ࡱ> qsnop5@ bjbj22 XX6%4NJNJNJhJBLujM@P"PPPPn:SS\ """"""$RkfF^PP^^FPP .iii^PP i^ i ii00XP^M .NJd(`$ 9<uѠgrѠHXѠXJTDW idY[~JTJTJTFF$B GD(iX GFINAL REPORT ECONOMIC MODELING TO EVALUATE THE POTENTIAL COSTEFFECTIVENESS OF MULTIPLE ABATEMENT OPTIONS FOR EMISSIONS SOURCES IN THE CEMENT MANUFACTURING INDUSTRY AND THE IRON AND STEEL MANUFACTURING INDUSTRY IN IMPROVING AIR QUALITY IN RESIDUAL PM2.5 NON-ATTAINMENT AREAS PREPARED FOR: Program Design Group Sector Policies and Programs Division Office of Air Quality Planning and Standards U.S. Environmental Protection Agency Research Triangle Park, NC 27711 PREPARED BY: CONSAD Research Corporation 121 North Highland Avenue Pittsburgh, PA 15206 Under Contract Number GS-23F-0259S June 6, 2008 TABLE OF CONTENTS  TOC \o "1-3" \h \z \u  HYPERLINK \l "_Toc198636115" 1.0 Introduction and Background  PAGEREF _Toc198636115 \h 1  HYPERLINK \l "_Toc198636116" 2.0 General Structure of the Expanded Linear Programming Model  PAGEREF _Toc198636116 \h 2  HYPERLINK \l "_Toc198636117" 2.1 Input variables  PAGEREF _Toc198636117 \h 3  HYPERLINK \l "_Toc198636118" 2.2 Initial Conditions  PAGEREF _Toc198636118 \h 5  HYPERLINK \l "_Toc198636119" 2.3 Decision Variables  PAGEREF _Toc198636119 \h 5  HYPERLINK \l "_Toc198636120" 2.4 Consequences of Control Application  PAGEREF _Toc198636120 \h 6  HYPERLINK \l "_Toc198636121" 2.5 Objective Function: Cost Minimization  PAGEREF _Toc198636121 \h 6  HYPERLINK \l "_Toc198636122" 2.6 Constraints on Cost Minimization  PAGEREF _Toc198636122 \h 8  HYPERLINK \l "_Toc198636123" 2.7 Outputs from Linear Programming Model  PAGEREF _Toc198636123 \h 9  HYPERLINK \l "_Toc198636124" 2.8 Data Flow  PAGEREF _Toc198636124 \h 13  HYPERLINK \l "_Toc198636125" 3.0 Data Sources and Data Used in Applying Linear Programming Model  PAGEREF _Toc198636125 \h 16  HYPERLINK \l "_Toc198636126" 3.1 PM2.5 Non-attainment Areas  PAGEREF _Toc198636126 \h 16  HYPERLINK \l "_Toc198636127" 3.2 Transfer Coefficients  PAGEREF _Toc198636127 \h 17  HYPERLINK \l "_Toc198636128" 3.3 Source Emissions  PAGEREF _Toc198636128 \h 18  HYPERLINK \l "_Toc198636129" 3.4 Emission Control Measures and Costs  PAGEREF _Toc198636129 \h 19  HYPERLINK \l "_Toc198636130" 4.0 Structure of the Analysis  PAGEREF _Toc198636130 \h 20  HYPERLINK \l "_Toc198636131" 5.0 Results from the Analysis  PAGEREF _Toc198636131 \h 21  HYPERLINK \l "_Toc198636132" 6.0 Conclusions  PAGEREF _Toc198636132 \h 23  HYPERLINK \l "_Toc198636133" 7.0 References  PAGEREF _Toc198636133 \h 23  HYPERLINK \l "_Toc198636134" 8.0 Quality Assurance Report  PAGEREF _Toc198636134 \h 25  HYPERLINK \l "_Toc198636135" 8.1 Data Acquisition  PAGEREF _Toc198636135 \h 25  HYPERLINK \l "_Toc198636136" 8.2 Data Management  PAGEREF _Toc198636136 \h 26  1.0 Introduction and Background In this economic modeling project, CONSAD Research Corporation has performed a comparative screening analysis of multiple abatement options for emissions sources in the cement manufacturing industry and the iron and steel manufacturing industry. The analysis has been performed using a linear programming model that calculates the allocation of control responsibility among emissions sources that will achieve specified improvements in ambient concentrations of fine particulate matter (particles with aerodynamic diameters of 2.5 microns or less: PM2.5) in multiple PM2.5 non-attainment areas at minimum total abatement cost. The analysis has therefore considered: emissions of PM2.5 and its precursors [i.e., oxides of nitrogen (NOx) and sulfur dioxide (SO2)] from individual sources in both of those industries; emissions controls that are available for reducing emissions of PM2.5, NOx, and/or SO2 from those types of sources, the control efficiencies and the total annualized costs of the controls; and the impacts of reductions in emissions of PM2.5, NOx, and/or SO2 from individual sources on ambient PM2.5 concentrations in specific non-attainment areas. The screening analysis has been expressly designed to provide valid comparison of results produced by the linear programming model for different plausible scenarios for the cement manufacturing industry and the iron and steel manufacturing industry separately and jointly. The linear programming model used in the project has been developed progressively in two previous studies conducted for the U.S. Department of Energy (DOE). In the first study, CONSAD (2004) has developed and applied a computer-based model that simulates the development of state implementation plans (SIPs) for reducing annual average ambient concentrations of PM2.5 in individual non-attainment areas. In the second study, CONSAD and Booz Allen Hamilton, Inc. (2006) have collaborated in extending the model to incorporate results from recent atmospheric research, conducted under the joint sponsorship of the National Energy Technology Laboratory (NETL) of the DOE and the Supersites Program of the U.S. Environmental Protection Agency (EPA), that has found that the major portion of the ambient concentration of PM2.5 in an air quality control region (AQCR) is caused by long-range transport of the pollutant and its precursors from sources outside the AQCR. In the second study, the computer-based model developed and used in the first study has been revised by suitably expanding the objective that the model seeks to optimize. The revised objective is minimization of the total abatement cost that is required to achieve, or come as close as possible to achieving, the annual national ambient air quality standard (NAAQS) for PM2.5 in all non-attainment areas concurrently. Minimization of the revised objective function has required application of standard linear programming methods. In that initial application of the linear programming model, the number of industries and associated emission sources that have been considered candidates for control has been sufficiently large that achieving the annual NAAQS for PM in all, or almost all, non-attainment areas might be a realistic prospect. Also, only one emission control measure has been specified as a candidate for adoption by any individual source, although a variety of different measures has been specified among all of the sources. For the screening analysis conducted in the current project, the linear programming model has been further refined to overcome these limitations. The refinements enable analysis of cost-effective air quality improvement for the relatively small numbers of emission sources that typically are contained within individual industries or small groups of industries, and for situations in which two or more different control measures are candidates for application separately or jointly to individual emission sources. The structure of the expanded model is specified and explained in Section 2.0. Section 3.0 then discusses the data sources and data that have been used to apply the model in the screening analysis. The structure of the analysis is described in Section 4.0. Results from the analysis are presented in Section 5.0. Conclusions are presented in Section 6.0. References used in the projects are listed in Section 7.0. The quality assurance report for the project is contained in Section 8.0. 2.0 General Structure of the Expanded Linear Programming Model The linear programming model that has been applied in this project has been expressly designed to determine the cost-effective allocation of responsibility for controlling emissions of PM2.5 and its precursors among sources in specific industries that are located in specific origination areas. Thus, the allocation of control responsibility that is determined by the model will achieve at minimum cost specified improvements in the ambient concentration of PM2.5 in all non-attainment areas where air quality is affected by emissions from those sources. To facilitate clear description of the general structure of the linear programming model, the description has been divided into eight components. Those components are discussed successively below. First, the input variables for the model are specified in Section 2.1. The input variables define the detailed information that must be compiled or estimated to enable derivation of the specific data about individual emission sources or groups of equivalent sources that are entered into the linear programming model. Second, the initial air quality conditions that are described by the values developed for the input variables are designated in Section 2.2. Third, the decision variables for which values are calculated in the solution to the linear programming model are specified in Section 2.3. Fourth, in Section 2.4 estimators for the consequences of implementing the decisions indicated by the values estimated for the decision variables are presented, and their practical interpretations are explained. Fifth, in Section 2.5 the objective function for the linear programming model is specified as the sum of a pertinent subset of the consequence estimators. Sixth, the constraints that restrict the ability of the linear programming model to identify solutions that attain even more inexpensive allocations of control responsibility among emission sources are specified, and their practical interpretations are explained in Section 2.6. The constraints also are specified in terms of sums of different pertinent subsets of the consequence estimators. Seventh, in Section 2.7, outputs that can be computed from the solution to the linear programming model are described, and practical options for presenting them at different levels of detail are discussed. Finally, the flow of data into, through, and out of the model is depicted graphically and explained in Section 2.8. 2.1 Input variables Six basic sets of input data are required for calibrating and solving the linear programming model. They are: IEMpsr = Initial annual emissions of pollutant p (PM2.5 or a precursor) from source s in origination area r (either a specific PM2.5 non-attainment area designated by the EPA, or a specific area outside any of those non-attainment areas that contains one or more pertinent sources) (p=1,,P; r=1,,R; and s=1,,Sr) EFFmpsr = Control efficiency (proportional reduction in emissions) of control measure m (m=1,,M) applied to pollutant p emitted from source s in origination area r ACmpsr = Total annualized cost incurred when applying control measure m to decrease emissions of pollutant p from source s in origination area r. The total annualized cost of a control measure includes the capital cost and the annually recurring costs involved in purchasing, installing, operating, and maintaining the control equipment. The control efficiency and the average annualized cost per ton of emissions for a particular control measure applied to a particular pollutant are the same in all situations involving that control measure. Different versions of a general type of control measure are distinguished by different values of m, and have separate values for their control efficiencies and their average annualized cost per ton of emissions. The specific control measures that are pertinent to specific sources, s, in specific origination areas, r, are specified in a separate data table that links to the data file containing the control efficiencies and the average annualized cost per ton of emissions for the various control measures. Those links are subsumed within the subscripts specified for the pertinent variables in this specification of the overall linear programming model. TRCpsrn = Transfer coefficient specifying the decrease in ambient concentration of PM2.5 (in g/m3) in non-attainment area n (n=1,& ,N) that is realized by reducing annual emissions of pollutant p from source s in origination area r by one unit (typically, one ton). Arrays