E. Approach to Power Sector Emission Variability
1. Introduction to Power Sector Variability	
      Variability is an inherent aspect of the production and delivery of electricity.  It follows that variations in state emissions are not only a result of variations in the level of emission control, but also are caused by the inherent variability in power generation.  In particular, emission variability is built into the design of power systems, which use a wide mix of power generation sources with varying use and emissions patterns to ensure reliability in electric power generation.  Variations in weather, demand due to changes in the level of economic activity, the portion of electric generation that is fossil-fuel-fired, the length and number of outages at power generation units, and other factors can lead  to significant variations in the load levels of different power generation sources.  Variations in the load levels of sources in any given state cause variations in the level of emissions in that state.  Variations in state emissions thus are not only a result of variations in the level of emission control, but also are caused by the inherent variability in power generation.  Thus, EPA believes it is appropriate, in this rule, to take into account the variations that are caused by inherent variability in power generation.  More specifically, variations in these external variables can cause significant fluctuations in state emissions, even when action has been taken to prohibit all emissions within a state that significantly contribute to nonattainment or interfere with maintenance in another state.  For this reason, EPA considers variability when determining the state specific requirements in this rule.  EPA does so by developing variability limits and assurance levels for each state, as described in this section, that are consistent with the statutory mandate of CAA section 110(a)(2)(D)(i)(I).        
      
Loads on a power system, and thus on power generation sources in a given state that are on the power system, vary over every time interval, changing not only in the short term and seasonally, but also annually.  As noted above, load patterns and levels are determined by a multiplicity of factors, including weather, economic activity, the portion of electric generation that is fossil-fuel-fired, and the length and number of outages at power generation units, which vary over time.  In particular, weather obviously varies not just from season-to-season but also from year-to-year, and even small changes in annual weather patterns can affect how the power system and power generation sources on the power system operate during a year.  For example, load, and the resulting use of generation sources on an interconnected grid to meet load, depend not only on how hot a summer day is, but also on where a heat wave occurs and how long it lasts.  Similarly, a relatively cold winter that drives up winter load may also change what generation sources are used to address the increased demand for heat.  Thus, the pattern of generation may shift geographically as a weather pattern moves across the country.  Because weather and other factors affecting loads, and the patterns of generation used to meet loads, vary over time and from state to state, the resulting level of emissions also varies over time and from state to state.  This variability in emissions is not a result of variation in emission rates, emission controls, or emission control strategies, but instead is a result of the inherent variability in power generation.  Patterns of generation change to ensure demand for electricity is met and to ensure continued reliability of the power system.  This results in temporal and geographic fluctuations in emissions.  In the final Transport Rule, like the proposed rule, EPA explicitly takes account of these changing patterns of generation and the resultant variability in power sector emissions.  As discussed previously, EPA identified a specific amount of emissions that must be prohibited by each state to meet the requirements of CAA section 110(a)(2)(D)(i)(I).  EPA also developed state baseline emissions for power generation sources based on projections of state emissions in an average year before the elimination of prohibited emissions, and state budgets for power generation sources based on projections of state emissions in an average year after the elimination of such emissions.  However, because of the inherent variability in state-level baseline emissions  - - resulting from the inherent variability in loads and power system and power generation source operations -- state-level emissions will fluctuate from year to year even after all significant contribution to nonattainment and interference with maintenance that EPA identified in this final rule are eliminated.  In an above average year, emissions may exceed state budgets based on an analysis of projected emissions in an average year.  EPA believes that, because baseline emissions are variable for reasons unrelated to the degree of emission control in a state and emissions after the elimination of all significant contribution to nonattainment and interference with maintenance are therefore also variable, it is appropriate to take this variability into account in developing the remedy for meeting the requirements of CAA section 110(a)(2)(D)(i)(I).  The variability limits and assurance levels in the final rule account for this inherent variability, while ensuring that emissions within each state that significantly contribute to nonattainment or interfere with maintenance in another state are prohibited.  EPA believes this approach is both reasonable in that it reflects the operation of the power system generation in order to maintain electric reliability and consistent with the statutory mandate of CAA section 110(a)(2)(D)(i)(I).  For these reasons, EPA is finalizing variability limits for each state budget to identify the range of emissions that EPA believes is likely to occur in each state following the elimination of all the state's significant contribution to nonattainment and interference with maintenance.  
