Household Vehicle Carbon Emissions in California Cities

Transcription

1 Salon, Cook, and Williams 1 Household Vehicle Carbon Emissions in California Cities Submission Date: August 1, 2013 Word Count: 3928 plus 10 tables and figures Deborah Salon (corresponding author) Institute of Transportation Studies University of California, Davis 1605 Tilia Street Davis, CA ddsalon@ucdavis.edu Jonathan Cook Agricultural and Resource Economics Department University of California, Davis Davis, CA jcook@ucdavis.edu Jeffrey Williams Agricultural and Resource Economics Department University of California, Davis Davis, CA williams@primal.ucdavis.edu

2 Salon, Cook, and Williams 2 ABSTRACT Estimating jurisdiction-level carbon emissions from the transport sector is challenging for the simple reason that vehicles are mobile. For criteria air pollutant emissions estimation, the emissions counted are those that are physically emitted in the relevant geographic area. This makes sense because criteria air pollutants have localized health effects. Since carbon emissions have global rather than local environmental consequences, however, it is less obvious which transport emissions should count toward the emissions inventory for a particular jurisdiction. Thus, jurisdiction-level transport greenhouse gas emissions inventories are done in a variety of ways and are generally not comparable to one another. This paper reports on a project that calculated household vehicle greenhouse gas emissions inventories for eight cities in California for the years 2000 and Common data sources and methods were used for all cities, resulting in inventories that are comparable across time, across jurisdictions, and also between zip codes within a jurisdiction. These emissions estimates are discussed and compared to official inventory estimates.

3 Salon, Cook, and Williams 3 INTRODUCTION AND LITERATURE REVIEW Estimating emissions from vehicles for small geographic areas is challenging for the simple reason that vehicles are mobile. There are many choices that must be made, and it is not obvious which vehicles should be counted in the estimate of emissions associated with a particular geographic area. For criteria air pollutant emissions estimation, the emissions counted are those that are physically emitted in the relevant geographic area. This makes sense because criteria air pollutants have localized health effects. However, the disadvantage of this is that to the extent that emissions in a particular area come from pass-through traffic, localities have little power to reduce them. Since greenhouse gas emissions have global rather than local environmental consequences, it is reasonable to count emissions in a greenhouse gas emissions inventory from the vehicles that a local government might be able to impact through policy instruments whether or not they physically occurred within jurisdiction boundaries. To this end, the current recommended method for conducting transportation emissions inventories at the scale of the jurisdiction is to use a regional travel model and count emissions on a trip basis, allocating half of the emissions from each trip to the origin jurisdiction and half to the destination (1). This requires having access to a reliable travel model, however, which not all jurisdictions have. In light of this, alternative emissions allocation methods have been discussed in the literature (2,3,4). These include allocating regional transport emissions according to consumption in the jurisdictions and allocating emissions according to vehicles registered in the jurisdiction. Many jurisdictions have recognized this inherent ambiguity as to which method is most appropriate, and have made different choices of method to employ. As a result, jurisdiction-level transport greenhouse gas emissions inventories are done in a variety of ways and are not comparable to one another. Table 1 provides official emissions transportation inventory information for seven of the eight cities highlighted in this paper (the eighth city has not released an official greenhouse gas emissions inventory). As is apparent from the Notes column, different cities have taken different approaches to obtain their official city estimates of transportationrelated carbon emissions. While comparability of emissions inventories across jurisdictions is not necessary for tracking change over time within a jurisdiction, it is useful to be able to compare emissions across cities to identify jurisdiction characteristics and policies that may lead to lower emissions profiles for household vehicles. This paper reports on a project that calculated household vehicle greenhouse gas emissions inventories for eight cities in California for the years 2000 and 2005 following the emissions allocation method suggested in (2). Common data sources and methods were used for all cities, resulting in inventories that are comparable across time, across jurisdictions, and also between zip codes within a jurisdiction. This provides a useful starting point for beginning to understand some of the factors that may cause one jurisdiction to have lower carbon emissions from household vehicles than another.