of transfer coefficients have been developed for sources in specific origination areas. A separate coefficient has been estimated for each combination of a specific pollutant, origination area, and non-attainment area that is pertinent to the sources under consideration. Each source is linked to the array of transfer coefficients for its origination area. IDVn = Initial design value for the annual average ambient concentration of PM2.5 (in g/m3) in non-attainment area n. The design value for each area is  & the mathematically determined pollutant concentration at a particular site that must be reduced to, or maintained at or below the National Ambient Air Quality Standards (NAAQS) in order to assure compliance. (Chu, undated, p. 1) The design value for any non-attainment area is the highest annual average ambient concentration of PM2.5 computed by the U.S.EPA for any monitoring site where complete data have been compiled within the counties that are included in the area. Bn = Background level of the annual average ambient concentration of PM2.5 (in g/m3) in non-attainment area n. The background level is the portion of the annual average ambient concentration of PM2.5 in non-attainment area n that is caused by emissions from sources other than those in the origination areas r (r = 1, & , R) that are candidates for additional control in the scenario under consideration in the linear programming model. Bn is the annual average ambient concentration of PM2.5 that would occur in non-attainment area n if all emissions from the sources that are candidates for additional control were simultaneously reduced to zero. IMPn = Specified improvement sought in the annual average ambient concentration of PM2.5 in non-attainment area n. 2.2 Initial Conditions The values that are compiled or estimated for the input variables that have been identified and described in the Section 2.1 can be used to compute the initial conditions that are described below. IEMpsr % TRCpsrn = Increase in the annual average ambient concentration of PM2.5 in non-attainment area n that is caused by the initial annual emissions of pollutant p from source s in origination area r. Summing the increases for all pollutants, sources, and origination areas produces: R Sr P    IEMpsr % TRCpsrn = Increase in the annual average ambient r=1 s=1 p=1 concentration of PM2.5 in non-attainment area n that is caused by the initial annual emissions of all pertinent pollutants (p=1,, P) from all sources (s=1,,Sr) in all origination areas (r=1,,R). Then, adding the background level yields: R Sr P    IEMpsr % TRCpsrn + Bn = IDVn, the initial design value for the annual average r=1 s=1 p=1 ambient concentration of PM2.5 in non-attainment area n. 2.3 Decision Variables The linear programming model includes one decision variable for each emission source or group of equivalent sources that is a candidate for a specific additional control in the scenario that is being analyzed. The decision variable and its interpretations are: Dmpsr Degree (or probability) of application of control measure m to pollutant p emitted from source s in origination area r (0 < Dmpsr < 1). The interpretation of the value of Dmpsr as the degree of application of a control measure is more appropriate when the decision variable relates to applying a specific proportion of the total achievable control to a single emission source. Its interpretation as the probability that a control measure is applied to a particular source is apt when the decision variable relates to a group of equivalent sources. Under those circumstances, cost-effective allocation of control responsibility would likely involve, in practice, full application of the control measure to a portion of the emission sources in the group. Hereafter, whenever interpretations of mathematical expressions that include the decision variable Dmpsr are presented, the word degree should be interpreted as alternatively meaning either proportion or probability. 2.4 Consequences of Control Application If control measure m is only applied to reduce emissions of pollutant p from source s in origination area r, the following three consequences will occur: IEMpsr % EFFmpsr % Dmpsr = Reduction in annual emissions of pollutant p from source s in origination area r that is achieved by applying control measure m to the degree Dmpsr. IEMpsr % EFFmpsr % TRCpsrn % Dmpsr = Decrease in the annual average ambient concentration of PM2.5 in non-attainment area n that is caused by the reduction in annual emissions of pollutant p from source s in origination area r that is achieved by applying control measure m to the degree Dmpsr. ACmpsr % Dmpsr = Total annualized cost of applying control measure m to the degree Dmpsr to reduce emissions of pollutant p from source s in origination area r. 2.5 Objective Function: Cost Minimization The objective of the linear programming model is to determine the degrees to which specific control measures should be applied to reduce emissions of specific pollutants from specific sources in specific origination areas, so as to minimize the cost of achieving specified reductions of the annual average ambient concentration of PM2.5 in all non-attainment areas that have been designated by the EPA throughout the eastern U.S. In December 2004, the EPA designated its final list of PM2.5 non-attainment areas that includes 47 areas containing 216 counties in 21 states. Four of the areas containing 14 counties are located in the West (three in California and one in Montana). The remaining 43 non-attainment areas, containing 202 counties in 19 states in the East, are the areas for which transfer coefficients have been estimated by CONSAD and Booz Allen Hamilton, Inc. (2006). Thus, they are the AQCRs for which the linear programming model can determine how to achieve the annual PM2.5 NAAQS cost-effectively. The counties in those 43 areas plus a small number of additional counties in the eastern U.S. for which transfer coefficients have been estimated comprise the origination areas where control measures may be applied to achieve that objective. The objective function therefore is to minimize: R Sr P M     ACmpsr % Dmpsr r=1 s=1 p=1 m=1 = Total annualized cost of applying all pertinent control measures (m=1,& ,M) to the specified degrees Dmpsr (m=1,& ,M; p=1,& ,P; r=1,& ,R; and s=1,& ,Sr) to reduce emissions of all pertinent pollutants (p=1,,P) from all pertinent sources (s=1,,Sr) in all pertinent origination areas (r=1,,R). In the initial application of the objective function in the second study conducted for DOE, it was assumed that only one control measure was pertinent to apply to any pollutant that was emitted from any source. Thus, a single, specific value of m was associated with each pertinent combination of p, s, and r. No cost, no emission reduction, and no improvement in air quality would be associated with any other value of m for that combination of p, s, and r. Under those conditions, including summation over the various values of m in the mathematical statement of the objective function would have been redundant and was omitted. In the current project, two or more suitable control measures have been identified that might be determined to be cost-effective means for reducing emissions of specific pollutants from many specific types of sources. Whenever two or more control measures are available for reducing emissions of any pollutant from any source, it is first necessary to determine whether any of the control measures has both lower or equal control efficiency, EFFmpsr, and higher or equal average annualized cost per ton, ACTmp, than another measure. If so, the first measure is dominated by the second measure and should be eliminated from further consideration. The remaining control measures must then be arranged in increasing order of control efficiency and average annualized cost per ton. Next, for each control measure, EFFmpsr and ACmpsr should be measured as the incremental control efficiency and incremental total annualized cost in excess of the total control efficiency and total annualized cost of the next most efficient control measure. Thus, for the least efficient measure, the values of EFFmpsr and ACmpsr will remain unchanged; and for each other measure that, if adopted, would supplant a less efficient measure, the values of EFFmpsr and ACmpsr will become equal to the difference between their initial total values and the initial total values of the next most efficient measure. Specifically: EFFmpsr should be set equal to the initial [EFFmpsr EFF(m-1)psr], and ACmpsr should be set equal to the initial [ACmpsr - AC(m-1)psr] An exception to this general calculation must be used for control measures (such as increased monitoring frequency or continuous emission monitoring, which reduce the time during which a leak or other failure of extant emission control remains undetected, and hence before which repair can be initiated) that supplement rather than supplant the other control measure. Under these circumstances: ACmpsr should be set equal to the product of the average annualized cost per ton of the supplemental measure, ACTmp, and the initial annual emissions of the pollutant from the source, IEMpsr ,and EFFmpsr should be set equal to the smaller of (a) the specified control efficiency for the supplemental control measure and (b) the difference between the maximum realistically achievable control efficiency and the initial control efficiency for the other control measure that is supplemented (i.e., the sum of the incremental control efficiencies of all of the control measures that have been successively adopted). 2.6 Constraints on Cost Minimization The minimization of the objective function must be accomplished, however, while achieving the specified improvements in the ambient concentration of PM2.5 in all non-attainment areas where air quality is affected by emissions from those sources. For each non-attainment area n (n=1,& ,43), the pertinent constraint is: R Sr P M     ( IEMpsr % EFFmpsr % TRCpsrn % Dmpsr ) > IMPn r=1 s=1 p=1 m=1 As explained previously, the expression in parentheses represents the decrease in the annual average ambient concentration of PM2.5 in non-attainment area n that is caused by the reduction in annual emissions of pollutant p from source s in origination area r that is achieved by applying control measure m to the degree Dmpsr. Thus, the summation over r, s, p, and m yields the decrease in the annual average ambient concentration of PM2.5 in non-attainment area n that is caused by the reductions in annual emissions of all pertinent pollutants (p=1,..,P) from all pertinent sources (s=1,,Sr) in all pertinent origination areas (r=1,,R) that are achieved by applying all pertinent control measures (m=1,,M) to the degrees Dmpsr. In situations where the number of industries and associated emission sources that have been considered candidates for control is sufficiently large that achieving the annual NAAQS for PM2.5 in all, or almost all, non-attainment areas might be a realistic prospect, the value specified for IMPn becomes the difference between the initial design value for the annual average concentration of PM2.5 in non-attainment area n (IDVn) and the NAAQS. Under those circumstances, the pertinent constraint is: R Sr P    ( IEMpsr % EFFmpsr % TRCpsrn % Dmpsr ) > (IDVn - NAAQS) r=1 s=1 p=1 In this form of the constraint, the term on the right-hand side, (IDVn - NAAQS), represents the amount by which the annual average ambient concentration of PM2.5 in non-attainment area n must be reduced to achieve the NAAQS in that area. In addition to these 43 constraints, there are two additional constraints that apply to each decision variable, Dmpsr (m=1,,M; p=1,,P; r=1,,R; and s=1,,Sr). They are: Dmpsr > 0 and Dmpsr < 1. If only one control measure is available for application to any pollutant that is emitted from any source, these constraints are the only restrictions on the values that may be determined for the decision variables. If two or more control measures are available for application to a source either alternatively or concurrently, an additional set of constraints must be imposed. For each control measure that might supplant or supplement a control measure with lower control efficiency, the following constraint must be established: Dmpsr < D(m-1)psr, in addition to the usual constraints that 0 < Dmpsr < 1. 