Gulf War Illnesses: DOD's Conclusions about U.S. Troops' Exposure
Cannot Be Adequately Supported (01-JUN-04, GAO-04-159). 	 
                                                                 
Since the end of the Gulf War in 1991, many of the approximately 
700,000 U.S. veterans have experienced undiagnosed illnesses.	 
They attribute these illnesses to exposure to chemical warfare	 
(CW) agents in plumes--clouds released from bombing of Iraqi	 
sites. But in 2000, the Department of Defense (DOD) estimated	 
that of the 700,000 veterans, 101,752 troops were potentially	 
exposed. GAO was asked to evaluate the validity of DOD, 	 
Department of Veterans Affairs (VA), and British Ministry of	 
Defense (MOD) conclusions about troops' exposure.		 
-------------------------Indexing Terms------------------------- 
REPORTNUM:   GAO-04-159 					        
    ACCNO:   A10329						        
  TITLE:     Gulf War Illnesses: DOD's Conclusions about U.S. Troops' 
Exposure Cannot Be Adequately Supported 			 
     DATE:   06/01/2004 
  SUBJECT:   Chemical and biological agents			 
	     Data collection					 
	     Data integrity					 
	     Diseases						 
	     Health hazards					 
	     Health research programs				 
	     Health statistics					 
	     Health surveys					 
	     Infectious diseases				 
	     Statistical data					 
	     Veterans						 
	     Gulf War Syndrome					 
	     Persian Gulf War					 
	     Iraq						 

******************************************************************
** This file contains an ASCII representation of the text of a  **
** GAO Product.                                                 **
**                                                              **
** No attempt has been made to display graphic images, although **
** figure captions are reproduced.  Tables are included, but    **
** may not resemble those in the printed version.               **
**                                                              **
** Please see the PDF (Portable Document Format) file, when     **
** available, for a complete electronic file of the printed     **
** document's contents.                                         **
**                                                              **
******************************************************************
GAO-04-159

United States General Accounting Office

GAO

                       Report to Congressional Requesters

June 2004

                                    GULF WAR
                                   ILLNESSES

  DOD's Conclusions about U.S. Troops' Exposure Cannot Be Adequately Supported

GAO-04-159

June 2004

GULF WAR ILLNESSES

DOD's Conclusions about U.S. Troops' Exposure Cannot Be Adequately Supported

DOD's and MOD's conclusions about troops' exposure to CW agents, based on
DOD and CIA plume modeling, cannot be adequately supported. The models
were not fully developed for analyzing long-range dispersion of CW agents
as an environmental hazard. The modeling assumptions as to source term
data-quantity and purity of the agent-were inaccurate because they were
uncertain, incomplete, and nonvalidated.

The plume heights used in the modeling were underestimated, and so were
the hazard areas. Postwar field testing used to estimate the source term
did not realistically simulate the actual conditions of bombings or
demolitions. Finally, the results of all models-DOD and non-DOD
models-showed wide divergences as to plume size and path.

DOD's and VA's conclusions about no association between exposure to CW
agents and rates of hospitalization and mortality, based on two
epidemiological studies conducted and funded by DOD and VA, also cannot be
adequately supported because of study weaknesses. In both studies, flawed
criteria-DOD's plume model and DOD's estimation of potentially exposed
troops based on this model-were used to determine exposure. This may have
resulted in large-scale misclassification.

Troops under the path of the plume were classified as exposed; those not
under the path, as not exposed. But troops classified as not exposed under
one DOD model could be classified as exposed under another DOD model.
Under non-DOD models, however, a larger number of troops could be
classified as exposed. Finally, as an outcome measure, hospitalization
rate failed to capture the types of chronic illnesses that Gulf War
veterans report but that typically do not lead to hospitalization

Highlights of GAO-04-159, a report to Congressional Requesters

Since the end of the Gulf War in 1991, many of the approximately 700,000
U.S. veterans have experienced undiagnosed illnesses. They attribute these
illnesses to exposure to chemical warfare (CW) agents in plumes-clouds
released from bombing of Iraqi sites. But in 2000, the Department of
Defense (DOD) estimated that of the 700,000 veterans, 101,752 troops were
potentially exposed. GAO was asked to evaluate the validity of DOD,
Department of Veterans Affairs (VA), and British Ministry of Defense (MOD)
conclusions about troops' exposure.

GAO recommends that the Secretary of Defense and the Secretary of Veterans
Affairs not use the plume-modeling data for any other epidemiological
studies of the 1991 Gulf War, since VA and DOD cannot know from the flawed
plume modeling who was and who was not exposed. VA concurred with GAO's
recommendation, but DOD did not concur.

GAO also recommends that the Secretary of Defense require no additional
plume modeling of Khamisiyah and other sites. DOD concurred with GAO's
recommendation.

The Central Intelligence (CIA) did not concur with the report, stating
that it could not complete its review of the draft report in the time
allotted.

                           The Plume-Modeling Process

                                  Source: GAO.

Plume hazard area

www.gao.gov/cgi-bin/getrpt?GAO-04-159.

To view the full product, including the scope and methodology, click on
the link above. For more information, contact Keith Rhodes at (202)
512-6412 or [email protected].

Contents

  Letter 1

Scope and Methodology 3
Results in Brief 4
Background 7
Modeling the Environmental and Health Effects of Fallout from

CW Agents 10
DOD's Conclusion about U.S. Troops' Exposure to CW Agents

Cannot Be Adequately Supported 12
Major Unresolved Issues Concerning Modeling 19
DOD's Modeling Efforts Were Flawed 30
The MOD Relied on U.S. Plume Modeling to Determine Its Troops'

Exposure 47
to CW Agents 47
Total U.S. Plume-Modeling Costs 47
DOD's and VA's Epidemiological Conclusions on CW Exposure and

Hospitalization and Mortality Rates Cannot Be Adequately

Supported 49
Conclusions 61
Recommendations for Executive Action 62
Agency Comments and Our Evaluation 62

Appendix I DOD's Chronology of Khamisiyah Modeling Events

Appendix II Power-Law Formula

Appendix III DOD's Model Divergences

Appendix IV Divergence and Wind Field Models 69

Appendix V Comments from the Department of Veterans Affairs 73

  Appendix VI	Comments from the Department of Defense 74
  GAO Comments 92

         Appendix VII Comments from the Central Intelligence Agency 109

Tables

Table 1: The Meteorological and Dispersion Models DOD Used to

Model Khamisiyah Table 2: Gulf War Suspected Chemical Weapon Sites Table
3: Detector and Agent Identification Systems Deployed by

Coalition Forces Reporting the Detection of CW Agents Table 4: Principal
Reported Detections of Chemical Agents in Saudi Arabia, January 17-23,
1991 Table 5: U.S. Direct Costs for Modeling Gulf War Illnesses

                                     12 22

                                       24

                                     26 48

Figures

Figure 1: Timeline of Events Following the U.S. Demolition of CW

Agents at Khamisiyah, 1991-2000 9 Figure 2: The Plume-Modeling Process 11
Figure 3: NOAA-11 Meteorological Imagery of Areas Occupied by

Coalition Forces, January 18-24, 1991 25 Figure 4: Meteorological
Satellite Photography, Iraq, January 19,

1991 28 Figure 5: Plume Height Trend Line by Weight of Explosive 31 Figure
6: Boundary Layer Characteristics 32 Figure 7: Three Types of Plume
Geometry 33 Figure 8: The Impact of Nocturnal Jets on a Plume at Higher

Altitudes 34 Figure 9: Lawrence Livermore National Laboratory Projection
39 Figure 10: DOD Composite Projection 41 Figure 11: DOD Composite
Projection and Lawrence Livermore

National Laboratory Projection 43 Figure 12: Validation Runs of Various
Models 44 Figure 13: Divergence in Models Used to Construct DOD and CIA

Composite Analyses 67 Figure 14: Divergence in DOD Muhammadiyat Models 68
Figure 15: Lawrence Livermore National Laboratory Diagnostic

Wind Model, Based on Observational Data 70 Figure 16: Lawrence Livermore
National Laboratory Diagnostic Wind Model, Based on ECMWF Projections 71
Figure 17: Wind Field Vector Model, Based on COAMPS 72

Abbreviations

ADPIC Atmospheric Dispersion by Particle-in-Cell
AFTAC Air Force Technical Applications Center
BSS body-system symptom
CAPS Clinician Administered Posttraumatic Stress
CDC Centers for Disease Control and Prevention
CES Combat Exposure Scale
CIA Central Intelligence Agency
COAMPS Coupled Ocean-Atmosphere-Mesoscale Prediction System
CW chemical warfare
DIA Defense Intelligence Agency
DMDC Defense Manpower Data Center
DOD Department of Defense
DOE Department of Energy
DTRA Defense Threat Reduction Agency
GDAS Global Data Assimilation System
HPAC Hazard Prediction and Assessment Capability
IDA Institute for Defense Analyses
LLNL Lawrence Livermore National Laboratory
MM5 Mesoscale Model Version 5
MOD Ministry of Defense
NOAA National Oceanic and Atmospheric Administration
NOGAPS Naval Operational Global Atmospheric Prediction System
NUSSE Non-Uniform Simple Surface Evaporation
OMEGA Operational Multiscale Environmental Model with Grid

Adaptivity PON paraoxonase/arylesterase enzyme PTSD Post-Traumatic Stress
Disorder RAMS Regional Atmospheric Modeling System SAIC Science
Applications International Corporation SCIPUFF Second-order Closure
Integrated Puff UNMOVIC United Nations Monitoring, Verification, and
Inspection

Commission UNSCOM United Nations Special Commission VA Department of
Veterans Affairs VLSTRACK Vapor, Liquid, and Solid Tracking WMO World
Meteorological Organization

This is a work of the U.S. government and is not subject to copyright
protection in the United States. It may be reproduced and distributed in
its entirety without further permission from GAO. However, because this
work may contain copyrighted images or other material, permission from the
copyright holder may be necessary if you wish to reproduce this material
separately.

United States General Accounting Office Washington, DC 20548

June 1, 2004

The Honorable Robert C. Byrd Ranking Minority Member Committee on
Appropriations United States Senate

The Honorable Christopher Shays

Chairman

Subcommittee on National Security, Emerging Threats, and International
Relations Committee on Government Reform House of Representatives

Many of the approximately 700,000 U.S. veterans of the Gulf War have
experienced undiagnosed illnesses since the war's end in 1991. Some
veterans fear they are suffering from chronic disabling conditions because
of exposure to chemical warfare (CW) agents, as well as vaccines,
pesticides, and other hazardous substances with known or suspected adverse
health effects. They believe that their exposure may have been caused by
the Coalition forces' bombing of several sites in Iraq, including storage
and production facilities for nuclear, biological, or chemical warfare
agents. DOD's estimates based on available bomb damage assessment during
the war are that 16 of the 21 sites that were bombed were destroyed. Many
U.S. and British troops were located near some of these sites. In
addition, in March 1991, after the end of the war, U.S. troops conducted
large-scale demolition operations to destroy munitions and facilities at
Khamisiyah, a forward-deployed site in Iraq. These munitions were later
found to have been filled with CW agents.

When the possible exposure of U.S. troops to low levels of CW agents first
became an issue, during summer 1993, the Department of Defense (DOD) and
the Central Intelligence Agency (CIA) concluded that no troops had been
exposed because (1) there were no forward-deployed CW agent munitions and
(2) plumes-clouds of CW agents released from the bombing that destroyed
the chemical facilities-could not have reached the troops.

This conclusion was maintained until June 1996, when DOD publicly
acknowledged that U.S. troops had destroyed stockpiles of chemical
munitions at Khamisiyah after the war. Khamisiyah was a large

ammunition storage depot that contained nearly 100 ammunition storage
bunkers covering 25 sq km. Earlier, in October 1991, United Nations
Special Commission (UNSCOM) inspectors had found evidence of munitions
containing CW agents at Khamisiyah. Specifically, among the nearly 100
bunkers at Khamisiyah, remnants of 122 mm rockets were identified at
Bunker 73. The rockets were found to have been filled with a combination
of sarin and cyclosarin nerve agents. Several hundred 122 mm rockets
containing the same nerve agents were also found at a pit area close to
Bunker 73. It was not until 1996 that UNSCOM conclusively determined that
CW agents were in Bunker 73.

Since DOD's 1996 recognition that the bombing and demolition of Iraqi
facilities during the war did result in the release of plumes, DOD has
conducted numerous investigations, studies, and analyses based on computer
modeling. DOD has sought to determine the potential hazard area-the path
of the plume-and the U.S. troops who may have been exposed to
contamination from the bombing and demolition of storage facilities
containing CW agents (see appendix. I).1 In June 1996, DOD estimated that
300 to 400 U.S. troops participated in the demolition of Khamisiyah Bunker
73. In August 1996, the CIA, from the results of its computer modeling,
stated that around Khamisiyah, an area as large as 25 km downwind and 8 km
wide could have been contaminated.2

In September 1996, DOD estimated that 5,000 troops had been within a 25 km
radius of Khamisiyah. In October 1996, DOD extended this radius to 50 km:
It estimated that 20,000 U.S. troops had been within this zone. In July
1997, from the first plume analyses, DOD estimated that 98,910 U.S. troops
had potentially been exposed. In 2000, additional analyses led DOD to
reestimate that 101,752 U.S. troops had potentially been exposed.

In response to your request, we evaluated how well conclusions-about the
extent of exposure of U.S. troops and the association between CW exposure
and troops' hospitalization and mortality rates-are supported by available
evidence. We presented our preliminary results to you in our

1Appendix I contains a detailed chronology of DOD's modeling events.

2Central Intelligence Agency, CIA Report on Intelligence Related to Gulf
War Illnesses (McLean, Va.: Aug. 2, 1996).

Scope and Methodology

testimony in June 2003.3 In this report, we present our final results.
Specifically, we have assessed the following:

1. 	How valid is the DOD and British Ministry of Defense (MOD)
conclusion-based on CIA and DOD plume-modeling results-about U.S. and
British troops' exposure to CW agents?

2. 	What were the total costs for the CIA's and DOD's various
plumemodeling efforts?

3. 	How valid are the DOD and Department of Veterans Affairs (VA)
conclusions from epidemiological studies, based on DOD's plumemodeling
results, that there was no association between CW exposure at Khamisiyah
and the troops' hospitalization and mortality rates?

To determine the validity of DOD's conclusion, based on its plumemodeling
analysis, that U.S. troops' exposure to CW agents was as low as it said it
was, we examined the meteorological and dispersion models DOD used to
model chemical agent releases from the U.S. demolition of Khamisiyah and
Coalition bombings of Al Muthanna and Muhammadiyat in Iraq during the Gulf
War. We evaluated the basis for technical and operational assumptions DOD
made in the models and the specific data and information on source term,
meteorological conditions, and other key parameters used for modeling
chemical releases from Iraqi sites. We also evaluated the efforts of the
CIA and DOD to collect and develop data on source term and other key
parameters for the modeling.

We interviewed DOD and CIA modelers and officials involved with the
modeling and obtained documents and reports from DOD's Deployment Health
Support Directorate. We also interviewed and received documents from
Department of Energy (DOE) officials who were involved with the modeling
at DOE's Lawrence Livermore National Laboratory (LLNL). In addition, we
interviewed officials and obtained documents from the

o  	Institute for Defense Analyses (IDA) concerning the IDA expert panel
assessment of CIA's modeling of Khamisiyah,

3U.S. General Accounting Office, Gulf War Illnesses: Preliminary
Assessment of DOD Plume Modeling for U.S. Troops' Exposure to Chemical
Agents, GAO-03-833T (Washington, D.C.: June 2, 2003). www.gao.gov.

o  	U.S. Army Dugway Proving Ground in Utah to determine how CW agents in
Iraq's rockets might have been released during the Khamisiyah pit area
demolitions,

o  	U.S. Army Center for Health Promotion and Preventive Medicine to
determine how specific troop unit exposures were identified, and

o  	United Nations Monitoring, Verification, and Inspection Commission
(UNMOVIC) to obtain information on source term from UNSCOM's analyses and
investigations on Khamisiyah, Al Muthanna, and Muhammidiyat.

To determine the validity of DOD's and VA's conclusions-based on
epidemiological studies-that there was no association between Khamisiyah
exposure and the rates of hospitalization or mortality, we reviewed
published epidemiological studies in which hospitalization and mortality
among exposed and nonexposed U.S. troops were analyzed. We also
interviewed the study authors and researchers involved with the studies
and examined the Gulf War Khamisiyah population databases DOD provided in
support of these studies. We interviewed Veterans Benefits Administration
officials and obtained documents and reports on their analysis of DOD's
population databases. We did not examine whether plume modeling data were
being used by VA to determine eligibility for treatment or compensation.

To identify total costs associated with modeling and analysis of CW agent
releases during the Gulf War, we interviewed officials and collected cost
data from various DOD agencies and contractors who supported the modeling
efforts. However, total costs incurred could not be determined because DOD
agencies and other organizations could provide only direct costs, not
their indirect costs, associated with the modeling.

To determine the extent of British troops' exposure to CW agent releases
during the Gulf War, we interviewed MOD officials in London and Porton
Down, and we reviewed MOD reports concerning the effect of exposure to CW
agent releases on British forces.

We conducted our work from May 2002 through May 2004 in accordance with
generally accepted government auditing standards.

Results in Brief	DOD's and MOD's conclusions, based on DOD's plume
modeling, about their troops' exposure to CW agents cannot be adequately
supported.

Given the inherent weaknesses associated with the specific models they
used and the lack of accurate and appropriate meteorological and source
term data in their analyses, we found five major reasons to question their
conclusions. First, the models were not fully developed for analyzing
longrange dispersion of CW agents as an environmental hazard. Second,
assumptions regarding source term data used in the modeling-such as the
quantity and purity of the agent-were inaccurate, since they were based on
(1) uncertain and incomplete information and (2) data that were not
validated. Third, the plume heights from the Gulf War bombings were
underestimated in DOD's models. Fourth, postwar field testing at the U.S.
Army Dugway Proving Ground to estimate the source term data did not
reliably simulate the actual conditions of either the bombings or the
demolition at Khamisiyah. Fifth, there is a wide divergence in results
among the individual models DOD selected, as well as in the DOD and
non-DOD models, with regard to the size and path of the plume and the
extent to which troops were exposed. Therefore, given these inherent
weaknesses, DOD and MOD cannot know which troops were and which troops
were not exposed.

The total costs for the various plume-modeling efforts to analyze the
potential exposure of U.S. troops-from the demolition at Khamisiyah and
the bombing of several sites in Iraq-cannot be estimated. DOD
organizations and other entities involved with the plume-modeling efforts
could provide only their direct costs (that is, contractors' costs), which
totaled about $13.7 million. However, this amount does not include an
estimate of the considerable indirect costs associated with the salaries
of DOD, VA, and contractors' staff or costs of facilities, travel, and
equipment. We requested, but DOD could not provide, this estimate. In
addition, the CIA would not provide direct and indirect costs for Gulf War
plume modeling because, in its view, our request constituted oversight of
an intelligence matter and is beyond our scope of authority. The CIA's
contractor, the Science Applications International Corporation (SAIC),
also did not respond to our request.

DOD's and VA's conclusions-that there was no association between exposure
to CW agents at Khamisiyah and U.S. troops' rates of hospitalization and
mortality-also cannot be adequately supported. DOD and VA based their
conclusions on two epidemiological studies they

funded, one conducted by DOD researchers, the other by VA researchers.4 In
both of these studies, flawed criteria were used to determine the troops'
exposure. For example, the studies' criteria were based on (1) the DOD
plume model of exposure from postwar demolition of the Khamisiyah
munitions depot and (2) DOD's estimation, using this modeling, of which
troops might have been under the path of the plume. Troops under the path
of the plume were classified as exposed, those not under the path as
nonexposed. However, troops classified as nonexposed under one DOD model
could be classified as exposed under another DOD model, thereby
confounding the results. In the DOD models, a small area was identified as
being under the path of the plume, resulting in a small number of troops
identified as exposed. But in a model DOD used from LLNL, for example, a
much larger area was identified under the path of the plume, resulting in
a large number of troops exposed; in addition, these exposed troops
included different troops from those in the DOD models.

These flaws may have resulted in large-scale misclassification of the
exposure groups-that is, a number of exposed veterans may have been
classified as nonexposed, and a number of nonexposed veterans may have
been misclassified as exposed. In addition, in the hospitalization study,
the outcome measure-number of hospitalizations-failed to capture the
chronic illnesses of Gulf War veterans, which are commonly reported but
typically do not lead to hospitalization. Some published scientific
studies of exposure involving Gulf War veterans, studies of genetic
anomalies, and animal exposure-response studies suggest an association
between lowlevel exposure to CW agents and chronic illnesses.

We recommend that the Secretary of Defense and the Secretary of Veterans
Affairs not use the plume-modeling data for future epidemiological studies
of the 1991 Gulf War, since VA and DOD cannot know from the flawed plume
modeling who was and who was not exposed.

We recommend that the Secretary of Defense require no further plume
modeling of Khamisiyah and the other sites bombed during the 1991 Gulf War
in order to determine troops' exposure. Given the uncertainties in the

4G. C. Gray and others, "The Postwar Hospitalization Experience of Gulf
War Veterans Possibly Exposed to Chemical Munitions Destruction at
Khamisiyah, Iraq," American Journal of Epidemiology 150 (1999); H. K. Kang
and T.A. Bullman, "Mortality among U.S. Veterans of the Persian Gulf War:
7 Year Follow-Up, "American Journal of Epidemiology 154 (2001): 399-409.

Background

source term and meteorological data, additional modeling of the various
sites bombed would likely result in additional costs while still not
providing DOD with any definitive data on who was or was not exposed.

We obtained comments on a draft of this report from VA, DOD, and CIA. VA
concurred with our first recommendation (see appendix V). DOD did not
concur with our first recommendation, indicating that it apparently
represents a blanket prohibition of plume modeling in the future. The
intent of our recommendation is only directed at epidemiological studies
involving the DOD and CIA plume modeling data from the 1991 Gulf War and
not a blanket prohibition of plume modeling in the future (see appendix
VI). We have clarified the recommendation along these lines. DOD concurred
with our second recommendation, indicating that despite enhancements in
the modes, uncertainties will remain (see appendix VI). CIA did not concur
with the report, indicating that it could not complete their review in the
time allotted (see appendix VII).

