Speaker
Description
Geomagnetic storms pose substantial risks to modern technological infrastructure, making accurate forecasting critical for mitigating their impacts on essential systems. Traditionally, space weather research has emphasized forecasting global geomagnetic indices. While these global indices provide valuable insights, local geomagnetic indices remain relatively understudied, despite their crucial role in capturing regional disturbances. A recent event is the May 2024 geomagnetic storm, which triggered auroral activity at low latitudes, yet with considerable longitudinal differences. This event highlighted the necessity of advancing forecasts for local geomagnetic indices to capture regional variations in space weather impacts.
To address this need, our study focuses on forecasting local geomagnetic indices, specifically the Local Disturbance Index (LDi), using data from five globally distributed geomagnetic observatories. We have developed a neural network model that uses historical geomagnetic observations alongside real-time solar wind data from the ACE spacecraft. Model inputs include previous LDi values specific to each observatory, solar wind parameters such as interplanetary magnetic field components, and plasma characteristics including density, velocity, and temperature. To represent geographical and temporal variations in geomagnetic activity, we also consider the magnetic local time (MLT) and observatory latitude. Moreover, our forecasts include prediction intervals at a 90% confidence level, quantifying the uncertainty and enhancing their operational utility.
To validate the network’s performance, we compare its forecasting errors against the inherent discrepancy between local indices and the global SYM-H index. Specifically, we evaluate whether the error in predicting future LDi values is smaller than the actual difference between the observed LDi and SYM-H at the forecast’s initial time. Our results demonstrate that the neural network reduces forecasting errors compared to using SYM-H as a proxy for local disturbances. This improvement highlights the limitations of relying on global indices to estimate regional geomagnetic activity and emphasizes the value of localized forecasting frameworks. By prioritizing regional specificity and uncertainty quantification, our approach advances the capacity to predict space weather impacts with the precision required to safeguard critical infrastructure.
| Do you plan to attend in-person or online? | In-person |
|---|
