2
Trace Gaseous Emissions from Agent Incineration

HISTORY OF THE REGULATION OF TRACE EMISSIONS

Trace emissions were essentially unregulated until 1977 when polychlorinated dibenzo-p-dioxins and dibenzofurans (collectively known as "dioxins") were first detected in the stack gases of a municipal waste incinerator in Sweden (Aslander, 1987). This was quickly followed by a similar discovery in the United States (New York State Legislative Commission on Solid Waste Management, 1986). Shortly thereafter, a moratorium on the construction of incinerators was imposed in Sweden. The Swedish moratorium was lifted in 1986 with the promulgation of a standard equal to 0.1 ng/dnm3 at 11 percent O2 ITEQ1 dioxin. In the units commonly used by the Environmental Protection Agency (EPA), this is approximately 0.3 ng/dsm3 at 7 percent O2.

In the United States, the regulation of particulates from municipal and hazardous waste incinerators began in 1972 (40 CFR 60c).2 Hazardous-waste incinerators were also required to control emissions of hydrogen chloride and demonstrate at least 99.99 percent destruction and removal efficiency (DRE) of hazardous organic compounds (40 CFR 263). For wastes contaminated with polychlorinated biphenyls (PCBs) and dioxins, the DRE requirement was 99.9999 percent.

In 1991, regulations covering boilers and industrial furnaces that burn hazardous waste were adopted (the BIF Rule) (EPA, 1991). The BIF Rule included maximum allowable concentrations for a host of trace organic and inorganic compounds in ambient air. Because ground-level atmospheric concentrations are related to stack concentrations through site-specific dispersion modeling, the BIF Rule effectively created emissions standards for all regulated trace organic and inorganic emissions. However, they establish a common ambient impact for specified emissions constituents rather than a constant emitted concentration standard for each source, which results in different localized ground-level concentrations.3 This has led to considerable confusion and charges that facilities regulated under the BIF Rule may not be equipped with the best available control technologies, thus exposing some areas to more pollution than others.

In 1995, the Clean Air Act set standards for maximum achievable emissions control technology (MACT) for municipal waste combustors that incinerate more than 250 tons per day (40 CFR 60 Subpart Eb). These regulations restrict total dioxin emissions (i.e., the sum of the dioxin and furan homologues with 4 to 8 chlorine atoms per molecule) to 13 ng/dsm3 at

1  

 ITEQ (international toxic equivalency) dioxin is the amount of 2,3,7,8 TCDD (2,3,7,8-tetrachlorodibenzo-p-dioxin) with toxicity equivalent to the complex mixture of 210 dioxin and furan isomers with 4 to 8 chlorine atoms found in flue gases. This equivalency is based on the International Toxic Equivalence Factor scheme adopted by the Environmental Protection Agency and most countries to simplify the reporting of dioxin emissions.

2  

 CFR citations refer to the U.S. Code of Federal Regulations with the volume number preceding CFR and the section number following. Copies of volumes of the U.S. Code of Federal Regulations are available through the Government Printing Office outlets and commercial document and regulatory services.

3  

 Examples of constant emitted concentration standards include the EPA's New Source Performance Standards (NSPS) and Maximum Achievable Emissions Control Technology (MACT) standards.



