Default arguments for flatten overfit in TESS light curves

Problem description

While removing long term trends of TESS light curves, I’ve noticed that the current default arguments for flatten usually overfit transits.

Example

import lightkurve as lk

pixels = pixels = lk.search_targetpixelfile("wasp-124b").download()
lightcurve = pixels.to_lightcurve()
flat_lightcurve, trend  = lightcurve.flatten(return_trend = True)
trend.plot()
exoplanet = flat_lightcurve.fold(period=3.372)
exoplanet.plot()

Expected behavior

Increasing the break_tolerance and the window length helps a lot:

flat_lightcurve, trend  = lightcurve.flatten(window_length = 1001, break_tolerance = 50, return_trend = True)
trend.plot()
exoplanet = flat_lightcurve.fold(period=3.372)
exoplanet.plot()

I didn’t see this overfit behavior in Kepler/K2 lightcurves.

Environment:

• platform (e.g. Linux, OSX, Windows): OSX
• lightkurve version (e.g. 1.0b6): I’m living on the bleeding edge (1.1.dev) 😄
• installation method (e.g. pip, conda, source): source

[gully]

Note this is near-duplicate of #486

[orionlee]

On a related note, some other methods’ default parameters might be tailored to Kepler as well.

1. transit_duration for LightCurve.estimate_cdpp() https://docs.lightkurve.org/api/lightkurve.lightcurve.LightCurve.html#lightkurve.lightcurve.LightCurve.estimate_cdpp