De-trending Light Curves¶

AltaiPony currently features three different de-trending aproaches that aim at removing astrophysical and instrumental trends in the time series while preserving the flares.

Call the de-trender on a FlareLightCurve with

detrended_flc = flc.detrend(<detrending method>, **kwargs)

For <detrending method> you can pass one the following:

“k2sc” - DEPRECATED AND NO LONGER MAINTAINED, WILL BE REMOVED IN A FUTURE VERSION! If you wish to de-trend a K2 light curve, this is the best way to deal with, among other issues in K2, the Kepler spacecraft roll (see [1], [2] and [3] for details)
“savgol” - applies a Savitky-Golay [4] filter to the light curve. This method is quick and gives good results in both Kepler and TESS light curves, but you may have to play with the window_length keyword to get optimal results.
“custom” - you can use your own de-trending pipeline. Pass a function that takes a FlareLightCurve and returns a FlareLightCurve to flc.detrend("custom", func=<your custom de-trending function>) .

You can use the custom de-trending method detailed in [6] that can be used to de-trend Kepler and TESS light curves as follows with AltaiPony

from funcs.customdetrend import custom_detrending

flcd = flc.detrend("custom", func=custom_detrending)

The custom_detrending function takes the following keyword arguments:

spline_coarseness (default=30) - time in hours to average over for the spline fit
spline_order (default=3) - the order of the spline fit
savgol1 (default=6) - window length of the first Savitzky-Golay filter
savgol2 (default=3) - the window length of the second Savitzky-Golay filter
max_sigma (default=2.5) - the number of standard deviations above the median to clip the light curve at for de-trending (not for flare searching!)
longdecay (default=6) - the number of consecutive clipped points times this parameter is additionally clipped at the end of the clipped points to avoid smoothing away flare decay
pad (default=3) - outliers in Savitzky-Golay filter are padded with this number of data points to avoid cutting flares after applying the longdecay expansion. The padding is applied before and after the outliers.