Table SI.1: Urban and total GHG emissions by region and sector
(million metric tons CO2-eq.)
Latin America / NorthAfrica / Asia / & Caribbean / Europe / America / Oceania / World
Sector / Urban / Total / Urban / Total / Urban / Total / Urban / Total / Urban / Total / Urban / Total / Urban / Total
Agriculture / 12 / 487 / 148 / 2,458 / 19 / 846 / 59 / 651 / 21 / 419 / 8 / 163 / 266 / 5,024
Energy (low) / 233 / 738 / 2,572 / 6,749 / 296 / 833 / 2,132 / 4,225 / 1,589 / 3,836 / 106 / 299 / 6,927 / 16,681
Energy (high) / 372 / 3,710 / 412 / 2,981 / 3,351 / 229 / 11,055
Industry / 30 / 75 / 476 / 1,565 / 67 / 202 / 444 / 935 / 178 / 350 / 8 / 30 / 1,204 / 3,158
Residential / 19 / 133 / 320 / 1,296 / 41 / 152 / 403 / 1,010 / 443 / 736 / 7 / 20 / 1,234 / 3,346
Transportation / 48 / 159 / 425 / 1,242 / 161 / 414 / 587 / 1,240 / 1,288 / 1,884 / 52 / 92 / 2,561 / 5,030
Waste / 23 / 121 / 250 / 768 / 60 / 149 / 105 / 259 / 159 / 248 / 9 / 17 / 606 / 1,563
Total (Low) / 366 / 1,713 / 4,192 / 14,079 / 644 / 2,597 / 3,730 / 8,320 / 3,679 / 7,473 / 188 / 621 / 12,798 / 34,802
Total (High) / 504 / 5,330 / 760 / 4,579 / 5,440 / 312 / 16,926
Values represent million metric tons CO2-eq. for both "Urban" areas within the region and for all ("Total") GHG emissions for the region. "World" includes all urban extents (“Urban”) in the database and all GHG emissions globally ("Total"). We exclude large biomass burning and aviation and navigation over oceans. Note that for the energy sector and the total GHG emissions levels there is only one value for each region.
Figure SI.1: Global and Urban GHG Emissions (low and high estimates), 2000 (billion tons CO2-eq.). "Urban (low)" includes total direct GHG emissions from all urban extents in the database (Scope 1), "Urban (high)' includes total direct GHG emissions and those from thermal power plants for all urban extents in the database (Scope 2). "World" includes all anthropogenic GHG emissions except biomass burning and aviation and navigation related emissions over oceans.
Figure SI.2: Percent Urban Share of Total GHG emissions byDevelopment Status, 2000. The blue shaded bars represent the low estimates for urban CO2-eq emission shareand the red shaded bars represent the high estimates for urban CO2-eq emission share
Table SI.2: Urban and total GHG emissions by sector and development status, 2000
(million metric tons CO2-eq.)
Developing / DevelopedSector / Urban / Total / Urban / Total
Agriculture / 162 / 3,754 / 104 / 1,271
Energy (low) / 2,629 / 7,478 / 4,298 / 9,203
Energy (high) / 3,903 / 7,152
Industry / 443 / 1,580 / 761 / 1,577
Residential / 263 / 1,405 / 971 / 1,941
Transportation / 473 / 1,497 / 2,088 / 3,533
Waste / 309 / 1,002 / 297 / 561
Total (Low) / 4,279 / 16,717 / 8,519 / 18,086
Total (High) / 5,553 / 11,372
Values represent million metric tons CO2-eq for both "Urban" areas within the region and for all ("Total") GHG emissions for developed and developing countries. "Energy (low)" includes only the energy-related GHG emissions from within urban extents. "Energy (high)" includes the GHG emissions within urban extents and from thermal power plants outside urban extents apportioned by national share of the urban area size.
Table SI.3: Urban and total transportation GHG emissions by sub-sector and development status, 2000
(million metric tons CO2-eq.)
Developing / Developed / WorldUrban / Total / Urban / Total / Urban / Total
Aviation / 10 / 115 / 55 / 344 / 65 / 459
Navigation / 0 / 2 / 3 / 8 / 3 / 9
Road / 431 / 1,247 / 1,906 / 2,945 / 2,337 / 4,192
Non-road / 32 / 133 / 124 / 236 / 156 / 369
Values represent million metric tons CO2-eq for both "Urban" areas within the region and for all ("Total") GHG emissions for developed and developing countries. . "World" includes all urban extents in the database ("Urban") and all GHG emissions globally ("Total"). We exclude large biomass burning and aviation and navigation over oceans.
