‘Review of OHC Building Energy and Water Audits’
(2000) Technical Series 00-110, Research Highlights Canadian Mortgage and Housing
Corporation <http://cmhc.ca/odpub/pdf/62587.pdf?lang=en> at 22 July 2012.
93
[26] 26 Ibid 1.
[27] 27 Ibid.
[28] 28 A Hakim Elmahdy, Building Research Note: Annual Consumption Data on Apartment
Buildings (Ottawa, 1982) Division of Building Research, National Research Council of
Canada < http://www.nrc-cnrc.gc.ca/obj/irc/doc/pubs/brn/brn181/brn181.pdf> at 22 July
2012.
[29] 29 Ibid 3, Table 1.
[30] 30 Ibid 6, 7, Table 3.
[31] 31 Ibid 3, Table 1.
[32] 32 Ibid 15, 16, Figure 1 - 4.
[33] 33 Ibid 17, 18, Figure 5 – 7.
[34] 34 Ibid 3, Table 1.
[35] 35 Ibid.
[36] 36 Ronggui Liu, ‘Energy Consumption and Energy Intensity in Multi-Unit Resnidential
Buildings (MURBs) in Canada’ (2007) 04, Canadian Building Energy End-Use Data and
Analysis Centre < http://www.cbeedac.com/publications/documents/MURBsrp04.pdf> at 22
July 2012.
[37] 37 Ibid 2.
[38] 38 Ibid 5, Table 3.1.1.
[39] 39 Ibid 6, Table 3.1.2.
[40] 40 Enermodal Engineering Limited, above n 25, 2.
[41] 41Ibid.
[42] 42 Ronggi Liu, above n 36, 10.
[43] 43 RDH Building Engineering Limited, above n 12.
[44] 44 Ibid 60.
[45] 45 Ibid 60, Figure 3.5.1.
[46] 46 Environment Canada, Canadian Climate Normals 1971-2000 Vancouver Int’l A (2012)
National Climate Data and Information Archive
94
<http://www.climate.weatheroffice.gc.ca/climate_normals/results_e.html?stnID=889&lang=e
&dCode=1&StationName=VANCOUVER&SearchType=Contains&province=ALL&provB
ut=&month1=0&month2=12> at 24 July 2012.
[47] 47 Environment Canada, Canadian Climate Normals 1971-2000 Toronto Lester B.
Pearson Int’l A (2012) National Climate Data and Information Archive
<http://www.climate.weatheroffice.gc.ca/climate_normals/results_e.html?stnID=5097&lang
=e&dCode=1&StationName=TORONTO&SearchType=Contains&province=ALL&provBut
=&month1=0&month2=12> at 24 July 2012.
[48] 48 Environment Canada, Daily Data Report for January 1975(2012) National Climate
Data and Information Archive
<http://www.climate.weatheroffice.gc.ca/climateData/dailydata_e.html?timeframe=2&Prov=
ONT&StationID=5051&dlyRange=1840-03-01|2012-04-
08&Year=1975&Month=1&Day=01> at 24 July 2012.
[49] 49 Ibid.
[50] 50 Statistics Canada, Report on Energy Supply and Demand in Canada (2009) <
http://www.statcan.gc.ca/pub/57-003-x/57-003-x2009000-eng.pdf> at 24 July 2012.
[51] 51 Ibid 124.
[52] 52 Natural Resources Canada, Survey of Household Energy Use 2007 Detailed Statistical
Report, above n 4.
[53] 53 Ibid 194.
[54] 54 Ibid 197.
[55] 55 Ibid 196.
[56] 56 Natural Resources Canada, Energy Use Data Handbook 1990-2009 (2012) Office of
Energy Efficiency, 19 <
http://oee.nrcan.gc.ca/publications/statistics/handbook11/pdf/handbook11.pdf> at 24 July
2012.
[57] 57 Natural Resources Canada, Survey of Household Energy Use 2007 Detailed Statistical
Report, above n 4, 162, Table 11.2.
[58] 58 Natural Resources Canada, Survey of Household Energy Use 2003 Detailed Statistical
Report (2006) Office of Energy Efficiency, 117 (Table 8.1), 121 (Table 8.2) <
http://oee.nrcan.gc.ca/Publications/statistics/sheu03/pdf/sheu03.pdf > at 30 July 2012.
[59] 59 Natural Resources Canada, Energy Use Data Handbook Table 4 (16 April 2012) Office
of Energy Efficiency
95
<http://oee.nrcan.gc.ca/corporate/statistics/neud/dpa/tableshandbook2/res_00_4_e_5.cfm?attr
=0> at 10 August 2012.
[60] 60 Natural Resources Canada, Comprehensive Energy Use Database – Residential Sector
Canada Table 45 (16 April 2012) Office of Energy Efficiency,
<http://oee.nrcan.gc.ca/corporate/statistics/neud/dpa/tablestrends2/res_ca_45_e_4.cfm?attr=0
> at 9 August 2012.
[61] 61 Natural Resources Canada, Comprehensive Energy Use Database – Residential Sector
Ontario Table 38, above n 9.
