We present a general concept for evolutionary, collaborative, multiscale inversion of geophysical data, specifically applied to the construction of a first‐generation Collaborative Seismic Earth Model. This is intended to address the limited resources of individual researchers and the often limited use of previously accumulated knowledge. Model evolution rests on a Bayesian updating scheme, simplified into a deterministic method that honors today's computational restrictions. The scheme is able to harness distributed human and computing power. It furthermore handles conflicting updates, as well as variable parameterizations of different model refinements or different inversion techniques. The first‐generation Collaborative Seismic Earth Model comprises 12 refinements from full seismic waveform inversion, ranging from regional crustal‐ to continental‐scale models. A global full‐waveform inversion ensures that regional refinements translate into whole‐Earth structure.

Growing evidence from a variety of geologic indicators points to significant topography maintained convectively by viscous stresses in the mantle. However, while gravity is sensitive to dynamically supported topography, there are only small free‐air gravity anomalies (<30 mGal) associated with Earth's long‐wavelength topography. This has been used to suggest that surface heights computed assuming a complete isostatic equilibrium provide a good approximation to observed topography. Here we show that the apparent paradox is resolved by the well‐established formalism of global, self‐gravitating, viscously stratified Earth models. The models predict a complex relation between dynamic topography, mass, and gravity anomalies that is not summarized by a constant admittance—i.e., ratio of gravity anomalies to surface deflections—as one would infer from analytic flow solutions formulated in a half‐space. Our results suggest that sizable dynamic topography may exist without a corresponding gravity signal.

The South Atlantic region displays (1) a topographic gradient across the basin, with Africa elevated relative to South America, (2) a bimodal spreading history with fast spreading rates in Late Cretaceous and Eo‐Oligocene, and (3) episodic regional uplift events in the adjacent continents concentrated in Late Cretaceous and Oligocene. Here we show that these observations can be linked by dynamic processes within Earth's mantle, through temporal changes in asthenosphere flow beneath the region. The topographic gradient implies westward, pressure‐driven mantle flow beneath the basin, while the rapid spreading rate changes, on order 10 million years, require significant decoupling of regional plate motion from the large‐scale mantle buoyancy distribution through a mechanically weak asthenosphere. Andean topographic growth in late Miocene can explain the most recent South Atlantic spreading velocity reduction, arising from increased plate boundary forcing associated with the newly elevated topography. But this mechanism is unlikely to explain the Late Cretaceous/Tertiary spreading variations, as changes in Andean paleoelevation at the time are small. We propose an unsteady pressure‐driven flow component in the asthenosphere beneath the South Atlantic region to explain the Late Cretaceous/Tertiary spreading rate variations. Temporal changes in mantle flow due to temporal changes in regional mantle pressure gradients imply a correlation of horizontal and vertical motions: we find that this prediction from our models agrees with geologic and geophysical observations of the South Atlantic region, including episodes of passive margin uplift, regional basin reactivation, and magmatic activity.

The analysis of Gravity Recovery and Climate Experiment (GRACE) Level 2 data time series from the Center for Space Research (CSR) and GeoForschungsZentrum (GFZ) allows us to extract a new estimate of the co‐seismic gravity signal due to the 2004 Sumatran earthquake. Owing to compressible self‐gravitating Earth models, including sea level feedback in a new self‐consistent way and designed to compute gravitational perturbations due to volume changes separately, we are able to prove that the asymmetry in the co‐seismic gravity pattern, in which the north–eastern negative anomaly is twice as large as the south–western positive anomaly, is not due to the previously overestimated dilatation in the crust. The overestimate was due to a large dilatation localized at the fault discontinuity, the gravitational effect of which is compensated by an opposite contribution from topography due to the uplifted crust. After this localized dilatation is removed, we instead predict compression in the footwall and dilatation in the hanging wall. The overall anomaly is then mainly due to the additional gravitational effects of the ocean after water is displaced away from the uplifted crust, as first indicated by de Linage et al. (2009). We also detail the differences between compressible and incompressible material properties. By focusing on the most robust estimates from GRACE data, consisting of the peak‐to‐peak gravity anomaly and an asymmetry coefficient, that is given by the ratio of the negative gravity anomaly over the positive anomaly, we show that they are quite sensitive to seismic source depths and dip angles. This allows us to exploit space gravity data for the first time to help constraining centroid‐momentum‐tensor (CMT) source analyses of the 2004 Sumatran earthquake and to conclude that the seismic moment has been released mainly in the lower crust rather than the lithospheric mantle. Thus, GRACE data and CMT source analyses, as well as geodetic slip distributions aided by GPS, complement each other for a robust inference of the seismic source of large earthquakes. Particular care is devoted to the spatial filtering of the gravity anomalies estimated both from observations and models to make their comparison significant.