Comparison of Manual and Machine Learning Assisted HF Amateur Radio LSTID Observations

Large scale traveling ionospheric disturbance (LSTID) signatures manifest as quasi-periodic variations in contact ranges in high frequency (HF) amateur radio communication reports recorded by automated monitoring systems such as the Weak Signal Propagation Reporting Network (WSPRNet), Reverse Beacon Network (RBN) and PSK Reporter. These datasets cover continental scale geographic regions and are available back to 2008 and in near real time. To effectively study LSTID activity in these datasets, an automated approach is necessary. Here, members of the Ham Radio Science Citizen Investigation (HamSCI) present results from a Tensor Flow-based machine learning algorithm for detection of LSTIDs. These results are compared with results from a manual visual search technique for validation purposes. Results are presented as a function of longitudinal sector (North America and Europe), season, and geomagnetic activity level.