2.7 Outputs from Linear Programming Model The data that will be used to develop the outputs that will be reported to summarize the results from the solution to any scenario that is analyzed in the linear programming model will consist of the values of two sets of variables: the input variables and the decision variables in the model. If the value computed for a particular decision variable, Dmpsr, is between zero and one, costs will be minimized by applying the corresponding control measure to emissions of the corresponding pollutant from the corresponding source to the specified degree. If the computed value is one, costs will be minimized by applying that control measure fully to that source and pollutant. If the computed value is zero, costs will be minimized by not applying that control measure to that source and pollutant at all. To specify equations for computing the outputs that will be reported from the values of the input and decision variables, it will often be useful to add one or more additional subscripts to the input variables that measure initial annual emissions and total annualized costs for specific pollutants from specific sources. Those subscripts will indicate other pertinent attributes of the source, such as the industry that operates it. Accordingly, the input variables for specific sources will generally have the following form: IEMpskr = Initial annual emissions of pollutant p (PM2.5 or a precursor) from source s in category k (e.g., in source classification k or operated by industry k) in origination area r (p=1,,P; s=1,,Sr; k=1,,K; and r=1,,R), and ACmpskr = Total annualized cost incurred when applying control measure m to decrease emissions of pollutant p from source s in category k in origination area r (m=1,,M; p=1,,P; s=1,,Sr; k=1,,K; and r=1,,R). The solution for any scenario analyzed using the linear programming model will specify the degrees to which specific control measures should be applied to reduce emissions of specific pollutants from specific sources or source categories in specific origination areas, so as to minimize the cost of achieving the specified reductions of the annual average ambient concentrations of PM2.5 in all non-attainment areas that have been designated by the EPA throughout the eastern U.S. The types of outputs that can be provided on the basis of those results and the associated input data are discussed below. Total Abatement Cost in All Non-attainment Areas The value calculated for the objective function will be the minimum value achievable for: R Sr P M TACN =     ACmpskr % Dmprs r=1 s=1 p=1 m=1 = Total annualized abatement cost for all pertinent origination areas in the eastern U.S. = Total annualized cost of applying all pertinent control measures (m=1,,M) to the specified degrees Dmprs (m=1,,M; p=1,,P; r=1,,R; and s=1,,Sr) to reduce emissions of all pertinent pollutants (p=1,,P) from all pertinent sources (s=1,,Sr) in all pertinent origination areas (r=1,,R) sufficiently to attain the specified reductions in the annual average ambient concentration of PM2.5 in all designated non-attainment areas throughout the eastern U.S. In this expression: ACmpskr % Dmprs = Total annualized cost of applying control measure m to the degree Dmprs to reduce emissions of pollutant p from source s in category k (e.g., in source classification k or operated by industry k) in origination area r. Ambient Air Quality in Individual Non-attainment Areas For each non-attainment area n (n=1,& ,43), the projected design value for its ambient concentration of PM2.5 after implementation of the cost-effective incremental controls is: R Sr P M IDVn -     ( IEMpskr % EFFmpsr % TRCpsrn % Dmprs ) r=1 s=1 p=1 m=1 The expression in parentheses represents the decrease in the annual average ambient concentration of PM2.5 in non-attainment area n that is caused by the reduction in annual emissions of pollutant p from source s in origination area r that is achieved by applying control measure m to the degree Dmprs. Thus, the summation over r, s, p and m yields the total decrease in the annual average ambient concentration of PM2.5 in non-attainment area n that is caused by the reductions in annual emissions of all pertinent pollutants (p=1,..,P) from all pertinent sources (s=1,,Sr) in all pertinent origination areas (r=1,,R) that are achieved by applying all pertinent control measures (m=1,,M) to the degrees Dmprs. Subtracting that total from the initial level of the annual average ambient concentration of PM2.5 in non-attainment area n (IDVn) produces the projected design value for the annual average ambient concentration of PM2.5 in non-attainment area n after applying those control measures to those sources to those degrees. In addition, the total decrease in the annual average ambient concentration of PM2.5 in non-attainment area n can be separated into the portions that are caused by the reductions in annual emissions of specific pollutants, or the reductions from all sources in specific categories (e.g., specific source classifications or specific industries), or the reductions from specific origination areas, or the reductions from application of specific control measures. The separate portions can be estimated by computing appropriate sums of the individual decreases in the annual average ambient concentration of PM2.5 in the non-attainment area that are caused by the reduction in annual emissions of specific pollutants from specific sources in specific origination areas that are achieved by applying specific control measures to specified degrees, Dmprs. Abatement Costs in Origination Areas Applying Incremental Control Measures For each origination area in which incremental control measures are applied to any sources, the total annualized cost of applying the cost-effective control measures in that origination area are: Sr P M TACr =    ACmpskr % Dmprs s=1 p=1 m=1 = Total annualized cost of applying all pertinent control measures to the specified degrees Dmprs to reduce emissions of all pertinent pollutants from all pertinent sources in origination area r sufficiently to attain the specified reductions in the annual average ambient concentration of PM2.5 in all designated non-attainment areas throughout the eastern U.S. Abatement Costs in Industries Applying Incremental Control Measures Similarly, the total annualized abatement cost incurred in each industry (or in any other category specified on the basis of the attributes of individual emission sources) within an origination area can be calculated by summing the incremental abatement costs for all emission sources operated by the industry (or the sources in the specified category) within the origination area. Sr P M TACkr =    ACmpskr % Dmprs s=1 p=1 m=1 sTk = Total cost of applying all pertinent control measures to the specified degrees Dmprs to reduce emissions of all pertinent pollutants from all pertinent sources in category k (e.g., operated by industry k) in origination area r sufficiently to attain the specified reductions in the annual average ambient concentration of PM2.5 in all designated non-attainment areas throughout the eastern U.S. Then, the total annualized abatement cost incurred in each industry (or other specified category) throughout all origination areas in which incremental control measures are applied to any sources can be calculated by summing the values calculated for the industry (or category) in the individual origination areas. R TACk =  TACkr r=1 R Sr P =    ACmpskr % Dmprs r=1 s=1 p=1 sTk = Total annualized cost of applying all pertinent control measures to the specified degrees Dmprs to reduce emissions of all pertinent pollutants from all pertinent sources in category k (e.g., operated by industry k) in all pertinent origination areas sufficiently to attain the specified reductions in the annual average ambient concentration of PM2.5 in all designated non-attainment areas throughout the eastern U.S. It is important to note that it is not possible to develop a non-arbitrary estimate of the total cost of attaining the NAAQS for PM2.5 in any individual non-attainment area. Because long-range transport generally conveys primary PM2.5, secondary PM2.5, and the various precursors from an emission source through several non-attainment areas, the control measures applied and the associated costs incurred to reduce emissions from a source will typically reduce the ambient PM2.5 concentrations in several non-attainment areas. The abatement costs are therefore joint costs of achieving all of the reductions in ambient PM2.5 concentrations coincidentally. Any allocation of those costs among the individual non-attainment areas will perforce be arbitrary. Each dollar spent on abatement of emissions from a source achieves some reduction in the ambient PM2.5 concentration in each of the non-attainment areas affected by those emissions. Most importantly, it is not possible to forgo the improvement in air quality in any of the non-attainment areas and thereby save some of the abatement cost. Either all of the improvements or none of the improvements are realized. The costs of abatement can be meaningfully assigned to individual sources and to the origination areas where the sources are located, but they cannot be meaningfully apportioned among the non-attainment areas where the abatement produces coincidental reductions in ambient PM2.5 concentrations. Summary of Control Measures Applied to Sources in Specific Source Categories, Industries, and Origination areas The linear programming solution estimates values for the cost-effective degrees to which specific control measures should be applied to specific sources in specific origination areas. Those data can be combined with the corresponding input data that describe the corresponding control measures and specify their estimated control efficiencies and total annualized costs of emission reduction. That consolidated information can be suitably compiled to produce a summary of the control measures that have been determined to be cost-effective, their control efficiencies, and their total annualized costs for specific source categories, industries, and origination areas. Summaries can be prepared for source categories within individual origination areas or among several or all origination areas, for industries within origination areas or among several or all origination areas, and for origination areas individually or collectively. 