Near real-time construction of solar coronal parameters based on MAS by deep learning.

In this study, we apply the Pix2PixCC model to reproduce the 3D distribution of solar coronal parameters (density, magnetic field, and temperature) from 2D synoptic photospheric magnetic fields. 5280 pairs of inputs and outputs are considered for training, validation, and testing from 2010 June to 2023 January, which is simulated with the MHD Algorithm outside a Sphere (MAS) model. To cover each parameter’s range of 1 to 30 solar radii, we train 54 distinct deep-learning models. Our study’s key findings are as follows. Firstly, our model reproduces 3D coronal parameter structures from 1 to 30 solar radii with an average correlation coefficient value of about 0.96. Secondly, during both the solar active and quiet periods, the AI-generated data are consistent with the target MAS simulation data. Thirdly, compared to the usual MAS simulation time, the 54 deep-learning models take about 41 seconds to generate the results with an NVIDIA Titan XP GPU. As the MAS simulation is a regularization model, we may significantly reduce the simulation time by using our results together with the coronal density data produced in a similar way as an initial magnetic configuration to obtain an equilibrium condition. Furthermore, we hope that the generated solar coronal parameters can be used for near-real-time forecasting of heliospheric propagation of solar eruptions in the future.