Residential Land-Use Density and Building Energy Consumption

MISSING IMAGE

Material Information

Title:
Residential Land-Use Density and Building Energy Consumption A Case Study of the City of Gainesville, Florida
Physical Description:
1 online resource (96 p.)
Language:
english
Creator:
Tjindra, Djundi
Publisher:
University of Florida
Place of Publication:
Gainesville, Fla.
Publication Date:

Thesis/Dissertation Information

Degree:
Master's ( M.S.A.S.)
Degree Grantor:
University of Florida
Degree Disciplines:
Architecture
Committee Chair:
WALTERS,BRADLEY SCOTT
Committee Co-Chair:
JONES,PIERCE H

Subjects

Subjects / Keywords:
conservation -- consumption -- electricity -- energy -- gainesville-fl -- residential -- urban
Architecture -- Dissertations, Academic -- UF
Genre:
Architecture thesis, M.S.A.S.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract:
To date, studies investigating the energy consumption associated with urban development patterns have focused mainly on the transportation sector. These studies suggest that more efficient land use and high-density urban development is more likely to lead to shorter travel times, more energy-efficient transportation, more cost-effective delivery of services, and an overall lower carbon footprint. Yet, there is growing interest in better understanding the correlations between urban form and residential building energy consumption. In service to this emerging body of knowledge, this study examines the correlation between net land use density and annual energy consumption per residential dwelling unit within the city of Gainesville, Florida. Spatial and residential energy consumption datasets used for the analysis model are based on Alachua County Property Appraiser and Gainesville Regional Utilities (GRU) records. A 3D geospatial visualization model was developed; an inferential and a descriptive analysis were conducted to investigate the correlation between annual per dwelling unit energy consumption (in equivalent kWh per utility meter) and net land use density (in DU/acre permeter associated parcel). Correlative insights and suggested research next steps were derived from the results of the statistical and graphical interpretation.
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility:
by Djundi Tjindra.
Thesis:
Thesis (M.S.A.S.)--University of Florida, 2013.
Local:
Adviser: WALTERS,BRADLEY SCOTT.
Local:
Co-adviser: JONES,PIERCE H.

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Classification:
lcc - LD1780 2013
System ID:
UFE0046365:00001


This item is only available as the following downloads:


Full Text

PAGE 1

1 RESIDENTIAL LAND USE DENSITY A ND BUILDING ENERGY CONSUMPTION: A CASE STUDY OF THE CITY OF GAINESVILLE, FLORIDA By DJUNDI TJINDRA A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE IN ARCHITECTURAL STUDIES UNIVERSITY OF FLORIDA 2013

PAGE 2

2 2013 Djundi Tjindra

PAGE 3

3 To my family and friends

PAGE 4

4 ACKNOWLEDGMENTS This thesis would not have been possible without the support and assistance of several individuals who in one way or another contributed and extended their valuable help in the preparation and completion of this research. For that very reason, I wo uld like to express my thank to B radley Walters and Pierce Jones who have work ed with me very closely until the c ompletion of this research work; to Ruth Steiner who has encouraged me as well as connec ted me with UF Program for Resource Effici ent Communities (UF PERC) team; to Hal Knowles Nicholas Taylor and Lynn Jarrett of UF PREC; to William Tilson, Martin Gold, and Albertus Wang who provided me their relentless encou ragement to complete this study; and to Edith Williams for helping me in editing this thesis. Also, I would like to convey my appreciation to Thomas Smith and Hui Zou for their recommendations; to Michael Kung for technical support; to all of my professors who have contributed their kno wledge in various forms; to my classmates with whom I walked this study path together; and last but not least, to my family who stood behind me and are the reason for everything I do.

PAGE 5

TABLE OF CONTENTS page ACKNOWLEDGMENTS .................................................................................................. 4 LIST OF TABLES ............................................................................................................ 7 LIST OF FIGURES .......................................................................................................... 8 LIST OF ABBREVIATIONS ........................................................................................... 10 ABSTRA CT ................................................................................................................... 13 CHAPTER 1 INTRODUCTION .................................................................................................... 15 Background ............................................................................................................. 15 Objectives ............................................................................................................... 15 Hypothesis .............................................................................................................. 15 Scope and Facts ..................................................................................................... 16 GRU's Residential Energy Consumption Profile ..................................................... 16 2 LITERATURE REVIEW .......................................................................................... 20 Housing Type and Energy Consumption ................................................................ 21 Critiques .................................................................................................................. 23 Urban Density and Energy Consumption ................................................................ 23 Geospatial Energy Consumption Modeling ............................................................. 27 Brief Summary ........................................................................................................ 29 3 METHODOLOGY ................................................................................................... 35 Strategy .................................................................................................................. 35 Data Sources .......................................................................................................... 36 Data Preparation ..................................................................................................... 37 GRU Consumption Database (GRU_CY2012) ................................................. 37 ACPA/CAMA 2012 Database ........................................................................... 38 NAL 2013 Database ......................................................................................... 38 Outliers Det ection ................................................................................................... 38 Data Grouping and Aggregation ............................................................................. 39 Visualization Tasks and Method ............................................................................. 41 Regression Method ................................................................................................ 41 4 VISUALIZATION ..................................................................................................... 47 Land -Use Density ................................................................................................... 47 5

PAGE 6

Energy Consumption per Residential Unit Density ................................................. 47 Normal Distribution of the Energy Consumption ..................................................... 48 5 ANALYSIS .............................................................................................................. 54 Examining the Loess Residuals .............................................................................. 54 Arithmetic Means and Median Methods Comparison ............................................. 55 Confidence Interval (CI) .......................................................................................... 55 Inferential Analysis .................................................................................................. 56 Residential Energy Consumption per House Type ................................................. 57 Case Studies .......................................................................................................... 58 6 CONCLUSIONS AN D DISCUSSION .................................................................... 72 Key Findings and Conclusions ................................................................................ 72 Suggestions for Further Studies ............................................................................. 74 Discussions ............................................................................................................. 75 Residential Energy Consumption by Building Vintage ...................................... 75 Seasonal Residential Energy Consumption ..................................................... 76 Energy load analysis for additional residential units in various scenarios ........ 77 APPENDIX A LIST OF DATABASES, TABLES AND FIELDS ...................................................... 86 B CAMA 2012 TABLE METADATA: BUILDING ......................................................... 88 C LIST OF IMAGES ................................................................................................... 90 Gainesville Land-Use Density Map ......................................................................... 90 3D Geospatial Model of Average ekWh per Land -Use Density .............................. 91 3D Geospatial Normal Distribution of the Energy Consumption ............................. 92 LIST OF REFERENCES ............................................................................................... 93 BIOGRAPHICAL SKETCH ............................................................................................ 96 6

PAGE 7

LIST OF TABLES Table page 2-1 Overview of studies on building energy consumpt ion in reviewed literatures ..... 31 3-1 Some examples of discrepancies between premise count and the value from NO_RES_UNTS of NAL1 1P201302 ................................................................... 44 5-1 Overview of studies on building energy consumption in reviewed literatures ..... 60 6-1 GRUs Historical and Projected Net Energy Load (NEL) and Summer/Winter Peak Demands (MW) ......................................................................................... 80 6-2 Summer energy load forecasts of seasonal residential energy load based on 0-1 5 and >1550 du/ac ....................................................................................... 80 6-3 Winter energy load forecasts of seasonal residential energy load based on 015 and >1550 du/ac ........................................................................................... 80 6-4 NEL calculation table with 1K and 2K extra residential units added ................... 81 7

PAGE 8

LIST OF FIGURES Figure page 1-1 GRU electric facilities and service boundary, Alach ua County, FL ..................... 18 1-2 GRUs residential energy consumption and number of customers ..................... 18 1-3 GRUs and Floridas avera ge kWh p er customer. ............................................... 19 2-1 U.S. Energy Consumption per Sector 2010 ..................................................... 34 3-1 Process diagram of data preparation from the raw format to output results ....... 45 3-2 A chart to quickly spot outliers ............................................................................ 45 3-3 A one to -many relationship between Parcel ID Premise ID tables and Premise ID monthly consumption data tables .................................................. 46 3-4 Process diag ram of data use for 3D modeling .................................................... 46 4-1 Gainesville map generated by ArcMap color-codes each residential land parcel based on its DU/ac density (map scale 1:40,000). ................................... 49 4-2 The ArcMap rendering shows the land-use density of Cabana Beach Apartment and Rockwood Villas complex (map scale 1: 25,000) ........................ 49 4-3 A 3D representation of average annual energy consumption per residential unit density of the City of Gainesville based on 2012 energy consumption data .................................................................................................................... 50 4-4 The color legends of density in DU/ac (on the left) and avg ekWh (on the right) ................................................................................................................... 50 4-5 A closer look at southwest residential area, the corner between SW Archer r oad and I -75 in Gainesville, FL .......................................................................... 51 4-6 A scene from the SW overlooking the NW area of the city ................................. 51 4-7 A detailed view of Haile Plantation subdivision, in the SW area of Gainesville, FL ....................................................................................................................... 52 4-8 The colors on the map show the standard deviation values of the average energy c onsumption per land-use density .......................................................... 52 4-9 gend, and the ekWh distribution curve .................................................. 53 410 A close-up look at the spatial distribut ............... 53 8

PAGE 9

5 1 Bivariate plot graph of median ekWh (2012) pe r DU/ac ................................ ..... 61 5 2 Fitted loess curve of median ekWh (20 12) per DU/ac (up to 153 DU/ac) ........... 62 5 3 Fitted loess curve of median ekWh (2012) per DU/ac (limited to 50 DU/ac) ....... 63 5 4 Residuals of loess fitted median ekWh (2012) per land use density with (on the left) and with ................................ ............................... 64 5 5 Residuals of loess fitted median ekWh 2012 per land use density with ....... 64 5 6 Comparison of energy consumption profiles normalized using ar ithmetic mean and median method ................................ ................................ .................. 65 5 7 Median ekWh 2012 per DU/ac plotted wit h 95% confidence interval bands ....... 66 5 8 Res. energy consumption (2012) vs DU/ac scatt erplot by house type ............... 67 5 9 Residential energy consumptio n (2012) vs DU/ac by house type ....................... 67 5 10 T he selected areas for case s tudy on the geospatial 3D model ......................... 68 5 11 Detailed view on selected area of study: R ockwood Villas and Cabana Beach 68 5 12 Detailed view of selected area of study : Haile Plantation (SFD and SFA) .......... 69 5 13 Aerial views of Rockwood Villas and Cabana Beach Apt areas ......................... 69 5 14 Street level view of Rockwood Vi llas and Cabana Beach Apt areas .................. 70 5 15 Aerial view of Hail e Plantation, U nit 15 and Unit 25 Ph. II areas ........................ 70 5 16 Street level view of Haile Plantatio n unit 15 and unit 25/II areas ........................ 70 5 17 Average annual energy consumptio n per studied residential units ..................... 71 6 1 Residential energy consumption (2012) by land use density, split by built year ................................ ................................ ................................ .................... 82 6 2 Averag e floor area (heated area) by built year ................................ ................... 82 6 3 R esidential energ y consumption (2012) by seasons ................................ .......... 83 6 4 GRU's Historical and Projected Net Energy Load (NEL) ................................ .... 84 6 5 Summer/Winter residential energ y consumption in ekWh vs. du/ac ................... 84 6 6 Addi tional NEL in various scenarios ................................ ................................ ... 85 9

PAGE 10

10 LIST OF ABBREVIATIONS 2D Two dimensional. 3D Three dimensional. ACPA Alachua County Property Appraiser. AHS American Housing Survey from U.S. Census Bureau. BTU British Thermal Unit. A traditional unit measurement of the energy required to cool or heat one pound of water by one degree Fahrenheit. CAMA Computer Assisted Mass Appraisal system. A generic term referring to software systems that provide real estate appraisal services for property tax purposes. CBECS Commercial Building Energy Consumption Survey from EIA. CO2 Carbon dioxide. CREEDAC Canadian Residential Energy End use Data Analysis Centre. DOE U.S. Department of Energy. DU/acre Dwelling Units/acre. Housing or residential land use density unit measurement. eGRID Emissions & Generation Resource Integrated Database from EPA. EIA U.S. Energy Information Administration (DOE EIA). EPA U.S. Environmental Protection Agency. EU 27 European Union with 27 member states (from Jan 1, 2007 to June 30, 2013): EU 25 + Bulgaria and Romania. EUI Energy Use Intensities (in kWh/sq m/year or kBtu/sq ft/year). FAR Floor A rea Ratio. FDOR Florida Department of Revenue. FRCC Florida Reliability Coordinating Council GHG Greenhouse Gas.

