A simulation published in 2025 in Nature changes the game for space telescopes. If current satellite constellation projects reach their planned size, more than nine out of ten images taken by certain observatories will bear the trace of a satellite.
The essentials
- In the scenario of 560,000 planned satellites, 39.6% of Hubble images and more than 92% of exposures from SPHEREx, ARRAKIHS, and Xuntian would contain at least one light streak (Borlaff, Marcum & Howell, Nature, 2025).
- Xuntian, the large Chinese space telescope, would undergo an average of 92(+11/−10) streaks per exposure in the scenario of approximately 560,000 satellites.
- The luminous visibility of a streak depends on several physical and geometric parameters, including the satellite-telescope distance, and does not mechanically increase with lower altitude. A lower orbital altitude of the observatory can increase the number of streaks, while longer exposures separately increase the probability that a satellite crosses the field of view.
- As of November 2, 2026, the ESA counts approximately 15,800 functional satellites still in Earth orbit; projects filed with regulators plan to bring this number to 560,000.
- No international framework currently constrains operators to reduce the reflectivity of their satellites or to share orbital data enabling observatories to plan their observations accordingly.
Low Earth orbit, new friction zone between commerce and science
Low Earth orbit, between 200 and 2,000 kilometers altitude, has become the primary deployment zone for broadband satellite constellations over the past decade. SpaceX already operates thousands of Starlink satellites there. Amazon, OneWeb, the Chinese GW consortium, and about twenty other operators have filed authorization requests that, combined, reach 560,000 units. According to ESA statistics updated on November 2, 2026, approximately 15,800 satellites were still functional in Earth orbit across all orbital regions.
This densification is not a problem for ground-based telescopes alone. It directly affects space observatories, whose instruments can observe many directions of the sky, within limits imposed by their pointing constraints, including toward zones where satellites reflect sunlight with measurable intensity. The simulation published on December 3, 2025 by Alejandro S. Borlaff, Pamela M. Marcum, and Steve B. Howell in Nature quantifies this effect for large current and planned space telescopes.
The main result is stark: in the 560,000-satellite scenario, 39.6% of Hubble telescope images would contain at least one streak. For the three wide-field observatories, SPHEREx (a NASA instrument dedicated to infrared cosmology), ARRAKIHS (a European Space Agency project targeting galaxy halos), and Xuntian (the Chinese space telescope under assembly), this rate would exceed 92%. Xuntian would experience an average of 92(+11/−10) streaks per exposure in the scenario of approximately 560,000 satellites, mainly due to its low orbital altitude.
Space telescopes are more vulnerable than previous estimates indicated
Common intuition holds that satellites only hinder ground-based telescopes, which see them crossing the night sky. Space observatories would benefit from a “clean” sky without atmosphere. The study by Borlaff and co-authors shows this reasoning is incomplete.
Three factors explain the vulnerability of orbiting telescopes. First, geometry: a satellite in low Earth orbit and a telescope a few hundred kilometers in altitude cross much more often, and at far more varied angles, than if the observatory were on the ground. Second, exposure times: to detect distant galaxies or minute variations in brightness, instruments like SPHEREx maintain exposures of several hours. The longer the exposure, the greater the probability that a satellite crosses the field of view, mechanically increasing. Finally, field of view: Xuntian and SPHEREx each cover a much larger angular area than Hubble, multiplying opportunities for intersection with satellite orbits.
The combination of these three factors produces a nonlinear effect. Doubling the number of satellites does not double the rate of affected images: in certain configurations, the rate spirals because streaks overlap and become indistinguishable from one another.
Data lost, not merely degraded
One might think image-processing software suffices to correct the problem. On Hubble, algorithms already remove certain streaks by combining multiple exposures of the same field. But this method has two limitations the study highlights.
The first is statistical: contamination of 39.6% of images does not mechanically imply loss of 39.6% of observation time. Removing contaminated images can result in telescope time loss that no one recovers. On an instrument as in-demand as Hubble, whose time is allocated through international competition among researchers, each lost hour reduces a limited observation resource and diminishes scientific yield.
The second limitation concerns wide-field missions. With an increasing number of streaks per exposure, current algorithms struggle to separate signals. Certain more restrictive pointing configurations could reduce the time and sky area observable, particularly for regions near constellation orbital planes. For programs mapping the entire visible sky, as SPHEREx does in infrared, reduced sky coverage represents significant gaps in the data.
This data loss has a direct scientific cost. Programs like SPHEREx aim to map the distribution of galaxies over millions of light-years to constrain dark matter and dark energy models. Incomplete sky coverage limits the statistical reach of measurements, equivalent to being unable to complete a survey because part of the population is hard to access.
