Tooele Chemical Agent Disposal Facility: Update on National Research Council Recommendations (1999)

In the areas of systems performance and plant operations, the Stockpile Committee recommended that the following conditions be satisfied:

• mandatory Army Preoperational Survey requirements prior to the start of agent operations [S-8]

• all Resource Conservation and Recovery Act (RCRA) and Toxic Substance Control Act (TSCA) trial-burn requirements for the LICs and DFS [S-9, S-11]

• testing and certification of the BRA and DUN or implementation of a satisfactory alternative [S-12]

• demonstration of the slag-removal system for the LICs [S-13]

• active pursuit of continual improvements in monitoring systems [S-17]

• continued evaluation of the proposed addition of a carbon-bed filter to the PAS [S-18] (the subject of a separate NRC report, Carbon Filtration for Reducing Emissions from Chemical Agent Incineration [NRC, 1999])

The TOCDF began agent operations on August 22, 1996. As of May 19, 1999, 20,001 GB M55 rockets, 2,710 GB ton containers, 137,754 GB 105 mm projectiles, and 4,463 GB MC-1 bombs had been destroyed. The destruction schedule for M55 rockets had fallen behind the original timetable because of a delay in obtaining the TSCA permit; and more projectiles and fewer ton containers had been processed than was projected by the TOCDF QRA schedule. Approximately 2,751 tons of GB have been destroyed, more than 20 percent of the total DCD stockpile.

Every year the Army submits a report to Congress on the CSDP that includes a description of "other events" and a summary of significant events that resulted in plant shutdowns, of which there have been two each year. The most recent shutdown, which occurred on December 13, 1998, was caused by improper reassembly of an in-line filter after maintenance that resulted in 140 gallons of GB leaking into the toxic cubicle sump. Although all agent was contained by the safeguards built into the facility, this significant maintenance error suggests that there are problems in training and the implementation of a safety culture throughout the organization. This event also suggests insufficient communication between control room operations and maintenance personnel. None of the events resulted in exposure of personnel to chemical agent or its release to the environment.

RCRA trial burns have been satisfactorily completed with GB for LIC-1 and LIC-2, the MPF, and the DFS. The TSCA trial burn for the DFS had to be redone, however, which delayed the processing of M55 rockets. The second TSCA trial burn was successful.

The BRA did not pass its initial compliance test because of excessive particulate emissions, but the probable cause of the problem was identified. However, because economics favor the off-site disposal of brine, the Army has decided not to retest the BRA at this time. This has raised concerns on the committee about what would happen if the off-site shipping of brine becomes unavailable. TOCDF site managers have discussed alternatives to the off-site disposal of brine, and the BRA is presently in a long-term lay-up configuration, which means the equipment will be protected while it is inactive. Approximately four weeks would be necessary for the equipment to be made operational. The state of Utah has verbally agreed that, in the event of a change to requirements for brine management, it would allow the Army time to effect the transition. This could include authorizing the temporary storage of brines in isolation containers (as is done at JACADS) until the equipment in the BRA can be brought on line and tested to demonstrate compliance with regulatory requirements.

The DUN at the TOCDF has not been used because contaminated wastes that were scheduled for destruction in the DUN are being disposed of at qualified hazardous-waste management facilities. Although off-site disposal was always an option, the DUN was originally designed as part of the overall waste-minimization program required by the Environmental Protection Agency (EPA) and endorsed by the committee. The major contaminated waste stream scheduled for destruction in the DUN is the activated carbon from the facility's ventilation system. As an alternative, the Army is studying the installation of a micronizer and burner designed to dispose of activated carbon in the DFS. A prototype unit will be tested at JACADS during the closure phase of that facility (calendar year 2001).

Modifications to improve the LIC slag-removal system have been successful. As of December 1998, slag had been tapped approximately 45 times, almost all from LIC-1, which has an improved slag-removal system. During a recent maintenance shutdown, the slag-removal system for LIC-2 was also upgraded, but LIC-2 has not been operated long enough since then to demonstrate the performance of the upgraded system. To date, a total of approximately 22,000 lbs of slag has been drained from both incinerators; this has avoided approximately three maintenance shutdowns that would have been necessary to remove slag manually. A recurrent problem in the slag-removal system has been the failure of the heater, and the Army is evaluating ways to extend heater life.

