[[Page 48342]]

Studies of coke oven workers who were exposed to higher levels of coke oven emissions than the populations affected by these proposed amendments have reported an increase in cancer of the lung, trachea, bronchus, kidney, prostate, and other sites. Chronic (longterm) exposure of workers to coke oven emissions has also been associated with conjunctivitis, severe dermatitis, and lesions of the respiratory system and digestive system. We have classified coke oven emissions as a Group A, known human carcinogen.

One of the more important constituents of coke oven emissions (from a health effects point of view) is the trace metal arsenic, a known human carcinogen. Studies of humans occupationally exposed to higher levels of arsenic than the populations affected by these proposed amendments have found increased incidence of lung cancers. Chronic (longterm) exposure to inorganic arsenic has also been associated with irritation of the skin and mucous membranes, and with neurological injury. Animal studies of inhalation exposure have indicated developmental effects.

Another important constituent of coke oven emissions, benzene, is a known human carcinogen. Increased incidence of leukemia (cancer of the tissues that form white blood cells) has been observed in humans occupationally exposed to benzene, and we have derived a range of inhalation cancer unit risk estimates for benzene. The value at the high end of the range was used in this assessment. Chronic (longterm) inhalation exposure has caused various disorders in the blood, including reduced numbers of red blood cells, in occupationally exposed humans. Reproductive effects have been reported in women exposed by inhalation to high levels of benzene, and adverse effects for high dose exposures on the developing fetus have been observed in animal tests. III. Summary of the Proposed Amendments

A. What Are the Affected Sources and Emission Points?

The affected sources would be each coke oven battery subject to the emission limitations in 40 CFR 63.302 or 40 CFR 63.303 (i.e., the MACT track batteries). As noted above, the proposed amendments would cover emissions from doors, topside port lids, offtake systems, and charging on existing byproduct coke oven batteries and emissions from doors and charging on new and existing nonrecovery batteries.

B. What Are the Proposed Requirements?

For existing byproduct batteries, the proposed amendments would limit visible emissions from coke oven doors to 4 percent leaking doors for tall batteries and for batteries owned or operated by a foundry coke producer. Short batteries would be limited to 3.3 percent leaking doors. Visible emissions from other emission points would be limited to 0.4 percent leaking topside port lids and 2.5 percent leaking offtake systems. No change would be made in the limit for chargingemissions must not exceed 12 seconds of visible emissions per charge. Each of these visible emission limits would be based on a 30day rolling average. The proposed amendments would replace the less stringent limits that became effective on January 1, 2003, for MACT track batteries and are equivalent to the limits that will become effective on January 1, 2010, for LAER track batteries. We are not proposing to amend the standards for new byproduct batteries.

The monitoring, reporting, and recordkeeping requirements in the existing MACT standards would continue to apply to existing byproduct coke oven batteries on the MACT track. These requirements include daily performance tests to determine compliance with the visible emission limits. Each performance test must be conducted by a visible emissions observer certified according to the test method requirements. A daily inspection of the collecting main for leaks is also required. Specific work practice standards must also be implemented if required by the provisions in 40 CFR 63.306(c). Under the existing standards, companies must make semiannual compliance certifications; report any uncontrolled venting episodes or startup, shutdown, or malfunction events; and keep records of information needed to demonstrate compliance.

We are also proposing amendments for the improved control of charging emissions from a new nonrecovery battery (i.e., constructed or reconstructed on or after August 9, 2004. Fugitive charging emissions would be subject to an opacity limit of 20 percent. A weekly performance test would be required to determine the average opacity of five consecutive charges for each charging emissions capture system. Emissions from a charging emissions control device would be limited to 0.0081 pounds of PM per ton (lb/ton) of dry coal charged. A performance test using EPA Method 5 (40 CFR part 60, appendix A) would be required to demonstrate initial compliance with subsequent performance tests at least once during each title V permit term. If any visible emissions are observed from a charging emissions control device, the owner or operator would be required to take corrective action and followup with a visible emissions observation by EPA Method 9 (40 CFR part 60, appendix A) to ensure that the corrective action had been successful. Any Method 9 observation greater than 10 percent opacity would be reported as a deviation in the semiannual compliance report. The proposed amendments would also require the owner or operator to implement a new work practice standard designed to ensure that the draft on the oven is maximized during charging.

