| Presentation Title: | Full Sun Spectral Analysis of Hinode/EIS data using Machine Learning |
| Author(s): | James McKevitt, Sarah Matthews, Deborah Baker, Hamish Reid |
Abstract:
Throughout its mission, Hinode/EIS has performed over 20 full disk observations, capturing a wide range of active regions at multiple wavelengths. This provides a large dataset from which many observables like intensity, and fitted parameters like non-thermal velocity, can be derived. As the physical origins of non-thermal velocity remain ambiguous, these full disk observations are a valuable resource, useful to help us better understand the statistical behaviour of non-thermal velocity in active regions alongside other key parameters.
In this study, we analyse three Hinode/EIS full-disk observations from January 2013, April 2015 and October 2015, and use machine learning techniques to analyse the relationship between intensity, Doppler velocity and non-thermal velocity at multiple wavelengths, and FIP, in the quiet Sun and different active region types. In addition to considering these parameters alongside one another, we systematically apply clustering algorithms and use the result to characterise the significance of each parameter to each active region type and plasma condition (e.g. quiet Sun or coronal hole). We also apply machine learning techniques to capture the importance of the non-linear relationships between the parameters.
We use this study to consider known models of solar physics (e.g. FIP and I-FIP) alongside large sets of data, to demonstrate the usefulness of machine learning techniques in this data’s analysis, and to pave the way for an extended analysis of the other full disks observed by Hinode/EIS during its mission.