4 Salon, Cook, and Williams 4 Table 1: Official City Emissions Inventory Estimates for Household Vehicles/Transportation City Official City Emissions Inventory Estimates (MT CO2) Notes s Berkeley , (5) , Davis (6,7) 75, , Counts cars and light-duty trucks driving within the city limits Fresno Los 25 million Based on rough scaled estimate of Angeles (8) Sacramento (9) San Francisco (10,11) Stockton (12) Ventura (13) 1. 1,849, ,716, ,842, ,910,402 transport fuel use Counts all on-road vehicles driving within the city limits Based on a Fehr and Peers revised estimate of VKT 1,834, (1) inventory counts all on-road vehicles driving within the city limits 1990 (2) and 2010 inventories count VKT from private vehicle trips originating and/or terminating within the city limits. Only half of VKT counted for trips with only one in-city endpoint. Pass-through traffic is not included. 829,326 1,123, Counts VKT from on-road transportation trips originating and/or terminating within the city limits. Only half of VKT counted for trips with only one in-city endpoint. Pass-through traffic is not included emissions a backcast 380, Counts all vehicles driving within the city limits and all other transport modes. DATA AND METHODS The method employed here to estimate household vehicle carbon emissions is a multi-step process, incorporating data from the US Census, the California Department of Motor Vehicles, the California Bureau of Automotive Repair s SmogCheck program, and the US Environmental Protection Agency. The calculation is straightforward, given by Equation 1. To clearly describe how the data are used to perform this calculation, this section of the paper is divided into subsections that identify the data used to estimate each piece of Equation 1. Equation 1: CO2 /capita= (CO2/Liter gas *Liters gas /Vehicle*Vehicles)/capita where: Liters gas /Vehicle = Avg(VKTperVehicle/KPL of gasoline)

5 Salon, Cook, and Williams 5 and: VKT = Vehicle Kilometers Traveled KPL = Kilometers Per Liter The data that are the basis of the per-vehicle average fuel use estimates are odometer readings available in the records of the California Smog Check inspection and maintenance program and the state's Department of Motor Vehicle (DMV) registration database, paired with fuel economy ratings from the US Environmental Protection Agency (EPA). The US Census provides estimates of aggregate household vehicles available in each zip code tabulation area and in each city (14). CO 2 /Liter of Gasoline In this analysis, a carbon dioxide emissions factor of 2.35 kg CO 2 per liter of motor gasoline is used. This is the emissions factor provided by the US Department of Energy s Energy Information Administration (15). It does not include emissions of methane (CH 4 ) and nitrous oxide (N 2 O), both of which are greenhouse gases that are emitted by vehicles in small amounts. It also does not include any upstream emissions from the transport and refining of gasoline, and it does not take into account that a small portion of the fuel used by household vehicles in California is actually diesel fuel rather than gasoline (diesel fuel has a higher emissions factor of 2.69 kg CO 2 per liter). Finally, this emissions factor does not take into account the fact that for much of the year, the gasoline blend in California is actually 10 percent ethanol a fuel that has a lower tailpipe emissions factor than gasoline, but also has a lower energy density than gasoline. Liters of Gasoline/Vehicle To estimate the average annual liters of gasoline used by a vehicle, this value is calculated for a large sample of vehicles in each geographic area based on actual odometer readings and EPA estimates of vehicle fuel economy, and then averaged. This subsection details the estimation method for annual VKT per vehicle and explains how this is combined with fuel economy estimates to obtain average annual fuel used by vehicles registered in a particular zip code. The zip code portion of the physical address of the vehicle registration is used here as the geographic identifier, and vehicle odometer readings taken when vehicles are inspected or have a change of title are used to estimate VKT. The advantage of this method is that odometer readings provide precise and reliable VKT information, making the estimates of average and total VKT presented here quite accurate for a given geographic area. The disadvantage of the method is that where the vehicles are actually traveling remains unknown, so all of the VKT and associated emissions for each vehicle are assigned to the zip code where the vehicle is registered. There are two places where the State of California holds collected odometer readings. The first is the Smog Check inspection data (16). Smog Check inspections are conducted every two years after a vehicle reaches a certain age (under current rules, this age is 6 years). These inspections are also conducted on some newer vehicles when they first enter California from another state or when they are sold. Every Smog Check inspection record includes an odometer reading. The second place where odometer readings are recorded is the DMV registration data (17). Vehicles are registered annually in California, but odometer readings are only recorded in this data in three special situations: (1) the first time a new vehicle is registered, (2) the first time a used vehicle is registered in the State, and (3) when a vehicle has a change of ownership. Note that in situations (2) and (3), a Smog Check inspection record will usually be generated close to

