In the realm of solar research, a team of scientists from the Indian Institute of Astrophysics, led by Dibya Kirti Mishra, has made significant strides in leveraging machine learning to analyze historical solar data. Their work focuses on the Kodaikanal Solar Observatory (KoSO), one of the oldest solar observatories, which has amassed a vast archive of solar observations spanning over a century.
The researchers have developed a machine learning model using Convolutional Neural Networks (CNNs) to identify solar disks and plages from hand-drawn suncharts at KoSO. These suncharts, dating from 1904 to 2022, contain detailed markings of various solar features such as sunspots, plages, filaments, and prominences. The team’s primary goal was to digitize and analyze this extensive dataset to fill gaps in the observatory’s multi-wavelength observations.
The study involved training the CNN model with manually identified solar disks and plages from the suncharts. The model first detects the solar limb and the North-South line, enabling the extraction of disk center coordinates, radius, and P-angle. Subsequently, the model performs accurate image segmentation to identify plages. The researchers compared the plage areas derived from the suncharts with those obtained from Ca II K full-disk observations and found a good agreement, validating the effectiveness of their machine learning approach.
This research, published in the journal Astronomy & Astrophysics, demonstrates the potential of machine learning techniques to analyze historical solar data. By filling in the gaps in KoSO’s multi-wavelength observations, this work contributes to constructing a comprehensive composite series of solar observations over the last century. For the energy sector, particularly solar energy, understanding long-term solar activity patterns is crucial for predicting solar output and improving the efficiency and reliability of solar power systems. This research provides a valuable tool for enhancing our understanding of solar behavior and its implications for solar energy generation.
This article is based on research available at arXiv.