Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Tracking and Analysis of Stockpile Leakers Leaking munitions or containers (leakers) represent a unique class of stockpile anomaly that affects both storage and disposal operations. For this reason, the occurrence of leakers in storage igloos containing stockpiled chemical munitions and containers has long been subject to attention and tracking. The Army has had programs in place to monitor and inspect stored chemical munitions and containers for leakers since their manufacture. The original intent of the program was to identify problems that would affect the battle- field readiness of the munitions. Both the U.S. Army Soldier and Biological Chemical Command (SBCCOM) at Aberdeen Proving Ground, Maryland, and the U.S. Army Technical Center for Explosives Safety of the Defense Ammunition Center in McAlester, Oklahoma, have been responsible for monitoring and inspecting the stockpile and maintain- ing databases. Members of the Stockpile Committee visited both of the above organizations to review and gather infor- mation on these databases. This chapter discusses the Destruction of the chemical agent and munitions stockpile on Johnston Island was completed at Johnston Atoll Chemical Agent Disposal System (JACADS) in November 2000. By March 2002, destruction of all GB munitions stored at Deseret Chemical Depot (DCD) was completed at Tooele Chemical Agent Disposal Facility (TOCDF). Leaker data from these destroyed munitions were in- cluded in the database used for the statistical analyses in this report. 19 occurrence and measurement of leaking munitions in the chemical stockpile. The committee primarily con- sidered the database provided by SBCCOM but also referred to the Defense Ammunition Center data as needed. The committee also developed several new data fields to facilitate its analyses. It used a commer- cial statistics software program (Minitab) to analyze the leaker data and to determine if any significant trends were discernible. In a briefing to committee members in July 2002, SBCCOM explained how data had been obtained and how the Stockpile Tracking System (STS ~ database has been used. Since 1973, there have been three distinct periods when different inspection protocols were used. Until 1984 (Period I), the primary focus of Army moni- toring was to obtain data on the frequency of leakers, with the goal of ascertaining the combat readiness of the munitions rather than identifying the leakers per se or their sources.2 In 1984, when the Resource Conser- vation and Recovery Act (RCRA) became applicable to the stockpile, new regulations required more infor- mation to be gathered and maintained at facilities where Identification of a leaker required the parent ammunition lot to be declared unserviceable, adversely impacting the Army's readi- ness posture. Currently, the Army monitors for leakers and miti- gates the consequences associated with their occurrence because such efforts are directly linked to safe storage of the stockpile.
20 EFFECTS OF DEGRADED AGENT AND MUNITIONS ANOMALIES ON CHEMICAL STOCKPILE DISPOSAL OPERATIONS chemicals and various classes of chemical wastes were stored.3 At stockpile sites, the implementation of the RCRA regulations coincided with the replacement of relatively unsophisticated techniques, such as paper tape (which was able to detect only relatively high levels of liquid agent), by much more sensitive means. From 1984 to 1991 (Period II), the Army also con- ducted a program of limited intrusive sampling for de- tection of internal munition leakers (the program was called SUPLECAM). Approximately 0.0145 percent of stockpiled munitions were tested in this manner (see Chapter 2~.4 At the same time, the Army continued to monitor the overall stockpile externally using the more sensitive instruments that were becoming available. Dur- ing this period, leakers were recorded in the STS data- base in one of two ways: SUPLECAM or "standard." After 1991 (Period III), the Army developed new inspection protocols under its storage monitoring and inspection (SMI) program, in accordance with proce- dures detailed in Army Supply Bulletin SB 742- 1 (U.S. Army, 1998a). These protocols have been used ever since for monitoring the chemical stockpile for the presence of leakers. Thus, monitoring and inspection protocols changed during three distinct periods. In each period, the motivation for the protocols used was dif- ferent. As discussed below, the committee primarily used data developed during Period III, although some analyses were performed using data from Periods I and II to help provide a historical context. The committee performed various analyses that en- compassed all types of munitions, even as it recognized that the Army has placed appropriate emphasis on monitoring and inspecting GB M55 rockets for leakers in recent years (U.S. Army, 1997; SAIC, 2002~. It should be noted that the Army continued to monitor all munitions for leaks while performing more compre- hensive inspections of GB M55 rockets.5 This empha- sis on inspecting GB rockets is a prominent feature of 3In 1982, M55 chemical rockets were declared obsolete and of no military value. These rockets were declared hazardous waste in August 1984 (SAIC, 2002~. 4Kevin Dolan, U.S. SBCCOM, e-mail to the committee, Febru- ary 20, 2003. sIn this discussion, monitoring refers to examination of the mu- nitions stockpile with leak detection instruments to determine whether a leak occurs. Inspection refers to visual examination of leaking munitions once they are detected during monitoring opera- tions. some of the data presented later in the chapter. In addi- tion, the committee focused on the leaker data for stockpile munitions at one storage site, the Anniston Chemical Activity (ANCA), near Anniston, Alabama. This site, which has 661,529 stockpile items contain- ing 2,254 tons of agent, or 7.4 percent of the original stockpile tonnage, has experienced the second highest incidence of leakers since 1991 (Period III) (U.S. Army, 1995b). (The largest number of leakers occurred at the Deseret Chemical Depot (DCD) near Tooele, Utah, which originally had 1,138,488 items containing 13,616 tons of agent, or 44.5 percent of the original stockpile tonnage. All GB munitions at DCD have been destroyed.) THE ARMY'S STORAGE MONITORING AND INSPECTION PROGRAM The SBCCOM Stockpile Tracking System (STS) da- tabase is the primary source of agent and munition infor- mation (Studdert, 2002~. It has been used by the Army as a basis for analyses of the stockpile to generate data reports and to maintain specific information on the stock- pile. The parameters included in the database (before any additions by the committee) are chemical activities (storage locations of the munitions), munition models (M104, M110, etc.), munition types (rocket, bomb, car- tridge, mine, etc.), munition classes (105-mm, 155-mm, etc.), component types with associated component mod- els (burster casings, burster charge, etc.), munition names, and numerical identifiers. Another source of data is the STS Lot Book (U.S. Army, 2001a), which provides complementary muni- tion and agent information pertinent to the STS data- base and which is periodically updated by the Army. The STS Lot Book provides information on leakers and other suspected munition anomalies based on process- ing experience at JACADS and TOCDF. The informa- tion is largely (but not exclusively) lot-specific and in- cludes some one-of-a-kind occurrences. The STS database covers the geographic distribu- tion of munitions (eight sites in the continental United States and one on Johnston Island in the Pacific Ocean). In concert with the STS Lot Book, it describes the anomalies that have been observed in the chemical stockpile over time. Not all data are directly compa- rable because of the differences in monitoring proto- cols during the three periods noted previously: Period I, before 1984; Period II, 1984 to 1991; and Period III, after 1991.