PAGE 11

11 GIS Geographical Information System. GRU Gainesville Regional Utilities. GWh Gigawatt Hour = 1 million kWh. kWh Kilowatt Hour. Unit measurement of energy equals to 1,000 watt hours or 3.6 megajoules. LCA Life Cycle Assessment. LOESS Locally Weighted Regression. LOWESS Locally Weighted Scatterplot Smoother. MFR Multi Family Residential home. NAL Name, Address, Legal file. Certified tax roll file. NCDC National Climate Data Center. NCEP/NCAR National Center for Environme ntal Prediction National Center for Atmospheric Research. NEL Net Energy for Load NNR NCEP/NCAR's 50 year Reanalysis (data). NRCan Natural Resources Canada. NRI National Resources Inventory (statistic reports) from U.S. Department of Agriculture (USDA). NRTEE National Round Table on the Environment and the Economy (Canada). NSDI Cadastral National Spatial Data Infrastructure from the Federal Geographic Data Committee (FGDC). OLS Ordinary Least Squares (Linear Least Squares) regression method. PREC Program for Resource Efficient Communities. PUM Public Us e Microdata Samples from U.S. Census Bureau. RDBMS Relational Database Management System.

PAGE 12

12 RECS Residential Energy Consumption Survey from EIA. SFA Single Family Attached home. SFD Single Family Detached home. Sq m Square meter or m2. Sq mile Square mile or mile2. Sqft Square Foot. T&D Electricity transmission and distribution. TYSP Ten Year Site Plan. Submitted to Florida Public Service Commission. UHI Urban Heat Island. USDA U.S. Department of Agriculture.

PAGE 13

13 Abstract of Thesis P resented to the Graduate School of the University of Florid a in Partial Fulfillment of the Master of Science in Architectural Studies RESIDENTIAL LAND USE DENSITY AND BUILDING ENERGY CONSUMPTION: A CASE STUDY OF THE CITY OF GAINESVILLE, FLORIDA By Djundi Tjindra December 2013 Chair: Bradley Walters Cochair: Pierce Jones Major: Architecture To date, studies investigating the energy con sumption associated with urban development patterns have focused mainly on the transportation sector. These studies suggest that more efficient land use and high density urban development is more likely to lead to shorter travel times, more energy efficien t transportation, more cost effective delivery of services, and an overall lower carbon footprint. Yet, there is growing interest in better understanding the correlations between urban form and residential building energy consumption. In service to this em erging body of knowledge, this study examines the correlation between net land use density and annual energy consumption per residential dwelling unit within the city of Gainesville, Florida. Spatial and residential energy consumption datasets used for the analysis model are based on Alachua County Property Appraiser and Gainesville Regional Utilities (GRU) records. A 3D geospatial visualization model was developed; an inferential and a descriptive analysis were conducted to investigate the correlation betw een annual per dwelling unit energy consumption (in equivalent kWh per utility meter) and net land use density (in DU/acre

PAGE 14

14 per meter associated parcel). Correlative insights and suggested research next steps were derived from the results of the statistical and graphical interpretation

PAGE 15

15 CHAPTER 1 INTRODUCTION Background As the cities around the world continue to grow economically as well as spatially, the global rising demand of energy has caused the anthropogenic Greenhouse Gas (GHG) emissions to increase at a steep rate. Managing urban energy consumption and future energy demand, as a part of energy conservation efforts, is one of many strategies to mitigate GHG emissions. Efficient land use patterns and high density urban development is more likely to lead to efficient energy utilization. The direct impact b etween the urban form and energy efficiency, mostly regarding transportation, has long been studied. This study looks into the significance of how urban land use density contributes to community energy consumption, particularly in the City of Gainesville, FL. Objectives The main objective of this study is to determine if a statistically significant correlation exists between land use density and annual energy consumption per residential unit in Gainesville, Florida. Other objectives include creating visual representations of land use density and annual energy consumption of the studied geospatial area, so that insights of the relationship can be derived from the statistical and graphical interpretation Hypothesis The null hypothesis ( ! to be tested in this research states that there is no statistically significant correlation between land use density and annual energy consumption per residential unit in the city of Gainesville, FL. The alternative hypothesis

PAGE 16

16 ( ! states otherwise, that there is a s tatistically significant correlation between land use density and annual energy consumption per residential unit in the city of Gainesville FL Scope and Facts Gainesville Regional Utilities (GRU) is a municipal electric, natural gas, water, wastewater, a nd telecommunications utility system, owned and operated by the City of Gainesville, Florida and is the 5th largest municipal electric utility in Florida. The City of Gainesville is the county seat of Alachua County, located in North Central Florida. It ha s population of 123,903 as of April 2012 (City Of Gainesville, FL, 2012) and contains approximately 55,989 housing units (U.S. Census Bureau, 2011) Gainesville is also home to the University of Fl orida and Santa Fe College. As a part of the energy conservation effort, the City of Gainesville, Florida has committed to adopt clean energy strategies. GRU has the most solar energy installed in Florida, and is currently pursuing biomass as a renewable e nergy sourc e to meet the city energy needs for the next three decades. This study was conducted using the GRU monthly electric billing data from residential utility customer within the GRU service area. Figure 1 1 shows GRU electric facilities and service boundary. Its service area includes the City of Gainesville and the surrounding urban area with the zip codes from 32601 to 32609, 32614, 32615, 32641 and 32653. GRU's R esidential E nergy C onsumption P ro file Currently, residential consumers are responsible for 30 percent of the total energy consumption in Florida (EERE, 2013; EIA, 2012) compared to 37% of overall residential electricity consumption in the U.S. (EIA, 2012) The 2012 energy

PAGE 17

17 consumption for residential use in Gainesville, Florida alone was about 44.5% (GRU, 2013) Based on 2011 statistics, residential electricity consumed in the state of Florida is 1 13,554 GWh (Florida Public Service Commission, 2012, p. 12) while GRU's residential consumption is 805 GWh (GRU, 2013, p. 32) At the same time, GRU's 2011 average kWh per residential cu stomer is 9,829 kWh, and is much lower compared to the rest of Florida, which is 13,567 kWh per residential customer. GRU generates approximately 0.9% of the total energy load in the state of Florida (Florida Public Service Commiss ion, 2012) Figure 1 2 shows history and forecast of GRU's residential energy consumption and number of customers from 2003 to 2022. Although the number of GRU customers is increasing from year to year, overall the average energy consumption per customer has been declining. GRU has managed and continue s to reduce its average residential energy consumption more than the overall Florida's residential consumption, as shown in the graph 's trend line I n Figure 1 3 FL trend line slope, around 0.36% per year, has been far more moderate than GRU's, which is around 1.06% per year.

PAGE 18

18 Figure 1 1. GRU electric facilities and service boundary, Alachua C ounty, FL (GRU, 2013, p. 9) Figure 1 2. GRU's residential energy consumption and number of customers (GRU, 2013, p. 32) !"#$$$% &$#$$$% &"#$$$% '$#$$$% '"#$$$% ($#$$$% !"$% &$$% &"$% '$$% '"$% ($$% )$$*% )$$+% )$$"% )$$!% )$$&% )$$'% )$$(% )$,$% )$,,% )$,)% )$,*% )$,+% )$,"% )$,!% )$,&% )$,'% )$,(% )$)$% )$),% )$))% !"#$ !%&'($)*(+,*-./0$*-*)12$34-(567.4-$ /-,$-568*)$49$35(:46*)($$ %-./% %012345672%

PAGE 19

19 Figure 1 3. GRU's and Florida's average kWh per customer (GRU, 2013; FRCC, 2012, p. 2) !"###$ %"###$ &#"###$ &&"###$ &'"###$ &("###$ &)"###$ &*"###$ '##($ '##)$ '##*$ '##+$ '##,$ '##!$ '##%$ '#&#$ '#&&$ '#&'$ '#&($ '#&)$ '#&*$ '#&+$ '#&,$ '#&!$ '#&%$ '#'#$ '#'&$ !"#$ %&'$!"#$()*+,)-.%/$)-)('0$12-*345.2-$5)($13*624)($ $-./$ $0123456$ 748963$:$-./;$ 748963$:$0123456;$ <4=>23?$$$$$$$$$$$$$$$$$$$$$$$@239A6=>$

PAGE 20

20 CHAPTER 2 LITERATURE REVIEW This literature review looks into the theory based and empirical research that exists on the correlation between urban density and energy consumption. I n the 1970s, the research on urban form and energy consumption gained momentum due to the energy crises that took place in that decade. This circumstance propelled the research on energy consumption efficiency in relationship to the impact of urban form. However, the interest subsided along with the fall of oil prices in the 1980s. Currently, the return of h igh energy prices and the growing concerns over global climate change has put the research in this area back on the agenda (Safirova, Houde, & Harrington, 2007; Norman, MacLean, & Kennedy, 2006) B ased on the 2010 stati stics, as shown in Figure 2 1, buildings in the U.S. account ed for 41% of primary energy consumption (19% commercial and 22% residential), while industry and transportation sectors consumed only approximately 31% and 28 % respectively (EERE, 2012) Past studies on the relationship between urban form and energy consumption are mostly focused on transportation, travel patterns and urban heating. Most researchers would agree that efficient land use and high density urban development is likely to lead to shorter travel times, energy efficient transportation, cost effective delivery of services, lower energy consumption and overall lower carbon footprint. Despite the fact that overall energy consumption in buildings is higher than the energy consumption in transportation sector, there was not much attention given to th is related issue According to Ewing and Rong (2008) this is understandable due to the dependency of the transportation sector on oil as source of energy and the geopolitics

PAGE 21

21 that make headlines. In fact, energy consumption in buildings plays a significant role in global Greenhouse Gas (GHG) emissions The direct impact of the urban form and energy efficiency, or more specifically, the s ignificance of urban development density towards the community energy consumption, has yet to be studied extensively (Safirova, Houde, & Harrington, 2007) Housing Type and Energy Consumption Ewing and Rong (2008) studied the impact of urban form on residential energy consumption in the U.S. based on county sprawl index es The authors argued that there are direct and indirect causal paths and that urban form can affect residential energy use. The direct impact is the loss of electric ity through transmission and distribution (T&D) from generating sites to the residential area, where the loss in less dense urban area is expected to be higher due to the longer transmission line. However this impact is less significant, since the T&D loss only represent s 7% of the total electricity generated in the U.S. The other indirect impacts are through intermediate variables such as housing type and size, formation of urban heat island s (UHIs), house ownership, household income, number of members and ethnic background. The housing size relates to the required space heating and cooling areas; more energy is needed for a bigger house compared to a smaller one. Likewise, more energy is nee ded for a detached house compared to an attached one of the same size, due to the exposed surface area. The UHI effect is caused by object surfaces that absorb heat and the lack of natural cooling properties, such as tree shading and evapotranspiration. Ex cess heat from space heating and cars also contribute to higher temperatures, and this translates to higher energy demand for summertime cooling, especially in the Southern region. However,

PAGE 22

22 the housing effect is more dominant than the UHI effect, and the a uthors came to conclusion that, urban sprawl can be said to inflate residential energy consumption and associated greenhouse gas emissions regardless of location. (p. 22) Other studies also showed that sin gle family detached housing (SFD) have consistently consumed more energy compared to multi family housing (Kaza, 2010; Owens, 1990) However, SFD normalized energy consumption per area is consistently lower than other h ousing types. On the one hand, SFDs are more efficient in term of energy consumption per area, but on the other hand, they tend to be larger in size and thus higher in total energy consumption. This dichotomy illustrates the "rebound effect", where the ga ins in efficiency are offset by the consumption increases, as described by Haas, Auer, and Biermayr (1998) : They [several economists] argue that increases in energy efficiency will lead to cheaper prices for service prov ided and to a substantial increase in service and energy demand. This increase will outweigh the conservation effect to a large extent and, hence, make conservation programs useless. (p. 195) Kaza found that m ulti family units within large apartment blocks consumed 50% to 60% less energy, but only 10% to 20% less within smaller blocks, compared to single family house (Kaza, 2010, p. 6576) and later concluded that, Changing housing type mix makes a difference only when replacing single family residences with multi family units in large apartment blocks. For the most part, other types of housing types do not promote savings in energy use across the consumption spectrum for all uses. (p. 6582) Additionally, this finding is also in agreement with other author s such as Owens (1990) D ue to the lack of data, Kaza (201 0) is inconclusive in regard to how the influence of climatic factors at the micro scale affects residential electricity consumption.

