In October 2025 SpaceX launched its 10-thousandth satellite into low Earth orbit, putting the number of satellites overhead at about 15-thousand. If all launch filings currently lodged with the US Federal Communications Commission head into space, there will be half a million artificial satellites orbiting our planet within 15 years.
Not only is this proliferation of space real estate fast increasing the risk of space collisions, the light and noise pollution that satellites emit may put the very future of ground-based astronomy at risk.
Impacts on astronomy across the wave spectrum
For optical astronomy carried out by observatories such as the European Southern Observatory’s Very Large Telescope the challenge is light pollution. Most efforts to reduce this have centered around dimming satellites through changes in surface design.
It’s a different story in radio astronomy, a discipline that captures faint, non-visible radio waves, allowing scientists to observe phenomena like pulsars, quasars, and the cosmic microwave background. Today’s modern satellite constellations transmit powerful radio signals back to Earth which are vastly stronger than the radio waves from distant galaxies that astronomers are trying to detect.
Even designated radio quiet zones can’t currently solve this growing problem because, while terrestrial transmitters can be restricted within a certain radius of observatories, satellites pass directly overhead and their transmissions can’t be geographically fenced off.
“This rapidly increasing radio frequency interference or RFI from above is like trying to look at the stars with a lamp post right in front of you,” explains SKACH’s Dr Chris Finlay, a High Performance Computing specialist at EPFL. “The light from the lamp post is so bright you can’t see the stars behind it.”
AI to the rescue?
Now, a new data-analysis algorithm, TABASCAL – short for Trajectory-Based Subtraction and Calibration – is offering radio astronomers hope in the face of this rapidly intensifying challenge.
Developed by Finlay and colleagues, TABASCAL first emerged almost a decade ago in response to an earlier satellite interference problem. Before today’s mega-constellations began filling the sky, astronomers working for South Africa’s Square Kilometre Array precursor programs were already losing 30 to 40 percent of their data in certain frequency bands.
In particular, the L-band, a range critical to radio astronomy, overlaps with signals from Global Navigation Satellite Systems such as GPS, Europe’s Galileo, and China’s BeiDou that emit continuously, contaminating astronomical data. At the time, the only available strategy was to identify corrupted data and throw it away.
“It was a high-impact problem, and we were simply flagging contaminated data and deleting it because we couldn’t model the signal well enough to subtract it,” Finlay recalls. “TABASCAL takes a radically different approach.”
Instead of discarding data, the TABASCAL algorithm models astronomical signals and the interfering satellite signals simultaneously. By incorporating information about a satellite’s trajectory and behavior, it predicts how the interference should appear in the telescope’s measurements. It then subtracts that modeled interference, ideally leaving behind only the true cosmic signal.
The approach is Bayesian – rooted in probabilistic modelling, treating both astronomical sources and interference as components of a statistical model, allowing uncertainties to be rigorously quantified.
From Simulation to Reality
After seven years of development, TABASCAL is now transitioning from simulations to real-world data. Researchers are testing it on observations from the Engineering Development Array 2 (EDA2), a 256-antenna prototype station in Australia associated with the SKA’s low-frequency array.
So far, results are promising. In extreme test cases, interference thousands of times stronger than the brightest astrophysical sources has been modeled and significantly reduced.
Yet challenges remain. Radio interferometry at SKA scales uses huge computational resources so TABASCAL needs to be optimized to run efficiently across computing clusters. Over the next two years, the team plans to refine documentation, improve scalability and validate performance limits.
“Modeling radio frequency interference forces you back to first principles,” Finlay says. “It’s one of the hardest problems in interferometry so TABASCAL is quite a big step forward.”
Protecting Dark and Quiet Skies
TABASCAL’s development comes at a time of growing global concern about preserving Earth’s “dark and quiet skies” with the International Astronomical Union (IAU) leading a global advocacy effort to protect astronomical observations from satellite constellations.
Through its Centre for the Protection of the Dark and Quiet Sky from Satellite Constellation Interference – co-hosted with the SKAO – the IAU works with policymakers, regulators and satellite operators to balance technological progress with scientific preservation, yet potential regulation is complicated and may create financial and political tensions.
Reducing unintended “out-of-band” radiation to levels safe for astronomy would require more advanced, and more expensive, satellite engineering. Satellite companies face cost pressures and global demand for connectivity. At the same time, astronomers need to protect multibillion-dollar scientific investments and fear that unless the challenge of RFI, and satellite reflections, is overcome, the pathways to scientific discovery with future ground-based astronomy facilities may become far more challenging than anticipated.
For now, Finlay is focusing on the task at hand – ensuring TABASCAL is battle ready when the SKAO becomes fully operational in 2029.
“I just love interferometry. I love the problem. I love the challenge. I always wanted to work on something that is applied and that could have a high impact. The fact that this problem has remained so open is amazing to me and indicates how challenging it is. When I first started as a master’s student my supervisor and I thought, ‘cool, let’s do something big’. And seven years later, here we are!”