BENGALURU: Artificial intelligence can help scientists uncover a century of the sun’s history by analyzing handwritten observations from the Kodaikanal Solar Observatory, turning fragile paper records into a valuable digital archive and improving our understanding of long-term solar activity and future space weather.In the study published in The Astrophysical Journal, researchers led by Divya Kirti Mishra from the Aryabhata Institute of Observational Sciences (Aries), along with collaborators from Isro’s Indian Institute of Space Science and Technology (IIST), the Southwest Research Institute in Boulder, USA, and the Indian Institute of Astrophysics (IIA), used machine learning to analyze daily solar maps produced at the Kodaikanal Solar Observatory (KoSO). 1916 and 2007.This study shows how historical observations, once considered difficult to analyze systematically, can now be transformed into reliable scientific data. Aries and IIA are autonomous institutions of the Department of Science and Technology (DST).For more than a century, scientists have studied the Sun’s magnetic activity, which waxes and wanes on a roughly 11-year cycle. These cycles can cause sunspots, solar flares, and eruptions that can interfere with Earth’s satellites, navigation systems, radio communications, and power grids. Modern telescopes provide accurate digital observations, but records from earlier decades are often incomplete or inconsistent, making it difficult to understand how the Sun behaved over long periods of time.One of the world’s oldest solar observatories, KoSO maintains an excellent archive of daily “Suncharts” from 1904 to 2022. “Before digital images became commonplace, astronomers carefully sketched features such as sunspots, plages, filaments, and prominences on a standard grid. “While of scientific value, differences in drawing styles, paper deterioration, and variations in scan quality make these records difficult to analyze using traditional methods,” DST said.To overcome these hurdles, researchers employed a supervised machine learning model known as “U-Net.” The system first identified the sun’s disk in each scanned drawing and precisely determined its center, size, and orientation. They then detected and mapped bright magnetic regions known as plages over nine solar cycles from 1916 to 2007.Plages are bright spots in the Sun’s atmosphere associated with strong magnetic fields. Because they closely reflect the Sun’s magnetic activity, scientists consider them one of the most reliable indicators of long-term changes. Extracting them from the historical record helps bridge the gap between early observations and ongoing measurements available in the space age.The AI-generated data allowed the team to construct the classic “butterfly diagram.” This shows how solar activity moves from high latitudes toward the equator during each solar cycle. “They also found that the plage regions identified from the drawings matched well with measurements obtained from KoSO’s Ca II K full-disk images, confirming that the 100-year-old sketches provide reliable scientific information,” DST said.The researchers say such long-term records are essential for comparing the strength and structure of different solar cycles, reconstructing past changes in the Sun’s magnetic influence, and improving models of long-term space weather.
