Global magnetic field extrapolation using solar backside magnetograms generated by deep learning

In this study, we have improved solar coronal magnetic field extrapolations by using more realistic backside magnetograms generated by deep learning. For this, we apply a deep learning model based on cGAN by training a pairs of SDO/AIA Extreme-Ultraviolet (EUV) three channel (17.1, 19.3 and 30.4 nm ) images and SDO/HMI magnetograms. For testing data sets, we demonstrate that the generated magnetic field distributions are consistent with the real ones. Then we generate the backside magnetograms from the corresponding three channel images of STEREO/EUVI-A and -B by the model. We make global magnetic field synchronic maps in which conventional backside magnetic data are replaced by backside ones generated by our model. The synchronic map shows much better the evolution of active regions (appearance or disappearance) than before. We use these synchronized data to extrapolate the coronal magnetic fields using Potential Field Source Surface (PFSS) model. The results of coronal fields, calculated from both the conventional and the synchronized data, are compared with coronal EUV observations from SDO and STEREO spacecrafts. Our results are more consistent with the observations than those of the conventional method in view of solar active regions and open field regions (coronal holes). And the results show more consistently the sequences of coronal structure changes over several solar rotation. Finally we will discuss the application of our new methodology and future prospects.

This work was supported by Institute for Information & communications Technology Promotion(IITP) grant funded by the Korea government(MSIP) (2018-0-01422, Study on analysis and prediction technique of solar flares).