[62] 62 Natural Resources Canada, Comprehensive Energy Use Database – Residential Sector
British Columbia Table 38 (16 April 2012) Office of Energy Efficiency,
<http://oee.nrcan.gc.ca/corporate/statistics/neud/dpa/tablestrends2/res_bc_38_e_4.cfm?attr=0
> at 9 August 2012.
[63] 63 Enermodal Engineering Limited, above n 25.
[64] 64 Ekaterina Tzekova, Kim D Pressnail, David De Rose, and Kevin Day, Evaluating the
effectiveness of energy-efficient retrofits on multi-unit residential buildings: two case studies.
Winnipeg: 13th
Canadian conference on buildings science and technology, 2011.
[65] 65 Arup, Community energy plan for pilot sites.(2010) Toronto: Mayor’s Tower Renewal,
<http://www.toronto.ca/city_manager/pdf/tr_arup_cep.pdf> at 10 August 2012.
[66] 66 Bob Krawczyk, above n 18.
[67] 67Natural Resources Canada, Energy Use Data Handbook Table 11 (16 April 2012)
Office of Energy Efficiency
<http://oee.nrcan.gc.ca/corporate/statistics/neud/dpa/tableshandbook2/res_00_11_e_5.cfm?at
tr=0> at 10 August 2012.
[68] 68 Neil B Hutcheon and Gustav O P Handegord, Building Science for a Cold Climate
(Institute for Research in Construction, 1995).
[69] 69Natural Resources Canada, Gas Heating Terms and Conversions (2 July 2010) Office
of Energy Efficiency, <http://oee.nrcan.gc.ca/equipment/heating/11280> at 10 August 2012.
[70] 70ENERGY STAR, ENERGY STAR Portfolio Manager Methodology for Accounting for
Weather, United States Environmental Protection Agency, 3,
<http://www.energystar.gov/ia/business/evaluate_performance/Methodology_Weather_2011
0224.pdf> at 19 September 2012.
[71] 71 Natural Resources Canada, Comprehensive Energy Use Database – Residential Sector
Ontario Table 38, above n 9.
[72] 72 Natural Resources Canada, Survey of Household Energy Use 2007 Detailed Statistical
Report, above n 4.
96
[73] 73 J K Latta, Building Research Note: Computed Energy Consumption for New and
Existing High-Rise Residential Buildings Suggested Norms and Potential Reductions
(Ottawa, 1983) Division of Building Research, National Research Council of Canada
<http://www.nrc-cnrc.gc.ca/obj/irc/doc/pubs/brn/brn201/brn201.pdf> at 22 July 2012.
[74] 74 Toronto Atmospheric Fund, Emissions Quantification (2008) Toronto
<http://www.toronto.ca/taf/quant_policy_approach.htm> at 10 August 2012.
[75] 75 Natural Resources Canada, Comprehensive Energy Use Database – Residential Sector
Ontario Table 38, above n 9.
[76] 76National Round Table on the Environment and the Economy, Geared for Change:
Energy Efficinecy in Canada’s Commercial Building Sector (2009) Sustainable
Development Technology Canada, < http://www.sdtc.ca/uploads/documents/en/nrtee-
report.pdf> at August 10, 2012.
[77] 77
Demand Side Management Working Group Sub-Committee on Building and
Housing Energy Labelling, Council of Energy Ministers, Building Energy Benchmarking:
Recommendation and Work Outline for a System for Canada (1 September 2009) Office of
Energy Efficiency,
<http://oee.nrcan.gc.ca/sites/oee.nrcan.gc.ca/files/files/pdf/publications/cem-
cme/benchmark_e.pdf> at 10 August 2012.
97
Appendix A Calculations of Statistics for Toronto Dwellings
The majority of the calculations shown below are based on 2006 and 2007 statistics. No
adjustment has been made to bring the estimates into the current year. According to 2011 census
data, the population of the City of Toronto grew by 4.5% between 2006 and 2011.i Therefore, it
would be reasonable to expect an increase in all of the given statistics. However, no inflation
factor has been applied since the change within each dwelling type cannot be readily estimated.
1. Number of apartment dwellings in Toronto
Statistics Canada defines a dwelling as:
…a separate set of living quarters with a private entrance either from outside the
building or from a common hall, lobby, vestibule or stairway inside the building. The
entrance to the dwelling must be one that can be used without passing through the living
quarters of some other person or group of persons.ii
Based on 2006 census data, the City of Toronto contains a total of 979,440 dwellings: 379,700 of
which are in apartment buildings of five or more storeys, and 162,984 of which are in apartment
buildings of four or fewer storeys.iii
Therefore, the statistics provided on page 1 were calculated
as follows:
% of dwellings in mid and high-rise apartment buildings: ���,���
���,��� = 38.8% → 39%
% of dwellings in low-rise apartment buildings: ���,���
���,��� = 16.6% → 16%
% of dwellings in apartment buildings: 38.8% + 16.6% = 55.4% → 55%
98
2. Greenhouse gas emissions of Toronto MURBs
According to the 2007 Survey of Household Energy Use (SHEU), there are 776,286 Ontario
households in apartment buildings of five or more storeys. In total, these buildings use
18,796,181GJ/year of electricity and 15,247,961GJ/year of natural gas. There are 456,393
Ontario households in apartment buildings of four or fewer storeys that use a total of
12,288,400GJ/year of electricity and 9,929,064GJ/year of natural gas.iv
The emissions factors
assigned to electricity and natural gas use are 0.187kg/kWh and 1.879kg/m3 respectively.
v There
are 162,984 low-rise apartment dwellings and 379,700 mid and high-rise apartment dwellings in
Toronto.vi
Table A1 provides a summary of the estimated greenhouse gas emissions for the
MURB sector in Toronto.