      As discussed above, the air quality-assured trading remedy's state-specific budgets represent each state's emissions in an average year after elimination of significant contribution to nonattainment and interference with maintenance.  Because actual base case emissions are likely to vary from projected base case emissions, this remedy incorporates provisions that account for such variability.  While the primary purpose of this remedy is to address variability resulting from factors external to emissions controls, EPA believes variability limits also satisfy several other objectives.  The remedy provides the flexibility to deal with real-world variability in the operation of the power system through limited interstate trading and reduces costs of compliance with emission reduction requirements, while still providing assurance for downwind states that significant contribution to nonattainment and interference with maintenance by upwind states will be eliminated.  EPA believes the limited fluctuation in state level emissions that this approach permits is consistent with the statutory mandate of section 110(a)(2)(D)(i)(I) because some geographic and temporal shifting of emissions necessarily results from the inherent variability in power generation and is caused by factors unrelated to the degree of emission control such as weather, economic activity, and unit availability.  Far from excusing any state from addressing emissions within the state that significantly contribute to nonattainment or interfere with maintenance in other states, these variability limits ensure that the system can accommodate the inherent variability in the power sector while ensuring that each state eliminates the amount of emissions within the state, in a given year, that must be eliminated to meet the statutory mandate of section 110(a)(2)(D)(i)(I).  
      Moreover, the structure of the program, which ensures that there is no overall increase in emissions through state-specific assurance limits as well as a penalty mechanism, ensures that the variability limits only allow the temporal and geographic shifting of emissions that is a result of the inherent variability in power generation, and not decisions to avoid or delay the installation of necessary controls.  Under the remedy, an individual state can have emissions up to its budget plus the variability limit.  However, the requirement that all sources hold allowances covering emissions, and the fact that those allowances are allocated based on state-specific budgets without variability, ensure that the total emissions from the states do not exceed the sum of the state budgets.  The remedy, therefore, ensures both that total emissions do not exceed the total of the state budgets and that the required emission reductions occur in each state.
      This section describes how EPA calculated variability limits for each state to achieve this goal.
2. Transport Rule Variability Limits
      EPA performed analyses using historical data to demonstrate that there is year-to-year variability in base case emissions (even when emission rates for all units are held constant) and to quantify the magnitude of this variability.  
      The focus of the analysis is on quantifying the magnitude of the inherent year-to-year variability in state-level EGU emissions independent of measures taken to control those emissions (and thus due only to changes in electricity generation within each state).  EPA used this analysis to set variability limits as part of the remedy to ensure that states are eliminating their significant contribution to nonattainment and interference with maintenance to protect air quality.
      As discussed in detail below, EPA is finalizing the Transport Rule with 1-year variability limits calculated using a modified approach from the one described in the proposal.  EPA is not including the proposal's 3-year variability limits in the final Transport Rule.  EPA received comments that the 3-year variability limits increased program costs and diminished compliance flexibility without delivering any additional air quality benefits.  EGU owners and operators expressed concern that 3-year variability limits would be impracticable to implement and that the 1-year variability limits themselves would be adequately stringent to ensure elimination of significant contribution to nonattainment and interference with maintenance in each state.
      After further consideration, EPA has concluded that 3-year variability limits would be unnecessary, would be difficult to anticipate, and would not have a measurable impact on air quality benefits. EPA has determined that annual limits are sufficient to eliminate significant contribution to nonattainment and interference with maintenance in all upwind states while accommodating the historically observed year-to-year fluctuation in state-level EGU emissions even at the same rate of emissions control in a given state.
