# Forecasting Periods of Strong Southward Magnetic Field Following Interplanetary Shocks

## TL;DR

This study develops a probabilistic forecasting model for long-duration southward magnetic fields following interplanetary shocks, improving early warning capabilities for geomagnetic storms by analyzing solar wind and shock data over 22 years.

## Contribution

The paper introduces a novel shock-based probabilistic forecast method that significantly enhances prediction accuracy and warning time for geomagnetic storms compared to previous models.

## Key findings

- 76% of strong southward Bz periods are preceded by shocks within 48 hours
- The model provides an average 14-hour warning for intense Bz events
- It outperforms random chance and previous forecasting approaches

## Abstract

Long periods of strong southward magnetic fields are known to be the primary cause of intense geomagnetic storms. The majority of such events are caused by the passage over Earth of a magnetic ejecta. Irrespective of the interplanetary cause, fast-forward shocks often precede such strong southward B$_{z}$ periods. Here, we first look at all long periods of strong southward magnetic fields as well as fast-forward shocks measured by the \textit{Wind} spacecraft in a 22.4-year span. We find that 76{\%} of strong southward B$_{z}$ periods are preceded within 48 hours by at least a fast-forward shock but only about 23{\%} of all shocks are followed within 48 hours by strong southward B$_{z}$ periods. Then, we devise a threshold-based probabilistic forecasting method based on the shock properties and the pre-shock near-Earth solar wind plasma and interplanetary magnetic field characteristics adopting a `superposed epoch analysis'-like approach. Our analysis shows that the solar wind conditions in the 30 minutes interval around the arrival of fast-forward shocks have a significant contribution to the prediction of long-duration southward B$_{z}$ periods. This probabilistic model may provide on average a 14-hour warning time for an intense and long-duration southward B$_{z}$ period. Evaluating the forecast capability of the model through a statistical and skill score-based approach reveals that it outperforms a coin-flipping forecast. By using the information provided by the arrival of a fast-forward shock at L1, this model represents a marked improvement over similar forecasting methods. We outline a number of future potential improvements.