Table A1: Calculated Greenhouse Gas Emissions for Toronto MURBs
Low-rise apartments Mid and high-rise apartments
Electricity Natural gas Electricity Natural gas
Energy use per Ontario
household (GJ)
26.93 21.76 24.21 19.64
Energy use for all Toronto
households (GJ)
4,388,000 3,546,000 9,194,000 7,458,000
Emissions factor (kg/GJ) 51.94 50.67 241.7 50.67
Tonnes of CO2e 228,000 179,700 447,600 377,900
Sub-total 407,600 855,500
Total 1,263,000
3. Number of MURBs in Toronto
The number of MURBs in Toronto has been estimated based on the number of dwellings in the
City of Toronto as given in Appendix A, Section 1. The number of dwellings per building has
been taken as the median number of suites from the 130 MURBs included in this research. The
median number of suites in mid and high-rise MURBs was 181 suites and the median number of
suites in low-rise MURBs was 36 suites.
Estimated number of mid and high-rise MURBs: ���,���
��� = 2,092 → 2,100
Estimated number of low-rise MURBs: ���,���
�� = 4,527 → 4,500
99
4. Cities with the largest number of high-rise buildings
The 20 cities with the largest number of high-rise buildings in the world are listed in Table A2.
The number of high-rise buildings in each city was obtained from an online database and the
population of each city was obtained from United Nations population data for the latest available
year. In the online database, high-rise buildings are defined as having 12 or more storeys, which
is different from the eight or more storeys definition used throughout the rest of this thesis.
Table A2: List of cities with the largest number of high-rise buildings and their population
Rank City Number of high-
rise buildingsvii
Populationviii
1 Hong Kong 7,896 6,977,700
2 New York 6,508 8,363,710
3 Sao Paulo 6,467 10,990,249
4 Singapore 4,764 4,839,400
5 Caracas 3,864 4,196,514
6 Moscow 3,760 10,489,644
7 Seoul 2,963 10,031,719
8 Rio de Janeiro 2,947 6,161,047
9 Tokyo 2,780 8,489,653
10 Toronto 2,505 2,615,060
11 Istanbul 2,439 11,372,613
12 London 2,128 7,556,900
13 St. Petersburg 2,108 4,574,950
14 Buenos Aires 1,868 2,965,403
15 Chicago 1,625 2,853,114
16 Kiev 1,588 2,698,926
17 Mexico City 1,533 8,836,045
18 Mumbai 1,517 11,978,450
19 Santiago 1,498 4,988,252
20 Osaka 1,484 2,628,811
i City of Toronto: Policy and Research, City Planning Division, and Social Policy Analysis & Research,
Backgrounder 2011 Census: Age and Sex Counts (29 May 2012) 1,
<http://www.toronto.ca/demographics/pdf/censusbackgrounder_ageandsex_2011.pdf> at 10 August 2012.
iiStatistics Canada, Private Dwelling Universe Definition (25 July 2008) Definitions Data Sources and Methods <
http://www.statcan.gc.ca/concepts/definitions/priv-eng.htm> at 10 August 2012
iii
City of Toronto: Policy and Research, City Planning Division, and Social Policy Analysis & Research, Release of
2006 Census Results Marital Status, Families, Households and Dwelling Characteristics (September 2007) 8
<http://www.toronto.ca/demographics/pdf/2006_familieshouseholds_backgrounder.pdf> at 21 July 2012.
100
iv Natural Resources Canada, 2007 Survey of Household Energy Use (SHEU-2007) – Data Tables (29 March 2010)
Office of Energy Efficiency, Table 1.1, 11.3, 11.5,
<http://oee.nrcan.gc.ca/corporate/statistics/neud/dpa/data_e/sheu07/tables.cfm?attr=0> at 10 August 2012.
v Toronto Atmospheric Fund, Emissions Quantification (2008) Toronto
<http://www.toronto.ca/taf/quant_policy_approach.htm> at 10 August 2012.
vi City of Toronto: Policy and Research, City Planning Division, and Social Policy Analysis & Research, Release of
2006 Census Results Marital Status, Families, Households and Dwelling Characteristics (September 2007) 8,
<http://www.toronto.ca/demographics/pdf/2006_familieshouseholds_backgrounder.pdf> at 21 July 2012.
vii
Emporis, Toronto Buildings, < http://www.emporis.com/> at 10 July 2012.
viii United Nations Statistics Division, Demographic Yearbook 2005 (2011) Table 8,
<http://unstats.un.org/unsd/demographic/products/dyb/dyb2005.htm> at 10 July 2012.
