Thus, with or without tests, ranges of reasonable and defensible source term factors
should be selected (although field tests should help to narrow some of these ranges). Best
estimates and reasonable bounding values for at least the following characteristics should
be determined:
While the number of combinations could be large, a fairly rich representative sampling should be chosen to ensure that the range of possible outcomes is investigated. This is particularly important since different cases will result in different coverage patterns, and units that might be exposed in some cases might not be in others. This approach can be used to differentiate among the units that would not have been exposed in any case, the units that would have been exposed in all cases, and the units that may have been exposed in some cases.
Fortunately, this aspect of the simulation is less computer intensive than the meteorological reconstructions, and a large number of different cases can be examined for each meteorological analysis. It is important, however, to use an accepted transport and diffusion model that can model the key phenomena associated with the source term and that can accept the output of a meteorological analysis.
As we have noted above, there are sparse local observational data to 'support mesoscale modeling at the time of the Khamisiyah pit demolition. Moreover, reconstructing the meteorological conditions at the time requires considerable computer resources, making the exploration of variations in the meteorology time-consuming and expensive.
The Panel recommends that models well known and accepted in the
community be used for this purpose. Among the models mentioned in this regard are MM5,
RAMS, and ETA. The COAMPS and OMEGA models are not as well-established in the
meteorological community, although both appear to be capable models. Comparisons among
MM5, COAMPS, and OMEGA indicate that these models produce similar reconstructions of the
meteorology at Khamisiyah. The MATHEW model used by LLNL was viewed to be less capable
because it models atmospheric phenomena with less fidelity. Because all models have
relative strengths and weaknesses, using several models to reconstruct the meteorology
will help bound model-induced (as opposed to data- induced) uncertainties in the results.