Space Weather!
See more about my work below!
Reconstructing Equatorial Electron Flux Measurements from low-Earth-orbit: A Conjunction Based Framework
D.L. Stumbaugh, J. Bortnik, and S.G. Claudepierre
Published in Space Weather
Summary:
We present a machine learning based model trained on a dataset that uses the global coherent nature of the radiation belts to reconstruct electron flux measurements near the geomagnetic equator. We establish conjunctions, or times, when the low-Earth-orbit (LEO) National Oceanic and Atmospheric Administration’s Polar Orbiting Environmental Satellites (POES) and the National Aeronautics and Space Administration’s near-equatorial Van Allen Probes are connected along the same magnetic field line and measuring the same electron population. Our conjunction dataset contains electron flux measurements, positional coordinates, and geomagnetic activity measurements. We use the conjunction dataset to train our machine learning model to reconstruct equatorial electron flux measurements. We show that the model performs well for data it was not trained on giving confidence that it can effectively predict fluxes. Our current work demonstrates that we can monitor in situ radiation belt fluxes using only relatively smaller and cost-effective satellites with a neural network model instead of the more traditional high-altitude satellites. The ability to predict radiation belt dynamics, and thus space weather, has become increasingly important for the broader society due to an increasing satellite infrastructure that is vulnerable to energetic electrons.
Reconstructing Equatorial Electron Flux Measurements From Data collected at Low-Earth-Orbit resolved in Magnetic Local Time
D.L. Stumbaugh and J. Bortnik
In Review in Space Weather
Summary:
We present five machine learning models that builds on an earlier model to reconstruct, predict equatorial electron flux measurements. We are able to do so by using the physical understanding of the global coherent nature between the equatorial and low-altitude fluxes along each field line of the radiation belts’ electron populations. We map the Polar Orbiting Environmental Satellites (POES) in low-Earth-orbit (LEO) to that of the Van Allen Probes (or Radiation Belt Storm Probes; RBSP) in equatorial, geostationary transfer orbit (GTO). Using the mapping process, we identify a total of over 76,500 conjunction events between POES and RBSP. Since each POES spacecraft orbits in its own magnetic local time swath, we train five separate machine learning models on roughly 27,000 conjunctions, one for each POES spacecraft. The training dataset for each POES spacecraft contains electron measurements, positional data, and geomagnetic activity indices. The model performs well on out-of-sample data, data it was not trained on, indicating that we can reconstruct GTO observations from LEO. By adding the MLT component, we show the potential for future means to in real-time, monitor the radiation belts from LEO satellites.