101
Appendix B Sample Letter of Request
Appendix B contains a copy of the letter that was sent to the 42 building owners during the
Refined Data Set data collection process. Each letter was personalized with the building owner’s
name and address and mailed out to the Green Condo Champions participants. The FAQs sheet
also provided in this appendix was sent later in the data acquisition process.
35 St. George Street
Toronto, ON, M5S 1A4
January 27, 2012
Subject: REQUEST FOR ELECTRICITY DATA
<<CONTACT NAME>>
<<BUILDING ADDRESS>>
Dear <<CONTACT NAME>>,
The Sustainable Building Group at the University of Toronto is working with the Toronto Atmospheric Fund to
develop a better understanding of energy use in multi-unit residential buildings (MURBs).
As part of the Green Condo Champions initiative, members of your board participated in training and engagement
programs focused on cutting energy costs and reducing greenhouse gas emissions and received an audit report
and recommendations for strategies to reduce the energy use in your building. These audit reports, used
anonymously, have been invaluable in establishing the foundations of a comprehensive MURB database that will
help inform policy makers and direct energy retrofit efforts.
To complete the data set, we are requesting access to your historical electricity records in order to match these
data with the existing natural gas data in the audit report. The anonymity of all of your energy performance data
will be strictly maintained. When entered into the energy use database, your building name and/or address will be
replaced with a unique number for the data analysis procedures.
Following completion of the study in September 2012, you will be provided with a ranking of your building within
the rest of the data set to determine your building’s level of energy performance compared to other similar
buildings.
Kindly complete both pages of the attached letter and return it in the self-addressed stamped envelope at your
earliest convenience. If you have any questions or concerns, please do not hesitate to contact me at
[email protected] or (416) 432-1535.
Your prompt attention and cooperation are greatly appreciated!
Kind Regards,
Clarissa Binkley
Sustainable Building Group Researcher
University of Toronto
Toronto MURB Energy Study – FAQs
WHAT is this project?
This is a research project to
establish the baseline energy use
characteristics of Toronto multi-
unit residential buildings
(MURBs). Before proceeding to
the next stage of developing
strategies to improve new and
existing MURBs, it is vital that
current energy consumption and
the key variables that drive
energy performance are well
understood. With more than
2,700 MURBs in the City of
Toronto housing over one third
of the population and producing
over 40% of the city’s residential
sector greenhouse gas emissions,
maintaining these buildings is a
paramount concern.
WHO are the project partners?
The Toronto Atmospheric Fund (TAF) is sponsoring this study
undertaken by researchers at the University of Toronto.
This project is separate from the Green Condo Champions
Initiative and has different objectives.
HOW will the information be used?
Information provided about your building will be added to a
larger pool of data being used to study baseline energy use
characteristics of Toronto MURBs. The building address
associated with the information will be replaced with a
unique identification tag and building addresses and
defining characteristics will be kept strictly confidential.
This study examines the characteristics of the entire pool of
data, rather than focusing on any one particular building.
WHAT are the incentives
to participate?
Following completion of the
study in September 2012,
participants will be provided
with a ranking of your building
within the data set. The
“Toronto MURB Energy
Intensity” graph is a sample
ranking. In addition, you will be
contributing to an important
area of research with wide
spread benefits.
WHAT is required to participate?
A board member or building manager must provide the study partners with permission to access
your historical electricity records. This can be done by completing the letter providing your Toronto
Hydro account number and your permission to release the data. The letter of release has been sent
by paper mail or electronic email for your completion. To obtain a letter of release please contact
Clarissa Binkley at 416-432-1535 or [email protected].
104
Appendix C Individual Building Characteristics
The date of construction, number of floors, gross floor area and the number of suites for each of
the buildings in the examined data sets are provided in the tables below. The Meta-Analysis data
is divided into three tables based on the original source of the data. Table C1 shows the data
from the HiSTAR database, Table C2 provides the data derived from the Green Condo
Champions data set, and Table C3 details the data obtained from the Tower Renewal
Benchmarking Initiative. The characteristics of the buildings included in the Refined Data Set
are provided in Table C4.
Although 20 of the buildings from the Meta-Analysis were also included in the Refined Data Set,
these buildings were assigned new reference letters for the Refined Data Set. Also note that the
reference letter for buildings in the Refined Data Set in Table C4 does not correspond to the
building number assigned to each building based on its ranked energy intensity. Different
identification tags have been used to protect the anonymity of the buildings so that a specific
building cannot be linked to an energy consumption value. However, the reference letters are
consistent with the reference letters in Appendix F, which shows the pictures of each building’s
window configuration.
