Classification of Infrasonic Atmospheric Events Using Electromagnetic Pulse Analysis

The Comprehensive Nuclear-Test-Ban Treaty Organization (CTBTO) operates the international monitoring system (IMS), consisting of four complementary technologies: Seismic, hydroacoustic, infrasound, and radionuclide, to detect events that might indicate a treaty violation. Infrasound is the main technology aimed at detecting atmospheric nuclear explosions. However, there are many other sources of infrasound signals, which are low frequency inaudible sound-waves in the atmosphere, that are detected. To deny that the source is of a nuclear nature, requires a lot of resources. As opposed to most other sources of infrasound, nuclear explosions also emit an electromagnetic pulse (EMP). Thus, if an infrasonically detected event is not accompanied by an EMP, it is certain that it has not originated from a nuclear explosion. In this research we use electromagnetic data recorded by a single antenna to examine coincidence of infrasonically detected events with detected EMPs. We show that a large fraction of infrasonically detected events are not accompanied by any EMP and thus the need to analyze them in the context of the IMS is eliminated. For those infrasonically detected events that do coincide with EMPs, we use machine learning techniques to classify the EMPs as either lightning or a potential explosion. We show that one can identify the lightning out of the EMPs. By filtering out all lightning signals one can almost always verify that there is no EMP other than lightning, which coincided with the infrasonically detected event. This way we can dramatically reduce the number of infrasound signals which require manual analysis.