Brittany Zwicker
Ripe for an Energy Efficiency Retrofit? An Analysis of Residential Energy
Consumption in Somerville
Data Sources
M: drive> State> Mass GIS> Census 2000> CEN2K_B_POP_AGE_GEN.dbf (Senior)
M: drive> State> Mass GIS> Census 2000> CEN2K_BG_HOUS_RENT_STATS.dbf
(percent renter occupied)
M: drive> State> Mass GIS> Census 2000> CEN2K_BG_INCOME_HH.dbf (Median
income)
M: drive> State> Mass GIS> Census 2000> CEN2K_BG_HOUSING_AMEN_TEN.dbf
(fuel source)
M: drive> State> Mass GIS> Census 2000> CEN2K_BG_HOUS_STRUCT_AGE.dbf
(home year built)
M: drive>City>Somerville>Assessor> Parcels_03_2009.shp (parcel data)
Project Background
As the United States works to reduce its greenhouse gas emissions and fossil fuel
dependence, the role of residential energy efficiency in reducing energy consumption
must be recognized. In 2006, residential buildings were responsible for 37 percent of the
country’s electricity consumption1 and 2,236 million metric tons of carbon dioxide
emissions.2 Most of the energy consumed in residential buildings is used for lighting,
heating and cooling services. The high level of energy consumption in the U.S.’s
residential building sector is largely a result of a national housing stock that does not
efficiently consume energy.
For a municipal government interested in promoting residential energy efficiency, it is
important to know how much energy is being consumed. Unfortunately, utility data
documenting each household’s energy use is sometimes impossible to obtain.
Consequently, this analysis relied on a statistical regression model that was developed by
Tufts University Master’s student, Neil Veilleux, using the national Residential Energy
Consumption survey (http://www.eia.doe.gov/emeu/recs/) to predict the influence of
several socioeconomic variables on residential energy consumption. The model used
coefficients for the following variables: percent of homes built per decade, percent of
rental units, percent of seniors, median household income, heating fuel type, and average
number of people per household. Somerville has an old housing stock, a high renter
population, and a relatively low median income, all of which has an effect on its energy
consumption.
GIS is an ideal medium for this kind of analysis for two reasons. First, because it requires
that the residential units are isolated from the total buildings and then that they are
1
U.S. Department of Energy. (n.d.b). Carbon Dioxide Emissions for U.S. Buildings. Retrieved March 26, 2010, from
Buildings Energy Databook:
http://buildingsdatabook.eren.doe.gov/TableView. aspx?table=1.4.1
2 U.S. Department of Energy. (n.d.b). Carbon Dioxide Emissions for U.S. Buildings. Retrieved March 26, 2010, from
Buildings Energy Databook:
http://buildingsdatabook.eren.doe.gov/TableView. aspx?table=1.4.1
spatially aggregated to the census block group level, and second because it allows the
results of the analysis to be visually displayed in a map that can be shared with multiple
stakeholders.
Process and Findings
The data sources for this analysis were the 2000 U.S. Census, the Somerville Assessor’s
Database, and MassGIS’ Somerville census block group (CBG) layer, which were chosen
for their public availability and broad accessibility. ArcGIS was used to isolate the
Somerville parcels that were zoned “residential.” The Assessor’s data also had to be
aggregated to the CBG level using a point to polygon spatial join in ArcGIS:
1. Selecting by attribute, where class = residential.
2. Make parcels points by “Feature to point” tool
3. Do a point to polygon join, asking to see averages.
4. Pull data from census tables and parcel data averages for each block group
into an excel spreadsheet, retaining the LOGRECNO identifier.
At first I attempted to do all of the math using the spreadsheets in ArcGIS but it soon
became apparent that that platform did not have the capacity to do what I wanted. Since
the data for each of the variables I needed was in a different Census table, there were
dozens of columns in the ArcGIS which made it difficult to work with. I later found out
that there is a tool called Hawth’s Analyst tools that allows one to delete multiple
columns, but I had already imported my data into Excel.
Using Census information joined to MassGIS’ census block group shape file I derived the
following maps of the variables used in the analysis:
Percentage renters – homes that are renter occupied/total occupied
Median household income
Once in Excel, the first column contained the LOGRECNO number for each census block
group as the first column, and then each sequential column represented the data for each
for the variables for each CBG. Each of the variables was multiplied by its respective
energy consumption regression coefficient for each CBG, two variables called heating
degree days and cooling degree days which control for Somerville’s climate, a New
England variable to account for factors such as the region’s unique fuel mix, and a
control variable to account for the fact that the regression model only explains 57 percent
of the variability in residential energy consumption. All of the variables were multiplied
together and then divided by the average housing unit square footage for the relevant
block group to produce a normalized estimation of residential energy consumption in
kBtu/sqft/year for each CBG. (Excel file of analysis is also attached)
The resulting Energy Usage Index (EUI) for Somerville showed that its homes consume
an average of 33 kBtu/sqft/year, with a minimum consumption of 4.8 kBtu/sqft/year and
a maximum of 52 kBtu/sqft/year. Homes built before 1960, which includes 82 percent of
Somerville’s housing stock, had a very high effect on energy consumption. The overall
distribution of variation in residential energy consumption by CBG is shown in the
“Annual Residential Energy Consumption” map at the top right.