Table C1: Characteristics of the HiSTAR buildings included in the Meta-Analysis
Reference
Letter
Date of
Construction
Number
of Floors
Number
of Suites
Gross Floor
Area (m2)
H1 unknown 10 100 8,410
H2 unknown 13 125 7,423
H3 unknown 6 67 3,931
H4 1991 9 125 11,186
H5 1989 9 96 8,779
H6 1973 7 74 3,345
H7 1982 6 60 12,465
H8 1969 17 242 17,156
H9 1976 23 336 23,212
H10 1970 26 201 16,486
H11 1971 22 321 22,203
H12 1966 12 171 12,110
H13 1968 17 223 15,520
H14 1969 15 218 15,138
H15 1970 20 400 18,116
105
Reference
Letter
Date of
Construction
Number
of Floors
Number
of Suites
Gross Floor
Area (m2)
H16 1969 24 400 21,726
H17 1986 29 223 32,244
H18 1974 11 45 4,640
H19 1965 12 510 43,051
H20 1992 7 123 7,840
H21 1989 35 210 20,656
H22 1984 17 193 22,583
H23 1985 11 101 13,146
H24 1980 14 101 17,306
H25 1964 15 224 9,418
H26 1992 9 297 21,350
H27 1991 11 212 12,135
H28 1984 13 179 17,621
H29 1977 10 330 18,774
H30 1967 11 199 11,165
H31 1993 8 155 11,715
H32 1992 7 160 13,297
H33 1988 6 106 13,297
H34 1994 5 35 3,638
H35 1980 7 114 10,341
H36 1984 6 44 4,089
H37 1979 7 189 11,122
H38 1975 7 200 25,951
H39 1959 11 298 19,991
H40 1991 5 126 9,233
H41 1978 5 131 9,019
H42 1941 4 57 3,701
H43 1941 4 36 2,521
H44 1941 4 57 3,707
H45 1978 10 186 20,851
H46 1978 18 178 23,120
H47 1990 23 224 31,459
H48 1990 19 184 26,426
H49 1979 20 236 31,266
H50 1991 23 224 35,869
H51 1989 12 134 12,214
H52 1971 14 104 14,229
H53 1971 22 216 36,795
H54 1971 22 216 36,795
H55 1971 22 216 36,795
106
Table C2: Characteristics of the Condo Champs buildings included in the Meta-Analysis
Reference
Letter
Date of
Construction
Number
of Floors
Number
of Suites
Gross Floor
Area (m2)
C1 1991 20 206 36,323
C2 1986 4 16 4,459
C3 2001 12 105 16,687
C4 1998 25 190 16,620
C5 2003 28 319 30,217
C6 2004 16 317 33,337
C7 2001 12 116 19,632
C8 1986 4 14 3,345
C9 1999 24 174 15,134
C10 1990 14 169 19,251
C11 2002 27 339 33,776
C12 1990 15 unknown 25,920
C13 2007 22 158 14,910
C14 1998 16 132 15,236
C15 1988 22 326 39,019
C16 2009 46 420 60,783
C17 1973 37 542 73,393
C18 1984 8 119 20,940
C19 1990 33 434 43,381
C20 1998 18 174 22,817
C21 1977 24 193 21,650
C22 1980 18 100 19,846
C23 2000 12 82 9,094
C24 1984 8 187 32,758
C25 1999 8 46 9,749
C26 1999 8 480 101,733
C27 2006 36 342 70,606
C28 1985 5 74 11,487
C29 1988 9 61 11,074
C30 1985 5 70 11,487
C31 1976 12 117 16,318
C32 1998 10 86 7,246
C33 2005 6 72 7,580
C34 2003 10 381 44,085
C35 2007 8 98 14,307
C36 1999 16 160 15,474
C37 2000 4 19 1,997
C38 1971 21 188 21,554
C39 1971 18 188 30,119
C40 2003 11 143 19,168
C41 1991 20 216 42,271
C42 1973 21 216 20,067
107
Table C3: Characteristics of the Tower Renewal buildings included in the Meta-Analysis
Reference
Letter
Date of
Construction
Number
of Floors
Number
of Suites
Gross Floor
Area (m2)
T1 1978 23 228 28,986
T2 1978 23 227 28,986
T3 1969 17 192 20,700
T4 1968 17 310 22,389
T5 1961 9 116 10,202
T6 1964 unknown 299 25,826
T7 1964 unknown 300 25,826
T8 1964 unknown 300 25,826
T9 1966 unknown 264 32,581
T10 1970 unknown 719 51,110
T11 1969 unknown 626 50,222
Table C4: Characteristics of all buildings in the Refined Data Set
Reference Letter
Date of Construction
Number of Floors
Gross Floor Area (m
2)
Number of Suites
A 1991 9 11,186 125
B 1973 7 3,345 74
C 1966 12 12,110 171
D 1976 23 23,212 336
E 1982 6 12,465 84
F 1968 17 15,520 223
G 1969 15 15,138 218
H 1989 9 8,779 96
I 1981 21 20,656 210
J 1991 23 35,869 225
K 1992 7 13,297 160
L 1985 11 13,146 101
M 1990 23 31,459 224
N 1975 7 25,951 200
O 1990 19 26,426 184
P 1992 13 10,464 124
Q 1995 16 10,997 138
R 1992 8 8,842 92
S 1960 5 6,304 80
T 1992 9 13,877 115
U 1993 5 5,938 68
V 1993 5 6,076 69
W 1993 12 25,623 243
X 1974 13 14,493 152
Y 1974 14 15,608 164
Z 1991 6 8,872 83
AA 1999 16 10,829 160
BB 1977 24 21,650 193
CC 1973 21 20,067 216
DD 2002 27 33,776 339
EE 2003 28 30,217 336
FF 1963 23 28,986 230
GG 1967 17 22,390 216
108
Reference Letter
Date of Construction
Number of Floors
Gross Floor Area (m
2)
Number of Suites
HH 1967 17 23,000 92
II 1976 22 11,378 123
JJ 1967 15 6,008 84
KK 1967 7 7,943 75
LL 1964 6 7,411 74
MM 1973 12 19,036 184
NN 1974 9 13,006 90
109
Appendix D Characteristics of Toronto MURB Population
Information on the number of Toronto mid and high-rise MRUBs in certain height and vintage
categories was obtained by searching the TOBuilt databasei and limiting the query ranges to fit
the categories listed. Table D1 summarized the information obtained from TOBuilt. The total
number of buildings in the categories sums 1,734, which is greater than the total number of
buildings in the vintage categories. This is because information on building height is available
for more buildings than the date of construction.
