Data-Driven Solar Surface Flux Transport Modeling with Uncertainty Quantification

Solar active regions (ARs) are areas of increased magnetic flux on the Sun's surface. Their downstream effects, including phenomena such as solar flares, Solar Energetic Particle events (SEPs) and coronal mass ejections (CMEs), can impair Earth- and space-based human infrastructure. To mitigate these dangers, the prediction of such events and their precursors --namely the ARs-- are necessary, and uncertainty quantification (UQ) of these predictions are crucial for subsequent human decision-making. In this manuscript, we present Active Region Characterization and Analysis of Dynamics and Evolution (ARCADE), an ongoing effort to provide AI-driven and UQ-equipped forecasting pipeline for solar surface magnetic field evolution. ARCADE is trained by propagating a ResNet through a numerical simulation of the solar surface magnetic field. In this work, we validate our numerical integrator against the state-of-the-art Advective Flux Transport model (AFT) and demonstrate that ARCADE offers high-accuracy forecasts with consistent uncertainty estimations. Moreover, ARCADE is the first model of surface magnetic evolution to produce forecast of the emergence of flux several hours into the future. Finally, we provide a convenient user interface for forecasting with UQ, accessible here: http://34.10.144.174:4000/. This work is a promising first step for hybrid physics-ML methods for solar magnetic field forecasting with UQ, offering clear interpretability and usability for heliophysicists and stakeholders in space weather prediction.