Accelerated Reduction in \(SO_2\) Emissions from the U.S. Power Sector Triggered by Changing Prices of Natural Gas

Emissions of sulfur dioxide (SO 2 ) from the US power sector decreased by 24% in 2009 relative to 2008. The logarithmic mean Divisia index (LMDI) approach was applied to isolate the factors responsible for this decrease. It is concluded that 15% of the decrease can be attributed to the drop in demand for electricity triggered by the economic recession, 28% to switching of fuel from coal to gas responding to the decrease in prices for the latter. The largest factor in the decrease, close to 57% resulted from an overall decline in emissions per unit of power generated from coal. This is attributed in part to selective idling of older, less efficient coal plants that generally do not incorporate technology for sulfur removal, in part to continued investments by the power sector in removal equipment in response to the requirements limiting emissions imposed by the US Environmental Protection Agency (US EPA). The paper argues further that imposition of a modest tax on emissions of carbon would have ancillary benefits in terms of emissions of SO 2 .


Introduction
Overall generation of electricity in the US decreased by 4.1% between 2008 and 2009. Carbon emissions from the power sector, which accounts for approximately 40% of total US greenhouse gas emissions, declined over the same period by an even larger factor, by 8.76 % in 2009 relative to 2008.
Part of this reduction was attributed to the recession that set in during late 2008 with an important additional contribution due to a price-induced shift in the generation of electricity from coal to gas. Lu et al (1), using an econometric model, concluded that the increase in the use of natural gas relative to coal was responsible for a 4.3% reduction in CO 2 emissions from the US power sector over and above the 4.1% reduction attributed to the recession.
We focus in this paper on the implications of the shift in the power mix and other factors for sulfur dioxide (SO 2 ). SO 2 is an important air pollutant responsible for production of sulfate aerosols impacting not only public health (2,3), but also acid rain and potentially climate (4). Recent studies (5,6) suggest that warming of the climate by increasing level of greenhouse gases has been significantly offset by cooling due to the direct and indirect impact of aerosols, with particular attention to the role of sulfur. Combustion of fossil fuels in the electric power sector represents the dominant anthropogenic source of SO 2 emissions in the US. Emissions of SO 2 in the US amounted to 10.4 million metric tons (MT) in 2008 with 66% associated with the generation of electricity using sulfur-containing fuels (mainly coal), 28% from varied industrial activities, with the balance from a combination of onroad and off-road transportation (4.6%) and other miscellaneous sources (1.4%) (7,8). Emissions from the US power sector decreased by 24% in 2009 as compared to 2008, from 7.8 MT to 6.0 MT (9). The decrease was significantly greater than the corresponding drop either in total power production or in emissions of CO 2 . We argue here that the primary factor responsible for the reduction in SO 2 emissions in 2008 relative to 2009 involved a decrease in emissions of SO 2 per unit of electricity produced (we refer to this as the SO 2 emission intensity). Price induced switching from coal to gas also contributed, as did the decrease in total power production. As we shall indicate in what follows, the reduction in SO 2 emissions occurred despite the fact that the sulfur content of coal consumed by the power sector was actually higher in 2009 than in 2008.
Our analysis makes use of the Logarithmic Mean Divisia Index (LMDI) approach as described in Section 2. Results are presented and discussed in Section 3. We propose to treat not only what took place on a national scale but also what developed regionally. For the latter purpose, we chose to focus on census regions, building on the earlier CO 2 analysis (1). Implications for SO 2 of a potential tax on emissions of carbon from the power sector are discussed in Section 4, with discussion included in Section 5.

Methodology and Data
A number of factors is expected to contribute to changes in emissions of SO 2 from the US power sector, specifically changes in electricity production, differences in fuel mix, in the sulfur content of fuels (mainly coal and oil), and in the efficiencies for removal of SO 2 . The fuel mix for electricity generation varies to a significant extent across the US. The bar graphs in Figure 1 illustrate the fractions of electricity that were generated using coal, natural gas and other fuels for different census regions in 2008 with grayscales indicating the strengths of related emissions of SO 2 . As illustrated in Figure 1, coal is the dominant fuel employed for electricity generation in the West North Central, East North Central and East South Central regions. As a result, SO 2 emissions are relatively high in these regions (10) . Other generating sources (including natural gas, hydro power, nuclear, oil and renewables) are more important for New England and for the contiguous Pacific region. Emissions of SO 2 are much lower as a consequence in these regions. Alaska and Hawaii were excluded from the Pacific census region since generation of electricity from coal is negligible in these states.