In March 1991, after the Gulf War had ended, U.S. Army demolition units
destroyed munitions in Bunker 73 and in an open pit at the Khamisiyah
ammunition storage depot in southeastern Iraq. In October 1991, it was
discovered, from inspections conducted by UNSCOM that hundreds of 122 mm
rockets destroyed at Khamisiyah had contained nerve agents sarin and
cyclosarin.

U.S. and Coalition forces also bombed many other known or suspected Iraqi
CW research, materiel, storage, and production sites. According to the CIA
and DOD, Coalition aerial bombings resulted in damage to filled chemical
munitions at only two facilities in central Iraq-Al Muthanna Bunker 2 and
Muhammadiyat-and at one facility in southern Iraq-the Ukhaydir ammunition
storage depot. During these aerial bombings, munitions were damaged at Al
Muthanna containing an estimated 17 metric tons of sarin and cyclosarin
and at Muhammadiyat containing an estimated 2.9 metric tons of sarin and
cyclosarin and 15 metric tons of the chemical agent mustard.

According to DOD, the connection between the CW agent munitions UNSCOM
found at Khamisiyah and U.S. demolition operations there had not been
immediately made. However, in 1996, concerns that the Presidential
Advisory Committee on Gulf War Illnesses raised prompted

the CIA to examine this issue.5 In early 1996, after linking Khamisiyah to
the presence of CW agents, based on UNSCOM and other reporting, the CIA
contracted with SAIC to conduct an initial analysis and modeling of the
bombing of chemical munitions in Khamisiyah Bunker 73.

The CIA issued two reports. The first report, in August 1996, modeled a
potential release of agents from Khamisiyah Bunker 73 and from Al Muthanna
and Muhammadiyat.6 However, when modeling of the pit area at Khamisiyah
began, the CIA realized that the source term data-such as the quantity and
the purity of the agent and data on meteorological conditions, including
the wind and the weather patterns-were not available. Because of these
uncertainties, DOD and the CIA asked IDA to convene an independent panel
of experts to review the modeling. The IDA expert panel conducted its
review from November 1996 to February 1997 and (1) reported that the
initial analyses of the pit area were inadequate and (2) recommended
taking different approaches to improve the modeling.

The CIA issued its second report jointly with DOD in September 1997, this
time focusing on an open pit area at Khamisiyah.7 This report combined the
results of five different meteorological and dispersion models that the
CIA and DOD used to project the size and path of the plume from the
demolition operations. In 2000, DOD remodeled the Khamisiyah pit site,
using updated CIA source assessments and troop location data, which
changed the projected hazard area. The 1991-2000 timeline of major events
at Khamisiyah is shown in figure 1.

5The Presidential Advisory Committee on Gulf War Veterans' Illnesses was
established by Executive Order 12961 on May 26, 1995, to provide oversight
for Gulf War illness investigations; it terminated in November 1997.
http://www.gwvi.ncr.gov/ (Apr. 26, 2004).

6Central Intelligence Agency, CIA Report on Intelligence.

7Central Intelligence Agency and the Department of Defense, Modeling the
Chemical Warfare Agent Release at the Khamisiyah Pit (McLean, Va.: Sept.
1997).

Figure 1: Timeline of Events Following the U.S. Demolition of CW Agents at
                             Khamisiyah, 1991-2000

                                  Source: GAO.

Modeling the Environmental and Health Effects of Fallout from CW Agents

According to the CIA, modeling is the art and science of using
interconnected mathematical equations to predict the activities of an
actual event. In this case, modeling was used to determine the direction
and extent of the plume from CW agents. In environmental hazard modeling,
simulations recreate or predict the size and path (that is, the direction)
of the plume, including the potential hazard area, and potential exposure
levels are generated.

Modeling requires several components of accurate information:

o  	the characteristics or properties of the material that was released
and the rate of release (for example, quantity and purity, the vapor
pressure, the temperature at which the material burns, particle size, and
persistency and toxicity),

o  	temporal information (for example, whether chemical agent was
initially released during daylight hours, when it might rapidly disperse
into the surface air, or at night, when a different set of breakdown and
dispersion characteristics would pertain, depending on terrain, plume
height, and rate of agent degradation),

o  	data that drive meteorological models during the modeled period (for
example, temperature, humidity, barometric pressure, dew point, wind
velocity and direction at varying altitudes, and other related measures of
weather conditions), and

o  	data from global weather models to simulate large-scale weather
patterns and from regional and local weather models to simulate the
weather in the area of the chemical agent release and throughout the area
of dispersion.

In addition, in modeling, information is required on the potentially
exposed populations, animals, crops, and other assets that may be affected
by the agent's release.

The process flow for chemical plume modeling, to estimate the plume hazard
area, is shown in figure 2.

                      Figure 2: The Plume-Modeling Process

Plume hazard area

                             Meteorological Models

Various plumes during the 1991 Gulf War were modeled with global-scale
meteorological models, such as the National Centers for Environmental
Prediction Global Data Assimilation System (GDAS) and the Naval
Operational Global Atmospheric Prediction System (NOGAPS). Regional and
local weather models were also used, including the Coupled
Ocean-Atmosphere Mesoscale Prediction System (COAMPS), the Operational
Multiscale Environmental Model with Grid Adaptivity (OMEGA), and the
Mesoscale Model Version 5 (MM5). Outputs from global weather models are
mainly used as initial and boundary conditions for regional weather
models.

                                  Source: GAO.

Transport and Diffusion Models

Transport and diffusion models were also used during the Gulf War.8 These
models project both the path of a plume and the degree of potential hazard
posed by the agents. Dispersion models used during the Gulf War included
the Hazard Prediction and Assessment Capability (HPAC) model, along with
the HPAC component, the Second-order Closure Integrated Puff (SCIPUFF)
model; the Vapor, Liquid, and Solid Tracking (VLSTRACK) model; the
Non-Uniform Simple Surface Evaporation (NUSSE) model; and the Atmospheric
Dispersion by Particle-in-Cell (ADPIC) model.

The meteorological and dispersion models DOD used to model Khamisiyah are
shown in table 1.

8We use dispersion in this report to refer to both transport and diffusion
models.

Table 1: The Meteorological and Dispersion Models DOD Used to Model
Khamisiyah

                        Model type Developer or sponsor

Meteorological

COAMPS: Coupled Ocean-Atmosphere Mesoscale Prediction System U.S. Navy

MATHEW: Mass Consistent Wind Fielda	Department of Energy, Lawrence
Livermore National Laboratory

    MM5: Mesoscale Model Version 5 National Center for Atmospheric Research

NOGAPS: Naval Operational Global Atmospheric Prediction System, U.S. Navy

OMEGA: Operational Multiscale Environment Model with Grid Adaptivity
Defense Threat Reduction Agency

Dispersion

ADPIC: Atmospheric Dispersion by Particle-in-Cella	Department of Energy,
Lawrence Livermore National Laboratory

HPAC/SCIPUFF: Hazard Prediction and Assessment Capability/Second-Defense
Threat Reduction Agency order Closure Integrated Puff

NUSSE4: Non-Uniform Simple Surface Evaporation, Version 4 U.S. Army

VLSTRACK: Vapor, Liquid, Solid Tracking U.S. Navy

Source: Department of Defense, Technical Report: Modeling and Risk
Characterization of U.S. Demolition Operations at the Khamisiyah Pit
(Washington, D.C.: Apr. 2002).

DOD's Conclusion about U.S. Troops' Exposure to CW Agents Cannot Be
Adequately Supported

aInstitute of Defense Analysis used this model in its analysis of
Khamisiyah modeling, an analysis done at DOD's request.

DOD's conclusion about the extent of U.S. troops' exposure to CW agents
from the Gulf War, based on CIA and DOD plume models, cannot be adequately
supported because of uncertainty associated with the source term data and
meteorological data. The models are neither sufficiently certain nor
sufficiently precise to draw definitive conclusions about the size or path
(that is, the direction) of the plumes.

In particular, we found five reasons to question DOD's conclusion. First,
the models DOD selected were not fully developed for analyzing longrange
dispersion of CW agents as environmental hazards. Second, the assumptions
about the source term data used in the models were inaccurate. Third, in
most of the models, the plume height was significantly underestimated.
Fourth, postwar field testing at the U.S. Army Dugway Proving Ground in
Utah to estimate the source term data did not realistically simulate the
actual conditions of the demolition operations at Khamisiyah or the
effects of the bombings at any of the other sites in Iraq. And fifth,
there are wide divergences among the individual models DOD selected with
regard to the size and path of the plume and the extent to which troops
were exposed.

DOD's Models Were Not Fully Developed for Analyzing Long-Range
Environmental Hazards

COAMPS and OMEGA Meteorological Models

VLSTRACK and HPAC Dispersion Models

CIA and DOD officials selected several in-house models for plume models.
For Khamisiyah and the other Iraqi sites, DOD selected the COAMPS and
OMEGA meteorological models and the HPAC/SCIPUFF and VLSTRACK dispersion
models. However, at the time, these models were not fully developed for
long-range environmental hazards. In particular, they were inappropriate
for the long-range tracking of chemical agents.

Although COAMPS was accepted by the DOD peer review panel, OMEGA was still
under development. The DOD 1997 peer review panel that was reviewing the
models chosen for the 1997 Khamisiyah analysis reported problems with
OMEGA that resulted in major errors in its simulations. In the analyses of
Khamisiyah, as well as Al Muthanna, according to an IDA technical review
panel, OMEGA consistently underpredicted surface wind speeds by a factor
of 2 to 3, compared with actual observations collected at five World
Meteorological Organization (WMO) stations in the area.

VLSTRACK had been developed primarily to predict immediate health hazards
(that is, lethal or incapacitating effects) associated with troops' direct
exposure to CW agents. It was not developed for predicting longterm health
effects from indirect exposure to low levels of these agents. According to
modeling experts at the Naval Surface Warfare Center, the 2-month IDA
panel reanalysis and modeling was a developmental effort because existing
models did not have the capability to perform the required predictions.
Considerations of dispersion areas associated with low-level exposure to
CW agents released in Iraq, such as nerve and blister agents, required
these experts to make many extensions and modifications to some of the
methodology in VLSTRACK.

HPAC was developed jointly by the Defense Intelligence Agency (DIA) and
the Defense Special Weapons Agency-now the Defense Threat Reduction Agency
(DTRA)-and was specifically tailored for counterproliferation contingency
planning. In a 1998 scientific review and evaluation of the SCIPUFF
dispersion model (an integral component of HPAC), the National Oceanic and
Atmospheric Administration's (NOAA) Air Resources Laboratory reported that
SCIPUFF was better suited for short-range dispersion applications of about
10 km than for long-range transport modeling. One of HPAC's weaknesses is
that it does not provide definitive answers because of uncertainties about
agent transport, plume location, and weather.

It is also evident, from NOAA's review, that a group using VLSTRACK might
receive significantly divergent results from a group using HPAC. Further,
neither model is sufficiently accurate to permit a conclusion that the
path of the plume is confined to the hazard area that the model predicts.

In a September 1998 memorandum, the Deputy to the Secretary of Defense for
Counterproliferation and Chemical and Biological Defense cited a DOD panel
study team finding that the VLSTRACK and HPAC models generate hazard
predictions that differ significantly from each other: "This occurred even
when the source terms and weather inputs are as simple and as identical as
possible. In operational deployment, the average model user could obtain
different answers for the same threat."

A former Modeling and Simulation Adviser to the Deputy for
Counterproliferation and Chemical and Biological Defense told us that the
reliability of these models was of extreme concern. Also, in 1998, at IDA,
a panel study's initial comparison of the hazard-prediction models HPAC
and VLSTRACK documented substantial differences-by factors between 5 and
1,000-between the models for the prediction of the same event. The most
significant errors in the coding and the potential for misuse were found
in HPAC and its subcomponent models. Given these problems with the
analyses conducted up to 1998, HPAC could not be considered reliable.

In 1997, the director of NOAA's Air Resources Laboratory stated that DOD's
model selection resulted in the use of in-house models that were not well
known outside DOD. As to the meteorological models, he noted that three
mainstream mesoscale models were available, such as MM5 and Regional
Atmospheric Modeling System (RAMS), that were well accepted for deriving
site-specific flow conditions from large-scale meteorological information.
DOD used MM5 in both its 1997 and 2000 modeling.

DOD's Models Included Source Term Assumptions That Cannot Be Verified

The source term data DOD used in the modeling for sites at Khamisiyah, as
well as Al Muthanna and Muhammadiyat, contain significant unreliable
assumptions. DOD and the CIA based assumptions on field testing,
intelligence information, imagery, UNSCOM inspections, and Iraqi
declarations to UNSCOM. However, these assumptions were based on limited,
nonvalidated, and unconfirmed data concerning (1) the nature of the
Khamisiyah pit demolition, (2) meteorology, (3) agent purity, (4) amount
of agent released, and (5) other CW agent data. In addition, DOD and the
CIA excluded, for modeling, many other sites and potential hazards
associated with the destruction of binary chemical weapons, CW

agent precursor materials, and the potential release of CW agents and
toxic byproducts from other sites.9

During the initial modeling of the demolition of CW agent munitions at the
Khamisiyah pit, the CIA did not have the data-the number of rockets
present, agent purity, amount of agent released into the atmosphere, agent
reaction in an open-pit demolition, and prevailing meteorological
conditions-on how rockets with chemical warheads would be affected by
open-pit demolition compared with bunker demolition. A 1996-97 IDA panel
found substantial uncertainty about the number of damaged rockets that
might have released CW agents and how rapidly they were released. The
panel also found that the CIA and SAIC had used what were, "essentially,
guesses" to make up for the lack of data on how much agent was released
and over what period of time. For example, the CIA based the number of
rockets on the number known to have been there before the demolition and
the number UNSCOM found during inspections. But, according to the IDA
panel, the difference between what was estimated and what UNSCOM found
varied by a factor of 5 or 6.

Meteorological data at Khamisiyah were lacking because relatively few
observations had been made, according to DOD modelers, and those that had
been made were far from the site. This lack of meteorological observation
applied also to the modeling of other sites. Observations were few because
Iraq stopped reporting weather station measurement information to WMO in
1981. As a result, data on the meteorological conditions in Iraq during
the Gulf War were not available.

At Khamisiyah, the only data available were for the surface wind
observation site, 80 to 90 km away, and for the upper atmosphere, about
200 km away. At other sites modeled, the nearest data were at even greater
distances. The IDA panel recognized that wind patterns could contain areas
of bifurcation-lines where winds move in one direction on one side and in
another direction on another side-which migrate over time and are
different at varying altitudes.

Assumptions about the purity of the CW agents sarin and cyclosarin
established for Khamisiyah, Al Muthanna, Muhammadiyat, and Ukhaydir
differed widely. In each case, agent purity was a key factor in the CIA
and

9A binary weapon mixes two less toxic materials to create a toxic nerve
agent within the weapon when it is fired or dropped.

DOD methodology for determining the amount of agent released. For example,
for modeling purposes, 10 tons of agents with a purity of 18 percent would
be represented as 1.8 tons of agent. The CIA relied on UNSCOM reporting on
the amount of CW agents Iraq produced. But to establish these rates,
UNSCOM relied primarily on Iraqi declarations and Iraqi production
records, other UNSCOM testing, and assumptions about the extent of agent
degradation.

For example, according to Iraqi production records UNSCOM obtained, the
agent purity at Khamisiyah in early January 1991 was about 55 percent. The
agent degraded to 10 percent purity by October 1991, when laboratory
analysis had been completed on samples taken by UNSCOM from one of the
rockets. On the basis of the initial sample purity and the projected
degradation rate for sarin and cyclosarin, the CIA assessed that the ratio
of sarin to cyclosarin when the munitions were blown up, in March 1991,
was the same as that sampled in October 1991-3:1. According to the CIA,
"assuming a conservative exponential degradation" of the sarin and
cyclosarin, the purity on the date of demolition, 2 months after
production, was calculated at about 50 percent.

At Al Muthanna, where sarin was stored in a bunker, the CIA estimated that
it had deteriorated to approximately 18 percent purity by early February
1991, when Bunker 2 was destroyed, leaving the equivalent of about 1,600
kg (1.6 metric tons) of viable sarin. Iraqi records recovered by UNSCOM
inspectors suggested that the agent in Bunker 2 was from 1988 production
runs. UNMOVIC confirmed that in UNSCOM testing of other sarin samples,
produced during 1998, the maximum purity of agent had degraded to a range
of 18 percent to 2 percent by 1991.

According to a 2002 CIA assessment, only 2.5 percent of sarin was released
from the demolition of rockets in Bunker 73 at Khamisiyah.10 This
assessment was based on comparisons with U.S. testing in the 1960s at
Black Hills, South Dakota, on sarin-filled rockets.11 The CIA considered
the 2.5 percent estimate conservative because agent-heating conditions
were

10Central Intelligence Agency, Intelligence Update: Chemical Warfare Agent
Issues during the Persian Gulf War (McLean, Va.: Apr. 2002).

11U.S. Army Dugway Proving Ground, Hazard Classification Tests of Igloo
Storage of GBand VX-Filled M55 Rockets, RTD&E Project 1B650312D624,
USATECOM Project 5-3-013502, DPGDR C-505 (Dugway, Utah: Apr. 1966).

harsher in Bunker 73 than in the Black Hills tests. The 2002 CIA
assessment included this statement:

Far less agent (a maximum of 0.01 percent) would have been released in the
Al Muthanna bunker incident than the 2.5 percent indicated by 1960s US
field tests at Black Hills, South Dakota. The Black Hills tests used
simulated bunkers that had a wood slat, sand ceiling, and earth walls.
Those bunkers did not allow heat to build up as rapidly as in an Iraqi
bunker with thick reinforced concrete ceiling and walls. However, we have
chosen 10 kilograms as the release amount to account for unmodeled
releases from rocket flyouts or

12

transients at the beginning of the fire.

The CIA assessed that far less agent would have been released in the Al
Muthanna bunker because, based on U.S. field testing with simulated
bunkers, heat would build up rapidly in Iraqi bunkers made of thick
reinforced concrete ceiling and walls, destroying most of the agent.

DOD's Postwar Field Testing Did Not Realistically Simulate Actual
Conditions of Demolition Operations

During the 1960s Black Hills testing, rockets filled with sarin and VX
nerve agents were intentionally ignited by thermite grenades-alone and
with the addition of diesel fuel-as well as by fused initiation of the
burster explosive charge. According to the testing report, on the first
trial, the 11 crates of rockets were stacked 18 in. from the front wall on
the right side of the igloo. Because of safety considerations, a centrally
located burster was ignited with a 10-min. safety time fuse instead of an
electric blasting cap. Once detonated, there were 104 major explosions
during the 3 hrs. and 4 min. of observation. The maximum pressure recorded
during the initial detonation was 19 lbs./sq. in. above atmospheric
pressure. The maximum pressure during the trial was 621-oC, which was
recorded 15 min. after the ignition of the burster. The igloo sustained
three wide cracks in the arch and two large holes on either side of the
door. The only major debris found outside the igloo was the partial body
of one M55 rocket, located approximately 200 ft. away. Because of the
damage to the igloo, the test was deemed an official disaster.

During a second trial, crated rockets, with bursters removed, were stacked
as in the first trial. Two thermite grenades were set on top of the crates
at the motor end and were ignited with a 10-min. safety time fuse. The
grenades functioned properly but failed to ignite the rocket motors or
crate material. One day later, 12 thermite grenades were placed in three

12Central Intelligence Agency, Intelligence Update, p. 62.

groups of four, at the motor end of the crates, and 5 gal. of diesel fuel
were poured over the 11 crates just before ignition with a safety fuse.
From 2 crates, 12 rocket motors were ignited and hit the rear wall of the
igloo, but the only damage found was a slight bulge in the bottom of the
igloo door. The igloo withstood the fire and contained the rockets.

According to a later DOD study, based on the analysis in the 1960s of
Dugway Proving Ground Trial C505, from a fire within an igloo containing
GB-filled M55 rockets, the estimated amount of agent released was assumed
to be 2.52 percent over a 60-min. interval.13 This appears to be the basis
for the conclusion that only approximately 2.5 percent of the agent would
have been released.

However, the bunkers at Al Muthanna and other locations in Iraq were not
deliberately set on fire with incendiary devices. They were targeted with
high-explosive weapons such as Tomahawk missiles and laser-guided and
nonguided bombs, which detonate and produce instantaneous and extreme
blast forces, as well as shock and pressure waves and heat. A high
explosive is one in which the speed of reaction (typically 5,000 to 8,000
m/s) is faster than the speed of sound in the explosive. High explosives
produce a shock wave along with gas, and the characteristic duration of a
high-explosive detonation is measured in microseconds (10-6 s). Further,
if an explosion is confined within a chamber or room, the gas pressure
increases rapidly to a sustained level and then decays by venting out.
Under these conditions, shock reflections occur and the overall effect can
be greater than that of the incident shock.14 While the CIA analysts gave
much credibility to blast heat, no consideration was given to either the
shock blast effects of the munitions or the higher altitude plumes
generated from the transfer of mass associated with the shock waves.
Further, the CIA stated that its conclusion was supported by UNSCOM's
physical inspections of Bunker 2. UNMOVIC, however, informed us that
UNSCOM had not physically inspected this bunker for safety reasons
relating to structural instability.

13U.S. Army Armament, Munitions, and Chemical Command, Chemical Research
Development and Engineering Center, Source Characteristics of a Fire
within an Igloo Filled with M55 Chemical Munitions, CRDEC-TR-87056 (Rock
Island, Ill.: May 1987).

14National Academy of Sciences, Commission on Engineering and Technical
Systems, Protecting Buildings from Bomb Damage: Transfer of Blast-Effects
Mitigation Technologies from Military to Civilian Applications
(Washington, D.C.: 1995), pp. 28-29.

At Muhammadiyat, DOD, using test data from Dugway Proving Ground, provided
details about how source term characterizations for agent released were
derived. However, the type and quantity of explosives used in the Dugway
testing-and, therefore, the resulting effects-are not comparable with the
type and quantity of munitions that were actually used at Muhammadiyat. At
Dugway Proving Ground, small explosive charges were placed on boxed
rockets; at Muhammadiyat, the munitions storage bunkers were targeted with
multiple high-explosive bombs. Agent purity at Muhammadiyat was estimated
at 15 percent. In addition, according to UNSCOM, there are many questions
as to the accuracy of Iraqi records for the number of munitions filled
with sarin at Muhammadiyat.

Major Unresolved The major unresolved issues concerning DOD's modeling
include

assumptions about (1) CIA's modeling of Khamisiyah Bunker 73; (2)Issues
Concerning repeated aerial bombings of storage facilities; (3) repeated
aerial bombings Modeling of other storage, as well as research and
production, facilities; (4) binary

munitions and combustion by-products; and (5) detection of CW agents.

