Speaker
Description
Solar photospheric line-of-sight magnetograms are easier to infer than vector magnetograms since the line-of-sight component (Blos) can be obtained from total intensity and circular polarization signals only, unlike the perpendicular component, which depends on harder-to-measure linear polarization. However, Blos is generally not physically meaningful although it is used in a variety of Space Weather models and applications. To produce an estimate of the radial component (Br) a common “correction” is often applied that includes an assumption which is nearly always false. We present "SSLOS", an ML-based model to estimate the full vector-field information from Blos, building on the recent SuperSynthIA approach that used Stokes vectors as input for simultaneous inversion and disambiguation (Wang+ 2024). As input, SSLOS accepts one or more line-of-sight magnetograms and associated metadata; as output, it estimates the full vector field in heliographic components, meaning that the physically-relevant vector components (including Br) are returned without need for further disambiguation steps or component transforms. SSLOS can allow for more complete evaluation of active regions for the likelihood of producing Solar energetic events when only using Blos, and enable significantly increased limits of distance-from-disk-center for data assimilation into models that rely on Br. We demonstrate estimates of the full vector field on unseen examples from both HMI and GONG, including examples that predate the Solar Dynamics Observatory mission. Our results show that while supervised learning is not a replacement for dedicated vector-field observing facilities, SSLOS may serve to unlock information from past data and, at the very least, provide more accurate Br maps from Blos than are created otherwise.
This work was primarily supported by NASA/MIRO 80NSSC24M0174 and NASA/LWSSC grant 80NSSC22K0892.
