Space Weather by the Seasons express
The Sun doesn't burn steadily: it erupts in series. Like a campfire that dies down and then suddenly throws out showers of sparks, these "fiery seasons" last from six months to a year and a half, and during them the star shows an unusually high number of sunspots. A team of researchers took sunspot data going back to 1878 and taught a computer to recognize when the next fiery outburst is approaching.
Instead of looking at the Sun as a whole, they split it into north and south. Each hemisphere follows its own rhythm: the north peaked at about 70 spots in March 2025, while the south went wilder — up to 130 spots at nearly the same time. The difference was captured using a mathematical model similar to predicting seasonal sales in a store: it accounts for both smooth trends and sharp jumps.
Then came the "forest predictor"—the random forest algorithm, which learns from examples. It was fed hundreds of past surges and learned to guess how strong the next one would be. This hybrid of statistics and machine learning already correctly pointed to the rise in activity in late 2024 and early 2025. The next flare-up is expected in December 2025, with the southern hemisphere again setting the tone. The overall strength of the storms depends mostly on it.
Original abstract
as written by the authors arXiv:2605.27057
We present a hybrid forecasting strategy that combines numerical modeling, statistical forecasting, and machine learning methods to predict enhanced bursts of solar activity. These bursts, referred to here as space weather seasons, occur on intermediate timescales of approximately 6 to 18 months. We analyze monthly smoothed sunspot number data from 1878 to 2025 and use Gaussian fitting to identify burst events and characterize their amplitudes and durations. The sunspot number data are divided into training, testing, and forecasting intervals to evaluate both hindcast performance and future predictions. Each solar hemisphere is modeled separately using a seasonal autoregressive integrated moving average approach. This baseline forecast is refined using an asymmetric Gaussian override to represent the rapid rise and gradual decay of burst activity. Burst amplitudes and durations are then estimated using a random forest regression model. The hybrid method successfully reproduces burst timing between November 2024 and May 2025, with the Northern Hemisphere reaching a peak sunspot number of about 70 around March 2025. The next Northern Hemisphere burst is forecast to occur around December 2025, with a slightly lower peak sunspot number of about 60. In contrast, the Southern Hemisphere shows more complex behavior, with multiple burst amplitudes beginning around October 2024 and continuing until October 2025. The dominant Southern Hemisphere burst reaches an amplitude of about 130 in sunspot number, and the next burst is forecast to occur around December 2025. When the hemispheric forecasts are combined, the total sunspot number is found to be mainly controlled by the stronger activity cycle in the Southern Hemisphere.
A sunspot the size of a small city can trigger a magnetic storm on Earth that knocks out navigation systems.