Raw Seismic Waveform Datasets for Real-time Discrimination

Six raw seismic waveform datasets have been compiled from a portion of the localized University of Utah Seismograph Stations (UUSS) data for modeling and testing, the global Stanford Earthquake Dataset (STEAD) for generalization testing, and the Russia-Ukraine conflict data for case study analysis.Additionally, a code package is provided to demonstrate the application of tiny deep learning on microcontroller units.

The model, named MCU-Quake, can distinguish seismological signals from ambient noise and explosions in real-time using low-cost microcontrollers and short-duration data. This capability is particularly valuable for monitoring catastrophic natural hazards.