We are also proposing a work practice standard for the control of door leaks from all nonrecovery coke oven batteries on the MACT track. The owner or operator would be required to observe each coke oven door after each charge and record the oven number of any door from which visible emissions occur. If a coke oven door leak is observed at any time during the coking cycle, the owner or operator would be required to take corrective action and stop the leak within 15 minutes from the time the leak is first observed. No additional leaks would be allowed from doors on that oven for the remainder of that oven's coking cycle. However, we are also proposing to allow up to 45 minutes instead of 15 minutes to stop the leak for no more than two occurrences per battery during each semiannual reporting period. The limit of two occurrences per battery would not apply if a worker must enter a cokeside shed to take corrective action to stop a door leak. In this case, 45 minutes would be allowed to stop the leak, and the evacuation system and control device for the cokeside shed must be operated at all times that there is a leaking door under the cokeside shed. The owner or operator would also be required to identify malfunctions that might cause a door to leak, establish preventative measures, and specify types of corrective actions for such events in its startup, shutdown, and malfunction plan. Recordkeeping and reporting requirements necessary to demonstrate initial and continuous compliance are also proposed.

We are also proposing an amendment to clarify that the work practice standard for charging in 40 CFR 63.303(a)(2) that applies to existing nonrecovery batteries also applies to new nonrecovery batteries. These work
[[Page 48343]]

practices are described in 40 CFR 63.306(b)(6).

As specified in the CAA section 112(f)(4)(A), the owner or operator of an existing byproduct coke oven battery on the MACT track would have to comply with the proposed amendments within 90 days of the effective date of the final rule amendments. We are also proposing that nonrecovery coke oven batteries on the MACT track comply within 90 days (or upon startup for a new nonrecovery battery which comes into existence after August 9, 2004).
IV. Rationale for the Proposed Amendments

A. How Did We Estimate Risks?

Cancer and noncancer health impacts caused by environmental exposures generally cannot be isolated and measured directly. Even if it were possible to do so, we would not be able to use measurements to assess the impacts of future or alternative regulatory control strategies. As a result, modelingbased risk assessment is used as a tool to estimate health risks for many EPA programs. In risk assessments, there are many possible levels of analysis from the most basic screening approach to the more refined, detailed assessment.

Our ``Residual Risk Report to Congress'' (EPA453/R99011) provides the general framework for conducting risk assessments to support decisions made under the residual risk program. The 1999 Report to Congress acknowledged that each risk assessment design would have some common elements. In general, each assessment would contain a problem formulation phase where the content and scope of each assessment would be specified, an analysis phase where the exposure and effects relationship would be evaluated, and the risk characterization phase where the risks would be calculated and interpreted. While the final risk assessment used to support the decisions in these proposed amendments used advanced modeling of sitespecific data for many modeling parameters and population characteristics derived from census data, we also used default assumptions for exposure parameterssome of which are assumed to be health protective (e.g., exposure frequency and exposure duration, 70year constant emission rates).\6,\ \7\ However, in keeping with the tiered approach laid out in the Report to Congress, we decided that a quantitative description of uncertainty in the final risk characterization was not necessary for this assessment because it likely would not have altered the decision to propose further standards. The approach used to assess the risks associated with our coke oven standards is consistent with the technical approach and policies described in the Report to Congress.
\6\ Additional details are provided in Table 210 of the risk assessment document in the rulemaking docket.
\7\ Residual Risk Report to Congress, pp. B18 and B22.

