Computers can now play an important role for space weather alerts using a new machine-learning technique. The technique developed by a team of scientists at the National Centers for Environmental Information and the University of Colorado Boulder explores vast amount of satellite data to throw light on conditions significant for space weather.
Meanwhile, programming of computers to find solar flares in large streams of images is the basis of the technique. Therefore, it helps scientists to announce timely space weather alerts, says the research published in the Journal of Space Weather and Space Climate.
Characteristically, changing conditions in the space and on the sun can impact technologies on Earth. Such changes can impact radio communications, damage power grids, and diminish accuracy of navigation system. This necessitates processing of solar data in real-time, as flares breaking out on the sun impacts Earth in minutes.
New Algorithm addresses shortcomings of Predecessor Techniques
At present, the technique involves summarizing conditions on the sun two times every day to predict approaching space weather. This includes use of hand-drawn maps label marked with various solar features. The marking include active regions, coronal hole boundaries, and filaments.
However, solar imagers release a new set of data every few minutes. The Solar Ultraviolet Imager, for example positioned on GOES-R Series satellite programed for a four-minute cycle. Precisely, each cycle collects data in six different wavelengths.
This requires tools to process solar data and make it utilizable. To address this, lead researcher of the study developed a computer algorithm. The algorithm can view images released by Solar Ultraviolet Imager simultaneously and identify patterns in the data.
Meanwhile, the development of algorithm and associated techniques are extension of a NSF summer research.