The present study focuses both on the influence of impact scale on ejecta expansion and on specific features of ejecta deposits around relatively small craters (i.e., those a few kilometers in width). The numerical model is based on the SOVA multimaterial multidimensional hydrocode, considering subaerial vertical impacts only, applying a 2-D version of the code to projectiles of 100, 300, and 1000 m diameter. Ejecta can roughly be divided into two categories: “ballistic” ejecta and “convective” ejecta; the ballistic ejecta are the ejecta with which the air interacts only slightly, while the convective ejecta motion is entirely defined by the air flow. The degree of particle/air interaction can be defined by the time/length of particle travel before deceleration. Ejecta size-distributions for the impacts modeled can be described by the same power law, but the size of maximum fragment increases with scale. There is no qualitative difference between the 100 m diameter projectile case and the 300 m diameter projectile impact. In both cases, fine ejecta decelerate in the air at a small distance from launching point and then rise to the stratosphere by air flows induced by the impacts. In the 1000 m-scale impact, the mass of ejecta is so large that it moves the atmosphere itself to high altitudes. Thus, the atmosphere cannot decelerate even the fine ejecta and they consequently expand to the rarefied upper atmosphere. In the upper atmosphere, even fine ejecta move more or less ballistically and therefore may travel to high altitudes.