Removal of telluric lines

Telluric lines disturbe more and more with increasing resolution (R > 8000), but they are then also resolved and thus be eliminable in principle.

They are also included in low-resolution spectra, but they are not resolved there. That is, in a broad H alpha line of a star, they are hidden and not resolved and therefore cannot be eliminated. In this case we cannot remove in no way the effects of the telluric lines from the measured spectrum.

The intensities (equivalent width W = area under a line in the normalized spectrum) of the telluric lines are dependent on

Therefore, professional spectrographs will be used on telescopes, which are on high mountains in territories of low humidity (eg at the periphery of a desert).

The intensity ratio of water lines (abbreviated as WL) is constant among each other. The same is true for the oxygen bands. If you have a model derived from the practice range of telluric lines and folds it with the apparatus function (in practice, by modeling a modifiable Gaussian function), we can apply the theoretical spectrum as a normalization spectrum to the measured spectrum and thus removes telluric lines - loosely speaking - we dehumidify the spectrum.

VSpec for example, offers a tool that is operated with the mouse and is very intuitive. For MIDAS, I have a script that does this task with a command.

Link to telluric lines webpage of C. Buil.



The graph on the left side shows the H alpha line from Mizar A. The bottom spectrum with the sharp spikes of the water lines is measured. In the upper spectrum the WL's are minimized to a few small remnants( using MIDAS). The remnants are produced by

  • Inaccuracies in the chosen (adopted) apparatus function,
  • Inaccuracies in the calibration,
  • The line profiles are not exaktly Gauss function like (blends).

In case of high standards of accuracy in the determination of the equivalent widths (W), the telluric lines need to be eliminated, because they mimic typically increased EW's by about 5%. The 5% is only a fist value. In reality, it depends on the air humidity.

Here is a second example:

The graph shows the used reference spectrum of the telluric lines in a narrow wavelength window. This is the graphical interpretation of a table of the ESO, which contains the wavelengths and relative intensities of each line input.

Such a table of telluric lines is also implemented in VSpec (file H2O.dat).

To use the above reference spectrum the 1-pixel sharp lines in the diagram are transformed into Gaussian profiles. This occurs within MIDAS by a function that is determined from the measured spectrum:

  • One parameter is the measured FWHM of the water lines to be specified by the user.
  • The second parameter is an intensity factor (represents the observed intensities of WL in the measured spectrum depending on air humidity conditions).

The "theoretical" water lines are "shaped" that they coincide approximately with the measured water lines. Exact is of course impossible, as measured lines have only approximately a Gaussian profile.

Is shown on the leftside the tailored reference spectrum to an individual spectrum. It shows the theoretical spectrum now as if I had measured it at a star, which itself has a normalized continuum with no own photospheric lines, so that only the telluric lines would be measured.

Left, a measured spectrum of eps Aur is shown (black line). In this above adaptive reference spectrum was used (standardization). The result is in red. The water lines are significantly reduced. The remaining residues would play in a EW calculation not matter anymore.

back to content