Using Machine Learning Technique To Process Solar Data in Real-Time

A team of scientists has developed a machine learning technique to improve space weather forecasts and provide a better understanding of solar data.

Flares erupting from the Sun can impact the Earth within minutes.

For example, a massive solar storm barely missed Earth back in July 2012. According to Pete Riley of Predictive Science, there’s a 12 percent chance of such an event occurring.

As such, real-time processing of solar data is essential to stay on top of the event. However, that has never been easy.

Currently, forecasters summarize conditions on the Sun twice a day to predict incoming space weather. As you can imagine, this entails collecting tons of data.

Aside from the hand-drawn maps labeled with solar features, solar imagers produce a new set of observations every few minutes. For example, the Solar Ultraviolet Imager runs on a 4-minute cycle and collects data in six different wavelengths in each cycle.

Keeping up with all the data could be a bit challenging. So, it became necessary to develop a tool that could process solar data into digestible chunks.

The scientists at NOAA‘s National Centers for Environmental Information (NCEI) and CIRES developed a machine learning tool that does just that.

Developing an Algorithm to Process Solar Data in Real-Time

A computer scientist in NCEI, J. Marcus Hughes, developed a computer algorithm for the SUVI images. Not only could the system analyze the images simultaneously, but it could also spot patterns in the data.

Next, Hughes and his colleagues had to create a database of expert-labeled maps of the Sun. Using these images, they trained the system to recognize solar features that are essential for forecasting.

Hughes noted:

“We didn’t tell it how to identify those features, but what to look for — things like flares, coronal holes, bright regions, filaments, and prominences. The computer learns how through the algorithm.”

To identify solar features, the algorithm examines an image one pixel at a time. Then, it decides if a specific pixel is either brighter or dimmer than a certain threshold, before sending it to a category – like a flare.

After undergoing the initial training, the system can classify millions of pixels within seconds. So, forecasters will understand what’s happening on the Sun faster than before.

Also, the algorithm can help scientists evaluate long-term solar data, which in turn could help improve models of the Sun.