Figure 1
The composition of electricity generation for the nine US census regions in 2008. The grayscales for each region indicate the amounts of SO 2 emissions from their power sectors for the same year.
As indicated by Figure 1, the magnitude of SO 2 emissions from the power sector is not determined solely by total power production or by the fraction of electricity generated using coal. The fraction of the total electricity produced using coal in the South Atlantic region is roughly the same as for the Mountain region. While the former generates less than twice as much as electricity as the latter, its emissions of SO 2 are more than seven times greater. Other factors, including the SO 2 emission intensity must also contribute to the distinction between emissions from the two regions. Similar considerations must be invoked to account for the fact that the Middle Atlantic region is responsible for 12% of total US power sector SO 2 emissions while producing 10% of the nation's electricity, despite the fact that only 34% of total power in this region is generated using coal.
Percent changes in electricity generation and SO 2 emissions for 2009 relative to 2008 for the nine census regions and for the entire contiguous US are summarized in Table 1. The changes in electricity generation by energy source varied significantly from region to region between 2008 and 2009 and their impacts on SO 2 emissions are similarly complex. The South Atlantic region, for example, was distinguished by an 18% decrease in generation of electricity using coal, a 21% increase in production from gas, with a 0.7% increase from other fuels, accompanied by the largest drop in SO-2 emissions (about 34%) as compared with other regions. In the contiguous Pacific region, electricity generated using coal, gas and other sources decreased by 16%, 3.8% and 1.7% respectively. Despite this, emissions of SO 2 increased by 5.4%. Although emissions of SO 2 and electricity generation from coal declined simultaneously between 2008 and 2009 in most regions except for the Contiguous Pacific, their change rates exhibited significant differences throughout the nine census regions. The primary purpose of this paper is to evaluate the relative importance of the different where j S refers to the total emissions of SO 2 from the power sector in region j; j E denotes the corresponding production of electricity;  Table 2.
The LMDI approach was adopted in order to separate the impacts of the different factors on the overall changes in emissions of SO 2 . This approach has been widely applied in analyses of energy demand and supply, carbon dioxide emissions, and efficiencies in energy related studies (14,15), and has proved to have advantages over other methodologies in terms of theoretical foundation, adaptability, ease of use, and ease in interpretation of results (14,15,16). With the logarithmic division approach, the unexplained residual terms may be allocated as contributions to the individual factors (17,18,19). If the residual terms are large, this could lead to unavoidable ambiguity in the assignment of influence to specific factors. Arguments supporting the validity of the approach in the present context, in which that the residual terms are relatively small, are presented in the supporting information (SI).
The changes in SO 2 emissions for each census region from 2008 to 2009 ( j S ∆ ) may be expressed in the additive form of the LMDI decomposition analysis (14) as a sum of impacts from the individual influential factors as follows: Here superscripts 0 and t refer to the beginning and ending years of interest, 2008 and 2009 for the present analysis.    (20). Average removal efficiencies for SO 2 control systems were computed using the methodology described by Zhao et al. (21). Summaries of the data used in this study are presented in Tables S1 and S2 of the SI.

Results
As described in the previous section, we considered three factors in the first-level LMDI analysis: electricity generation (    The results in Figure 4 indicate that the sulfur content of coal consumed in 2009 was marginally higher than that used in 2008, contributing to an increase in the net intensity emission factor. On a national basis, the sulfur content of the coal consumed in 2009 was 3.2% higher than in 2008 while the heat content decreased by 1% (21). The decrease in heat content reflected a modest switch in percentage from higher-energy bituminous coal to lower-energy sub-bituminous fuel. The overall decrease in electricity consumption, in combination with an overall decrease in production using coal, may have reduced the pressure on utilities to limit emissions of sulfur. This could account for the larger fraction of higher sulfur, and potentially cheaper bituminous coal consumed in 2009 as compared to 2008 as indicated in Figure S5 of SI.
Results from the application of the second-level LMDI analysis to the individual census regions are summarized in Figure 5. For most regions, the relative contributions of the different factors are similar to those inferred for the nation as a whole. The increase in the removal efficiency was greatest as expected in the regions where most of the older coal fired plants were located, notably in the East North Central, East South Central, Middle Atlantic and South Atlantic regions (26). Emissions actually increased in the New England and Pacific regions due primarily in the former case to the use of higher sulfur coal with a further contribution in the latter case associated with a decrease in the removal efficiency. .

