Radio propagation models provide an estimate of the power loss in a communication link caused by the surface of the ground, atmospheric refraction, foliage, and other environmental factors. Many of the models rely on digital topographic databases to provide information about the terrain, and generally the databases are sparsely sampled relative to the electromagnetic wavelengths used for communication systems. This work primarily develops a technique to evaluate the effects of that sparsity on the uncertainty of propagation models.That is accomplished by accurately solving the electromagnetic fields over many randomly rough surfaces which pass through the sparse topographic data points, many possible communication links, all of which fit the underlying data, are represented. The power variation caused by the different surface realizations is that due to the sparse sampling. Additionally, to verify that this solution technique is a good model, experimental propagation measurements were taken, and compared to the computations. The experimental work presented in Chapter 7 required the help of many individuals from VT, NSWC, and NRAO. In addition to Ben and Dr. Brown, Wes Sizemore at NRAO, and Bruce Naley, Punk Chilton and Natasha Lagoudous from NSWC all put in lots of work during the weeks conducting the experiment. Bruce was invaluable for climbing 100 feet up the water tower, with the antenna and spectrum analyzer, to take measurements. Additionally, equipment was provided by NSWC, and many at NRAO provided help in experiment planning, regulatory approval, and scrounging for parts, and fixing broken equipment.iii