Because risk to the public is directly related to the existence of the stockpile, its rate of destruction is of key concern to the Stockpile Committee. The faster the stockpile can be safely destroyed, the lower the overall risk to the public becomes, and the Army has organized the disposal schedule to maximize risk reduction. The first campaigns, therefore, were focused on the disposal of M55 GB rockets, with co-processing of GB ton containers. At the start of agent processing, the expected value of the public acute fatality risk as calculated in the QRA was 1.4 x 10^-3 per year.¹

According to the schedule issued at the start of agent destruction operations, all GB M55 rockets were to have been processed within the first nine months of operation. In actuality, after about one-third of the rockets (11,592 units) had been processed, rocket processing was stopped because some of the exhaust gas samples collected during the first TSCA trial burn contained a specific polychlorinated biphenyl (PCB) cogener that later proved to be a random sampling or analysis artifact. Thus, results of the first PCB destruction and removal efficiency test were ambiguous, and the TSCA permit for processing M55 rockets at the full rate was delayed pending a successful retest. The recovery efficiencies of surrogate spikes during the TSCA trial burns were low, which was probably due to the severe weather conditions during testing in January 1997. (Severe weather can affect the sampling procedures.) When the trial burns were repeated in November 1998, the results met regulatory requirements, and the processing of M55 rockets was resumed. In the interim, ton containers were processed, and GB MC-1 bombs and 105 mm projectiles were moved up in the schedule to make the most effective use of the facility.

At the end of calendar year 1998 (after 28 months of agent operations), the TOCDF had processed 71,771 items (rockets, bombs, projectiles, and ton containers) containing approximately 2,495 tons of agent. The public acute fatality risk calculated in the QRA for the condition at the end of 1998 was 2.5 x 10^-4 expected fatalities per year. According to the operations schedule in the QRA, by this time 47,162 items were to have been processed containing approximately 4,004 tons of agent. In percentage terms, 52 percent more items had been processed by the end of calendar year 1998, but 37 percent less agent had been destroyed than originally scheduled. The difference reflects that more projectiles and fewer ton containers have actually been processed than were projected in the QRA schedule.

Thus, the TOCDF is ahead of the original QRA schedule in the number of items processed but behind in the tonnage of agent destroyed. The changes in the order of agent disposal operations have reduced the

overall risk and enabled efficient utilization of the facility, which is processing three munitions (GB-filled rockets, ton containers, and projectiles) at the same time. Because of the delay, the stacking height of stored VX rockets was lowered to reduce the storage risk. The current schedule allows for a constant rate of agent processing during the overall GB campaign, but the delay in processing GB-filled M55 rockets has slowed the rate of risk reduction. At the completion of the GB processing campaign (third quarter of calendar year 2001), the TOCDF is now projected to have destroyed 929,865 items containing approximately 6,097 tons of agent. At a similar point in the original schedule, the TOCDF was projected to have destroyed a total of 942,561 items containing approximately 6,683 tons of agent, including some non-GB agent.

At the start of agent processing, the public acute-fatality risk calculated in the QRA for accidental agent release was 1.4 x 10^-3 per year. This was based on five phases of disposal: (1) disposal of GB rockets and ton containers; (2) disposal of VX rockets and spray tanks; (3) processing of remaining GB items; (4) processing of remaining VX items; and (5) disposal of HD. Because of the delay in the processing of GB rockets, the Army decided to complete disposal of all other GB items first, followed by all VX items. Thus, the public acute-fatality risk at the end of 1998 was 2.5 × 10^-4 per year (18 percent of the original rate at the start of operations). This risk is based on the disposal of the GB munitions and ton containers and the reconfiguration (by reducing the stacking height and banding rockets together) of the stored VX rockets. At the same time in the original QRA schedule, the calculated public acute-fatality risk was to have been 7.0 × 10^-5 per year, or 5 percent of the original risk at the start of operations, based on the assumption that all GB rockets, VX rockets, spray tanks, MC1 bombs, weteye bombs, and a little more than half of the GB ton containers had been processed.

The TOCDF destruction program was behind schedule by approximately one month (33 days) as of the end of calendar year 1998. Given the recent regulatory approvals for the operation of both of the LICs and the DFS at the full rate and the successful completion of the TSCA trial burn for the DFS, the committee believes that the current schedule delay can be made up. The processing of GB rockets is expected to resume after the disposal of the M360 projectiles (which are processed in the MPF) has been completed in the third quarter of calendar year 2001. The remaining GB ton containers and munitions can be coprocessed during this same time, and GB rockets are being processed, as the system allows. Their disposal is expected to be completed in calendar year 1999. GB ton containers are processed whenever there is enough capacity in the LICs. This overall strategy is the shortest pathway through the TOCDF operations schedule that is consistent with the principle of processing the items with the highest storage risk as soon as practical.

