Atmospheric test fallout data have been used to determine admissible dust particle size distributions for nuclear winter and nuclear fallout studies. The research was originally motivated by extreme differences noted in the magnitude and longevity of dust effects predicted by particle size distributions routinely used in fallout predictions versus those used for nuclear winter studies. Three different sets of historical data have been analyzed:

1. Stratospheric burden of Strontium-90 and Tungsten-185, 1954-1967 (97 contributing events) 2. Continental U.S Strontium-90 fallout through 1958 (75 contributing events) 3. Local Fallout from selected Nevada tests (16 events)

The contribution of dust to possible long term climate effects following a nuclear exchange depends strongly on the particle size distribution. The distribution affects both the atmospheric residence time and optical depth. One dimensional models of stratospheric/tropospheric fallout removal were developed and used to identify optimum particle distributions. Results indicate that particle distributions which properly predict bulk stratospheric activity transfer tend to be somewhat smaller than number size distributions used in initial nuclear winter studies. In addition, both 90Sr and 185W fallout behavior is better predicted by the lognormal distribution function than the prevalent power law hybrid function.

It is shown that the power law behavior of particle samples may well be an aberration of gravitational cloud stratification. Results support the possible existence of two independent particle size distributions in clouds generated by surface or near surface bursts. One distribution governs late time stratospheric fallout, the other governs early time fallout. A bimodal lognormal distribution is proposed to describe the cloud particle population. The distribution predicts higher initial sunlight attenuation and lower late time attenuation than the power law hybrid function used in initial nuclear winter studies.