PAGE 23

23 Critiques Randolph (2008) commented on Ewing and Rong's (2008) findings and argued that the reduction in energy consumption is better achieved through residential energy efficiency improvement, rather than through land use, density and housing type modification. Likewise, Staley (2008) argued that although Ewing and Rong's analysis is empirically sound, the data and methods used do not justify the conclusions. Furthermore, Staley's critiques focused more on the policy analysis. He argued that policies, which promote energy conservation and innovation incentives leading to energy efficiency improvements are more preferable than fostering density and smaller housing type. Urban Density and Energy Consumption While most of the studies in the literature focused only on t he transportation sector, and showed that the high density urban areas consume less energy per capita than in low density suburban areas, Larivire & Lafrance (1999) found that electricity consumption in high density urb an areas is lower than in low density areas. Focusing specifically on electricity consumption and using a dataset of 45 densely populated cities in QuŽbec, Canada, the authors modeled electricity consumption per capita as a function of urban density, demog raphic, economic, and climatic factors Larivire & Lafrance's (1999) model showed that by increasing the density of an urban area by the factor of 3, from 360 persons/sq mile to 1,080 persons/sq, the electricity consumption would decrease by 7%. T he authors concluded that compared to gasoline use, the effect of the populat ion density is less significant, and pointed out that the change in urban design and land use policy in those cities would not have a huge impact on reducing overall electricity consumption (p. 62) This sugges ted that density is

PAGE 24

24 only one of many other important factors in lowering the electricity consumption per capita. Another study was conducted by Norman et al. (2006) employing Life Cycle Assessment (LCA) approach on the analysis of the energy consumption and GHG emissions associated with the residential density in the c ity of Toronto, Canada. According to the authors, Toronto 's development trends and housing styles are quite similar to those of the relatively compact cent ral core and suburban sprawl patterns common to many major cities in North America. LCA represents a broader approach to examining the environmental impacts of an entire building, in which the material and energy flows are quantified and analyzed, includin g upstream and downstream processes. Upstream processes of the building include raw material extraction, production, transportation and building construction, while downstream refers to operation, maintenance and demolition of the building (Treloar, Love, Faniran, & Iyer Raniga, 2000) Newton, Tucker, & Ambrose (2000) characterized this as embodied (embedded) energy, operating energy and maintenance energy (pp. 76 77) In Norman et al.'s (2006) model, the analytical framework utilized for estimating energy use and GHG emissions includes three diverse components ; construction materials, building operations, and public/private transportation. After summing the results of each individual component, the authors concluded that the low density suburban development consumes 2.0 to 2.5 times more energy and GHG emissions compared to the high density urban core developme nt per capita. At the same time, low urban density consumes 1.0 to 1.5 times more energy and GHG emissions

PAGE 25

25 proportionally, than the high density urban core development per unit of living space (p. 19) While ov erall added result s of the three components shows the difference in energy consumption between the low and high density urban areas, on the building operations component alone low density and high density developments in term s of energy and GHG emissions per unit of living space, are nearly equal. On the other hand, the low density development consumes energy twice as much as the high density, if the functional unit is changed from per unit of living space to per capita (p. 19) In a more recent study, Wilson (2013) examined the correlation between residential electricity consumption and building characteristics using a unique dataset. The objective of his study, according to the author is to provide empirical evidence that supports or refutes the assertion that more compact urban form matters with respect to residential electricity consumption, including how the way that residential subdivisions are configured. Focusing on SF D housing units, the author raised two questions. First, whether there is a relationship between urban form and electricity consumption for SFD housing in Illinois, post covariates adjustment (i.e. household and structural characteristics, climatic and oth er mitigating factors), and second, what are the implications of the first findings for how residential subdivisions are designed and built (p. 63) The required data for the model were collected from three Ill inois counties (Adams County, Macon County, Champaign County). They are based on local residential utilities consumer mail surveys including residential energy consumption

PAGE 26

26 records over the last 12 months, as well as the data from !"#$%"&'$()*+&",-.*/0&# 123'$0&* 41,5".*6 RECS) 2005. Using linear regression method, the seasonal and annual electricity consumption pattern was determined as a function of demographic, climatic, structural, technological, behavioral and urban form factors. Previous literature suggested that the Empirical evidence establishing the influence of climatic factors at the micro scale on residential electricity consumption is sparse (Ewing & Rong, 2008), despite strong conceptual arguments in favor of a connection ." Other findings by Holden (2004) and also by Holden and Norland (2005) discovered that there was evidence of a density effect with lower consumption observed in more densely developed areas; therefore the authors argued in favor of more compact urban form (Wilson, 2013, p. 64) Based on the model, Wilson (2013) observed that the most consistent predictors of residential electric ity consumption are climate days, household size, number of bedrooms, and heating equipment. Moreover, The general relationship between the net density of the subdivision and summer electricity usage is negative, but when cooling degree days are lower tha n average the downward slope of this relationship is steeper than when the number of cooling degree days is higher than average. Therefore the author concluded that "t he negative relationship observed between net density at the subdivision level and summer electricity usage is consistent with arguments in favor of more compact development patterns and is interpreted in the context of the heat island effect." Finally, urban form characteristics matter at the micro scale and reiterates the potential value of compact residential development for managing residential electricity consumption, and by extension, greenhouse gas emissions. (p. 70)

PAGE 27

27 The author argued that, "homes in subdivisions that are more compact and less peripheral are likely to reap benefits in the form of reduced electricity consumption." Geospatial Energy Consumption Modeling The advance of geospatial technologies allow energy consumption studies to utilize Geographical Information System (GIS) in mapping the related analytical data ( energy consumption profiles, base electricity loads, heating and cooling load s) to their actual geographical locations. The result is a geo visual analytic model, in which the data can be explored and validated geospati ally in two dimensional (2D), as well as three dimensional (3D) map s Kraak (2003) argued that this interactive and dynamic environment is helpful to generate hypotheses, develop problem solutions, construct knowledge an d stimulate the visual thinking. To study the anthropogenic heating effects in building sector, Heiple and Sailor (2008) developed a modeling technique to predict energy consumption for residential and commercial sectors at a high spatial resolution. A high resolution spatial data of any given day's hourly results at individual parcel level can be obtained when detailed building energy simulati ons of prototypical buildings with GIS geospatial data are integrated. The data should contain the building characteristic's information such as building types sizes and geographical location The city of Houston, TX, was selected as t he case study city since it has historical extreme heat and pollution in summertime, and it is one of the densest U.S. cities with a high level of energy consumption per capita. Thus it is likely that Houston's local climate is influenced by waste heat emitted from its build ings. The authors claimed that the same method could be applied to virtually any major city in the United States to

PAGE 28

28 estimate day specific residential and commercial electricity and natural gas consumption. Min, Hausfather, and Lin (2010) presented a novel approach, using five digit zip code high resolution level information to model residential energy by end use and fuel type for the entire territory of the United States. The authors developed four models for space heati ng, space cooling, water heating, and also appliance energy end uses, based on the !+/4* 2005 dataset. They found large variations between residential energy use characteristics both within and across different regions of the country, with significant diffe rence of and distribution by fuel of residential energy consumption in urban and rural areas. Model outputs by fuel type for each zip code include electricity, natural gas, fuel oil, propane and other (e.g. wood, solar, coal, etc.) were spatially mapped us ing GIS software. T he authors suggested possible research applications for the developed models, including examining factors in rural urban residential energy consumption and fuel use pattern s ; estimating residential end use costs in combination with fuel specific price data and examining the impact of various residential energy conservation actions, based on zip codes or regions. Howard et al. (2012) proposed a model to estimate the building sector energy consumption for four primary end use intensities including base electric, space cooling, space heating, and water heating in New York City. The end use intensity is measured in kWh/sq m floor area. With the annual electricity and natural gas consumption data provided by the New York City Mayor's Office of Long Term Planning and Sustainability and also information derived from Residential Energy Consumption Survey (RECS)

PAGE 29

29 2005 and Commercial Building Energy Consumption Survey (CBECS) 2003, the annual end use energy consumpt ion data were developed by performing a robust multiple linear regression resulting in information about total fuel intensities for 8 different building types ( residential 1 4 family, residential multi family, office, store, education, health, warehouse an d other commercial ) Finally, based on the zip code level clusters for four primary end use rs the annual end use energy intensities were applied to building floor area across New York City to determine the spatial distribution of the energy consumption. B rief Summary Table 2 1 summarizes the studies in reviewed literatures of residential energy consumption. Studies in the reviewed literature on urban density and its relationship with building energy consumption have been inconclusive. Several authors argue d that high density development reduces energy consumption (Ewing & Rong, 2008; Holden, 2004; Holden & Norland, 2005; Wilson, 2013) while others believed that the link between urban density an d energy consumption is sparse (Larivire & Lafrance, 1999; Norman, MacLean, & Kennedy, 2006) This reflects the fact that this relationship between urban density and aspects of sustainability remains a debatable issue (Kaza, 2010; Echenique, Hargreaves, Mitchell, & Namdeo, 2012; Wilson, 2013) While energy consumption in buildings is influenced by some spatial and non spatial factors, such as building types, attributes, us e, occupancy and weather t he complex interactions between those factors make it difficult to conclude with confidence that any one specific urban form will be more energy efficient than another (Owens, 1990, p. 63; Doherty, Nakanishi, Bai, & Meyers, 2009)

PAGE 30

30 Moreover, according to Doherty et al. (2009) there is no one size fits all solution for optimizing the energy consumption within the built environment per se, where compar ison between locations will be ambiguous (i.e. comparison between countries or even within a country). On the other hand, by analyzing the forms of energy consumption separately, it may be possible to assess differences in performance at different levels.