B. What Did We Analyze in the Risk Assessment?

We performed a detailed risk assessment for the four byproduct coke facilities (five MACT track batteries). Given the small number of facilities, we chose to analyze each of these facilities in a site specific manner. As described earlier, there are multiple source categories associated with coke ovens, each with its own standards. There are two MACT standards that affect this industry (i.e., the 1993 national emission standards for charging, topside leaks, and door leaks and the 2003 NESHAP for pushing, quenching, and battery stacks), as well as the 1989 NESHAP for coke byproduct recovery plants and the 1990 NESHAP for benzene waste operations. Using an iterative assessment approach, we assessed emissions and estimated risks from all emission points at each coke facility. The initial screeninglevel analysis considered all emission points to determine if a more refined analysis was necessary and to determine the focus of such an analysis. A more refined analysis was then performed to determine the maximum individual risk and the risk distribution around the facilities. Results from the refined analysis are presented in this preamble.

Emission points associated with the coking process include charging, door leaks, topside leaks, pushing, quenching, battery stacks, and the byproduct recovery plant. To estimate baseline risks (both baseline facilitywide emissions and baseline of 1993 MACT emission points), we assumed that each battery was in compliance with its required performance level and that emission rates were equivalent to those allowed by the national emission standards. We modeled emissions at the rate allowed by the national emission standards because it represents the source's potential emissions and risks, and is, therefore, consistent with the language in section 112(f)(2), which states that ``if standards promulgated pursuant to subsection (d) * * * do not reduce lifetime risk * * * to less than one in a million, the Administrator shall promulgate standards under this subsection * * *'' We specifically request comments on this interpretation of section 112(f)(2).

Emission estimates for individual batteries were based on battery specific data such as coking time; the number of doors, lids, and offtakes on each battery; and the number of charges per year, as well as the performance standards for those emission points (5 percent leaking doors, 0.6 percent leaking lids, 3 percent leaking offtakes, and 12 seconds of visible emissions per charge). For the facility with two operating coke batteries, emission estimates for both batteries were combined to yield a risk estimate from the facility. The battery characteristics were obtained from a survey of the industry and from an EPA report that assessed control performance for these emission points at a coke facility that is similar to those included in this assessment. Information on the tons of coke produced and the tons of coal charged were also obtained from the industry survey. Emission estimates were based on emission factors for each emissions point and the applicable regulatory emissions limit. Our uncertainty analysis shows that the use of sitespecific data and emission factors results in an uncertainty range for the emission estimates for leaks from doors, lids, and offtakes that may be a factor of 2 lower or a factor of 3 higher for these combined emission points. The uncertainty is dominated by the emissions from leaking doors, which comprise approximately 90 percent of the total emissions. We did not evaluate the uncertainty in estimates of charging emissions, which contribute less than 7 percent of the total emissions. Additional information on the uncertainty analysis is included in the risk assessment document.

Emissions from pushing, quenching, and battery stacks were derived from two EPA tests, one at a battery producing foundry coke and one at a battery producing furnace coke. Pushing emission estimates included fugitive emissions and emissions from control devices. Because emissions vary depending on the type of push experienced (e.g., ``green'' pushes result when coal is not fully coked), emission factors were used for the range of pushes experienced. Supporting data for estimating the number and frequency of green pushes were obtained from visible emission observations at several facilities. We then calculated an overall pushing emissions rate based on the frequency of green pushes and emission factors for each type of push. Emissions farom quenching and battery stacks were based on emissions tests.

Emissions from the byproduct recovery plant were estimated from [[Page 48344]]
information on the type of processes at each facility, emission factors for each process, and the facility capacity. Emissions from equipment leaks were based on the number of equipment components at each facility, the composition of process liquids, and emission factors for each component. Emissions from benzene waste operations were estimated from sitespecific data on the quantity of benzene in wastewater. In assessing risk from all of the emission points mentioned above, we used a combination of sitespecific data and estimation techniques as inputs to the models used to evaluate risk and hazard.

