Precipitation particles which fall from a source aloft through a wind shear are sorted as to size, the largest particles reaching the ground closest to the generating source, the smaller particles further from it. If precipitation is assumed to form continuously in cloud, with a fixed size distribution, this sorting affects significantly the size distributions to be observed below the cloud, and so the relationship between the precipitation rate R and the radar scattering-parameter Z (which is ∑D6, where D is the diameter of a raindrop or the water drop to which a snowflake would melt, and the summation is over unit volume). As an approximation to a small isolated shower, a horizontal generating element has been taken, of linear extent 1.6 kilometers in the direction of the wind shear. The quantities R and Z tend to be less at the ground than in the generating region, the size distributions remaining the same except for upper and lower limits of size imposed by the sorting. Several values of R and Z in the generating region have been considered, all obeying Z = aRb. The Z/R data below the showers have been found to be widely scattered about a locus Z = a′Rb′, where a′ > a and b′ < b. If a given R is obtained at the ground on many occasions involving widely varying values of R aloft, the corresponding values of Z are found to differ by as much as a factor of twelve. Findings regarding an isolated shower agree with the observations of Atlas and Plank (1953). As an approximation to “continuous” snow and rain, a regular array of snow-generating elements was considered. In this, there was one element of linear extent one mile every five miles. A size sample taken at the ground over a time interval less than a minute would then yield a discontinuous distribution, with a small range of sizes contributed by each cell. The distribution of a sample collected over several minutes would be fairly smooth and resemble that aloft except as to scale. In any case, Z and R at the ground would be reduced by a factor of approximately five, compared with the values in the generating elements. This introduces a scatter in the Z/R data at the ground, and a shift in the Z/R locus, similar to those noted for the case of an isolated shower.