Table D1: Number of mid and high-rise MURBs in height and vintage categories based on TOBuilt data
Height Categories Vintage Categories
Number of
Floors
Number of
Buildings
Time Period Number of
Buildings
5 -8 131 Before 1946 71
9 – 12 255 1946 – 1960 62
13 – 16 497 1961 – 1970 428
17 – 20 360 1971 – 1980 475
21 – 24 216 1981 – 1990 204
25 – 28 108 1991 – 2000 141
29 – 32 73 2001 – 2010 316
33 – 36 38 TOTAL 1,697
37 – 40 24
41 – 44 9
45 – 48 10
49 – 52 13
TOTAL 1,734
The TObuilt data provided in Table D1 was adjusted to more accurately reflect the actual
population of Toronto MURBs. The adjusted data is provided in Table D2. Within the
residential sector, the database is focused on high-rises, and is estimated by the authors of
TOBuilt to be 95% accurate. There are entries for a total of 1,609 high-rise MURBs and 131
mid-rise MURBs. The total number of mid and high-rise MURBs in Toronto is 2,100; however
the split between mid-rise MURBs and high-rise MURBs in unknown. Since the focus of
TObuilt is on high-rise buildings, this sector was considered 95% complete, while the mid-rise
sector was assumed to be more incomplete. Therefore, after the number of high-rise buildings
was increased, the remaining buildings were added to the mid-rise category to bring the total
110
number of mid-rise and high-rise MURBs to 2,100. The calculation for this adjustment is shown
below:
�,��������� �������������
��%=1,693 high-rise MURBs
2,100midandhigh-riseMURBs–1,693high-riseMURBs=406mid-riseMRUBs
Therefore, the number of buildings in each height category was increased proportionally based
on the assumption that Toronto contains 406 mid-rise MURBs and 1,693 high-rise MURBs. It
was assumed that the vintage categories were represented equally the number of buildings were
increased proportionally without making special adjustments to certain categories.
Table D2: Estimated number of Toronto mid and high-rise MURBs based on adjusted TOBuilt data
Height Categories Vintage Categories
Number of
Floors
Number of
Buildings
% of
Population
Time Period Number of
Buildings
% of
Population
5 -8 406 19% Before 1946 88 4.2%
9 – 12 270 13% 1946 – 1960 77 3.7%
13 – 16 525 25% 1961 – 1970 530 25%
17 – 20 380 18% 1971 – 1980 588 28%
21 – 24 228 11% 1981 – 1990 252 12%
25 – 28 114 5.4% 1991 – 2000 174 8.3%
29 – 32 77 3.7% 2001 – 2010 391 19%
33 – 36 40 1.9% TOTAL 2,100
37 – 40 25 1.2%
41 – 44 10 0.5%
45 – 48 11 0.5%
49 – 52 14 0.7%
TOTAL 2,100
i Bob Krawczyk, A Database of Buildings in Toronto, Canada (2010) TOBuilt < http://www.tobuilt.ca/> at 5 May
2012.
111
Appendix E Heat Equations Related to Energy Flows
Conduction heat loss is governed by Equation 1:
q = UA∆T (1)
Where:
q= rate of heat flow (W)
U = heat loss coefficient of material (W/m2K)
A = Area of material transverse to heat flow (m2)
∆T = Temperature differential across the material (°C)
Between opaque and transparent building envelope components, glazing is typically the weakest
part of the building envelope because of a high heat loss coefficient relative to walls sections.
The governing equation also indicates the important of glazing area in determining heat loss.
Air leakage which affects heat transferred across the building envelope, is governed by Equation
2:
Q =CA[(2/ρ)(∆P)]1/2
(2)
Where:
C = orifice coefficient
A = opening area (m2)
ρ = density of air (kg/m3)
∆P = pressure differential across envelope (Pa)
Though the pressure differential across the building envelope is influenced by mechanical
equipment, the bulk of this differential is generated by stack pressure due to building height and
112
wind pressure. Therefore air leakage is influenced by the building construction in terms of the
type and area of openings. Again, the glazed components of the envelope will dominate this heat
loss component because they are often the air-leakiest parts of the buildings.