PAGE 31

31 Table 2 1. Overview of studies on building energy consumption in reviewed literatures. Author Variables Dataset Analysis Methodology Ewing & Rong (2008) Housing type and size, formation of UHIs, house ownership, household size, income and ethnicity, HDDs, CDDs. Sprawl index as measure of urban form. Sprawl index: U.S. Census, NRI reports from USDA; Residential energy use: RECS 2001 from EIA; Housing stoc k: PUMS 2000, AHS 1998, 2002. Climatic data: HDD, CDD from RECS including NCEP/NCAR Reanalysis (NNR); Others: ESRI GIS Data & Map 2005. OLS (Ordinary Least Squares) regression Kaza (2010) Housing type, size and year built, house ownership, household inco me, urban index, avg energy price, climate: HDDs, CDDs. RECS 2005 derived from householder interviews, mailed questionnaires and energy consumption data provided by utilities companies. Quantile regression Larivire & Lafrance (1999) Electricity consumption, economy and demography (population density, population, household income, std land wealth per capita, etc ), housing type, energy profile, climate, etc. Demography data: Statistics Canada 1992; Electricity consumption data: Hydro QuŽbec 1991; other data from various organizations. Linear regression

PAGE 32

32 Table 2 1. Continued Author Variables Dataset Analysis Methodology Norman et al. (2006) Building materials, building operations, public/private transportation, urban density, GHG emissions. Various data sources for urban infrastructure materials; Residential energy use/GHG emissions: NRCan 2003; SFD & apartments data: NRCan 2003, CREEDA C 2000, Statistics Canada 1999. Trans portation: NRTEE 2003, University of Toronto 2003, Kennedy 2002, City of Toronto, etc. Life Cycle Assessment (LCA) analysis Wilson (2013) Housing and household characteristics, electricity consumption, demographic, climatic, structural, technological, behavioral, urban form. RECS 2005; Electricity consumption from local utilities and other data from residential mail survey. Linear regression Heiple & Sailor (2008) Building type (commercial and residential), building uses, floor space, weekly occupancy duration, building age; climate: annual HDDs, CDDs; Energy consumptions, incl. Energy Use Intensities (EUI). RECS and CBECS for building prototypes. Parcel GIS data from NSDI. eQuest building simulation program from U.S. DOE.

PAGE 33

33 Table 2 1. Continued Author Variables Dataset Analysis Methodology Heiple & Sailor (2008) Building type (commercial and residential), building uses, floor space, weekly occupancy duration, building age; climate: annual HDDs, CDDs; Energy consumptions, incl. Energy Use Intensities (EUI). RECS and CBECS for building prototypes. Parcel GIS data from NSDI. eQuest building simulation program from U.S. DOE. Min et al. (2010) Housing type, size and year built, urban index, household demography; Climate: HDDs, CDDs; Electric price. RECS 2005; Climatic data: NCDC 2009; Emission data: eGRID from EPA; Zip codes, household demography, housing types, etc.: U.S. Census 2000; Fuel price from EIA 2005. OLS regression Howard, et al. (2012) Building characteristics: building type, total floor area. Electricity consumption: base electric, space cool ing, space heating, and water heating. Total fuel consumption. RECS 2005; CBECS 2003; annual electricity & gas consumption data from NYC Mayor's Office of Long Term Planning and Sustainability. Robust multiple linear regression.

PAGE 34

34 Figure 2 1. U.S. Energy Consumption per Sector 2010 (EERE, 2012) !"#$%&'() *+,) &'-"%./'&-0/") 12,) '3%!#3"0-4) 11,) 5/663'5!-4) +7,) 8$!4#!"9%) :+,) !"#$%&'()#*+%,-./0+%#/&'#,&12+'#345657#

PAGE 35

35 CHAPTER 3 METHODOLOGY Strategy The spatial and residential energy consumption datasets used for the visualization and analysis model are based on Al achua County Property Appraiser (ACPA) and Gainesville Regional Utilities (GRU) records. The GRU's energy consumption data used in this thesis is based only on 2012 consumption data, provided by Program for Resource Efficient Communities (PREC) at the Univ ersity of Florida. This thesis study is a part of PREC's larger research project, which relates to PREC's joint cooperation with GRU. The raw data was provided in comma separated t ext file format, and was later processed, cleaned and stored in a relational database management system (RDBMS). It is common for a logically interconnected relational data model to be stored in several tables where each has a relationship to the other tables in the form of one to one, one to many and many to many. The data mining multi step process includes data extraction, filtering, transformation, merging and integration. These steps are required to generate the data in the correct format for visualization and analysis. Conventional database tools available in MS Office 2013, s uch as Power Query in MS Excel and Query feature in MS Access, are used for the data mining purposes. The processed results are later stored in mySQL database as cleaned and ready to use data. This will subsequently be used for further queries using MS Acc ess and MS Excel to generate the results for the graphs and ArcGIS data feed. The diagram in Figure 3 1 shows the required block

PAGE 36

36 processes, from data mining to the generation of the clean data format, ready for the model. A geospatial visualization model w as developed using ArcGIS latest available version 10.2. Subsequently, a descriptive analysis based on the graphs generated from the data was conducted to investigate the correlation between annual per dwelling unit energy consumption (in equivalent kWh pe r utility meter) and net land use density (in DU/acre per meter associated parcel). Data Sources The following databases were used in building the analysis data model. A complete list of the database tables and their data fields can be found in Appendix A, and the metadata of CAMA 2012 database related table is listed in Appendix B. 1. CAMA 2012 database is Alachua County Property Appraiser (ACPA), public downloadable Computer Assisted Mass Appraisal (CAMA) database. The database was downloaded in August 201 3 from ACPA's GIS Service Center online website (ACPA, 2012) The dataset tables required for the analysis data model are as follows: Parcel: This table has 109 176 unique parcel records. The PIN field is used as primary iden tifier for each unique parcel of the data model. Address: This table has 123,854 parcel related addresses Building: Building table has 88,927 records associated with Parcel in one to many relationship. 2. ACPA's "Public GDB" file is similar to CAMA 2012, on ly it is in ArcGIS GDB file format instead of MS Access. 3. GRU consumption database (GRU_CY2012) with datasets comprised of approximately 82,122 GRU residential electric customers with individual utility

PAGE 37

37 billing records from total 85,500 premises; and 2 051 908 consumption billing records of the year 2012. GRU Premises table: This table has 85,500 unique premises associated with 51,462 unique parcels. Each premise has a physical address consists of premise house number, street name, premise unit, city, state and zipcode. meter_type: This field contains the meter type: E, G or W. "E" stands for Electricity, "G" for Gas, and "W" for Water consumption. date_to: This field stands for Meter Read Date. is_cancellation: This field stands for Cancelled Bill Status. C ancelled if marked with an "X". 4. NAL 2013 database is Name, Address, Legal (NAL) database of Alachua County. Parcel_ID: The table has 100 441 unique parcel records. LND_SQFOOT: This stands for Land Size in Square Feet. Later converted to acre. TOT_LVG_ARE A : This stands for Total Living Area in Square Feet. Data P reparation Data transformation and cleansing are required prior to importing the data from above mentioned sources into the analysis data structure. The following lists the required steps. GRU C on sumption D atabase (GRU_CY2012) Excluded all non electric billing records. The GRU residential utility consumption database consists of all utility billing records from 2008 to 2012, including electric, water and gas totaling about 3.6 million records ; only the electric billing records are used. Upon examination of the electric billing records, there are some records, which do not have any Parcel ID associated with them. These records were excluded along with the records that have "is_cancellation" status ma rked with an "X".

PAGE 38

38 The street addresses in the Premise dataset were converted into the geospatial address locations in geodetic latitude and longitude coordinate format and respective columns were added ("latitude", "longitude"). All natural gas consumptio ns were converted to its kWh equivalent (expressed in ekWh units) by multiplying it with 29.30722. ACPA/CAMA 2012 D atabase Excluded all records in Parcel table, which has no PIN number (PIN is empty) Determined the quantity of units per parcel (Dwelling U nits or DU) by grouping the parcels in the Address table together. NAL 2013 D atabase Converted the "Parcel _ID" format from "00000 000 000" to "00000 000 000". Converted the value "LND_SQFOOT" from s quare feet to acre by dividing it with 43,560 and store the new value in a new "size_acre" column. Following fields were selected for the calculation of DU/acre: "Parcel_ID", "LND_SQFOOT", "size_acre" and "TOT_LVG_AREA". For DU/ac calculation, acre unit can be deriv ed from the field LND_SQFOOT value by dividing it with 43,560. Outliers D etection A simple deviation method can be used to filter out premises energy consumption outliers within each parcel. Outliers records could be easily detected by examining their st andard deviation values, Records with values, which lie within times its standard deviations ! !" of the median, are included for further calculation, otherwise they are discarded. The threshold value used here is 4, so that the !"# ! va lue must be within four times its standard deviations, which corresponding to the 99.99 percentile of the normal distribution

PAGE 39

39 The total !"# ! of a parcel is equal the sum of all energy consumption of its premises, and can be expressed as follow: !"# ! ! !"# ! ! ! (3 1) If ! is the median value of energy consumptions of all premises within the parcel, and is the number of premises within the parcel, then the parcel's standard deviation ! can be expressed as follow: ! ! ! ! !"# ! ! ! ! ! (3 2) To determine the test value of the !"# ! : ! ! ! !"# ! ! ! ! ! (3 3) All premise records with its value > 4 are considered to be outliers and will be discarded. Figure 3 2 below shows an overview of the each record's standard deviation value on a chart. The purpose of this chart is to provide a quick eye level observation of the whole dataset to easily spot o utliers if they exist. Later these outliers will be removed from the dataset using common SQL statement. As an example, there are a few records, which exceeded the 4 threshold in the dataset. One of them is premise #7000052152 which has > 12, as shown in Figure 3 2. Data G rouping and A ggregation GRU Premises table has 85,500 unique premises associated with 51,462 unique parcels in a one to many relationship, where one Parcel is linked to one or many

PAGE 40

40 Premises (see Figure 3 3). At least one premise record exists and can be linked to the related parcel record. After the process of data preparation steps described above, to determine the number of the premises of each parcel, the grouping and counting of parcels is necessary. This can be done using SQL stat ement i.e. SELECT DISTINCT (Parcel_ID), count(*) as premise_count FROM GRU_Premises; Although NAL11P201302 data table has a NO_RES_UNITS field, that should indicate the number of residential units of the parcel, there are some discrepancies between these numbers and the numbers obtained from the GRU_Premises' premise count as shown in Table 3 1 Since the energy consumption is directly derived from GRU database, it is more relevant to use the number resulted from premise count rather than those from NAL11 P201302 data table. To calculate the parcel's density ( !"#$%&' ! expressed in DU/ac unit, the premise count of each parcel ( !"#$%&# !"#$% ! is divided by the acre value obtained from converting LND_SQFOOT ( !"# !"#$$% ! as described in data preparation above. !"#$%&' ! !"#$%&# !"#$% !"# !"#$$% ! !" !"# ! !" !" ( 3 4) To reflect the energy consumption per dwelling unit of each parcel, the sample median method is used to calculate the average energy consumption (ekWh) of all premises on each respective parcel. The aggregated dataset required for the rendering of the 3D model in ArcGIS should at least have the following fields: Parcel_ID, du_acre, avg_ekWh Data used for scatter plots only require two fields, du_acre and avg_ekWh In order to obtain a single

PAGE 41

41 du_ac data point, all du_acre rows need to be grouped together b ased on their values with respective avg_ekWh values averaged. Visualization Tasks and Method The visualization tasks for generating the data model include: 1. A visual representation of land use density and 2. a visual representation of average energy consump tion per residential unit density in Gainesville, Florida. The visualization method utilizes the aggregated data in combination with the GIS database (Public GDB File) downloaded from ACPA's GIS Service Center online website (ACPA, 2 012) to generate the 3D model in ArcGIS, as shown in the process diagram in Figure 3 4 Using the Parcel layer from the Public GDB database, the aggregated data table "Parcel_ID" field was joined with the "PIN" field with a INNER JOIN method, where only matching records from both tables are selected. The symbology of the Parcel layer was set to show the value of "avg_ekWh" based on the graduated colors from green to red in 25 steps. This will map the average energy consumption value to its designated color. Regression Method Locally Weighted Regression (LOESS) method was used to fit the residential energy consumption and densit y model to data Loess is a non parametric regression method, which does not require a predetermined definition of the relationship between the dependent and independent variables, but is defined based on the information derived from the data. Loess method is a generalization of the preceding LOWESS

