Author: Dr. Harriet Turner
The solar wind is a constant flow of charged particles that flow out of the sun and fill the solar system. This is an important aspect of space weather, a term we use to describe changing conditions in near-Earth space. Weather in space can cause effects on the planet, such as radio communication problems, damage to spacecraft, and potentially damage the health of astronauts and people in high altitude flights. Severe space weather is driven by coronal mass ejection (CME), which are brief eruptions of materials and magnetic fields from the sun. They propagate through the solar wind, and their conditions affect the CME speed and arrival time to the Earth. The solar wind itself can also be a source of recurring space weather, as its fast and slow flow structure rotates as the sun rotates.
Current solar wind prediction methods use observations of solar magnetic fields to initialize models of the solar external atmosphere. The outer boundary of the coronal model can then be used as the inner boundary of the solar system model, which propagates solar energy to Earth and later (called Earth model). After the initial magnetic field observation, the modeling system does not contain further observation constraints, thus limiting the prediction accuracy. Data assimilation (DA) is a technique that combines model output with system observations to find the best estimate of reality. It has been widely used in ground weather forecasts and has led to a large forecast improvement, but is inadequate in space weather forecasts. Here, at the University of Reading, we developed a novel solar wind DA scheme (Hamburg Radius Difference Data Assimilation or BRAVDA) that uses in-situ observations of solar wind speed to update the internal boundaries of the Earth stratigraphic model.

Figure 1. Schematic diagram of the BRAVDA scheme. This is a view overlooking the ecliptic plane above the North Pole of the Sun. The earth is a black circle with its orbital radius shown in the black line. Observing in the Red Cross (a), it can be mapped to the internal boundary (b) at some point in the past. This information can be used to update the internal boundaries and can propagate outward (C), resulting in improved solar wind state. Then it can be run forward to produce predictions (d).
The observations are taken from spacecraft located in orbits similar to Earth. They contain information about the conditions close to the Sun at some time in the past, and this can be used to improve the inner boundary of the heliospheric model, as shown in Figure 1. Previous work (eg Turner 2022, Turner 2023) has shown that BRaVDA is effective in improving the inner boundary of the heliospheric model when using the Magnetohydrodynamics Around a Sphere (MAS) coronal model, which is used primarily for research purposes due to its long and ready-to-use archives. In the operating environment, the Wang-Sheeley-Arge (WSA) model is more widely used when updated every day. The Metropolitan Office Space Weather Operations Centre (MOSWOC) and the National Oceanic and Atmospheric Management Space Weather Forecasting Centre (NOAA SWPC) use it.
I've been using the WSA model lately using Bravda. Due to its use in government organizations, data is not readily available, so we are limited to our analysis in 2020. In the first half of this year, WSA output produced unusually high solar wind speeds, so solar wind forecasts on Earth were too high. The reason is not yet clear. However, it provides an interesting opportunity to see if DA can improve predictions. As shown in Figure 2, DA provides a significant improvement in prediction error, reducing the mean absolute error (MAE) by about 50%. This suggests that the use of DA in operational predictions can not only provide important prediction improvements, but also help correct system bias caused by errors under input conditions.

Figure 2. The predicted average absolute error (MAE) varies with the predicted delivery time. Black lines show predictions driven by WSA before any data assimilation occurs. The red line shows the prediction after DA occurs.
Currently, this work is being written into a paper and hopefully will be submitted soon. The study also shows the best way to process BRAVDA output for use in the Earth model used in the study. This is a necessary condition for use in any Heliosphere model and sets up a framework for how best to use BRAVDA's output. The next study was to investigate the effect of using DA on CME speed and time to arrive estimation. As the background solar wind improves, it is hoped that this can lead to more accurate CME predictions.
refer to
Turner, H., Owens, MJ, Lang, MS, Gonzi, S. , & Riley, P. (2022). Quantify the effect of ICME removal and observe the effect of age on in situ solar wind data assimilation. Space Weather, 20 years old, E2022SW003109. https://doi.org/10.1029/2022sw003109.
Turner (H. Solar wind data assimilation in operating environments: using near real-time data and predicted values of L5 monitors. Space Weather, 21 years old, E2023SW003457. https://doi.org/10.1029/2023sw 003457.