A common good without a custodian
Low Earth orbit harbors a paradox economists recognize well: a shared resource exploited by private actors who do not internalize the cost of their presence for other users. Commercial satellites generate value for their operators and clients; the streaks they produce are borne, without compensation, by public space agencies and their scientific programs.
This imbalance also structures debates around orbital debris, a problem researchers follow closely as orbital congestion progresses. The Nature study gives it a new quantitative translation for space astronomy.
No international regulation currently imposes reflectivity constraints on constellation operators. The only significant voluntary commitment is SpaceX’s, which has deployed less reflective surface treatments on its satellites since 2020, the “VisorSat” and then “DarkSat” variants. Ground-based measurements showed a real reduction in optical magnitude. But these voluntary efforts remain unilateral, concern only one constellation among dozens, and were not designed for the specific constraints of space telescopes.
The International Telecommunication Union, which allocates frequencies and registers orbits, has no mandate to impose reflectivity criteria. The International Astronomical Union has published non-binding recommendations since 2020. The UN’s Committee on the Peaceful Uses of Outer Space (COPUOS) examines the subject without reaching enforceable standards.
Low Earth orbit on the horizon to 2040
Borlaff and co-authors’ study goes beyond present assessment. By 2040, low Earth orbit must function both as global connectivity infrastructure and as usable space for science. Public missions risk absorbing alone the costs of congestion they did not create.
Two trajectories are discernible, not as predictions, but as different regimes of orbital governance.
In the first regime, preventive regulation organizes itself before major constellations reach their nominal size. The tools exist in principle: differentiated altitude caps according to use, maximum reflectivity requirements inscribed in operating licenses, and most importantly real-time distribution of ephemerides—that is, orbital position data—enabling observatories to plan exposures while avoiding satellite passes. This last measure is technically accessible today; constellations have the data, telescopes have planning software. What is missing is the obligation to share.
This scenario assumes difficult convergence: national telecommunications regulators, space agencies, and private operators must gather around a common framework. Precedents exist; coordination of radio frequencies between countries has been mandatory for decades, but they took time and crises to impose themselves.
In the second regime, congestion establishes itself by default. Each operator deploys according to its own standards, constellations reach sizes close to current projections, and each observatory separately absorbs the cost of planning, algorithmic correction, and lost data. Agencies like NASA, ESA, or China’s National Space Administration strengthen their software, reschedule their observations, and publish results on accessible sky zones. Contamination varies according to orbital planes and pointing constraints, without demonstration of permanent blind spots.
This second regime is not catastrophic for all space sciences. Targeted missions focused on specific objects—particular exoplanets, supernovae, asteroids—remain possible with fine planning. What becomes difficult are large statistical surveys of the entire sky, those for which SPHEREx and Xuntian are prototypes. These programs produce data serving thousands of scientists across different disciplines; their impoverishment is a diffuse cost, invisible in a single invoice, but real over time.
The signal that would distinguish the two trajectories before 2030 is relatively precise: adoption or rejection, by the International Telecommunication Union or by an agreement among major space agencies, of a minimum standard for real-time ephemerides sharing. If this norm emerges, the preventive regime remains achievable. If authorizations continue being granted without this requirement, default congestion becomes the baseline scenario.
Real technical solutions, political will to build
Astronomy is not without resources facing this challenge. Several concrete paths are documented, even if none suffices alone.
On the instruments side, next telescopes can be designed with adapted exposure strategies: short and stacked sequences rather than continuous long exposures, spectral filters attenuating reflected sunlight in the most affected wavelengths. These adjustments have a cost—more computing time, less photometric depth—but they reduce contamination without awaiting regulatory reform.
On the satellite side, the most promising direction remains reduced reflectivity from design onward. Less bright surfaces exist. Absorbent paints and geometries that diffuse light rather than reflect it are documented. Additional cost per satellite is marginal at constellation scale. What is missing is the obligation.
Ephemerides sharing is probably the measure with the best impact-to-cost ratio. Operators possess orbital information, but commonly accessible data are not precise enough to finely predict all streaks. Making this information accessible in real time would allow observatories to plan observations accordingly. SpaceX has had technical exchanges with the IAU center and publishes Starlink ephemerides via Space-Track; generalizing and standardizing this sharing is an operational decision, not an engineering challenge.
Borlaff and co-authors’ study raises a governance problem of a shared resource. Low Earth orbit belongs to no single actor, but many operators crowd it. Scientific instruments depending on it, from fundamental research to agricultural applications using satellite data, do not participate in launch authorization procedures. Granting them a place in these decisions is a political choice.