Interestingly, there are approximately 50 urban extents where the aviation GHG emissions are greater than 3% of total GHG emissions for the urban center and approximately 25 urban extents where they are greater than 5%. The high percentage emissions for some urban extents, such as Atlanta (USA) and Geneva (Switzerland) can be explained because they are international airport hubs. Others urban extents, such as Anchorage (USA) are larger cities with significant levels of GHG emissions, but are on highly used flight paths and we pick up the emissions of planes flying over these city. Finally, others urban areas with high percentages of aviation GHG emissions have small population and low levels of other emissions, but are close to centers of high plane activity, such as Montargis, France, which is just north of Paris. More detailed work on this issue needs to be performed.
Table SI.4: Comparative GHG emissions per capita ranges, urban, non-urban and total by region, 2000
GHG emissions(tons CO2-eq./capita)
Africa / 2.1
Urban / 1.5-2.1
Non-urban / 2.1-2.3
Asia / 3.8
Urban / 3.3-4.2
Non-urban / 3.6-4.1
Latin America / 5.0
Urban / 2.4-2.8
Non-urban / 7.3-7.8
Europe / 11.4
Urban / 8.7-10.7
Non-urban / 12.4-15.3
North America / 23.8
Urban / 15.9-23.5
Non-urban / 24.5-45.7
Oceania / 20.6
Urban / 11.6-19.2
Non-urban / 22.2-31.0
World / 5.7
Urban / 5.2-6.9
Non-urban / 4.9-6.1
Developing / 3.4
Urban / 2.6-3.3
Non-urban / 3.5-3.9
Developed / 15.0
Urban / 10.9-14.5
Non-urban / 16.0-22.8
Figure SI. 3: Comparative GHG emissions per capita (tons CO2-eq./cap) from Urban (low and high estimates) and Non-urban (low and high estimates) areas, by Region, 2000. The dark and light blue shaded bars represent the urban CO2-eq emission levels from all urban areas (low = dark, high = light). The red and pink shaded bars represent the urban CO2-eq emission levels from all Non-urban areas (red = low, pink = high). The green shaded bars represent regionalCO2-eq emission per capita levels.
Table SI.5: Percent distribution of GHG emissions by sector, region and distance from urban area, 2000Africa / Asia
within UE / UE-20 km / 20-40 km / 40-80 km / remainder / within UE / UE-20 km / 20-40 km / 40-80 km / remainder
Agriculture / 2.4 / 12.2 / 13.4 / 24.8 / 47.2 / 6.0 / 41.7 / 23.8 / 16.3 / 12.2
Energy / 31.7 / 27.4 / 11.1 / 8.8 / 21.1 / 38.1 / 44.3 / 8.6 / 5.8 / 3.1
Industry / 40.5 / 23.3 / 8.8 / 15.7 / 11.7 / 30.4 / 51.2 / 10.5 / 6.1 / 1.9
Residential / 14.5 / 26.7 / 16.6 / 20.6 / 21.6 / 24.7 / 44.7 / 16.4 / 9.6 / 4.6
Transportation / 30.4 / 28.2 / 10.7 / 12.7 / 18.0 / 34.3 / 35.7 / 12.5 / 9.1 / 8.5
Waste / 18.7 / 28.6 / 13.9 / 19.0 / 19.8 / 32.6 / 45.8 / 12.6 / 6.5 / 2.5
Total / 21.4 / 23.0 / 12.2 / 15.7 / 27.8 / 29.8 / 44.0 / 12.8 / 8.3 / 5.1
Latin America and the Caribbean / Europe
within UE / UE-20 km / 20-40 km / 40-80 km / remainder / within UE / UE-20 km / 20-40 km / 40-80 km / remainder
Agriculture / 2.2 / 15.5 / 17.6 / 27.4 / 37.3 / 9.0 / 35.6 / 25.6 / 20.0 / 9.8