TRACKING AND ANALYSIS OF STOCKPILE LEAKERS The committee concluded that Period III data for the 10-year period from 1992 through 2001 are the most reliable for a number of reasons: . In Period I, some rocket lots were designated as leaker lots if only one or a few leaking munitions were found. This designation did not result in an overcount of leakers. It did, however, result in closer monitoring of those lots because they were considered more likely to harbor future leakers. · Since 1991, improved detection methods and more sensitive equipment have contributed to bet- ter detection. · Also, since 1991, no intrusive inspections have been conducted on internal parts of munitions. Prior to 1992, internal (within a munition) and external (exterior to a munition) leakers were not differentiated in the database. A significant aspect of the database is that the moni- toring protocols for both Period II and Period III are more reliable than those for Period I. Early on, much less sensitive monitoring (paper tape) was used that might not have been able to detect leakers that could be detected by more sophisticated inspections. (New monitoring techniques were introduced in 1984 that improved detection sensitivity by a factor of 8,000.) In Period II, there was a strong emphasis on intrusive SUPLECAM sampling. In Period III, there has been a strong emphasis on inspecting GB munitions (particu- larly M55 rockets), which have been the source of the majority of leakers. The Period III monitoring proce- dures are more rigorous than those in earlier periods, ensuring that all leakers are identified in the year they begin to leak. In addition, the data do not include infor- mation on munitions that were inspected but not found to leak; in this sense, all the data sets are incomplete for other than leakage information. Table 3-1 provides the Army's current SMI require- ments for all munitions and containers. Under the cur- rent SMI program (Period III), all structures holding the entire chemical stockpile (including the site hold- ing the limited outdoor stockpile of ton containers at Aberdeen Proving Ground, Maryland) are subject to quarterly monitoring and, if necessary, inspection. If agent is not detected in an igloo, it is assumed that none of the munitions in that igloo leak (if agent is detected, the igloo will be entered by personnel in protective gear). Specific guidance states that each item is to be visually inspected for evidence of leakage, condition 21 TABLE 3-1 SMI Requirements for Toxic Chemical Items Item Not Overpacked Overpacked Ton containers Mines Projectiles/cartridges Bombs Spray tanks DOT bottles Rockets SUPLECAM samples M56 warheads Binary components Quarterly Annual Annual Annual Quarterly Quarterly Quarterly N/A Quarterly N/A Quarterly Quarterly Quarterly Quarterly Quarterly Quarterly Quarterly Quarterly N/A Source: U.S. Army (1998a). of the outer pack, dunnage, or any other condition af- fecting its suitability for continued safe storage. Items, whether in the original shipping and storage container or in overpacks, are to have the outer pack inspected without opening (U.S. Army, 1998a). Once identified as leaking, a munition is segregated from nonleaking munitions, overpacked, and stored separately. While a new generation of overpacks is now in use, the early overpack containers were "acceptable containers of convenience," and it was not uncommon for the over- packs themselves to leak, sometimes more than once. Even though only the leaking munition in the overpack is the ultimate source of a leak from an overpack, a leaking overpack is always reported as a separate leak. Rocket lots that are identified as "leaker lots" are in- spected more often. Under the monitoring protocols of Periods I and II, a procedure similar to that described above for Period III was used namely, igloos were entered to determine if a leak had occurred and if it had, the leaker was identi- fied and moved to an overpack igloo. In principle, these protocols were able to identify all leakers that occurred in a given year. Nevertheless, it is probable that many leakers were not detected during Period I. During SMI inspections conducted according to the frequencies shown in Table 3-1, the storage structures are monitored to an 8-hour time-weighted average (TWA) exposure limits for the type of agent stored 6The TWA exposure limit represents the mean allowable level to which workers can be exposed for 8 hours over the course of a 40 hour/week working lifetime with no adverse health effects.
