Improved AI-generated Solar Farside Magnetograms by STEREO EUV Observations and Their Applications

We have greatly improved Artificial Intelligence-generated Solar Farside Magnetograms (AISFMs) from STEREO EUV observations than before. We have modified our previous deep learning model and configuration of input datasets to generate more realistic magnetograms. First, our model uses an updated loss function which includes correlation coefficients between the real and generated data. Second, we construct input data sets of our model: solar farside EUV observations together with the frontside data pairs of EUV observations and magnetograms. Our results show that the present model is much better than our previous model (Jeong et al. 2020, ApJ Letter) in view of several metrics. We can monitor the temporal evolution of active regions using our AISFMs together with the frontside SDO/HMI magnetograms. In addition, the AISFMs produce consistent polar field strengths and magnetic field polarities with those of nearby frontside ones for solar cycles 24 and 25. Our AISFMs are publicly available at http://sdo.kasi.re.kr. We present several applications of the AISFMs. We construct synchronic global magnetic field maps with the SDO/HMI and AISF magnetograms, and extrapolate coronal magnetic fields from the maps. We show that our results are much more consistent with EUV observations than those of the conventional method in view of solar active regions and open field regions. The results show more consistently the sequences of coronal structure changes over a solar rotation. Finally we suggest several prospects to study global magnetic connectivity with multi-view point observations, e.g., STEREO, Parker Solar Probe, and Solar Orbiter.