TESS Back ML

TL,DR

This is a Python package to create specialized support data from the TESS Full Frame Image (FFI). With tess-backml you can isolate the scatter light (SL) contribution to FFIs and create downsize versions of the cubes for multiple uses:

• One use is to correct the SL signal in any FFI cutout by interpolating from a downsize version of the cube.
• Another use is to create a machine learning training set with the SL cube and pixel maps of how the Earth and Moon angles change during sector observations. These two are the main contributions to the background signal.
• This package, also will enable ML model evaluation to correct the scatter light on the fly.

Install

Install from this GitHub repository with

pip install git+https://github.com/jorgemarpa/tess-backml

PyPI will available shortly.

Scatter Light Cube

This animation shows the scattered light of a TESS FFI. The original 2048 x 2048 pixel image was downsized to 128 x 128 pixels.

Angle Maps

The next figure shows the vector maps (distance, elevation, and azimuth angles) for Earth and Moon with respect to the camera boresight. These maps have the same shape as the scatter light cube shown above.

Usage

Create a SL Cube

From the original FFI cube size (~1200 times, 2k, 2k), we can extract and compute a downsize version of the SL by doing pixel binning and time binning. The resulting SL cube has typically shape (~300, 256, 256).

from tess_backml import BackgroundCube

bkg_data = BackgroundCube(
    sector=2, camera=1, ccd=1, img_bin=8, downsize="binning"
)

bkg_data.get_scatter_light_cube(frames=None, mask_straps=True, plot=True, rolling=True, errors=True)

bkg_data.bin_time_axis(bin_size=2.0)

fname = f"./data/ffi_sl_cube_sector{sector:03}_{camera}-{ccd}.fits"
_ = bkg_data.save_to_fits(out_file=fname, binned=True)

This saves a FITS file that can be used later to correct any TPF. For more details see the SL Cube Tutorial.

Correct SL in TPF

Once we have a FITS file with the SL cube, we can use the corrector object to interpolate to the desired pixels and times, and correct the background SL.

import numpy as np
import lightkurve as lk

from tesscube import TESSCube
from tess_backml.corrector import ScatterLightCorrector

tcube = TESSCube(sector=2, camera=1, ccd=1)
# this is pseudo code
tpf = tcube.get_tpf(<TIC coords>, shape=(25, 25))

# Corrector object
slcorr = ScatterLightCorrector(sector=2, camera=1, ccd=1, fname=fname)

# pixel array for evaluation
row_eval = np.arange(tpf.row, tpf.row + tpf.shape[1])
col_eval = np.arange(tpf.column, tpf.column + tpf.shape[2])
# time array for evaluation
time_eval = tpf.time.value + slcorr.btjd0

# evaluate the SL model at given pixel/times
sl_eval = slcorr.evaluate_scatterlight_model(
    row_eval=row_eval, col_eval=col_eval, times=time_eval
)

# subtract the SL model
flux_corr = tpf.flux - sl_eval * tpf.flux.unit

For more details see the Corrector Tutorial.

Create ML Training Set

To get the training data follow the steps:

from tess_backml import Background_Data

# initialize the object for given sector/camera/ccd
# will do 16x16 pixel binning
tess_bkg = BackgroundCube(
    sector=1, camera=1, ccd=1, img_bin=16, downsize="binning"
)

# get the flux data from MAST/AWS, compute scatter light and downsize
tess_bkg.get_scatter_light_cube(plot=False)
# compute the vector maps for the Earth and Moon
tess_bkg.get_vector_maps(ang_size=True)

# make an animation of the scatter light cube
tess_bkg.animate_data(data="sl", save=False, step=10);

# save data to disk
tess_bkg.save_to_npz(save_maps=True)

Or you can run a Python script build_dataset.py in the terminal (plotting flag is optional and will add run time):

usage: build_dataset.py [-h] [--sector SECTOR] [--camera CAMERA] [--ccd CCD] [--downsize DOWNSIZE] [--image-bin IMG_BIN] [--plot]
                        [--out-dir OUT_DIR]

Build TESS FFI background dataset for a Sector/Camera/CCD.

options:
  -h, --help           show this help message and exit
  --sector SECTOR      TESS sector.
  --camera CAMERA      TESS camera.
  --ccd CCD            TESS camera.
  --downsize DOWNSIZE  Method for downsizing the image, one of [sparse, binning].
  --image-bin IMG_BIN  Image binning size, must divide 2048.
  --plot               Plot target light curve.
  --out-dir OUT_DIR    Outputh directory path where files and figures will be saved.

For example

python build_dataset.py --sector 3 --camera 1 --ccd 2 --image-bin 16 --downsize binning --plot

The script will save npz files with the following dictionary:

"scatter_cube":   the data cube with the scatter light, shape [nt, 128, 128] depending on the binsize (16)
"time":           time array in JD
"cadenceno":      cadence number array
"earth_alt":      Earth altitude angle from the camera boresight
"earth_az":       Earth azimuth angle from the camera boresight
"earth_dist":     Earth distance from the camera boresight
"moon_alt":       Moon altitude angle from the camera boresight
"moon_az":        Moon azimuth angle from the camera boresight
"moon_dist":      Moon distance from the camera boresight
"earth_alt_map":  Earth altitude angle map for every pixel, same shape as `scatter_cube`
"earth_az_map":   Earth azimuth angle map for every pixel 
"earth_dist_map": Earth angular size map for every pixel 
"moon_alt_map":   Moon altitude angle map for every pixel 
"moon_az_map":    Moon azimuth angle map for every pixel 
"moon_dist_map":  Moon angular size map for every pixel 

For more details see the Data Cubes Tutorial.