22 EFFECTS OF DEGRADED AGENT AND MUNITIONS ANOMALIES ON CHEMICAL STOCKPILE DISPOSAL OPERATIONS therein. The follow-up inspections require that person- nel enter the storage structures in appropriate personal protective equipment. Moreover, monitoring to the TWA limit from external inspection ports is conducted once daily for igloos containing leaker lots of nonoverpacked GB M55 rockets. Igloos with nonleaker lots or overpacked leakers are monitored at least once a week (Studdert, 2003~. Figures 3-1, 3-2, and 3-3 were prepared by the Army. As far as the committee knows, no statistical analysis of the data was carried out by the Army. Figure 3-1 depicts leaker data gathered from moni- toring during Periods I, II, and III (see also Table 3-2~. The data for overpacks and rockets are presented by year; they include leaking munitions from all sites. The more than 500 leakers detected in 1981 resulted prima- rily from a 100 percent inspection of one type of muni- tions at DCD. This inspection was carried out because the Army had reason to believe that there were a large number of leakers among these munitions. While it was suspected at the time that GB M360 105-mm projec- tiles were the principal source of leaks, M55 rockets were subsequently determined to be of greater concern in this regard. The data for 1981 show, as indicated above, that when the occurrence of leakers was more intensively investigated during Period I (e.g., by in- specting 100 percent of a munition type), more leakers could be found. Figure 3-2 shows data similar to the data in Figure 3-1, but for the Anniston stockpile only. It is probable that many of the leakers detected in 1981 had started leaking earlier but were missed because of the relatively poor sensitivity of the detection equip- ment and the incomplete monitoring that took place at that time. Figure 3-3 shows the number of leakers by munition type in the overall stockpile. The data in Figure 3-3 and the summary data in Table 3-3 include all types of leakers (internal, external, and overpack). Figure 3-3 shows that GB M55 rockets have the highest number of leakers, almost equal to the number of leakers from all other munition types combined. The highest num- ber of leakers occurs in GB munitions, as indicated in Table 3-3. H munitions show a higher percentage of leakers than GB munitions, but these H munitions com- prise a relatively small number of projectiles, located primarily at DCD. Furthermore, H is much less volatile than GB. The GB in the stockpile comprises several subtypes 500- 400- 300 200 100 o 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 Overpack All Others ~ Rockets FIGURE 3-1 Distribution of leaking munitions in the U.S. chemical weapons stockpile from 1973 to June 30, 2002. Source: Studdert (2002).
TRACKING AND ANALYSIS OF STOCKPILE LEAKERS 200- 180- 160- 140- 120- 100- 80- 60- 4n- o~ FIGURE 3-2 (2002). 23 I 1 1 1 1 1 1 T I I I 1 T 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 200 · Overpack ~ All Others Rockets Distribution of leaking munitions in the Anniston stockpile from 1973 to June 30, 2002. Source: Studdert 2250- 2000 1750 1500 1250 1 000- 750 500 250 o ~1 1 ,,,,,,,,R,.... T r I.............. 1 R ~3 1 1 1 1 1 1 1 1 1 A A' Atop heir ~~ <,~Z'~ 0' . FIGURE 3-3 Number of leaks by munition type, all sites. Key: TC, ton container; WHD, warhead; Projo, projectile; Ctg, cartridge. Source: Studdert (2002~. Category Information: GB rockets, M55: includes rockets (H520) and rocket warheads; GB projectiles, 105-mm M360 (C766, formerly 1315-7906~; GB cartridges, 105-mm, M360 (C441~; GB projectiles, 155-mm: includes 155-mm M122 (D483) and 155-mm M121/M121Al(D542~; GB bombs: includes Weteye (E832), 500 lb MK94-0 (E384), and 750 lb MC-1 (E388~; VX rockets, M55: includes rockets (H521) and rocket warheads; and overpacks: collated quantity for all overpack failures: 451 GB overpacks, 51 VX overpacks, and 4 each of H/HD/HT overpacks failed.
24 EFFECTS OF DEGRADED AGENT AND MUNITIONS ANOMALIES ON CHEMICAL STOCKPILE DISPOSAL OPERATIONS TABLE 3-2 Total Number of Leakers for All Years, All Sites, and All Categoriesa Year Category 1 Category 2 Category 3 Category4 1973 180 0 0 0 1974 88 0 0 0 1975 56 0 0 0 1976 57 0 0 0 1977 78 0 0 0 1978 58 0 0 0 1979 63 0 0 0 1980 211 0 2 0 1981 516 0 0 0 1982 68 0 0 0 1983 59 0 0 0 1984 244 0 0 0 1985 223 0 3 1986 131 0 15 , 1987 63 4 17 1988 105 1 5 1989 161 0 2 5 1990 111 5 0 0 1991 110 0 2 1992 126 1 0 138 1993 127 2 0 1 1994 145 2 0 27 1995 175 0 0 11 1996 237 2 0 1 1997 70 5 0 6 1998 141 3 1 24 1999 249 73 0 36 2000 103 87 0 16 2001 260 15 0 19 2002 (through dune) 25 11 0 0 All 4,240 211 47 291 180 88 56 57 78 58 63 213 516 68 59 244 227 149 85 112 168 116 113 265 130 174 186 240 81 169 358 206 294 36 4,789 aSee Box 3-1 for category descriptions. Period I, 1973-1984; Period II, 1984-1991; Period III, 1992 forward. TABLE 3-3 Munition Leakers by Type of Agent Agent GB H VX HD HT Stockpile quantity 1,546,387 77,498 497,175 931,945 270,135 Number of leakers 3,859 360 273 236 61 Percent leakers 0.25 0.46 0.05 0.03 0.02 Source: Studdert (2002~.