      In the proposal, EPA used statistical methods to derive the 3-year variability limit directly from the 1-year variability limit, meaning that the two are statistically equivalent in the long run under certain statistical assumptions.  Primarily, these assumptions were that the variation in electric demand around the budget is random from year-to-year and that, when the annual emissions are averaged over a multi-year time period, the average emissions per year will equal the state's budget.  The first assumption was also made in the assessment of the historical year-to-year variation in heat input in developing the 1-year limit (see section 2 of the "Power Sector Variability Final Rule TSD" for more details).  Regarding the second assumption, since the state-by-state emission budgets are based on the availability of emission reductions at an equal marginal cost level, EPA expects the sources in each of the upwind states to make these cost-effective reductions and to meet the emission budgets each year, on average. 
      Since the 3-year variability limit was based on average year-to-year variability over a longer time horizon, EPA notes that a random ordering of those years could yield 2 above-average years in a row.  If, by chance, a third above-average year were to follow, the state could face violation of the 3-year limit, even if over a time period longer than 3 years, that state would never have exceeded the statistically-equivalent 1-year variability limit and its annual emissions would have averaged to the level of its budget.  Effectively, this means that imposing a multi-year variability limit would erode the 1-year variability limit's ability to accommodate historically observed year-to-year variability in state-level EGU emissions (due only to generation changes), and it would do so without providing any additional air quality benefits or protection for downwind areas (since the average emissions over the long time horizon equal the level of the budget).
      For more details about the relationship between the 1- and 3-year limits, see the discussions in section 3 of the "Power Sector Variability" TSD from the proposed Transport Rule, which describes the derivation of the 3-year limit from the 1-year variability and section 3 of the "Power Sector Variability Final Rule TSD", which describes the results of a numerical simulation showing that the 1- and 3-year limits are statistically indistinguishable and, thus, redundant over the course of the program to accommodate year-to-year variability.  
      While EPA expects the yearly emissions in each state, on average, to equal the level of the budgets, EPA also estimated the air quality impacts of 5, 10, 15, and 20 percent emission variability using the air quality assessment tool (see section 4 of the "Power Sector Variability Final Rule TSD" for more details).  This analysis shows that year-to-year fluctuations of up to 20 percent in SO2 emissions from upwind states linked to a given downwind receptor do not undermine the ability of the Transport Rule programs to resolve nonattainment or maintenance concerns at that receptor.  This analysis was designed to examine the sensitivity of downwind air quality to upwind EGU emission levels.  The share of total SO2 emitted by EGUs is significantly larger than the share of total NOX emitted by EGUs.  For example, in the states for which EPA modeled base case contributions of these pollutants, EGUs accounted for 74 percent of total SO2, 14 percent of total annual NOX, and 15 percent of total ozone-season NOX emissions.  Therefore, when varying EGU emissions only, downwind air quality would be most sensitive to upwind variations in SO2, because relative variations in EGU SO2 emissions have a greater impact on total SO2 emissions than the same relative variation in EGU NOX emissions would have on total NOX emissions affecting downwind air quality. 
      Furthermore, because the state budgets are based directly on IPM modeling of electric generation when cost-effective emission reductions have been achieved, sources within each state should have the same incentive to meet that budget, on average, in any given year.  Additional EPA analysis supports the claim that states would be no more likely to exceed 1-year variability limits without the 3-year limits than with the 3-year limits.  See the "Power Sector Variability Final Rule TSD" for more details on this statistical analysis.  Finally, because the state budgets (and thus the total amount of allowances available) are fixed and every covered source must hold allowances covering its emissions, it is not feasible for all, or even many, states to repeatedly exceed their budgets.   
