Publication Date:
2015-09-21
Description:
Wave injection experiments provide an opportunity to explore and quantify aspects of nonlinear wave-particle phenomena in a controlled manner. Waves are injected into space from ground-based ELF/VLF transmitters and the modified waves are measured by radio receivers on the ground in the conjugate hemisphere. These experiments are expensive and challenging projects to build and to operate, and the transmitted waves are not always detected in the conjugate region. Even the powerful transmitter located at Siple Station, Antarctica in 1986, estimated to radiate over 1 kW, only reported a reception rate of ∼40%, indicating that a significant number of transmissions served no observable scientific purpose and reflecting the difficulty in determining suitable conditions for transmission and reception. Leveraging modern machine learning classification techniques, we apply two statistical techniques, a Bayes and a SVM classifier, to predict the occurrence of detectable 1-hop transmissions from Siple data with accuracies on the order of 80%-90%. Applying these classifiers to our 1986 Siple data set, we detect 406 receptions of Siple transmissions which we analyze to generate more robust statistics on nonlinear growth rates, 3 dB/s - 270 dB/s, and nonlinear total amplification, 3 dB - 41 dB.
Print ISSN:
0148-0227
Topics:
Geosciences
,
Physics