PAGE 42

42 method, an acronym for Locally Weighted Scatterplot Smoother, and i s used most frequently as bivariate scatterplot smoother (Cleveland & Devlin, 1988; Jacoby, 2000) Compared to parametric regression methods used in the reviewed literature, such as Simple Linear Regression, Ordinary Least Squares (OLS), Quantile r egression methods etc., the non parametric method used in LOESS addresses some limitations faced by the parametric regression methods. P arametric fitting is broadly used, and is a very effective way to construct a relationship between the variables if the structure in the data conforms to the type of function that is fitted by the smoothing algorithm. However, the exact functional form is mostly unknown. Non parametric approach such as Loess can be used to "locate a smooth curve among the data points without requiring any advance specification of the functional relationship between the variables." (Jacoby, 2000) Suppose that number of observations on two variables, and and the iteration ! ! !" ! The plotted points in a bivariate scatterplot are the ordered pairs ! ! ! The whole model can be expressed in the following relationship: ! ! ! ! ! (3 5) Where is the regression function or observed fitted value, and ! is the random error or the residual value. The ! should approximate the true unobserved value, in which the approximation is attained by fitting a regression li ne within a chosen neighborhood of the data point ! Loess is not strictly a descriptiv e tool. According to Jacoby (2000) : "the statistical theory for local regression models has been worked out, so it is possible to incorporate an inferential component into a loess analysis." (p. 594) Moreover,

PAGE 43

43 Statistical inference with a loess smooth curve is usually gr ounded in least squares theory (Cleveland & Devlin, 1988) and it require s several assumptions. Specific ally, the observed fitte d values, ! are now viewed as estimates that should approximate, as closely as possible, the true but unobserved fitted values, ! ! Fur thermore, the residuals about these fitted values should be gaussian. That is, the ! ! ! should be indepe ndently and identically distributed according to a normal distribution, with a mean of zero and a constant variance. When these assumptions are met, direct generalizations of traditional least squares methods can be employed to perform statistical tests. T o test the null hypothesis ( ! of no functional dependence between Y and X variables on a single loess curve, it employs an distribution (or ratio of standard deviation) with !" !"#$$ ! and ! !" !"#$$ degrees of freedom: ! ! !"" ! !"" !"#$$ ! !" !"#$$ ! ! !"" !"#$$ ! ! !" !"#$$ (3 6) In the above equation, !"" !"#$$ is the sum of squared loess residuals !" !"#$$ i s the degree of freedom associated with the predicted curves and is the number of data points.

PAGE 44

44 Table 3 1. Some examples of discrepancies between premise count and the value from NO_RES_UNTS of NAL11P201302 Parcel_id premise_count NO_RES_UNTS 03812 007 000 2 1 04305 003 001 247 176 04314 102 096 3 1 04333 001 002 201 1 06014 008 002 3 4 06655 053 036 9 1 06655 200 000 173 15 06680 022 000 249 792 06680 023 000 254 20

PAGE 45

45 Figure 3 1. Process diagram of data preparation from the raw format to output results. Figure 3 2. A chart to quickly spot outliers.

PAGE 46

46 Figure 3 3. A o ne to many relationship between Parcel ID Premise ID tables and Premise ID monthly consumption data tables Figure 3 4 Process diagram of data use for 3D modeling

PAGE 47

47 CHAPTER 4 VISUALIZATIO N Land Use D ensity A 2D visual representation of land use density was generated using ArcMap. Projected coordinate system used for this rendering was the State Plane coordinate system ( NAD_1983_StatePlane_Florida_North_FIPS_0903_Feet ), and as Basemap layer, ESRI_Imagery_World_2D and ESRI_StreetMap_World_2D map service was used. The symbology of the Parcel layer was set to map the value of "du_ac" based on the graduated colors from green to red in 25 steps. Both services are connected to ArcGIS online server on http://services.arcgisonline.com/ArcGIS/rest/services The hi resolution residential density map generated by ArcMap provided an overall sense of the residenti al land use density of the city of Gainesville, FL, as shown in Figure 4 1. Each parcel was mapped to a specific color based on its calculated density value. The color legend of the density (DU/ac) values is presented in Figure 4 4 The figure 4 2 shows th e Cabana Beach Apartments and Rockwood Villas private condos areas on the map using ESRI_StreetMap_World_2D as Basemap E nergy Consumption per Residential Unit Density A 3D map representation of the average annual energy consumption per residential unit de nsity was modeled using ArcScene. The intensity of the energy consumption of each residential land parcel is represented by a spectrum of colors, from green to red to indicate low to high energy use. Additionally, the height of the column bar extruded from the parcel represents the overall unit density, expressed in DU/ac (see Figure 4 3) The color legends, both for density and ekWh values, are shown in Figure 4 4

PAGE 48

48 The 3D model provides a quick observation method to identify and to evaluate an entire or partial spatial area of interest. Information can be read and analyzed directly from the model including the parcel size, the parcel land use density by the heights o f the bar extruded from the parcel, and the parcel's energy consumption (see Fig ure 4 5, 4 6, and 4 7) Normal Distribution of the Energy Consumption A geospatial visualization of the normal distribution of residential average annual energy consumption val ues was rendered, to provide a better insight into the actual per parcel distribution of energy consumption values As shown in Figure 4 8 and a close up look as in Figure 4 10 the colors on the visualized map indicate the standard deviation values of the average energy consumption per land use density The or standard deviation values are represented in the color legend in Figure 4 9, and the curve on the right shows the normal distribution of the ekWh values ( average energy consumption values), with th e dotted vertical lines denoting the standard deviations With the geospatial mapping of the ekWh normal distribution values, it is much easier to identify the spatial areas with the average annual energy consumption based on their deviation values from th e mean

PAGE 49

49 Figure 4 1. Gainesville map generated by ArcMap color codes each residential land parcel based on its DU/ac density (map scale 1:40,000) Figure 4 2. The ArcMap rendering shows the land use density of Cabana Beach Apartment and Rockwood V illas complex (map scale 1:25,000)

PAGE 50

50 Figure 4 3. A 3D representation of average annual energy consumption per residential unit density of the City of Gainesville based on 2012 energy consumption data. Figure 4 4. The color legends of density in DU/ac (on the left) and avg ekWh (on the right)

PAGE 51

51 Figure 4 5 A closer look at southwest residential area, the corner between SW Archer road and I 75 in Gainesville, FL. Figure 4 6. A scene from the SW overlooking the NW area of the city.

PAGE 52

52 Figure 4 7. A detailed view of Haile Plantation subdivision in the SW area of Gainesville, FL Figure 4 8. The colors on the map show the standard deviation values of the average energy consumption per land use density.

PAGE 53

53 Figure 4 9. The legend, and the ekWh distribution curve. Figure 4 10. A close up look at the spatial distribution of the values of the ekWh.

PAGE 54

54 CHAPTER 5 ANALYSIS A nonlinear, nonparametric regression model based on Loess fitting was constructed with the pre processed data. The GRU residential energy consumption 2012 data (GRU_CY2012) was transformed, and the premises energy consumption values were aggregated to calculate the median ekWh of each parcel, as the dependent variable of the model. The density values, expressed in DU/ac as the independent variable of the model, were calculated from the GRU_CY2012 by linking and merging with the CAMA2012 and NAL2013. A bivariate plot of the median of ekWh per land use density can be generated based on the processed data as shown in Figure 5 1 Based on eye level observation of the plotted graph as in Figure 5 1 and 5 2, it can be seen that the density of the data points beyond 50 DU/ac is sparse. Further calculation of the data points' normal distribution within three standard deviations ! ! ! from the median (or 99.7%), shows that ( + 3 ) = 48.32 and can be rounded up to 50 DU/ac : n = 1218 = 15.69751 = 1.223255 ( + 3 ) = 48.31579 The same f itted loess curve with adjusted density of 50 DU/ac is shown in Figure 5 3. Examining the Loess R esiduals As in traditional regression methods, the purpose of plotting the loess residuals is to examine whether the smooth curve has sufficiently incorporated all feasible structure of the data points. From the equation 3 5, the loess residual values are defined as the

PAGE 55

55 difference between the each observed Y values, and each of the corresponding fitted values of X : ! ! ! ! (5 1) In finding the best fit parameter of the loess curve # (lambda) which is the degree of the loess polynomial: when # =0 the mean method is used, # =1 the linear equation method is used, and # =2, the quadratic equation method Among all the loess fitted residual plots in Figure 5 4 and 5 5 the latter with # =2 (Quadratic equation) shows the lowest residuals and thus the best fit. Arithmetic M eans and Median M ethods C omparison To better reflect the energy consumption per land parcel, the total energy consumption value is normalized against the number of the residential uni ts on the respective parcel. There are two ways to average the sum of the energy consumptions of all residential units, as defined in equation 3 1, over the number of the respective parcel. The median method is chosen because it better represents the avera ge value of the skewed distribution of the residential unit's energy consumptions. A comparison curves for both arithmetic means and median methods on the overall energy consumption per unit density is shown in Figure 5 6 Confidence Interval (CI) The 95% confidence interval (CI) bands, as described in Jacoby (2000) are plotted around the loess fitting curve of median ekWh (2012) per DU/ac, as shown in Figure 5 7. The confidence bands are calculated based on the curve's s tandard error: CI = 95% and ! ! 1.96 $ 2.

PAGE 56

56 Inferential A nalysis The loess curve graph s are generated using R statistical software version 3.0.2, and the fitted loess curve in Figure 5 3 has parameter information as follow : Call: loess(formula = y ~ x, data = data.frame(x = unit_acre, y = med_ekwh), span = 0.4205214, degree = 2) Number of observations: 1189 Equivalent Number of Parameters: 9.35 Residual Standard Error: 7436 Trace of smoother matrix: 10.32 Control settings: normalize: TRUE span : 0.4205214 degree : 2 family : gaugasian surface : interpolate cell = 0.2 To test the hypothesis ! by applying the equation 3 6, following values are calculated from the data and residuals tables: n = 1,189 !"" = 98 882 024 133 !"" !"#$$ = 65 125 684 362 and !" !"#$$ = 9.35 Using these values, the distribution value and its distribution value can be determined: ! ! !"" ! !"" !"#$$ ! !" !"#$$ ! ! !"" !"#$$ ! ! !" !"#$$ ! ! ! !" !!" !"# !"" ! !" !"# !"# !"# ! ! !" ! !" !"# !"# !"# ! !!"# ! !" ! ! !" !!"#$ and distribution's value = 6.81875E 98 < 0.0001

PAGE 57

57 It is apparent that the value of the distribution is far less than 0.0001, thus the ! of no statistic ally significant correlation between land use density and annual energy consumption per residential unit in the city of Gainesville, FL can be rejected, and at the same time it is possible to accept the alternative hypothesis ! Residential Energy Consu mption per House Type Additionally, from the GRU residentia l energy consumption 2012 and CAMA2012 databases, further data extraction can be performed to obtain the dataset for residential energy consumption per house type The scatterplot graph in Figure 5 8 and smoothed Loess curves in Figure 5 9 show the energy consumption of Single Family Attached (SFA), Single Family Detached (SFD) and Multi Family house types and the their distribution s based on the land use density are depicted in Figure 5 8 Follo wing records quantity per house type are used for the calculation, SFA = 4,900 units SFD = 36,965 units Multi Family = 33,801 units SFD has the highest energy consumption from all house types; its average energy consumption per unit is higher than overall per unit's energy consumption. On the other hand, SFD energy consumption continue s to decline as the land use density increases and it is starting to match that from SFA starting about 13 du/ac From all house types, Multi Family has the lowest unit consumption, and in this study, it is well represented through their normal distribution with 33,801 housing units.