      The approach calculated the standard deviation in state-level heat input from units expected to be covered by the final Transport Rule over an 11-year time period (2000 through 2010), from which the 95th percent confidence level was calculated.  EPA divided this value by the mean to get the percentage variation in heat input.  The two-tailed 95th percent confidence level is the equivalent of the 97.5 percent upper (single-tailed) confidence level.  This approach yielded an average year-to-year heat input variability for each state, as a proxy for historic year-to-year variability in state-level EGU emissions while holding emission rates constant.  The result, expressed as a percentage, conveys the maximum degree to which EGU emissions at the state level may be expected with 95th percent confidence to vary around a given target (i.e., budget) from year-to-year, on average, based on the statistical analysis of historic heat input over the 2000 through 2010 time period.
      From the state-by-state variability calculations, EPA identified a single variability level (percentage) for each of the annual and ozone-season programs based on the historic variability measured at units in covered states on an annual basis and an ozone-season basis, respectively.  In the proposal, EPA "identified a single set of variability levels . . . to apply to all states in order to make the application of the variability limits straightforward rather than developing state-by-state percentage variability values" (75 FR 45293).  In the final rule, EPA is taking the straightforward approach of identifying a single set of variability levels to apply to all states because EPA has determined that it is reasonable to afford all states under the Transport Rule programs the extent of measured historic variability experienced by any Transport Rule state during 2000 through 2010.  In the variability analysis for the final rule, EPA identified Tennessee as having the highest measured historic variability of annual heat input of 18 percent, and Virginia as having the highest measured historic variability of ozone-season heat input of 21 percent.  Because the percentage of variability in Tennessee on an annual basis and in Virginia on an ozone-season basis are reasonably likely to occur in each of the other states in the future, EPA believes it is appropriate to apply an 18 percent annual variability limit to all states covered by the annual SO2 and NOX programs and a 21 percent ozone-season variability limit to all states covered by the ozone-season NOX program.
      EPA's analysis of historic heat input variability in multiple states over the 2000 to 2010 baseline yields a range of potential year-to-year variability values for state-level EGU emissions.  As discussed above, any one state's measured variability (in this case, from 2000 to 2010) is due to a multiplicity of factors.  These factors include, but are not limited to, variation in weather, variation in demand due to increased or decreased level of economic activity, variation in the portion of electric generation that is fossil-fuel-fired, and variation in the length and number of outages at power generation units, and these individual factors may sometimes act in concert and may other times be offsetting.  
      The mix and levels of factors present in a state from year-to-year can lead to variation of state-level emissions above and below the level for the state under average conditions.  Because the levels of the various factors are difficult to predict on a year-to-year basis for an individual state, the resulting variability in state-level emissions is difficult to predict.  Moreover, because the electric generation, transmission, and distribution system in the eastern half of the U.S. is highly integrated, year-to-year variation in these factors in one state can cause year-to-year variability in state-level emissions both in that state and in other states on the system.  For example, increased demand due to extreme weather or increased economic activity in one state can be met through increased generation and emissions a number of states away?/in a number of other states?.  
