Throngs of satellites are streaking across the Hubble Space Telescope’s field of view, leaving what look like scratch marks on space photos and hindering scientists’ work. Teeming swarms of these satellites, which reflect sunlight and mimic astronomical objects, threaten to gradually transform the night sky and affect how astronomy can be done.
“We see these satellite trails in Hubble data, and really in all astronomical data, and they’re a bit of a nuisance,” said David Stark, an astronomer at the Space Telescope Science Institute in Baltimore, speaking last week at the American Astronomical Society conference in Albuquerque, New Mexico. In fact, he said, his team used a new detection method to measure that the rate of satellite trails is doubling. But Stark was presenting his team’s idea for a Band-Aid fix: new software they described in a recent report that is five to 10 times more sensitive at finding trails than previous software, and then masking them out. “It’s particularly good at finding satellite trails that can be missed by eye,” he said.
It’s standard procedure for astronomers to try to clean images of “artifacts,” like the effects of cosmic rays hitting Hubble’s camera detectors, or diffraction spikes, which make bright stars look like crosses. Occasionally, a pesky nearby Milky Way star might get in the way of one’s view of a faraway object. The new technique, called Median Radon Transform, examines every linear path across an image at every possible angle. When a particular path aligns with a satellite trail, the system notes a deviation from the average flux—or the brightness at a particular wavelength in a pixel—measured across areas with apparently empty sky. It can spot short streaks too, but they have to be slightly brighter to be identified, since they cover fewer pixels.
This software then allows astronomers to mask out satellite trails, so that the affected pixels are ignored in their analysis of the data. It’s like dog-earing a few misprinted pages in a book, so that you skip them while studying the rest of the volume.
But it’s better not to lose those pages. When there are multiple exposures of the same field, an astronomer can use additional software tools to completely remove the line from the final combined image. That part of the sky will then appear as it should, though the signal-to-noise ratio in that line’s pixels will be lower than if the satellite had never skimmed in front of the telescope that day. Stark and his team have included their code in a standard software pack called “acstools” that they manage.
Still, this fix has a big limitation: It’s designed for Hubble, which is orbiting 332 miles above the Earth, and it’s less plagued by satellite streaks than observatories on the ground. Ground-based optical telescopes with wide-field imaging—which often don’t take multiple exposures—stand to be far more affected. There have been a few instances already of satellites photobombing images taken by telescopes at the Cerro Tololo Inter-American Observatory in Chile and the Lowell Observatory in Arizona, for example.
The problem will be much, much worse for the long-awaited National Science Foundation–funded Vera Rubin Observatory, which is being assembled in the Chilean Andes and will begin taking images next year. Its incredibly sensitive camera will detect faint and changing objects, like a star going supernova or a near-Earth asteroid, and the telescope will automatically send alerts to astronomers when it spots such things. But the Rubin collaboration has expressed concern about the possibility of false alerts thanks to light reflecting off of satellites or space junk in orbit and has warned that as many as 30 percent of its images could be affected by satellite streaks. For example, a glint of sunlight off a small piece of insulation shed by a satellite could appear in a telescope image like a flaring star. Unless an astronomer can also measure the light’s spectrum, they might be fooled, says John Barentine, a Tucson, Arizona, astronomer who recently authored a study about light pollution from low Earth orbit objects.
A second problem is that the number of satellite streaks is growing. Stark tested the software on 20 years of data from Hubble’s Advanced Camera for Surveys. While the satellite trails haven’t changed in brightness, their rate has approximately doubled. The team found trails in every three to four hours of Hubble data taken in 2002. But in 2022, satellites photobombed Hubble every one to two hours. That means 5 percent of the images taken 20 years ago were affected, and now it’s about 10 percent.
The rate will surely continue rising, says Sandor Kruk, an astronomer at the Max Planck Institute for Extraterrestrial Physics in Munich, Germany. “You expect, as time goes, more streaks in images. It goes proportionately with the number of satellites you have above the observatory,” says Sandor Kruk, lead author of a recent study that used crowdsourced classifications and machine learning to track satellite trails in images.
Kruk and his colleagues observed a lower rate, rising from about 2.5 to 5 percent over the past two decades. They found that this trend ramped up rapidly starting in about 2018, around the same time companies began deploying satellite mega-constellations, connecting hundreds or thousands of them in networks. (Stark and Kruk attribute their studies’ different percentages to using different measuring techniques.)
Those mega-constellations have obvious benefits for their operators. The satellites are smaller and therefore cheaper to produce and launch, and networked services are less vulnerable to disruption, such as by space weather or anti-satellite weapons. SpaceX’s Starlink makes up the biggest satellite network by far, with some 4,000 in orbit and plans to increase that to 42,000. OneWeb’s constellation has more than 600 satellites, but they’re situated at a higher orbit, lessening the effect on astronomical observations. And Amazon is poised to launch its Project Kuiper this summer, lofting its first broadband-service-providing satellites on the inaugural flight of the United Launch Alliance’s Vulcan Centaur rocket. The company plans to populate that constellation with more than 3,000 satellites.
SpaceX and some other companies have tested possible solutions, like covering a satellite with a thin film to darken it so it reflects less light, or adding a visor to reflect light away from Earth. These limited efforts have fallen far short of the International Astronomical Union’s brightness targets, and some of these designs have caused problems for the satellites themselves, by heating them too much or interfering with inter-satellite communication.
Hubble’s deployment by the Space Shuttle Discovery in 1990.
Photograph: NASA
A NASA concept to work with its commercial partners to boost Hubble to a higher orbit might inadvertently mitigate the photobombing problem. Atmospheric drag has gradually pulled the spacecraft closer to Earth. Nudging it back is intended to extend its lifespan—but would also move it away from a small fraction of passing satellites.
None of that will resolve problems for Earth-based observatories, which have to peer through the whole atmosphere, including all satellite orbits. And Barentine worries that although companies haven’t found any tech fixes yet, they haven’t slowed down the brisk pace of satellite launch. “People in industry have this unerring faith in innovation,” he says, “and my response to that has been: The history of science and technology and the environment is just littered with instances of running headlong into technology we did not understand that created a lot of negative side effects.”
