The cumulative propagation effects of weather and ground surface on sound exposure levels at long distances from an impulsive sound source are not well established. Measurements of sound exposure and peak pressure at distances up to about 20 km from explosions were made by the U.S. Army for three types of ground conditions. These data, along with measured weather profiles and ground impedance parameters, were used as inputs to the FFP model, for comparison of measurement to prediction. The 1/3-octave band attenuations were calculated for band center frequencies from 1 Hz to 1 kHz, for the measurement microphone locations in the study. A great deal of the variation in signal level from event to event was predicted by the FFP model, without resort to nonlinear propagation effects or to turbulent scattering. Groups of cases with similar signal attenuation with distance and similar weather profiles may provide a reasonable basis for classifying long-distance propagation conditions. Such a classification would have merit as an efficient way to assess long-term noise averages or to estimate the error in short-term predictions of noise levels. [Work supported by USACERL and AMSAA.]