The first thing DoD did for their modeling was to collect all available weather observations. Although global weather information exists for this period, atmospheric conditions were generally poorly recorded for this region during the Gulf War. For its analysis, DoD used important declassified data (e.g., from US Air Force Special Forces) not available to the CIA during the preliminary modeling. However, even this data was not sufficiently detailed to predict accurate local and regional dispersion. To address this problem, the modeling team adopted a two step process similar to that used to model possible releases at Khamisiyah to reconstruct the weather conditions in the vicinity of Ukhaydir at the time of the release:

• global scale modeling to determine general weather patterns affecting the area; and,
• regional and local (i.e., mesoscale) weather modeling using the results of the global scale models and all available observations of specific local weather. Regional and local weather models incorporate information about local weather effects specific to the particular region in question.^{[65, 66]} For example, the weather at Ukhaydir was affected not only by the local terrain but possibly also by breezes from several lakes to the north. (See Figure 2.)

In keeping with the ensemble approach, the Ukhaydir modeling effort used four weather models:

• two global scale models:
  -  The National Centers for Environmental Prediction Global Data Assimilation System (GDAS);
  -  The Naval Operational Global Atmospheric Prediction System (NOGAPS);

• two regional and local weather models:
  -  The Coupled Ocean-Atmosphere Mesoscale Prediction System (COAMPS);
  -  The Operational Multiscale Environment Model with Grid Adaptivity (OMEGA).

After running these models, the modeling team had sufficiently detailed information about specific weather conditions for use in the transport and diffusion models. For example, DoD estimated that the near surface winds (approximately 10 meters above ground level) at Ukhaydir were generally light and variable for the time in question (approximately midnight on February 13, 1991, through February 16, 1991).

DoD used two different transport/diffusion models:

• The Second-order Closure Integrated Puff (SCIPUFF) model; and
• The Vapor, Liquid, and Solid Tracking (VLSTRACK) model.

In addition to weather data, transport/diffusion models require source characterization data and dosage information.

The source characterization was the same as that developed by the CIA for its preliminary modeling, i.e., an instantaneous release of seven gallons followed by a sustained release of 83 gallons over three days. Although the evaporation of agent from the crater would have taken place over a number of days, the decreasing release rate, as well as the generally calm weather conditions, would have limited any potential hazard areas after the first 72 hours. Dosage information determines which parts of the hazard areas represent the greatest potential danger for exposed personnel. Dosage is defined as concentration integrated over a specific time period of exposure. DoD used the general population limit to model the area of potential exposure. The general population limit is the average concentration below which the general population, including children and the elderly, could remain indefinitely with no effects. The general population limit for mustard over a 72 hour period is 0.432 milligram-minute per cubic meter of air (0.0001 x 72 hours x 60 minutes).^[67] The hazard areas shown in Figures 15, 16, and 17 depict the area where the level of mustard is above the general population limit.

The modelers did not take into account certain mechanisms that could have limited the extent of any exposure area (e.g., agent decay over time such as that caused by the possible effects of exposure to sunlight). Because there was no specific information about how these mechanisms might affect the amount of agent degraded during transport and dispersion, the modelers used a more conservative approach and did not use them to limit the extent of the hazard areas.

To produce the ensemble hazard areas, the modeling team created a composite of three different combinations of weather and diffusion model simulations:

• NOGAPS/COAMPS/SCIPUFF;
• GDAS/OMEGA/SCIPUFF; and
• NOGAPS/COAMPS/VLSTRACK.

The modelers then overlaid the hazard areas produced by these three simulations and used the outermost perimeter of the union to define an ensemble hazard area.

Any assessment of the likelihood of exposure of US troops to mustard agent released from Ukhaydir depends on how close the modeled hazard areas came to known troop locations. The US Armed Services Center for Unit Records Research conducted an extensive review of Gulf War unit records to determine where units were located during February 14, 15, and 16, 1991. While impossible to recreate precisely where each servicemember was during every moment of every day, Figures 15, 16, and 17 generally reflect locations where the majority of unit members were as they performed their missions.^[68]

| First Page | Prev Page | Next Page |