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 11
Carbon Filtration for Reducing Emissions from Chemical Agent Incineration 2 Trace Gaseous Emissions from Agent Incineration HISTORY OF THE REGULATION OF TRACE EMISSIONS Trace emissions were essentially unregulated until 1977 when polychlorinated dibenzo-p-dioxins and dibenzofurans (collectively known as "dioxins") were first detected in the stack gases of a municipal waste incinerator in Sweden (Aslander, 1987). This was quickly followed by a similar discovery in the United States (New York State Legislative Commission on Solid Waste Management, 1986). Shortly thereafter, a moratorium on the construction of incinerators was imposed in Sweden. The Swedish moratorium was lifted in 1986 with the promulgation of a standard equal to 0.1 ng/dnm3 at 11 percent O2 ITEQ1 dioxin. In the units commonly used by the Environmental Protection Agency (EPA), this is approximately 0.3 ng/dsm3 at 7 percent O2. In the United States, the regulation of particulates from municipal and hazardous waste incinerators began in 1972 (40 CFR 60c).2 Hazardous-waste incinerators were also required to control emissions of hydrogen chloride and demonstrate at least 99.99 percent destruction and removal efficiency (DRE) of hazardous organic compounds (40 CFR 263). For wastes contaminated with polychlorinated biphenyls (PCBs) and dioxins, the DRE requirement was 99.9999 percent. In 1991, regulations covering boilers and industrial furnaces that burn hazardous waste were adopted (the BIF Rule) (EPA, 1991). The BIF Rule included maximum allowable concentrations for a host of trace organic and inorganic compounds in ambient air. Because ground-level atmospheric concentrations are related to stack concentrations through site-specific dispersion modeling, the BIF Rule effectively created emissions standards for all regulated trace organic and inorganic emissions. However, they establish a common ambient impact for specified emissions constituents rather than a constant emitted concentration standard for each source, which results in different localized ground-level concentrations.3 This has led to considerable confusion and charges that facilities regulated under the BIF Rule may not be equipped with the best available control technologies, thus exposing some areas to more pollution than others. In 1995, the Clean Air Act set standards for maximum achievable emissions control technology (MACT) for municipal waste combustors that incinerate more than 250 tons per day (40 CFR 60 Subpart Eb). These regulations restrict total dioxin emissions (i.e., the sum of the dioxin and furan homologues with 4 to 8 chlorine atoms per molecule) to 13 ng/dsm3 at 1    ITEQ (international toxic equivalency) dioxin is the amount of 2,3,7,8 TCDD (2,3,7,8-tetrachlorodibenzo-p-dioxin) with toxicity equivalent to the complex mixture of 210 dioxin and furan isomers with 4 to 8 chlorine atoms found in flue gases. This equivalency is based on the International Toxic Equivalence Factor scheme adopted by the Environmental Protection Agency and most countries to simplify the reporting of dioxin emissions. 2    CFR citations refer to the U.S. Code of Federal Regulations with the volume number preceding CFR and the section number following. Copies of volumes of the U.S. Code of Federal Regulations are available through the Government Printing Office outlets and commercial document and regulatory services. 3    Examples of constant emitted concentration standards include the EPA's New Source Performance Standards (NSPS) and Maximum Achievable Emissions Control Technology (MACT) standards.

