The results of these assessments differed so significantly that it was not possible to choose the most likely affected areas; hence, plots of the results did not provide predictions that were useful for helping to investigate possible effects on US forces. Much of the uncertainty emanated from differences in the way the meteorological models reconstructed the winds in the area at the time, as well as from the lack of definitive information about the source term. Another major uncertainty stemmed from the possibility that the models treated vertical diffusion differently. The wind in the area varied rapidly with height; thus, model differences in vertical mixing could result in large differences in the transport of agent.
More specifically, the DSWA and LLNL models showed similar initial directions for agent transport; however, the DSWA model predicted agent eventually turning to the west, while the LLNL model showed the agent continuing in a southeasterly direction. The NRL/NSWC models showed a more southwesterly initial vector but, like the DSWA model, the agent contours turned toward the west and then back toward the north as the wind changed. Review of the wind patterns reconstructed by the models suggests that part of the difference has to do with the altitude of the winds that control agent dispersion. In the case of the LLNL model, agent is moved by higher altitude winds whereas, for the other models, the boundary layer holds the agent nearer the surface.
Finally, in order to gain insight into the validity of the meteorological reconstruction based on so few observations in the immediate vicinity of Khamisiyah, the Panel asked that analyses that perturbed the weather inputs be conducted. The Naval Research Laboratory performed a number of reassessments of the meteorology with COAMPS and the global-scale model NOGAPS. These included "data denial" runs, where observations in the vicinity of Khamisiyah were ignored; a "constant baseline" run, where the synoptic input to COAMPS was held constant; and a "random perturbation" run, where local observations were randomly changed to represent observational error (a single component of an "ensemble" analysis).
These analyses suggest that the lack of local observations may be less of a concern than initially thought-that the meteorology was driven primarily by synoptic weather activity.