At its meeting in September 1998, the committee was informed that the recent program-wide audit performed by the Arthur Anderson Company indicated that the present schedule and budget estimates were probably optimistic (Evans, 1998a; Arthur Anderson, 1998). Although safety is the committee's highest priority, the prompt destruction of the stockpile is the primary factor in risk reduction. A strong commitment program-wide and by site management to meeting schedules without compromising operational safety is essential to meeting the overall goal of safe and expeditious destruction of the stockpile.

Trial burns are conducted to demonstrate that incinerator systems perform as designed and meet applicable state and federal regulations and permit restrictions. The specific purpose of a trial burn is to demonstrate permissible emissions while processing at maximum allowable chemical agent feed rates under projected worst-case operating conditions for both the combustion chamber(s) and the air-pollution control equipment. The demonstrated worst-case operating conditions then become the operating limits in the operating permit. The facility operator is allowed to operate the incinerators at conditions equal to or better than the worst-case conditions. Hence, normal incinerator performance should always be as good or better than the performance demonstrated during the trial burn.

The TOCDF's LICs, DFS, and MPF were first tested using agent surrogates (i.e., chemicals that behave similarly to agents in incinerators but are not nearly as toxic at the same concentration). Once the surrogate trial burns demonstrated that the incinerators met the Army's performance standards, chemical agent trial burns were conducted to satisfy RCRA and TSCA requirements.

The sections that follow summarize the results of the surrogate and agent trial burns, discuss the implications of the agent trial-burn data for the HRA, and describe the problem with the TSCA trial-burn data that delayed

the processing of M55 rockets. If compounds of concern were present in concentrations below the detection limits, the practical quantification limits (PQLs) were reported for most tests.² Consequently, the maximum amount of a compound of concern that might have been present is overstated by a factor of at least 3.3.

The TOCDF DFS, MPF, and one of two identical LICs were tested using agent surrogates. The DUN was not tested because DUN operations are no longer planned. The purpose of a surrogate trial burn is to demonstrate that an incinerator system (combustor plus air-pollution control system) can efficiently destroy and remove typically hard-to-burn compounds. The Army set a target destruction and removal efficiency (DRE) of 99.9999 percent, which is more stringent than the federal DRE requirement for all substances that do not contain polychlorinated dibenzo-p-dioxins and dibenzofurans (PCDD/F). Surrogates were selected to meet the Utah Division of Solid and Hazardous Waste criteria. The surrogate trial burn for LIC-1 was successfully conducted between June 30 and July 7, 1995 (the results are summarized in NRC, 1996a). The results of the other three surrogate trial burns are summarized below.

The TOCDF operates under RCRA permit UT5210090002 issued by the state of Utah. Under the requirements of this permit, the incinerator systems must demonstrate that they meet performance standards that ensure effective and safe destruction of chemical agents before beginning routine operations. The primary objective of the surrogate trial burns was to demonstrate that the incinerators meet the following performance criteria:

• DRE of at least 99.9999 percent for the surrogates, also known as principal organic hazardous constituents

• emissions of total particulate matter lower than the federal requirement of 180 milligrams per dry standard cubic meter (mg/dscm), which is equivalent to 0.08 grains per dry standard cubic foot (gr/dscf) at 7 percent oxygen (O₂); and the state requirement of 0.016 gr/dscf at 7 percent O₂ for particulate matter smaller than 10 microns³ (PM ₁₀)

• hydrogen chloride (HCI) emissions, measured downstream of the pollution control equipment, less than four pounds per hour (lbs/hr) or less than 1 percent of the total organically-bound chlorine input to the furnace (i.e., chlorine in the surrogate, not salts that might contaminate the fuels)

• minimal emissions of products of incomplete combustion evidenced by 60-minute moving average carbon monoxide (CO) concentrations of less than 100 parts per million (ppm) on a dry, volumetric basis corrected to 7 percent O₂

LIC-2 surrogate trial burns were conducted on January 29 and 30, 1996. The surrogates selected to simulate the chemical agents were 1,2,4-trichlorobenzene and tetrachloroethylene (also known as perchloroethylene), which contain a lot of organically bound chlorine to challenge the PAS and have chemical bonds similar to those in the agents. The results should be reasonably representative of chemical agent operations.

Table 2-1 is a summary of the particulate matter, HCI, and CO emissions and DREs for the LIC-2 surrogate trial burns. Total particulate emissions were significantly lower than the PM₁₀ requirement, showing that the fraction of emissions of sub-10 micron particulates was lower than the requirement. A greater than 99.9999 percent DRE was

demonstrated. Limitations for particulate matter, HCI, and CO emissions were met during the test.

