Identifying the Signatures of Sporadic-E layers in SuperDARN backscatter Using Machine Learning

The Super Dual Auroral Radar Network (SuperDARN) is an international network of High Frequency (HF) radars which are used for continuously monitoring ionospheric space weather phenomena. Most SuperDARN backscatter comes from either ionospheric plasma irregularities in the F-region or reflections from the ground/sea surface. In addition, SuperDARN radars also observe “grainy” near- range echoes which have been attributed to backscatter from meteor trails left by meteoroids as they enter the Earth's atmosphere. However, previous studies have shown that near-range SuperDARN backscatter can also occur from other sources. In particular, backscatter from Sporadic-E (Es) layers can produce visually distinctive near-range signatures that can oftentimes be wrongly classified as meteor echoes. The primary objective of this study is to develop a machine learning classification algorithm to distinguish between meteor echoes and backscatter from Es layers. This algorithm will then allow SuperDARN to be used as a tool for monitoring Es over vast geographical regions, continuously, with an unprecedented resolution for the longitudinal coverage. We employ a two-stage algorithm, namely, clustering and sorting. In the clustering stage, we apply a HDBSCAN (Hierarchical Density-Based Spatial Clustering of Applications with Noise) algorithm with weighted inputs to separate the data into clusters. In the sorting stage, the near-range clusters are classified into categories based on differences in the distributions of the input parameters. Finally, we will validate the ability of our algorithm to remote sense Es layers using independent and collocated observations from Digisondes and GPS RO observations.