-5

Like many aspects of the SIP process, the compilation of an emissions inventory is not an exact science. We are learning more about what kinds of sources emit what amounts of VOCS. For example, just a few years ago we did not realize that publicly owned treatment works (POTWs) and hazardous waste treatment, storage, and disposal facilities (TSDFs) could emit relatively large amounts of VOCs. Consequently, they often were not included in emissions inventories. Now they will be. By overlooking important emissions sources in the past, we also overestimated the improvement in air quality that would result from emissions reductions from other sources.

Even when we could identify emissions sources clearly, it has sometimes been difficult to characterize their emissions accurately. This is particularly true for small, widely dispersed sources like degreasing facilities. Thus, the States' ability to estimate the quantity and location of emissions has been inexact at best. Furthermore, SIP inventories are compiled for a "typical summer day," and they do not necessarily incorporate fluctuations in source production rates or operating schedules. Consequently, imprecision is inevitable.

After air quality has been monitored and the sources of emissions inventoried, the emissions reductions set forth in the SIP are incorporated into a mathematical model to see whether post-control air quality will meet the national standard. In one sense, this is the most important step in the SIP process. The ability to accurately predict the

results of State and EPA control actions is essential to

the success of an air quality control program.

-6

Modelling the emissions of some pollutants--carbon

monoxide, for example--is relatively straightforward. About
90 percent of the CO emitted in typical urban areas comes from
cars and trucks. It tends to remain in the same "hot spots"
where it is emitted, and it is not converted over time into
another pollutant. A reduction in CO emissions will tend to

result in a proportionate reduction in CO concentrations in
the same area where the emissions reductions are made. Thus,
in the case of CO, the use of a relatively simple proportional
model tends to give reasonably accurate predictions of air
quality over time.

But the situation is much more complex with ozone. VOCS are emitted into the atmosphere and then undergo photochemical transformations with NOx in the presence of sunlight, resulting in the creation of ozone.

Because VOCS, NOx, and ozone all

can be transported dozens of miles during and after transformation, it is much more difficult to predict where and to what extent reductions in ozone concentrations will result from the reduction of its precursor emissions.

Because a simple proportional model is not sufficient for predicting changes in air quality over wide areas, EPA has developed several other models that improve our predictive capabilities. For example, in the late 1970s EPA developed a photochemical model called the Empirical Kinetics Modelling Approach, or EKMA. Because the EKMA model factors in the effects of sunlight, chemical reactivity, and the amount of

-7

ozone being transported into an area, it provides a much more realistic representation of air quality. If the maximum ozone concentration and the ratio of non-methane organic compounds (NMOC) and nitrogen oxides (NOx) are known, EKMA provides an estimate of how much VOC and NOx emissions need to be reduced to cut ozone concentrations to meet the national standard. One of the major advantages of EKMA is that it minimizes the data-gathering burden while still factoring in complex atmospheric chemistry. EKMA has been used widely by State air quality programs since the early 1980s.

EKMA, however, has its own problems. For example, it uses simplified meteorological conditions, and it can only analyze maximum ozone concentrations. Since it is not reliable for predicting the distribution of ozone concentrations across a given area, it is not very useful for determining how the boundaries of a particular non attainment problem might change over time due to reductions in VOC and NOx emissions.

To overcome these deficiencies, EPA sponsored the development of the Urban Airshed Model (UAM). The UAM is much more sophisticated than EKMA. It can consider meteorological conditions, and it can predict differences in air quality at different places in an area. Thus it can demonstrate progress in terms of declining peak ozone concentrations as well as indicate of how peak ozone concentrations will change across an entire urban area.

-8

This additional sophistication comes at a cost, of course. UAM requires much more data input, data validation, and computer capacity. Consequently, it has not yet been used very widely. But as a reliable tool for predicting the possible effects of Federal and State control programs over an entire urban area, it represents a significant improvement in the state of the art of ozone modelling.

EPA is now in the process of broadening the Urban Airshed Model to cover larger geographical areas. When fully developed, this new modelling tool, the Regional Oxidant Model (ROM), will be capable of modelling the transportation of air pollutants over a large region; e.g., the U.S. northeastern corridor. Thus ROM will be able to predict the amount of ozone moving into and out of a particular area. Used in conjunction with EKMA or UAM, ROM will greatly improve our ability to predict the transboundary flow of ozone and its precursors and thus help the States refine their ozone control plans.

I have described these four parts of the air quality planning process to give you some sense of the complexity and imperfection of the tools we use. Each of these steps in the process introduces some element of possible inaccuracy or inexactness. The artificial boundaries of air quality control areas do not exactly match the constantly shifting mass of air pollutants that cause non attainment. Because of their cost, the number of air quality monitors is limited and the placement of those monitors requires good judgment. While improving, our knowledge of the existence and location

-9

of different VOC sources is incomplete, and the models we use to predict the hoped-for effects of emissions control can never produce an exact replica of the real world.

The widely recognized imperfections in the air quality planning process reflect the fact that EPA and the States must work in the real world. And in the real world our tools will never be perfect, no matter how much they are improved. In the real world, we always have to make our decisions using the best possible tools at hand, while simultaneously recognizing their inadequacies.

In the case of ozone, the past inaccuracies and inexactness of our technical tools sometimes have led us to underestimate the level of control needed in a particular area to attain the national ambient health standard. It has been difficult for States and EPA to develop and run the more complex computer models for all problem areas. The availability of more air quality monitors would have helped us better characterize the actual air quality in different areas, and that might have affected an area's SIP planning target.

But no one should be deluded into believing that by merely fine tuning these technical tools a significant part of the ozone non attainment problem will go away. Marginal improvements in the technical components of the planning process may make a difference in those areas with marginal air quality. But in the 25 to 30 most severely polluted areas, tinkering with