Note: This analysis does have some data quality limitations. In some cases, the Census
reported that a home used no fuel for heat, and some reported using coal or another
source not specifically accounted for in the regression model, so those were all grouped
into an “other fuel” category. Also, although the Census reports the actual year a home
was built, some homes have already been retrofitted for energy efficiency or otherwise
upgraded and so do not consume energy at the same rate as their contemporaries. These
kinds of case-by-case issues may have an unaccounted-for effect on energy consumption.
Implications for Energy Efficiency
This model has great potential as a tool for promoting widespread residential energy
efficiency. If the data limitations associated with the model can be solved, the next step is
to verify its accuracy using real utility data. If it could be vetted, this analysis has
potential as a tool to circumnavigate the difficulty of accessing utility data, which would
allow Somerville or other cities to understand residential energy consumption patterns
and develop targeted energy efficiency programs to reduce energy consumption and
greenhouse gas emissions.
For example, if the city can determine a baseline of residential energy consumption for
each census block group (CBG), their proposed retrofit program’s effectiveness can be
tracked and a specific reduction goal for the program can be set. A community like
Somerville could also take the results of this analysis and look at the impact of specific
variables on the city’s residential energy consumption. As mentioned before, the city has
an exceptionally high rental rate, an old housing stock, and many houses use inefficient
fuel oil for heating. Knowing how much each of these variables affect energy
consumption will allow the city to target certain homes or block groups for extra support
through the retrofit process.
The advantage of having energy consumption aggregated to the census block group level
is that that the CBG contains a convenient number of households for focused,
community-based energy efficiency retrofit initiatives. Neighborhoods with the highest
annual energy consumption can receive targeted outreach from the city, neighbors can
share information among themselves, and many small retrofit projects can be “bundled”
together to reach a scale that would interest a contractor.
Finally, this analysis model’s use of GIS underscores the importance of collecting quality
data that can be widely applied. Many cities and states still do not maintain GIS data that
can be easily accessed and used for analyses such as these.
Bibliography
Veilleux, Neil. (2010).Energy Mapping for Community Energy Efficiency Initiatives.
Tufts University Master’s Thesis - Department of Urban and Environmental Policy and
Planning.
The author does utilizes energy mapping and a statistical regression model to determine
how energy-use is tied to particular socioeconomic factors. The models use information
from the Residential Energy Consumption Survey, census data, and assessors data. The
results showed that larger homes, those with higher income, and the elderly tend to
consume more energy.
Etzion, B.A. Portnov, E. Erell, I. Meir, D. Pearlmutter. (2001). An open GIS framework
for recording and analysing post-occupancy changes in residential buildings — a
climate-related case study. Building and Environment, 36, 1075–1090.
The authors of this article studied the structural changes made to residential buildings
once they were occupied. They assumed that those structural changes were made to
compensate for climate-related inadequacies in the home and the type of structural would
reflect the particular inadequacy in the home. A coding system was developed that used
GIS to record the type of changes made in each home. The results showed that the
modifications made were dependent on the orientation and level of the building the
housing unit occupied. The implications for my study are that even parcel-level analysis
may not be of a high enough resolution to make assumptions for the specific energy
efficiency upgrades.
Bennett, M., & Newborough, M. (2001). Auditing energy use in cities. Energy Policy,
29(2), 125‐134.
The authors of this article discuss the importance of city-scale energy efficiency efforts
and of auditing a city’s energy consumption. It also discusses the ways to break down the
sectors in a city (industrial, commercial, residential, etc) to prioritize the audit processes.
The authors also explain the advantages and disadvantages of top-up versus bottom-down
audits and suggest using a hybrid of the two.
The U.S. DOE tables are very useful for determining trends and projected changes in
carbon dioxide emissions for specific sectors, which is good for providing support for the
argument that energy efficiency needs to be a priority for city governments.
U.S. Department of Energy. (n.d.b). Carbon Dioxide Emissions for U.S. Buildings.
Retrieved March 26, 2010, from Buildings Energy Databook:
http://buildingsdatabook.eren.doe.gov/TableView.aspx?table=1.4.1
U.S. Department of Energy. (n.d.b). Carbon Dioxide Emissions for U.S. Buildings.
Retrieved March 26, 2010, from Buildings Energy Databook:
http://buildingsdatabook.eren.doe.gov/TableView.aspx?table=1.4.1