Background

Geomagnetic activity indices such as Kp and the higher-resolution Hpo (Hp60 and Hp30) are essential for forecasting, as solar-driven geomagnetic disturbances can significantly impact technological systems and human activities on Earth and in near-Earth space. Our new forecast model (Kervalishvili et al., 2025) improves prediction accuracy by incorporating data from individual Kp-observatories, capturing local physical effects often overlooked by traditional models based on a single global index. By first predicting local indices and then combining them using standard procedures to generate global indices like Kp and Hpo, the model bridges the gap between localized measurements and global forecasts, preserving site-specific physical influences. This localized approach enhances the accuracy of the Kp, Hp60, and Hp30 indices, effectively capturing key trends and overall geomagnetic behaviour even when solar wind data is sparse or incomplete.

Note that unlike the Kp index (Matzka et al., 2021), which is limited to a maximum value of 9 and has a three-hour resolution, the Hpo index (Yamazaki et al., 2022) is open-ended and available at higher temporal resolutions, one hour (Hp60) and half an hour (Hp30), providing a more detailed and scalable representation of geomagnetic storm intensity.

Forecast Model

We have developed a machine learning-based forecast model that leverages solar wind parameters from the high-resolution OMNI dataset and the Sunspot number from the low-resolution OMNI dataset. Minute-by-minute data is used for parameters such as IMF Bt, proton density, and solar wind speed, while hourly data is utilized for the Sunspot number. This model is built on the simplified framework of Kervalishvili et al. (2025), focusing on IMF Bt and a constant Sunspot number (SSN) to generate forecasts.

Figure Description

The figures below present a 3-day forecast of geomagnetic activity Hs30, Hs60, and Ks indices. The first panel displays the Hs30 index, the second shows Hs60, and the third presents the Ks index — each corresponding to individual Kp-observatories. All values are colour-coded using a scale based on the NOAA Space Weather Scales, ranging from quiet conditions to extreme geomagnetic storms.