A new approach for predicting geomagnetic Kp and Hpo indices using machine learning techniques

Widely used geomagnetic activity indices such as Kp or Dst are derived from the combined data from several geomagnetic observatories that are distributed over the globe to provide a global index. Forecasting such indices is crucial as solar-driven geomagnetic activity can significantly affect both technology and human activities on Earth and in the near-Earth space environment.

We developed a new model to forecast geomagnetic indices by incorporating predicted data from individual observatories. Unlike previous models that relied directly on an index and ignored diverse physical effects at individual observatories, this approach considers each observatory separately in the forecasting process. It thus produces predictions of global geomagnetic indices that integrate the same physical principles as in the original calculations of the Kp index.

We demonstrate the performance of the model for the Kp index along with the recently derived Hpo indices, which all measure planetary geomagnetic disturbances caused by solar activity. The Hpo indices, Hp60 and Hp30, provide high-resolution (hourly and half-hourly, respectively) representations of these disturbances, similar to the 3-hourly Kp index but without the upper limit of 9. The model demonstrates good agreement, accurately capturing trends and overall behaviour, even with sparse solar wind data.