SIML-HSS

This model forecasts the solar wind speed at the Earth using a machine learning algorithm based on solar images. It focuses on high-speed solar wind streams (HSSs) and their solar source regions, coronal holes. As input, it uses the coronal hole area, extracted from solar extreme ultraviolet (EUV) images and mapped on a fixed grid, as well as the solar wind speed 27 days before. A polynomial regression algorithm is employed to compute the solar wind speed with a cadence of one hour and a lead time of four days (SIML-SW). Additionally, the model applies a distribution transformation to the predictions such that they match the underlying observed solar wind speed distribution, which improves the accuracy of HSS peak predictions (SIML-HSS).

The model is described in more detail in the publication: Forecasting high-speed solar wind streams from solar images (see Publications section).

It is implemented in Python and the source code is publicly accessible on GitHub.

The model is currently not running in real time. It does not include CME predictions.

• The coronal hole area coupled to its location on the solar disk from 4, 5, 6, and 7 days before the predicted time point (extracted from a 4x3 grid on coronal hole segmentation maps of solar images).
• The observed solar wind speed at Earth from 26, 27, and 28 days before the predicted time point.
• The smoothed monthly sunspot number and the slope of the smoothed monthly sunspot number.

The solar wind speed at the Earth (near the bow shock as provided by OMNI) in an hourly cadence four days in advance.