2.8 Data Flow The flow of data into and out of the components of the linear programming model is depicted graphically in Figure 2.1. The representation of the model in Figure 2.1 contains nine components. Each component is shown as a rectangular cell and corresponds to either a source of data or an analytic procedure. Each cell contains an index number in square brackets. Data flows are represented as lines between pairs of cells. Information is passed downward through the figure, from lower-numbered to higher-numbered cells. The information that is transmitted is specified beside each connecting line. The portions of the data flow that relate to estimation of the effects that controlling emissions from specific sources have on ambient PM.25 concentrations in specific non-attainment areas appear on the left-hand side of the figure. The portions that relate to estimation of the effects that controlling those emissions have on costs incurred in specific origination areas appear on the right-hand side. Both portions of the data flow originate in cell [1] with data describing individual emission sources that are compiled in the EPA's final 2002 National Emission Inventory (NEI). Both portions use data indicating the type of source, the pollutant emitted, and the annual emission rate for the source. In addition, the portion on the left uses data identifying the location of the source. Proceeding down the left-hand side of the figure to cell [2], for each emission source the transfer coefficients for all non-attainment areas affected by its emissions are determined on the basis of the type of source, its location, and the pollutant emitted. Multiplying those transfer coefficients by the annual emission rate of that pollutant from the source yields, in cell [4], the portions of the initial annual average ambient PM2.5 concentrations in those non-attainment areas that are attributable to those emissions (i.e., the incremental ambient concentrations attributable to those emissions). In cell [3], the control measures that are suitable for the source are specified on the basis of the type of source and the pollutant emitted. The control efficiency for each measure (i.e., the proportional reduction in the annual emission rate of the pollutant that will be achieved if the measure is applied to the source) is then transmitted to cell [5]. Those control efficiencies are multiplied by the corresponding incremental ambient concentrations in specific non-attainment areas that have been computed in cell [4]. Those calculations produce the estimated reductions in annual average ambient PM2.5 concentrations in those non-attainment areas that will occur if the control measure is applied to the source. Shifting to the right-hand side of Figure 2.1, average annualized costs per ton of emissions are also specified for each pertinent control measure in cell [3]. In cell [6], the annual emission rate of the pollutant from the source, from cell [1], is multiplied by the average annualized cost per ton of emissions for a particular control measure to compute the estimated total annualized cost of applying that control measure to that source. For the various emission sources and associated pollutants, these estimated total annualized abatement costs and the corresponding estimated reductions of ambient concentrations in specific non-attainment areas that have been computed in cell [5] are the main inputs into the linear programming model. If, however, the values estimated for the total annualized abatement cost and the reductions in ambient concentrations in specific non-attainment areas for any emission source are a uniform proportion of the corresponding values estimated for any other sources, the arrays of inputs for those sources in the linear programming problem are not mathematically unique. Rather, there are exact linear relationships among their arrays of inputs. Under those circumstances, the linear programming algorithm is not able to compute separate values for the decision variables for those sources, and hence is unable to compute a solution for the linear programming problem. In such situations, in general, the group of mathematically equivalent emission sources has been represented as a single aggregate source in the linear programming problem by summing, in cell [7], the values of the estimated total annualized abatement costs and the estimated reductions of ambient concentrations in specific non-attainment areas for all of the sources in the group. Those total values are then used as the array of inputs for the aggregate source, and the value estimated for the decision variable for the aggregate source is applied to each of the individual sources in the group when computing the outputs from the linear programming model. When, however, a group of mathematically equivalent emission sources includes some sources for which two or more emission control measures are suitable, it is necessary to create a separate subgroup for each set of sources for which a different set of control measures is suitable. This has been accomplished by adding a different, unique, small amount to the average annualized cost per ton of emissions for the common control in all but one of the subgroups. These adjustments commonly introduce some minor inaccuracy into the calculation of the aggregate annualized costs for the individual subgroups of sources and control measures, and for aggregations that include any of those subgroups. The adjustments are useful because they enable determination of the cost-effective control measures for each of the subgroups. In cell [8], the arrays of inputs that have been developed in cells [5], [6], and [7] for the aggregate sources and the remaining individual sources are entered into the linear programming algorithm, and the solution to the linear programming model is computed. The solution indicates the cost-effective allocation of control responsibility among the various individual and aggregate sources. More specifically, it estimates the degree to which a pertinent control measure for a pollutant emitted from a source or group of sources should be applied to each of those sources in order to achieve the objective annual ambient PM2.5 concentration in each non-attainment area at minimum total annualized abatement cost for all sources in the aggregate. Finally, in cell [9], the degrees of control responsibility estimated in cell [8] are multiplied by the arrays of inputs for the corresponding emission sources from cells [5] and [6]. The products from those multiplications are then appropriately summed to produce the desired outputs from the linear programming model. For individual origination areas or for all origination areas in total, those outputs can include estimates of the aggregate total annualized abatement costs and reductions in emissions of individual pollutants that will occur for specific source categories, specific industries, or specific control measures. Analogously, for individual non-attainment areas, the outputs can include estimates of the reductions in annual average ambient PM2.5 concentrations that will be realized from cost-effective reductions in emissions from specific origination areas, specific source categories, specific industries, or specific control measures. Also, for specific source categories or specific industries, the control measures that will be cost-effective to apply and the corresponding aggregate total annualized abatement costs and emission reductions for specific pollutants can be tabulated. 3.0 Data Sources and Data Used in Applying Linear Programming Model Four distinct sets of data have been acquired and used in applying the linear programming model described in Section 2.0 in this economic modeling project. They include data on: PM2.5 non-attainment areas in the eastern United States, Transfer coefficients estimating the increases in annual average ambient PM2.5 concentrations in those non-attainment areas that are caused by emissions of PM2.5 and its precursors from sources located in specific origination areas, Emissions of PM2.5 and its precursors from sources in the cement manufacturing industry and the iron and steel manufacturing industry that are located in origination areas for which transfer coefficients have been estimated, Control efficiencies and average annualized costs per ton of emissions for control measures that are suitable for abating emissions of PM2.5 and its precursors from sources in the cement manufacturing industry or the iron and steel manufacturing industry. The sources from which the four sets of data have been obtained and the data acquired from those sources are described in Section 3.1 through 3.4 below. 3.1 PM2.5 Non-attainment Areas In December 2004, the U.S.EPA designated its final list of non-attainment areas in relation to the NAAQS for PM2.5. All of the areas failed to achieve ambient concentrations of PM2.5 lower than the annual NAAQS of 15 g.m3, based on the three-year averages of their annual arithmetic mean concentrations in 2001, 2002, and 2003. The list of non-attainment areas includes 47 areas that, in total, contain 216 counties in 21 states. Four of the areas containing 14 counties are located in the West. They include three areas in California and one area in Montana. The remaining 43 areas, containing 202 counties in 19 states, are located in the East. The 43 areas in the eastern U.S., their design values for 2001 through 2003, and their initial excess ambient concentrations are presented in Table 3.1. The design value for any non-attainment area is the highest annual average ambient concentration of PM2.5 computed by the U.S.EPA for any monitoring site where complete data have been compiled within the counties that are included in the area. The design values computed for 2001 through 2003 for each county in each designated non-attainment area and their specified completeness have been obtained from U.S.EPA (2005). The excess ambient concentration is the amount by which the design value for the area must be reduced to attain the NAAQS. The 43 areas in Table 3.1 are the PM2.5 non-attainment areas for which transfer coefficients have been estimated in the previous study conducted for DOE by Booz Allen Hamilton, Inc. and CONSAD (2006). The counties in those 43 areas plus all other counties in the eastern U.S. that contain coal-fired power plants comprise the origination areas associated with those transfer coefficients. To enable use of the estimated transfer coefficients in this project, these origination areas are the locations for which data on emissions of PM2.5 and its precursors have been obtained for sources in the cement manufacturing industry and the iron and steel manufacturing industry for use in the linear programming model. 3.2 Transfer Coefficients A transfer coefficient quantifies the relationship between the emissions released from a source and the ambient concentration measured at a monitoring site. It represents a composite of the effects of wind, chemical changes in the atmosphere, deposition, and other factors on the emissions as they are transported through space and time. The transfer coefficients that have been used to compute values for input variables in the linear programming model have been estimated in the previous study conducted for DOE by Booz Allen Hamilton, Inc. and CONSAD (2006). In that study, a pollution transport equation has first been developed that combines background concentrations of PM2.5, NOx, and SO2 with estimated emissions from specific sources, including natural sources, and chemical transformation and deposition actions that occur inside the transport plume as the emissions flow through time and space. Transfer coefficient values have been developed by grouping upwind emission sources with similar locations in relation to downwind monitoring sites. These normalized upwind emissions have been summed and the resulting downwind ambient concentrations have been estimated using the pollution transport equation and the distances between the source groups and the receptor sites. Finally, to produce estimates of the contributions of the normalized upwind emissions to downwind ambient PM2.5 concentrations that, first, are realistically attributable to the measured emissions from controllable sources and, second, are consistent with the design values calculated for the downwind monitoring sites by the EPA, the raw transfer coefficients have also been normalized. The normalization has been based on estimates of the portions of the ambient PM2.5 concentration in a typical eastern urban area that are attributable to sulfate formation, nitrate formation, and emission of primary PM2.5, and the portions of those constituents that have been emitted from controllable sources. The normalized transfer coefficients that have been estimated through this process for origination areas that contain sources in the cement manufacturing industry or the iron and steel manufacturing industry are presented in Table 3.2. Separate transfer coefficients are reported for primary PM2.5, NOx, and SO2 for all origination areas from the previous study that contain one or more emission sources in the cement manufacturing industry or the iron and steel manufacturing industry, and for all non-attainment areas whose ambient PM2.5 concentrations are affected by emissions of PM2.5, NOx, and SO2 from each of those origination areas. 