Improving Geomagnetic Perturbation Forecasts in the Southern Hemisphere via Kernel‐Based Encoding of Mirror Symmetry

Abstract

Accurate forecasting of geomagnetic perturbations is essential for assessing space weather risk of geomagnetic activities. Thanks to decades of magnetometer observations from world‐wide distributed networks and upstream solar wind observations at L1, data‐driven modeling of global geomagnetic perturbations has become increasingly feasible. Such models are promising for operational use because they typically run at lower computational cost and can offer higher predictive skill than first‐principles, physics‐based simulations when observations available for training are abundant. However, this potential has not been fully realized as sparse observational coverage in the Southern Hemisphere (SH) has long limited model generalizability. In this work, we use a zeroth‐order mirror‐symmetry assumption in ionospheric electrodynamics under the joint sign reversal of magnetic latitude , interplanetary magnetic field and dipole tilt to improve SH predictions using information from the densely observed Northern Hemisphere (NH). Rather than enforcing exact symmetry, we build on our previous GeoDGP model by introducing a symmetry‐aware kernel for a deep Gaussian process that learns the symmetry‐asymmetry balance from data. For comparison, we also experiment with an alternative approach that imposes symmetry as a hard constraint via data augmentation. We evaluate the models on 22 geomagnetic storms and present a case study of the 2024‐05‐10 Gannon extreme storm. The results show that both methods significantly improve the SH predictions, while the kernel‐based approach achieves the best performance and maintains NH accuracy comparable to GeoDGP.