Abstract Details

Name: Aditya Priyadarshi
Affiliation: Aryabhatta Research Institute of Observational Sciences 263002
Conference ID: ASI2021_177
Title : Implementation of machine learning for analysis and extraction of solar filaments from the hand-drawn KoSO data
Authors and Co-Authors : Aditya Priyadarshi, Aryabhatta Research Institute of Observational-Sciences, Nainital, India Subhamoy Chatterjee, Southwest Research Institute, Boulder, CO, United States Manjunath Hegde, Aryabhatta Research Institute of Observational-Sciences, Nainital, India Bibhuti Kumar Jha, Aryabhatta Research Institute of Observational-Sciences, Nainital, India Sudip Mandal, Max Planck Institute for Solar System Research, Katlenburg-Lindau, Germany Dipankar Banerjee, Aryabhatta Research Institute of Observational-Sciences, Nainital, India and Indian Institute Astrophysics, Bangalore, India
Abstract Type : Oral
Abstract Category : Sun and the Solar System
Abstract : Kodaikanal Solar Observatory (KoSO) is an abundant source of solar archival data. It includes observation taken at different wavelengths such as in white light, Ca II K, and Hα beginning from 1905 to till date. These observations are taken in photographic film/plates. In addition to that, KoSO has stored so-called ‘suncharts’ (hand-drawn solar observations) that constitute composite data from all wavelengths. These drawings are made on Stonyhurst latitude and longitude grid, and all solar features are marked on it. For our analysis, we have used data for a period of 23 years (1954-1976). These suncharts are digitized using industry level scanner and are stored in digital format (‘.tif’). We started with filaments and implemented the detection algorithm. Filaments are solar features that are formed along polarity inversion lines. Here, we have extracted these features (generally marked in red color) using an unsupervised machine learning algorithm known as the k-means clustering technique. Apart from this, the solar disc, with its corresponding center and radius, are extracted from the suncharts. Binary maps of extracted filaments were further used to form the Carrington maps, to have a detailed comparison with previously generated Kodaikanal Halpha Carrington maps from plates. They had a remarkable match, and the results were quite promising. Further, we will compare these maps with the Meudon database. Filament parameters like tilt angle, length, perimeter, and area were deducted for further analysis. Butterfly diagram is confirmation of our detection technique. A clear sign of a polar rush is observed in the butterfly diagram. The study shows a clear increase in filament length and area, as we move higher in latitude. Our next step would be to go for other solar features.