Machine Learning Methods to Catalog Nuclear Sources From Diverse, Widely Distributed Sensors

Session: Explosion Seismology Advances [Poster]
Type: Poster
Date: 4/29/2020
Time: 08:00 AM
Room: Ballroom
Description: 

With nuclear tests potentially ongoing in the world, it is important to develop a robust method to automatically detect, classify and locate sources from diverse and widely distributed sensors around the globe using a diverse network of sensors. Using a rich list of declassified nuclear events reported globally generated from the combined reports of the SIPRI catalog, latest DOE catalogue report and USGS reports for the North Korea nuclear tests, we created an underground nuclear explosion catalogue and retrieved waveforms from all available International Federation of Digital Seismograph Network (FDSN) stations. Our dataset is supplemented by the Lawrence Livermore National Laboratory nuclear explosion dataset, the former Soviet Union’s digitized analog seismic records, as well as the Japan National Research Institute for Earth Science and Disaster Resilience (NIED) Hi-Net data for the 6 North Korea nuclear tests. Since nuclear explosions often have no discernable S-wave arrival, we use only P-wave arrival picks for the labeled dataset. A seismic and nuclear event classifier was built focusing initially on regional events of magnitude 3 or greater using a USGS dataset composed of 149,326 regional earthquake and 3523 nuclear explosion observations for the 10-layer convolutional neural network (CNN) training set.The CNN algorithm was trained for three classes: earthquake P-wave, nuclear P-wave and noise. In order to address the effect of low signal-to-noise ratio on CNN performance, an energy filter was developed to calculate the ratio of mean energy of a P-wave to the noise and an energy ratio threshold was set to 5 for data pre-processing. Our initial results are promising, with the accuracy of the validation set rising rapidly to accuracy of over 95% after a few epochs. We anticipate that this classifier system will be useful beyond automatic nuclear detections, potentially being applied to detect many types of seismic signals associated with geological hazards such as volcanic eruptions, tsunamigenic earthquakes, glacier/ice-quakes or landslides.

Presenting Author: Louisa Barama

Authors