CIA's Modeling Assumptions about Khamisiyah Bunker 73

In July 1996, the CIA briefed the Presidential Advisory Committee on Gulf
War Veterans' Illnesses on the results of its Bunker 73 model. In the
August 2, 1996, report, identifying its modeling assumptions, the CIA
concluded that any hazard area resulting from the demolition of Bunker 73
had moved east and northeast.15According to CIA reporting, Bunker 73 and a
number of other bunkers were destroyed on March 4, 1991. The pit,
warehouses, and most of the remaining bunkers were destroyed on March 10,
1991. The CIA revised some of the source term assumptions for Bunker 73 in
its 2002 report, but Bunker 73 was never remodeled.16 Among the CIA's more
significant assumptions in the 1996 modeling of the demolition of Bunker
73 were these:

o  	Bunker 73 contained approximately 1,060 rockets filled with sarin
(this figure was modified to 910 rockets in 2002).

o  Each rocket was filled with 8 kg of a 2:1 ratio of sarin to cyclosarin.

15Central Intelligence Agency, CIA Report on Intelligence. 16Central
Intelligence Agency, Intelligence Update.

o  	The demolition ejected an estimated 10 percent of the rockets from the
bunker.

o  	U.S. tests showed that heat from the explosion, as well as burning
motors and crates, degraded all but 2.5 percent of the agent in the
bunker.

o  Winds were light, northeast to east.

o  	The modeling did not include the effect of thermal energy released by
the simultaneous burning and detonation of the other 32 to 37 bunkers at
the site.

The models used to arrive at these conclusions, however, were not
identified in the 1996 CIA report. According to the SAIC analyst who
conducted this modeling, the models used, the NUSSE4 and DP2C, were
dispersion models. (They were among the models that the IDA expert panel
later evaluated as having been inadequate for assessing the pit.)

In addition, the potential for greater contamination than indicated by the
initial models exists, given (1) UNMOVIC's assertion that UNSCOM did not
physically inspect this bunker through 1996 for safety reasons related to
contamination and structural instability and given DOD's conclusion that
all but 2.5 percent of the agent was degraded and (2) the lack of
correlation between the igloo testing studies and actual events.

Assumptions about Repeated Aerial Bombings of Storage Facilities

DOD reported that the Al Muthanna storage, as well as research and
production, facilities for CW agents was repeatedly attacked. Despite its
repeated bombing, however, on only one occasion did the CIA and DOD
express any concern about agent release. According to DOD analysis of the
destruction of Bunker 2 at Al Muthanna on February 8, 1991, no trace
residue from CW agents was detected in or around the bunker during UNSCOM
inspections. However, UNMOVIC told us that UNSCOM never inspected this
bunker for safety reasons. A low-level vapor hazard, however, probably
emanating from damaged and leaking 122 mm rockets stored in the open, was
detected around some of the other facilities. UNSCOM also reported
contamination in several other facilities at Al Muthanna that produced CW
agents. These observations suggest that additional releases may have
resulted from the repeated attacks at this site. However, DOD did not
analyze or model additional direct or incidental destruction from the
numerous bombings at the site.

There were 17 discrete Coalition aerial bombings of the Muhammadiyat
munitions storage facility. For Muhammadiyat models, we identified two
issues: time for agent dissipation and number of aerial bombings. For
agent dissipation, DOD assumed that agent concentration would have been
reduced to miniscule amounts during the first 24 hours. Therefore, for
this model, DOD chose a duration of 24 hours.

For number of aerial bombings, DOD made the assumption that all sarin at
Muhammadiyat was released at one time; therefore, DOD modeled each aerial
bombing as if it were the only bombing that caused a release. According to
DOD, each model produced a freeze frame of the largest hazard area. The
models suggest that the hazard area grows until it reaches its maximum
size, which is about 10 to 12 hours after the release. However, in these
models, DOD focused on a single bombing, failing to consider the
cumulative effects of exposure resulting from multiple aerial bombings.

Assumptions about Repeated Aerial Bombings of Other Storage, as Well as
Research and Production, Facilities

CW agents could have been released from a number of suspected storage
sites targeted for Coalition bombing, in addition to Al Muthanna and
Muhammadiyat. As shown in table 2, the intelligence community identified
numerous sites alleged to include research, production, or storage
facilities for CW agents. Another element of uncertainty in the modeling
site selection process results from the possibility that low-level
chemical agent exposures might have originated from bombing these other
sites during the Gulf War. Many of these sites were entirely discounted
for the purposes of the CIA's and DOD's models because the intelligence
community's assessment was that potential plumes from these sites could
not have reached U.S. troops.

Table 2: Gulf War Suspected Chemical Weapon Sites

       Site type         Site name and place          Site name and place     
       Facility         Al Muthanna (Samarra)     Fallujah III (Habbaniyah I) 
                     Fallujah I (Habbaniyah III)   Khamisiyah (Tall al Lahm)  
                     Fallujah II (Habbaniyah II)       Muhammadiyat (Qubaysah 
                                                               Storage Depot) 
Airbase, airfield  Al Bakr Airfield (Samarra     Murasana Airbase (H3 NW   
                            East Airfield)                 Airfield)          
                      Al Taba'at Airstrip (H3 SW  Qadisiyah Airbase (Al Asad  
                              Airfield)                    Airfield)          
                         Al Taqaddum Airfield       Qayyarah West Airfield    
                         Al Tuz Airfield (Tuz        Saddam Airbase (Qayyarah 
                          Khurmatu Airfield)                   West Airfield) 
                         Al Walid Airbase (H3           Tallil Airfield       
                              Airfield)           
                             K-2 Airfield         Tammuz Airbase (Al Taqaddum 
                                                                    Airfield) 

         Kirkuk Airfield Ubaydah Bin al Jarrah Airfield Mosul Airfield

Ammo Ad Diwaniyah Ammo Depot Kirkuk Ammo Depot West

Al Fallujah Ammo Depot South 	Qabatiyah Ammo Storage (Wadi al Jassiyah
Ammo Storage)

An Nasiriyah Ammo Storage Depot SW Qayyarah West Ammo Storage Depot

         Ash Shuaybah Ammo Storage Depot Tikrit Ammo Depot (Salahadin)

Baghdad Ammo Depot Taji Ukhaider (Karbala Depot and Ammo Storage)

Other Al Qaim Superphosphate Fertilizer Plant 	Fallujah Chem Proving Gnd
(Habbaniyah CW Training Center)

Dujayl/Awarah (Sumaykah SSM Support Facility SE)

Source: Central Intelligence Agency, CIA Report on Intelligence Related to
Gulf War Illnesses (McLean, Va.: Aug. 2, 1996).

Assumptions about Binary During our interview with UNSCOM and UNMOVIC
officials, we were

Munitions and Combustion informed that Iraq had extensively deployed
binary CW munitions

By-Products 	components at a number of sites within range of the theater
of operation. The chemical munitions were deployed unfilled but with
chemical components that when mixed would make the nerve agent sarin.
While there is no evidence that Iraq used these munitions offensively, it
remains unknown whether these materials were mixed intentionally or
inadvertently during combat engagements. In addition, Iraq declared that
823 metric tons of CW agent precursor materials, deployed throughout Iraq,
were destroyed during the Gulf War. The hazards associated with
environmental fallout from the destruction of these materials remain
unknown. CIA assessments and DOD modeling specifically did not address
potential exposure to binary CW agents, exposure to CW agent precursors,
or the hazardous environmental consequences associated with the

combustion by-products of sulfur mustard from the bombings at Muhammadiyat
and other sites.17

Assumption about the Detection of CW Agents

Credible scientific evidence suggests that the reported detections of CW
agents were reliable. These detections, during the early phases of the air
war, included the use of an array of scientific methods. The CIA and DOD
assert that the various detections are not valid because the source of the
agents cannot be detected. Available evidence, however, can explain the
detections and suggests that troop exposure may be far more widespread
than that projected from the CIA and DOD modeling of the Khamisiyah
release alone.

Coalition forces possessed a diverse array of CW detection and
identification equipment. Czechoslovak chemical detection troops also used
Warsaw Pact equipment, such as the GSP-1, GSP-11, and Czech mobile
chemical agent laboratories. This broad array of equipment included
various technologies to detect and confirm the presence of CW agents, as
well as to identify specific agents. The different physical principles
that were employed included wet chemistry, mass spectrometry, ion mobility
spectrometry, chemical reaction, biochemical enzyme reactivity, flame
photometry, and ionization. Detector (that is, sensor) and agent
identification systems that Coalition forces deployed, reporting the
detection of CW agents, are shown in table 3.

17Precursors are chemicals and other materials used in producing CW agents

Table 3: Detector and Agent Identification Systems Deployed by Coalition
Forces Reporting the Detection of CW Agents

Nation System Chemical agent Sensitivity Method or technology
Czech Automatic Nerve Agent G-and V-series nerve agents 0.05 mg/m3 Air
sampling, biochemical enzyme

Republic

Detector Alarm GSP-11

                (cholinesterase reactivity) 0.1 mg/m3 2.0 mg/m3

          Mobile Laboratory Most CW agents  Agent           Field portable    
          CHP-71,                           identification  chemical agent    
            PCHL-90, and                                          laboratory, 
               PCHL-90                        through wet            chemical 
                                                                    reagents, 
                                            chemistry         wet chemistry   
                                            analysis            analysis      
France    Detalac F1      G-and V-series  Not available  Biochemical       
                               nerve agents                 enzyme detector   
          Chemical                                                   Chemical 
          Detection Control   Tabun/sarin                         biochemical 
                                (GA/GB)         1 mg/m3              detector 
             Kit (TDCC)        Hydrogen        350 mg/m3    
                             cyanide (AC)                   
                               Cyanogen       2000 mg/m3    
                             chloride (CK)                  
          Chemical                                                            
          Detection Unit      Sarin/soman    Not available        Flame       
          for                   (GB/GD)                       spectrometry
          Fixed                                             
          Installations                                     
          (ADLIF)                                           

            Chemical Agent  G-and                           Ion mobility      
United      Monitor      V-series       0.1 mg/m3        spectrometry      
                            nerve agents               
Kingdom                     Blister     2.0 mg/m3   (quantitative feature) 
                             agents (H)                
             Nerve Agent      G-series                     Biochemical enzyme 
             Immobilized    nerve agents   0.05mg/m3                 detector 
           Enzyme Alarm and   V-series    0.005 mg/m3     (cholinesterase     
           Detector         nerve agents                    reactivity)       
               (NAIAD)                                 
                MARK I        G-series        Not        Biochemical chemical 
                            nerve agents   available               reactivity 
                               Blister                 
                             agents (H)                

United    M8A1 Alarm    G-series nerve 0.1 mg/m3         Ionization        
                               agents                
States                  V-series nerve 0.2 mg/m3      Automatic alarm      
                               agents                
            M8 Paper; M9   G-and V-series                                     
               Paper       nerve agents     Yes/no     Chemical reactivity
                           Blister agents   Yes/no     Color interpretation   
                                (H)                  
          Chemical Agent   G-series nerve                  Biochemical enzyme 
          Detector Kit         agents     0.05 mg/m3                 detector 
          M256 and M256A1  V-series nerve 0.15 mg/m3 (cholin-esterase         
                               agents                reactivity)              
                           Blister agents 3.0 mg/m3    Chemical reactivity    
                                (H)                  
           Chemical Agent  G-and V-series            
              Monitor      nerve agents              
                           Blister agents            
                                (H)                  

Ion mobility spectrometry (quantitative feature)

            MM1 FOX NBC vehicle G-series nerve agents Several mg/m3

Quadrapole gas chromatography
C-mass spectrometryS
Full GC-MS

60 other preprogrammed agent spectra

M8A1 (M43) ionization backup unit (early warning)

Sources: DOD FM Series 3; Chemical Research, Engineering, and Development
Command, Aberdeen Proving Grounds, Md.; manufacturers; Jane's NBC
Protection Equipment (1991-92 and 1995-96); CIA, Chemical Warfare Agent
Issues during the 1991 Persian Gulf War (Apr. 2002); video footage of
Czech chemical detection unit activity during the 1991 Gulf War; Czech
Chemical Warfare Report, Intelligence Assessment of Chemical and
Biological Warfare in the Gulf, prepared by U.S. Army Foreign Science and
Technology Center and DIA for the Defense Science Board investigating the
Desert Storm Syndrome.

Shortly after the air war began, the period between January 18 and January
23, 1991, was marked by a stationary frontal pattern and the development
of low-level cloud and fog activity directly over the areas occupied by
Coalition forces. Figure 3, based on NOAA-11 AVHRR 1B meteorological
imagery, shows contemporaneous visual evidence of this anomaly.

Figure 3: NOAA-11 Meteorological Imagery of Areas Occupied by Coalition
Forces, January 18-24, 1991

Source: NOAA.

Note: NOAA-11 AVHRR 1B meteorological imagery, processed by Earth
Satellite Corporation, Rockville, Md., was obtained from the National
Center for Atmospheric Research, Ashville, N.C.

During this prolonged period, a significant number of reported CW agent
detections were reported in northern Saudi Arabia where Coalition forces
were deployed (see table 4).

Table 4: Principal Reported Detections of Chemical Agents in Saudi Arabia,
                              January 17-23, 1991

    Date   Nation and unit    Place        Agent       Method or technology   
                                         detected    
Jan. 17 United States   NW Hafir Al Unknown nerve Ionization, biochemical  
           2/5TH SFG       Batin           agent     reaction; ion            
                                                     mobility spectrometry;   
                                                     M8A1, M256,              
                                                               CAM            
Jan. 19 Czech Republic  N Hafir Al                 Biochemical reactivity; 
           Chemical           Batin     Sarin (GB)             wet chemistry; 
           Detection Unit                             GSP-1(11), mobile lab   
                           NE Hafir Al                Biochemical reactivity; 
                           Batin        Sarin (GB)             wet chemistry; 
                                                      GSP-1(11), mobile lab   
                           King Khalid Unknown nerve  Biochemical reactivity; 
                           Military        agent               wet chemistry; 

                       City (KKMC) GSP-1(11), mobile lab

          France        30 km from  Unknown nerve    Biochemical reactivity   
                        KKMC            agent      
      Czech Republic    30 km from Confirm French                             
         Chemical       KKMC          detection     Wet chemistry mobile lab
      Detection Unit                               
                           KKMC    Sulfur mustard   Wet chemistry mobile lab  
                                        (HD)       
                                   Unknown blister Chemical reactivity; ion   
      United Kingdom      Jubayl       (after      mobility                   
                                     unexplained     spectrometry; M-9, CAM   
                                     explosions)   
    United States 24th             Unknown blister Chemical reactivity; M-256 
          Naval           Jubayl       (after                     (2/3 tests) 
Reserve Construction              unexplained   
                                     explosions)   
Battalion (Seabees)                             

Jan. 20	Czech Republic Chemical Near KKMC Sulfur mustard (HD) for Wet
chemistry mobile lab Detection Unit 2 hours

French sector Sarin (GB)/tabun (GA) Biochemical reactivity; wet chemistry
KKMC mobile lab

               France      Near KKMC    Unknown nerve        Biochemical      
                                            agent            reactivity       
            United States                                    Ionization,      
              800th MP     NW of KKMC   Unknown nerve        biochemical      
                                            agent            reactivity;      
                 BDE                                         M8A1, M256       
                             Dharan   Unknown nerve          Biochemical      
           United Kingdom             agent (after      reactivity (separate  
                                        SCUD attack)      devices); NAIAD,    
                                                                MARKI         
                                                                  Biochemical 
Jan. 21 Czech Republic     KKMC     Sarin (GB)/tabun       reactivity; wet 
           Chemical                        (GA), sulfur             chemistry 
           Detection Unit               mustard (HD)         mobile lab       
               France         KKMC      Unknown nerve        Biochemical      
                                            agent            reactivity       
                                                        Chemical or           
                              KKMC                      biochemical           
                                      Unknown CW agent  reactivity            

Jan. 22 United States RAFHA Unknown nerve agent 	Ionization, biochemical
reactivity; M8A1, M256

Jan. 23	Czech Republic Chemical KKMC Unknown CW agent Wet chemistry mobile
lab Detection Unit

Near KKMC Patch of sulfur mustard (HD) Wet chemistry mobile lab

Sources: CENTCOM CCJ3-X log (partially declassified 1995), Defense Science
Advisory Board report (June 1994), Czech government reports on detection
activity during the Gulf War, and declassified Defense Intelligence Agency
reports on chemical detection activity.

This period of reporting on chemical agent detection is attributable to
the use of many of the instruments and methodologies cited in tables 3 and
4 and coincides directly with the initial Coalition bombings of confirmed
and suspected Iraqi CW research, production, and storage facilities during
January 17-24, 1991.

That these detections were reported raises continued and unresolved
concern that U.S. troop exposures may have been more frequent and more
widespread than the CIA and DOD believe. In addition, these exposures may
have involved different populations than can be assumed from the limited
simulations performed with the CIA and DOD models we refer to in this
report. For example, figure 4 shows NOAA-11 collected infrared satellite
imagery data for January 19, 1991, that NOAA published, indicating a line
of instability and an inversion coincident with, and possibly
precipitating, the detection of CW agents in the vicinity of Hafir al
Batin and King Khalid Military City, reported by Czech chemical detection
units and other Coalition units in the region (as shown in table 4).

     Figure 4: Meteorological Satellite Photography, Iraq, January 19, 1991

     Muhammadiyat  Fallujah               Muthanna Baghdad                   
                                          Ukhaider                           
                                (Karbala Depot and                           
                                          Storage)                           
                                                                Khamisiyah   
                      Rafha                                                  
                                                            Hafir al Batin   
                                                               King Khalid   
                                                             Military City   

                       Source: GAO analysis of NOAA data.

The lighter areas in the image represent warmer air, the darker areas
colder air. In the image, a collision is shown between cold and warm air
masses, as well as what appears to be evidence of inversion activity. That
is, colder upper air moves below the warmer air, which is normally closer
to the surface, possibly precipitating the detections.

The assessment by the CIA that the detections of the Czech chemical
detection units are not credible-despite earlier DOD technical assessments
that Czech detections of sarin and mustard agent on two occasions were
credible-are unsound because these assessments do not refute the
underlying scientific evidence, according to DOD and DIA

18

experts.

The presence of both sarin and mustard agents at subacute levels, and in
such close proximity, could reasonably be explained as the result of
fallout from Coalition bombings of Iraqi weapons facilities and storage
bunkers. This possibility is supported not only by the atmospheric data
but also by observing that the Czech nerve agent detections were made,
according to DIA reporting, by multiple teams over a range of 20 to 50 km,
within 30 minutes of one another, during a strong weather inversion.19
According to U.S. reporting, during the period immediately before the
detections, the following known Iraqi CW agent research, production, and
storage sites (with geocoordinates and dates) were bombed:

o  Al Muthanna (3350 N, 04348 E, January 17, 1991),

o  Al Nasiriyah (3058 N, 04611 E, January 17, 1991), and

o  Qabatiyah (3353 N, 04239 E, January 19, 1991).

In addition, production sites for CW agent precursors were bombed during
this period.20 The potential remains that sites that may have been
incorrectly assessed as not containing CW agents were also bombed

18See Czech chemical warfare report, Intelligence Assessment of Chemical
and Biological Warfare in the Gulf, prepared by the U.S. Army Foreign
Science and Technology Center and DIA for the Defense Science Board
investigating Desert Storm Syndrome.

19Defense Intelligence Agency, Detection of Chemical Warfare Agents by
Czechoslovak Unit during Desert Storm, Part III (U), IIR 6 284 0008 94
(Washington, D.C.: Oct. 14, 1993).

20U.S. Army Operations Group, UNSCOM, Inspection of Chemical Warfare
Facilities, IIR 2 201 0022 92 (Washington, D.C.: Oct. 3, 1991,
declassified 1995).

during this period. These activities may have resulted in additional
releases of CW agents.

In this section, we discuss our findings about DOD's modeling efforts in
more detail, explaining the ways in which we believe them to be lacking in
credibility.

DOD's Modeling

Efforts Were Flawed

DOD's Models Significantly Underestimated Plume Height

Actual plume height might have been significantly higher than the height
DOD estimated from its models of demolition operations and bombings, given
the assumptions DOD's modeling used for Khamisiyah and the other Iraqi
storage facilities. The plume height estimates that the CIA provided for
demolition operations at the Khamisiyah pit were 0 to 100 meters. However,
neither the CIA nor DOD conducted testing to support estimated plume
height associated with the bombings of Al Muthanna, Muhammadiyat, or
Ukhaydir. According to DOD modelers, neither plume height nor any other
heat or blast effects associated with these bombings were calculated;
instead, these data were taken from DOD's Office of the Special Assistant
for Gulf War Illnesses.

In addition, according to a principal DTRA modeler, DOD's data on plume
height were inconsistent with other test data for the types of facilities
bombed. This expert cited test studies conducted at White Sands Proving
Ground demonstrating that plume height would range from 300 to 400 m. The
CIA maintains, however, that the plume occurred near ground level because
there was no altitude, burst of munitions, or any kinetic force such as
bomb blast to force the agent to become airborne initially. DOD maintains
that during the bombing of chemical agents, liquid agent absorbs the
energy of the blast, greatly reducing plume heights. We asked the CIA and
DOD to provide test data in support of their assertions but neither agency
provided any evidence.

Modeling experts from LLNL who participated in only the initial modeling
of the Khamisiyah pit site also said that they questioned the plume height
estimates. In a prewar analysis, LLNL projected that the plume immediately
following the bombing of Iraqi storage facilities for CW agents would be a
surface-based plume with a horizontal radius of 54 m (177 ft.) and a
height of 493 m (1,617 ft.). A 1969 Sandia National Laboratories empirical
study established a power-law formula for calculating plume heights
attributable to high-explosive detonations (see appendix II). By this
formula, a conventional MK-84 or GBU-24 bomb (containing 942.6 lb.

of high explosives) of the type used to bomb sites other than those at
Khamisiyah would generate a plume of approximately 421 m.

Impact by Weight of Explosive 	Figure 5 shows the trend line for a plume
height predicted on the basis of the formula for calculating plume height
resulting from detonating high explosives ranging from 100 to 2,000 lbs.
in weight.

Figure 5: Plume Height Trend Line by Weight of Explosive

Height in meters 600

500

400

300

200

100

0

100lb 500lb 1000lb 2000lb

Weight of high explosives Source: GAO.