Energy / 35.5 / 30.6 / 15.0 / 9.0 / 10.0 / 50.5 / 29.6 / 10.8 / 4.7 / 4.4
Industry / 33.3 / 36.3 / 14.6 / 9.6 / 6.3 / 47.5 / 34.9 / 9.6 / 5.3 / 2.7
Residential / 27.1 / 31.8 / 14.7 / 14.4 / 12.0 / 40.0 / 37.2 / 11.9 / 7.1 / 3.8
Transportation / 38.9 / 28.5 / 10.8 / 11.2 / 10.6 / 47.3 / 32.4 / 9.9 / 6.1 / 4.2
Waste / 40.4 / 30.6 / 10.7 / 10.6 / 7.7 / 40.5 / 35.7 / 11.3 / 7.9 / 4.5
Total / 24.8 / 25.9 / 14.9 / 15.8 / 18.7 / 44.8 / 32.2 / 11.9 / 6.6 / 4.6
North America / Oceania
within UE / UE-20 km / 20-40 km / 40-80 km / remainder / within UE / UE-20 km / 20-40 km / 40-80 km / remainder
Agriculture / 5.0 / 20.2 / 16.5 / 22.9 / 35.4 / 4.9 / 8.2 / 6.8 / 12.7 / 67.4
Energy / 41.4 / 27.0 / 11.9 / 10.5 / 9.2 / 35.3 / 17.3 / 11.9 / 11.8 / 23.7
Industry / 50.9 / 32.0 / 6.7 / 6.5 / 3.8 / 25.4 / 24.1 / 27.5 / 6.2 / 16.8
Residential / 60.3 / 19.8 / 7.4 / 7.0 / 5.5 / 33.3 / 21.2 / 7.0 / 8.5 / 30.0
Transportation / 68.4 / 13.1 / 5.7 / 6.1 / 6.8 / 56.3 / 14.7 / 2.8 / 4.3 / 21.8
Waste / 64.1 / 18.6 / 6.7 / 6.3 / 4.3 / 50.9 / 22.1 / 4.0 / 5.4 / 17.6
Total / 49.2 / 22.4 / 9.7 / 9.4 / 9.3 / 30.3 / 15.1 / 9.6 / 10.4 / 34.6
Note: "within UE" refers to the within urban extents. Values for within UE only include energy emissions within the urban extent (low estimates). "UE-20 km" includes the buffer from the border of the UE to 20 km. "20-40 km" includes the buffer that starts 20 km from the UE to 40 km of the UE". "40 - 80 km" includes the buffer that starts 40 km from the UE to 80 km of the UE". "Remainder" includes all GHG emissions from all other areas.
Figure SI.4: Share of Total GHG Emissions along urban-to-rural gradient, by Sector and Development Status, 2000. The rose color bars from lightest to darkest represent agriculture (lightest), energy, industry, residential and transportation (darkest) sector share of all CO2-eq emissions. The green bar represents the waste sector share of all CO2-eq emissions and the black bar represents total emissions of all CO2-eq emission. "UE" includes within the urban extent, "UE-20 km" is the buffer from the UE to 20 km, "20-40 km" is the buffer from 20 to 40 km from the urban extent, "40-80 km" is the buffer from 40 to 80 km from the urban extent, "remainder" is the remaining GHG emissions
Table SI.6: Comparison of values between “top-down” and “bottom-up” estimates for individual urban areas
GHG emissions per capita (tons CO2e/capita) / ActualBottom-up estimate / Top-down estimate / Percent / Different
City / Country / Year / Value / Low / High / Difference / (tons co2e/capita)
Bangkok / Thailand / 2005 / 10.70 / 3.73 / 3.82 / -64.3 / -6.88
Athens / Greece / 2005 / 10.40 / 3.94 / 4.02 / -61.4 / -6.38
Geneva / Switzerland / 2005 / 7.80 / 3.09 / 3.11 / -60.2 / -4.69
Beijing / China / 2006 / 10.10 / 3.85 / 4.53 / -55.2 / -5.57
Stuttgart / Germany / 2005 / 16.00 / 7.96 / 8.66 / -45.9 / -7.34
Porto / Portugal / 2005 / 7.30 / 4.33 / 4.38 / -40.0 / -2.92
Frankfurt / Germany / 2005 / 13.70 / 8.22 / 8.48 / -38.1 / -5.22
Tianjin / China / 2006 / 11.10 / 6.26 / 6.94 / -37.5 / -4.16