Solar radiation heat gains through glazing are governed by equation 3:
qt = SHGC x SHGF + U(to – ti) (3)
Where:
SHGC = solar heat gain coefficient
SHGF = solar heat gain factor (W/m2)
U = glazing U-value (W/m2K)
to = outdoor temperature (°C)
ti = indoor temperature (°C)
Though indoor and outdoor temperatures contribute to solar radiation heat gains, the
characteristics of the glazing that govern are the SHCG and the U-value.
113
Appendix F Refined Data Set Building Pictures
Table F1 shows a small portion of the photograph of each building used to estimate the
fenestration ratio for the MURBs in the Refined Data Set. It is important to note that the
photograph of the entire building was used to revise the estimates of fenestration ratio, not the
close-ups shown here.
Table F1: Snapshot of buildings and associated fenestration ratio
Building Original
Fenestration
Ratio
Revised
Fenestration
Ratio
Snapshot
A 42% 60%
B 22% 32%
C 53% 53%
D 53% 53%
114
Building Original
Fenestration
Ratio
Revised
Fenestration
Ratio
Snapshot
E 45% 45%
F 52% 52%
G 53% 53%
H 30% 35%
I 27% 35%
J 69% 75%
115
Building Original
Fenestration
Ratio
Revised
Fenestration
Ratio
Snapshot
K 24% 35%
L 37% 35%
M 77% 75%
N 20% 20% No Picture available
O 66% 75%
P 40% 45%
116
Building Original
Fenestration
Ratio
Revised
Fenestration
Ratio
Snapshot
Q 45% 45%
R 30% 40%
S 25% 35%
T 40% 50%
U 25% 25%
V 25% 25%
117
Building Original
Fenestration
Ratio
Revised
Fenestration
Ratio
Snapshot
W 45% 70%
X 25% 30%
Y 25% 30%
Z 30% 60%
AA 50% 50%
BB 65% 65%
118
Building Original
Fenestration
Ratio
Revised
Fenestration
Ratio
Snapshot
CC 25% 25%
DD 55% 70%
EE 70% 70%
FF 30% 52%
GG 30% 52%
HH 35% 52%
119
Building Original
Fenestration
Ratio
Revised
Fenestration
Ratio
Snapshot
II 17% 25%
JJ 18% 25%
KK 35% 35%
LL 35% 35%
MM 50% 50%
NN 40% 40% No Photo Available
120
Appendix G Additional Refined Data Set Correlations
The correlations explored in the Meta-Analysis have been reproduced using the Refined Data
Set. These correlations were not included in the body of the thesis because they do not lead to
any new conclusions.
1. Date of Construction
The Meta-Analysis equivalent to Figure G1 and Figure G2 can be found in Section 6.2.1 and
Section 6.2.2 respectively.
Figure G1: Energy intensity versus building vintage
0
100
200
300
400
500
600
1950 1960 1970 1980 1990 2000 2010
En
erg
y I
nte
nsi
ty (
ek
Wh
/m2)
Date of Construction
The Influence of Building Vintage on Energy
Intensity1961-1970 1971-1980 1981-1990 1991-2000 2001-2010 Average
121
Figure G2: Weather-dependant loads versus building vintage
2. Building height
The Meta-Analysis equivalent to Figure G3 and Figure G4 can be found in Section 6.2.3
Figure G3: Total energy use versus number of storeys
0
200
400
600
800
1000
1950 1960 1970 1980 1990 2000 2010
Va
ria
ble
En
erg
y I
nte
nsi
ty
(ek
Wh
/m2)
Date of Construction
The Influence of Building Vintage on Variable
Energy Intensity
1961-1970 1971-1980 1981-1990 1991-2000 2001-2010 Average
R² = 0.5039
0
2000
4000
6000
8000
10000
12000
14000
0 5 10 15 20 25 30To
tal
An
nu
al
En
erg
y (
1,0
00
s o
f e
kW
h)
Number of Storeys
The Influence of Number of Storeys on Total
Energy
122
Figure G4: Energy intensity versus building height by building vintage
3. Number and size of suites
The Meta-Analysis equivalent to Figure G5, Figure G6 and Figure G7 can be found in Section
6.2.4.
Figure G5: Total energy use versus number of suites
0
100
200
300
400
500
600
0 5 10 15 20 25 30
En
erg
y I
nte
nsi
ty (
ek
Wh
/m2)
Number of Storeys
The Influence of Number of Storeys on Energy
Intensity by Building Vintage1961-1970 1971-1980 1981-1990 1991-2000 2001-2010
R² = 0.5764
0
2000
4000
6000
8000
10000
12000
14000
0 100 200 300 400To
tal
an
nu
al
En
erg
y (
1,0
00
s o
f e
kW
h)
Number of Suites
The Influence of Number of Suites on Total
Energy
123
Figure G6: Energy intensity versus number of suites
Figure G7: Attributed suite size versus date of construction
4. Ownership Type
The Meta-Analysis equivalent to Figure G8, Figure G9 and Figure G10 can be found in Section
6.2.5.