MPF surrogate trial burns were conducted on June 4, 5, and 6, 1996. The surrogates selected to simulate the chemical agents were a combination of monochlorobenzene and hexachloroethane. This combination was recommended by the Utah DSHW as one that would be more difficult to destroy than the chemical agents and would provide a maximum challenge to the PAS.

Six of the first seven runs were invalidated because of sampling and analytical problems, such as the inadvertent use of an incorrectly spiked resin or a sampling system leak. Another run, Run 6, was aborted because of operating difficulties with the MPF. Because the sampling problems are not associated with the ability of the incinerator to meet performance standards, and because the operating difficulty during Run 6 involved ancillary equipment that was not likely to affect emissions, the Utah DSHW, with the guidance of the EPA, agreed that additional performance runs could be conducted. The next few runs, Runs 8 through 10, were completed without incident.

Table 2-2 summarizes the particulate matter, HCI, and CO emissions and DREs for the MPF surrogate trial burns. The particle-size distribution was not measured so no information is available on the amount of PM₁₀ actually emitted, but compliance with the PM₁₀ standard (see performance criteria given earlier) was demonstrated because the total particulate emissions were less than the PM₁₀ performance standard. The Army's 99.9999 percent DRE requirement was also demonstrated. Hence, the MPF surrogate trial burns demonstrated that the system could safely proceed to the second phase of the RCRA demonstration and testing requirements—the chemical agent trial burn (ATB).

The DFS surrogate trial burns were conducted between September 30, 1995, and October 6, 1995. The tests included one run using only supplementary fuel and five performance runs with surrogates. The surrogate compounds selected by the Utah DSHW were monochlorobenzene and hexachloroethane. An error in sample recovery voided run 1. Run 2 was not completed because of a mechanical failure in a feed chute that interrupted incinerator operations. Incinerator performance was assessed using runs 3, 4, and 5.

Table 2-3 summarizes the particulate matter, HCI, and CO emissions and DREs for the DFS surrogate trial burns. Although particulate size was not measured, total

particulate emissions were less than the PM₁₀ emissions standard. Therefore, the fraction of emissions smaller than 10 microns (10^-6 m) meets the requirement. The Army's 99.9999 percent DRE requirement was also demonstrated. Hence, the DFS surrogate trial burns demonstrated that the system could proceed to the second phase of the RCRA demonstration and testing requirements—the ATBs.

The agent trial burns (ATBs) at the TOCDF site demonstrated that the incineration systems meet emissions requirements when burning chemical munitions. The ATBs are conducted (1) to demonstrate a DRE requirement for agent in accordance with the state of Utah permit, the Code of Federal Regulations (Title 40 Part 264), and RCRA regulations, and (2) to demonstrate system performance and the control of emissions. The results of the ATBs conducted to date for LIC-1, LIC-2, the DFS, and the MPF using agent GB are summarized below.⁴ The following performance standards were characterized:

• DRE for the incinerator using agent GB as the principal organic hazardous constituent for fulfillment of RCRA requirements (i.e., 99.99 percent)

• compliance with the particulate-matter emission-rate limits in both the RCRA permit UT5210090002 and the Approval Order issued by the state of Utah

• compliance with the HCI emission-rate limits in the RCRA permit

• emission rates for phosphorus and the 20 metals estimated by the state of Utah for the screening HRA conducted by the Department of Environmental Quality DSHW (Utah DSHW, 1996)

• emissions of PCDD/F

• emissions of certain semivolatile organic compounds (SVOCs) and volatile organic compounds (VOCs)

• exhaust gas concentrations of O₂ and CO using the TOCDF continuous emission-monitoring systems (CEMS) to document one aspect of combustion conditions in the system and show compliance with the CO concentration limits in the RCRA permit

LIC-1 ATBs were conducted on February 26, 27, and 28, 1997, and LIC-2 ATBs, on August 20, 22, and 23, 1997 (EG&G, 1997a, 1997b). During these performance trials, agent GB was processed. The results presented in Table 2-4 show that emissions of particulate matter, HCI, agent GB, and CO were within the permit limits established by the state of Utah for liquid incinerator systems. Agent destruction was better than the minimum DRE requirement of 99.99 percent.