3.3 Source Emissions Estimates of the annual rates of emission of primary PM2.5, NOx, and SO2 from individual sources in the cement manufacturing industry or the iron and steel manufacturing industry that are located in each of the origination areas for which estimated transfer coefficients have been developed as described in Section 3.2 have been acquired from the point source component of the final 2002 NEI. Individual records in the NEI database contain estimates of the annual emissions of individual pollutants from specific release points at specific facilities. Each record also contains data fields that indicate the location of the facility, the industry in which it operates, the type of source emitting the pollutant, the control status for the pollutant, and other factors relating to their release. Pertinent emission sources have been identified in the NEI database on the basis of their source classification code (SCC) and their location in one of the 281 counties in 20 states and the District of Columbia that are included in the origination areas for which transfer coefficients have been estimated. The types of sources for which records have been acquired from the final 2002 NEI database and the corresponding ranges of SCCs include: By-product coke manufacturing (SCCs 30300302 30300399) Ferroalloy manufacturing (SCCs 30300601-30300704) Iron production (SCCs 30300801-30300899) Steel manufacturing (SCCs 30300901-300300999) Iron Foundries (SCCs 30400301-30400399) Steel foundries (SCCs 30400701-30400799) Malleable iron manufacturing (SCCs 30400901-30400999) In-process fuel use - coke related (SCCs 39000701, 39000702, 39000789, 39000899) Cement manufacturing dry process (SCCs 30500606-30500699) Cement manufacturing wet processes (SCCs 30500706-30500799) In-process fuel use cement kiln/dryer (SCC 39000201, 39000402, 39000502, 39000602) As reported in Table 3.3, the records acquired on the basis of these criteria include 47,707 sources of PM2.5 emissions, 25,627 sources of NOx emissions, and 20,537 sources of SO2 emissions. 3.4 Emission Control Measures and Costs The primary database from which estimates of the control efficiencies and average annualized costs per ton of emissions have been obtained for control measures that are suitable for abating emissions of PM2.5, NOx, and SO2 from sources in the cement manufacturing industry or the iron and steel manufacturing industry is AirControlNET, a PC-based database tool that contains information about control measures that are available for reducing emissions of specific pollutants from specific categories of emission sources. AirControlNET contains data for ten pollutants, including: PM2.5, particulate matter with aerodynamic diameters of 10 microns or less (PM10), elemental carbon (EC), organic carbon (OC), NOx, SO2, volatile organic compounds (VOC), carbon monoxide (CO), ammonia (NH3), and mercury (Hg). For each control measure that is available for a specific source category, Appendix A in E.H. Pechan & Associates, Inc. (2006) reports its control efficiency in reducing the emissions (the proportions of the initial emissions that are abated by the measures), pollutants whose emissions are increased by the measure (if any), and the estimated average annualized cost of applying the measure per ton of the primary pollutant that the measure controls. The SCCs included in each source category are specified in Chapter III of E.H. Pechan & Associates, Inc. (2006). In combination, the information in Appendix A and Chapter III provides, for each SCC, a list of the control measures that are suitable for controlling emissions of each pertinent pollutant (PM2.5, NOx, or SO2) that is released from that SCC, and estimates of the control efficiency and the average annualized cost per ton of pollutant for each control measure. The list compiled for each SCC has been edited by deleting any control measure that has both the same or lower control efficiency and the same or higher average annualized cost per ton than another control measure. The former control measure is dominated by the latter measure, and will never be included in a cost-effective solution computed using the linear programming model. In addition, because the AirControlNET database does not include information about some potentially cost-effective control measures (especially, the CemStar process) for emission sources in the cement manufacturing industry, and does not report estimates of average annualized cost per ton for any control measures for SO2 emissions in either the cement manufacturing industry or the iron and steel manufacturing industry, a focused search has been conducted for information about those topics that can be used to develop input data for those control measures. As a result of this search, reports prepared for the Lake Michigan Air Directors Consortium (LADCO) by MACTEC Federal Programs, Inc. (2005a, 2005b, 2006a, 2006b), for the Texas Commission on Environmental Quality by ERG, Inc. (2006), and by State and Territorial Air Pollution Program Administrators / Association of Local Air Pollution Control Officials (STAPPA/ALAPCO, 2006) have been identified and reviewed. Data reported in STAPPA/ALAPCO (2006) have been used to produce estimates of control efficiency and average annualized cost per ton for the CemStar process, and for wet, dry, and advanced flue gas desulfurization (FGD) control measures for the cement manufacturing industry. The data indicate that, on average, advanced FGD controls dominate both wet FGD controls and dry FGD controls in that industry. Accordingly, only advanced FGD has been included as a potentially cost-effective control measure in this project. No estimates of average annualized cost per ton have been found for SO2 control measures for emission sources in the iron and steel manufacturing industry. Finally, some alternative estimates of control efficiency and average annualized cost per ton have been found in MACTEC Federal Programs, Inc. (2006b) and in STAPPA/ALAPCO (2006) for NOx control measures in the cement manufacturing industry for which estimates are contained in the AirControlNET database. Because the sources of information cited for the alternative estimates are largely the same sources cited in E.H. Pechan & Associates, Inc. (2006) for the estimates in the AirControlNET database, and because the specific SCCs to which the alternative estimates pertain are not indicated in the other two reports, the AirControlNET data have been used to describe the control efficiency and average annualized cost of those control measures in this project. A summary of the control measures that have been included as options for abating emissions of PM2.5, NOx , and SO2 from sources in the cement manufacturing industry or the iron and steel manufacturing industry is presented in Table 3.4. The table contains estimates of control efficiencies and average annualized costs per ton of pollutant emitted for 17 control groups, where each control group is suitable for application to specific SCCs in either or both of the industries. Ten of the control groups contain only one control measure; four groups contain two control measures; one group contains three control measures; and two groups contain four control measures. For each control measure, the table indicates the pollutant or pollutants that the measure controls, and estimates of its control efficiency, its typical average annualized cost per ton and, where available, low and high estimates of its cost per ton. In addition, for control measures in groups that contain more than one measure, the table reports the incremental control efficiency and the incremental cost per ton for each measure in relation to the next most efficient control measure in the group. 4.0 Structure of the Analysis Because the screening analysis performed in this project considered emissions of PM2.5, NOx, and SO2 and their effects on ambient PM2.5 concentrations in non-attainment areas for only two industries, it has not been realistic to establish improvement of those concentrations sufficiently to achieve the annual average NAAQS for PM2.5 in all of those areas as constraints when applying the linear programming model to determine the allocation of control responsibility among sources in those industries that would minimize the total annualized cost of reducing emissions from those sources in compliance with those constraints. An attempt to fulfill those constraints would likely require application of the most stringent available control measures to all, or almost all, emission sources in both industries. Such a result would not provide constructive evidence about the usefulness of the linear programming model as a tool for screening analysis. To avert this likely outcome, arrays of less ambitious, achievable improvements in ambient PM2.5 concentrations in the individual non-attainment areas have been developed as bounds for those constraints. The first step in developing those arrays has involved computing the total reduction in ambient PM2.5 concentration that would be achieved in each non-attainment area if the most stringent control measure specified in the AirControlNET database for the sources SCC were applied to each emission source in each industry. Those values represent the maximum reductions that are achievable by applying any of the suitable control measures specified in the AirControlNET database to sources in those industries. Each of those maximum achievable values has then been multiplied by a common proportion between zero and one to produce a uniform, feasible goal for improvement of the ambient PM2.5 concentration in all of the non-attainment areas concurrently. In the screening analysis, separate arrays of reductions in ambient PM2.5 concentrations have been computed for the proportions 0.75, 0.50, and 0.25 of the maximum achievable reductions for both industries jointly, and for the proportion 0.75 of the maximum achievable reductions for each industry separately. In addition, for both industries jointly, a separate arrays has been computed for the proportion 0.75 of the maximum achievable reductions for both industries jointly when the control measures in the AirControl NET database are augmented in the cement manufacturing industry with both the CemStar process for NOx emissions and advanced flue gas desulfurization (AFGD) for SO2 emissions. Thus, in total, six separate arrays of feasible improvements in ambient PM2.5 concentrations have been developed and used as constraining values in the screening analysis conducted in this project. 