Our analysis of nonrecovery batteries on the MACT track indicates that emissions from charging and door leaks are relatively low. There are no emissions from lids and offtakes because existing nonrecovery batteries in the U.S. do not have these emission points. There are no emissions from door leaks during most normal operations because the ovens usually operate under negative pressure. Our modeling approach based on allowable emissions under MACT (zero percent leaking doors for nonrecovery batteries) would estimate no door leak emissions at all. However, we recently obtained information that indicates certain equipment failures or operating problems can temporarily create a positive pressure in an oven and cause a door to leak. These events are considered to be short in duration and the problem can be quickly remedied (typically within 5 to 15 minutes). In order to ensure that door leak emissions are minimized, we have addressed these equipment failures and operating problems in our proposed amendments to the 1993 national emission standards. The proposed revisions would require that corrective actions be implemented promptly if such events occur.

With respect to emissions from charging, nonrecovery ovens are operated under maximum draft during charging, and the organic compounds that may be generated during the process are mostly contained within the oven and combustion system. A small amount of charging emissions may escape from an oven through the opening used for charging. However, all nonrecovery batteries have a capture hood and baghouse to control these emissions.

Consequently, we would not anticipate any adverse public health or environmental impacts due to emissions from charging and coke oven doors at nonrecovery batteries.

C. How Were Cancer and Noncancer Risks Estimated?

The primary HAP emitted by this category are coke oven emissions which include POM, PAH, benzene, and other air toxics known or suspected to cause cancer and other health problems. For estimating cancer health risk due to inhalation exposure, emissions were based on the benzene soluble organics (BSO) fraction that was used as the surrogate for coke oven emissions in the epidemiology study which established coke oven emissions as a human carcinogen. In the assessment of noninhalation risk, coke oven emissions were characterized and speciated (i.e., individual constituents were identified). A set of 13 constituents \8\ was selected based on an analysis of their persistence, bioaccumulation, and toxicity (PBT). Emission estimates were determined for all constituents identified based on measurements of the chemical composition of the emissions from various emission sources. For this risk assessment, emission estimates for coke oven emissions (as BSO) were determined for charging, door leaks, topside leaks, fugitive pushing, and quenching emission points for byproduct batteries. Emission rates for individual constituents were estimated for the pushing control device and battery stack emission points. Emission rates also were estimated for the HAP compounds known to be emitted from the byproduct recovery plant (benzene, xylene, and toluene).
\8\ Constituents of coke oven emissions selected for this assessment include: acenaphthene, anthracene, benz(a)anthracene, benzo(a)pyrene, benzo(b)fluoranthene, benzo(k)fluoranthene, cadmium, chrysene, fluoranthene, fluorene, indeno(1,2,3cd)pyrene, lead, and pyrene.

To characterize the risk from exposure to these HAP, toxicity information was integrated with results from the exposure assessment. For this assessment, we modeled exposures to the total population living within 50 kilometers (km) of each of these facilities and estimated the exposure concentrations where people live and the cancer risks associated with lifetime exposures to coke oven emissions and to the individual constituents for which we have cancer unit risk factors. Where reference values for noncancer effects were available, we also evaluated the potential hazard associated with those effects. The selection and use of cancer unit risk factors and reference dose or concentration values for this assessment follows the approach outlined in the 1999 ``Residual Risk Report to Congress.'' The approach used to assess the risks associated with our coke oven standards is likewise consistent with the technical approach and policies described in the report. Our assessment has also been peerreviewed to ensure that its methodology rests on sound scientific principles, and we have revised the assessment document to reflect comments made as part of the peer review process. The assessment document, comments made during the peer review, and a summary of our responses to those comments are included in the docket for the proposed amendments.
D. How Did We Estimate the Atmospheric Dispersion of Emitted Pollutants?