      Because these factors can vary year-to-year in every state in ways that are difficult to predict and can affect other states, EPA maintains that the maximum variability measured in one state for a discrete period (2000-2010) is reasonably likely to occur in the future in any of the states in the region.  Consequently, EPA believes that it is reasonable to use the maximum historic percentage variability figure as a proxy for the percentage variability that any of the states is likely to experience in the future.  Although EPA is therefore using a uniform percentage figure for variability, EPA applies that percentage figure to each state-specific budget so that variability in tons of emissions is determined on a state-specific basis.  That state-specific number is used in determining whether the assurance provisions and penalty are triggered in the specific state. EPA also believes that it is appropriate to accommodate this potential future variability at the state level if and only if it can be accommodated without undermining the programs' beneficial impacts on downwind air quality that eliminate significant contribution to nonattainment or interference with maintenance of the NAAQS assessed in this rulemaking (see the "Power Sector Variability Final Rule TSD" for more information on this analysis).  The Transport Rule identifies and quantifies, on a state-by-state basis, the emissions in each state that significantly contribute to nonattainment or interfere with maintenance in another state.  This is done by analyzing specific air pollution linkages between each upwind state and each downwind maintenance or nonattainment receptor.    Nonetheless, it is clear from the air quality analyses that the air quality outcome at a given downwind receptor is a function of the cumulative emissions from all upwind states and the receptor's home state.  Once the Transport Rule emission reduction requirements are implemented in all states subject to the programs, EPA's analysis shows that the impact on a downwind receptor of any single upwind state's year-to-year fluctuation of up to 20 percent in SO2 emissions would be so limited as to not disturb that receptor's ability to maintain or attain the NAAQS analyzed in this rulemaking.  Therefore, to the extent that such variability has been measured in historic data in any state subject to the Transport Rule programs, it is reasonable to provide for potential future variability in Transport Rule states within the scope of what EPA's analysis shows to preserve downwind air quality gains achieved by the Transport Rule programs.
      The approach to establishing variability limits in the final rule modifies the approach from the proposed rule in two ways.  First, EPA is applying only a percentage variability limit to each budget in the final rule, whereas the proposed rule applied the greater of a percentage or an absolute tonnage variability limit to each budget.  EPA explained in the proposal that it was necessary to impose both a percentage and a tonnage limit due to the inclusion of "states with small numbers of units where expected variability would be more pronounced in percentage terms" (75 FR 45293).  However, the states with the smallest numbers of units included at proposal (such as Connecticut and the District of Columbia) are not covered by any of the final Transport Rule's programs.  In the final rule's variability analysis, Tennessee has the highest measured annual variability percentage and Virginia has the highest measured ozone-season variability percentage.  Both of these states have a sufficient number of units for the percentage variability findings to be representative of variability in all of the Transport Rule states; therefore, it is not necessary to impose a tonnage limitation in the final rule.
      Second, EPA has expanded the historic baseline of the variability analysis to consider heat input data from 2000 through 2010, as compared to 2002 through 2008 at proposal, and EPA has also expanded the dataset to include all units expected to be covered by the final Transport Rule's programs.  EPA received a number of comments that the proposal's variability limits were too stringent in part because they relied on too short a historical baseline that failed to capture the full extent of long-run year-to-year variability.  EPA agrees with these comments and believes that the historic baseline modification described above supports variability limits in the final rule that are a better approximation of future potential year-to-year variability in state-level EGU emissions around the budgets as a function of inherent variability in baseline state-level EGU operations.  EPA believes the 2000 through 2010 historic baseline supports a more accurate approximation of year-to-year variability in state-level EGU operations than previously measured on a 2002 through 2008 baseline.
Some commenters expressed the view that allowing variability limits in addition to state budgets undermines the requirements of CAA section 110(a)(2)(D)(i)(I) to eliminate significant contribution to nonattainment and interference with maintenance of the NAAQS in downwind states.  EPA disagrees with these comments.  As explained above, EPA finds that year-to-year variability is an inherent characteristic of power sector emissions whether or not such emissions are controlled by state budgets; the future year-to-year variability is a component of the sector's emissions baseline before emission reductions are required.  As done for proposal, EPA has analyzed the impact of allowing emissions from upwind states in a given year to rise above the budgets but within the variability limits allowed in the final rule.  This analysis shows that emission fluctuations around the budgets but within the variability limits will not undermine the downwind air quality gains achieved by the implementation of the Transport Rule budgets, and therefore the variability limits cannot be said to undermine the elimination of significant contribution to nonattainment or interference with maintenance achieved under the Transport Rule programs.  Based on historical data and projected air quality impacts, the Agency believes that states will have sufficient flexibility and room to operate within the final rule's variability limits while addressing all emissions identified as significantly contributing to nonattainment or interfering with maintenance in other states.