PAGE 58

58 Case Studies Geospatial mapping of average annual energy consumption of the land use density provides a helpful and quick overview of the wh ole area of interest and reveals otherwise hidden information. The interactive nature of the 3D model creates a virtual panoramic image that can be panned, zoomed and rotated in 360 degree view. This allows the selected residential areas to be compared and studied further, from a bird's eye view down to the street level. Some areas in Gainesville were selected for case studies, as depicted in Figure 5 1 0 are as follow: Cabana Apartments, Phase I & II (Multi Family) Rockwood Villas (SFA) An SFA Condo in Hai le Plantation, Unit 15 An SFD in Haile Plantation, Unit 25 Ph. II T able 5 1 summarizes the studied residential areas and their density. For example, the subdivision of Rockwood Villas has 227 housing units E ach unit is in its own respective parcel with la nd size of 1,307 sqf t or 0.03 acre Since there is only one housing unit per parcel, it gains a relatively high land use density of ! !" = 33.33 DU/ac. For comparison, Cabana Beach I Apartment's parcel has a huge 30.3 acre land size area (because it is s hared across all units, includes all common areas ), but achieves a land use density of only !"# !" ! = 8.15 DU/ac. The 3D model of the average energy consumption per residential unit density in Figure 5 1 1 shows the differences in the land size of the parcels, between Rockwood

PAGE 59

59 Villas and Cabana Beach Apt I & II. From the eye-level observation of the model, the following information can be easily identified: The land size among the parcels. The parcels land-use density in DU/ac, indicated by the heights of the bars. The respective parcels average energy consumption, indicated by the color spectrum, from green (low consumption) to red (high consumption). The same as above, Figure 5-12 shows the close-up visualization of the areas of study in Haile Plantation, both the SFD area (development unit 25, Phase II) and SFA area (unit 15). Based on eye-level observation, the high-density areas, the lowdensity areas, and the parcel units that include their intensity of energy consumption can be identified. An aerial view of the two land parcels, Rockwood Villas and Cabana Beach Apt I & II can be seen in Figure 5-13, showing the layout of the parcels from above. Figure 514 shows the street-level view of the two land parcels. Subsequently, the aerial view of Haile Plantation, Unit 15 (an SFA) and Unit 25 Ph. II (an SFD) areas is shown in Figure 5-15 and Figure 5-16 their street-level view. Finally, a comparison of the average annual energy consumption per residential unit in those five studied areas is shown in Figure 5 -17. It can be seen, that the energy consumption of the SFD/Haile Plantation Unit 25 Ph. II is the highest among the other four parcels.

PAGE 60

60 Table 5 1. Overview of studies on building energy consumption in reviewed literatures. Name Parcel_id DU acres DU/ac LND_ SQFT TOT_ LVG_ AREA Cabana Beach Apts I 06680 022 000 247 30.3 8.15 1319432 322845 Cabana Beach Apts II 06680 023 000 252 25.2 10 1097712 299051 Rockwood Villas 06680 030 xxx 227 0.03 33.33 1307 1106 an SFD (Haile Plant.) 06860 252 046 1 0.54 1.85 23522 3052 an SFA (Haile Plant.) 06860 150 014 1 0.03 33.33 1307 1088

PAGE 61

61 Figure 5 1. Bivariate plot graph of median ekWh (2012) per DU/ac

PAGE 62

62 Figure 5 2. Fitted loess curve of median ekWh (2012) per DU/ac (up to 153 DU/ac)

PAGE 63

63 Figure 5 3. Fitted loess curve of median ekWh (2012) per DU/ac (limited to 50 DU/ac)

PAGE 64

64 Figure 5 4 Residuals of loess fitted median ekWh ( 2012 ) per land use density with =0 (on the left) and with =1 (on the right) Figure 5 5 Residuals of loess fitted median ekWh 2012 per land use density with =2.

PAGE 65

65 Figure 5 6 Comparison of e nergy consumption profiles normalized using arithmetic mean and median method.

PAGE 66

66 Figure 5 7 Median ekWh 2012 per DU/ac plo t t ed with 95% confidence interval bands.

PAGE 67

67 Figure 5 8 Res. energy consumption ( 2012 ) vs DU/ac scatterplot by house type. Figure 5 9 Residential energy consumption ( 2012 ) vs DU/ac by house type. !" #!!!!" $!!!!" %!!!!" &!!!!" '!!!!" (!!!!" !" $!" &!" (!" )!" #!!" #$!" #&!" #(!" *+,-" .*/01*2345"6*1"*+,-"$!#$"7/"84219:/*";*2/0<=">?*@"ABC/*"D=?*E" FG;" HC539G4605=" FGI"

PAGE 68

68 Figure 5 1 0 The selected areas for case study on the geospatial 3D model. Figure 5 1 1 Detailed view on selected area of study: Rockwood Villas and Cabana Beach.

PAGE 69

69 Figure 5 1 2 Detailed view of selected area of study: Haile Plantation (SFD and SFA) Rockwood Villas Cabana Beach Apt I & II Figure 5 1 3 Aerial views of Rockwood Villas and Cabana Beach Apt areas. [Source: Google Maps.]

PAGE 70

70 Rockwood Villas (typical SFA housing) Cabana Beach Apt I & II (typical Multi Family housing) Figure 5 1 4 Street level view of Rockwood Villas and Cabana Beach Apt areas. Haile Plantation Unit 15 (SFA housing) Haile Plantation Unit 25 Ph. II (SFD housing) Figure 5 1 5 Aerial view of Haile Plantation, Unit 15 and Unit 25 Ph. II areas. [Source: Google Maps.] Haile Plantation Unit 15 (typical SFA housing) Haile Plantation Unit 25 Ph. II (typical SFD housing) Figure 5 1 6 Street level view of Haile Plantation unit 15 and unit 25/II areas.

PAGE 71

71 Figure 5 17 Average annual energy consumption per studied residential units.

PAGE 72

72 CHAPTER 6 CONCLUSIONS AND DISCUSSION S Key Findings and Conclusions This thesis study attempts to find the correlation between land use density and annual energy consumption per residential dwelling unit within the city of Gainesville, Florida. The results provide a quantitative comparison and insights into the residential energy usage pattern based on the dwelling unit density, and house type. The application of statistical infer ence method helps to confirm the correlation that exists between annual energy consumption per residential unit and development density The 3D geospatial model helps to visualize the studied variables by mapping them to their respective geographical locat ions. The model provides not only a quick visual assessment of the analyzed data on spatial distribution (i.e. the size and location of the land parcels; their average annual energy consumption; and the parcel's land use density), but also its assessment a nd analytical insight can be applied to verify those results from the statistical analysis. Moreover, detailed investigation on the differences of the parcels' average annual energy consumptions is feasible through the additional visualization of each parc el's energy usage deviation values The results and conclusions presented in this thesis are limited to the use of the 2012 annual GRU residential energy consumption database as source data. The level of accuracy of the model will increase if more historic al energy consumption data is added, which may or may not change the overall profiles of the residential energy consumption per land use density. Based on the analysis result of the distribution of the energy consumption per house type, the study indicates that the overall residential

PAGE 73

73 average energy consumption pattern per land -use density in Gainesville, FL is unlikely to change much if more historical data is added to the model In conjunction with other publicly available data, i.e. CAMA 2012 and NAL 2013 it was possible to employ a data-mining algorithm to extract and dissect the data, and to construct the model for statistical analysis purposes as well as for the 3D geospatial modeling With the results of this thesis study, it can be concluded that: 1. Based on the statistical inference analysis, it is evident that a correlation exists between land -use density and annual energy consumption per residential dwelling unit within the city of Gainesville, Florida 2. The s tatistical regression model shows that there is a non-linear, non-parametric inverse relationship; up to a certain extent (about 15 -35 du/ac), with the increasing of land-use density, the level of energy consumption per residential dwelling unit is decreased. 3. Based on the 3D geospatial model, it c an be observed that land parcels with the high land -use density tend to have lower energy consumption compared to those parcels with lower land-use density. This characteristic can also be observed and verified in the normal distribution of energy consumpt ion model, as depicted in Figure 4-8 and 4 -10. 4. Based on the analysis of the energy consumption per house type, as shown in Figure 5-8 and 5 -9, the results suggest that, the SFD has the highest energy consumption per house type. This finding is also in accordance with the SFDs trait, which has a low land -use density value. Its average energy consumption per unit is higher than the overall average energy consumption per unit o f all residential housing stock in Gainesville, FL. the SFA average energy consumption is much lower than SFD. Aside from less significant fluctuations up to about 10 du/ac, the SFA values do

PAGE 74

74 not change much as the land use density increases. However, starting from around 13 du/ac, the SFD values match those from SFA Multi Family housing has the lowest energy consumption. 5. Based on the median averaged values of energy consumption per house type by land use density, the calculation yields the following results (in ekWh/[du/ac]): SFD = 17,300; SFA = 9,219; and Multi Family = 7 401. I n other words, o n average an SFD unit consumes 2.3 times as much energy as a Multi Family unit, and 1,88 times as much as an SFA unit. An SFA unit consumes only 1.2 times more energy than a Multi Family unit. Suggestions for Further Studies There are sever al areas in this study, which could be improved or extended in the future research. In particular : To increase sample size by adding more historical energy consumption data to the model to improve accuracy. To study in a more detailed manner the possible effect caused by the implementation of building codes and standards in the past on the residential energy consumption by comparing house energy consumption across building vintage. To further break down the seasonal residential energy consumption analysis into more sub dimensions, such as house type (to investigate whether there is any significant relationship exists between seasonal energy usage and house type) and seasonal energy usage per house vintage (to investigate whe ther there is any improvement i n home energy efficiency, due to new building codes etc.) To study the residential energy consumption related to Urban Heat Island effect. To study the residential energy consumption analysis based on Floor Area Ratio (FAR). To study the microclimatic ef fects, such as tree shades and canopy benefits on residential energy efficiency. To study the residential water consumption and its correlation with land use density.

PAGE 75

75 To study the potential GHG emissions related to Gainesville land use density. Discussi ons This thesis study provides empirical and quantitative analyses on the correlation between net land u se density and annual energy consumption per residential dwelling unit within the city of Gainesville, Florida. Based on the key findings and conclusion s in this study, additional analyses can be derived from the available datasets these include the analysis of residential energy c onsumption by building vintage, seasonal residential energy consumption and also future projected energy demand and forecasts in different scenarios could be developed. Although these supplementary analyses are formally not a part of the main objectives of this thesis, however they are interesting and can provide a further insight into various dimensions of the correlation betwe en net land u se density and annual energy consumption for further discussions and recommendations. Residential E nergy C onsumption by B uilding V intage The graph in Figure 6 1 show s the residential avg ekWh 2012 consumption plotted against the density ( DU/ac ). It also shows the comparison of residential buildings energy consumptions for buildings built before 1990 (orange line) and those built in year 1990 and thereafter (blue line). The prevailing wisdom suggests that buildings built in the 1990s and therea fter should consume less energy due to their energy efficiency requirements imposed by building codes, such as better insulation, window glazing, Energy Star rated equipment, etc. However, surprisingly the graph above shows that this assumption is reversed The assumption to respond to this deviation is that the average residential unit size (sq ft) has gradually increased since 1980s, as shown in Figure 6 2

PAGE 76

76 Since the energy consumption is only related to the heated area, the average heated area (htd area) is used in the calculation instead of average total floor area (heated area + non heated area). The pre 1930 buildings are not included since their numbers are sparse and therefore less representative. Seasonal Residential Energy Consumption Based on the GRU residentia l energy consumption 2012 data, further insight into the energy consumption patterns can be obtained. Figure 6 3 shows the seasonal residential energy consumption plotted against the land use density ( DU/ac ). Prior to aggregating the usage va lues of the residential units (premises) to determine the average usage per land parcel, the energy usage value in each record must be split and assigned proportionally or pro rated to its respective usage months (monthly binning method) based on its meter reading date. For example, a premise has a usage value = 1,250 kWh, meter reading duration is equal 30 days and meter reading date of May 20, 2012 ( !"#$ !"#$ !" ), and let monthly ekWh usage bins = ! where ! ! !" !" denoting the month number. This is calculated as: 1. Calculate the normalized daily usage: ! ! !"#$% !"#$%&'( ! ! !"# !" !" !!" !" ! !"# 2. If read date < duration: Add May 2012 pro rated usage to current month's bin (m=5): ! ! ! ! !"#$ !"#$ ! !" !!" !" ! ! !" !"" !! !" 3. If read date < duration: Then add the remaining pro rated usage to (m 1)'s bin: ! ! ! ! !"#$%&'( ! !"#$ !"#$ ! !" !!" !" ! ! !" !"# !" !" !