Emission rates of VOCs, SVOCs, PCDD/F, phosphorus, and metals were compared to the emission rates used in the HRA (Utah DSHW, 1996). The results of this comparison are summarized below and shown in Table 2-5:

• Emission rates for 20 of the metals were below the rates used in the screening HRA. The highest measurement for LIC-1 lead is a statistical outlier indicating a potential sampling problem (which, had it been confirmed prior to the publication of the test report, would have invalidated that particular run and indicated compliance). The phosphorus concentration measured for LIC-1 was above the HRA estimated rate. Mercury was not detected, but the detection limit was above the rate used in the HRA.

• The international toxic equivalent concentrations (ITEQ) for the PCDD/F averaged 0.00034 ng/dscm and 0.00053 ng/dscm (at 7 percent O₂) for LIC-1 and LIC-2, respectively. These are lower than the federal hazardous-waste incinerator regulatory limit of 0.2 ng/dscm (at 7 percent O₂) for new sources.

• Emission rates for two VOCs, ethylbenzene and m,p-xylene, were above the emission rates used in the HRA in at least one run on LIC-1. The other VOCs were either not detected or their emission rates were below the emission rates used in the HRA.

• The majority of the 141 target SVOCs were below measurement method detection limits. The measured emission rate for one SVOC, Bis (2-ethylhexyl) phthalate, was above the assumed HRA

emission rate. Measurement method detection limits were above the equivalent HRA emission rates for dimethylphthalate, however, so conclusions cannot be drawn about the relation of actual and projected emissions for this SVOC.

The list in Table 2-5 includes compounds for which a measured emission rate from LIC-1 or LIC-2 was higher than the value used in the HRA or for which the detection limit was too high to draw a meaningful conclusion.

DFS ATBs with GB were conducted on January 7, 10, and 11, 1997. During these performance runs, M55 rockets were processed at an average rate of 35 rockets per hour. The rockets were punched and drained of GB prior to entering the DFS, although some residual agent remained after the draining operation. The test results are summarized below and in Tables 2-6 and 2-7:

• Emissions of particulate matter, HCl, GB, and CO were below the state of Utah permit limits established for the DFS.

• The measured 99.999981 percent DRE was better than the minimum 99.99 percent DRE requirement.

• Emission rates for 16 metals were below the HRA estimated values. Cadmium, lead, zinc, and phosphorus were higher than the HRA estimated emission rates. The detection limit for mercury was too

high to make a definitive statement. The measured concentration for lead plus cadmium was less than 20 percent of the 24 µqg/dsm³ corrected to 7 percent O₂ limit for hazardous waste incinerators.

• The ITEQ concentrations for the PCDD/F emissions averaged 0.00055 ng/dscm (at 7 percent O₂), compared to the new source performance standard of 0.2 ng/dscm for hazardous waste incinerators.

• Detection limits for seven VOCs and three SVOCs were higher than the estimated values in the HRA in at least one sample set. The measured emission rates or detection limits for the other VOCs and SVOCs were below those used in the HRA or were not detected at all.

Table 2-7 lists compounds for which measured emission rates or detection limits from the DFS were higher than the value used in the HRA.

ATBs of GB in the MPF were conducted on April 4, 15, and 17, 1997. During these performance runs, ton containers with residual GB were spiked with metals to represent the worst case of munitions feed containing heavy metals and agent-contaminated dunnage. In addition, 75 pounds of GB were added to each ton container. The agent feed rate for the MPF was nominally 110 lbs/hr, including both undrained heels (of gelled agent) and added agent. Packages of metal spiking compounds were placed on the feed cradle adjacent to each ton container. The results shown in Tables 2-8 and 2-9 are summarized below:

• Emissions of particulate matter, HCl, GB, and CO were within the state of Utah permit limits established for the MPF.

• The measured DRE was 99.9999 percent, which is better than the required minimum 99.99 percent DRE.

• Metals emission rates were below the rates used in the HRA. Phosphorus emission rates were higher than the HRA estimates.

• Emission rates for the PCDDs were below the rates used in the HRA. Emission rates for tetra-, penta-, and hexa-chlorodibenzofurans in two runs were higher than the HRA rates for these homologues. However, the ITEQ concentrations for the PCDD/F emissions averaged 0.025 ng/dscm (corrected to

7 percent oxygen), which is well below the new source performance standard for hazardous waste incinerators of 0.2 ng/dscm corrected to 7 percent oxygen.

• Two VOCs, m,p-xylene and o-xylene, were measured at levels slightly above the HRA estimated emission rate.

• The detection limits for four SVOCs and 12 VOCs were too high to verify that the maximum emission rates were lower than the assumed HRA emission rate.

Table 2-9 lists compounds for which measured emission rates or detection limits from the MPF were higher than the values used in the HRA.

