How to make mock noises?

Noise-added mock cubes for Monte Carlo fitting

Tokult can add spatially correlated noise maps to the observed data. These noise maps have the same rms with the observed data. Tokult internally uses these noise-added data cubes for the Monte Carlo fitting method on the image plane.

from tokult import Tokult

tok = Tokult.launch('tokult_mockcube_dirty.fits', 'tokult_cube_dirty.psf.fits')
tok.use_region((32, 96), (32, 96), (5, 12))

cube_noise_added = tok.datacube.perturbed(tok.dirtybeam.fullconvolve)

The lensing parameters are not needed only to produce mock noises. Let’s compare the created noise-added cubes with the original data.

import matplotlib.pyplot as plt
fig = plt.figure(figsize=[6.28 * 2, 6.28])
ax0, ax1 = fig.subplots(1, 2)

ax0.imshow(tok.datacube.imageplane[0, :, :], origin='lower')
ax1.imshow(cube_noise_added[0, :, :], origin='lower')

<matplotlib.image.AxesImage at 0x7f77deea62b0>

Note that the shape of tok.datacube.imageplane is (7, 64, 64), so tok.datacube.imageplane[0, :, :] contains the first channel map of the object.

Spatially correlated noise maps

Tokult can output the pure noise maps, with a bit long commands.

rms = tok.datacube.rms(is_originalsize=True)[..., np.newaxis, np.newaxis]
noisecube = tok.datacube.create_noise(rms, tok.datacube.original.shape, tok.dirtybeam.fullconvolve)

Let’s compare again the created pure noise map with the original data.

fig = plt.figure(figsize=[6.28 * 2, 6.28])
ax0, ax1 = fig.subplots(1, 2)

ax0.imshow(tok.datacube.original[0, :, :], origin='lower')
ax1.imshow(noisecube[0, :, :], origin='lower')

<matplotlib.image.AxesImage at 0x7f77ddc4fb80>

Since the noises of tokult_mockcube_dirty.fits is created using this method, the noise patterns looks really similar. Note that the shape of tok.datacube.original is (47, 128, 128), so tok.datacube.original[0, :, :] contains only a noise map without any objects.