Hamburg / Germany / 2005 / 9.70 / 6.58 / 6.65 / -31.4 / -3.05
London (Gr. London) / UK / 2003 / 9.60 / 7.11 / 7.15 / -25.5 / -2.45
Shanghai / China / 2006 / 11.70 / 8.81 / 9.10 / -22.2 / -2.60
Madrid / Spain / 2005 / 6.90 / 5.84 / 5.96 / -13.6 / -0.94
Denver / USA / 2005 / 21.50 / 16.22 / 18.84 / -12.4 / -2.66
Seattle / USA / 2005 / 13.68 / 12.54 / 12.83 / -6.2 / -0.85
Amman / Jordan / 2008 / 3.25 / 3.06 / 3.13 / -3.8 / -0.12
Prague / Czech Rep. / 2005 / 9.40 / 9.01 / 10.33 / 0.0 / 0.00
Singapore / Singapore / 1994 / 7.86 / 7.70 / 9.75 / 0.0 / 0.00
Ljubljana / Slovenia / 2005 / 9.50 / 6.09 / 11.39 / 0.0 / 0.00
Portland / USA / 2005 / 12.41 / 12.99 / 13.63 / 4.7 / 0.58
Los Angeles / USA / 2000 / 13.00 / 14.56 / 14.74 / 12.0 / 1.56
Barcelona / Spain / 2006 / 4.20 / 4.87 / 4.91 / 15.9 / 0.67
Rio de Janeiro / Brazil / 1998 / 2.10 / 2.53 / 2.58 / 20.4 / 0.43
Toronto (Metro area) / Canada / 2005 / 11.60 / 14.63 / 16.64 / 26.1 / 3.03
Glasgow / UK / 2004 / 8.80 / 11.59 / 11.73 / 31.6 / 2.79
Helsinki / Finland / 2005 / 7.00 / 9.82 / 10.28 / 40.3 / 2.82
Paris / France / 2005 / 5.20 / 7.64 / 7.65 / 46.9 / 2.44
New York City / USA / 2005 / 10.50 / 16.65 / 16.71 / 58.5 / 6.15
Tokyo / Japan / 2006 / 4.89 / 8.44 / 8.44 / 72.6 / 3.55
Sao Paulo / Brazil / 2000 / 1.40 / 2.82 / 2.86 / 101.4 / 1.42
Brussels / Belgium / 2005 / 7.50 / 15.15 / 16.23 / 102.0 / 7.65
Stockholm / Sweden / 2005 / 3.60 / 7.71 / 7.75 / 114.2 / 4.11
Oslo / Norway / 2005 / 3.50 / 7.55 / 7.55 / 115.6 / 4.05
Source: values for bottom up studies from Hoornweg et al, 2011, “Year” is the year for the “bottom-up” study
Figure SI.5: Percent differences in values between bottom-up and top-down urban GHG per capita emission estimates
Figure SI.6: Absolute value differences between bottom-up and top-down estimates of urban GHG emissions per capita
Figure SI.7: Absolute differences in regional and global averages between “bottom-up” and “top-down” urban GHG per capita estimates
SI.7: Distribution of GHGs by compound by region, 2000
(Million tons CO2-eq)
Methane (CH4) / Carbon Dioxide (CO2) / Nitrous Oxide (N2O) / Sulfur Hexaflouride (SF6)Region / Urban / Total / Urban (Low) / Urban (High) / Total / Urban / Total / Urban / Total
Asia / 553 / 3,468 / 3,540 / 4,679 / 9,756 / 83 / 825 / 15 / 30
Africa / 83 / 713 / 273 / 412 / 811 / 8 / 186 / 1 / 2
Europe / 370 / 1,261 / 3,271 / 3,394 / 6,641 / 75 / 383 / 15 / 34
Latin America & Carb. / 106 / 928 / 520 / 636 / 1,415 / 18 / 252 / 1 / 2
North America / 270 / 780 / 3,296 / 5,057 / 6,367 / 79 / 280 / 34 / 47
Oceania / 20 / 159 / 164 / 288 / 404 / 4 / 57 / 0 / 0
World / 1,400 / 7,310 / 11,064 / 15,192 / 25,393 / 268 / 1,983 / 66 / 116
Note: Values are for emissions from urban extents (Urban) and for the entire region(Total). All emissions from thermal power plants outside of urban extents were counted as CO2emissions. In Asia, urban extents emitted approximately 553 million tons CO2-eq. of CH4emissions in 2000, which were approximately 15.9% of total CH4emissions in that region (3,468 million tons CO2-eq.).
