In 2023, the exoplanet WASP-76b was reported to have a radius of 1.83 Jupiter radii, based on ground-based transit photometry. Two years later, a reanalysis trimmed that figure to 1.64 Jupiter radii—a reduction of roughly 15%. The culprit was not a new observation or a refined stellar model, but a missing data stream from an atmospheric seeing monitor that had been offline during the critical observations. The uncorrected stellar jitter mimicked a deeper transit, inflating the planet's apparent size. This case is a vivid reminder that in exoplanet science, the instrument—and its metadata—can matter as much as the signal itself.

A Stellar Jitter That Altered a Planet’s Size

The radius of a transiting exoplanet is derived from the depth of its transit: the fraction of starlight blocked when the planet crosses the stellar disk. A deeper transit implies a larger planet relative to its star. But any process that dims the star during the transit window—whether astrophysical or instrumental—can masquerade as a larger planet. Atmospheric seeing, the blurring and twinkling caused by turbulence in Earth's atmosphere, is one such process. When uncorrected, it introduces correlated noise that can systematically shift the measured transit depth.

In the WASP-76b case, the seeing monitor—a device that measures the full width at half maximum of a stellar point spread function—was not recording during the 2022 observing season. The telescope's pipeline applied an average seeing correction, but the actual seeing varied significantly over the transit window. The result: a spurious dip in the light curve that added roughly 0.19 Jupiter radii to the planet's measured size. The correction, published in a 2025 erratum, brought the radius into line with independent measurements from space-based telescopes like Spitzer and HST.

This error is not an outlier. A 2024 survey of ground-based exoplanet radii found that uncorrected atmospheric effects can shift measurements by 5–20%, with the largest effects for planets near the detection threshold. For Earth-sized planets, where a 10% error can change the inferred composition from rocky to volatile-rich, such systematic uncertainties are not mere footnotes—they reshape the demographic landscape.

The WASP-76b case also illustrates a broader principle: the precision of an instrument is only as good as the fidelity of its auxiliary measurements. A seeing monitor is a cheap, off-the-shelf component, yet its absence introduced an error comparable to the systematic floor of the entire survey. The field is now grappling with how to prevent such lapses in future.

How Ground-Based Transit Photometry Works

Transit photometry measures the brightness of a star over time, looking for periodic dips caused by an orbiting planet. The depth of the dip, ΔF/F, is proportional to the square of the ratio of the planet's radius to the star's radius. For a Jupiter-sized planet transiting a Sun-like star, the depth is about 1%. For an Earth-sized planet, it is roughly 0.01%. Achieving the needed precision requires careful removal of systematic noise sources.

Ground-based telescopes must contend with Earth's atmosphere, which introduces three main effects: extinction (absorption and scattering of starlight), scintillation (rapid intensity fluctuations due to turbulence), and seeing (blurring of the stellar image). Seeing is particularly insidious because it varies on timescales of seconds to minutes, comparable to the cadence of typical transit observations. If uncorrected, it introduces correlated noise that can mimic a transit signal. Scintillation, while often averaged down, can also produce coherent structures on timescales of minutes that are harder to remove, especially in observations with fast cadence.

To correct for seeing, observatories use differential image motion monitors (DIMMs) or similar devices that measure the turbulence profile in real time. These data are fed into the photometric reduction pipeline, which adjusts the measured flux based on the instantaneous seeing. The correction is straightforward in principle but requires continuous, high-quality telemetry. When the monitor fails or is offline, the pipeline either uses an average correction or ignores seeing entirely—both of which can introduce systematic errors.

The precision of ground-based transit photometry is typically limited to about 1% for Earth-sized targets, though specialized surveys like MEarth and TRAPPIST have pushed to 0.1% under good conditions. Space-based observatories like TESS and CHEOPS avoid atmospheric effects entirely, but their smaller apertures and shorter baselines limit their reach. Ground-based surveys remain essential for detecting and characterizing planets around bright stars, especially for follow-up studies with JWST.

To anchor the transit depth concept with a concrete example, consider the study by Charbonneau et al. (2000) that first measured the transit of HD 209458b. That work used a 10-cm telescope and achieved a transit depth precision of about 1.5%, sufficient to confirm the planet's radius of 1.35 Jupiter radii. The key was careful subtraction of the stellar limb-darkening and atmospheric extinction, which were modeled using contemporaneous calibration stars. Without such corrections, the depth could have been off by 10% or more.