Immediate Co-benefits of a Carbon Tax
The potential impact of a tax on carbon emissions imposed on the power sector was discussed by Lu et al (1). A tax on carbon would have an affect similar to that associated with a reduction in the price of natural gas relative to coal, responding to the fact that, per unit of electricity produced, emissions of carbon from consumption of gas are only about half of those from coal. Lu et al (1) argued that a modest price on carbon, as low as $5 a ton, could result in a significant additional reduction in CO 2 emissions. We discuss here the implications of a carbon tax for emissions of sulfur. As in the earlier study, our focus is on the immediate impacts of such a tax. That is to say, we assume that the demand for electricity remains the same and that the response of the generating system to a change in the gascoal price differential is the same as that simulated using the econometric model developed by Lu et al (1).
Regional results from the SO 2 analysis are presented in Figure 6. As expected, the additional savings in SO 2 emissions are greatest when the difference between gas and coal prices is at a minimum.
As indicated in the figure, a tax of $10 per ton of CO 2 imposed for example in the East South Central region would result in an additional reduction in SO 2 emissions of 27,000 tons (3.9% of total regional emissions) under conditions where the gas-coal cost differential (i.e., the annual cost of gas-fired minus coal-fired electricity) was at a level of 2 cents/kWh. The reduction would be negligibly small if the cost difference for generation of electricity using gas versus coal were to exceed 6 cents/kWh.
In practice, prices for both gas and coal vary across the country. Adopting reported regional prices for gas and coal in 2009 (11), the results in Figure 6 can be used to calculate the additional savings In estimating the change in SO 2 emissions that would have resulted from a carbon tax we assumed that sulfur emissions would be reduced in proportion to the overall reduction in the generation of electricity from coal. In practice we would expect the carbon tax to result in idling of the oldest, least efficient, and probably most polluting, coal power plants. The present analysis most likely underestimates in this case the savings in SO 2 emissions that could be realized as an ancillary benefit of a carbon tax.
A carbon tax is expected to promote an increase of demand for natural gas, resulting in an increase in the price of gas and thus an increase in the gas-coal price differential. The reduction in SO 2 emissions would be expected in this case to be less than our present estimate. On the other hand, a decrease in demand for electricity responding to the tax-induced increase in electricity prices would result in a reduction in emissions. The effects are obviously offsetting. A detailed analysis of the consequences is however beyond the scope of the present paper.

Figure 6
Savings in emissions of SO 2 relative to 2008 as a function of the pre-tax difference in costs for power generation using gas rather than coal (gas-coal) as estimated to result from imposition of a carbon tax on the power sector at three levels ($5/ton CO 2 , $10/ton CO 2 and $20/ton CO 2 ). The vertical axes to the left indicate the magnitude of the SO 2 reductions measured in thousands of tons SO 2 . The scales to the right present these data in terms of percentage reductions relative to 2009 for each region.

Discussion
The present paper focused on analysis of the changes in emissions of SO 2 that took place between 2008 and 2009, a period over which recession in the economy led to a reduction in demand for electricity and when prices for natural gas resulted in significant switching of fuel use in the power sector from coal to gas. We have applied the LMDI methodology also to the more extended time period from 1995 to 2010. Results from this analysis are presented in Figure S6 of SI. It suggests that the trend in emissions of SO 2 over this period closely followed trends in SO 2 emission intensities. Relative The results in Figure S6 indicate that the regulations promulgated by the US EPA played an important role in triggering the 67% decrease in emission intensity inferred to have taken place between 1995 and 2010 (28). The suggestion in this paper is that the recent decrease in prices for natural gas relative to coal is introducing a new consideration into this dynamics. The decrease is prompting replacement of older, less efficient coal-fired plants by newer, more efficient gas combined cycle plants.
The overall increase in removal of sulfur from the coal sector in 2009 reflected in part elimination of the older plants which generally did not incorporate technology for removal of sulfur. If gas prices remain low in the future, we may anticipate that this trend will continue with mothballing of increasing numbers of the more inefficient coal plants, with future increase in the overall efficiency for removal of sulfur from the coal segment of the power generation system. If on the other hand, gas prices recover, the power industry will be compelled to make further, expensive, investments to reduce overall emissions from the remaining coal-fired component of the power sector.
summarized in Figure S3 and related discussion. The percentages of total electricity generation using coal in the US in 2008 and 2009 are presented in Figure S4 as a function of the ages of the generating plants. In Figure S5, the percentages of coal consumed in US electric power generation in 2008 and 2009 by sulfur content are illustrated for four categories of coal. Results of applying the LDMI methodology to a more extended time period, from 1995 to 2000, are presented in Figure S6. Tables S1 and S2 summarize the data used for the first-level and second-level LMDI analyses, and Table S3 summarizes the immediate effect of carbon taxes in reducing SO 2 emissions for the nine census regions and for the contiguous US as a whole.