Solving the air pollution control puzzle - Environmental Science

Solving the air pollution control puzzle. Ellison S. Burton, Edward H. Pechan, and William. Sanjour. Environ. Sci. Technol. , 1973, 7 (5), pp 412–41...
0 downloads 0 Views 3MB Size
Solving the air pollution control puzzle Ellison S. Burton, Edward H. Pechan, and William Sanjour

U.S.Environmental Protection Agency, Washington, D.C. 20460 Air quality in a given area is based on the emission characteristics and geographic locations of pollution sources, on the meteorological factors affecting pollutant transport, and on emission control strategies imposed by regulatory agencies. In regions containing many sources, the systematic examination of proposed abatement strategies can be aided by computer modeling systems. Such systems require a large data input and contain several separate models. Development of a methodology for analysis of regional air pollution abatement should be guided by several key requirements: The method must apply to a wide range of environments with different pollution sources, fuel use and cost structures, meteorology, and different abatement objectives. The method must be generalized to accommodate abatement analysis of several different pollutants. 0 The method must be useful immediately with present data and become increasingly more powerful as research provides more and better data concerning air pollution. 0 The method must be useful at the federal, state, and local levels for examining the cost-effectiveness relationships implied by a given abatement policy, for judging the efficacy of competing incentives for abatement, and for indicating productive avenues of research in pollution control. Given the large number of individual emission sources in most regions, the scale of any multiple-source abatement analysis is combinatorially so great that computerassisted simulation is the only analytical method available which can practically meet the above requirements.

Framework for air pollution control Here's how modeling fits into an overall framework for analysis of air pollution control policy. Figure 1 shows a macrolevel flow chart of a system to generate and evaluate pollution control strategies within a geographic region. The system utilizes regional and nonregional data and other inputs based on costs, benefits, possible control alternatives, and others to produce an enumeration of all possible pollution control strategies. Each possible strategy (or only those strategies meeting standards or other constraints) is characterized by costs, benefits, and nonquantitative factors. The decision process then consists of selecting the "best" or most appropriate strategy. In blocks 1, 2, 6, and 7, data on the current emission inventory and regional growth trends are used to project a future uncontrolled inventory. The addition of costs and control alternatives in blocks 3 and 8 produces a data file containing all feasible control alternatives, effects, and costs, for all sources. Then, each strategy (where strategy is defined as a control alternative or none for each source) which meets any required constraints is evaluated. Evaluation starts by converting emissions to air quality data with a meteorological model (blocks 4 and 11) Standards compliance is checked (block 1 3 ) . At the same time, demographic data and air quality data are used to determine health 412

Environmental Science & Technology

damages and benefits from abatement (blocks 5, 12, and 17). Finally, economic impacts on the regions are determined (blocks 14 and 151, and decision inputs are used to make the strategy selection decision (block 1 8 ) . Currently, models and data exist for some of the blocks of the framework. Some areas in which additional work is needed are in detailed growth projections, doseresponse and benefit models, and determination of detailed economic impact. Existing models which have been implemented to estimate costs and air quality for regions of Air Quality Control Region (AQCR) size can be considered as subsets of the framework. However, the type of model is at least as important as the particular framework features incorporated. Two types of models currently exist: the first starts with a mathematical optimization technique, and the real system is simplified to meet the requirements of the model; the second begins with as accurate a description as possible of the real system and uses whatever techniques can be found to approach an optimal solution. The first technique finds an exact solution to an approximate problem, while the second finds an approximate solution to an exact problem. To date, no one has proposed a practical technique for finding the exact solution to the exact problem for large systems such as this. The first type of model is classified as mathematical and the second type as heuristic. Two models which illustrate the difference are a mathematical model of air quality in St. Louis (Kohn, 1970), and a heuristic model including elements of the Implementation Planning Program (IPP) model (developed by TRW Systems in 1970) and the Direct Cost of Implementation Model (DCIM)-developed by CONSAD Research Corp. in 1972. In comparing the two models, a major emphasis will be placed on which type of model is of most help in assisting the policy decisionmaker. The Kohn in St. Louis model is a macrolevel model in the sense that it measures air quality at a single point and treats sources in groups. This means that the resultant quality computed by the model is based directly on total emissions regardless of their geographic location. An advantage of this model is that it is in linear programming format and can easily be run using a standard linear programming system. The IPP model, a microlevel heuristic model, treats each point source in the AQCR individually, simulating application of each feasible control alternative (of which there are 50) in turn. The effects of emissions on air quality are simulated by an atmospheric diffusion model which estimates pollutant concentrations at a number of receptor points. The original IPP did not possess an optimization capability; an .inexact integer program technique was later added to the system to determine near-optimal solutions. A further disadvantage of IPP was that it treated only two pollutants simultaneously. Model characteristics are listed in Table 1. Clearly the Kohn model offers a number of advantages. Similar approaches have also been developed.