According to DOD officials, the Sandia National Laboratories study is not
accurate or scientifically valid because it did not account for weather
effects. Further, they said, it based plume height on the detonation of
conventional explosives, but the liquid agent of chemical-filled munitions
would have absorbed much of the energy of the blast if these had been
bombed; therefore, plume height would have been greatly reduced. However,
DOD could not provide us with any data, studies, or testing of the
explosive aspects and buoyancy of chemical agents to corroborate these
observations.

At Muhammadiyat, DOD established a plume height of 0.5 m (roughly half the
bomb height) for nerve agent and a plume height of 1.0 m (roughly half the
median height of the various bomb stacks) for mustard agent destroyed at
this location. Moreover, according to one internal DOD memorandum, an
"initial cloud size" of 10 m, in both lateral and vertical directions, was
"arbitrarily" established. According to DOD, no effort was made to
validate these estimates by analyzing video images that showed

some of the plume data, such as (1) those taken from ground level at
Khamisiyah and (2) the available footage from the aircraft and munitions
used to bomb the other sites.

Impact by Wind Speed	Figure 6 shows that disparity in source term data for
plume height could lead to vastly divergent results as to how far the
plume travels and disperses. This observation is particularly relevant
during nighttime periods, when a stable nocturnal boundary layer emerges.

Figure 6: Boundary Layer Characteristics

Source: Roland B. Stull, An Introduction to Boundary Layer Meteorology
(Boston, MA: Kluwer Academic Publishers, 1988), p. 11. Used with kind
permission of Kluwer Academic Publishers.

Figure 6 shows the stable nocturnal boundary layer where winds often
accelerate to higher speeds, in a phenomenon referred to as the low-level
or nocturnal jet-that is, winds are aloft in the nighttime hours. At
altitudes on the order of 200 m above ground, winds may reach 10 m to 30 m
per second (22 to 67.5 mph) in the nocturnal jet. Higher plumes than those
DOD postulated, coupled with this phenomenon, could result in chemical
agents being transported rapidly until disturbed by turbulence or the
return of the mixed layer, some time after dawn. However, this possibility
was not taken into consideration in any of the modeling DOD performed.
Consequently, the models may have dramatically underestimated the extent
of plume coverage.

According to DOD, such winds are not known to be present in Iraq. DOD
confirmed, however, that it has no available data on prevailing wind
conditions in the region, over the varying terrain, or during the time
period in question to rule them out. DOD also stated that data are not
available for determining the presence of a low-level or nocturnal jet at
the time of the Khamisiyah demolition. However, DOD acknowledged
uncertainty as to whether a low-level jet was present on any specific date
or at any specific location.

Impact by Plume Geometry	According to plume geometry, the majority of
plume mass associated with high-explosive discharges is located toward
higher altitudes (see figure 7). This suggests that the majority of the
plume mass would move to higher altitudes to be transported by higher wind
speeds.

Figure 7: Three Types of Plume Geometry

Source: Department of Energy, Lawrence Livermore National Laboratory.

Similarly, the distribution of plume geometry may be affected by nocturnal
jets, as shown in figure 8.

Figure 8: The Impact of Nocturnal Jets on a Plume at Higher Altitudes

Source: Department of Energy, Lawrence Livermore National Laboratory.

Empirical studies and observed events tend to refute the assumptions with
which the CIA and DOD discounted the alternative assumption that the plume
was transported by low-level jets. First, empirical testing suggests that
the plume heights were much higher than postulated in the source term
data. Second, no massive casualties or effects were claimed, reported, or
observed in areas immediately surrounding the Iraqi CW research,
production, and storage facilities bombed by Coalition forces. In the
absence of an alternative explanation, acute effects should have been
observed in areas near the bombed sites.

Third, since many of the bombings were at night, the explosive effects-
coupled with higher-altitude plumes and a nocturnal boundary layer capable
of moving hazardous materials hundreds of miles-could easily account for
the phenomena reported above. These effects could also account for the
numerous reports of CW agents detected in sites occupied by U.S. and
Coalition forces. Fourth, these effects may account for reported nighttime
detections of low levels of CW agents, associated with
turbulence-resulting from aircraft-related sonic booms and incoming
missiles and artillery-mixing the upper-level and lower-level atmospheric
layers.

DOD's Field Testing Did Not Realistically Simulate Actual Conditions

Demolition and Bombing Simulations Were Not Realistic

Extrapolation Data Contained Acknowledged Limitations

To simulate the effects of demolition on chemical nerve agent stockpiles,
the CIA and DOD conducted postwar field testing at Dugway Proving Ground.
They explored what percentages of agent would be deposited on the ground
as a liquid, consumed by the demolition, and aerosolized. To obtain these
source term data for their models of Khamisiyah, the Dugway Proving Ground
testing center conducted seven field tests and two laboratory studies from
May 1997 through November 1999.

For field testing to be effective, conditions have to be as close to the
actual event as possible, but these tests did not provide more definitive
data for the CIA's and DOD's models. The tests did not realistically
simulate the conditions of the demolition of 122 mm chemical-filled
rockets in Khamisiyah. The simulations took place under conditions that
were not comparable with those at Khamisiyah. There were differences in
meteorological and soil conditions; the construction material of munitions
crates; rocket construction (including the use of concrete-filled pipes as
rocket replacements to provide inert filler to simulate larger stacks);
and the numbers of rockets, using far fewer rockets and, therefore,
explosive materials. Additionally, the tests used an agent stimulant whose
physical properties differed from those of the actual agent.

For example, of the 32 test rockets with simulant-filled warheads, a small
sample was used to conduct all seven field tests: five tests were
singlerocket demolitions and two involved multiple-rocket demolitions. One
multiple-rocket field test demolition used 9 functional rockets plus 3
dummy rockets; the other multiple-rocket field test used 19 functional
rockets and 5 dummy rockets. In contrast, at the Khamisiyah pit, stacks of
122 mm rockets were detonated, estimated at about 1,250 rockets. Dugway
officials acknowledged that detonating a larger number of rockets would
have made a significant difference to the testing. In addition, aerial
bombing with a heavy explosive load, such as had occurred at the sites
other than Khamisiyah, would have had a far greater effect than was
achieved with the Dugway testing.

According to DOD officials, SAIC developed projection data, in support of
the Dugway tests, to extrapolate the results to the larger stockpiles
identified at Khamisiyah. However, in our review of these data, we found
that the SAIC analysts acknowledged limitations to their study. They noted
in their report:

Testing Did Not Use Iraqi Soil and Wood

This is the first attempt to develop an understanding of these processes
and we lack sufficient data to either validate or completely calibrate our
models. The models in many cases have diverged from first principles
models and have become "engineering" models in the sense that explicit
calculations have been based on the observations of others and these have
not been well documented as yet. In some cases, it may be that further
calculations and/or tests will determine that for different configurations
than those calculated/tested, some of the understandings developed from
examining the existing data

21

may have to be modified.

Finally, SAIC acknowledged that calculations were performed in two
dimensions, stating that full, three-dimensional calculations were not
feasible.22

SAIC's conclusion was that the test explosions did not produce gases
likely to endure long enough to cook off-that is, ignite-a motor, even in
a large stack. But SAIC pointed out that "the real world often holds
surprises in chaotic situations of this sort." For example, if a rocket
motor were to ignite and burn in place, its energy might sensitize
adjacent motors or cause cook-off initiation: "these processes could walk
their way through the aisles [stacks] with minutes between dramatic
events."23 The potential for real world events like these points to the
inadequacy of attempting to extrapolate from the small-scale controlled
testing at Dugway Proving Ground to the large-scale and relatively
uncontrolled conditions at Khamisiyah.

According to DOD and CIA analysts, the type of soil and wood can have a
significant effect on the dispersion of the agent. Their estimates of the
evaporation and retention rates of the chemical agent spilled on the soil
may not be similar to what actually evaporated from and was retained in
the pit sand at Khamisiyah. Although Iraqi soil was available and used in
the laboratory testing, it was not used in field testing. Similarly,
assessments of DOD and CIA estimates of the amount of spilled agent
evaporated from and retained in wooden crates are uncertain. This is
because Dugway testing officials could not obtain actual wood from the
Khamisiyah pit site. The aged and possibly damp wood at Khamisiyah

21G. I. Kent and others, Interim Report: Calculations and Measurements in
Support of the Rocket Testing at Dugway Proving Ground (Dugway, Utah: June
17, 1997), p. 2.

22Kent and others, Interim Report: Calculations and Measurements, p. 2.

23G. I. Kent and others, Interim Report: Rocket Motor Ignition/Explosives
Issue (Dugway, Utah: June 17, 1997), p. 15.

might have absorbed less agent than the new wood used at Dugway. DOD and
CIA determined that only about 32 percent of the agent was released and
that most leaked into the soil and wood, with 18 percent of the leakage
becoming part of the plume (2 percent through aerosolization and 16
percent through evaporation).

Field tests were also conducted at a time of year and time of day
different from the actual Khamisiyah pit event. According to Dugway
officials, testing was done in May and in the early morning hours, when
drainage conditions prevail.24 When the March 10, 1991, U.S. demolition
operations took place at the Khamisiyah pit, it was late afternoon with a
mixed layer. The bombings of the other modeled storage and production
sites often took place during evening and nighttime hours, when a stable
nocturnal boundary layer emerges.

Despite the uncertainties in approximating the conditions at Khamisiyah,
DOD and the CIA used these data not only for modeling Khamisiyah but also
for modeling other sites. At the other sites, the CW munitions would have
been destroyed by aerial bombings with much greater quantities of
high-explosive charges and under significantly different meteorological
conditions.

All Models, Including DOD's Composite Model, Showed Divergent and
Unresolved Results

The models DOD used to predict the fallout from Khamisiyah and the other
sites showed great divergence, even with the same source term data. While
the models' divergences included plume size and path, DOD made no effort
to reconcile them (see appendix III). The IDA expert panel determined that
the results were so divergent that it would not be possible to choose the
most exposed areas or which U.S. troops might potentially have been
exposed. IDA therefore recommended a composite model, which DOD adopted.25
However, this approach masked differences in individual model projections
with respect to plume size and path.

24Drainage winds, also called mountain or gravity winds, are caused by the
cooling air along the slopes of a mountain. Drainage wind periods were
intentionally chosen for Dugway testing to minimize the dispersion of the
test materials to the surrounding areas. According to Dugway officials,
these winds are common to Dugway Proving Ground in the morning before the
development of the mixing layer.

25The composite approach DOD used is also known as the ensemble approach.

In addition, DOD chose not to include in the composite model the results
of the LLNL model, created at the IDA expert panel's request, which showed
a different and larger plume size and path than DOD's models showed. The
IDA panel regarded the LLNL model as less capable than other models
because it modeled atmospheric phenomena with less fidelity. Finally,
modeling of Khamisiyah that the Air Force Technical Applications Center
(AFTAC) had done also showed differences from DOD's composite model.

LLNL's Model 	To determine plume size and path, LLNL conducted analyses
using DOE's MATHEW meteorological model with the ADPIC dispersion model.
During LLNL's presentations to the IDA panel in November 1996 and February
1997, LLNL provided a 72-hour projection, assuming an instantaneous
release of the contents of 550 rockets containing sarin (see figure 9).
The results of LLNL's analyses show the plume covering an area extending
south southeast from the release point to the Persian Gulf, then turning
eastward at the Gulf coast, and then turning northeast over the Gulf and
extending northeastward across central Iran.

Figure 9: Lawrence Livermore National Laboratory Projection

Source: Department of Energy, Lawrence Livermore National Laboratory.

LLNL's modeling assessment showed that the 72-hour exposure projection-for
the instantaneous release of sarin from 550 rockets- covered a large
hazard area. According to LLNL, 2,255 sq km were covered

by agent concentration in excess of the dosage expected to cause a person
minimal effects.26 This area extended approximately 130 km south southeast
from the release point. Dosages in excess of the amount that would be
allowed for a worker exposed to sarin in the workplace, the occupational
limit, were predicted over 114,468 sq km-including Kuwait City, an area
approximately 200 km wide across the Persian Gulf, and the higher
elevations of the Zagros Mountain range of western Iran.27 The remaining
area was determined to be at the general population limit.28

DOD's Composite Model 	DOD's composite model was based on OMEGA and COAMPS
meteorological models and HPAC/SCIPUFF and VLSTRACK dispersion models. In
contrast to LLNL's modeling simulations, it showed the plume moving first
southerly and then turning to the west southwest. The 72-hour plume
overlay for DOD's composite model, resulting from using the VLSTRACK and
HPAC/SCIPUFF dispersion models with COAMPS and OMEGA meteorological
models, is shown in figure 10.

26"Minimal effect" is the lowest concentration expected to have noticeable
effects on human beings.

27"Occupational limit" is about one-tenth of the minimal effects value and
the maximum concentration level that would be allowed for a worker who
could become exposed to sarin in the course of performing job duties.

28"General population limit" represents the limit below which any member
of the general population could be exposed-by, for example, exhaling-7
days a week, every week, for a lifetime without experiencing any adverse
health effects.

  Source: Department of Defense, Office of the Special Assistant for Gulf War
                                    Issues.

                     Page 41 GAO-04-159 Gulf War Illnesses

Composite modeling may be an appropriate methodology, but DOD's composite
model understated the number of troops potentially exposed by not
including the LLNL model. If LLNL's model were included, a far larger
number of forces would potentially be shown as having been exposed (see
figure 11). DOD's exclusion of this model seriously skewed the outcome of
any epidemiological studies done thus far.

Source: GAO analysis of Department of Defense and Department of Energy,
Lawrence Livermore National Laboratory, models.

                     Page 43 GAO-04-159 Gulf War Illnesses

Explaining the Models' Divergent Results

Further research was conducted to determine whether the models' divergent
results could be explained. It was observed that the divergence in the
modeling outcomes could be explained by a directional split-a line of
diffluence-in the independently modeled 10 m wind field data near
Khamisiyah in the first 2 days of the modeling period. While the precise
location of this line is critical in determining which way the wind would
have transported the CW agent, its precise location cannot be resolved
with any degree of certainty. (Appendix IV illustrates this diffluence
with three different data sets.)

In addition, DTRA officials said at the time of the modeling that they had
conducted data-validation analyses of the various models against visible
smoke plumes from the oil well fires in Kuwait. These analyses indicated a
definite directional bias, shown in figure 12. This validation
demonstrates that the actual area covered could have been from 10 to 50
degrees to the east and could have affected a different population from
that indicated by the model results.

                  Figure 12: Validation Runs of Various Models

COAMPS MM5 OMEGA Visible plume VLSTRACK HPAC HPAC

Source: GAO.

aEast bias compared with OMEGA/HPAC. bEast bias compared with MM5/HPAC.
cEast bias compared with COAMPS/VLSTRACK.

In addition, in the 1997 DOD peer review panel report on the Khamisiyah
models, a panel of experts in meteorological and turbulent diffusion
modeling stated that the VLSTRACK and SCIPUFF/HPAC results were
complicated by the use of significantly different source term inputs.29
According to DOD, the 2000 modeling used a consistent source term for
SCIPUFF and VLSTRACK. For VLSTRACK, internal source algorithms had been
used, while for SCIPUFF, source term inputs from the Dugway experiments
had been used. These differences would lead to significant divergences in
the dosage contours the two models predicted, which were then used to
generate the composite.

In addition, the panel noted, while using a composite model is a valid
method, successfully applied to other atmospheric modeling problems, using
a composite model to reconstruct the dosage requires more advanced,
state-of-the-art, high-resolution models, with the fewest physical
limitations and assumptions. Furthermore, in 1998, the Air Force Human
Effectiveness Directorate at Wright-Patterson Air Force Base reviewed the
Muhammadiyat modeling and found that the work may have been flawed.
According to the Chief of the Human Effectiveness Directorate, the
protocol was not correct in that it constituted a "plume of plumes, rather
than a plume based upon data."

AFTAC Modeling 	AFTAC modeling of Khamisiyah also showed differences from
DOD's composite model. After DOD's Khamisiyah models were published in
1997, the Senate Committee on Veterans' Affairs asked AFTAC to conduct an
analysis of the event.30 AFTAC is a principal modeling agency for DOD and,
according to experts, the quality of its modeling system is among the
highest. To conduct the analysis, AFTAC meteorologists used four models

29Richard A. Anthes and others, "Comments by Review Panel on Khamisiyah
Modeling Report and Presentations on November 4-5, 1997," report for the
Directorate for Deployment Health Support of the Special Assistant to the
Under Secretary of Defense (Personnel and Readiness) for Gulf War
Illnesses, Medical Readiness, and Military Deployments, Fairfax, Virginia,
December 11, 1997.

30In early 1997, the Senate Committee on Veterans' Affairs established a
Special Investigation Unit on Gulf War Illnesses to conduct a bipartisan
review of the U.S. government's response to the unexplained illnesses
suffered by veterans of the Gulf War. The year-long effort produced a
detailed report on the actions by DOD before and during the war and by VA
in its aftermath, relating to the current health of Gulf War veterans. See
Arlen Specter, William F. Tuerk, and James R. Gottlieb, Report of the
Special Investigation Unit on Gulf War Illnesses, U.S. Senate, Committee
on Veterans' Affairs (Washington, D.C.: U.S. Government Printing Office,
1998).

from their suite of atmospheric models, including two primary transport
and diffusion models and the RAMS meteorological model.

While AFTAC's analysis was meteorologically consistent with the modeling
efforts of the CIA and DOD, the results showed that some additional areas
might have been exposed to at least low-level exposure dosages from the
Khamisiyah plume. In particular, it showed the plume differing in several
aspects from the plumes generated by the DOD and CIA analyses. Most
significantly, the AFTAC plume is shown drifting across Kuwait and the
northwestern Persian Gulf coast, an area not covered by the DOD and CIA
plume. AFTAC's analysis was published in a report to the Senate Committee
on Veterans' Affairs and was provided to DOD's Office of the Special
Assistant to the Deputy Secretary of Defense for Gulf War Illnesses.31 The
report recommended that AFTAC's results be integrated into the DOD and CIA
modeling results to depict these additional areas of potential exposure.

However, in 1998, a DOD expert panel reviewed the AFTAC modeling
simulations and recommended that AFTAC's modeling results not be included
in DOD's composite plume model, given that (1) AFTAC's simulations were
generally consistent with DOD's composite model results and that the
effect of any differences would only have resulted in the additional
notification of a small number of individuals, (2) continuing to refine
DOD's modeling process rather than including AFTAC's modeling results
would be the best use of DOD's resources, and (3) the decay capability of
the agent in AFTAC's model was still immature and would have limited any
efforts to identify potentially exposed individuals.

31U.S. Air Force, Technical Applications Center, Report on Atmospheric
Modeling of the 10 Mar 91 Chemical Warfare Agent Release at the Khamisiyah
(Iraq) Munitions Pit

(Patrick Air Force Base, Fla.: Dec. 15, 1997).

The MOD Relied on U.S. Plume Modeling to Determine Its Troops' Exposure to
CW Agents

Total U.S. Plume-Modeling Costs

According to British officials, the MOD did not collect any source term or
meteorological data during the Gulf War. It also did not independently
model the plume from Khamisiyah, relying instead on the 1997 DOD and CIA
modeling of Khamisiyah. According to British MOD officials, however, they
were reassessing the extent of British troops' exposure, based on DOD's
revised 2000 remodeling of Khamisiyah. We requested from the British MOD
but did not receive information on the findings from this reassessment.

Responding to parliamentary concerns and questions raised in 1997, the
British MOD reviewed the U.S. modeling of demolition operations at
Khamisiyah, publishing a report in December 1999.32 The MOD concluded from
the 1997 DOD and CIA composite model of the Khamisiyah demolitions that
the maximum concentration of agent that British troops might have been
exposed to was below the level that the most sensitive British warning
device could have been expected to detect. Moreover, according to the MOD,
the highest theoretical dosage troops received would have been 3.6 times
lower than the level at which the first noticeable symptoms occur.
Finally, the MOD said, this level of exposure would have had no detectable
effect on health.

The MOD also determined that a number of British troops were within the
boundary of the plume in the DOD and CIA composite model, and it estimated
that the total number of British troops potentially exposed was about
9,000. The total number of troops definitely within the path of the plume,
however, was estimated to be about 3,800. In addition, of 53,500 British
troops deployed, at least 44,000 were definitely not within the path of
the plume. However, since the MOD relied exclusively on DOD's modeling and
since we found that DOD could not know who was and who was not exposed,
the MOD cannot know the extent of British troops' exposure.

The CIA and DOD were the primary agencies in the modeling and analysis of
U.S. troops' exposure from the demolition at Khamisiyah and bombing of
chemical facilities at Al Muthanna, Muhammadiyat, and Ukhaydir, but
several other agencies and contractors also participated. Funding to

32"Review of Events Concerning 32 Field Hospital and the Release of Nerve
Agent Arising from U.S. Demolition of Iraqi Munitions at the Khamisiyah
Depot in March 1991," December 1999.

support the modeling efforts was provided to various DOD agencies and
organizations, the military services, and non-DOD agencies and
contractors. We collected data on the direct costs these agencies incurred
or funds they spent. Table 5 shows direct costs to the United States for
modeling the Gulf War of about $13.7 million.

