To avoid the worst possible effects of climate change, either greenhouse gases must somehow be removed from the atmosphere or we must drastically reduce the amount of these gases entering the atmosphere. Mitigation is the term used to describe “measures to reduce the amount and speed of future climate change by reducing emissions of heat-trapping gases or removing carbon dioxide from the atmosphere” are (National Climate Assessment). The majority of greenhouse gas emissions is due to the burning of fossil fuels for energy, so any climate change mitigation effort must begin with energy-related CO2 emissions.
Every year the EPA releases a detailed national greenhouse gas emission inventory, but there is no corresponding inventory for the state of Georgia. However, the U.S. Energy Information Administration (EIA) does release statewide energy-related CO2 emission numbers for total emissions and five subareas: commercial, electric power, residential, industrial, and transportation. In 2016 (the latest year available) Georgia’s total emissions were 136.2 million metric tons of CO2, or about 29,000 pounds per person.
State, regional, and local mitigation efforts could be much more successful if we had any knowledge about the spatial distribution of emissions across the state. To fill this gap we have developed a set of basic allocation models to assign the different categories of EIA statewide emissions to the 1,969 small-area census tracts across the state. The result is a statewide, neighborhood-level, sector-specific, set of energy CO2 estimates. These values should be regarded as benchmark estimates that can provide a starting-point for mitigation, but they cannot substitute for a full local or regional greenhouse gas inventory.
The allocation methodology makes use of the best available small-area proxies for emissions in each of the sectors. For employment-related emissions in the commercial and industrial sectors, the analysis begins with the 2016 Census LODES datasets, which provide employment estimates at the Census block level. We aggregate two-digit NAICS employment to match the EIA commercial and industrial aggregations, then aggregate block-level employment to census tracts. After calculating each tract’s share of statewide employment industrial and commercial employment we allocate statewide emissions in each sector according to tract employment shares.
For residential emissions we use a more complex model since we have better energy-related data for sub-state areas. We begin with a 2015-2018 Census American Community Survey Public-Use Microdata Sample (PUMS) dataset of individual household electricity, natural gas, and fuel oil expenditures. We aggregate the individual household expenditures (and several other items) in each category to the lowest available level of PUMS geography: the Public Use Microdata Area (PUMA), which has a minimum population of 100,000.
The next step is fitting a PUMA-level multiple regression model to predict energy use in the three categories based upon the PUMA’s mean values for household income, number of rooms, units in structure, year built, and house value.
Finally, we use the same variables at the census tract level to predict mean tract-level residential energy expenditures, and we multiply the mean value times the number of households in each tract. To partially account for the differences between the larger PUMAs and the smaller tracts (which have different means and standard deviations for the two levels), we fit the PUMA model to standardized Z values and apply the prediction model to tract-level standardized Z values.
For the transportation model we, again, use a simpler allocation model based upon a single variable: tract-level commuting time to work. We sum all commuting minutes within each tract, determine the tracts’ shares of statewide commuting minutes, and allocate statewide transportation emissions based upon tract shares. One major weakness of this approach is that all transportation emissions are assigned to residents’ place of residence. In the future we may alter the model to assign approximately half of emissions to place of work by using the LODES origin-destination commuting dataset.
The preview map below shows total tract-level emissions per acre, which is necessary since tracts vary widely in size.
Preview maps are also available for the individual sectors:
Commercial emissions per acre
Industrial emissions per acre
Residential emissions per acre
Transportation emissions per acre
The full tract-level emissions dataset is available for download as a zipped GIS shapefile or text csv file. The length of variable names in shapefiles is limited. These are the shapefile variables added to the original Census cartographic boundary shapefile with descriptions.
acres: Area of tract in acres pop: Population hh: Households totemp: Total employment reselcm: Residential electricity emissions resfulm: Residential heating fuel emissions resemi: Residential emissions tranemi: Transportation emissions comemi: Commercial emissions indemi: Industrial emissions totemi: Total emissions empmprm: Employment emissions per employee rsmprpr: Residential emissions per capita emiprcr: Emissions per acre cmmprcr: Commercial emissions per acre indmprc: Industrial emissions per acre rsmprcr: Residential emissions per acre trnmprc: Transportation emissions per acre