OCR for page 11
Carbon Filtration for Reducing Emissions from Chemical Agent Incineration 7 percent O2 for new combustors. Using the EPA's rough equivalency factor (1 ng ITEQ dioxin equals 60 ng of total dioxins), this standard is 13/60 or 0.2 ng/dsm3 at 7 percent O2 for ITEQ dioxins (EPA, 1994). New municipal waste combustors that can incinerate more than 250 tons per day must also limit mercury emissions to 80 μg/dsm3, lead to 200 μg/dsm3, and cadmium to 20 μg/dsm3, all concentrations corrected to 7 percent O2. The EPA has scheduled its release of MACT standards for hazardous waste combustors for the third quarter of 1999. The April 19, 1996, proposal included emissions limitations for dioxins expressed as the equivalent amount of 2,3,7,8 TCDD using the ITEQ Factors (EPA, 1989, 1996). The EPA proposed an ITEQ dioxin emissions limitation of 0.2 ng/dsm3 at 7 percent 02 in the May 2, 1997, Notice of Data Availability (NODA) for this rule (EPA, 1997a). Total hydrocarbon and carbon monoxide limitations were proposed to act as surrogates for the other trace organics listed as hazardous air pollutants in Clean Air Act Section 112. Regulations were also proposed for limiting emissions of mercury, ''semivolatile'' metals (defined as the sum of cadmium and lead), and "low volatile" metals (defined as the sum of arsenic, beryllium, and chromium). The May 2, 1997, NODA proposed limiting new hazardous waste incinerators to 40 μg/dsm3 of mercury, 100 μg/dsm3 of semivolatile metals (lead and cadmium), and 55 μg/dsm 3 of low volatile metals (arsenic, beryllium, and chromium), all at 7 percent O2. Emissions limitations for some trace SOPCs are specified in site-specific construction and operating permits. Although environmental impact and HRA guidelines adopted by regulatory authorities are intended to limit emissions to levels that are unlikely to harm human health or the environment, public perception sometimes induces regulatory agencies to adopt even lower emission levels. MEASURING TRACE EMISSIONS All combustion systems necessarily emit some trace materials as a direct consequence of the laws of thermodynamics. Under less than ideal mixing conditions, trace organic emissions increase. Too much air in part of the gas stream reduces the temperature and slows the reaction rate; too little air produces reducing conditions that prevent complete oxidation of the fuel. In either case, these effects prevent portions of the gas from being burned to completion. Emitted SOPCs may include: some of the POHCs fed to the incinerator (e.g., agent or energetics) PICs (i.e., organic compounds formed during the combustion process itself) dioxins and other substances that form downstream of the combustion zone inherent equilibrium products at very low concentrations small quantities of noncombustible materials such as ash, metal oxides, or salts, that penetrate the air pollution control system Regardless of the source of SOPCs, properly designed, operated, and controlled combustion systems produce very low concentrations of these substances, typically well below the few parts per million level that can be reliably measured by (near real-time) continuous emissions monitors. SOPC concentrations, including concentrations of agent, are frequently below the detection limits of slower but more sensitive manual methods of monitoring. Near real-time monitors that produce an alarm if agent concentrations approach levels of concern have been developed, but even these monitors are not capable of quantifying the actual concentrations. Incinerator gas samples are drawn through a series of traps in which the targeted materials are selectively removed and concentrated: Volatile organics (i.e., compounds with boiling points below 125ºC [~255ºF]) are sampled using volatile organics sampling trains. Most of these organics are sorbed in activated carbon and Tenex® traps, but some are separated with water vapor in a condensate trap. After sampling, the traps are desorbed, cleaned to separate co-collected interferants, and the organics are analyzed using a combination of a high-resolution gas chromatograph followed by a mass spectrometer. The method and procedures are described in SW-846, Method 0030 (EPA, 1997b). Semivolatile organics (i.e., compounds with boiling points between 125 and 300ºC [~255 and 570ºF]) include dioxins. Semivolatile organics are concentrated by means of a Modified EPA Method 5 particulate sampling train that has a sorbent trap between the filter and liquid-filled impingers

OCR for page 11
Carbon Filtration for Reducing Emissions from Chemical Agent Incineration (scrubbers)(40 CFR 60, Appendix A Method 23; EPA, 1997b). In a procedure similar to the one used for volatile organic sampling, the probe, filters, and resin traps are recovered and extracted, and the extract is cleaned and analyzed using high-resolution gas chromatography coupled with high-resolution mass spectrometry to quantify the organics. The procedures are described in SW-846, Method 0010, 40 CFR 60, Appendix A, Method 23 (EPA, 1997b, 1997c) and 40 CFR 263, Appendix X, Method 23A. In Method 23, samples below the detection limit are treated as zeros for calculating total dioxins (the sum of the tetra-through octa-substituted dioxin and furan homologue totals). SW-846, Method 8290 specifies that estimated maximum possible concentration (EMPC) values (i.e., detection responses that do not meet all quality control criteria) be treated as zeros when calculating ITEQ dioxin concentrations (EPA, 1997b). Metals are sampled using a Method 5 train that has various liquids in the impingers following the filter. Solids deposited in the probe and caught on the filter are recovered and digested (dissolved) prior to analysis. Liquids in the impingers are also recovered and analyzed. The sampling method is described in 40 CFR 60 Appendix A, Method 29, and 40 CFR 263, Method 29. Metals are analyzed using a number of techniques described in SW-846 (EPA, 1997b). Mercury levels are usually quantified using a cold-vapor or graphite-furnace atomic absorption spectrophotometer. The balance of the metals of interest are generally analyzed using an ion-coupled plasma atomic emission spectrophotometer or mass spectrometer. EMISSIONS CONCENTRATIONS IN EXHAUST GAS FROM JACADS AND THE TOCDF Emissions have been tested at JACADS and the TOCDF during the incineration of agents (GB, HD, or VX), surrogate waste, and fossil fuel (oil or natural gas). All four types of incineration systems—liquid incinerator (LIC), deactivation furnace system (DFS), metal parts furnace (MPF), and dunnage incinerator (DUN)—have been tested. Table 2-1 shows the types and numbers of tests conducted through November 1998. Since 1988, all three chemical agents have been tested in the LIC at JACADS, but only agent GB has been tested in the LICs at the TOCDF, which began agent disposal operations in August 1996. In addition to testing for particulates and hydrogen chloride—the two SOPCs regulated by the Resource Conservation and Recovery Act (RCRA) regulations for hazardous waste incinerators (40 CFR 263)—emissions at JACADS and the TOCDF have also been analyzed for the following substances: other halogen-containing (C1 and F) gaseous species 22 elements, including all 11 elemental hazardous air pollutants covered in the Clean Air Act 204 trace organics, including the three agents being destroyed, 54 organic compounds classified as hazardous air pollutants by the Clean Air Act, and 147 other organic chemicals light and total nonvolatile hydrocarbons (i.e., hydrocarbons with boiling points lower than 100ºC (212ºF) and higher than 300ºC (572ºF), respectively) TABLE 2-1 Emissions Tests at the Two Operational Baseline Incineration Facilities, JACADS and the TOCDF   JACADS TOCDF Type of Waste LIC MPF DUN LIC DFS MPF GB 2   1 2 4 2 HD 2 2     1   VX 2       1   Agent surrogate       1     Fossil fuel         1  