The purpose of a screening HRA is to estimate an upper bound of health risks to people outside the facility fence-line who could be exposed to facility emissions under worst-case conditions. The HRA is not intended to represent actual risk but to indicate whether risk thresholds have been exceeded and further investigation is warranted. Because the estimated emission rates generated by the Utah Department of Environmental Quality and used in the HRA (Utah DSHW, 1996) differ from several of the actual emission rates, the risks in the HRA would certainly be different if they were recalculated today. Many of the measured emission rates are lower

than those used in the HRA—particularly for major risk contributors, such as dioxins, furans, arsenic, and hexavalent chromium. A few are either higher than the estimated values or are measured with a technique whose detection limits are too high to determine that actual emission rates were below the estimated values. Therefore, to determine the net effect, the calculations will have to be revised using the original HRA model and actual emissions. To assess the potential effect of revised emission rates on the HRA, the committee members made preliminary computations based on the human health medium-specific screening levels established by the EPA (EPA, 1998). The committee found that the revised risk estimates would probably be lower than the original HRA values. Thus, the committee believes that the Army could facilitate use of the measured emission rates in HRAs in the following ways:

• The Army does not have jurisdictional authority for the TOCDF HRA, which was performed by the state of Utah. However, the committee believes the Army, which provided the initial trial burn data (from JACADS), should take the initiative in revising the HRA by issuing a brief update of HRA results based on measured emissions concentrations/upper limits. If and when these revisions are made, the committee urges that the revised figures be widely distributed to the public.

• Emissions estimates for future incineration facilities should take into consideration data from all existing facilities and not just JACADS, which was the only operating facility when the TOCDF emission rate estimates were prepared. New estimates should be based on appropriate statistical bounds scaled to the feed rates of the new facilities and should take into account differences in air pollution control technologies and measurement techniques. Upper confidence limits should be used for assessing latent risks; tolerance limits should be used for assessing acute risks.⁵

• Every effort should be made to ensure that the trial burn conditions and measurement techniques are consistent with the assumptions used for developing the emissions estimates and preliminary operating plans.

Public confidence in the risk estimates is eroded when actual emission rates are higher than those used in the initial assessment. Consequently, every effort should be made to use reasonable upper-bound emissions estimates at the outset of the HRA process, and the consequences of deviations should be explained in the HRA, not after the fact. In addition to design differences, estimates must account for differences in testing techniques and laboratory detection limits between the data used to prepare the projections and the testing procedures that will be used to demonstrate compliance and establish actual emissions rates.

A TSCA trial burn was required for the DFS because PCBs were used as lubricants inside the shipping and firing tubes of M55 rockets. During these trial burns, M55 rockets were processed at an average rate of 35 rockets per hour. The first TSCA trial burn was conducted in January 1997 and the second in November 1998. Results from both agent trial burns are presented in Table 2-10.

Analyses of some of the January 1997 trial burn samples found a tetra-chlorinated PCB congener (four

chlorine atoms in the PCB molecule) in Runs 3 and 4. The tetra-chlorinated congener peak was not present in the samples for Run 1 or in one of the two scrubber liquor samples taken during Run 3. The tetra-chlorinated PCB congener appeared randomly throughout other process samples.

The PCB test series from the January 1997 trial burn resulted in calculated DREs that were better than 99.9995 but averaged slightly below the required 99.9999 regulatory limit for dioxin-containing wastes. During the trial burn, PCDD/F and PCB samples were taken simultaneously using the same sampling, recovery, and cleanup and analysis procedures. The PCDD/F sampling train was spiked with PCDD/F field and recovery surrogates, but not with PCB surrogates, and vice versa. Therefore, quality assurance indicators for the PCB test method cannot be calculated for PCB analyses performed on the archived portion of the PCDD/F samples. Archived PCDD/F samples were analyzed for PCBs and did not exhibit the tetra-chlorinated PCB congener peak. Because the tetra-chlorinated PCB congener only appeared randomly in the first PCB test series and was not found in the simultaneous PCDD/F sampling train, it is probably a sampling or analysis artifact that invalidates the PCB sampling train results. Consequently, the actual DRE for PCBs using the complete required methodology is unknown. PCB DRE results calculated from the PCDD/F samples (better than 99.99994 percent) are probably more representative of actual incinerator performance.

