AI and the World's Hard Problems · Earth & climate science · arXiv · 2026-10-21
HypoNet Nankai PS: rapid hypocenter determination for the Nankai Trough using physics-informed neural networks with three-dimensional P- and S-wave velocity models
Rapid hypocenter determination based on 3D seismic velocity models is computationally demanding but essential for monitoring subduction zones. HypoNet Nankai PS extends a physics-informed neural network travel-time framework from P- to both P- and S-waves, and a revised training procedure for the collocation-point distribution, network architecture, and optimizer substantially reduced root-mean-square deviations compared with the original algorithm in benchmarks against the fast marching method. Applied to 223 earthquakes recorded by DONET stations off the Kii Peninsula, it agreed with the JMA catalog beneath the network and placed events farther seaward substantially shallower, in the oceanic crust and uppermost mantle, consistent with previous relocations using DONET data and 3D velocity models.