Scholarlink

<p>Effective discrimination between earthquakes and explosions is pivotal, particularly in the context of the Comprehensive Nuclear-Test-Ban Treaty (CTBT) verification regime. This paper introduces the usage of a Support Vector Machine (SVM) algorithm tailored to discern seismic records produced by natural earthquakes from those caused by underground nuclear tests, wherein the registered values of mb and Ms magnitudes (body-wave and surface-wave magnitudes respectively) of each event are selected as feature vectors. These magnitude values are directly provided in official bulletins for each seismic event, therefore, no preliminary calculations were necessary, making our method easy to implement. By harnessing a diverse dataset and employing state-of-the-art machine learning algorithms, our approach demonstrates remarkable accuracy in discriminating these events. Also, we provide a posterior probability that estimates the correctness of the prediction performed by the classification algorithm. This work represents a significant stride towards enhancing the capabilities of seismic monitoring systems, thereby reinforcing international efforts towards nuclear non-proliferation and global stability.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Deciphering earth's tremors: a machine learning approach to distinguish earthquakes from explosions

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年3月1日

查看全文
  • A. Pignatelli,
  • C. Petrucci,
  • V. Vignoli,
  • F. D’Ajello Caracciolo,
  • R. Console

摘要

Effective discrimination between earthquakes and explosions is pivotal, particularly in the context of the Comprehensive Nuclear-Test-Ban Treaty (CTBT) verification regime. This paper introduces the usage of a Support Vector Machine (SVM) algorithm tailored to discern seismic records produced by natural earthquakes from those caused by underground nuclear tests, wherein the registered values of mb and Ms magnitudes (body-wave and surface-wave magnitudes respectively) of each event are selected as feature vectors. These magnitude values are directly provided in official bulletins for each seismic event, therefore, no preliminary calculations were necessary, making our method easy to implement. By harnessing a diverse dataset and employing state-of-the-art machine learning algorithms, our approach demonstrates remarkable accuracy in discriminating these events. Also, we provide a posterior probability that estimates the correctness of the prediction performed by the classification algorithm. This work represents a significant stride towards enhancing the capabilities of seismic monitoring systems, thereby reinforcing international efforts towards nuclear non-proliferation and global stability.