PAGE 77

77 4. In special cases, if read date > durat ion (e.g. for month May): 5 ! !"#$% Note that the index above indicates the premise index of a parcel (refer to equation 3 1). The subsequent step is to sum all the pro rated monthly usage bins of respective premise together. The final season al usage value of each premise is obtained by aggregating the monthly values based on following rule: Spring months = March, April, May Summer months = June, July, Aug Fall months = Sep, Oct, Nov Winter months = Dec, Jan, Feb Finally, the density of each p arcel is calculated based on equation 3 4 and its premise seasonal usage is aggregated based on the sample median method. The result in Figure 6 3 shows that the highest seasonal residential energy consumption in 2012 occurred during the summer months. On the other hand, in the winter months, the energy consumption is the lowest compared to the rest of the seasons. This seasonal energy consumption pattern is expected, where the energy consumption profile matches that of the City of Gainesville and Florida's seasonal energy consumption profiles. In Florida generally and in Gainesville specifically, the summer electricity peak demand is usually higher compared to winter peak demand. Energy load analysis for additional residen tial units in various scenarios Ass uming that there are 1,000 to 2,000 new residential units to be added to Gainesville's housing stock th ese units will add additional GWh of Net Energy Load ( NEL ) to the original GRU's forecasted energy demand. The potential energy

PAGE 78

78 consumption profiles of the new built residential units will depend on their land -use density. Table 6 -1 shows history and forecast of NEL and Seasonal Peak Demands, and was compiled from the GRUs 2013-Ten -Year Site Plan (GRU, 2013). Based on the data, the GRU predicted NEL for the year of 2014 up to 2022 graph can be plotted (see Figure 6-4). The NEL forecast line has a linear function: y = 7.33x 12838, and using this function, the total NEL in various additional residential units scenarios can be extrapolated. Based on the previous calculation of the Seasonal Residential Energy Consumption vs land-use density above, as shown in Figure 6-3 and Figure 6-5, the maximum median ekWh of both summer and winter consumption for density 0-15 du/ac and >15-50 du/ac can be determined. From the previous assumption above, a new development project will add a few thousand new residential units to Gainesvilles housing stock, and thus the additional energy loads as shown in various scenarios, including summer energy load and winter energy load for the additional housing units, as shown in Table 6-2 and 6 -3. Note that, to simplify the calculation, the density values are grouped to -15 and >15-50 respectively. Table 6-4 shows the additional energy loa ds of 1K-2K residential units, added to the GRUs forecast NEL, and based on the calculated numbers, the additional energy loads to GRUs previous forecasted NEL can be added. The graph in Figure 6-6 shows NEL in various scenarios: original NEL; NEL of 1K additional res units with 0-5 du/ac; NEL of 2K additional res units with 0-5 du/ac; NEL of 1K additional res units with >15-50 du/ac and NEL of 2K additional res units with

PAGE 79

79 >15 50 du/ac Based on the se different scenarios, a maximum of NEL that will add to the original GRU's forecasted energy demand is approximately 37.84 GWh

PAGE 80

80 Table 6 1. GRU's Historical and P rojected Net Energy Load (NEL) and Summer/Winter Peak Demand s ( MW ) 7"(, 80'()*$&#'())"%* 9(3(9$'. /(3(9$'. :5($)(;)"* 4122", 4122",*<"(=* >"2(&% /(3(9$'.* (5($)(;)"* ?$&'", ?$&'",*<"(=* >"2(&% @AAB CDA CAE FDE C@G BHA @AAF CDD CAG FB@ C@I BEE @AAH CDD CAG FCH C@I BGC @AAC CDD CAG FCF CB@ BC@ @AAE CDD CDD FGD CBD BCD @AAG CDA CHI FHE EDD F@D @AAI CAG EAI FCH EAH FCF @ADA CAG EDA FEA CGD FAI @ADD CAG CCF FFH CGB BED @AD@ CDA CCE FDH CE@ BBG @ADB HIG CHE FFD E@F BF@ @ADF HIG EDA FDC E@F BFB @ADH HIG EDD FDE EAD BFH @ADC HEH CIA FDI EA@ BFE @ADE HEH CIA F@D EA@ BFG @ADG HCD CEC F@@ CE@ BHA @ADI HBB CFG F@F CHE BHD @A@A HBB CFG F@H CHE BHB @A@D HBB CFG F@E CHE BHH @A@@ FHG HEB F@G HG@ BHE Data c ompiled from GRU's 2013 Ten Year Site Plan (GRU, 2013) Ta ble 6 2 Summer energy load f orecasts of seasonal residential energy load based on 0 15 and >15 50 du/ac du/ac max. med ekWh x 1,000 new residential units x 2,000 new residential units 0 15 5,468.23 5,468,230 ekWh 10,936,469.55 ekWh >15 50 2,072.99 2,072,990 ekWh 4,141,973.06 ekWh Ta ble 6 3 Winter energy load forecasts of seasonal residential energy load based on 0 15 and >15 50 du/ac du/ac max. med ekWh x 1,000 new residential units x 2,000 new residential units 0 15 4,899.61 4,899,610 ekWh 9,797,211.95 ekWh >15 50 1,494.22 1,494,220 ekWh 2,988,434.19 ekWh

PAGE 81

81 Table 6 4 NEL calculation table with 1K and 2K extra residential units added. year NEL NEL+1K =0 15 NEL+1K >15 50 NEL+2K =0 15 NEL+2K >15 50 2014 1,932 1,949.72 1,939.08 1,968.18 1,946.89 2015 1,938 1,957.06 1,946.41 1,975.51 1,954.22 2016 1,946 1,964.39 1,953.74 1,982.84 1,961.55 2017 1,953 1,971.72 1,961.08 1,990.18 1,968.89 2018 1,960 1,979.06 1,968.41 1,997.51 1,976.22 2019 1,967 1,986.39 1,975.74 2,004.84 1,983.55 2020 1,975 1,993.72 1,983.08 2,012.18 1,990.89 2021 1,982 2,001.06 1,990.41 2,019.51 1,998.22 2022 1,991 2,008.39 1,997.74 2,026.84 2,005.55

PAGE 82

82 Figure 6 1 Residential energy consumption ( 2012 ) by land use density, split by built year Figure 6 2 Average floor area (heated area) by built year. !" #!!" $!!!" $#!!" %!!!" %#!!" &!!!" &#!!" '!!!" '#!!" #!!!" $(&!" $('!" $(#!" $()!" $(*!" $(+!" $((!" %!!!" %!$!" !"#$# %&'()#*+,-# ,./#0)1#,-+,# ,-./0123453" 6788109,-./0123453:"

PAGE 83

83 Figure 6 3 Residential energy consumption ( 2012 ) by seasons.

PAGE 84

84 Figure 6 4 GRU's Historical and P rojected Net Energy Load (NEL) Figure 6 5 Summer/Winter residential energy consumption in ekWh vs. du/ac !"#$%& !"#'%& !"#(%& !"#)%& !"#*%& !"#+%& !"#,%& !"##%& $"%%%& $%!'& $%!(& $%!)& $%!*& $%!+& $%!,& $%!#& $%$%& $%$!& $%$$& $%$'& -./& 0123& !"#$%&'$()&*+,$-./0$12*&034'&05$ 456&73189:;18&<=39>9?2@A&

PAGE 85

85 Figure 6 6 Additional NEL in various scenarios.

PAGE 86

86 APPENDIX A LIST OF DATABASES, TABLES AND FIELDS Tables (Database) Fields 1. GRU_CY2012 customer_id premise_id premise_parcelid read_cycle rate_code service_point_id meter_type meter_units usage_value duration is_estimate is_cancellation BELNR date_to 2. GRU_PREMISES (PREC) premise_id premise_house_number premise_street_name premise_unit premise_city premise_state premise_zip_code premise_parcelid premise_type 3. NAL11P201302 ID, CO_NO, PARCEL_ID, FILE_T, ASMNT_YR, BAS_STRT, ATV_STRT, GRP_NO, DOR_UC, PA_UC, SPASS_CD, JV, JV_CHNG, JV_CHNG_CD, AV_SD, AV_NSD, TV_SD, TV_NSD, JV_HMSTD, AV_HMSTD, JV_NON_HMSTD_RESD, AV_NON_HMSTD_RESD, JV_RESD_NON_RESD, AV_RESD_NON_RESD, JV_CLASS_USE, AV_CLASS_USE, JV_H2O_RECHRGE, AV_H2O_RECHRGE, JV_CONSRV_LND, AV_CONSRV_LND, JV_HIST _COM_PROP, AV_HIST_COM_PROP, JV_HIST_SIGNF, AV_HIST_SIGNF, JV_WRKNG_WTRFNT, AV_WRKNG_WTRFNT, NCONST_VAL, DEL_VAL, PAR_SPLT, DISTR_CD, DISTR_YR, LND_VAL, LND_UNTS_CD, NO_LND_UNTS, LND_SQFOOT, DT_LAST_INSPT, IMP_QUAL, CONST_CLASS, EFF_YR_BLT, ACT_YR_BLT, TOT _LVG_AREA, NO_BULDNG, NO_RES_UNTS, SPEC_FEAT_VAL, MULTI_PAR_SAL1, QUAL_CD1, VI_CD1, SALE_PRC1, SALE_YR1, SALE_MO1, OR_BOOK1, OR_PAGE1, CLERK_NO1, SAL_CHNG_CD1, MULTI_PAR_SAL2, QUAL_CD2, VI_CD2, SALE_PRC2, SALE_YR2, SALE_MO2, OR_BOOK2, OR_PAGE2, CLERK_NO2, SAL_CHNG_CD2, OWN_NAME, OWN_ADDR1, OWN_ADDR2, OWN_CITY, OWN_STATE, OWN_ZIPCD, OWN_STATE_DOM, FIDU_NAME, FIDU_ADDR1, FIDU_ADDR2, FIDU_CITY, FIDU_STATE,

PAGE 87

87 FIDU_ZIPCD, FIDU_CD, S_LEGAL, APP_STAT, CO_APP_STAT, MKT_AR, NBRHD_CD, PUBLIC_LND, TAX_AUTH_CD, TWN, RNG, SEC, CENSUS_BK, PHY_ADDR1, PHY_ADDR2, PHY_CITY, PHY_ZIPCD, ALT_KEY, ASS_TRNSFR_FG, PREV_HMSTD_OWN, ASS_DIF_TRNS, CONO_PRV_HM, PARCEL_ID_PRV_HMSTD, YR_VAL_TRNSF, EXMPT_01, EXMPT_02, EXMPT_03, EXMPT_04, EXMPT_05, EXMPT_06, EXMPT_07, EXMPT_08, EXMPT_09, EXMP T_10, EXMPT_11, EXMPT_12, EXMPT_13, EXMPT_14, EXMPT_15, EXMPT_16, EXMPT_17, EXMPT_18, EXMPT_19, EXMPT_20, EXMPT_21, EXMPT_22, EXMPT_23, EXMPT_24, EXMPT_25, EXMPT_26, EXMPT_27, EXMPT_28, EXMPT_29, EXMPT_30, EXMPT_31, EXMPT_32, EXMPT_33, EXMPT_34, EXMPT_35, EXMPT_36, EXMPT_37, EXMPT_38, EXMPT_39, EXMPT_40, EXMPT_80, EXMPT_81, SEQ_NO, RS_ID, MP_ID, STATE_PAR_ID, SPC_CIR_CD, SPC_CIR_YR, SPC_CIR_TXT 4. Address (CAMA 2012) OBJECTID, SHAPE, ADDRESSID, PARCEL, FULLADDR, ADDRNUM, ROADPREDIR, ROADNAME, ROADTYPE, UNIT, CITY, STATE, ZIP, ESN, MSAGCOMM, LANDMARK, E911TYPE, APR_NO, ADDRCOMMENTS, ISSUED_DATE, STATUS, LRSIDE, EDITOR_NAME, EDIT_DATE, ADDRFLAG, CAPTUREMETH, MAPSAG_EXC, OLDADDRESSID, USNGCOORD, NO_MSAG, C1_EXCEPTION, LASTUPDATE, LASTEDITOR, Field0 5. Par cel (CAMA 2012) OBJECTID, SHAPE, CODE, PIN, REDEVELOPMENT, X, Y, LOT, LINK_ACPA, ACCURACY, SOURCE, ACRES_CALC, ACRES_CAMA, EDITOR_NAME, EDIT_DATE, LASTUPDATE, LASTEDITOR, SRCREF, OWNTYPE, Vincent_ACPA_parcel_AREA, SHAPE_Length, SHAPE_Area 6. Building (CAMA 2012) OBJECTID, parcel, seq, style, effyr, beds, baths, buse, qual, extw1, extw2, actyr, cond, roof, ac, heat, fuel, stories, floor1, floor2, intw1, intw2, totarea, htdarea 7. Code01 (CAMA 2012) OBJECTID, type, code, description

