The Earth’s magnetosphere is a complex system composed of about a dozen major subsystems that are not only complex in-and-of themselves but also interconnected in complex, sometimes nonlinear ways. In other words, a ‘system of systems’. The inner magnetosphere is home to several coupled subsystems including the ring current, plasmasphere, and radiation belts which are, in turn, also coupled to the magnetopause/solar wind and the ionosphere/thermosphere/mesosphere system.
Modern space missions are designed to focus on one system and one, limited, set of science questions at a time. Typically, missions do not overlap in time and are not planned to complement one another in any deliberate way. Numerical models can be key to developing system-level understanding. A model, however, is often declared ‘successful’ if it reproduces a limited set of data to some level of fidelity showing that the observations are ‘consistent with’ our current understanding of the physics of the system - or at least one set of processes.
Several modeling techniques that incorporate data to produce more accurate predications are in widespread use and have proven extremely successful. They include artificial intelligence (e.g. neural networks), data assimilation, and others.
In this talk I present some ideas that, to me, are more exciting than prediction. It has been demonstrated that models and data can be used together, synergistically, to discover previously unknown relationships and underlying physical processes. We might describe these models as ‘data-intelligence’. I will give some examples of data-intelligent (DI) model successes and talk about ways we might push the capabilities of existing techniques in new, perhaps unusual ways. The talk will focus on applications to the physics of the inner magnetosphere but the ideas should be generally applicable to the broader solar-terrestrial system-of-systems.
- Invited by Yuri Shprits