6 Salon, Cook, and Williams 6 this time as well, so there will be two odometer readings for the same vehicle that are close in time. In both datasets, the Vehicle Identification Number (VIN) uniquely identifies each vehicle and each record represents a single inspection/registration for a particular vehicle. Putting these datasets together using the VIN, we obtain a panel dataset that includes a large percentage of the vehicles in the state with odometer readings at multiple points in time. The key variables used to estimate VKT for an individual vehicle are the date of inspection/registration and the vehicle s odometer reading at that time. For a given year of interest, the basic strategy is to identify two odometer readings, one closest to the start of the year and the other closest to the end of the year, and then use the difference as a basis to calculate annualized VKT for that year. The population of vehicles for which we estimate VKT is all vehicles that had a current registration in that year with the California DMV. It is important to note that the VKT estimates are approximations of the number of kilometers individual cars traveled between January 1 and December 31 of the year of interest, but are not usually based on actual distance traveled between these two dates. Two rules regarding VKT estimation were created with the intent of improving the reliability of these estimates. Rule 1: In order to minimize the effects of seasonal driving patterns, every vehicle is required to have at least 1 year between the two dates that are used to estimate VKT. Rule 2: At least half a year of the date range used for the VKT estimation must occur within the year of interest. Using 2005 as an example, the process of estimating VKT is described in more detail. A vehicle must have at least two odometer readings in order to estimate VKT using this method. The readings can either be from a Smog Check inspection or from a vehicle registration. In the vast majority of cases, two Smog Check inspections were used as the source of the odometer readings used to calculate VKT. Keeping in mind the rules outlined above, there are six distinct scenarios that must be considered regarding the history of Smog Check inspections for a particular vehicle. 1. Last inspection occurs before 1/1/05 2. First inspection occurs after 12/31/05 3. First inspection occurs before 1/1/05 and last inspection occurs after 12/31/05 4. All inspections occur between 1/1/05 and 12/31/05 5. First inspection occurs between 1/1/05 and 12/31/05, last inspection occurs after 12/31/05 6. First inspection occurs before 1/1/05, Last inspection occurs between 1/1/05 and 12/31/05 Of these, 2005 VKT can be estimated for Scenarios 3, 4 and 5 using only the Smog Check data. For Scenario 2 vehicles, we can estimate 2005 VKT using the first Smog Check inspection and an odometer reading from the DMV registration data as long as the registration date is before July 2, 2005 (midpoint of 2005). Similarly, for Scenario 6 VINs, we can estimate 2005 VKT using the last Smog Check inspection if it occurs after July 2, The critical task is to identify the two observations (each containing a unique odometer reading and date) that will be used to estimate VKT and also satisfy Rules 1 and 2. The basic strategy is to look for the observations that are closest to 1/1/05 and 12/31/05 respectively. One of the goals of this VKT estimation exercise was to analyze the spatial variation in VKT throughout California. To accomplish this, we utilized the home zip code of each VIN