As described in our Report to Congress, risk assessments may use a variety of models to describe the fate and transport of HAP released to the atmosphere. The models chosen must be appropriate for the intended application. In the fairly unique case of coke ovens, the collective heat rising from various emission points can significantly enhance the rise of the emissions plume, functioning like a ``representative'' stack. In order to include this aspect in the modeling, we used the Buoyant Line and Point Source (BLP) dispersion model. The BLP model, however, was not designed to consider the effects of the surrounding terrain on dispersion nor to model deposition of HAP as the plume disperses. To allow consideration of these parameters, we coupled the BLP model with the Industrial Source Complex Short Term (ISCST3) model. In this application, we used the BLP model to estimate the plume height and then used that value as an input to the ISCST3 model. The ISCST3 model was used to simulate the subsequent dispersion and transport of the emissions. Sitespecific inputs to the BLP model such as facility location, battery layout, dimensions, orientation, and operating temperatures were provided by the industry.

Both the BLP and the ISCLT3 models have undergone standard scientific peer reviews prior to this assessment. The concept of coupling these two models together was peerreviewed for the first time as part of this assessment. The reviewers agreed with the modeling concept and approach. Monitoring data may be useful for evaluating modeling approaches used to estimate ambient concentrations (see the risk assessment document for discussion of when this is appropriate). For the sites and pollutants included in this risk assessment, no ambient monitoring data were available. Therefore, it was not possible to evaluate the modeling
[[Page 48345]]
approach beyond what was done in the peer review. Moreover, even if comprehensive and high quality monitoring data were available, they would not be adequate by themselves for evaluating the impacts of alternative control strategies.

E. What Factors Are Considered in the Risk Assessment?

The risk assessment was designed to generate a series of risk metrics that would provide information for a regulatory decision. The metrics consider both the maximum individual risk and the total population risk, the latter providing perspective on the potential public health impact by addressing each of the following questions:

In addition, we are to determine if any adverse environmental effects exist.

Consistent with standard atmospheric dispersion modeling practice, we assessed inhalation risks within 50 km (about 30 miles) of each of the four facilities. The annual average concentrations at the area weighted centers of census blocks or block groups were estimated using the ISCST3 model for each emission point. Based on the number of people residing in each block or block group along with the estimated concentrations in each block or block group, we generated an estimate of risk for all people living within 50 km (about 30 miles) of each coke facility, including an identification of which census block group had the estimated maximum air concentration. For this estimate, we assumed that the individual is exposed to the maximum level of coke oven emissions allowed by the 1993 national emission standards, and, as prescribed in the 1989 Benzene NESHAP, that they are exposed to these emissions 24 hours a day for 70 years. Where risk estimates exceeded 1 in a million, we identified the number of people at the various risk levels exceeding 1 in a million (i.e., the population risk distribution). For this estimate, we also assumed exposure occurred 24 hours a day for 70 years because we wanted a conservative upperbound estimate of the population at risk.

Because of their chemical and physical properties, some HAP are known to present potential health risks as a result of deposition, persistence, and bioaccumulation in environmental media other than air. As a result, exposure to these HAP may occur by ingestion as well as by inhalation. Thirteen constituents of coke oven emissions were identified as PBT chemicals (i.e., they are environmentally persistent, they may bioaccumulate, and are toxic). Emissions of these pollutants are transported from the emission site by atmospheric processes and removed from the air by both wet and dry deposition. Upon deposition, they may cycle through various environmental compartments, such as soil, plants, animals, and surface water. The movement of these constituents through these compartments can be modeled using a fate and transport model in order to estimate human exposure through the ingestion pathway.