TRACKING AND ANALYSIS OF STOCKPILE LEAKERS TABLE 3-4 M55 GB Rocket Sampling Plan and M55 GB Rocket Storage Monitoring Inspectiona Quarterlyb Sample Size by Lot CategoryC Category C Lot Size Category A Category B (leaker Lots) 0-25 84 204 100% 26-150 164 444 2Se 151-300 6 16 40e 301-750 9 23 64 751-3,000 12 30 75 >3,000 15 35 85 aStorage inspection entails monitoring the interior of M441 shipping and firing tubes. bUnless otherwise indicated. CCategory A - no leaks occurred; Category B - one or more leaks occurred but not considered a leaker lot; Category C - numerous leaks occurred and considered a leaker lot. Quantity indicated is to be sampled every 3 years. This sampling may be accomplished at one time within the 3 years or in increments as determined by the QASAS-in-Charge. eWhen total annual sample size exceeds the lot size, the lot will be sampled 100% each year. This 100% sampling may be accomplished at one time within the year or in increments as determined by the QASAS- in-Charge. Source: Adapted from Studdert (20021. based on the manufacturing history, the stabilization techniques employed, and the type of stabilizer used (see Chapter 2 for details). The frequency of leaking munitions containing GB is a function of subtype. In 1984 and 1985 (onset of Period II), the Army increased both the frequency and number of its inspec- tions to detect leaking GB rockets. At that time, the Army was beginning to recognize that the largest num- ber of leakers was occurring in GB munitions. This resulted in more aggressive monitoring using instru- ments that were more sensitive that is, the new in- struments were able to detect much lower levels of agent than what had been previously used.7 The in- creased inspections were undertaken to make certain that the problem of leaking GB munitions was under- stood. Subsequently, entries became more frequent and more munitions were inspected for possible leaks. Here, the inspection protocol led to an increase in the number of GB rockets inspected from GB lots (batches of munitions) that had a propensity to develop leakers 7These actions also produced a substantial increase in the num- ber of leakers detected in 1984 and 1985 relative to 1983 (see Fig- ure 3-1 and Table 3-21. Again, it is probable that many of these leakers occurred before 1984 but went undetected at the time. 25 (Studdert, 2002~. The SBCCOM-specific SMI proto- col for inspecting these rockets, where lots were cat- egorized as A, B. or C on the basis of leaker history, is summarized in Table 3-4. Category A lots are those in which no leakers occurred. Category B lots are those that experienced one or more leakers but were not des- ignated as leaker lots. Category C lots are those that experienced numerous leakers and have been desig- nated as leaker lots. Sampling protocols mentioned in Table 3-4 refer to inspections, not to the basic leaker monitoring that is used to identify leaks in the first place. The sampling covered in Table 3-4 does not relate specifically to the statistical analysis discussed below. By way of background, the Army uses two types of reports to report leakers. The first is a "chemical event report" (U.S. Army, 2001b), which reports any chemi- cal event that releases agent outside of engineering con- trols. The second is a "leaker report" (U.S. Army, 1998a), which reports any munitionts) that are leaking upon inspection. The leaker reports contain all the tech- nical information available about each leak. They are the basis for compiling the STS database and for re- ports derived from that database. The two types of re- ports serve different purposes and are administered under different regulations by different organizations.
26 EFFECTS OF DEGRADED AGENT AND MUNITIONS ANOMALIES ON CHEMICAL STOCKPILE DISPOSAL OPERATIONS Thus, not all leakers reported in the leaker reports are reported as a chemical event and vice versa. Conse- quently, the leaker reports and the STS database were the primary sources of the information gathered on leakers for this report because they are more specific to the phenomena of interest. Technical data in leaker reports are used as a man- agement tool to minimize risk associated with the stor- age of the chemical stockpile (Studdert, 2002~. They do this by · Identifying leakage trends among specific muni- tions families, manufacturers' lots, and agent populations; Establishing or modifying the scope and fre- quency of surveillance inspections and special studies; Prioritizing resources to maximize benefits (e.g., installation of engineering controls on selected magazines or during specific operations); · Sharing information with PMCD to facilitate lot scheduling within disposal campaigns for specific munition types and for developing site-specific quantitative risk assessments; and Providing the basis for response to queries from DOD, DA, and AMC; response to media queries and FOIA requests; response to requests from non-DOD agencies (GAO, CDC, NRC, etc.~; ne- gotiation with state environmental agencies; and legal defense during litigation. . . STOCKPILE LEAKER DATA This section describes the STS leaker database for the entire stockpile for the period from 1973 through June 2002, with particular emphasis on the data ob- tained since 1992 (Period III). The data for Periods I and II are included and addressed in certain instances to assist in placing the Period III materials in appropri- ate perspective. For reasons that will be discussed be- low, data from Periods I and II are limited to time- independent analyses, i.e., the Pareto charts. The section also discusses how additional information on the temperature history in storage igloos could make the database more useful for elucidating trends in leaker frequency. SBCCOM provided the committee with access to the STS database, wherein 4,789 leakers had been documented in 2,792 separate reports from 1973 through June 2002 through its various monitoring pro- grams. These data are shown in Table 3-2, where the leakers are arranged into four categories. Category 1 contains leaks attributable to deterioration of the origi- nal item; Category 2 contains leaks from an overpack of an item that leaked in the past; Category 3 contains leaks detected during SUPLECAM or other reliability programs in items that had not previously leaked; and Category 4 contains leaks associated not with deterio- ration but with, for example, ineffective maintenance and