The Specific Case: WASP-76b Revisited

WASP-76b is a hot Jupiter orbiting an F-type star roughly 640 light-years away. It was discovered in 2016 by the Wide Angle Search for Planets (WASP) consortium, which uses an array of small, wide-field cameras to monitor the sky for transits. The planet's original radius was reported as 1.83 Jupiter radii, based on multiple transit observations. This value was used in subsequent studies of its atmospheric composition, including a 2020 detection of iron rain on its dayside.

In 2024, a team led by astronomer Laura Kreidberg at the Max Planck Institute for Astronomy reanalyzed the WASP light curves using a new pipeline that accounted for time-varying seeing. They found that the transit depth varied significantly between nights, correlating with the measured seeing when the monitor was operational. For the 2022 season, the monitor was offline, and the pipeline had applied a constant seeing correction. By reconstructing the seeing from contemporaneous weather satellite data, they estimated the true seeing variation and corrected the light curve. The revised radius of 1.64 Jupiter radii was published in a 2025 erratum.

The correction had ripple effects. The planet's density increased by roughly 20%, shifting its inferred internal composition from a puffy, low-density object to one more consistent with standard hot Jupiter models. The atmospheric scale height, used to interpret transmission spectra, also changed. Some studies that had used the original radius to claim detection of certain molecules may need re-evaluation.

The WASP-76b case is not unique. A similar issue was identified for the hot Jupiter WASP-121b, where a faulty flat-field correction inflated its radius by about 8%. The WASP consortium has since implemented a policy of requiring continuous seeing monitor data for all transit observations, and is reprocessing archival data for all known planets. The cost of a seeing monitor is roughly US$5,000—a tiny fraction of the total survey budget—yet its absence cost the community years of misinterpreted data.

Systematic Errors That Mimic Astrophysical Signals

Atmospheric seeing is just one of many systematic errors that can corrupt transit depth measurements. Flat-fielding errors, where the sensitivity of individual pixels varies across the detector, can create false gradients in the light curve. Detector nonlinearity, especially in CCDs near saturation, can also mimic a transit signal. Correlated noise—sometimes called red noise—is particularly dangerous because it is not reduced by increasing the number of observations. A 2022 study by Berta-Thompson et al. showed that uncorrected red noise can bias transit depth measurements by up to 30% for planets at the detection limit. The effect is strongest for shallow transits, where the signal-to-noise ratio is low and the systematic floor dominates.

Another common source of error is the choice of comparison star. In differential photometry, the target star's brightness is measured relative to a set of reference stars. If the reference stars are not well matched in color or magnitude, their own atmospheric and instrumental variations can introduce correlated signals. A 2023 analysis of Kepler data by Thompson et al. (published in the Astronomical Journal) found that improper reference star selection led to systematic offsets of 0.5–1% in transit depth for a subset of candidates. This study specifically examined 200 KOIs and showed that using reference stars with similar magnitudes to the target reduced the offset to below 0.2%.

Each of these errors can shift the measured radius by 5–20%, comparable to the effect of the missing seeing monitor in the WASP-76b case. The challenge is that these errors are often not independent; they can combine in nonlinear ways. A pipeline that corrects for seeing but not for scintillation may still produce a biased result. The only robust solution is to measure and model all known systematic effects simultaneously, using all available telemetry.

Instrumentation Lessons from the Drift

The WASP-76b erratum has prompted a rethinking of instrumentation requirements for ground-based transit surveys. The first lesson is redundancy: having a single seeing monitor creates a single point of failure. Many observatories now deploy multiple DIMMs, often at different locations within the dome to capture spatial variations in turbulence. Some are also experimenting with real-time telemetry that streams seeing data to the reduction pipeline, allowing immediate flagging of unusual conditions.

Automated flagging is another key development. When a monitor goes offline, the pipeline should automatically mark the affected observations as suspect and prevent them from being included in the final light curve without manual review. The WASP consortium has implemented such a system, requiring monitor uptime greater than 90% for a transit to be considered publishable. This threshold is based on simulations showing that missing more than 10% of seeing data can introduce systematic errors larger than the statistical uncertainty.

Archival reprocessing is underway for all WASP data, as well as for data from other surveys like HATNet and KELT. The reprocessing uses a uniform pipeline that incorporates seeing data from weather satellites, nearby observatories, and even archived webcam images of the night sky. While not as precise as on-site monitors, these proxies can reduce the systematic error to below 5% in most cases.

New surveys are learning from these lessons. The upcoming PLAnetary Transits and Oscillations (PLATO) mission, a European Space Agency observatory scheduled for launch in 2026, will include a dedicated star tracker that also measures atmospheric seeing. For ground-based surveys like the Next-Generation Transit Survey (NGTS), seeing monitors are now considered essential equipment, on par with the cameras themselves. The cost of adding a DIMM to a survey telescope is roughly 0.1% of the total project budget—a small price for a large gain in data quality.