SI.8: Distribution of urban GHGs by compound by source, 2000
(Million tons CO2-eq)
Carbon / Nitrous / SulfurMethane / Dioxide / Oxide / Hexafluoride
Source / CH4 / CO2 / N2O / SF6
Agriculture / 186 / 7 / 72 / 0
Energy (low) / 621 / 6,274 / 32 / 0
Energy (high) / 621 / 10,402 / 32 / 0
Industry / 2 / 1,062 / 73 / 66
Residential / 22 / 1,207 / 5 / 0
Transportation / 9 / 2,497 / 56 / 0
Waste / 561 / 17 / 29 / 0
Total (Low) / 1,400 / 11,064 / 268 / 66
Total (High) / 1,400 / 15,192 / 268 / 66
Note: Values are for emission from each source for all urban extents combined. All emissions from thermal power plants outside of urban extents were counted as CO2 emissions. Methane (CH4) from agriculture sources for all urban extents accounted for 186 million tons CO2-eq. for 2000. All sulfur hexafluoride (SF6) emissions were from industrial sources.
SI. Qualifications to the data and methods
This study uses a variety of spatial datasets to obtain figures for GHG emissions levels, shares and intensities for urban areas. Beside the complexities of defining a reliable protocol, the study is limited by the uncertainties in the data and methods. We identify several of these in the body of the article, and elaborate on other issues here.
Uncertainties introduced with GRUMP data
The GRUMP data are not the only available for urban areas. A recent study compared ten global urban and urban-related mapping efforts (Schneider, Friedl and Potere 2009). That study demonstrates that different mapping techniques result in considerably different total global land area estimates. Schneider et al. (2009) suggest that the method applied in GRUMP results in the largest area classified as urban at the global scale (over 3.5 million km2). In contrast, the MODIS urban land cover mapping effort finds urban areas covering 20% of that area (657,000 km2). We choose GRUMP because it has been used and validated by the Millennium Ecosystem Assessment (McGranahan et al. 2005), because the GRUMP boundaries are consistent with urban areas in several developed countries including the United States, and because the large areas defined by GRUMP are more closely aligned to what urbanists call “urban fields” (Friedmann 1973) or “functional urban economic areas” (Fox and Kumar 1965). These areas bound a geographic space where, arguably, a large percentage of urban activities occur.
The UN also produces both urbanization and urban growth data by country and city-level population data. We believe the UN data are less useful than GRUMP for a variety of reasons. First, the list of urban agglomerations provided by the UN only includes large cities (those 750,000 pop or greater) and as (Grubler et al. 2012) point out most of the urban growth in the future will happen in small- to medium-sized urban centers. Hence, these small urban areas are important to analyze, but not well represented in the UN Publications. Second, the UN database on large cities only includes up to 750 urban areas worldwide, or a bit more than 10% of our sample. Third, the definitions of urban in the UN database vary significantly (Satterthwaite 2007). Fourth, there is no spatial application of the topology of these data provided by the UN. That is, they are only city names and do not always correspond to administrative or political units. The spatial orientation of population is beyond the UN’s historical or current purview. At the same time, GRUMP population data for 2000 is based upon the UN population data at the national scale, but it is downscaled/spatialized.
There are uncertainties related to the use of GRUMP. These are introduced with the data used to develop GRUMP (i.e., the nighttime lights data, Elvidge et al., 1997a, 1997b). The nighttime lights data do not indicate built-up area, but rather represent stable sources of light produced by electricity and permanent fires. The biases introduced with these data associate with differential levels of economic activity. Poor cities may be underrepresented as compared to wealthier cities of equal geographic size, as poorer cities use less electricity. On the other hand, wealthier urban areas may be larger than expected due to “blooming effects” or the overestimation of the true extents of urban areas due to the intrinsic characteristics of the sensor. For details on the construction of GRUMP urban extents and the relationship between these data and the UN data see (Balk et al. 2004, Balk and Yetman 2004, Balk 2009).
