Probabilistic linear inversion of satellite gravity gradient data applied to the northeast Atlantic

Abstract

Lateral density variations in Earth's structure are reflected in gravity data which can be used, together with other types of geological and geophysical information, to reveal Earth's structure. The well-known limitation of potential field data is the lack of information about the depth of structural heterogeneity. The gravity signal caused by individual lithospheric layers, such as sediments, crystalline crust, or mantle lithosphere, can be much larger than their sum (as well as the observed data) due to the isostatic compensation (e.g., Stacey and Davis (2008)). It follows that the lithospheric-scale gravity models directly reflect the type and accuracy of a priori information used as constraints.Regional or global crustal thickness models have been traditionally calculated by inverting gravity data which usually reflect variations in surface topography and Moho depth (Lebedeva-Ivanova et al., 2019;Reguzzoni & Sampietro, 2015;Wieczorek & Phillips, 1998). On the other hand, the gravity inversion can also be used to infer the top basement geometry given that the crustal thickness is constrained by seismic data and the region of interest is small enough, so that the contribution of mantle density anomalies can be ignored or removed assuming a certain model of isostatic compensation (e.g., Martinec & Fullea, 2015). In the case where the thickness of both sedimentary layers and the crystalline crust is constrained by seismic reflection and refraction data, the residual gravity signal can be computed from the observed gravity data and used to infer the distribution of density heterogeneity of the upper mantle (Kaban et al., 2010).In this study, we have chosen the northeast Atlantic region which is well covered by active-source seismic data, and where mantle density heterogeneities must be particularly strong. The uncertainty of a priori information is quantified in the form of the data covariance matrix. We address the problem of limited information on source depth by the following modifications applied to the standard 3-D gravity inversion approach (Li &