           Table 5: U.S. Direct Costs for Modeling Gulf War Illnesses

     Agency or contractor    Direct costs a             Work done             
           BAHR Inc.                $11,796    Reviewed (1) processes and     
                                               technology used to produce     
                                            estimates of U.S. forces          
                                            potentially exposed and (2) draft 
                                            reports                           
                                                      on Khamisiyah           
     Central Intelligence                 b    Computer-modeling analysis     
            Agency                          
      Chemical Biological                   Wood-surface evaporative modeling 
       Defense Command,             140,000      and environmental data       
    Aberdeen Proving Ground                          support efforts          
Defense Threat Reduction                  Computer-modeling analyses with  
            Agency                  870,000      HPAC/SCIPUFF dispersion      
                                                and OMEGA weather models      
     Institute for Defense                  Convened a panel of experts to    
           Analyses                 149,429 review Khamisiyah pit modeling    
                                                        analyses              
      Lawrence Livermore                     Computer-modeling analyses with  
      National Laboratory            60,000       ADPIC dispersion and        
                                                  MATHEW weather models       
      National Center for           308,000   Computer-modeling simulations   
     Atmospheric Research                        using MM5 weather model      
                                            Meteorological analysis to        
                                            identify downwind hazard          
Naval Research Laboratory      1,090,000 assessment                        
                                             with NOGAPS and COAMPS weather   
                                                         models.              
     Naval Surface Warfare                   Computer-modeling analyses with  
            Center                  522,000      VLSTRACK dispersion and      
                                                  COAMPS weather models       
     Office of the Special                  Internal costs for producing case 
    Assistant to the Deputy       7,980,000    narratives for Al Muthanna,    
Secretary of Defense for                   Khamisiyah, Muhammadiyat, and   
      Gulf War Illnesses                                Ukhaydir              
     Science Applications                 c    Computer-modeling analysis     
International Corporation                
                                            Exposure assessment and           
     U.S. Army Center for                   environmental modeling to         
     Health Promotion and           731,000 determine                         

                             Preventative Medicine

U.S. troops' exposed to chemical releases from multiple incidents during
the Gulf War

U.S. Army Dugway Proving   1,861,950  Field trials and laboratory testing  
            Ground                             using 122 mm chemical-         
                                           simulant filled rockets to develop 
                                                source term data for modeling 
     White Sands Missile          2,600    Missiles for testing at Dugway     
            Range                                  Proving Ground             
            Total           $13,726,775 

Sources: Agency and contractor responses provided to GAO on their modeling
and analysis costs.

aDirect costs for agencies includes funding for contracts provided by the
Office of the Special Assistant to the Deputy Secretary of Defense for
Gulf War Illnesses.

bThe CIA denied our request for its costs for modeling chemical releases
from Khamisiyah, as well as Al Muthanna, Muhammadiyat, and Ukhaydir.

cSAIC did not respond to our requests for information.

Indirect costs were much more difficult to obtain because the modeling
efforts involved many people, and some were full-time while others were
part-time. However, these estimates do not include, and DOD could not
provide, an estimate of the considerable indirect costs involved for
salaries of DOD and VA staff, contractors, facilities, travel, and
equipment. In addition, the CIA would not provide us with the direct and
indirect costs for modeling Gulf War plume and determinations of source
term because, in its view, our request constituted oversight of an
intelligence matter and is beyond our scope of authority. The CIA's
contractor, SAIC, also would not respond to our request for cost
information.

DOD's and VA's Epidemiological Conclusions on CW Exposure and
Hospitalization and Mortality Rates Cannot Be Adequately Supported

DOD and VA each funded an epidemiological study on CW exposure- DOD's on
hospitalization rates and VA's on mortality rates. From the
hospitalization study conducted by DOD researchers and the mortality study
conducted by VA researchers, DOD and VA each concluded that there was no
significant difference in the rates of hospitalization and mortality
between exposed and nonexposed troops. These conclusions, however, cannot
be adequately supported by the available evidence. These studies contained
two inherent weaknesses: (1) flawed criteria for classifying exposure,
resulting in classification bias, and (2) an insensitive outcome measure,
resulting in outcome bias. In addition, several other studies of Gulf War
veterans, genetics, and animals found a strong association between
exposure and illnesses.

DOD and VA Used Flawed Criteria for Determining Troops' Exposure

In the two epidemiological studies, DOD and VA researchers used DOD's 1997
plume model for determining which troops were under the path of the
plume-that is, were exposed-and which troops were not-that is, were not
exposed. However, this classification is flawed, given their inappropriate
criteria for inclusion and exclusion.

In the hospitalization study, the DOD researchers included 349,291 Army
troops "coded" as being in the Army on February 21, 1991. However, the
researchers did not report a cutoff date for inclusion in the study-that
is, they did not indicate whether these troops were in the Persian Gulf
between January 17, 1991, and March 13, 1991, the period during which
bombing and demolition took place. Although we requested this date, DOD
researchers failed to provide it. Finally, the total number of 349,291
troops is misleading because many troops left the service soon after
returning from the Persian Gulf. Moreover, the researchers did not conduct
any analyses to determine whether those who left were in the exposed or
nonexposed group (including uncertain low-dose exposure or

estimated subclinical exposure). Given all the methodological problems in
this study, it is not possible to accurately determine the total number of
the exposed and nonexposed groups.

In the mortality study, the VA researchers included 621,902 Gulf War
veterans who arrived in the Persian Gulf before March 1, 1991. Troops who
left before January 17, 1991-the beginning of the bombing of Iraqi
research, production, and storage facilities for CW agents-were included
in the study. This group was not likely to have been exposed. Therefore,
including them resulted in VA's overestimation of the nonexposed group.

Troops who came after March 1, 1991-the period during which Khamisiyah
demolition took place-were excluded from the VA study. The Defense
Manpower Data Center (DMDC) identified 696,000 troops deployed to the
Persian Gulf, but the mortality study included only the 621,902 troops
deployed there before March 1, 1991. This decision excluded more than
74,000 troops, approximately 11 percent of the total deployed. However,
these troops were most likely to have been exposed. Therefore, excluding
them resulted in underestimation of the size of the exposed group. In
addition, 693 troops who were in the exposed group were excluded because
identifying data, such as Social Security numbers, did not match the DMDC
database. But VA researchers did not conduct follow-up analysis to
determine whether those who were excluded differed from those who were
included in ways that would affect the classification.

DOD and VA Used an Insensitive Outcome Measure for Determining
Hospitalization Rate

Hospitalization rate-the outcome measure used in the hospitalization
study-was insensitive because it failed to capture the chronic illnesses
that Gulf War veterans commonly report but that typically do not lead to
hospitalization. Studies that rely on this type of outcome as an end point
are predetermined to overlook any association between exposure and
illness.

Based on DOD's 1997 plume model, DOD's hospitalization study compared the
rates for Gulf War veterans who were exposed with the rates for those who
were not exposed. This study included 349,291 active duty Army troops who
were deployed to the Persian Gulf. However, DOD researchers did not
determine the resulting bias in their analyses, because they did not
account for those who left the service.

The Institute of Medicine noted that the hospitalization study was limited
to Army troops remaining on active duty and to events occurring in

military hospitals. Conceivably, those who suffered from Gulf War-related
symptoms might leave active duty voluntarily or might take a medical
discharge. Hospitalization for this group would be reflected in VA or
private sector databases but not in DOD databases. The health or other
characteristics of active duty troops could differ from those of troops
who left active duty and were treated in nonmilitary hospitals. Moreover,
economic and other factors not related to health are likely to affect the
use of nonmilitary hospitals and health care services.33

This limiting of the study to troops remaining on active duty produced a
type of selection bias known as the healthy warrior effect.34 It strongly
biased the study toward finding no excess hospitalization among the active
duty Army troops compared with those who left the service after the war.

Some Studies Suggest an Association between CW Exposure and Gulf War
Illnesses

Gulf War Veterans Studies

We found some studies that suggest an association between CW exposure and
Gulf War illnesses. They are Gulf War veterans studies, genetics studies,
and animal studies. Each of these studies, described below, has strengths
and limitations.

In a privately funded study, Haley and colleagues reported an association
between a syndromic case definition of Gulf War illnesses, developed to
model the ill veterans' symptomatic complaints, with exposure to CW
agents.35 In this study, the authors developed questionnaires on symptoms
and environmental exposure identified in pilot studies of ill Gulf War
veterans, similar to epidemic investigations by the Centers for Disease
Control and Prevention (CDC).36 These questionnaires were given to 249
troops from a U.S. Navy Mobile Construction Battalion that participated in
the Gulf War. Factor analysis of the data on symptoms was used to derive

33Institute of Medicine, Gulf War Veterans: Measuring Health (Washington,
D.C.: National Academy Press, 1999), p. 36.

34R. W. Haley, "Point: Bias from the `Healthy-Warrior Effect' and Unequal
Follow-Up in Three Government Studies of Health Effects of the Gulf War,"
American Journal of Epidemiology 148 (1998): 315-38.

35R. W. Haley and T. L. Kurt, "Self-Reported Exposure to Neurotoxic
Chemical Combinations in the Gulf War," JAMA 277 (1997): 231-37.

36See Michael B. Gregg, ed., Field Epidemiology, 2nd ed. (New York: Oxford
University Press, 2002).

a case definition identifying six syndrome factors.37 Three syndrome
factor variants found to be the most significant were (1) impaired
cognition, (2) confusion-ataxia, and (3) arthro-myo-neuropathy.

Impaired cognition (syndrome 1) was associated with troops' having worn
flea collars that contained chlorpyrifos.38 Confusion-ataxia (syndrome 2),
the most severe clinically, was associated with three risk factors.39 The
first was likely CW exposure; the second was the geographic location near
the Saudi-Kuwaiti border around the fourth day of the air war, conducted
January 18-23, 1991, when Czech chemical detection units detected sarin
and mustard in ambient air near the Saudi-Kuwaiti border; and the third
was side effects experienced after taking pyridostigmine. There was also a
significant synergistic association between likely exposure to CW agents
and the number of side effects from pyridostigmine.40 Arthro-myoneuropathy
(syndrome 3) was associated with the amount of exposure to 95 percent DEET
in ethanol insect repellent and with the number of side effects of
pyridostigmine.41

37R. W. Haley and others, "Is There a Gulf War Syndrome? Searching for
Syndromes by Factor Analysis of Symptoms," JAMA 277 (1997): 215-22. The
six syndrome factors were impaired cognition, confusion-ataxia,
arthro-myo-neuropathy, phobia-apraxia, feveradenopathy, and
weakness-incontinence.

38Impaired cognition is characterized by problems with attention, memory,
and reasoning, as well as insomnia, depression, daytime sleepiness, and
headache. (Study results showed relative risk 8.2, 95 percent, CI
2.9-23.5, p = 0.001.)

39Confusion-ataxia is characterized by problems with thinking,
disorientation, balance disturbances, vertigo, and impotence.

40(1) CW exposure, relative risk 7.8, 95 percent, CI 2.3-25.9, p < 0.0001;
(2) geographic location, relative risk 4.3, 95 percent, CI 1.9-10.0, p =
0.004; (3) pyridostigmine side effects, dose-response trend up to relative
risk 32.4, 95 percent, CI 7.8-135.0, p < 0.0001; (4) synergistic
association, Rothman synergy statistic 5.3, 95 percent, CI 1.04-26.7, p <
0.05. See Jonathan B. Tucker, "Evidence Iraq Used Chemical Weapons during
the 1991 Persian Gulf War," The Nonproliferation Review 4:3 (Spring-Summer
1997): 114-22. Center for Nonproliferation Studies, Monterey Institute of
International Studies, http://cns.miis.edu/pubs (Apr. 28, 2004); and U.S.
Department of Defense, Office of the Special Assistant for Gulf War
Illnesses, Coalition Chemical Detections and Health of Coalition Troops in
Detection Area (Washington, D.C.: Aug. 5, 1996).
http://www.gulflink.osd.mil/czech_french/czfr_refs/n08en011/coalitn.html
(Apr. 28, 2004).

41Arthro-myo-neuropathy is characterized by joint and muscle pains, muscle
fatigue, difficulty lifting, and paresthesias of the extremities. (Results
showed for exposure, doseresponse effect to relative risk 7.8, 95 percent,
CI 2.4-24.7, p < 0.0001; for side effects, doseresponse effect to relative
risk 3.9, 95 percent, CI 1.3-12.1, p < 0.0001.)

The inference that these risk-factor associations represent causal effects
is supported by (1) the large, highly significant relative risks; (2) the
doseresponse effects; and (3) the synergistic effect of the risk factor
associations with the syndromic case definition. Risk factors found not to
be significantly associated with the case definition include environmental
pesticides, pesticides in uniforms, antibiotic or antimalarial
prophylaxis, multiple immunizations, smoke from oil well fires, fumes from
jet fuel, fumes from burning jet fuel in tents, petroleum in drinking
water, depleted uranium munitions, smoking, alcohol use, and combat
exposure.

Another study of Gulf War veterans by Nisenbaum and colleagues, funded by
CDC, examined the risk factors in 1,002 Air Force reservists.42 They
found, first, that the case definition of Fukuda and colleagues of
"multisymptom illness" was strongly associated with at least one of the
three chronic symptom groups fatigue, mood/cognition, and musculoskeletal
pain. And, next, they found that reported exposure to CW agents was most
strongly associated with the "severe illness" case definition of Fukuda
and colleagues and less strongly associated with their "mild-moderate
illness" case definition.43

Both case definitions were less strongly associated with the use of insect
repellent (p = 0.006), the use of pyridostigmine (p = 0.01), and having an
injury requiring medical attention (p = 0.03). But neither case definition
was associated with smoke from oil well fires, coming under attack, seeing
casualties, or having adverse health events in the family. The findings
were attributed to the effects of stress but offered no empirical support
for the explanation.

In a study that VA funded, Proctor and colleagues compared the exposure
histories of 186 Gulf War veterans from Fort Devens, Massachusetts, and 66
from New Orleans, including 48 who deployed only to Germany. Collectively,
the 252 veterans are known as the Massachusetts-New

42R. Nisenbaum and others, "Deployment Stressors and a Chronic
Multisymptom Illness among Gulf War Veterans," Journal of Nervous and
Mental Disease 188 (2000): 259-66.

43Association with "severe illness," adjusted OR 3.46, 95 percent, CI
1.73-6.91, p < 0.0001; association with "mild-moderate illness," adjusted
OR 2.25, 95 percent, CI 1.54-3.27, p < 0.0001. See K. Fukuda and others,
"A Chronic Multisymptom Illness Affecting Air Force Veterans of the
Persian Gulf War," JAMA 280 (1998): 981-88.

Orleans cohort.44 The case definition was a set of eight body-system
symptom scores (BSS, distributed from 0 to 4), each constructed by summing
the 5-point frequency-of-occurrence scales (0 = occurs never, 4 = occurs
almost every day) for three symptoms typical of a particular body system.
The eight body systems were cardiac, dermatological, gastrointestinal,
musculoskeletal, neurological, neuropsychological, psychological, and
pulmonary. Post-traumatic stress disorder (PTSD) was diagnosed with the
structural clinical interviews, Clinician Administered Posttraumatic
Stress (CAPS) disorder scale, or a Mississippi Scale score of >89. The
symptoms were obtained from the 52-item Expanded Health Symptom Checklist,
the exposure measures from an environmental exposure questionnaire and an
Expanded Combat Exposure Scale (CES) questionnaire. Multiple regression
analysis-controlling for age, sex, education, study site, Expanded CES
score, and PTSD status-was used to develop a risk-factor model for each
BSS scale.

Exposure to CW agents and debris from SCUD missiles was associated with
four BSS scales; exposure to smoke from tent heaters, with three BSS
scales; exposure to pesticides, vehicle exhaust, and burning human waste,
with two BSS scales; the Expanded CES, with only one BSS scale; and
exposure to pyridostigmine bromide (antinerve gas pills) and smoke from
oil well fires, with no BSS scale. Controlling for depression scores and
excluding veterans diagnosed with PTSD did not substantially affect the
associations.

Three additional studies conducted with VA and DOD funding extended the
risk-factor research for the Massachusetts-New Orleans cohort. The
association of self-reported CW agent (nerve agent) exposure was tested
with different formulations of the case definition. White and colleagues
used psychological and neuropsychological tests to define illness. They
found that exposure to CW agents was associated with abnormal measures of
mood, memory, and attention or executive function.45 Associations remained
significant after controlling for age, sex, race, years of education,
repeated grade in school, head injury, medication use, diagnosis of
current

44S. P. Proctor and others, "Health Status of Persian Gulf War Veterans:
Self-Reported Symptoms, Environmental Exposures, and the Effect of
Stress," International Journal of Epidemiology 27 (1998): 1000-10.

45R. F. White and others, "Neuropsychological Function in Gulf War
Veterans: Relationships to Self-Reported Toxicant Exposures," American
Journal of Industrial Medicine 40 (2001): 42-54.

PTSD (by CAPS), diagnosis of current depression (by structural clinical
interviews), active duty versus Reserve or Guard status, seeking
disability rating, and Vietnam service.

Lindem and colleagues developed multiple regression models for
neuropsychological test measures as case definitions of Gulf War
illnesses.46 Chemical warfare agent exposure was found to be associated
with attention and executive function (continuous performance test),
delayed verbal recall (California Verbal Learning Test and Visual
Reproduction Test), and confusion and fatigue (Profile of Mood States).
These associations remained significant when controlling for age,
education, and PTSD diagnosis (by CAPS).

Wolfe and colleagues, studying 945 troops from the Massachusetts-New
Orleans cohort, found that the CDC case definition of multisymptom illness
was most strongly associated with having smelled a chemical odor, having
taken up to 21 antinerve gas pills, or having experienced up to 10 formal
alerts for CW agent attack.47

Kang and colleagues conducted a random sample mail survey that VA funded.
Obtaining responses from 11,441 Gulf War veterans and 9,476 nondeployed
Gulf War era veterans, they developed a case definition by factor analysis
of symptoms measured by their questionnaire.48 The first three syndrome
factors closely resembled those that Haley and others derived (noted
earlier). Finding that syndrome 2 was unique to the sample that had been
deployed in the Gulf War (found in the deployed, but not the nondeployed,
sample) and that the component symptoms were neurological in character,
the researchers termed their syndrome 2 a possible unique Gulf War
neurological syndrome. Four symptoms- blurred vision, loss of balance or
dizziness, tremor or shaking, and speech difficulties-were associated with
syndrome 2 only in the deployed sample. Consequently, Kang and colleagues
established their case

46K. Lindem and others, "Neuropsychological Performance in Gulf War Era
Veterans: Traumatic Stress Symptomatology and Exposure to
Chemical-Biological Warfare Agents, Journal of Psychopathology and
Behavioral Assessment 25:2 (2003): 105-19.

47Chemical odor, OR = 6.2, 95 percent, CI 3.9-9.9; antinerve gas pills, OR
= 3.7, 95 percent, CI 2.4-5.6; formal alerts for CW attack, OR = 2.7, 95
percent, CI 2.0-3.7. See J. Wolfe and others, "Risk Factors for
Multisymptom Illness in U.S. Army Veterans of the Gulf War," Journal of
Occupational and Environmental Medicine 44:3 (2002): 271-81.

48H. K. Kang and others, "Evidence for a Deployment-Related Gulf War
Syndrome by Factor Analysis," Archives of Environmental Health 57:1
(2002): 61-68.

definition as having all four of these symptoms. In the deployed sample,
277 met the case definition and 6,730 who had none of the four symptoms
constituted the control group. Of a large number of risk factors analyzed,
only nine were associated with the case definition, with an odds ratio
greater than 3.0. Of these, perceived exposure to nerve agent had the
strongest association (odds ratio 15.1, 95 percent, CI 11.5-19.9, p <
0.000001). This finding-a neurological syndrome appearing as the second
factor in a factor analysis and being the most strongly associated risk
factor, 15 times more common in ill veterans meeting the case definition
than in controls-closely parallels the findings of Haley and colleagues.
The finding received little notice, however, because the VA-funded mail
survey did not (1) provide the odds ratio values in the table reporting
the risk factor analysis results and (2) describe the finding in the text
or abstract of the paper. When we noticed the finding, we manually
calculated the odds ratios from the raw data in the table.

Smith and colleagues showed that hospitalization rates for several ICD-9
diagnoses were higher in veterans categorized in the Khamisiyah 2000 plume
than in those not in the plume, and the association for cardiac
arrhythmias was statistically significant. However, this study suffers
from the same deficiencies as the earlier study that we cited: use,
inappropriately, of hospitalization outcome measures rather than measures
of Gulf War illness, which usually do not result in hospitalization, and
use of plume modeling based on flawed data.49

The 2002 Kang and Bullman study has not been published in a peerreviewed
journal and therefore should not have been included in a review of the
scientific epidemiologic literature. The DOD studies were invalid for two
reasons: (1) Hospitalization and mortality were inappropriate outcomes
because they do not measure Gulf War illnesses, which often do not lead to
hospitalization, and (2) The DOD studies, no matter how powerful their
techniques, could not control for the selection bias that resulted from
the disproportionate early discharge of the ill Gulf War veterans soon
after the Gulf War. Including only DOD hospital records of service members
remaining on active duty resulted in the exclusion of veterans who left
service for poor health. No amount of sophisticated

49T. C. Smith and others, "Gulf War Veterans and Iraqi Nerve Agents at
Khamisiyah: Postwar Hospitalization Data Revisited," American Journal of
Epidemiology 158 (2003): 457-67.

Genetics Studies

techniques can correct for this selection bias toward finding no

50

difference.

In one genetics study, Haley and colleagues found an association between
the case definition of Gulf War illnesses in U.S. Gulf War veterans and
low blood levels of the Q-type isoenzyme of the paraoxonase/arylesterase
enzyme group (PON).51 The PON group of enzymes is a potentially important
predisposing factor in Gulf War illnesses because one of its major
functions in normal body physiology is to protect the nervous system from
organophosphate chemical toxins, such as pesticides and nerve agents. This
finding was remarkable because the only function of Q type of the PON
enzyme group is to protect the nervous system from nerve agents sarin,
soman, tabun, and VX. The R-type isoenzyme has as its main function
protection from organophosphate pesticides, such as diazinon, malathion,
and parathion. Thus, an association between Gulf War illnesses and blood
levels of only the Q-type isoenzyme of PON points specifically to nerve
agent exposure. In addition, the total PON level-that is, the sum of the Q
and R isoenzyme levels-was not associated with the illnesses. And the
genotype (QQ, QR, or RR) was only marginally associated with them, as
expected, because the level of the Q-type isoenzyme is a more important
determinant of susceptibility to nerve agents than the genotype.

In another genetics study, Mackness and colleagues reported lower blood
levels of total PON in ill British Gulf War veterans than in civilian
controls in a previously published study; however, they did not measure
the blood levels of the Q and R isoenzymes of PON, needed for a definitive
study of Haley's hypothesis.52 This finding could indicate that ill
British Gulf War veterans represented a mixture of some with low Q-type
PON and others with low R-type PON. In some veterans, the illness would be
associated with exposure to nerve agents; in others, with exposure to
pesticides. Alternatively, the difference in total PON levels may have
resulted from differences in the assays or in the veterans, since (1) the
enzyme assays in

50H. K. Kang and T. A. Bullman, Mortality among U.S. Gulf War Veterans Who
Were Potentially Exposed to Nerve Gas at Khamisiyah, Iraq (Washington, DC:
Department of Veterans Affairs, May 2002).

51R. W. Haley and others, "Association of Low PON1 Type Q (Type A)
Arylesterase Activity with Neurologic Symptom Complexes in Gulf War
Veterans," Toxicology and Applied Pharmacology 157 (1999): 227-33.

