It is proposed that the transport of material from Io to the plasma torus requires the presence of an atmosphere as an intermediate buffer. Direct escape of volcanic SO2 is expected to be negligible. The sputtering of surface SO2 frost by corotating ions (if they reach the surface) or by magnetospheric MeV ions would not provide sufficient escape flux. Instead, such sputtering process produces an ~97% fraction with energy less than the escape energy; hence this process alone could produce an atmosphere thick enough to prevent the corotating ions from reaching the surface and thick enough to disallow even the energetic sputtered products created at the surface to escape. The escape from atmosphere could take place by several mechanisms, of which secondary sputtering at the exobase is dominant. The escape rate from a thin atmosphere increases with increasing atmospheric SO2 content. By preparing photochemical models of SO2 atmospheres for varying total atmospheric pressure and calculating the escape rates, it is concluded that the globally averaged atmospheric SO2 pressure on Io at Voyager 1 encounter must have been >1.4¿10-10 bar (surface density n0(SO2)>1.2¿1010 cm-3, or column density of >1¿1016 cm-2), as the escape rates in this limiting case are just adequate to populate the torus if the Brown et al. (1983) estimates of the supply rate are applicable. However, if the higher estimates of Pontius and Hill (1983) are appropriate, then more SO2 would be needed in the Io atmosphere.