This paper investigates the variability of the monthly time fraction of excess of atmospheric propagation parameters. In the first part of the paper a theoretical investigation is carried out regarding the relationship between the variability of monthly time fraction of excess vis-a-vis the variability of the annual-worst-month time fraction of excess. It is shown that the tail of the distribution of the annual-worst-month time fraction of excess can be derived from the average distribution of the monthly time fraction of excess without requiring detailed knowledge of the seasonal behavior of the propagation. In the second part of the paper the modeling of the average distribution of the monthly time fraction of excess (the parent distribution) is investigated using some 92 site years of data (from slant path attenuation and transhorizon measurements). Three competing models were investigated: the conditional exponential, the conditional lognormal, and the shifted gamma distribution. For the testing of the suitability of the model a novel data pooling method is applied. The tests results show the shifted gamma model as the best model, followed by the conditional exponential model and then the conditional lognormal model.¿ 1997 American Geophysical Union