Details of the eleanor Pipeline

These slides were presented virtually at the Earth 2.0 workshop I on December 10, 2020.

Transcript

1. Unpacking eleanor, a Python tool
   for TESS FFI light curve extraction
   Adina Feinstein, Benjamin Montet, Dan Foreman-Mackey, Megan Bedell,
   Nicholas Saunders, Jacob Bean, Jessie Christiansen, Christina Hedges, Rodrigo
   Luger, Daniel Scolnic, José Vinícius de Miranda Cardoso
   !1
   December 10, 2020
   ET Workshop @afeinstein20
   

   View Slide

2. !2 Image Credit: Ethan Kruse
   The TESS FFIs cover ~107 stars brighter than Tmag < 16.
   

   View Slide

3. eleanor was really designed with the
   user in mind.
   !3
   • Flexibility
   • Durability
   • Open source
   • Well documented
   • Easy to integrate with other software
   

   View Slide

4. What steps are taken to extract and
   create a light curve?
   !4
   • Background estimation & subtraction
   • Aperture selection
   • Creation of quality flags
   • Detrending using co-trending basis vectors (CBVs)
   • Optimization of light curve for transit searches
   

   View Slide

5. What steps are taken to extract and
   create a light curve?
   !5
   • Background estimation & subtraction
   • Aperture selection
   • Creation of quality flags
   • Detrending using co-trending basis vectors (CBVs)
   • Optimization of light curve for transit searches
   

   View Slide

6. The eleanor pipeline takes the FFIs and breaks them
   into smaller, more space manageable “postcards.”
   !6
   

   View Slide

7. Background approach #1: We mask the stars and estimate a
   constant background from each cadence across the postcard.
   !7
   

   View Slide

8. Background approach #2: After masking bright sources, we
   linearly interpolate across the postcard to model the expected
   background.
   !8
   

   View Slide

9. The modeled pixel values are subtracted from
   the TPF before the light curve is created.
   !9
   

   View Slide

10. Background approach #3: We estimate a constant
    background from each cadence from the Target Pixel
    File (TPF).
    !10
    

    View Slide

11. There is a difference between the postcard and TPF
    estimated 1D backgrounds, but both capture large
    features nicely.
    !11
    

    View Slide

12. We store each of the three backgrounds in the
    eleanor FITS file and indicate the best background
    through the header keyword `BKG_LVL`.
    !12
    

    View Slide

13. What steps are taken to extract and
    create a light curve?
    !13
    • Background estimation & subtraction
    • Aperture selection
    • Creation of quality flags
    • Detrending using co-trending basis vectors (CBVs)
    • Optimization of light curve for transit searches
    

    View Slide

14. We test out a variety of aperture shapes, sizes,
    and pixel weightings for each target.
    !14
    

    View Slide

15. We limit faint (Tmag > 13) targets to the
    smallest apertures in the library.
    !15
    

    View Slide

16. And we do the reverse for the brightest 

    (Tmag < 8) targets as well.
    !16
    

    View Slide

17. What steps are taken to extract and
    create a light curve?
    !17
    • Background estimation & subtraction
    • Aperture selection
    • Creation of quality flags
    • Detrending using co-trending basis vectors (CBVs)
    • Optimization of light curve for transit searches
    

    View Slide

18. Quality flags are taken from the 2-minute targets and we
    create our own from regions with large, non-astrophysical
    centroid motions and bad background.
    !18
    

    View Slide

19. The centroid motion is tracked from the
    pointing model we create per each sector.
    !19
    

    View Slide

20. What steps are taken to extract and
    create a light curve?
    !20
    • Background estimation & subtraction
    • Aperture selection
    • Creation of quality flags
    • Detrending using co-trending basis vectors (CBVs)
    • Optimization of light curve for transit searches
    

    View Slide

21. The co-trending basis vectors are provided by the SPOC
    pipeline. We bin them down to 30-minutes to work with the
    FFIs.
    !21
    

    View Slide

22. The eleanor light curves are optimized for transit searches by
    calculating the combined differential photometric precision (CDPP).
    !22
    eleanor corrected flux
    light curve for CDPP calculation
    

    View Slide

23. !23
    There are 4 light curves set as the
    default per each method in eleanor.
    

    View Slide

24. Each light curve was processed with a different technique.
    !24
    

    View Slide

25. The raw flux has only the background removed and no
    other processing.
    !25
    

    View Slide

26. The corrected flux is a combination of CBVs and regressing
    against a linear model of position, background, and time.
    !26
    

    View Slide

27. The PCA flux detrends the raw flux by only applying the
    CBVs.
    !27
    

    View Slide

28. The eleanor software allows for PSF modeling of light
    curves. This is not available in the data product.
    !28
    

    View Slide

29. There are pros and cons to each of the flux
    options created through eleanor.
    !29
    Raw Corrected PCA PSF
    Transits/eclipses x x x
    Long term trends x x
    Short term variability x x
    Solar system science x
    

    View Slide

30. eleanor was really designed with the
    user in mind.
    !30
    • Flexibility
    • Durability
    • Open source
    • Well documented
    • Easy to integrate with other software
    

    View Slide

31. We provide every raw & corrected light curve for
    each aperture tested.
    !31
    

    View Slide

32. Software users are also able to input their own
    apertures and apply all of the eleanor detrending
    techniques.
    !32
    

    View Slide

33. The eleanor software also includes visualization tools
    developed to make light curve vetting easier.
    !33
    

    View Slide

34. eleanor was really designed with the
    user in mind.
    !34
    • Flexibility
    • Durability
    • Open source
    • Well documented
    • Easy to integrate with other software
    

    View Slide

35. Maintaining an open source package leads to
    community-driven improvements.
    !35
    

    View Slide

36. eleanor was really designed with the
    user in mind.
    !36
    • Flexibility
    • Durability
    • Open source
    • Well documented
    • Easy to integrate with other software
    

    View Slide

37. By creating thorough example notebooks, and updating
    them as we learned, makes it easy for users to explore their
    own science with our tool.
    !37
    

    View Slide

38. eleanor was really designed with the
    user in mind.
    !38
    • Flexibility
    • Durability
    • Open source
    • Well documented
    • Easy to integrate with other software
    

    View Slide

39. One of the biggest lessons learned is to check different
    light curves to make sure your source is truly astrophysical.
    !39
    OS19 = Oelkers & Stassun (2019)
    TASOC = TESS Asteroseismic Science Consortium
    

    View Slide

40. We’ve incorporated functions to easily input
    eleanor light curve outputs to lightkurve objects.
    !40
    

    View Slide

41. eleanor Resources
    !41
    • PASP paper: https://ui.adsabs.harvard.edu/abs/2019PASP..131i4502F/abstract
    • Documentation: http://adina.feinste.in/eleanor
    • GitHub: https://github.com/afeinstein20/eleanor
    • YouTube Tutorial: https://www.youtube.com/watch?v=xpvniFrA6V0
    • ET Google Colab Notebook: https://bit.ly/3mXY9ex
    

    View Slide