Reading Spectra from a FITS file

FITS is one of the most common ways to store spectra. In most cases the flux values are stored as the pixel values in a 1-, 2- or 3-D arrays. The mapping of axes indices to physically meaningful values (e.g. to wavelength) is encoded in so FITS keywords (see WCS). This document describes how to extract this information and read it in to the various spectral classes. A lot of information about the keyword header storage is taken from The IRAF/NOAO Spectral World Coordinate Systems. As the FITS keyword headers have a complex way of storing information (often redundantly), we have created a FITS WCS parser (FITSWCSSpectrum) that initializes with a FITS header and can extract/validate the information stored in the various keyword headers making them available in a simple API. All of the FITS readers use this parser object.


FITS keywords can encode two mappings from physical to logical to dispersion. A general approach to logical coordinate systems is in development. Currently, the FITS readers ignore this information and might give wrong results.

Reading simple linear 1D WCS

One of the most common and simple ways that dispersion is encoded in FITS files is linear dispersion using four keywords:

CRVAL1  =    4402.538203477947 # Dispersion at reference pixel
CRPIX1  =                    1 # Reference pixel
CDELT1  =      1.3060348033905 # Dispersion by pixel
CUNIT1  = 'Angstrom'

One can easily create a simple wcs file from this information:

>>> from specutils.wcs import specwcs
>>> from import fits
>>> from astropy import units as u
>>> header = fits.getheader('myfile.fits')
>>> dispersion_start = header['CRVAL1'] - (header['CRPIX1'] - 1) * header['CDELT1']
>>> linear_wcs = specwcs.Spectrum1DPolynomialWCS(degree=1, c0=dispersion_start,
>>>                                              c1=header['CDELT1'],
>>>                                              unit=u.Unit(header['CUNIT1']))
>>> flux = fits.getdata('myfile.fits')
>>> myspec = Spectrum1D(flux=flux, wcs=linear_wcs)

However, this process requires too much code for the user. As this is such a common format, there exists a reader that generates a Spectrum1D object from a fits file directly

>>> from import read_fits
>>> myspec = read_fits.read_fits_spectrum1d('myfile.fits')

read_fits_spectrum1d is the only method required to read various types of FITS formats listed in this page. This method detects the type of WCS, and uses the correct reader accordingly.

Currently only linear one-dimensional WCS is implemented, but the examples should give a guide to implement more complex or WCS created from different keywords.

Reading FITS “multispec” format WCS

The multispec format holds multiple one-dimensional spectra in a single file. The current 1-D readers can be used to read such a file. A list of spectra is returned whenever the input file is a multispec file. Here is an example of reading a simple FITS multispec format:

>>> from import read_fits
>>> spectra_list = read_fits.read_fits_spectrum1d('mymultispec.fits')
>>> spectra_list[0]
Spectrum1D([ 338.06109619,  395.59234619,  326.0012207 , ...,
             660.0098877 ,  686.54498291,  689.58374023])
>>> spectra_list[1].dispersion
<Quantity [ 8293.44875263, 8293.40459999, 8293.36044556,...,
            8166.53073537, 8166.48250242, 8166.43426803] Angstrom>

The multispec format supports various functions to map the pixel indices to dispersion. Currently the following multispec formats are supported:

  • Linear Dispersion Function
  • Log-linear Dispersion Function
  • Legendre Polynomial Dispersion Function
  • Chebyshev Polynomial Dispersion Function
  • Linear Spline Dispersion Function
  • Cubic Spline Polynomial Dispersion Function


Linear and cubic spline functions implementations haven’t been tested due to lack of test files. If you have one of these files, please post at astropy-dev.

Reference/API Module


multispec_wcs_reader(wcs_info[, dispersion_unit]) Extracting multispec information out of WAT header keywords and
read_fits_spectrum1d(filename[, ...]) 1D reader for spectra in FITS format. This function determines what format
read_fits_wcs_linear1d(wcs_info[, ...]) Read very a very simple 1D WCS mainly comprising of CRVAL, CRPIX, ...


FITSWCSSpectrum(fits_header) This class is designed to extract all known spectroscopic WCS keywords from a FITS keyword header.