OCR for page 11
Carbon Filtration for Reducing Emissions from Chemical Agent Incineration indicators of good combustion, such as acceptable levels of carbon monoxide and total hydrocarbons Much of this testing has been done under trial-burn conditions designed to show emissions under simulated worst-case operating conditions. Organic emissions are maximized in some tests by reducing combustion temperature. Metals are maximized in others by spiking excess metals into the feed, maximizing combustion temperatures to volatilize as much metal as possible, and even operating the air pollution control equipment at below optimal levels. The concentrations of emittants in the stack gas listed in Appendix B of this report for JACADS and TOCDF incinerators are among the lowest reported for all hazardous waste incinerators in the database of hazardous waste combustion emissions maintained by the EPA's Office of Solid Waste and Emergency Response (EPA, 1997d). The EPA's graphical summaries consistently show that TOCDF results either set the lower bound or are among the lowest in emissions of dioxins, mercury, semivolatile metals (cadmium and lead), and low volatile metals (arsenic, beryllium, and chromium). Appendices IV and V of the BIF Rule (40 CFR 266) provide ambient concentration limits for risk of 1 in 100,000 (i.e., 10-5) of an adverse health effect caused by breathing these ambient concentrations for 70 years (EPA, 1991). These concentration limits are called reference air concentrations (RACs) for noncarcinogenic materials and risk-specific doses (RsDs) for carcinogenic materials. The relationship between ambient and stack concentrations is determined by a facility's design, operating profile, and site-specific dispersion characteristics. For example, based on the dispersion modeling results reported in the TOCDF HRA (Utah DSHW, 1996), an MPF stack concentration of 7,850,000 ng/dsm3 at 7 percent O2 produces a 100 ng/m3 concentration at the point of maximum ground level impact. Similar values for the LIC and DFS are 36,000,000 and 22,000,000 ng/dsm3, respectively. Comparing either the average of the detected concentrations or the lowest detection limit for all test results below detection limits, the acid gases equal 3 percent of their RACs. All measured metals are present at less than 1 percent of their RACs and less than 0.15 percent of their RsDs. The maximum ambient contribution of any trace organic is less than 4 percent of its RAC or RsD, as appropriate, with a median ambient contribution of 0.03 percent. The specific contribution for ITEQ dioxins is 0.025 percent of the RsD. The RAC and RsD do not fully quantify risk because they are based only on inhalation, and many chemicals can translocate between media and bio-accumulate in the food chain. Quantifying the significance of these emitted concentrations would require an updated, multipathway HRA, which would include various paths of exposure. The emission rates used in the TOCDF HRA were generally higher than the rates indicated by the data now available. A few SOPCs are emitted at rates higher than the estimated rates in the HRA; however, they contribute less to the health risk than many other SOPCs, including dioxins, that bioaccumulate and translocate and whose actual emission rates are lower than the estimated rates in the HRA. Consequently, it is logical to conclude that the actual TOCDF risk is even lower than the HRA estimate, which was below the level of regulatory concern. Emissions Sampling And Analysis Methodology The emissions sampling and analysis at JACADS and the TOCDF were performed following the standard sampling methodologies found in EPA documents SW-846 and 40 CFR 60, Appendix A (EPA, 1997b, 1997c). Because analytic laboratory procedures have been evolving and detection limits have improved since the original JACADS testing, detection limits for early tests were higher than they would be today—sometimes 10 to 1,000 times higher. Moreover, in a typical analysis of flue gas, emittants are captured in several different portions of the sampling train, which are separately recovered and prepared for analysis. The extracts and digestates can frequently be combined before analysis, however, to reduce the combined detection limit for the replicate test run to the lowest practical level. The reported concentration is the sum of the results of individual analyses. The reported concentration for results below detection limits is the sum of all of the masses, assuming concentrations equal to the detection limit for each separately analyzed component of the sampling train. This practice differs from standard stack-testing practices where individual nondetects are treated as zeroes except that the largest nondetect is used to characterize test replicates with no detected values. Adding

