Body waves traversing the Earth’s interior from a seismic source to receivers on the surface carry rich information about its internal structures. Their travel time measurements have been widely used in seismology to constrain Earth’s interior at the global scale by mapping the time anomaly along their ray paths. However, picking the travel time of global seismic waves, suitable for studying Earth’s fine-scale structures, requires highly skilled personnel and is often fairly subjective. Here, we report the development of an automatic picker for PKIKP waves, traversing the Earth nearly along its diameters and through the inner core, based on the latest advances in supervised deep learning. A convolutional neural network (CNN) we developed automatically determines the PKIKP onset on vertical seismograms near its theoretical prediction of cataloged earthquakes. As high-quality manual onset picks of global seismic phases are limited, we employed a scheme to generate a synthetic supervised training dataset containing 300,000 waveforms. The PKIKP onsets picked by our trained CNN automatic picker exhibit a mean absolute error of ~0.5 s compared to 1,503 manual picks, comparable to the estimated human-picking error. In an integration test, the CNN automatic picks obtained from an extended waveform dataset yield a cylindrically anisotropic inner core model that agrees well with the models inferred from manual picks, which illustrates the success of this pilot model. This is a significant step closer to harvesting an unprecedented volume of travel time measurements for studying the inner core or other regions of the Earth’s deep interior.

The violent eruption of the Hunga (Tonga) submarine volcano on 15 January 2022 caused a 58 km-heigh ash plume, catastrophic tsunami, and significant global seismic and infrasound waves. However, the physical mechanism underpinning its multiple-explosive events remains unclear, and its resolvability relies on the seismic waveform source inversion. The studies of two different point-source models, the seismic moment tensor (MT) and the single force (SF), have been performed separately for this eruption, which, interestingly, can explain the seismic data adequately. Here, we use a joint inversion of MT and SF to unravel a composite source of an explosive MT and a significant upward force for the first major explosive event. Regarding the direction and magnitude, we propose that the upward force is likely a rebound force in response to the pressure drop on the seafloor because the water body above the volcano was abruptly uplifted by the shallow underwater explosion.

Determining the seismic moment tensors (MT) from the observed waveforms, known as full-waveform seismic MT inversion, remains challenging for small to moderate-size earthquakes at regional scales. Firstly, there is an intrinsic difficulty due to a tradeoff between the isotropic (ISO) and compensated linear vector dipole (CLVD) components of MT that impedes resolving shallow explosive sources, e.g., underground nuclear explosions. It is caused by the similarity of long-period waveforms radiated by ISO and CLVD at regional distances. Secondly, regional scales usually bear complex geologic structures; thus, inaccurate knowledge of Earth’s structure should be considered a theoretical error in the MT inversion. However, this has been a challenging problem. So far, only the uncertainty of the 1D Earth model (1D structural error), apart from data errors, has been explored in the source studies. Here, we utilize a hierarchical Bayesian MT inversion to address the above problems. Our approach takes advantage of affine-invariant ensemble samplers to explore the ISO-CLVD tradeoff space thoroughly and effectively. Furthermore, we invert for station-specific time shifts to treat the structural errors along specific source-station paths (2D structural errors). We present synthetic experiments demonstrating the method’s advantage in resolving the ISO components. The application to nuclear explosions conducted by the Democratic People’s Republic of Korea (DPRK) shows highly similar source mechanisms, dominated by a high ISO, significant CLVD components, and a small DC component. The recovered station-specific time shifts from the nuclear explosions present a consistent pattern, which agrees well with the geological setting surrounding the event location.

The structure of the lowermost mantle and the core-mantle boundary (CMB) has profound implications for Earth’s evolution and current-day dynamics. Whilst tomographic studies of Vs show good agreement in the lowermost mantle, consensus as to Vp and especially CMB radius has not yet been reached. We perform a hierarchical Bayesian inversion for Vp in the lowermost 300 km of the mantle and the radius of the core-mantle boundary using differential travel time data. Concurrent with finding Vp perturbations of 0.56% RMS amplitude that spatially agree with previous studies in areas of low posterior variance, we find 4.5 km RMS amplitude core-mantle boundary radius perturbations with a broadly north-south hemispherical character, with spherical harmonic power evenly distributed between degrees 1-3. These results suggest that CMB radial processes are set by a longer scale process than the Vp perturbations.