Abstract/Summary

Accurate space weather forecasting requires knowledge of the solar wind conditions. Current forecasting methods are initialised with photospheric magnetic field observations but contain no information from observations of the solar wind. Data assimilation (DA) is a method that combines a model of a system with observations to find an optimum estimation of reality. DA has led to large forecast improvements in terrestrial weather forecasting but has been underused in space weather forecasting, especially in the solar wind. The first part of this thesis investigates the errors present in solar wind forecasts due to the latitudinal separation of spacecraft observations and the forecast location. It is important to understand how the observation locations affect the forecast accuracy to better inform the data assimilation scheme in the future. It was found that up to a separation of 6 degrees in latitude, there was no significant impact on forecast accuracy. Secondly, the impact of observational age and the presence of CMEs on solar wind speed forecasts produced using DA is examined. This allows us to understand how the forecast error is expected to change across the solar cycle and under different observing configurations. It was found that assimilating multiple sources of observations improves the forecast accuracy. Removing CMEs caused a marginal improvement in forecast accuracy, but it would be difficult to achieve in real time. The final part of this thesis considers the use of near-real-time (NRT) data in the DA scheme. This is in preparation for it to be used for operational forecasting. The DA still performed well when using NRT data, showing that it is suitable for operational use. Through simulations of similarly configured spacecraft, it was also shown that the future Vigil mission would be useful for solar wind DA.