A group of researchers is using artificial intelligence technology to calibrate some NASA solar images to help improve the data scientists use for solar research. This new technology was published in the journal Astronomy and Astrophysics on April 13, 2021. The work of the
Solar Telescope is difficult. Staring at the sun can cause severe losses, with
continuous streams of solar particles and intense bombardments of sunlight. Over time, the sensitive lenses and sensors of the solar telescope began to degrade. To ensure that the data sent by these instruments remains accurate, scientists regularly recalibrate to ensure that they understand the changes in the instruments.
NASA’s Solar Dynamics Observatory (SDO) was launched in 2010 and has been providing high-resolution images of the sun for more than a decade. Their images have given scientists a detailed understanding of various solar phenomena that may trigger space weather and affect our astronauts and the earth and space technology. The Atmospheric Imaging Assembly (AIA) is one of the two imaging instruments in SDO. It constantly looks at the sun and takes images with 10 wavelengths of ultraviolet light every 12 seconds. This creates a lot of information from the sun, but like all Sunstaring instruments, AIA degrades over time, so the data must be calibrated frequently.
Seven of the ultraviolet wavelengths observed by AIA at NASA’s SDO. The top row is taken from May 2010 and the bottom row shows the 2019 data, without corrections, showing how the instrument has degraded over time.
This image shows seven ultraviolet wavelengths observed by the atmospheric imaging component at NASA’s Solar Dynamics Observatory. The top row shows the results of the observation in May 2010, and the bottom row shows the results of the observation in 2019, without any correction, showing how the instrument has degraded over time.
Credit: Luiz Dos Santos / NASA GSFC
Since the launch of SDO, scientists have used sounding rockets to calibrate the AIA. Sounding rockets are smaller rockets that generally carry only a few instruments and fly short distances into space. The
generally only takes 15 minutes. Fundamentally, the ringing rocket flies over most of Earth’s atmosphere so that instruments on board can see the ultraviolet wavelengths measured by AIA. Light of these wavelengths is absorbed by the Earth’s atmosphere and cannot be measured from the ground. To calibrate the AIA, they connected the ultraviolet telescope to the probe rocket and compared the data with the results of the AIA measurement. The scientist can then adjust any changes to the AIA data.
handlebeforeafter
only shows the left side only shows the right side
This slider shows the sun seen by AIA under 304 angstroms in 2021, before degradation correction (left) and sonic rocket calibration (right).
Credit: NASA GSFC
The sonic rocket calibration method has some disadvantages. Ringing rockets can only be launched often, but AIA has been watching the sun. This means that there will be a period of downtime between calibrations of each ringing rocket, and there will be a slight deviation in calibration.
“This is also important for deep space missions where sonic rockets cannot be calibrated,” Dr. Luis dos Santos,
heliophysicist and lead author of NASA’s Goddard Space Flight Center in Greenbelt, Maryland Say. “We are solving two problems at the same time.”
Virtual Calibration
Considering these challenges, the scientists decided to find other options to calibrate the instrument in order to perform continuous calibration. Machine learning is a technique used in artificial intelligence, and it seems very appropriate.
As the name suggests, machine learning requires computer programs or algorithms to learn how to perform its tasks.
First, researchers need to train machine learning algorithms to recognize the structure of the sun and how to use AIA data for comparison. To this end, they provided an algorithm image of the calibration flight of the sonic rocket and told it the correct calibration amount they needed. After enough of these examples, they provided similar images to the algorithm to see if it could recognize the correct calibration required. With enough data, the algorithm will learn how much calibration is needed for each image.
Because AIA observes the sun in multiple wavelengths of light, researchers can also use the algorithm to compare specific structures at different wavelengths and enhance their evaluation.
First, they will show the algorithm all solar flares of the AIA wavelength until it recognizes all solar flares in different types of light, so as to teach the algorithm to understand what the solar flares look like. Once the program can identify solar flares without any degradation, the algorithm can determine the degree of degradation that affects the current AIA image and how much calibration is required for each image.
“This is the most important thing,” Dos Santos said. “We don’t just identify it at the same wavelength, but identify the structure along the wavelength.”
This means that researchers can have more confidence in the calibration identified by the algorithm. In fact, when comparing its virtual calibration data with sonic rocket calibration data, the machine learning program is correct.
Two rows of sun images. The top row becomes darker and harder to see, while the bottom row is still a bright and consistent visible image. The top row of
shows the degradation of the AIA 304 Angstrom wavelength channel over the years since the introduction of SDO. A machine learning algorithm is used to correct the bottom image row for this degradation.
Credit: Luiz Dos Santos / NASA GSFC
Through this new process, researchers are ready to continuously calibrate AIA images between calibration rocket flights to improve the accuracy of researchers’ SDO data.
Machine Learning Beyond the Sun
Researchers have also been using machine learning to better understand situations closer to home.
A team of researchers led by Dr. Ryan McGranaghan, Chief Data Scientist and Aerospace Engineer of ASTRA LLC and NASA Goddard Space Flight Center
Use machine learning to better understand this connection

By Peter

Leave a Reply

Your email address will not be published. Required fields are marked *