2018/10/03 by S. Mostafa Mousavi, Weiqiang Zhu, Mousavi, S. Mostafa +5 · 5 citations
Computer Science · Earth and Planetary Sciences · #Earthquake Detection and Analysis #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Seismic Waves and Analysis #Seismology and Earthquake Studies
paper · pdf · doi:10.48550/arxiv.1810.01965
openalex publication_date 2018/10/03 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Earthquake signal detection is at the core of observational seismology. A\ngood detection algorithm should be sensitive to small and weak events with a\nvariety of waveform shapes, robust to background noise and non-earthquake\nsignals, and efficient for processing large data volumes. Here, we introduce\nthe Cnn-Rnn Earthquake Detector (CRED), a detector based on deep neural\nnetworks. The network uses a combination of convolutional layers and\nbi-directional long-short-term memory units in a residual structure. It learns\nthe time-frequency characteristics of the dominant phases in an earthquake\nsignal from three component data recorded on a single station. We train the\nnetwork using 500,000 seismograms (250k associated with tectonic earthquakes\nand 250k identified as noise) recorded in Northern California and tested it\nwith an F-score of 99.95. The robustness of the trained model with respect to\nthe noise level and non-earthquake signals is shown by applying it to a set of\nsemi-synthetic signals. The model is applied to one month of continuous data\nrecorded at Central Arkansas to demonstrate its efficiency, generalization, and\nsensitivity. Our model is able to detect more than 700 microearthquakes as\nsmall as -1.3 ML induced during hydraulic fracturing far away than the training\nregion. The performance of the model is compared with STA/LTA, template\nmatching, and FAST algorithms. Our results indicate an efficient and reliable\nperformance of CRED. This framework holds great promise in lowering the\ndetection threshold while minimizing false positive detection rates.\n