5.0 Results from the Analysis As explained in Section 2.7, a large variety of output measures can be computed to describe the detailed consequences of the cost-effective allocation of control responsibility among sources of emissions of specific pollutants that is determined using the linear programming model. In this section, a selection of possible output measures have been used to provide comparative displays of the results obtained when the linear programming model has been applied to the six arrays of feasible improvements in ambient PM2.5 concentrations in the individual non-attainment areas that have been described in Section 4.0. In the majority of the results presented in the remainder of this section, two sets of comparisons are displayed. First, in Tables 5.1, 5.2, 5.3, and 5.4, comparisons are provided between the results derived for cost-effective control of emissions from the cement manufacturing industry and the iron and steel manufacturing industry separately, and for cost-effective control of emissions from both industries jointly. Second, in Tables 5.5, 5.6, 5.7, and 5.8, comparisons are provided among the results derived for cost-effective control of emissions from both industries jointly when seeking improvements in ambient PM2.5 concentrations that are progressively smaller proportions (0.75. 0.50, and 0.25) of the maximum feasible improvements. In Tables 5.1 and 5.5, comparisons are presented at the level of the two industries, individually and in total. In Tables 5.2 and 5.6, comparisons are presented at the level of ten individual source groups, which correspond to distinct segments of the iron and steel manufacturing industry and the cement manufacturing industry. In Tables 5.3 and 5.7, comparisons are presented at the level of eleven types of control measures. In those tables, the output measures displayed are: the total annualized costs of controlling emissions, the number of control options that are available, the number of control options that have been exercised in the solution, and the percentage of the available control options that have been exercised. A control option is a combination of an emission source and a suitable control measure. A source for which four control measures are suitable represents four control options. In Tables 5.3 and 5.7, the baseline emissions from the sources to which each type of control measure has been applied is also presented. Note that these tables do not report any use of the CemStar process or AFGD to control emissions of NOx or SO2 in the cement manufacturing industry. When these control measures have been included as additional options in the linear programming model, the solution determined by the model has not changed. The control measures adopted to control NOx emissions in the cement manufacturing industry are mainly selective catalytic reduction (SCR) and selective non-catalytic reduction (SNCR), both of which have higher control efficiency and higher average annualized cost per ton than the CemStar process, and apparently have supplanted the CemStar process in the cost-effective set of control measures. AFGD is a more expensive option for improving ambient PM2.5 concentrations than the control measures that are adopted, and hence is not included among the cost-effective measures. Collectively, the results in Tables 5.1, 5.2, and 5.3 indicate that the specified proportional improvements in ambient PM2.5 concentrations can be obtained at much lower total cost (a decrease of $3.07 billion) if control responsibility is determined for both industries jointly than for each industry separately. Further, the total cost increase consists of a small decrease in cost (-$19 million) for the iron and steel manufacturing industry and a large increase in cost (+$3.09 billion) for the cement manufacturing industry. The bulk of the cost increase is attributable to the use of selective non-catalytic reduction (SNCR) to control NOx emissions in dry process cement manufacturing. The results in Tables 5.5, 5.6, and 5.7 indicate that the incremental cost of achieving successive 0.25 increases in the proportion of the maximum feasible improvement in ambient PM2.5 concentrations that is concurrently realized in different non-attainment areas increases at an increasing rate (from $525 million to $1.90 billion to $4.23 billion). Those incremental costs are incurred primarily to control NOx emissions in dry process cement manufacturing, first to apply SNCR and then to apply selective catalytic reduction (SCR). In Tables 5.4 and 5.8, comparisons are presented among the individual PM2.5 non-attainment areas. The output measures displayed in the tables are the reductions in PM2.5 concentrations that are attained in each non-attainment area, and the percentages of the corresponding maximum achievable reductions that those cost-effective reductions represent. The results indicate that attaining a specified minimum proportion of the maximum feasible improvement in ambient PM2.5 concentrations in different non-attainment areas involves achieving considerably larger proportional reductions in the ambient PM2.5 concentrations of a substantial number of the areas. It should also be noted that, in a small number of non-attainment areas, the reported proportional improvement is smaller than the minimum required improvement. This result is a computational artifact of the independent truncation of control cost data that is caused by the restrictive input format that is required by the linear programming solver. An approach for mitigating or avoiding this computational error is discussed in Section 8.0. Finally, Table 5.9 presents estimates of the reductions in emissions of a co-pollutant, PM10, that are obtained as a by-product of intentionally controlling emissions of PM2.5 as part of a cost-effective allocation of control responsibility among sources designed for simultaneously reducing ambient PM2.5 concentrations in multiple non-attainment areas by specified amounts. 6.0 Conclusions The screening analysis conducted in this project has clearly demonstrated that the linear programming model developed by CONSAD provides credible, economically coherent estimates of the cost-effective allocation of emission control responsibility among sources for the purpose of simultaneously achieving specified improvements in ambient PM2.5 concentrations in multiple non-attainment areas. Further, the analysis demonstrates that useful estimates are obtained when the analysis is restricted to a small number of industries, and when multiple control options are considered for use on individual emission sources. Previous applications of the model did not examine its capabilities under such conditions. Doubtless, the model is capable of providing similarly useful and informative estimates under other circumstances that have not yet been attempted. 7.0 References Booz Allen Hamilton, Inc. and CONSAD Research Corporation (2006), Economic Impact of PM2.5 Transport Between Nonattainment Areas, report prepared for Energy Information Administration, U.S. Department of Energy, Washington, DC. February 15. Chu, Shao-Hang (undated), Attachment A: Critical Design Value Estimation and Its Application, U.S. Environmental Protection Agency, Office of Air Quality Planning and Standards, Air Quality Strategies and Standards Division, Research Triangle Park, NC. http://www.epa.gov/ttn/oarpg/t1/memoranda/cdv.pdf CONSAD Research Corporation (2004). Economic Impacts Of Alternative Regulatory Proposals. report prepared for the National Energy Technology Laboratory, U.S. Department of Energy. Pittsburgh, PA, February. Czyzyk, Joseph, Sanjay Mehrotra, Michael Wagner, and Stephen J. Wright (1997), PCx User Guide, Technical Report OTC 96/01, Optimization Technology Center, Argonne National Laboratory, Argonne, Illinois, and Northwestern University, Chicago, Illinois, November 3. E.H. Pechan & Associates, Inc.(2006), AirControlNET 4.1 Documentation Report, May. E.H. Pechan & Associates, Inc. (2005) AirControlNET Version 4.1 User Guide, September. ERG, Inc. (2006), Assessment of NOx Emissions Reduction Strategies for Cement Kilns Ellis County, Final Report prepared for Texas Commission on Environmental Quality. Cincinnati, OH, July 14. MACTEC Federal Programs, Inc. (2006a), Identification and Evaluation of Candidate Control Measures Phase II Final Report, prepared for Lake Michigan Air Directors Consortium (LADCO). Herndon, VA, June. MACTEC Federal Programs, Inc. (2006b), Interim White Paper Midwest RPO Candidate Control Measures, Source Category: Cement Kilns, prepared for Lake Michigan Air Directors Consortium (LADCO). Herndon, VA, March 6. MACTEC Federal Programs, Inc. (2005a), Midwest Regional Planning Organization (RPO) Cement Best Available Retrofit Technology (BART) Engineering Analysis, prepared for Lake Michigan Air Directors Consortium (LADCO), March 30. MACTEC Federal Programs, Inc. (2005b), Midwest Regional Planning Organization (RPO) Iron and Steel Mills Best Available Retrofit Technology (BART) Engineering Analysis, prepared for Lake Michigan Air Directors Consortium (LADCO), March 30. State and Territorial Air Pollution Program Administrators / Association of Local Air Pollution Control Officials (STAPPA/ALAPCO) (2006), Controlling Fine Particulate Matter Under the Clean Air Act: A Menu of Options, March. U.S. Environmental Protection Agency, Air Quality and Analysis Division, Emission Inventory and Analysis Group (2006), Documentation for the Final 2002 Point Source National Emissions Inventory, February 10. U.S. Environmental Protection Agency (2005), Technical Support Document for PM2.5 Designations-Supplemental Notice, Office of Air Quality Planning and Standards, April 5.  HYPERLINK "http://www.epa.gov/pmdesignations/documents/Apr05/tsd/tsd.pdf" http://www.epa.gov/pmdesignations/documents/Apr05/tsd/tsd.pdf 8.0 Quality Assurance Report The procedures that CONSAD has applied to ensure adequate quality control in data acquisition and data management are summarized in the Sections 8.1 and 8.2 below. 8.1 Data Acquisition CONSAD has acquired three sets of data for this analysis. The data describe: emissions from specific sources, the cost and efficiency of specific emission control measures, and transfer coefficients estimating the impacts of source emissions on ambient PM2.5 concentrations. Data have been obtained for both the cement manufacturing industry and the iron and steel manufacturing industry. Both industries have been included in the screening analysis. The quality assurance activities that have been conducted by CONSAD to ensure that the data was not compromised are discussed in the following paragraphs. Emissions Data. CONSAD has used the point source component of the final 2002 NEI as the source for data on the emissions of PM2.5, NOx, and SO2 from individual sources in the cement manufacturing industry and the iron and steel manufacturing industry. The data have been stored and manipulated in an Access database. Separate tables have been developed containing the emissions data and the relationships between the emissions data and the locations of their sources, the industries operating the sources, and control measures that would be effective in abating the emissions. These relationship tables have been used to select the appropriate entries from the source data set to be included in the final data set that has been analyzed. Emissions Control Data. The principal source of data on the emissions control measures that are available for reducing emissions of PM2.5, NOx, or SO2 from individual sources in the cement manufacturing industry and the iron and steel manufacturing industry is E.H. Pechan & Associates, Inc. (2006). Data from the Source at a Glance sections in Chapter III that report have been converted into a table containing the SCCs for the emission sources that are controllable by each pertinent measure, the name of the control measure, its average annualized cost per ton, and its control efficiency. A table of pertinent SCCs has been compiled manually and used to select applicable controls from the complete control table. This table has been reviewed and edited to remove redundant listings (entries with the same control name, cost, and efficiency). The condensed table has then been stored in the Access database and linked to the emissions data by the relationship table that links SCCs to control measures. Data on Impacts of Emissions on PM2.5 Concentrations. Transfer coefficients that estimate the impacts that an increase or decrease in emissions of specific pollutants from specific sources or groups of sources in specific locations or geographic areas has on ambient concentrations of those pollutants or successor pollutants in specific PM2.5 non-attainment areas have been obtained from Booz Allen Hamilton, Inc. and CONSAD (2006). These transfer coefficients are specific to an origination area, affected non-attainment area, and pollutant type. After these coefficients have been acquired, they have been recorded in a table in the database and linked to the source data by a table that relates FIPS codes to origination areas. 