NCAR-WY - Mehdi Nourelahi
Using Machine Learning to Find a Mapping Between States of Polarization and Atmospheric Model Variables
Due to the Sun's constant effect on Earth’s space environment, e.g. solar wind and CMEs, it would be highly beneficial to understand the physical causes of these phenomena. To study and measure magnetic fields on the Sun we use polarized light which provides a wealth of information that can be used to infer the state of the solar atmosphere. Inversion procedures infer the solar atmospheric temperature, density, velocity, and magnetic field by using states of polarization, i.e. Stokes profiles. Large quantities of polarized spectra will soon be provided by Daniel K Inouye Solar Telescope (DKIST). The inversion procedure is computationally expensive and is unable to keep pace with the large data volumes from DKIST. We explore the use of clustering techniques on maps of physical quantities from polarimetric inversion codes, as well as the raw Stokes spectra, with the goal of finding a meaningful relation that can be used in place of the prohibitively expensive inversion.
Mentors: Ricky Egeland, Rebecca Centeno, Roberto Casini