New capabilities for prediction of high‐latitude ionospheric scintillation: Novel exploration through machine learning

Abstract:

As societal dependence on trans‐ionospheric radio signals grows, space weather impact on these signals becomes increasingly important yet our understanding of the effects remains inadequate. This challenge is particularly acute at high‐latitudes where the effects of space weather are most direct and no reliable predictive capability exists. We take advantage of a large volume of data from Global Navigation Satellite Systems (GNSS) signals, increasingly sophisticated tools for data‐driven discovery, and machine learning algorithms that span the spectrum of complexity (from `simple’ and interrogable algorithms such as the Support Vector Machine [McGranaghan et al., 2018] to complex and less easily explained approaches such as Deep Learning) to develop novel predictive models for high‐latitude ionospheric phase scintillation. We find that machine learning approaches significantly outperform current predictive capabilities, which, at high-latitudes, only consist of climatology and persistence.

This talk will feature the details of our machine learning exploration, emphasizing the critical techniques required to bring together diverse and voluminous data across the solar-terrestrial connection. We will discuss the broader impact of our work on space weather forecasting, including the importance of data science innovation and the use of robust evaluation techniques, and spark discussion about data science in Heliophysics.

McGranaghan, R.M., A.J. Mannucci, B.D Wilson, C.A. Mattmann, and R. Chadwick. (2018), New capabilities for prediction of high‐latitude ionospheric scintillation: A novel approach with machine learning, Space Weather, 16. https://doi.org/10.1029/2018SW002018.