7 Salon, Cook, and Williams 7 found in the DMV registration data. In order to account for ownership changes and changes of address, we took advantage of the panel nature of the registration data by selecting the registration that occurred between the dates of the two observations used as the basis of the VKT estimate. In addition to the zip code, average fuel economy from the EPA was also attached to each VIN based on the make and trim line of the vehicle. This is not a straightforward database merging exercise, since the way in which the vehicle make, model, model year, and trimline are documented in the EPA fuel economy database is not compatible with the VIN-based information. To identify the fuel economy ratings of vehicles, a web-scraping program was used to obtain both the VIN and EPA fuel economy ratings from a large number of used vehicles. Because the first 10 digits of the VIN identify the make, model, model year, and trimline for each vehicle, it was then possible to generate a database of fuel economy ratings for each VIN prefix code, resulting in approximately 10,000 distinct vehicle types with their associated fuel economy ratings. This data was then linked to the VKT estimates for each VIN. Vehicles older than the fuel economy program (1978 model year for cars and 1979 model year for light-duty trucks) and particularly unusual vehicle types were not included, and were therefore dropped from this analysis. To obtain the average liters of gasoline used per vehicle, VKT was divided by the fuel economy for each vehicle and the result was averaged over vehicles in the relevant geographic area. Number of Vehicles and Population The US Census is the source of data for both vehicle and population counts. The reason to use Census rather than Department of Motor Vehicle registration records for the vehicle count is that the DMV registration list includes not only household vehicles, but all vehicles. For vehicle counts, the variable Aggregate Vehicles Available is used in the areas of interest. For 2000, these come from the Summary File 3 data based on the Census long form. For 2005, the values used are based on a linear interpolation between the 2000 decennial census data and the 2011 American Community Survey 5-year average data. For population counts, the 100 percent count from Summary File 1 of the 2000 decennial census is used, and again a linear interpolation method between that and the 2010 decennial census count is used to arrive at population estimates for Calculating Emissions by Zip Code and by City The calculation of per capita emissions by zip code was relatively straightforward, once all of the data that has been described above was assembled. The calculation represented in Equation 1 was done for each zip code, using the Zip Code Tabulation Areas (ZCTA) from the census to display the results in Figures 1 through 6. Calculating per capita emissions for each city took a few more steps because city boundaries are not coincident with zip code boundaries (Figures 1 through 6 clip the zip code map using each city s actual boundary). Since census tracts are substantially smaller than zip codes, tracts were used to estimate the number of household vehicles available, as explained below. This allowed us to continue to make use of the information we have about the spatial variation in average fuel use per vehicle, but to approximate the true number of vehicles in the city more accurately than we could using only the ZCTA data. For the city population data, census city-level population estimates were used directly.

8 Salon, Cook, and Williams 8 To accomplish this, census tract-level vehicle population information was obtained for both 2000 and Then, the percent of each census tract s area that was inside each city s boundaries was calculated, as well as the percent of each census tract s area that was in each ZCTA. For those tracts that were either not entirely within the city or split between multiple ZCTA s, the percent of the tract s area within each of these other areas was used to split the vehicle count data appropriately. Then, the numerator of Equation 1 was used to estimate the total CO2 emissions from household vehicles in each area census tract within each zip code area that was inside of the city limits. Finally, these emissions were summed and then divided by the city s population to obtain CO2 per capita. RESULTS Table 4 presents estimates of the annual CO2 emitted by household vehicles used by people who live in each of our case study cities for 2000 and Both total household vehicle emissions and per capita emissions from household vehicles. The latter provides a useful metric to compare the carbon intensity of household travel of the residents of each city, since most non-automobilebased travel has a low carbon intensity (with the important exception of air travel). Table 2 and Table 3 detail the average base data for each city for 2000 and 2005, respectively. Table 2: Base Data for CO2 Emissions Calculation, 2000 Average VKT Per Vehicle Average Vehicle Fuel Economy (KPL/MPG) Number of Available Household Vehicles Total Population Berkeley 9, / , ,743 Davis 11, / ,280 60,308 Fresno 10, / , ,652 LosAngeles 10, /28.7 1,830,268 3,694,820 Sacramento 11, / , ,018 SanFrancisco 9, / , ,733 Stockton 11, / , ,771 Ventura 11, / , ,916 Table 3: Base Data for CO2 Emissions Calculation, 2005 Average VKT Per Vehicle Average Vehicle Fuel Economy (KPL/MPG) Number of Available Household Vehicles Total Population Berkeley 8, / , ,662 Davis 11, / ,493 62,965 Fresno 10, / , ,159 LosAngeles 10, /28.5 1,919,767 3,743,721 Sacramento 10, / , ,753 SanFrancisco 9, / , ,984 Stockton 11, / , ,739 Ventura 10, / , ,675

