Derivation of Boundary Conditions for Data-Driven Simulations of Active Regions and QS
Our objective is to develop insight into the spatial properties of solar coronal heating using a statistical analysis of the emission from observed and simulated active regions. To this end, we plan to perform data-driven MHD simulations of active regions. The MURaM simulation is being modified to work with photospheric inputs as boundary conditions, including observed vector magnetograms, and electric field maps and flow maps inferred from observations. We focus on photospheric flow maps derived through deep learning. More specifically, we train a convolutional neural network to emulate the MURaM simulation and infer MURaM-like flows from observational data, including flows in the granulation surrounding active regions. We present derivations of boundary conditions (i.e., electric field maps, flows maps) from SDO/HMI observations of selected active regions, and discuss the limitations and challenges associated with the methods. We detail planned efforts for driving the MURaM simulation from derived boundary conditions. Finally, we discuss how these data-driven simulations will be used to study the structuring of the emission of active regions statistically and identify which scenario of coronal heating best matches observations.
Benoit Tremblay received in 2019 his Ph.D. in Solar Physics from the Université de Montréal in Montréal, Canada, under the guidance of Professors Alain Vincent and Paul Charbonneau. His thesis explored the use of deep learning and neural networks in conjunction with magnetohydrodynamics simulations to generate synthetic observations of quantities like plasma velocities at the surface of the Sun. He then joined in 2019 the Laboratory for Atmospheric & Space Physics in Boulder, Colorado, as a George Ellery Hale Postdoctoral fellow to work with Professors Maria Kazachenko and Benjamin Brown. Finally, in mid-September 2022 (a few weeks away), Benoit will be joining HAO to work with Anna Malanushenko and Matthias Rempel on the derivation of boundary conditions for data-driven simulations via deep learning.