Our Sun is dynamic. Solar activity peaks approximately every 11 years and this rise in activity increases the chance of rare solar wind events (e.g. coronal mass ejections). The most recognizable effect of this rise in solar activity is the increase of planetary aurorae. It was recognized in October 2024 by NASA, NOAA and the international Solar Cycle Prediction Panel that we have entered the solar maximum period of Solar Cycle 25. Accordingly, in the last year Vancouverites have been fortunate to be able to observe visible aurora. These low-latitude aurora were caused by the interaction of Earth’s magnetosphere with several large coronal mass ejections (e.g. space weather).

Such space weather effects are not solely limited to Earth. Spacecraft assets throughout the solar system observe our Sun and the solar wind. Unfortunately, these datasets can only provide discontinuous spatiotemporal observations of a very large and dynamic system. To bridge our understanding of this sparse data and our physical understanding of space weather impacts, these data are often combined with high-fidelity physical models (e.g. magnetohydrodynamics, MHD). 

Like other global fluid dynamics-based models (e.g. GCMs), MHD models can be computationally expensive. This limits the potential number of simulations for comparison to observations. This combined bottleneck of sparse datasets and expensive forward physics-based models is a ubiquitous challenge for inverse problems in the Earth and space sciences. In the case of Mars, this methodological bottleneck has limited our understanding of how the rise, and fall, of solar cycle activity affects planetary habitability.

In this presentation I will briefly introduce the solar cycle and space weather before discussing a rare space weather event we observed at Mars with the NASA MAVEN mission known as the “the day the solar wind disappeared”. I will then show how this event can be posed as an inverse problem, a solution estimated using a machine learning low-fidelity emulator (Gaussian processes) of a physics-based (MHD) model, and discuss the scientific conclusions we are gaining about Mars’ space environment from this effort. 

I will conclude with an outlook for how this process incorporated and quantified uncertainties, and is an emerging solution for addressing the challenge of bridging physical models, machine learning, and inference problems in the Earth and space sciences.