Researchers from Aryabhatta Research Institute of Observational Sciences (ARIES), an autonomous institute under the Department of Science and Technology, along with collaborators from Indian Institute of Space Science and Technology, Southwest Research Institute, and Indian Institute of Astrophysics, have successfully applied artificial intelligence to analyze 100 years of hand-drawn solar observations from the Kodaikanal Solar Observatory. The study, published in the Astrophysical Journal, used a supervised machine learning approach (U-Net) to digitize daily 'suncharts' recorded from 1904 to 2022, specifically focusing on the period from 1916 to 2007 covering nine solar cycles (Solar Cycles 15-23).
The research addressed the challenge of inconsistent historical records by employing AI in two key steps: first automatically identifying the Sun's disk in each scanned drawing to pinpoint center, size, and tilt for accurate feature placement, and then identifying and tracing plages (magnetically active patches on the Sun) across the drawings. The resulting data enabled the creation of time-latitude 'butterfly' maps showing plage area distribution measured in 1° latitude bands for each day, revealing how magnetic activity shifts with latitude and solar-cycle phase.
The study demonstrated that plage areas derived from hand-drawn suncharts match well with those from KoSO's Ca II K full-disk observations, proving that historical drawings can help fill gaps and improve long-term solar data consistency. This approach allows scientists to connect modern space-age measurements with historical solar activity patterns, which is crucial for understanding how solar cycles vary in strength and structure, improving reconstructions of the Sun's energy output and magnetic influence changes over time, and better assessing long-term space weather risks that can affect Earth's technology systems.