We conducted multimedia, multipathway exposure modeling (using the EPA's Indirect Exposure Model) to determine if emissions from coke ovens present potential risks by routes of exposure other than inhalation. Sitespecific modeling was performed for all four facilities using information collected on land use, population, soil types, farming activity, and watershed/waterbody locations and areas. The assessment was based on a subsistence farmer scenario located where landuse data identified actual farming activity around each of the four facilities (agricultural lands were identified at distances ranging from 1.7 to 11 km from the four coke facilities). This scenario reflects an adult living on a farm and consuming meat, dairy products, and vegetables that the farm produces. The animals raised on the farm subsist primarily on forage that is grown on the farm. We also assumed that the farm family fishes in nearby waters at a recreational level, and that they eat the fish they catch. These results allow for comparison of risks by ingestion with those presented by inhalation. F. How Did We Calculate Risks?

Cancer risks were characterized for the inhalation exposure pathway using lifetime excess cancer risk estimates which are calculated as the product of the unit risk estimate (URE) (the unit risk estimate is an upperbound estimate of the probability of developing cancer over a lifetime) and the exposure concentration estimated for each HAP. The cancer risk estimates for each HAP are summed across all carcinogenic HAP. These estimates represent the probability of developing cancer over a lifetime as a result of exposure to emissions from these coke ovens.

Noncancer risks were characterized through the use of hazard quotient (HQ) and hazard index (HI). An HQ is calculated as the ratio of the exposure concentration of a pollutant to its benchmark concentration. An HI is the sum of HQ for HAP that target the same organ or system.

The maximum individual risk was estimated deterministically. More probabilistic presentations and analyses (ranging from simple risk distributions to more quantitative Monte Carlo simulations) \9\ may be done to better understand the assessment uncertainty and variability. As our Residual Risk Report to Congress suggested, we would consider doing a probabilistic analysis after considering the needs and scope of the assessment. This is consistent with the policy of EPA as stated in the 1997 ``Policy for Use of Probabilistic Analysis in Risk Assessment,'' which states ``* * * it is not the intent of this policy to recommend that probabilistic analysis be conducted for all risk assessments supporting risk management decisions.'' \10\ The policy also states ``* * * probabilistic methods should be used wherever the circumstances justify these approaches.'' As discussed earlier in this preamble, we determined that this level of refinement was not necessary for this risk assessment because the results of a probabilistic analysis are unlikely to affect the proposed risk management decisions. \9\ Residual Risk Report to Congress, pp. 94128.
\10\ Policy for Use of Probabilistic Analysis in Risk
Assessment, EPA Science Policy Council. May 15, 1997.

G. How Did We Assess Environmental Impacts?

In order to assess whether the continuing emissions from these four coke oven facilities could contribute to adverse environmental effects, we performed a screeninglevel ecological risk assessment. We intentionally designed this assessment to be protective of the health of ecological receptors. It was not intended to be used in predicting specific types of effects to individuals, species, populations, or communities or to the structure and function of the ecosystem. We used the assessment to identify HAP or sources which may pose potential risk or hazard to ecological receptors and, if so, would need to be evaluated in a more refined level of risk assessment.

The screening endpoints were the structure and function of generic aquatic and terrestrial populations and communities, including threatened and endangered species, that might be
[[Page 48346]]
exposed to HAP emissions from these four facilities. The assessment endpoints were relatively generic with respect to descriptions of the environmental values that are to be protected and the characteristics of the ecological entities and their attributes. We assumed in the assessment that these ecological receptors were representative of sensitive individuals, populations, and communities that may be present near these facilities.

The HAP included in the ecological assessment were the metals cadmium and lead and 11 PAH: Acenaphthene, anthracene, benzo(a)pyrene, benzo(a)anthracene, chrysene, benzo(b)fluoranthene,
benzo(k)fluoranthene, fluoranthene, fluorene, pyrene, and indeno 123(cd)pyrene. We derived estimated media concentrations for each of these HAP from the media concentrations estimated in the multipathway exposures assessment. We chose exposure pathways to reflect the potential routes of exposure through sediment, soil, water, and air. We selected these environments because they are considered representative of locations of generic populations and communities most likely to be exposed to the HAP. Within these environments the receptors evaluated consisted of two distinct groups: Terrestrial and aquatic (i.e., including aquatic, benthic, and soil organisms; terrestrial plants and wildlife; and herbivorous, piscivorus, and carnivorous wildlife).