52B. Mackness and others, "Low Paraoxonase in Persian Gulf War Veterans
Self-Reporting Gulf War Syndrome," Biochemical and Biophysical Research
Communications 276 (2000): 729-33.

the controls were performed years before those for the ill veterans and
(2) the controls were civilians studied in an entirely different setting.

In yet a third genetics study, Hotopf and colleagues reported results of
tests for total PON levels in blood samples-obtained in a study by Unwin
and colleagues-for four groups of British troops: (1) ill veterans of the
Gulf War, (2) well veterans of the Gulf War, (3) ill nondeployed veterans
of the Gulf War era, and (4) ill veterans of the Bosnian conflict.53 The
case definition of illness was a score below 72.2 on the SF-36 Physical
Status questionnaire. Again, the researchers did not measure the levels of
the Q and R isoenzymes of PON, making the findings difficult to interpret.
The researchers found a low mean level of total PON in both ill and well
groups deployed to the Gulf War and higher levels in the Gulf War era and
ill Bosnian groups.

The depressing of the total PON level, the researchers suggested, might be
the result of some deployment-related exposures. However, instead of
looking at exposure to CW agents, the researchers investigated the
possible effect of multiple immunizations on total PON levels and found no
evidence for it. An alternative explanation is that total PON level in
both ill and well deployed veterans was the result of misclassification of
veterans by the case definition. A score of 72.2 on the SF-36 scale is not
a very low score, particularly in ill Gulf War veterans, and it is a
nonspecific measure of illness, given that a low score indicates illness
from any cause.54 Consequently, many veterans ill from causes unrelated to
the war would be misclassified as cases of Gulf War illness and,
conversely, many ill from the war but with less disability would be
misclassified as controls. This conclusion is supported by a
nonsignificant trend showing that ill veterans who had been deployed to
the Gulf War had a lower median total PON level than well veterans who had
also been deployed to the Gulf War.

The many flaws of design and methodology in both British studies of PON
levels do not contribute to an understanding of the PON hypothesis and
leave the finding of Haley and colleagues in need of better replication.

53See Matthew Hotopf and others, "Paraoxonase in Persian Gulf War
Veterans," Journal of Occupational and Environmental Medicine 45 (2003):
668-75, and C. Unwin and others, "Health of UK Servicemen Who Served in
the Persian Gulf War," Lancet 353 (1999): 169-78.

54R. W. Haley and others, "Severely Reduced Functional Status in Veterans
Fitting a Case Definition of Gulf War Syndrome," American Journal of
Public Health 92 (2002): 46-47.

Animal Studies

A series of laboratory studies with animals have established the
biological plausibility that brain cell damage results from low-level
exposure to sarin. Husain and colleagues demonstrated in two studies at
the Division of Pharmacology and Toxicology at the Defense Research and
Development Establishment in Gwalior, India, that repetitive
administration of low-dose sarin (approximately 0.25 LD50) daily for 10
days caused delayed onset damage to neurons in the spinal cords and brains
of mice exposed by inhalation and of hens exposed by subcutaneous
injection.55

Privately funded studies by Abou-Donia and colleagues demonstrated that
combinations of organophosphates and similar cholinesterase-inhibiting
chemicals in hens produce greater neurotoxic effect on brain and nerve
tissue than any of the agents alone.56 Abou-Donia's subsequent work,
funded by DOD, extended the findings to synergistic combinations involving
sarin at moderate concentrations (0.5 LD50).57 A similar study by Husain
and Somani, also funded by DOD, on the delayed brain effects of low-dose
sarin (0.05 LD50) in combination with pyridostigmine and exercise,
confirmed these findings. In particular, it demonstrated that the neuronal
damage from very low doses of sarin affected primarily the basal ganglia
region of the brain (striatum).58

A study by Henderson and colleagues, with DOD funding, found that repeated
inhalation exposure to low-level sarin at subsymptomatic doses

55K. Husain and others, "Delayed Neurotoxic Effect of Sarin in Mice after
Repeated Inhalation Exposure," Journal of Applied Toxicology 13 (1993):
143-45, and "A Comparative Study of Delayed Neurotoxicity in Hens
Following Repeated Administration of Organophosphorus Compounds," Indian
Journal of Physiology and Pharmacology 39 (1995): 47-50.

56Mohamed B. Abou-Donia and others, "Neurotoxicity Resulting from
Coexposure to Pyridostigmine Bromide, DEET, and Permethrin," Journal of
Toxicology and Environmental Health 48 (1996): 35-56, and "Increased
Neurotoxicity Following Concurrent Exposure to Pyridostigmine Bromide,
DEET, and Chlorpyrifos," Fundamentals of Applied Toxicology 34 (1996):
201-22.

57Mohamed B. Abou-Donia and others, "Combined Exposure to Sarin and
Pyridostigmine Bromide Increased Levels of Rat Urinary 3-Nitrotyrosine and
8-Hydroxy-2'-Deoxyguanosine, Biomarkers of Oxidative Stress," Toxicology
Letters 123 (2001): 51-58; "Disruption of the Blood-Brain Barrier and
Neuronal Cell Death in Cingulate Cortex, Dentate Gyrus, Thalamus, and
Hypothalamus in a Rat Model of Gulf-War Syndrome," Neurobiology of Disease
10 (2002): 306-26; and "Sarin: Health Effects, Metabolism, and Methods of
Analysis," Food and Chemical Toxicology 40 (2002): 1327-33.

58K. Husain and S. Somani, "Delayed Toxic Effects of Nerve Gas Sarin and
Pyridostigmine under Physical Stress in Mice," Journal of Burns and
Surgical Wound Care 2 (2003): 2-19.

(0.1 LCt50) for 5 or 10 days, with or without heat stress, produced no
immediate effects.59 But at 30 days after exposure to sarin, damage was
produced to cholinergic receptors in several brain regions, including the
basal ganglia. In the same study, Henderson and colleagues identified
evidence of an autonomic nervous system injury affecting the function of
T-cells in the immune system as well.60 In addition, chronic abnormalities
of neuronal metabolism in the basal ganglia have been implicated in ill
Gulf War veterans by several investigators through the use of magnetic

61

resonance spectroscopy.

Two recent laboratory studies at the U.S. Army Medical Research Institute
of Chemical Defense, Aberdeen Proving Ground, support the animal studies.
Scremin and colleagues demonstrated that moderate doses of sarin (0.5
LD50) in combination with pyridostigmine bromide produced prolonged
elevations in rats' cerebral blood flow but that neither agent alone had a
prolonged effect on cerebral blood flow.62 A companion study, by Roberson
and colleagues, demonstrated that repeated administration of sarin to
guinea pigs in doses of 0.2 or 0.4 LD50 produced no immediate ill effects
on behavior, weight, body temperature, flinch threshold, or EEG brain wave
activity. But at 100 days postdosing, abnormal brain function was found,
indicating neurochemical or pathological brain cell changes

63

that affect behavior.

59R. F. Henderson and others, "Response of F344 Rats to Inhalation of
Subclinical Levels of Sarin: Exploring Potential Causes of Gulf War
Illness," Journal of Toxicology and Industrial Health 17 (2001): 294-97
and 18:1 (2002): 48.

60See Henderson and others, "Response of Rats to Low Levels of Sarin," and
"Subclinical Doses of the Nerve Gas Sarin Impair T Cell Responses through
the Autonomic Nervous System," Journal of Toxicology and Applied
Pharmacology 184 (2002): 82-87.

61See R. W. Haley and others, "Brain Abnormalities in Gulf War Syndrome:
Evaluation by 1H Magnetic Resonance Spectroscopy," Radiology 215 (2000):
807-17, and "Effect of Basal Ganglia Injury on Central Dopamine Activity
in Gulf War Syndrome: Correlation of Proton Magnetic Resonance
Spectroscopy and Plasma Homovanillic Acid," Archives of Neurology 57
(2000): 1280-85, as well as D. J. Meyerhoff and others, "Reduced
N-Acetylaspartate in the Right Basal Ganglia of Ill Gulf War Veterans by
Magnetic Resonance Spectroscopy," Proceedings of the International Society
of Magnetic Resonance Medicine 9 (2001): 994.

62O. U. Scremin and others, "Effects of Chronic Exposure to Low Levels of
Cholinesterase Inhibitors on Cerebral Blood Flow," paper for the Society
for Neuroscience Meeting, Orlando, Florida, 2002.

63Melinda Roberson and others, "Depression of Cholinesterase Activity by
Low-Dose Sarin Exposure May Lead to Persistent Changes That Influence
Behavior," Society for Neuroscience, Washington, D.C., Program no. 205.3
(Abstract, 2002).

Conclusions

In evaluating the weaknesses of the plume models, we conclude that the
results from the CIA and DOD modeling can never be definitive. Plume
models can allow only estimates of what happens when CW agents are
released in the environment. Such estimates are based on mathematical
equations, which are used to predict an actual event-in this case, the
direction and extent of the plume. However, in order to predict precisely
what happens, one needs to have accurate data on source term and
meteorological conditions. DOD had neither of these.

Given the unreliability of the input data, the lack of troop location
data, and the divergent results of modeling, DOD's analyses cannot
adequately determine the extent of U.S. troops' exposure. In particular,
the models selected were not fully developed for projecting long-range
environmental fallout, and the assumptions used to provide the source term
data were flawed. Even when models used the same source term data, their
results diverged. In addition, the models did not include many potential
exposure events and some key materials-for example, binary chemical
weapons, mustard agent combustion by-products, and CW agent precursor
materials. It is likely that if models were more fully developed, and if
more credible data for source term and meteorological conditions were
included in them, particularly with respect to plume height as well as
level and duration of exposure, the hazard area would be much larger and
most likely would cover most of the areas where U.S. troops and Coalition
forces were deployed. However, given the lack of verifiable data for
analysis, it is unlikely that any further modeling efforts would be more
accurate or helpful.

The results of DOD's modeling efforts were, nonetheless, used in
epidemiological studies to determine the troops' CW exposure
classification-exposed versus nonexposed. As we noted in 1997, to
ascertain the causes of veterans' illnesses, it is imperative that
investigators have valid and reliable data on exposure, especially for
lowlevel or intermittent exposures to CW agents.64 To the extent that
veterans are misclassified as to exposure, relationships will be obscured
and conclusions misleading. In addition, DOD combined the results of
individual models that showed smaller plume size and ignored the results

64U.S. General Accounting Office, Gulf War Illnesses: Improved Monitoring
of Clinical Progress and Reexamination of Research Emphasis Are Needed,
GAO/NSIAD-97-163 (Washington, D.C.: June 23, 1997).

of the LLNL model, which showed much larger plume size. Given the
uncertainties in source term data and divergences in model results, DOD
cannot determine-with any degree of certainty-the size and path of the
plumes or who was or who was not exposed.

Recommendations for Executive Action

Agency Comments and Our Evaluation

We recommend that the Secretary of Defense and the Secretary of Veterans
Affairs not use the plume-modeling data for future epidemiological studies
of the 1991 Gulf War, since VA and DOD cannot know from the flawed plume
modeling who was and who was not exposed.

We recommend that the Secretary of Defense require no further plume
modeling of Khamisiyah and the other sites bombed during the 1991 Gulf War
in order to determine troops' exposure. Given the uncertainties in the
source term and meteorological data, additional modeling of the various
sites bombed would likely result in additional cost while still not
providing DOD with any definitive data on who was or was not exposed.

We obtained comments on a draft of this report from VA, DOD, and the CIA.
VA concurred with our first recommendation (see appendix V). Nevertheless,
VA stated that it has already completed three studies that incorporated
the DOD plume model as part of the parameters for the research and has
submitted these studies to scientific journals for publication. In
addition, VA is currently collaborating with other research groups that
may have used the DOD plume model. These studies are under way and will be
completed as planned. Given our assessment, it is important that VA inform
the researchers to include appropriate caveats, indicating the limitation
of the conclusions based on flawed modeling data.

DOD did not concur with our first recommendation, indicating that the "GAO
recommendation apparently represents a blanket prohibition against plume
modeling in the future, where the limitations of the 1991 Gulf War may not
apply" (see page 77). The intent of our recommendation was directed only
at epidemiological studies involving the DOD and CIA plume modeling data
from the 1991 Gulf War and was not a blanket prohibition of plume modeling
in the future (see appendix VI). We have clarified the recommendation
along these lines.

The CIA did not concur with the report, indicating that it could not
complete a review in the time allotted. The CIA indicated that a

comprehensive review would require 3 to 4 weeks. Nevertheless, from its
preliminary review of our report, the CIA identified several statements
that it considered inaccurate, such as those about source term data. The
CIA contended that the agent source term is complete and accurate to a
known certainty. Since our initiation of this inquiry in late 2002, we
have met with the CIA on a number of occasions, most recently on April 7,
2004, to identify the methodologies pursued in establishing source term
parameters used in the modeling. At the suggestion of the CIA, we met with
UNMOVIC officials to determine what UNSCOM inspections disclosed and the
methodologies used in determining source term data.

Our point-by-point evaluation of the detailed comments provided by DOD are
presented in appendix VI.

As we agreed with your offices, unless you publicly announce the contents
of this report earlier, we plan no further distribution until 30 days from
its issue date. We will then send copies of the report to other interested
congressional members and committees. In addition, the report will be
available at no charge on GAO's Web site at http://www.gao.gov.

If you or your staff have any questions about this report or would like
additional information, please contact me at (202) 512-6412 or Sushil
Sharma, Ph.D., Dr.PH., at (202) 512-3460. We can also be reached by e-mail
at [email protected] and [email protected].

Individuals who made key contributions to this report were Venkareddy
Chennareddy, Susan Conlon, Neil Doherty, Jason Fong, Penny Pickett, Laurel
Rabin, Katherine Raheb, and Joan Vogel. James J. Tuite III, a GAO
consultant and recognized expert on Gulf War issues, provided technical
expertise.

Keith Rhodes, Chief Technologist Center for Technology and Engineering
Applied Research and Methods

Appendix I: DOD's Chronology of Khamisiyah Modeling Events

                                   Date Event

1995

June	DOD formed the Persian Gulf Illnesses Investigation Team; by October,
it had identified some of the U.S. forces that had occupied the area
around Khamisiyah during the Gulf War, including the 37th Engineer
Battalion

Aug. President Bill Clinton created the Presidential Advisory Committee on
Gulf War Veterans' Illnesses

1996

Spring 	The Presidential Advisory Committee on Gulf War Veterans'
Illnesses directed the CIA to model chemical agent release from Bunker 73

                        May UNSCOM inspected Khamisiyah

June 	DOD confirmed publicly that "US soldiers from the 37th Engineer
Battalion destroyed ammunition bunkers [at Khamisiyah] in early March
1991.... It now appears that one of these destroyed bunkers contained
chemical weapons"

July The CIA briefed the Presidential Advisory Committee on Bunker 73
modeling results

Oct. 	The CIA requested LLNL to perform atmospheric dispersion
calculations, using a hypothetical release scenario; the Deputy Secretary
of Defense sent a memorandum to 21,000 veterans who had been identified as
being within 50 km of Khamisiyah

Nov. 	The Secretary of Defense established the Office of the Special
Assistant for Gulf War Illnesses (OSAGWI) to focus ongoing DOD
investigations and expand the investigation into Gulf War veterans'
complaints of undiagnosed illnesses

Dec. 	IDA released its interim report, critical of the model the CIA used;
it recommended rocket demolition testing to determine how rockets behaved
without high explosives and an ensemble approach with prognostic models

1997

Jan. 	The Deputy Secretary of Defense sent letters with a survey to
veterans, saying that chemical weapons had been present at Khamisiyah when
the demolitions occurred and urging them to call the Persian Gulf Incident
Hotline with any additional information they had about Khamisiyah

Jan.-Feb. 	The Special Assistant agreed to remodel Khamisiyah, as well as
Al Muthanna, Muhammadiyat, and Ukhaydir, following entreaties by the
Presidential Advisory Committee on Gulf War Veterans' Illnesses and IDA's
recommendations

May 	DOD and the CIA conducted open-field demolition tests on 122 mm
rockets similar to those destroyed in the Khamisiyah pit at Dugway Proving
Ground, Utah, to determine how CW agents in Iraq's rockets might have been
released

May (to June The Department of the Army and OSAGWI hosted G3/S3
conferences to elicit more correct information on unit 1998) locations
during the war

               June-July CW agent release was modeled at the pit

July	IDA released its final report, "Report of the Panel Reviewing
Analysis of the Khamisiyah Pit Release of Nerve Agent, March 1991"; DOD
and the CIA jointly announced the results of Khamisiyah dispersion
modeling. Given the unit locations available then, the modeling indicated
a hazard area where troops may have been exposed to low levels of nerve
agent. DOD sent written notices to 98,910 veterans in the potential hazard
area and approximately 10,000 notices to those who had received the Deputy
Secretary of Defense's letter and survey but were not in the potential
hazard area

Dec. 	An independent scientific panel (Anthes and others) reviewed and
commented on the methodology used to complete the Khamisiyah modeling,
making recommendations for improvements in future modelinga

           Appendix I: DOD's Chronology of Khamisiyah Modeling Events

Date Event

1998: Jan.	President Clinton created the Presidential Special Oversight
Board for the Department of Defense Investigations of Gulf War Chemical
and Biological Incidents to provide recommendations, based on its review
of DOD investigations into possible detection of, and exposures to,
chemical or biological weapons agents, as well as environmental and other
factors that might have contributed to Gulf War illnesses

2000

Jan. 	DOD completed efforts to remodel the Khamisiyah release using
updated meteorological and dispersion models, with revised source terms
from the CIA and with a better understanding of where U.S. forces had
been. As a result, DOD's estimate of the number possibly exposed increased
by about 2,000; almost 35,000 troops who had previously been notified of
possible exposure were no longer in the possible hazard area, whereas
about 37,000 newly identified troops probably were

Mar. 	The scientific review committee completed its review of the revised
methodology, commenting that "the methodologies are sound" and "the
results ... very likely overestimate the dosages actually received by
personnel"

2002: Apr. DOD published its final Khamisiyah report, with complete
technical report documenting methodologies

Source: Department of Defense, Office of the Assistant Secretary of
Defense for Health Affairs, Deployment Health Support Directorate.

a Richard A. Anthes and others, "Comments by Review Panel on Khamisiyah
Modeling Report and Presentations on November 4-5, 1997," report for the
Directorate for Deployment Health Support of the Special Assistant to the
Under Secretary of Defense (Personnel and Readiness) for Gulf War
Illnesses, Medical Readiness, and Military Deployments, Fairfax, Virginia,
December 11, 1997.

                         Appendix II: Power-Law Formula

A 1969 Sandia National Laboratories empirical study established a powerlaw
formula for calculating plume heights attributable to high-explosive
detonations. The power-law formula was derived from data on 23 test shots,
ranging from 140 lbs. to 2,242 lbs. of high explosives at DOE's Nevada
Test Site (the National Exercise, Test, and Training Center), and it
provides a cloud-top height at 2 minutes after detonation. Most of the
shots were detonated during near-neutral conditions, where the clouds
continued to rise after 2 minutes; data for 5 minutes after detonation on
some shots show tops rising to nearly double the 2-minute values. The
2minute values better represent the final cloud-top heights during stable
conditions.

This formula is represented as

h = 76 (w1/4)

where

                       h = height of plume in meters and

                       w = weight of explosives in pounds

According to this formula, an MK-84 or GBU-24 bomb (942.6 lbs. of high
explosives) would generate a plume of 421 meters:

h = 76 (942.6 pounds of high explosives)1/4

h = 76 (5.541)

h  o  421 meters

                     Appendix III: DOD's Model Divergences

Even among the models selected for use by the DOD panel, widely divergent
directional outcomes were observed. For example, in figure 13, the
differences among various models for hazard areas during the first 2 days
of the modeling period for Khamisiyah can be seen.

  Khamisiyah Pit Demolition - March 10, 1991 Modeled Exposure (Deposition and
           Degradation) Modeled Exposure (Deposition and Degradation)

                                       N

                  VLSTRACK/COAMPS VLSTRACK/COAMPS HPAC/COAMPS

HPAC/COAMPS

N

                                    HPAC/MM5

Angle approximately equal Angle approximately equal to 40-45 degrees to 80
degrees

Each dot represents a troop unit's location or part of a unit's location.
Each unit can range in size from a company to a batallion.

Source: Naval Surface Warfare Center.

In the March 10, 1991, section of the figure, an approximately 40 to 45
degree divergence between the HPAC/OMEGA and the HPAC/COAMPS models can be
seen; in the March 11, 1991, section, an approximately 80 degree
divergence can be seen.

The uncertainty attributed to this divergence is not limited to the
Khamisiyah modeling. According to a modeling analyst involved with the
modeling of Al Muthanna, the COAMPS and OMEGA weather models showed the
plume going in different directions, at a difference of 110 to 120
degrees. According to the analyst, COAMPS showed the plume going north
northwest, while OMEGA showed it going south. Similar divergence

                     Appendix III: DOD's Model Divergences

among model projections was also observed in the modeling of Muhammadiyat,
as shown in figure 14.

                         Source: Department of Defense.

Appendix IV: Divergence and Wind Field Models

In figure 15, LLNL projections for divergence of wind field vectors 6.0 m
above terrain are based on observational data the Meteorological Data
Interpolation Code (MEDIC) model processed.

Appendix IV: Divergence and Wind Field Models

Source: Lawrence Livermore National Laboratory.

Page 70 GAO-04-159 Gulf War Illnesses

Appendix IV: Divergence and Wind Field Models

Figure 16 shows the wind field vector model, based on European Centre for
Medium-Range Weather Forecast (ECMWF) projections and processed by the
MEDIC model.

Source: Lawrence Livermore National Laboratory.

Page 71 GAO-04-159 Gulf War Illnesses

                 Appendix IV: Divergence and Wind Field Models

The wind field vector model in figure 17 is based on COAMPS simulations at
the U.S. Naval Research Laboratories.

                Source: Lawrence Livermore National Laboratory.

                     Page 73 GAO-04-159 Gulf War Illnesses

Appendix VI: Comments from the Department of Defense

Note: GAO comments supplementing those in the report text appear at the
end of this appendix.

              Appendix VI: Comments from the Department of Defense

                                 See comment 1.

              Appendix VI: Comments from the Department of Defense

                                 See comment 2.

                                 Now on p. 60.

                                 See comment 3.

              Appendix VI: Comments from the Department of Defense

See comment 4.

See comment 5. See comment 6. See comment 7.

See comment 8.

See comment 9.

See comment 10.

See comment 11.

Now on pp. 50-55.

              Appendix VI: Comments from the Department of Defense

Now on p. 1.
See comment 12.