8.2 Data Management After the data on pollutant emissions from specific sources, on emission control measures available for specific categories of sources, and on transfer coefficients for sources in specific geographic areas have been acquired, compiled, and linked together in the Access database, as described above, as separate entity tables with corresponding relationship tables, specific data elements from the tables have been combined to produce estimated values of data items required as inputs to the linear programming model for each individual source. Those data items include: the total annualized cost of applying each pertinent control measure to each source of emissions, the estimated reduction in emissions of each pertinent pollutant that is achieved when each pertinent control measure is applied to each source, and the estimated reductions in the ambient concentration of PM2.5 that is attained in specific PM2.5 non-attainment areas as a result of the reduction in emissions of PMa g n t      ʼʼrarVrʌʼh{nmHnHu j}he0UmHnHujhe0UmHnHuhe0mHnHuhe0PJmHnHtHu2jh|he0>*B*UmHnHphuhe0mHnHuh|he00JmHnHu$jh|he00JUmHnHuhjm#jhjm#U hRhRhe0h& hhR! =       + , A g $a$gdR % : ; ^ _ ` a n o p q r s t  3 b  ! gdRgdR$a$gdR ! " # $ ' ( b c d ~  詚~d詚S~ jqhe0UmHnHu2jh|he0>*B*UmHnHphuh{nmHnHu jwhe0UmHnHujhe0UmHnHuhe0mHnHuhe0PJmHnHtHu$jh|he00JUmHnHu&jh|he0UmHnHuh|he00JmHnHuhe0mHnHu    - . / 0 1 2 3 4 5 Q R S T W X j k l үҤy_Ҥ2jh|he0>*B*UmHnHphuh{nmHnHu jkhe0UmHnHujhe0UmHnHuhe0mHnHu2jh|he0>*B*UmHnHphuhe0mHnHuh|he00JmHnHuhe0PJmHnHtHu$jh|he00JUmHnHu     µ§§yyhµ§N2jh|he0>*B*UmHnHphu j_he0UmHnHuhe0mHnHu2jh|he0>*B*UmHnHphuhe0mHnHuh|he00JmHnHuhe0PJmHnHtHu$jh|he00JUmHnHuh{nmHnHujhe0UmHnHu jehe0UmHnHu   @ A B \ ] ^ _ ` a b c d ٹʮʛxgʮʛ jShe0UmHnHu2jh|he0>*B*UmHnHphuhe0mHnHu$jh|he00JUmHnHuh{nmHnHu jYhe0UmHnHujhe0UmHnHuhe0mHnHuhe0PJmHnHtHuh|he00JmHnHu"b 5 o$QV()$a$gdM4$a$gdM4gdR !  /01234567STUVYZcdeŸŭӸy_ŸŭN jG he0UmHnHu2jh|he0>*B*UmHnHphuhe0mHnHuh{nmHnHu jMhe0UmHnHujhe0UmHnHuhe0mHnHuhe0PJmHnHtHuh|he00JmHnHu$jh|he00JUmHnHu2jh|he0>*B*UmHnHphu    +,-.1247LMNҵҪeUҪh|he00JH*mHnHu2j h|he0>*B*UmHnHphuh{nmHnHu jA he0UmHnHujhe0UmHnHuhe0mHnHu&j h|he0UmHnHuhe0mHnHuh|he00JmHnHuhe0PJmHnHtHu$jh|he00JUmHnHu!NhijlmnopqշyhշN2j h|he0>*B*UmHnHphu j5 he0UmHnHu2j h|he0>*B*UmHnHphuhe0mHnHuh|he00JmHnHuhe0PJmHnHtHu$jh|he00JUmHnHuh{nmHnHujhe0UmHnHu j; he0UmHnHuhe0mHnHu!"#$%&BCDEHIlmnǸǧߓyǸhߓ j)he0UmHnHu2j h|he0>*B*UmHnHphuhe0mHnHuh{nmHnHu j/ he0UmHnHujhe0UmHnHuhe0mHnHuhe0PJmHnHtHuh|he00JmHnHu$jh|he00JUmHnHu#./0JKLNOŸŭӸy_ŸŭN jhe0UmHnHu2jh|he0>*B*UmHnHphuhe0mHnHuh{nmHnHu j#he0UmHnHujhe0UmHnHuhe0mHnHuhe0PJmHnHtHuh|he00JmHnHu$jh|he00JUmHnHu2jh|he0>*B*UmHnHphuOPQRSopqruvүҤy_Ҥ2jh|he0>*B*UmHnHphuh{nmHnHu jhe0UmHnHujhe0UmHnHuhe0mHnHu2jh|he0>*B*UmHnHphuhe0mHnHuh|he00JmHnHuhe0PJmHnHtHu$jh|he00JUmHnHu 345OPQSTUVWXtuvwµ§§yyhµ§N2jh|he0>*B*UmHnHphu j he0UmHnHuhe0mHnHu2jh|he0>*B*UmHnHphuhe0mHnHuh|he00JmHnHuhe0PJmHnHtHu$jh|he00JUmHnHuh{nmHnHujhe0UmHnHu jhe0UmHnHuwz{ ٹʮʛxgʮʛ_[W[hV>[hjm#jhjm#U jhe0UmHnHu2jh|he0>*B*UmHnHphuhe0mHnHu$jh|he00JUmHnHuh{nmHnHu jhe0UmHnHujhe0UmHnHuhe0mHnHuhe0PJmHnHtHuh|he00JmHnHu" 'MORbe !"JMQRTX]^ag8;^@K`!IRe . G hh$h8h2 hr>H*hr>h< hH*hr h%H*hh"n hlH*h%hr>hlGZ[]L$ & FEƀ3Ʀa$gdM4$a$gdM4L$ & FEƀ3Ʀa$gdM4 ""$$$$$&&'$a$gd($a$gdM4L$ & FEƀ3Ʀa$gdM4G H h o y C"J""h#####$$>$?$$$$$$%%&&&&&&&&'P'p'x''''((()) ****%*-****+++,,--$-,-/-ÿÿÿÿÿhf1 h?Ih?Ih?IhX hXH* hH*hhjm#hYmh"nhdh/4hfh[chh$ h$h$E''(())]L$ & FEƀ3Ʀa$gdM4L$ & FEƀ3Ʀa$gdM4$a$gdM4)****]L$ & FEƀ3Ʀa$gdM4$a$gdM4L$ & FEƀ3Ʀa$gdM4*++--]L$ & FEƀ3Ʀa$gdM4$a$gdM4L$ & FEƀ3Ʀa$gdM4/-N-W-g---N.U.Y.].../ /j/m/q/r///!0"0#0%0(0,0y00000011112222)30333=4^44444555556666<678888888999 9#9$9,9x9::::::L:hh9 hHKH*h Jhh hHKH* hHKh?Ihf1h?IhHKQ---\.].q.r...]UI $^a$gdM4$a$gdM4L$ & FEƀ3Ʀa$gdM4$a$gdM4L$ & FEƀ3Ʀa$gdM4.$0%000j1k1559 9L:N:%=&=N>P>AABAAAAA:C>>>:>>??$?S@f@n@o@@@@@AABAEAFAAAAAAAAABBB&CBCHCTC\CCCD DDD4D EEEEEEEEFF6FDFYF\F`F뼸 hHKCJhXhYmhV>[ h9hV>[ h9H* hH*hhp6, hHKH*h9 hHKH*hHKhf1 h;H*hh;F*hh{8eh; hHKCJhYm hHKH*hHKhhhXMJRKSKNNNNZO[OQQSSTTTTVVYY$a$gdM4 $^a$gdM4 $^a$gdM4$ ^` a$gdM4$a$gdM4$a$gdM4$`^``a$gdM4QQQQQQQQRRXR`R|RSSSSSSSTWT^TpTtTTTUUMUTUUUU4V6V9VAVOVxVVVVVVVWWWvX~XXXXX#Y5YAYuYYYYZZZ$Z,Z0Z2Z8Z:Z@ZBZмh"nhHKmHsHhp6,mHsHhehHKH*mHsHhehHKmHsHh-hp6,hrh;hhh hHKH*hHKFYZZ.ZnZZZv\w\^^uavahdiddL$ & FEƀ3Ʀa$gdM4$0^`0a$gdM4 $^a$gdM4$a$gdM4$a$gdM4BZFZHZJZNZRZZZ`ZbZjZlZnZpZZZZ[&[r[z[[[E\F\w\\\\\\\\\]]/]P]n]r] ^?^^^^^^^^_$_+_____t`ö晦晦hyh5 hKhh3 hHKH*hhHK hp6,CJ hHKCJh"nhHKCJmHsHh"nhHKH*mHsHhehHKH*mHsHhehHKmHsHhp6,h"nhp6,mHsHh"nhHKmHsH7t`{``````````aaa8a?aaabbb!b&b(b,bGbRbjbub1c4c8c=c?cCcccccccZdidldpddddddddddddcf~fffffff:g=gBgCgDgGgKghhhȼȸhjl h}8H* h"nh}8h}8h5 mHsHh"nh5 6h"nh5 H* h"nh5 h5 H*hh5 6h5 hHdddd~ffCgDgQL$ & FEƀ3Ʀa$gdM4 $^a$gdM4L$ & FEƀ3Ʀa$gdM4$a$gdM4Dghh i ijjjjj4k\k^knn~~$@ ^@ `a$gdM4$a$gdM4 $^a$gdM4$a$gdM4$a$gdM4L$ & FEƀ3Ʀa$gdM4hhhhfiiijjjj4jjjjjjjjjjjjjjjjjk kkk"k$k(k0k2k4kPkXk/l2l6lDlllllmmdmgmkmymnndnknlnnnnnnIo֨hBhp6,h h9CJ hHKCJ h9hHK h9H* hf1>* hHKH*h9h hHKaJ hf1H*hf1hHKh; h"nh5 h}8@IoLooooo.p4p*h hHKH*hHK hBH*hBh9 hyH*Cnqqqqqqrrssssstttuuuuvv$`^``a$gdM4 $`a$gdM4$^`a$gdM4$@ ^@ `a$gdM4$a$gdM4uuuuuuuuuuuuuuuuuvvv&vvvwwwwxx&y-yd{e{i{m{{{6|9|W|\|||}}=}\}a}t}}}}U~~~~~~~6ʹʴ h>* hHK>*h-hp6,h hHKH*hHKhh; hHK5\hJn#hHnh5 >* h5 H*h5 h"nh5 hJn#h5 @v-v.vSyTy{{e{f{T|U|/}0}24\܀$^`a$gdM4$@ `^@ ``a$gdM4$a$gdM4$`^``a$gdM4$a$gdM4$a$gdM468>@BLNXZ\bdxz~Ҁڀ܀^*12W[涪ztzplhc hHKH*hp6,hh- hf}CJ hHKCJh"nhHKH*mHsHhf}h"nhf}mHsHhHKh"nhHKmHsHhehHKH*mHsHhehHKmHsHhf}CJmHsHh"nhf}CJmHsHhehHKCJH*mHsHh"nhHKCJmHsHhehHKCJmHsH&܀ހЃӃ΅Ѕ҅@BևR$^`a$gdM4$@ ^@ `a$gdM4$a$gdM4$p^p`a$gdM4$a$gdM4$`^``a$gdM4$0^`0a$gdM4[qtwՃlz΅Ѕ҅@BZ\Ƈȇ·ԇ܇އ  (4<DLPRjpx^ew{hp6,h hJn#CJ hHKCJh h>*aJ hHK>*aJ h>*aJ hHKaJhh-hJn# hHKH*hHKHwz+23#ƏǏ078ǑΑϑґӑ (08ŻŲū眫 h"nhf}h"nhHKH* h"nhHKh"nhf}CJh"nhHKCJH*h"nhHKCJ hHKCJ hp6,>* h>* hHK>*hhhf}hp6,hhJn# hHKH*hHK=XYϑ<XZޖ`~$^`a$gdM4$`^``a$gdM4$@ `^@ ``a$gdM4$a$gdM4$a$gdM48:<LVX:Bdg "$*,046>BLT\^`p| )- 8Xþ᫡Б h"nhf}h"nhf}CJh"nhHKCJH*h"nhHKCJ hp6,>* h>* hHK>*hq h- hf}H*hf} hHKH*hp6,hhHK hf}CJ hHKCJh"nhHKH* h"nhHK9dfBDP2HJޞߞȤ $^`a$gdM4$@ `^@ ``a$gdM4$`^``a$gdM4$a$gdM4Xe DLPVXrt~ H,4Cݹ||pkf hf}H* hHKH*hehHKH*mHsHhjm#hHKH*mHsHhjm#hHKmHsHhehHKmHsHhjm#hHKCJH*mHsHhehHKCJmHsHhjm#hHKCJmHsHhehHKH*mHsHhehHKmHsHhehHKCJmHsH hHKCJhp6,hhHKhf}(Ҟݞbe{|Ɵɟןڟ"#AHI۠ܠ+.NQao8;\]01t{|'./ -57\dftΫ28<¾ͺh\+h#Y h;hHKhh; hp6,>* h>* hHK>* hHK6]hp6,h hHKH*hHKh-hf}I ¨12abz{VW$`^``a$gdM4$a$gdM4$a$gdM4<X^װ?@Get ?Fҳ Yiƴ ۵ζܶʷ-H+=ACEh\( hHKCJ h;hHKhDg3h$h;h#Yh hHKH*hh\+hHKP*8Krxy'9<AF*x2ؾdp#&=K,T 1?RYZUVWX\PQSŽŹh!ehx6Ohjm#hy hHKCJhp6,h\+ hHKH*hYh5h2#Fhh\(h$hhHKhDg3HWXPQtuUL$ & F Eƀ3Ʀa$gdM4L$ & F Eƀ3Ʀa$gdM4$a$gdM4$a$gdM4 SV),rtuVWVW46V\^_}~8CKӾӾӺӾӾӾ׾ӭh8Z<h= hOAH*hFhhhOA hJhJ hJH* hJH*hJh-h h!eH* h-H* hx6OH*hx6Oh!ehy hy H*BuVWWX]U$a$gdM4L$ & F Eƀ3Ʀa$gdM4$a$gdM4L$ & F Eƀ3Ʀa$gdM4 JKefmn*+$a$gdM4$a$gdM4Smu/0IJKde kADBQTXYabCFux|} 綱h!E hUzTH*hUzT hq hq hq hM(H*hM( h9H*h9 h&H*h- h&h&h&hJ h,H*h,h8Z<h=CFH;fo.678 ^dru"#AZ_iĹİİİĩĩĩ h'~sh^ h^H*hO~ h^h^hh^ hDH*h'7h#hx6O hUzThDh/hDh_hUzThGth\+h!Eh6CA eL$ & F Eƀ3Ʀa$gdM4L$ & F Eƀ3Ʀa$gdM4 2`eL$ & F Eƀ3Ʀa$gdM4L$ & F Eƀ3Ʀa$gdM4`eL$ & F Eƀ3Ʀa$gdM4L$ & F Eƀ3Ʀa$gdM48eL$ & F Eƀ3Ʀa$gdM4L$ & F Eƀ3Ʀa$gdM48teL$ & F Eƀ3Ʀ a$gdM4L$ & F Eƀ3Ʀ a$gdM44567$a$gdM4$a$gdM4L$ & F Eƀ3Ʀ a$gdM47:<12:d3T{^!?DQU':=>JhKhh6hKhhKh6hKhhO~ hU)H*hU)hnxh^hoah h'~sh^hh'~shH* h'~shGJYdhpv2345@vw 9IOPa^_:i UZ<+568W$Ekؽعصh6 ht~H*ht~hI?h4evhnh5* hmH*h} h? H* h LtH*h"hmh Lt h+?H*h+h? h+? hKhhKhhKhh>78VWMNOmn    klFGtub$a$gdM4$a$gdM4 lo =@c~ S\}>B.C;<mtz{LMOnvwF i  ˽˽¸°¬°hI hTEyhOh4ev he6H* h+AH*he6 hK!3H*h+A hp+H*hK!3hp+ hH*h hH*hI?hhhh6 h6H*B  s v       G P Z [ _ ` a b d e f g m n o p r  . / e                 C F V _             -9:;<ABCDFÿÿÿh9h&h'7 hwH*h(hg,h O-hwh!lhS\hI? hOhI? hI H*hI hTEyKIOjlKPVWXY^_`ap;<BC01bEFklpqrsyz{|$(2X!ʽʽʽh e h"eH*h"ehs hQ(H*hQ( hqH*hq hnPhH*hnPhhdVh oCh5Sh!lh&hg,h9hS\G!"#$2Eduvy !"'(L]`stnqabrsþúþúǥ hwlH* hbH*hbhs^ hqH*hqh O-h5S h&H*h&hRDhS\hef hdVH* hdVhdVhdV h \H*h \hehs h9h"eh e=bcH#I#J#Y#Z#K$L$|%}%L&M&T'U'''(($ H0^`0a$gdM4$0^`0a$gdM4$a$gdM4$a$gdM4L  !C!F!v!!!!!!!!"$"X"G#H#I#J#Z###f$$I%z%|%}%%%%%%K&L&&&z'{'~'''''(óϝϥԕԎφh'~sht~6 h'~sht~h ht~6hrw-ht~6 hrw-ht~ hrht~ht~>*B* OJQJ^Jph ht~>* ht~6] ht~6ht~h h:H*h:hs^hI?h4ev hbhbhwlhb4(#($(c((<))***f*g**+q++:,;,,-----.`.......$/%/&/(/,/E/a/j//////00 0'0(0?00011/1W1X1˾и˴ﴰ heZH*h: hY;hM4heZ hhM4hhM4hI? ht~0Jj|ht~>*U ht~>*jht~>*U ht~6] h% ht~ hzht~ht~ ht~6H* ht~69((())f*g*I+J+:,;,----&/'/(/E/F///0$a$gdM4$a$gdCM$a$gdM4$a$gdM4$0^`0a$gdM400_2`2B5C55969<<'<(<bc`a $`a$gdM4$a$gdM4$a$gdM4 $^a$gdCM$a$gdCM $$`a$gdM4X11111272;2]2`2o2w2{22222222222E3n3t333333 44!42444X4e4m4r444445053585C5Y5555555'6F6g6z6}666666666 7737L777777778'88888889h:h dhM4>*hH hM4H* hM4>*heZhM4X93969X9[9k99999:5:r::::::::;@;E;K;W;`;;;;;;;<<<=:=>=S=W=s=====>>>>????@cqu:DEUx+1AINah~DU hM4H*h J h:H* hM4>*H* hM4>*hM4h:U2.5, NOx, or SO2 that is achieved when each pertinent control measure is applied to each source. The joining of data in the various tables and the calculation of required values of various inputs to the linear programming model have been accomplished in Access using SQL commands. The formatting and creation of input files for the linear programming solver program has been accomplished using Perl scripts. Then the PCx Linear Programming Solver has been used to produce a solution for each scenario described by an input file. The reformatting of the solvers output file has also been accomplished in Perl. The reformatted output file has then been incorporated into the Access database and the appropriate data have been joined and the required aggregations have been performed using SQL. Unexpected problems that have arisen in this process have included the lack of data for iron foundries, the inability of the linear programming model to be solved for specific input files due to collinear data, errors caused by truncation of input data for the linear programming solver, the lack of cost data for SO2 control measures, and the lack of cost and efficiency data for the CemStar process. The lack of data for iron foundries has arisen due to the limitations of the NAICS coding system. This problem has been solved by switching to the SCC coding system as the basis for linking the source data to the data on appropriate control measures. The problem of collinear data has arisen due to the equivalence of data inputs for sources that are located in the same origination area and for which the same control measure with the same average annualized cost per ton and the same efficiency is applicable. This problem of collinearity has been solved by aggregating sources that are located in the same origination area and