The chronic ecological toxicity screening values used in the assessment were estimates of the maximum concentrations that should not affect survival, growth, or reproduction of sensitive species after longterm (more than 30 days) exposure to HAP. We screened HAP, pathways, and receptors using the ecological HQ method, which simply calculates the ratio of the estimated environmental concentrations to the selected ecological screening values.

H. What Are the Results of the Risk Assessment?

Table 1 of this preamble summarizes the estimated maximum individual risk using the modeled ambient air concentrations from the refined air modeling assessment and risk distribution for the four facilities at the baseline emissions level (i.e., risks based on MACT allowable emission levels allowed by the three regulations for all emission points assessed across the four coke facilities). Table 1 of this preamble also shows the estimated risks attributable to emissions from only charging, door, and topside leaks under the 1993 national emission standards. These latter emissions contribute about 38 percent of total facility HAP emissions.
Table 1.Baseline Risk Estimates Due to HAP Exposure Based on 70Year Exposure Duration \1\
1993 national Parameter Facility emission standards Maximum individual risk from 500 in a million.. 200 in a million. facility with highest risk.
Annual cancer incidence summed 0.1............... 0.04
for all four facilities (cases/
year).
Population at risk across all
four facilities (modeled to 50
km):
> 1 in a million............ 900,000........... 300,000 > 10 in a million........... 50,000............ 8,000 > 100 in a million.......... 300............... 8
Total modeled........... 4,000,000......... 4,000,000 \1\ All risk, cancer incidence, and population estimates are rounded to one significant figure.

The maximum individual facilitylevel risk (i.e., modeled risk based on emission levels allowed by the three regulations for all emission points assessed) is 500 in a million compared to 200 in a million for emissions only from those processes associated with the 1993 national emission standards. This level of risk was seen at only one of the four facilities assessed. The maximum individual facility level risk values for the other three facilities were 50, 100, and 100 in a million compared with risks of 20, 50, and 70 in a million, respectively, for emissions associated with only the 1993 national emission standards.

The annual cancer incidence (the number of cancer cases estimated to occur) for all facilities combined is 0.1 and 0.04 cases per year based on the facility level versus the emissions level from sources subject to the 1993 national emission standards, respectively. Across all four facilities, and assuming the entire population is exposed for 70 years, approximately 900,000 persons (approximately 20 percent of total population) are estimated to be exposed to risks greater than 1 in a million for the total facility emissions compared to 300,000 persons (approximately 7 percent) for the emission points subject to the 1993 national emission standards.

We also evaluated potential risks for adverse health effects other than cancer. The estimated maximum inhalation HI for any noncancer effect from an entire facility is 0.4 for hematologic (blood) effects due to benzene. In addition, results from a multipathway risk assessment presented in the risk assessment document shows that cancer risks from inhalation exposures exceed cancer risks due to ingestion, generally, by an order of magnitude. In this same assessment, the noncancer ingestion HI was estimated to be 0.001. This level was seen at two facilities assessed with highend exposure factors.

The results of a screeninglevel ecological assessment show that each of the coke plants had ecological HQ values less than 1 for all pollutants assessed. Therefore, it is not likely that the HAP emitted would pose an ecological risk to ecosystems near any of these facilities. It is also not likely that any threatened and endangered species, if they exist around these facilities, would be adversely affected by these HAP emissions because they are not likely to be any more sensitive to the effects of these HAP than the species evaluated.