R² = 0.0006
0
100
200
300
400
500
600
0 100 200 300 400
En
erg
y I
nte
nsi
ty (
ek
Wh
/m2)
Number of Suites
The Influence of Number of Suites on Energy
Intensity
0
20
40
60
80
100
120
140
160
180
1950 1960 1970 1980 1990 2000 2010
Att
rib
ute
d S
uit
e A
rea
(m
2)
Date of Construction
The Influence of Construction Date on Suite Size1961-1970 1971-1980 1981-1990 1991-2000 2001-2010 Average
124
Figure G8: Total energy intensity by area for each type of building ownership
Figure G9: Total energy intensity by suite for each type of building ownership
0
100
200
300
400
500
600
En
erg
y I
nte
nsi
ty (
ek
Wh
/m2
)
Energy Intensity by Ownership TypeOwned, No Subsidy Owned, Subsidized Rented, No SubsidyRented, Subsidized Average
0
10000
20000
30000
40000
50000
60000
Su
ite
En
erg
y U
se (
ek
Wh
/su
ite
)
Suite Energy Use by Ownership TypeOwned, No Subsidy Owned, Subsidized Rented, No SubsidyRented, Subsidized Average
125
Figure G10: Suite size versus building ownership type
0
50
100
150
200
Su
ite
Siz
e (
m2
)
Suite Size by Ownership TypeOwned, No Subsidy Owned, Subsidized Rented, No SubsidyRented, Subsidized Average
126
Appendix H Multi-Variable Linear Regression Results
This results of the multi-variable regression conducted on the Refined Data Set are presented
below. The only two analyses that considered all the variables available were the total energy
consumption analysis and the total energy intensity analysis. For the rest of the analyses the list
of variables was reduced as specified in the results. The list of variables available for each
regression analysis includes: number of floors, number of suites, gross floor area, year of
construction, heating boiler capacity, heating boiler efficiency, DHW boiler capacity, DHW
boiler efficiency, MAU ventilation capacity, air conditioner cooling capacity, presence of
balcony or through-wall slab, wall R-value, glazing U-value, fenestration ratio and boiler age.
1. Total energy consumption: systematic forward-selection
a. Variables considered: All 15
b. Variables selected and coefficients
Variable Coefficients
y –intercept -6,650,000
Heating boiler efficiency 4,390,000
MAU ventilation capacity 412
Gross floor area 83.8
Presence of balcony or through-wall slab 1,680,000
c. Adjusted R2: 0.845
2. Variable gas consumption: systematic forward-selection
a. Variables considered: 9
• Gross floor area, year of construction, fenestration ratio, heating boiler capacity, heating
boiler efficiency, presence of balcony or through-wall slabs, wall R-value, glazing U-
value, MAU ventilation capacity.
b. Variables selected and coefficients:
127
Variable Coefficients
Y –intercept 197,000
Wall R-value -183,000
Gross floor area 44.71
MAU ventilation
capacity 218
c. Adjusted R2: 0.81
3. Base gas consumption: systematic forward-selection
a. Variables considered: 6
• Gross floor area, year of construction, DHW boiler capacity, DHW boiler efficiency,
number of floors, number of suites
b. Variables selected and coefficients:
Variables Coefficients
Y-intercept 2,110,000
DHW boiler
efficiency -2,340,000
Gross floor area 40.4
c. Adjusted R2: 0.17
4. Variable electricity consumption: systematic forward-selection
a. Variables considered: 7
• Gross floor area, year of construction, air conditioner cooling capacity, number of floors,
number of suites, fenestration ratio, MAU ventilation capacity
b. Variables selected and coefficients:
None, all R-square values were less than 0.1
128
5. Base electricity consumption: systematic forward-selection
a. Variables considered: 7
• Gross floor area, year of construction, air conditioning cooling capacity, number of
floors, number of suites, fenestration ratio, MAU ventilation capacity
b. Variables selected and coefficients:
Variables Coefficients
y-intercept -146,000
MAU ventilation capacity 20.5
Gross floor area 58.2
Air Conditioning cooling capacity 159,000
c. Adjusted R2: 0.67
6. Total energy intensity: systematic forward-selection
a. Variables considered: All 15
b. Variables in order of selection and relative weighting
Order
Selected Variable Coefficient
Relative
Weighting
Adjusted R-Squared at
Time Variable was Added
y-Intercept -16,200
1 Fenestration ratio 335 1% 0.09
2
MAU ventilation
capacity 0.01 1% 0.19
3 Number of suites -1.2 1% 0.24
4 Year of Construction 8.3 94% 0.31
5 Wall R-value -25.0 1% 0.46
6 Number of floors 8.8 1% 0.57
7 Fenestration ratio -123 0% 0.58
7. Total energy intensity: logical method
a. Variables considered: 4
129
• Glazing U-value, fenestration ratio, boiler efficiency, boiler age
b. Variables selected and coefficients:
Order
Selected Variable Coefficients
Adjusted
R-Square
y-Intercept 200
1 Glazing U-value 137 0.097
c. Although the adjusted R2 value has already been maximized, these are the results if the rest of
the variables continue to be added
Order
Added Variable Coefficients
Relative
Weighting
Adjusted R-Square at
Time Variable was
Added
Intercept 529
1 Glazing U 166 24% 0.097
2
Fenestration
Ratio -57 5% 0.073
3 Boiler Efficiency -379 63% 0.057
4 Boiler Age (yrs) -2 8% 0.045