An isostatic correction is commonly made to Bouguer anomaly gravity data to remove the gravity effect of isostatic compensation of topographic loads. In the USSR a ''decompensative'' correction has then been made to the isostatic gravity anomaly to remove the gravity effect of isostatic compensation of geologic loads as well. Under an hypothesis of local (as opposed to regional) compensation the gravity effect of a shallow geological body can be separated from that of its inferred deep compensating root by deconvolution. By contrast with the procedure originally defined by Zorin, we employ here calculations in the wave number domain, leading to an efficient and exact solution. The decompensative transfer function acts to increase amplitude of the isostatic anomaly at all wavelengths, but especially at the low end of the isostatic anomaly spectrum in the 200--600 km wavelength region, where anomaly amplitudes may increase by as much as a factor of 2. In a 1200¿1200 km region centered on the Rio Grande rift the decompensative correction ranges from about -35 to +25 mGal. The rift is not particularly apparent in the decompensative correction nor is the rift anymore apparent in the decompensative anomaly than in the isostatic anomaly. The decompensative anomaly, however, highlights an arcuate gravity low and a system of gravity highs inferred to reflect prerift welts of mass concentration which have indirectly influenced the position of the rift and its segmentation and zones of accommodation.

Under the assumptions made, if the decompensative anomaly is subtracted from the Bouguer anomaly, then the residual is the gravity anomaly field of deep structure, without gravity effects of shallow sources in the upper crust. We suppose this gravity field to be dominated by the gravity effects of the density contrast across the crust-mantle boundary (Moho) and the density contrast at the lithosphere-asthenosphere boundary. Using available seismic data to (weakly) constrain the Moho surface, we invert the residual gravity field for topography of the base of the lithosphere. Lithosphere is found to be 200 km thick in the High Plains: 40--50 km in the eastern Great Basin: 75--100 km in the Colorado Plateau, and as thin as 40 km in the southern Rio Grande rift. In the area studied, the thickness of the lithosphere is everywhere greater than that of the crust. An asthenosphere bulge closely tracks the axis of the rift and is symmetrical to it, by contrast with asymmetry that might be inferred from broad features of the Bouguer anomaly. The separation of gravity effects made possible by the decompensative correction shows how the rift is fundamentally controlled by thinning of the lithosphere, yet in detail is deflected by long-lived tectonic welts in the shallow, brittle crust. ¿1991 American Geophysical Union