dropped pallets. The committee, with SBCCOM assistance, added several new fields for additional data. The data were entered into a Microsoft Excel spreadsheet, which was then exported into a statistical software program (Minitab). As discussed previously, statistical analyses were conducted for Period III data obtained during the years 1992 through 2001. As noted above, during that period, the Army recorded 2,103 leakers for all agent and munition types in the stockpile; these leakers gen- erated 1,258 leaker reports. Twenty-five fields in the STS database as modified by the committee were applicable to the committee's efforts. These fields, with explanatory text given in parentheses where necessary, are listed in Box 3-1. Various mechanisms for leak formation were dis- cussed in Chapter 2. In general, and especially for GB, leakers can be a result of chemical attack or a result of mechanical failure from bad welds, poor machining, and so forth. The several GB agent subtypes (PRO, PR-RS, RO-RS, RD-RS) in the stockpile have varying purities and stabilizer constituents and concentrations (see Chapter 2). The GB subtype reflects the manufac- turing lot and subsequent treatment history of the agent. The composition of the GB agent subtypes affects the degree of acid attack on the metal components of mu- nitions and thus the propensity for leaks to develop. As discussed in Chapter 2, temperature can affect the rate at which the stabilizer in GB degrades. The rate of agent degradation can also be expected to differ among GB agent subtypes and will also be time and temperature dependent. The time dependence is a func- tion of the concentration of stabilizer. The average ambient temperature at different storage sites is likely to vary. The interior temperatures of storage igloos at each site over time could be a significant variable in the rate of degradation and the extent to which degra- dation reactions proceed, whatever the GB agent sub- type or the type of stabilizer. The STS database contains reasonably good informa- tion on the age of the more than 3 million items recorded.
TRACKING AND ANALYSIS OF STOCKPILE LEAKERS 27 However, information over time on the interior tempera- tures of storage igloos is sparse to nonexistent. Figure 3-4 is a representation of the only available data that relate storage igloo temperature to leaker rates; the data are for leakers of one type of munition (155-mm projec- tiles) at one site. To complicate matters, the degradation rate is related to the temperature of the agent in the mu- nitions, and the igloo temperature may only be a surro- gate for that parameter. Based on the very limited data presented in Figure 3-4, it is not possible to infer any correlation between degradation rate or leaker formation and temperature. Having more detailed information on the temperature inside igloos and on a wider range of munitions over a period of several years could be useful for ascertaining if there is a correlation. STATISTICAL APPROACH OF THE STOCKPILE COMMITTEE Approximately 3,300,000 munitions were in the original stockpile prior to destruction operations at JACADS and TOCDF. From the monitoring protocols used by the Army from 1973 to June 2002 to inspect the stockpile, 4,789 leakers were reported in the four cat- egories described in the preceding section (see Item 9 in Box 3-1 and Table 3-2). This number includes some
28 EFFECTS OF DEGRADED AGENT AND MUNITIONS ANOMALIES ON CHEMICAL STOCKPILE DISPOSAL OPERATIONS 70 - 60 50 - 40 30 ~1 LEAK IN STORAGE\ ( JAN 96 o | LEAKS | Deg F = ~~- ~ FEB 96 ) ~ LEAS D ISCO HER ED DURING SPEC INsPECT IoN ~ ~ INTER IoR ~ ~ ~ WP P CONTAIN ERS - F EB 88 CONTROLLED ENVIRONMENT - NO RELEASE )- _ ~ ~ ~e ~-~1 / ~ ~ \> -~ \ /! ~ JAN WEB | MA~PR ~ MAY T JUNE | JUL | AUG | SEP | OCT | NOV | DEC | 1 ~ 6 1 0 1 0 ~ ~ 11 1 2 1 12 1 18 1 3 1 1 1 0 1 0 1 _ ~ 42 T 41 T 45 ,~; 51 T 56 T 64 T 68 1 68 T 62 1 55 T 48 1 WEAK IN sro RAGE MAY 79 ) gAVG INTER IOR IGLOO T EM P;\\ REPRESENTATIVE UMCD IGLOO ) MAR01- FEB 02 ~ ~~ If- OCT84 INTRUSIVE INSPECT ION ACCOUNTS FOR DISCO VER Y OF 46 THIRSTER W ELL LEAKS (INTERNAL LEAKS) CONTROLLED ENVIRONMENT - NO R ELEASE FIGURE 3-4 Seasonal distribution of 155-mm projectiles leaker occurrences at Umatilla by month and igloo temperature. Source: Studdert (20021. Overpack leakers, which in the statistical context of mu- nitions per se would constitute a double count. There are 4,240 leakers in Category 1, which does not include overpacks, SUPLECAM data, or leakers due to mecha- nisms other than deterioration. For Period III (after 1991), there are 1,633 Category 1 leaks, and these are predominately in munitions containing GB. Of these leakers, 578 are at ANCA and 846 were at DCD (the latter were destroyed). Of the 578 leakers at ANCA, 547 are GB-filled (across all GB subtypes); of these, 543 are M55 rockets. The number of M55 leakers by GB sub- type are 101 PRO, 378 PR-RS, and 64 for which the data are incomplete and the identification uncertain. The committee limited the data used for the analysis to Category 1 leaks to eliminate the double counting that would occur because of leakers in overpacks and to concentrate on leaks resulting from deterioration. Statistical tools included (1) Pareto charts; (2) descrip- tive statistics packages; (3) a chi-squared test; and (4) analysis of variance. For statistical analyses of trends and variance by the committee (see below), only Period III data were used. For the Pareto charts, data from all three periods were used in order to gain perspective on the distribution of leaker types over the entire time the stockpile was monitored. Because the Pareto charts encompass the total number of leakers found during the entire 30-year period and are not dependent on the year a leaker was discovered, they are probably reasonable representa- tions of the distribution of total leakers at Anniston. As argued above, it is very likely that leakers that went undetected early in the Army's monitoring activities (primarily in Period I) were eventually detected when detection sensitivity improved and monitoring became comprehensive. 8 Thus, any error introduced on a year- ~This observation was confirmed by Lis Wachutka, Quality Assurance Specialist, SBCCOM, Aberdeen Proving Ground, Mary- land, who noted in a private communication on December 22, 2003, that at least 99 percent of the external leakers over the 30-year period have been detected.