PAGE 88

88 APPENDIX B CAMA 2012 TABLE METADATA: BUILDING # Name Alias Description 1 parcel Parcel Number Parcel Identification Number 2 seq Sequence Sequence Number Used To Indicate Number Of Buildings 3 style Style Building Style 4 effyr Effective Year Year Improvement Was Built As Indicated By Its Condition. May Not Be The Actual Year Built. 5 beds Bedrooms Number of Bedrooms 6 baths Bathrooms Number of Bathrooms 7 buse Building Use Building Use. See BUSE Fields In GIScode01 Table For Details. 8 qual Quality Quality Level Of Building 1=Lowest, 3=Average, 6=Highest 9 extw1 Exterior Wall 1 Primary Material Used In Exterior walls. See EXTW Fields in GIScode 01 Table For Details. 10 extw2 Exterior Wall 2 Secondary Material Used In Exterior Walls. See EXTW Fields in GIScode 01 Table For Details. 11 actyr Actual Year Actual Year Improvement Was Built 12 Cond Condition Code To Indicate If Additional Depreciation Is Warranted Due To Special Condition. See OBSCC Fields In Code01 Table For Details. 13 roof Roof Type of Roof Covering. See RFCVR Fields In Code02 Table. 14 ac AC Type of Air Conditioning System. See AC Fields In Code01 Table 15 heat Heat Type of Heating System. See HTG Fields In Code01 Table. 16 fuel Fuel Type of Heating Fuel. See HTFL Fields In Code01 Table. 17 stories Stories Number of Building Stories.

PAGE 89

8 9 18 floor1 Floor 1 Primary Type of Flooring Material. See FLR Fields In Code01 Table. 19 f loor2 Floor 2 Secondary Type of Flooring Material. See FLR Fields In Code01 Table. 20 intw1 Interior Wall 1 Primary Type of Material Used in Interior Walls. See INTW Fields In Code01 Table. 21 i ntw2 Interior Wall 2 Secondary Type of Material Used in Interior Walls. See INTW Fields In Code01 Table. 22 t otarea Total Area Total Area (Heated + Sub Areas) o f Building in Square Feet. 23 h tdarea Heated Area Area of Building in Square Feet Considered To Be Enclosed And Subject To Heating And Cooling

PAGE 90

90 APPENDIX C LIST OF IMAGES Gainesville Land Use Density Map

PAGE 91

91 3D Geospatial M odel of A verage ekWh per L and U se D ensity

PAGE 92

92 3D Geospatial Normal Distribution of the Energy Consumption

PAGE 93

93 LIST OF REFERENCES ACPA. (2012). GIS information and services Retrieved Aug 2013, from Alachua County Property Appraiser: http://kate.acpafl.org/ServiceCenter/gis_main.aspx Bertoldi, P., Hirl, B., & Labanca, N. (2012). Energy Efficiency Status Report 2012. European Commission Joint Research Centre, Institute for Energy and Transport. Publications Office of the European Union. doi:10.2788/37564 City Of Gainesville, FL. (2012). About Gainesville, Florida Retrieved Aug 2013, from City of Gainesville, Florida: http://www.cityofgainesville.org/VISITOR/AboutGainesville/GeneralFacts/tabid/34 1/Default.aspx Cleveland, W. S., & Devlin S. J. (1988, 09). Locally Weighted Regression: An Approach to Regression Analysis by Local Fitting. Journal of the American Statistical Association, 83 (403), 596 610. Doherty, M., Nakanishi, H., Bai, X., & Meyers, J. (2009). Relationships between form, m orphology, density and energy in urban environments. Global Energy Assessment Working Paper, CSIRO Sustainable Ecosystems, Canberra, Australia. Echenique, M., Hargreaves, A., Mitchell, G., & Namdeo, A. (2012). Growing Cities Sustainably: Does Urban Form Re ally Matter? Journal of the American Planning Association, 78 (2), 121 137. doi:10.1080/01944363.2012.666731 EERE. (2012, Mar). Buildings Sector Energy Consumption Retrieved Aug 2013, from Buildings Energy Data Book, Energy Efficiency & Renewable Energy, U .S. Department of Energy: http://buildingsdatabook.eren.doe.gov/TableView.aspx?table=1.1.3 EERE. (2013, Mar 27). Florida Energy Consumption Retrieved Jul 2013, from Buildings Energy Data Book, Energy Efficiency & Renewable Energy, U.S. Department of Energ y: http://apps1.eere.energy.gov/states/consumption.cfm/state=FL EIA. (2012, Apr). Use of Electricity Retrieved Aug 2013, from Energy Information Administration, U.S. Department of Energy: http://buildingsdatabook.eren.doe.gov/ChapterIntro1.aspx Ewing, R., & Rong, F. (2008). The Impact of Urban Form on U.S. Residential Energy Use. Housing Policy Debate, 19 (1). doi:10.1080/10511482.2008.9521624 Florida Public Service Commission. (2012). Review of the 2012 Ten Year Site Plans. For Florida's Electric Utilities. Florida Public Service Commission. Tallahassee, FL: Florida Public Service Commission.

PAGE 94

94 FRCC. (2012). 2012 Regional Load & Resource Plan. Tampa. Retrieved from http://www.psc.state.fl.us/utilities/electricgas/docs/FRCC_2012_Load_Resource_ Plan.pdf GRU. (2013). 2013 Ten -Year Site Plan. Submitted to: The Florida Public Service Commission. Gainesville Regional Utilities. Haas, R., Auer, H., & Biermayr, P. (1998). The impact of consumer behavior on residential energy demand for space heating. Energy and Buildings, 27, 195 -205. doi:10.1016/S0378-7788(97)00034-0 Heiple, S., & Sailor, D. (2008). Using building energy simulation and geospatial modeling techniques to determine high resolution building sector energy consumption profiles. Energy and Buildings 40 1426 1436. doi:10.1016/j.enbuild.2008.01.005 Holden, E. (2004). Ecological footprints and sustainable urban form. Journal of Housing and the Built Environment, 19(1), 91-109. doi:10.1023/B:JOHO.0000017708.98013.cb Holden, E., & Norland, I. (2005). Three Challenges for the Compact City as a Sustainable Urban Form: Household Consumption of Energy and Transport in Eight Residential Areas in the Greater Oslo Region. Urban Studies, 42(12), 21452166. doi:10.1080/00420980500332064 Howard, B., Parshall, L., Thompson, J., Hammer, S., Dickinson, J., & Modi, V. (2012). Spatial distribution of urban building energy consumption by end use. Energy and Buildings, 45, 141 -151. doi:10.1016/j.enbuild.2011.10.061 Jacoby, W. (2000). Loess: A nonparametric, graphical tool for depicting relationships between variables. Electoral Studies, 19(4), 577613. doi:10.1016/S02613794(99)00028-1 Kaza, N. (2010). Understanding the spectrum of residential energy consumption: A quantile regression approach. Energy Policy, 38(11), 6574-6585. doi:10.1016/j.enpol.2010.06.028 Kraak, M. (2003). Geovisualization illustrated. ISPRS Journal of Photogrammetry & Remote Sensing, 57, 390 -399. doi:10.1016/S0924-2716(02)00167-3 Larivire, I., & Lafrance, G. (1999). Modelling the electricity consumpti on of cities: effect of urban density. Energy Economics, 21, 53 -66. doi:10.1016/S01409883(98)00007-3 Min, J., Hausfather, Z., & Lin, Q. (2010). Statistical Model of Residential Energy Use Characteristics for the United States. Journal of Industrial Ecology, 14(5). doi:10.1111/j.1530 -9290.2010.00279.x

PAGE 95

95 Newman, P., & Kenworthy, J. (1989). Cities and Automobile Dependence: An International Sourcebook. Aldershot, England: Gower. Newton, P., Tucker, S., & Ambrose, M. (2000). Housing form, energy use and greenho use gas emissions. In K. Williams, E. Burton, & M. Jenks (Eds.), Achieving Sustainable Urban Form (pp. 74 84). London: Spon Press, Taylor and Francis Group. Norman, J., MacLean, H., & Kennedy, C. (2006, Mar). Comparing High and Low Residential Density: Lif e Cycle Analysis of Energy Use and Greenhouse Gas Emissions. Journal Of Urban Planning And Development 10 21. doi:10.1061/(ASCE)0733 9488(2006)132:1(10) Owens, S. (1990). Land Use Planning for Energy Efficiency. In J. B. Cullingworth (Ed.), Energy, Land, and Public Policy (pp. 53 98). New Brunswick, N.J., U.S.A: Transaction Publishers. Randolph, J. (2008). Comment on Reid Ewing and Fang Rong's "The impact of urban form on U.S. residential energy use". Housing Policy Debate, 19 (1). doi:10.1080/10511482.2008 .9521626 Safirova, E. A., Houde, S., & Harrington, W. (2007). Spatial Development and Energy Consumption. Resources For the Future, Discussion Papers, 2007. Staley, S. (2008). Missing the forest through the trees? Comment on Reid Ewing and Fang Rong's "the impact of urban form on U.S. residential energy use". Housing Policy Debate, 19 (1), 31 43. doi:10.1080/10511482.2008.9521625 Treloar, G., Love, P., Faniran, O., & Iyer Raniga, U. (2000). A hybrid life cycle assessment method for construction. Construction Management and Economics, 18 (1), 5 9. doi:10.1080/014461900370898 U.S. Census Bureau. (2011). ACS Demographic and Housing Estimates Retrieved Aug 2013, from American Fact Finder Community Facts: http://factfinder2.census.gov/faces/tableservices/jsf/pages/productview.xhtml?pid =ACS_11_5YR_DP05 Wilson, B. (2013). Urban form and residential electricity consumption: Evidence from Illinois, USA. Landscape and Urban Planning, 115 62 71. doi:10.1016/j.la ndurbplan.2013.03.011

PAGE 96

96 BIOGRAPHICAL SKETCH Djundi T jindra received his Diplom Ingenieur Elektrotechnik Automatisierungs technik degree from MŠrkische FH, H agen, Germany ( currently FH SŸdwestfalen). Professionally, h e has 20+ years of experience in both i nformation t echnology and creative fields. Djundi has lived and work in Europe, Asia and the United States He speak s fluently several languages, including Indonesian, German, English and Chinese. Djundi is interested in how i nformation t echnology can contribute to the sustainable design study in the built environment. In 2012, he pursued this interest by enrolling in the Master's in Sustainable Design program College of Design, Construction and P lanning at University of Florida. He hopes that the research of this thesis c an unfold into further studies. He is a photographer in his free time.