9 Salon, Cook, and Williams 9 Table 4: City-Level Carbon Dioxide Emissions from Household Vehicles, 2000 and 2005 City Metric Tons CO2 Household Vehicles Total, 2000 Metric Tons CO2 Household Vehicles Per Capita, 2000 Metric Tons CO2 Household Vehicles Total, 2005 Metric Tons CO2 Household Vehicles Per Capita, 2005 Berkeley 174, , Davis 125, , Fresno 768, , Los Angeles 6,698, ,543, Sacramento 878, , San Francisco 1,098, ,024, Stockton 461, , Ventura 187, , Comparing across cities for each year, San Francisco and Berkeley stand out as having much lower carbon emissions per capita from household vehicles than the other cities in our study. This makes sense, since these two cities are both located centrally in the Bay Area the only region in the state with high levels of public transit access and usage. The per capita emissions levels in these two cities stand out from the rest even more in 2005 than in This is because these are the only cities that actually had a reduction in the total number of household vehicles available between 2000 and 2005, even as their populations were growing. Interestingly, Davis and Sacramento both have higher per-capita emissions from vehicles in 2000 than the other cities in this study, but by 2005, emissions in Davis had dropped substantially while Sacramento s per capita emissions remained relatively high. This is due to a combination of factors the average fuel economy stayed constant in Davis, but went down in Sacramento, the average vehicle kilometers traveled by each vehicle went down in both cities, but it went down much more in Davis than in Sacramento, and the number of vehicles increased faster than the population in Sacramento and slower than the population in Davis. Comparison with official emissions inventories As quickly becomes clear from a comparison of Table 1 and Table 4, our estimates of household vehicle emissions are quite different from the official city emissions inventory estimates for most of our case study cities. This is largely because we are not only using different data sources, but we are actually counting different emissions! City inventories often do not differentiate between vehicle types, often including all road vehicles in the same category, and sometimes only listing combined emissions for the transport sector as a whole. Also, as is apparent from the Notes column of Table 1, official emissions inventories usually count either emissions that occur within the city limits or count emissions from trips that originate or end within the city limits. In contrast, our emissions estimates count emissions from all travel of vehicles that are used by residents living within the city limits whether that travel actually happens within the city or not and include only those household vehicles that are reported to the US Census. Although limited in this way, we submit that our estimates are interesting and useful because they are calculated using the same data sources and methodology for all of the cities in our study. This quality of our estimates makes it meaningful to directly compare per capita emissions estimates across cities something that is not possible to do using the official inventory numbers.

10 Salon, Cook, and Williams 10 Spatial variation in emissions within cities In addition to calculating our whole-city estimates of carbon emissions from vehicles used by people living in our case study cities, we also made these calculations for each city zip code. The results are illustrated in map form in Figure 1 through Figure 6. Each map has a star that indicates roughly where the city s central business district (CBD) is located, and each zip code area is shaded to represent the level of CO2 emissions per capita come from vehicles in that area. The scale of this shading is the same for all 7 maps, so they can be viewed together and zip code per capita emission levels can be compared across cities. Maps of Davis and Ventura are not included because Davis has only a single zip code and Ventura has only two zip codes, so the spatial pattern of results for these cities is not apparent from the zip code-based map. The first point to note is that in most of our case study cities, there is substantial spatial variation in the carbon emissions per capita coming from household vehicles. Los Angeles presents an especially varied picture. The second thing to note in these maps is that in all of the cities except Ventura and Davis (which are both quite small), the zip code areas that contain the CBD have quite low vehicle emissions per capita. This means that to reduce these emissions, it may make sense to target particular areas of each city generally those on the outskirts rather than in the center of the city.

