This multiyear project addresses the long-standing challenge of imaging Earth’s interior based on full waveform inversion on a global scale, which has so far remained a challenge mainly due to computational limitations.
The team has produced the first global adjoint tomography model, the results of 15 iterations for a global mantle and crustal model with transverse isotropy confined to the upper mantle, using seismic waves with a shortest period of ~ 17 s. Additional petascale resources will enable continued iterative updating and an increase in the resolution of seismic images of the entire planet based on massive data assimilation accommodated by adjoint techniques.
The team started with a global dataset of 253 earthquakes with magnitudes in the range of 5.8 ≤ Mw ≤ 7.0. Since then, the team has expanded its dataset to include 1,200 earthquakes and is using this larger dataset to understand challenges in data assimilation.
Their ultimate goal is to use all earthquakes recorded by permanent and temporary seismographic networks since 1995, using the global Centroid Moment-Tensor catalogue for earthquakes in the magnitude range of 5.5 ≤ Mw ≤ 7.0. To date, the team has gathered data for more than 4,200 earthquakes and is working on their source inversions. They started with 27 s waves, and since the 12th iteration were using 17 s waves. The current aim is to go down to ~ 9 s and include more earthquakes in the inversion.
Additionally, they will continue optimizing the solver and improving the adjoint tomography workflow in conjunction with preparations for running higher-resolution simulations (~ 1 s) on the next-generation supercomputer Summit.
|Source||Hours||Start Date||End Date|