The risk analysis assumed that all emission points from the batteries are leaking or emitting at the maximum rate allowable under the 1993 national emission standards for charging, doors, and topside leaks, since it is theoretically possible that these amounts of emissions could occur. However, this assumption (although theoretically possible) overstates actual emission levels. We analyzed 1,000 to 2,600 daily compliance determinations for each battery to compare the actual average emissions to the maximum rate allowed under the 1993 national emission standards as modeled.\11\ The
[[Page 48347]]
results of this analysis indicate that average performance is better than the current MACT limits and is closer to the more stringent 2010 LAER limits. The five MACT track batteries average 44 percent of the MACT limit for doors leaks, 16 percent of the limit for lid leaks, 21 percent of the limit for offtake leaks, and 27 percent of the limit for charging. An average performance that is better than the limit is to be expected because if batteries were to operate on average at the level of the 1993 national emission standards, they would likely exceed the standards a high percent of the time. Consequently, facility owners and operators consistently operate below the standards to avoid violations. \11\ We updated the database to include inspections in 2003. There was only a small change from the previous database used in the risk analysis for actual emissions, and the update did not have a significant impact on the estimate of emissions and risks.

Table 2 of this preamble repeats (from Table 1) the estimated risks attributable to charging, doors, lids, and offtakes at the baseline level (i.e., the level of risk assuming emissions from the batteries are at the maximum allowed by the 1993 national emission standards). Table 2 of this preamble further projects risks at the 2010 LAER level. Table 2.Risk Estimates Due to HAP Exposure Based on 70Year Exposure Duration
1993 national
Parameter emission standards 2010 LAER Maximum individual risk at 200 in a million.. 180 in a facility with highest risk. million.\1\ Annual cancer incidence summed 0.04.............. 0.03
for all four facilities (cases/
year).
Population at risk across all
four facilities (modeled to 50
km):
> 1 in a million............ 300,000........... 200,000 > 10 in a million........... 8,000............. 7,000 > 100 in a million.......... 8................. 6
Total modeled........... 4,000,000......... 4,000,000 \1\ The maximum individual risk estimate of 180 in a million is presented with two significant figures in order to show the risk reduction expected by the 10 percent decrease in emissions we anticipate seeing between the 1993 and 2010 emission levels.

The maximum individual risk is 200 in a million for the baseline and 180 in a million for the 2010 LAER limits. For the baseline, 93 percent of the total modeled population is exposed to risk levels less than 1 in a million compared to 95 percent for the 2010 LAER limits (based on 70year exposure duration). However, because these facilities are in fact performing better than the limits in the 1993 national emission standards (i.e., they could already meet the 2010 LAER limits), the difference in risk between the two scenarios may be smaller than the table indicates (and could be as small as zero).

We acknowledge that there are uncertainties in various aspects of risk assessment due to the use of some modeling and exposure assumptions. In this risk assessment, the use of these assumptions is likely to result in our overestimating the maximum individual risk and the magnitude of risk experienced by individual members of the population. For example, Tables 1 and 2 of this preamble present estimates of the number of people whose individual risk exceeds various levels (e.g., 1 in a million, 10 in a million, 100 in a million) under different scenarios (e.g., 1993 national emission standards, 2010 LAER). We based these estimates on an assumption that everyone in the modeled population (4 million people) is exposed to the maximum level of coke oven emissions allowed by the MACT standard rather than the actual emissions known to occur now, and that they were exposed to these emissions in one place of residence for 70 years. Such a scenario is very unlikely because individuals typically do not occupy the same residence for such a long period of time (e.g., the median residential occupancy period is approximately 9 years, and less than 0.1 percent of the population is estimated to occupy the same residence for greater than 70 years). Because EPA typically assumes that an individual's excess lifetime risk of cancer is directly proportional to their duration of exposure to the carcinogen(s) in question, reducing the duration of exposure for individuals in the modeled population would reduce the estimates of their risk. To illustrate this, we performed an addi

FOR FURTHER INFORMATION CONTACT Ms. Lula Melton, Emission Standards Division, Office of Air Quality Planning and Standards (C43902), Environmental Protection Agency, Research Triangle Park, NC 27711, telephone number (919) 5412910, fax number (919) 5413207, email address: melton.lula@epa.gov.