2020/02/04 by Othniel J. E. Y. Konan, Konan, Othniel J. E. Y., Amit Kumar Mishra +3
Physics and Astronomy · Earth and Planetary Sciences · #Ionosphere and magnetosphere dynamics #Earthquake Detection and Analysis #Seismic Waves and Analysis
paper · pdf · doi:10.48550/arxiv.2002.01244
Lightning strokes create powerful electromagnetic pulses that routinely cause\nvery low frequency (VLF) waves to propagate across hemispheres along\ngeomagnetic field lines. VLF antenna receivers can be used to detect these\nwhistler waves generated by these lightning strokes. The particular\ntime/frequency dependence of the received whistler wave enables the estimation\nof electron density in the plasmasphere region of the magnetosphere. Therefore\nthe identification and characterisation of whistlers are important tasks to\nmonitor the plasmasphere in real-time and to build large databases of events to\nbe used for statistical studies. The current state of the art in detecting\nwhistler is the Automatic Whistler Detection (AWD) method developed by\nLichtenberger (2009). This method is based on image correlation in 2 dimensions\nand requires significant computing hardware situated at the VLF receiver\nantennas (e.g. in Antarctica). The aim of this work is to develop a machine\nlearning-based model capable of automatically detecting whistlers in the data\nprovided by the VLF receivers. The approach is to use a combination of image\nclassification and localisation on the spectrogram data generated by the VLF\nreceivers to identify and localise each whistler. The data at hand has around\n2300 events identified by AWD at SANAE and Marion and will be used as training,\nvalidation, and testing data. Three detector designs have been proposed. The\nfirst one using a similar method to AWD, the second using image classification\non regions of interest extracted from a spectrogram, and the last one using\nYOLO, the current state of the art in object detection. It has been shown that\nthese detectors can achieve a misdetection and false alarm of less than 15% on\nMarion's dataset.\n