Artificial intelligence helps improve NASA studies in the sun and beyond

A group of researchers is using artificial intelligence techniques to calibrate some of the images of the Sun obtained by NASA, to improve the data that scientists use in solar research. Esta nueva técnica se publicó en la revista Astronomy & Astrophysics el 13 de abril de 2021.

The work of a solar telescope is difficult.Observe the sun is complicated, since from it an endless current of solar particles and intense sunlight emanates.Therefore, over time, highly sensitive lenses and solar telescope sensors begin to degrade.To ensure that the data that these instruments send us remain precise, scientists must periodically emphasize to adapt to the changes in the instrument.

NASA's Solar Dynamics Observatory, or SDO, has provided high -definition images for more than a decade since its launch in 2010.His images have provided scientists with a detailed vision of several solar phenomena that can modify the spatial climate and affect astronauts and technology on earth and in space.Atmospheric Imagery Assembly, or AIA, is one of the two image instruments in SDO that constantly points to the sun, taking images in 10 wavelengths of ultraviolet light every 12 seconds.This creates a large amount of exceptional sun information, but like all the instruments that observe the sun, the AIA degrades over time and the data must frequently calibrate.

Since SDO launch, scientists have used calibration rockets to calibrate the AIA.These rockets are small and usually carry few instruments, make short flights to space, usually only 15 minutes.Fundamentally, what they do is fly above the atmosphere of the Earth, which allows the instruments on board to detect the ultraviolet wavelengths that measures the AIA.As these light wavelengths are absorbed by the earth's atmosphere, they cannot be measured from the ground.To calibrate the AIA, they connect an ultraviolet telescope to a calibration rocket and compare that data with the AIA measurements.Then, scientists make adjustments taking into account any change in AIA data.

There are some inconveniences in the calibration rocket method.These rockets can only be launched very occasionally, although the AIA is constantly looking in the sun.Therefore, there is a time of inactivity in which the calibration is slightly diverted until a calibration rocket again allows adequate calibration.

La inteligencia artificial ayuda a mejorar los estudios de la NASA en el Sol y más allá

"It is also important in missions in deep space, which do not have the option of using calibration rockets," said DR.Luiz dos Santos, solar physicist of the Goddard Space Flight Center of NASA in Greenbelt, Maryland, and main author of the article."We are addressing two problems at the same time".

Calibración virtual

With these challenges in mind, scientists decided to look for other options to calibrate the instrument, with the aim of achieving constant calibration.Automatic learning, a technique used in artificial intelligence, seemed to fit perfectly.

As the name implies, automatic learning requires a computer program, or algorithm, which learns to perform your task.

First, researchers needed to train an automatic learning algorithm to recognize solar structures and how to compare them using AIA data.To do this, give the algorithm images of calibration rocket flights and tell him the correct amount of calibration they need.After many of these examples, they give similar images to the algorithm and see if it would identify the necessary correct calibration.With enough data, the algorithm learns to identify how much calibration is needed for each image.

Because the AIA observes the sun in multiple wavelengths of light, researchers can also use the algorithm to compare structures in specific wavelengths and strengthen their evaluations.

To begin with, they would show the algorithm what a solar flare was showing solar flares in all aia wavelengths until the solar flares in all different types of light.Once the program can recognize a solar flare without any degradation, the algorithm can determine how much degradation is affecting the images of the current AIA and how much calibration is needed for each.

"This was the most important thing," said Dos Santos."Instead of simply identifying it in the same wavelength, we are identifying structures along the wavelengths".

This means that researchers can be safer of calibration that identified the algorithm.In fact, when comparing the virtual calibration data with the data of the calibration rockets, the automatic learning program was right.

With this new process, researchers are prepared to constantly calibrate the images of the AIA between calibration rocket flights, improving the accuracy of SDO data for researchers.

Aprendizaje automático más allá del Sol

Researchers have also been using automatic learning to better understand the closest conditions.

A group of researchers led by DR.Ryan McGranaghan, main data scientist and Airospace Engineer from Astra LLC and the Goddard Space Flight Center of NASA, used automatic learning to better understand the connection between the magnetic field of the Earth and the ionosphere (the upper part of the atmosphere ofthe earth electrically loaded).By using techniques for large volumes of scientific data, they could apply automatic learning techniques to develop new models that help them better understand how space energized particles travel to the earth's atmosphere, where they permeate the space climate.

As automatic learning progresses, your scientific applications will expand to more and more missions.For the future, this can mean that missions in deep space, which travel to places where calibration rocket flights are not possible, can be calibrate and continue to provide precise data, even gradually moulating from the earth or any star from any star.

English version of this news.

Edition: R..Castro.