We have observed about 16 absorption lines of the ν SO 2 vibrational band on Io, in disk-integrated 19-μm spectra taken with the TEXES high spectral resolution mid-infrared spectrograph at the NASA Infrared Telescope Facility in November 2001, December 2002, and January 2004. These are the first ground-based infrared observations of Io's sunlit atmosphere, and provide a new window on the atmosphere that allows better longitudinal and temporal monitoring than previous techniques. Dramatic variations in band strength with longitude are seen that are stable over at least a 2 year period. The depth of the strongest feature, a blend of lines centered at 530.42 cm -1, varies from about 7% near longitude 180° to about 1% near longitude 315° W, as measured at a spectral resolution of 57,000. Interpretation of the spectra requires modeling of surface temperatures and atmospheric density across Io's disk, and the variation in non-LTE ν vibrational temperature with altitude, and depends on the assumed atmospheric and surface temperature structure. About half of Io's 19-μm radiation comes from the Sun-heated surface, and half from volcanic hot spots with temperatures primarily between 150 and 200 K, which occupy about 8% of the surface. The observations are thus weighted towards the atmosphere over these low-temperature hot spots. If we assume that the atmosphere over the hot spots is representative of the atmosphere elsewhere, and that the atmospheric density is a function of latitude, the most plausible interpretation of the data is that the equatorial atmospheric column density varies from about 1.5×10 cm near longitude 180° W to about 1.5×10 cm near longitude 300° W, roughly consistent with HST UV spectroscopy and Lyman- α imaging. The inferred atmospheric kinetic temperature is less than about 150 K, at least on the anti-Jupiter hemisphere where the bands are strongest, somewhat colder than inferred from HST UV spectroscopy and millimeter-wavelength spectroscopy. This longitudinal variability in atmospheric density correlates with the longitudinal variability in the abundance of optically thick, near-UV bright SO 2 frost. However it is not clear whether the correlation results from volcanic control (regions of large frost abundance result from greater condensation of atmospheric gases supported by more vigorous volcanic activity in these regions) or sublimation control (regions of large frost abundance produce a more extensive atmosphere due to more extensive sublimation). Comparison of data taken in 2001, 2002, and 2004 shows that with the possible exception of longitudes near 180° W between 2001 and 2002, Io's atmospheric density does not appear to decrease as Io recedes from the Sun, as would be expected if the atmosphere were supported by the sublimation of surface frost, suggesting that the atmosphere is dominantly supported by direct volcanic supply rather than by frost sublimation. However, other evidence such as the smooth variation in atmospheric abundance with latitude, and atmospheric changes during eclipse, suggest that sublimation support is more important than volcanic support, leaving the question of the dominant atmospheric support mechanism still unresolved.