Uncertainties introduced with EDGAR data
The EDGAR data have been developed over the past several decades (Olivier et al. 1998, Olivier et al. 1994, Olivier et al. 2005). According to the most recent description of the independent variable data, the classification of emitting sources in EDGAR v4, are based upon the IPCC National Greenhouse Gas Inventories Programme (Reporting guidelines in the revised 1996 IPCC guidelines), with the main sectors as 1) Energy, 2) Industrial processes, 3) Solvents and other produce use; 4) Agriculture; 5) Land use change and forestry; 6) Waste and 7) Other.
It is important to note that GHG compounds are emitted differentially by region and source. For example, at the global scale, SF6 is only calculated for industrial emissions. The largest sources of CH4 and N2O are from agriculture and waste management. The CO2 emissions are highest from energy, industry and transportation sources. Tables SI.7 and SI.8 provide an account of the distribution of the various GHG gases by region (urban and total) and by source.
Within EDGAR, the emissions from area (diffuse), line (road, rail, water and airways) and point (stacks) sources are calculated as country totals and are distributed on a 0.1 x 0.1 resolution grid map. For point sources, emissions are distributed in the grid cells where they are located using the geographic coordinates of each facilities. For area, line and some point sources, the emission distribution of a given substance on a grid covering the country is performed using proxy data. The EDGAR development team uses a set of a hundred proxy datasets as geospatial surrogates for gridding the sector and country specific emission totals(Janssens-Maenhout et al. 2012).
The concern among urban analysts is the use of the proxy data for the allocation procedures. Indeed, the EDGAR developers state(Janssens-Maenhout et al. 2012, p 24) that “the spatial uncertainty remains very large and it cannot be expected that with these surrogates each single point source is represented with its emission as accurately as in those databases which trace locally the emission sources.”
As such there are several points of uncertainty in this study. First, some of the information we use to construct the dependent variable (GHG emissions) is also included in the model to understand variation. In particular we note that EDGAR 4.0 uses population and population density (from GPWv3) to spatially allocate emissions for some sectors where more detailed information is not available (i.e., residential and waste). Incorporating population as a condition for allocating emissions inherently implicates population as an important explanatory variable.
Second, EDGAR does not account for affluence when using population or density to spatially allocate emissions. Therefore, EDGAR may systematically underestimate emissions in high-density areas where urbanites have higher incomes and use more energy than their rural counterparts. Nevertheless, the presence of high-density slums in other urban areas of the region, where residents use considerably less energy than their rural counterparts, may counter this effect.
Third, EDGAR uses country specific emission factors when estimating emissions for each sector and for each technology. While the addition of emission factors for different technologies provides more detailed information than using average sector emissions factors, most urban GHG researchers agree that local emissions factors would be more accurate than national level factors.
Fourth, uncertainty estimates for some compounds in EDGAR could be large. Studies suggest that there are is a very high degree of similarity between EDGAR and other databases, such as those from the Oak Ridge National Laboratory (ORNL), but relative differences are largest for small emitters, especially from developing countries in Africa (Marland, Brenkert and Olivier 1999). Uncertainty estimates for GHG emissions based upon the EDGAR Version 4.0 are not available. EDGAR uncertainty estimates for Version 2.0, based upon expert judgment ranges, suggest uncertainties vary by source but for total emissions uncertainties for CO2 are small (10% or less), for CH4 are medium (10-50%), and for N2O are large (50-100%) (see For EDGAR version 3, uncertainties are in the order of 30% for methane, 50% for all other gases except CO2, for which it is 10% (van Amstel, Olivier and Janssen 1999).
Finally, research findings differ from EDGAR estimates for some compounds. For example, Kim et al. (2010) finds good agreement between their extrapolations from China and EDGAR emissions for HFC-23, but finds higher emissions levels for HFC-134a, HFC-152a, HFC-32, HFC-125 and SF6 than reported by EDGAR. At the same time, the Chinese HFC-23 emissions map for EDGAR is smooth and follows population distribution, a result considered unlikely (Stohl et al. 2010). Muhle et al. (2010) finds that EDGAR emissions levels for C2F6 are in general agreement with their model, but EDGAR CF4 and C3F8 levels are significantly lower than their estimates, particularly after 1991. Comparison of recent observation-inferred studies and modeling exercises suggest general agreement with EDGAR up until the mid-1990s. After that, estimated uncertainty levels for SF6 EDGAR data increased from 10% prior to 1995 to over 15% in 2005, and to 20% by 2008 (Rigby et al. 2010).