Speaker
Description
The north–south displacement of the cosmic-ray Sun shadow has been shown to precede variations in near-Earth IMF $B_{y}$ by approximately 3.3 days. Whether this leading signal improves predictions beyond existing near-Earth observations, however, remains to be established.
In this work, we use multiple data sources from KM2A Sun shadow and OMNI data between 2020 and 2024, and apply Ridge regression to compare with and without Sun shadow information. Models incorporate magnetic-field history, 27-day recurrence, and additional solar-wind parameters. Training used the data from 2020 to 2022, with separate evaluations in 2023 and 2024. Uncertainty is estimated using a calendar-block bootstrap. For non-ICME samples in 2024, adding Sun shadow information to the baseline containing magnetic-field history and 27-day recurrence reduces the RMSE by 10.2% and increases the Pearson correlation coefficient from 0.235 to 0.425. Positive gains persist in 2024 with an enhanced near-Earth observational baseline and in a retrospective analysis incorporating photospheric and PFSS magnetic-field features.
These results provide empirical support for combining ground-based Sun shadow measurements with other observations to predict near-Earth magnetic-field variations, although the physical origin of the incremental information and its applicability to real-time forecasting require further investigation.
