AUTHOR=Swiger Brian M. , Liemohn Michael W. , Ganushkina Natalia Y. 

TITLE=Improvement of Plasma Sheet Neural Network Accuracy With Inclusion of Physical Information

JOURNAL=Frontiers in Astronomy and Space Sciences

VOLUME=Volume 7 - 2020

YEAR=2020

URL=https://www.frontiersin.org/journals/astronomy-and-space-sciences/articles/10.3389/fspas.2020.00042

DOI=10.3389/fspas.2020.00042

ISSN=2296-987X

ABSTRACT=There is a strong connection between the variations in the Sun’s extended and supersonically outflowing atmosphere, called the solar wind, and the Earth’s space environment. Behind the Earth is a region called the plasma sheet, a layer of electrically charged particles that is the source of electrons for the inner regions of near-Earth space, and we know that there is a dependence of the plasma sheet upon the solar wind. Many studies have used statistical models to help describe the relationship because of the relatively sparse measurements available in the plasma sheet. Presently, there are over two 11-year solar cycles of observations of the plasma sheet and upstream solar wind, and we have enough computational resources to use machine learning tools to help us more thoroughly understand the connection between these two observational regions. Therefore, we report the development of a machine learning, neural network model to help predict the electron flux in the plasma sheet from solar wind observations. Several authors have been inquisitive about the utility using “black- or grey-box” types of approaches to modeling in space science. Our preliminary findings suggest that using physics informed neural networks can help improve overall understanding.