A second TSCA ATB with GB was conducted November 17 to 21, 1998. The uncertified November 1998 test results (the final report was not available when this report was prepared) showed no detectable dioxins, and only near-detection-limit values of dichlorobiphenyls (1.2 to 4.6 ng versus a 1 ng detection limit). Trichlorobiphenyls (2.1 to 2.7 ng versus a 1 ng detection limit) were also observed. The reported concentrations were lower than the concentrations found in the field-blank train (11 ng⁶ and 2.7 ng⁷ for dichlorobiphenyls and trichlorobiphenyls, respectively); however, regulatory practice prohibits deducting field-blank train results from sample measurements to correct for contamination (a common practice for analytical chemists). Consequently, the reported concentrations are likely too large. If these reported concentrations are simply extreme realizations of measurement uncertainty (i.e., data noise) or the result of undetected sample contamination, real PCB emissions may be zero and the calculated DREs significantly understated. The resulting PCB DREs (shown in Table 2-10) calculated from these test results range from 99.999984 to 99.999986 percent, all better than the 99.9999 percent DRE requirement for PCDD/F-contaminated wastes. Consequently, on December 23, 1998, the facility was authorized to process rockets at a rate equal to one-half the rate demonstrated during the November trial burn.

In 1994, after reviewing monitoring systems for the detection and quantification of chemical agents and the by-products of agent and nonagent destruction at JACADS and proposed for the TOCDF, the Stockpile Committee issued the Review of Monitoring Activities Within the Army Chemical Stockpile Disposal Program (NRC, 1994b). This report included a wide range of recommendations for supplementing the ACAMS active alarms and passive DAAMS sampling systems routinely used at chemical demilitarization facilities for agent detection. It also recommended revising the operating procedures of on-site chemical laboratories that analyze DAAMS sample tubes for agent on a daily basis, as well as an aggressive program of the monitoring and analysis of stack emissions for a wide range of products of incomplete combustion at the TOCDF.

Progress made by the CSDP in addressing those recommendations was reviewed in the Systemization report (NRC, 1996a), which generally endorsed the Army's ongoing efforts to improve monitoring instruments and procedures at the TOCDF. The following additional recommendation was made in the Systemization report: "An active program for continual improvement of monitoring instrumentation, including techniques for more rapid recognition of significant levels of agent release, should be pursued" [S-17].

This section reviews the experience at the TOCDF with agent and nonagent (i.e., products of incomplete combustion) monitoring since the beginning of agent

operations and the Army's progress in improving agent monitoring technology. EG&G and the Army have responded to the issue of monitoring products of incomplete combustion by installing a reasonable suite of CEMS on the common stack and feed ducts and have provided an active stack-sampling protocol for ongoing analysis of a wide range of SVOCs (EG&G, 1994).⁸

The major issues that required attention were both agent related: (1) the problem of sporadic, but too frequent, false positive ACAMS alarms; and (2) the selection, testing, and eventual deployment of advanced technology capable of more rapid (< 10 sec) detection of the release of significant levels of agent in the plant or through the common stack.

The problem of sporadic false positive alarms from plant and exhaust stack ACAMS monitors is apparent in operational data from the TOCDF (Holmes, 1998a). Between August 22, 1996, when agent operations began, and October 20, 1998, the seven ACAMS monitors associated with the PAS, including those sampling the common stack (PAS701A,B,C) and those sampling the ducts between individual furnaces and the common PAS (PAS702-PAS705), registered 98 false positive alarms. (In a false positive alarm, an ACAMS response indicates the possible presence of agent above threshold values although no agent is subsequently detected in the much more sensitive and discriminating analyses of material desorbed from the associated DAAMS tubes.) Of these, 39 were attributed to probable interference compounds, 35 were attributed to furnace upsets (which may include responses to odorant compounds in unburned natural gas), 18 were attributed to alarm malfunctions, and 6 were attributed to operator error (Holmes, 1998a). (False positive responses to sulfur-based natural gas odorant compounds may become more frequent when the ACAMS are switched from their current phosphorus-detection mode to the sulfur-detection mode used for mustard agent operation.)

Twenty-two of these PAS ACAMS false alarms automatically shut down agent feed to the LIC, interrupting operations for about an hour each time. Although the false alarm rate was lower than the rate during early JACADS operation, the committee believes that these disruptions are unnecessary and that the Army should continue to improve the instrument specificity and robustness of the monitoring systems.

The committee notes with approval the steps taken by the Army in response to this problem. First, they have defined specifications for an improved ACAMS instrument, which includes improved chromatography to increase specificity, better quantification algorithms to improve accuracy, and more modern electronics to improve signal processing. A competitive procurement for the development and demonstration of this improved ACAMS is planned. Second, the Army has instituted a parallel effort to upgrade the microprocessor and signal-processing software of the existing ACAMS and has initiated plans to test a prototype of the enhanced ACAMS at the TOCDF. Finally, the Army is investigating enhancements to the commercial gas chromatograph-mass spectrometric detector (GC-MSD) units deployed in the laboratories at CAMDS, JACADS, and TOCDF. These units are currently being used to identify interferant compounds that trigger false positive ACAMS alarms so that they can be eliminated from the plant and/or exhaust stream.

