Satellite trails touch many HETDEX observations—but mask little survey area

Finding moving objects in a galaxy survey

HETDEX uses spectra—light separated by wavelength—to map the large-scale structure of the Universe. Satellite sunlight can enter those measurements, creating unwanted signals rather than information about distant galaxies.

Erin Mentuch Cooper and colleagues examined 6,778 observations in the survey’s first public data release, spanning January 2017 to July 2024 and roughly 90 square degrees of sky. They asked how often satellite trails appear, which orbiting objects produce them and how their brightness should be interpreted. The arXiv record lists the study as published in The Astronomical Journal, volume 172, article 303, rather than solely as an unpublished preprint.

The team found 527 streaks across 492 observations: 7.3% of the total. That overall figure hides a substantial change over time. Monthly contamination rose from a few percent early in the survey to more than 20% by 2023, peaking at 22.7% in July that year.

From dotted tracks to orbital matches

HETDEX collects spectra through arrays of optical fibers, with gaps between the instrument’s observing patches. A moving satellite therefore leaves a sampled, sometimes dotted-looking track rather than a continuous photographic line.

The researchers used an image-classification system to flag potential satellite detections in images made by adding together light across a range of wavelengths. They then manually selected aligned tracks and fitted straight lines to their sky positions. Five candidates were removed after their spectra revealed meteor emission rather than reflected sunlight.

To identify the remaining objects, the team calculated predicted paths from archival two-line elements: compact records describing satellite orbits. Matching considered position, track angle and timing, using precise exposure times from internal survey records.

They assigned cataloged objects to 475 streaks, or 90.1%. An independent service agreed with 161 of 166 matches it could resolve in a checked subset. Identification was not flawless: five assigned orbits placed the objects inside Earth’s dark shadow at the crossing time, indicating false matches. Another 52 streaks remained unidentified.

Frequent does not mean a whole exposure is lost

The streaks’ frequency and their footprint tell different stories. Although many observations contained a trail, the flagged patches totaled only 0.09% of the effective survey area.

That number depends on the analysis rules. The pipeline masks a strip extending six arcseconds on either side of each track; choosing a wider strip would increase the masked area. Some patches also carry flags for other problems, so the authors describe the estimate as an upper limit on additional survey area lost to satellite contamination under those rules—not a universal measure of all affected light.

In 2023–2024, contamination reached 20.4% in the first hour after astronomical twilight, when the Sun is 18 degrees below the horizon, and averaged 10.5% through the remaining night. Trails were therefore not confined to twilight.

Why distant hardware still shows up

Among identified streaks, 254 were associated with payloads, 172 with rocket bodies and 49 with debris. Nearly half came from geostationary or highly elliptical orbits. Only about 35% were attributed to objects launched since 2010; 34 streaks matched Starlink spacecraft.

Those proportions are not a ranking of today’s satellite populations or their impact on every telescope. The observations began before Starlink launches, giving older hardware more opportunities to appear. The telescope’s large collecting area and roughly six-minute exposures also detect faint, distant objects.

Exposure averaging further complicates brightness comparisons. A fast satellite contributes light only while crossing instrument-covered patches, but its recorded flux is calibrated over the full exposure. Correcting for crossing time and instrument gaps changed the median low-Earth-orbit brightness from magnitude 16.2 to 7.4, compared with 16.7 to 13.0 for geostationary objects. Smaller magnitudes mean brighter sources: apparently similar recorded streaks can represent very different instantaneous brightnesses.

Sunlight, with unresolved details

The combined spectra show solar absorption features, consistent with reflected sunlight. Fainter streaks appear somewhat bluer, but the researchers could not establish whether that reflects different materials or measurement uncertainties, including uncertain blue-wavelength signals and possible sky-subtraction residuals.

The central result is a distinction: satellite contamination became more frequent, while typical measured streak brightness showed no strong evolution. For HETDEX, masking removes only a small area, yet sensitive spectroscopy records the accumulated legacy of decades of launch activity alongside the growth of newer satellite populations.