A statistical analysis of 33,396 TESS sector observations to quantify errors in the pipeline's CROWDSAP metric. White dwarfs serve as stable photometric flux standards.
Presented at TASC7 / KASC14 in Hawaii and TASC NYC. Outstanding Research Award in Physics.
TESS monitors millions of stars, but in dense fields, light from neighboring stars contaminates each target's measurement aperture. This crowding affects every astrophysical measurement that depends on flux: planet transit depths, stellar variability amplitudes, asteroseismic signals.
The TESS pipeline estimates contamination automatically with a metric called CROWDSAP (cp). For most stars in clean fields, it works fine. But for faint targets or complex scenes, the metric can be substantially wrong, and that error propagates silently into every downstream analysis.
We needed a way to empirically validate CROWDSAP across the entire sky. White dwarfs gave us the leverage: intrinsically stable, well-characterized stellar remnants that work as photometric flux standards.
We anchored each observation's measured flux to the expected flux from the TESS Input Catalog. The discrepancy gave us a "ground truth" crowding estimate to compare directly against the pipeline's CROWDSAP value.
For each white dwarf observation, we computed an estimated crowding metric cest from the ratio of expected to observed flux:
Stars where R ≈ 1 confirm the pipeline got it right. Stars where R deviates significantly are flagged for investigation. The discrepancy distribution tells us about systematic failure modes.
We pulled 8,132 white dwarfs across 33,396 sector observations, parsing TESS pipeline data headers in Python.
White dwarfs were chosen for stability: they don't pulsate at amplitudes that contaminate the analysis, they're well-characterized in the catalog, and they cover the full TESS magnitude range.
We used iterative 5-sigma clipping on the discrepancy distribution to identify statistically significant outliers from the bulk.
Each iteration recomputed the median and standard deviation excluding previously flagged points. Convergence took 3-4 passes for the full dataset.
The TESS pipeline is generally reliable. We identified roughly 200 specific failure modes where CROWDSAP was substantially inaccurate, falling into four mechanistic categories:
Near the detection limit, the pipeline misattributes background noise as source flux. The signal-to-noise floor breaks the assumptions in CROWDSAP's calculation.
Cataclysmic variables and binary systems violate the stable-flux assumption baked into the calibration. Their physics, not the crowding, drives the discrepancy.
When a nearby bright star saturates the detector, charge bleeds into adjacent pixels and contaminates the target aperture in ways the pipeline's contamination model doesn't anticipate.
Solar system objects, mostly asteroids, pass through the aperture during long observations. They temporarily inflate the measured flux without altering the catalog flux.
The output: a list of flagged TESS observations where CROWDSAP should not be trusted, organized by failure mode. Future TESS analyses can apply the categorization to identify when a target's photometry needs special handling, which matters most for borderline-detection signals where pipeline systematics can swamp real physics.