TRACKING AND ANALYSIS OF STOCKPILE LEAKERS 4,000 - tn au ~ 3,000- o au 2,000- z 1 ,000 - O- (j~ ~ I ~ Obey Count 3,395 359 229 220 37 Percent 80.1 8.5 5.4 5.2 o.s Cumulative (DO) 80.1 88.5 93.9 99.1 100.0 29 100 80 60 ~ au 40 2n FIGURE 3-5 Pareto chart indicating Category 1 leaks by agent (across all sites and including all munition types). Note: The term "other" refers to data for which there is no information as to the particular munition or to data for an agent not listed on the horizontal axis, such as lewisite. to-year basis by imprecise monitoring procedures is probably not a significant factor over the course of a period. Similar analyses can be done for other sites, for vari- ous combinations of sites, and for any number of sub- sets of data. One example of another analysis is con- tained in the PMCD report Assessment of the Storage Monitoring Inspection Program for M55 GB Rockets, which found that the limited inspection conducted un- der the SMI program was sufficient to determine the stability of the GB stockpile as a whole (U.S. Army, 1997~. That report further concluded that the Army would be justified in continuing the SMI program, but that a more rigorous and robust protocol could enhance the SMI results (U.S. Army, 1997~. Another example is a more focused and detailed assessment of a particular leaker lot of M55 rockets stored at the Anniston site. This is the subject of the report Statistical Analysis of the Leakage History for Anniston M55 GB Rocket Lot 1033-45-181, by the Science Applications International Corporation (SAIC, 2002~. Plotting data in Pareto charts (vertical bar graphs with bars in order of size from left to right) enables easy visualization of the data on leakers. Figure 3-5 is a Pareto chart showing the number of Category 1 leaks by agent type across all sites and all munition types for all the years covered by the STS database (Periods I, II, and III). It includes destroyed munitions as well as those still in storage. As indicated, 80.1 percent of the leakers contained GB; 8.5 percent, H; 5.4 percent, HD; and 5.2 percent, VX. Figure 3-6 shows total Category 1 leaks by Depart- ment of Defense Identification Codes (DODICs) for lots and lot clusters of munitions at all storage loca- tions.9 Six of the first seven DODICs (H520, C766, D543, D483, K725, and D542) identify GB-filled M55 rockets, which account for over 80 percent of the leakers. The information contained in Figures 3-5 and 3-6 again illustrates that by far the largest number of leakers in the stockpile involve GB-filled munitions. These two figures aptly demonstrate the rationale of the Army for its increased inspections in recent years of GB-filled munitions, particularly M55 rockets, rela- tive to other munitions. Over three-quarters of Category 1 leaks to date have occurred at just two sites: Deseret Chemical Depot and Anniston Chemical Activity, as shown in Figure 3-7. This observation points to the desirability of a site- centric approach for the conduct of statistical analy- 9DoDICs are definitive identifiers of a type of munition.
4,000 - ~n ~ 3,000- ct 0 2,000- n Z 1,000- 1 / an_ O- 1 1 1 1 1 1 1 1 1 1 1 1 1 1 ~5~ G566 o~ o~ ~5 ~~ o~ ~~ Gas o~ ~5 ~3~ ou~ i Count 2,O76 680 306 157 155 106 102 100 95 83 82 76 Percent 49 16 7 4 4 3 2 2 Cumulative (DO) 49 65 72 76 80 82 69 153 2 2 2 2 2 4 84 87 89 91 93 95 96 100 2 2 2 FIGURE 3-6 Pareto chart of Category 1 leaks by DOD Identification Code for all sites. 4,000 - ~D 3,000- 0 2,000- Z 1,000- O- - 100 - 80 - 60 - 40 - 20 - O AGE EGG ~~ S\ ALGA ALGA 0~ Cou nt 2,271 984 364 201 195 89 136 Percent 53.6 23.2 8.6 4.7 4.6 2.1 3.2 Cumulative (°/0) 53.6 76.8 85.4 90.1 94.7 96.8 100.0 FIGURE 3-7 Pareto chart of Category 1 leaks by site for all sites. See List of Acronyms - 100 - 80 - 60 - 40 - 20 - O in front matter for acronyms.