OCR for page 11
Carbon Filtration for Reducing Emissions from Chemical Agent Incineration the detection limits together results in an overall detection limit substantially higher than would have been achieved by combining them before analyzing all components of the sampling train. The value inflation of the detection limit for the overall replicate is a function of the number of sampling train components analyzed separately and the detection limit of each. For example, when the separate components of a volatile organic compound sampling train are analyzed, seven pairs of sorbent tubes and two condensate samples—16 individual samples—all produce results. For typical laboratory detection limits (10 ng for each of the 14 sorbent tubes and 80 ng for each condensate trap), the reported detection limit concentration is 300 ng or [(14)(10) + (2)(80)]. If, however, the individual sorbent tubes are desorbed, concentrated, and combined with the condensate before analysis, the detection limit becomes 10 ng, which is one-thirtieth of the value typically reported for volatile organic compounds at JACADS and the TOCDF. Footnotes in some test reports (see, for example, Tables 5-9 and 5-19 in EG&G, 1997) state that practical quantitation limits (PQLs) were reported when results were below detection limits. When the concentration of a sample with three to five times the estimated detection limit is measured repeatedly, the replicates show some scatter, usually characterized by a bell-shaped, Gaussian distribution. When the variance is constant, the standard deviation of this distribution (S0) is used to define the detection limit as three times So, and the PQL is defined as 10 times S0 (EPA, 1997b). Based on these definitions, the PQL is 3.3 times the detection limit. For measurements at the detection limit, the analyst can be confident that the analyte is present but cannot confirm the amount. At or above the PQL, however, the analyst can be confident about the quantitation. Consequently, by reporting the PQL for results below detection limits, the maximum amount of SOPCs that might have been in the sample is overstated by at least a factor of 3. When multiple tests are performed, there is no statistically meaningful chance that all of the actual concentrations are at the detection limit without half of them being detected. Because multiple tests are performed for all SOPCs and because many sets are all below the detection limit, the best scientific estimate of the average concentration is half the detection limit, about 15 percent of the reported PQL (Hass and Sheff, 1990). Clearly, the projections of the harmful effects and benefits of additional technological controls will necessarily be overstated for undetected SOPCs when their detection limits are used in an HRA. The vast majority of the emissions data for JACADS and the TOCDF show a few emissions with detected values and many below the detection limit. These data sets are "left-censored" (i.e., only large values are quantified, and for the values below the detection limit, the concentration is assumed to be somewhere between zero and the detection limit). The average of the detected values is larger than the true average because a number of unknown, but by definition smaller, values are excluded from the calculations. The committee estimated the amount of positive bias for this type of data analysis by computing the ratio of the average of the detected values to estimates of maximum likelihood of the average (Cohen, 1959). The median positive bias introduced by using the average of the detected values to represent the emitted concentration or rate is 175 percent. The positive bias in JACADS and TOCDF analyses is actually higher because the highest measured value, rather than the statistically derived confidence limit,4 was used as a bounding value for the average. Based on the reported data characteristics at JACADS and the TOCDF, the averages are either representative (i.e., all detected values) or very conservative (117 to 660 times the most likely value, depending on whether left-censoring or the chemical analysis methodology controls the result). Characteristics Of Exhaust Gas Emissions For JACADS and the TOCDF, the vast majority of emittants that were analyzed are below the analytic detection limit. The committee prepared box-plots of the detected concentrations showing mean, central 50 percent, and extreme values. Based on these box-plots, the characteristics of SOPC emissions at both sites for all agents and incinerator types were the same, even though JACADS uses jet fuel and the TOCDF uses natural gas as supplemental fuel. The oxygen content of the flue gas was lower and the moisture and temperature of the exhaust gas higher at the TOCDF, 4    The highest measured value is the nonparametric estimate of the likely mean if there are nine samples. For the 35 sample data set on dioxins now available, the upper confidence interval is the 24 th highest value. Instead of a 95 percent confidence interval, the highest dioxin measurement is at the 99.99999994 percent confidence level based on the nonparametric estimating equations in Hahn and Meeker (1991).