A GC-MSD unit with a parallel atomic emission detector designed to recognize phosphorus and sulfur-containing compounds that can trigger the ACAMS flame photometric detectors has been developed and is being tested at CAMDS. In addition, laboratory GC units, with and without MSDs, are being equipped and tested with recently developed pulsed-flame photometric detectors (PFPDs), which promise better, more reliable performance than the flame photometric detectors currently used (DAAMS tube analysis) (Amirav and Jing, 1996). These GC-PFPD and GC-MSD-PFPD units

could also be used to identify interferant species that lead to ACAMS false positive alarms.

The desirability of real-time or near real-time (< 10 sec) detection of significant agent releases from the viewpoint of both worker and resident safety has been discussed in two previous NRC reports (1994b, 1996a). The Army has responded to the Stockpile Committee's concerns in several ways. First, it has installed multiple ACAMS units on the common stack at the TOCDF. By phasing the sampling and chromatography cycles of these units, the intrinsic response time of the ACAMS has been cut from about eight to ten minutes to four to five minutes, providing significantly shorter response times for most releases. The Army has also made shorter intrinsic ACAMS response times a design specification for the improved ACAMS system.

Finally, the Army is supporting a project contracted to the University of Denver to investigate using Fourier transform infrared (FTIR) spectrometers as true real-time detectors. The initial FTIR project by the University of Denver investigated agent-detection limits of a commercial FTIR spectrometer with a conventional open-path, multipass absorption cell and spectral signal-processing techniques. The prototype unit was calibrated for GB and HD and tested at CAMDS. Under laboratory conditions, the system demonstrated an absolute detection limit for GB of ~0.005 mg/m³ (Stedman and McLaren, 1996). Detection limits in the initial field trial at CAMDS, which were affected by the cleanliness of the multipass mirrors and their alignment, were significantly worse than the laboratory values. A second field trial designed to test the feasibility of detecting both agent and products of incomplete combustion in exhaust gases from the CAMDS incinerator stack was unsuccessful because of spectral interference from the high concentration of water vapor in the exhaust samples (Stedman and McLaren, 1996). Further field trials of the FTIR technology at CAMDS are planned.

The committee believes that the theoretical one to ten second FTIR spectral measurement times are encouraging enough that further development and testing of this technology for high-risk venues, such as the munitions unpacking area and the common stack, are warranted. The committee also encourages the Army to monitor published research that may result in new methods of fast-response agent detection.

The Stockpile Committee believes that the Army is pursuing a wise course in upgrading the current ACAMS monitors and simultaneously funding the development of a faster, more specific, more reliable ACAMS. In addition, the promise of combined GC MSD-atomic emission detector, GC-PFPD, and GC-MSD-PFPD for improving the laboratory identification and quantification of both agents and interferants is encouraging and should be vigorously pursued. Finally, FTIR spectroscopy, coupled with state-of-the-art multipass absorption cells and spectral signal-processing algorithms, is a promising technology for real-time monitoring of higher agent concentrations. The committee urges the Army to continue to support its development.

Although the Army has not fulfilled all of the Stockpile Committee's recommendations related to system performance and plant operations, it has completed the period of start-up operations, and its operating mode indicates that program destruction goals will be met (Holmes, 1998b). However, the delay in the processing of M55 rockets has significantly slowed the rate of risk reduction from stockpile storage. Some of the problems in early operation linked to safety management are addressed in Chapters 3 and 4. Although these problems, and the investigations that were necessary to follow up on them, have taken time and management resources that might otherwise have been applied to improving operations, the committee believes the management problems were of much higher priority.

LIC-1, LIC-2, MPF, and DFS RCRA trial burns have been passed satisfactorily, and the DFS TSCA permit is expected in 1999. Unresolved issues involving the management of dunnage, the slag-removal heater, and the need for a BRA are not critical to safe plant performance, although their prompt resolution remains a priority. The renewal application for the RCRA permit was submitted in late 1998. Thus, regulatory authorities had at least six months for review before the permit expired in June 1999.

The Army appears to be making progress in addressing the committee's previous recommendations for upgrading the ACAMS and DAAMS agent-monitoring systems and developing new technologies for real-time detection of agent release.