apply the same pertinent control measure. This aggregation has collapsed the data set down to a set of combined sources that are not collinear. Whenever this complication has been revealed by the linear programming solver, the aggregated sources have been reassembled from the data tables and the solution table. Additional collinearity has been encountered within the set of groups of multiple control options that are applicable to sources in specific SCCs. This collinearity has been broken by adding a different small amount to the average annualized cost per ton of each one of the control measures that has been found to have the same cost and efficiency as a control measure in another group. The small amount that has been added is inconsequential in comparison to the variance of the data that are being used, but is large enough to provide the linear programming solver with a solvable matrix. An error in precision has also been found and tracked back to the need to truncate the input data so that they fit into a eight character field in the PCx input file. This problem can be dealt with in at least two different ways. First, all input data that relate to a problematic non-attainment area can be multiplied by an amount that increases the number of significant digits in the data for that area sufficiently that truncation does not set any values equal to zero. After the solution based on the modified data is acquired, the appropriate division can be performed to yield the correct answer. Secondly, the PCx program may be used as a procedure library, and the need for truncation of its input file can be avoided by passing to PCx a matrix taken directly from the Perl script. The AirControlNet database has been found to lack any data on the cost of control measures for abating SO2 emissions. Due to these missing data, no sources of SO2 emissions have been included in the data set initially sent to the linear programming solver. To resolve this problem, cost data from STAPPA/ALAPCO (2006) have been used to estimate the average annualized cost per ton for flue gas desulfurization control measures applied in the cement manufacturing industry. Similarly, to include the CemStar process in the available array of NOx control measures, cost and efficiency data had to be estimated because the effects of the CemStar process are notably different from those of other control measures. The CemStar process not only reduces NOx emissions by 30 percent, but also increases production efficiency by 15 percent. Therefore, it has been necessary to adjust the control efficiency reported for the CemStar process to take into account both its reduction of NOx emissions and its increase of production efficiency. The unadjusted data on the CemStar process has been obtained from STAPPA/ALAPCO (2006). Finally, redundancies in the set of input data entered into the linear programming solver confounded solution of the linear programming model. These redundancies have all been traced to either errors in the code used to generate the set of input data or to redundancies in the final 2002 NEI database. All errors that have been discovered have been repaired. The redundancies in the NEI data have been investigated and have been found to result from use of a model of a representative plant to estimate emissions from plants that do not have actual data recorded for them. This redundancy has been retained, and has been resolved by suitable aggregating the data for different sources as described above.     PAGE  PAGE i 5=?^:<MXYlt| "'8e!BC',OV&Jkn h>PhCMhEh~DhM4H*h/h~DhM4X "2<cz   )4Ng0d  &+-GTYIJgy #(023>KXjkuvb h'EH*h'EhhM4H*hh hM4h>PX*+ &`#$gdR$a$gdCMbj$2EM^rtvy  UY]fu")+4Habdr~%@\a1xy|hO5h/hf2M h'EH*h'EhM4[    h{n0JmHnHuh{n h{n0Jjh{n0JUhyjhyU hM4hM40001hP:pe0/ =!"#$%0001hP:pe0/ =!"#$%-01hP:pV>[/ =!"#$%}DyK _Toc198636115}DyK _Toc198636115}DyK _Toc198636116}DyK _Toc198636116}DyK _Toc198636117}DyK _Toc198636117}DyK _Toc198636118}DyK _Toc198636118}DyK _Toc198636119}DyK _Toc198636119}DyK _Toc198636120}DyK _Toc198636120}DyK _Toc198636121}DyK _Toc198636121}DyK _Toc198636122}DyK _Toc198636122}DyK _Toc198636123}DyK _Toc198636123}DyK _Toc198636124}DyK _Toc198636124}DyK _Toc198636125}DyK _Toc198636125}DyK _Toc198636126}DyK _Toc198636126}DyK _Toc198636127}DyK _Toc198636127}DyK _Toc198636128}DyK _Toc198636128}DyK _Toc198636129}DyK _Toc198636129}DyK _Toc198636130}DyK _Toc198636130}DyK _Toc198636131}DyK _Toc198636131}DyK _Toc198636132}DyK _Toc198636132}DyK _Toc198636133}DyK _Toc198636133}DyK _Toc198636134}DyK _Toc198636134}DyK _Toc198636135}DyK _Toc198636135}DyK _Toc198636136}DyK _Toc198636136mDyK >http://www.epa.gov/pmdesignations/documents/Apr05/tsd/tsd.pdfyK |http://www.epa.gov/pmdesignations/documents/Apr05/tsd/tsd.pdf@@@ NormalCJ_HaJmH sH tH Z@Z jm# Heading 1$<@&5CJ KH OJQJ\^JaJ \@\ jm# Heading 2$<@& 56CJOJQJ\]^JaJDA@D Default Paragraph FontRiR  Table Normal4 l4a (k(No List2B@2 HK Body TextaJH@H  Balloon TextCJOJQJ^JaJ4 @4 #Footer  !.)@!. # Page Number6U@16 t~ Hyperlink >*B*ph8@8 RTOC 1$ ! a$.@. jm#TOC 2 ^4@b4 V>[Header  !HC@rH M4Body Text Indenthx^hB'B  eComment ReferenceCJaJ<<  e Comment TextCJaJ@j@  eComment Subject5\t '62d =+,Ag %:;^_`anopqrst3b5 o$Q V    ( ) Z[ !!""""##%%%%\&]&q&r&&&$(%(((j)k)--0 00022 3 3#5$55555666g7h777788899&999999::x;y;>>>>??I@J@xAyADBEBoBpBDD\G]GGGGGGGHHuKvKMMPP8Q9QyQzQSSSSlUmUUUVVVVV!W5W6WZZ \ \\[\k\l\Z][]^^ ^!^^^7`8`O`P`````cceeeeffggjjEjFjjjjjjjckdkBmCmWmZmHnInJnnn3o4oLooooss wwYwZwxx0x]xkxlxyy/z0z{{{{{||}}~~~ KXcd./׆؆HIUVpqϒВ[\DEݣޣۨܨɩʩ012QRܯݯ_`z{Ͷζ?@¿ÿؿٿGuM-.IJKLMkl#$bcd[\)*wx]^_no`aabij      { | ^ _ O P 12;<=Z[tuWXJK'(<=w!x!0$1$%%&&u)v)++..00?3@3666666#6$6(600000000000000000000p00000000000000000000000000p0p00000000000000000000000 000 p0 p0 p 0 p0 p 0 p0  0 p0 0 0 p0 p0 0 p0 0 p0 p0 p0 p0 p0p00p0p0p0 0p0p 0p0p 0p0  0p0p 0p0p 0p0 0p0p 00p0p0]&0]&p0]&p0]&p0]&p0]&p0]&p0]&p0]&0]&p0]&p0]&p0]&p0]&p0]&p0]&p0]&p0]&0]&p0]&p0]&p0]&p0]&00p05p05p05p05p05p05p0505p05p05p05p05p05p05p05p0505p05009p0909p0909p09 09p0p0>p0>p0>p0>p0>p0>p0>0>0>0p0EBp0EBp0EBp0EB0EBp0EBp0EBp0EBp0EBp0EBp0EBp0EB0EBp0EBp0EBp0EBp0EBp0EB0EBp 0EB0EB 0EB0EB0EB 0EBP 0EBp0EBp 0EBp0EBp0p0mUp0mUp0mUp0mU0mUp0mUp0mUp0mUp0mUp0mUp0mUp0mUp0mUp0mUp0mUX0mUp0mUp0mUp0mUp0mUp0mU0mU0mUp0mUp0mU0mUp0mUp0mU0mUp0mUp0mU0p0p0`0`p0`p0`p0`p0``0`p0`p0`p0`p0`p0`p0`0`p0`p0`0`p0`p0`0`0`p0`p0`p0`p0`p0`p0`p0`p0`0` 0`h0`p0`p0`p0`p0`p0`p0`0`p0`0`0`0`0`p0`p0`p0`p0`p0`p0`p0`p0`p0`p0`p0`p0`p0`0`p0`0`p0`p0`0`0`p0`p0`p0`p0`p0`p0`p0`p0`p0`p0`p0`p0`0`p0`p0`p0`p0`p0`p0`p0`0p0p0p00p0p00p0p0p0p0p0p00p0p0p0p0p0p0p0p0p000p00p0p 0p0p 0p0p 0p0p 00p0p00p0p020202p0202p0202p02p02p0p0`p0`p0`0`p0`p0`0`p0`p0`p0`p0`0p0ÿp0ÿp0ÿp0ÿp0ÿp 0ÿ 0ÿp 0ÿp 0ÿ 0ÿp 0ÿp 0ÿp 0ÿp 0ÿp 0 ÿp 0 ÿp0ÿp0ÿp0ÿp0p00p0p0p0p0p0p00p00p0p0p0p0p0Mp0Mp0Mp0Mp0Mp0Mp0Mp0Mp00d0d0dp0d0d0d0dp0d0dp0d0d0d0dp0d0dp0dp0dp0dp0dp0d0p0p00p0p00_p0_p0_p0_p0_p0_p0_p0_p0_0_0_p0_p0_0_p0_p0_0_p0_p0_p0_p0_p0_p0_p0_p0_p0_0_p0_p0_p0_p0p0=p0=p0=p0=0p00p0p0p0p0p0p0p0=p0(p0(p0(p0(0(0(p0(p0(p0(p0(p0(p0(p0(0(0(p0(p0(0(p0(p0Oy00Oy00Oy00P+My00My00@0@0 0 %:;^_`and[\)*6(6M90dM90M90M90M90M90M90O900O900M900M900M900M900O9000O900M900O900M90 M90 L&M90XM90d $$$$$$$$$'  NOw G /-L:`FQBZt`hIou6[8X<SJ !(X19 b b ')*-.EBmU`2`ÿMd_=((6 ' p&59>nBU`ڨPy׿jmY;(66B.l((B.B.dB./(B. *B.4?B.$|B.dlB.tP+B.=B."B.B./(B.D_)B.'B.T(B.B.nB.]B."B.hB.iB."B.D"B."B."B.d,B.,B.4{(B.t{(B.L) B.) B.) B. ) B.L) B.)B.,B.$-B.d-B.-B.$.B.-B.+B.+B.,+B.l+B.+B.+B.dB.B.B.$B.dB.  %%14aDDEEFF]k]kyyaa||aBBZGGS       e e n 22(6      !#"$&%')(*,+-.0/12345$$/399mEEEEFFakakyyǩǩkkdX\\QUU       l p p 66(6    !#"$&%')(*,+-.0/12345>1*urn:schemas-microsoft-com:office:smarttags PostalCode:/*urn:schemas-microsoft-com:office:smarttagsStreet83*urn:schemas-microsoft-com:office:smarttagsCityB5*urn:schemas-microsoft-com:office:smarttagscountry-region;0*urn:schemas-microsoft-com:office:smarttagsaddress92*urn:schemas-microsoft-com:office:smarttagsState96*urn:schemas-microsoft-com:office:smarttagsplace8**urn:schemas-microsoft-com:office:smarttagsdate 20086DayMonthYear6563210/6321*62626565656562626565663263263263266326565GGGGGGGGGGGGGGGGNNNNOOOOFPMPRPXPPP9Q?QdQjQtQwQRSS SSSSSjjjjjjjjjjjjjjjjjjjj x"x0x4xCxDxHxIxLxSxVx[x{{{{{{{{{{{{~~~  /ABIGTAG(+cpqt^d)6jn:ANQr%NQCF4;hofjbhyou6 9 bh04!!""%%''))****],`,0.3.....h0w00000Z1a11111s2z2223 366666666 6 6 6 6%6(6! ( (((..66667777 9%9/98999i;q;GGGGGGPPRSVVVVVV$W(W\\\\^\b\^^9`A`P`R`ffjjjjjjdmlmnnCoHoSoWooo_xcx{{{|~ -JQU_b % G`uXoK>,?,66666666 6 6 6 6%6(6333333333333333333333333333333333333333333333333333agc1kA`3d Mm"m/ O 4 T  666666666 6 6 6 6%6(666666666 6 6 6 6%6(6 Alex Botkinu4X0J(C.N#5(VBN,Cj!sJ!JRv>;AS`'[6etj)x(y\7N8|m* hh^h`OJQJo(hH^`OJQJ^Jo(hHopp^p`OJQJo(hH@ @ ^@ `OJQJo(hH^`OJQJ^Jo(hHo^`OJQJo(hH^`OJQJo(hH^`OJQJ^Jo(hHoPP^P`OJQJo(hH0^`05o(.0T0T^T`05o(.$ 0$ ^$ `05o(.. 0 ^ `05o(... ,,^,`5o( .... ^`5o( ..... 4`4^4``5o( ...... `^``5o(....... <<^<`5o(........^`OJQJo(hH^`OJQJ^Jo(hHo  ^ `OJQJo(hH  ^ `OJQJo(hHxx^x`OJQJ^Jo(hHoHH^H`OJQJo(hH^`OJQJo(hH^`OJQJ^Jo(hHo^`OJQJo(hHhh^h`OJQJo(hH^`OJQJ^Jo(hHopp^p`OJQJo(hH@ @ ^@ `OJQJo(hH^`OJQJ^Jo(hHo^`OJQJo(hH^`OJQJo(hH^`OJQJ^Jo(hHoPP^P`OJQJo(hH^`OJQJo(hH^`OJQJ^Jo(hHo  ^ `OJQJo(hH  ^ `OJQJo(hHxx^x`OJQJ^Jo(hHoHH^H`OJQJo(hH^`OJQJo(hH^`OJQJ^Jo(hHo^`OJQJo(hHhh^h`OJQJo(hH^`OJQJ^Jo(hHopp^p`OJQJo(hH@ @ ^@ `OJQJo(hH^`OJQJ^Jo(hHo^`OJQJo(hH^`OJQJo(hH^`OJQJ^Jo(hHoPP^P`OJQJo(hHhh^h`OJQJo(hH^`OJQJ^Jo(hHopp^p`OJQJo(hH@ @ ^@ `OJQJo(hH^`OJQJ^Jo(hHo^`OJQJo(hH^`OJQJo(hH^`OJQJ^Jo(hHoPP^P`OJQJo(hHhh^h`OJQJo(hH^`OJQJ^Jo(hHo  ^ `OJQJo(hH  ^ `OJQJo(hHxx^x`OJQJ^Jo(hHoHH^H`OJQJo(hH^`OJQJo(hH^`OJQJ^Jo(hHo^`OJQJo(hHhh^h`OJQJo(hH^`OJQJ^Jo(hHo  ^ `OJQJo(hH  ^ `OJQJo(hHxx^x`OJQJ^Jo(hHoHH^H`OJQJo(hH^`OJQJo(hH^`OJQJ^Jo(hHo^`OJQJo(hHhh^h`OJQJo(hH^`OJQJ^Jo(hHopp^p`OJQJo(hH@ @ ^@ `OJQJo(hH^`OJQJ^Jo(hHo^`OJQJo(hH^`OJQJo(hH^`OJQJ^Jo(hHoPP^P`OJQJo(hH hh^h`OJQJo( ^`OJQJo(o pp^p`OJQJo( @ @ ^@ `OJQJo( ^`OJQJo(o ^`OJQJo( ^`OJQJo( ^`OJQJo(o PP^P`OJQJo(hh^h`OJQJo(hH^`OJQJ^Jo(hHopp^p`OJQJo(hH@ @ ^@ `OJQJo(hH^`OJQJ^Jo(hHo^`OJQJo(hH^`OJQJo(hH^`OJQJ^Jo(hHoPP^P`OJQJo(hH^`OJQJo(hH^`OJQJ^Jo(hHo  ^ `OJQJo(hH  ^ `OJQJo(hHxx^x`OJQJ^Jo(hHoHH^H`OJQJo(hH^`OJQJo(hH^`OJQJ^Jo(hHo^`OJQJo(hHhh^h`OJQJo(hH^`OJQJ^Jo(hHopp^p`OJQJo(hH@ @ ^@ `OJQJo(hH^`OJQJ^Jo(hHo^`OJQJo(hH^`OJQJo(hH^`OJQJ^Jo(hHoPP^P`OJQJo(hHhh^h`OJQJo(hH^`OJQJ^Jo(hHopp^p`OJQJo(hH@ @ ^@ `OJQJo(hH^`OJQJ^Jo(hHo^`OJQJo(hH^`OJQJo(hH^`OJQJ^Jo(hHoPP^P`OJQJo(hH0u'[C.etVByCJ5R0N8|)xsJ0;AS%X        i        j        i        i        46        n 6        DN                j        i        D"i        t~=6__O!E5 & ? q nxJH8- s jm#Jn#&TV&M(Q(\(U)5*\+p+p6, O-/e0f1K!3Dg3/4M4O5X6'7}88Z<+?I?+AB6C oC~DEy#E'E2#F?If2MNx6O>PRUzTFUu!UdVV>[ \S\?]s^oa[c3&d"e{8efKhnPhJVk!ljlwlYmn{noGt Lt4evTEyf}O~@LY9q 5SXOAm JyFeZ%($6bn, e}F{HKe&Xy#Yef9r>hy"n5_:r/d$"lRDTR^CM:wr>#<y 2;h I 9q3Zg,D!e+e6@dޔ-.d0d123567&9&:?@AGHVW[\jkamanaoap?x?y?{?|?~??!'6@@@6@p@@:@x@@>@@@B@D@@@H@@@P@R@@@Z@@@j@@@p@@@@@@@@@@@@(@@@0@@@@<@@@@@@Unknown Alex BotkinGz Times New Roman5Symbol3& z Arial;SimSun[SO5& z!Tahoma?5 z Courier New;Wingdings"1h3Ʀ3Ʀ3Ʀ D.2D.2#4dc5c5 2qHX ?l1Frederick H Rueter Alex BotkinH          Oh+'0   < H T `lt|1ssFrederick H Rueterredred Normal.dot  Alex Botkin3exMicrosoft Word 10.0@vA@x*@@֩D.՜.+,D՜.+,, hp|   92c5A 1 Title 8@ _PID_HLINKSAY>http://www.epa.gov/pmdesignations/documents/Apr05/tsd/tsd.pdf;_Toc198636136;z_Toc198636135;t_Toc198636134;n_Toc198636133;h_Toc198636132;b_Toc198636131;\_Toc198636130;V_Toc198636129;P_Toc198636128;J_Toc198636127;D_Toc198636126;>_Toc198636125;8_Toc198636124;2_Toc198636123;,_Toc198636122;&_Toc198636121; _Toc198636120;_Toc198636119;_Toc198636118;_Toc198636117;_Toc198636116;_Toc198636115  !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz{|}~     !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]_`abcdeghijklmrRoot Entry FtData 1Table WordDocumentSummaryInformation(^DocumentSummaryInformation8fCompObjj  FMicrosoft Word Document MSWordDocWord.Document.89q