Search for: All records

SUMMARY Applications of machine learning in seismology have greatly improved our capability of detecting earthquakes in large seismic data archives. Most of these efforts have been focused on continental shallow earthquakes, but here we introduce an integrated deep-learning-based workflow to detect deep earthquakes recorded by a temporary array of ocean-bottom seismographs (OBSs) and land-based stations in the Tonga subduction zone. We develop a new phase picker, PhaseNet-TF, to detect and pick P- and S-wave arrivals in the time–frequency domain. The frequency-domain information is critical for analysing OBS data, particularly the horizontal components, because they are contaminated by signals of ocean-bottom currents and other noise sources in certain frequency bands. PhaseNet-TF shows a much better performance in picking S waves at OBSs and land stations compared to its predecessor PhaseNet. The predicted phases are associated using an improved Gaussian Mixture Model Associator GaMMA-1D and then relocated with a double-difference package teletomoDD. We further enhance the model performance with a semi-supervised learning approach by iteratively refining labelled data and retraining PhaseNet-TF. This approach effectively suppresses false picks and significantly improves the detection of small earthquakes. The new catalogue of Tonga deep earthquakes contains more than 10 times more events compared to the reference catalogue that was analysed manually. This deep-learning-enhanced catalogue reveals Tonga seismicity in unprecedented detail, and better defines the lateral extent of the double-seismic zone at intermediate depths and the location of four large deep-focus earthquakes relative to background seismicity. It also offers new potential for deciphering deep earthquake mechanisms, refining tomographic models, and understanding of subduction processes. 

Abstract The Alaska Peninsula has a long history of plate subduction with along‐arc variations in volcanic eruption styles and geochemistry. However, the sub‐arc melting processes that feed these volcanoes are unclear. The Alaska slab morphology below 200 km depth remains debated due to limited seismic data and thus low tomography resolution in this region. Here we utilize the newly available regional and teleseismic data to build 3‐D high‐resolutionV_PandV_Smodels to 660 km depth. We find that the high‐velocity Pacific Plate subducts to the bottom of the mantle transition zone (MTZ) with complex deformation and gaps. In the southwest, we observe a wide gap in the high‐velocity slab at 200–500 km depths. Toward the northeast, the slab becomes more continuous extending to the MTZ with a few holes below 200 km. We interpret these gaps as a slab tear that coincides with the subducted ancient Kula‐Pacific Ridge. We also invert for 3‐DV_PandV_P/V_Smodels to 200 km depth with higher resolution and find strong along‐strike changes in slab dehydration and sub‐arc melting, indicated by lowV_Pand highV_P/V_Sanomalies. Slab dehydration and sub‐arc melting are most extensive below the Pavlof and Shumagin segments in the southwest, becoming limited below the Chignik and Chirikof segments in the northeast, and extensive again beneath the Kodiak segment further to the northeast. We propose that the variations of slab hydration at the outer rise significantly influence slab dehydration at greater depths and further control sub‐arc melting beneath the Alaska Peninsula. 

SUMMARY Despite progress in tomographic imaging of Earth's interior, a number of critical questions regarding the large-scale structure and dynamics of the mantle remain outstanding. One of those questions is the impact of phase-boundary undulations on global imaging of mantle heterogeneity and on geodynamic (i.e. convection-related) observables. To address this issue, we developed a joint seismic-geodynamic-mineral physical tomographic inversion procedure that incorporates lateral variations in the depths of the 410- and 660-km discontinuities. This inversion includes S-wave traveltimes, SS precursors that are sensitive to transition-zone topography, geodynamic observables/data (free-air gravity, dynamic surface topography, horizontal divergence of tectonic plates and excess core-mantle boundary ellipticity) and mineral physical constraints on thermal heterogeneity. Compared to joint tomography models that do not include data sensitivity to phase-boundary undulations in the transition zone, the inclusion of 410- and 660-km topography strongly influences the inference of volumetric anomalies in a depth interval that encompasses the transition zone and mid-mantle. It is notable that joint tomography inversions, which include constraints on transition-zone discontinuity topography by seismic and geodynamic data, yield more pronounced density anomalies associated with subduction zones and hotspots. We also find that the inclusion of 410- and 660-km topography may improve the fit to the geodynamic observables, depending on the weights applied to seismic and geodynamic data in the inversions. As a consequence, we find that the amplitude of non-thermal density anomalies required to explain the geodynamic data decreases in most of the mantle. These findings underline the sensitivity of the joint inversions to the inclusion of transition-zone complexity (e.g. phase-boundary topography) and the implications for the inferred non-thermal density anomalies in these depth regions. Finally, we underline that our inferences of 410- and 660-km topography avoid a commonly employed approximation that represents the contribution of volumetric heterogeneity to SS-wave precursor data. Our results suggest that this previously employed correction, based on a priori estimates of upper-mantle heterogeneity, might be a significant source of error in estimating the 410- and 660-km topography.