OCR for page 11
Carbon Filtration for Reducing Emissions from Chemical Agent Incineration but all other detected concentrations were similar. However, there were too few detected concentrations to verify this observation statistically. Despite emissions tests that exceed the minimum sample volume requirements given in the federal regulations (40 CFR 60, Appendix A and 40 CFR 266, Appendices IX through XIV), the data do not show detectable concentrations. Because the true concentrations are not detectable, the results are a qualitative indication of low emitted concentrations. Upper bounds for the SOPC emission rates can be developed by combining the available information from all of the incinerators at JACADS and the TOCDF. That is, although the units and agent feeds are different, there is no reason to believe from the available information that emitted concentrations of individual chemical compounds are different. Or, stated another way, regardless of the agent being fed (GB, VX, or HD) or the incinerator being used (LIC, MPF, or DFS), the emitted concentrations of nonagent chemical compounds can reasonably be assumed to be similar. This result is in large measure attributable to the high DRE achieved and the effectiveness of the existing air pollution control systems. Appendix B lists the tested emittants; the percentage of tests in which each emittant was found; the minimum, mean, and maximum of the detected concentrations; and the detection limit for undetected concentrations. Concentrations of gaseous emittants are expressed in conventional regulatory units, as well as parts per billion (dry volume) (ppbdv) to facilitate engineering calculations. Carbon absorption systems remove trace organics in proportion to their partial pressure (see Chapter 3), which can be easily determined by multiplying ppb dv, by local barometric pressure. For engineering purposes, ppbdv is more useful, but the standardized regulatory (mass per unit volume) concentrations are necessary for comparing emitted concentrations reported at other sources. The emission rates actually used for risk assessments of SOPCs are usually deliberately high to yield a conservative assessment. Consequently, the concentrations that correspond to these emission rates are higher than the concentration data summarized in Appendix B.5 As discussed in Chapters 3 and 4 of this report, activated carbon adsorbs different amounts of individual organic vapors based on their chemical structures, their vapor pressures, and the operating temperature and pressure of the sorbent bed. The amount of any organic chemical in the bed at any time is determined by the amount introduced and the fraction adsorbed minus the amount that has degraded while adsorbed. Because the amount introduced is determined by the inlet concentration and flow rate, bounding concentration estimates are necessary to determine the bed life, benefits, and risks of activated carbon. These estimates can be derived from the existing emissions test data, including the observed below detection limit concentrations from JACADS and the TOCDF. EMISSION RATES The JACADS and TOCDF incineration systems have been extensively tested, and the results have consistently shown that emissions of materials regulated under Sections 111 (as criteria pollutants) and 112 (as hazardous air pollutants) of the Clean Air Act and noncriteria (trace) emittants subject to regulation are either the lowest or among the lowest in the EPA's Hazardous Waste Combustor Emissions Database (EPA, 1997a). Although emissions test results can only provide an estimate of the true mean and standard deviation for any emittant, the statistical characteristics of distributions can be used to establish bounding values likely to contain the true population parameter. For example: • Long-term, multiyear exposures are characterized by the average emission rate. The 95 percent statistical confidence level, the upper confidence limit (UCL), for the mean is the bounding value: where is the arithmetic average, is the t-statistic, N is the number of replicates used to estimate the average, a is the statistical significance level (e.g., a is 0.05 for the 95 percent statistical confidence level), and S is the data standard deviation (Hahn, 1970). • Short-duration events caused by normal fluctuations in emissions are characterized by the upper tolerance limit (UTL) designed to cover 99 percent of future events. The UTL is: 5    The TOCDF HRA and QRA used the highest concentrations measured at JACADS through mid-1994 (Utah DSHW, 1996; U.S. Army, 1996a). Because these values are above actual experience, they also have been used in subsequent evaluations to minimize confusion.

