One of the most dramatic displays of aurora borealis (northern light) is called “substorm onset”. You will first see auroral band moving toward south and then the band suddenly break into multiple bands of light moving and dancing to the North.  This sudden break of aurora is the substorm onset during which energy is released from the Earth’s magnetosphere into the ionosphere to brighten aurora. The energy is stored earlier during the 30-60 minute period when the auroral band is moving southward. We call this period “substorm growth phase”. If we understand how the energy is stored and distributed in the Earth’s magnetosphere, then we can get a better picture on how and where the energy can be released. Our project is therefore to use observations measured by satellites in space and utilize some basic physical principles and state-of-art numerical techniques to map and predict the energy storage.
The energy originally comes from the Sun and can be stored in magnetosphere by heating plasma, as boiling water, and bending magnetic field, as stretching an elastic band. Several satellites have been launched into space to measure plasma and magnetic field at different locations of the magnetosphere. So we first to build a big database of these measurement that have been collected from many years. This is like to know for example how different cities in US receives different sunlight for storing solar power. Since we want to find out from this big data set if there is a pattern in the energy storage, such as the location where the largest energy is stored, as if to find out from a huge data base of stock market which companies have the largest invested money, we take advantage of new technique that is called computer machine learning to analyze the data. From the analysis, we have found out that the energy for substorms is mostly stored at a location about 10-14 Earth radii (RE) from the Earth on the nightside. But this location and how much energy is stored can vary depending on the Sun’s activity. We thus build a model by applying the machine learning and physical principles to empirically predict this energy storage by presenting it in the form of 3-dimensional configuration of plasma energy and magnetic field shape. Like accuracy is important to a weather forecast model, we also tested our model and found that it can provide quite reliable prediction to most of Sun’s activity. This empirical substorm model is the first of its kind.
We also use our model to study many physical processes associated with substorms, including the process that leads ion plasma from the magnetosphere to the ionosphere to generate proton aurora and the process that can possibly destroys the balance between heating plasma and stretched magnetic field and leads to substorm onset, as if a blown balloon eventually blows up. Since our model is reliable, we believe that we have found more convincing answers to these physical processes than previous research.

Last Modified: 12/30/2015
Modified by: Chih-Ping Wang