Predicted inventory fils

9

IB I

Ai ter na!ive ccntrols file

Costs 2nd air quality

6 Alternative controls file

7

Current air quality fila

The decisionmaker must use some additional criteria in selecting a modeling technique. For example, if a model is used in developing standards and the standards are challenged, the agency setting the standard may be required to demonstrate convincingly its logic to nonexperts in a courtroom situation. I f this occurs, the microlevel model has numerous advantages, particularly its treatment of individual pollutant sources. While pollution can be viewed easily as a macrolevel phenomenon and strategies defined by tons of emissions reduced, implementation of any strategy must involve a microlevel application of each control strategy sourceby-source. Thus, aggregating data on sources in the macrolevel does not really represent the physical realities of implementing controls on sources with different characteristics. For example, for some industrial processes, switching fuel types is relatively easily accomplished; while for others, it is close to impossible. The micromodel may better reflect this difference. In pollution control, there has been much more experience in designing and implementing models and setting standards than in actually implementing standards. Thus, there is little experience in evaluating the accuracy of the models available. One method of ensuring the accuracy of a computer model is to have the model operate on a detail level.

least cost solution meeting particular air quality standards. DClM added this capability. A macroflow chart of IPP-DCIM is given in Figure 2 . Each of the programs includes: 0 Emission inventory-corresponds to block 1 of the general framework 0 Regional cost data-correspond to block 3 0 Meteorological data-correspond to block 4 0 Control technologies and costs-correspond to block 8 0 Meteorological model-a diffusion model which computes existing air quality based on current emissions and meteorological data 0 Alternative controls, each source-relate to block 9 0 Current air quality-called the source contribution file; this file contains a table indicating the pollutant concentration at each simulated receptor from each source 0 Optimization (standards model)-is used to apply emission standards selected by the user or minimize control cost while meeting fuel and air quality constraints. This model thus performs a decision process producing a selected strategy consisting of a control alternative for each source. A comparison of the IPP-DCIM structure given in Figure 2 with the general framework of Figure 1 shows that the IPP-DCIM system can be considered as a subset of the overall model. Two decision criteria are built into IPP-DCIM. I t can simulate source standards by selecting the least cost solution meeting the applicable standard for each source, or it can select the overall least cost meeting fuel and air quality constraints. The "least cost" solution is not guaranteed to be optimum because a heuristic technique is used. The overall least cost solution can help the policy planner in selecting standards to simulate since, in effect, this solution applies a standard to each source. IPP-DCIM does not provide for computation of emissions growth, benefits, or economic impacts. However, despite these weaknesses, the model has been able to provide results useful to federal, state. and urban agencies in developing abatement implementation plans.

Describing a micromodel

Using a microlevel model

One attempt which has been made to assist in analysis of control strategies is the IPP-DCIM system mentioned earlier. The stated purpose of IPP was to assist state governments in preparing implementation plans for control of sulfur oxides and particulate matter. IPP simulated proposed or existing standards but could not generate a

Results are presented below for the following strategies simulated for the New York and Philadelphia AQCR's: Least cost subject to primary air quality standards. In this strategy, pollutants at each receptor were constrained so as not to exceed concentrations of 80 p g / m 3

TABLE 2

Results of Applying Strategies NEW YORK AQCR Least cost subject to standards

Maximum control

658 99 X I O 3 tons SO,

414

Environmental Science & Technology

. .

