Creating/reading spectral cubes¶
SpectralCube class is used to
represent 3-dimensional datasets (two positional dimensions and one spectral
dimension) with a World Coordinate System (WCS) projection that describes the
mapping from pixel to world coordinates and vice-versa. The class is imported
>>> from spectral_cube import SpectralCube
Reading from a file¶
In most cases, you are likely to read in an existing spectral cube from a
file. The reader is designed to be able to deal with any
arbitrary axis order and always return a consistently oriented spectral cube
(see Accessing data). To read in a file, use the
read() method as follows:
>>> cube = SpectralCube.read('L1448_13CO.fits')
This will always read the Stokes I parameter in the file. For information on accessing other Stokes parameters, see Stokes components.
In most cases, the FITS reader should be able to open the file in
memory-mapped mode, which means that the data is not immediately
read, but is instead read as needed when data is accessed. This
allows large files (including larger than memory) to be accessed.
However, note that certain FITS files cannot be opened in
memory-mapped mode, in particular compressed (e.g.
files. See the Handling large datasets page for more details about dealing
with large data sets.
>>> cube = SpectralCube(data=data, wcs=wcs)
data can be any Numpy-like array, including memory-mapped Numpy
arrays (as mentioned in Reading from a file, memory-mapping is a technique
that avoids reading the whole file into memory and instead accessing it from
the disk as needed).
Hacks for simulated data¶
If you’re working on synthetic images or simulated data, where a location on the sky is not relevant (but the frequency/wavelength axis still is!), a hack is required to set up the world coordinate system. The header should be set up such that the projection is cartesian, i.e.:
CRVAL1 = 0 CTYPE1 = 'RA---CAR' CRVAL2 = 0 CTYPE2 = 'DEC--CAR' CDELT1 = 1.0e-4 //degrees CDELT2 = 1.0e-4 //degrees CUNIT1 = 'deg' CUNIT2 = 'deg'
Note that the x/y axes must always have angular units (i.e., degrees). If your
data are really in physical units, you should note that in the header in other
spectral-cube doesn’t care about this.
If the frequency axis is irrelevant,
spectral-cube is probably not the
right tool to use; instead you should use astropy.io.fits or some other file reader