11 Salon, Cook, and Williams 11 Figure 1: Berkeley: Metric Tons CO2 Per Capita from Vehicles by Zip Code

12 Salon, Cook, and Williams 12 Figure 2: Fresno: Metric Tons CO2 Per Capita from Vehicles by Zip Code

13 Salon, Cook, and Williams 13 Figure 3: Los Angeles: Metric Tons CO2 Per Capita from Vehicles by Zip Code

14 Salon, Cook, and Williams 14 Figure 4: Sacramento: Metric Tons CO2 Per Capita from Vehicles by Zip Code

15 Salon, Cook, and Williams 15 Figure 5: San Francisco: Metric Tons CO2 Per Capita from Vehicles by Zip Code

16 Salon, Cook, and Williams 16 Figure 6: Stockton: Metric Tons CO2 Per Capita from Vehicles by Zip Code

17 Salon, Cook, and Williams 17 FUTURE RESEARCH In future research, we hope to expand this method to include emissions estimates for all jurisdictions in California and for multiple years. The resulting jurisdiction-level household vehicle emissions estimates could then be used as the basis for a multivariate analysis to explain the differences in per-vehicle and per-capita emissions between cities. REFERENCES (1) ICLEI USA, U.S. Community Protocol for Accounting and Reporting of Greenhouse Gas Emissions, October (2) Salon, D., D. Sperling, A. Meier, S. Murphy, R. Gorham, and J. Barrett. City Carbon Budgets: A Proposal to Align Incentives for Climate-Friendly Communities. Energy Policy, Vol. 38, Issue 4, 2010, pp (3) Duduta, Nicolae and Bishins, Allison Citywide Transportation Greenhouse Gas Emissions Inventories: A Review of Selected Methodologies. WRI Working Paper. World Resources Institute, Washington, DC. Available at (4) Ganson, Christopher The Transportation Greenhouse Gas Inventory: A First Step toward City-Driven Emissions Rationalization. Berkeley: University of California Transportation Center, UCTC Research Paper no. 879, p. 8. (5) City of Berkeley Climate Action Plan. June Available at (6) City of Davis Greenhouse Gas Inventory and Forecast Update. June Presentation format. Available at community-development.cityofdavis.org. (7) City of Davis Greenhouse Gas Inventory and Forecast Report. March Prepared by Christa Clark Jones, Public Works Department, City of Davis. (8) Green LA: An Action Plan to Lead the Nation In Fighting Global Warming. May Available at environmentla.org/pdf/greenla_cap_2007.pdf. (9) ICF Jones & Stokes GHG Emissions Inventory for Incorporated and Unincorporated Sacramento County. June. (ICF J&S ) Sacramento, CA. Prepared for: Sacramento County Department of Environmental Review and Assessment. (10) San Francisco Department of the Environment and San Francisco Public Utilities Commission. September Climate Action Plan for San Francisco: Local Actions to Reduce Greenhouse Gas Emissions. (11) San Francisco Municipal Transportation Agency (SFMTA). Draft 2011 Climate Action Strategy for San Francisco s Transportation System. (12) ICF International City of Stockton Climate Action Plan. Draft. February. (ICF ) Sacramento, CA. Prepared for City of Stockton, Stockton, CA. (13) Community Greenhouse Gas Inventory for Ventura, CA. Available at (14) US Census, 2000 Decennial Census Data. (15) US Census, American Community Survey year Average Data. (16) California Bureau of Automotive Repair. Smog Check dataset. (17) California Department of Motor Vehicles. Registration records dataset.