Experimental results are presented indicating that only a few parameters describing capping inversion of the marine atmospheric boundary layer (MABL) need to be considered deterministically for accurate estimation of the MABL's effects on propagation loss, provided that (1) the right deterministic parameters are chosen and (2) other parameters are accounted for statistically. To perform this experiment, the refractivity structure associated with the capping inversion of the MABL is modeled using shifted empirical orthogonal functions (SEOFs) developed from the second-order statistics of radiosondes. Propagation estimates calculated using the SEOFs as well as radiosondes and the heuristic trilinear model are compared with measured propagation. Using only two deterministic parameters, the results obtained using the SEOFs are essentially as good as can be obtained with a fading channel. Why this result is important in data assimilation for refractivity assessments is discussed. ¿ 1998 American Geophysical Union