Classification of High-resolution Solar Hα Spectra Using t-distributed Stochastic Neighbor Embedding

Abstract

The Hα spectral line is a well-studied absorption line revealing properties of the highly structured and dynamic solar chromosphere. Typical features with distinct spectral signatures in Hα include filaments and prominences, bright active-region plages, superpenumbrae around sunspots, surges, flares, Ellerman bombs, filigree, and mottles and rosettes, among others. This study is based on high-spectral resolution Hα spectra obtained with the Echelle spectrograph of the Vacuum Tower Telescope (VTT) located at Observatorio del Teide, Tenerife, Spain. The t-distributed stochastic neighbor embedding (t-SNE) is a machine-learning algorithm, which is used for nonlinear dimensionality reduction. In this application, it projects Hα spectra onto a two-dimensional map, where it becomes possible to classify the spectra according to results of cloud model (CM) inversions. The CM parameters optical depth, Doppler width, line-of-sight velocity, and source function describe properties of the cloud material. Initial results of t-SNE indicate its strong discriminatory power to separate quiet-Sun and plage profiles from those that are suitable for CM inversions. In addition, a detailed study of various t-SNE parameters is conducted, the impact of seeing conditions on the classification is assessed, results for various types of input data are compared, and the identified clusters are linked to chromospheric features. Although t-SNE proves to be efficient in clustering high-dimensional data, human inference is required at each step to interpret the results. This exploratory study provides a framework and ideas on how to tailor a classification scheme toward specific spectral data and science questions.

Cloud model inversions of strong chromospheric absorption lines using principal component analysis

“…Indeed, the data processing pipeline was successfully tested on several VTT datasets. presented high-resolution Hα spectroscopy of active region NOAA 12722, investigating the temporal evolution of a pore, whereas Verma, Matijevič, et al (2019) use machine learning and statistical techniques to classify Hα spectra according to results of CM inversions. The collaborative research environment and GREGOR archive 1 at AIP provides the solar physics community already with access to GFPI and HiFI data.…”

Machine learning in solar physics

“…t-SNE can reliably distinguish between profiles associated with flaring regions and non-flaring regions. Additionally, it has been used by (Verma et al 2021) for classifying Ha profiles and identifying those that are suitable for a simple inversion method based on the cloud model. Both works demonstrate that t-SNE is promising for understanding the general picture of large observations.…”

1.5D non-LTE spectral synthesis of a 3D filament and prominence simulation

“…Heinzel et al 2014;Labrosse & Rodger 2016;Levens & Labrosse 2019). Even for such idealised stratified atmospheres, there is degeneracy when inverting from only the shape of the associated spectra, and this hampers the use of more advanced models (current efforts to minimise this are expanding to include t-distributed stochastic neighbouring, Verma et al (2021), and principle component analysis, Dineva et al 2020). Gunár & Mackay (2015) took a more consistent approach by constructing model threads under the magnetohydrostatic (MHS) assumption according to a magnetic arcade topology derived from non-linear force-free field (NLFFF) extrapolations Gunár et al , 2018Gunár et al , 2019.…”