Reconstruction of inner magnetospheric density, waves, and particle fluxes based on a neural network technique

Bortnik, Jacob

Reconstruction of inner magnetospheric density, waves, and particle fluxes based on a neural network technique

Wednesday, 13 February 2019: 08:45

Fountain I/II (Westin Pasadena)

Jacob Bortnik, University of California Los Angeles, Atmospheric and Oceanic Sciences, Los Angeles, CA, United States, Xiangning Chu, University of California Los Angeles, Department of Atmospheric and Oceanic Sciences, Los Angeles, CA, United States, Qianli Ma, UCLA, Los Angeles, United States, Chao Yue, Department of Atmospheric and Oceanic Sciences, University of California, Los Angeles, Los Angeles, CA, United States, Wen Li, Boston University, Boston, MA, United States, Richard Eugene Denton, Dartmouth College, Department of Physics and Astronomy, Hanover, NH, United States, Richard M Thorne, UCLA, Los Angeles, CA, United States and Vassilis Angelopoulos, University of California, Los Angeles, Department of Earth, Planetary, and Space Sciences and Institute of Geophysics and Planetary Physics, Los Angeles, CA, United States

Abstract:

The volume of space physics data continues to rise exponentially, and promises to accelerate its growth in the near future to the point that individual projects return on the order of a petabyte of data. At the same time, our analysis techniques have not kept pace with the rapid growth of data, and often do not exploit the capabilities of the data to their fullest potential. In this talk, we present a novel method based on machine learning technology, that aims to convert a sequence of point measurements of some given quantity Q made over a long period of time (for example observations made on a satellite), into a 3-dimensional dynamic spatiotemporal model of that quantity. As an example, we show a three-dimensional dynamic electron density (DEN3D) model in the inner magnetosphere, that can provide full coverage of the inner magnetosphere and in fact is sufficiently accurate that it points the way to new physical discoveries. For instance, we report, an unexpected plasmaspheric density increase at low L shell regions on the nightside during the main phase of a moderate storm during 12-16 October 2004, as opposed to the expected density decrease due to storm-time plasmaspheric erosion. Since plasmaspheric density values have been shown to be the largest source of error in radiation belt models, we also show reconstructions of whistler-mode chorus and plasmaspheric hiss waves, and show how these models can be used as inputs to downstream models, that can subsequently predict the dynamics of ‘data starved’ quantities, such as ultra-relativistic electron fluxes.