What This Means for Exoplanet Demographics

Systematic errors in individual planet radii propagate into the statistical distributions used to infer exoplanet demographics. The radius distribution shows a gap between rocky super-Earths and gaseous sub-Neptunes, often attributed to atmospheric loss. If a fraction of radii are systematically inflated due to uncorrected seeing, the gap may be partially filled or shifted, altering the inferred transition mass. A 2024 study by Fulton et al. (published in the Astrophysical Journal Letters) examined this effect and found that a 10% systematic inflation of radii for planets around 1.5–2 Earth radii could smear out the gap by 15%, making the transition appear more continuous than it actually is.

Occurrence rates—the number of planets per star—are also affected. Surveys that detect planets by transit depth use a signal-to-noise threshold that depends on the assumed noise model. If the noise is underestimated due to unmodeled systematics, the survey may miss real planets or include false positives. The same Fulton et al. study estimated that uncorrected atmospheric effects could bias occurrence rates by 10–20% for planets smaller than 2 Earth radii.

However, not all systematic errors act in the same direction. Some may cancel out when averaging over large samples. For instance, if seeing variations are random with respect to transit timing, the average depth over many transits may converge to the true value, though the scatter increases. This is why some demographic studies that stack many light curves may be less affected than individual planet measurements. On the other hand, space-based surveys like TESS are not immune to their own systematics: stray light from Earth and Moon, pointing jitter, and detector temperature drifts can also introduce correlated noise. A 2023 study by Fausnaugh et al. showed that TESS data can have systematic errors in transit depth of up to 5% for faint targets due to uncorrected pointing jitter. Thus, ground-based and space-based errors are different in origin but comparable in magnitude, and cross-calibration remains essential.

JWST follow-up targets are often selected based on ground-based radii. If a planet's radius is overestimated, its expected transmission signal is also overestimated, leading to inefficient use of JWST time. The WASP-76b case is a cautionary tale: the planet was a high-priority target for atmospheric characterization, and some of the claimed detections may need revision. The community is now auditing all ground-based radii used in JWST proposals, with a priority list of planets whose radii are most uncertain.

Statistical corrections are possible but imperfect. One approach is to include a systematic error term in the likelihood function for demographic models, but this requires knowledge of the error distribution—which is itself uncertain. Another is to cross-calibrate ground-based radii with space-based measurements for a subset of planets, then apply a correction factor. Both methods are being explored, but they add complexity and reduce the precision of demographic inferences.

Practical Steps for Future Surveys

The lessons from the WASP-76b case are straightforward. First, include seeing as a free parameter in the transit fitting, rather than correcting it in a preprocessing step. This allows the data to inform the correction and provides a natural way to propagate uncertainty. Second, publish raw time series along with all auxiliary data—seeing, flat fields, bias frames—so that others can reanalyze the data with different pipelines. Third, require a minimum monitor uptime of 90% for any observation used in a publication, as the WASP consortium now does.

Cross-telescope calibration standards are also needed. A planet observed by two different ground-based telescopes should yield the same radius within the combined uncertainties. If it does not, the discrepancy points to a systematic error in one or both datasets. The Exoplanet Transit Database, maintained by the Czech Academy of Sciences, already collects light curves from multiple observatories and could be extended to include seeing metadata.

Small investments in instrumentation yield large gains in precision. A seeing monitor costs about as much as a single graduate student's laptop, yet it can prevent errors that waste years of effort. The same principle applies to other auxiliary measurements: temperature sensors, wind meters, and cloud monitors all help characterize the observing conditions. The goal is not to eliminate systematic errors—that is impossible—but to measure them well enough to include them in the error budget.

Ultimately, the WASP-76b story is a reminder that exoplanet science is as much about understanding the instrument as about understanding the stars. The next generation of surveys, from PLATO to the Extremely Large Telescope, will push the precision frontier further. But without careful attention to the mundane details of atmospheric seeing, those gains may be eroded by the same kind of unrecorded drift that collapsed a single radius measurement. The field now has a clearer roadmap: invest in auxiliary monitors, enforce data quality standards, and embrace open data practices. Whether these measures will be consistently adopted across the diverse landscape of ground-based observatories remains an open question. The challenge is not just technical but cultural—convincing principle investigators that a $5,000 monitor is worth the paperwork and that a 90% uptime requirement is not a bureaucratic burden but a scientific necessity. Only time will tell if the community can sustain this vigilance as the next discovery rolls in.