Now on p. 2.
See comment 13.

Now on p. 2.
See comment 14,

Now on p. 2.
See comment 15.

Now on p. 3.
See comment 16.

Now on p. 6.
See comment 17.

Now on p. 5.
See comment 18.

              Appendix VI: Comments from the Department of Defense

See comment 19.

Now on p. 10. See comment 20.

Now on p. 11. See comment 21.

Now on p. 11. See comment 22.

Now on p. 12. See comment 23.

Now on p. 12. See comment 24.

Now on pp. 13-14. See comment 25.

              Appendix VI: Comments from the Department of Defense

Now on p. 13. See comment 26.

Now on p. 14. See comment 27.

Now on p. 14. See comment 28.

Now on p. 14. See comment 29.

Now on p. 15. See comment 30.

              Appendix VI: Comments from the Department of Defense

Now on p. 16. See comment 31.

Now on p. 18. See comment 32.

Now on p. 18. See comment 33.

Now on p. 19. See comment 34.

Now on p. 19. See comment 35.

Now on p.19. See comment 36.

Now on p. 20 See comment 37.

Now on p. 20. See comment 38.

              Appendix VI: Comments from the Department of Defense

Now on p. 20. See comment 39.

Now on p. 21. See comment 40.

Now on p. 23. See comment 41.

Now on p. 25. See comment 42.

Now on p. 27. See comment 43.

Now on p. 27. See comment 44.

              Appendix VI: Comments from the Department of Defense

Now on pp. 27 and 29. See comment 45.

Now on p. 28. See comment 46.

Now on p. 29. See comment 47.

Now on p. 30. See comment 48.

Now on pp. 30-34. See comment 49.

              Appendix VI: Comments from the Department of Defense

Now on p. 31-32. See comment 50.

Now on p. 32. See comment 51.

Now on p. 34. See comment 52.

Now on p. 34. See comment 53.

Now on p. 34. See comment 54.

              Appendix VI: Comments from the Department of Defense

                               Now on pp. 35-37.

                                See comment 55.

Now on p. 37.

              Appendix VI: Comments from the Department of Defense

See comment 56.

Now on p. 38. See comment 57.

Now on pp. 41 and 43. See comment 58.

              Appendix VI: Comments from the Department of Defense

Now on p. 44. See comment 59.

Now on p. 44. See comment 60.

Now on p. 45. See comment 61.

Now on p. 45. See comment 62.

Now on pp. 49-60. See comment 63.

              Appendix VI: Comments from the Department of Defense

                               Now on pp. 51-60.

                                See comment 64.

              Appendix VI: Comments from the Department of Defense

Now on pp. 59-60. See comment 65.

Now on pp. 50-51. See comment 66.

              Appendix VI: Comments from the Department of Defense

Now on pp. 49-50. See comment 67.

Now on pp. 70-72. See comment 68.

              Appendix VI: Comments from the Department of Defense

Now on p. 72. See comment 68.

              Appendix VI: Comments from the Department of Defense

                                  GAO Comments

1. 	The use of the phrase "flawed plume modeling" in the report refers to
the use of DOD models that were not fully developed for analyzing
long-range dispersion of CW agents as an environmental hazard. In
addition, the uncertain source term data used resulted in a flawed
modeling outcome. Meteorological and dispersion modeling, as predictive
and diagnostic tools, can have significant value in cases where detailed
meteorological data are unavailable and in providing warning for potential
environmental hazards, assuming that the necessary input data supplied to
the model are accurate. DOD now asserts that it has made significant
improvements in its models; however, we have not evaluated DOD's
assertion, since it was beyond the scope of this study.

2. 	We revised the recommendation to state: "We recommend that the
Secretary of Defense and the Secretary of VA not use the plume modeling
data as a basis for future epidemiological studies of Gulf War Illnesses
in Iraq, since DOD and VA cannot know who was or who was not exposed."

DOD correctly states that the necessary input data (i.e., source term and
meteorological data) were not available. However, the models that DOD used
were not fully developed for long-range dispersion of CW agents as an
environmental hazard. Consequently, the modeling results were not fully
reliable for determining which troops were exposed or were not exposed.

The report did not intend to suggest that modeling, in general, is a
flawed approach for predicting the hazard potential resulting from the
release of toxic materials; rather, it intends to suggest that for a
retrospective event, such as events at Khamisiyah, the use of models that
were not fully developed for deriving long-range environmental hazards, in
conjunction with the uncertain source term data used, resulted in a flawed
modeling outcome. As we mentioned before, DOD now asserts that it has made
significant improvements in its models; however, we have not evaluated
DOD's assertion, since this was beyond the scope of this study.

3. 	An ensemble approach can be a useful tool in addressing issues of
uncertainty. However, the process DOD used discounted at least two
simulations using models that resulted in plume footprints that either
were much larger or traveled in different directions or both. To properly
account for uncertainty, and the untested ability of the several DOD
models to estimate long-range environmental fallouts, other more mature
models such as the MATHEW/ADPIC models from LLNL should have been included
in the model. In the absence of their inclusion, evidence that the plumes
did not travel in a divergent direction needs to be produced.

Page 92 GAO-04-159 Gulf War Illnesses

              Appendix VI: Comments from the Department of Defense

4. 	Again, our conclusion was not intended to suggest that modeling, or
even ensemble modeling, is a flawed approach to predicting the hazard
potential resulting from the release of toxic materials. Rather, it was
intended to suggest that using models that were not fully developed for
deriving environmental fallout estimates in conjunction with uncertain
source term data resulted in a flawed modeling outcome and unsupported
results. In addition, the selection of data and sites for modeling has a
profound impact on model predictions. It is this area of the use of
incomplete and improbable data and uncertainty in the selection criteria
for sites and times to model that has resulted in flawed modeling
outcomes.

5. 	The objectives of this report were not to evaluate DOD's modeling
capabilities in general but, rather, to (1) determine the validity of DOD
and British Ministry of Defense (MOD) conclusions-based on CIA and DOD
plume-modeling results- regarding U.S. and British troops' exposure to CW
agents during the 1991 Gulf War; (2) determine the total costs for the
CIA's and DOD's various plumemodeling efforts related to these exposures;
and (3) determine DOD and VA conclusions from epidemiological studies,
based on DOD's plume-modeling results, that there was no association
between CW exposure at Khamisiyah and the troops' hospitalization and
mortality rates. In other words, we examined how DOD's capabilities were
applied to this specific case.

6. This comment is dealt with in detail in addressing specific DOD
comments below.

7. 	Our observation regarding inaccurate, inappropriate, and incomplete
source term data and assumptions would apply equally to simulations
conducted throughout the period from 1996 to the present. Because flawed
data were fed into those models, the fact remains that the modeling
results are unsupported. As we mentioned before, DOD now asserts that it
has made significant improvements in its models; however, we have not
evaluated DOD's assertion, since it was beyond the scope of this study.

8. 	The report refers to panels of experts who reviewed DOD reports and
made comments and recommendations regarding the DOD modeling efforts. In
at least one case, we have documented where an expert made a
recommendation regarding the potential presence of meteorological
phenomena not addressed in the DOD modeling studies.

9. 	We reviewed all published studies as well as technical reports DOD and
VA prepared. We agree with DOD that "scientifically peer-reviewed and
published" is considered a high standard of validity when it implies
anonymous review by scientists coordinated by the editor of a reputable
scientific journal leading to publication of findings in that scientific
journal. The peer review and 2000 publication to which DOD refers here had
not gone through that process. The 2002 Kang and Bullman study has not
been published in a peer-reviewed journal

Page 93 GAO-04-159 Gulf War Illnesses

              Appendix VI: Comments from the Department of Defense

and therefore should not have been included in a review of the scientific
epidemiologic literature. This subject is discussed in further detail in
our response to DOD's comments on the section "DOD's and VA's
Epidemiological Conclusions on CW Exposure and Hospitalization and
Mortality Rates Cannot Be Adequately Supported."

10. As mentioned in response to comment 7 above, the inaccurate,
inappropriate, and incomplete source term data and assumptions that we
describe in the report would apply equally to simulations conducted
throughout the period from 1996 to the present. The statement in appendix
I reflects DOD's chronology of modeling events, not our assessment or
conclusion.

11. One of the central conclusions of our report is that DOD's plume
modeling was flawed, and this conclusion applied to the 2000 plume
modeling as well as to the 1997 plume modeling. As for the comment that
"GAO never mentions several completed epidemiology studies that used the
results of the 2000 plume modeling," these studies were not mentioned for
varying reasons. Smith and others, showed that hospitalization rates for
several ICD-9 diagnoses were higher in veterans categorized in the
Khamisiyah 2000 plume than in those not in the plume, and the association
for cardiac arrhythmias was statistically significant. However, that study
suffered from the same deficiencies as the earlier study that we
cited-namely, inappropriate use of hospitalization outcome measures rather
than appropriate use of measures of Gulf War illness, which usually do not
result in hospitalization, and use of plume modeling based on flawed data.
The 2002 Kang and Bullman study has not been published in a peer-reviewed
journal and therefore should not have been included in a review of the
scientific epidemiologic literature. This subject is discussed in further
detail in our response to DOD's comments on the section "DOD's and VA's
Epidemiological Conclusions on CW Exposure and Hospitalization and
Mortality Rates Cannot Be Adequately Supported."

In response to the comment that "GAO's review of the literature is
selective [and] incomplete," we do not agree with this characterization of
the literature review. We reviewed all published literature. The review of
the literature was focused on assessing the validity of DOD and VA
conclusions from the epidemiological studies, based on DOD's
plume-modeling results that there was no association with CW exposures at
Khamisiyah and troops' hospitalization and mortality rates. We address
specific reasons for excluding reports DOD identified in our response to
DOD's comments in the section "DOD's and VA's Epidemiological Conclusions
on CW Exposure and Hospitalization and Mortality Rates Cannot Be
Adequately Supported."

With respect to the comment that we ignored the Department of Defense
Low-Level Chemical Warfare Agents (CWAs) Research Master Plan, June 2003,
or the results of any of the research indicated in the plan, we repeat
that the review of

Page 94 GAO-04-159 Gulf War Illnesses

              Appendix VI: Comments from the Department of Defense

the literature was focused on assessing the validity of DOD and VA
conclusions from the epidemiological studies, based on DOD's
plume-modeling results that there was no association with CW exposures at
Khamisiyah and troops' hospitalization and mortality rates.

With regard to the comment that "GAO does not explain what it thinks would
be a better method to assess exposure," we believe that some of the
methodologies we cite in this report provide valuable insight into the
identification of both the effects of exposure and those who were likely
to have been affected by exposure. The methodologies include the
identification of biomarkers or genetic polymorphisms, animal model
studies that attempt to recreate the suspected event and evaluate the
appearance of a similar outcome, epidemiological studies in which the
cohort classifications can be safely made (such as deployed, nondeployed,
or deployed outside the period of potential exposure).

Regarding the comment that we "appear to favor some studies of 1991 Gulf
War veterans that are mentioned on pages 52 to 61 [now 50 to 55], which
seems to imply that GAO thinks these studies used a superior method of
exposure assessment, compared to the DOD modeling," we believe that in the
absence of more reliable meteorological and source term data relating to
the 1991 Gulf War, the DOD plume modeling must be considered inferior to
other methodologies that avoid the selection bias likely to be introduced
as a result of using the DOD Khamisiyah ensemble plume modeling, for the
reasons cited in this report.

Concerning the comment "some of the studies GAO mentioned did not evaluate
possible chemical warfare exposure at all," we evaluated only studies that
examined possible CW agent exposure, genetic polymorphisms believed to be
associated with CW agent exposure, animal model studies associated with CW
agent exposure, and studies that used the DOD ensemble plume models (1997,
2000).

12. The initiation of the bombing of Iraq's CW research, production, and
storage sites began on January 17, 1991. The release of environmental
hazards associated with the open-air destruction of these facilities would
have commenced on that date rather than on the ground invasion that began
on February 25, 1991. U.S .and British forces did not have to penetrate
these sites to be at risk from the potential health consequences of the
fallout of material released as a result of these bombings. In addressing
fallout, we were referring to hazardous materials thrust into the air and
potentially exposing troops to CW agents at subacute levels at significant
distances downwind. "Located near" is a relative term intended only to
reflect that the troops were close enough to be at risk for exposure. This
is not only our observation; it was also a concern of the war planners
before the onset of Operation Desert Storm, as reflected by requests for
hazard assessments modeling the U.S. Air Force made to Lawrence Livermore
National Laboratories.

Page 95 GAO-04-159 Gulf War Illnesses

              Appendix VI: Comments from the Department of Defense

13. We concur and the report has been clarified accordingly.

14. We concur and the report has been clarified accordingly.

15. We concur with this comment and have clarified the report accordingly.

16. We concur with this comment and have clarified the report accordingly.

17. Our statement addresses not how DOD classified the troops during the
modeling process but, rather, how researchers later used these data to
classify troops into exposed and not exposed study cohorts in conducting
health-related studies.

18. Our determination that conclusive classification criteria are
unsupported does not mean or assume that the entire population was
exposed. But the classification by researchers is not supported by
reliable scientific, or any other, evidence since the determination of who
was and who was not exposed is based on flawed and inaccurate data and the
exclusion from the ensemble modeling estimates of modeling simulations
that projected larger and directionally divergent potential exposure
areas. Given the uncertainty associated with determining who was and who
was not exposed and with determining whether or not the demolition of the
Khamisiyah pit represents a single exposure event, researchers will have
to assess, independent of the modeling performed and the ensemble
projections, who was and who was not exposed.

The observation that not all those who were exposed are ill applies
equally, whether the DOD models and the ensemble estimates are viewed as
accurate or flawed. This phenomenon is much more likely to be attributable
to the genetic susceptibility of certain individuals to be physically
affected by exposure at varying exposure levels.

19. The recommendation is intended to apply only to the 1991 Gulf War, and
we have amended the recommendation to clarify its intent.

20. While it is not part of modeling methodology, it was part of the
process DOD used in examining this issue. The report has been clarified.

21. We concur with this comment and have clarified the report accordingly.

22. DOD is correct in characterizing SCIPUFF as a component of HPAC and we
have changed the language accordingly. DOD comments regarding NUSSE and
ADPIC, however, are contradicted by the record. On September 4, 1997, the
CIA and DOD issued a report entitled "Modeling the Chemical Warfare Agent
Release at the Khamisiyah Pit (U)." In that report, the CIA identified
transport and diffusion models used in this effort to include SCIPUFF,
VLSTRACK, and NUSSE4. The MATHEW/ADPIC suite of models were, based on
documentation supplied to us by

Page 96 GAO-04-159 Gulf War Illnesses

              Appendix VI: Comments from the Department of Defense

officials at the National Atmospheric Release and Advisory Capability,
Lawrence Livermore National Laboratory, used in simulations of the
Khamisiyah pit performed at the request of IDA in late 1996 and early
1997.

23. The section DOD cites does not imply that DOD or the CIA suggested
that there can be "definitive conclusions" from the modeling process.
Rather, it states that "The models are neither sufficiently certain nor
precise to draw definitive conclusions about the size or path (that is,
the direction) of the plumes." It was the use of this information to
define study cohorts in follow-on government funded health-related
research that resulted in our comment.

24. We concur with this comment and have clarified the report accordingly.
However, the MATHEW/ADPIC models were used in the simulations of the
Khamisiyah pit, performed at IDA's request in late 1996 and early 1997.

25. DOD is correct that MM5 and COAMPS are commonly used mesocale
meteorological models. However, when they were used with dispersion models
(such as VLSTRACK), which were not fully developed for long-range
environmental hazards and in conjunction with uncertain meteorological and
source term data, meteorological models such as MM5 could not overcome the
limitations of the dispersion models. The validation of VLSTRACK at 220 km
still falls far short of the distances contemplated in this report.

26. This statement is based on an internal DOD memorandum, dated December
18, 1998, with the subject "Discounting the Results of the Omega Version
3.5 for the Khamisiyah Reanalysis and Al Muthanna Analysis." The
memorandum states that the Omega consistently underpredicts surface wind
speeds by a factor of 2 to 3 from actual observation collected at the five
world meteorological stations in the area.

27. This statement is based on a 1999 IDA report evaluating the variance
of VLSTRACK and HPAC predictions for the dispersion of chemical and
biological warfare agents. The IDA report noted that for chemical
releases, the HPAC and VLSTRACK predictions of areas of hazard differed
substantially and that for biological releases they differed by factors of
5 and 1,000.

28. The statement DOD quotes is based on a 1998 memorandum from DOD's
Deputy for Counterproliferation and Chemical and Biological Defense. This
memo states: "VLSTRACK and HPAC generate hazard predictions that are
significantly different from an operational perspective." Correspondence
to GAO from the Deputy's Modeling and Simulation Advisor noted that "the
1998 project found significant errors in the coding of one of the models
such that for analyses conducted prior to that date, I would not consider
that model reliable for use."

29. We concur with this comment and have clarified the report accordingly.

Page 97 GAO-04-159 Gulf War Illnesses

              Appendix VI: Comments from the Department of Defense

30. It is, in part, this observation that causes us to view as uncertain
the plume data DOD subsequently presented. Even the most elegant and
precise model will provide inaccurate results if it uses inaccurate data.

31. We reported that "Assumptions about the purity of the CW agents sarin
and cyclosarin established for Khamisiyah, Al Muthanna, Muhammadiyat, and
Ukhaydir differed widely. In each case, agent purity was a key factor in
the CIA and DOD methodology for determining the amount of agent released.
For example, for modeling purposes, 10 tons of agents with a purity of 18
percent would be represented as 1.8 tons of agent. The CIA relied on
UNSCOM reporting on the amount of CW agents Iraq produced. But to
establish these rates, UNSCOM relied primarily on Iraqi declarations and
Iraqi production records, as well as assumptions about the extent of agent
degradation." This section was included to demonstrate that varying rather
than consistent methodologies of differing levels of credibility were used
in deriving the estimated agent purity. It was also used to explain the
methodology used by the CIA and DOD in determining the maximum amount of
agent available for dispersion. The report was clarified to address DOD's
comment.

32. We were not suggesting that complete transfer occurred. Rather, we
were providing other science-based evidence that is contrary to the
earlier DOD observation that little or no transfer occurred. The
overpressures generated by high explosives have to go somewhere. In a
building, the structure would be destroyed. In a bunker, if the structure
were not destroyed, this overpressure release would occur through the
openings in the structure.

33. The photographs DOD referred to show the view from the opening in the
top of Bunker 2, as well as several aerial views. According to UNMOVIC,
however, UNSCOM did not physically inspect this bunker for safety reasons
relating to structural instability. This observation seems to be confirmed
by the photos referred to in the report DOD cited.

34. Given the images available, while the munitions may not have been
targeted directly, they certainly seem to have been hit.

35. We are aware of the design of the Dugway testing and its purpose to
simulate the demolitions at the Khamisiyah pit. Contrary to the DOD's
assertion, the use of Dugway testing data for evaporation and degradation
at Muhammadiyat is inappropriate for either leakage or destruction by
high-explosive bombs, neither of which were approximated by the conditions
at Dugway or at Khamisiyah. For example, the type and quantity of
explosives used in the Dugway testing and, therefore, the resulting
effects were not comparable to the type and quantity of munitions that
were actually used at Muhammadiyat. At Dugway Proving Ground, small
explosive charges were placed on boxed rockets; at Muhammadiyat, the
munitions were targeted with multiple high-explosive bombs.

Page 98 GAO-04-159 Gulf War Illnesses

              Appendix VI: Comments from the Department of Defense

36. The CIA Report on Intelligence Related to Gulf War Illnesses, dated
August 2, 1996, included modeling of Bunker 73 and identified the effort
as being done by the CIA in parallel with DOD's Persian Gulf Investigative
Team, to determine whether U.S. troops were exposed to chemical and
biological warfare agents during the Gulf War. The CIA's effort did not
seek to duplicate DOD's; however, CIA analysts drew on and examined DOD
information to clarify intelligence, obtain leads, and ensure a thorough
and comprehensive intelligence assessment. We have clarified the report to
reflect that this was a CIA modeling effort.

37. According to Iraqi declarations and UNSCOM, the stocks at Bunker 73
were part of the same lot as that discovered at the Khamisiyah pit, and
the munitions in the pit were estimated to have a purity of up to 50
percent when demolitions occurred. Therefore, a 2.5 percent purity rate is
not supported. We concur that the earlier modeling and source term were
performed by the CIA; however, later, after the IDA panel judged the
models used in the early modeling efforts to be inappropriate, no effort
was made to reexamine the releases from this site.

38. Our statement was taken out of context. The entire statement reads
"DOD reported that the Al Muthanna research, production, and storage
facility for CW agents was repeatedly attacked. Despite its repeated
bombing, however, on only one occasion did the CIA and DOD express any
concern about agent release. According to DOD analysis of the destruction
of Bunker 2 at Al Muthanna on February 8, 1991 . . . ." This is to suggest
not that the CIA and DOD are unconcerned about this issue but only that
they were only sufficiently concerned to publish the results of their
modeling in connection with this singular event at Al Muthanna, despite
the repeated bombings of this principal Iraqi CW agent research,
production, and storage facility.

39. We do not agree with DOD. According to UNMOVIC, UNSCOM never inspected
this bunker for safety reasons.

40. We explained the methodology used by DOD and have clarified the report
by deleting the comparison with Khamisiyah.

41. This section refers to the detections identified in table 4, of the
report, now on p.

26. The comment that no confirmatory testing was conducted is not
accurate. The Czech chemical detection unit reported the detections to
U.S. command officials immediately, as is reflected in both the Czech and
CENTCOM NBC logs, but the responding units were unable to confirm their
findings when they arrived hours after the initial detections on January
19, 1991, and were unable to confirm these reports. However, in addition
to the field detections, the Czech chemical detection units conducted
reagent based wet chemistry confirmation tests that supported the findings
of the initial detections. The French never officially acknowledged the
detections attributed to their units; however, the CENTCOM NBC logs again
noted that the French reported detecting chemical nerve agents on January
19,

Page 99 GAO-04-159 Gulf War Illnesses

              Appendix VI: Comments from the Department of Defense

1991, and that the Czechs confirmed the French detections. In addition,
according to an Agence France Presse report, on February 4, 1991, General
Raymond Germanos, a spokesperson for the French Ministry of Defense, was
attributed as having confirmed that chemical fallout, "probably
neurotoxins" had been detected in small quantities, "a little bit
everywhere," from allied air attacks of Iraqi CW facilities and the depots
that stored them.