OCR for page 11
Carbon Filtration for Reducing Emissions from Chemical Agent Incineration where φ-1( ) is the inverse normal distribution value, P is the fraction included, and χ2N-1,1-α is the chi squared statistic (Hahn, 1970). Fifteen-minute events have different maximum likely concentrations than 3, 8, or 24-hour events because the sampling distribution of the mean is different for each averaging time (Taylor, 1990). To make this correction, the square root of T/D has been added where D is the event duration, and T is the sampling time associated with the replicates used to estimate the average and standard deviation. • Short-duration events caused by equipment failures and random high concentrations caused by, for example, combustion instability are very unlikely to occur simultaneously. Therefore, short-duration events are characterized by the average inlet concentration to the pollution control train. This average inlet concentration should be combined with the reduced control efficiency associated with the failure mode. • Compliance test limits are described by the upper prediction limit (UPL)6 for the number of times a facility will be tested. The UPL is: where k is the number of future tests to be performed, and q is the number of replicates averaged to determine a test result (Hahn, 1970). Prediction limits are bounds that are unlikely to be exceeded by the next specified number of tests. The number of tests that should be included in the prediction limit calculation is a risk management decision. Although the number of tests has no effect on average emission rates, it produces higher and higher emissions limitations as the frequency of finding a compliant operation in violation of the standards is reduced. For a facility like the TOCDF, each incinerator may be tested annually over its seven and one-half year operating life. If the limit is for one unit, q is 7, but if no exceedance at any of the four operating TOCDF incinerators is desired, q is 28. When the number of baseline incinerator facilities in the United States is considered, q increases accordingly. For a probabilistic HRA, Bayesian statistics can be used as an alternative treatment of the nondetects. This would eliminate the need to make assumptions about the nature of the distribution describing the likely concentrations of undetected species. Bayesian statistics could be used in refined multipathway HRAs because they provide information on the distribution of outcomes, rather than a single point estimate intended to be higher than the real risk. Table 2-2 shows the relationship between the arithmetic average and the bounds discussed above for different coefficients of variance (i.e., the standard deviation to average ratio). The data from JACADS and the TOCDF indicate that the coefficients of variance for emittants of interest range between 0.25 and 2 for most of the measured emissions. When the coefficient of variance is larger than 1, the distribution is probably not normal, and advanced statistical techniques (e.g., normalizing transformations) are required to provide meaningful estimates. Detailing these techniques is beyond the scope of this report. Maximum likelihood estimates of the arithmetic average and standard deviation are used when some measurements are below detection limits (Cohen, 1959). When all the measurements are nondetects, the best point estimate for the average is half the detection limit. However, there is no known estimate for the standard deviation, and there is no assurance that the detection limit is not several orders of magnitude above the true emittant concentration. Consequently, it is impossible to establish meaningful emissions limits statistically. Statistical practice is to use half the detection limit as the annual average and the detection limit for all other averaging times.7 If the coefficient of variance is available for another analyte believed to behave like the undetected emittant, then half the detection limit for that analyte can be scaled using an appropriate value from Table 2-2 or calculated using the preceding equations. For chemical agents (which have never been detected in any emissions test at a baseline system 6    The UPL is a bound below which an achievable emissions limitation cannot reside. The emissions limitation to avoid false exceedances when a facility is operating exactly as it was when the limit was set may be considerably higher. 7    EPA Region III has published similar guidelines for the treatment of nondetects in stack tests for risk assessments used to establish permit limits.