Least cost emission charges

PHILADELPHIA AQCR Least cost subject to standards

Maximum control

Least cost emission charges

SO, and 75 p g / m 3 particulate matter. The heuristically obtained least cost solution meeting these constraints was then determined. Maximum reduction efficiency. This strategy maximizes the total reduction of pollutants by weight. Computed by summing the SO, and particulate matter reductions, the alternative is selected for each point source regardless of cost. 0 Least cost with emission charges. This strategy involves selection of the least cost solution for each point source based on the sum of the costs of the device plus a cost for emissions produced. A uniform cost of 10Q/lb for emissions of each pollutant was used as the penalty cost. These data (Table 2) and other results lead to a number of conclusions. One which is obvious from Table 2 is that, while costs and total emission reductions were greater under maximum control than least costs subject to air quality standards, ambient air quality at the worst receptor was not significantly different. The New York and Philadelphia runs showed a large increase in the quantities of low-sulfur fuels which occurred because switching to low-sulfur fuels was the least expensive option for many sources. A series of IPPDClM runs was made, for the Buffalo. New York area AQCR-constraining fuels. Comparisons were made with the existing State Implementation Plan regulations. These results, while very preliminary, showed a large increase in the amount of hig'h-sulfur coal burned, a number of point sources which had required additional control in the State Implementation Plan were left in their original state, high-sulfur residual oil consumption increased, and lowsulfur distillate oil and gas consumption remained the same (by constraint). A variety of conclusions can be obtained from analysis of the micromodels output. By analyzing the devices selected by industry type for a specific strategy, a series of industry-based standards can be developed. Alternatively, standards could be based on the geographic location since the model indicated that "hot spots" occurred at several receptors. Some of the more basic problems with use of a microtype model are difficulty of obtaining detailed data required, problem of projecting growth of emissions, difficulty of dealing with area sources at a microlevel, difficulties in setting up and running a large model on a production basis, and accuracy of all data and ability of the model to accurately reflect all alternatives. The problem of obtaining data wiii probably be solved with the passage of time. Data collection efforts are under way in a number of AQCR's to define the emissions inventory. The problem of how to e p n a t e growth is difficult at a microlevei. An interim solution may be to use macrolevel techniques until better methods are developed. Area sources are a significant contribution to ambient air pollution, but detailed data are not available to permit them to be handled as easily as point sources. Most area sources are either transportation (mobile sources) or fixed, but small heating and incineration sources. Ttansportation sources will continue to be treated at an aggregate level. Heating sources can be characterized by the amounts and types of fuels burned, and control strategies would consist of fuel switching. Incineration Sources are difficult to define exactly. More work is needed here. Setting up and operating micromodels is difficult and requires sophisticated personnel. Generally. manual analysis of certain parts of the data base is required, and computer programs must often be modified to devise and

test new strategies. For example, tests on the IPP-DCIM fuel constraint code indicated the great difficulty of constraining air quality and fuels simultaneously. Additionally, the dimensionality of the problem was further increased. These difficulties suggest the need for a largescale mixed integer linear program. While the IPP type of micromodel has many weaknesses, it still provides a powerful tool to the policy planner. Work in expanding the usefulness and capability of such models will be rewarded by making strategy development and implementation more efficient and accurate. Additional reading Kohn. R. E., "Abatement Strategy and Air Quality Standards in Development of Air Quality Standards," Arthur Atkisson and Richard S. Gaines, Eds., Charles E. Merrili Publishing Co., Columbus, Ohio, 1970. TRW Systems. Air Quality Implementation Planning Program Contract PH 22-68-60, Environmental Protection Agency, 1970. Norsworthy, J. R.. Teller, A,, "The Evaluation of the Cost of Aiternative Strategies for Air Pollution Control." APCA Paper 69172. NewYork. N.Y.. June22-26,1969. CONSAD Research Corp., "The Direct Cost of Implementation Model." Vol. 1, Environmental Protection Agency, 1972. Burton, E. S., Sanjour Wm., "A Simulation Approach to Air Pollution~Abatement Program Planning," J. Socio-Economic Plannkgsci., VOI. 4, pp 147-59, 1970.

Ellison S. Burton is director, Standards and Regulations Evaluation Division, Office of Planning and Evaluation for the U.S. Environmental Protection Agency. Mr. Burton has managed and participated in the deveiopment of a number of environmental modeis and has authored over 20 papers and professional journal articles on pollution control and management.

Edward H. Pechan, a director of the Management Research institute and a consultant in computer systems and operations research, is currently engaged by the federal €PA. Mr. Pechan has made several contributions i n the development of air poiiution monitoring systems and has extensive experience in systems management and operations research. Address inquiries to Mr. Pechan in the Office of Planning and Evaluation, Environmental Protection Agency.

William Sanjour is also a director of the Management Research institute and a consultant in operations research currently engaged by the Office of Planning and Evaluation. U.S. €PA. Mr. Sanjour has been instrumental in developing some of the models discussed in this article as well as many other environmental studies and operations research.

Volume 7, Number 5, May 1973

415