42. Since this NOAA-11 AVHRR-1B imagery was being used to demonstrate
meteorological activity between January 18 and January 23, 1991, and not
at some specific time, the exact time each image was captured will not add
to or detract from its evaluation.

43. The coincidence between these and other events involving the
destruction of Iraq's CW agent research, production, and storage
infrastructure using aerial bombs and cruise missiles, and the reported
airborne detections of CW agents by Czech, French, U.S., and British
forces (see table 4) suggest that DOD and the CIA should have reassessed
their positions regarding the potential for additional exposure events.

44. We concur and the figure has been clarified accordingly.

45. The discussion in this section of the report deals with temperature
inversions. Unlike a capping inversion, which is almost a constant feature
in the atmosphere, a temperature inversion is not. When a temperature
inversion occurs, air pollutants can be trapped near the surface of the
Earth.

46. Figure 4 has been corrected to axis labels and time of day 0008Z.

47. In the April 2002 Gulf War Illness Task Force Report, "Intelligence
Update: Chemical Warfare Agent Issues during the Persian Gulf War," the
CIA assessed that "the Czech detections were unlikely to be from a
chemical agent." The Czech chemical detection units conducted wet
chemistry confirmation tests supporting the initial detections. We have
clarified the report to remove the reference to DOD in the relevant
sentence.

48. We agree that the initial modeling referred to was that performed by
LLNL for IDA and not included in the DOD ensemble model. This modeling
effort actually assumed fewer rockets destroyed than were later assessed
to be present at Khamisiyah and that produced a plume path considerably
larger and more divergent from those selected for use in the DOD ensemble.

49. The use of the power-law formula was intended to illustrate the
unreasonable heights assumed during the DOD modeling efforts. Clearly not
all the agent would be released simultaneously into the atmosphere and the
agent released would be distributed throughout the plume geometry, but the
power-law formula also

Page 100 GAO-04-159 Gulf War Illnesses

              Appendix VI: Comments from the Department of Defense

projects the plume height at time = 2 minutes. After that time, the plume
continues to grow in height. In some cases, at time = 5 minutes the plume
heights nearly double. Further, the distribution used in the Khamisiyah
pit is based on field and laboratory testing conducted at Dugway Proving
Ground that inadequately simulated the conditions at the pit and did not
simulate conditions at the other sites modeled at all. We concur that the
Khamisiyah pit demolition, which occurred around 4:15 pm local time, when
the atmospheric boundary layer was convective and well mixed, would have
been insensitive to the argument of nocturnal low-level jet. The other
sites, however, were bombed during the nighttime, and therefore this
argument remains valid. The power-law formula demonstrates the
relationship between the amount of explosives detonated and the resultant
plume height. Whether this occurs by demolition or aerial bombing is
irrelevant. The issue of top boundary versus centroid of the plume is
addressed in our report (see figure 8). Regarding buoyant puff at
Khamisiyah, DOD based its comment on Dugway field testing, which did not
realistically simulate actual conditions at the Khamisiyah site.

50. Videos of the demolition operations at Khamisiyah have been widely
released. While we concur that videos do not show data, they can certainly
demonstrate that DOD data assumptions, such as plume height estimates, are
inaccurate, because the plume height was higher than DOD assumptions.

51. Figure 6 shows the layers in which these activities occur. We have
changed the report to clarify this reference.

52. This statement should read "Empirical studies and observed events tend
to refute the assumptions with which the CIA and DOD discounted the
alternative assumption that the plume was transported by low-level jets."
The empirical studies are those involving the likely plume heights reached
in high-explosive explosions. The observed events are the reported
detections of CW agents associated with temperature inversion activity and
atmospheric turbulence. We changed the report accordingly.

53. We did not make this assumption; rather, it is a possibility we were
obliged to consider when evaluating the potential for the long-range
transport of CW agents. It is precisely the absence of on-site
measurements leading to this additional element of uncertainty for a
phenomenon that is far from rare that has resulted in our questioning
DOD's and CIA's discounting this phenomenon, despite having been cautioned
to consider the possibility by a DOD expert consultant.

54. Nothing in this section of the report suggests that either long-range
advection (transport of pollution) or turbulence events occurs independent
of dispersion or dilution. In fact, exposures occurring at these distances
would almost by necessity be at low or subacute levels. The report also
does not suggest that the low-level jets function independent of
turbulence. But aircraft and artillery would

Page 101 GAO-04-159 Gulf War Illnesses

              Appendix VI: Comments from the Department of Defense

produce a directional shift in turbulence, possibly resulting in mixing to
the surface. The characterization of the MATHEW wind field model as one we
favored is not accurate. The MATHEW/ADPIC suite simulations are simply
demonstrative of the uncertainty associated with the modeling process.

55. We understand the financial and practical limitations in conducting
this sort of testing. Still, differences in experimental conditions can
result in profound differences in outcome. For example, there may have
been more agent dispersed immediately, leaving less to evaporate over time
had the simulations been conducted under different conditions.

56. We agree with this comment. Predictive modeling is a crucial asset and
should be so considered. Retrospective modeling, however, in the absence
of robust data is far more easily criticized as "deficient," not
necessarily because of deficiencies in the models or the approach, but
because of the lack of validated input data and a selection process that
is subject to limited available data, an inexact intelligence assessment
process, and the potential for individual bias.

57. We do not agree with DOD's characterization. Despite not having
modeled the same quantities, LLNL modeled a variety of release scenarios
and used the meteorological data available for the region. This modeling
effort actually assumed a similar number of rockets destroyed as were
later assessed to have been destroyed at Khamisiyah by the CIA in 2002,
yet it produced a plume path considerably larger and quite divergent from
the models selected for use in the DOD ensemble. We do not understand why
IDA characterized the model as less capable. The LLNL models had an
established history of modeling the release of hazardous materials,
including the 1991 Kuwaiti oil fires.

In February 1991, during the last few days of the Gulf War, the Iraqis
ignited about 605 oil wells, causing an unprecedented environmental
disaster in the region. During spring and summer 1991, two working groups,
one sponsored by the World Meteorological Organization in conjunction with
the World Health Organization and the other consisting of the U.S.
government's scientific community, conducted airborne sampling programs to
evaluate the local and global consequences of these fires. The Atmospheric
Release Advisory Capability (ARAC), incorporating the MATHEW/ADPIC
modeling suite, provided daily forecasts of the location and density of
the smoke plumes in support of these aircraft missions, and concurrently
to all the countries affected.

Page 102 GAO-04-159 Gulf War Illnesses

              Appendix VI: Comments from the Department of Defense

The modeling was performed in the same region and during the same general
time

as the Khamisiyah event and the 1986 nuclear reactor accident at
Chernobyl.1 In

conjunction with the Chernobyl event, LLNL's long-range particle-in-cell
model

accurately simulated the spread of the radioactive cloud over the entire
northern

hemisphere, as verified later by radiological measurements. The NARAC

emergency response central modeling system LLNL currently uses consists of
a

coupled suite of meteorological and dispersion models. The data
assimilation

model ADAPT constructs fields of such variables as the mean winds,
pressure,

precipitation, temperature, and turbulence, using a variety of
interpolation

methods and atmospheric parameterizations. Nondivergent wind fields are

produced by an adjustment procedure based on the variational principle and
a

finite-element discretization. The dispersion model LODI solves the 3D
advection

diffusion equation using a Lagrangian stochastic; LODI includes methods
for

simulating the processes of mean wind advection, turbulent diffusion,
radioactive

decay and production, bioagent degradation, first-order chemical
reactions, wet

deposition, gravitational settling, dry deposition, and buoyant/momentum
plume

rise. The models are coupled to NARAC databases providing topography,

geographical data, chemical-biological-nuclear agent properties and health
risk

levels, real-time meteorological observational data, and global and
mesoscale

forecast model predictions. The NARAC modeling system also includes an in

house version of the Naval Research Laboratory's mesoscale weather
forecast

model COAMPS. This is a mesoscale meteorological model that LLNL has

incorporated into its modeling suite.

58. Regarding the comment, "the source term is different," LLNL modeled a
variety of release scenarios and used the meteorological data available
for the region. This modeling effort actually assumed a similar number of
rockets destroyed as were later assessed to have been destroyed at
Khamisiyah by DOD in 2002, yet it produced a plume path considerably
larger and quite divergent from the models selected for use in the DOD
ensemble. According to DOD's technical report, "Modeling and Risk
Characterization of U.S. Demolition Operations at the Khamisiyah Pit"
(April 16, 2002), released by William Winkenwerder, Jr., Assistant
Secretary of Defense (Health Affairs) and Special Assistant to the Under
Secretary of Defense (Personnel and Readiness) for Gulf War Illnesses,
Medical Readiness, and Military Deployments, the input source parameters
used in the 1997 Khamisiyah pit modeling included a best estimate of 500
rockets damaged in

1F. W. Whicker and others, "PATHWAY: A Dynamic Food-Chain Model to Predict
Radionuclide Ingestion after Fallout Deposition," Health Physics 52
(1987): 717-37; L. R. Anspaugh and others, "The Global Impact of the
Chernobyl Reactor Accident," Science 242 (1988): 1513-19; T. Straume and
others, "The Feasibility of Using I129 to Reconstruct I131 Deposition from
the Chernobyl Reactor Accident," Health Physics 71:5 (1996): 733-40; K. T.
Bogen and others, Uncertainty and Variability in Updated Estimates of
Potential Dose and Risk at a U.S. Nuclear Test Site Bikini Atoll,
UCRL-JC-122616 (Livermore, Calif.: Lawrence Livermore National Laboratory,
1995).

Page 103 GAO-04-159 Gulf War Illnesses

              Appendix VI: Comments from the Department of Defense

demolition, based on UNSCOM reporting and intelligence information.
According to the report, this number could be as high as 650 or as low as
170, based on number of rockets minus rockets found by UNSCOM that were
undamaged. The same report estimated that the total number of rockets in
the pit was 1,250. Again, according to the report, this number could be as
high as 1,400 or as low as 1,100, based on size of crates and stacks of
rockets.

Regarding the comment "area-defining thresholds are different," we concur.
However, the legend in figure 10 acknowledges this difference. Regarding
the comment "dosage accumulation times are different," we concur. Again,
however, this issue is addressed in the footnote defining these factors.
Regarding the comment "meteorological fields are different," again, we
concur with this observation. However, this comment is self-evident in
that the plume is moving in a different direction. It is important to note
that the corrections above do not have any effect on our conclusions. The
use of the composite image is intended to illustrate that the use of
different models, based on different underlying principles and
assumptions, can result in different outcomes. In that regard, it is not
misleading.

The fact that Iraq is a data-sparse region only serves to strengthen our
observation that the uncertainty associated with attempting to model the
fallout from the Khamisiyah pit, and to an equal or greater degree the
other sites modeled, is too great to provide meaningful results. If an
ensemble approach is to be attempted, then a range of methodologies needs
to be incorporated.

59. That our language "could be explained" suggests that it is only one
possible reason. When we interviewed researchers at LLNL, they noted that
this diffluence may account for the different outcomes.

60. Figure 12 not only illustrates intermodel bias; it also illustrates
model, and potential ensemble, bias. In the case illustrated, an ensemble
of the three models would still not incorporate the area containing the
actual hazard. We concur that ensemble modeling is essential in minimizing
uncertainty and providing hazard warning. But as illustrated by figure 12,
even in using an ensemble approach, significant uncertainty remains.

61. Our report has been clarified to reflect the use of consistent source
term for SCIPUFF and VLSTRACK in the 2000 modeling. This reconciliation in
source term does not, however, change any of our other observations or
recommendations regarding the uncertainty associated with the source term,
including the observation that, even with harmonized source, model
projections still differ.

62. We do not agree that this is a misleading statement. While each of
DOD's modeled plumes was based on data, the composite or ensemble plume
was based on a

Page 104 GAO-04-159 Gulf War Illnesses

              Appendix VI: Comments from the Department of Defense

simple graphic overlay of the projection of the three component plumes.
The relationship between these plumes, therefore, was not based on data.

63. One of the central conclusions of our report is that DOD's plume
modeling was flawed, and this conclusion applied to the 2000 plume
modeling as well as to the 1997 plume modeling, because both suffered from
the same weaknesses. That is, the models were not fully developed for
long-range environmental hazards, and source term and input data were
incomplete. Regarding the comment that "Several relevant studies have been
published . . . that GAO did not mention," we have now incorporated these
studies in our report.

64. As we explained in our report, the PON group of enzymes is a
potentially important predisposing factor in Gulf War illnesses because
one of its major functions in normal body physiology is to protect the
nervous system from organophosphate chemical toxins, such as pesticides
and nerve agents. This finding was remarkable because the only function of
Q type of the PON enzyme group is to protect the nervous system from nerve
agents sarin, soman, tabun, and VX. The R-type isoenzyme has as its main
function protection from organophosphate pesticides, such as diazinon,
malathion, and parathion. Thus, an association between Gulf War illnesses
and blood levels of only the Q-type isoenzyme of PON points specifically
to nerve agent exposure. Therefore, they were being cited to illustrate
studies that have examined paraoxonase deficiencies in veterans reporting
Gulf War Syndrome.

DOD's second paragraph is inaccurate. We did review the three studies
cited in DOD, footnote 14. First, we rejected the Greenberg and others
reference (second citation in DOD footnote 14). It is a letter to the
editor merely commenting on possible errors in the recall of vaccine
receipt in British Gulf War veterans; it did not deal with self-reports of
CW agents and was not peer reviewed.

The two other studies DOD cited, McCauley and others (the first citation
in DOD's footnote 14) and Wessely and others (the third citation in DOD's
footnote 14) both reported studies that measured the level of agreement on
self-reported endorsement of various wartime environmental exposures in
ill and well Gulf War veteran populations. These studies do not
demonstrate recall bias. In fact, they contradict that claim. DOD has
overlooked that recall bias results from nondifferential misclassification
of exposure measurements in the case and control groups. Simply finding
that some veterans misreported their exposures does not establish recall
bias; other criteria are required to establish that. Both studies showed
moderate to good levels of test-retest agreement (kappa >0.4) on
self-reports of CW agent exposure, but, more importantly, the level of
test-retest agreement (kappa statistic) did not differ between the ill and
well groups. This means that whatever errors in recall occurred were
nondifferential (occurred at the same rate in both groups), and therefore
they did not bias the estimates of the

Page 105 GAO-04-159 Gulf War Illnesses

              Appendix VI: Comments from the Department of Defense

relative risks for CW agent exposure. This means also that the errors in
recall did not result in recall bias.

65. In suggesting that "Most" of the animal studies "cannot be directly
extrapolated to the possible health effects of low-level sarin exposure in
1991 Gulf War veterans," DOD has missed the important developments in
recent studies cited in our report that can be directly extrapolated to
the health effects of low-level sarin exposure in the 1991 Gulf War. The
best examples are the DOD-funded experiments by Henderson and others from
the Lovelace Respiratory Research Institute and from the U.S. Army Medical
Institute of Chemical Defense (Toxicol Appl Toxicol 184 (2002): 67-76).
This study modeled the low-level inhalation exposure in rodents to sarin,
with and without heat stress. No immediate health effects were measured by
clinical indicators and brain pathological examination. However, 30 days
after cessation of exposure, the rodents were found to have developed
physical evidence of brain cell damage, demonstrated using sophisticated
microscopic, neurochemical brain examination techniques. Once structural
changes are found, the capability of sarin to cause chronic brain illness
is established, and following the animals in the longer term is
unnecessary. Other studies we cited add depth to the significant findings
of Henderson and others and establish a body of evidence demonstrating the
potential of subacute sarin exposures to cause brain cell damage and
related chronic symptoms.

66. DOD's response fails to distinguish between veterans having only
symptoms relating to different organ systems and veterans having classic
diseases of those organ systems to which physicians assign ICD-9 codes and
for which they often hospitalize patients. The public health problem of
1991 Gulf War veterans is characterized by a collection of symptoms of
various organ systems in which physicians typically do not recognize
classic, diagnosable diseases for which ICD9 codes exist. And since
veterans with this problem are generally not critically ill enough to
require hospitalization, their physicians do not hospitalize them more
commonly than they do veterans without the condition. This is why DOD and
VA studies with hospitalization as the outcome have not been productive.

The Gulf War illness manifested by serious symptoms but no physical signs
is the example of chronic illness in 1991 Gulf War veterans that would not
lead to hospitalization. Recent studies by several investigators at
different institutions have shown that this set of conditions stems from
physical damage to brain cells in deep parts of the brain.

By stressing that DOD "hospitalization data are very complete for
active-duty personnel," DOD overlooks the significant selection bias that
results from studies that rely on DOD hospitalization data alone as an
outcome measure. Since most of the more severely ill Gulf War veterans
left the military soon after the 1991 Gulf War, they were no longer
eligible for hospitalization in DOD hospitals and, thus, their further
hospitalizations were no longer counted. This "attrition" of the most

Page 106 GAO-04-159 Gulf War Illnesses

              Appendix VI: Comments from the Department of Defense

ill veterans creates a strong selection bias in these studies toward
falsely negative findings.

Gray and others (DOD's footnote 16) fails to address the observation that
Gulf War illness does not satisfy the diagnostic criteria for ICD-9 codes
for the illness categories Gray and others studied-for example, infectious
diseases, neoplasms, endocrine diseases, and blood diseases. Gray's
finding that these classic diagnoses, defined by ICD-9 codes, were no more
common in those who left the military soon after the war does not address
the issue of selection bias from using DOD hospital data. To the contrary,
other data Gray and others provided in their 1996 paper, reviewed by Haley
(Am J Epidemiol 148 (1998): 325-23), demonstrated conclusively that
military personnel discharged soon after the war did have largely
different reasons for being discharged from the service, which points to a
selection bias.

The references to the studies involving the California Office of Statewide
Health Planning and Development and VA are not relevant in assessing the
health effects of CW agent exposure, because these studies did not relate
illness to CW agent exposure measures.

67. We did review all the studies cited but found them unsuitable for
consideration for the following reasons. Smith and others (DOD's footnote
17) used inappropriate outcome measures (ICD-9 diagnoses made in
hospitalized patients) and unsupported measures of CW agent exposure-that
is, the 2000 Khamisiyah plume model addressed in the report. Consequently,
no useful conclusions can be derived from it. Kang and Bullman (DOD's
footnote 18) is an internal VA technical report that has not been peer
reviewed or published in a scientific journal and therefore is not
appropriate for inclusion in a review of the scientific epidemiologic
literature. McCauley and others (the first citation in DOD's footnote

19) found that 1991 Gulf War veterans who witnessed the Khamisiyah
demolition (a more valid measure of Khamisiyah exposure than the flawed
plume models) had more chronic symptoms on average than those who did not
witness it. This study suggests a causal role of Khamisiyah-associated
sarin exposure in chronic Gulf War illness. We did not include the paper
by Shapiro, Lasarev, and McCauley (second citation in DOD's footnote 19)
because an anomalous problem with its factor analysis method appeared to
disqualify its findings on a methodological basis. The investigators
performed factor analysis of symptoms in a random sample of 1991 Gulf War
veterans and found that those who actually witnessed the Khamisiyah
demolition were significantly more likely to have their syndrome factor 2
("dysesthesia syndrome"). While this appears to support a causal role of
Khamisiyah-associated sarin exposure in chronic Gulf War illness, the
authors also reported that their factor analysis method was unreliable
when applied to randomly generated variables. If this study were included,
however, it would suggest a causal link with CW agent exposure. As for the
two studies by Smith and

Page 107 GAO-04-159 Gulf War Illnesses

              Appendix VI: Comments from the Department of Defense

others (DOD's footnote 20), we did not include these because both used
participation in the DOD and VA registries as a proxy case-definition for
Gulf War illness. Since military personnel and veterans were free to
participate in the registry, regardless of whether or not they were ill
or, if ill, the nature of their illness, registry participation is an
entirely nonspecific measure and not suitable for scientific research on
the problem. The two studies showed that registry participation and
hospitalization were more common in veterans who were present at the
Khamisiyah risk area than those who were not. Although the findings would
support a causal link of CW agent exposure and illness, their methods do
merit inclusion in a scientific literature review.

68. We believe that the images in figure 17 are adequate to demonstrate
the modeled diffluence in wind field data. While we cannot improve the
quality of the figure, we have added an arrow to show the diffluence.

Page 108 GAO-04-159 Gulf War Illnesses

Appendix VII: Comments from the Central Intelligence Agency

(460530) Page 109 GAO-04-159 Gulf War Illnesses

GAO's Mission

Obtaining Copies of GAO Reports and Testimony

The General Accounting Office, the audit, evaluation and investigative arm
of Congress, exists to support Congress in meeting its constitutional
responsibilities and to help improve the performance and accountability of
the federal government for the American people. GAO examines the use of
public funds; evaluates federal programs and policies; and provides
analyses, recommendations, and other assistance to help Congress make
informed oversight, policy, and funding decisions. GAO's commitment to
good government is reflected in its core values of accountability,
integrity, and reliability.

The fastest and easiest way to obtain copies of GAO documents at no cost
is through the Internet. GAO's Web site (www.gao.gov) contains abstracts
and fulltext files of current reports and testimony and an expanding
archive of older products. The Web site features a search engine to help
you locate documents using key words and phrases. You can print these
documents in their entirety, including charts and other graphics.

Each day, GAO issues a list of newly released reports, testimony, and
correspondence. GAO posts this list, known as "Today's Reports," on its
Web site daily. The list contains links to the full-text document files.
To have GAO e-mail this list to you every afternoon, go to www.gao.gov and
select "Subscribe to e-mail alerts" under the "Order GAO Products"
heading.

Order by Mail or Phone 	The first copy of each printed report is free.
Additional copies are $2 each. A check or money order should be made out
to the Superintendent of Documents. GAO also accepts VISA and Mastercard.
Orders for 100 or more copies mailed to a single address are discounted 25
percent. Orders should be sent to:

U.S. General Accounting Office 441 G Street NW, Room LM Washington, D.C.
20548

To order by Phone: 	Voice: (202) 512-6000 TDD: (202) 512-2537 Fax: (202)
512-6061

To Report Fraud,	Contact: Web site: www.gao.gov/fraudnet/fraudnet.htm

Waste, and Abuse in E-mail: [email protected]

Federal Programs Automated answering system: (800) 424-5454 or (202)
512-7470

Jeff Nelligan, Managing Director, [email protected] (202) 512-4800

Public Affairs 	U.S. General Accounting Office, 441 G Street NW, Room 7149
Washington, D.C. 20548
*** End of document. ***