The relative positions of the test masses in gravitational-wave detectors will be influenced not only by astrophysical gravitational waves, but also by the fluctuating Newtonian gravitational forces of moving masses in the ground and air around the detector. These effects are often referred to as gravity gradient noise. This paper considers the effects of gravity gradients from density perturbations in the atmosphere, and from massive airborne objects near the detector. These have been discussed previously by Saulson (1984 Phys. Rev. D 30 732), who considered the effects of background acoustic pressure waves and of massive objects moving smoothly past the interferometer; the gravity gradients he predicted would be too small to be of serious concern even for advanced interferometric gravitational-wave detectors. In this paper, I revisit these phenomena, considering transient atmospheric shocks, and estimating the effects of sound waves or objects colliding with the ground or buildings around the test masses. I also consider another source of atmospheric density fluctuations: temperature perturbations that are advected past the detector by the wind. I find that background acoustic noise and temperature fluctuations still produce gravity gradient noise that is below the noise floor even of advanced interferometric detectors, although temperature perturbations carried along non-laminar streamlines could produce noise that is within an order of magnitude of the projected noise floor at 10 Hz. A definitive study of this effect may require better models of the wind flow past a given instrument. I also find that transient shockwaves in the atmosphere could potentially produce large spurious signals, with signal-to-noise ratios in the hundreds in an advanced interferometric detector. These signals could be vetoed by means of acoustic sensors outside of the buildings. Massive wind-borne objects such as tumbleweeds could also produce gravity gradient signals with signal-to-noise ratios in the hundreds if they collide with the interferometer buildings, so it may be necessary to build fences preventing such objects from approaching within about 30 m of the test masses.