Enhanced lead time for geomagnetic storms

In this project we combine in an innovative way existing simulations with observations by current and possible future spacecraft to improve the lead time for the prediction of geomagnetic storms due to impacts of solar coronal mass ejections (CMEs). Forecasting the solar wind that interacts with the Earth’s magnetosphere is currently severely limited, with accurate forecasts only ranging from 1 to 3 hours. This situation should clearly be enhanced in order to be able to predict the time evolution of geomagnetic storms, which has many consequences ranging from geomagnetically induced currents to energetic particles disturbing satellites or simply pushing the location of the auroral oval to lower latitudes, which is of high relevance for aurora tourism and public space weather outreach.

The main idea behind this proposal is to use L1 data as a real-time constraint on the physical modeling on coronal mass ejection flux ropes (CMEs). Their organized magnetic flux rope fields at the cores of CMEs must make it possible to use them as a key to unlock their magnetic field time evolution up to 48 hours in advance. The other driver behind this proposal is a convergence of the recent availability of 40 years of high-resolution solar wind data at 1 AU (Wind, STEREO, ACE, DSCOVR), and the availability of machine learning codes. Taken together, ensemble runs of our 3DCORE model for CME flux ropes, producing synthetic solar wind observations, will be downselected by machine learning algorithms based on training data from L1 observations. This will lead to a forecast of the L1 solar wind from > 1 hour up to 48 hours in advance during CMEs.

Figure 1: Example of a coronal mass ejection observed by coronagraphs (upper left) and in situ at Solar Orbiter and Wind (upper right). High speed solar wind streams can cause aurorae but the brightest northern lights (bottom panels) are related to the impact of such solar storms.

Figure 2: Plot of (a) geomagnetic variations at FUR (normalised to around zero by subtracting the mean field strength), (b) modeled GICs at two substations, and (c) cumulative hourly GICs on 2000-09-17 as an example of a day that likely had no extreme GIC values but large cumulative hourly GICs. [Bailey+ 2022].

A pipeline combining 3DCORE, machine learning, a prediction of Dst from the solar wind and auroral modeling will be established for the first time as an open source code. The solar wind prediction output will be connected to geomagnetically induced current modeling for central Europe at a national collaborating institution. We will then proceed to test the implementation with skill scores, first with hindsight data and later in real time. Additionally, two different concepts of interplanetary CubeSats acting as solar wind monitors will be tested with existing data from STEREO, Venus Express and MESSENGER.

PI: Christian Möstl
Postdoc: Rachel L. Bailey
Postdoc: Martin A. Reiss
Duration: 03/2019 – 02/2023
Link to FWF project page: P 31659

Acknowledgements: This research is funded by the Austrian Science Fund (FWF) [P 31659].