The tsunami-generated magnetic field is a magnetic field that appears with movement of seawater by tsunamis. In the previous studies, researchers found that the tsunami-generated magnetic field arrives earlier than the tsunami sea level change based on analytical solutions and numerical simulations. In this study, we used the world's first simultaneous data of sea level change and magnetic field in the 2009 Samoa and 2010 Chile tsunamis to study the relation between these two physical quantities. We found that the vertical component of tsunami magnetic field arrives earlier than the sea level change. Moreover, the horizontal component of tsunami magnetic field arrives even earlier than the vertical component. We also revealed that the tsunami magnetic field can be used to estimate the tsunami wave height very accurately. We investigated the observed tsunami magnetic field by the 3-D time domain simulation. However, the currently available tsunami source models were unable to reproduce the observation in our research area. We confirmed that a better source model can improve the simulation. It follows that our high precision tsunami wave height data calculated from the magnetic field can improve the existing tsunami source models.

Plate tectonic theory is the framework that describes everything from the formation of the continents billions of years ago to natural disasters such as volcanoes, earthquakes, and tsunamis today. Even climate change estimates over geologic timescales rely on accurate plate tectonic reconstructions to understand the paleo-oceans. Despite the intricate links between plate tectonics and life on Earth, exactly what makes a plate “plate-like” is debated. In other words, what properties define the transition from the rigid plate, or lithosphere, to the weaker, convecting asthenosphere, and where does this transition occur? Classically, the lithosphere-asthenosphere boundary is defined thermally, with a gradual transition from the cold conductively cooling lithosphere to the warmer, convecting asthenosphere beneath. Overall, lithospheric thickening with age is observed beneath the oceans and toward the continental interiors suggesting that temperature and conductive cooling play a first-order role in controlling lithospheric thickness. However, within any given tectonic age interval a wide range of lithospheric thicknesses have been reported. Observations of sharp changes with depth in seismic wave speed and strong anomalies in seismic wave speed and electrical resistivity are similarly inconsistent with the smooth variations predicted by simple conductive cooling. Other properties or processes must define the tectonic plate. The lithosphere may be relatively dehydrated, which would enhance its strength. In contrast, asthenospheric hydration could make it relatively weak and also reduce its melting temperature. A small amount of partial melt beneath the plate in the asthenosphere may exist, which could further ease convection and therefore define the plate. Melt provides a simple explanation for a host of observations with large implications for plate tectonics, mantle dynamics, and Earth's evolution. So far reports of melt are variable in location and character. The variability in lithospheric thickness and also melt location and character suggests that the lithosphere-asthenosphere boundary is likely dynamic and dictated by mantle dynamics including melt generation and migration.

The first step in many seismological workflows is identifying if a signal contains an earthquake, and at which time which type of seismic wave arrived. These steps are known as event detection, phase identification and phase picking. In recent years, machine learning methods, in particular deep learning methods have been developed, showing promising performance on these tasks. However, so far these models have not been compared systematically in a quantitative way. Here we evaluate the performance of six deep learning models on eight datasets. Additionally, we compare them to a traditional picking algorithm not using machine learning. From our results we identify that the models EQTransformer, GPD and PhaseNet perform best. As in many use cases no picker trained on the target region will be available, we further evaluated how well models are transferable across regions. We identified that transfer across regions works well as long as the distance ranges stay similar. To foster application of the results, we make all our trained models available through the SeisBench framework.