TRACKING AND ANALYSIS OF STOCKPILE LEAKERS TABLE 3-5 Period III Category 1 Leaks at Each Site by Year Year ANCA BGCA DCD JI PBCA PCD UMCD All 1992 45 24 44 5 2 1 5 126 1993 33 11 72 2 1 0 8 127 1994 50 2 88 0 3 2 0 145 1995 99 5 62 6 0 0 3 175 1996 107 5 46 64 1 12 2 237 1997 40 0 27 0 1 0 2 70 1998 21 2 113 0 2 0 3 141 1999 75 9 161 0 1 0 3 249 2000 52 7 37 2 2 1 2 103 2001 56 2 196 0 3 0 3 260 All 578 67 846 79 16 16 31 1,633 Note: See List of Acronyms in the front matter for site acronyms. TABLE 3-6 Period III Category 1 Leaks at Each Site by DODIC DODIC ANCA BGCA DCD JI PBCA PCD UMCD All C441 C442 C698 C703 C766 D483 D484 D542 D543 E382 E384 E388 H520 K257 K655 K665 K725 K732 NA All 4 3 1 11 o o o o o o o o 543 12 4 o o o o 578 o o o o o o o o 46 o o o 21 o o o o o o 67 2 o 2 o 217 153 s 28 79 7 o 33 248 o 13 o 39 8 12 846 o 4 o o o o o o 69 1 1 2 o o o o o 79 o o o o o o o o o o o o 15 o o o o o 16 o 16 o o o o o o o o o o o o o o o o o 16 o o o o o o o 2 o o o 21 o o o o 31 20 3 15 217 153 s 30 125 7 69 35 849 14 21 42 8 2 ,633 Note: See List of Acronyms in the front matter for site acronyms. ses that is, an approach in which the particular mix of agent and munition types at a site is considered in de- termining whether apparent leaker rates are indicative of a trend over time. Tables 3-5 through 3-9 present some examples of how Period III (1992-2001) data can be arrayed to high- light areas and munitions of concern and to help direct 31 monitoring and management of the stockpile and pro- cessing sequence. In these cases, the data are from the inspections done according to Army protocols during Period III. Tables 3-5, 3-6, and 3-7 offer site-centric perspectives of leaker occurrences according to year, DODIC, and agent type, respectively. Leaker occur- rences could be related to the ages of the leaker muni-
32 EFFECTS OF DEGRADED AGENT AND MUNITIONS ANOMALIES ON CHEMICAL STOCKPILE DISPOSAL OPERATIONS TABLE 3-7 Period III Category 1 Leaks at Each Site by Agent Type Agent ANCA BGCA DCD JI PBCA PCD UMCD All GB 547 21 739 72 15 0 27 1,421 H 0 46 84 0 0 0 0 130 HD 18 0 13 5 0 16 4 56 HT 1 0 2 0 1 0 0 4 VX 12 0 8 2 0 0 0 22 All 578 67 846 79 16 16 31 1,633 Note: See List of Acronyms in the front matter for site acronyms. TABLE 3-8 Period III Category 1 Leaks by Agent Type and DODIC TABLE 3-9 Period III Category 1 Leaks by Agent Type and DODIC for ANCA Site DODIC GB H HD HT VX All DODIC GB HD HT VX All C441 C442 C698 C703 C766 D483 D484 D542 D543 E382 E384 E388 H520 K257 K655 K665 K725 K732 NA All 7 o o o 217 153 o 30 o 7 69 35 849 o o o 42 o 12 1,421 o o o o o o s o 125 o o o o o o o o o o 130 o 20 o 15 o o o o o o o o o o 21 o o o o 56 o o 3 o o o o o o o o o o o o o o o 4 o o o o o o o o o o o o o 14 o o o 7 20 3 15 217 153 s 30 125 7 69 35 849 14 21 1 42 8 8 0 12 22 1,633 tions for each year through additional analysis. Table 3-8 presents recorded leakers during Period III by agent type across all sites. The contribution from the ANCA stockpile to the population of leaking munitions for Period III is shown in Table 3-9. The raw data shown in Table 3-5 were analyzed us- ing an appropriate regression analysis and analysis of variance. The data can be analyzed using individual data points or totals by year. The regression analysis of total number of leakers versus year, as seen in Figure 3-8, shows no significant increase in the number of leakers over time. If the data are analyzed differently, C441 C442 C698 C703 H520 K257 K655 All 4 o o o 543 o o 547 o 3 o 11 o o 4 8 o o 1 o o o o 1 o o o o o 2 o 2 11 543 2 4 578 however, a trend might appear. For example, data for all sites rather than just ANCA suggest an increase of leakers with time.l° Figure 3-9 is a Pareto chart showing the frequency of GB M55 rocket leakers by GB subtype at ANCA. Figure 3-10 is a Pareto chart of the GB M55 rocket population in the ANCA stockpile by GB subtype. These charts were based on Period III data for Cat- egory 1 leaks. The percentage of leakers by GB sub- type and the percentage of GB rockets by subtype in the total population do not match. Specifically, the pro- portion of PR-RS rockets that were observed to have leaked in Period III is 3.07 percent (378/12,325), com- pared with the 1.38 percent (543/39,210) in the total GB M55 rocket inventory at ANCA in the same pe- riod. This is a statistically significant difference in the proportion of leaks (p < 0.001) between PR-RS i°~t should be noted, however, that this apparent trend is prima- rily a result of including DCD weapons, where, of course, all GB munitions have been destroyed.