OCR for page 11
Carbon Filtration for Reducing Emissions from Chemical Agent Incineration TABLE 2-2 Effect of Data Characteristics on Emissions Characteristics for Different Averaging Timesa Emissions Multiplier—Average to Likely Upper Bound Coefficient of Variance 25% 50% 75% 100% 150% 200% 6 runs in database             15 minute (UTL-0.25) 5.12 9.24 13.37 17.49 25.73 33.98 1 hour (UTL-1) 3.06 5.12 7.18 9.24 13.37 17.49 3 hour (UTL-3) 2.19 3.38 4.57 5.76 8.14 10.52 8 hour (UTL-8) 1.73 2.46 3.19 3.92 5.37 6.83 24 hour (UTL-24) 1.42 1.84 2.26 2.68 3.52 4.37 annual (UCL) 1.32 1.65 1.97 2.29 2.94 3.58 3-run average (UPL-3) 1.53 2.05 2.58 3.11 4.16 5.22 18 runs in database             15 minute (UTL-0.25) 3.62 6.24 8.86 11.48 16.73 21.97 1 hour (UTL-1) 2.31 3.62 4.93 6.24 8.86 11.48 3 hour (UTL-3) 1.76 2.51 3.27 4.03 5.54 7.05 8 hour (UTL-8) 1.46 1.93 2.39 2.85 3.78 4.71 24 hour (UTL-24) 1.27 1.54 1.80 2.07 2.61 3.14 annual (UCL) 1.14 1.29 1.43 1.58 1.87 2.16 3-run average (UPL-3) 1.30 1.61 1.91 2.22 2.82 3.43 200 runs in database             15 minute (UTL-0.25) 2.99 4.98 6.97 8.96 12.95 16.93 1 hour (UTL-1) 2.00 2.99 3.99 4.98 6.97 8.96 3 hour (UTL-3) 1.57 2.15 2.72 3.30 4.45 5.60 8 hour (UTL-8) 1.35 1.70 2.06 2.41 3.11 3.82 24 hour (UTL-24) 1.20 1.41 1.61 1.81 2.22 2.63 annual (UCL) 1.04 1.08 1.12 1.16 1.24 1.32 3-run average (UPL-3) 1.25 1.50 1.74 1.99 2.49 2.98 a The upper bound will exceed the average by the factors shown. incinerator), the average and short-term peak concentrations derived from the detection limit in ng/dsm3 at 7 percent O2 are: • GB 1.8 and 3.6 • HD 115 and 230 • VX8 1.8 and 3.6 For dioxins and furans, expressed as ITEQs, the average loading to a carbon bed filter is 0.01 ng/dsm3 at 7 percent 02. The 15-minute short-term variability induced value is 0.41 ng/dsm3 at 7 percent 02. SUMMARY Trial bums have been performed at JACADS and the TOCDF to test the incinerators at each site, as well as the combustion of the various agents (although not all agents were tested with all incinerators. The reported emission concentrations are among the lowest for all hazardous waste incinerators in the EPA's Hazardous Waste Combustor Emissions Database. Data for most of the SOPCs consisted of a few measurements at a low concentration level, with many more below the detection level. Analyses of human health risks (see Chapter 5) were based on the highest values recorded during trial bums at JACADS for each measured SOPC. A statistical evaluation of the data, 8    Not measured but assumed to be equal to GB.

OCR for page 11
Carbon Filtration for Reducing Emissions from Chemical Agent Incineration including allowances for variances in measurement, indicates that this approach is extremely conservative. Based on the most recently developed analytical techniques, detection limits and nondetectable concentrations are now lower than when the JACADS trials were run. For example, mustard (HD) was not observed in the exhaust gases of any trial burn. Yet the concentration used in the HRA analyses for Tooele, Anniston, and Umatilla, based on the JACADS test data, was 8,700 ng/dsm3. This can be compared to the average value derived from the trial burn detection limit of 115 ng/dsm3.