Researchers use AI to help improve NASA's SDO data calibration work

2024-03-12

According to foreign media reports, a group of researchers are using AI technology to calibrate some of NASA's solar images to help improve the data scientists use for solar research. This new technology was published in Astronomy&Astrophysics on April 13, 2021. The work of a solar telescope is quite arduous, and staring at the sun comes at a painful cost. It faces an endless stream of solar particles and intense sunlight.

As time goes by, the sensitive lenses and sensors of the solar telescope begin to degrade. To ensure that the data sent back by these instruments is still accurate, scientists need to regularly recalibrate to ensure they understand how the instruments are changing.

NASA's Solar Dynamics Observatory (SDO) was launched in 2010 and has been providing high-definition images of the Sun for over a decade. Its images allow scientists to observe various solar phenomena in detail, which can trigger space weather and affect our astronauts and technology on Earth and in space. The Atmospheric Image Assembly (AIA) is one of the two imaging instruments on SDO, which continuously observes the sun and captures images every 12 seconds through 10 wavelengths of ultraviolet light. This creates rich solar information, but like all instruments that gaze at the sun, AIA will degrade over time and data needs to be calibrated regularly.

Since the launch of SDO, scientists have used sounding rockets to calibrate AIA. Sounding rocket is a small rocket that carries only a small amount of instruments and conducts a brief space flight - usually only 15 minutes. It is crucial that the sounding rocket flies over most of the Earth's atmosphere so that the instruments above can see the ultraviolet wavelength measured by AIA. These wavelengths of light are absorbed by the Earth's atmosphere and cannot be measured from the ground. To calibrate AIA, scientists will install a UV telescope on a sounding rocket and compare this data with AIA's measurement data. Then, scientists can adjust for any changes in AIA data.

However, the calibration method for sounding rockets has some shortcomings. Although it can be launched many times, AIA is constantly observing the sun, which means there is a calibration time between each sounding rocket calibration.

"Thisalso very important for deep space missions because there is no option for rocket calibration," Luiz Dos Santos pointed out. "We are solving two problems simultaneously." Santos is a solar physicist at NASA's Goddard Space Flight Center and the lead author of this paper.

Virtual calibration

Considering these challenges, scientists have decided to look for other methods to calibrate instruments and focus on continuous calibration. The machine learning techniques used in AI seem to be a better choice.

Machine learning, as the name suggests, requires a computer program or algorithm to learn how to perform its tasks.

Firstly, researchers need to train a machine learning algorithm to recognize solar structures and how to use AIA data for comparison. To achieve this, they provide the algorithm with images from the calibration flight of the detection rocket and tell them the correct calibration amount needed. After these examples are sufficient, they provide similar images to the algorithm to see if it can recognize the correct calibration required. With sufficient data, the algorithm can identify how much calibration is required for each image.

Due to AIA observing the sun in multiple wavelengths of light, researchers can also use this algorithm to compare specific structures of different wavelengths and enhance their evaluation.

Researchers teach algorithms to determine what a solar flare looks like by displaying solar flares of all wavelengths on AIA, and continue until it identifies all types of solar flares of light. Once the program can identify solar flares without any degradation, the algorithm can determine how much degradation affects the current image of AIA and how much calibration is needed for each image.

"Thisa big deal," Dos Santos said. "We don't just recognize it at the same wavelength, but recognize the structure at different wavelengths."

This means that researchers can be more certain about the calibration of algorithm recognition. In fact, when comparing their virtual calibration data with sounding rocket calibration data, the machine learning program is correct.

With this new process, researchers can continuously calibrate AIA images and improve the accuracy of SDO data between rocket flights.

Machine learning beyond the sun

In fact, researchers have been using machine learning to better understand situations closer to home.

A group of researchers led by Dr. Ryan McGranaghan, Chief Data Scientist and Aerospace Engineer at ASTRA LLC and NASA Goddard Space Flight Center, used machine learning to better understand the connection between the Earth's magnetic field and the ionosphere. The ionosphere is the charged part of the Earth's upper atmosphere. By using data science techniques on a large amount of data, they can apply machine learning techniques to develop an updated model to help them better understand how charged particles from space enter the Earth's atmosphere, where they drive space weather.

With the progress of machine learning, its scientific applications will be extended to more and more tasks. In the future, this may mean that deep space missions - going to places where parallel rocket flight cannot be calibrated - can still be calibrated and continue to provide accurate data, even at increasingly distant locations from Earth or any star.

Scientists develop futuristic biomimetic arms using micro robots to restore natural reflexes in ampu