TRACKING AND ANALYSIS OF STOCKPILE LEAKERS 150 100 u' a) <~ 50 o _ w _ _ . - C,,. i~r~ so ~ . , ~ _ ~ ~ 'or ~ _ ,,,, _ ~ ~ ~ [~Ql KIEV . it. a, i.. via , ~ Fill ~ ~ _ . _ .c w 0~ ~ ~ ~~ ~ ~ ~ · wry _. . .~ - Hi. ~ ~~h se:K so ~ ~ is, · ~ ~ . _- . _ ,~ 5 ~ x 09 1992 1993 1994 1995 1996 1997 Year Detected Regression 95% Cl 95% Pi 1998 1999 2000 2001 FIGURE 3-8 Regression analysis of aggregated-by-year-of-leak totals at ANCA versus year detected. 500 - `,' 400 - Cal ~ 300 - o ~ 200 - it 00 - O- - - _w ..... ...~. ?~,~e ?~0 ~ ~ Count 378 101 62 2 Percent 69.6 1 8.6 1 1.4 0.4 Cumulative (DO) 69.6 88.2 99.6 100.0 33 - 100 - 80 - 60 ~ c' - 40 - 20 - O FIGURE 3-9 GB rocket leaks by agent subtype at ANCA from 1992 through July 2002. Note: *indicates no information available to specify subtype; "Others" indicates subtypes other than PRO, PR-RS.
34 40,000 - c~ 30,000- ° 20,000- n Z 1 0,000 - O- EFFECTS OF DEGRADED AGENT AND MUNITIONS ANOMALIES ON CHEMICAL STOCKPILE DISPOSAL OPERATIONS / ~~ ~~ ~5 Count 25,2 35 12,325 1,650 Percent 64.4 31.4 4.2 Cumulative (°/O) 64.4 95.8 100.0 - 100 - 80 - 60 a, - 40 - 20 - O FIGURE 3-10 GB rocket population by agent subtype at ANCA prior to start of disposal operations (20031. Note: "Others" indicates subtypes other than RO, PR-RS. and the total population. Similarly, the 0.40 percent (101/25,235) of leakers for PRO is significantly lower (p < 0.001) than the 1.38 percent of leakers in the total population. More important, perhaps, is the ob- servation that PR-RS leakers are more than seven times as frequent as PRO leakers. These findings give management further information to pinpoint the muni- tions that should be of greatest concern. Since all of the munitions currently in the stockpile are more than 30 years old, any analysis of time of leakage onset for individual munitions based on data from Periods I, II, and III would be of limited use for monitoring the remaining life of the stockpile regard- less of the quality of the early data. It is important to note, however, that the time since redistillation or restabilizing of the GB is about 30 or 35 years. Chapter 2 discusses the possibility that autocatalytic reactionts) might occur. One possible concern is that the rapidly rising portion of the curve for the autocatalytic reac- tion has not been reached (see Figure 2-3~. If, in fact, it is still to be reached, the stability of the stored muni- tions could be significantly compromised and leakage rates could rise substantially. Statistical treatment of the information in the STS database by available software tools such as Minitab can be used to examine a very large number of possible interactions and to develop findings that could provide qualitative guidance for monitoring the existing stock- pile until its destruction is complete. However, any such findings cannot be used for gelled and/or foaming munitions because they are not represented by leakers. With requisite data, Minitab (or equivalent statistical software programs) could be used to examine whether or not there are correlations between DODICs and the number of gelled rockets. SUMMARY Statistical analysis offers a means for gaining under- standing of and insights into the condition of stock- piled chemical munitions. It can enable the Army to recognize developing trends in leaker frequencies and other manifestations of anomalies. Such analyses gen- erally involve a three-step process: · Compiling a good database, Comparing graphical depictions of the data, and Employing more rigorous statistical analysis techniques to learn whether a conclusion drawn from a graph is valid. . .
TRACKING AND ANALYSIS OF STOCKPILE LEAKERS Such a process is not easy to establish at the beginning of a program unless the objectives are clear and a stan- dard and consistent methodology for data recording and monitoring of the populations) of interest is estab- lished. In this case, the objectives of the Army's muni- tion inspection program have changed over time, lead- ing to changes in data requirements and means of collection. This chapter has discussed the various monitoring and inspection protocols used by the Army over the years to identify leaking munitions in the stockpile. Of these approaches, the SMI procedures (provided in Army Supply Bulletin SB742-1) used over the past 10 years have provided a consistent data set (U.S. Army, 1998a). In addition, SBCCOM and the U.S. Army Technical Center for Explosives Safety of the Defense Ammunition Center have maintained separate data- bases on the occurrence of leakers in the chemical stockpile. The committee examined both but used the SBCCOM database for this study, adding several new data fields in the process (such as age of munitions and type of agent) to facilitate the analyses. The results of its work are as follows: 35 · The committee looked at the chemical stockpile as a whole, and the Anniston chemical stockpile in particular, for evidence that the leaker rate is increasing with time. As for the rate of leakers increasing with time, the possibility of the onset of autocatalytic reactions of the stabilizer makes it important to continually monitor the stockpile for possible upturns in the leakage rates. A relatively small number of munition types con- tain the bulk of the leakers. Most of these are GB- filled munitions, the bulk of which are M55 rock- ets. While H munitions appear to leak at a higher rate than GB munitions, the former constitute a much smaller number of total munitions in the stockpile. Also, VX, HD, and HT munitions leak much less often than either GB or H munitions. A significantly higher proportion of M55 rocket subtype PR-RS leaks than of subtype PRO. Fur- ther inspections should focus on this disparity and attempt to determine its causes and significance for monitoring and inspecting the stockpile for its ~ ~ r remalmng llIe.
